Biomedical Engineering Handbook

“FrontMatter.” The Biomedical Engineering HandBook, Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2...

Author: Joseph D. Bronzino

1457 downloads 6022 Views 41MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

“FrontMatter.” The Biomedical Engineering HandBook, Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

Library of Congress Cataloging-in-Publication Data Catalog record is available from the Library of Congress.

This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials or for the consequences of their use. Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, microfilming, and recording, or by any information storage or retrieval system, without prior permission in writing from the publisher. All rights reserved. Authorization to photocopy items for internal or personal use, or the personal or internal use of specific clients, may be granted by CRC Press LLC, provided that $.50 per page photocopied is paid directly to Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923 USA. The fee code for users of the Transactional Reporting Service is ISBN 0-8493-0461-X/00/$0.00+$.50. The fee is subject to change without notice. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. The consent of CRC Press LLC does not extend to copying for general distribution, for promotion, for creating new works, or for resale. Specific permission must be obtained in writing from CRC Press LLC for such copying. Direct all inquiries to CRC Press LLC, 2000 Corporate Blvd., N.W., Boca Raton, Florida 33431. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation, without intent to infringe. © 2000 by CRC Press LLC No claim to original U.S. Government works International Standard Book Number 0-8493-0461-X Printed in the United States of America 1 2 3 4 5 6 7 8 9 0 Printed on acid-free paper

Introduction and Preface

As we enter the new millennium, the prospects for the field of Biomedical Engineering are bright. Individuals interested in pursuing careers in this field continue to increase and the fruits of medical innovation continue to yield both monetary rewards and patient well being. These trends are reflected in this second edition of the Biomedical Engineering Handbook. When compared to the first edition published in 1995, this new two-volume set includes new sections on “Transport Phenomena and Biomimetic Systems” and “Ethical Issues Associated with Medical Technology”. In addition, over 60% of the chapters has been completely revised, incorporating the latest developments in the field. therefore, this second edition is truly an updated version of the “state-of-the-field of biomedical engineering”. As such, it can serve as an excellent reference for individuals interested not only in a review of fundamental physiology, but also in quickly being brought up to speed in certain areas of biomedical engineering research. It can serve as an excellent textbook for students in areas where traditional textbooks have not yet been developed, and serve as an excellent review of the major areas of activity in each biomedical engineering subdiscipline, such as biomechanics biomaterials, clinical engineering, artificial intelligence, etc., and finally it can serve as the “bible” for practicing biomedical engineering professionals by covering such topics as a “Historical Perspective of Medical Technology, the Role of Professional Societies and the Ethical Issues Associated with Medical Technology”. Biomedical Engineering is no longer an emerging discipline; it has become an important vital interdisciplinary field. Biomedical engineers are involved in many medical ventures. They are involved in the design, development and utilization of materials, devices (such as pacemakers, lithotripsy, etc.) and techniques (such as signal processing, artificial intelligence, etc.) for clinical research and use; and serve as members of the health care delivery team (clinical engineering, medical informatics, rehabilitation engineering, etc.) seeking new solutions for difficult heath care problems confronting our society. To meet the needs of this diverse body of biomedical engineers, this handbook provides a central core of knowledge in those fields encompassed by the discipline of biomedical engineering as we enter the 21st century. Before presenting this detailed information, however, it is important to provide a sense of the evolution of the modern health care system and identify the diverse activities biomedical engineers perform to assist in the diagnosis and treatment of patients.

Evolution of the Modern Health Care System Before 1900, medicine had little to offer the average citizen, since its resources consisted mainly of the physician, his education, and his “little black bag.” In general, physicians seemed to be in short supply, but the shortage had rather different causes than the current crisis in the availability of health care professionals. Although the costs of obtaining medical training were relatively low, the demand for doctors’ services also was very small, since many of the services provided by the physician also could be obtained from experienced amateurs in the community. The home was typically the site for treatment and recuperation, and relatives and neighbors constituted an able and willing nursing staff. Babies were delivered by midwives, and those illnesses not cured by home remedies were left to run their natural, albeit frequently fatal, course. The contrast with contemporary health care practices, in which specialized physicians and nurses located within the hospital provide critical diagnostic and treatment services, is dramatic.

© 2000 by CRC Press LLC

The changes that have occurred within medical science originated in the rapid developments that took place in the applied sciences (chemistry, physics, engineering, microbiology, physiology, pharmacology, etc.) at the turn of the century. This process of development was characterized by intense interdisciplinary cross-fertilization, which provided an environment in which medical research was able to take giant strides in developing techniques for the diagnosis and treatment of disease. For example, in 1903, Willem Einthoven, the Dutch physiologist, devised the first electrocardiograph to measure the electrical activity of the heart. In applying discoveries in the physical sciences to the analysis of biologic process, he initiated a new age in both cardiovascular medicine and electrical measurement techniques. New discoveries in medical sciences followed one another like intermediates in a chain reaction. However, the most significant innovation for clinical medicine was the development of x-rays. These “new kinds of rays,” as their discoverer W. K. Roentgen described them in 1895, opened the “inner man” to medical inspection. Initially, x-rays were used to diagnose bone fractures and dislocations, and in the process, x-ray machines became commonplace in most urban hospitals. Separate departments of radiology were established, and their influence spread to other departments throughout the hospital. By the 1930s, x-ray visualization of practically all organ systems of the body had been made possible through the use of barium salts and a wide variety of radiopaque materials. X-ray technology gave physicians a powerful tool that, for the first time, permitted accurate diagnosis of a wide variety of diseases and injuries. Moreover, since x-ray machines were too cumbersome and expensive for local doctors and clinics, they had to be placed in health care centers or hospitals. Once there, x-ray technology essentially triggered the transformation of the hospital from a passive receptacle for the sick to an active curative institution for all members of society. For economic reasons, the centralization of health care services became essential because of many other important technological innovations appearing on the medical scene. However, hospitals remained institutions to dread, and it was not until the introduction of sulfanilamide in the mid-1930s and penicillin in the early 1940s that the main danger of hospitalization, i.e., cross-infection among patients, was significantly reduced. With these new drugs in their arsenals, surgeons were able to perform their operations without prohibitive morbidity and mortality due to infection. Furthermore, even though the different blood groups and their incompatibility were discovered in 1900 and sodium citrate was used in 1913 to prevent clotting, full development of blood banks was not practical until the 1930s, when technology provided adequate refrigeration. Until that time, “fresh” donors were bled and the blood transfused while it was still warm. Once these surgical suites were established, the employment of specifically designed pieces of medical technology assisted in further advancing the development of complex surgical procedures. For example, the Drinker respirator was introduced in 1927 and the first heart-lung bypass in 1939. By the 1940s, medical procedures heavily dependent on medical technology, such as cardiac catheterization and angiography (the use of a cannula threaded through an arm vein and into the heart with the injection of radiopaque dye for the x-ray visualization of lung and heart vessels and valves), were developed. As a result, accurate diagnosis of congenital and acquired heart disease (mainly valve disorders due to rheumatic fever) became possible, and a new era of cardiac and vascular surgery was established. Following World War II, technological advances were spurred on by efforts to develop superior weapon systems and establish habitats in space and on the ocean floor. As a by-product of these efforts, the development of medical devices accelerated and the medical profession benefited greatly from this rapid surge of “technological finds.” Consider the following examples: 1. Advances in solid-state electronics made it possible to map the subtle behavior of the fundamental unit of the central nervous system — the neuron — as well as to monitor various physiologic parameters, such as the electrocardiogram, of patients in intensive care units. 2. New prosthetic devices became a goal of engineers involved in providing the disabled with tools to improve their quality of life.


3. Nuclear medicine — an outgrowth of the atomic age — emerged as a powerful and effective approach in detecting and treating specific physiologic abnormalities. 4. Diagnostic ultrasound based on sonar technology became so widely accepted that ultrasonic studies are now part of the routine diagnostic workup in many medical specialties. 5. “Spare parts” surgery also became commonplace. Technologists were encouraged to provide cardiac assist devices, such as artificial heart valves and artificial blood vessels, and the artificial heart program was launched to develop a replacement for a defective or diseased human heart. 6. Advances in materials have made the development of disposable medical devices, such as needles and thermometers, as well as implantable drug delivery systems, a reality. 7. Computers similar to those developed to control the flight plans of the Apollo capsule were used to store, process, and cross-check medical records, to monitor patient status in intensive care units, and to provide sophisticated statistical diagnoses of potential diseases correlated with specific sets of patient symptoms. 8. Development of the first computer-based medical instrument, the computerized axial tomography scanner, revolutionized clinical approaches to noninvasive diagnostic imaging procedures, which now include magnetic resonance imaging and positron emission tomography as well. The impact of these discoveries and many others has been profound. The health care system consisting primarily of the “horse and buggy” physician is gone forever, replaced by a technologically sophisticated clinical staff operating primarily in “modern” hospitals designed to accommodate the new medical technology. This evolutionary process continues, with advances in biotechnology and tissue engineering altering the very nature of the health care delivery system itself.

The Field of Biomedical Engineering Today, many of the problems confronting health professionals are of extreme interest to engineers because they involve the design and practical application of medical devices and systems — processes that are fundamental to engineering practice. These medically related design problems can range from very complex large-scale constructs, such as the design and implementation of automated clinical laboratories, multiphasic screening facilities (i.e., centers that permit many clinical tests to be conducted), and hospital information systems, to the creation of relatively small and “simple” devices, such as recording electrodes and biosensors, that may be used to monitor the activity of specific physiologic processes in either a research or clinical setting. They encompass the many complexities of remote monitoring and telemetry, including the requirements of emergency vehicles, operating rooms, and intensive care units. The American health care system, therefore, encompasses many problems that represent challenges to certain members of the engineering profession called biomedical engineers.

Biomedical Engineering: A Definition Although what is included in the field of biomedical engineering is considered by many to be quite clear, there are some disagreements about its definition. For example, consider the terms biomedical engineering, bioengineering, and clinical (or medical) engineering which have been defined in Pacela’s Bioengineering Education Directory [Quest Publishing Co., 1990]. While Pacela defines bioengineering as the broad umbrella term used to describe this entire field, bioengineering is usually defined as a basic research–oriented activity closely related to biotechnology and genetic engineering, i.e., the modification of animal or plant cells, or parts of cells, to improve plants or animals or to develop new microorganisms for beneficial ends. In the food industry, for example, this has meant the improvement of strains of yeast for fermentation. In agriculture, bioengineers may be concerned with the improvement of crop yields by treatment of plants with organisms to reduce frost damage. It is clear that bioengineers of the future


will have a tremendous impact on the quality of human life. The potential of this specialty is difficult to imagine. Consider the following activities of bioengineers: • • • • • • • •

Development of improved species of plants and animals for food production Invention of new medical diagnostic tests for diseases Production of synthetic vaccines from clone cells Bioenvironmental engineering to protect human, animal, and plant life from toxicants and pollutants Study of protein-surface interactions Modeling of the growth kinetics of yeast and hybridoma cells Research in immobilized enzyme technology Development of therapeutic proteins and monoclonal antibodies

In reviewing the above-mentioned terms, however, biomedical engineering appears to have the most comprehensive meaning. Biomedical engineers apply electrical, mechanical, chemical, optical, and other engineering principles to understand, modify, or control biologic (i.e., human and animal) systems, as well as design and manufacture products that can monitor physiologic functions and assist in the diagnosis and treatment of patients. When biomedical engineers work within a hospital or clinic, they are more properly called clinical engineers. Activities of Biomedical Engineers The breadth of activity of biomedical engineers is significant. The field has moved significantly from being concerned primarily with the development of medical devices in the 1950s and 1960s to include a more wide-ranging set of activities. As illustrated below, the field of biomedical engineering now includes many new career areas, each of which is presented in this Handbook. These areas include: • Application of engineering system analysis (physiologic modeling, simulation, and control) to biologic problems • Detection, measurement, and monitoring of physiologic signals (i.e., biosensors and biomedical instrumentation) • Diagnostic interpretation via signal-processing techniques of bioelectric data • Therapeutic and rehabilitation procedures and devices (rehabilitation engineering) • Devices for replacement or augmentation of bodily functions (artificial organs) • Computer analysis of patient-related data and clinical decision making (i.e., medical informatics and artificial intelligence) • Medical imaging, i.e., the graphic display of anatomic detail or physiologic function • The creation of new biologic products (i.e., biotechnology and tissue engineering) Typical pursuits of biomedical engineers, therefore, include: • • • • • • • •

Research in new materials for implanted artificial organs Development of new diagnostic instruments for blood analysis Computer modeling of the function of the human heart Writing software for analysis of medical research data Analysis of medical device hazards for safety and efficacy Development of new diagnostic imaging systems Design of telemetry systems for patient monitoring Design of biomedical sensors for measurement of human physiologic systems variables


• • • • • • • • •

Development of expert systems for diagnosis of disease Design of closed-loop control systems for drug administration Modeling of the physiologic systems of the human body Design of instrumentation for sports medicine Development of new dental materials Design of communication aids for the handicapped Study of pulmonary fluid dynamics Study of the biomechanics of the human body Development of material to be used as replacement for human skin

Biomedical engineering, then, is an interdisciplinary branch of engineering that ranges from theoretical, nonexperimental undertakings to state-of-the-art applications. It can encompass research, development, implementation, and operation. Accordingly, like medical practice itself, it is unlikely that any single person can acquire expertise that encompasses the entire field. Yet, because of the interdisciplinary nature of this activity, there is considerable interplay and overlapping of interest and effort between them. For example, biomedical engineers engaged in the development of biosensors may interact with those interested in prosthetic devices to develop a means to detect and use the same bioelectric signal to power a prosthetic device. Those engaged in automating the clinical chemistry laboratory may collaborate with those developing expert systems to assist clinicians in making decisions based on specific laboratory data. The possibilities are endless. Perhaps a greater potential benefit occurring from the use of biomedical engineering is identification of the problems and needs of our present health care system that can be solved using existing engineering technology and systems methodology. Consequently, the field of biomedical engineering offers hope in the continuing battle to provide high-quality health care at a reasonable cost; if properly directed toward solving problems related to preventive medical approaches, ambulatory care services, and the like, biomedical engineers can provide the tools and techniques to make our health care system more effective and efficient.

Joseph D. Bronzino Editor-in-Chief



Editor-in-Chief

Joseph D. Bronzino, Ph.D., P.E., Vernon Roosa Professor of Applied Science at Trinity College, Hartford, Connecticut, and director of the Biomedical Engineering Alliance for Connecticut (BEACON), teaches graduate and undergraduate courses in biomedical engineering in the fields of clinical engineering, electrophysiology, signal analysis, and computer applications in medicine. He earned his B.S. in electrical engineering from Worcester Polytechnic Institute, M.S. in electrical engineering from the Naval Postgraduate School, and Ph.D. in electrical engineering also from Worcester Polytechnic Institute. Deeply concerned with the discipline of biomedical engineering, as well as with ethical and economic issues related to the application of technology to the delivery of health care, Dr. Bronzino has written and lectured internationally. He is the author of over 200 articles and 8 books: Technology for Patient Care (C.V. Mosby, 1977), Computer Applications for Patient Care (Addison-Wesley, 1982), Biomedical Engineering: Basic Concepts and Instrumentation (PWS Publishing Co., 1986), Expert Systems: Basic Concepts and Applications (Research Foundation of the State University of New York, 1989), Medical Technology and Society: An Interdisciplinary Perspective (MIT Press, 1990), Management of Medical Technology (Butterworth/Heinemann, 1992), The Biomedical Engineering Handbook (CRC Press, 1995), The Introduction to Biomedical Engineering (Academic Press, 1999), and The Biomedical Engineering Handbook, 2nd Edition (CRC Press, 2000). Dr. Bronzino is a fellow of both the Institute of Electrical and Electronic Engineers (IEEE) and the American Institute of Medical and Biological Engineering (AIMBE), a past president of the IEEE Engineering in Medicine and Biology Society (EMBS), a past chairman of the Biomedical Engineering Division of the American Society for Engineering Education, and a charter member of the American College of Clinical Engineering (ACCE) and the Connecticut Academy of Science and Engineering (CASE). Dr. Bronzino has extensive experience in the formulation of public policy regarding the utilization and regulation of medical technology. He has served as a past chairman of both the IEEE Health Care Engineering Policy Committee (HCEPC) and the IEEE Technical Policy Council in Washington, D.C.


Advisory Board

Jean Louis Coatrieux

Banu Onaral

Université de Rennes I Rennes, France

Drexel University Philadelphia, Pennsylvania

Dov Jaron

Robert Plonsey

Drexel University Philadelphia, Pennsylvania

Duke University Durham, North Carolina

Swamy Laxminarayan

Alfred R. Potvin

New Jersey Institute of Technology Newark, New Jersey

MEECO Sarasota, Florida

Karen M. Mudry

Charles J. Robinson

Formerly of The Whitaker Foundation Washington, D.C.

Louisiana Tech University Ruston, Louisiana

Michael R. Neuman

Daniel J. Schneck

Joint Program in Biomedical Engineering The University of Memphis and University of Tennessee Memphis, Tennessee

Virginia Polytechnic Institute and State University Blacksburg, Virginia

John Webster University of Wisconsin Madison, Wisconsin


Contributors

Joseph Adam

G. Faye Boudreaux-Bartels

A. Enis Çetin

Premise Development Corporation Hartford, Connecticut

University of Rhode Island Kingston, Rhode Island

Bilkent University Ankara, Turkey

Kai-Nan An

Joseph D. Bronzino

K. B. Chandran

Biomechanics Laboratory Mayo Clinic Rochester, Minnesota

Trinity College/The Biomedical Engineering Alliance for Connecticut (BEACON) Hartford, Connecticut

University of Iowa Iowa City, Iowa

Gary J. Baker Stanford University Stanford, California

D. C. Barber University of Sheffield Sheffield, United Kingdom

Berj L. Bardakjian University of Toronto Toronto, Canada

Roger C. Barr Duke University Durham, North Carolina

Khosrow Behbehani The University of Texas at Arlington and The University of Texas, Southwestern Medical Center at Dallas Arlington, Texas

Edward J. Berbari Indiana University/Purdue University at Indianapolis Indianapolis, Indiana

Wei Chen

University of North Carolina Chapel Hill, North Carolina

Center for Magnetic Resonance Research and the University of Minnesota Medical School Minneapolis, Minnesota

Thomas F. Budinger

David A. Chesler

University of California Berkeley, California

K. J. L. Burg

Massachusetts General Hospital and Harvard University Medical School Boston, Massachusetts

Carolinas Medical Center Charlotte, North Carolina

C. K. Chou

Thomas J. Burkholder

Motorola, Inc. Plantation, Florida

The Georgia Institute of Technology Atlanta, Georgia

Chih-Chang Chu

Robert D. Butterfield

Cornell University Ithaca, New York

IVAC Corporation San Diego, California

Ben M. Clopton

Joseph P. Cammarota

University of Washington Seattle, Washington

Richard P. Buck

Naval Air Warfare Center, Aircraft Division Warminster, Pennsylvania

Robin Coger

Anna M. Bianchi

Argonne National Laboratory Argonne, Illinois

Massachusetts General Hospital, Harvard University Medical School, and the Shriners Burns Institutes Cambridge, Massachusetts

St. Raffaele Hospital Milan, Italy

Ewart R. Carson

Arnon Cohen

W. G. Billotte

City University London, United Kingdom

Ben-Gurion University Beer Sheva, Israel

University of Dayton Dayton, Ohio

Sergio Cerutti

Steven Conolly

Bruce R. Bowman

Polytechnic University Milan, Italy

Stanford University Stanford, California

EdenTec Corporation Eden Prairie, Minnesota


Thomas R. Canfield

Derek G. Cramp

Edward Elson

Craig S. Henriquez


Walter Reed Army Institute of Research Washington, D.C.


Cristina Cristalli Instrumentation Laboratory Milano, Italy

Barbara Y. Croft National Institutes of Health Kensington, Maryland

Ian A. Cunningham Victoria Hospital, the John P. Robarts Research Institute, and the University of Western Ontario London, Canada

Roy B. Davis Motion Analysis Laboratory Shriners Hospitals for Children Greenville, South Carolina

K. Whittaker Ferrara Riverside Research Institute New York, New York

Ross Flewelling Nellcor Incorporation Pleasant, California

Michael Forde Medtronic, Inc. Minneapolis, Minnesota

Kenneth R. Foster University of Pennsylvania Philadelphia, Pennsylvania

Michael J. Furey

Robert M. Hochmuth Duke University Durham, North Carolina

Xiaoping Hu Center for Magnetic Resonance Research and the University of Minnesota Medical School Minneapolis, Minnesota

Bernard F. Hurley University of Maryland College Park, Maryland

Arthur T. Johnson University of Maryland College Park, Maryland


Christopher R. Johnson

Leslie A. Geddes

G. Allan Johnson

Purdue University West Lafayette, Indiana

Duke University Medical Center Durham, North Carolina

Hines VA Hospital and Loyola University Medical Center Hines, Illinois

V. K. Goel

Millard M. Judy

University of Iowa Iowa City, Iowa

Baylor Research Institute Dallas, Texas

Cathryn R. Dooly

Richard L. Goldberg

Philip F. Judy

University of Maryland College Park, Maryland

University of North Carolina Chapel Hill, North Carolina

Brigham and Women’s Hospital and Harvard Medical School Boston, Massachusetts

Gary Drzewiecki

Boris Gramatikov

Rutgers State University Piscataway, New Jersey

Johns Hopkins School of Medicine Baltimore, Maryland

Edwin G. Duffin

Wallace Grant

Medtronic, Inc. Minneapolis, Minnesota


Peter A. DeLuca Gait Analysis Laboratory Connecticut Children’s Medical Center Hartford, Connecticut

Philip B. Dobrin

Dominique M. Durand Case Western Reserve University Cleveland, Ohio

Jeffrey L. Eggleston Valleylab, Inc. Boulder, Colorado

Walter Greenleaf Greenleaf Medical Palo Alto, California

Alan R. Hargen

University of Utah Salt Lake City, Utah

J. Lawrence Katz Case Western Reserve University Cleveland, Ohio

Kenton R. Kaufman Biomechanics Laboratory, Mayo Clinic Rochester, Minnesota

J. C. Keller University of Iowa Iowa City, Iowa

Gilson Khang

Jeffrey T. Ellis

University of California, San Diego and NASA Ames Research Center San Diego, California

Georgia Institute of Technology Atlanta, Georgia

Kaj-Åge Henneberg

Albert I. King

University of Montreal Montreal, Canada

Wayne State University Detroit, Michigan


Chonbuk National University Chonju, Korea

Young Kon Kim

Albert Macovski

Sylvia Ounpuu

Inje University Kyungnam, Korea


Gait Analysis Laboratory Connecticut Children’s Medical Center Hartford, Connecticut

Hayrettin Köymen Bilkent University Ankara, Turkey

Kenneth K. Kwong

Luca T. Mainardi Polytechnic University Milan, Italy

Joon B. Park University of Iowa Iowa City, Iowa

Massachusetts General Hospital and Harvard University Medical School Boston, Massachusetts

Jaakko Malmivuo Ragnar Granit Institute, Tampere University of Technology Tampere, Finland

S.-H. Park

Roderic Lakes

Angelo Manzoni

University of Wisconsin-Madison Madison, Wisconsin

Instrumentation Laboratory Milano, Italy

Maqbool Patel

Christopher G. Lausted

Andrew D. McCulloch

University of Maryland College Park, Maryland

University of California San Diego, California

Hai Bang Lee

Yitzhak Mendelson

Korea Research Institute of Chemical Technology Chungnam, Korea

Worcester Polytechnic Institute Worcester, Massachusetts

Jin Ho Lee

Evangelia MicheliTzanakou

Hannam University Taejon, Korea

Rutgers University Piscataway, New Jersey

Jack D. Lemmon

Jack G. Mottley


University of Rochester Rochester, New York

Shu-Tung Li

Karen M. Mudry

Collagen Matrix, Inc.

Formerly of The Whitaker Foundation Washington, D.C.

Baruch B. Lieber State University of New York at Buffalo Buffalo, New York

Richard L. Lieber

Robin Murray University of Rhode Island Kingston, Rhode Island

University of California and Veterans Administration Medical Centers San Diego, California

Joachim H. Nagel

Chung-Chiun Liu

Michael R. Neuman

Electronics Design Center and Edison sensor Technology Center Case Western Reserve University Cleveland, Ohio

Joint Program in Biomedical Engineering The University of Memphis and University of Tennessee Memphis, Tennessee

A. Llinás Pontificia Universidad Javeriana Bogota, Colombia


University of Stuttgart Stuttgart, Germany

Banu Onaral Drexel University Philadelphia, Pennsylvania

University of Southern California Los Angeles, California


Robert Patterson University of Minnesota Minneapolis, Minnesota

A. William Paulsen Emory University Atlanta, Georgia

John Pauly Stanford University Stanford, California

P. Hunter Peckham Case Western Reserve University and Veterans Affairs Medical Center Cleveland, Ohio

Athina P. Petropulu Drexel University Philadelphia, Pennsylvania

Tom Piantanida Greenleaf Medical Palo Alto, California

Roland N. Pittman Virginia Commonwealth University Richmond, Virginia

Robert Plonsey Duke University Durham, North Carolina

Charles Polk University of Rhode Island Kingston, Rhode Island

Aleksander S. Popel

Artin A. Shoukas

Kamil Ugurbil

Johns Hopkins University Baltimore, Maryland

Johns Hopkins University Baltimore, Maryland


Rulong Ren City of Hope National Medical Center Duarte, California

Robert E. Shroy, Jr.

Pat Ridgely

Stephen W. Smith

Medtronic, Inc. Minneapolis, Minnesota


Richard L. Roa

Francis A. Spelman

Baylor University Medical Center Dallas, Texas


Eric Rosow

Charles R. Steele

Hartford Hospital and Premise Development Corporation Hartford, Connecticut


Bradley J. Roth Oakland University Rochester, Michigan

Carl F. Rothe Indiana University Indianapolis, Indiana

John Schenck General Electric Corporate Research and Development Center Schenectady, New York

Geert W. SchmidSchönbein

Picker International Highland Heights, Ohio

George Stetten Duke University Durham, North Carolina

Primoz Strojnik Case Western Reserve University Cleveland, Ohio

Maria A. Stuchly University of Victoria Victoria, Canada

Willis A. Tacker Purdue University West Lafayette, Indiana

Henry F. VanBrocklin University of California Berkeley, California

Michael S. Van Lysel University of Wisconsin Madison, Wisconsin

Anthony Varghese University of Minnesota Minneapolis, Minnesota

David C. Viano Wayne State University Detroit, Michigan

Wolf W. von Maltzahn Whitaker Foundation Washington, D.C.

Gregory I. Voss IVAC Corporation San Diego, California

Richard E. Waugh University of Rochester Rochester, New York

James C. Weaver

Nitish V. Thakor

Massachusetts Institute of Technology Cambridge, Massachusetts

Johns Hopkins School of Medicine Baltimore, Maryland

Martin J. Yaffe


Jason A. Tolomeo

University of Toronto Toronto, Canada


Ajit P. Yoganathan

Edward Schuck

Mehmet Toner


EdenTec Corporation Eden Prairie, Minnesota

Massachusetts General Hospital, Harvard University Medical School, and the Shriners Burns Institutes Cambridge, Massachusetts

University of California San Diego, California

Daniel J. Schneck

S. W. Shalaby Poly-Med, Inc. Anderson, South Carolina

David Sherman Johns Hopkins School of Medicine Baltimore, Maryland


Benjamin M.W. Tsui University of North Carolina Chapel Hill, North Carolina

Deborah E. Zetes-Tolomeo Stanford University Stanford, California

Xiaohong Zhou Duke University Medical Center Durham, North Carolina

Contributors

James J. Abbas

Joseph D. Bronzino

Yadin David

University of Kentucky Lexington, Kentucky

Texas Children’s Hospital Houston, Texas

Joseph Adam

Trinity College/Biomedical Engineering Alliance for Connecticut (BEACON) Hartford, Connecticut

Premise Development Corporation Hartford, Connecticut

Susan V. Brooks

Vanderbilt University Nashville, Tennessee

Patrick Aebischer

University of Michigan Ann Arbor, Michigan

Paul D. Drumheller

Lausanne University Medical School Lausanne, Switzerland

Mark E. Bruley

Gore Hybrid Technologies, Inc. Flagstaff, Arizona

Robert C. Allen

ECRI Plymouth Meeting, Pennsylvania

G. M. Drzewiecki

Emory University Atlanta, Georgia

Ewart R. Carson


John G. Aunins


Karen A. Duca

Merck Research Laboratories Rahway, New Jersey

Andrea Caumo

University of Wisconsin Madison, Wisconsin

Dennis D. Autio

San Raffaele Scientific Institute Milan, Italy

Graham A. Dunn

Dybonics, Inc. Portland, Oregon

Joseph A. Chinn

King’s College London, United Kingdom

James W. Baish

Sulzer Carbomedics Austin, Texas

John Denis Enderle

Bucknell University Lewisburg, Pennsylvania

Vivian H. Coates

University of Connecticut Storrs, Connecticut

Pamela J. Hoyes Beehler

ECRI Plymouth Meeting, Pennsylvania

John A. Faulkner

University of Texas at Arlington Arlington, Texas

Claudio Cobelli


Ravi Bellamkonda

University of Padova Padua, Italy

Stanley M. Finkelstein

Lausanne University Medical School Lausanne, Switzerland

Clark K. Colton

University of Minnesota Minneapolis, Minnesota

Jan E. W. Beneken Eindhoven University of Technology Eindhoven, the Netherlands

François Berthiaume Surgical Services, Massachusetts General Hospital, Harvard Medical School, and the Shriners Hospital for Children Cambridge, Massachusetts

Jeffrey S. Blair IBM Health Care Solutions Atlanta, Georgia © 2000 by CRC Press LLC


Rory A. Cooper

Benoit M. Dawant

Robert J. Fisher University of Connecticut Storrs, Connecticut

University of Pittsburgh, VA Pittsburgh Health Care System Pittsburgh, Pennsylvania

J. Michael Fitzmaurice

Derek G. Cramp

L.E. Freed


Harvard University Cambridge, Massachusetts

Department of Health and Human Services Rockville, Maryland

Catherine Garbay

Craig T. Jordan

Joseph M. Le Doux

Laboratoire TIMC/IMAG Grenoble, France

Somatix Therapy Corp. Alameda, California

Leslie A. Geddes

Thomas M. Judd

Purdue University West Lafayette, Indiana

Kaiser Permanente Atlanta, Georgia

Center for Engineering in Medicine, and Surgical Services, Massachusetts General Hospital, Harvard Medical School, and the Shriners Burns Hospital Cambridge, Massachusetts

John W. Goethe

Kurt A. Kaczmarek

Ann L. Lee

The Institute of Living Hartford, Connecticut

University of Wisconsin at Madison Madison, Wisconsin


Michael L. Gullikson

Robert Kaiser

John K-J. Li

Texas Children’s Hospital Houston, Texas



Katya Hill

Peter L.M. Kerkhof

E.N. Lightfoot

Edinboro University of Pennsylvania and University of Pittsburgh Edinboro, Pennsylvania

Medwise Working Group Maarssen, the Netherlands


Tao Ho Kim

Michael W. Long

University of Pittsburgh Pittsburgh, Pennsylvania

Harvard University and Boston Children’s Hospital Boston, Massachusetts


Jeffrey A. Hubbell

Manfred R. Koller

California Institute of Technology Pasadena, California

Oncosis San Diego, California

H. David Humes

George V. Kondraske


University of Texas at Arlington, Human Performance Institute Arlington, Texas

Douglas Hobson

Sheik N. Imrhan

Marilyn Lord King’s College London, United Kingdom

Michael J. Lysaght Brown University Providence, Rhode Island

Vasilis Z. Marmarelis

David N. Ku

University of Southern California Los Angeles, California


Kenneth J. Maxwell

Casimir Kulikowski

Departure Technology Fort Worth, Texas


Joseph P. McClain

Université de Technologie de Compiègne Compiègne, France

Luis G. Kun

Walter Reed Army Medical Center Washington, D.C.

CIMIC/Rutgers University New Brunswick, New Jersey

Larry V. McIntire

Hugo O. Jauregui

Robert S. Langer

Rice University Houston, Texas

Rhode Island Hospital Providence, Rhode Island


Evangelia MicheliTzanakou

Douglas A. Lauffenburger



Marcos Intaglietta University of California La Jolla, California

Michel Jaffrin

Stephen B. Johnson Columbia University New York, New York

Richard D. Jones Christchurch Hospital and University of Otago Christchurch, New Zealand



Swamy Laxminarayan New Jersey Institute of Technology Newark, New Jersey

David J. Mooney Massachusetts Institute of Technology Cambridge, Massachusetts

John Moran

Dejan B. Popovic´

Rangarajan Sampath

Vasca, Inc. Tewksbury, Massachusetts

University of Belgrade Belgrade, Yugoslavia


Jeffrey R. Morgan

T. Allan Pryor

Niilo Saranummi

Center for Engineering in Medicine, and Surgical Services, Massachusetts General Hospital, Harvard Medical School, and the Shriners Burns Hospital Cambridge, Massachusetts

University of Utah Salt Lake City, Utah

VIT Information Technology Tampere, Finland

Yuchen Qiu

Soumitra Sengupta

The Pennsylvania State University University Park, Pennsylvania

Columbia University New York, New York

Robert L. Morris

Gerard Reach

Yasuhiko Shimizu

Dybonics, Inc. Portland, Oregon

Hôpital Hotel-Dieu Paris Paris, France

Kyoto University Kyoto, Japan

Tatsuo Nakamura

Charles J. Robinson

Michael L. Shuler

Kyoto University Kyoto, Japan

Louisiana Tech University Ruston, Louisiana The Overton Brooks VA Medical Center Shreveport, Louisiana


Brian A. Naughton Advanced Tissue Sciences, Inc. La Jolla, California

Robert M. Nerem Georgia Institute of Technology Atlanta, Georgia

A. Noordergraaf University of Pennsylvania Philadelphia, Pennsylvania

Patrick R. Norris Vanderbilt University Nashville, Tennessee

Pirkko Nykänen

Barry Romich Prentke Romich Company Wooster, Ohio

Gerson Rosenberg Pennsylvania State University Hershey, Pennsylvania

Eric Rosow Hartford Hospital Hartford, Connecticut

Charles M. Roth

Srolamtj Simdara Rutgers University Piscataway, New Jersey

Steven M. Slack The University of Memphis Memphis, Tennessee

Susan S. Smith Texas Woman’s University Dallas, Texas

Giovanni Sparacino University of Padova Padua, Italy

Ron Summers

Joseph L. Palladino

Center for Engineering in Medicine, Massachusetts General Hospital, Harvard Medical School, and the Shriners Burns Hospital Cambridge, Massachusetts

Trinity College Hartford, Connecticut

Alan J. Russell


Bernhard Ø. Palsson


Karl Syndulko


Maria Pia Saccomani

UCLA School of Medicine Los Angeles, California

Mohamad Parnianpour


John M. Tarbell

Ohio State University Columbus, Ohio

Pamela S. Saha

The Pennsylvania State University University Park, Pennsylvania

Charles W. Patrick, Jr.

Clemson University Clemson, South Carolina

William D. Timmons


Subrata Saha

University of Akron Akron, Ohio

Chi-Sang Poon


Gianna Maria Toffolo

W. Mark Saltzman


VIT Information Technology Tampere, Finland

Massachusetts Institute of Technology Cambridge, Massachusetts © 2000 by CRC Press LLC


Loughborough University Leicestershire, United Kingdom

Srikanth Sundaram

Mehmet Toner

Paul J. Vasta

Ioannis V. Yannas

Massachusetts General Hospital, Harvard Medical School, and the Shriners Burns Hospital Cambridge, Massachusetts



Chenzhao Vierheller

Elaine Trefler



David B. Volkin

Alan Turner-Smith


King’s College London, United Kingdom

G. Vunjak-Novakovic

Shiro Usai Toyohashi University of Technology Toyohashi, Japan

Joseph P. Vacanti Harvard University and Boston Children’s Hospital Boston, Massachusetts

Robert F. Valentini Brown University Providence, Rhode Island

Gary Van Zant


Martin L. Yarmush Massachusetts General Hospital, Harvard Medical School, and the Shriners Burns Hospital Cambridge, Massachusetts

Ajit P. Yoganathan Georgia Institute of Technology Atlanta, Georgia

Columbia University New York, New York

Daniel A. Zahner

S. Patrick Walton


Center for Engineering in Medicine, Massachusetts General Hospital, Harvard Medical School, and the Shriners Burns Hospital Cambridge, Massachusetts

Robert M. Winslow

Gregg Vanderheiden

Daniel E. Wueste



Alvin Wald

University of Kentucky Medical Center Lexington, Kentucky


David M. Yarmush

SANGART, Inc. La Jolla, California


Craig Zupke Massachusetts General Hospital and the Shriners Burns Institute Cambridge, Massachusetts

Andrew L. Zydney University of Delaware Newark, Delaware

Contents

SECTION I Introduction

Physiologic Systems Robert Plonsey

1

An Outline of Cardiovascular Structure and Function

2

Endocrine System

3

Nervous System

4

Vision System

5

Auditory System

6

The Gastrointestinal System

7

Respiratory System

Daniel J. Schneck

Derek G. Cramp, Ewart R. Carson Evangelia Micheli-Tzanakou

George Stetten Ben M. Clopton, Francis A. Spelman Berj L. Bardakjian

Arthur T. Johnson, Christopher G. Lausted,

Joseph D. Bronzino

Historical Perspectives 1: Cardiac Pacing — Historical Highlights Leslie A. Geddes

SECTION II Introduction

Bioelectric Phenomena

Craig S. Henriquez

8

Basic Electrophysiology

9

Volume Conductor Theory

Roger C. Barr Robert Plonsey

10

The Electrical Conductivity of Tissues

11

Membrane Models

12

Numerical Methods for Bioelectric Field Problems

13

Principles of Electrocardiography

14

Principles of Electromyography


Bradley J. Roth

Anthony Varghese

Edward J. Berbari Kaj-Åge Henneberg

Christopher R. Johnson

15

Principles of Electroencephalography

16

Biomagnetism

17

Electric Stimulation of Excitable Tissue

Section III Introduction

Jaakko Malmivuo Dominique M. Durand

Biomechanics Daniel J. Schneck

18

Mechanics of Hard Tissue

19

Mechanics of Blood Vessels

20

Joint-Articulating Surface Motion

21

Joint Lubrication

22

Joseph D. Bronzino

J. Lawrence Katz Thomas R. Canfield, Philip B. Dobrin Kenton R. Kaufman, Kai-Nan An

Michael J. Furey

Musculoskeletal Soft Tissue Mechanics

Richard L. Lieber,

Thomas J. Burkholder

23

Mechanics of Head/Neck

24

Biomechanics of Chest and Abdomen Impact

Albert I. King, David C. Viano David C. Viano,

Albert I. King,

25

Analysis of Gait

26

Exercise Physiology

27

Factors Affecting Mechanical Work in Humans

Roy B. Davis, Peter A. DeLuca, Sylvia Õunpuu Arthur T. Johnson, Cathryn R. Dooly Arthur T. Johnson,

Bernard F. Hurley

28 29

Cardiac Biodynamics

Andrew D. McCulloch

Heart Valve Dynamics

Aijt P. Yoganathan, Jack D. Lemmon,

Jeffrey T. Ellis

30

Arterial Macrocirculatory Hemodynamics

31

Mechanics and Transport in the Microcirculation

Baruch B. Lieber Aleksander S. Popel,

Roland N. Pittman

32

Mechanics and Deformability of Hematocytes Robert M. Hochmuth

33

The Venous System


Artin A. Shoukas, Carl F. Rothe

Richard E. Waugh,

34

Mechanics of Tissue/Lymphatic Transport

Alan R. Hargen,

Geert W. Schmid-Schönbein

35

Cochlear Mechanics

Charles R. Steele, Gary J. Baker, Jason A. Tolomeo,

Deborah E. Zetes-Tolomeo

36

Vestibular Mechanics

Wallace Grant

Section IV Biomaterials Introduction

Joon B. Park

37

Metallic Biomaterials

Joon B. Park, Young Kon Kim

38

Ceramic Biomaterials

W. G. Billotte

39

Polymeric Biomaterials

40

Composite Biomaterials

41

Biodegradable Polymeric Biomaterials: An Updated Overview

Hai Bang Lee, Gilson Khang, Jin Ho Lee Roderic Lakes

Chih-Chang Chu

42

Biologic Biomaterials: Tissue-Derived Biomaterials (Collagen) Shu-Tung Li

43

Soft Tissue Replacements 43.1 Blood Interfacing Implants K. B. Chandran 43.2 Non-Blood-Interfacing Implants for Soft Tissues K.J.L Burg, S. W. Shalaby

44

Hard Tissue Replacements 44.1 Bone Repair and Joint Implants S-H. Park, A. Llinás, V. K. Goel 44.2 Dental Implants: The Relationship of Materials Characteristics to Biologic Properties J. C. Keller

45

Preservation Techniques for Biomaterials

46

Hip Joint Prosthesis Fixation—Problems and Possible Solutions Joon B. Park

Section V Biomedical Sensors Introduction

Michael R. Neuman

47

Physical Measurements

Michael R. Neuman

48

Biopotential Electrodes

Michael R. Neuman


Robin Coger, Mehmet Toner

49

Electrochemical Sensors

50

Optical Sensors

51

Bioanalytic Sensors

Chung-Chiun Liu

Yitzhak Mendelson Richard P. Buck

Historical Perspectives 2: The Elecrocardiograph

Section VI Introduction

Leslie A. Geddes

Biomedical Signal Analysis Banu Onaral

52

Biomedical Signals: Origin and Dynamic Characteristics; Frequency-Domain Analysis Arnon Cohen

53

Digital Biomedical Signal Acquisition and Processing

Luca T. Mainardi,

Anna M. Bianchi, Sergio Cerutti

54

Compression of Digital Biomedical Signals

A. Enis Cetin, ¸

Hayrettin Köymen

55

Time-Frequency Signal Representations for Biomedical Signals G. Faye Boudreaux-Bartels, Robin Murray

56

Wavelet (Time-Scale) Analysis in Biomedical Signal Processing Nitish V. Thakor, Boris Gramatikov, David Sherman

57

Higher-Order Spectral Analysis

58

Neural Networks in Biomedical Signal Processing

Athina P. Petropulu

Evangelia Micheli-Tzanakou

59

Complexity, Scaling, and Fractals in Biomedical Signals

Banu Onaral,

Joseph P. Cammarota

60

Future Directions: Biomedical Signal Processing and Networked Multimedia Communications Banu Onaral

Section VII Imaging Introduction

61

Karen M. Mudry

X-Ray 61.1 X-Ray Equipment Robert E. Shroy, Jr. 61.2 X-Ray Projection Angiography Michael S. Van Lysel 61.3 Mammography Martin J. Yaffe


62

Computed Tomography 62.1 Instrumentation Ian A. Cunningham 62.2 Reconstruction Principles Philip F. Judy

63

Magnetic Resonance Imaging 63.1 63.2 63.3 63.4

Acquisition and Processing Steven Conolly, Albert Macovski, John Pauly Hardware/Instrumentation John Schenck Functional MRI Kenneth K. Kwong, David A. Chesler Chemical-Shift Imaging: An Introduction to Its Theory and Practice Xiaoping Hu, Wei Chen,

Maqbool Patel, Kamil Ugurbil

64

Nuclear Medicine 64.1 Instrumentation Barbara Y. Croft 64.2 SPECT (Single-Photon Emission Computed Tomography)

65

Benjamin M. W. Tsui

Ultrasound 65.1 Transducers Richard L. Goldberg, Stephen W. Smith 65.2 Ultrasonic Imaging Jack G. Mottley 65.3 Blood Flow Measurement Using Ultrasound K. Whittaker Ferrara

66

Magnetic Resonance Microscopy

67

Positron-Emission Tomography (PET)

Xiaohong Zhou, G. Allan Johnson

67.1 Radiopharmaceuticals Thomas E. Budinger, Henry F. Van Brocklin 67.2 Instrumentation Thomas E. Budinger

68

Electrical Impedance Tomography

69

Medical Applications of Virtual Reality Technology

D. C. Barber Walter Greenleaf,

Tom Piantanida

Section VIII Medical Instruments and Devices Introduction

Wolf W. von Maltzahn

70

Biopotential Amplifiers

71

Noninvasive Arterial Blood Pressure and Mechanics

72

Cardiac Output Measurement

73

Bioelectric Impedance Measurements

74

Respiration

75

Clinical Laboratory: Separation and Spectral Methods


Joachim H. Nagel Gary Drzewiecki

Leslie A. Geddes Robert Patterson

Leslie A. Geddes Richard L. Roa

76

Clinical Laboratory: Nonspectral Methods and Automation Richard L. Roa

77

Implantable Cardiac Pacemakers

78

Implantable Stimulators for Neuromuscular Control

Michael Forde, Pat Ridgely Primoz Strojnik,

P. Hunter Peckham,

79

External Defibrillators

80

Implantable Defibrillators

81

Electrosurgical Devices

Jeffrey L. Eggleston , Wolf W. von Maltzahn

82

Mechanical Ventilation

Khosrow Behbehani

83

Parenteral Infusion Devices

84

Essentials of Anesthesia Delivery

85

Biomedical Lasers

86

Noninvasive Optical Monitoring

87

Medical Instruments and Devices Used in the Home

Willis A. Tacker Edwin G. Duffin

Gregory I. Voss, Robert D. Butterfield A. William Paulsen

Millard M. Judy Ross Flewelling Bruce R. Bowman,

Edward Schuck

88

Virtual Instrumentation

Eric Rosow, Joseph Adam

Historical Perspectives 3: Recording of Action Potentials

Leslie A. Geddes

Section IX Biologic Effects of Nonionizing Electromagnetic Fields Introduction

Charles Polk

89

Dielectric Properties of Tissues

90

Low-Frequency Magnetic Fields: Dosimetry, Cellular, and Animal Effects

Kenneth R. Foster

Maria A. Stuchly

91

Therapeutic Applications of Low-Frequency Sinusoidal and Pulsed Electric and Magnetic Fields Charles Polk

92

Biologic Effects of Radiofrequency and Microwave Fields: In Vivo and In Vitro Experimental Results Edward Elson


93

Radio Frequency Hyperthermia in Cancer Therapy

C. K. Chou,

Rulong Ren

94

Electroporation of Cells and Tissues

Appendix A

Basics of Blood Gas Instrumentation

Angelo Manzoni


James C. Weaver Christina Cristalli,

Contents

SECTION X Transport Phenomena and Biomimetic Systems Introduction

Robert J. Fisher

95

Biomimetic Systems

96

Diffusional Processes and Engineering Design

97

Animal Surrogate Systems

98

Microvascular Heat Transfer

99

Interstitial Transport in the Brain: Principles for Local Drug Delivery

Robert J. Fisher E. N. Lightfoot

Michael L. Shuler James W. Baish .

W. Mark Saltzman

100

Arterial Wall Mass Transport: The Possible Role of Blood Phase Resistance in the Localization of Arterial Disease John M. Tarbell, Yuchen Qiu

SECTION XI Biotechnology Introduction

Martin L. Yarmush, Mehmet Toner

101

Protein Engineering

102

Monoclonal Antibodies and Their Engineered Fragments

Alan J. Russell, Chenzhao Vierheller

Srikanth Sundaram, David M. Yarmush

103

Antisense Technology

S. Patrick Walton, Charles M. Roth,

Martin L. Yarmush

104

Tools for Genome Analysis

105

Vaccine Production

106

Gene Therapy

107

Cell Engineering

108

Metabolic Engineering


Robert Kaiser

John G. Aunins, Ann L. Lee, David B. Volkin

Joseph M. Le Doux, Jeffrey R. Morgan, Martin L. Yarmush Douglas A. Lauffenburger Craig Zupke

SECTION XII Introduction

Tissue Engineering

Bernhard Ø. Palsson, Jeffrey A. Hubbell

109

Tissue Engineering

110

Surface Immobilization of Adhesion Ligands for Investigations of Cell-Substrate Interactions Paul D. Drumheller, Jeffrey A. Hubbell

111

Biomaterials: Protein-Surface Interactions

François Berthiaume, Martin L. Yarmush

Joseph A. Chinn, Steven M. Slack

112

Engineering Biomaterials for Tissue Engineering: The 10–100 Micron Size Scale David J. Mooney, Robert S. Langer

113

Regeneration Templates

114

Ioannis V. Yannas

Fluid Shear Stress Effects on Cellular Function

Charles W. Patrick, Jr.,

Rangarajan Sampath, Larry V. McIntire

115

The Roles of Mass Transfer in Tissue Function

Edwin N. Lightfoot,

Karen A. Duca

116

The Biology of Stem Cells

117

Cell Motility and Tissue Architecture

118

Tissue Microenvironments

119

The Importance of Stromal Cells

120

Tissue Engineering of Bone Marrow

Craig T. Jordan, Gary Van Zant Graham A. Dunn

Michael W. Long Brian A. Naughton Manfred R. Koller,

Bernhard Ø. Palsson

121

Tissue Engineering of the Liver

122

Tissue Engineering in the Nervous System

Tae Ho Kim, Joseph P. Vacanti Ravi Bellamkonda,

Patrick Aebischer

123

Tissue Engineering of Skeletal Muscle

Susan V. Brooks,

John A. Faulkner

124

Tissue Engineering of Cartilage

125

Tissue Engineering of the Kidney


L.E. Freed, G. Vunjak-Novakovic H. David Humes

SECTION XIII Prostheses and Artificial Organs Introduction

Pierre M. Galletti, Robert M. Nerem

126

Artificial Heart and Circulatory Assist Devices

127

Cardiac Valve Prostheses

128

Vascular Grafts

129

Gerson Rosenberg

Ajit P. Yoganathan

David N. Ku, Robert C. Allen

Artificial Lungs and Blood-Gas Exchange Devices

Pierre M. Galletti,

Clark K. Colton

130

Artificial Kidney

131

Peritoneal Dialysis Equipment

132

Therapeutic Apheresis and Blood Fractionation

133

Liver Support Systems

134

Pierre M. Galletti, Clark K. Colton, Michael J. Lysagh

Artificial Pancreas

Michael J. Lysaght, John Moran Andrew L. Zydney

Pierre M. Galletti, Hugo O. Jauregui

Pierre M. Galletti, Clark K. Colton, Michel Jaffrin,

Gerard Reach .

135

Nerve Guidance Channels

136

Tracheal, Laryngeal, and Esophageal Replacement Devices

Robert F. Valentini

Tatsuo Nakamura, Yasuhiko Shimizu

137

Artificial Blood

138

Artificial Skin and Dermal Equivalents

Marcos Intaglietta, Robert M. Winslow Ioannis V. Yannas

SECTION XIV Rehabiliation Engineering Introduction

139

Charles J. Robinson

Rehabilitation Engineering, Science, and Technology Charles J. Robinson

140

Orthopedic Prosthetics and Orthotics in Rehabilitation

Marilyn Lord,

Alan Turner-Smith

141

Wheeled Mobility: Wheelchairs and Personal Transportation Rory A. Cooper


142

Externally Powered and Controlled Orthotics and Prosthetics Dejan B. Popović

143 144

Sensory Augmentation and Substitution

Kurt A. Kaczmarek

Augmentative and Alternative Communication

Barry Romich,

Gregg Vanderheiden, Katya Hill

145

Measurement Tools and Processes in Rehabilitation Engineering George V. Kondraske

146

Rehabilitation Engineering Technologies: Principles of Application Douglas Hobson, Elaine Trefler

Historical Perspectives 4: Electromyography

Leslie A. Geddes

SECTION XV Human Performance Engineering Introduction

147

George V. Kondraske

A Working Model for Human System-Task Interfaces George V. Kondraske

148

Measurement of Neuromuscular Performance Capacities Susan S. Smith

149

Measurement of Sensory-Motor Control Performance Capacities: Tracking Tasks Richard D. Jones

150

Measurement of Information-Processing Performance Capacities George V. Kondraske, Paul J. Vasta

151

High-Level Task Analysis: Mental Components

152

Task Analysis and Decomposition: Physical Components

Kenneth J. Maxwell

Sheik N. Imrhan

153

Human-Computer Interface Design Issues

154

Applications of Human Performance Measurements to Clinical Trials to Determine Therapy Effectiveness and Safety Pamela J. Hoyes Beehler,

Kenneth J. Maxwell

Karl Syndulko

155

Applications of Quantitative Assessment of Human Performance in Occupational Medicine Mohamad Parnianpour

156

Human Performance Engineering: Computer-Based Design and Analysis Tools Paul J. Vasta, George V. Kondraske


157

Human Performance Engineering: Challenges and Prospects for the Future George V. Kondraske

SECTION XVI Introduction

158

Physiological Modeling, Simulation, and Control

Joseph L. Palladino

Modeling Strategies in Physiology

J. L. Palladino, G. M. Drzewiecki,

A. Noordergraaf

159

Compartmental Models of Physiologic Systems

Claudio Cobelli,

Giovanni Sparacino, Andrea Caumo, Maria Pia Saccomani, Gianna Maria Toffolo

160

Cardiovascular Models and Control

161

Respiratory Models and Control

162

Neural Networks for Physiological Control

163

Methods and Tools for Identification of Physiologic Systems

William D. Timmons

Chi-Sang Poon James J. Abbas

Vasilis Z. Marmarelis

164

Autoregulating Windkessel Dynamics May Cause Low Frequency Oscillations Gary Drzewiecki, John K-J. Li, Abraham Noordergraaf

165

Control of Movements

166

The Fast Eye Movement Control System

SECTION XVII Introduction

Dejan B. Popovi c´ John Denis Enderle

Clinical Engineering

Yadin David

167

Clinical Engineering: Evolution of a Discipline

168

Management and Assessment of Medical Technology

Joseph D. Bronzino Yadin David,

Thomas M. Judd

169

Risk Factors, Safety, and Management of Medical Equipment Michael L. Gullikson

170

Clinical Engineering Program Indicators

Dennis D. Autio,

Robert L. Morris

171

Quality of Improvement and Team Building


Joseph P. McClain

172

A Standards Primer for Clinical Engineers

173

Regulatory and Assessment Agencies

174

Applications of Virtual Instruments in Health Care

Alvin Wald

Mark E. Bruley, Vivian H. Coates Eric Rosow,

Joseph Adam

SECTION XVIII Introduction

175

Medical Informatics

Luis G. Kun

Hospital Information Systems: Their Function and State T. Allan Pryor

176

Computer-Based Patient Records

177

Computer Networks in Health Care

178

Overview of Standards Related to the Emerging Health Care Information Infrastructure Jeffrey S. Blair

179

Non-AI Decision Making

J. Michael Fitzmaurice Soumitra Sengupta

Ron Summers, Derek G. Cramp,

Ewart R. Carson

180

Design Issues in Developing Clinical Decision Support and Monitoring Systems John W. Goethe, Joseph D. Bronzino

SECTION XIX Artificial Intelligence Introduction

Stanley M. Finkelstein

181

Artificial Intelligence in Medical Decision Making: History, Evolution, and Prospects Casimir A. Kulikowski

182

Artificial Neural Networks: Definitions, Methods, Applications Daniel A. Zahner, Evangelia Micheli-Tzanakou

183

Clinical Decision Systems

184

Expert Systems: Methods and Tools

Pirkko Nykänen, Niilo Saranummi Ron Summers, Derek Cramp,

Ewart R. Carson

185

Knowledge Acquisition and Representation

186

Knowledge-Based Systems for Intelligent Patient Monitoring and Management in Critical Care Environments Benoit M. Dawant, Patrick R. Norris


Catherine Garbay

187

Medical Terminology and Diagnosis Using Knowledge Bases Peter L. M. Kerkhof

188

Natural Language Processing in Biomedicine

Historical Perspectives 5: Electroencephalography

Stephen B. Johnson

Leslie A. Geddes

SECTION XX Ethical Issues Associated with the Use of Medical Technology Introduction

Subrata Saha, Joseph D. Bronzino

189

Professional Ethics in Biomedical Engineering

190

Beneficence, Nonmaleficence, and Technological Progress

Daniel E. Wueste

Joseph D. Bronzino

191

Ethical Issues of Animal and Human Experimentation in the Development of Medical Devices Subrata Saha, Pamela S. Saha

192

Regulation of Medical Device Innovation

Appendix A

Joseph D. Bronzino

The Role of Professional Societies in Biomedical Engineering

Swamy Laxminarayan, Joseph D. Bronzino, Jan E.W. Beneken, Shiro Usai, Richard D. Jones


Robert Plonsey. “Physiologic Systems.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

A view of the human cerebral cortex showing the underlying auditory cortex.


I Physiologic Systems Robert Plonsey Duke University 1 An Outline of Cardiovascular Structure and Function Daniel J. Schneck The Working Fluid: Blood • The Pumping Station: The Heart • The Piping Network: Blood Vessels • Cardiovascular Control

2 Endocrine System

Derek G. Cramp, Ewart R. Carson

Endocrine System: Hormones, Signals, and Communication Between Cells and Tissues • Hormone Action at the Cell Level: Signal Recognition, Signal Transduction, and Effecting a Physiological Response • Endocrine System: Some Other Aspects of Regulation and Control

3 Nervous System

Evangelia Micheli-Tzanakou

Definitions • Functions of the Nervous System • Representation of Information in the Nervous System • Lateral Inhibition • Higher Functions of the Nervous System

4 Vision System George Stetten Fundamentals of Vision Research • A Modular View of the Vision System

5 Auditory System

Ben M. Clopton, Francis A. Spelman

Physical and Psychological Variables • The Peripheral Auditory System • The Central Auditory System • Pathologies • Models of Auditory Function

6 The Gastrointestinal System

Berj L. Bardakjian

gastrointestinal Electrical Oscillations • A Historical Perspective • The Stomach • The Small Intestine • The Colon • Epilogue

7 Respiratory System Joseph D. Bronzino

Arthur T. Johnson, Christopher G. Lausted,

Respiration Anatomy • Lung Volumes and Gas Exchange • Perfusion of the Lung • Gas Partial Pressures • Pulmonary Mechanics • Respiratory Control • The Pulmonary Function Laboratory

T

HE CONTENT OF THIS HANDBOOK is devoted to the subject of biomedical engineering. We understand biomedical engineering to involve the application of engineering science and technology to problems arising in medicine and biology. In principle, the intersection of each engineering discipline (i.e., electrical, mechanical, chemical, etc.) with each discipline in medicine (i.e., cardiology, pathology, neurology, etc.) or biology (i.e., biochemistry, pharmacology, molecular biology,


cell biology, etc.) is a potential area of biomedical engineering application. As such, the discipline of biomedical engineering is potentially very extensive. However, at least to date, only a few of the aforementioned “intersections” contain active areas of research and/or development. The most significant of these are described in this Handbook. While the application of engineering expertise to the life sciences requires an obvious knowledge of contemporary technical theory and its applications, it also demands an adequate knowledge and understanding of relevant medicine and biology. It has been argued that the most challenging part of finding engineering solutions to problems lies in the formulation of the solution in engineering terms. In biomedical engineering, this usually demands a full understanding of the life science substrates as well as the quantitative methodologies. This section is devoted to an overview of the major physiologic systems of current interest to biomedical engineers, on which their work is based. The overview may contain useful definitions, tables of basic physiologic data, and an introduction to the literature. Obviously these chapters must be extremely brief. However, our goal is an introduction that may enable the reader to clarify some item of interest or to indicate a way to pursue further information. Possibly the reader will find the greatest value in the references to more extensive literature. This section contains seven chapters, and these describe each of the major organ systems of the human body. Thus we have chapters describing the cardiovascular, endocrine, nervous, visual, auditory, gastrointestinal, and respiratory systems. While each author is writing at an introductory and tutorial level, the audience is assumed to have some technical expertise, and consequently, mathematical descriptions are not avoided. All authors are recognized as experts on the system which they describe, but all are also biomedical engineers. The authors in this section noted that they would have liked more space but recognized that the main focus of this Handbook is on “engineering.” Hopefully, readers will find this introductory section helpful to their understanding of later chapters of this Handbook and, as noted above, to at least provide a starting point for further investigation into the life sciences.


Schneck, D. J. “An Outline of Cardiovascular Structure and Function.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

1 An Outline of Cardiovascular Structure and Function Daniel J. Schneck Virginia Polytechnic Institute and State University

1.1 1.2 1.3 1.4

The Working Fluid: Blood The Pumping Station: The Heart The Piping Network: Blood Vessels Cardiovascular Control

Because not every cell in the human body is near enough to the environment to easily exchange with it mass (including nutrients, oxygen, carbon dioxide, and the waste products of metabolism), energy (including heat), and momentum, the physiologic system is endowed with a major highway network—organized to make available thousands of miles of access tubing for the transport to and from a different neighborhood (on the order of 10 µm or less) of any given cell whatever it needs to sustain life. This highway network, called the cardiovascular system, includes a pumping station, the heart; a working fluid, blood; a complex branching configuration of distributing and collecting pipes and channels, blood vessels; and a sophisticated means for both intrinsic (inherent) and extrinsic (autonomic and endocrine) control.

1.1 The Working Fluid: Blood Accounting for about 8 ± 1% of total body weight, averaging 5200 ml, blood is a complex, heterogeneous suspension of formed elements—the blood cells, or hematocytes—suspended in a continuous, strawcolored fluid called plasma. Nominally, the composite fluid has a mass density of 1.057 ± 0.007 g/cm3, and it is three to six times as viscous as water. The hematocytes (Table 1.1) include three basic types of cells: red blood cells (erythrocytes, totaling nearly 95% of the formed elements), white blood cells (leukocytes, averaging > λ .

(10.13)

The leading factor is the parallel combination of the intracellular and extracellular resistances. If, on the other hand, L is very small compared to λ , the extracellular voltage drop becomes

∆Ve = re L I

L 0 for – – some nonzero positive constant α and (AΦ, Φ) = (Φ, AΦ), respectively], then we can define a space E – – – with an inner product defined as (Φ, Φ)E = (AΦ, Φ) a(Φ, Φ) and norm (the so-called energy norm) equal to

Φ


E

 = 

∫( ) Ω

1

( )

2 ∇Φ d Ω = Φ, Φ  2

1 2

E

(12.14)

The solution Φ of Eq. (12.9) satisfies

( AΦ, ψ ) = −(I , ψ ) (∀ψ ∈S) i

v

i

(12.15)

i

and the approximate Galerkin solution obtained by solving Eq. (12.13) satisfies

( Aφ, ψ ) = −(I , ψ ) (∀ψ ∈S) i

v

i

(12.16)

i

Subtracting Eq. (12.15) from Eq. (12.16) yields

( A(φ − Φ), ψ ) = (ϕ − ϕ, ψ ) i

i

E

(∀ψ ∈S)

=0

(12.17)

i

The difference φ – Φ denotes the error between the solution in the finite dimensional space V and the N + 1 dimensional space S. Equation (12.17) states that the error is orthogonal to all basis functions spanning the space of possible Galerkin solutions. Consequently, the error is orthogonal to all elements in S and must therefore be the minimum error. Thus the Galerkin approximation is an orthogonal projection of the true solution Φ onto the given finite dimensional space of possible approximate solutions. Therefore, the Galerkin approximation is the best approximation in the energy space E. Since the operator is positive definite, the approximate solution is unique. Assume for a moment that there are two solutions, φ1 and φ2, satisfying

( Aφ , ψ ) = −(I , ψ ) ( Aφ , ψ ) = −(I , ψ ) (∀ψ ∈S) 1

i

v

i

2

i

v

i

i

(12.18)

respectively. Then the difference yields

( A(φ − φ ), ψ ) = 0 (∀ψ ∈S) 1

2

i

i

(12.19)

The function arising from subtracting one member from another member in S also belongs in S; hence the difference function can be expressed by the set of A orthogonal basis functions spanning S:

∑ ∆φ ( A(ψ , ψ )) = 0 (∀ψ ∈) N

j

j

i

i

(12.20)

j =0

When i ≠ j, the terms vanish due to the basis functions being orthogonal with respect to A. Since A is positive definite,

( AΦ , Φ ) > 0 i

i

i = 0,…, N

(12.21)

Thus ∆φ i = 0, i = 0, … , N, and by virtue of Eq. (12.20), δφ = 0, such that φ1 = φ 2. The identity contradicts the assumption of two distinct Galerkin solutions. This proves the solution is unique [28].

The Finite-Difference Method Perhaps the most traditional way to solve Eq. (12.1) utilizes the finite-difference approach by discretizing the solution domain Ω using a grid of uniform hexahedral elements. The coordinates of a typical grid


point are x = lh, y = mh, z = nh (l, m, n = integers), and the value of Φ(x, y, z) at a grid point is denoted by Φl,m,n . Taylor’s theorem can then be used to provide the difference equations. For example,

  ∂Φ 1 2 ∂2Φ 1 3 ∂ 3Φ Φl +1,m ,n =  Φ + h + h + h + L  ∂x 2 ∂x 2 6 ∂x 3   l ,m ,n

(12.22)

with similar equations for Φl–1,m,n, Φl,m+1,n, Φl,m–1,n,…. The finite-difference representation of Eq. (12.1) is

Φl +1,m ,n − 2Φl ,m ,n + Φl −1,m ,n h +

2

+

Φl ,m +1,n − 2Φl ,m ,n + Φl ,m −1,n

Φl ,m ,n+1 − 2Φl ,m ,n + Φl ,m ,n−1 h2

h2

(12.23)

()

= − I l ,m ,n v

or, equivalently,

()

Φl +1,m ,n + Φl −1,m ,n + Φl ,m +1,n + Φl ,m −1,nΦl ,m ,n+1 + Φl ,m ,n−1 − 6Φl ,m ,n = −h2 I l ,m ,n v

(12.24)

If one defines the vector Φ to be [Φ1,1,1 L Φ1,1,N–1 L Φ1,N–1,1 L ΦN–1,N–1,N–1]T to designate the (N – 1)3 unknown grid values and pull out all the known information from (24), one can reformulate Eq. (12.1) by its finite-difference approximation in the form of the matrix equation AΦ = b, where b is a vector that contains the sources and modifications due to the Dirichlet boundary condition. Unlike the traditional Taylor’s series expansion method, the Galerkin approach utilizes basis functions, such as linear piecewise polynomials, to approximate the true solution. For example, the Galerkin approximation to sample problem (12.1) would require evaluating Eq. (12.13) for the specific grid formation and specific choice of basis function:



∫  σ Ω

x

∂φ ∂ψ i ∂φ ∂ψ i ∂φ ∂ψ i  dΩ = − + σy + σz ∂x ∂x ∂y ∂y ∂z ∂z 

∫ I ψ dΩ Ω

v

i

(12.25)

Difference quotients are then used to approximate the derivatives in Eq. (12.25). Note that if linear basis functions are utilized in Eq. (12.25), one obtains a formulation that corresponds exactly with the standard finite-difference operator. Regardless of the difference scheme or order of basis function, the approximation results in a linear system of equations of the form AΦ = b, subject to the appropriate boundary conditions.

The Finite-Element Method As seen earlier, in the classic numerical treatment for partial differential equation—the finite-difference method—the solution domain is approximated by a grid of uniformly spaced nodes. At each node, the governing differential equation is approximated by an algebraic expression that references adjacent grid points. A system of equations is obtained by evaluating the previous algebraic approximations for each node in the domain. Finally, the system is solved for each value of the dependent variable at each node. In the finite-element method, the solution domain can be discretized into a number of uniform or nonuniform finite elements that are connected via nodes. The change of the dependent variable with regard to location is approximated within each element by an interpolation function. The interpolation function is defined relative to the values of the variable at the nodes associated with each element. The original boundary value problem is then replaced with an equivalent integral formulation [such as


Eq. (12.13)]. The interpolation functions are then substituted into the integral equation, integrated, and combined with the results from all other elements in the solution domain. The results of this procedure can be reformulated into a matrix equation of the form AΦ = b, which is subsequently solved for the unknown variables [20, 29]. The formulation of the finite-element approximation starts with the Galerkin approximation, (σΦ, — — — Φ) = –(Iv , Φ), where Φ is our test function. Now one can use the finite-element method to turn the continuous problems into a discrete formulation. First, one —discretizes the solution domain, — — Ω = UEe=1Ω e, and defines a finite dimensional subspace Vh ⊂ V = {Φ: Φ is continuous on Ω, Φ is piece— — wise continuous on Ω}. One usually defines parameters of the function Φ ∈ Vh at node points αi = Φ(xi), i = 0, 1, … , N. If one now defines the basis functions ψi ∈ Vh as linear continuous piecewise functions that take the value 1 at node points and zero at other node points, then one can represent the — function Φ ∈ Vh as

( ) ∑ d ψ (x ) N

Φx =

i

(12.26)

i

i=0

—

such that each Φ ∈ Vh can be written in a unique way as a linear combination of the basis functions Ψi ∈ Vh . Now the finite-element approximation of the original boundary value problem can be stated as

(

) (

Find Φ h ∈Vh such that σ∇Φh , ∇Φ = − I v , Φ

)

(12.27)

Furthermore, if Φh ∈ Vh satisfies problem (12.27), then we have (σΦh , Ψi ) = –(Iv , Ψi) [30]. Finally, since Φh itself can be expressed as the linear combination

∑ ξ Ψ (x )

( )

N

Φh =

i

ξi = Φh xi

i

i=0

(12.28)

one can then write problem (12.27) as

∑ ξ ( σ ∇ Ψ , ∇ Ψ ) = −( I , Ψ ) N

i

ij

i

j

v

j

j = 0,…, N

(12.29)

i=0

subject to the Dirichlet boundary condition. Then the finite-element approximation of Eq. (12.1) can equivalently be expressed as a system of N equations with N unknowns ξi , …, ξN (the electrostatic potentials, for example). In matrix form, the preceding system can be written as Aξ = b, where A = (aij ) is called the global stiffness matrix and has elements (aij) = (σij Ψi, Ψj), while bi = –(Iv , Ψi) and is usually termed the load vector. For volume conductor problems, A contains all the geometry and conductivity information of the model. The matrix A is symmetric and positive definite; thus it is nonsingular and has a unique solution. Because the basis function differs from zero for only a few intervals, A is sparse (only a few of its entries are nonzero). Application of the FE Method for 3D Domains Now let us illustrate the concepts of the finite-element method by considering the solution of Eq. (12.1) using linear three-dimensional elements. One starts with a 3D domain Ω that represents the geometry of our volume conductor and breaks it up into discrete elements to form a finite dimensional subspace Ω h . For 3D domains, one has the choice of representing the function as either tetrahedra


˜ = α + α x + α y + α z, Φ 1 2 3 4

(12.30)

˜ = α + α x + α y + α z + α xy + α yz + α xz + α xyz Φ 1 2 3 4 5 6 7 8

(12.31)

or hexahedra

Because of space limitations, let us restrict the development to tetrahedra, knowing that it is easy to modify the formulas for hexahedra. Take out a specific tetrahedra from the finite dimensional subspace and apply the previous formulations for the four vertices:

˜  1 x Φ 1 1 ˜   1 Φ x 2  2 =  ˜  1 x Φ 3 3 ˜   1 Φ x  4  4

y1 z1   α1    y2 z 2   α 2  y3 z3   α3   y4 z4   α4 

(12.32)

or

Φi = Cα

(12.33)

˜ α = C −1Φ i

(12.34)

which define the coordinate vertices, and

˜ at any point within the which defines the coefficients. From Eqs. (12.30) and (12.34) one can express Φ tetrahedra,

[

]

˜ = 1, x , y , z α = Sα = SC −1Φ ˜ Φ i

(12.35)

or, most succinctly,

˜= Φ

∑ N Φ˜

(12.36)

i

i

˜ i is the solution value at node i, and N = SC–1 is the local shape function or basis function. This can be Φ expressed in a variety of ways in the literature (depending, usually, on whether you are reading engineering or mathematical treatments of finite element analysis):

( )

(

) (

)

Φ j N i = N i x , y , z = fi x , y , z ≡

ai + bi x + ci y + di z 6V

(12.37)

where

1 1 6V = 1 1 defines the volume of the tetrahedra V.


x1 x2 x3 x4

y1 y2 y3 y4

z1 z2 z3 z4

(12.38)

Now that a suitable set of basis functions is available, one can find the finite-element approximation to the 3D problem. The original problem can be formulated as

( ) ( )

a u, v = I v , v

∀v ∈Ω

(12.39)

where

( ) ∫

a u, v = ∇u ⋅∇vdΩ

(12.40)

(I , v ) = ∫ I ⋅ vdΩ

(12.41)

Ω

and

v

Ω

v

The finite-element approximation to the original boundary value problem is

( ) ( )

a uh , v = I v , v

∀v ∈Ωh

(12.42)

which has the equivalent form

∑ ξ a (Φ , Φ ) = ( I , Φ ) N

i

i

j

v

(12.43)

j

i =1

where

(

) ( ( ) ( ))

a Φi , Φ j = a Φi , N j , Φ j N i

(12.44)

which can be expressed by the matrix and vector elements.

 ∂N ∂N j ∂N ∂N j ∂N ∂N j  i i i + +   dΩ Ω E  ∂x ∂x ∂y ∂y ∂z ∂z 

(a ) = ∫ ij

(12.45)

and

Ii =

∫

ΩE

N i I v dΩ

(12.46)

Fortunately, these quantities are easy to evaluate for linear tetrahedra. As a matter of fact, there are closedform solutions for the matrix elements (aij):

∫

Ωh

N1a N 2b N 3c N 4d dΩ = 6V

a! b! c! d!

(a + b + c + d + 3)!

(12.47)

Therefore,

(a ) = ∫ ij


ΩE

bib j + cic j + did j 6V

2

dΩ =

bib j + cic j + did j 6V

(12.48)

and, for the right hand side, one has, assuming constant sources,

Ii =

ai + bi x + ci y + di z VI I v dΩ = v 6V 4

(12.49)

1  (n) (n) (n) (n) (n) (n)  bi b j + ci c j + di + d j  6V 

(12.50)

∫

which have the compact forms

aij( ) = n

and

I i( ) =

VI v 4

n

for constant sources

(12.51)

Now one adds up all the contributions from each element into a global matrix and global vector: Nel

∑  a( ) (ξ ) =  I ( ) n ij

n

i

(12.52)

i

n =1

where Nel is equal to the total number of elements in the discretized solution domain, and i represents the node numbers (vertices). This yields a linear system of equations of the form AΦ = b, where Φ is the solution vector of voltages, A represents the geometry and conductivity of the volume conductor, and b represents the contributions from the current sources and boundary conditions. For the finite-difference method, it turns out that the Dirichlet boundary condition is easy to apply, while the Neumann condition takes a little extra effort. For the finite-element method, it is just the opposite. The Neumann boundary condition

∇Φ ⋅ n = 0

(12.53)

is satisfied automatically within the Galerkin and variational formulations. This can be seen by using Green’s divergence theorem,

∫ ∇⋅ Adx = ∫ A ⋅ n dS Ω

(12.54)

Γ

and applying it to the left-hand side of the Galerkin finite-element formulation:

 ∂v ∂w ∂v ∂w  + dΩ  Ω  ∂x1 ∂x1 ∂x2 ∂x2 

∫ ∇v ⋅∇w dΩ ≡ ∫ Ω


 ∂w ∂w   v ∂x n1 + v ∂x n2  dS −   1 2

=

∫

=

∫ v ∂n dS − ∫

Γ

∂w

Γ

Ω

v∇2w dΩ

 ∂ 2w ∂ 2w  v  2 + 2 n2  dΩ Ω  ∂x ∂x2  1

∫

(12.55)

If one multiples the original differential equation, 2Φ = –Iv , by an arbitrary test function and integrates, one obtains

(I , v ) = −∫ (∇ Φ)v dΩ = −∫ 2

v

Ω

Γ

∂Φ v dS + ∂n

∫

Ω

( )

∇Φ ⋅∇v dΩ = a Φ, v

(12.56)

where the boundary integral term ∂Φ/∂n vanishes, and one obtains the standard Galerkin finite-element formulation. To apply the Dirichlet condition, one has to work a bit harder. To apply the Dirichlet boundary condition directly, one usually modifies the (aij) matrix and bi vector such that one can use standard linear system solvers. This is accomplished by implementing the following steps. Assuming that the ith value of ui is known, 1. Subtract from the ith member of the right-hand side the product of aij and the known value of — — ˆ =b –aΦ Φi (call it Φ i); this yields the new right-hand side, bi i i ij j. 2. Zero the ith row and column of A: aîj = aˆji = 0. 3. Assign aîi = 1. 4. Set the jth member of the right-hand side equal to Φi . 5. Continue for each Dirichlet condition. ˆ = bˆv . 6. Solve the augmented system AΦ

The Boundary-Element Method The bioelectric field problems with isotropic domains (and few inhomogeneities), another technique, called the boundary-element method, may be used. This technique utilizes information only on the boundaries of interest and thus reduces the dimension of any field problem by one. For differential operators, the response at any given point to sources and boundary conditions depends only on the response at neighboring points. The FD and FE methods approximate differential operators defined on subregions (volume elements) in the domain; hence direct mutual influence (connectivity) exists only between neighboring elements, and the coefficient matrices generated by these methods have relatively few nonzero coefficients in any given matrix row. As is demonstrated by Maxwell’s laws, equations in differential forms often can be replaced by equations in integral forms; e.g., the potential distribution in a domain is uniquely defined by the volume sources and the potential and current density on the boundary. The boundary-element method uses this fact by transforming the differential operators defined in the domain to integral operators defined on the boundary. In the boundary-element method [31–33], only the boundary is discretized; hence the mesh generation is considerably simpler for this method than for the volume methods. Boundary solutions are obtained directly by solving the set of linear equations; however, potentials and gradients in the domain can be evaluated only after the boundary solutions have been obtained. Since this method has a rich history in bioelectric field problems, the reader is referred to some of the classic references for further information regarding the application of the BE method to bioelectric field problems [6, 42–44].

Solution Methods and Computational Considerations Application of each of the previous approximation methods to Eq. (12.1) yields a system of linear equations of the form AΦ = b, which must be solved to obtain the final solution. There is a plethora of available techniques for the solutions of such systems. The solution techniques can be broadly categorized as direct and iterative solvers. Direct solvers include Gaussian elimination and LU decomposition, while iterative methods include Jacobi, Gauss-Seidel, successive overrelaxation (SOR), and conjugate gradient (CG) methods, among others. The choice of the particular solution method is highly dependent on the approximation technique employed to obtain the linear system, on the size of the resulting system, and


on accessible computational resources. For example, the linear system resulting from the application of the FD or FE method will yield a matrix A that is symmetric, positive definite, and sparse. The matrix resulting from the FD method will have a specific band-diagonal structure that is dependent on the order of difference equations one uses to approximate the governing equation. The matrix resulting from the FE method will be exceedingly sparse and only a few of the off-diagonal elements will be nonzero. The application of the BE method, on the other hand, will yield a matrix A that is dense and nonsymmetric and thus requires a different choice of solver. The choice of the optimal solver is further complicated by the size of the system versus access to computational resources. Sequential direct methods are usually confined to single workstations, and thus the size of the system should fit in memory for optimal performance. Sequential iterative methods can be employed when the size of the system exceeds the memory of the machine; however, one pays a price in terms of performance, since direct methods are usually much faster than iterative methods. In many cases, the size of the system exceeds the computational capability of a single workstation, and one must resort to the use of clusters of workstations and/or parallel computers. While new and improved methods continue to appear in the numerical analysis literature, my studies comparing various solution techniques for direct and inverse bioelectric field problems have resulted in the conclusion that the preconditioned conjugate gradient methods and multigrid methods are the best overall performers for volume conductor problems computed on single workstations. Specifically, the incomplete Choleski conjugate gradient (ICCG) method works well for the FE method,1 and the preconditioned biconjugate gradient (BCG) methods are often used for BE methods. When clusters of workstations and/or parallel architectures are considered, the choice is less clear. For use with some highperformance architectures that contain large amounts of memory, parallel direct methods such as LU decomposition become attractive; however, preconditioned conjugate gradient methods still perform well. A discussion of parallel computing methods for the solution of biomedical field problems could fill an entire text. Thus the reader is directed to the following references on parallel scientific computing [45–47].

Comparison of Methods Since there is not enough space to give a detailed, quantitative description of each of the previously mentioned methods, an abbreviated summary is given of the applicability of each method in solving different types of bioelectric field problems. As outlined earlier, the FD, FE, and BE methods can all be used to approximate the boundary value problems that arise in biomedical research problems. The choice depends on the nature of the problem. The FE and FD methods are similar in that the entire solution domain must be discretized, while with the BE method only the bounding surfaces must be discretized. For regular domains, the FD method is generally the easiest method to code and implement, but the FD method usually requires special modifications to define irregular boundaries, abrupt changes in material properties, and complex boundary conditions. While typically more difficult to implement, the BE and FE methods are preferred for problems with irregular, inhomogeneous domains and mixed boundary conditions. The FE method is superior to the BE method for representing nonlinearity and true anisotropy, while the BE method is superior to the FE method for problems where only the boundary solution is of interest or where solutions are wanted in a set of highly irregularly spaced points in the domain. Because the computational mesh is simpler for the BE method than for the FE method, the BE program requires less bookkeeping than an FE program. For this reason, BE programs are often considered easier to develop than FE programs; however, the difficulties associated with singular integrals in the BE method are often highly underestimated. In general, the FE method is preferred for problems where the domain is highly heterogeneous, whereas the BE method is preferred for highly homogeneous domains. 1This is specifically for the FE method applied to elliptic problems. Such problems yield a matrix that is symmetric and positive definite. The Choleski decomposition only exists for symmetric, positive-definite matrices.


12.4 Adaptive Methods Thus far how one formulates the problem, discretizes the geometry, and finds an approximate solution have been discussed. Now one is faced with answering the difficult question pertaining to the accuracy of the solution. Without reference to experimental data, how can one judge the validity of the solutions? To give oneself an intuitive feel for the problem (and possible solution), consider the approximation of a two-dimensional region discretized into triangular elements. The finite-element method will be applied to solve Laplace’s equation in the region. First, consider the approximation of the potential field Φ(x, y) by a two-dimensional Taylor’s series expansion about a point (x, y):

( )

( )

 ∂Φ x , y ∂Φ x , y   Φ x + h, y + k = Φ x , y + h +k  ∂x ∂y   

(

) ( )

( )

( )

( )

 ∂2 Φ x , y ∂2 Φ x , y ∂2 Φ x , y  1 2  +L + h2 + hk k + 2 ∂x∂y 2!  ∂2 x ∂2 y   

(12.57)

where h and k are the maximum x and y distances within an element. Using the first two terms (up to first-order terms) in the preceding Taylor’s expansion, one can obtain the standard linear interpolation function for a triangle:

(

∂Φ xi , yi ∂x

)=

[ (

)

(

)

(

1 Φi y j − ym + Φm yi − y j + Φ j ym − yi 2A

)]

(12.58)

where A is the area of the triangle. Likewise, one could calculate the interpolant for the other two nodes and discover that

(

∂Φ xi , yi ∂x

) = ∂Φ (x , y ) = ∂Φ (x j

j

∂x

m

, ym

∂x

)

(12.59)

is constant over the triangle (and thus so is the gradient in y as well). Thus one can derive the standard linear interpolation formulas on a triangle that represents the first two terms of the Taylor’s series expansion. This means that the error due to discretization (from using linear elements) is proportional to the third term of the Taylor’s expansion:

( )

( )

( )

2  ∂2 Φ x , y ∂2 Φ x , y  1  2 ∂ Φ x, y 2  + 2hk ≈ h +k ∂x∂y 2!  ∂2 x ∂2 y   

(12.60)

where Φ is the exact solution. One can conjecture, then, that the error due to discretization for firstorder linear elements is proportional to the second derivative. If Φ is a linear function over the element, then the first derivative is a constant and the second derivative is zero, and there is no error due to discretization. This implies that the gradient must be constant over each element. If the function is not linear or the gradient is not constant over an element, the second derivative will not be zero and is proportional to the error incurred due to “improper” discretization. Examining Eq. (12.60), one can easily see that one way to decrease the error is to decrease the size of h and k. As h and k go to zero, the error tends to zero as well. Thus decreasing the mesh size in places of high errors due to high gradients decreases the error. As an aside, note that if one divided Eq. (12.9) by hk, one can also express the error © 2000 by CRC Press LLC

in terms of the elemental aspect ratio h/k, which is a measure of the relative shape of the element. It is easy to see that one must be careful to maintain an aspect ratio as close to unity as possible. The problem with the preceding heuristic argument is that one has to know the exact solution a priori before one can estimate the error. This is certainly a drawback considering that one is trying to accurately approximate Φ.

Convergence of a Sequence of Approximate Solutions Let’s try to quantify the error a bit further. When one considers the preceding example, it seems to make sense that if one increases the number of degrees of freedom used to approximate the function, the accuracy must approach the true solution. That is, one would hope that the sequence of approximate solutions will converge to the exact solution as the number of degrees of freedom (DOF) increases indefinitely:

()

()

˜ x →0 Φ x −Φ n

as n → ∞

(12.61)

This is a statement of pointwise convergence. It describes the approximate solution as approaching arbitrarily close to the exact solution at each point in the domain as the number of DOF increases. Measures of convergence often depend on how the closeness of measuring the distance between functions is defined. Another common description of measuring convergence is uniform convergence, ˜ n(x) in the domain vanish as N → ∞. This is which requires that the maximum value of Φ(x) – Φ stronger than pointwise convergence because it requires a uniform rate of convergence at every point in the domain. Two other commonly used measures are convergence in energy and convergence in mean, which involve measuring an average of a function of the pointwise error over the domain [40]. In general, proving pointwise convergence is very difficult except in the simplest cases, while proving the convergence of an averaged value, such as energy, is often easier. Of course, scientists and engineers are often much more interested in ensuring that their answers are accurate in a pointwise sense than in an energy sense because they typically want to know values of the solution Φ(x) and gradients Φ(x) at specific places. One intermediate form of convergence is called the Cauchy convergence. Here, one requires the sequences of two different approximate solutions to approach arbitrarily close to each other:

()

()

˜ x →0 Φm x − Φ n

as m, n → ∞

(12.62)

While the pointwise convergence expression would imply the preceding equation, it is important to note that the Cauchy convergence does not imply pointwise convergence, since the functions could converge to an answer other than the true solution. While one cannot be assured of pointwise convergence of these functions for all but the simplest cases, there do exist theorems that ensure that a sequence of approximate solutions must converge to the exact solution (assuming no computational errors) if the basis functions satisfy certain conditions. The theorems can only ensure convergence in an average sense over the entire domain, but it is usually the case that if the solution converges in an average sense (energy, etc.), then it will converge in the pointwise sense as well.

Energy Norms The error in energy, measured by the energy norm, is defined in general as [37–39]

 e =  © 2000 by CRC Press LLC

1

2 e Le dΩ  Ω

∫

T

(12.63)

˜ n(x), and L is the differential operator for the governing differential equation [i.e., it where e = Φ(x) – Φ contains the derivatives operating on Φ(x) and any function multiplying Φ(x)]. For physical problems, this is often associated with the energy density. Another common measure of convergence utilizes the L2 norm. This can be termed the average error and can be associated with errors in any quantity. The L2 norm is defined as

 e =  L2

1

2 e e dΩ  Ω

∫

T

(12.64)

While the norms given above are defined on the whole domain, one can note that the square of each can be obtained by summing element contributions: M

2

e =

∑e i =1

2 i

(12.65)

where i represents an element contribution and m the total element number. Often for an optimal finiteelement mesh, one tries to make the contributions to this square of the norm equal for all elements. While the absolute values given by the energy or L2 norms have little value, one can construct a relative percentage error that can be more readily interpreted:

η=

e Φ

× 100

(12.66)

This quantity, in effect, represents a weighted RMS error. The analysis can be determined for the whole domain or for element subdomains. One can use it in an adaptive algorithm by checking element errors against some predefined tolerance η0 and increasing the DOF only of those areas above the predefined tolerance. Two other methods, the p and the hp methods, have been found, in most cases, to converge faster than the h method. The p method of refinement requires that one increase the order of the bias function that was used to represent the interpolation (i.e., linear to quadratic to cubic, etc.) The hp method is a combination of the h and p methods and has recently been shown to converge the fastest of the three methods (but, as you might imagine, it is the hardest to implement). To find out more about adaptive refinement methods, see [27, 30, 34, 35, 37, 40].

Acknowledgments This work was supported in part by awards from the Whitaker Foundation, the NIH, and the NSF. I would like to thank K. Coles, J. Schmidt, and D. Weinstein for their helpful comments and suggestions. An expanded, educational case studies version of this chapter entitled, “Direct and Inverse Bioelectric Field Problems,” may be found as part of the DOE-sponsored Computational Science Education Program (CSEP). The chapter is available freely via Mosaic at http://csep1.phy.ornl.gov/csep.html and contains exercises, projects, and C and Fortran computer codes.


References 1. Miller CE, Henriquez CS. Finite element analysis of bioelectric phenomena. Crit Rev Biomed Eng 18:181, 1990. This represents the first review paper on the use of the finite-element method as applied to biomedical problems. As the authors note, bioengineers came to these methods only fairly recently as compared with other engineers. It contains a good survey of applications. 2. Nenonen J, Rajala HM, Katilia T. Biomagnetic Localization and 3D Modelling. Report TKK-FA689. Espoo, Finland, Helsinki University of Technology, 1992. 3. Johnson CR, MacLeod RS, Ershler PR. A computer model for the study of electrical current flow in the human thorax. Comput Biol Med 22(3):305, 1992. This paper details the construction of a three-dimensional thorax model. 4. Johnson CR, MacLeod RS, Matheson MA. Computer simulations reveal complexity of electrical activity in the human thorax. Comput Phys 6(3):230, 1992. This paper deals with the computational and visualization aspects of the forward ECG problem. 5. Kim Y, Fahy JB, Tupper BJ. Optimal electrode designs for electrosurgery. IEEE Trans Biomed Eng 33:845, 1986. This paper discusses an example of the modeling of bioelectric fields for applications in surgery. 6. Plonsey R. Bioelectric Phenomena. New York, McGraw-Hill, 1969. This is the first text using physics, mathematics, and engineering principles to quantify bioelectric phenomena. 7. Henriquez CS, Plonsey R. Simulation of propagation along a bundle of cardiac tissue. I. Mathematical formulation. IEEE Trans Biomed Eng 37:850, 1990. This paper and the companion paper below describe the bidomain approach to the propagation of electrical signals through active cardiac tissue. 8. Henriquez CS, Plonsey R. Simulation of propagation along a bundle of cardiac tissue: II. Results of simulation. IEEE Trans Biomed Eng 37:861, 1990. 9. Keener JP. Waves in excitable media. SIAM J Appl Math 46:1039, 1980. This paper describes mathematical models for wave propagation in various excitable media. Both chemical and physiologic systems are addressed. 10. Hadamard J. Sur les problemes aux derivees parielies et leur signification physique. Bull Univ Princeton, pp 49–52, 1902. This is Hadamard’s original paper describing the concepts of well- and ill-posedness, in French. 11. Greensite F, Huiskamp G, van Oosterom A. New quantitative and qualitative approaches to the inverse problem of electrocardiography: Their theoretical relationship and experimental consistency. Med Phy 17(3):369, 1990. This paper describes methods for constraining the inverse problem of electrocardiology in terms of sources. Methods are developed which put bounds on the space of acceptable solutions. 12. Tikhonov A, Arsenin V. Solution of Ill-Posed Problems. Washington, Winston, 1977. This is the book in which Tikhonov describes his method of regularization for ill-posed problems. 13. Tikhonov AN, Goncharsky AV. Ill-Posed Problems in the Natural Sciences. Moscow, MIR Publishers, 1987. This is a collection of research papers from physics, geophysics, optics, medicine, etc., which describe ill-posed problems and the solution techniques the authors have developed. 14. Glasko VB. Inverse Problems of Mathematical Physics. New York, American Institute of Physics, 1984. This book has several introductory chapters on the mathematics of ill-posed problems followed by several chapters on specific applications. 15. Hansen PC. Analysis of discrete ill-posed problems by means of the L-curve. SIAM Rev 34(4):561, 1992. This is an excellent review paper which describes the various techniques developed to solve illposed problems. Special attention is paid to the selection of the a priori approximation of the regularization parameter.


16. Hansen PC, Regularization tools: A Matlab package for analysis and solution of discrete ill-posed problems. Available via netlib in the library numeralgo/no4. This is an excellent set of tools for experimenting with and analyzing discrete ill-posed problems. The netlib library contains several Matlab routines as well as a postscript version of the accompanying technical report/manual. 17. Yamashita Y. Theoretical studies on the inverse problem in electrocardiography and the uniqueness of the solution. IEEE Trans Biomed Eng 29:719, 1982. The first paper to prove the uniqueness of the inverse problem in electrocardiography. 18. Bertrand O. 3D finite element method in brain electrical activity studies. In J Nenonen, HM Rajala, T Katila (eds), Biomagnetic Localization and 3D Modeling, pp 154–171. Helsinki, Helsinki University of Technology, 1991. This paper describes the inverse MEG problem using the finite-element method. 19. Bowyer A. Computing Dirichlet tesselations. Comput J 24:162, 1981. One of the first papers on the Delaunay triangulation in 3-space. 20. Hoole SRH. Computer-Aided Analysis and Design of Electromagnetic Devices. New York, Elsevier, 1989. While the title wouldn’t make you think so, this is an excellent introductory text on the use of numerical techniques to solve boundary value problems in electrodynamics. The text also contains sections on mesh generation and solution methods. Furthermore, it provides pseudocode for most of the algorithms discussed throughout the text. 21. MacLeod RS, Johnson CR, Matheson MA. Visualization tools for computational electrocardiography. In Visualization in Biomedical Computing, pp 433–444, 1992. This paper and the paper which follows concern the modeling and visualization aspects of bioelectric field problems. 22. MacLeod RS, Johnson CR, Matheson MA. Visualization of cardiac bioelectricity—A case study. IEEE Visualization 92:411, 1992. 23. Pilkington TC, Loftis B, Thompson JF, et al. High-Performance Computing in Biomedical Research. Boca Raton, Fla, CRC Press, 1993. This edited collection of papers gives an overview of the state of the art in high-performance computing (as of 1993) as it pertains to biomedical research. While certainly not comprehensive, the text does showcase several interesting applications and methods. 24. Thompson J, Weatherill NP. Structed and unstructed grid generation. In TC Pilkington, B Loftis, JF Thompson, et al. (eds), High-Performance Computing in Biomedical Research, pp 63–112. Boca Raton, Fla, CRC Press, 1993. This paper contains some extensions of Thompson’s classic textbook on numerical grid generation. 25. George PL. Automatic Mesh Generation. New York, Wiley, 1991. This is an excellent introduction to mesh generation. It contains a survey of all the major mesh generation schemes. 26. Thompson JF, Warsi ZUA, Mastin CW. Numerical Grid Generation. New York, North-Holland, 1985. This is the classic on mesh generation. The mathematical level is higher than that of George’s book, and most of the applications are directed toward computational fluid dynamics. 27. Schmidt JA, Johnson CR, Eason JC, MacLeod RS. Applications of automatic mesh generation and adaptive methods in computational medicine. In JE Flaherty and I. Babuska (eds), Modeling, Mesh Generation, and Adaptive Methods for Partial Differential Equations. New York, Springer-Verlag, 1994. This paper describes three-dimensional mesh generation techniques and adaptive methods for bioelectric field problems. 28. Henriquez CS, Johnson CR, Henneberg KA, et al. Large scale biomedical modeling and simulation: From concept to results. In N Thakor (ed), Frontiers in Biomedical Computing. Philadelphia, IEEE Press, 1994. This paper describes the process of large-scale modeling along with computational and visualization issues pertaining to bioelectric field problems. 29. Akin JE. Finite Element Analysis for Undergraduates. New York, Academic Press, 1986. This is an easy-to-read, self-contained text on the finite-element method aimed at undergraduate engineering students.


30. Johnson C. Numerical Solution of Partial Differential Equations by the Finite Element Method. Cambridge, Cambridge University Press, 1990. An excellent introductory book on the finite-element method. The text assumes mathematical background of a first-year graduate student in applied mathematics and computer science. An excellent introduction to the theory of adaptive methods. 31. Brebbia CA, Dominguez J. Boundary Elements: An Introductory Course. New York, McGraw-Hill, 1989. This is an introductory book on the boundary element method by one of the foremost experts on the subject. 32. Jawson MA, Symm GT. Integral Equation Methods in Potential Theory and Elastostatics. London, Academic Press, 1977. An introduction to the boundary integral method as applied to potential theory and elastostatics. 33. Beer G, Watson JO. Introduction to Finite and Boundary Element Methods for Engineers. New York, Wiley, 1992. This is an excellent first book for those wishing to learn about the practical aspects of the numerical solution of boundary value problems. The book covers not only finite and boundary element methods but also mesh generation and the solution of large systems. 34. Johnson CR, MacLeod RS. Nonuniform spatial mesh adaption using a posteriori error estimates: Applications to forward and inverse problems. Appl Num Math 14:331, 1994. This is a paper by the author which describes the application of the h-method of mesh refinement for large-scale twoand three-dimensional bioelectric field problems. 35. Flaherty JE. Adaptive Methods for Partial Differential Equations. Philadelphia, SIAM, 1989. This is a collection of papers on adaptive methods, many by the founders in the field. Most of the papers are applied to adaptive methods for finite-element methods and deal with both theory and applications. 36. Golub GH, Van Loan CF. Matrix Computations. Baltimore, Johns Hopkins, 1989. This is a classic reference for matrix computations and is highly recommended. 37. Zienkiewicz OC. The Finite Element Method in Engineering Science. New York, McGraw-Hill, 1971. This is a classic text on the finite-method. It is now in its fourth edition, 1991. 38. Zienkiewicz OC, Zhu JZ. A simple error estimate and adaptive procedure for practical engineering analysis. Int J Num Meth Eng 24:337, 1987. This is a classic paper which describes the use of the energy norm to globally refine the mesh based on a priori error estimates. 39. Zienkiewicz OC, Zhu JZ. Adaptivity and mesh generation. Int J Num Meth Eng 32:783, 1991. This is another good paper on adaptive methods which describes some more advanced methods than the 1987 paper. 40. Burnett DS. Finite Element Method. Reading, Mass, Addison-Wesley, 1988. This is an excellent introduction to the finite-element method. It covers all the basics and introduces more advanced concepts in a readily understandable context. 41. Barber CB, Dobkin DP, Huhdanpaa H. The quickhull algorithm for convex hull. Geometry Center Technical Report GCG53. A public domain two- and three-dimensional Delaunay mesh generation code. Available via anonymous ftp from geom.umn.edu:/pub/software/qhull.tar.Z. There is also a geometry viewer available from geom.umn.edu:/pub/software/geomview/geomview-sgi.tar.Z. 42. Barr RC, Pilkington TC, Boineau JP, Spach MS. Determining surface potentials from current dipoles, with application to electrocardiography. IEEE Trans Biomed Eng 13:88, 1966. This is the first paper on the application of the boundary-element method to problems in electrocardiography. 43. Rudy Y, Messinger-Rapport BJ. The inverse solution in electrocardiography: Solutions in terms of epicardial potentials. CRC Crit Rev Biomed Eng 16:215, 1988. An excellent overview on the inverse problem in electrocardiography as well as a section on the application of the boundary-element method to bioelectric field problems. 44. Gulrajani RM, Roberge FA, Mailloux GE. The forward problem of electrocardiography. In PW Macfarlane and TD Lawrie (eds), Comprehensive Electrocardiology, pp 197–236. Oxford, England, Pergamon Press, 1989. This contains a nice overview of the application of the boundary-element method to the direct ECG problem.


45. Golub G, Ortega JM. Scientific Computing: An Introduction with Parallel Computing. New York, Academic Press, 1993. An excellent introduction to scientific-computing with sections on sequential and parallel direct and iterative solvers, including a short section on the multigrid method. 46. Freeman TL, Phillips C. Parallel Numerical Algorithms. New York, Prentice-Hall, 1992. A good introduction to parallel algorithms with emphasis on the use of BLAS subroutines. 47. Van de Velde EF. Concurrent Scientific Computing. New York, Springer-Verlag, 1994. This text contains an excellent introduction to parallel numerical algorithms including sections on the finitedifference and finite-element methods.


Berbari, E. J. “Principles of Electrocardiography.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

13 Principles of Electrocardiography 13.1 13.2 13.3

The Ambulatory ECG • Patient Monitoring • High Resolution ECG

Edward J. Berbari Indiana University/Purdue University at Indianapolis

Physiology Instrumentation Applications

13.4

Conclusions

The electrocardiogram (ECG) is the recording on the body surface of the electrical activity generated by heart. It was originally observed by Waller in 1889 [1] using his pet bulldog as the signal source and the capillary electrometer as the recording device. In 1903, Einthoven [2] enhanced the technology by employing the string galvanometer as the recording device and using human subjects with a variety of cardiac abnormalities. Einthoven is chiefly responsible for introducing some concepts still in use today, including the labeling of the various waves, defining some of the standard recording sites using the arms and legs, and developing the first theoretical construct whereby the heart is modeled as a single timevarying dipole. We also owe the EKG acronym to Einthoven’s native Dutch language, where the root word cardio is spelled with a k. In order to record an ECG waveform, a differential recording between two points on the body are made. Traditionally, each differential recording is referred to as a lead. Einthoven defined three leads numbered with the Roman numerals I, II, and III. They are defined as

I = VLA − VRA II = VLL − VRA III = VLL − VLA where RA = right arm, LA = left arm, and LL = left leg. Because the body is assumed to be purely resistive at ECG frequencies, the four limbs can be thought of as wires attached to the torso. Hence lead I could be recorded from the respective shoulders without a loss of cardiac information. Note that these are not independent, and the following relationship holds: II = I + III. The evolution of the ECG proceeded for 30 years when F.N. Wilson added concepts of a “unipolar” recording [3]. He created a reference point by tying the three limbs together and averaging their potentials so that individual recording sites on the limbs or chest surface would be differentially recorded with the same reference point. Wilson extended the biophysical models to include the concept of the cardiac source enclosed within the volume conductor of the body. He erroneously thought that the central terminal was a true zero potential. However, from the mid-1930s until today, the 12 leads composed of the 3 limb


FIGURE 13.1 The 12-lead ECG is formed by the 3 bipolar surface leads: I, II, and III; the augmented Wilson terminal referenced limb leads: aVR , aVL , and aVF ; and the Wilson terminal referenced chest leads: V1 , V2 , V3 , V4 , V5 , and V6 .

leads, 3 leads in which the limb potentials are referenced to a modified Wilson terminal (the augmented leads [4]), and 6 leads placed across the front of the chest and referenced to the Wilson terminal form the basis of the standard 12-lead ECG. Figure 13.1 summarizes the 12-lead set. These sites are historically based, have a built-in redundancy, and are not optimal for all cardiac events. The voltage difference from any two sites will record an ECG, but it is these standardized sites with the massive 90-year collection of empirical observations that have firmly established their role as the standard. Figure 13.2 is a typical or stylized ECG recording from lead II. Einthovin chose the letters of the alphabet from P to U to label the waves and to avoid conflict with other physiologic waves being studied at the turn of the century. The ECG signals are typically in the range of ±2 mV and require a recording bandwidth of 0.05 to 150 Hz. Full technical specification for ECG equipment has been proposed by both the American Heart Association [5] and the Association for the Advancement of Medical Instrumentation [6]. There have been several attempts to change the approach for recording the ECG. The vectorcardiogram used a weighted set of recording sites to form an orthogonal xyz lead set. The advantage here was minimum lead set, but in practice it gained only a moderate degree of enthusiasm among physicians. © 2000 by CRC Press LLC

FIGURE 13.2 This is a stylized version of a normal lead II recording showing the P wave, QRS complex, and the T and U waves. The PR interval and the ST segment are significant time windows. The peak amplitude of the QRS is about 1 mV. The vertical scale is usually 1 mV/cm. The time scale is usually based on millimeters per second scales, with 25 mm/s being the standard form. The small boxes of the ECG are 1 × 1 mm.

Body surface mapping refers to the use of many recording sites (>64) arranged on the body so that isopotential surfaces could be computed and analyzed over time. This approach still has a role in research investigations. Other subsets of the 12-lead ECG are used in limited-mode recording situations such as the tape-recorded ambulatory ECG (usually 2 leads) or in intensive care monitoring at the bedside (usually 1 or 2 leads) or telemetered within regions of the hospital from patients who are not confined to bed (1 lead). The recording electronics of these ECG systems have followed the typical evolution of modern instrumentation, e.g., vacuum tubes, transistors, integrated chips, and microprocessors. Application of computers to the ECG for machine interpretation was one of the earliest uses of computers in medicine [7]. Of primary interest in the computer-based systems was replacement of the human reader and elucidation of the standard waves and intervals. Originally this was performed by linking the ECG machine to a centralized computer via phone lines. The modern ECG machine is completely integrated with an analog front end, a 12- to 16-bit analog-to-digital (A/D) converter, a computational microprocessor, and dedicated input-output (I/O) processors. These systems compute a measurement matrix derived from the 12 lead signals and analyze this matrix with a set of rules to obtain the final set of interpretive statements [8]. Figure 13.3 shows the ECG of a heartbeat and the types of measurements that might be made on each of the component waves of the ECG and used for classifying each beat type and the subsequent cardiac rhythm. The depiction of the 12 analog signals and this set of interpretive statements form the final output, with an example shown in Fig. 13.4. The physician will overread each ECG and modify or correct those statements which are deemed inappropriate. The larger hospital-based system will record these corrections and maintain a large database of all ECGs accessible by any combination of parameters, e.g., all males, older than age 50, with an inferior myocardial infarction. There are hundreds of interpretive statements from which a specific diagnosis is made for each ECG, but there are only about five or six major classification groups for which the ECG is used. The first step in analyzing an ECG requires determination of the rate and rhythm for the atria and ventricles. Included here would be any conduction disturbances either in the relationship between the various chambers or within the chambers themselves. Then one would proceed to identify features that would relate to the presence or absence of scarring due to a myocardial infarction. There also may be evidence of acute events that would occur with ischemia or an evolving myocardial infarction. The ECG has been a primary tool for evaluating chamber size or enlargement, but one might argue that more accurate information in this area would be supplied by noninvasive imaging technologies. © 2000 by CRC Press LLC

FIGURE 13.3 The ECG depicts numerous measurements that can be made with computer-based algorithms. These are primarily durations, amplitudes, and areas. (Courtesy of Hewlett Packard Company, Palo Alto, Calif.)

FIGURE 13.4 This is an example of an interpreted 12-lead ECG. A 2-1/2-s recording is shown for each of the 12 leads. The bottom trace is a continuous 10-s rhythm strip of lead II. Patient information is given in the top area, below which is printed the computerized interpretive statements (Courtesy of the Hewlett Packard Company, Palo Alto, Calif.)


More recently, the high-resolution ECG has been developed, whereby the digitized ECG is signalaveraged to reduce random noise [9,10]. This approach, coupled with postaveraging high-pass filtering, is used to detect and quantify low-level signals (~1.0 µV) not detectable with standard approaches. This computer-based approach has enabled the recording of events that are predictive of future life-threatening cardiac events [11,12].

13.1 Physiology The heart has four chambers; the upper two chambers are called the atria, and the lower two chambers are called the ventricles. The atria are thin-walled, low-pressure pumps that receive blood from the venous circulation. Located in the top right atrium are a group of cells that act as the primary pacemaker of the heart. Through a complex change of ionic concentration across the cell membranes (the current source), an extracellular potential field is established which then excites neighboring cells, and a cell-to-cell propagation of electrical events occurs. Because the body acts as a purely resistive medium, these potential fields extend to the body surface [13]. The character of the body surface waves depends on the amount of tissue activating at one time and the relative speed and direction of the activation wavefront. Therefore, the pacemaker potentials that are generated by a small tissue mass are not seen on the ECG. As the activation wavefront encounters the increased mass of atrial muscle, the initiation of electrical activity is observed on the body surface, and the first ECG wave of the cardiac cycle is seen. This is the P wave, and it represents activation of the atria. Conduction of the cardiac impulse proceeds from the atria through a series of specialized cardiac cells (the A-V node and the His-Purkinje system) which again are too small in total mass to generate a signal large enough to be seen on the standard ECG. There is a short, relatively isoelectric segment following the P wave. Once the large muscle mass of the ventricles is excited, a rapid and large deflection is seen on the body surface. The excitation of the ventricles causes them to contract and provides the main force for circulating blood to the organs of the body. This large wave appears to have several components. The initial downward deflection is the Q wave, the initial upward deflection is the R wave, and the terminal downward deflection is the S wave. The polarity and actual presence of these three components depend on the position of the leads on the body as well as a multitude of abnormalities that may exist. In general, the large ventricular waveform is generically called the QRS complex regardless of its makeup. Following the QRS complex is another short relatively isoelectric segment. After this short segment, the ventricles return to their electrical resting state, and a wave of repolarization is seen as a low-frequency signal known as the T wave. In some individuals, a small peak occurs at the end or after the T wave and is the U wave. Its origin has never been fully established, but it is believed to be a repolarization potential.

13.2 Instrumentation The general instrumentation requirements for the ECG have been addressed by professional societies through the years [5,6]. Briefly, they recommend a system bandwidth between 0.05 and 150 Hz. Of great importance in ECG diagnosis is the low-frequency response of the system, because shifts in some of the low-frequency regions, e.g., the ST segment, have critical diagnosis value. While the heart rate may only have a 1-Hz fundamental frequency, the phase responses of typical analog high-pass filters are such that the system corner frequency must be much smaller than the 3-dB corner frequency where only the amplitude response is considered. The system gain depends on the total system design. The typical ECG amplitude is ±2 mV, and if A/D conversion is used in a digital system, the enough gain to span the full range of the A/D converter is appropriate. To first obtain an ECG the patient must be physically connected to the amplifier front end. The patientamplifier interface is formed by a special bioelectrode that converts the ionic current flow of the body to the electron flow of the metallic wire. These electrodes typically rely on a chemical paste or gel with


a high ionic concentration. This acts as the transducer at the tissue-electrode interface. For short-term applications, silver-coated suction electrodes or “sticky” metallic foil electrodes are used. Long-term recordings, such as for the monitored patient, require a stable electrode-tissue interface, and special adhesive tape material surrounds the gel and an Ag+/Ag+Cl– electrode. At any given time, the patient may be connected to a variety of devices, e.g., respirator, blood pressure monitor, temporary pacemaker, etc., some of which will invade the body and provide a low-resistance pathway to the heart. It is essential that the device not act as a current source and inject the patient with enough current to stimulate the heart and cause it to fibrillate. Some bias currents are unavoidable for the system input stage, and recommendations are that these leakage currents be less than 10 µA per device. This applies to the normal setting, but if a fault condition arises whereby the patient comes in contact with the high-voltage side of the alternating current (ac) power lines, then the isolation must be adequate to prevent 10 µA of fault current as well. This mandates that the ECG reference ground not be connected physically to the low side of the ac power line or its third-wire ground. For ECG machines, the solution has typically been to AM modulate a medium frequency carrier signal (≈400 kHz) and use an isolation transformer with subsequent demodulation. Other methods of signal isolation can be used, but the primary reason for the isolation is to keep the patient from being part of the ac circuit in the case of a patient-to-power-line fault. In addition, with many devices connected in a patient monitoring situation, it is possible that ground loop currents will be generated. To obviate this potential hazard, a low-impedance ground buss is often installed in these rooms, and each device chassis will have an external ground wire connected to the buss. Another unique feature of these amplifiers is that they must be able to withstand the high-energy discharge of a cardiac defibrillator. Older-style ECG machines recorded one lead at a time and then evolved to three simultaneous leads. This necessitated the use of switching circuits as well as analog weighting circuits to generate the various 12 leads. This is usually eliminated in modern digital systems by using an individual single-ended amplifier for each electrode on the body. Each potential signal is then digitally converted, and all the ECG leads can be formed mathematically in software. This would necessitate a 9-amplifier system. By performing some of the lead calculations with the analog differential amplifiers, this can be reduced to an 8-channel system. Thus only the individual chest leads V1 through V6 and any 2 of the limb leads, e.g., I and III, are needed to calculate the full 12-lead ECG. Figure 13.5 is a block diagram of a modern digital based ECG system. This system uses up to 13 single-ended amplifiers and a 16-bit A/D converter, all within a small lead wire manifold or amplifier lead stage. The digital signals are optically isolated and sent via a high-speed serial link to the main ECG instrument. Here the 32-bit CPU and DSP chip perform all the calculations, and a hard-copy report is generated (Fig. 13.4). Notice that each functional block has its own controller and that the system requires a real-time, multitasking operating system to coordinate all system functions. Concomitant with the data acquisition is the automatic interpretation of the ECG. These programs are quite sophisticated and are continually evolving. It is still a medical/legal requirement that these ECGs be overread by the physician.

13.3 Applications Besides the standard 12-lead ECG, there are several other uses of ECG recording technology that rely on only a few leads. These applications have had a significant clinical and commercial impact. Following are brief descriptions of several ECG applications which are aimed at introducing the reader to some of the many uses of the ECG.

The Ambulatory ECG The ambulatory or Holter ECG has an interesting history, and its evolution closely followed both technical and clinical progress. The original, analog, tape-based, portable ECG resembled a fully loaded backpack and was developed by Dr. Holter in the early 1960s [14]. It was soon followed by more compact devices © 2000 by CRC Press LLC

FIGURE 13.5 This is a block diagram of microprocessor-based ECG system. It includes all the elements of a personal computer class system, e.g., 80386 processor, 2 Mbytes of RAM, disk drive, 640 × 480 pixel LCD display, and battery operable. In addition, it includes a DSP56001 chip and multiple controllers which are managed with a real-time multitasking operating system. (Courtesy of the Hewlett Packard Company, Palo Alto, Calif.)

that could be worn on the belt. The original large-scale clinical use of this technology was to identify patients who developed heart block transiently and could be treated by implanting a cardiac pacemaker. This required the secondary development of a device that could rapidly play back the 24 hours of taperecorded ECG signals and present to the technician or physician a means of identifying periods of time where the patient’s heart rate became abnormally low. The scanners had the circuitry not only to play back the ECG at speeds 30 to 60 times real time but also to detect the beats and display them in a superimposed mode on a CRT screen. In addition, an audible tachometer could be used to identify the periods of low heart rate. With this playback capability came numerous other observations, such as the identification of premature ventricular complexes (PVCs), which lead to the development of techniques to identify and quantify their number. Together with the development of antiarrhythmic drugs, a marriage was formed between pharmaceutical therapy and the diagnostic tool for quantifying PVCs. ECG tapes were recorded before and after drug administration, and drug efficacy was measured by the reduction of the number of PVCs. The scanner technology for detecting and quantifying these arrhythmias was originally implemented with analog hardware but soon advanced to computer technology as it became economically feasible. Very sophisticated algorithms were developed based on pattern-recognition techniques and were sometimes implemented with high-speed specialized numerical processors as the tape playback speeds became several hundred times real time [15]. Unfortunately, this approach using the ambulatory ECG for identifying and treating cardiac arrhythmias has been on the decline because the rationale of PVC suppression was found to be unsuccessful for improving cardiac mortality. However, the ambulatory ECG is still a widely used diagnostic tool, and modern units often have built-in microprocessors with considerable amounts of random access memory and even small disk drives with capacities greater than 400 Mbytes. Here the data can be analyzed on-line, with large segments of data selected for storage and later analysis with personal computer-based programs. © 2000 by CRC Press LLC

Patient Monitoring The techniques for monitoring the ECG in real time were developed in conjunction with the concept of the coronary care unit (CCU). Patients were placed in these specialized hospital units to carefully observe their progress during an acute illness such as a myocardial infarction or after complex surgical procedures. As the number of beds increased in these units, it became clear that the highly trained medical staff could not continually watch a monitor screen, and computerized techniques were added that monitored the patient’s rhythm. These programs were not unlike those developed for the ambulatory ECG, and the high-speed numerical capability of the computer was not taxed by monitoring a single ECG. The typical CCU would have 8 to 16 beds, and hence the computing power was taken to its limit by monitoring multiple beds. The modern units have the CPU distributed within the ECG module at the bedside, along with modules for measuring many other physiologic parameters. Each bedside monitor would be interconnected with a high-speed digital line, e.g., Ethernet, to a centralized computer used primarily to control communications and maintain a patient database.

High-Resolution ECG High-resolution (HR) capability is now a standard feature on most digitally based ECG systems or as a stand-alone microprocessor-based unit [16]. The most common application of the HRECG is to record very low level (~1.0-µV) signals that occur after the QRS complex but are not evident on the standard ECG. These “late potentials” are generated from abnormal regions of the ventricles and have been strongly associated with the substrate responsible for a life-threatening rapid heart rate (ventricular tachycardia). The typical HRECG is derived from 3 bipolar leads configured in an anatomic xyz coordinate system. These 3 ECG signals are then digitized at a rate of 1000 to 2000 Hz per channel, time aligned via a realtime QRS correlator, and summated in the form of a signal average. Signal averaging will theoretically improve the signal-to-noise ratio by the square root of the number of beats averaged. The underlying assumptions are that the signals of interest do not vary, on a beat-to-beat basis, and that the noise is random. Figure 13.6 has four panels depicting the most common sequence for processing the HRECG to measure the late potentials. Panel a depicts a 3-second recording of the xyz leads close to normal resolution. Panel b was obtained after averaging 200 beats and with a sampling frequency of 10 times that shown in panel a. The gain is also 5 times greater. Panel c is the high-pass filtered signal using a partially time-reversed digital filter having a second-order Butterworth response and a 3-dB corner frequency of 40 Hz [12]. Note the appearance of the signals at the terminal portion of the QRS complex. A common method of analysis, but necessarily optimal, is to combine the filtered xyz leads into a vector magnitude, that is, (X2 + Y2 + Z2)1/2. This waveform is shown in panel d. From this waveform, several parameters have been derived such as total QRS duration (including late potentials), the RMS voltage value of the terminal 40 ms, and the low-amplitude signal (LAS) duration from the 40-µV level to the end of the late potentials. Abnormal values for these parameters are used to identify patients at high risk of ventricular tachycardia following a heart attack.

13.4 Conclusions The ECG is one of the oldest instrument-bound measurements in medicine. It has faithfully followed the progression of instrumentation technology. Its most recent evolutionary step, to the microprocessorbased system, has allowed patients to wear their computer monitor or has provided an enhanced, highresolution ECG that has opened new vistas of ECG analysis and interpretation.


FIGURE 13.6 The signal-processing steps typically performed to obtain a high-resolution ECG are shown in panels A through D. See text for a full description.

References 1. Waller AD. One of the electromotive changes connected with the beat of the mammalian heart, and the human heart in particular. Phil Trans B 180:169, 1889. 2. Einthoven W. Die galvanometrische Registrirung des menschlichen Elektrokardiogramms, zugleich eine Beurtheilung der Anwendung des Capillar-Elecktrometers in der Physiologie. Pflugers Arch Ges Physiol 99:472, 1903. 3. Wilson FN, Johnson FS, Hill IGW. The interpretation of the falvanometric curves obtained when one electrode is distant from the heart and the other near or in contact with the ventricular surface. Am Heart J 10:176, 1934. 4. Goldberger E. A simple, indifferent, electrocardiographic electrode of zero potential and a technique of obtaining augmented, unipolar, extremity leads. Am Heart J 23:483, 1942. 5. Voluntary standard for diagnostic electrocardiographic devices. ANSI/AAMI EC11a. Arlington, Va: Association for the Advancement of Medical Instrumentation, 1984. 6. Bailey JJ, Berson AS, Garson A, et al. Recommendations for standardization and specifications in automated electrocardiography: bandwidth and digital signal processing: A report for health professionals by an ad hoc writing group of the committee on electrocardiography and cardiac electrophysiology of the Council on Clinical Cardiology, American Heart Association. Circulation 81:730, 1990.


7. Jenkins JM. Computerized electrocardiography. CRC Crit Rev Bioeng 6:307, 1981. 8. Pryor TA, Drazen E, Laks M (eds). Computer Systems for the Processing of diagnostic electrocardiograms. Los Alamitos, Calif, IEEE Computer Society Press, 1980. 9. Berbari EJ, Lazzara R, Samet P, Scherlag BJ. Noninvasive technique for detection of electrical activity during the PR segment. Circulation 48:1006, 1973. 10. Berbari EJ, Lazzara R, Scherlag BJ. A computerized technique to record new components of the electrocardiogram. Proc IEEE 65:799, 1977. 11. Berbari EJ, Scherlag BJ, Hope RR, Lazzara R. Recording from the body surface of arrhythmogenic ventricular activity during the ST segment. Am J Cardiol 41:697, 1978. 12. Simson MB. Use of signals in the terminal QRS complex to identify patients with ventricular tachycardia after myocardial infarction. Circulation 64:235, 1981. 13. Geselowitz DB. On the theory of the electrocardiogram. Proc IEEE 77:857, 1989. 14. Holter NJ. New method for heart studies: Continuous electrocardiography of active subjects over long periods is now practical. Science 134:1214, 1961. 15. Ripley KL, Murray A (eds). Introduction to Automated Arrhythmia Detection. Los Alamitos, Calif, IEEE Computer Society Press, 1980. 16. Berbari EJ. High-resolution electrocardiography. CRC Crit Rev Bioeng 16:67, 1988.


Henneberg, K. “Principles of Electromyography.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

14 Principles of Electromyography

Kaj-Åge Henneberg University of Montreal

14.1 14.2 14.3

The Structure and Function of Muscle The Origin of Electromyograms Electromyographic Recordings Concentric Electrode EMG • Single-Fiber EMG • Macro EMG

Movement and position of limbs are controlled by electrical signals traveling back and forth between the muscles and the peripheral and central nervous system. When pathologic conditions arise in the motor system, whether in the spinal cord, the motor neurons, the muscle, or the neuromuscular junctions, the characteristics of the electrical signals in the muscle change. Careful registration and study of electrical signals in muscle (electromyograms) can thus be a valuable aid in discovering and diagnosing abnormalities not only in the muscles but also in the motor system as a whole. Electromyography (EMG) is the registration and interpretation of these muscle action potentials. Until recently, electromyograms were recorded primarily for exploratory or diagnostic purposes; however, with the advancement of bioelectric technology, electromyograms also have become a fundamental tool in achieving artificial control of limb movement, i.e., functional electrical stimulation (FES) and rehabilitation. This chapter will focus on the diagnostic application of electromyograms, while FES will be discussed in Chapter 17. Since the rise of modern clinical EMG, the technical procedures used in recording and analyzing electromyograms have been dictated by the available technology. The concentric needle electrode introduced by Adrian and Bronk in 1929 provided an easy-to-use electrode with high mechanical qualities and stable, reproducible measurements. Replacement of galvanometers with high-gain amplifiers allowed smaller electrodes with higher impedances to be used and potentials of smaller amplitudes to be recorded. With these technical achievements, clinical EMG soon evolved into a highly specialized field where electromyographists with many years of experience read and interpreted long paper EMG records based on the visual appearance of the electromyograms. Slowly, a more quantitative approach emerged, where features such as potential duration, peak-to-peak amplitude, and number of phases were measured on the paper records and compared with a set of normal data gathered from healthy subjects of all ages. In the last decade, the general-purpose rack-mounted equipment of the past have been replaced by ergonomically designed EMG units with integrated computers. Electromyograms are digitized, processed, stored on removable media, and displayed on computer monitors with screen layouts that change in accordance with the type of recording and analysis chosen by the investigator. With this in mind, this chapter provides an introduction to the basic concepts of clinical EMG, a review of basic anatomy, the origin of the electromyogram, and some of the main recording procedures and signal-analysis techniques in use.


FIGURE 14.1

Schematic illustration of different types of muscles: (a) fusiform, (b) unipennate, and (c) bipennate.

14.1 The Structure and Function of Muscle Muscles account for about 40% of the human mass, ranging from the small extraocular muscles that turn the eyeball in its socket to the large limb muscles that produce locomotion and control posture. The design of muscles varies depending on the range of motion and the force exerted (Fig. 14.1). In the most simple arrangement (fusiform), parallel fibers extend the full length of the muscle and attach to tendons at both ends. Muscles producing a large force have a more complicated structure in which many short muscle fibers attach to a flat tendon that extends over a large fraction of the muscle. This arrangement (unipennate) increases the cross-sectional area and thus the contractile force of the muscle. When muscle fibers fan out from both sides of the tendon, the muscle structure is referred to as bipennate. A lipid bilayer (sarcolemma) encloses the muscle fiber and separates the intracellular myoplasma from the interstitial fluid. Between neighboring fibers runs a layer of connective tissue, the endomysium, composed mainly of collagen and elastin. Bundles of fibers, fascicles, are held together by a thicker layer of connective-tissue called the perimysium. The whole muscle is wrapped in a layer of connective tissue called the epimysium. The connective tissue is continuous with the tendons attaching the muscle to the skeleton. In the myoplasma, thin and thick filaments interdigitate and form short, serially connected identical units called sarcomeres. Numerous sarcomeres connect end to end, thereby forming longitudinal strands of myofibrils that extend the entire length of the muscle fiber. The total shortening of a muscle during contraction is the net effect of all sarcomeres shortening in series simultaneously. The individual sarcomeres shorten by forming cross-bridges between the thick and thin filaments. The cross-bridges pull the filaments toward each other, thereby increasing the amount of longitudinal overlap between the thick and thin filaments. The dense matrix of myofibrils is held in place by a structural framework of intermediate filaments composed of desmin, vimetin, and synemin [Squire, 1986]. At the site of the neuromuscular junction, each motor neuron forms collateral sprouts (Fig. 14.2) and innervates several muscle fibers distributed almost evenly within an elliptical or circular region ranging from 2 to 10 mm in diameter. The motor neuron and the muscle fibers it innervates constitute a functional unit, the motor unit. The cross section of muscle occupied by a motor unit is called the motor unit territory (MUT). A typical muscle fiber is only innervated at a single point, located within a crosssectional band referred to as the end-plate zone. While the width of the end-plate zone is only a few millimeters, the zone itself may extend over a significant part of the muscle. The number of muscle fibers per motor neuron (i.e., the innervation ratio) ranges from 3:1 in extrinsic eye muscles where fine-graded contraction is required to 120:1 in some limb muscles with coarse movement [Kimura, 1981]. The fibers © 2000 by CRC Press LLC

FIGURE 14.2 Innervation of muscle fibers. (a) Two normal motor units with intermingled muscle fibers. (b) Reinnervation of muscle fibers. The second and fourth muscle fibers have lost their motor neuron (2) and subsequently have become reinnervated by newly formed sprouts from the motor neuron (1) innervating adjacent muscle fibers. Not drawn to scale.

of one motor unit are intermingled with fibers of other motor units; thus several motor units reside within a given cross section. The fibers of the same motor unit are thought to be randomly or evenly distributed within the motor unit territory; however, reinnervation of denervated fibers often results in the formation of fiber clusters (see Fig. 14.2).

14.2 The Origin of Electromyograms Unlike the myocardium, skeletal muscles do not contain pacemaker cells from which excitations arise and spread. Electrical excitation of skeletal muscle is initiated and regulated by the central and peripheral nervous systems. Motor neurons carry nerve impulses from the anterior horn cells of the spinal cord to the nerve endings, where the axonal action potential triggers the release of the neurotransmitter acetylcholine (Ach) into the narrow clefts separating the sarcolemma from the axon terminals. As Ach binds to the sarcolemma, Ach-sensitive sodium channels open, and miniature end-plate potentials arise in the sarcolemma. If sufficient Ach is released, the summation of miniature end-plate potentials, i.e., the endplate potential, reaches the excitation threshold, and sarcolemma action potentials propagate in opposite directions toward the tendons. As excitation propagates down the fiber, it spreads into a highly branched transverse network of tubules (T system) which interpenetrate the myofibrils. The effective radial conduction velocity (~4 cm/s) is about two orders of magnitude slower than the longitudinal conduction velocity (2 to 5 m/s). This is due to the fact that the main portion of the total membrane capacitance is located in the T system and that the lumen of the T system constitutes a higher electrical resistance than the myoplasma. The slower tubular conduction velocity implies an increasingly delayed onset of tubular action potentials toward the center of the fiber relative to that of the sarcolemmal action potential (Fig. 14.3a). However, compared with the time course of the subsequent contraction, the spread of excitation along and within the muscle fiber is essentially instantaneous, thereby ensuring simultaneous release of calcium from the sarcoplasmic reticulum throughout the entire volume of the muscle. If calcium release were restricted to a small longitudinal section of the muscle fiber, only sarcomeres in this region would contract, and sarcomeres in the rest of the fiber would stretch accordingly. Similarly, experiments in detubulated muscle fibers, i.e., fibers in which the continuation between the sarcolemmal and the tubular membrane has been disrupted, have demonstrated that only a thin layer of superficial myofibrils contracts when tubular action potentials fail to trigger calcium release deep in the muscle fiber. It is well known that the shape of the skeletal muscle action potential differs from that of nerve action potentials with regard to the repolarization phase. In skeletal muscle, the falling phase of the action


FIGURE 14.3 Simulated sarcolemmal and tubular action potentials of frog sartorius muscle fiber. (a) Temporal membrane action potentials calculated in a transverse plane 5 mm from the end-plate zone of a fiber with radius a = 50 µm. Curve 1: Sarcolemmal action potential; curves 2 to 4: action potentials in tubular membrane patches at r = a (2), r = a/2 (3), and r = a/20 (4). (b) Sarcolemmal action potentials for fibers with radius 70 (1), 50 (2), and 30 (3) µm. The time axes have been expanded and truncated.

potential is interrupted by a long, slowly decaying potential known as the afterpotential. This late potential is caused by two opposing forces, the repolarization force due to the efflux of potassium through the sarcolemma and a depolarization force due to an influx of current from the interstitial space into the tubular lumen. The latter current is drawn in through the tubular openings by the repolarizing tubular membrane. Large muscle fibers have a higher tubular-sarcolemma area ratio; hence the inward surge of tubular current increases with fiber size. Figure 14.3b illustrates this by comparing sarcolemmal action potentials for fibers of increasing diameter. The small fiber has only a small amount of tubular membrane; hence the sarcolemmal potassium current has sufficient strength to rapidly repolarize the membrane. For the large fiber, inward tubular current actually depolarizes the sarcolemma slightly during repolarization of the tubular membrane system, thereby producing a small hump on the afterpotential. Since the hump on the afterpotential is influenced primarily by fiber size, this feature is more typical for large frog fibers than for smaller human fibers. Experiments have demonstrated that the sarcolemma action potential of detubulated fibers hyperpolarizes in a manner similar to that of a nerve action potential. In Fig. 14.4a, the time course of the sarcolemmal current density is compared with that of the current passing through the tubular mouth during the time course of the sarcolemmal action potential. The positive (outward) peak of the tubular current overlaps in time with the small capacitive sarcolemmal

FIGURE 14.4 Simulated currents and extracellular potentials of frog sartorius muscle fiber (radius a = 50 µm). (a) The net fiber current density is the summation of the current density through the sarcolemma and that passing the tubular mouth. (b) Extracellular action potentials calculated at increasing radial distances (in units of fiber radius) using a bidomain volume conductor model and the net current source in panel a. The time axes have been expanded and truncated.


FIGURE 14.5 Schematic cross-sectional view of overlapping single-fiber action potential distributions in a normal (a) and reinnervated (b) motor unit. Muscle fibers are represented by filled black circles. Concentric rings represent axisymmetric isopotential lines of the individual single-fiber potentials at 2 (gray) and 5 radii from the fiber center, respectively. Compare with radial decline of extracellular potentials in Fig. 14.4. See text for discussion of simplifying assumptions.

displacement current (initial positive peak) and the negative peak (inward sarcolemmal sodium current). As a result, the net current has a much larger positive peak than that of the sarcolemma alone, and the negative peak of the net current is only about half that of the sarcolemmal current. The outward sarcolemmal potassium current (late positive phase) is almost completely opposed by an antisymmetric inward tubular current, i.e., the current drawn into the tubular lumen by the repolarizing T system. The combined effect of the sarcolemmal and tubular interaction is a net current with an almost biphasic waveform and with similar amplitudes of the positive and negative peaks. As the net current source propagates toward the tendon, an extracellular potential field arises and follows the action potential complex. At a given field point, this phenomenon is observed as a temporal potential waveform (Fig. 14.4b); however, a more complete picture of the phenomenon is obtained by considering the spatial distribution of potentials in the cross section of the motor unit. Figure 14.5 shows schematically how the concentric isopotential lines of individual fibers overlap with those of other fibers in the same motor unit. In a typical healthy motor unit (Fig. 14.5a), the mean interfiber distance is on the order of a few hundred microns. Taking into account the steep radial decline of the potential amplitudes illustrated in Fig. 14.4b, it is evident that single-fiber action potential (SFAP) overlapping occurs between low-magnitude isopotential lines. Figure 14.5b illustrates how spatial overlapping between SFAPs might look like in a motor unit with extensive fiber grouping. In this case, higher-level isopotential lines overlap within the clusters, while regions with no fibers would appear as electrically silent. Several factors ignored in Fig. 14.5 need further discussion. All fibers are assumed to be of the same size and with identical, perfectly axisymmetric potential distributions. The net single-fiber current source is an increasing function of fiber size; thus the magnitude of the potential distribution will vary with varying fiber size. Fibers can to a good approximation be considered as constant current sources; hence if the resistivity of the muscle tissue increases, e.g., due to increased fiber packing density, the potential difference between an observation point and a reference point also will increase. It follows that local variations in fiber packing density in the region of an active fiber will destroy the axisymmetric appearance of its potential distribution. Muscle fibers are not perfect cylinders, and angular variation in the shape of the sarcolemma must be expected to create angular variations in the potential distribution. However, due to the relatively high conductivity of the volume conductor, it is plausible that such variations become increasingly insignificant as the distance to the fiber is increased.


A very important factor not considered in Fig. 14.5 concerns the degree of longitudinal alignment of SFAPs in the motor unit. As illustrated in Fig. 14.2, SFAPs are usually dispersed along the motor unit axis, and their potential distributions do not sum up in a simple manner. SFAPs can be misaligned for several reasons. The end plates are not aligned; hence some SFAPs originate ahead of others. Variations in fiber size and packing density cause variations in conduction velocity; thus the dispersion of SFAPs may actually increase with increasing distance from the end-plate zone. Neuropathologic conditions can affect the alignment of SFAPs. Complete or partial failure of collateral nerve sprouts to conduct nerve action potentials can abolish or delay the onset of action potentials, and immature sprouts formed during reinnervation may cause significant variability (jitter) in the onset of muscle action potentials. Denervated muscle fibers shrink; hence a newly reinnervated muscle fiber is likely to have a slower conduction velocity. An increased dispersion of SFAPs creates a very complex and irregular spatial potential distribution. The superimposed temporal potential waveform in a fixed observation point, i.e., the motor unit potential (MUP), is therefore comprised of several peaks rather than a simple bi- or triphasic waveform. MUPs with five or more peaks are classified as polyphasic. A satellite potential is an SFAP that is so misaligned from the main potential complex that its waveform is completely separated from the MUP.

14.3 Electromyographic Recordings A considerable amount of information regarding the bioelectrical state of a muscle is hidden in the timevarying spatial distribution of potentials in the muscle. Unfortunately, it is not clinically feasible to obtain high-resolution three-dimensional samples of the spatial potential distribution, since this would require the insertion of hundreds of electrodes into the muscles. In order to minimize the discomfort of the patient, routine EMG procedures usually employ only a single electrode that is inserted into different regions of the muscle. As the SFAPs of an active motor unit pass by the electrode, only their summation, i.e., the MUP, will be registered by the electrode. The electrode is effectively integrating out the spatial information hidden in the passing potential complex, leaving only a time-variant potential waveform to be recorded and interpreted. It goes without saying that such a constraint on the recording procedure puts the electromyographist at a considerable disadvantage. To partially circumvent this setback, technical innovations and new procedures have continued to refine EMG examinations to such a level that some of the spatial information originally obscured by the electrode can be extracted intuitively from the temporal waveforms by an experienced electromyographist. With a detailed understanding of the bioelectric principles involved, e.g., the bioelectric sources, the volume conductor, and the recording properties of the electrode, the electromyographist can quickly recognize and explain the waveform characteristics associated with various neuromuscular abnormalities. To increase the amount of diagnostic information, several sets of EMG investigations may be performed using electrodes with different recording characteristics. Figure 14.6 illustrates three of the most popular EMG needle electrodes. The concentric and monopolar electrodes have an intermediate pickup range and are used in conventional recordings. The single-fiber electrode is a more recent innovation. It has a very small pickup range and is used to obtain recordings from only one or two muscle fibers. The macro electrode, which is the cannula of either the concentric or single-fiber electrode in combination with a remote reference electrode, picks up potentials throughout the motor unit territory. This section will review these EMG electrodes, the waveforms they produce, and the signal analysis performed in each case.

Concentric Electrode EMG Adrian and Bronk developed the concentric electrode (see Fig. 14.6) in order to obtain a pickup range that is smaller than that of wire electrodes. The modern version of the concentric electrode consists of a platinum or stainless steel wire located inside the lumen of a stainless steel cannula with an outer diameter of about 0.5 mm. The tip is beveled at 15 to 20 degrees, thereby exposing the central wire as an oblique elliptical surface of about 150 × 580 µm. The central wire is insulated from the cannula with araldite or epoxy. © 2000 by CRC Press LLC

FIGURE 14.6 Needle electrodes for subcutaneous EMG recordings. For the single-fiber and concentric electrodes, the cannula of the hypodermic needle acts as reference electrode. The monopolar electrode is used with a remote reference electrode.

The concentric electrode is connected to a differential amplifier; thus common-mode signals are effectively rejected, and a relatively stable baseline is achieved. The cannula cannot be regarded as an indifferent reference electrode because it is located within the potential distribution and thus will pick up potentials from active fibers. Simulation studies [Henneberg and Plonsey, 1993] have shown that the cannula shields the central wire from picking up potentials from fibers located behind the tip. The sensitivity of the concentric electrode is therefore largest in the hemisphere facing the oblique elliptical surface. Due to the asymmetric sensitivity function, the waveshape of the recorded potentials will vary if the electrode is rotated about its axis. This problem is not observed with the axisymmetric monopolar electrode; however, this electrode has a more unstable baseline due to the remote location of the reference electrode. Both the concentric and monopolar electrodes (see Fig. 14.6) are used in conventional EMG. Because of the differences in recording characteristics, however, concentric and monopolar recordings cannot be compared easily. A particular EMG laboratory therefore tends to use the one and not the other. During the concentric needle examination, the investigator searches for abnormal insertional activity, spontaneous activity in relaxed muscles, and motor unit potentials with abnormal appearance. The waveshape of motor unit potentials is assessed on the basis of the quantitative waveform features defined in Fig. 14.7: • Amplitude is determined by the presence of active fibers within the immediate vicinity of the electrode tip. Low-pass filtering by the volume conductor attenuates the high-frequency spikes of remote SFAPs; hence the MUP amplitude does not increase for a larger motor unit. However, MUP amplitude will increase if the tip of the electrode is located near a cluster of reinnervated fibers. Large MUP amplitudes are frequently observed in neurogenic diseases. • Rise time is an increasing function of the distance between the electrode and the closest active muscle fiber. A short rise time in combination with a small MUP amplitude might therefore indicate that the amplitude is reduced due to fiber atrophy rather than to a large distance between the electrode and the closest fiber. • Number of phases indicates the complexity of the MUP and the degree of misalignment between SFAPs. In neurogenic diseases, polyphasic MUPs arise due to slow conduction velocity in immature nerve sprouts or slow conduction velocity in reinnervated but still atrophied muscle fibers. Variation in muscle fiber size also causes polyphasic MUPs in myopathic diseases. To prevent noisy baseline fluctuations from affecting the count of MUP phases, a valid baseline crossing must exceed a minimum absolute amplitude criterion.


FIGURE 14.7 Definition of quantitative waveform features of MUPs recorded with a concentric EMG electrode. MUP area is defined as the rectified area under the curve in the interval between the onset and end. The number of MUP phases is defined as the number of baseline crossings () plus one. MUP turns are marked by a . MUP rise time is the time interval between the 10% and 90% deflection () of the main negative-going slope. As by convention, negative voltage is displayed upward.

• Duration is the time interval between the first and last occurrence of the waveform exceeding a predefined amplitude threshold, e.g., 5 µV. The MUP onset and end are the summation of lowfrequency components of SFAPs scattered over the entire pickup range of the electrode. As a result, the MUP duration provides information about the number of active fibers within the pickup range. However, since the motor unit territory can be larger than the pickup range of the electrode, MUP duration does not provide information about the total size of a large motor unit. MUP duration will increase if a motor unit has an increased number of fibers due to reinnervation. MUP duration is affected to a lesser degree by SFAP misalignment. • Area indicates the number of fibers adjacent to the electrode; however, unlike MUP amplitude, MUP area depends on MUP duration and is therefore influenced by fibers in a larger region compared with that of MUP amplitude. • Turns is a measure of the complexity of the MUP, much like the number of phases; however, since a valid turn does not require a baseline crossing like a valid phase, the number of turns is more sensitive to changes in the MUP waveshape. In order to distinguish valid turns from signal noise, successive turns must be offset by a minimum amplitude difference. Based on the complimentary information contained in the MUP features defined above, it is possible to infer about the number and density of fibers in a motor unit as well as the synchronicity of the SFAPs. However, the concentric electrode is not sufficiently selective to study individual fibers, nor is it sufficiently sensitive to measure the total size of a motor unit. The following two techniques were designed with these objectives in mind.

Single-Fiber EMG The positive lead of the single-fiber electrode (see Fig. 14.6) is the end cap of a 25-µm wire exposed through a side port on the cannula of a steel needle. Due to the small size of the positive lead, bioelectric sources, which are located more than about 300 µm from the side port, will appear as common-mode signals and be suppressed by the differential amplifier. To further enhance the selectivity, the recorded signal is high-pass filtered at 500 Hz to remove low-frequency background activity from distant fibers. Due to its very small pickup range, the single-fiber electrode rarely records potentials from more than one or two fibers from the same motor unit. Because of the close proximity of the fibers, potentials are of large amplitudes and with small rise times. When two potentials from the same motor unit are picked up, the slight variation in their interpotential interval (IPI) can be measured (Fig. 14.8). The mean IPI


FIGURE 14.8 Measurement of interpotential interval (IPI) between single-fiber potentials recorded simultaneously from two fibers of the same motor unit.

(jitter) is normally 5 to 50 µs but increases when neuromuscular transmission is disturbed. When the single-fiber electrode records potentials from an increased number of fibers, it usually indicates that the side port is close to either a cluster of fibers (reinnervation) or that the positive lead is close to fibers in the process of splitting.

Macro EMG For this electrode, the cannula of a single-fiber or concentric electrode is used as the positive lead, while the reference electrode can be either a remote subcutaneous or remote surface electrode. Due to the large lead surface, this electrode picks up both near- and far-field activity. However, the signal has very small amplitude, and the macro electrode must therefore be coupled to an electronic averager. To ensure that only one and the same MUP is being averaged, the averager is triggered by a SFAP picked up from that motor unit by the side port wire of the single-fiber electrode or by the central wire of the concentric electrode. Since other MUPs are not time-locked to the triggering SFAP, they will appear as random background activity and become suppressed in the averaging procedure. Quantitative features of the macro MUP include the peak-to-peak amplitude, the rectified area under the curve, and the number of phases.

Defining Terms Concentric electrode EMG: Registration and interpretation of motor unit potentials recorded with a concentric needle electrode. Electromyograms (EMGs): Bioelectric potentials recorded in muscles. Jitter: Mean variation in interpotential interval between single-fiber action potentials of the same motor unit. Macro EMG: The registration of motor unit potentials from the entire motor unit using the cannula of the single-fiber or concentric electrode. Motor unit: The functional unit of an anterior horn cell, its axon, the neuromuscular junctions, and the muscle fibers innervated by the motor neuron. Motor unit potential (MUP): Spatial and temporal summation of all single-fiber potentials innervated by the same motor neuron. Also referred to as the motor unit action potential (MUAP). Motor unit territory (MUT): Cross-sectional region of muscle containing all fibers innervated by a single motor neuron. Satellite potential: An isolated single-fiber action potential that is time-locked with the main MUP. Single-fiber action potential (SFAP): Extracellular potential generated by the extracellular current flow associated with the action potential of a single muscle fiber. Single-fiber EMG (SFEMG): Recording and analysis of single-fiber potentials with the single-fiber electrode.


References Henneberg K, Plonsey R. 1993. Boundary element analysis of the directional sensitivity of the concentric EMG electrode. IEEE Trans Biomed Eng 40:621. Kimura J. 1981. Electrodiagnosis in Diseases of Nerve and Muscle: Principles and Practice. Philadelphia, FA Davis. Squire J. 1986. Muscle: Design, Diversity and Disease. Menlo Park, Calif, Benjamin/Cummings.

Further Information Barry DT. 1991. AAEM minimonograph no. 36: Basic concepts of electricity and electronics in clinical electromyography. Muscle Nerve. 14:937. Buchthal F. 1973. Electromyography. In Handbook of Electroencephalography and Clinical Neurophysiology, vol 16. Amsterdam, Elsevier Scientific. Daube JR. 1991. AAEM minimonograph no. 11: Needle examination in clinical electromyography. Muscle Nerve 14:685. Dumitru D, DeLisa JA. 1991. AAEM minimonograph no. 10: Volume conduction. Muscle Nerve 14:605. Stålberg E. 1986. Single fiber EMG, macro EMG, and scanning EMG: New ways of looking at the motor unit. CRC Crit Rev Clin Neurobiol. 2:125.


Bronzino, J. d. “Principles of Electroencephalography.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

15 Principles of Electroencephalography

Joseph D. Bronzino Trinity College/The Biomedical Engineering Alliance for Connecticut (BEACON)

15.1 15.2 15.3

Historical Perspective EEG Recording Techniques Use of Amplitude Histographs to Quantify the EEG

15.4 15.5

Frequency Analysis of the EEG Nonlinear Analysis of the EEG

Mean • Standard Amplitude • Skewness • Kurtosis

Electroencephalograms (EEGs) are recordings of the minute (generally less that 300 µV) electrical potentials produced by the brain. Since 1924, when Hans Berger reported the recording of rhythmic electrical activity from the human scalp, analysis of EEG activity has been conducted primarily in clinical settings to detect gross pathologies and epilepsies and in research facilities to quantify the central effects of new pharmacologic agents. As a result of these efforts, cortical EEG patterns have been shown to be modified by a wide range of variables, including biochemical, metabolic, circulatory, hormonal, neuroelectric, and behavioral factors. In the past, interpretation of the EEG was limited to visual inspection by an electroencephalographer, an individual trained to qualitatively distinguish normal EEG activity from localized or generalized abnormalities contained within relatively long EEG records. This approach left clinicians and researchers alike buried in a sea of EEG paper records. The advent of computers and the technologies associated with them has made it possible to effectively apply a host of methods to quantify EEG changes. With this in mind, this chapter provides a brief historical perspective followed by some insights regarding EEG recording procedures and an in-depth discussion of the quantitative techniques used to analyze alterations in the EEG.

15.1 Historical Perspective In 1875, Richard Caton published the first account documenting the recording of spontaneous brain electrical activity from the cerebral cortex of an experimental animal. The amplitude of these electrical oscillations was so low (i.e., in the microvolt range) that Caton’s discovery is all the more amazing because it was made 50 years before suitable electronic amplifiers became available. In 1924, Hans Berger, of the University of Jena in Austria, carried out the first human EEG recordings using metal strips pasted to the scalps of his subjects as electrodes and a sensitive galvanometer as the recording instrument. Berger was able to measure the irregular, relatively small electrical potentials (i.e., 50 to 100 µV) coming from the brain. By studying the successive positions of the moving element of the galvanometer recorded on a continuous roll of paper, he was able to observe the resultant patterns in these brain waves as they varied with time. From 1924 to 1938, Berger laid the foundation for many of the present applications of


electroencephalography. He was the first to use the word electroencephalogram in describing these brain potentials in humans. Berger also noted that these brain waves were not entirely random but instead displayed certain periodicities and regularities. For example, he observed that although these brain waves were slow (i.e., exhibited a synchronized pattern of high amplitude and low frequency, standard amplitude N

(15.5)

A normal distribution will have a value of 0.5 for this measures. Figure 15.1 illustrates the sensitivity of these measures in analyzing the effect of systemic (IP) administration of morphine sulfate (30 mg/kg) on the cortical EEG. It will be noted that the skewness measure changes abruptly only immediately after the morphine injection, when the EEG was dominated by the appearance of spindles. However, the index of kurtosis characterizes the entire extent of the drug effect from onset to its return to baseline.

FIGURE 15.1 Plot of the indices of the amplitude distribution, i.e., the mean, standard amplitude, skewness, and kurtosis, of the EEG recorded from a rat prior to and for 3 hours following intraperitoneal injection of morphine sulfate (30 mg/kg). Arrow (inj) indicates time of injection. © 2000 by CRC Press LLC

The central moments of the EEG amplitude histogram, therefore, are capable of (1) characterizing the amplitude distributions of the EEG and (2) quantifying alterations in these electrical processes brought about by pharmacologic manipulations. In addition, use of the centile index for skewness and kurtosis provides a computer-efficient method for obtaining these measures in real-time.

15.4 Frequency Analysis of the EEG In early attempts to correlate the EEG with behavior, analog frequency analyzers were used to examine single channels of EEG data. Although disappointing, these initial efforts did introduce the use of frequency analysis to the study of gross brain wave activity. Although power spectral analysis, i.e., the magnitude square of the Fourier transform, provides a quantitative measure of the frequency distribution of the EEG, it does so, as mentioned above, at the expense of other details in the EEG such as the amplitude distribution and information concerning the presence of specific EEG patterns. The first systematic application of power spectral analysis by general-purpose computers was reported in 1963 by Walter; however, it was not until the introduction of the fast Fourier transform (FFT) by Cooley and Tukey in 1965 that machine computation of the EEG became commonplace. Although an individual FFT is ordinarily calculated for a short section of EEG data (e.g., from 1 to 8 s), such signal segmentation with subsequent averaging of individual modified periodograms has been shown to provide a consistent estimator of the power spectrum. An extension of this technique, the compressed spectral array, has been particularly useful for evaluating EEG spectra over long periods of time. A detailed review of the development and use of various methods to analyze the EEG is provided by Bronzino et al. [1984] and Givens and Remond [1987]. Figure 15.2 provides an overview of the computational proFIGURE 15.2 Block diagram illustratcesses involved in performing spectral analysis of the EEG, i.e., ing the steps involved in conventional including computation of auto- and cross-spectra [Bronzino (linear) spectral analysis of EEG activity. et al., 1984]. It is to be noted that the power spectrum is the autocorrelellogram, i.e., the correlation of the signal with itself. As a result, the power spectrum provides only magnitude information in the frequency domain; it does not provide any data regarding phase. The power spectrum is computed by

()

[ ( )] [ ( )]

P f = Re2 X f + I m2 X f

(15.6)

where X( f ) is the Fourier transform of the EEG signal. Power spectral analysis not only provides a summary of the EEG in a convenient graphic form but also facilitates statistical analysis of EEG changes that may not be evident on simple inspection of the records. In addition to absolute power derived directly from the power spectrum, other measures calculated from absolute power have been demonstrated to be of value in quantifying various aspects of the EEG. Relative power expresses the percentage contribution of each frequency band to the total power and is calculated by dividing the power within a band by the total power across all bands. Relative power has the benefit of reducing the intersubject variance associated with absolute power that arises from intersubject differences in skull and scalp conductance. The disadvantage of relative power is that an © 2000 by CRC Press LLC

increase in one frequency band will be reflected in the calculation by a decrease in other bands; for example, it has been reported that directional shifts between high and low frequencies are associated with changes in cerebral blood flow and metabolism. Power ratios between low (0 to 7 Hz) and high (10 to 20 Hz) frequency bands have been demonstrated to be an accurate estimator of changes in cerebral activity during these metabolic changes. Although the power spectrum quantifies activity at each electrode, other variables derivable from the FFT offer a means of quantifying the relationships between signals recorded from multiple electrodes or sites. Coherence (which is a complex number), calculated from the cross-spectrum analysis of two signals, is similar to cross-correlation in the time domain. The cross-spectrum is computed by

() ()

Cross spectrum = X f Y * f

(15.7)

where X( f ) and Y( f ) are Fourier transforms, and * indicates the complex conjugate. Coherence is calculated by

Coherence =

cross spectrum

()

()

PX f − PY f

(15.8)

The magnitude squared coherence (MSC) values range from 1 to 0, indicating maximum and no synchrony, respectively. The temporal relationship between two signals is expressed by the phase angle, which is a measure of the lag between two signals of common frequency components or bands. Since coherence is a complex number, the phase is simply the angle associated with the polar expression of that number. MSC and phase then represent measures that can be employed to investigate interactions of cerebral activity recorded from separate brain sites. For example, short (intracortical) and long (corticocortical) pathways have been proposed as the anatomic substrates underlying the spatial frequency and patterns of coherence. Therefore, discrete cortical regions linked by such fiber systems should demonstrate a relatively high degree of synchrony, while the temporal difference between signals, represented by the phase measure, quantifies the extent to which one signal leads another.

15.5 Nonlinear Analysis of the EEG As mentioned earlier, the EEG has been studied extensively using signal-processing schemes, most of which are based on the assumption that the EEG is a linear, gaussian process. Although linear analysis schemes are computationally efficient and useful, they only utilize information retained in the autocorrelation function (i.e., the second-order cumulant). Additional information stored in higher-order cumulants is therefore ignored by linear analysis of the EEG. Thus, while the power spectrum provides the energy distribution of a stationary process in the frequency domain, it cannot distinguish nonlinearly coupled frequency from spontaneously generated signals with the same resonance condition [Nikias and Raghvveer, 1987]. There is evidence showing that the amplitude distribution of the EEG often deviates from gaussian behavior. It has been reported, for example, that the EEG of humans involved in the performance of mental arithmetic tasks exhibits significant nongaussian behavior. In addition, the degree of deviation from gaussian behavior of the EEG has been shown to depend on the behavioral state, with the state of slow-wave sleep showing less gaussian behavior than quiet waking, which is less gaussian than rapid eye movement (REM) sleep [Ning and Bronzino, 1989a,b]. Nonlinear signal-processing algorithms such as bispectral analysis are therefore necessary to address nongaussian and nonlinear behavior of the EEG in order to better describe it in the frequency domain. © 2000 by CRC Press LLC

But what exactly is the bispectrum? For a zero-mean, stationary process {X(k)}, the bispectrum, by definition, is the Fourier transform of its third-order cumulant (TOC) sequence: α

(

α

) ∑ ∑C(m, n) e

B ω1, ω 2 =

(

− j w1m + w 2n

)

(15.9)

m = −α n= −α

The TOC sequence [C(m, n)} is defined as the expected value of the triple product

( ) { () (

) ( )}

C m, n = E X k X k + m X k + n

(15.10)

If process X(k) is purely gaussian, then its third-order cumulant C(m, n) is zero for each (m, n), and consequently, its Fourier transform, the bispectrum, B(ω1, ω2) is also zero. This property makes the estimated bispectrum an immediate measure describing the degree of deviation from gaussian behavior. In our studies [Ning and Bronzino, 1989a, b], the sum of magnitude of the estimated bispectrum was used as a measure to describe the EEG’s deviation from gaussian behavior that is,

D=

∑ B (ω , ω ) 1

(ω1ω 2 )

(15.11)

2

Using bispectral analysis, the existence of significant quadratic phase coupling (QPC) in the hippocampal EEG obtained during REM sleep in the adult rat was demonstrated [Ning and Bronzino, 1989a,b, 1990]. The result of this nonlinear coupling is the appearance, in the frequency spectrum, of a small peak centered at approximately 13 to 14 Hz (beta range) that reflects the summation of the two theta frequency (i.e., in the 6- to 7-Hz range) waves (Fig. 15.3). Conventional power spectral (linear) approaches are incapable of distinguishing the fact that this peak results from the interaction of these two generators and is not intrinsic to either. To examine the phase relationship between nonlinear signals collected at different sites, the crossbispectrum is also a useful tool. For example, given three zero-mean, stationary processes {xj(n)j = 1, 2, 3}, there are two conventional methods for determining the cross-bispectral relationship, direct and indirect. Both methods first divide these three processes into M segments of shorter but equal length. The direct method computes the Fourier transform of each segment for all three processes and then estimates the cross-bispectrum by taking the average of triple products of Fourier coefficients over M segments, that is,

(

)

B x1 x 2 x3 ω1, ω 2 =

1 M

∑ X (ω ) X (ω ) X (ω + ω ) M

m 1

m 2

1

m* 3

2

1

2

(15.12)

m =1

where Xmj (ω) is the Fourier transform of the mth segment of {xj(n)}, and * indicates the complex conjugate. The indirect method computes the third-order cross-cumulant sequence for all segments:

C xmx

1 2 x3

(k, l) = ∑ x (n) x (n + k) x (n + l) m 1

m 2

m 3

(15.13)

n⋅

where τ is the admissible set for argument n. The cross-cumulant sequences of all segments will be averaged to give a resultant estimate:

( )

C x1 x 2 x3 k, l =


1 M

M

∑C m =1

m x1 x 2 x 3

(k, l)

(15.14)

FIGURE 15.3 Plots a and b represent the averaged power spectra of sixteen 8-s epochs of REM sleep (digital sampling rate = 128 Hz) obtained from hippocampal subfields CA1 and the dentate gyrus, respectively. Note that both spectra exhibit clear power peaks at approximately 8 Hz (theta rhythm) and 16 Hz (beta activity). Plots c and d represent the bispectra of these same epochs, respectively. Computation of the bicoherence index at f(1) = 8 Hz, f(2) = 8 Hz showed significant quadratic phase coupling (QPC), indicating that the 16-Hz peak seen in the power spectra is not spontaneously generated but rather results from the summation of activity between the two recording sites.

The cross-bispectrum is then estimated by taking the Fourier transform of the third-order crosscumulant sequence:

(

α

α

) ∑ ∑C

B x1 x 2 x3 ω1, ω 2 =

k = −α

l = −α

x1 x 2 x 3

(k, l)

(

− j ω1k +ω 2l

)

.

(15.15)

Since the variance of the estimated cross-bispectrum is inversely proportional to the length of each segment, computation of the cross-bispectrum for processes of finite data length requires careful consideration of both the length of individual segments and the total number of segments to be used. The cross-bispectrum can be applied to determine the level of cross-QPC occurring between {x1(n)} and {x2(n)} and its effects on {x3(n)}. For example, a peak at Bx1x2x3(ω1, ω2) suggests that the energy component at frequency ω1 + ω2 of {x3(n)} is generated due to the QPC between frequency ω1 of {x1(n)} and frequency ω2 of {x2(n)}. In theory, the absence of QPC will generate a flat cross-bispectrum. However, due to the finite data length encountered in practice, peaks may appear in the cross-bispectrum at locations where there is no significant cross-QPC. To avoid improper interpretation, the cross-bicoherence index, which indicates the significance level of cross-QPC, can be computed as follows: © 2000 by CRC Press LLC

FIGURE 15.4 Cross-bispectral plots of BCA1-DG-CA1 (ω1, ω2) computed using (a) the direct method and (b) the indirect method.

(

(

) (ω ) P (ω ) P (ω + ω ) B x1 x 2 x3 ω1, ω 2

)

bic x1 x 2 x3 ω1, ω 2 =

Px1

1

x2

2

x3

1

(15.16)

2

where Pxj (ω) is the power spectrum of process {xj(n)}. The theoretical value of the bicoherence index ranges between 0 and 1, and i.e., from nonsignificant to highly significant. In situations where the interest is the presence of QPC and its effects on {x(n)}, the cross-bispectrum equations can be modified by replacing {x1(n)} and {x3(n)} with {x(n)} and {x2(n)} with {y(n)}, that is,

(

)

B xyz ω1, ω 2 =

1 M

∑ X (ω ) Y (ω ) X (ω + ω ) . M

m

m*

m

1

2

1

2

(15.17)

m =1

In theory, both methods will lead to the same cross-bispectrum when data length is infinite. However, with finite data records, direct and indirect methods generally lead to cross-bispectrum estimates with different shapes (Fig. 15.4). Therefore, like power spectrum estimation, users have to choose an appropriate method to extract the information desired.

Defining Terms Bispectra: Computation of the frequency distribution of the EEG exhibiting nonlinear behavior. Cross-spectra: Computation of the energy in the frequency distribution of two different electrical signals. Electroencephalogram (EEG): Recordings of the electrical potentials produced by the brain. Fast Fourier transform (FFT): Algorithms that permit rapid computation of the Fourier transform of an electrical signal, thereby representing it in the frequency domain. Magnitude squared coherence (MSC): A measure of the degree of synchrony between two electrical signals at specified frequencies. Power spectral analysis: Computation of the energy in the frequency distribution of an electrical signal. Quadratic phase coupling (QPC): A measure of the degree to which specific frequencies interact to produce a third frequency.


References Brazier M. 1968. Electrical Activity of the Nervous System, 3d ed. Baltimore, Williams & Wilkins. Bronzino JD, Kelly M, Cordova C, et al. 1981. Utilization of amplitude histograms to quantify the EEG: Effects of systemic administration of morphine in the chronically implanted rat. IEEE Trans Biomed Eng 28(10):673. Bronzino JD. 1984. Quantitative analysis of the EEG: General concepts and animal studies. IEEE Trans Biomed Eng 31(12):850. Cooley JW, Tukey JS. 1965. An algorithm for the machine calculation of complex Fourier series. Math Comput 19:267. Givens AS, Remond A (eds). 1987. Methods of analysis of brain electrical and magnetic signals. In EEG Handbook, vol 1. Amsterdam, Elsevier. Kay SM, Maple SL. 1981. Spectrum analysis—A modern perspective. Proc IEEE 69:1380. Kondraske GV. 1986. Neurophysiological measurements. In JD Bronzino (ed), Biomedical Engineering and Instrumentation, pp 138–179. Boston, PWS Publishing. Nikias CL, Raghuveer MR. 1987. Bispectrum estimation: A digital signal processing framework. Proc IEEE 75:869. Ning T, Bronzino JD. 1989a. Bispectral analysis of the rat EEG during different vigilance states. IEEE Trans Biomed Eng 36(4):497. Ning T, Bronzino JD. 1989b. Bispectral analysis of the EEG in developing rats. In Proc Workshop HigherOrder Spectral Anal, Vail, CO, pp 235–238. Ning T, Bronzino JD. 1990. Autoregressive and bispectral analysis techniques: EEG applications. Special Issue on Biomedical Signal Processing. IEEE Eng Med Biol Mag 9:47. Smith JR. 1986. Automated analysis of sleep EEG data. In Clinical Applications of Computer Analysis of EEG and Other Neurophysiological Signals, EEG Handbook, revised series, vol 2, pp 93–130. Amsterdam, Elsevier.

Further Information See The Electroencephalogram: Its Patterns and Origins, by J.S. Barlow (Cambridge, Mass., MIT Press, 1993). See also the journals, IEEE Transactions in Biomedical Engineering and Electroencephalography and Clinical Neurophysiology.


Malmivuo, J. “Biomagnetism.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

16 Biomagnetism 16.1

Theoretical Background Origin of Bioelectric and Biomagnetic Signals • Measurement of the Biomagnetic Signals • Independence of Bioelectric and Biomagnetic Signals

16.2

Sensitivity Distribution of Dipolar Electric and Magnetic Leads Concepts of Lead Vector and Lead Field • Lead Fields of Leads Detecting the Electric and Magnetic Dipolar Moments of a Volume Source • Independence of Dipolar Electric and Magnetic Leads

16.3

Magnetocardiography (MCG) Selection of the Source Model for MCG • Detection of the Equivalent Magnetic Dipole of the Heart • Diagnostic Performance of ECG and MCG • Technical Reasons to Use the MCG • Theoretical Reasons to Use the MCG

16.4

Jaakko Malmivuo Ragnar Granit Institute, Tampere University of Technology

Magnetoencephalography (MEG) Sensitivity Distribution of the Axial Magnetometer • Sensitivity Distribution of the Planar Gradiometer • HalfSensitivity Volumes of Electro- and Magnetoencephalography Leads • Sensitivity of EEG and MEG to Radial and Tangential Sources

Since the first detection of the magnetocardiogram (MCG) in 1963 by Baule and McFee [Baule and McFee, 1963], new diagnostic information from biomagnetic signals has been widely anticipated. The first recording of the magnetoencephalogram (MEG) was made in 1968 by David Cohen [Cohen, 1968], but it was not possible to record biomagnetic signals with good signal quality before the invention of the superconducting quantum interference device (SQUID) in 1970 [Zimmerman et al., 1970].

16.1 Theoretical Background Origin of Bioelectric and Biomagnetic Signals In 1819, Hans Christian Örsted demonstrated that when an electric current flows in a conductor, it generates a magnetic field around it (Örsted, 1820]. This fundamental connection between electricity and magnetism was expressed in exact form by James Clerk Maxwell in 1864 [Maxwell, 1865]. In bioelectromagnetism, this means that when electrically active tissue produces a bioelectric field, it simultaneously produces a biomagnetic field as well. Thus the origin of both the bioelectric and the biomagnetic signals is the bioelectric activity of the tissue.


The following equations describe the electric potential field (16.1) and the magnetic field (16.2) of a volume source distribution J i in an inhomogeneous volume conductor. The inhomogeneous volume conductor is represented by a piecewise homogeneous conductor where the regions of different conductivity σ are separated by surfaces S. →

 

 

( ) ∫ J ⋅∇  1r  dv + ∑ ∫ (σ′′− σ′ ) Φ∇  1r  dS

4 πσΦ r =

i

v

( ) ∫J

4 πH r =

v

Sj

j

i

 1 × ∇   dv +  r

j

j

j

(16.1)

j

(16.2)

 

∑ ∫ (σ′′− σ′ ) Φ∇  1r  dS j

Sj

j

j

The first term on the right-hand side of Eqs. (16.1) and (16.2) describes the contribution of the volume source, and the second term describes the contribution of boundaries separating regions of different conductivity, i.e., the contribution of the inhomogeneities within the volume conductor. These equations were developed by David Geselowitz [Geselowitz, 1967, 1970].

Measurement of the Biomagnetic Signals The amplitude of the biomagnetic signals is very low. The strongest of them is the MCG, having an amplitude on the order of 50 pT. This is roughly one-millionth of the static magnetic field of the earth. The amplitude of the MEG is roughly 1% of that of the MCG. This means that, in practice, the MEG can only be measured with the SQUID and that the measurements must be done in a magnetically shielded room. The MCG, instead, can be measured in the clinical environment without magnetic shielding.

Independence of Bioelectric and Biomagnetic Signals The source of the biomagnetic signal is the electric activity of the tissue. Therefore, the most interesting and most important question in biomagnetism is whether the biomagnetic signals contain new information that cannot be obtained from bioelectric signals; in other words, whether the bioelectric and biomagnetic signals are fully independent or whether there is some interdependence. If the signals were fully independent, the biomagnetic measurement would possibly give about the same amount of new information as the bioelectric method. If there were some interdependence, the amount of new information would be reduced. Helmholtz’s theorem states that “A general vector field, that vanishes at infinity, can be completely represented as the sum of two independent vector fields, one that is irrotational (zero curl) and another that is solenoidal (zero divergence)” [Morse and Feshbach, 1953; Plonsey and Collin, 1961]. The impressed current density J i is a vector field that vanishes at infinity and, according to the theorem, may be expressed as the sum of two components: —

J i = J Fi + J Vi

(16.3)

where the subscripts F and V denote flow and vortex, respectively. By definition, these vector fields satisfy × J Fi = 0 and · J Vi = 0. We first examine the independence of the electric and magnetic signals in the infinite homogeneous case, when the second term on the right-hand side of Eqs. (16.1) and (16.2), caused by inhomogeneities, is zero. The equation for the electric potential may be rewritten as —


—

4 πσΦ =

 1 ∇   ⋅ J i dv = v  r

∫

∫

v

∇⋅ J i dv r

(16.4)

∇× Ji dv . r

(16.5)

and that for the magnetic field may be rewritten as

 1 4 πH = − ∇   × J i dv = − v  r

∫

∫

v

Substituting Eq. (16.3) into Eqs. (16.4) and (16.5) shows that under homogeneous and unbounded conditions, the bioelectric field arises from · J Fi , which is the flow source, and the biomagnetic field arises from × J Vi , which is the vortex source. For this reason, in the early days of biomagnetic research it was generally believed that the bioelectric and biomagnetic signals were fully independent. However, it was soon recognized that this could not be the case. For example, when the heart beats, it produces an electric field recorded as the P, QRS, and T waves of the ECG, and it simultaneously produces the corresponding magnetic waves recorded as the MCG. Thus the ECG and MCG signals are not fully independent. There have been several attempts to explain the independence/interdependence of bioelectric and biomagnetic signals. Usually these attempts discuss different detailed experiments and fail to give a satisfying general explanation. This important issue may be easily explained by considering the sensitivity distributions of the ECG and MCG lead systems, and this will be discussed in the next section. →

→

16.2 Sensitivity Distribution of Dipolar Electric and Magnetic Leads Concepts of Lead Vector and Lead Field Lead Vector Let us assume that two electrodes (or sets of electrodes) are placed on a volume conductor to form a lead. Let us further assume that inside the volume conductor in a certain location Q there is placed a unit dipole consecutively in the x, y, and z directions (Fig. 16.1a). Due to the sources, we measure from the lead the signals cx , cy , and cz , respectively. Due to linearity, if instead of the unit dipoles we place in the source location dipoles that are px , py , and pz times the unit vectors, we measure signals that are cx px , cy py , and cz pz , respectively. If these dipoles are placed simultaneously to the source location, due to the principle of superposition, we measure from the lead a voltage that is

V = c x p x + c y p y + c z pz .

(16.6) –

–

–

–

–

These dipoles can be considered to be components of a dipole p, that is, p = px i + py j + pz k. We may – – – – – understand the coefficients cx , cy , and cz to be components of a vector c, that is, c = cx i + cy j + cz k. Now – – we may express the lead voltage Eq. (16.6) as the scalar product of the vector c and the dipole p as

V =c⋅p. –

(16.7) –

The vector c is a three-dimensional transfer coefficient that describes how a dipole source p at a fixed point Q inside a volume conductor influences the voltage measured from the lead and is called the lead vector.


FIGURE 16.1

The concepts of (a) lead vector and (b) lead field. See the text for more details.

–

The lead vector c describes what is the sensitivity of the lead to a source locating at the source location. It is self-evident that for another source location the sensitivity may have another value. Thus the sensitivity, i.e., the lead vector, varies as a function of the location, and we may say that it has a certain distribution in the volume conductor. This is called the sensitivity distribution. Lead Field We may define the value of the lead vector at every point in the volume conductor. If we then place the lead vectors to the points for which they are defined, we have –a field of lead vectors throughout the volume conductor. This field of lead vectors is called the lead field JL. The lead field illustrates the behavior of the sensitivity in the volume conductor and is a very powerful tool in analyzing the properties of electric and magnetic leads (see Fig. 16.1b). It follows from the principle of reciprocity, described by Hermann von Helmholtz in 1853 [Helmholtz, 1853], that the lead field is identical to the electric current field that arises in the volume conductor if a IR, is fed to the lead. unit current, called reciprocal current – When we –know the lead field JL, we can determine the signal VL in the lead due to the volume source distribution J i . For each source element the signal is, of course, proportional to the dot product of the © 2000 by CRC Press LLC

FIGURE 16.1 (continued)

source element and the lead field at the source location, as shown in Eq. (16.7). The contributions of the whole volume source is obtained by integrating this throughout the volume source. Thus the signal the volume source generates to the lead is

VL =

∫ σ J ⋅J 1

L

i

dv .

(16.8)

The lead field may be illustrated either with lead vectors in certain locations in the volume conductor or as the flow lines of the distribution of the reciprocal current in the volume conductor. This is called the lead current field. In the latter presentation, the lead field current flow lines are oriented in the direction of the sensitivity, and their density is proportional to the magnitude of the sensitivity. © 2000 by CRC Press LLC

Lead Fields of Leads Detecting the Electric and Magnetic Dipole Moments of a Volume Source Electric Lead The sensitivity of a lead system that detects the electric dipole moment of a volume source consists of three orthogonal components (Fig. 16.2a). Each of these is linear and homogeneous. In other words, one component of the electric dipole moment is detected when the corresponding component of all – elements of the impressed current density J i are detected with the same sensitivity throughout the source area. Magnetic Lead The sensitivity distribution of a lead system that detects the magnetic dipole moment of a volume source also consists of three orthogonal components (Fig. 16.2b). Each of these has such a form that the sensitivity is always tangentially oriented around the symmetry axis (the coordinate axis). The magnitude of the sensitivity is proportional to the radial distance from the symmetry axis and is zero on the symmetry axis.

Independence of Dipolar Electric and Magnetic Leads Electric Lead The sensitivity distributions of the three components of the lead system detecting the electric dipole moment of a volume source are orthogonal. This means that none of them can be obtained as a linear combination of the two other ones. (Note that any fourth measurement having a similar linear sensitivity distribution would always be a linear combination of the three previous ones.) Thus the sensitivity distributions, i.e., the leads, are orthogonal and thus independent. However, because the three electric signals are only different aspects of the same volume source, they are not (fully) independent. Magnetic Lead The sensitivity distributions of the three components of the lead system detecting the magnetic dipole moment of a volume source are also orthogonal, meaning that no one of them can be obtained as a linear combination of the two other ones. Thus, similarly, as in measurement of the electric dipole moment, the sensitivity distributions, i.e., the leads, are orthogonal and thus independent. However, because the three magnetic signals are only different aspects of the same volume source, they are not (fully) independent. On the basis of the sensitivity distributions, we also can similarly explain the independence between the electric and magnetic signals. According to Helmholtz’s theorem, the electric leads are orthogonal to the three magnetic leads. This means that none of these six leads can be obtained as a linear combination of the other five. However, the six signals, which they measure, are not independent because they arise from the same electrically active volume source.

16.3 Magnetocardiography (MCG) Selection of the Source Model for MCG In ECG and MCG it is the clinical problem to solve the inverse problem, i.e., to solve the source of the detected signal in order to get information about the anatomy and physiology of the source. Although the actual clinical diagnostic procedure is based on measuring certain parameters, such as time intervals and amplitudes, from the detected signal and actually not to display the components of the source, the selection of the source model is very important from the point of view of available information. In clinical ECG, the source model is a dipole. This is the model for both the 12-lead ECG and vectorcardiography (VCG). In 12-lead ECG, the volume conductor (thorax) model is not considered, which causes considerable distortion of the leads. In VCG, only the form of the volume conductor is


FIGURE 16.2 Sensitivity distributions, i.e., lead fields of lead systems detecting (a) electric and (b) magnetic dipole moments of a volume source. The lead field of the electric lead is shown both with vectors representing the magnitude and direction of the lead field (on the left) and with lead field current flow lines (on the right).


modeled. This decreases the distortion in the lead fields but does not eliminate it completely. Note that today the display systems used in these ECG and VCG systems do not play any role in the diagnostic procedure because the computerized diagnosis is always based on the signals, not on the display. In selection of the source model for MCG, it is logical, at least initially, to select the magnetic source model to be on the same theoretical level with the ECG. Only in this way is it possible to compare the diagnostic performance of these methods. It is clear, of course, that if the source model is more accurate, i.e., has more independent variables, the diagnostic performance is better, but when comparing ECG and MCG, the comparison is relevant only if their complexity is similar [Malmivuo and Plonsey, 1995].

Detection of the Equivalent Magnetic Dipole of the Heart The basic detection method of the equivalent magnetic dipole moment of a volume source is to measure the magnetic field on each coordinate axis in the direction of that axis. To idealize the sensitivity distribution throughout the volume source, the measurements must be made at a distance that is large compared with the source dimensions. This, of course, decreases the signal amplitude. The quality of the measurement may be increased considerably if bipolar measurements are used; i.e., measurements are made on both sides of the source. Measurement of the magnetic field on each coordinate axis is, however, difficult to perform in MCG due to the geometry of the human body. It would require either six sequential measurements with one magnetometer (dewar) or six simultaneous measurements using six dewars (Fig. 16.3). It has been shown [Malmivuo, 1976] that all three components of the magnetic dipole also can be measured from a single location. Applying this unipositional method symmetrically so that measurements are made on both the anterior and posterior sides of the thorax at the same distance from the heart, only two dewars are needed and a very high quality of lead fields is obtained. Figure 16.4 illustrates the sensitivity distributions in nonsymmetrical and symmetrical measurements [Malmivuo and Plonsey, 1995].

FIGURE 16.3 Measurement of the three orthogonal components of the magnetic dipole moment of the heart (a) on the coordinate axis (xyz lead system) and (b) at a single location over and under the chest (unipositional lead system).


FIGURE 16.4 Sensitivity distributions in the measurement of the magnetic dipole moment of the heart. (a) Nonsymmetrical and (b) symmetrical measurements of the x component. (c) Symmetrical measurement of the y and z components.

Diagnostic Performance of ECG and MCG The diagnostic performances of ECG and MCG were compared in an extensive study made at the Ragnar Granit Institute [Oja, 1993]. The study was made using the asymmetrical unipositional lead system, i.e., making measurements only on the anterior side of the thorax. The patient material was selected, however, so that myocardial changes were located dominantly on the anterior side. This study consisted of 290 normal subjects and 259 patients with different myocardial disorders. It was found that the diagnostic performance of ECG and MCG is about the same (83%). Diagnostic parameters were then selected from both ECG and MCG. With this combined method, called electromagnetocardiogram (EMCG), a diagnostic performance of 90% was obtained. This improvement in diagnostic performance was obtained without increasing the number of parameters used in the diagnostic procedure. Moreover, this improvement is significant because it means that the number of incorrectly diagnosed patients was reduced by approximately 50%. This important result may be explained as follows: The lead system recording the electric dipole moment of the volume source has three independent leads. (This is also the case in the 12-lead ECG



system.) Similarly, the lead system detecting the magnetic dipole moment of the volume source has three independent leads. Therefore, the diagnostic performances of these methods are about the same. However, because the sensitivity distributions of electric and magnetic leads are different, the patient groups diagnosed correctly with both methods are not identical.



As stated before, the electric leads are independent of the magnetic leads. If the diagnostic procedure simultaneously uses both the ECG and the MCG leads, we obtain 3 + 3 = 6 independent leads, and the correctly diagnosed patient groups may be combined. Thus the diagnostic performance of the combined method is better than that of either method alone. This is the first large-scale statistically relevant study of the clinical diagnostic performance of biomagnetism.


Technical Reasons to Use the MCG The technical differences between ECG and MCG include the MCG’s far better ability to record static sources, sources on the posterior side of the heart, monitor the fetal heart, and perform electrodeless recording. As a technical drawback, it should be mentioned that the MCG instrument costs 2 to 3 times more. An important feature of MCG is that, unlike the MEG instrument, it does not need a magnetically shielded room. This is very important because the shielded room is not only very expensive but also limits application of the technique to a certain laboratory space.

Theoretical Reasons to Use the MCG It has been shown that MCG has clinical value and that it can be used either alone or in combination with ECG as a new technique called the electromagnetocardiogram (EMCG). The diagnostic performance of the combined method is better than that of either ECG or MCG alone. With the combined method, the number of incorrectly diagnosed patients may be reduced by approximately 50%.

16.4 Magnetoencephalography (MEG) Similarly as in the cardiac applications, in the magnetic measurement of the electric activity of the brain, the benefits and drawbacks of the MEG can be divided into theoretical and technical ones. First, the theoretical aspects are discussed. The two main theoretical aspects in favor of MEG are that it is believed that because the skull is transparent for magnetic fields, the MEG should be able to concentrate its measurement sensitivity in a smaller region than the EEG, and that the sensitivity distribution of these methods are fundamentally different. These questions are discussed in the following: The analysis is made using the classic spherical head model introduced by Rush and Driscoll [Rush and Driscoll, 1969]. In this model, the head is represented with three concentric spheres, where the outer radii of the scalp, skull, and brain are 92, 85, and 80 mm, respectively. The resistivities of the scalp and the brain are 2.22 Ω·cm, and that of the skull is 80 times higher, being 177 Ω·cm. The two basic magnetometer constructions in use in MEG are axial and planar gradiometers. In the former, both coils are coaxial, and in the latter, they are coplanar. The minimum distance of the coil from the scalp in a superconducting magnetometer is about 20 mm. The coil radius is usually about 10 mm. It has been shown [Malmivuo and Plonsey, 1995] that with this measurement distance, decreasing the coil radius does not change the distribution of the sensitivity in the brain region. In the following the sensitivity distribution of these gradiometer constructions is discussed. To indicate the magnetometer’s ability to concentrate its sensitivity to a small region, the concept of half-sensitivity volume has been defined. This concept means the region in the source area (brain) where the detector sensitivity is one-half or more from the maximum sensitivity. The smaller the half-sensitivity volume, the better is the detector’s ability to focus its sensitivity to a small region. In magnetocardiography, it is relevant to detect the magnetic dipole moment of the volume source of the heart and to make the sensitivity distribution within the heart region as independent of the position in the axial direction as possible. In magnetoencephalography, however, the primary purpose is to detect the electric activity of the cortex and to localize the regions of certain activity.

Sensitivity Distribution of the Axial Magnetometer In a cylindrically symmetrical volume conductor model, the lead field flow lines are concentric circles and do not cut the discontinuity boundaries. Therefore, the sensitivity distribution in the brain area of the spherical model equals that in an infinite, homogeneous volume conductor. Figure 16.5 illustrates the sensitivity distribution of an axial magnetometer. The thin solid lines illustrates the lead field flow lines. The dashed lines join the points where the sensitivity has the same value, being thus so-called isosensitivity lines. The half-sensitivity volume is represented by the shaded region. © 2000 by CRC Press LLC

FIGURE 16.5

Sensitivity distribution of the axial magnetometer in measuring the MEG (spherical lead model).

Sensitivity Distribution of the Planar Gradiometer Figure 16.6 illustrates the sensitivity distribution of a planar gradiometer. Again, the thin solid lines illustrate the lead field flow lines, and the dashed lines represent the isosensitivity lines. The half-sensitivity volume is represented by the shaded region. The sensitivity of the planar gradiometer is concentrated under the center of the two coils and is mainly linearly oriented. Further, there exist two zero-sensitivity lines.

Half-Sensitivity Volumes of Electro- and Magnetoencephalography Leads The half-sensitivity volumes for different EEG and MEG leads as a function of electrode distance and gradiometer baselines are shown in Fig. 16.7. The minimum half-sensitivity volume is, of course, achieved with the shortest distance/baseline. For three- and two-electrode EEG leads, the half-sensitivity volumes at 1 degree of electrode distance are 0.2 and 1.2 cm3, respectively. For 10-mm-radius planar and axial gradiometer MEG leads, these volumes at 1 degree of coil separation (i.e., 1.6-mm baseline for axial gradiometer) are 3.4 and 21.8 cm3, respectively. © 2000 by CRC Press LLC

FIGURE 16.6

Sensitivity distribution of the planar gradiometer (half-space model).

The 20-mm coil distance from scalp and 10-mm coil radii are realistic for the helmet-like whole-head MEG detector. There exist, however, MEG devices for recording at a limited region where the coil distance and the coil radii are on the order of 1 mm. Therefore, the half-sensitivity volumes for planar gradiometers with 1-mm coil radius at 0- to 20-mm recording distances are also illustrated in Fig. 16.7. These curves show that when the recording distance is about 12 mm and the distance/baseline is 1 mm, such a planar gradiometer has about the same half-sensitivity volume as the two-electrode EEG. Short separation will, of course, also decrease the signal amplitude. An optimal value is about 10 degrees of separation. Increasing the separation to 10 degrees increases the EEG and MEG signal amplitudes to approximately 70 to 80% of their maximum value, but the half-sensitivity volumes do not increase considerably from their values at 1 degree of separation. Thus, contrary to general belief, the EEG has a better ability to focus its sensitivity to a small region in the brain than the whole-head MEG. At about 20 to 30 degrees of separation, the two-electrode EEG lead needs slightly smaller separation to achieve the same half-sensitivity volume as the planar gradiometer. The sensitivity distributions of these leads are, however, very similar. Note that if the sensitivity distributions of two different lead systems, whether they are electric or magnetic, are the same, they © 2000 by CRC Press LLC

FIGURE 16.7 Half-sensitivity volumes of different EEG leads (dashed lines) and MEG leads (solid lines) as a function of electrode distance and gradiometer baseline, respectively.

detect exactly the same source and produce exactly the same signal. Therefore, the planar gradiometer and two-electrode EEG lead detect very similar source distributions.

Sensitivity of EEG and MEG to Radial and Tangential Sources The three-electrode EEG has its maximum sensitivity under that electrode which forms the terminal alone. This sensitivity is mainly directed radially to the spherical head model. With short electrode © 2000 by CRC Press LLC

distances, the sensitivity of the two-electrode EEG is directed mainly tangentially to the spherical head model. Thus with the EEG it is possible to detect sources in all three orthogonal directions, i.e., in the radial and in the two tangential directions, in relation to the spherical head model. In the axial gradiometer MEG lead, the sensitivity is directed tangentially to the gradiometer symmetry axis and thus also tangentially to the spherical head model. In the planar gradiometer, the sensitivity has its maximum under the center of the coils and is directed mainly linearly and tangentially to the spherical head model. The MEG lead fields are oriented tangentially to the spherical head model everywhere. This may be easily understood by recognizing that the lead field current does not flow through the surface of the head because no electrodes are used. Therefore, the MEG can only detect sources oriented in the two tangential directions in relation to the spherical head model.

References Baule GM, McFee R. 1963. Detection of the magnetic field of the heart. Am Heart J 55(7):95. Cohen D. 1968. Magnetoencephalography: Evidence of magnetic fields produced by alpha-rhythm currents. Science 161:784. Geselowitz DB. 1967. On bioelectric potentials in an inhomogeneous volume conductor. Biophys J 7(1):1. Geselowitz DB. 1970. On the magnetic field generated outside an inhomogeneous volume conductor by internal current sources. IEEE Trans Magn MAG-6(2):346. Helmholtz HLF. 1853. Ueber einige Gesetze der Vertheilung elektrischer Ströme in körperlichen Leitern mit Anwendung auf die thierisch-elektrischen Versuche. Ann Physik Chem 89:211. Malmivuo J, Plonsey R. 1995. Bioelectromagnetism: Principles and Applications of Bioelectric and Biomagnetic Fields. New York, Oxford University Press. Malmivuo JA. 1976. On the detection of the magnetic heart vector: An application of the reciprocity theorem. Acta Polytechnol Scand 39:112. Maxwell J. 1865. A dynamical theory of the electromagnetic field. Phil Trans R Soc (Lond) 155:459. More PM, Feshbach H. 1953. Methods of Theoretical Physics, part I. New York, McGraw-Hill. Oja OS. 1993. Vector Magnetocardiogram in Myocardial Disorders, MD thesis, University of Tampere, Medical Faculty. Örsted HC. 1820. Experimenta circa effectum conflictus electrici in acum magneticam. J F Chem Phys 29:275. Plonsey R, Collin R. 1961. Principles and Applications of Electromagnetic Fields. New York, McGraw-Hill. Rush S, Driscoll DA. 1969. EEG-electrode sensitivity: An application of reciprocity. IEEE Trans Biomed Eng BME-16(1):15. Zimmerman JE, Thiene P, Hardings J. 1970. Design and operation of stable rf biased superconducting point-contact quantum devices. J Appl Phys 41:1572.


Durand, D. M. “Electric Stimulation of Excitable Tissue .” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

17 Electric Stimulation of Excitable Tissue 17.1 17.2 17.3

Electric Stimulation of Neural Tissue Physiology of Excitation Electric Fields in Volume Conductors Quasi-State Formulation • Equivalence Between Dielectric and Conductive Media • Potential from a Monopole Source • Potential from Bipolar Electrodes and Dipoles • Inhomogeneous Volume Conductors

17.4

Electric Field Interactions with Excitable Tissue Discrete Cable Equation for Myelinated Axons • Equivalent Cable Equation for Myelinated Fibers • Unmyelinated/Myelinated Fibers • Anodic/Cathodic Stimulation • Large/Small-Diameter Axons • Spatial Derivative of the Electric Field • Activating Function • Net Driving Function • Current-Distance Relationship • Longitudinal/Transverse Field • Anodal Surround Block • Unidirectionally Propagated Action Potentials

17.5

Electrode-Tissue Interface Effect of Stimulation Waveform on Threshold • Electrochemistry of Stimulation • Stimulation of Brain Tissue • Stimulation of the Peripheral Nerve • Stimulation of Muscle Tissue • Corrosion • Tissue Damage • Biphasic Waveforms

Dominique M. Durand Case Western Reserve University

17.6

Conclusion

17.1 Electric Stimulation of Neural Tissue Functional electric stimulation (FES) of neural tissue provides a method to restore normal function to neurologically impaired individuals. By inserting electrodes inside or near nerves, it is possible to activate pathways to the brain or to muscles. Functional nerve stimulation (FNS) is often used to describe applications of electric stimulation in the peripheral nervous system. Neural prostheses refer to applications for which electric stimulation is used to replace a function previously lost or damaged. Electric stimulation has been used to treat several types of neurologic dysfunction with varying amounts of success [see Hambrecht, 1979]. For example, electric stimulation of the auditory nerves to restore hearing in deaf patients has proved to be not only feasible but also clinically useful [Clark et al., 1990]. Similarly, the phrenic nerve of patients with high-level spinal cord injury can be stimulated to produce diaphragm contractions and restore ventilation [Glenn et al., 1984]. Electric stimulation of the visual cortex produces visual sensations called phosphenes [Brindley and Lewin, 1968], and a visual prosthesis for the blind is currently being tested. Electric stimulation in the peripheral nerves of paralyzed patients can restore partial function of both the upper extremities for hand function [Peckham et al.,


1977] and lower extremities for gait [Marsolais and Kobetic, 1988]. Several other attempts were not so successful. Electric stimulation of the cerebellum cortex for controlling epileptic seizures has been tried but was not reliable. However, a method involving stimulation of the vagus nerve has been successful in decreasing seizure frequency in patients with epilepsy [Rutecki, 1990; Fisher et al., 1997]. There are many other applications of electric stimulation of the nervous system and many problems associated with each one, but it is now clear that the potential and the limits of the technique have not yet been realized. Other excitable tissues such as cardiac muscle can also be excited by externally applied electric fields. The underlying principles are similar to those reviewed below for neural tissue. Stimulation of the nervous system also can be achieved with magnetic fields [Chokroverty, 1990]. A coil is placed near the excitable tissue, and a capacitor is rapidly discharged into the coil. Large magnetic fluxes are generated, and the induced electric fields can generate excitation. Magnetic stimulation has several advantages over electric stimulation. Magnetic fields can easily penetrate low-conductivity tissues such as bone, and the stimulation is completely noninvasive. However, magnetic stimulation requires a large amount of energy, and the magnetic field is difficult to localize. Magnetic stimulation of excitable tissue shares many aspects with electric stimulation, since the electric field is, in both cases, the source of the stimulus [Roth and Basser, 1990]. However, there are several important differences [Durand et al., 1992; Nagarajan et al., 1997] that will not be reviewed below. What is the basis for the effect of electric stimulation of the nervous system? Clearly, it comes from the fact that a propagated action potential can be triggered by applying a rapidly changing electric field near excitable tissue. This fact was demonstrated early in this century, and clinical experimental applications in the 1950s resulted in the highly successful cardiac pacemaker. Other clinical applications have been slow in coming for several reasons to be discussed below. One of the difficulties is that the fundamental principles of the interaction of electric fields and neurons is not completely understood. In order to understand how applied currents can generate excitation, it will necessary to describe the mechanisms underlying excitation (Sec. 17.2), the distribution of currents inside the volume conductor (Sec. 17.3), and the interaction between the axon and applied electric fields (Sec. 17.4). Another difficulty lies at the interface between electrodes applying the current and neural tissue to be stimulated. Current in the wires to the electrodes is carried by electrons, whereas current in the volume conductor is carried by ions. Chemical reasons at the interface will take place, and these reactions are still poorly understood. The waveforms used to apply the current can significantly affect the threshold, the electrochemistry at the electrode site, and tissue damage. These issues will be reviewed in Sec. 17.5.

17.2 Physiology of Excitation Electric stimulation of excitable tissue is mediated by artificially depolarizing membrane-containing channels capable of producing action potentials. Action potentials are normally elicited by synaptic currents, which in turn produce depolarization of the somatic membrane. The sodium channels are responsible for the generation of the depolarizing phase of the action potential and are sensitive to membrane voltage [Ferreira and Marshal, 1985]. Once initiated in the soma, the action potential is carried unattenuated along the axon to its intended target, such as another neuron or neuromuscular junction for muscle contraction (Fig. 17.1). Unmyelinated axons have ionic channels distributed throughout their membranes, and the action potential is carried smoothly along their length. The membranes of the axons containing the channels behave as resistive and capacitive elements, limiting the conduction velocity. Myelinated axons are surrounded by a insulation sheath of myelin that significantly decreases the membrane capacitance, thereby increasing the speed of propagation. Conduction of the action potential is no longer smooth but takes place in discrete steps (saltatory conduction). The myelin sheath is broken at regularly spaced intervals (nodes of Ranvier) to allow the current to flow through the membrane [Aidley, 1978]. Action potentials can be generated artificially by placing electrodes directly inside a cell. Current flowing from the inside to the outside will produce depolarization followed by excitation, provided that the current amplitude is large enough. This technique cannot be used for functional stimulation because we do not yet have the technology to interface electrodes with large numbers of single axons. Therefore, © 2000 by CRC Press LLC

FIGURE 17.1 Electric stimulation of the myelinated fiber. An electrode is located near the axon, and a cathodic stimulus is applied to the electrode. The current flow in and around the axon is described in this chapter and causes depolarization of the membrane at the sites closest to the electrodes. Action potentials are generated underneath the electrode and propagate orthodromically and antidromically.

electrodes must be placed in the extracellular space near the tissue to be excited. It is then possible to activate many cells simultaneously. However, stimulation must be selective. Selectivity is defined as the ability of a stimulation system to activate any chosen set of axons. For example, the nervous system chooses to recruit small fibers connected to small motor units followed by large fibers connected to large motor units for smooth motor control. Applied current pulses activate first large fibers and then small fibers with increasing current (reverse recruitment) for reasons described in Sec. 17.4. The reverse recruitment order as well as our inability to recruit a chosen set of axons within a nerve bundle makes electric stimulation a powerful but difficult-to-control tool for activation of the nervous system.

17.3 Electric Fields in Volume Conductors The excitable tissue to be stimulated (peripheral nerves, motor neurons, CNS neurons) are in all cases surrounded by an extracellular fluid with a relatively high conductivity (80 to 300 Ω ·cm). The electrodes used for electric stimulation are always placed in this “volume conductor” and it is essential to understand how the currents and the electric fields are distributed [Heringa et al., 1982]. The calculation of current density and electric fields can be done easily in simple cases such as a homogeneous (conductivity the same everywhere) and isotropic (conductivity the same in all directions) medium.

Quasi-Static Formulation The calculations of electric fields generated by an electrode located in a conducting volume conductor can be done by solving Maxwell equations. The frequencies used for electric stimulation are generally under 10 kHz, and therefore, a simplified set of equations known as the quasi-static formulation can be used [Plonsey, 1969]:

Conservation of charge:

∇⋅ J = 0

Gauss’ law:

∇⋅ E =


ρ

(17.1) (17.2)

Ohm’s law for conductors:

J = σE

(17.3)

Electric field:

E = −∇φ

(17.4)

where E is the electric field (V/m) defined as gradient of the scalar potential φ ; J is the current density (defined as the current crossing a given surface, in A/m2); σ is the conductivity (inverse of resistivity), in s/m; ρ is the charge density, in C/m3 ; ε is the permittivity of the medium; and · A is the divergence of vector A.

Equivalence Between Dielectric and Conductive Media Assuming an infinite homogeneous conducting medium with conductivity σ with a single point source as shown in Fig. 17.2a, the current density J at any point is the sum of a source term Js and an ohmic term σE:

J = σE + J s

(17.5)

∇⋅ J = ∇⋅ σE + ∇⋅ J s = 0

(17.6)

Using Eq. (17.1),

FIGURE 17.2 Voltage and electric field along an axon. (a) The current density J within an infinite, homogeneous, and isotropic volume conductor is radial and is inversely proportional to the square of the distance. (b) Voltage along an axon located 1 cm from an anode with 1 mA of current. (c) Electric field along the same axon.


Since the volume conductor is homogeneous, ·(σE) = σ · E, and we then have

σ∇⋅ E = −∇⋅ J s

(17.7)

Since E = –φ and ·(A) is by definition the laplacian of A, 2A, we then have

∇2φ = ∇⋅ J s σ = − I v σ

(17.8)

where Iv is a source term (in A/m3). The source term Iv is zero everywhere except where the sources are located. This equation is nearly identical to the Poisson equation derived from Eq. (17.2) [Kraus and Carver, 1973];

∇2 φ = −

ρ

(17.9)

derived for dielectric media. Using the following equivalence,

ρ → Iv →σ

the solution of the Poisson equation for dielectric problems can then be used for the solution of the current in volume conductors.

Potential from a Monopole Source Given a point source (monopolar) in an infinite homogeneous and isotropic volume conductor connected to a current source I, the potential and currents anywhere can be derived easily. Using spherical symmetry, the current density J at a point P located at a distance r from the source is equal to the total current crossing a spherical surface with radius r (see Fig. 17.2):

J=

I ur 4 πr 2

(17.10)

where ur is the unit radial vector, and r is the distance between the electrode and the measurement point. The electric field is then obtained from Eq. (17.3):

E=

I ur 4 πσr 2

(17.11)

The electric field is the gradient of the potential. In spherical coordinates,

E=−

dφ ur dr

(17.12)

Therefore, the potential at point P is obtained by integration:

φ=


I 4 πσr

(17.13)

It can be shown easily that this solution satisfies the Poisson equation. For a monopolar electrode, the current distribution is radial and is inversely proportional to the conductivity of the medium and the distance to the source. The potential decays to zero far away from the electrode and goes to infinity on the electrode. The singularity at r = 0 can be eliminated by assuming that the electrode is spherical with finite radius a. Equation (17.13) is then valid on the surface of the electrode for r = a and for r > a [Nunez, 1981]. Equation (17.13) can be generalized to several monopolar electrodes. Assuming n electrodes with a current Ii located at a distance ri from the recording point, the voltage is then given by

φ=

1 4 πσ

∑r

Ii

n

(17.14)

i

For an axon located in the volume conductor as shown in Fig. 17.2, the voltage along the axon located 1 cm away from an anode with a 1-mA current sources is given by the following equation and is plotted in Fig. 17.2b :

φ=

I 4 πσ d 2 + x 2

(17.15)

The electric field along the axon can be obtained by taking the spatial derivative of Eq. (17.15) and is plotted in Fig. 17.2c.

Potential From Bipolar Electrodes and Dipoles In the derivation of the potential generated by a monopolar electrode, the current enters the volume conductor at the tip of the electrode and exits at infinity. However, in the following example (saltwater tank), a current is applied through two monopolar electrodes separated by a distance d as shown in Fig. 17.3a. The potential generated by this bipolar configuration can be calculated at point P (assuming that the voltage reference is at infinity) as follows:

φ=

I  1 1 − 4 πσ  r1 r2 

(17.16)

When the distance d between the two electrodes is small compared with the distance r, the equation for the potential is given by the following dipole equation valid for d :

φ=

Id cos θ 4 πσr 2

(17.17)

where the angle θ is defined as in Fig. 17.3. The current distribution for a current dipole is no longer radial and is shown in Fig. 17.3b. The voltage along a line perpendicular to the axis of the dipole and passing through a point equidistant from the electrodes is zero. Therefore, an axon located in that region would not be excited regardless of the current amplitude. The potential generated by a dipole is inversely proportional to the square of the distance and therefore decays faster than the monopole. Therefore, the dipole configuration will have a more localized excitation region. However, the excitation thresholds are higher because most of the current flows around the electrode. For an observation point located at a distance r from a dipole or monopole with the same current, the ratio of the monopole voltage to that of the dipole voltage is proportional to r/d. Since r is assumed to be much larger than d, then the monopole voltage is much larger than the voltage generated by the dipole.


FIGURE 17.3 Voltage generated by bipolar electrodes. (a) Saltwater tank. Two electrodes are located within a large tank filled with a conduction solution. The voltage generated by the bipolar arrangement can be measured as shown provided that the reference electrode is located far away from the stimulation electrode. The effect of the stimulation electrode on the reference potential also can be taken into account. (b) Current lines and equipotential voltage distributions (dashed lines) for a dipole. The distance r between the observation P and the electrodes is much greater that the distance d between the electrodes.

Inhomogeneous Volume Conductors For practical applications of electric stimulation, electrodes are placed in various parts of the body. The volume conductors are clearly not homogeneous because we must consider the presence of bone with a conductivity significantly higher than that of extracellular space or even air above the skin with a conductivity of zero. How do those conductivities affect the potentials generated by the electrode? This question usually can only be answered numerically by computer models that take into account these various compartments such as finite-differences, finite-elements, or boundary-elements methods. A simple solution, however, can be obtained in the case of a semi-infinite homogeneous volume conductor using the method of images. Consider two volume conductors with conductivities σ1 and σ2 separated by an infinite plane as shown in Fig. 17.4. A monopolar stimulating electrode is placed in region 1. Potential recordings are made in that same region. It can be shown that the inhomogeneous volume conductor can be replaced by a homogeneous volume by adding another current source located on the other side of the plane (see Fig. 17.4) with an amplitude equal to [Nunez, 1981]

I′ =


σ1 − σ 2 I σ1 + σ 2

(17.18)

FIGURE 17.4 Method of images. The method of images can be used to calculate the voltage generated by an electrode in a homogeneous volume conductor. The two semiinfinite volume conductors with conductivities σ1 and σ2 (a) are replaced by a single infinite volume conductor with conductivity σ1 and an additional image electrode (b).

The voltage at point P is then given by Eq. (17.14). The mirror-image theory is only applicable in simple cases but can be useful to obtain approximations when the distance between the recording electrode and the surface of discontinuity is small, thereby approximating an infinite surface [Durand et al., 1992].

17.4 Electric Field Interactions with Excitable Tissue When axons are placed inside a volume conductor with a stimulation electrode, current flows according to equations derived in the preceding section. Some of the current lines enter and exit the axon at different locations and will produce excitation or inhibition. An example of this current distribution is shown in Fig. 17.5a for a monopolar anodic electrode. A length ∆x of the axonal membrane can be modeled at rest by a capacitance Cm in parallel with a series of combination of a battery (Er) for the resting potential and a resistance Rm stimulating the combined resistance at rest of all the membrane channels (see Fig. 17.5b). Nonlinear ionic channel conductances can be added in parallel with the membrane resistance and capacitance. Their contribution at rest is small and becomes significant only around threshold. Since we are interested here in how to drive the membrane potential toward threshold from resting values, their contribution is ignored. When current enters the membrane flowing from the outside to the inside, the membrane is hyperpolarized (moved closer to threshold), as illustrated in Fig. 17.5c. Similarly, when the current exits the membrane, depolarization is generated (the membrane voltage is moved closer to the threshold). In the case of an anodic electrode, illustrated in Fig. 17.5, stimulation should not take place directly underneath the electrode but further along the axon where the membrane is depolarized. A more quantitative approach to this problem can be obtained by modeling the interaction of the model with the applied current. The applied current I generates a voltage distribution in the extracellular space that can be calculated using Eq. (17.15) assuming a homogeneous and isotropic medium. An unmyelinated fiber is modeled as a one-dimensional network by linking together electrical models of the membrane with a resistance Ra to account for the internal resistance of the axon, as shown in Fig. 17.6a. The circuit can be stimulated using numerical methods [Koch and Segev, 1989] or by using already available general software packages such as Pspice or neuronal simulation packages such as Neuron [Hines, 1984]. The variable of interest is the transmembrane potential Vm , since the sodium channels are sensitive to the voltage across the membrane. Vm is defined as the difference between the intracellular


FIGURE 17.5 Effect of extracellular current on axon membrane polarization. (a) A monopolar anode is located near an axon, and current enters and exits the membrane. (b) At rest, the membrane can be modeled by a simple RC network. (c) When current enters the membrane, charge is added to the membrane capacitance, and additive membrane polarization is generated. Therefore, the membrane is hyperpolarized. (d) When current exits the axon, the membrane is depolarized.

voltage Vi and the extracellular voltage Ve minus the resting potential Er in order to reflect the change from resting values. Stimulation of the fiber can occur when the extracellular voltage difference between two nodes is large enough to generate transmembrane voltage greater than the firing threshold. Applying Kirchoff ’s law at each node and taking the limit when the length of membrane ∆x goes to zero, one obtains the following inhomogeneous cable equation [Rall, 1979; Clark and Plonsey, 1966; Altman and Plonsey, 1988; Rattay, 1989]:

λ2

∂2Vm ∂x 2

+ τm

∂Vm ∂2V − Vm = − λ2 2e ∂t ∂x

(17.19)

Vm and Ve are the transmembrane and extracellular voltages, respectively. λ is the space constant of the fiber and depends only on the geometric and electric properties of the axon:

λ=

1 Rmsd 2 Ras

(17.20)

where R sm is the specific membrane resistance, R sa is the axoplasmic-specific resistance, and d is the diameter of the axon. τ m is the time constant of the axon and is given by

τm = RmCm


(17.21)

FIGURE 17.6 Model of extracellular voltage and axon interactions. (a) The effect of the extracellular voltage generated by the stimulating electrode on the axon can be modeled by directly connecting the membrane compartment to the extracellular source. The effect of the ion channels can be taken into account by adding nonlinear conductances and an equilibrium battery (not shown). (b) The axon with an extracellular voltage is a is equivalent to an axon inside an infinite conductance medium (zero voltage) and an equivalent voltage source inside the membrane.

The term on the right-hand side of the equation is called the source term or forcing function and is the product of the square of the space constant with the second spatial derivative of the extracellular voltage. In order to explain the significance of this term, it can be shown easily that the model in Fig. 17.6a is equivalent to a model in which the extracellular space with its electrode and voltage has been replaced by a set of equivalent voltage sources inside the nerve given by

Veq = λ2

∆2Ve ∆x

2

= λ2

d 2Ve dx 2

(17.22) ∆x →0

The amplitude of the equivalent voltage sources is plotted in Fig. 17.7a for a 10-µm axon stimulated by a 1-mA anodic current located 1 cm away from the axon. A positive value for the equivalent voltage source indicates membrane depolarization, while a negative value indicates hyperpolarization. The peak value of the depolarization indicates the excitation site. Therefore, one would predict that for anodic stimulation, excitation would take place at the peak of the two depolarized regions, and two action potentials would propagate toward the ends of the axon away from the electrode. Note that the shape of Veq calculated is similar to that predicted by simply examining the current flow in and out of the axon in Fig. 17.5. This analysis is valid only at the onset of the pulse, since during the pulse currents will be distributed throughout the cable and will affect the transmembrane potential [Warman et al., 1994]. However, it has been shown that for short pulses these effects are small, and the shape of the transmembrane voltage can be predicted using the source term of the cable equation.

Discrete Cable Equation for Myelinated Axons Electric stimulation of myelinated fibers is significantly different from that of unmyelinated fibers because the presence of a myelin sheath around the axon forces the current to flow in and out of the membrane


FIGURE 17.7 Eqivalent source voltage (Veq ). A positive value of the Veq indicates membrane depolarization, and a negative value indicates membrane hyperpolarization. (a) Equivalent source voltage plotted along the axon for an anodic stimulus. The peak value is negative, and therefore, the membrane is hyperpolarized underneath the electrode. Depolarization occurs at the sites of maximum value of Veq (arrows). (b) Cathodic stimulation generates depolarization underneath the electrode, and the amplitude of Veq is larger than for anodic stimulation. Therefore, a cathodic stimulus has a lower threshold than an anodic stimulus. (c) Veq is larger for fibers with larger diameters; therefore, large-diameter fibers are recruited first.

only at the nodes (see Fig. 17.1). The action potential propagates from one node to the next (saltatory conduction). The effect of an applied electric field on a myelinated fiber can be described using a discrete model of the axon and extracellular voltage [McNeal, 1976; Rattay, 1989]. Such a model is shown in Fig. 17.8, in which voltages are applied at each node. The resistance Rn represents the membrane resistance at the node of Ranvier only because the myelin resistance is considered very high for this model and is neglected. Cn is the capacitance at the node of Ranvier. Ra represents the resistance between two nodes. The battery for the resting potential is removed for simplification by shifting the resting potential to zero. Rn, Cn, and Ri are given by

Rn =

Ra =


Rns πdl

4 Ras L πd 2

(17.23)

(17.24)

FIGURE 17.8 Discrete model for myelinated fibers. The resistance Rn and capacitance Cn of the node of Ranvier is modeled with an additional internodal resistance Ra . The resistance of the myelin is assumed to be infinite in this model. Extracellular voltages are calculated or measured at each node, and the transmembrane voltage can be estimated.

Cn = Cns πdl

(17.25)

where d is the inner fiber diameter, l is the width of the node, and L is the internodal distance (see Fig. 17.8). The cable equation for this model can be derived using Kirchoff ’s law:

∂V Rn 2 R ∆ Vm − RnCn m − Vm = − n ∆2Ve ∂t Ra Rd

(17.26)

∆2 is the second different operator: ∆2V = Vn–1 – 2Vn + Vn+1. Using Eqs. (17.23) to (17.25), one can show that RnCn and Rn /Ra are independent of the diameter, since L/D = 100, d/D = 0.7 and l is constant. Therefore, the left side of Eq. (17.26) is independent of the diameter. The source term, however, does depend on the diameter but only implicity. Since the distance between the node increases with the diameter of the fiber, the voltage across the nodes also increases, suggesting that fibers with larger diameters are more easily excitable because they “see” larger potentials along the nodes.

Equivalent Cable Equation for Myelinated Fibers The diameter dependence of these fibers can be expressed explicitly by using an equivalent cable equation recently derived [Basser, 1993] and adapted here to take into effect the extracellular voltage:

λ2my

∂2Vm ∂x 2

− τmy

∂Vm ∂2V − Vm = − λ2my 2e ∂t ∂x

(17.27)

where λmy is the equivalent space constant for the case of an axon with myelin sheath of infinite resistance and is defined as

λ my =


1 RnsdL 2 Ras l

(17.28)

Unmyelinated/Myelinated Fibers The equivalent cable equation for the myelinated fibers is similar to Eq. (17.20) derived for unmyelinated axons. The forcing function is also proportional to the first derivative of the extracellular electric field along the nerve. The dependence on the diameter can be observed directly by expressing the equivalent voltage source for myelinated fibers (Veqmy) as function of the inner diameter d:

Veqmy = 35.7

Rns d 2 ∂2Ve Ras l ∂x 2

(17.29)

Equation (17.29) shows that Veqmy is proportional to the square of the diameter of fiber d, while the equivalent voltage source for an unmyelinated fiber is proportional to its diameter (17.23). Therefore, the change in membrane voltage at t = 0 for myelinated and unmyelinated fibers is proportional to d2 and d respectively.

Anodic/Cathodic Stimulation It has been demonstrated experimentally that in the case of electric stimulation of peripheral nerves, cathodic stimulation has a lower threshold (less current required) than anodic stimulation. This experimental result can be explained directly by plotting Veq for a cathodic and an anodic electrode (1 mA) located 1 cm away from a 10-µm unmyelinated fiber (see Fig. 17.7). The maximum value of Veq for anodic stimulation is 0.05 mV at the two sites indicated by the arrow. However, the maximum depolarization for the cathodic electrode is significantly larger at 0.2 mV, with the site of excitation located directly underneath the electrode. In special cases, such as an electrode located on the surface of a cortex, cathodic stimulation can have a higher threshold [Ranck, 1975].

Large/Small-Diameter Axons The equivalent voltage source of the cable equation is proportional to the square of the space constant λ . λ2 is proportional to the diameter of the fiber (17.20) for myelinated fibers and to the square of the diameter (17.29) for myelinated fibers. Therefore, in both cases, Veq is higher for fibers with larger diameters (see Fig. 17.7c), and large-diameter fibers have a lower threshold. Since the physiologic recruitment order by the CNS is to first recruit the small fibers followed by large ones, electric stimulation produces a reverse recruitment order. However, techniques have been developed to recruit small fibers before large fibers by using a different stimulation waveform [Fang and Mortimer, 1991]. Since λ2 is also dependent on the electrical properties of the axons, it is then possible to predict that fibers with a larger membrane resistance or lower axoplasmic resistance also will have lower thresholds.

Spatial Derivative of the Electric Field The first spatial derivative (or the second spatial derivative of the voltage along the nerve) is responsible for electric excitation of the nerve. Therefore, an electric field with a nonzero second spatial derivative is required for excitation. An axon with a linearly decreasing voltage distribution would not be excited despite the presence of a large voltage difference along the axon. This is due to the fact that a linear voltage distribution gives a constant electric field, and therefore, the spatial derivative of the field is zero.

Activating Function The second spatial derivative term of the equivalent voltage is also known as the activation function [Rattay, 1990]:

funmy =


d 2Ve dx 2

(17.30)

This function can be evaluated from knowledge of the extracellular voltage alone and can be used to predict the location of excitation. For unmyelinated fibers, the activation function does not contain any information about the axon to be stimulated. In the case of myelinated fibers, where the voltage is evaluated at the nodes of Ranvier, the activating function becomes

fmy =

∆2Ve ∆x 2

(17.31)

The new function contains implicit information about the fiber diameter, since the distance L between the nodes of Ranvier is directly proportional to the fiber diameter D (L = 100 × D). the diameter dependence can be made explicit in Eq. (17.29).

Net Driving Function The equivalent voltage source or the activating function represents only the source term in the electrical model of the axon (see fig. 17.6). However, the transmembrane voltage is determined by a weighted sum of the currents flowing at all the nodes. A net driving function that takes into account both the source term at each node and the passive redistribution from sources at other nodes has been defined and found useful for accurate prediction of the excitation threshold for any applied field [Warman et al., 1994].

Current-Distance Relationship The amount of current required to activate a fiber with a given diameter depends on its geometry but also on its distance from the electrode. The farther away the fiber, the lower is the voltage along the fiber (17.15); therefore, larger current will be required to reach threshold. This effect is illustrated in Fig. 17.9 for myelinated fibers. The distance as a function of current amplitude at threshold is plotted for several experiments [Ranck, 1975]. With a current of 1mA, all fibers within 2 mm are activated. The calculated current-distance relationship for a 10-µm fiber has been shown to approximate well the experimental data (see dashed line in Fig. 17.9) [Rattay, 1989]. The current-distance relationship is linear only for small distances. For distances above 1mm, doubling the distance will require four times the current amplitude.

FIGURE 17.9 Current-distance relationship for monopolar cathodic stimulation of myelinated axons. The distance between the axon and the electrode is plotted as a function of current threshold amplitude for many experiments from several authors. The dashed line shows the current-distance relationship calculated for a 10-µm fiber stimulated with a 200-µs pulse. [From Rattay, 1989, with permission.]


FIGURE 17.10

(a) Longitudinal/transverse electrode placement. (b) Anodal surround block.

Longitudinal/Transverse Field The equivalent voltage source is also proportional to the second spatial derivative of the potential present at each point along the axon. It is important to note that it is the longitudinal component of the electric field that is responsible for exciting the nerve. Therefore, electrodes placed longitudinally generate the most efficient stimulus, since this placement would produce a large longitudinal electric field component. Conversely, electrodes placed transversely (on each side of the axon) require a much higher current, since the largest component of the field does not contribute to excitation of the nerve (Fig. 17.10a).

Anodal Surround Block As shown by the current-distance relation, the current amplitudes required for excitation decrease with distance. This is not entirely true for cathodic stimulation. Cathodic stimulation produces membrane depolarization underneath the electrode and membrane hyperpolarization on both sides of the electrodes (see Fig. 17.7b). As the current amplitude is increased, the hyperpolarization also increases and can block propagation of the action potential along the axon. This effect is known as anodal surround block and is shown in Fig. 17.10b. It is possible to identify three regions around the electrode each giving different responses. There is a spherical region close to the electrode (I) in which no excitation will take place due to the surround block. Fibers located in the region (II) are excited, and fibers still further away (III) from the electrode are below threshold and are not excited.


Unidirectionally Propagated Action Potentials Electric stimulation of nerves with electrodes normally depolarizes the membrane to threshold, producing two action potentials propagating in opposite directions, as shown in Fig. 17.1. Stimulation techniques have been developed to generate action potentials propagating in one direction only [Sweeney and Mortimer, 1986]. The techniques rely on the fact that bipolar stimulation generates depolarization under the cathode and hyperpolarization under the anode. By increasing the amount of hyperpolarization relative to the depolarization, the active potential generated underneath the cathode and traveling toward the anode can be blocked, while the action potential traveling in the other direction can escape.

17.5 Electrode-Tissue Interface At the interface between the electrode and the tissue, the shape of the waveform can influence the threshold for activation as well as the corrosion of the electrode and the tissue damage generated.

Effect of Stimulation Waveform on Threshold Strength-Duration Curve It has been known for a long time that it is the time change in the applied current and not the continuous application of the external stimulus that can excite. Direct current (dc current) cannot excite and even in small amplitude can cause significant tissue damage. It also has been observed experimentally that the relationship between the pulse width and the amplitude suggests that it is the total charge injected that is the important parameter. This relationship between the amplitude and the width of a pulse required to bring an excitable tissue to threshold is shown in Fig. 17.11a. The amplitude of the current threshold stimulus Ith decreases with increasing pulse width W and can be modeled by the following relationship derived experimentally [Lapicque, 1907]:

I th =

I rh  W 1 − exp  −   T

(17.32)

The smallest current amplitude required to cause excitation is known as the rheobase current Irh . T is the membrane time constant of the axon is stimulated intracellularly. For extracellular stimulation, T is a time constant that takes into account the extracellular space resistance. The relationship between current amplitude and pulse width also can be derived theoretically using the cable equation by assuming that total charge on the cable for excitation is constant [Jack et al., 1983]. Charge Duration Curve The threshold charge injected Qth = Ith × W is plotted in Fig. 17.11b and increases as the pulse width increases:

Qth =

I rh W  W 1 − exp  −   T

(17.33)

The increase in the amount of charge required to fire the axon with increasing pulse width is due to the fact that for long pulse duration, the charge is distributed along the cable and does not participate directly in raising the membrane voltage at the excitation site. The minimum amount of charge Qmin required for stimulation is obtained by taking the limit of Qth (17.33) when W goes to zero is equal to IrhT. In practice, this minimum charge can be nearly achieved by using narrow, current pulses.


FIGURE 17.11 Effect of pulse width on excitation threshold. (a) Strength-duration curve. The amplitude of the current required to reach threshold decreases with width of the pulse. This effect can be derived theoretically by assuming that the charge on the cable at threshold is constant. (b) The threshold charge (amplitude times pulsewidth) injected into the tissue increases with the pulsewidth. Narrow pulses are recommended to minimize charge injection.

Anodic Break Excitation generated by cathodic current threshold normally takes place at the onset of the pulse. However, long-duration, subthreshold cathodic or anodic current pulses have been observed experimentally to generate excitation at the end of the pulse. This effect has been attributed to the voltage sensitivity of the sodium channel. The sodium channel is normally partially inactivated at rest (see Chap. 1). However, when the membrane is hyperpolarized during the long-duration pulse, the inactivation of the sodium channel is completely removed. Upon termination of the pulse, an action potential is generated, since the inactivation gate has a slow time constant relative to the activation time constant and cannot recover fast enough [Mortimer, Chap. 3 in Agnew and McCreery, 1990]. This effect can be observed with an anodic or cathodic pulse, since both can generate hyperpolarization. Anodic break can be prevented by avoiding abrupt termination of the current. Pulse shapes with slow decay phases such as exponential or trapezoidal decay shapes have been issued successfully [Fang and Mortimer, 1991].

Electrochemistry of Stimulation Conduction in metal is carried by electrons, while in tissue, current is carried by ions. Although capacitive mechanisms have been tested, electrodes have not yet been developed that can store enough charge for stimulation. Therefore, most electric stimulation electrodes rely on faradaic mechanisms at the interface


between the metal and the tissue. Faradaic mechanisms require that oxidation and reduction take place at the interface [Roblee and Rose, Chap. 2 in Agnew and McCreery, 1990]. Faradaic stimulation mechanisms can be divided into reversible and nonreversible mechanisms. Reversible mechanisms occur at or near the electrode potential and include oxide formation and reduction and hydrogen plating. Irreversible mechanisms occur when the membrane is driven far away from its equilibrium potential and include corrosion and hydrogen or oxygen evolution. These irreversible processes can cause damage to both the electrode and the tissue, since they alter the composition of the electrode surface and can generate toxic products with pH changes in the surrounding tissue. During charge injection, the electrode potential is modified by an amount related to the charge density (total charge divided by the surface area). In order to maintain the electrode potential within regions producing only minimal irreversible changes, this charge density must be kept below some values. The maximum charge density allowed depends on the metal used for the electrode, the stimulation waveform, the type of electrode used, and the location of the electrode within the body.

Stimulation of Brain Tissue Electrodes can be placed on the surface of the brain and directly into the brain to activate CNS pathways. Experiments with stimulation of electrode arrays made of platinum and placed on the surface of the brain indicate that damage was produced for charge densities between 50 to 300 µC/cm2 but that the total charge per phase also was an important factor and should be kept below 3 µC [Pudenz et al., 1975a, b]. Intracortical electrodes with small surface areas can tolerate a charge density as high as 1600 µC/cm2, provided that the charge per phase remains below 0.0032 µC [Agnew et al., 1986].

Stimulation of the Peripheral Nerve The electrodes used in the peripheral nerves are varied and include extraneural designs such as the spiral cuffs or helix electrodes [Naples et al., 1991] with electrode contacts directly on the surface of the nerve or intraneural designs placing electrodes contacts directly inside the nerve [Nannini and Horch, 1991; Rutten et al., 1991]. Intraneural electrodes can cause significant damage, since electrodes are inserted directly into the nerve through the perineurium. However, these electrodes can display good selectivity. Extraneural electrodes are relatively safe because the newer designs such as the spiral or helix are selfsizing and allow for swelling without compression but display poor selectivity. New designs are aimed at producing selective stimulation by recruiting only small portions of the nerve and little damage [Veraart et al., 1990; Tyler and Durand, 1997]. Damage in peripheral nerve stimulation can be caused by the constriction of the nerve as well as by neuronal hyperactivity and irreversible reactions at the electrode [McCreery et al., 1992].

Stimulation of Muscle Tissue Muscle tissue can best be excited by electrodes located on the nerve supplying the muscle [Popovic, 1991]. However, for some applications, electrodes can be placed directly on the surface of the skin (surface stimulation) [Myklebust et al., 1985], directly into the muscle (intramuscular electrode) [Caldwell and Reswick, 1975], or on the surface of the muscle (epimysial electrode) [Grandjean and Mortimer, 1985]. The current thresholds are higher when compared with nerve stimulation unless the electrode is carefully placed near the point of entry of the nerve [Mortimer, 1981]. Stainless steel is often used for these electrodes and is safe below 40 µC/cm2 for coiled wire intramuscular electrodes [Mortimer et al., 1980].

Corrosion Corrosion of the electrode is a major concern because it can cause electrode damage, metal dissolution, and tissue damage. However, corrosion occurs only during the anodic phase of the stimulation. Therefore, by using monophasic waveforms as shown in Fig. 17.12b, corrosion can be avoided. Conversely, the monophasic anodic waveform (Fig. 17.12a) must be avoided because it will cause corrosion. (This not


FIGURE 17.12 Comparison of stimulation waveforms. The waveforms are ranked for their ability to generate low threshold stimulation, low corrosion, and low tissue damage.

true, however, for capacitive electrode metals such as tantalum, for which a dielectric layer of tantalum pentoxide is formed during the anodic phase and reduced during the cathodic phase.) For most applications, cathodic stimulation has a lower threshold than anodic stimulation. It appears, therefore, that monophasic cathodic waveforms (Fig. 17.11a) would be a preferred stimulation waveform, since it minimizes both the current to be injected and the corrosion. However, since the current only flows in one direction, the chemical reactions at the interface are not reversed, and the electrode is driven in the irreversible region.

Tissue Damage Electrodes operating in the irreversible region can cause significant tissue damage because irreversible processes can modify the pH of the surrounding tissue and generate toxic products. Balanced biphasic waveforms are preferred because the second phase can completely reverse the charge injected into the tissue. Provided that the amplitude of the current is small, the electrode voltage can then be maintained within the reversible region. Waveforms that have the most unrecoverable charge are the most likely to induce tissue damage. Tissue damage also can be caused by generating a high rate of neural activity


[Agnew et al., Chap. 6 in Agnew and McCreery, 1990]. The mechanisms underlying this effect are still unclear but could include damage to the blood-nerve barrier, ischemia, or a large metabolic demand on the tissue leading to changes in ionic concentration both intracellularly and extracellularly.

Biphasic Waveforms Common biphasic waveforms for stainless steel or platinum use a cathodic pulse followed by an anodic phase. An example of a square balanced biphasic waveform is shown in Fig. 17.12c. Another commonly used biphasic waveform easily implemented with capacitor and switches is shown in Fig. 17.12e. This waveform ensures that the charge is exactly balanced, since a capacitor is inserted in series with the tissue to be stimulated and the charge injected is then reversed by discharging the capacitor [Mortimer, 1981]. Biphasic cathodic-first waveforms have a higher threshold than monophasic waveforms because the maximum depolarization induced by the cathodic pulse is decreased by the following anodic pulse [Mortimer, Chap. 3 in Agnew and McCreery, 1990]. A delay can then be inserted between the cathodic and anodic phases, as shown in Fig. 17.12d. However, the time delay also can prevent adequate charge reversal and can be dangerous to the electrode and the tissue. An alternative method is to decrease the maximum amplitude of the anodic phase but increase its length as shown in Fig. 17.12f. However, this also can be damaging to the electrode because the charge is not reversed fast enough following the cathodic pulse. The various waveforms shown in Fig. 17.12 are ranked for their effect on tissue damage, corrosion, and threshold of activation.

17.6 Conclusion Electric stimulation of excitable tissue has been used for over a century, and important discoveries have been made concerning the mechanisms underlying the interaction between the applied fields with the tissue and the electrochemistry at the electrode-tissue interface. It is now clear that electric stimulation is a powerful technique that could potentially activate any excitable tissue in the body and replace damaged functions. It is also clear, however, that if our goal is to provide intimate interfacing between excitable tissue and electrodes in order to reproduce the normal function of the nervous system, present technology is not adequate. The electrodes are too big and can only access either a few elements separately or a large number of elements simultaneously. Moreover, it is also clear that our understanding of the electrochemistry at the interface between the electrode and tissue is limited as well as the mechanisms underlying tissue damage. However, the substantial gains to be made are worth the efforts required to solve these problems.

Acknowledgment I am grateful to Srikantan Nagarajan for critically reviewing this manuscript. Supported by NIH grant RO1 NS 32845-01.

References Agnew WF, McCreery DB. 1990. Neural Prostheses: Fundamental Studies. Englewood Cliffs, NJ, PrenticeHall. Agnew WF, Yuen TGH, McCreery DB, Bullara LA. 1986. Histopathologic evaluation of prolonged intracortical electrical stimulation. Exp Neurol 92:162. Aidley DJ. 1978. The Physiology of Excitable Cells. Cambridge, England, Cambridge University Press. Altman KW, Plonsey R. 1988. Development of a model for point source electrical fibre bundle stimulation. Med Biol Eng Comput 26:466. Basser PJ. 1993. Cable equation for a myelinated axon derived from its microstructure. Med Biol Eng Comput 31:S87.


Brindley GS, Lewin WS. 1968. The sensations produced by electrical stimulation of the visual cortex. J Physiol 106:479. Caldwell CW, Reswick JB. 1975. A percutaneous wire electrode for chronic research use. IEEE Trans Biomed Eng 22:429. Chokroverty S. 1990. Magnetic Stimulation in Clinical Neurophysiology. London, Butterworth. Clark GM, Tong YC, Patrick JF. 1990. Cochlear Prostheses. New York: Churchill-Linvinston. Clark J, Plonsey R. 1966. A mathematical evaluation of the core conductor model. Biophys J 6:95. Durand D, Ferguson ASF, Dalbasti T. 1992. Effects of surface boundary on neuronal magnetic stimulation. IEEE Trans Biomed Eng 37:588. Fang ZP, Mortimer JT. 1991. A method to effect physiological recruitment order in electrically activated muscle. IEEE Trans Biomed Eng 38:175. Ferreira HG, Marshal MW. 1985. The Biophysical Basis of Excitibility. Cambridge, England, Cambridge University Press. Fischer RS, Krauss GL, Ramsey E, Laxer K, Gates K. 1997. Assessment of vagus nerve stimulation for epilepsy: report of the Therapeutic and Technology Assessment Subcommittee of the American Academy of Neurology. Neurology 49(1):293. Glenn WWL, Hogan JF, Loke JSO, et al. 1984. Ventilatory support by pacing of the conditioned diaphragm in quadraplegia. N Engl J Med 310:1150. Grandjean PA, Mortimer JT. 1985. Recruitment properties of monopolar and bipolar epimysial electrodes. Ann Biomed Eng 14:429. Hambrecht FT. 1979. Neural prostheses. Annu Rev Biophys Bioeng 8:239. Heringa A, Stegeman DF, Uijen GJH, deWeerd JPC. 1982. Solution methods of electrical field in physiology. IEEE Trans Biomed Eng 29:34. Hines M. 1984. Efficient computation of branched nerve equations. Int J Biol Med Comput 15:69. Jack JJB, Noble D, Tsien RW. 1983. Electrical Current Flow in Excitable Cells. Oxford, Clarendon Press. Koch C, Segev I. 1989. Methods in Neural Modeling. Cambridge, Mass, MIT Press. Kraus JD, Carver KR. 1973. Electromagnetics. New York, McGraw-Hill. Lapicque L. 1907. Recherches quantitatives sur l’excitation electrique des nerfs traites comme une polarization. J Physiol (Paris) 9:622. Marsolais EB, Kobetic R. 1988. Development of a practical electrical stimulation system for restoring gait in the paralyzed patient. Clin Orthop 233:64. McCreery DB, Agnew WF. Yuen TGH, Bullara LA. 1992. Damage in peripheral nerve from continuous electrical stimulation: Comparison of two stimulus waveforms. Med Biol Eng Comput 30:109. McNeal DR. 1976. Analysis of a model for excitation of myelinated nerve. IEEE Trans Biomed Eng 23:329. Mortimer JT, Kaufman D, Roessmann U. 1980. Coiled wire electrode intramuscular electrode: Tissue damage. Ann Biomed Eng 8:235. Mortimer JT. 1981. Motor prostheses. In VB Brooks (ed), Handbook of Physiology—The Nervous System, vol 3, pp 155–187. New York, American Physiological Society. Myklebust J, Cusick B, Sances A, Larson SLJ. 1985. Neural Stimulation. Boca Raton, Fla, CRC Press. Nagarajan SS, Durand DM, Hsuing-Hsu K. 1997. Mapping location of excitation during magnetic stimulation. Effect of coil position. Ann Biomed Eng 25:112. Nannini N, Horch K. 1991. Muscle recruitment with intrafascicular electrodes. IEEE Trans Biomed Eng 38:769. Naples et al. 1990. Overview of peripheral nerve electrode design and implementation. In WF Agnew and DB McCreery (eds), Neural Prostheses: Fundamental Studies. Englewood Cliffs, New Jersey, Prentice-Hall. Nunez PL. 1981. Electric Fields in the Brain: The Neurophysics of EEG. Oxford, England, Oxford University Press. Peckham PH, Keith MW, Freehofe AA. 1987. Restoration of functional control by electrical stimulation in the upper extremity of the quadriplegic patient. J Bone and Joint Surg. 70A(1):144–148. Plonsey R. 1969. Bioelectric Phenomena. New York, McGraw-Hill.


Popovic´ D, Gordon T, ReeJuse VF, Prochazka A. 1991. Properties of implanted electrodes for functional electrical stimulation. Ann Biomed Eng 19:303–316. Pudenz RHL, Bullara A, Dru D, Tallala A. 1975a. Electrical stimulation of the brain: II. Effects on the blood-brain barrier. Surg Neurol 4:2650. Pudenz RH, Bullara LA, Jacques P, Hambrecht FT. 1975b. Electrical stimulation of the brain: III. The neural damage model. Surg Neurol 4:389. Rall W. 1979. Core conductor theory and cable properties of neurons. In Handbook of Physiology—The Nervous System, vol 1, chap 3, pp 39–96. Bethesda, Maryland, American Physiological Society. Ranck JB. 1975. Which elements are excited in electrical stimulation of mammalian central nervous system: A review. Brain Res 98:417. Rattay F. 1989. Analysis of models for extracellular fiber stimulation. IEEE Trans Biomed Eng 36:676. Rattay F. 1990. Electrical Nerve Stimulation, Theory, Experiments and Applications. New York, SpringerVerlag. Roth BJ. 1979. Core conductor theory and cable properties of neurons. In Handbook of Physiology: The Nervous System, vol 1, pp 39–96. Bethesda, Maryland, American Physiological Society. Roth BJ, Basser PJ. 1990. A model for stimulation of a nerve fiber by electromagnetic induction. IEEE Trans Biomed Eng 37:588. Rutecki P. 1990. Anatomical, physiological and theoretical basis for the antiepileptic effect of vagus nerve stimulation. Epilepsia 31:S1. Rutten WLC, van Wier HJ, Put JJM. 1991. Sensitivity and selectivity of intraneural stimulation using a silicon electrode array. IEEE Trans Biomed Eng 38:192. Sweeney JD, Mortimer JT. 1986. An asymmetric two electrode cuff for generation of unidirectionally propagated action potentials. IEEE Trans Biomed Eng 33:541. Tyler D, Durand DM. 1997. Slowly penetrating intrafascicular electrode for electrical simulation of nerves. IEEE Trans Rehab 5:51. Veraart C, Grill WM, Mortimer JT. 1990. Selective control of muscle activation with a multipolar nerve cuff electrode. IEEE Trans Biomed Eng 37:688. Warman NR, Durand DM, Yuen G. 1994. Reconstruction of hippocampal CA1 pyramidal cell electrophysiology by computer simulation. IEEE Trans Biomed Eng 39:1244.

Further Information Information concerning functional electrical stimulation can be obtained from FES Information Center 11000 Cedar Av., Cleveland, Ohio 44106-3052 Tel: 800 666 2353 or 216 231 3257 Articles concerning the latest research can be found mainly in the following journals: IEEE Transactions in Biomedical Engineering IEEE Transactions on Rehabilitation Engineering Annals of Biomedical Engineering Medical & Biological Engineering and Computing


Schneck, D. J. “Biomechanics.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

III Biomechanics Daniel J. Schneck Virginia Polytechnic Institute and State University 18 Mechanics of Hard Tissue J. Lawrence Katz Structure of Bone • Composition of Bone • Elastic Properties • Characterizing Elastic Anisotropy • Modeling Elastic Behavior • Viscoelastic Properties • Related Research

19 Mechanics of Blood Vessels Thomas R. Canfield, Philip B. Dobrin Assumptions • Vascular Anatomy • Axisymmetric Deformation • Experimental Measurements • Equilibrium • Strain Energy Density Functions

20 Joint-Articulating Surface Motion

Kenton R. Kaufman, Kai-Nan An

Ankle • Knee • Hip • Shoulder • Elbow • Wrist • Hand • Summary

21 Joint Lubrication

Michael J. Furey

Tribology • Lubrication • Synovial Joints • Theories on the Lubrication of Natural and Normal Synovial Joints • In Vitro Cartilage Wear Studies • Biotribology and Arthritis: Are There Connections? • Recapitulation and Final Comments • Conclusions

22 Musculoskeletal Soft Tissue Mechanics Richard L. Lieber, Thomas J. Burkholder Fundamentals of Soft Tissue Biomechanics

23 Mechanics of Head/Neck Albert I. King, David C. Viano Mechanisms of Injury • Mechanical Response • Regional Tolerance of the Head and Neck to Blunt Impact • Human Surrogates of the Head and Neck

24 Biomechanics of Chest and Abdomen Impact David C. Viano, Albert I. King Chest and Abdomen Injury Mechanisms • Injury Tolerance Criteria • Biomechanical Responses During Impact • Injury Risk Assessment

25 Analysis of Gait

Roy B. Davis, Peter A. DeLuca, Sylvia Õunpuu

Fundamental Concepts • Gait Data Reduction • Illustrative Clinical Case • Gait Analysis: Current Status

26 Exercise Physiology

Arthur T. Johnson, Cathryn R. Dooly

Muscle Energetics • Cardiovascular Adjustments • Maximum Oxygen Uptake • Respiratory Responses • Optimization • Thermal Responses • Applications

27 Factors Affecting Mechanical Work in Humans Arthur T. Johnson, Bernard F. Hurley Exercise Biomechanics • Exercise Training • Age • Gender • Ergogenic Aids

28 Cardiac Biodynamics Andrew D. McCulloch Cardiac Geometry and Structure • Cardiac Pump Function • Myocardial Material Properties • Regional Ventricular Mechanics: Stress and Strain


29 Heart Valve Dynamics Ajit P. Yoganathan, Jack D. Lemmon, Jeffrey T. Ellis Aortic and Pulmonic Valves • Mitral and Tricuspid Valves

30 Arterial Macrocirculatory Hemodynamics Baruch B. Lieber Blood Vessel Walls • Flow Characteristics • Wave Propagation • Velocity Profiles • Pathology

31 Mechanics and Transport in the Microcirculation Roland N. Pittman

Aleksander S. Popel,

Mechanics of Microvascular Blood Flow • Mass Transport in the Microcirculation • Regulation of Blood Flow

32 Mechanics and Deformability of Hematocytes Richard E. Waugh, Robert M. Hochmuth Fundamentals • Red Cells • White Cells

33 The Venous System

Artin A. Shoukas, Carl F. Rothe

Definitions • Methods to Measure Venous Characteristics • Typical Values

34 Mechanics of Tissue and Lymphatic Transport Geert W. Schmid-Schönbein

Alan R. Hargen,

Basic Concepts of Tissue and Lymphatic Transport

35 Cochlear Mechanics Charles R. Steele, Gary J. Baker, Jason A. Tolomeo, Deborah E. Zetes-Tolomeo Anatomy • Passive Models • The Active Process • Active Models • Fluid Streaming? • Clinical Possibilities

36 Vestibular Mechanics Wallace Grant Structure and Function • Otolith Distributed Parameter Model • Nondimensionalization of the Motion Equations • Otolith Transfer Function • Otolith Frequency Response • Semicircular Canal Distributed Parameter Model • Semicircular Canal Frequency Response

M

ECHANICS IS THE ENGINEERING SCIENCE that deals with studying, defining, and mathematically quantifying “interactions” that take place among “things” in our universe. Our ability to perceive the physical manifestation of such interactions is embedded in the concept of a force, and the “things” that transmit forces among themselves are classified for purposes of analysis as being solid, fluid, or some combination of the two. The distinction between solid behavior and fluid behavior has to do with whether or not the “thing” involved has disturbance-response characteristics that are time-rate-dependent. A constant force transmitted to a solid material will generally elicit a discrete, finite, time-independent deformation response, whereas the same force transmitted to a fluid will elicit a continuous, time-dependent response called flow. In general, whether or not a given material will behave as a solid or a fluid often depends on its thermodynamic state (i.e., its temperature, pressure, etc.). Moreover, for a given thermodynamic state, some “things” are solid-like when deformed at certain rates but show fluid behavior when disturbed at other rates, so they are appropriately called viscoelastic, which literally means, “fluid-solid.” Thus a more technical definition of mechanics is the science that deals with the action of forces on solids, fluids, and viscoelastic materials. What makes mechanics biomechanics is the fact that biomechanics is the science that deals with the time and space response characteristics of biologic solids, fluids, and viscoelastic materials to imposed systems of internal and external forces. As early as the fourth century B.C., we find in the works of Aristotle (384–322 B.C.) attempts to describe through geometric analysis the mechanical action of muscles in producing locomotion of parts or all of the animal body. Nearly two thousand yeas later, in his famous anatomic drawings, Leonardo da Vinci (A.D. 1452–1519) sought to describe the mechanics of standing, walking up and down hill, rising from a sitting position, and jumping, and Galileo (A.D. 1564–1643) followed a hundred years later with some of the earliest attempts to mathematically analyze physiologic function. Because of his pioneering efforts in defining the anatomic circulation of blood, William Harvey (A.D. 1578–1657) is credited by many as being the father of modern-day biofluid mechanics, and Alfonso Borelli (A.D. 1608–1679) shares the same honor for contemporary biosolid mechanics because of his efforts to explore the amount of force produced by various muscles and his theorization that bones serve as levers that are operated and © 2000 by CRC Press LLC

controlled by muscles. The early work of these pioneers of biomechanics was followed up by the likes of Sir Isaac Newton (A.D. 1642–1727), Daniel Bernoulli (A.D.• 1700–1782), Jean L. M. Poiseuille (A.D. 1799–1869), Thomas Young (A.D. 1773–1829), Euler (whose work was published in 1862), and others of equal fame. To enumerate all their individual contributions would take up much more space than is available in this short introduction, but there is a point to be made if one takes a closer look at the names involved. In reviewing the preceding list of biomechanical scientists, it is interesting to observe that many of the earliest contributions to our ultimate understanding of the fundamental laws of physics and engineering (e.g., Bernoulli’s equation of hydrodynamics, the famous Young’s modulus in elasticity theory, Poiseuille flow, and so on) came from physicians, physiologists, and other health care practitioners seeking to study and explain physiologic structure and function. The irony in this is that as history has progressed, we have just about turned this situation completely around. That is, more recently, it has been biomedical engineers who have been making the greatest contributions to the advancement of the medical and physiologic sciences. These contributions will become more apparent in the chapters that follow in this section. The individual chapters address the subjects of biosolid mechanics and biofluid mechanics as they pertain to various subsystems of the human body. Since the physiologic organism is 60 to 75% fluid, it is not surprising that the subject of biofluid mechanics should be so extensive, including—but not limited to—lubrication of human synovial joints (Chapter 21), cardiac biodynamics (Chapter 28), mechanics of heart valves (Chapter 29), arterial macrocirculatory hemodynamics (Chapter 30), mechanics and transport in the microcirculation (Chapter 31), venous hemodynamics (Chapter 33), mechanics of the lymphatic system (Chapter 34), cochlear mechanics (Chapter 35), and vestibular mechanics (Chapter 36). The area of biosolid mechanics is somewhat more loosely defined—since all physiologic tissue is viscoelastic and not strictly solid in the engineering sense of the word. Also generally included under this heading are studies of the kinematics and kinetics of human posture and locomotion, i.e., biodynamics, so that under the generic section on biosolid mechanics in this Handbook you will find chapters addressing the mechanics of hard tissue (Chapter 18), the mechanics of blood vessels (Chapter 19) or, more generally, the mechanics of viscoelastic tissue, mechanics of joint articulating surface motion (Chapter 20), musculoskeletal soft tissue mechanics (Chapter 22), mechanics of the head/neck (Chapter 23), mechanics of the chest/abdomen (Chapter 24), the analysis of gait (Chapter 25), exercise physiology (Chapter 26), biomechanics and factors affecting mechanical work in humans (Chapter 27), and mechanics and deformability of hematocytes (blood cells) (Chapter 32). In all cases, the ultimate objectives of the science of biomechanics are generally twofold. First, biomechanics aims to understand fundamental aspects of physiologic function for purely medical purposes, and second, it seeks to elucidate such function for mostly nonmedical applications. In the first instance above, sophisticated techniques have been and continue to be developed to monitor physiologic function, to process the data thus accumulated, to formulate inductively theories that explain the data, and to extrapolate deductively, i.e., to diagnose why the human “engine” malfunctions as a result of disease (pathology), aging (gerontology), ordinary wear and tear from normal use (fatigue), and/or accidental impairment from extraordinary abuse (emergency medicine). In the above sense, engineers deal directly with causation as it relates to anatomic and physiologic malfunction. However, the work does not stop there, for it goes on to provide as well the foundation for the development of technologies to treat and maintain (therapy) the human organism in response to malfunction, and this involves biomechanical analyses that have as their ultimate objective an improved health care delivery system. Such improvement includes, but is not limited to, a much healthier lifestyle (exercise physiology and sports biomechanics), the ability to repair and/or rehabilitate body parts, and a technology to support ailing physiologic organs (orthotics) and/or, if it should become necessary, to replace them completely (with prosthetic parts). Nonmedical applications of biomechanics exploit essentially the same methods and technologies as do those oriented toward the delivery of health care, but in the former case, they involve mostly studies to define the response of the body to “unusual” environments—such as subgravity conditions, the aerospace milieu, and extremes of temperature, humidity, altitude, pressure, acceleration, deceleration, impact, shock, and vibration, and so on. Additional applications include vehicular safety © 2000 by CRC Press LLC

considerations, the mechanics of sports activity, the ability of the body to “tolerate” loading without failing, and the expansion of the envelope of human performance capabilities—for whatever purpose! And so, with this very brief introduction, let us take somewhat of a closer look at the subject of biomechanics.

Free body diagram of the foot.


Katz, J. L. “ Mechanics of Hard Tissue.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

18 Mechanics of Hard Tissue

J. Lawrence Katz Department of Biomedical Engineering, Case Western Reserve University

18.1 18.2 18.3 18.4 18.5 18.6 18.7

Structure of Bone Composition of Bone Elastic Properties Characterizing Elastic Anisotropy Modeling Elastic Behavior Viscoelastic Properties Related Research

Hard tissue, mineralized tissue, and calcified tissue are often used as synonyms for bone when describing the structure and properties of bone or tooth. The hard is self-evident in comparison with all other mammalian tissues, which often are referred to as soft tissues. Use of the terms mineralized and calcified arises from the fact that, in addition to the principle protein, collagen, and other proteins, glycoproteins, and protein-polysaccherides, comprising about 50% of the volume, the major constituent of bone is a calcium phosphate (thus the term calcified) in the form of a crystalline carbonate apatite (similar to naturally occurring minerals, thus the term mineralized). Irrespective of its biological function, bone is one of the most interesting materials known in terms of structure-property relationships. Bone is an anisotropic, heterogeneous, inhomogeneous, nonlinear, thermorheologically complex viscoelastic material. It exhibits electromechanical effects, presumed to be due to streaming potentials, both in vivo and in vitro when wet. In the dry state, bone exhibits piezoelectric properties. Because of the complexity of the structure-property relationships in bone, and the space limitation for this chapter, it is necessary to concentrate on one aspect of the mechanics. Currey [1984] states unequivocally that he thinks, “the most important feature of bone material is its stiffness.” This is, of course, the premiere consideration for the weight-bearing long bones. Thus, this chapter will concentrate on the elastic and viscoelastic properties of compact cortical bone and the elastic properties of trabecular bone as exemplar of mineralized tissue mechanics.

18.1 Structure of Bone The complexity of bone’s properties arises from the complexity in its structure. Thus it is important to have an understanding of the structure of mammalian bone in order to appreciate the related properties. Figure 18.1 is a diagram showing the structure of a human femur at different levels [Park, 1979]. For convenience, the structures shown in Fig. 18.1 will be grouped into four levels. A further subdivision of structural organization of mammalian bone is shown in Fig. 18.2 [Wainwright et al., 1982]. The individual figures within this diagram can be sorted into one of the appropriate levels of structure shown on Fig. 18.1 as described in the following. At the smallest unit of structure we have the tropocollagen molecule and


FIGURE 18.1 Park).

Hierarchical levels of structure in a human femur [Park, 1979] (Courtesy of Plenum Press and Dr. J.B.

the associated apatite crystallites (abbreviated Ap). The former is approximately 1.5 by 280 nm, made up of three individual left-handed helical polypeptide (alpha) chains coiled into a right-handed triple helix. Ap crystallites have been found to be carbonate-substituted hydroxyapatite, generally thought to be nonstoichiometric. The crystallites appear to be about 4 × 20 × 60 nm in size. This level is denoted the molecular. The next level we denote the ultrastructural. Here, the collagen and Ap are intimately associated and assembled into a microfibrilar composite, several of which are then assembled into fibers from approximately 3 to 5 µm thick. At the next level, the microstructural, these fibers are either randomly arranged (woven bone) or organized into concentric lamellar groups (osteons) or linear lamellar groups (plexiform bone). This is the level of structure we usually mean when we talk about bone tissue properties. In addition to the differences in lamellar organization at this level, there are also two different types of architectural structure. The dense type of bone found, for example, in the shafts of long bone is known as compact or cortical bone. A more porous or spongy type of bone is found, for example, at the articulating ends of long bones. This is called cancellous bone. It is important to note that the material and structural organization of collagen-Ap making up osteonic or haversian bone and plexiform bone are the same as the material comprising cancellous bone. Finally, we have the whole bone itself constructed of osteons and portions of older, partially destroyed osteons (called interstitial lamellae) in the case of humans or of osteons and/or plexiform bone in the case of mammals. This we denote the macrostructural level. The elastic properties of the whole bone results from the hierarchical contribution of each of these levels.

18.2 Composition of Bone The composition of bone depends on a large number of factors: the species, which bone, the location from which the sample is taken, and the age, sex, and type of bone tissue, e.g., woven, cancellous, cortical. However, a rough estimate for overall composition by volume is one-third Ap, one-third collagen and other organic components, and one-third H2O. Some data in the literature for the composition of adult human and bovine cortical bone are given in Table 18.1.


FIGURE 18.2 Diagram showing the structure of mammalian bone at different levels. Bone at the same level is drawn at the same magnification. The arrows show what types may contribute to structures at higher levels [Wainwright et al., 1982] (Courtesy Princeton University Press). (a) Collagen fibril with associated mineral crystals. (b) Woven bone. The collagen fibrils are arranged more or less randomly. Osteocytes are not shown. (c) Lamellar bone. There are separate lamellae, and the collagen fibrils are arranged in “domains” of preferred fibrillar orientation in each lamella. Osteocytes are not shown. (d) Woven bone. Blood channels are shown as large black spots. At this level woven bone is indicated by light dotting. (e) Primary lamellar bone. At this level lamellar bone is indicated by fine dashes. ( f ) Haversian bone. A collection of Haversian systems, each with concentric lamellae round a central blood channel. The large black area represents the cavity formed as a cylinder of bone is eroded away. It will be filled in with concentric lamellae and form a new Haversian system. (g) Laminar bone. Two blood channel networks are exposed. Note how layers of woven and lamellar bone alternate. (h) Compact bone of the types shown at the lower levels. (i) Cancellous bone.

TABLE 18.1 Composition of Adult Human and Bovine Cortical Bone Species % H2O Ap % Dry Weight Collagen GAG* Bovine Human

9.1 7.3

76.4 67.2

* Glycosaminoglycan † Not determined


21.5 21.2

N.D 0.34

†

Reference

Herring, 1977 Pellagrino and Blitz, 1965; Vejlens, 1971

18.3 Elastic Properties Although bone is a viscoelastic material, at the quasi-static strain rates in mechanical testing and even at the ultrasonic frequencies used experimentally, it is a reasonable first approximation to model cortical bone as an anisotropic, linear elastic solid with Hooke’s law as the appropriate constitutive equation. Tensor notation for the equation is written as:

σ ij = Cijkl ekl

(18.1)

where σij and εkl are the second-rank stress and infinitesimal second rank strain tensors, respectively, and Cijkl is the fourth-rank elasticity tenor. Using the reduced notation, we can rewrite Eq. (18.1) as

σ i = Cij j

i , j = 1 to 6

(18.2)

where the Cij are the stiffness coefficients (elastic constants). The inverse of the Cij , the Sij , are known as the compliance coefficients. The anisotropy of cortical bone tissue has been described in two symmetry arrangements. Lang [1969], Katz and Ukraincik [1971], and Yoon and Katz [1976a,b] assumed bone to be transversely isotropic with the bone axis of symmetry (the 3 direction) as the unique axis of symmetry. Any small difference in elastic properties between the radial (1 direction) and transverse (2 direction) axes, due to the apparent gradient in porosity from the periosteal to the endosteal sides of bone, was deemed to be due essentially to the defect and did not alter the basic symmetry. For a transverse isotropic material, the stiffness matrix [Cij] is given by

C11  C12 C Cij =  13 0 0   0

[ ]

C12 C11 C13

C13 C13 C33

0 0 0

0 0 0

0 0 0 C44 0 0

0 0 0 0 C44 0

0   0  0   0  0  C66 

(18.3)

where C66 = 1/2 (C11 – C12). Of the 12 nonzero coefficients, only 5 are independent. However, Van Buskirk and Ashman [1981] used the small differences in elastic properties between the radial and tangential directions to postulate that bone is an orthotropic material; this requires that 9 of the 12 nonzero elastic constants be independent, that is,

C11  C12 C Cij =  13 0 0   0

[ ]

C12 C22 C23

C13 C23 C33

0 0 0

0 0 0

0 0 0 C44 0 0

0 0 0 0 C55 0

0   0  0   0  0  C66 

(18.4)

Corresponding matrices can be written for the compliance coefficients, the Sij , based on the inverse equation to Eq. (18.2): i = Sij σ j


i , j = 1 to 6

(18.5)

where the Sijth compliance is obtained by dividing the [Cij] stiffness matrix, minus the ith row and jth column, by the full [Cij] matrix and vice versa to obtain the Cij in terms of the Sij . Thus, although S33 = 1/E3 , where E3 is Young’s modulus in the bone axis direction, E3 ≠ C33, since C33 and S33 , are not reciprocals of one another even for an isotropic material, let alone for transverse isotropy or orthotropic symmetry. The relationship between the compliance matrix and the technical constants such as Young’s modulus (Ei) shear modulus (Gi) and Poisson’s ratio (vij) measured in mechanical tests such as uniaxial or pure shear is expressed in Eq. (18.6).

 1   E1  −ν  12  E1  −ν  13 E Sij =  1  0    0    0 

[ ]

− ν21 E2 1 E2 − ν23 E2

− ν31 E3 − ν32 E3 1 E3

0

0

0

0

0

0

0

0

1 G31

0

0

0

0

1 G31

0

0

0

0

 0    0    0    0    0   1   G12 

(18.6)

Again, for an orthotropic material, only 9 of the above 12 nonzero terms are independent, due to the symmetry of the Sij tensor:

ν12 ν21 = E1 E2

ν13 ν31 = E1 E 3

ν23 ν32 = E2 E3

(18.7)

For the transverse isotropic case, Eq. (18.5) reduces to only 5 independent coefficients, since

E1 = E2

ν12 = ν21 G23 = G31

ν31 = ν32 = ν13 = ν23 G12 =

(

E1

2 1 + ν12

)

(18.8)

In addition to the mechanical tests cited above, ultrasonic wave propagation techniques have been used to measure the anisotropic elastic properties of bone [Lang, 1969; Yoon and Katz, 1976a,b; Van Buskirk and Ashman, 1981]. This is possible, since combining Hooke’s law with Newton’s second law results in a wave equation which yields the following relationship involving the stiffness matrix:

ρV 2 U m = Cmrns N r N s U n

(18.9)

where ρ is the density of the medium, V is the wave speed, and U and N are unit vectors along the particle displacement and wave propagation directions, respectively, so that Um, Nr, etc. are direction cosines. Thus to find the five transverse isotropic elastic constants, at least five independent measurements are required, e.g., a dilatational longitudinal wave in the 2 and 1(2) directions, a transverse wave in the 13 (23) and 12 planes, etc. The technical moduli must then be calculated from the full set of Cij . For © 2000 by CRC Press LLC

TABLE 18.2 Elastic Stiffness Coefficients for Various Human and Bovine Bones; All Measurements Made with Ultrasound except for Knets [1978] Mechanical Tests Experiments (Bone Type) Van Buskirk and Ashman [1981] (bovine femur) Knets [1978] (human tibia) Van Buskirk and Ashman [1981] (human femur) Maharidge [1984] (bovine femur haversian) Maharidge [1984] (bovine femur plexiform)

C11 (GPa)

C22 (GPa)

C33 (GPa)

C44 (GPa)

C55 (GPa)

C66 (GPa)

C12 (GPa)

C13 (GPa)

C23 (GPa)

14.1

18.4

25.0

7.00

6.30

5.28

6.34

4.84

6.94

11.6 20.0

14.4 21.7

22.5 30.0

4.91 6.56

3.56 5.85

2.41 4.74

7.95 10.9

6.10 11.5

6.92 11.5

21.2

21.0

29.0

6.30

6.30

5.40

11.7

12.7

11.1

22.4

25.0

35.0

8.20

7.10

6.10

14.0

15.8

13.6

improved statistics, redundant measurements should be made. Correspondingly, for orthotropic symmetry, enough independent measurements must be made to obtain all 9 Cij ; again, redundancy in measurements is a suggested approach. One major advantage of the ultrasonic measurements over mechanical testing is that the former can be done with specimens too small for the latter technique. Second, the reproducibility of measurements using the former technique is greater than for the latter. Still a third advantage is that the full set of either five or nine coefficients can be measured on one specimen, a procedure not possible with the latter techniques. Thus, at present, most of the studies of elastic anisotropy in both human and other mammalian bone are done using ultrasonic techniques. In addition to the bulk wave type measurements described above, it is possible to obtain Young’s modulus directly. This is accomplished by using samples of small cross sections with transducers of low frequency so that the wavelength of the sound is much larger than the specimen size. In this case, an extensional longitudinal (bar) wave is propagated (which experimentally is analogous to a uniaxial mechanical test experiment), yielding

V2 =

E ρ

(18.10)

This technique was used successfully to show that bovine plexiform bone was definitely orthotropic while bovine haversian bone could be treated as transversely isotropic [Lipson and Katz, 1984]. The results were subsequently confirmed using bulk wave propagation techniques with considerable redundancy [Maharidge, 1984]. Table 18.2 lists the Cij (in GPa) for human (haversian) bone and bovine (both haversian and plexiform) bone. With the exception of Knet’s [1978] measurements, which were made using quasi-static mechanical testing, all the other measurements were made using bulk ultrasonic wave propagation. In Maharidge’s study [1984], both types of tissue specimens, haversian and plexiform, were obtained from different aspects of the same level of an adult bovine femur. Thus the differences in Cij reported between the two types of bone tissue are hypothesized to be due essentially to the differences in microstructural organization (Fig. 18.3) [Wainwright et al., 1982]. The textural symmetry at this level of structure has dimensions comparable to those of the ultrasound wavelengths used in the experiment, and the molecular and ultrastructural levels of organization in both types of tissues are essentially identical. Note that while C11, almost equals C22 and that C44 and C55 are equal for bovine haversian bone, C11 and C22 and C44 and C55 differ by 11.6 and 13.4%, respectively, for bovine plexiform bone. Similarly, although C66 and ½ (C11 – C12) differ by 12.0% for the haversian bone, they differ by 31.1% for plexiform bone. Only the differences between C13 and C23 are somewhat comparable: 12.6% for haversian bone and 13.9% for plexiform. These results reinforce the importance of modeling bone as a hierarchical ensemble in order to understand the basis for bone’s elastic properties as a composite material-structure system in


FIGURE 18.3 Diagram showing how laminar (plexiform) bone (a) differs more between radial and tangential directions (R and T) than does haversian bone (b). The arrows are vectors representing the various directions [Wainwright et al., 1982] (Courtesy Princeton University Press).

which the collagen-Ap components define the material composite property. When this material property is entered into calculations based on the microtextural arrangement, the overall anisotropic elastic anisotropy can be modeled. The human femur data [Van Buskirk and Ashman, 1981] support this description of bone tissue. Although they measured all nine individual Cij , treating the femur as an orthotropic material, their results are consistent with a near transverse isotropic symmetry. However, their nine Cij for bovine femoral bone clearly shows the influence of the orthotropic microtextural symmetry of the tissue’s plexiform structure. The data of Knets [1978] on human tibia are difficult to analyze. This could be due to the possibility of significant systematic errors due to mechanical testing on a large number of small specimens from a multitude of different positions in the tibia. The variations in bone’s elastic properties cited earlier above due to location is appropriately illustrated in Table 18.3, where the mean values and standard deviations (all in GPa) for all g orthotropic Cij are given for bovine cortical bone at each aspect over the entire length of bone. Since the Cij are simply related to the “technical” elastic moduli, such as Young’s modulus (E), shear modulus (G), bulk modulus (K), and others, it is possible to describe the moduli along any given direction. The full equations for the most general anisotropy are too long to present here. However, they can be TABLE 18.3 Mean Values and Standard Deviations for the Cij Measured by Van Buskirk and Ashman [1981] at Each Aspect over the Entire Length of Bone (all values in GPa)

C11 C22 C33 C44 C55 C66 C12 C13 C23


Anterior

Medial

Posterior

Lateral

18.7 ± 1.7 20.4 ± 1.2 28.6 ± 1.9 6.73 ± 0.68 5.55 ± 0.41 4.34 ± 0.33 11.2 ± 2.0 11.2 ± 1.1 10.4 ± 1.4

20.9 ± 0.8 22.3 ± 1.0 30.1 ± 2.3 6.45 ± 0.35 6.04 ± 0.51 4.87 ± 0.35 11.2 ± 1.1 11.2 ± 2.4 11.5 ± 1.0

20.1 ± 1.0 22.2 ± 1.3 30.8 ± 1.0 6.78 ± 1.0 5.93 ± 0.28 5.10 ± 0.45 10.4 ± 1.0 11.6 ± 1.7 12.5 ± 1.7

20.6 ± 1.6 22.0 ± 1.0 30.5 ± 1.1 6.27 ± 0.28 5.68 ± 0.29 4.63 ± 0.36 10.8 ± 1.7 11.7 ± 1.8 11.8 ± 1.1

found in Yoon and Katz [1976a]. Presented below are the simplified equations for the case of transverse isotropy. Young’s modulus is

1

( )

E γ3

(

)

= S33 ′ = 1 − γ 23 2S11 + γ 43 S33

(

+ γ 1− γ 2 3

2 3

) (2S

13

+ S44

(18.11)

)

where γ3 = cos φ, and φ is the angle made with respect to the bone (3) axis. The shear modulus (rigidity modulus or torsional modulus for a circular cylinder) is

( ) ( 1

G γ3

=

)

(

(

)

1 1 S44 ′ + S55 ′ = S44 + S11 − S12 − S44 1 − γ 23 2 2

) (

(

+ 2 S11 + S33 − 2S13 − S44 γ 1 − γ 2 3

2 3

)

(18.12)

)

where, again γ3 = cos φ. The bulk modulus (reciprocal of the volume compressibility) is

(

)

C + C + 2C33 − 4C13 1 = S33 + 2 S11 + S12 + 2S13 = 11 12 K C33 C11 + C12 − 2C132

(

)

(18.13)

Conversion of Eqs. (18.11) and (18.12) from Sij to Cij can be done by using the following transformation equations:

S11 =

2 C22 C33 − C23 ∆

S22 =

C33 C11 − C132 ∆

S33 =

C11 C22 − C122 ∆

S12 =

C13 C23 − C12 C33 ∆

C C −C C S13 = 12 23 13 22 ∆ S44 =

1 C44

S55 =

1 C55

(18.14)

C C − C23 C11 S23 = 12 13 ∆ S66 =

1 C66

where

C11  ∆ = C12 C  13

C12 C22 C23

C13   2 + C22 C132 + C33 C122 C23  = C11 C22 C33 + 2 C12 C23 C13 − C11 C23 C33  

(

)

(18.15)

In addition to data on the elastic properties of cortical bone presented above, there is also available a considerable set of data on the mechanical properties of cancellous (trabecullar) bone including measurements of the elastic properties of single trabeculae. Indeed as early as 1993, Keaveny and Hayes (1993)


TABLE 18.4 Elastic Moduli of Trabecular Bone Material Measured by Different Experimental Methods Study Townsend et al. (1975) Ryan and Williams (1989) Choi et al. (1992) Ashman and Rho (1988) Rho et al. (1993) Rho et al. (1999) Turner et al. (1999) Bumrerraj (1999)

Method

Average Modulus

Buckling Buckling Uniaxial tension 4-point bending Ultrasound Ultrasound Ultrasound Tensile Test Nanoindentation Nanoindentation Acoustic microscopy Nanoindentation Acoustic microscopy

11.4 14.1 0.760 5.72 13.0 10.9 14.8 10.4 19.4 15.0 17.5 18.1 17.4

(GPa) (Wet) (Dry)

(Human) (Bovine)

(Longitudinal) (Transverse)

presented an analysis of 20 years of studies on the mechanical properties of trabecular bone. Most of the earlier studies used mechanical testing of bulk specimens of a size reflecting a cellular solid, i.e., of the order of cubic mm or larger. These studies showed that both the modulus and strength of trabecular bone are strongly correlated to the apparent density, where apparent density, ρa , is defined as the product of individual trabeculae density, ρt, and the volume fraction of bone in the bulk specimen, Vf , and is given by ρa = ρtVf . Elastic moduli, E, from these measurements generally ranged from approximately 10 MPa to the order of 1 GPa depending on the apparent density and could be correlated to the apparent density in g/cc by 2 a power law relationship, E = 6.13P144 a , calculated for 165 specimens with an r = 0.62 [Keaveny and Hayes, 1993]. With the introduction of micromechanical modeling of bone, it became apparent that in addition to knowing the bulk properties of trabecular bone it was necessary to determine the elastic properties of the individual trabeculae. Several different experimental techniques have been used for these studies. Individual trabeculae have been machined and measured in buckling, yielding a modulus of 11.4 GPa (wet) and 14.1 GPa (dry) [Townsend et al., 1975], as well as by other mechanical testing methods providing average values of the elastic modulus ranging from less than 1 GPa to about 8 GPa (Table 18.4). Ultrasound measurements [Ashman and Rho, 1988; Rho et al., 1993] have yielded values commensurate with the measurements of Townsend et al. (1975) (Table 18.4). More recently, acoustic microscopy and nanoindentation have been used, yielding values significantly higher than those cited above. Rho et al. (1999) using nanoindentation obtained average values of modulus ranging from 15.0 to 19.4 GPa depending on orientation, as compared to 22.4 GPa for osteons and 25.7 GPa for the interstitial lamellae in cortical bone (Table 18.4). Turner et al. (1999) compared nanoindentation and acoustic microscopy at 50 MHz on the same specimens of trabecular and cortical bone from a common human donor. While the nanoindentation resulted in Young’s moduli greater than those measured by acoustic microscopy by 4 to 14%, the anisotropy ratio of longitudinal modulus to transverse modulus for cortical bone was similar for both modes of measurement; the trabecular values are given in Table 18.4. Acoustic microscopy at 400 MHz has also been used to measure the moduli of both human trabecular and cortical bone [Bumrerraj, 1999], yielding results comparable to those of Turner et al. (1999) for both types of bone (Table 18.4). These recent studies provide a framework for micromechanical analyses using material properties measured on the microstructural level. They also point to using nano-scale measurements, such as those provided by atomic force microscopy (AFM), to analyze the mechanics of bone on the smallest unit of structure shown in Figure 18.1.


18.4 Characterizing Elastic Anisotropy Having a full set of five or nine Cij does permit describing the anisotropy of that particular specimen of bone, but there is no simple way of comparing the relative anisotropy between different specimens of the same bone or between different species or between experimenters’ measurements by trying to relate individual Cij between sets of measurements. Adapting a method from crystal physics [Chung and Buessem, 1968], Katz and Meunier [1987] presented a description for obtaining two scalar quantities defining the compressive and shear anisotropy for bone with transverse isotropic symmetry. Later, they developed a similar pair of scalar quantities for bone exhibiting orthotropic symmetry [Katz and Meunier, 1990]. For both cases, the percentage compressive (Ac*) and shear (As*) elastic anisotropy are given, respectively, by

( )

KV − KR

( )

GV − G R

Ac * % = 100 As * % = 100

KV + KR

(18.16)

G V + GR

where KV and KR are the Voigt (uniform strain across an interface) and Reuss (uniform stress across an interface) bulk moduli, respectively, and GV and GR are the Voigt and Reuss shear moduli, respectively. The equations for KV, KR , GV, and GR are provided for both transverse isotropy and orthotropic symmetry in App. 20A. Table 18.5 lists the values of KV, KR , GV, GR , Ac*, and As* for the five experiments whose Cij are given in Table 18.2. It is interesting to not that haversian bones, whether human or bovine, have both their compressive and shear anisotropy factors considerably lower than the respective values for plexiform bone. Thus, not only is plexiform bone both stiffer and more rigid than haversian bone, it is also more anisotropic. The higher values of Ac* and As*, especially the latter at 7.88% for the Knets [1978] mechanical testing data on human haversian bone, supports the possibility of the systematic errors in such measurements suggested above.

18.5 Modeling Elastic Behavior Currey [1964] first presented some preliminary ideas of modeling bone as a composite material composed of a simple linear superposition of collagen and Ap. He followed this later [1969] with an attempt to take into account the orientation of the Ap crystallites using a model proposed by Cox [1952] for fiberreinforced composites. Katz [1971a] and Piekarski [1973] independently showed that the use of Voigt and Reuss or even Hashin-Shtrikman [1963] composite modeling showed the limitations of using linear combinations of either elastic moduli or elastic compliances. The failure of all these early models could be traced to the fact that they were based only on considerations of material properties. This is comparable to trying to determine the properties of an Eiffel Tower built using a composite material by simply TABLE 18.5 Values of KV, KR, GV, and GR, (all in GPa), and Ac* and As* (%) for the Bone Specimens Given in Table 18.2 Experiments (Bone Type) Van Buskirk and Ashman [1981] (bovine femur) Knets [1978] (human tibia) Van Buskirk and Ashman [1981] (human femur) Maharidge [1984] (bovine femur haversian) Maharidge [1984] (bovine femur plexiform)


KV

KR

GV

GR

Ac*

As*

10.4 10.1 15.5 15.8 18.8

9.87 9.52 15.0 15.5 18.1

6.34 4.01 5.95 5.98 6.88

6.07 3.43 5.74 5.82 6.50

2.68 2.68 1.59 1.11 1.84

2.19 7.88 1.82 1.37 2.85

FIGURE 18.4 Variation in Young’s modulus of bovine femur specimens (E) with the orientation of specimen axis to the long axis of the bone, for wet (o) and dry x) conditions compared with the theoretical curve (———) predicted from a fibre-reinforced composite model [Bonfield and Grynpas, 1977] (Courtesy Nature 1977 270:453. © Macmillan Magazines Ltd.).

modeling the composite material properties without considering void spaces and the interconnectivity of the structure [Lakes, 1993]. In neither case is the complexity of the structural organization involved. This consideration of hierarchical organization clearly must be introduced into the modeling. Katz in a number of papers [1971b, 1976] and meeting presentations put forth the hypothesis that haversian bone should be modeled as a hierarchical composite, eventually adapting a hollow fiber composite model by Hashin and Rosen [1964]. Bonfield and Grynpas [1977] used extensional (longitudinal) ultrasonic wave propagation in both wet and dry bovine femoral cortical bone specimens oriented at angles of 5, 10, 20, 40, 50, 70, 80, and 85 degrees with respect to the long bone axis. They compared their experimental results for Young’s moduli with the theoretical curve predicted by Currey’s model [1969]; this is shown in Fig. 18.4. The lack of agreement led them to “conclude, therefore that an alternative model is required to account for the dependence of Young’s modulus on orientation” [Bonfield and Grynpas, 1977]. Katz [1980, 1981], applying his hierarchical material-structure composite model, showed that the data in Fig. 18.4 could be explained by considering different amounts of Ap crystallites aligned parallel to the long bone axis; this is shown in Fig. 18.5. This early attempt at hierarchical micromechanical modeling is now being extended with more sophisticated modeling using either finiteelement micromechanical computations [Hogan, 1992] or homogenization theory [Crolet et al., 1993]. Further improvements will come by including more definitive information on the structural organization of collagen and Ap at the molecular-ultrastructural level [Wagner and Weiner, 1992; Weiner and Traub, 1989].

18.6 Viscoelastic Properties As stated earlier, bone (along with all other biologic tissues) is a viscoelastic material. Clearly, for such materials, Hooke’s law for linear elastic materials must be replaced by a constitutive equation which includes the time dependency of the material properties. The behavior of an anisotropic linear viscoelastic material may be described by using the Boltzmann superposition integral as a constitutive equation: © 2000 by CRC Press LLC

FIGURE 18.5 Comparison of predictions of Katz two-level composite model with the experimental data of Bonfield and Grynpas. Each curve represents a different lamellar configuration within a single osteon, with longitudinal fibers; A, 64%; B, 57%; C, 50%; D, 37%; and the rest of the fibers assumed horizontal. (From Katz JL, Mechanical Properties of Bone, AMD, Vol. 45, New York, American Society of Mechanical Engineers, 1981, with permission.)

() ∫

σ ij t =

t

−∞

d τ ( ) dτ( ) dτ

Cijkl t − τ

kl

(18.17)

where σij(t) and εkl(τ) are the time-dependent second rank stress and strain tensors, respectively, and Cijkl(t – τ) is the fourth-rank relaxation modulus tensor. This tensor has 36 independent elements for the lowest symmetry case and 12 nonzero independent elements for an orthotropic solid. Again, as for linear elasticity, a reduced notation is used, i.e., 11 → 1, 22 → 2, 33 → 3, 23 → 4, 31 → 5, and 12 → 6. If we apply Eq. (18.17) to the case of an orthotropic material, e.g., plexiform bone, in uniaxial tension (compression) in the 1 direction [Lakes and Katz, 1974], in this case using the reduced notation, we obtain

for all t, and


t

( ) ()

 ( ) ( )  dτ

( ) ()

( ) ()

 ( ) ( )  = 0

 d τ d2 τ d 3 τ C t − τ 1 + C t − τ + C23 t − τ 21 22 −∞  dτ dτ dτ 

() ∫

σ2 t =

( ) ()

 d τ d2 τ d 3 τ C t − τ 1 + C t − τ + C t − τ 11 12 13 −∞  dτ dτ dτ 

() ∫

σ1 t =

t

(18.18)





(18.19)

( ) ()

( ) ()

 ( ) ( )  dτ = 0

 d τ d2 τ d 3 τ C t − τ 1 + C t − τ + C33 t − τ 31 32 −∞  dτ dτ dτ 

() ∫

σ3 t =

t

(18.20)



for all t. Having the integrands vanish provides an obvious solution to Eqs. (18.19) and (18.20). Solving them simultaneously for

[d( ) ] and [d( ) ] and substituting these values in Eq. (18.17) yields τ 2

dτ

τ 3

dτ

d τ ( ) ∫ E (t − τ) d(τ ) dτ

σ1 t =

t

1

−∞

(18.21)

1

where, if for convenience we adopt the notation Cij Cij (t – τ), then Young’s modulus is given by

( )

E1 t − τ = C11 + C12

[C − (C C C )] + C [C − (C C C )] [(C C C ) − C ] [(C C C ) − C ] 31

21 33

23

21

31 22

32

13

21 33

23

32

22

33

32

(18.22)

23

In this case of uniaxial tension (compression), only nine independent orthotropic tensor components are involved, the three shear components being equal to zero. Still, this time-dependent Young’s modulus is a rather complex function. As in the linear elastic case, the inverse form of the Boltzmann integral can be used; this would constitute the compliance formulation. If we consider the bone being driven by a strain at a frequency ω, with a corresponding sinusoidal stress lagging by an angle δ, then the complex Young’s modulus E*(ω) may be expressed as

() ()

()

E * ω = E ′ ω + iE ′′ ω

(18.23)

where E′(ω), which represents the stress-strain ratio in phase with the strain, is known as the storage modulus, and E″(ω), which represents the stress-strain ratio 90 degrees out of phase with the strain, is known as the loss modulus. The ratio of the loss modulus to the storage modulus is then equal to tan δ. Usually, data are presented by a graph of the storage modulus along with a graph of tan δ, both against frequency. For a more complete development of the values of E′(ω) and E″(ω), as well as for the derivation of other viscoelastic technical moduli, see Lakes and Katz [1974]; for a similar development of the shear storage and loss moduli, see Cowin [1989]. Thus, for a more complete understanding of bone’s response to applied loads, it is important to know its rheologic properties. There have been a number of early studies of the viscoelastic properties of various long bones [Sedlin, 1965; Smith and Keiper, 1965; Laird and Kingsbury, 1973; Lugassy, 1968; Black and Korostoff, 1973]. However, none of these was performed over a wide enough range of frequency (or time) to completely define the viscoelastic properties measured, e.g., creep or stress relaxation. Thus it is not possible to mathematically transform one property into any other to compare results of three different experiments on different bones [Lakes and Katz, 1974]. In the first experiments over an extended frequency range, the biaxial viscoelastic as well as uniaxial viscoelastic properties of wet cortical human and bovine femoral bone were measured using both dynamic and stress relaxation techniques over eight decades of frequency (time) [Lakes et al., 1979]. The results of these experiments showed that bone was both nonlinear and thermorheologically complex, i.e., timetemperature superposition could not be used to extend the range of viscoelastic measurements. A nonlinear constitutive equation was developed based on these measurements [Lakes and Katz, 1979a].


FIGURE 18.6 Comparison of relaxation spectra for wet human bone, specimens 5 and 6 [Lakes et al., 1979] in simple torsion; T = 37°C. First approximation from relaxation and dynamic data. Human tibial bone, specimen 6. Human tibial bone, specimen 5, Gstd = G(10 s). Gstd(5) = G(10 s). Gstd(5) = 0.590 × 106 lb/in2. Gstd(6) × 0.602 × 106 lb/in2 (Courtesy Journal of Biomechanics, Pergamon Press).

In addition, relaxation spectrums for both human and bovine cortical bone were obtained; Fig. 18.6 shows the former [Lakes and Katz, 1979b]. The contributions of several mechanisms to the loss tangent of cortical bone is shown in Fig. 18.7 [Lakes and Katz, 1979b]. It is interesting to note that almost all the major loss mechanisms occur at frequencies (times) at or close to those in which there are “bumps,” indicating possible strain energy dissipation, on the relaxation spectra shown on Fig. 18.6. An extensive review of the viscoelastic properties of bone can be found in the CRC publication Natural and Living Biomaterials [Lakes and Katz, 1984]. Following on Katz’s [1976, 1980] adaptation of the Hashin-Rosen hollow fiber composite model [1964], Gottesman and Hashin [1979] presented a viscoelastic calculation using the same major assumptions.

18.7 Related Research As stated earlier, this chapter has concentrated on the elastic and viscoelastic properties of compact cortical bone and the elastic properties of trabecular bone. At present there is considerable research activity on the fracture properties of the bone. Professor William Bonfield and his associates at Queen Mary and Westfield College, University of London and Professor Dwight Davy and his colleagues at Case Western Reserve University are among those who publish regularly in this area. Review of the literature is necessary in order to become acquainted with the state of bone fracture mechanics. An excellent introductory monograph which provides a fascinating insight into the structure-property relationships in bones including aspects of the two areas discussed immediately above is Professor John Currey’s The Mechanical Adaptations of Bones, published in 1984 by Princeton University Press.


FIGURE 18.7 Contributions of several relaxation mechanisms to the loss tangent of cortical bone. A: Homogeneous thermoelastic effect. B: Inhomogeneous thermoelastic effect. C: Fluid flow effect. D: Piezoelectric effect [Lakes and Katz, 1984]. (Courtesy CRC Press)

Defining Terms Apatite: Calcium phosphate compound, stoichiometric chemical formula Ca5(PO4)3 ·X, where X is OH– (hydroxyapatite), F– (fluorapatite), Cl– (chlorapatite), etc. There are two molecules in the basic crystal unit cell. Cancellous bone: Also known as porous, spongy, trabecular bone. Found in the regions of the articulating ends of tubular bones, in vertebrae, ribs, etc. Cortical bone: The dense compact bone found throughout the shafts of long bones such as the femur, tibia, etc. also found in the outer portions of other bones in the body. Haversian bone: Also called osteonic. The form of bone found in adult humans and mature mammals, consisting mainly of concentric lamellar structures, surrounding a central canal called the haversian canal, plus lamellar remnants of older haversian systems (osteons) called interstitial lamellae. Interstitial lamellae: See Haversian bone above. Orthotropic: The symmetrical arrangement of structure in which there are three distinct orthogonal axes of symmetry. In crystals this symmetry is called orthothombic. Osteons: See Haversian bone above. Plexiform: Also called laminar. The form of parallel lamellar bone found in younger, immature nonhuman mammals. Transverse isotropy: The symmetry arrangement of structure in which there is a unique axis perpendicular to a plane in which the other two axes are equivalent. The long bone direction is chosen as the unique axis. In crystals this symmetry is called hexagonal.


References Ashman RB, Rho JY. 1988. Elastic modulus of trabecular bone material. J Biomech 21:177. Black J, Korostoff E. 1973. Dynamic mechanical properties of viable human cortical bone. J Biomech 6:435. Bonfield W, Grynpas MD. 1977. Anisotropy of Young’s modulus of bone. Nature, London 270:453. Bumrerraj S. 1999. Scanning acoustic microscopy studies of human cortical and trabecular bone. M.S. (BME) Project (Katz JL, advisor) Case Western Reserve University. Choi K, Goldstein SA. 1992. A comparison of the fatigue behavior of human trabecular and cortical bone tissue. J Biomech 25:1371. Chung DH, Buessem WR. 1968. In Anisotropy in Single-crystal Refractory Compounds, vol 2, p 217, FW Vahldiek and SA Mersol (eds) New York, Plenum Press. Cowin SC. 1989. Bone Mechanics. Boca Raton, Fla, CRC Press. Cox HL. 1952. The elasticity and strength of paper and other fibrous materials. British Appl Phys 3:72. Crolet JM, Aoubiza B, Meunier A. 1993. Compact bone: Numerical simulation of mechanical characteristics. J Biomech 26:(6)677. Currey JD. 1964. Three analogies to explain the mechanical properties of bone. Biorheology (2):1. Currey JD. 1969. The relationship between the stiffness and the mineral content of bone. J Biomech (2):477. Currey J. 1984. The Mechanical Adaptations of Bones. New Jersey, Princeton University Press. Gottesman T, Hashin Z. 1979. Analysis of viscoelastic behavior of bones on the basis of microstructure. J Biomech 13:89. Hashin Z, Rosen BW. 1964. The elastic moduli of fiber reinforced materials. J Appl Mech (31):223. Hashin Z, Shtrikman S. 1963. A variational approach to the theory of elastic behavior of multiphase materials. J Mech Phys Solids (11):127. Hastings GW, Ducheyne P (eds). 1984. Natural and Living Biomaterials, Boca Raton, Fla, CRC Press. Herring GM. 1977. Methods for the study of the glycoproteins and proteoglycans of bone using bacterial collagenase. Determination of bone sialoprotein and chondroitin sulphate. Calcif Tiss Res (24):29. Hogan HA. 1992. Micromechanics modeling of haversian cortical bone properties. J Biomech 25(5):549. Katz JL. 1971a. Hard tissue as a composite material: I. Bounds on the elastic behavior. J Biomech 4:455. Katz JL. 1971b. Elastic properties of calcified tissues. Isr J Med Sci 7:439. Katz JL. 1976. Hierarchical modeling of compact haversian bone as a fiber reinforced material. In Mates, R.E. and Smith, CR (eds), Advances in Bioengineering, pp. 17–18. New York, American Society of Mechanical Engineers. Katz JL. 1980. Anisotropy of Young’s modulus of bone. Nature 283:106. Katz JL. 1981. Composite material models for cortical bone. In Cowin SC (ed), Mechanical Properties of Bone vol 45, pp 171–184. New York, American Society of Mechanical Engineers. Katz JL, Meunier A. 1987. The elastic anisotropy of bone. J Biomech 20:1063. Katz JL, Meunier a. 1990. A generalized method for characterizing elastic anisotropy in solid living tissues. J Mat Sci Mater Med 1:1. Katz JL, Ukraincik K. 1971. On the anisotropic elastic properties of hydroxyapatite. J Biomech 4:221. Katz JL, Ukraincik K. 1972. A fiber-reinforced model for compact haversian bone. Program and Abstracts of the 16th Annual Meeting of the Biophysical Society, 28a FPM-C15, Toronto. Keaveny TM, Hayes WC. 1993. A 20-year perspective on the mechanical properties of trabecular bone. J Biomech Eng 115:535. Knets IV. 1978. Mekhanika Polimerov 13:434. Laird GW, Kingsbury HB. 1973. Complex viscoelastic moduli of bovine bone. J Biomech 6:59. Lakes RS. 1993. Materials with structural hierarchy. Nature 361:511. Lakes RS, Katz JL. 1974. Interrelationships among the viscoelastic function for anisotropic solids: Application to calcified tissues and related systems. J Biomech 7:259. Lakes RS, Katz JL. 1979a. Viscoelastic properties and behavior of cortical bone. Part II. Relaxation mechanisms. J Biomech 12:679. © 2000 by CRC Press LLC

Lakes RS, Katz JL. 1979b. Viscoelastic properties of wet cortical bone: III. A nonlinear constitutive equation. J Biomech 12:689. Lakes RS, Katz JL. 1984. Viscoelastic properties of bone. In GW Hastings and P Ducheyne (eds), Natural and Living Tissues, pp 1–87. Boca Raton, Fla, CRC Press. Lakes RS, Katz JL, Sternstein SS. 1979. Viscoelastic properties of wet cortical bone: I. Torsional and biaxial studies. J Biomech 12:657. Lang SB. 1969. Elastic coefficients of animal bone. Science 165:287. Lipson SF, Katz JL. 1984. The relationship between elastic properties and microstructure of bovine cortical bone. J Biomech 4:231. Lugassy AA. 1968. Mechanical and Viscoelastic Properties of Bone and Dentin in Compression, thesis, Metallurgy and Materials Science, University of Pennsylvania. Maharidge R. 1984. Ultrasonic properties and microstructure of bovine bone and Haversian bovine bone modeling. Thesis. Rensselaer Polytechnic Institute, Troy, NY. Park JB. 1979. Biomaterials: An Introduction. New York, Plenum. Pellegrino ED, Biltz RM. 1965. The composition of human bone in uremia. Medicine 44:397. Piekarski K. 1973. Analysis of bone as a composite material. Int J Eng Sci 10:557. Reuss A. 1929. Berechnung der fliessgrenze von mischkristallen auf grund der plastizitatsbedingung fur einkristalle, A. Zeitschrift fur Angewandte Mathematik und Mechanik 9:49–58. Rho JY, Ashman RB, Turner CH. 1993. Young’s modulus of trabecular and cortical bone material; ultrasonic and microtensile measurements. J Biomech 26:111. Rho JY, Roy ME, Tsui TY, Pharr GM. 1999. Elastic properties of microstructural components of human bone tissue as measured by indentation. J Biomed Mat Res 45:48. Ryan SD, Williams JL. 1989. Tensile testing of rodlike trabeculae excised from bovine femoral bone. J Biomech 22:351. Sedlin E. 1965. A rheological model for cortical bone. Acta Orthop Scand 36 (suppl 83). Smith R, Keiper D. 1965. Dynamic measurement of viscoelastic properties of bone. Am J Med Elec 4:156. Townsend PR, Rose RM, Radin EL. 1975. Buckling studies of single human trabeculae. J Biomech 8:199. Turner CH, Rho JY, Takano Y, Tsui TY, Pharr GM. 1999. The elastic properties of trabecular and cortical bone tissues are simular: results from two microscopic measurement techniques. J Biomech 32:437. Van Buskirk WC, Ashman RB. 1981. The elastic moduli of bone. In SC Cowin (ed), Mechanical Properties of Bone AMD vol 45, pp 131–143. New York, American Society of Mechanical Engineers. Vejlens L. 1971. Glycosaminoglycans of human bone tissue: I. Pattern of compact bone in relation to age. Calcif Tiss Res 7:175. Voigt W. 1966. Lehrbuch der Kristallphysik Teubner, Leipzig 1910; reprinted (1928) with an additional appendix. Leipzig, Teubner, New York, Johnson Reprint. Wagner HD, Weiner S. 1992. On the relationship between the microstructure of bone and its mechanical stiffness. J Biomech 25:1311. Wainwright SA, Briggs WD, Currey JD, Gosline JM. 1982. Mechanical Design in Organisms. Princeton, N.J., Princeton University Press. Weiner S, Traub W. 1989. Crystal size and organization in bone. Conn Tissue Res 21:259. Yoon HS, Katz JL. 1976a. Ultrasonic wave propagation in human cortical bone: I. Theoretical considerations of hexagonal symmetry. J Biomech 9:407. Yoon HS, Katz JL. 1976b. Ultrasonic wave propagation in human cortical bone: II. Measurements of elastic properties and microhardness. J Biomech 9:459.

Further Information Several societies both in the United States and abroad hold annual meetings during which many presentations, both oral and poster, deal with hard tissue biomechanics. In the United States these societies include the Orthopaedic Research Society, the American Society of Mechanical Engineers, the Biomaterials Society, the American Society of Biomechanics, the Biomedical Engineering Society, and the Society © 2000 by CRC Press LLC

for Bone and Mineral Research. In Europe there are alternate year meetings of the European Society of Biomechanics and the European Society of Biomaterials. Every four years there is a World Congress of Biomechanics; every three years there is a World Congress of Biomaterials. All of these meetings result in documented proceedings; some with extended papers in book form. The two principal journals in which bone mechanics papers appear frequently are the Journal of Biomechanics published by Elsevier and the Journal of Biomechanical Engineering published by the American Society of Mechanical Engineers. Other society journals which periodically publish papers in the field are the Journal of Orthopaedic Research published for the Orthopaedic Research Society, the Annals of Biomedical Engineering published for the Biomedical Engineering Society, and the Journal of Bone and Joint Surgery (both American and English issues) for the American Academy of Orthopaedic Surgeons and the British Organization, respectively. Additional papers in the field may be found in the journal Bone and Calcified Tissue International. The 1984 CRC volume, Natural and Living Biomaterials (Hastings GW and Ducheyne P, Eds.) provides a good historical introduction to the field. A more advanced book is Bone Mechanics (Cowin SC, 1989); a second edition is due out in late 1999 or early 2000. Many of the biomaterials journals and society meetings will have occasional papers dealing with hard tissue mechanics, especially those dealing with implant-bone interactions.


Appendix The Voigt and Reuss moduli for both transverse isotropic and orthotropic symmetry are given below:

Voigt Transverse Isotropic KV = GV =

(

) (

2 C11 + C12 + 4 C13 + C33

(

)

9

)

(

C11 + C12 − 4C13 + 2C33 + 12 C44 + C66

(18.A1)

)

30

Reuss Transverse Isotropic KR =

(

(C

11

GR =

)

C33 C11 + C12 − 2C132 + C12 − 4C13 + 2C33

)

[ ( ) ] ) − 2C ] (C + C ) + [C C (2C

(18.A2)

5 C33 C11 + C12 − 2C132 C44C66

{[ (

2 C33 C11 + C12

2 13

44

66

44 66

11

] }

)

+ C12 + 4C13 + C33 3

Voigt Orthotropic

KV = GV =

(

C11 + C22 + C33 + 2 C12 + C13 + C23 9

[

(

) ) (

C11 + C22 + C33 + 3 C44 + C55 + C66 − C12 + C13 + C23

)]

(18.A3)

15

Reuss Orthotropic KR =

(

∆ − 2 C11C23 + C22C13 + C33C12 C11C22 + C22C33 + C33C11

(

) (

2 + 2 C12C23 + C23C13 + C13C12 − C122 + C132 + C23

{(

)

)

G R = 15  4 C11C22 + C22C33 + C33C11 + C11C23 + C22C13 + C33C12

[ (

)

(

)

(

− C12 C12 + C23 + C23 C23 + C13 + C13 C13 + C12

(

+ 3 1 C44 + 1 C55 + 1 C66 where ∆ is given in Eq. (18.15).


))

)]}

∆

)

(18.A4)

Canfield, T. R., Dobrin, P. B. “ Mechanics of Blood Vessels.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

19 Mechanics of Blood Vessels 19.1

Assumptions Homogeneity of the Vessel Wall • Incompressibility of the Vessel Wall • Inelasticity of the Vessel Wall • Residual Stress and Strain

Thomas R. Canfield Argonne National Laboratory

Philip B. Dobrin Hines VA Hospital and Loyola University Medical Center

19.2 19.3 19.4 19.5 19.6

Vascular Anatomy Axisymmetric Deformation Experimental Measurements Equilibrium Strain Energy Density Functions Isotropic Blood Vessels • Anisotropic Blood Vessels

19.1 Assumptions This chapter is concerned with the mechanical behavior of blood vessels under static loading conditions and the methods required to analyze this behavior. The assumptions underlying this discussion are for ideal blood vessels that are at least regionally homogeneous, incompressible, elastic, and cylindrically orthotropic. Although physiologic systems are nonideal, much understanding of vascular mechanics has been gained through the use of methods based upon these ideal assumptions.

Homogeneity of the Vessel Wall On visual inspection, blood vessels appear to be fairly homogeneous and distinct from surrounding connective tissue. The inhomogeneity of the vascular wall is realized when one examines the tissue under a low-power microscope, where one can easily identify two distinct structures: the media and adventitia. For this reason the assumption of vessel wall homogeneity is applied cautiously. Such an assumption may be valid only within distinct macroscopic structures. However, few investigators have incorporated macroscopic inhomogeneity into studies of vascular mechanics [17].

Incompressibility of the Vessel Wall Experimental measurement of wall compressibility of 0.06% at 270 cm of H2O indicates that the vessel can be considered incompressible when subjected to physiologic pressure and load [2]. In terms of the mechanical behavior of blood vessels, this is small relative to the large magnitude of the distortional strains that occur when blood vessels are deformed under the same conditions. Therefore, vascular

Work sponsored by the US Department of Energy order Contract W-31-109-Eng-38.


compressibility may be important to understanding other physiologic processes related to blood vessels, such as the transport of interstitial fluid.

Inelasticity of the Vessel Wall That blood vessel walls exhibit inelastic behavior such as length-tension and pressure-diameter hysteresis, stress relaxation, and creep has been reported extensively [1, 10]. However, blood vessels are able to maintain stability and contain the pressure and flow of blood under a variety of physiologic conditions. These conditions are dynamic but slowly varying with a large static component.

Residual Stress and Strain Blood vessels are known to retract both longitudinally and circumferentially are excision. This retraction is caused by the relief of distending forces resulting from internal pressure and longitudinal tractions. The magnitude of retraction is influenced by several factors. Among these factors are growth, aging, and hypertension. Circumferential retraction of medium-caliber blood vessels, such as the carotid, iliac, and bracheal arteries, can exceed 70% following reduction of internal blood pressure to zero. In the case of the carotid artery, the amount of longitudinal retraction tends to increase during growth and to decrease in subsequent aging [5]. It would seem reasonable to assume that blood vessels are in a nearly stress-free state when they are fully retracted and free of external loads. This configuration also seems to be a reasonable choice for the reference configuration. However, this ignores residual stress and strain effects that have been the subject of current research [4, 11–14, 16]. Blood vessels are formed in a dynamic environment which gives rise to imbalances between the forces that tend to extend the diameter and length and the internal forces that tend to resist the extension. This imbalance is thought to stimulate the growth of elastin and collagen and to effectively reduce the stresses in the underlying tissue. Under these conditions it is not surprising that a residual stress state exists when the vessel is fully retracted and free of external tractions. This process has been called remodeling [11]. Striking evidence of this remodeling is found when a cylindrical slice of the fully retracted blood vessel is cut longitudinally through the wall. The cylinder springs open, releasing bending stresses kept in balance by the cylindrical geometry [16].

19.2 Vascular Anatomy A blood vessel can be divided anatomically into three distinct cylindrical sections when viewed under the optical microscope. Starting at the inside of the vessel, they are the intima, the media, and the adventitia. These structures have distinct functions in terms of the blood vessel physiology and mechanical properties. The intima consists of a thin monolayer of endothelial cells that line the inner surface of the blood vessel. The endothelial cells have little influence on blood vessel mechanics but do play an important role in hemodynamics and transport phenomena. Because of their anatomical location, these cells are subjected to large variations in stress and strain as a result of pulsatile changes in blood pressure and flow. The media represents the major portion of the vessel wall and provides most of the mechanical strength necessary to sustain structural integrity. The media is organized into alternating layers of interconnected smooth muscle cells and elastic lamellae. There is evidence of collagen throughout the media. These small collagen fibers are found within the bands of smooth muscle and may participate in the transfer of forces between the smooth muscle cells and the elastic lamellae. The elastic lamellae are composed principally of the fiberous protein elastin. The number of elastic lamellae depends upon the wall thickness and the anatomical location [18]. In the case of the canine carotid, the elastic lamellae account for a major component of the static structural response of the blood vessel [6]. This response is modulated by the smooth-muscle cells, which have the ability to actively change the mechanical characteristics of the wall [7]. The adventitia consists of loose, more disorganized fiberous connective tissue, which may have less influence on mechanics. © 2000 by CRC Press LLC

FIGURE 19.1 Cylindrical geometry of a blood vessel: top: stress-free reference configuration; middle: fully retracted vessel free of external traction; bottom: vessel in situ under longitudinal tether and internal pressurization.

19.3 Axisymmetric Deformation In the following discussion we will concern ourselves with deformation of cylindrical tubes, see Fig. 19.1. Blood vessels tend to be nearly cylindrical in situ and tend to remain cylindrical when a cylindrical section is excised and studied in vitro. Only when the vessel is dissected further does the geometry begin to deviate from cylindrical. For this deformation there is a unique coordinate mapping

(R, Θ, Z ) → (r, θ, z )

(19.1)

where the undeformed coordinates are given by (R, Θ, Z) and the deformed coordinates are given by (r, θ, z). The deformation is given by a set of restricted functions

r=r R

()

(19.2)

θ = βΘ

(19.3)

z = µZ + C1

(19.4)

where the constants µ and β have been introduced to account for a uniform longitudinal strain and a symmetric residual strain that are both independent of the coordinate Θ. © 2000 by CRC Press LLC

If β = 1, there is no residual strain. If β ≠ 1, residual stresses and strains are present. If β > 1, a longitudinal cut through the wall will cause the blood vessel to open up, and the new cross-section will form a c-shaped section of an annulus with larger internal and external radii. If β < 1, the cylindrical shape is unstable, but a thin section will tend to overlap itself. In Choung and Fung’s formulation, β = π/Θo, where the angle Θo is half the angle spanned by the open annular section [4]. For cylindrical blood vessels there are two assumed constraints. The first assumption is that the longitudinal strain is uniform through the wall and therefore

λ z = µ = a constant

(19.5)

for any cylindrical configuration. Given this, the principal stretch ratios are computed from the above function as

λr =

dr dR

(19.6)

r R

(19.7)

λθ = β

λz = µ

(19.8)

The second assumption is wall incompressibility, which can be expressed by

λ r λ θλ z ≡ 1

(19.9)

or

βµ

r dr =1 R dR

(19.10)

1 RdR βµ

(19.11)

1 2 R + c2 βµ

(19.12)

1 2 Re βµ

(19.13)

and therefore

rdr = Integration of this expression yields the solution

r2 = where

c2 = re2 −

As a result, the principal stretch ratios can be expressed in terms of R as follows:


λr =

(

R

βµ R + βµc2

λθ =

2

c 1 + 2 βµ R2

)

(19.14)

(19.15)

19.4 Experimental Measurements The basic experimental setup required to measure the mechanical properties of blood vessels in vitro is described in [7]. It consists of a temperature-regulated bath of physiologic saline solution to maintain immersed cylindrical blood vessel segments, devices to measure diameter, an apparatus to hold the vessel at a constant longitudinal extension and to measure longitudinal distending force, and a system to deliver and control the internal pressure of the vessel with 100% oxygen. Typical data obtained from this type of experiment are shown in Figs. 19.2 and 19.3.

19.5 Equilibrium When blood vessels are excised, they retract both longitudinally and circumferentially. Restoration to natural dimensions requires the application of internal pressure, pi , and a longitudinal tether force, FT . The internal pressure and longitudinal tether are balanced by the development of forces within the vessel wall. The internal pressure is balanced in the circumferential direction by a wall tension, T. The longitudinal tether force and pressure are balanced by the retractive force of the wall, FR

FIGURE 19.2

T = pi ri

(19.16)

FR = FT + pi π ri2

(19.17)

Pressure-radius curves for the canine carotid artery at various degrees of longitudinal extension.


FIGURE 19.3

Longitudinal distending force as a function of radius at various degrees of longitudinal extension.

The first equation is the familiar law of Laplace for a cylindrical tube with internal radius ri . It indicates that the force due to internal pressure, pi , must be balanced by a tensile force (per unit length), T, within the wall. This tension is the integral of the circumferentially directed force intensity (or stress, σθ) across the wall:

T=

∫

re

ri

σ θdr = σθ h

(19.18)

–

where σθ is the mean value of the circumferential stress and h is the wall thickness. Similarly, the longitudinal tether force, FT , and extending force due to internal pressure are balanced by a retractive internal force, FR , due to axial stress, σz , in the blood vessel wall:

FR = 2 π

∫

re

ri

(

σ z rdr = σ z πh re + ri

)

(19.19)

–

where σz is the mean value of this longitudinal stress. The mean stresses are calculated from the above equation as

σθ = pi

σz =

(

FT

πh re + ri

ri h

)

(19.20)

+

pi ri 2 h

(19.21)

The mean stresses are a fairly good approximation for thin-walled tubes where the variations through the wall are small. However, the range of applicability of the thin-wall assumption depends upon the material properties and geometry. In a linear elastic material, the variation in σθ is less than 5% for r/h > 20. When the material is nonlinear or the deformation is large, the variations in stress can be more severe (see Fig. 19.10). © 2000 by CRC Press LLC

The stress distribution is determined by solving the equilibrium equation,

( )

σ 1 d rσ r − θ = 0 r dr r

(19.22)

This equation governs how the two stresses are related and must change in the cylindrical geometry. For uniform extension and internal pressurization, the stresses must be functions of a single radial coordinate, r, subject to the two boundary conditions for the radial stress:

( )

σ r ri ,µ = − pi

(19.23)

( )

(19.24)

σ r re ,µ = 0

19.6 Strain Energy Density Functions Blood vessels are able to maintain their structural stability and contain steady oscillating internal pressures. This property suggests a strong elastic component, which has been called the pseudoelasticity [10]. This elastic response can be characterized by a single potential function called the strain energy density. It is a scalar function of the strains that determines the amount of stored elastic energy per unit volume. In the case of a cylindrically orthotropic tube of incompressible material, the strain energy density can be written in the following functional form:

(

)

W = W * λ r , λ θ , λ z + λ r λ θ λ z p,

(19.25)

where p is a scalar function of position, R. The stresses are computed from the strain energy by the following:

σi = λ i

∂W * +p ∂λ i

(19.26)

We make the following transformation [3]

λ=

βr

(

βµ r 2 − c2

(19.27)

)

which upon differentiation gives

r

(

dλ = β −1 βλ − µλ3 dr

)

(19.28)

After these expressions and the stresses in terms of the strain energy density function are introduced into the equilibrium equation, we obtain an ordinary differential equation for p * * * dp β W,λθ − W,λ r dW,λ r = − dλ dλ βλ = µλ3


(19.29)

subject to the boundary conditions

( )

(19.30)

( )

(19.31)

p Ri = pi p Re = 0

Isotropic Blood Vessels A blood vessel generally exhibits anisotropic behavior when subjected to large variations in internal pressure and distending force. When the degree of anisotropy is small, the blood vessel may be treated as isotropic. For isotropic materials it is convenient to introduce the strain invariants:

I1 = λ2r + λ2θ + λ2z

(19.32)

I 2 = λ2r λ2θ + λ2θ λ2z + λ2z λ2r

(19.33)

I 3 = λ2r λ2θ λ2z

(19.34)

These are measures of strain that are independent of the choice of coordinates. If the material is incompressible

I3 = j2 ≡ 1

(19.35)

and the strain energy density is a function of the first two invariants, then

( )

W = W I1, I 2 .

(19.36)

The least complex form for an incompressible material is the first-order polynomial, which was first proposed by Mooney to characterize rubber:

W* =

[( ) ( )]

G I1 − 3 + k I 2 − 3 2

(19.37)

It involves only two elastic constants. A special case, where k = 0, is the neo-Hookean material, which can be derived from thermodynamics principles for a simple solid. Exact solutions can be obtained for the cylindrical deformation of a thick-walled tube. In the case where there is no residual strain, we have the following:

(


)

 log λ 1  p = −G 1 + kµ 2  + 2 2  + c0 2µ λ   µ

(19.38)

 1  1 1  σr = G  2 2 + k  2 + 2   + p  µ λ    λ µ

(19.39)

   1 σ θ = G λ2 + k  2 + λ2µ 2   + p   µ 

(19.40)

FIGURE 19.4

Pressure-radius curves for a Mooney-Rivlin tube with the approximate dimensions of the carotid.

FIGURE 19.5

Longitudinal distending force as a function of radius for the Mooney-Rivlin tube.

  1  σ z = G µ 2 + k  λ2µ 2 + 2   + p λ    

(19.41)

However, these equations predict stress softening for a vessel subjected to internal pressurization at fixed lengths, rather than the stress stiffening observed in experimental studies on arteries and veins (see Figs. 19.4 and 19.5). An alternative isotropic strain energy density function which can predict the appropriate type of stress stiffening for blood vessels is an exponential where the arguments is a polynomial of the strain invariants. The first-order form is given by

W* =


[ ( ) ( )]

G0 exp k1 I1 − 3 + k2 I 2 − 3 2k1

(19.42)

FIGURE 19.6 Pressure-radius curves for tube with the approximate dimensions of the carotid calculated using an isotropic exponential strain energy density function.

FIGURE 19.7

Longitudinal distending force as a function of radius for the isotropic tube.

This requires the determination of only two independent elastic constants. The third, G0, is introduced to facilitate scaling of the argument of the exponent (see Figs. 19.6 and 19.7). This exponential form is attractive for several reasons. It is a natural extension of the observation that biologic tissue stiffness is proportional to the load in simple elongation. This stress stiffening has been attributed to a statistical recruitment and alignment of tangled and disorganized long chains of proteins. The exponential forms resemble statistical distributions derived from these same arguments.

Anisotropic Blood Vessels Studies of the orthotropic behavior of blood vessels may employ polynomial or exponential strain energy density functions that include all strain terms or extension ratios. In particular, the strain energy density function can be of the form © 2000 by CRC Press LLC

FIGURE 19.8 Pressure-radius curves for a fully orthotropic vessel calculated with an exponential strain energy density function.

FIGURE 19.9

Longitudinal distending force as a function of radius for the orthotropic vessel.

(

W * = qn λ r , λ θ , λ z

)

(19.43)

or

W* = e

(

qn λ r , λ θ , λ z

)

(19.44)

where qn is a polynomial of order n. Since the material is incompressible, the explicit dependence upon λr can be eliminated either by substituting λr = λ–1θ λ–1z or by assuming that the wall is thin and hence that the contribution of these terms is small. Figures 19.8 and 19.9 illustrate how well the experimental data can be fitted to an exponential strain density function whose argument is a polynomial of order n = 3. © 2000 by CRC Press LLC

Care must be taken to formulate expressions that will lead to stresses that behave properly. For this reason it is convenient to formulate the strain energy density in terms of the lagrangian strains

(

)

ei = 1 2 λ2i − 1

(19.45)

and in this case we can consider polynomials of the lagrangian strains, qn(er , eθ, ez). Vaishnav et al. [15] proposed using a polynomial of the form n

W* =

i

∑ ∑a i=2

i− j j i j−i θ z

e e

(19.46)

j =0

to approximate the behavior of the canine aorta. They found better correlation with order-three polynomials over order-two, but order-four polynomials did not warrant the addition work. Later, Fung et al. [10] found very good correlation with an expression of the form

W=

[(

) (

)

(

C exp a1 eθ2 − ez*2 + a2 ez2 − ez*2 + 2a 4 eθez − eθ*ez* 2

)]

(19.47)

for the canine carotid artery, where e*θ and e*z are the strains in a reference configuration at in situ length and pressure. Why should this work? One answer appears to be related to residual stresses and strains. When residual stresses are ignored, large-deformation analysis of thick-walled blood vessels predicts steep distributions in σθ and σz through the vessel wall, with the highest stresses at the interior. This prediction is considered significant because high tensions in the inner wall could inhibit vascularization and oxygen transport to vascular tissue. When residual stresses are considered, the stress distributions flatten considerably and become almost uniform at in situ length and pressure. Fig. 19.10 shows the radial stress distributions computed for a vessel with β = 1 and β = 1.11. Takamizawa and Hayashi have even considered the case where the strain

FIGURE 19.10


Stress distributions through the wall at various pressures for the orthotropic vessel.

distribution is uniform in situ [13]. The physiologic implications are that vascular tissue is in a constant state of flux. New tissue is synthesized in a state of stress that allows it to redistribute the internal loads more uniformly. There probably is no stress-free reference state [8, 11, 12]. Continuous dissection of the tissue into smaller and smaller pieces would continue to relieve residual stresses and strains [14].

References 1. Bergel DH. 1961. The static elastic properties of the arterial wall. J Physiol 156:445. 2. Carew TE, Vaishnav RN, Patel DJ. 1968. Compressibility of the arterial walls. Circ Res 23:61. 3. Chu BM, Oka S. 1973. Influence of longitudinal tethering on the tension in thick-walled blood vessels in equilibrium. Biorheology 10:517. 4. Choung CJ, Fung YC. 1986. On residual stresses in arteries. J Biomed Eng 108:189. 5. Dobrin PB. 1978. Mechanical properties of arteries. Physiol Rev 58:397. 6. Dobrin PB, Canfield TR. 1984. Elastase, collagenase, and the biaxial elastic properties of dog carotid artery. Am J Physiol 2547:H124. 7. Dobrin PB, Rovick AA. 1969. Influence of vascular smooth muscle on contractile mechanics and elasticity of arteries. Am J Physiol 217:1644. 8. Dobrin PD, Canfield T, Sinha S. 1975. Development of longitudinal retraction of carotid arteries in neonatal dogs. Experientia 31:1295. 9. Doyle JM, Dobrin PB. 1971. Finite deformation of the relaxed and contracted dog carotid artery. Microvasc Res 3:400. 10. Fung YC, Fronek K, Patitucci P. 1979. Pseudoelasticity of arteries and the choice of its mathematical expression. Am J Physiol 237:H620. 11. Fung YC, Liu SQ, Zhou JB. 1993. Remodeling of the constitutive equation while a blood vessel remodels itself under strain. J Biomech Eng 115:453. 12. Rachev A, Greenwald S, Kane T, Moore J, Meister J-J. 1994. Effects of age-related changes in the residual strains on the stress distribution in the arterial wall. In J Vossoughi (ed), Proceedings of the Thirteenth Society of Biomedical Engineering Recent Developments, pp 409–412, Washington, DC, University of District of Columbia. 13. Takamizawa K, Hayashi K. 1987. Strain energy density function and the uniform strain hypothesis for arterial mechanics. J Biomech 20:7. 14. Vassoughi J. 1992. Longitudinal residual strain in arteries. Proc of the 11th South Biomed Engrg Conf, Memphis, TN. 15. Vaishnav RN, Young JT, Janicki JS, Patel DJ. 1972. Nonlinear anisotropic elastic properties of the canine aorta. Biophys J 12:1008. 16. Vaishnav RN, Vassoughi J. 1983. Estimation of residual stresses in aortic segments. In CW Hall (ed), Biomedical Engineering II, Recent Developments, pp 330–333, New York, Pergamon Press. 17. Von Maltzahn W-W, Desdo D, Wiemier W. 1981. Elastic properties of arteries: a nonlinear twolayer cylindrical model. J Biomech 4:389. 18. Wolinsky H, Glagov S. 1969. Comparison of abdominal and thoracic aortic media structure in mammals. Circ Res 25:677.


Kaufman, K. R., Kai-Nan An. “Joint-Articulating Surface Motion.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

20 Joint-Articulating Surface Motion 20.1

Ankle Geometry of the Articulating Surfaces • Joint Contact • Axes of Rotation

20.2

Knee Geometry of the Articulating Surfaces • Joint Contact • Axes of Rotation

20.3

Hip Geometry of the Articulating Surfaces • Joint Contact • Axes of Rotation

20.4

Shoulder Geometry of the Articulating Surfaces • Joint Contact • Axes of Rotation

20.5

Elbow Geometry of the Articulating Surfaces • Joint Contact • Axes of Rotation

Kenton R. Kaufman Biomechanics Laboratory, Mayo Clinic

20.6 20.7

Hand Geometry of the Articulating Surfaces • Joint Contact • Axes of Rotation

Kai-Nan An Biomechanics Laboratory, Mayo Clinic

Wrist Geometry of the Articulating Surfaces • Joint Contact • Axes of Rotation

20.8

Summary

Knowledge of joint-articulating surface motion is essential for design of prosthetic devices to restore function; assessment of joint wear, stability, and degeneration; and determination of proper diagnosis and surgical treatment of joint disease. In general, kinematic analysis of human movement can be arranged into two separate categories: (1) gross movement of the limb segments interconnected by joints, or (2) detailed analysis of joint articulating surface motion which is described in this chapter. Gross movement is the relative three-dimensional joint rotation as described by adopting the Eulerian angle system. Movement of this type is described in Chapter 25: Analysis of Gait. In general, the threedimensional unconstrained rotation and translation of an articulating joint can be described utilizing the concept of the screw displacement axis. The most commonly used analytic method for the description of 6-degree-of-freedom displacement of a rigid body is the screw displacement axis [Kinzel et al. 1972; Spoor & Veldpaus, 1980; Woltring et al. 1985]. Various degrees of simplification have been used for kinematic modeling of joints. A hinged joint is the simplest and most common model used to simulate an anatomic joint in planar motion about a single axis embedded in the fixed segment. Experimental methods have been developed for determination


FIGURE 20.1

Three types of articulating surface motion in human joints.

of the instantaneous center of rotation for planar motion. The instantaneous center of rotation is defined as the point of zero velocity. For a true hinged motion, the instantaneous center of rotation will be a fixed point throughout the movement. Otherwise, loci of the instantaneous center of rotation or centrodes will exist. The center of curvature has also been used to define joint anatomy. The center of curvature is defined as the geometric center of coordinates of the articulating surface. For more general planar motion of an articulating surface, the term sliding, rolling, and spinning are commonly used (Fig. 20.1). Sliding (gliding) motion is defined as the pure translation of a moving segment against the surface of a fixed segment. The contact point of the moving segment does not change, while the contact point of the fixed segment has a constantly changing contact point. If the surface of the fixed segment is flat, the instantaneous center of rotation is located at infinity. Otherwise, it is located at the center of curvature of the fixed surface. Spinning motion (rotation) is the exact opposite of sliding motion. In this case, the moving segment rotates, and the contact points on the fixed surface does not change. The instantaneous center of rotation is located at the center of curvature of the spinning body that is undergoing pure rotation. Rolling motion occurs between moving and fixed segments where the contact points in each surface are constantly changing and the arc lengths of contact are equal on each segment. The instantaneous center of rolling motion is located at the contact point. Most planar motion of anatomic joints can be described by using any two of these three basic descriptions. In this chapter, various aspects of joint-articulating motion are covered. Topics include the anatomical characteristics, joint contact, and axes of rotation. Joints of both the upper and lower extremity are discussed.

20.1 Ankle The ankle joint is composed of two joints: the talocrural (ankle) joint and the talocalcaneal (subtalar joint). The talocrural joint is formed by the articulation of the distal tibia and fibula with the trochlea of the talus. The talocalcaneal joint is formed by the articulation of the talus with the calcaneus.

Geometry of the Articulating Surfaces The upper articular surface of the talus is wedge-shaped, its width diminishing from front to back. The talus can be represented by a conical surface. The wedge shape of the talus is about 25% wider in front than behind with an average difference of 2.4 mm ± 1.3 mm and a maximal difference of 6 mm [Inman, 1976].


TABLE 20.1

Talocalcaneal (Ankle) Joint Contact Area

Investigators Ramsey and Hamilton [1976] Kimizuka et al. [1980] Libotte et al. [1982] Paar et al. [1983] Macko et al. [1991] Driscoll et al. [1994] Hartford et al. [1995] Pereira et al. [1996]

Plantarflexion

Neutral

Dorsiflexion

5.01 (30°) 4.15 (10°) 3.81 ± 0.93 (15°) 2.70 ± 0.41 (20°)

4.40 ± 1.21 4.83 5.41 4.15 5.2 ± 0.94 3.27 ± 0.32 3.37 ± 0.52 1.67

3.60 (30°) 3.63 (10°) 5.40 ± 0.74 (10°) 2.84 ± 0.43 (20°)

1.49 (20°)

1.47 (10°)

Note: The contact area is expressed in square centimeters.

Joint Contact The talocrural joint contact area varies with flexion of the ankle (Table 20.1). During plantarflexion, such as would occur during the early stance phase of gait, the contact area is limited and the joint is incongruous. As the position of the joint progresses from neutral to dorsiflexion, as would occur during the midstance of gait, the contact area increases and the joint becomes more stable. The area of the subtalar articulation is smaller than that of the talocrural joint. The contact area of the subtalar joint is 0.89 ± 0.21 cm2 for the posterior facet and 0.28 ± 15 cm2 for the anterior and middle facets [Wang et al., 1994]. The total contact area (1.18 ± 0.35 cm2) is only 12.7% of the whole subtalar articulation area (9.31 ± 0.66 cm2) [Wang et al., 1994]. The contact area/joint area ratio increases with increases in applied load (Fig. 20.2).

Axes of Rotation Joint motion of the talocrural joint has been studied to define the axes of rotation and their location with respect to specific anatomic landmarks (Table 20.2). The axis of motion of the talocrural joint essentially passes through the inferior tibia at the fibular and tibial malleoli (Fig. 20.3). Three types of motion have been used to describe the axes of rotation: fixed, quasi-instantaneous, and instantaneous axes. The motion that occurs in the ankle joints consists of dorsiflexion and plantarflexion. Minimal or no transverse rotation takes place within the talocrural joint. The motion in the talocrural joint is intimately related to the motion in the talocalcaneal joint which is described next. The motion axes of the talocalcaneal joint have been described by several authors (Table 20.3). The axis of motion in the talocalcaneal joint passes from the anterior medial superior aspect of the navicular bone to the posterior lateral inferior aspect of the calcaneus (Fig. 20.4). The motion that occurs in the talocalcaneal joint consists of inversion and eversion.

20.2 Knee The knee is the intermediate joint of the lower limb. It is composed of the distal femur and proximal tibia. It is the largest and most complex joint in the body. The knee joint is composed of the tibiofemoral articulation and the patellofemoral articulation.

Geometry of the Articulating Surfaces The shape of the articular surfaces of the proximal tibia and distal femur must fulfill the requirement that they move in contact with one another. The profile of the femoral condyles varies with the condyle examined (Fig. 20.5 and Table 20.4). The tibial plateau widths are greater than the corresponding widths of the femoral condyles (Fig. 20.6 and Table 20.5). However, the tibial plateau depths are less than those


FIGURE 20.2 Ratio of total contact area to joint area in the (A) anterior/middle facet and (B) posterior facet of the subtalar joint as a function of applied axial load for three different positions of the foot. Source: Wagner UA, Sangeorzan BJ, Harrington RM, Tencer AF. 1992. Contact characteristics of the subtalar joint: load distribution between the anterior and posterior facets. J Orthop Res 10:535. With permission.

of the femoral condyle distances. The medial condyle of the tibia is concave superiorly (the center of curvature lies above the tibial surface) with a radius of curvature of 80 mm [Kapandji, 1987]. The lateral condyle is convex superiorly (the center of curvature lies below the tibial surface) with a radius of curvature of 70 mm [Kapandji, 1987]. The shape of the femoral surfaces is complementary to the shape of the tibial plateaus. The shape of the posterior femoral condyles may be approximated by spherical surfaces (Table 20.4). The geometry of the patellofemoral articular surfaces remains relatively constant as the knee flexes. The knee sulcus angle changes only ±3.4° from 15 to 75° of knee flexion (Fig. 20.7). The mean depth index varies by only ±4% over the same flexion range (Fig. 20.7). Similarly, the medial and lateral patellar facet angles (Fig. 20.8) change by less than a degree throughout the entire knee flexion range (Table 20.6). However, there is a significant difference between the magnitude of the medial and lateral patellar facet angles.

Joint Contact The mechanism of movement between the femur and tibia is a combination of rolling and gliding. Backward movement of the femur on the tibia during flexion has long been observed in the human knee. The magnitude of the rolling and gliding changes through the range of flexion. The tibial-femoral contact © 2000 by CRC Press LLC

TABLE 20.2

Axis of Rotation for the Ankle

Investigators

Axis*

Position

Elftman [1945] Isman and Inman [1969]

Fix. Fix.

Inman and Mann [1979] Allard et al. [1987]

Fix. Fix.

Singh et al. [1992]

Fix.

Sammarco et al. [1973] D’Ambrosia et al. [1976] Parlasca et al. [1979]

Ins. Ins. Ins.

Van Langelaan [1983]

Ins.

Barnett and Napier

Q-I

Hicks [1953]

Q-I

67.6° ± 7.4° with respect to sagittal plane 8 mm anterior, 3 mm inferior to the distal tip of the lateral malleolus; 1 mm posterior, 5 mm inferior to the distal tip of the medial malleolus 79° (68–88°) with respect to the sagittal plane 95.4° ± 6.6° with respect to the frontal plane, 77.7° ± 12.3° with respect to the sagittal plane, and 17.9° ± 4.5° with respect to the transverse plane 3.0 mm anterior, 2.5 mm inferior to distal tip of lateral malleolus, 2.2 mm posterior, 10 mm inferior to distal tip of medial malleolus Inside and outside the body of the talus No consistent pattern 96% within 12 mm of a point 20 mm below the articular surface of the tibia along the long axis. At an approximate right angle to the longitudinal direction of the foot, passing through the corpus tali, with a direction from anterolaterosuperior to posteromedioinferior Dorsiflexion: down and lateral Plantarflexion: down and medial Dorsiflexion: 5 mm inferior to tip of lateral malleolus to 15 mm anterior to tip of medial malleolus Plantarflexion: 5 mm superior to tip of lateral malleolus to 15 mm anterior, 10 mm inferior to tip of medial malleolus

* Fix. = fixed axis of rotation; Ins. = instantaneous axis of rotation; Q-I = quasi-instantaneous axis of rotation

FIGURE 20.3 Variations in angle between middle of tibia and empirical axis of ankle. The histogram reveals a considerable spread of individual values. Source: Inman VT. 1976. The Joints of the Ankle, Baltimore, Williams and Wilkins. With permission.

point has been shown to move posteriorly as the knee is flexed, reflecting the coupling of anterior/posterior motion with flexion/extension (Fig. 20.9). During flexion, the weight-bearing surfaces move backward on the tibial plateaus and become progressively smaller (Table 20.7). It has been shown that in an intact knee at full extension the center of pressure is approximately 25 mm from the anterior edge of the knee joint line [Andriacchi et al., 1986]. This net contact point moves posteriorly with flexion to approximately 38.5 mm from the anterior edge of the knee joint. Similar displacements have been noted in other studies (Table 20.8). © 2000 by CRC Press LLC

TABLE 20.3

Axis of Rotation for the Talocalcaneal (Subtalar) Joint

Investigators

Axis*

Position

Manter [1941]

Fix.

Shephard [1951] Hicks [1953] Root et al. [1966]

Fix. Fix. Fix.

Isman and Inman [1969] Kirby [1947]

Fix. Fix.

Rastegar et al. [1980]

Ins.

Van Langelaan [1983]

Ins.

Engsberg [1987]

Ins.

16° (8–24°) with respect to sagittal plane, and 42° (29–47°) with respect to transverse plane Tuberosity of the calcaneus to the neck of the talus Posterolateral corner of the heel to superomedial aspect of the neck of the talus 17° (8–29°) with respect to sagittal plane, and 41° (22–55°) with respect to transverse plane 23° ± 11° with respect to sagittal plane, and 41° ± 9° with respect to transverse plane Extends from the posterolateral heel, posteriorly, to the first intermetatarsal space, anteriorly Instant centers of rotation pathways in posterolateral quadrant of the distal articulating tibial surface, varying with applied load A bundle of axes that make an acute angle with the longitudinal direction of the foot passing through the tarsal canal having a direction from anteromediosuperior to posterolateroinferior A bundle of axes with a direction from anteromediosuperior to posterolateroinferior

* Fix. = fixed axis of rotation; Ins. = instantaneous axis of rotation

The patellofemoral contact area is smaller than the tibiofemoral contact area (Table 20.9). As the knee joint moves from extension to flexion, a band of contact moves upward over the patellar surface (Fig. 20.10). As knee flexion increases, not only does the contact area move superiorly, but it also becomes larger. At 90° of knee flexion, the contact area has reached the upper level of the patella. As the knee continues to flex, the contact area is divided into separate medial and lateral zones.

Axes of Rotation The tibiofemoral joint is mainly a joint with two degrees of freedom. The first degree of freedom allows movements of flexion and extension in the sagittal plane. The axis of rotation lies perpendicular to the sagittal plane and intersects the femoral condyles. Both fixed axes and screw axes have been calculated (Fig. 20.11). In Fig. 20.11, the optimal axes are fixed axes, whereas the screw axis is an instantaneous axis. The symmetric optimal axis is constrained such that the axis is the same for both the right and left knee. The screw axis may sometimes coincide with the optimal axis but not always, depending upon the motions of the knee joint. The second degree of freedom is the axial rotation around the long axis of the tibia. Rotation of the leg around its long axis can only be performed with the knee flexed. There is also an automatic axial rotation which is involuntarily linked to flexion and extension. When the knee is flexed, the tibia internally rotates. Conversely, when the knee is extended, the tibia externally rotates. During knee flexion, the patella makes a rolling/gliding motion along the femoral articulating surface. Throughout the entire flexion range, the gliding motion is clockwise (Fig. 20.12). In contrast, the direction of the rolling motion is counter-clockwise between 0° and 90° and clockwise between 90° and 120° (Fig. 20.12). The mean amount of patellar gliding for all knees is approximately 6.5 mm per 10° of flexion between 0° and 80° and 4.5 mm per 10° of flexion between 80° and 120°. The relationship between the angle of flexion and the mean rolling/gliding ratio for all knees is shown in Fig. 20.13. Between 80° and 90° of knee flexion, the rolling motion of the articulating surface comes to a standstill and then changes direction. The reversal in movement occurs at the flexion angle where the quadriceps tendon first contacts the femoral groove.

20.3 Hip The hip joint is composed of the head of the femur and the acetabulum of the pelvis. The hip joint is one of the most stable joints in the body. The stability is provided by the rigid ball-and-socket configuration.


FIGURE 20.4 (A) Variations in inclination of axis of subtalar joint as projected upon the sagittal plane. The distribution of the measurements on the individual specimens is shown in the histogram. The single observation of an angle of almost 70° was present in a markedly cavus foot. (B) Variations in position of subtalar axis as projected onto the transverse plane. The angle was measured between the axis and the midline of the foot. The extent of individual variation is shown on the sketch and revealed in the histogram. Source: Inman VT. 1976. The Joints of the Ankle, Baltimore, Williams and Wilkins. With permission.

Geometry of the Articulating Surfaces The femoral head is spherical in its articular portion which forms two-thirds of a sphere. The diameter of the femoral head is smaller for females than for males (Table 20.10). In the normal hip, the center of the femoral head coincides exactly with the center of the acetabulum. The rounded part of the femoral head is spheroidal rather than spherical because the uppermost part is flattened slightly. This causes the load to be distributed in a ringlike pattern around the superior pole. The geometrical center of the femoral head is traversed by the three axes of the joint, the horizontal axis, the vertical axis, and the anterior/posterior axis. The head is supported by the neck of the femur, which joins the shaft. The axis of the femoral neck is obliquely set and runs superiorly, medially, and anteriorly. The angle of inclination of the femoral neck to the shaft in the frontal plane is the neck-shaft angle (Fig. 20.14). In most adults, this angle is


FIGURE 20.5

TABLE 20.4

Geometry of distal femur. The distances are defined in Table 20.4.

Geometry of the Distal Femur Condyle Lateral

Parameter

Medial

Symbol

Distance (mm)

Symbol

Distance (mm)

Medial/lateral distance

K1

K2

Anterior/posterior distance

K3

Posterior femoral condyle spherical radii Epicondylar width

K6

31 ± 2.3 (male) 28 ± 1.8 (female) 72 ± 4.0 (male) 65 ± 3.7 (female) 19.2 ± 1.7

32 ± 31 (male) 27 ± 3.1 (female) 70 ± 4.3 (male) 63 ± 4.5 (female) 20.8 ± 2.4

Medial/lateral spacing of center of spherical surfaces

K4 K7

Overall Symbol

Distance (mm)

K5

90 ± 6 (male) 80 ± 6 (female) 45.9 ± 3.4

K8

Note: See Fig. 20.5 for location of measurements. Sources: Yoshioka Y, Siu D, Cooke TDV. 1987. The anatomy of functional axes of the femur. J Bone Joint Surg 69A(6):873–880. Kurosawa H, Walker PS, Abe S, Garg A, Hunter T. 1985. Geometry and motion of the knee for implant and orthotic design. J Biomech 18(7):487.

about 130° (Table 20.10). An angle exceeding 130° is known as coxa valga; an angle less than 130° is known as coxa vara. The femoral neck forms an acute angle with the transverse axis of the femoral condyles. This angle faces medially and anteriorly and is called the angle of anteversion (Fig. 20.15). In the adult, this angle averages about 7.5° (Table 20.10). The acetabulum receives the femoral head and lies on the lateral aspect of the hip. The acetabulum of the adult is a hemispherical socket. Its cartilage area is approximately 16 cm2 [Von Lanz & Wauchsmuth, 1938]. Together with the labrum, the acetabulum covers slightly more than 50% of the femoral head [Tönnis, 1987]. Only the sides of the acetabulum are lined by articular cartilage, which is interrupted inferiorly by the deep acetabular notch. The central part of the cavity is deeper than the articular cartilage and is nonarticular. This part is called the acetabular fossae and is separated from the interface of the pelvic bone by a thin plate of bone.


FIGURE 20.6

Contour of the tibial plateau (transverse plane). The distances are defined in Table 20.5.

TABLE 20.5

Geometry of the Proximal Tibia

Parameter Tibial plateau with widths (mm) Medial plateau Lateral plateau Overall width Tibial plateau depths (mm) AP depth, medial AP depth, lateral Interspinous width (mm) Intercondylar depth (mm)

Symbols

All Limbs

Male

Female

T1 T3 T 1 + T2 + T3

32 ± 3.8 33 ± 2.6 76 ± 6.2

34 ± 3.9 35 ± 1.9 81 ± 4.5

30 ± 22 31 ± 1.7 73 ± 4.5

T4 T5 T2 T6

48 ± 5.0 42 ± 3.7 12 ± 1.7 48 ± 5.9

52 ± 3.4 45 ± 3.1 12 ± 0.9 52 ± 5.7

45 ± 4.1 40 ± 2.3 12 ± 2.2 45 ± 3.9

Source: Yoshioka Y, Siu D, Scudamore RA, Cooke TDV. 1989. Tibial anatomy in functional axes. J Orthop Res 7:132.

Joint Contact Miyanaga et al. [1984] studied the deformation of the hip joint under loading, the contact area between the articular surfaces, and the contact pressures. They found that at loads up to 1000 N, pressure was distributed largely to the anterior and posterior parts of the lunate surface with very little pressure applied to the central portion of the roof itself. As the load increased, the contact area enlarged to include the outer and inner edges of the lunate surface (Fig. 20.16). However, the highest pressures were still measured anteriorly and posteriorly. Of five hip joints studied, only one had a pressure maximum at the zenith or central part of the acetabulum. Davy et al. [1989] utilized a telemetered total hip prosthesis to measure forces across the hip after total hip arthroplasty. The orientation of the resultant joint contact force varies over a relatively limited range during the weight-load-bearing portions of gait. Generally, the joint contact force on the ball of the hip prosthesis is located in the anterior/superior region. A three-dimensional plot of the resultant joint force during the gait cycle, with crutches, is shown in Fig. 20.17.


Sulcus Angle

WG DG

Femur 160

Sulcus Angle (°)

160

140 15

45

75 View Angle (°)

45

75 View Angle (°)

Depth Index 9 7 5 15

FIGURE 20.7 The trochlear geometry indices. The sulcus angle is the angle formed by the lines drawn from the top of the medial and lateral condyles to the deepest point of the sulcus. The depth index is the ratio of the width of the groove (WG) to the depth (DG). Mean and SD; n = 12. Source: Farahmand et al. Quantitative study of the quadriceps muscles and trochlear groove geometry related to instability of the patellofemoral joint, J Orthopaedic Res, 16:1, 140.

γm

γn

Patellar Equator

MED.

LAT.

FIGURE 20.8 Medial (γm) and lateral (γn) patellar facet angles. Source: Ahmed AM, Burke DL, Hyder A. 1987. Force analysis of the patellar mechanism. J Orthopedic Res 5:69–85. TABLE 20.6 Facet Angle γm (deg) γm (deg)

Patellar Facet Angles Knee Flexion Angle 0°

30°

60°

90°

120°

60.88 3.89a 67.76 4.15

60.96 4.70 68.05 3.97

61.43 4.12 69.36 3.63

61.30 4.18 68.39 4.01

60.34 4.51 68.20 3.67

a SD Source: Ahmed AM, Burke DL, Hyder A. 1987. Force analysis of the patellar mechanism. J Orthopedic Research 5:69–85.


FIGURE 20.9 Tibio-femoral contact area as a function of knee flexion angle. Source: Iseki F, Tomatsu T. 1976. The biomechanics of the knee joint with special reference to the contact area. Keio J Med 25:37. With permission. TABLE 20.7

Tibiofemoral Contact Area

Knee Flexion (deg)

Contact Area (cm2)

–5 5 15 25 35 45 55 65 75 85

20.2 19.8 19.2 18.2 14.0 13.4 11.8 13.6 11.4 12.1

Source: Maquet PG, Vandberg AJ, Simonet JC. 1975. Femorotibial weight bearing areas: Experimental determination. J Bone Joint Surg 57A(6):766.

TABLE 20.8 Posterior Displacement of the Femur Relative to the Tibia

Authors Kurosawa [1985] Andriacchi [1986] Draganich [1987] Nahass [1991]

Condition

A/P Displacement (mm)

In vitro In vitro In vitro In vivo (walking) In vivo (stairs)

14.8 13.5 13.5 12.5 13.9

Axes of Rotation The human hip is a modified spherical (ball-and-socket) joint. Thus, the hip possesses three degrees of freedom of motion with three correspondingly arranged, mutually perpendicular axes that intersect at the geometric center of rotation of the spherical head. The transverse axis lies in the frontal plane and controls movements of flexion and extension. An anterior/posterior axis lies in the sagittal plane and controls movements of adduction and abduction. A vertical axis which coincides with the long axis of the limb when the hip joint is in the neutral position controls movements of internal and external rotation. Surface motion in the hip joint can be considered as spinning of the femoral head on the acetabulum. © 2000 by CRC Press LLC

TABLE 20.9

Patellofemoral Contact Area

Knee Flexion (degrees)

Contact Area (cm2)

20 30 60 90 120

2.6 ± 0.4 3.1 ± 0.3 3.9 ± 0.6 4.1 ± 1.2 4.6 ± 0.7

Source: Hubert HH, Hayes WC. 1984. Patellofemoral contact pressures: The influence of Q-angle and tendofemoral contact. J Bone and Joint Surgery 66A(5):715–725. S 90

A

L

M 45 20 B 20°

45°

90°

S A

M

L

FIGURE 20.10 Diagrammatic representation of patella contact areas for varying degrees of knee flexion. Source: Goodfellow J, Hungerford DS, Zindel M, Patellofemoral joint mechanics and pathology. J Bone and Joint Surgery 58-B:3, 288.

B 135°

The pivoting of the bone socket in three planes around the center of rotation in the femoral head produces the spinning of the joint surfaces.

20.4 Shoulder The shoulder represents the group of structures connecting the arm to the thorax. The combined movements of four distinct articulations—glenohumeral, acromioclavicular, sternoclavicular, and scapulothoracic—allow the arm to be positioned in space.

Geometry of the Articulating Surfaces The articular surface of the humerus is approximately one-third of a sphere (Fig. 20.18). The articular surface is oriented with an upward tilt of approximately 45° and is retroverted approximately 30° with respect to the condylar line of the distal humerus [Morrey & An, 1990]. The average radius of curvature of the humeral head in the coronal plane is 24.0 ± 2.1 mm [Iannotti et al., 1992]. The radius of curvature © 2000 by CRC Press LLC

FIGURE 20.11 Approximate location of the optimal axis (case 1—nonsymmetric, case 3—symmetric), and the screw axis (case 2) on the medial and lateral condyles of the femur of a human subject for the range of motion of 0–90° flexion (standing to sitting, respectively). Source: Lewis JL, Lew WD. 1978. A method for locating an optimal “fixed” axis of rotation for the human knee joint. J Biomech Eng 100:187. With permission.

FIGURE 20.12 Position of patellar ligament, patella, and quadriceps tendon and location of the contact points as a function of the knee flexion angle. Source: van Eijden TMGJ, Kouwenhoven E, Verburg J, et al. A mathematical model of the patellofemoral joint. J Biomechanics 19(3):227.

in the anteroposterior and axillary-lateral view is similar, measuring 13.1 ± 1.3 mm and 22.9 ± 2.9 mm, respectively [McPherson et al., 1997]. The humeral articulating surface is spherical in the center. However, the peripheral radius is 2 mm less in the axial plane than in the coronal plane. Thus the peripheral contour of the articular surface is elliptical with a ratio of 0.92 [Iannotti et al., 1992]. The major axis is superior to inferior and the minor axis is anterior to posterior [McPherson et al., 1997]. More recently, the three-dimensional geometry of the proximal humerus has been studied extensively. The articular surface, which is part of a sphere, varies individually in its orientation with respect to inclination and retroversion, and it has variable medial and posterior offsets [Boileau and Walch, 1997]. These findings have great impact in implant design and placement in order to restore soft-tissue function. The glenoid fossa consists of a small, pear-shaped, cartilage-covered bony depression that measures 39.0 ± 3.5 mm in the superior/inferior direction and 29.0 ± 3.2 mm in the anterior/posterior direction [Iannotti et al., 1992]. The anterior/posterior dimension of the glenoid is pear-shaped with the lower © 2000 by CRC Press LLC

Rolling Gliding 0.4 0.2 0 20

40

60

-0.2

80

100 120

Angle of Flexion

-0.4 -0.6

FIGURE 20.13 Calculated rolling/gliding ratio for the patellofemoral joint as a function of the knee flexion angle. Source: van Eijden TMGJ, Kouwenhoven E, Verburg J, et al. A mathematical model of the patellofemoral joint. J Biomechanics 19(3):226.

TABLE 20.10

Geometry of the Proximal Femur

Parameter Femoral head diameter (mm) Neck shaft angle (degrees) Anteversion (degrees)

Females

Males

45.0 ± 3.0 133 ± 6.6 8 ± 10

52.0 ± 3.3 129 ± 7.3 7.0 ± 6.8

Source: Yoshioka Y, Siu D, Cooke TDV. 1987. The anatomy and functional axes of the femur. J Bone Joint Surg 69A(6):873.

FIGURE 20.14

The neck-shaft angle.

half being larger than the top half. The ratio of the lower half to the top half is 1:0.80 ± 0.01 [Iannotti et al., 1992]. The glenoid radius of curvature is 32.2 ± 7.6 mm in the anteroposterior view and 40.6 ± 14.0 mm in the axillary-lateral view [McPherson et al., 1997]. The glenoid is therefore more curved


FIGURE 20.15 The normal anteversion angle formed by a line tangent to the femoral condyles and the femoral neck axis, as displayed in the superior view.

FIGURE 20.16 Pressure distribution and contact area of hip joint. The pressure is distributed largely to the anterior and posterior parts of the lunate surface. As the load increased, the contact area increased. Source: Miyanaga Y, Fukubayashi T, Kurosawa H. 1984. Contact study of the hip joint: Load deformation pattern, contact area, and contact pressure. Arch Orth Trauma Surg 103:13. With permission.

superior to inferior (coronal plane) and relatively flatter in an anterior to posterior direction (sagittal plane). Glenoid depth is 5.0 ± 1.1 mm in the anteroposterior view and 2.9 ± 1.0 mm in the axillarylateral [McPherson et al., 1997], again confirming that the glenoid is more curved superior to inferior. In the coronal plane the articular surface of the glenoid comprises an arc of approximately 75° and in the transverse plane the arc of curvature of the glenoid is about 50° [Morrey & An, 1990]. The glenoid has a slight upward tilt of about 5° [Basmajian & Bazant, 1959] with respect to the medial border of the scapula (Fig. 20.19) and is retroverted a mean of approximately 7° [Saha, 1971]. The relationship of the


FIGURE 20.17 Scaled three-dimensional plot of resultant force during the gait cycle with crutches. The lengths of the lines indicate the magnitude of force. Radial line segments are drawn at equal increments of time, so the distance between the segments indicates the rate at which the orientation of the force was changing. For higher amplitudes of force during stance phase, line segments in close proximity indicate that the orientation of the force was changing relatively little with the cone angle between 30 and 40° and the polar angle between –25 and –15°. Source: Davy DT, Kotzar DM, Brown RH, et al. 1989. Telemetric force measurements across the hip after total arthroplasty. J Bone Joint Surg 70A(1):45, with permission.

FIGURE 20.18 The two-dimensional orientation of the articular surface of the humerus with respect to the bicondylar axis. By permission of Mayo Foundation.

dimension of the humeral head to the glenoid head is approximately 0.8 in the coronal plane and 0.6 in the horizontal or transverse plane [Saha, 1971]. The surface area of the glenoid fossa is only one-third to one-fourth that of the humeral head [Kent, 1971]. The arcs of articular cartilage on the humeral head


FIGURE 20.19 The glenoid faces slightly superior and posterior (retroverted) with respect to the body of the scapula. By permission of Mayo Foundation.

and glenoid in the frontal and axial planes were measured [Jobe and Iannotti, 1995]. In the coronal plane, the humeral heads had an arc of 159° covered by 96° of glenoid, leaving 63° of cartilage uncovered. In the transverse plane, the humeral arc of 160° is opposed by 74° of glenoid, leaving 86° uncovered.

Joint Contact The degree of conformity and constraint between the humeral head and glenoid has been represented by conformity index (radius of head/radius of glenoid) and constraint index (arc of enclosure/360) [McPherson, 1997]. Based on the study of 93 cadaveric specimens, the mean conformity index was 0.72 in the coronal and 0.63 in the sagittal plane. There was more constraint to the glenoid in the coronal vs. sagittal plane (0.18 vs. 0.13). These anatomic features help prevent superior–inferior translation of the humeral head but allow translation in the sagittal plane. Joint contact areas of the glenohumeral joint tend to be greater at mid-elevation positions than at either of the extremes of joint position (Table 20.11). These results suggest that the glenohumeral surface is maximum at these more functional positions, thus distributing joint load over a larger region in a more stable configuration. The contact point moves TABLE 20.11

Glenohumeral Contact Areas

Elevation angle (°)

Contact areas at SR (cm2)

Contact areas at 20° internal to SR (cm2)

0 30 60 90 120 150 180

0.87 ± 1.01 2.09 ± 1.54 3.48 ± 1.69 4.95 ± 2.15 5.07 ± 2.35 3.52 ± 2.29 2.59 ± 2.90

1.70 ± 1.68 2.44 ± 2.15 4.56 ± 1.84 3.92 ± 2.10 4.84 ± 1.84 2.33 ± 1.47 2.51 ± NA

SR = Starting external rotation which allowed the shoulder to reach maximal elevation in the scapular plane (≈40° ± 8°); NA = Not applicable Source: Soslowsky LJ, Flatow EL, Bigliani LU, Pablak RJ, Mow VC, Athesian GA. 1992. Quantitation of in situ contact areas at the glenohumeral joint: A biomechanical study. J Orthop Res 10(4):524. With permission.


FIGURE 20.20 Humeral contact positions as a function of glenohumeral motion and positions. Source: Morrey BF, An KN. 1990. Biomechanics of the Shoulder. In CA Rockwood and FA Matsen (eds), The Shoulder, pp 208–245, Philadelphia, Saunders. With permission.

forward and inferior during internal rotation (Fig. 20.20). With external rotation, the contact is posterior/inferior. With elevation, the contact area moves superiorly. Lippitt and associates [1998] calculated the stability ratio, which is defined as a force necessary to translate the humeral head from the glenoid fossa divided by the compressive load times 100. The stability ratios were in the range of 50 to 60% in the superior–inferior direction and 30 to 40% in the anterior–posterior direction. After the labrum was removed, the ratio decreased by approximately 20%. Joint conformity was found to have significant influence on translations of humeral head during active positioning by muscles [Karduna, et al. 1996].

Axes of Rotation The shoulder complex consists of four distinct articulations: the glenohumeral joint, the acromioclavicular joint, the sternoclavicular joint, and the scapulothoracic articulation. The wide range of motion of the shoulder (exceeding a hemisphere) is the result of synchronous, simultaneous contributions from each joint. The most important function of the shoulder is arm elevation. Several investigators have attempted to relate glenohumeral and scapulothoracic motion during arm elevation in various planes (Table 20.12). About two-thirds of the motion takes place in the glenohumeral joint and about one-third in the scapulothoracic articulation, resulting in a 2:1 ratio. Surface motion at the glenohumeral joint is primarily rotational. The center of rotation of the glenohumeral joint has been defined as a locus of points situated within 6.0 ± 1.8 mm of the geometric center of the humeral head [Poppen & Walker, 1976]. However, the motion is not purely rotational. The humeral head displaces, with respect to the glenoid. From 0–30°, and often from 30–60°, the humeral head moves upward in the glenoid fossa by about 3 mm, indicating that rolling and/or gliding has taken place. Thereafter, the humeral head has only about 1 mm of additional excursion. During arm elevation in the scapular plane, the scapula moves in relation to the thorax [Poppen & Walker, 1976]. From 0–30° the


TABLE 20.12 Arm Elevation: GlenohumeralScapulothoracic Rotation Investigator Inman et al. [1994] Freedman & Munro [1966] Doody et al. [1970] Poppen & Walker [1976] Saha [1971]

Glenohumeral/Scapulothoracic Motion Ratio 2:1 1.35:1 1.74:1 4.3:1 (24° elevation) 2.3:1 (30–135° elevation)

FIGURE 20.21 Rotation of the scapula on the thorax in the scapular plane. Instant centers of rotation (solid dots) are shown for each 30° interval of motion during shoulder elevation in the scapular plane from zero to 150°. The x and y axes are fixed in the scapula, whereas the X and Y axes are fixed in the thorax. From zero to 30° in the scapula rotated about its lower midportion; from 60° onward, rotation took place about the glenoid area, resulting in a medial and upward displacement of the glenoid face and a large lateral displacement of the inferior tip of the scapula. Source: Poppen NK, Walker PS. 1976. Normal and abnormal motion of the shoulder. J Bone Joint Surg 58A:195. With permission.

scapula rotates about its lower mid portion, and then from 60° onward the center of rotation shifts toward the glenoid, resulting in a large lateral displacement of the inferior tip of the scapula (Fig. 20.21). The center of rotation of the scapula for arm elevation is situated at the tip of the acromion as viewed from the edge on (Fig. 20.22). The mean amount of scapular twisting at maximum arm elevation is 40°. The superior tip of the scapula moves away from the thorax, and the inferior tip moves toward it.


FIGURE 20.22 (A) A plot of the tips of the acromion and coracoid process on roentgenograms taken at successive intervals of arm elevation in the scapular plane shows upward movement of the coracoid and only a slight shift in the acromion relative to the glenoid face. This finding demonstrates twisting, or external rotation, of the scapula about the x axis. (B) A lateral view of the scapula during this motion would show the coracoid process moving upward while the acromion remains on the same horizontal plane as the glenoid. Source: Poppen NK, Walker PS. 1976. Normal and abnormal motion of the shoulder. J Bone Joint Surg 58A:195. With permission.

20.5 Elbow The bony structures of the elbow are the distal end of the humerus and the proximal ends of the radius and ulna. The elbow joint complex allows two degrees of freedom in motion: flexion/extension and pronation/supination. The elbow joint complex is three separate synovial articulations. The humeralulnar joint is the articulation between the trochlea of the distal radius and the trochlear fossa of the proximal ulna. The humero-radial joint is formed by the articulation between the capitulum of the distal humerus and the head of the radius. The proximal radioulnar joint is formed by the head of the radius and the radial notch of the proximal ulna.

Geometry of the Articulating Surfaces The curved, articulating portions of the trochlea and capitulum are approximately circular in a crosssection. The radius of the capitulum is larger than the central trochlear groove (Table 20.13). The centers of curvature of the trochlea and capitulum lie in a straight line located on a plane that slopes at 45–50° anterior and distal to the transepicondylar line and is inclined at 2.5° from the horizontal transverse plane [Shiba et al., 1988]. The curves of the ulnar articulations form two surfaces (coronoid and olecranon) with centers on a line parallel to the transepicondylar line but are distinct from it [Shiba et al., 1988]. The carrying angle is an angle made by the intersection of the longitudinal axis of the humerus and the forearm in the frontal plane with the elbow in an extended position. The carrying angle is contributed to, in part, by the oblique axis of the distal humerus and, in part, by the shape of the proximal ulna (Fig. 20.23). © 2000 by CRC Press LLC

TABLE 20.13

Elbow Joint Geometry

Parameter

Size (mm)

Capitulum radius Lateral trochlear flange radius Central trochlear groove radius Medial trochlear groove radius Distal location of flexion/extension axis from transepicondylar line: Lateral Medial

10.6 ± 1.1 10.8 ± 1.0 8.8 ± 0.4 13.2 ± 1.4

6.8 ± 0.2 8.7 ± 0.6

Source: Shiba R, Sorbie C, Siu DW, Bryant JT, Cooke TDV, Weavers HW. 1988. Geometry of the humeral-ulnar joint. J Orthop Res 6:897.

FIGURE 20.23 Components contributing to the carrying angles: α + λ + ψ. Key: α, angle between C-line and TEL; γ, inclination of central groove (cg); λ, angle between trochlear notch (tn); ψ, reverse angulation of shaft of ulna; TLE, transepicondylar line; C-line, line joining centers of curvature of the trochlea and capitellum; cg, central groove; op, olecranon process; tr, trochlear ridge; cp, coronoid process. α = 2.5 ± 0.0; λ = 17.5 ± 5.0 (females) and 12.0 ± 7.0 (males); ψ = –6.5 ± 0.7 (females) and –9.5 ± 3.5 (males). Source: Shiba R, Sorbie C, Siu DW, Bryant JT, Cooke TDV, Weavers HW. 1988. Geometry of the humeral-ulnar joint. J Orthop Res 6:897. With permission.

Joint Contact The contact area on the articular surfaces of the elbow joint depends on the joint position and the loading conditions. Increasing the magnitude of the load not only increases the size of the contact area but shifts the locations as well (Fig. 20.25). As the axial loading is increased, there is an increased lateralization of the articular contact [Stormont et al., 1985]. The area of contact, expressed as a percentage of the total articulating surface area, is given in Table 20.14. Based on a finite element model of the humero-ulnar joint, Merz et al. [1997] demonstrated that the humero-ulnar joint incongruity brings about a bicentric distribution of contact pressure, a tensile stress exists in the notch that is the same order of magnitude as the compressive stress [Merz, 1997]. © 2000 by CRC Press LLC

FIGURE 20.24 Contact of the ulnohumeral joint with varus and valgus loads and the elbow at 90°. Notice only minimal radiohumeral contact in this loading condition. Source: Stormont TJ, An KN, Morrey BF, Chae EY. 1985. In Elbow joint contact study: Comparison of techniques. J Biomech 18(5):329. Reprinted with permission of Elsevier Science Inc.

FIGURE 20.25 Very small locus of instant center of rotation for the elbow joint demonstrates that the axis may be replicated by a single line drawn from the inferior aspect of the medial epicondyle through the center of the lateral epicondyle, which is in the center of the lateral projected curvature of the trochlea and capitellum. Source: Modified from Morrey BF, Chao EYS. 1976. In Passive motion of the elbow joint: a biomechanical analysis. J Bone Joint Surg 58A(4):501. Reprinted with permission of the Journal of Bone and Joint Surgery.

Axes of Rotation The axes of flexion and extension can be approximated by a line passing through the center of the trochlea, bisecting the angle formed by the longitudinal axes of the humerus and the ulna [Morrey & Chao, 1976]. © 2000 by CRC Press LLC

TABLE 20.14

Position Full extension 90° flexion Full flexion

Elbow Joint Contact Area Total Articulating Surface Area of Ulna and Radial Head (mm2)

Contact Area (%)

1598 ± 103 1750 ± 123 1594 ± 120

8.1 ± 2.7 10.8 ± 2.0 9.5 ± 2.1

Source: Goel VK, Singh D, Bijlani V. 1982. Contact areas in human elbow joints. J Biomech Eng 104:169.

The instant centers of flexion and extension vary within 2–3 mm of this axis (Fig. 20.25). With the elbow

FIGURE 20.26 During elbow flexion and extension, a linear change in the carrying angle is demonstrated, typically going from valgus in extension to varus in flexion. Source: Morrey BF, Chao EYS. 1976. In Passive motion of the elbow joint: A biomechanical analysis. J Bone Joint Surg 58A(4):501. Reprinted with permission of the Journal of Bone and Joint Surgery.

fully extended and the forearm fully supinated, the longitudinal axes of humerus and ulna normally intersect at a valgus angle referred to as the carrying angle. In adults, this angle is usually 10–15° and normally is greater on average in women [Zuckerman & Matsen, 1989]. As the elbow flexes, the carrying angle varies as a function of flexion (Fig. 20.26). In extension there is a valgus angulation of 10°; at full flexion there is a varus angulation of 8° [Morrey & Chao, 1976]. More recently, the three-dimensional kinematics of the ulno-humeral joint under simulated active elbow joint flexion-extension was obtained by using an electromagnetic tracking device [Tanaka et al. 1998]. The optimal axis to best represent flexion–extension motion was found to be close to the line joining the centers of the capitellum and the trochlear groove. Furthermore, the joint laxity under valgus–varus stress was also examined. With the weight of the forearm as the stress, a maximum of 7.6° valgus–varus and 5.3° of axial rotation laxity were observed.

20.6 Wrist The wrist functions by allowing changes of orientation of the hand relative to the forearm. The wrist joint complex consists of multiple articulations of eight carpal bones with the distal radius, the structures © 2000 by CRC Press LLC

of the ulnocarpal space, the metacarpals, and each other. This collection of bones and soft tissues is capable of a substantial arc of motion that augments hand and finger function. TABLE 20.15

Changes of Wrist Geometry with Grasp

Distal radioulnar joint space (mm) Ulnar variance (mm) Lunate, uncovered length (mm) Capitate length (mm) Carpal height (mm) Carpal ulnar distance (mm) Carpal radial distance (mm) Third metacarpal length (mm) Carpal height ratio Carpal ulnar ratio Lunate uncovering index Carpal radial ratio Radius—third metacarpal angle (degrees) Radius—capitate angle (degrees)

Resting

Grasp

Analysis of Variance (p = level)

1.6 ± 0.3 –0.2 ± 1.6 6.0 ± 1.9 21.5 ± 2.2 33.4 ± 3.4 15.8 ± 4.0 19.4 ± 1.8 63.8 ± 5.8 52.4 ± 3.3 24.9 ± 5.9 36.7 ± 12.1 30.6 ± 2.4 –0.3 ± 9.2 0.4 ± 15.4

1.8 ± 0.6 0.7 ± 1.8 7.6 ± 2.6 20.8 ± 2.3 31.7 ± 3.4 15.8 ± 3.0 19.7 ± 1.8 62.6 ± 5.5 50.6 ± 4.1 25.4 ± 5.3 45.3 ± 14.2 31.6 ± 2.3 –3.1 ± 12.8 –3.8 ± 22.2

0.06 0.003 0.0008 0.0002 0.0001 NS NS NS 0.02 NS 0.002 NS NS NS

Note: 15 normal subjects with forearm in neutral position and elbow at 90° flexion. Source: Schuind FA, Linscheid RL, An KN, Chao EYS. 1992. Changes in wrist and forearm configuration with grasp and isometric contraction of elbow flexors. J Hand Surg 17A:698.

Geometry of the Articulating Surfaces The global geometry of the carpal bones has been quantified for grasp and active isometric contraction of the elbow flexors [Schuind et al., 1992]. During grasping there is a significant proximal migration of the radius of 0.9 mm, apparent shortening of the capitate, a decrease in the carpal height ratio, and an increase in the lunate uncovering index (Table 20.15). There is also a trend toward increase of the distal radioulnar joint with grasping. The addition of elbow flexion with concomitant grasping did not significantly change the global geometry, except for a significant decrease in the forearm interosseous space [Schuind et al., 1992].

Joint Contact Studies of the normal biomechanics of the proximal wrist joint have determined that the scaphoid and lunate bones have separate, distinct areas of contact on the distal radius/triangular fibrocartilage complex surface [Viegas et al., 1987] so that the contact areas were localized and accounted for a relatively small fraction of the joint surface, regardless of wrist position (average of 20.6%). The contact areas shift from a more volar location to a more dorsal location as the wrist moves from flexion to extension. Overall, the scaphoid contact area is 1.47 times greater than that of the lunate. The scapho-lunate contact area ratio generally increases as the wrist position is changed from radial to ulnar deviation and/or from flexion to extension. Palmer and Werner [1984] also studied pressures in the proximal wrist joint and found that there are three distinct areas of contact: the ulno-lunate, radio-lunate, and radio-scaphoid. They determined that the peak articular pressure in the ulno-lunate fossa is 1.4 N/mm2, in the radioulnate fossa is 3.0 N/mm2, and in the radio-scaphoid fossa is 3.3 N/mm2. Viegas et al. [1989] found a nonlinear relationship between increasing load and the joint contact area (Fig. 20.27). In general, the distribution of load between the scaphoid and lunate was consistent with all loads tested, with 60% of the total contact area involving the scaphoid and 40% involving the lunate. Loads greater than 46 lbs were found to not significantly increase the overall contact area. The overall contact area, even at the highest loads tested, was not more than 40% of the available joint surface.


FIGURE 20.27 The nonlinear relation between the contact area and the load at the proximal wrist joint. The contact area was normalized as a percentage of the available joint surface. The load of 11, 23, 46, and 92 lbs. was applied at the position of neutral pronation/supination, neutral radioulnar deviation, and neutral flexion/extension. Source: Viegas SF, Patterson RM, Peterson PD, Roefs J, Tencer A, Choi S. 1989. The effects of various load paths and different loads on the load transfer characteristics of the wrist. J Hand Surg 14A(3):458. With permission.

Horii et al. [1990] calculated the total amount of force born by each joint with the intact wrist in the neutral position in the coronal plane and subjected to a total load of 143 N (Table 20.16). They found TABLE 20.16

Force Transmission at the Intercarpal Joints

Joint Radio-ulno-carpal Ulno-triquetral Ulno-lunate Radio-lunate Radio-scaphoid Midcarpal Triquetral-hamate Luno-capitate Scapho-capitate Scapho-trapezial

Force (N) 12 ± 3 23 ± 8 52 ± 8 74 ± 13 36 ± 6 51 ± 6 32 ± 4 51 ± 8

Note: A total of 143 N axial force applied across the wrist. Source: Horii E, Garcia-Elias M, An KN, Bishop AT, Cooney WP, Linscheid RL, Chao EY. 1990. Effect on force transmission across the carpus in procedures used to treat Kienböck’s Disease. J Bone Joint Surg 15A(3):393.

that 22% of the total force in the radio-ulno-carpal joint is dissipated through the ulna 14% through the ulno-lunate joint, and 18% through the ulno-triquetral joint) and 78% through the radius (46% through the scaphoid fossa and 32% through the lunate fossa). At the midcarpal joint, the scapho-trapezial joint transmits 31% of the total applied force, the scapho-capitate joint transmits 19%, the luno-capitate joint transmits 29%, and the triquetral-hamate joints transmits 21% of the load. © 2000 by CRC Press LLC

FIGURE 20.28 The location of the center of rotation during ulnar deviation (left) and extension (right), determined graphically using two metal markers embedded in the capitate. Note that during radial-ulnar deviation the center lies at a point in the capitate situated distal to the proximal end of this bone by a distance equivalent to approximately one-quarter of its total longitudinal length. During flexion-extension, the center of rotation is close to the proximal cortex of the capitate. Source: Youm Y, McMurty RY, Flatt AE, Gillespie TE. 1978. Kinematics of the wrist: an experimental study of radioulnar deviation and flexion/extension. J Bone Joint Surg 60A(4):423. With permission.

A limited amount of studies have been done to determine the contact areas in the midcarpal joint. Viegas et al. [1990] have found four general areas of contact: the scapho-trapezial-trapezoid (STT), the scapho-capitate (SC), the capito-lunate (CL), and the triquetral-hamate (TH). The high pressure contact area accounted for only 8% of the available joint surface with a load of 32 lbs and increased to a maximum of only 15% with a load of 118 lbs. The total contact area, expressed as a percentage of the total available joint area for each fossa was: STT = 1.3%, SC = 1.8%, CL = 3.1%, and TH = 1.8%. The correlation between the pressure loading in the wrist and the progress of degenerative osteoarthritis associated with pathological conditions of the forearm was studied in a cadaveric model [Sato, 1995]. Malunion after distal radius fracture, tear of triangular fibrocartilage, and scapholunate dissociation were all responsible for the alteration of the articulating pressure across the wrist joint. Residual articular incongruity of the distal radius following intra-articular fracture has been correlated with early osteoarthritis. In an in vitro model, step-offs of the distal radius articular incongruity were created. Mean contact stress was significantly greater than the anatomically reduced case at only 3 mm of step-off [Anderson et al., 1996].

Axes of Rotation The complexity of joint motion at the wrist makes it difficult to calculate the instant center of motion. However, the trajectories of the hand during radioulnar deviation and flexion/extension, when they occur in a fixed plane, are circular, and the rotation in each plane takes place about a fixed axis. These axes are located within the head of the capitate and are not altered by the position of the hand in the plane of rotation [Youm et al., 1978]. During radioulnar deviation, the instant center of rotation lies at a point in the capitate situated distal to the proximal end of this bone by a distance equivalent to approximately one-quarter of its total length (Fig. 20.28). During flexion/extension, the instant center is close to the proximal cortex of the capitate, which is somewhat more proximal than the location for the instant center of radioulnar deviation. Normal carpal kinematics were studied in 22 cadaver specimens using a biplanar radiography method. The kinematics of the trapezium, capitate, hamate, scaphoid, lunate, and triquetrum were determined during wrist rotation in the sagittal and coronal plane [Kobagashi et al., 1997]. The results were expressed


using the concept of the screw displacement axis and covered to describe the magnitude of rotation about and translation along three orthogonal axes. The orientation of these axes is expressed relative to the radius during sagittal plane motion of the wrist (Table 20.17). The scaphoid exhibited the greatest

TABLE 20.17

Individual Carpal Rotation Relative to the Radius (Degrees) (Sagittal Plane Motion of the Wrist) Axis of Rotation

Wrist Motiona Carpal Bone

X (+) Pronation; (–) Supination

Y (+) Flexion; (–) Extension

N-E60 N-E30 N-F30 N-F60 N-E60 N-E30 N-F60 N-E60

Trapezium (N=13) S.D. Capitate (N=22) S.D. Hamate (N=9) S.D. Scaphoid (N=22) S.D. Lunate (N=22) S.D. Triquetrum (N=22) S.D.

–0.9 2.8 0.9 2.7 0.4 3.4 –2.5 3.4 1.2 2.8 –3.5 3.5

–1.3 2.2 –1 1.8 –1 1.7 –0.7 2.6 0.5 1.8 –2.5 2

0.9 2.6 1.3 2.5 1.3 2.5 1.6 2.2 0.3 1.7 2.5 2.2

–1.4 2.7 –1.6 3.5 –0.3 2.4 2 3.1 –2.2 2.8 –0.7 3.7

–59.4 2.3 60.3 2.5 –59.5 1.4 –52.3 3 –29.7 6.6 –39.3 4.8

–29.3 1 –30.2 1.1 –29 0.8 –26 3.2 –15.4 3.9 –20.1 2.7

28.7 1.8 21.5 1.2 28.8 10.2 20.6 2.8 11.5 3.9 15.5 3.8

54.2 3 63.5 2.8 62.6 3.6 39.7 4.3 23 5.9 30.6 5.1

Z (+) Ulnar Deviation; (–) Radial Deviation N-E60

N-E30

N-F30

N-F60

1.2 4 0 2 2.1 4.4 4.5 3.7 4.3 2.6 0 2.8

0.3 2.7 0 1.4 0.7 1.8 0.8 2.1 0.9 1.5 –0.3 1.4

–0.4 1.3 0.6 1.6 0.1 1.2 2.1 2.2 3.3 1.9 2.4 2.6

2.5 2.8 3.2 3.6 1.8 4.1 7.8 4.5 11.1 3.4 9.8 4.3

a N-E60: neutral to 60° of extension; N-E30: neutral to 30° of extension; N-F30: neutral to 30° of flexion; N-F60: neutral to 60° of flexion. S.D. = standard deviation. Source: Kobayashi M, Berger RA, Nagy L, et al. 1997. Normal kinematics of carpal bones: A three-dimensional analysis of carpal bone motion relative to the radius. J Biomech 30:8, 787.

magnitude of rotation and the lunate displayed the least rotation. The proximal carpal bones exhibited some ulnar deviation in 60° of wrist flexion. During coronal plane motion (Table 20.18), the magnitude of radial-ulnar deviation of the distal carpal bones was mutually similar and generally of a greater magnitude than that of the proximal carpal bones. The proximal carpal bones experienced some flexion during radial deviation of the wrist and extension during ulnar deviation of the wrist.

20.7 Hand The hand is an extremely mobile organ that is capable of conforming to a large variety of object shapes and coordinating an infinite variety of movements in relation to each of its components. The mobility of this structure is possible through the unique arrangement of the bones in relation to one another, the articular contours, and the actions of an intricate system of muscles. Theoretical and empirical evidence suggest that limb joint surface morphology is mechanically related to joint mobility, stability, and strength [Hamrick, 1996].

Geometry of the Articulating Surfaces Three-dimensional geometric models of the articular surfaces of the hand have been constructed. The sagittal contours of the metacarpal head and proximal phalanx grossly resemble the arc of a circle [Tamai et al., 1988]. The radius of curvature of a circle fitted to the entire proximal phalanx surface ranges from 11–13 mm, almost twice as much as that of the metacarpal head, which ranges from 6–7 mm (Table 20.19). The local centers of curvature along the sagittal contour of the metacarpal heads are not fixed. The locus of the center of curvature for the subchondral bony contour approximates the locus of


TABLE 20.18

Individual Carpal Rotation to the Radius (Degrees) (Coronal Plane Motion of the Wrist) Axis of Rotation

Wrist Motiona Carpal Bone

X (+) Pronation; (–) Supination

Y (+) Flexion; (–) Extension

Z (+) Ulnar Deviation; (–) Radial Deviation

N-RD15

N-UD15

N-UD30

N-RD15

N-UD15

N-UD30

N-RD15

N-UD15

N-UD30

–4.8 2.4 –3.9 2.6 –4.8 1.8 0.8 1.8 –1.2 1.6 –1.1 1.4

9.1 3.6 6.8 2.6 6.3 2.4 2.2 2.4 1.4 0 –1 2.6

16.3 3.6 11.8 2.5 10.6 3.1 6.6 3.1 3.9 3.3 0.8 4

0 1.5 1.3 1.5 1.1 3 8.5 3 7 3.1 4.1 3

4.9 1.3 2.7 1.1 3.5 3.2 –12.5 3.2 –13.9 4.3 –10.5 3.8

9.9 2.1 6.5 1.7 6.6 4.1 –17.1 4.1 –22.5 6.9 –17.3 6

–14.3 2.3 –14.6 2.1 –15.5 2.4 –4.2 2.4 –1.7 1.7 –5.1 2.4

16.4 2.8 15.9 1.4 15.4 2.6 4.3 2.6 5.7 2.8 7.7 2.2

32.5 2.6 30.7 1.7 30.2 3.6 13.6 3.6 15 4.3 18.4 4

Trapezium (N=13) S.D. Capitate (N=22) S.D. Hamate (N=9) S.D. Scaphoid (N=22) S.D. Lunate (N=22) S.D. Triquetrum (N=22) S.D.

a N-RD15: neutral to 15° of radial deviation; N-UD30: neutral to 30° of ulnar deviation; N-UD15: neutral to 15° of ulnar deviation. S.D. = standard deviation. Source: Kobayashi M, Berger RA, Nagy L, et al. 1997. Normal kinematics of carpal bones: A three-dimensional analysis of carpal bone motion relative to the radius. J Biomech 30:8, 787.

the center for the acute curve of an ellipse (Fig. 20.29). However, the locus of center of curvature for the

FIGURE 20.29 The loci of the local centers of curvature for subchondral bony contour of the metacarpal head approximates the loci of the center for the acute curve of an ellipse. The loci of the local center of curvature for articular cartilage contour of the metacarpal head approximates the loci of the bony center of the obtuse curve of an ellipse. Source: Tamai K, Ryu J, An KN, Linscheid RL, Cooney WP, Chao EYS. 1988. In Three-dimensional geometric analysis of the metacarpophalangeal joint. J Hand Surg 13A(4):521. Reprinted with permission of Churchill Livingstone.

articular cartilage contour approximates the locus of the obtuse curve of an ellipse. The surface geometry of the thumb carpometacarpal (CMC) joint has also been quantified [Athesian et al., 1992]. The surface area of the CMC joint is significantly greater for males than for females (Table 20.20). The minimum, maximum, and mean square curvature of these joints is reported in


TABLE 20.19 Radius of Curvature of the Middle Sections of the Metacarpal Head and Proximal Phalanx Base Radius (mm)

MCH Index Long PPB Index Long

Bony Contour

Cartilage Contour

6.42 ± 1.23 6.44 ± 1.08 13.01 ± 4.09 11.46 ± 2.30

6.91 ± 1.03 6.66 ± 1.18 12.07 ± 3.29 11.02 ± 2.48

Source: Tamai K, Ryu J, An KN, Linscheid RL, Cooney WP, Chao EYS. 1988. Three-dimensional geometric analysis of the metacarpophalangeal joint. J Hand Surg 13A(4):521.

Table 20.20. The curvature of the surface is denoted by κ and the radius of curvature is ρ = 1/κ. The curvature is negative when the surface is concave and positive when the surface is convex.

Joint Contact The size and location of joint contact areas of the metacarpophalangeal (MCP) joint changes as a function of the joint flexion angle (Fig. 20.30) The radioulnar width of the contact area becomes narrow in the neutral position and expands in both the hyperextended and fully flexed positions [An & Cooney, 1991]. In the neutral position, the contact area occurs in the center of the phalangeal base, this area being slightly larger on the ulnar than on the radial side. The contact areas of the thumb carpometacarpal joint under the functional position of lateral key pinch and in the extremes of range of motion were studied using a stereophotogrammetric technique [Ateshian et al., 1995]. The lateral pinch position produced contact predominately on the central, volar, and volar–ulnar regions of the trapezium and the metacarpals (Fig. 20.31). Pelligrini et al. [1993] noted that the palmar compartment of the trapeziometacarpal joint was the primary contact area during flexion adduction of the thumb in lateral pinch. Detachment of the palmar beak ligament resulted in dorsal translation of the contact area producing a pattern similar to that of cartilage degeneration seen in the osteoarthritic joint.

Axes of Rotation Rolling and sliding actions of articulating surfaces exist during finger joint motion. The geometric shapes of the articular surfaces of the metacarpal head and proximal phalanx, as well as the insertion location of the collateral ligaments, significantly govern the articulating kinematics, and the center of rotation is not fixed but rather moves as a function of the angle of flexion [Pagowski & Piekarski, 1977]. The instant centers of rotation are within 3 mm of the center of the metacarpal head [Walker & Erhman, 1975]. Recently the axis of rotation of the MCP joint has been evaluated in vivo by Fioretti [1994]. The instantaneous helical axis of the MCP joint tends to be more palmar and tends to be displaced distally as flexion increases (Fig. 20.32). The axes of rotation of the CMC joint have been described as being fixed [Hollister et al., 1992], but others believe that a polycentric center of rotation exists [Imaeda et al., 1994]. Hollister et al. [1992] found that axes of the CMC joint are fixed and are not perpendicular to each other, or to the bones, and do not intersect. The flexion/extension axis is located in the trapezium, and the abduction/adduction axis is on the first metacarpal. In contrast, Imaeda et al. [1994] found that there was no single center of rotation, but rather the instantaneous motion occurred reciprocally between centers of rotations within the trapezium and the metacarpal base of the normal thumb. In flexion/extension, the axis of rotation was located within the trapezium, but for abduction/adduction the center of rotation was located distally to the trapezium and within the base of the first metacarpal. The average instantaneous center of circumduction was at approximately the center of the trapezial joint surface (Table 20.21).


FIGURE 20.30 (A) Contact area of the MCP joint in five joint positions. (B) End on view of the contact area on each of the proximal phalanx bases. The radioulnar width of the contact area becomes narrow in the neutral position and expands in both the hyperextended and fully flexed positions. Source: An KN, Cooney WP. 1991. Biomechanics, Section II. The hand and wrist. In BF Morrey (ed), Joint Replacement Arthroplasty, pp 137–146, New York, Churchill Livingstone. By permission of Mayo Foundation.

FIGURE 20.31 Summary of the contact areas for all specimens, in lateral pinch with a 25 N load. All results from the right hand are transposed onto the schema of a carpometacarpal joint from the left thumb. Source: Ateshian, GA, Ark JW, Rosenwasser MP, et al. Contact areas on the thumb carpometacarpal joint. J Orthop Res 13:450, 1995.


FIGURE 20.32 Intersections of the instantaneous helical angles with the metacarpal sagittal plane. They are relative to one subject tested twice in different days. The origin of the graph is coincident with the calibrated center of the metacarpal head. The arrow indicates the direction of flexion. Source: Fioretti S. 1994. Three-dimensional in-vivo kinematic analysis of finger movement. In F Schuind et al. (eds). Advances in the Biomechanics of the Hand and Wrist, pp 363–375, New York, Plenum Press. With permission.

TABLE 20.20

Curvature of Carpometacarpal Joint Articular Surfaces

Trapezium Female Male Total Female versus male Metacarpal Female Male Total Female versus male

–

–

–

Area (cm2)

κ min (m–1)

κ max (m–1)

κ rms (m–1)

8 5 13

1.05 ± 0.21 1.63 ± 0.18 1.27 ± 0.35 p ≤ 0.01

–61 ± 22 –87 ± 17 –71 ± 24 p ≤ 0.05

190 ± 36 114 ± 19 161 ± 48 p ≤ 0.01

165 ± 32 118 ± 6 147 ± 34 p ≤ 0.01

8 5 13

1.22 ± 0.36 1.74 ± 0.21 1.42 ± 0.40 p ≤ 0.01

–49 ± 10 –37 ± 11 –44 ± 12 p ≤ 0.05

175 ± 25 131 ± 17 158 ± 31 p ≤ 0.01

154 ± 20 116 ± 8 140 ± 25 p ≤ 0.01

n

Note: Radius of curvature: ρ = 1/k Source: Athesian JA, Rosenwasser MP, Mow VC. 1992. Curvature characteristics and congruence of the thumb carpometacarpal joint: differences between female and male joints. J Biomech 25(6):591.

The axes of rotation of the thumb interphalangeal and metacarpophalangeal joint were located using a mechanical device [Hollister et al., 1995]. The physiologic motion of the thumb joints occur about these axes (Fig. 20.33 and Table 20.22). The interphalangeal joint axis is parallel to the flexion crease of the joint and is not perpendicular to the phalanx. The metacarpophalangeal joint has two fixed axes: a fixed flexion–extension axis just distal and volar to the epicondyles, and an abduction–adduction axis related to the proximal phalanx passing between the sesamoids. Neither axis is perpendicular to the phalanges.


FIGURE 20.33 (A) The angles and length and breadth measurements defining the axis of rotation of the interphalangeal joint of the right thumb. (t/T = ratio of anatomic plane diameter; l/L = ratio of length). (B) The angles and length and breadth measurements of the metacarpophalangeal flexion-extension axis’ position in the metacarpal. (C) The angles and length and breadth measurements that locate the metacarpophalangeal abduction–adduction axis. The measurements are made in the metacarpal when the metacapophalangeal joint is at neutral flexion extension. The measurements are made relative to the metacarpal because the axis passes through this bone, not the proximal phalanx with which it moves. This method of recording the abduction–adduction measurements allows the measurements of the axes to each other at a neutral position to be made. The metacarpophalangeal abduction–adduction axis passes through the volar plate of the proximal phalanx. Source: Hollister A, Giurintano DJ, Buford WL, et al. The axes of rotation of the thumb interphalangeal and metacarpophalangeal joints. Clinical Orthopaedics and Related Research 320:188, 1995.

20.8 Summary It is important to understand the biomechanics of joint-articulating surface motion. The specific characteristics of the joint will determine the musculoskeletal function of that joint. The unique geometry of the joint surfaces and the surrounding capsule ligamentous constraints will guide the unique characteristics of the articulating surface motion. The range of joint motion, the stability of the joint, and the ultimate functional strength of the joint will depend on these specific characteristics. A congruent joint usually has a relatively limited range of motion but a high degree of stability, whereas a less congruent joint will have a relatively larger range of motion but less degree of stability. The characteristics of the joint-articulating surface will determine the pattern of joint contact and the axes of rotation. These © 2000 by CRC Press LLC

TABLE 20.21 Location of Center of Rotation of Trapeziometacarpal Joint Mean ± SD (mm) Circumduction X Y Z Flexion/Extension (in x-y plane) X Centroid Radius Y Centroid Radius Abduction/Adduction (in x-z plane) X Centroid Radius Z Centroid Radius

0.1 ± 1.3 –0.6 ± 1.3 –0.5 ± 1.4

–4.2 ± 1.0 2.0 ± 0.5 –0.4 ± 0.9 1.6 ± 0.5

6.7 ± 1.7 4.6 ± 3.1 –0.2 ± 0.7 1.7 ± 0.5

Note: The coordinate system is defined with the x axis corresponding to internal/external rotation, the y axis corresponding to abduction/adduction, and the z axis corresponding to flexion/extension. The x axis is positive in the distal direction, the y axis is positive in the dorsal direction for the left hand and in the palmar direction for the right hand, and the z axis is positive in the radial direction. The origin of the coordinate system was at the intersection of a line connecting the radial and ulnar prominences and a line connecting the volar and dorsal tubercles. Source: Imaeda T, Niebur G, Cooney WP, Linscheid RL, An KN. 1994. Kinematics of the normal trapeziometacarpal joint. J Orthop Res 12:197.


TABLE 20.22 Measurement of Axis Location and Values for Axis Position in the Bonea Interphalangeal joint flexion-extension axis (Fig. 20.33A) t/T 44 ± 17% l/L 90 ± 5% Θ 5 ± 2° β 83 ± 4° Metacarpophalangeal joint flexion-extension axis (Fig. 20.33B) t/T 57 ± 17% l/L 87 ± 5% α 101 ± 6° β 5 ± 2° Metacarpophalangeal joint abduction-adduction axis (Fig. 20.33C) t/T 45 ± 8% l/L 83 ± 13% α 80 ± 9° β 74 ± 8° M a The angle of the abduction-adduction axis with respect to the flexion-extension axis is 84.8° ± 12.2°. The location and angulation of the K-wires of the axes with respect to the bones were measured (Θ, α, β) directly with a goniometer. The positions of the pins in the bones were measured (T, L) with a Vernier caliper. Source: Hollister A, Giurintano DJ, Buford WL, et al. The axes of rotation of the thumb interphalangeal and metacarpophalangeal joints. Clinical Orthopaedics and Related Research, 320:188, 1995.

characteristics will regulate the stresses on the joint surface which will influence the degree of degeneration of articular cartilage in an anatomic joint and the amount of wear of an artificial joint.

Acknowledgment The authors thank Barbara Iverson-Literski for her careful preparation of the manuscript.

References Ahmed AM, Burke DL, Hyder A. 1987. Force analysis of the patellar mechanism. J Orthop Res 5:1, 69. Allard P. Duhaime M, Labelle H, et al. 1987. Spatial reconstruction technique and kinematic modeling of the ankle. IEEE Engr Med Biol 6:31. An KN, Cooney WP. 1991. Biomechanics, Section II. The hand and wrist. In BF Morrey (ed), Joint Replacement Arthroplasty, pp 137–146, New York, Churchill Livingstone. Anderson DD, Bell AL, Gaffney MB, et al. 1996. Contact stress distributions in malreduced intra-articular distal radius fractures. J Orthop Trauma 10:331. Andriacchi TP, Stanwyck TS, Galante JO. 1986. Knee biomechanics in total knee replacement. J Arthroplasty 1(3):211. Ateshian GA, Ark JW, Rosenwasser MD, et al. 1995. Contact areas in the thumb carpometacarpal joint. J Orthop Res 13:450. Athesian JA, Rosenwasser MP, Mow VC. 1992. Curvature characteristics and congruence of the thumb carpometacarpal joint: differences between female and male joints. J Biomech 25(6):591. Barnett CH, Napier JR. 1952. The axis of rotation at the ankle joint in man. Its influence upon the form of the talus and the mobility of the fibula. J Anat 86:1. Basmajian JV, Bazant FJ. 1959. Factors preventing downward dislocation of the adducted shoulder joint. An electromyographic and morphological study. J Bone Joint Surg 41A:1182. © 2000 by CRC Press LLC

Boileau P, Walch G. 1997. The three-dimensional geometry of the proximal humerus. J Bone Joint Surg 79B:857. D’Ambrosia RD, Shoji H, Van Meter J. 1976. Rotational axis of the ankle joint: Comparison of normal and pathological states. Surg Forum 27:507. Davy DT, Kotzar DM, Brown RH, et al. 1989. Telemetric force measurements across the hip after total arthroplasty. J Bone Joint Surg 70A(1):45. Doody SG, Freedman L, Waterland JC. 1970. Shoulder movements during abduction in the scapular plane. Arch Phys Med Rehabil 51:595. Draganich LF, Andriacchi TP, Andersson GBJ. 1987. Interaction between intrinsic knee mechanics and the knee extensor mechanism. J Orthop Res 5:539. Driscoll HL, Christensen JC, Tencer AF. 1994. Contact characteristics of the ankle joint. J Am Podiatr Med Assoc 84(10):491. Elftman H. 1945. The orientation of the joints of the lower extremity. Bull Hosp Joint Dis 6:139. Engsberg JR. 1987. A biomechanical analysis of the talocalcaneal joint in vitro. J Biomech 20:429. Farahmand F, Senavongse W, Amis AA. 1998. Quantitative study of the quadriceps muscles and trochlear groove geometry related to instability of the patellofemoral joint. J Orthop Res 16(1):136. Fioretti S. 1994. Three-dimensional in-vivo kinematic analysis of finger movement. In F Schuind et al. (eds), Advances in the Biomechanics of the Hand and Wrist, pp 363–375, New York, Plenum. Freedman L, Munro RR. 1966. Abduction of the arm in the scapular plane: scapular and glenohumeral movements. A roentgenographic study. J Bone Joint Surg 48A:1503. Goel VK, Singh D, Bijlani V. 1982. Contact areas in human elbow joints. J Biomech Eng 104:169–175. Goodfellow J, Hungerford DS, Zindel M. 1976. Patellofemoral joint mechanics and pathology. J Bone Joint Surg 58B(3):287. Hamrick MW. 1996. Articular size and curvature as detriments of carpal joint mobility and stability in strepsirhine primates. J Morphol 230:113. Hartford JM, Gorczyca JT, McNamara JL, et al. 1985. Tibiotalar contact area. Clin Orthop 320, 82. Hicks JH. 1953. The mechanics of the foot. The joints. J Anat 87:345–357. Hollister A, Buford WL, Myers LM, et al. 1992. The axes of rotation of the thumb carpometacarpal joint. J Orthop Res 10:454. Hollister A, Guirintano DJ, Bulford WL, et al. 1995. The axes of rotation of the thumb interphalangeal and metacarpophalangeal joints. Clin Orthop 320:188. Horii E, Garcia-Elias M, An KN, et al. 1990. Effect of force transmission across the carpus in procedures used to treat Kienböck’s Disease. J Bone Joint Surg 15A(3):393. Huberti HH, Hayes WC. 1984. Patellofemoral contact pressures: The influence of Q-angle and tendofemoral contact. J Bone Joint Surg 66A(5):715. Iannotti JP, Gabriel JP, Schneck SL, et al. 1992. The normal glenohumeral relationships: an anatomical study of 140 shoulders. J Bone Joint Surg 74A(4):491. Imaeda T, Niebur G, Cooney WP, et al. 1994. Kinematics of the normal trapeziometacarpal joint. J Orthop Res. 12:197. Inman VT, Saunders JB deCM, Abbott LC. 1944. Observations on the function of the shoulder joint. J Bone Joint Surg 26A:1. Inman VT. 1976. The Joints of the Ankle, Baltimore, Williams and Wilkins. Inman VT, Mann RA. 1979. Biomechanics of the foot and ankle. In VT Inman (ed), DuVrie’s Surgery of the Foot, St Louis, Mosby. Iseki F, Tomatsu T. 1976. The biomechanics of the knee joint with special reference to the contact area. Keio J Med 25:37. Isman RE, Inman VT. 1969. Anthropometric studies of the human foot and ankle. Pros Res 10–11:97. Jobe CM, Iannotti JP. 1995. Limits imposed on glenohumeral motion by joint geometry. J Shoulder Elbow Surg 4:281. Kapandji IA. 1987. The Physiology of the Joints, vol 2, Lower Limb, Edinburgh, Churchill-Livingstone, Edinburgh. © 2000 by CRC Press LLC

Karduna AR, Williams GR, Williams JI, et al. 1996. Kinematics of the glenohumeral joint: Influences of muscle forces, ligamentous constraints, and articular geometry. J Orthop Res 14:986. Kent BE. 1971. Functional anatomy of the shoulder complex. A review. Phys Ther 51:867. Kimizuka M, Kurosawa H, Fukubayashi T. 1980. Load-bearing pattern of the ankle joint. Contact area and pressure distribution. Arch Orthop Trauma Surg 96:45–49. Kinzel GL, Hall AL, Hillberry BM. 1972. Measurement of the total motion between two body segments: Part I. Analytic development. J Biomech 5:93. Kirby KA. 1947. Methods for determination of positional variations in the subtalar and transverse tarsal joints. Anat Rec 80:397. Kobayashi M, Berger RA, Nagy L, et al. 1997. Normal kinematics of carpal bones: A three-dimensional analysis of carpal bone motion relative to the radius. J Biomech 30:787. Kurosawa H, Walker PS, Abe S, et al. 1985. Geometry and motion of the knee for implant and orthotic design. J Biomech 18(7):487. Lewis JL, Lew WD. 1978. A method for locating an optimal “fixed” axis of rotation for the human knee joint. J Biomech Eng 100:187. Libotte M, Klein P, Colpaert H, et al. 1982. Contribution à l’étude biomécanique de la pince malléolaire. Rev Chir Orthop 68:299. Lippitts B, Vanderhooft JE, Harris SL, et al. 1993. Glenohumeral stability from concavity-compression: A quantitative analysis. J Shoulder Elbow Surg 2:27. Macko VW, Matthews LS, Zwirkoski P, et al. 1991. The joint contract area of the ankle: the contribution of the posterior malleoli. J Bone Joint Surg 73A(3):347. Manter JT. 1941. Movements of the subtalar and transverse tarsal joints. Anat Rec 80:397–402. Maquet PG, Vandberg AJ, Simonet JC. 1975. Femorotibial weight bearing areas: Experimental determination. J Bone Joint Surg 57A(6):766–771. McPherson EJ, Friedman RJ, An YH, et al. 1997. Anthropometric study of normal glenohumeral relationships. J Shoulder Elbow Surg 6:105. Merz B, Eckstein F, Hillebrand S, et al: 1997. Mechanical implication of humero-ulnar incongruity-finite element analysis and experiment. J Biomech 30:713. Miyanaga Y, Fukubayashi T, Kurosawa H. 1984. Contact study of the hip joint: Load deformation pattern, contact area, and contact pressure. Arch Orth Trauma Surg 103:13–17. Morrey BF, An KN. 1990. Biomechanics of the Shoulder. In CA Rockwood and FA Matsen (eds), The Shoulder, pp 208–245, Philadelphia, Saunders. Morrey BF, Chao EYS. 1976. Passive motion of the elbow joint: a biomechanical analysis. J Bone Joint Surg 58A(4):501. Nahass BE, Madson MM, Walker PS. 1991. Motion of the knee after condylar resurfacing—an in vivo study. J Biomech 24(12):1107. Paar O, Rieck B, Bernett P. 1983. Experimentelle untersuchungen über belastungsabhängige Drukund Kontaktflächenverläufe an den Fussgelenken. Unfallheilkunde 85:531. Pagowski S, Piekarski K. 1977. Biomechanics of metacarpophalangeal joint. J Biomech 10:205. Palmer AK, Werner FW. 1984. Biomechanics of the distal radio-ulnar joint. Clin Orthop 187:26. Parlasca R, Shoji H, D’Ambrosia RD. 1979. Effects of ligamentous injury on ankle and subtalar joints. A kinematic study. Clin Orthop 140:266. Pellegrini VS, Olcott VW, Hollenberg C. 1993. Contact patterns in the trapeziometacarpal joint: The role of the palmar beak ligament. J Hand Surg 18A:238. Pereira DS, Koval KJ, Resnick RB, et al. 1996. Tibiotalar contact area and pressure distribution: The effect of mortise widening and syndesmosis fixation. Foot Ankle 17(5):269. Poppen NK, Walker PS. 1976. Normal and abnormal motion of the shoulder. J Bone Joint Surg 58A:195. Ramsey PL, Hamilton W. 1976. Changes in tibiotalar area of contact caused by lateral talar shift. J Bone Joint Surg 58A:356.


Rastegar J, Miller N, Barmada R. 1980. An apparatus for measuring the load-displacement and loaddependent kinematic characteristics of articulating joints—application to the human ankle. J Biomech Eng 102:208. Root ML, Weed JH, Sgarlato TE, Bluth DR. 1966. Axis of motion of the subtalar joint. J Am Podiatry Assoc 56:149. Saha AK. 1971. Dynamic stability of the glenohumeral joint. Acta Orthop Scand 42:491. Sammarco GJ, Burstein AJ, Frankel VH. 1973. Biomechanics of the ankle: A kinematic study. Orthop Clin North Am 4:75–96. Sato S. 1995. Load transmission through the wrist joint: A biomechanical study comparing the normal and pathological wrist. Nippon Seikeigeka Gakkai Zasshi-J. of the Japanese Orthopedics Association 69:470. Schuind FA, Linscheid RL, An KN, et al. 1992. Changes in wrist and forearm configuration with grasp and isometric contraction of elbow flexors. J Hand Surg 17A:698. Shephard E. 1951. Tarsal movements. J Bone Joint Surg 33B:258. Shiba R, Sorbie C, Siu DW, et al. 1988. Geometry of the humeral-ulnar joint. J Orthop Res 6:897. Singh AK, Starkweather KD, Hollister AM, et al. 1992. Kinematics of the ankle: A hinge axis model. Foot and Ankle 13(8):439. Soslowsky LJ, Flatow EL, Bigliani LU, et al. 1992. Quantitation of in situ contact areas at the glenohumeral joint: A biomechanical study. J Orthop Res 10(4):524. Spoor CW, Veldpaus FE. 1980. Rigid body motion calculated from spatial coordinates of markers. J Biomech 13:391. Stormont TJ, An KA, Morrey BF, et al. 1985. Elbow joint contact study: comparison of techniques. J Biomech 18(5):329. Tamai K, Ryu J, An KN, et al. 1988. Three-dimensional geometric analysis of the metacarpophalangeal joint. J Hand Surg 13A(4):521. Tanaka S, An KN, Morrey BF. 1998. Kinematics and laxity of ulnohumeral joint under valgus-varus stress. J Musculoskeletal Res 2:45. Tönnis D. 1987. Congenital Dysplasia and Dislocation of the Hip and Shoulder in Adults, pp 1–12, Berlin, Springer-Verlag. Van Eijden TMGJ, Kouwenhoven E, Verburg J, et al. 1986. A mathematical model of the patellofemoral joint. J Biomech 19(3):219. Van Langelaan EJ. 1983. A kinematical analysis of the tarsal joints. An x-ray photogrammetric study. Acta Orthop Scand [Suppl] 204:211. Viegas SF, Tencer AF, Cantrell J, et al. 1987. Load transfer characteristics of the wrist: Part I. The normal joint. J Hand Surg 12A(6):971. Viegas SF, Patterson RM, Peterson PD, et al. 1989. The effects of various load paths and different loads on the load transfer characteristics of the wrist. J Hand Surg 14A(3):458. Viegas SF, Patterson RM, Todd P, et al. October 7, 1990. Load transfer characteristics of the midcarpal joint. Presented at Wrist Biomechanics Symposium, Wrist Biomechanics Workshop, Mayo Clinic, Rochester, MN. Von Lanz D, Wauchsmuth W. 1938. Das Hüftgelenk, Praktische Anatomie I Bd, pp 138–175, Teil 4: Bein und Statik, Berlin, Springer. Wagner UA, Sangeorzan BJ, Harrington RM, et al. 1992. Contact characteristics of the subtalar joint: load distribution between the anterior and posterior facets. J Orthop Res 10:535. Walker PS, Erhman MJ. 1975. Laboratory evaluation of a metaplastic type of metacarpophalangeal joint prosthesis. Clin Orthop 112:349. Wang C-L, Cheng C-K, Chen C-W, et al. 1994. Contact areas and pressure distributions in the subtalar joint. J Biomech 28(3):269. Woltring HJ, Huiskes R, deLange A, Veldpaus FE. 1985. Finite centroid and helical axis estimation from noisy landmark measurements in the study of human joint kinematics. J Biomech 18:379. Yoshioka Y, Siu D, Cooke TDV. 1987. The anatomy and functional axes of the femur. J Bone Joint Surg 69A(6):873. © 2000 by CRC Press LLC

Yoshioka Y, Siu D, Scudamore RA, et al. 1989. Tibial anatomy in functional axes. J Orthop Res 7:132. Youm Y, McMurty RY, Flatt AE, et al. 1978. Kinematics of the wrist: an experimental study of radioulnar deviation and flexion/extension. J Bone Joint Surg 60A(4):423. Zuckerman JD, Matsen FA: Biomechanics of the elbow. 1989. In M Nordine & VH Frankel (eds), Basic Biomechanics of the Musculoskeletal System, pp 249–260, Philadelphia, Lea & Febiger.


Furey, M. J. “Joint Lubrication.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

21 Joint Lubrication 21.1 21.2

Introduction Tribology

21.3

Lubrication

Friction • Wear and Surface Damage Hydrodynamic Lubrication Theories • Transition from Hydrodynamic to Boundary Lubrication

21.4 21.5 21.6 21.7 21.8

Terms and Definitions • Experimental Contact Systems • Fluids and Materials Used as Lubricants in In Vivo Studies • The Preoccupation with Rheology and Friction • The Probable Existence of Various Lubrication Regimes • Recent Developments

Michael J. Furey Mechanical Engineering Department and Center for Biomedical Engineering, Virginia Polytechnic Institute and State University

Synovial Joints Theories on the Lubrication of Natural and Normal Synovial Joints In Vitro Cartilage Wear Studies Biotribology and Arthritis: Are There Connections? Recapitulation and Final Comments

21.9

Conclusions

“The Fabric of the Joints in the Human Body is a subject so much the more entertaining, as it must strike every one that considers it attentively with an Idea of fine Mechanical Composition. Wherever the Motion of one Bone upon another is requisite, there we find an excellent Apparatus for rendering that Motion safe and free: We see, for Instance, the Extremity of one Bone molded into an orbicular Cavity, to receive the Head of another, in order to afford it an extensive Play. Both are covered with a smooth elastic Crust, to prevent mutual Abrasion; connected with strong Ligaments, to prevent Dislocation; and inclosed in a Bag that contains a proper Fluid Deposited there, for lubricating the Two contiguous Surfaces. So much in general.” The above is the opening paragraph of the classic paragraph of the classic paper by the surgeon, Sir William Hunter, “Of the Structure and Diseases of Articulating Cartilages” which he read to a meeting of the Royal Society, June 2, 1743 [1]. Since then, a great deal of research has been carried out on the subject of synovial joint lubrication. However, the mechanisms involved are still unknown.

21.1

Introduction

The purpose of this article is twofold: (1) to introduce the reader to the subject of tribology—the study of friction, wear, and lubrication; and (2) to extend this to the topic of biotribology, which includes the lubrication of natural synovial joints. It is not meant to be an exhaustive review of joint lubrication theories; space does not permit this. Instead, major concepts or principles will be discussed not only in


the light of what is known about synovial joint lubrication but perhaps more importantly what is not known. Several references are given for those who wish to learn more about the topic. It is clear that synovial joints are by far the most complex and sophisticated tribological systems that exist. We shall see that although numerous theories have been put forth to attempt to explain joint lubrication, the mechanisms involved are still far from being understood. And when one begins to examine possible connections between tribology and degenerative joint disease or osteoarthritis, the picture is even more complex and controversial. Finally, this article does not treat the (1) tribological behavior of artificial joints or partial joint replacements, (2) the possible use of elastic or poroplastic materials as artificial cartilage, and (3) new developments in cartilage repair using transplanted chondrocytes. These are separate topics, which would require detailed discussion and additional space.

21.2

Tribology

The word tribology, derived from the Greek “to rub,” covers all frictional processes between solid bodies moving relative to one another that are in contact [2]. Thus tribology may be defined as the study of friction, wear, and lubrication. Tribological processes are involved whenever one solid slides or rolls against another, as in bearings, cams, gears, piston rings and cylinders, machining and metalworking, grinding, rock drilling, sliding electrical contacts, frictional welding, brakes, the striking of a match, music from a cello, articulation of human synovial joints (e.g., hip joints), machinery, and in numerous less obvious processes (e.g., walking, holding, stopping, writing, and the use of fasteners such as nails, screws, and bolts). Tribology is a multidisciplinary subject involving at least the areas of materials science, solid and surface mechanics, surface science and chemistry, rheology, engineering, mathematics, and even biology and biochemistry. Although tribology is still an emerging science, interest in the phenomena of friction, wear, and lubrication is an ancient one. Unlike thermodynamics, there are no generally accepted laws in tribology. But there are some important basic principles needed to understand any study of lubrication and wear and even more so in a study of biotribology or biological lubrication phenomena. These basic principles follow.

Friction Much of the early work in tribology was in the area of friction—possibly because frictional effects are more readily demonstrated and measured. Generally, early theories of friction dealt with dry or unlubricated systems. The problem was often treated strictly from a mechanical viewpoint, with little or no regard for the environment, surface films, or chemistry. In the first place, friction may be defined as the tangential resistance that is offered to the sliding of one solid body over another. Friction is the result of many factors and cannot be treated as something as singular as density or even viscosity. Postulated sources of friction have included (1) the lifting of one asperity over another (increase in potential energy), (2) the interlocking of asperities followed by shear, (3) interlocking followed by plastic deformation or plowing, (4) adhesion followed by shear, (5) elastic hysteresis and waves of deformation, (6) adhesion or interlocking followed by tensile failure, (7) intermolecular attraction, (8) electrostatic effects, and (9) viscous drag. The coefficient of friction, indicated in the literature by µ or f, is defined as the ratio F/W where F = friction force and W = the normal load. It is emphasized that friction is a force and not a property of a solid material or lubricant.

Wear and Surface Damage One definition of wear in a tribological sense is that it is the progressive loss of substance from the operating surface of a body as a result of relative motion at the surface. In comparison with friction, very little theoretical work has been done on the extremely important area of wear and surface damage. This is not too surprising in view of the complexity of wear and how little is known of the mechanisms by which it can occur. Variations in wear can be, and often are, enormous compared with variations in friction. For example,


practically all the coefficients of sliding friction for diverse dry or lubricated systems fall within a relatively narrow range of 0.1 to 1. In some cases (e.g., certain regimes of hydrodynamic or “boundary” lubrication), the coefficient of friction may be 2 × 104 g/mol) and have excellent mechanical properties. More importantly, they display an excellent resistance to most chemicals and to water over wide temperature ranges. Polysulfones were developed by Union Carbide in the 1960s. These polymers have a high thermal stability due to the bulky side groups (therefore, they are amorphous) and rigid main backbone chains. They are also highly stable to most chemicals but are not so stable in the presence of polar organic solvents such as ketones and chlorinated hydrocarbons. Polycarbonates are tough, amorphous, and transparent polymers made by reacting bisphenol A and diphenyl carbonate. It is noted for its excellent mechanical and thermal properties (high Tg:150°C), hydrophobicity, and antioxidative properties.

Biodegradable Polymers Recently, several biodegradable polymers such as polylactide (PLA), polyglycolide (PGA), poly(glycolide-colactide) (PLGA), poly(dioxanone), poly(trimethylene carbonate), poly(carbonate), and so on are extensively © 2000 by CRC Press LLC

used or tested on a wide range of medical applications due to their good biocompatibility, controllable biodegradability, and relatively good processability [Khang et al., 1997]. PLA, PGA, and PLGA are bioresorbable polyesters belonging to the group of poly α-hydroxy acids. These polymers degrade by nonspecific hydrolytic scission of their ester bonds. The hydrolysis of PLA yields lactic acid which is a normal byproduct of anaerobic metabolism in the human body and is incorporated in the tricarboxylic acid (TCA) cycle to be finally excreted by the body as carbon dioxide and water. PGA biodegrades by a combination of hydrolytic scission and enzymatic (esterase) action producing glycolic acid which can either enter the TCA cycle or is excreted in urine and can be eliminated as carbon dioxide and water. The degradation time of PLGA can be controlled from weeks to over a year by varying the ratio of monomers and the processing conditions. It might be a suitable biomaterial for use in tissue engineered repair systems in which cells are implanted within PLGA films or scaffolds and in drug delivery systems in which drugs are loaded within PLGA microspheres. PGA (Tm: 225–230°C, Tg: 35–40°C) can be melt spun into fibers which can be converted into bioresorbable sutures, meshes, and surgical products. PLA (Tm: 173–178 oC, Tg: 60–65°C) exhibit high tensile strength and low elongation resulting in a high modulus suitable for load-bearing applications such as in bone fracture fixation. Poly-p-dioxanone (Tm: 107–112°C, Tg: ~10°C) is a bioabsorbable polymer which can be fabricated into flexible monofilament surgical sutures.

39.4 Sterilization Sterilizability of biomedical polymers is an important aspect of the properties because polymers have lower thermal and chemical stability than other materials such as ceramics and metals, consequently, they are also more difficult to sterilize using conventional techniques. Commonly used sterilization techniques are dry heat, autoclaving, radiation, and ethylene oxide gas [Block,1977]. In dry heat sterilization, the temperature varies between 160 and 190°C. This is above the melting and softening temperatures of many linear polymers like polyethylene and PMMA. In the case of polyamide (Nylon), oxidation will occur at the dry sterilization temperature although this is below its melting temperature. The only polymers which can safely be dry sterilized are PTFE and silicone rubber. Steam sterilization (autoclaving) is performed under high steam pressure at relatively low temperature (125–130°C). However, if the polymer is subjected to attack by water vapor, this method cannot be employed. PVC, polyacetals, PE (low-density variety), and polyamides belong to this category. Chemical agents such as ethylene and propylene oxide gases [Glaser,1979], and phenolic and hypochloride solutions are widely used for sterilizing polymers since they can be used at low temperatures. Chemical agents sometimes cause polymer deterioration even when sterilization takes place at room temperature. However, the time of exposure is relatively short (overnight), and most polymeric implants can be sterilized with this method. Radiation sterilization [Sato, 1983] using the isotopic 60Co can also deteriorate polymers since at high dosage the polymer chains can be dissociated or cross-linked according to the characteristics of the chemical structures, as shown in Table 39.6. In the case of PE, at high dosage (above 106 Gy) it becomes a brittle and hard material. This is due to a combination of random chain scission cross-linking. PP articles will often discolor during irradiation giving the product an undesirable color tint but the more severe problem is the embrittlement resulting in flange breakage, luer cracking, and tip breakage. The physical properties continue to deteriorate with time, following irradiation. These problems of coloration and changing physical properties are best resolved by avoiding the use of any additives which discolor at the sterilizing dose of radiation [Khang, 1996c].

39.5 Surface Modifications for Improving Biocompatability Prevention of thrombus formation is important in clinical applications where blood is in contact such as hemodialysis membranes and tubes, artificial heart and heart-lung machines, prosthetic valves, and © 2000 by CRC Press LLC

TABLE 39.6 Effect of Gamma Irradiation on Polymers Which Could be Cross-Linked or Degraded Cross-Linking Polymers Polyethylene Polypropylene Polystyrene Polyarylates Polyacrylamide Polyvinylchloride Polyamides Polyesters Polyvinylpyrrolidone Polymethacrylamide Rubbers Polysiloxanes Polyvinylalcohol Polyacroleine

Degradable Polymers Polyisobutylene Poly-α−methylstyrene Polymethylmetacrylate Polymethacrylamide Polyvinylidenechloride Cellulose and derivatives Polytetrafluoroethylene Polytrifluorochloroethylene

artificial vascular grafts. In spite of the use of anticoagulants, considerable platelet deposition and thrombus formation take place on the artificial surfaces [Branger, 1990]. Heparin, one of the complex carbohydrates known as mucopolysaccharides or glycosaminoglycan is currently used to prevent formation of clots. In general, heparin is well tolerated and devoid of serious consequences. However, it allows platelet adhesion to foreign surfaces and may cause hemorrhagic complications such as subdural hematoma, retroperitoneal hematoma, gastrointestinal bleeding, hemorrage into joints, ocular and retinal bleeding, and bleeding at surgical sites [Lazarus, 1980]. These difficulties give rise to an interest in developing new methods of hemocompatible materials. Many different groups have studied immobilization of heparin [Kim and Feijen, 1985; Park et al., 1988] on the polymeric surfaces, heparin analogues and heparin-prostaglandin or heparin-fibrinolytic enzyme conjugates [Jozefowicz and Jozefowicz, 1985]. The major drawback of these surfaces is that they are not stable in the blood environment. It has not been firmly established that a slow leakage of heparin is needed for it to be effective as an immobilized antithrombogenic agent, if not its effectiveness could be hindered by being “coated over” with an adsorbed layer of more common proteins such as albumin and fibrinogen. Fibrinolytic enzymes, urokinase, and various prostaglandins have also been immobilized by themselves in order to take advantage of their unique fibrin dissolution or antiplatelet aggregation actions [Ohshiro, 1983]. Albumin-coated surfaces have been studied because surfaces that resisted platelet adhesion in vitro were noted to adsorb albumin preferentially [Keogh et at., 1992]. Fibronectin coatings have been used in in vitro endothelial cell seeding to prepare a surface similar to the natural blood vessel lumen [Lee et al., 1989]. Also, algin-coated surfaces have been studied due to their good biocompatibility and biodegradability [Lee et al., 1990b; Lee et al., 1997b]. Recently, plasma gas discharge [Khang et al., 1997a] and corona treatment [Khang et al., 1996d] with reactive groups introduced on the polymeric surfaces have emerged as other ways to modify biomaterial surfaces [Lee et al., 1991; Lee et al., 1992]. Hydrophobic coatings composed of silicon- and fluorine-containing polymeric materials as well as polyurethanes have been studied because of the relatively good clinical performances of Silastic®, Teflon®, and polyurethane polymers in cardiovascular implants and devices. Polymeric fluorocarbon coatings deposited from a tetrafluoroethylene gas discharge have been found to greatly enhance resistance to both acute thrombotic occlusion and embolization in small diameter Dacron® grafts. Hydrophilic coatings have also been popular because of their low interfacial tension in biological environments [Hoffman, 1981]. Hydrogels as well as various combinations of hydrophilic and hydrophobic monomers have been studied on the premise that there will be an optimum polar-dispersion force ratio which could be matched on the surfaces of the most passivating proteins. The passive surface © 2000 by CRC Press LLC

TABLE 39.7

Physical and Chemical Surface Modification Methods for Polymeric Biomaterials

To modify blood compatibility

To influence cell adhesion and growth

To control protein adsorption

To improve lubricity

To improve wear resistance and corrosion resistance To alter transport properties To modify electrical characteristics

Octadecyl group attachment to surface Silicon containing block copolymer additive Plasma fluoropolymer deposition Plasma siloxane polymer deposition Radiation-grafted hydrogels Chemically modified polystyrene for heparin-like activity Oxidized polystyrene surface Ammonia plasma-treated surface Plasma-deposited acetone or methanol film Plasma fluoropolymer deposition Surface with immobilized polyethyelenglycol Treated ELISA dish surface Affinity chromatography particulates Surface cross-linked contact lens Plasma treatment Radiation-grafted hydrogels Interpenetrating polymeric networks Ion implantation Diamond deposition Anodization Plasma deposition (methane, fluoropolymer, siloxane) Plasma deposition Solvent coatings Parylene coatings

Source: Ratner et al., [1996], p106.

may induce less clot formation. Polyethylene oxide coated surfaces have been found to resist protein adsorption and cell adhesion and have therefore been proposed as potential “blood compatible” coatings [Lee et al., 1990a]. General physical and chemical methods to modify the surfaces of polymeric biomaterials are listed in Table 39.7 [Ratner et al., 1996]. Another way of making antithrombogenic surfaces is the saline perfusion method, which is designed to prevent direct contacts between blood and the surface of biomaterials by means of perfusing saline solution through the porous wall which is in contact with blood [Park and Kim, 1993; Khang et al., 1996a; Khang et al., 1996b]. It has been demonstrated that the adhesion of the blood cells could be prevented by the saline perfusion through PE, alumina, sulfonated/nonsulfonated PS/SBR, ePTFE (expanded polytetrafluoroethylene), and polysulfone porous tubes.

39.6 Chemogradient Surfaces for Cell and Protein Interaction The behavior of the adsorption and desorption of blood proteins or adhesion and proliferation of different types of mammalian cells on polymeric materials depends on the surface characteristics such as wettability, hydrophilicity/hydrophobicity ratio, bulk chemistry, surface charge and charge distribution, surface roughness, and rigidity. Many research groups have studied the effect of the surface wettability on the interactions of biological species with polymeric materials. Some have studied the interactions of different types of cultured cells or blood proteins with various polymers with different wettabilities to correlate the surface wettability and blood- or tissue-compatibility [Baier et al., 1984]. One problem encountered from the study using different kinds of polymers is that the surfaces are heterogeneous, both chemically and physically (different surface chemistry, roughness, rigidity, crystallinity, etc.), which caused widely varying results. Some others have studied the interactions of different types of cells or proteins with a range of methacrylate copolymers with different wettabilities and have the same kind of chemistry but are still physically


FIGURE 39.6 Schematic diagram showing corona discharge apparatus for the preparation of wettability chemogradient surfaces.

heterogeneous [van Wachem et al., 1987]. Another methodological problem is that such studies are often tedious, laborious, and time-consuming because a large number of samples must be prepared to characterize the complete range of the desired surface properties. Many studies have been focused on the preparation of surfaces whose properties are changed gradually along the material length. Such chemogradient surfaces are of particular interest in basic studies of the interactions between biological species and synthetic materials surfaces since the affect of a selected property can be examined in a single experiment on one surface preparation. A chemogradient of methyl groups was formed by diffusion of dimethyldichlorosilane through xylene on flat hydrophilic siliconedioxide surfaces [Elwing et al., 1989]. The wettability chemogradient surfaces were made to investigate hydrophilicity-induced changes of adsorbed proteins. Recently, a method for preparing wettability chemogradients on various polymer surfaces was developed [Lee et al., 1989 and 1990; Khang et al., 1997b]. The wettability chemogradients were produced via radio frequency (RF) and plasma discharge treatment by exposing the polymer sheets continuously to the plasma [Lee et al., 1991]. The polymer surfaces oxidized gradually along the sample length with increasing plasma exposure time and thus the wettability chemogradient was created. Another method for preparing a wettability chemogradient on polymer surfaces using corona discharge treatment has been developed as shown in Fig. 39.6 [Lee et al., 1992]. The wettability chemogradient was produced by treating the polymer sheets with corona from a knife-type electrode whose power was gradually changed along the sample length. The polymer surface gradually oxidized with the increasing power and the wettability chemogradient was created. Chemogradient surfaces with different functional gruops such as –COOH, –CH2OH, –CONH2, and –CH2NH2 were produced on PE surfaces by the above corona treatment followed by vinyl monomer grafting and substitution reactions [Kim et al., 1993; Lee et al.,1994a; Lee et al.,1994b]. We have also prepared chargeable functional groups [Lee et al., 1997c; Lee et al., 1997d; Lee et al., 1998a], comb-like polyethyleneoxide (PEO) [Jeong et al., 1996; Lee et al., 1997a] and phospholipid polymer chemogradient surfaces [Iwasaki et al., 1997] by the corona discharge treatment, followed by the graft copolymerization with subsequent substitution reaction of functional vinyl monomers as acrylic acid, sodium p-sulfonic styrene and N, N-dimethyl aminopropyl acrylamide, poly(ethyleneglycol) monomethacrylate, and ω-methacryloyloxyalkyl phosphorylcholine (MAPC), respectively. The water contact angles of the corona-treated PE surfaces gradually decrease along the sample length with increasing corona power (from about 95° to about 45°) as shown in Fig. 39.7. The decrease in contact angles, i.e., the increase in wettability along the sample length was due to the oxygen-based polar functionalities incorporated on the surface by the corona treatment. It was also confirmed also by fouriertransform infrared spectroscopy in the attenuated total reflectance mode and electron spectroscopy for chemical analysis (ESCA).


FIGURE 39.7 bers, n = 3.

Changes in water contact angle of corona-treated PE surface along the sample length. Sample num-

In order to investigate the interaction of different types of cells in terms of the surface hydrophilicity/hydrophobicity of polymeric materials, Chinese hamster ovaries (CHO), fibroblasts, and bovine aortic endothelial cells (EC) were cultured for 1 and 2 days on the PE wettability chemogradient surfaces. The maximum adhesion and growth of the cells appeared around a water contact angle of 50–55° as shown in Fig. 39.8. The observation of scanning electron microscopy (SEM) also verified that the cells are more adhered, spread, and grown onto the sections with moderate hydrophilicity as shown in Fig. 39.9.

FIGURE 39.8 CHO, fibroblast, and endothelial cell growth on wettability chemogradient PE surfaces after 2 days culture (number of seeded cells, 4 × 104/cm2). n = 3.


FIGURE 39.9 SEM microphotographs of CHO, fibroblast, and endothelial cells grown on PE wettability chemogradient surface along the sample length after 2 days culture (original magnification; ×400).

To determine the cell proliferation rates, the migration of fibroblasts on PE wettability chemogradient surfaces were observed [Khang et al., 1998b]. After the change of culture media at 24 h, cell growth morphology was recorded for 1 or 2 hr intervals at the position of 0.5, 1.5, 2.5, and 4.5 cm for the counting of grown cells and the observation of cell morphology with a video tape recorder. The proliferation rates of fibroblast cells were calculated from the slopes of Fig. 39.10 as given in Table 39.8. The proliferation rates on the PE surfaces with wettability chemogradient showed that as the surface wettability increased, it increased and then decreased. The maximum proliferation rate of the cells as 1,111 cells/h . cm2 appeared at around the position 2.5 cm. To observe the effect of serum proteins on the cell adhesion and growth behaviors, fetal bovine serum (FBS), which contains more than 200 kinds of different proteins, was adsorbed onto the wettability gradient PE surfaces for 1 h at 37°C. Figure 39.11 shows the relative adsorbed amount of serum proteins on the wettability gradient surfaces determined by ESCA. The maximum adsorption of the proteins appeared at around the 2.5 cm position, which is the same trend as the cell adhesion, growth, and migration behaviors. It can be explained that preferential adsorption of some serum proteins, like fibronectin and vitronectin from culture medium, onto the moderately wettable surfaces may be a reason for better cell adhesion, spreading, and growth. Proteins like fibronectin and vitronectin are well known as cell-adhesive proteins. Cells attached on surfaces are spread only when they are compatible on the surfaces. It seems that surface wettability plays an important role for cell adhesion, spreading, and migration. Also investigated were: (1) platelet adhesion on wettability chemogradient [Lee and Lee, 1998b], (2) cell interaction on microgrooved PE surfaces (groove depth, 0.5 µm; groove width, 0.45 µm; and pitch, 0.9 µm) with wettability chemogradient [Khang et al., 1997c], (3) detachment of human endothelial


FIGURE 39.10

Fibroblast cell proliferation rates on wettability chemogradient PE surfaces (24 to 60 h culture).

TABLE 39.8. Proliferation Rates of Fibroblast Cells on Wettability Gradient PE Surfaces Positions (cm)

Contact Angle (°)

Cell Proliferation Rate (#cell/h.cm2)

2.5 4.5 1.5 0.5

55 45 67 85

1,111 924 838 734

Note: 24 to 60 h culture.

FIGURE 39.11


Serum protein adsorption on PE wettability chemogradient surface (1 h adsorption). n = 3.

under flow from wettability gradient surface with different functional groups [Ruardy et al., 1997], (4) cell interaction on microporous polycarbonate membrane with wettability chemogradient [Lee et al., 1998c], and (5) cell interaction on poly(lactide-co-glycolide) surface with wettability chemogradient [Khang et al., 1998a]. During the last several years, “chemogradient surfaces” have evolved into easier and more popular tools for the study of protein adsorption and platelet or cell interactions continuously which relate to the surface properties such as wettability, chemistry and charge, or dynamics of polymeric materials. In many studies, different kinds of polymeric materials with widely varying surface chemistries are used and the explanation of the results is often in controversy due to the surface heterogeneity. In addition, these studies are tedious, laborious, and time-consuming, and biological variations are more likely to occur. The application of chemogradient surfaces for these studies can reduce these discomforts and problems, and eventually save time and money. Also, chemogradient surfaces are valuable in investigating the basic mechanisms by which complicated systems such as proteins or cells interact with surfaces, since a continuum of selected and controlled physical-chemical properties can be studied in one experiment on the polymeric surface. The possible applications of chemogradient surfaces in the near future are: (1) separation devices of cells and/or biological species by different surface properties, (2) column packing materials for separation, (3) biosensoring, etc.

Acknowledgments This work was supported by grants from the Korea Ministry of Health and Welfare (grant Nos. HMP95-G-2-33 and HMP-97-E-0016) and the Korea Ministry of Science and Technology (grant No. 97-N102-05-A-02).

Defining Terms Acetabulum: The socket portion of the hip joint. Addition (or free radical) polymerization: Polymerization in which monomers are added to the growing chains, initiated by free radical agents. Biocompatibility: Acceptance of an artificial implant by the surrounding tissues and as a whole. The implant should be compatible with tissues in terms of mechanical, chemical, surface, and pharmacological properties. Biomaterials: Synthetic materials used to replace part of a living system or to function in intimate contact with living tissue. Bone cement: Mixture of polymethylmethacrylate powder and methylmethacrylate monomer liquid to be used as a grouting material for the fixation of orthopedic joint implants. Branching: Chains grown from the sides of the main backbone chains. Chemogradient surface: The surface whose properties such as wettability, surface charge, and hydrophilicity/hydrophobicity ratio are changed gradually along the material length. Condensation (step reaction) polymerization: Polymerization in which two or more chemicals are reacted to form a polymer by condensing out small molecules such as water and alcohol. Copolymers: Polymers made from two or more monomers which can be obtained by grafting, block, alternating, or random attachment of the other polymer segment. Covalent bonding: Bonding of atoms or molecules by sharing valence electrons. Dacron®: Polyethyleneterephthalate polyester that is made into fiber. If the same polymer is made into a film, it is called Mylar®. Delrin®: Polyacetal made by Union Carbide. Elastomers: Rubbery materials. The restoring force comes from uncoiling or unkinking of coiled or kinked molecular chains. They can be highly stretched. Embolus: Any foreign matter, as a blood clot or air bubble, carried in the blood stream.


Fibrinogen: A plasma protein of high molecular weight that is converted to fibrin through the action of thrombin. This material is used to make (absorbable) tissue adhesives. Filler: Materials added as a powder to a rubber to improve its mechanical properties. Free volume: The difference in volume occupied by the crystalline state (minimum) and non-crystalline state of a material for a given temperature and a pressure. Glass transition temperature: Temperature at which solidification without crystallization takes place from viscous liquid. Grafts: A transplant. Heparin: A substance found in various body tissues, especially in the liver, that prevents the clotting of blood. Hydrogel: Polymer which can absorb 30% or more of its weight in water. Hydrogen bonding: A secondary bonding through dipole interactions in which the hydrogen ion is one of the dipoles. Hydroquinone: Chemical inhibitor added to the bone cement liquid monomer to prevent accidental polymerization during storage. Initiator: Chemical used to initiate the addition polymerization by becoming a free radical which in turn reacts with a monomer. Ionic bonding: Bonding of atoms or molecules through electrostatic interaction of positive and negative ions. Kevlar®: Aromatic polyamides made by DuPont. Lexan®: Polycarbonate made by General Electric. Oxygenator: An apparatus by which oxygen is introduced into blood during circulation outside the body, as during open-heart surgery. Plasticizer: Substance made of small molecules, mixed with (amorphous) polymers to make the chains slide more easily past each other, making the polymer less rigid. Refractive index: Ratio of speed of light in vacuum to speed of light in a material. It is a measure of the ability of a material to refract (bend) a beam of light. Repeating unit: Basic molecular unit which can represent a polymer backbone chain. The average number of repeating units is called the degree of polymerization. Repeating unit: The smallest unit representing a polymer molecular chain. Semi-crystalline solid: Solid which contains both crystalline and noncrystalline regions and usually occurs in polymers due to their long chain molecules. Side group: Chemical group attached to the main backbone chain. It is usually shorter than the branches and exists before polymerization. Steric hindrance: Geometrical interference which restrains movements of molecular groups such as side chains and main chains of a polymer. Suture: Material used in closing a wound with stitches. Tacticity: Arrangement of asymmetrical side groups along the backbone chain of polymers. Groups could be distributed at random (atactic), one side (isotactic), or alternating (syndiotactic). Teflon®: Polytetrafluoroethylene made by DuPont. Thrombus: The fibrinous clot attached at the site of thrombosis. Udel®: Polysulfone made by General Electric. Valence electrons: The outermost (shell) electrons of an atom. van der Waals bonding: A secondary bonding arising through the fluctuating dipole-dipole interactions. Vinyl polymers: Thermoplastic linear polymers synthesized by free radical polymerization of vinyl monomers having a common structure of CH2CHR. Vulcanization: Cross-linking of a (natural) rubber by adding sulfur. Ziegler-Natta catalyst: Organometallic compounds which have the remarkable capacity of polymerizing a wide variety of monomers to linear and stereoregular polymers.


References Baier, R.E., Meyer, A.E., Natiella, J.R., Natiella, R.R., and Carter, J.M. 1984. Surface properties determine bioadhesive outcomes; Methods and results, J. Biomed. Mater. Res., 18:337–355 Billmeyer, F.W. Jr. 1984. Textbook of Polymer Science, 3rd ed. John Wiley & Sons, NY. Block, S.S., Ed. 1977. Disinfection, Sterilization, and Preservation, 2nd ed., Rea and Febiger, Philadelphia, PA. Bloch, B. and Hastings, G.W. 1972. Plastic Materials in Surgery, 2nd ed. C.C. Thomas, Springfield, IL. Brandrup, J. and Immergut, E.H., Ed. 1989. Polymer Handbook, 3rd ed. A Wiley-Interscience Pub., NY. Branger, B., Garreau, M., Baudin, G., and Gris, J.C. 1990. Biocompatibility of blood tubings, Int. J. Artif. Organs, 13(10):697–703. Dumitriu, S., Ed. 1993. Polymeric Biomaterials, Marcell Dekker, Inc., NY. Elwing, E., Askendal, A., and Lundstorm, I. 1989. Desorption of fibrinogen and γ-globulin from solid surfaces induced by a nonionic detergent, J. Colloid Interface Sci., 128:296–300. Glaser, Z.R. 1979. Ethylene oxide: Toxicology review and field study results of hospital use. J. Environ. Pathol. and Toxic., 2:173–208. Hoffman, A.S. 1981. Radiation processing in biomaterials: A review, Radiat. Phys. Chem., 18(1):323–340. Ikada, Y., Ed. 1989. Bioresorbable fibers for medical use, In: High Technology Fiber, Part B., Marcel Dekker, NY. Iwasaki, Y., Ishihara, K., Nakabayashi, N., Khang, G., Jeon, J.H., Lee, J.W., and Lee, H.B. 1997. Preparation of gradient surfaces grafted with phospholipid polymers and evaluation of their blood compatibility, In: Advances in Biomaterials Science, vol. 1, T. Akaike, T. Okano, M. Akashi, M. Terano, and N. Yui, Eds. p 91–100, CMC Co., LTD., Tokyo. Jeong, B.J., Lee, J.H., and Lee, H.B. 1996. Preparation and characterization of comb-like PEO gradient surfaces. J. Colloid Interface Sci., 178:757–763. Jozefowicz, M. and Jozefowicz, J. 1985. New approaches to anticoagulation: Heparin-like biomaterials. J. Am. Soc. Art. Intern. Org. 8:218–222. Keogh, J.R., Valender, F.F., and Eaton J.W. 1992. Albumin-binding surfaces for implantable devices, J. Biomed. Mater. Res., 26:357–372. Khang, G., Park J.B., and Lee, H.B. 1996a. Prevention of platelet adhesion on the polysulfone porous catheter by saline perfusion, I. In vitro investigation, Bio-Med. Mater. Eng., 6(1):47–66. Khang, G., Park, J.B., and Lee, H.B. 1996b. Prevention of platelet adhesion on the polysulfone porous catheter by saline perfusion, II. Ex vivo and in vivo investigation, Bio-Med. Mater. Eng., 6(2):123–134. Khang, G., Lee, H.B., and Park, J.B. 1996c. Radiation effects on polypropylene for sterilization, Bio-Med. Mater. Eng., 6(5):323–334. Khang, G., Kang, Y.H., Park, J.B., and Lee, H.B, 1996d. Improved bonding strength of polyethylene/polymethylmetacrylate bone cement—a preliminary study, Bio-Med. Mater. Eng., 6(5):335–344. Khang, G., Jeon, J.H., Lee, J.W., Cho, S.C., and Lee, H.B. 1997a. Cell and platelet adhesion on plasma glow discharge-treated poly(lactide-co-glycolide), Bio-Med. Mater. Eng., 7(6):357–368. Khang, G., Lee, J.H., and Lee, H.B. 1997b. Cell and platelet adhesion on gradient surfaces, In: Advances in Biomaterials Science vol. 1, T. Akaike, T. Okano, M. Akashi, Terano, and N. Yui, Eds. p 63–70, CMC Co., LTD., Tokyo. Khang, G., Lee, J.W., Jeon, J.H., Lee, J.H., and Lee, H.B., 1997c. Interaction of fibroblasts on microgrooved polyethylene surfaces with wettabililty gradient, Biomat. Res., 1(1);1–6. Khang, G., Cho, S.Y., Lee, J.H., Rhee, J.M. and Lee, H.B., 1998. Interactions of fibroblast, osteoblast, hepatoma, and endothelial cells on poly(lactide-co-glycolide) surface with chemogradient, to appear. Khang, G., Jeon, J.H., and Lee, H.B., 1998. Fibroblast cell migration on polyethylene wettability chemogradient surfaces, to appear.


Kim, H.G., Lee, J .H., Lee, H.B., and Jhon, M.S. 1993. Dissociation behavior of surface-grafted poly(acrylic acid): Effects of surface density and counterion size, J. Colloid Interface Sci., 157(1):82–87. Kim, S.W., and Feijen, J. 1985. Surface modification of polymers for improved blood biocompatibility, CRC Crit. Rev. Biocompat., 1(3):229–260. Lazarus, J.M. 1980. Complications in hemodialysis: An overview, Kidney Int., 18:783–796. Lee, H.B. 1989. Application of synthetic polymers in implants. In: Frontiers of Macromolecular Science, T. Seagusa, T., Higashimura, and A. Abe, Eds. p.579–584, Blackwell Scientific Publications, Oxford. Lee, H.B. and Lee, J.H. 1995. Biocompatibility of solid substrates based on surface wettability, In: Encyclopedic Handbook of Biomaterials and Bioengineering: Materials, vol. 1., D.L. Wise, D.J. Trantolo, D.E. Altobelli, M.J. Yasemski, J.D. Gresser, and E.R. Schwartz, p 371–398, Marcel Dekker, New York. Lee, J.H., Khang, G., Park, K.H., Lee, H.B., and Andrade, J.D. 1989. Polymer surfaces for cell adhesion: I. Surface modification of polymers and ESCA analysis. J. Korea Soc. Med. Biolog. Eng., 10(1):43–51. Lee, J.H., Khang, G., Park, J. W., and Lee, H. B. 1990a. Plasma protein adsorption on polyethyleneoxide gradient surfaces, 33rd IUPAC Inter. Symp. on Macromolecules, Jul. 8–13, Montreal, Canada. Lee, J.H., Shin, B.C., Khang, G., and Lee, H.B. 1990b. Algin impregnated vascular graft: I. In vitro investigation. J. Korea Soc. Med. Biolog. Eng., 11(1):97–104. Lee, J.H., Park, J.W., and Lee, H.B. 1991. Cell adhesion and growth on polymer surfaces with hydroxyl groups prepared by water vapor plasma treatment, Biomaterials, 12:443–448. Lee, J.H., Kim, H.G., Khang, G., Lee, H.B., and Jhon, M.S. 1992. Characterization of wettability gradient surfaces prepared by corona discharge treatment, J. Colloid Interface Sci., 151(2):563–570. Lee, J.H., Kim, H.W., Pak, P.K., and Lee, H.B. 1994a. Preparation and characterization of functional group gradient surfaces, J. Polym. Sci., Part A, Polym. Chem., 32:1569–1579. Lee, J.H., Jung, H.W., Kang, I.K., and Lee, H.B. 1994b. Cell behavior on polymer surfaces with different functional groups, Biomaterials, 15:705–711. Lee, J.H., Jeong, B.J., and Lee, H.B. 1997a. Plasma protein adsorption and platelet adhesion onto comblike PEO gradient surface, J. Biomed. Mater. Res., 34:105–114. Lee, J.H., Kim W.G., Kim, S.S., Lee, J.H., and Lee, H.B., 1997b. Development and characterization of an alginate-impregnated polyester vascular graft, J. Biomed. Mater. Res., 36:200–208. Lee, J.H., Khang, G., Lee, J.H., and Lee, H.B. 1997c. Interactions of protein and cells on functional group gradient surfaces, Macromol. Symp., 118:571–576. Lee, J.H., Khang, G., Lee, J.H., and Lee, H.B. 1997d. Interactions of cells on chargeable functional group gradient surfaces, Biomaterials, 18(4):351–358. Lee, J.H., Khang, G., Lee, J.H., and Lee, H.B. 1998a. Platelet adhesion onto chargeable functional group gradient surfaces, J. Biomed. Mater. Res., 40:180–186. Lee, J.H. and Lee, H.B. 1998b. Platelet adhesion onto wettability gradient surfaces in the absence and presence of plasma protein, J. Biomed. Mater. Res., 41:304–311. Lee, J.H., Lee, S.J., Khang, G., and Lee, H.B. 1998c. Interactions of cells onto microporous polycarbonate membrane with wettability gradient surfaces, J. Biomat. Sci., Polm. Edn., in press. Leininger, R.I. and Bigg, D.M. 1986. Polymers, In: Handbook of Biomaterials Evaluation, p.24–37, Macmillian Publishing Co., NY. Oshiro,T. 1983. Thrombosis, antithrombogenic characteristics of immobilized urokinase on synthetic polymers, In: Biocompatible Polymers, Metals, and Composites, M. Szycher, Ed. p.275–299. Technomic, Lancaster, PA. Park, J.B. 1984. Biomaterials Science and Engineering, Plenum Publication, NY. Park, J.B. and Lakes, R. 1992. Biomaterials: An Introduction, 2nd ed. p.141–168, Plenum Press, NY. Park, J.B. and Kim S.S. 1993. Prevention of mural thrombus in porous inner tube of double-layered tube by saline perfusion. Bio-Medical Mater. Eng., 3:101–116. Park, K.D., Okano, T., Nojiri, C., and Kim S.W. 1988. Heparin immobilized onto segmented polyurethane effect of hydrophillic spacers. J. Biomed. Mater. Res., 22:977–992.


Ratner, B.D., Hoffman, A.S., Schoen, F.J., and Lemons, J.E. 1996. Biomaterials Science: An Introduction to Materials in Medicine, Academic Press, NY. Raurdy, T.G., Moorlag, H.E., Schkenraad, J.M., van der Mei, H.C., and Busscher, H.J. 1997. Detachment of human endothelial under flow from wettability gradient surface with different functional groups, Cell and Mat., 7:123–133. Sato, K. 1983. Radiation sterilization of medical products. Radioisotopes., 32:431–439. Shalaby, W.S. 1988. Polymeric materials, In: Encyclopedia of Med. Dev. Instr., J.G. Webster, Ed. p.2324–2335, Wiley-Interscience Pub., NY. Sharma, C.P. and Szycher, M. Eds., Blood Compatible Materials and Devices:Perspective Toward the 21st Century, Technomic Publishing Co. Inc., Lancaster, PA. van Wachem, P.B., Beugeling. T., Feijen, J., Bantjes, A., Detmers, J.P., and van Aken, W.G. 1985. Interaction of cultured human endothelial cells with polymeric surfaces of different wettabilities, Biomaterials, 6:403–408.


Lakes, R. “Composite Biomaterials.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

40 Composite Biomaterials

Roderic Lakes University of Wisconsin-Madison

40.1 40.2 40.3 40.4 40.5 40.6 40.7

Structure Bounds on Properties Anisotropy of Composites Particulate Composites Fibrous Composites Porous Materials Biocompatibility

Composite materials are solids which contain two or more distinct constituent materials or phases, on a scale larger than the atomic. The term “composite” is usually reserved for those materials in which the distinct phases are separated on a scale larger than the atomic, and in which properties such as the elastic modulus are significantly altered in comparison with those of a homogeneous material. Accordingly, reinforced plastics such as fiberglass as well as natural materials such as bone are viewed as composite materials, but alloys such as brass are not. A foam is a composite in which one phase is empty space. Natural biological materials tend to be composites. Natural composites include bone, wood, dentin, cartilage, and skin. Natural foams include lung, cancellous bone, and wood. Natural composites often exhibit hierarchical structures in which particulate, porous, and fibrous structural features are seen on different micro-scales [Katz, 1980; Lakes, 1993]. In this segment, composite material fundamentals and applications in biomaterials [Park and Lakes, 1988] are explored. Composite materials offer a variety of advantages in comparison with homogeneous materials. These include the ability for the scientist or engineer to exercise considerable control over material properties. There is the potential for stiff, strong, lightweight materials as well as for highly resilient and compliant materials. In biomaterials, it is important that each constituent of the composite be biocompatible. Moreover, the interface between constituents should not be degraded by the body environment. Some applications of composites in biomaterial applications are: (1) dental filling composites, (2) reinforced methyl methacrylate bone cement and ultrahigh molecular weight polyethylene, and (3) orthopedic implants with porous surfaces.

40.1 Structure The properties of composite materials depend very much upon structure. Composites differ from homogeneous materials in that considerable control can be exerted over the larger scale structure, and hence over the desired properties. In particular, the properties of a composite material depend upon the shape of the heterogeneities, upon the volume fraction occupied by them, and upon the interface among the constituents. The shape of the heterogeneities in a composite material is classified as follows. The principal inclusion shape categories are (1) the particle, with no long dimension, (2) the fiber, with one long dimension, and (3) the platelet or lamina, with two long dimensions, as shown in Fig. 40.1. The inclusions may vary in size and shape within a category. For example, particulate inclusions may be


FIGURE 40.1

Morphology of basic composite inclusions. (a) particle, (b) fiber, (c) platelet.

spherical, ellipsoidal, polyhedral, or irregular. If one phase consists of voids, filled with air or liquid, the material is known as a cellular solid. If the cells are polygonal, the material is a honeycomb; if the cells are polyhedral, it is a foam. It is necessary in the context of biomaterials to distinguish the above structural cells from biological cells, which occur only in living organisms. In each composite structure, we may moreover make the distinction between random orientation and preferred orientation.

40.2 Bounds on Properties Mechanical properties in many composite materials depend on structure in a complex way, however for some structures, the prediction of properties is relatively simple. The simplest composite structures are the idealized Voigt and Reuss models, shown in Fig. 40.2. The dark and light areas in these diagrams represent the two constituent materials in the composite. In contrast to most composite structures, it is easy to calculate the stiffness of materials with the Voigt and Reuss structures, since in the Voigt structure the strain is the same in both constituents; in the Reuss structure the stress is the same. The Young’s modulus, E, of the Voigt composite is:

[

E = Ei Vi + Em 1 − Vi

]

(40.1)

in which Ei is the Young’s modulus of the inclusions, and Vi is the volume fraction of inclusions, and Em is the Young’s modulus of the matrix. The Voigt relation for the stiffness is referred to as the rule of mixtures.

FIGURE 40.2 by arrows.

Voigt (a, laminar; b, fibrous) and Reuss (c) composite models, subjected to tension force indicated


FIGURE 40.3 Stiffness vs. volume fraction for Voigt and Reuss models, as well as for dilute isotropic suspensions of platelets, fibers, and spherical particles embedded in a matrix. Phase moduli are 200 GPa and 3 GPa.

The Reuss stiffness E,

(

 1 − Vi V E= i +  Ei Em 

) 

−1

 

,

(40.2)

is less than that of the Voigt model. The Voigt and Reuss models provide upper and lower bounds respectively upon the stiffness of a composite of arbitrary phase geometry [Paul, 1960]. The bounds are far apart if, as is commonplace, the phase moduli differ a great deal, as shown in Fig. 40.3. For composite materials which are isotropic, the more complex relations of Hashin and Shtrikman (1963) provide tighter bounds upon the moduli (Fig. 40.3); both the Young’s and shear moduli must be known for each constituent to calculate these bounds.

40.3 Anisotropy of Composites Observe that the Reuss laminate is identical to the Voigt laminate, except for a rotation with respect to the direction of load. Therefore, the stiffness of the laminate is anisotropic, that is, dependent on direction [Agarwal and Broutman, 1980; Nye, 1976; Lekhnitskii, 1963]. Anisotropy is characteristic of composite materials. The relationship between stress σij and strain εkl in anisotropic materials is given by the tensorial form of Hooke’s law as follows: 3

σ ij =

3

∑ ∑C

ε .

ijkl kl

(40.3)

k =1 l =1

Here Cijkl is the elastic modulus tensor. It has 34 = 81 elements, however since the stress and strain are represented by symmetric matrices with six independent elements each, the number of independent modulus tensor elements is reduced to 36. An additional reduction to 21 is achieved by considering elastic materials for which a strain energy function exists. Physically, C2323 represents a shear modulus


since it couples a shear stress with a shear strain. C1111 couples axial stress and strain in the 1 or x direction, but it is not the same as Young’s modulus. The reason is that Young’s modulus is measured with the lateral strains free to occur via the Poisson effect, while C1111 is the ratio of axial stress to strain when there is only one non-zero strain value; there is no lateral strain. A modulus tensor with 21 independent elements describes a triclinic crystal, which is the least symmetric crystal form. The unit cell has three different oblique angles and three different side lengths. A triclinic composite could be made with groups of fibers of three different spacings, oriented in three different oblique directions. Triclinic modulus elements such as C2311, known as cross-coupling constants, have the effect of producing a shear stress in response to a uniaxial strain; this is undesirable in many applications. An orthorhombic crystal or an orthotropic composite has a unit cell with orthogonal angles. There are nine independent elastic moduli. The associated engineering constants are three Young’s moduli, three Poisson’s ratios, and three shear moduli; the cross-coupling constants are zero when stresses are aligned to the symmetry directions. An example of such a composite is a unidirectional fibrous material with a rectangular pattern of fibers in the cross-section. Bovine bone, which has a laminated structure, exhibits orthotropic symmetry, as does wood. In a material with hexagonal symmetry, out of the nine C elements, there are five independent elastic constants. For directions in the transverse plane the elastic constants are the same, hence the alternate name transverse isotropy. A unidirectional fiber composite with a hexagonal or random fiber pattern has this symmetry, as does human Haversian bone. In cubic symmetry, there are three independent elastic constants, a Young’s modulus, E, a shear modulus, G, and an independent Poisson’s ratio, ν. Crossweave fabrics have cubic symmetry. Finally, an isotropic material has the same material properties in any direction. There are only two independent elastic constants, hence E, G, ν, and also the bulk modulus B are related in an isotropic material. Isotropic materials include amorphous solids, polycrystalline metals in which the grains are randomly oriented, and composite materials in which the constituents are randomly oriented. Anisotropic composites offer superior strength and stiffness in comparison with isotropic ones. Material properties in one direction are gained at the expense of properties in other directions. It is sensible, therefore, to use anisotropic composite materials only if the direction of application of the stress is known in advance.

40.4 Particulate Composites It is often convenient to stiffen or harden a material, commonly a polymer, by the incorporation of particulate inclusions. The shape of the particles is important [see, e.g., Christensen, 1979]. In isotropic systems, stiff platelet (or flake) inclusions are the most effective in creating a stiff composite, followed by fibers; and the least effective geometry for stiff inclusions is the spherical particle, as shown in Fig. 40.3. A dilute concentration of spherical particulate inclusions of stiffness Ei and volume fraction Vi, in a matrix (with Poisson’s ratio assumed to be 0.5) denoted by the subscript m, gives rise to a composite with a stiffness E:

E=

(

)

5 Ei − Em Vi 3+2

Ei Em

+ Em

(40.4)

The stiffness of such a composite is close to the Hashin-Shtrikman lower bound for isotropic composites. Even if the spherical particles are perfectly rigid compared with the matrix, their stiffening effect at low concentrations is modest. Conversely, when the inclusions are more compliant than the matrix, spherical ones reduce the stiffness the least and platelet ones reduce it the most. Indeed, soft platelets are suggestive of crack-like defects. Soft platelets, therefore result not only in a compliant composite, but also a weak


FIGURE 40.4 Microstructure of a dental composite. Miradapt® (Johnson & Johnson) 50% by volume filler: Barium glass and colloidal silica [Park and Lakes, 1992].

one. Soft spherical inclusions are used intentionally as crack stoppers to enhance the toughness of polymers such as polystyrene [high impact polystyrene], with a small sacrifice in stiffness. Particle reinforcement has been used to improve the properties of bone cement. For example, inclusion of bone particles in PMMA cement somewhat improves the stiffness and improves the fatigue life considerably [Park et al., 1986]. Moreover, the bone particles at the interface with the patient’s bone are ultimately resorbed and are replaced by ingrown new bone tissue. This approach is in the experimental stages. Rubber used in catheters, rubber gloves, etc. is usually reinforced with very fine particles of silica (SiO2) to make the rubber stronger and tougher. Teeth with decayed regions have traditionally been restored with metals such as silver amalgam. Metallic restorations are not considered desirable for anterior teeth for cosmetic reasons. Acrylic resins and silicate cements had been used for anterior teeth, but their poor material properties led to short service life and clinical failures. Dental composite resins have virtually replaced these materials and are very commonly used to restore posterior teeth as well as anterior teeth [Cannon, 1988]. The dental composite resins consist of a polymer matrix and stiff inorganic inclusions [Craig, 1981]. A representative structure is shown in Fig. 40.4. The particles are very angular in shape. The inorganic inclusions confer a relatively high stiffness and high wear resistance on the material. Moreover, since they are translucent and their index of refraction is similar to that of dental enamel, they are cosmetically acceptable. Available dental composite resins use quartz, barium glass, and colloidal silica as fillers. Fillers have particle size from 0.04 µm to 13 µm, and concentrations from 33 to 78% by weight. In view of the greater density of the inorganic filler phase, a 77% weight percent of filler corresponds to a volume percent of about 55%. The matrix consists of a polymer, typically BIS-GMA. In restoring a cavity, the dentist mixes several constituents, then places them in the prepared cavity to polymerize. For this procedure to be successful the viscosity of the mixed paste must be sufficiently low and the polymerization must be controllable. Low viscosity liquids such as triethylene glycol dimethacrylate are used to lower


TABLE 40.1

Properties of Bone, Teeth, and Biomaterials Young’s modulus E[GPa]

Density ρ (g/cm3)

Hard Tissue Tooth, bone, human compact bone, longitudinal direction

17

1.8

130 (tension)

Tooth dentin Tooth enamel Polymers Polyethylene (UHMW) Polymethyl methacrylate, PMMA PMMA bone cement Metals 316L Stainless steel (wrought) Co-Cr-Mo (cast) Co Ni Cr Mo (wrought) Ti6A14V Composites Graphite-epoxy (unidirectional fibrous, high modulus) Graphite-epoxy (quasi-isotropic fibrous) Dental composite resins (particulate) Foams Polymer foams

18 50

2.1 2.9

138 (compression)

1 3 2

0.94 1.1 1.18

30 (tension) 65 (tension) 30 (tension)

200 230 230 110

7.9 8.3 9.2 4.5

1000 (tension) 660 (tension) 1800 (tension) 900 (tension)

215

1.63

1240 (tension)

(Schwartz, 1997)

46 10-16

1.55

579 (tension) 170-260 (compression)

10-4–1

0.002–0.8

(Schwartz, 1997) (Cannon, 1988) (Gibson and Ashby, 1988)

Material

Strength (MPa)

Refs. (Craig, and Peyton, 1958; Peters et al., 1984; Park and Lakes, 1992; Reilly and Burstein, 1975)

(Park and Lakes, 1992)

(Park and Lakes, 1992)

0.01–1 (tension)

the viscosity and inhibitors such as BHT (butylated trioxytoluene) are used to prevent premature polymerization. Polymerization can be initiated by a thermochemical initiator such as benzoyl peroxide, or by a photochemical initiator (benzoin alkyl ether) which generates free radicals when subjected to ultraviolet light from a lamp used by the dentist. Dental composites have a Young’s modulus in the range 10 to 16 GPa, and the compressive strength from 170 to 260 MPa [Cannon, 1988]. As shown in Table 40.1, these composites are still considerably less stiff than dental enamel, which contains about 99% mineral. Similar high concentrations of mineral particles in synthetic composites cannot easily be achieved, in part because the particles do not pack densely. Moreover, an excessive concentration of particles raises the viscosity of the unpolymerized paste. An excessively high viscosity is problematical since it prevents the dentist from adequately packing the paste into the prepared cavity; the material will then fill in crevices less effectively. The thermal expansion of dental composites, as with other dental materials, exceeds that of tooth structure. Moreover, there is a contraction during polymerization of 1.2 to 1.6%. These effects are thought to contribute to leakage of saliva, bacteria, etc., at the interface margins. Such leakage in some cases can cause further decay of the tooth. Use of colloidal silica in the so-called “microfilled” composites allows these resins to be polished, so that less wear occurs and less plaque accumulates. It is more difficult, however, to make these with a high fraction of filler. All the dental composites exhibit creep. The stiffness changes by a factor of 2.5 to 4 (depending on the particular material) over a time period from 10 sec to three h under steady load [Papadogianis et al., 1985]. This creep may result in indentation of the restoration, but wear seems to be a greater problem. Dental composite resins have become established as restorative materials for both anterior and posterior teeth. The use of these materials is likely to increase as improved compositions are developed and in response to concern over long term toxicity of silver-mercury amalgam fillings.


40.5 Fibrous Composites Fibers incorporated in a polymer matrix increase the stiffness, strength, fatigue life, and other properties [Agarwal and Broutman, 1980; Schwartz, 1992]. Fibers are mechanically more effective in achieving a stiff, strong composite than are particles. Materials can be prepared in fiber form with very few defects which concentrate stress. Fibers such as graphite are stiff (Young’s modulus is 200 to 800 GPa) and strong (the tensile strength is 2.7 to 5.5 GPa). Composites made from them can be as strong as steel but much lighter, as shown in Table 40.1. The stiffness of a composite with aligned fibers, if it is loaded along the fibers, is equivalent to the Voigt upper bound, Eq. (40.1). Unidirectional fibrous composites, when loaded along the fibers, can have strengths and stiffnesses comparable to that of steel, but with much less weight (Table 40.1). However if it is loaded transversly to the fibers, such a composite will be compliant, with a stiffness not much greater than that of the matrix alone. While unidirectional fiber composites can be made very strong in the longitudinal direction, they are weaker than the matrix alone when loaded transversely, as a result of stress concentration around the fibers. If stiffness and strength are needed in all directions, the fibers may be oriented randomly. For such a three-dimensional isotropic composite, for a low concentration of fibers,

E=

Ei Vi + Em , 6

(40.5)

so the stiffness is reduced by about a factor of six in comparison with an aligned composite as illustrated in Fig. 40.3. However if the fibers are aligned randomly in a plane, the reduction in stiffness is only a factor of three. The degree of anisotropy in fibrous composites can be very well controlled by forming laminates consisting of layers of fibers embedded in a matrix. Each layer can have fibers oriented in a different direction. One can achieve quasi-isotropic behavior in the laminate plane; such a laminate is not as strong or as stiff as a unidirectional one, as illustrated in Table 40.1. Strength of composites depends on such particulars as the brittleness or ductility of the inclusions and the matrix. In fibrous composites failure may occur by (1) fiber breakage, buckling, or pullout, (2) matrix cracking, or (3) debonding of fiber from matrix. Short fiber composites are used in many applications. They are not as stiff or as strong as composites with continuous fibers, but they can be formed economically by injection molding or by in situ polymerization. Choice of an optimal fiber length can result in improved toughness, due to the predominance of fiber pull-out as a fracture mechanism. Carbon fibers have been incorporated in the high density polyethylene used in total knee replacements (Fig. 40.5). The standard ultra high molecular weight polyethylene (UHMWPE) used in these implants is considered adequate for most purposes for implantation in older patients. A longer wear-free implant lifetime is desirable for use in younger patients. It is considered desirable to improve the resistance to creep of the polymeric component, since excessive creep results in an indentation of that component after long term use. Representative properties of carbon reinforced ultra high molecular weight polyethylene are shown in Fig. 40.6 [Sclippa and Piekarski, 1973]. Enhancements of various properties by a factor of two are feasible. Polymethyl methacrylate (PMMA) used in bone cement is compliant and weak in comparison with bone. Therefore several reinforcement methods have been attempted. Metal wires have been used clinically as macroscopic “fibers” to reinforce PMMA cement used in spinal stabilization surgery [Fishbane and Pond, 1977]. The wires are made of a biocompatible alloy such as cobalt-chromium alloy or stainless steel. Such wires are not currently used in joint replacements owing to the limited space available. Graphite fibers have been incorporated in bone cement [Knoell et al., 1975] on an experimental basis. Significant improvements in the mechanical properties have been achieved. Moreover, the fibers have an added beneficial effect of reducing the rise in temperature which occurs during the polymerization of the PMMA


FIGURE 40.5

Knee prostheses with polyethylene tibial components reinforced with carbon fiber .

in the body. Such high temperature can cause problems such as necrosis of a portion of the bone into which it is implanted. Thin, short titanium fibers have been embedded in PMMA cement [Topoleski et al., 1992]; a toughness increase of 51% was observed with a 5% volumetric fiber content. Fiber reinforcement of PMMA cement has not found much acceptance since the fibers also increase the viscosity of the unpolymerized material. It is consequently difficult for the surgeon to form and shape the polymerizing cement during the surgical procedure. Metals are currently used in bone plates for immobilizing fractures and in the femoral component of total hip replacements. A problem with currently used implant metals is that they are much stiffer than bone, so they shield the nearby bone from mechanical stress. Stress-shielding results in a kind of disuse atrophy: the bone resorbs [Engh and Bobyn, 1988]. Therefore composite materials have been investigated as alternatives [Bradley et al., 1980; Skinner, 1988]. Fibrous composites can deform to higher strains (to about 0.01) than metals (0.001 for a mild steel) without damage. This resilience is an attractive characteristic for more flexible bone plates and femoral stems. Flexible composite bone plates are effective in © 2000 by CRC Press LLC

FIGURE 40.6 Properties of carbon fiber reinforced ultra high molecular weight polyethylene. Replotted from Sclippa and Piekarski [1973].

promoting healing [Jockish, 1992]. Composite hip replacement prostheses have been made with carbon fibers in a matrix of polysulfone and polyetherether ketone (PEEK). These prostheses experience heavy load with a static component. Structural metals such as stainless steel and cobalt chromium alloys do not creep significantly at room or body temperature. In composites which contain a polymer constituent, creep behavior is a matter of concern. The carbon fibers exhibit negligible creep, but polymer constituents tend to creep. Prototype composite femoral components were found to exhibit fiber dominated creep of small magnitude and are not expected to limit the life of the implant [Maharaj and Jamison, 1993]. Fibrous composites have also been used in external medical devices such as knee braces [Yeaple, 1989], in which biocompatibility is not a concern but light weight is crucial.

40.6 Porous Materials The presence of voids in porous or cellular solids will reduce the stiffness of the material. For some purposes, that is both acceptable and desirable. Porous solids are used for many purposes: flexible structures such as (1) seat cushions, (2) thermal insulation, (3) filters, (4) cores for stiff and lightweight sandwich panels, (5) flotation devices, and (6) to protect objects from mechanical shock and vibration; and in biomaterials, as coatings to encourage tissue ingrowth. Representative cellular solid structures are shown in Fig. 40.7. The stiffness of an open-cell foam is given by [Gibson and Ashby, 1988]

[ ]

E = Es Vs

2

(40.6)

in which Es is the Young’s modulus and Vs is the volume fraction of the solid phase of the foam; Vs is also called the relative density. The strength for crushing of a brittle foam and the elastic collapse of an elastomeric foam is given, respectively, by

[ ]

σ crush = 0.65 σ f ,s Vs


32

(40.7)

FIGURE 40.7 Cellular solids structures, after Gibson and Ashby [1988]. Left: synthetic cellular solids: (a) open-cell polyurethane, (b) closed-cell polyethylene, (c) foamed nickel, (d) foamed copper, (e) foamed zirconia, (f) foamed mullite, (g) foamed glass, (h) polyester foam with both open and closed cells. Right: natural cellular solids: (a) cork, (b) balsa wood, (c) sponge, (d) cancellous bone, (e) coral, (f) cuttlefish bone, (g) iris leaf, (h) plant stalk.

[ ]

2

σ coll = 0.05 Es Vs .

(40.8)

Here, σf,s is the fracture strength of the solid phase. These strength relations are valid for relatively small density. Their derivation is based on the concept of bending of the cell ribs and is presented by Gibson and Ashby [1988]. Most man-made closed cell foams tend to have a concentration of material at the cell edges, so that they behave mechanically as open cell foams. The salient point in the relations for the mechanical properties of cellular solids is that the relative density dramatically influences the stiffness and the strength. As for the relationship between stress and strain, a representative stress strain curve is shown in Fig. 40.8. The physical mechanism for the deformation mode beyond the elastic limit depends on the material from which the foam is made. Trabecular bone, for example, is a natural cellular solid, which tends to fail in compression by crushing. Many kinds of trabecular bone appear to behave mechanically as an open cell foam. For trabecular bone of unspecified orientation, the stiffness is proportional to the cube of the density and the strength as the square of the density [Gibson and Ashby, 1988], which indicates behavior dominated by bending of the trabeculae. For bone with oriented trabeculae, both stiffness and strength in the trabecular direction are proportional to the density, a fact which indicates behavior dominated by axial deformation of the trabeculae. Porous materials have a high ratio of surface area to volume. When porous materials are used in biomaterial applications, the demands upon the inertness and biocompatibility are likely to be greater than for a homogeneous material. Porous materials, when used in implants, allow tissue ingrowth [Spector et al., 1988a,b]. The ingrowth is considered desirable in many contexts, since it allows a relatively permanent anchorage of the implant to the surrounding tissues. There are actually two composites to be considered in porous implants: (1) the implant prior to ingrowth, in which the pores are filled with tissue fluid which is ordinarily of no © 2000 by CRC Press LLC

FIGURE 40.8 Representative stress-strain curve for a cellular solid. The plateau region for compression in the case of elastomeric foam (a rubbery polymer) represents elastic buckling; for an elastic-plastic foam (such as metallic foam), it represents plastic yield, and for an elastic-brittle foam (such as ceramic) it represents crushing. On the tension side, point ‘A’ represents the transition between cell wall bending and cell wall alignment. In elastomeric foam, the alignment occurs elastically, in elastic plastic foam it occurs plastically, and an elastic-brittle foam fractures at A.

mechanical consequence; and (2) the implant filled with tissue. In the case of the implant prior to ingrowth, it must be recognized that the stiffness and strength of the porous solid are much less than in the case of the solid from which it is derived. Porous layers are used on bone compatible implants to encourage bony ingrowth [Galante et al., 1971; Ducheyne, 1984]. The pore size of a cellular solid has no influence on its stiffness or strength (though it does influence the toughness), however pore size can be of considerable biological importance. Specifically, in orthopedic implants with pores larger than about 150 µm, bony ingrowth into the pores occurs and this is useful to anchor the implant. This minimum pore size is on the order of the diameter of osteons in normal Haversian bone. It was found experimentally that pores β-TCP >>> HA, whereas amorphous HA is more prone to biodegradation than crystalline HA [Koeneman et al., 1990]. Considerable attempts to investigate the effects of coating composition (relative percentages of HA, TCP) on bone integration have been undertaken. Using an orthopedic canine total hip model, Jasty and coworkers (1992) reported that by 3 weeks, a TCP/HA mixed coating resulted in significantly more woven bone apposition to the implants than uncoated implants. As determined by x-ray diffraction, the mixed coating consisted of 60% TCP, 20% crystalline HA, and 20% unknown Ca-PO4 materials. Jansen and colleagues [1993] reported that, using HA-coated implants (90% HA, 10% amorphous CP), bony apposition was extensive in a rabbit tibia model by 12 weeks. However, significant loss of the coating occurred as early as 6 weeks after implantation. Most recently, Maxian and coworkers [1993] reported that poorly crystallized HA (60% crystalline) coatings demonstrated significant degradation and poor bone apposition in vivo compared to amorphous coatings. Both these reports suggest that although considerable bioresorption of the coating occurred in the cortical bone, there was significant bone apposition (81 ± 2% for amorphous HA at 12 weeks, 77% for crystalline HA, respectively) which was not significantly affected by bioresorption. From these in vivo reports, it is clear that HA coatings with relatively low levels of crystallinity are capable of significant bone apposition. However, as reported in a 1990 workshop report, the FDA is strongly urging commercial implant manufacturers to use techniques to increase the postdeposition crystallinity and to provide adequate adhesion of the coating to the implant substrate [Filiaggi et al., 1993]. Although the biologic responses to HA coatings are encouraging, other factors regarding HA coatings continue to lead to clinical questions regarding their efficacy. Although the overall bone response to HA-coated implants occurs more rapidly than with uncoated devices, with time an equivalent bone contact area is formed for both materials [Jasty et al., 1992]. These results have questioned the true need for HA-coated implants, especially when there are a number of disadvantages associated with the coating concept. Clinical difficulties have arisen due to failures within the coating itself and with continued dissolution of the coating, and to catastrophic failure at the coating-substrate interface [Koeneman et al., 1988]. Recent progress is reported in terms of the improvements in coating technology. Postdeposition heat treatments are often utilized to control the crystallinity (and therefore the dissolution characteristics) of the coatings, although there is still debate as to the relationship between compositional variations associated with differing crystallinity and optimization of biologic responses. Additional coating-related properties are also under investigation in regard to their effects on bone. These include coating thickness, level of acceptable porosity in the coating, and adherence of the coating to the underlying substrate. However, until answers concerning these variables have been more firmly established, HA coatings used for dental implants will remain an area of controversy and interest. © 2000 by CRC Press LLC

FIGURE 44.13 Examples of current dental implant designs, illustrating the variety of macroscopic topographies which are used to encourage tissue ingrowth. Left to right: Microvent, Corevent, Screw-vent, Swede-vent, Branemark, IMZ implant.

Effects of Surface Properties Surface Topography The effects of surface topography are different than the overall three-dimensional design or geometry of the implant, which is related to the interaction of the host tissues with the implant on a macroscopic scale as shown in Fig. 44.13. This important consideration in overall biologic response to implants is discussed later in this chapter. In this discussion the concept of surface topography refers to the surface texture on a microlevel. It is on this microscopic level that the intimate cell and tissue interactions leading to osseointegration are based as shown in Fig. 44.14. The effects of surface topography on in vitro and in vivo cell and tissue responses have been a field of intense study in recent years. The overall goal of these studies is to identify surface topographies which mimic the natural substrata in order to permit tissue integration and improve clinical performance of the implant. In terms of cell attachment, the in vitro work by Bowers and colleagues [1992] established that levels of short-term osteoblast cell attachment were higher on rough compared to smooth surfaces and cell morphology was directly related to the nature of the underlying substrate. After initial attachment, in many cases, cells of various origin often take on the morphology of the substrate as shown in Fig. 44.15. Increased surface roughness, produced by such techniques as sand or grit blasting or by rough polishing, provided the rugosity necessary for optimum cell behavior. Work in progress in several laboratories is attempting to relate the nature of the implant surface to cell morphology, intracellular cytoskeletal organization, and extracellular matrix development. Pioneering work by Chehroudi and coworkers [1992] suggests that microtextured surfaces (via micromachining or other techniques) could help orchestrate cellular activity and osteoblast mineralization by several mechanisms including proper orientation of collagen bundles and cell shape and polarity. This concept is related to the theory of contact guidance and the belief that cell shape will dictate cell differentiation through gene expression. In Chehroudi’s work, both tapered pitted and grooved surfaces (with specific orientation and sequence patterns) supported mineralization with ultrastructural morphology similar in appearance to that observed by Davies and colleagues [1990]. However, mineralization was not observed on smooth surfaces in which osteoblastlike cells did not have a preferred growth orientation. © 2000 by CRC Press LLC

Thus the control of surface microtopography by such procedures as micromachining may prove to be a valuable technology for the control and perhaps optimization of bone formation on implant surfaces. It is apparent that macroscopic as well as microscopic topography may affect osteoblast differentiation and mineralization. In a recent study by Groessner-Schrieber and Tuan [1992], osteoblast growth, differentiation, and synthesis of matrix and mineralized nodules were observed on rough, textured, or porous coated titanium surfaces. It may be possible therefore, not only to optimize the interactions of host tissues with implant surfaces during the Phase I tissue responses but also to influence the overall bone responses to biomechanical forces during the remodeling phase (Phase II) of tissue responses. Based on these concepts, current implant designs employ microtopographically roughened surfaces with macroscopic grooves, threads, or porous surfaces to provide sufficient bone ingrowth for mechanical

FIGURE 44.14 Laboratory-prepared cpTi surfaces with (a–c, top to bottom) smooth (1 µm polish), grooved (600 grit polish), and rough (sandblasted) surfaces. © 2000 by CRC Press LLC

FIGURE 44.15 Osteoblastlike cell morphology after 2 hours’ attachment on (a–c, top to bottom) smooth, grooved, and rough cpTi surfaces.

stabilization and the prevention of detrimental micromotion as shown in Figs. 44.16 and 44.17 [Brunski, 1992; De Porter et al., 1986; Keller et al., 1987; Pilliar et al., 1991].

Surface Chemistry Considerable attention has focused on the properties of the oxide found on titanium implant surfaces following surface preparation. Sterilization procedures are especially important and are known to affect not only the oxide condition but also the subsequent in vitro [Stanford et al., 1994; Swart et al., 1992] and in vivo [Hartman et al., 1989] biologic responses. Interfacial surface analyses and determinations of surface energetics strongly suggest that steam autoclaving is especially damaging to titanium oxide surfaces. Depending upon the purity of the autoclave water, contaminants have been observed on the metal oxide and are correlated with poor tissue responses on a cellular [Keller et al., 1990, 1994] and tissue [Baier et al., 1984; Hartman et al., 1989; Meenaghan et al., 1979] level. The role of multiple sterilization regimens on the practice of implant utilization is also under scrutiny. Many implants and especially bone plate systems are designed for repackaging if the kit is not exhausted. However, early evidence indicates that this practice is faulty and, depending on the method of sterilization, may affect the integrity of the metal oxide surface chemistry [Vezeau et al., 1991]. In vitro experiments have verified that multiple-steam-autoclaved and ethFIGURE 44.16 Light microscopic photomicroylene-oxide-treated implant surfaces adversely affected graph of a bone–smooth cpTi interface with intercellular and morphologic integration. However, the vening layer of soft connective tissue. This implant effects of these treatments on long-term biological was mobile in the surgical site due to lack of tissue responses including in vivo situations remain to be ingrowth. (Original magnification = 50×.) clarified. Other more recently introduced techniques such as radiofrequency argon plasma cleaning treatments have succeeded in altering metal oxide chemistry and structure [Baier et al., 1984; Swart et al., 1992]. Numerous studies have demonstrated that PC treatments produce a relatively contaminant-free surface with improved surface energy (wettability), but conflicting biologic results have been reported with these surfaces. Recent in vitro studies have demonstrated that these highly energetic surfaces do not necessarily improve cellular responses such as attachment and cell expression. This has been confirmed by in vivo studies which indicate that the overall histologic and ultrastructural morphology of the bone-implant interface is similar for plasma-cleaned and dry-heat-sterilized implant surfaces [Albrektsson et al., 1983]. Another promising technique for the sterilization of implant materials is the exposure of the implant surface to ultraviolet light [Singh and Schaff, 1989] or gamma irradiation [Keller et al., 1994]. Both these methods of sterilization produce a relatively contaminant-free thin oxide layer which fosters high levels of cell attachment [Keller et al., 1994] and inflammatory-free long-term in vivo responses [Hartman et al., 1989]. Currently, gamma irradiation procedures are widely used for the sterilization of metallic dental implant devices.


FIGURE 44.17 Light microscopic photomicrograph of a bone–porous Ti alloy implant interface. Note significant bone ingrowth in open porosity at the apical end of the implant. (Original magnification = 10×.)

Metallic Corrosion Throughout the history of the use of metals for biomedical implant applications, electrochemical corrosion with subsequent metal release has been problematic [Galante et al., 1991]. Of the biomedical metal systems available today, Ti and its major medical alloy, Ti-6A1-4V, are thought to be the most corrosion resistant; however, Ti metals are not totally inert in vivo [Woodman et al., 1984]. Release of Ti ions from Ti oxides can occur under relatively passive conditions [Ducheyne, 1988]. Whereas other factors such as positioning of the implant and subsequent biomechanical forces may play important roles in the overall tissue response to implants, it is not unreasonable to predict that electrochemical interactions between the implant surface and host tissue may affect the overall response of host bone [Blumenthal and Cosma, 1989]. For example, it has been shown by several groups [De Porter et al., 1986; Keller et al., 1987] that the percentages of intimate bony contact with the implant is inconsistent, at best, and generally averages approximately 50% over a 5-year period. Continued studies involving the effects of dissolution products on both local and systemic host responses are required in order to more fully understand the consequences of biologic interaction with metal implants. Future Considerations for Implant Surfaces It is clear that future efforts to improve the host tissue responses to implant materials will focus, in large part, on controlling cell and tissue responses at implant interfaces. This goal will require continued acquisition of fundamental knowledge of cell behavior and cell response to specific materials’ characteristics. It is likely that a better understanding of the cellular-derived extracellular matrix-implant interface will offer a mechanism by which biologic response modifiers such as growth and attachment factors or hormones may be incorporated. Advancements of this type will likely shift the focus of future research from implant surfaces which as osseoconductive (permissive) to those which are osseoinductive (bioactive).


Defining Terms Amorphous zone: A region of the tissue-implant interface immediately adjacent to the implant substrate. This zone is of variable thickness (usually < 1000 Å), is free of collagen, and is comprised of proteoglycans of unknown composition. Calcium phosphate: A family of calcium- and phosphate-containing materials of synthetic or natural origin which are utilized for implants and bone augmentation purposes. The most prominent materials are the tricalcium-phosphate- and hydroxyapatite-based materials, although most synthetic implants are a mixture of the various compositions. Contact guidance: The theory by which cells attach to and migrate on substrates of specific microstructure and topographic orientation. The ability of the cell to attach and migrate on a substrate is related to the cytoskeletal and attachment molecules present on the cell membrane. Junctional epithelium: The epithelial attachment mechanism which occurs with teeth, and has been observed infrequently with implants by some researchers. Less than 10 cell layers thick, the hemidesmosomal attachments of the basal cells to the implant surface provide a mechanical attachment for epithelium and prevent bacterial penetration into the sulcular area. Osseointegration: A term developed by P.I. Branemark and his colleagues indicating the ability of host bone tissues to form a functional, mechanically immobile interface with the implant. Originally described for titanium only, several other materials are capable of forming this interface, which presumes a lack of connective tissue (foreign body) layer. Plasma spray: A high-temperature process by which calcium-phosphate-containing materials are coated onto a suitable implant substrate. Although the target material may be of high purity, the high-temperature softening process can dramatically affect and alter the resultant composition of the coating.

References Albrektsson T, Branemark PI, Hansson HA, et al. 1983. The interface zone of inorganic implants in vivo: Titanium implants in bone. Ann Biomed Eng 11:1. Albrektsson T, Hansson HA, Ivarsson B. 1985. Interface analysis of titanium and zirconium bone implants. Biomaterials 6:97. Baier RE, Meyer AE, Natiella JR, et al. 1984. Surface properties determine bioadhesive outcomes. J Biomed Mater Res 18:337. Blumenthal NC, Cosma V. 1989. Inhibition of appetite formation by titanium and vanadium ions. J Biomed Mater Res 23(A1):13. Bowers KT, Keller JC, Michaels CM, et al. 1992. Optimization of surface micromorphology for enhanced osteoblast responses in vitro. Int J Oral Maxillofac Implants 7:302. Branemark PI. 1983. Osseointegration and its experimental background. J Pros Dent 59(3):399. Brunski JB. 1992. Biomechanical factors affecting the bone-dental implant interface. Clin Mater 10:153. Chehroudi B, Ratkay J, Brunette DM. 1992. The role of implant surface geometry on mineralization in vivo and in vitro: A transmission and scanning electron microscopic study. Cells Materials 2(2):89–104. Cook SD, Kay JF, Thomas KA, et al. 1987. Interface mechanics and histology of titanium and hydroxylapatite coated titanium for dental implant applications. Int J Oral Maxillofac Implants 2(1):15. Cook SD, Thomas KA, Kay JF. 1991. Experimental coating defects in hydroxylapatite coated implants. Clin Orthop Rel Res 265:280. Davies JE, Lowenberg B, Shiga A. 1990. The bone-titanium interface in vitro. J Biomed Mater Res 24:1289–1306. De Bruijn JD, Flach JS, deGroot K, et al. 1993. Analysis of the bony interface with various types of hydroxyapatite in vitro. Cells Mater 3(2):115.


De Porter DA, Watson PA, Pilliar RM, et al. 1986. A histological assessment of the initial healing response adjacent to porous-surfaced, titanium alloy dental implants in dogs. J Dent Res 65(8):1064. Driskell TD, Spungenberg HD, Tennery VJ, et al. 1973. Current status of high density alumina ceramic tooth roof structures. J Dent Res 52:123. Ducheyne P. 1988. Titanium and calcium phosphate ceramic dental implants, surfaces, coatings and interfaces. J Oral Implantol 19(3):325. Filiaggi MJ, Pilliar RM, Coombs NA. 1993. Post-plasma spraying heat treatment of the HA coating/Ti6A1-4V implant system. J Biomed Mater Res 27:191. Galante JO, Lemons J, Spector M, et al. 1991. The biologic effects of implant materials. J Orthop Res 9:760. Groessner-Schreiber B, Tuan RS. 1992. Enhanced extracellular matrix production and mineralization by osteoblasts cultured on titanium surfaces in vitro. J Cell Sci 101:209. Hartman LC, Meenaghan MA, Schaaf NG, et al. 1989. Effects of pretreatment sterilization and cleaning methods on materials properties and osseoinductivity of a threaded implant. Int J Oral Maxillofac Implants 4:11. Herman H. 1988. Plasma spray deposition processes. Mater Res Soc Bull 13:60. Jansen JA, van der Waerden JPCM, Wolke JGC. 1993. Histological and histomorphometrical evaluation of the bone reaction to three different titanium alloy and hydroxyapatite coated implants. J Appl Biomater 4:213. Jarcho M, Kay JF, Gumaer KI, et al. 1977. Tissue, cellular and subcellular events at a bone-ceramic hydroxylapatite interface. J Bioeng 1:79. Jasty M, Rubash HE, Paiemont GD, et al. 1992. Porous coated uncemented components in experimental total hip arthroplasty in dogs. Clin Orthop Rel Res 280:300. Johansson CB, Lausman J, Ask M, et al. 1989. Ultrastructural differences of the interface zone between bone and Ti-6A1-4V or commercially pure titanium. J Biomed Eng 11:3. Katsikeris N, Listrom RD, Symington JM. 1987. Interface between titanium 6-A1-4V alloy implants and bone. Int J Oral Maxillofac Surg 16:473. Kay JF. 1992. Calcium phosphate coatings for dental implants. Dent Clinic N Amer 36(1):1. Keller JC, Draughn RA, Wightman JP, et al. 1990. Characterization of sterilized cp titanium implant surfaces. Int J Oral Maxillofac Implants 5:360. Keller JC, Lautenschlager EP. 1986. Metals and alloys. In A Von Recon (ed), Handbook of Biomaterials Evaluation, pp 3–23, New York, Macmillan. Keller JC, Niederauer GG, Lacefield WR, et al. 1992. Interaction of osteoblast-like cells with calcium phosphate ceramic materials. Trans Acad Dent Mater 5(3):107. Keller JC, Stanford CM, Wightman JP, et al. 1994. Characterization of titanium implant surfaces. J Biomed Mater Res. Keller JC, Young FA, Natiella JR. 1987. Quantitative bone remodeling resulting from the use of porous dental implants. J Biomed Mater Res 21:305. Koeneman J, Lemons JE, Ducheyne P, et al. 1990. Workshop of characterization of calcium phosphate materials. J Appl Biomater 1:79. Lucas LC, Lemons JE, Lee J, et al. 1987. In vivo corrosion characteristics of Co-Cr-Mu/Ti-6A1-4V-Ti alloys. In JE Lemons (ed), Quantitative Characterization and Performance of Porous Alloys for Hard Tissue Applications, pp 124–136, Philadelphia, ASTM. Maxian SH, Zawadsky JP, Durin MG. 1993. Mechanical and histological evaluation of amorphous calcium phosphate and poorly crystallized hydroxylapatite coatings on titanium implants. J Biomed Mater Res 27:717. Meenaghan MA, Natiella JR, Moresi JC, et al. 1979. Tissue response to surface treated tantalum implants: Preliminary observations in primates. J Biomed Mater Res 13:631. Orr RD, de Bruijn JD, Davies JE. 1992. Scanning electron microscopy of the bone interface with titanium, titanium alloy and hydroxyapatite. Cells Mater 2(3):241. Pilliar RM, DePorter DA, Watson PA, et al. 1991. Dental implant design—effect on bone remodeling. J Biomed Mater Res 25:647. © 2000 by CRC Press LLC

Puleo DA, Holleran LA, Doremus RH, et al. 1991. Osteoblast responses to orthopedic implant materials in vitro. J Biomed Mater Res 25:711. Singh S, Schaaf NG. 1989. Dynamic sterilization of titanium implants with ultraviolet light. Int J Oral Maxillofac Implants 4:139. Skalak R. 1985. Aspects of biomechanical considerations. In PI Branemark, G Zarb, T Albrektsson (eds), Tissue Integrated Prostheses, pp 117–128, Chicago, Quintessence. Smith DC. 1993. Dental implants: Materials and design considerations. Int J Prosth 6(2):106. Stanford CM, Keller JC. 1991. Osseointegration and matrix production at the implant surface. CRC Crit Rev Oral Bio Med 2:83. Stanford CM, Keller JC, Solursh M. 1994. Bone cell expression on titanium surfaces is altered by sterilization treatments. J Dent Res. Steflik DE, McKinney RV, Koth DL, et al. 1984. Biomaterial-tissue interface: A morphological study utilizing conventional and alternative ultrastructural modalities. Scanning Electron Microscopy 2:547. Steflik DE, Sisk AL, Parr GR, et al. 1993. Osteogenesis at the dental implant interface: High voltage electron microscopic and conventional transmission electric microscopic observations. J Biomed Mater Res 27:791. Swart KM, Keller JC, Wightman JP, et al. 1992. Short term plasma cleaning treatments enhance in vitro osteoblast attachment to titanium. J Oral Implant 18(2):130. Van Orden A. 1985. Corrosive response of the interface tissue to 316L stainless steel, Ti-based alloy and cobalt-based alloys. In R McKinney, JE Lemons (eds), The Dental Implant, pp 1–25, Littleton, PSG. Vezeau PJ, Keller JC, Koorbusch GF. 1991. Effects of multiple sterilization regimens on fibroblast attachment to titanium. J Dent Res 70:530. Vrouwenvelder WCA, Groot CG, Groot K. 1993. Histological and biochemical evaluation of osteoblasts cultured on bioactive glass, hydroxylapatite, titanium alloy and stainless steel. J Biomed Mater Res 27:465–475. Woodman JL, Jacobs JJ, Galante JO, et al. Metal ion release from titanium-based prosthetic segmental replacements of long bones in baboons: A long term study. J Orthop Res 1:421–430. Young FA. 1988. Future directions in dental implant materials research. J Dent Ed 52(12):770.


Coger, R., Toner, M. “Preservation Techniques for Biomaterials.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

45 Preservation Techniques for Biomaterials Robin Coger Massachusetts General Hospital, Harvard University Medical School, and the Shriners Burns Institutes

Mehmet Toner Massachusetts General Hospital, Harvard University Medical School, and the Shriners Burns Institute

45.1 45.2 45.3 45.4 45.5 45.6

Phase Behavior Nonfreezing Storage: Hypothermic Freeze-Thaw Technology Freeze-Drying Vitrification Summary

Biomaterials—i.e., proteins, cells, tissues, and organs—are used daily to preserve life. Uses such as blood transfusions, artificial insemination, burn repair, transplantation, and pharmaceuticals rely on their availability. Natural materials, however, are labile and often deteriorate over time. To counter this effect, preservation procedures for retarding deterioration rates have been developed. Furthermore, since each biomaterial is characterized by unique compositional and physical complexities, a variety of storage techniques exists. Table 45.1 lists examples of biomaterials that have been preserved successfully using various procedures. The list, although abbreviated, illustrates the wide range of cells, tissues, organs, and macromolecule structures that have been stored successfully and demonstrates how, in some cases (e.g., red blood cells), multiple preservation techniques may be appropriate. In the discussion that follows, four biomaterial storage procedures—nonfreezing, freeze-thaw, freeze-drying (lyophilization), and vitrification—are summarized. Nonfreezing techniques enable biomaterial storage by retarding metabolic processes and chemical reactions during cooling from physiologic to nonfreezing temperatures. With freeze-thaw techniques, the biomaterial is stored at low temperatures (usually less than—70°C) in crystalline form and then thawed for actual use. Lyophilized biomaterials are first frozen, then dehydrated by sublimation for storage at ambient temperature, and finally reconstituted for use. With vitrification, the biomaterials is cooled to subzero temperature in such a way that it is transformed to an amorphous solid for storage and then rewarmed for use. Each procedure—i.e., its applications and risks—will be described in more detail below. This discussion is not a comprehensive review but a general overview of the principles governing each of these four common biomaterial storage techniques.

45.1 Phase Behavior As mentioned previously, biomaterials in their naturally occurring forms tend to deteriorate over time. Hence to achieve safe long-term storage, it is generally necessary to alter the physicochemical state of the biomaterial. One approach is to promote the transformation of the original substance to a form that can be stored safely and then, when needed, restored to its original state. These types of transformations can best be represented using phase diagrams.


TABLE 45.1 Some Examples of Biomaterials Stored Using Various Storage Techniques Nonfreezing

Freeze-thaw

Lyophilization (freeze-dry)

Vitrification

FIGURE 45.1

Liver Kidney Heart valves Protein solutions Red blood cells Cartilage Bone marrow Skin Red blood cells Platelets Penicillin Collagen Liposomes Embryos Islets of Langerhans Corneal tissue Drosophila melanogaster

Phase diagram for a hypothetical solute additive. [Adapted from Fahy et al., 1984]

Figure 45.1 illustrates a temperature versus concentration phase diagram, where concentration corresponds to the quantity of a hypothetical solute additive present in the solution. The diagram is particularly useful for describing phase transformations in which a thermodynamic phase of reduced energy nucleates within a parent phase. Crystallization is one example of this type of phase transformation and is a twostep process of nucleation of the new phase and its subsequent growth. If the formation of the new phase is catalyzed from the surface of foreign particles or a substrate, the growth process is triggered by heterogeneous nucleation (THET). If, however, the new phase develops from the clustering of individual water molecules, growth of the new phase occurs from homogeneous nucleation (THOM) [Hobbs, 1974]. The latter type of nucleation requires higher activation energies than heterogeneous nucleation and thus occurs at lower temperatures. Figure 45.1 also illustrates the relative positions of the melting (TM) and the glass-transition (TG) temperature curves. The melting-temperature curve shows the melting-point depression of a crystalline sample relative to solute concentration. The glass-transition curves signifies the temperatures for which a supercooled solution becomes glassy during cooling. The final curve in © 2000 by CRC Press LLC

Fig. 45.1 (TD) demonstrates the devitrification temperature profile and illustrates the conditions for which a substance may experience crystallization damage during warming. For glasslike solids, damage is incurred as the glass transforms to a crystalline form at TD , while for previously crystalline or partially vitrified solids, recrystallization (the coalescence of small crystals during warming) produces damage. The described elements of the phase diagram are relevant because all the biomaterial storage techniques described in this chapter involve either a phase transformation or its avoidance. Furthermore, if a phase transformation is improperly controlled, the effects can be detrimental to the biomaterial. Four techniques relevant to biomaterial storage are now described.

45.2 Nonfreezing Storage: Hypothermic Given the high cost of organ transplantation, longer preservation times are urgently needed to make the procedure cost-effective [Evans, 1993] and expand the geographic area over which organs can be harvested and transplanted. Presently, hypothermic storage is the clinically employed organ preservation technique. There are essentially two techniques for hypothermic preservation of organs for subsequent transplantation. The first is static cold storage by ice immersion at ~4°C—to reduce metabolism as much as possible without the formation of deleterious ice crystals [Belzer and Southard, 1988]. The second procedure is continuous cold machine perfusion at ~10°C to provide diminished metabolism and to remove deleterious end products [Southard and Belzer, 1988]. The perfusate used for static cold storage mimics the intracellular ionic composition of the organ and has impermeant solutes added to prevent cell swelling. However, for perfusion preservation techniques in general, modified plasma or solutions of plasma-like composition usually have been preferred. This is presumably due to the slightly higher temperatures used in perfusion techniques that permit some degree of metabolism to occur during storage. In most practical instances, simple cold storage is the preservation mode of choice, since the storage time is only slightly increased by using continuous perfusion. Furthermore, the development of the University of Wisconsin (UW) solution has dramatically extended the preservation times. In the case of human liver, for example, the storage time increased from less than 10 to about 30 hours, thus alleviating the need for continuous-perfusion storage. As a result of extensive studies, it seems that lactobionate is an essential component of the UW solution, which is believed to act as an effective osmotic agent while suppressing cell swelling in metabolically depressed cells and tissues [Southard and Belzer, 1993]. Another nonfreezing technique that can be used to retard the chemical reaction rates of a biomaterial is undercooled storage. During undercooled storage, the biomaterial is exposed to subzero temperatures in the absence of deleterious ice crystal formation. Such a method may be particularly applicable to the mediumterm (months) storage of proteins. Small droplets of aqueous-based solutions are dispersed in an inert oil phase, and the final preparation is then put into test tubes for freezer storage at T ≥ –20°C. By keeping the quantity of aqueous droplets much greater than the number of heterogeneous nucleation sites in the bulk solution, one can effectively deter heterogeneous nucleation during storage. When needed, the dispersions are removed from the freezer and warmed to room temperature and reconstituted [Franks, 1988]. Unfortunately, a chief problem in using the undercooled storage technique in biochemical application lies in the scale-up of the procedure to the larger volumes needed for clinical and commercial applications. Furthermore, methods for “efficiently” separating the innocuous oil from the biomaterial are a problem that must be solved if the previously stored protein solutions are to be used in therapeutic applications. Apparently, it is also difficult to conduct all steps of the undercool procedure under sterile conditions [Franks, 1988].

45.3 Freeze-Thaw Technology Freeze-thaw technology is the most commonly used method for storing cells and tissues; hence it is an important cryopreservation methodology. When a transplantation emergency arises, the preserved biomaterial can be thawed and used to save lives. In addition, the development of cell and tissue transplantation procedures also benefits the growing field of tissue engineering—where frozen biomaterials are © 2000 by CRC Press LLC

utilized in organ replacement devices (e.g., bioartificial liver) [Borel-Rinkes et al., 1992; Karlsson et al., 1993]. Even the simple cell, however, is thermophysically complex. Consequently, the design of feasible cryopreservation protocols must incorporate knowledge from biology, engineering, and medicine to be successful. In 1949, Polge and coworkers made an important contribution to biomaterial preservation research. In work with sperm, they were the first to report the protective effects of additives or cryoprotectant agents (CPAs), i.e., glycerol, on biomaterials at low temperatures. The mechanisms by which CPAs protect cells from freeze injury are of fundamental importance but are, unfortunately poorly understood. There are four major protective actions of these compounds. First, CPAs act to stabilize the proteins of biomaterials under low-temperature conditions. Experimental evidence indicates that for cells, this effect results from the interaction of sugars with the polar head groups of phospholipids [McGann, 1978]. Recent thermodynamic analyses demonstrate that the preferential exclusion of CPAs from the protein hydration shell, at low temperatures, results in stabilization [Arakawa et al., 1990]. Second, CPAs lower the electrolyte concentration of the suspending medium of the cell at a given temperature by altering the phase relationship during cooling. Third, CPAs reduce the temperature at which cells undergo lethal intracellular ice formation. Fourth, CPAs promote the formation of the vitreous, rather than crystalline, phases inside the cell during cooling and help prevent intracellular ice formation (see Fig. 45.1) [Karlsson et al., 1994]. Unfortunately, the addition and removal of CPAs from the biomaterial introduce a separate set of problems. In the case of penetrating additives (i.e., dimethyl sulfoxide and glycerol), the permeability of the cell membrane is typically several orders of magnitude less than the permeability of the membrane to water [McGrath, 1988]. During the prefreeze addition of these compounds, the biomaterial recognizes the CPA as another extracellular solute. Hence the cell responds by initiating the rapid transport of water across the cell membrane and into the extracellular medium. Meanwhile, if the CPA is permeable, it gradually diffuses into the cell, thus contributing to the state of chemical nonequilibrium experienced by the cell. These transport processes continue (usually for minutes) until equilibrium is regained, at which time the volume of the biomaterial returns to normal. If the initial concentration of CPA is too large, the cell volume variations may be so severe that the biomaterial experiences osmotic stress damage. The reverse process, of cell swelling and the subsequent return to normal volume, is observed during the removal of CPAs. Osmotic injury from the addition or removal of penetrating compounds can be minimized by incrementally increasing or decreasing, respectively, their concentration. Unfortunately, the benefits of using stepwise CPA addition and removal in reducing osmotic damage is counterbalanced by the increased risk of toxic damage to the biomaterial from longer exposure times [Fahy, 1986]. Also, in some cases, a single-step CPA removal process has been shown to be effective [Friedler et al., 1988]. The balance between these considerations can be optimized if both the permeability of the biomaterial to CPAs and the effects of the CPA on the biomaterial are known. Methodologies for measuring these critical parameters are discussed in McGrath [1988]. In the case of impermeable CPAs (i.e., polyvinylpyrrolidone, hydroxyethyl starch), the potentially injurious effects of exosmosis are also experienced. However, because impermeable CPAs remain in the extracellular medium, they are relatively easier to remove from the biomaterial than permeable additives. Their major disadvantage is believed to be their inability to afford direct protection to intracellular structures during freezing. Figure 45.2 illustrates extrema observed when cooling biologic substances to subzero temperatures. To minimize cryoinjury, a suitable CPA is added to the biomaterial before the actual freeze-thaw protocol is begun. Cooling then proceeds at a controlled rate. As the material is cooled along the freezing path (Fig. 45.2), ice formation initiates first in the extracellular medium. This phase transformation is initiated either by external seeding or by heterogeneous nucleation from the walls of the container and is crucial because it corresponds to the initial temperature at which physicochemical changes occur within the biomaterial during cooling. The formation of extracellular ice results in a chemical potential gradient across the cell membrane that can be balanced by exosmosis of the intracellular water. If the freezing process occurs at slow cooling rates, the intracellular fluid has sufficient time to leave the cell. However, © 2000 by CRC Press LLC

FIGURE 45.2

Schematic of the physicochemical processes experienced by cells during cryopreservation.

if excessive exosmosis occurs, the result is excessive cell dehydration and shrinkage and subsequent cell death from the high concentration of solutes remaining within the cell after water diffusion. If the freezing rate is rapid, however, water entrapped in the cell becomes supercooled. Then, at some subzero temperature, thermodynamic equilibrium is achieved through intracellular ice formation. Unfortunately, intracellular ice formation is usually associated with irreversible cell damage. Although the exact damage mechanism is not known, mechanical effects and high electrolyte concentrations caused by the crystal formation are believed to be major modes of cell damage [Mazur, 1984]. Figure 45.3 schematically illustrates the relationship between cell survival and cooling rate during a freeze-thaw protocol. As shown, cell damage at suboptimal cooling rates is predominantly caused by “solution” effects (e.g., exposure to high electrolyte concentrations, decreased unfrozen fractions, excessive dehydration), until a range of peak survival is obtained at the optimal cooling rate. Experimental evidence suggests that the optimal cooling rate for maximum cellular survival is the fastest rate that will not result in intracellular ice formation. Beyond this peak, cells are exposed to supraoptimal cooling conditions, in which damage from intracellular ice formation dominates. Theoretical models have been developed to describe both the kinetics of water loss and the probability of intracellular ice formation [Mazur, 1984; Pitt, 1992; Muldrew and McGann, 1993]. The value of these models in designing effective freezing protocols for biomaterial preservation also has been demonstrated [e.g., Karlsson et al., 1993]. Once a cell or tissue sample has been frozen successfully, it can then be stored in liquid nitrogen in a –70°C freezer until needed. This step is usually not very crucial, since, as long as the freezer temperature is kept stable, storage does not cause cell damage. The next challenge in freeze-thaw technology is to thaw the biomaterial using a warming protocol that minimizes the risk of recrystallization, or devitrification if applicable, and promotes high survival. The optimal thawing rate correlates directly with the rate at which the sample was cooled before storage. If the material was frozen slowly at a suboptimal rate (Fig. 45.3), a wide range of thawing rates usually can be utilized with negligible effects on survival. If the biomaterial experienced slightly supraoptimal cooling rates (Fig. 45.3), rapid thawing is almost exclusively required to avoid recrystallization of small intracellular ice crystals formed during the initial cooling step. The rapid thawing of volumes of appreciable size is particularly difficult because the thermal mass of the specimen dictates the warming rate that can be achieved. © 2000 by CRC Press LLC

FIGURE 45.3 protocol.

A diagram illustrating the relationship between cell survival and cooling rate during a freeze-thaw

45.4 Freeze-Drying Freeze-drying, or lyophilization, is a dehydration storage technique that is advantageous because it produces a final product that can be stored at suprazero temperatures and reconstituted with the addition of an appropriate solvent. The procedure is commonly applied to protein products (e.g., collagen, penicillin) and less frequently applied to cells (e.g., platelets, red blood cells) [Doebbler et al., 1966; Goodrich and Sowemimo-Coker, 1993]. Freeze-drying techniques also have been applied to liposomes—vesicles utilized in drug delivery systems that can trap water-soluble substances within their lumina. Lyophilization is a dual stage process that consists of rapidly cooling the liquid material to its solid form and subsequent drying (or removing) of the solidified solvent. The intricacies of the procedures employed directly influence the shelf life of the final dehydrated product. The drying stage of the process utilizes vacuum sublimation (transformation of a solid phase directly to the vapor phase) and may itself consist of multiple steps. The complexities of the drying procedure are determined by the continuity of the ice phase throughout the frozen specimen [MacKenzie, 1976]. Franks [1982] explains that the quality of the final lyophilized product partially depends on the freezing protocol experienced by the original sample, the size and distribution of the resultant ice crystals, the degree of heterogeneity, the presence or absence of amorphous regions, and the conditions imposed on the sample during drying. As one might expect, these factors can produce differentiations between the desired goals of the freeze-drying process and the actual final product. The effects of the freezing protocol and the size and distribution of the resultant crystals are important for the reasons discussed previously. It should be mentioned that in the case of protein lyophilization, carbohydrates are often added for protection, while for cells, glycerol is a common lyoprotectant. The next two factors—the degree of heterogeneity and presence of amorphous regions—directly relate to the drying procedure. For simple cases in which large areas of ice (e.g., continuous channels) have formed in the sample, ice is removed by direct sublimation. However, for the more complex configurations


commonly encountered with biomaterials, sublimation alone is inadequate because the ice crystal formation is discontinuous and amorphous regions are present. The amorphous regions correspond to pockets of water bound by freeze-concentrated solutes and biomaterial components (e.g., cell membranes). Water bound in this way serves to satisfy the hydration shell requirements of the specimen [Steinbrecht and Müller, 1987], hence contributing to the stabilization of the biomaterial. It is now recognized that the formation of amorphous regions is necessary for the successful lyophilization of biomaterials [Crowe et al., 1993a; Levine and Slade, 1992]. Since material destabilization is one source of damage during freeze-drying, this is an important area of lyophilization research. Regardless of the complexity of the frozen matrix, biomaterial injury can occur during drying. Factors such as the effective drying temperature and the rate of drying are important in determining the degree of damage incurred. The drying step entails the removal of first the free water and then the bound water. However, the complete removal of this bound water during freeze-drying can be injurious to biomaterials and can promote protein denaturation and possibly aggregation into insoluble precipitates. Denaturation is the disruption of the folded structure of a protein and is affected by variables such as pH, surface interactions, and thermal and chemical changes [Darby and Creighton, 1993]. It is unfavorable because the final lyophilized product may no longer resemble the original biomaterial once denaturation occurs. Hence the avoidance of protein denaturation is crucial to effective biomaterial lyophilization. Investigations with liposomes and proteins reveal that the survival of biomaterials in the dry state is linked to the stabilizing influence of disaccharides (such as sucrose and trehalose) present in the system [Crowe et al., 1993b]. However, to minimize denaturation during freeze-drying protocols, lyoprotectants and CPAs should be used [Cleland et al., 1993]. It should be mentioned that the removal of these additives can be problematic in the large-scale industrial use of freeze-drying techniques. Once a biomaterial has been lyophilized successfully, it is stored for future use. For the special case of protein storage—an area that is of particular interest to the pharmaceutical industry—the stability of the drug or protein must be guaranteed throughout a reasonable shelf life (the time period, measured in years, for which the drug maintains its original physical and functional properties). Two means of biomaterial destabilization during storage are the occurrence of oxidative and chemical reactions after lyophilization. Oxidation effects can be reduced by the exclusion of oxygen from containers of the dried materials and the use of antioxidants. The chemical reactions can be inhibited through the maintenance of low residual moisture levels [Cleland et al., 1993]. The final step, in the use of freeze-drying techniques for biomaterial storage, is the reconstitution of the lyophilized product. If pure water is used as the rehydration solvent, concentration gradients and osmotic imbalances can result in severe injury. To counter these effects, biomaterials are commonly rehydrated using isotonic solutions or media. The effects of using additives in the reconstituting solvent of biomaterials has recently been addressed. Apparently sugars are also effective in reducing damage during this rehydration step [Crowe et al., 1993b].

45.5 Vitrification The hazards of ice crystal formation are significantly reduced by rapidly cooling the biomaterial to low temperatures at sufficient rates to produce an amorphous solid. This alternative, depicted in Fig. 45.2, was originally proposed in the 1930s and is called vitrification [Luyet, 1937; Goetz and Goetz, 1938]. Vitrification is the kinetic process by which a liquid solidifies into a glass. It requires the rapid cooling of the liquid to the glass-transition temperature TG and is most easily achieved in high-viscosity liquids [Doremus, 1973]. The molecular configuration of the supercooled liquid (T ≥ TG) is the same as that of the glass (T ≤ TG). Hence rapid cooling is necessary to prevent the supercooled liquid molecules from reorganizing into a regular (e.g., lattice) configuration. Vitrification is a second-order phase transition. Hence, by definition, the specific volumes of both phases (near TG) are identical, although the thermodynamic property values (i.e., heat capacity, coefficient of thermal expansion) are not [Kauzmann, 1948].


The difficulty in successfully vitrifying a material lies in reaching its glass transition temperature TG prior to crystal formation. Hence reducing the distance between TM and TG by increasing the solute concentration (Fig. 45.1) increases the probability that a given liquid will form a glass. Two alternative ways to achieve glassy state are (1) to cool biomaterials at ultrarapid rates such that TG is reached before nucleation can proceed and (2) to increase the pressure of the system such that the intersection of THOM and TG (Fig. 45.1) occurs at lower CPA concentrations [Fahy et al., 1984; Kanno and Angell, 1977]. Since glass formation circumvents the deleterious effects of freeze injury during cooling, it is becoming an increasingly important biomaterial storage technique. We have already mentioned the roles of viscosity and rapid cooling in vitrification. The more viscous a liquid, the slower it can be cooled to achieve its vitrified form. Fairly large cryoprotectant concentrations are necessary to obtain the vitrified form of aqueous solutions by slow cooling, and the requirement increases with specimen volume. This is undesirable, since high concentrations of CPAs are toxic to biomaterials. Hence, in practical applications of vitrification technology, a balance between the thermodynamic conditions necessary to achieve vitrification and the physicochemical conditions suitable for survival is crucial. The most common difficulty encountered in attempts to vitrify biologic materials is their susceptibility to ice crystal formation. As mentioned previously, the larger the sample volume, the higher the CPA concentration necessary to reduce the probability of crystallization [Coger et al., 1990]. Drosophila melanogaster cells, corneal tissue, and pancreatic islets of Langerhans are three examples of biomaterials that have been vitrified successfully [Steponkus et al., 1990; Bourne, 1986; Jutte et al., 1987]. Of these, only D. melanogaster has not been preserved previously by the freeze-thaw technique. In fact, the freeze-thaw technique remains the chief method of cells storage. For many cell and tissue types, the appropriate combination of effective temperature, cryoprotectant concentration, and cooling and warming rate conditions necessary for their vitrification have yet to be determined. For long-term organ storage, the freeze-thaw technique is not an alternative, because the mechanical stress and toxic damage that the organ would incur during freezing would be lethal [Bernard et al., 1988; Fahy et al., 1984]. Organ vitrification attempts, although unsuccessful, have demonstrated that the CPA toxicity limitation is especially evident in organ preservation such that toxic death prevention is an unavoidable challenge in organ vitrification [Fahy, 1986]. With respect to cooling rate, the relatively large volumes of organs require the use of slow cooling protocols to ensure a uniform thermal history throughout the total volume. Organs therefore require high-pressure, high-cryoprotectant concentrations and slow cooling conditions to achieve the vitreous state. Once a biomaterial has been vitrified successfully, it must be stored at temperatures below TG (Fig. 45.1) to encourage stability. When restoring the sample to physiologic conditions, special care must be taken to avoid crystallization during the warming protocol. Crystallization under these circumstances, termed devitrification, is possible (or even probable) since at temperatures greater than TG , the crystalline phase is more stable than the amorphous solid. If such a transformation occurs, cryoinjury is unavoidable. The probability of devitrification is significantly reduced by warming the biomaterial at rates equivalent to those imposed during the original cooling [Fahy, 1988; MacFarlane et al., 1992]. The use of vitrification as a biomaterial storage technique is an ongoing and important area of cryopreservation research. Presently, it is the only apparent solution for the long-term storage of organs.

45.6 Summary Biomaterial storage is an exciting research area whose advances are of value to a variety of fields, including medicine, biologic research, and drug design. However, the compositional and physical complexities of the various biosubstances (e.g., proteins, cells, tissues, and organs) require the development of specialized preservation procedures. In the present discussion, four stage techniques—nonfreezing, freeze-thaw, lyophilization, and vitrification—and their relevance to specific biomaterial examples have been reviewed. Although there have been definite advances, important challenges still remain for future investigations.


Defining Terms Cryopreservation: Techniques utilized to store cells, tissues, and organs under subzero conditions for future use in clinical applications. Cryoprotectant: Chemical additive used to protect biomaterials during cooling to low temperatures by reducing freezing injury. Devitrification: Crystallization of an amorphous substance during warming. Freeze-drying: Dehydration of a sample by vacuum sublimation. Lyoprotectant: Expedients added to lyophilization formulations to protect the biomaterial from the damaging effects of the process. Recrystallization: The coalescence of small ice crystals during warming. Shelf life: Length of time in which the stability of the components of a biomaterial (e.g., pharmaceutical drugs) is guaranteed during storage. Vitrification: Solidification process in which an amorphous (glasslike) solid, devoid of crystals, is formed.

References Arakawa T, Carpenter JF, Kita YA, Crowe JH. 1990. The basis for toxicity of certain cryoprotectants: A hypothesis. Cryobiology 27:401. Belzer FO, Southard JH. 1988. Principles of solid-organ preservation by cold storage. Transplantation 45:673. Bernard A, McGrath JJ, Fuller BJ, et al. 1988. Osmotic response to oocytes using a microscope diffusion chamber: A preliminary study comparing murine and human ova. Cryobiology 25:945. Borel-Rinkes IHM, Toner M, Tompkins RG, Yarmush ML. 1993. Long-term functional recovery of hepatocytes after cryopreservation in a three-dimensional culture configuration. Cell Transplant 1:281. Bourne WM. 1986. Clinical and experimental aspects of corneal cryopreservation. Cryobiology 23:88. Cleland JL, Powell MF, Shire SJ. 1993. The development of stable protein formulations: A close look at protein aggregation, deamidation, and oxidation. In S Bruck (ed), Crit Rev Ther Drug Carrier Syst 10:307. Coger R, Rubinsky B, Pegg DE. 1990. Dependence of probability of vitrification on time and volume. Cryo-Letters 11:359. Crowe JH, Crowe LM, Carpenter JF. 1993a. Preserving dry biomaterials: The water replacement hypothesis, part 1. Biopharm 6:28. Crowe JH, Crowe LM, Carpenter JF. 1993b. Preserving dry biomaterials: The water replacement hypothesis, part 2. Biopharm 6:40. Darby NJ, Creighton TE. 1993. Protein Structure. Oxford, Oxford University Press. Doebbler FG, Rowe AW, Rinfret AP. 1966. Freezing of mammalian blood and its constituents. In HT Meryman (ed), Cryobiology, pp 407–450. London, Academic Press. Doremus RH. 1973. Glass Science. New York, Wiley. Evans RW. 1993. A cost-outcome analysis of retransplantation the need for accountability. Transplant Rev 7:163. Fahy GM. 1988. Vitrification. In JJ McGrath, KR Diller (eds), Low Temperature Biotechnology, vols 10 and 98, pp 113–146. New York, ASME. Fahy GM. 1986. The relevance of cryoprotectant “toxicity” to cryobiology. Cryobiology 23:1. Fahy GM, MacFarlane DR, Angell CA, Meryman HT. 1984. Vitrification as an approach to cryopreservation. Cryobiology 21:407. Franks F. 1988. Storage in the undercooled state. In JJ McGrath, KR Diller (eds), Low Temperature Biotechnology, vols 10 and 98, pp 107–112. New York, ASME. Franks F. 1982. The properties of aqueous solutions at subzero temperatures. In F Franks (ed), Water: A Comprehensive Treatise, vol 7, pp 215–338. New York, Plenum Press.


Friedler S, Giudice L, Lamb E. 1988. Cryopreservation of embryos and ova. Fertil Steril 49:473. Goetz A, Goetz SS. 1938. Vitrification and crystallization of organic cells at low temperatures. J Appl Physiol 9:718. Goodrich RP, Sowemimo-Coker SO. 1993. Freeze drying of red blood cells. In PL Steponkus (ed), Advances in Low-Temperature Biology, vol 2, pp 53–99. London, JAI Press. Hobbs PV. 1974. Ice Physics. Oxford, Oxford University Press. Jutte NHPM, Heyse P, Jansen HG, et al. 1987. Vitrification of human islets of Langerhans. Cryobiology 24:403. Kanno H, Angell CA. 1977. Homogenous nucleation and glass formation in aqueous alkali halide solutions at high pressure. J Phys Chem 81(26):2639. Karlsson JOM, Cravalho EG, Borel-Rinkes IHM, et al. 1993. Nucleation and growth of ice crystals inside cultured hepatocytes during freezing in the presence of dimethyl sulfoxide. Biophys J 65:2524. Karlsson JOM, Cravalho EG, Toner M. 1994. A model of diffusion-limited ice growth inside biological cells during freezing. J Appl Physiol 75:4442. Kauzmann W. 1948. The nature of the glassy state and the behavior of liquids at low temperatures. Chem Rev. 43:219. Levine H, Slade L. 1992. Another view of trehalose for drying and stabilizing biological materials. Biopharm 5:36. Luyet B. 1937. The vitrification of organic colloids and of protoplasm. Biodynamica 1:1. MacFarlane DR, Forsyth M, Barton CA. 1992. Vitrification and devitrification in cryopreservation. In PL Steponkus (ed), Advances in Low Temperature Biology, vol 1, pp 221–278. London, JAI Press. MacKenzie AP. 1976. Principles of freeze-drying. Transplant Proc 8(suppl 1):181. Mazur P. 1984. Freezing of living cells: Mechanisms and implications. Am J Physiol 143:C125. McGann LE. 1978. Differing actions of penetrating and nonpenetrating cryoprotective agents. Cryobiology 15:382. McGrath JJ. 1988. Membrane transport properties. In JJ McGrath, KR Diller (eds), Low Temperature Biotechnology, vols 10 and 98, pp 273–330. New York, ASME. Muldrew K, McGann LE. 1994. The osmotic rupture hypothesis of intracellular freezing injury. Biophys J 66:532. Pitt RE. 1992. Thermodynamics and intracellular ice formation. In PL Steponkus (ed), Advances in LowTemperature Biology, vol 1, pp 63–99. London, JAI Press. Polge C, Smith AU, Parkes AS. 1949. Revival of spermatozoa after vitrification and dehydration at low temperatures. Nature 164:666. Rall WP, Mazur P, McGrath JJ. 1983. Depression of the ice-nucleation temperature of rapidly cooled mouse embryos by glycerol and dimethyl sulfoxide. Biophys J 41:1. Southard JH, Belzer FO. 1993. The University of Wisconsin organ preservation solution: components, comparisons, and modifications. Transplant Rev 7:176. Southard JH, Belzer FO. 1988. Kidney preservation by perfusion. In GJ Cerilli (ed), Organ Transplantation and Replacement, pp 296–311. Philadelphia, Lippincott. Steponkus PL, Myers SP, Lynch DV, et al. 1990. Cryopreservation of Drosophila melanogaster embryos. Nature 345:170. Steinbrecht RA, Muller M. 1987. Freeze substitution and freeze-drying. In RA Steinbrecht, K Zierold (eds), Cryotechniques in Biological Electron Microscopy, pp 149–172. Berlin, Springer-Verlag.

Further Information In addition to the review works cited throughout the text, the following publications also may be of interest. For more detailed information concerning lyophilization, as applied specifically to proteins, see The development of stable protein formulations: A close look at protein aggregation, deamidation, and oxidation, by Cleland et al., an especially helpful article (1993. Crit. Rev. Ther. Drug Carrier Syst 10:307). © 2000 by CRC Press LLC

A good introduction to the fields of cryobiology and cryopreservation is presented in The Clinical Applications of Cryobiology, by B. J. Fuller and B. W. W. Grout (CRC Press, Boca Raton, Fla., 1991). For review work relevant to vitrification and organ preservation, The Biophysics of Organ Cryopreservation, edited by D. E. Pegg and A. M. Karow Jr. (Plenum Press, New York, 1987) should be consulted. For additional information on the physicochemical processes encountered during cooling of biomaterials, refer to Korber’s 1988 review (Q. Rev. Biophys. 21:229). For an overview of bioheat transfer processes, refer to Diller’s review in Bioengineering Heat Transfer 22:157, 1992. For additional information on the fundamental principles of intracellular ice formation, consult the 1993 review by Toner (pp. 1–51 in Advances in Low Temperature Biology, edited by P. L. Steponkus, vol. 2, JAI Press, London).


Park, J. B. “Hip Joint Prosthesis Fixation-Problems and Possible Solutions.” The Biomedical Engineering Handbook: Second Edition Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

46 Hip Joint Prosthesis Fixation-Problems and Possible Solutions 46.1 46.2

Acetabular Cup Femoral Stem Cemented and Uncemented Fixation • Possible Solutions for Cemented Fixation • Uncemented Fixation

Joon B. Park The University of Iowa

46.3

Articulating Surface of the Acetabular Cup and Femoral Head

Total hip joint replacement (THR) has been very successful mainly due to the introduction of bone cement for the fixation by Dr. John Charnley on the advice of Dr. Dennis Smith in the late 1950s [Charnley, 1970; Charnley, 1972; Charnley and Cupic, 1973; Eftekhar, 1978; Wroblewski, 1986]. One of the inherent problems of orthopedic joint prosthesis implantation is the fixation and maintenance of a stable interface between the device and the host tissue at the cellular and organ levels [Crowninshield, 1988; Ducheyne, 1988; Park, 1992; Park, 1995]. The fixation can be classified into several categories as given in Table 46.1. Basically, the hip joint can be considered as three parts, (1) acetabular cup, (2) cup-femoral head-articulating surface, and (3) femoral stem (Fig. 46.1). The problems and possible solutions related to each of the three parts will be examined. Most frequent fixation problems are related to (1) infection, (2) wear and wear particulate, (3) migration and failure of implants, and (4) loosening of which the “long-term loosening” of the implant is especially important [Ducheyne, 1988; Park, 1992]. These problems manifest into osteolysis in the bone bed which is the major cause of long-term loosening mostly for the femoral stem [Anthony et al., 1990; Harris, 1995; Maloney et al., 1990; Maloney and Smith, 1995]. Some major factors related to (late) loosening are (1) mismatch of the physical properties between tissues and implant, (2) biocompatibility of the implant, (3) deterioration of physical properties of implant materials, (4) surgical techniques, (5) design of the implant, (6) selection of patients, and (7) post surgical care, etc. [Amstutz et al., 1976; Carlsson et al., 1983; Crowninshield, 1988; Dowd et al., 1995; Gruen et al., 1979; Harris and McGann, 1984; Havelin et al., 1995; Lu et al., 1993; Mohler et al., 1995; Mulroy and Harris, 1990; Nilsen and Wiig, 1996; RiegelsNielsen et al., 1995; Stromberg et al., 1996]. In this review, recently reported clinical cases based on the large studies will be examined, especially the Norwegian Arthroplasty Register which will shed some light on the cemented vs. uncemented fixation of the total hip arthroplasty. This is mainly due to the fact that the individuals reporting and their clinical results could be widely varied and biased. Variables related to the total (hip) joint replacement are (1) materials (Table 46.2), (2) design, and (3) fixation method (Table 46.3). One should keep in mind that any particular type of prosthesis is made to have a particular fixation method, e.g., the Charnley prosthesis is designed to be used with bone


TABLE 46.1

Summary of Various Methods of Prosthesis Fixation

Methods of Fixation

Examples

1. Mechanical Fixation a. Active—use of screws, bolts, nuts, wires, etc. b. Passive—interference fit and noninterference fit 2. Bone Cement Fixation a. Pure cement b. Modified cement—composite cement 3. Biological Fixation a. Porous ingrowth b. Modified porous ingrowth—electrical and pulsed electromagnetic field (PEMF) stimulation 4. Direct (Chemical) Bonding Fixation a. Osteogenic/inductive-glass-ceramics b. Osteoconductive hydroxyapatite

FIGURE 46.1

Pre-Charnley implants [Williams and Roaf, 1973] [Moore, 1952; Mittlemeier, 1975, 1976] [Charnley, 1970; 1972] [Dai et al., 1991; Henrich et al., 1993] resorbable particle impregnated cement, [Liu et al., 1987; Park et al., 1986] [Klawitter and Hulbert, 1972; Sauer et al., 1974; Hirshhorn et al., 1972; Homsy et al., 1972; Smith, 1963] [Park, 1983; Park and Kenner, 1975; Weinstein et al., 1976]

[Blencke et al., 1978; Hench and Paschall, 1973] [de Groot, 1983; Kay, 1988]

Schematic diagram showing the three important areas of THR.

cement at both the cup and femoral prosthesis sites while some were designed to be used as uncemented such as the BIAS® femoral stem prosthesis [Onsten et al., 1995]. Also, sometimes the outcome of the THR may depend on the type of bone cement used; (1) low- vs. high-viscosity, (2) plain vs. antibioticsimpregnated, and (3) low vs. high heat of polymerization cement, etc. A more mitigating problem is that the outcome of the THR could not be clearly attributed to any one specific factor, e.g., the osteolysis may be due to the particulates which are due to wear and corrosion products which in turn could be initiated between the articulating surfaces or between the metal backing and polyethylene. However, the critical stage may be reached due to the accelerated rate of particulate production due to the loosening of the implant, which causes a larger and unnatural motion of the articulation. Once the lysis starts, the cycle becomes more vicious—further aggravating the situation. The design and fixation related variables are summarized in Table 46.3. It is not that uncommon to find that a prosthesis designed for uncemented fixation is used with cement, making the outcome study more complicated.


TABLE 46.2 Physical Properties of Materials Used for Joint Prosthesis and Bone (Park and Lakes, 1992) Materials

UTSa (MPa)

Elongation (%)

Density (g/cm3)

200 230 230 110

1000 660 1800 900

9 8 8 10

7.9 8.3 9.2 4.5

400 200 120

260 200 200

2 × 106 g/mole)

TABLE 46.3 THR

List of Design and Fixation Related Variables in

Part

Materials Used

Cup

All PE Metal-backed Porous-coated Screw-holed All Ceramic PE/Metal PE/Ceramic Ceramic/ceramic Metal/metal Metal Composite Smooth Textured Porous-coated PMMA-coated HA-coated Metal Ceramic

Articulating surfaces

Femoral stem

Head

Fixation Methods Cemented Uncemented (mechanical) Uncemented (mechanical) Uncemented (mechanical) Cemented/uncemented

Cemented/uncemented Uncemented Threaded(rare) Uncemented Uncemented Cemented Uncemented Fixed Rotated (trunnion/modular)

PE: Polyethylene (ultra high molecular weight)


46.1 Acetabular Cup The advent of the uncemented acetabular cup necessitated the metal-backing of the polyethylene liner although some use all polyethylene without cement [Ritter, 1995]. Due to the relatively new advocacy of the so-called “hybrid” THR, where the femoral prosthesis is cemented and the cup is uncemented, there are fewer cases of comparing the outcome of the cemented vs. uncemented acetabular cup performance. Havelin et al. compared 11 of the most widely used uncemented cups (4352 cases) [Havelin et al., 1995]. After 5 years, the overall cumulative revision rate was 3.2% but increased drastically to 7.1% after 6 years with large differences among the designs. Hydroxyapatite and porous (metal) coated cups had a failure rate of less than 0.1%. It is interesting that all polyethylene cups had a cumulative revision rate of 14% while threaded uncoated metal-backed cups varied from 0% (PM® cups) to 21% (Ti-Fit®) after 6 years. They concluded that the coating of the surface is the most important factor for the uncemented cups for the first 6 years of implantation. As is the case with all tissue ingrowth fixation, the most important factor for short-term results with uncemented cups is the immediate stability and possibility for the tissue ingrowth. They also concluded “It remains to be seen if the results with some uncemented cups, in some groups of patients, are better than for the best-cemented. Until more is known about their long-term results, uncemented acetabular components should be used as part of randomized trials.” Of course, the performance of the acetabular cup may be closely interrelated to the femoral stem. However, Havelin’s study could not detect any outstanding difference with different combinations of the cup and femoral stem. Espehaug et al. studied 12,179 hip prostheses comparing 10 different brands reported to the Norwegian Arthroplasty Register during 1987–1993 [Espehaug et al., 1995]. The overall revision rate after 5 years was 2.5% while the Charnley prosthesis was 2.9%. Elite cup and Charnley stem combination resulted in much poorer performance (9.84% after 5 years). However, other combinations gave better results. The main cause of revision was aseptic loosening after 5 years (1.8%) followed by infection (0.5% after 5 years). In conclusion, they “…observed good overall results of cemented hip prostheses. However, clinically important differences in revision rates were demonstrated among the various cemented total hip prosthesis brands.” Some studied the aseptic loosening of the cemented cups and concluded that the mechanism of late loosening is biological (osteolysis) not mechanical, which is the opposite of the femoral stem [Harris, 1995; Schmalzried et al., 1992]. Therefore, the metal-backed cup which is designed to reduce the stresses in the bonebone cement interface would be ineffective. Some clinical studies showed that the all polyethylene cup performed better than the metal-backed cemented cups [Dall et al., 1993; Harris and Penenberg, 1987; Ritter, 1995; Ritter et al., 1990]. Malchau and Herberts reported revision and re-revision rates in THR in Sweden. The reasons for revision were similar to the Norwegian study (Table 46.4), and are as follows: (1) aseptic loosening, which was the major factor (72.3%) (2) infection (7.2%), (3) fracture (4.7%), and (4) dislocation (4.2%), etc. [Malchau and Herberts, 1998]. The loosening of metal-backed and


TABLE 46.4 Reason for Revision of THR in Sweden (Malchau and Herberts, 1998) Reason

Numbe r

Aseptic loosening Primary deep infection Fracture only Dislocation 2-stage procedure Technical error Implant fracture Secondary infection Pain Polyethylene wear Miscellaneous Missing Total

6965 690 4.7 403 386 372 161 94 37 26 33 13 9634

Percent(%) 72.3 7.2 4.2 4.0 3.9 1.7 1.0 0.4 0.3 0.3 0.1 100.0

all-polyethylene acetabular cups were compared. Mixed results were obtained, where in one design the all-poly was clearly better, while in others both performed equally well. One would be tempted to advocate the use of all-poly since it is a lot less costly (1/2 to 1/3) and equal or better in clinical performance compared to the metal-backed acetabular cup. Basically, the Charnley type all polyethylene cup has been the ‘gold standard’ for long-term clinical results. The metal-backed, either cemented or uncemented polyethylene lined acetabular cup is still considered as experimental until long-term follow-up studies (more than 10 years) become available [Malchau and Herberts, 1998]. Also, the benefits of biological fixation due to the lack of long-term followup studies are inconclusive at this time, although some designs gave promising results. Other variations such as flanged vs. unflanged, elevated vs. conventional cups, etc., should be proven clinically through long-term trials although the flanged cup performed better in one analysis [Wroblewski, 1993]. In our laboratory we have been developing polymethylmethacrylate pre-coated acetabular cups and tibial plateau of the total knee replacement (TKR) for cemented fixation [Dowd et al., 1995; Gruen et al., 1979; Harris and McGann, 1984; Havelin et al.] It is very difficult to adhere bone cement to the surface of ultra high molecular weight polyethylene (UHMWPE) with PMMA bone cement. Therefore, the acetabular cup is deeply grooved to make a mechanical bonding and thus causes a weak implant. The ultra high molecular weight polyethylene (UHMWPE) surface is treated with xylene, polymethylmethacrylate, and a methylmethacrylate monomer for chemical bonding between the two polymers. Preliminary results show that the interface tensile strength between the UHMWPE and bone cement resulted in about 10 MPa which is about one-third of the tensile strength of bone cement and UHMWPE [Kang, 1998; Kang and Park, 1998; Park and Lakes, 1992]. Recent studies on the UHMWPE powders treated with the methylmethacrylate (MMA) monomers without the use of xylene showed a high degree of adhesion after sintering at a high temperature (165°C) and pressure (77.4 MPa) [Park and Park, 1998]. Increased interfacial tensile strength by further treating the MMA treated powders with “PMMA + MMA” and “PMMA + MMA + BPO” is evidenced as shown in Table 46.5. The sintering of UHMWPE with a mold is one way of making an acetabular cup and tibial plateau, therefore, the latest results open the possibility of making a graded composite structure of the UHMWPE implants from pure PE articulating surface to the pure PMMA surface to which the acrylic bone cement can be attached directly via chemical bonding as shown in Fig. 46.2 [Park, 1998]. It is also possible to sinter UHMWPE powers which were cross-linked, which also showed improved wear properties [Gul et al.; Marrs et al., 1998; McKellop et al., 1998; Oonishi et al., 1998]. Glow discharge technique could be used for the high and low density polyethylene if one chooses to use this material for the cup as was the case for the early (before the mid 1980s, HDPE) Charnley prostheses [Foerch et al., 1994; Khang et al., 1996]. However, the interfacial strength is not as high as the present values with the UHMWPE. TABLE 46.5

The Tensile Strength for Molded and Original Specimens (Park, 1998)

Sintered Layers PMMA rod/PMMA powder/treated PE powder/PMMA powder/PMMA rod PE rod/PE powder/treated PE powder/PE powder/PE rod PE rod/pure PE powder/PE rod PMMA rod PE rod


Sintering Temp(°C)

PE Powder Treatments

Group

165

MMA only MMA+MMA/PMMA MMA+MMA/PMMA+BPO

165M 165PM 165PMB

145 165 185 205 205 — —

MMA+MMA/PMMA+BPO MMA+MMA/PMMA+BPO MMA+MMA/PMMA+BPO MMA+MMA/PMMA+BPO — — —

145PMB 165PMB 185PMB 205PMB 205PE PMMA UHMWPE

Interfacial Strength (MPa) 2.1 ± 1.6 10.6 ± 3.6 17.7 ± 2.53 12.4 ± 1.67 14.95 ± 0.85 22.6 ± 1.26 22.8 ± 2.5 32.5 ± 1.20 75.9 ± 1.00 36.9 ± 1.90

FIGURE 46.2 1998].

Schematic diagram showing the graded UHMWPE and PMMA for better fixation and wear [Park,

46.2 Femoral Stem Cemented and Uncemented Fixation The design of the femoral stem also involves the articulating surface with the acetabular cup which could be a more important factor for the longevity of the stem function. The fixation of the stem is largely divided into two categories, i.e., cemented and uncemented. The uncemented can be classified into interference fit and porous-coated for tissue ingrowth fixation. The porous-coated type can be further coated with a hydroxyapatite layer to aid tissue ingrowth. A large study involving 14,009 cases of cemented and 1,326 uncemented total hip arthroplasties were reported [Havelin et al., 1994]. It was found that the cumulative revision rate for the cemented hip was 2.7% after 4.5 years and 6.5% for the uncemented. For the cup, it was 0.6% for the cemented and 1.7% for the uncemented, and for the femoral components 1.7% and 3.9% after 4.5 years. The results for the uncemented prostheses were less favorable in young patients. In men and women under 60 the revision rates were 6 and 3% for the uncemented and cemented respectively. There were large variations in performance among the individual designs of the uncemented prostheses. However, the early results warrant further clinical trials for the uncemented implants. Malchau and Herberts reported revision and re-revision rates in THR in a large Swedish population consisting of 148,359 people. The reasons for revision were similar to the Norwegian study (Table 46.4): (1) aseptic loosening was the major factor (72.3%), (2) infection (7.2%), (3) fracture (4.7%), and (4) dislocation (4.2%), etc. [Malchau and Herberts, 1998]. The survival at 17 years (1979–86) was 81.8% and 9 years (1987–96) was 95.5% for the cemented implants. The survival at 13 years (1979–86) was 68.9% and at 9 years (1987-96) was 87.2% for the uncemented implants [Malchau and Herberts, 1998]. Again the results show a definitively better surgical outcome where cements were used regardless of the various factors involved, such as types of implants, materials, patients, etc. Using less stiff materials and filling the bone cavity with a custom-designed prosthesis has not met researchers’ expectations [Lombardi et al., 1995; Morscher, 1995]. It is believed that the bone reacts to the presence of implants as if they are part of the supporting member of the body weight. Thus, the


filling of the femoral cavity would cause a greatly different behavior of the bone since the prosthesis alters the stress pattern completely. Incorporation of the bony tissue into the interstices of pores would also alter the stress pattern of the femur. This also complicates the situation. Any stiff material contacting the surface of the bone will cause bone density change due to the altered stress pattern and bone lysis due largely to the particulate generated at the articulating surface between the acetabular cup and femoral head [Maloney et al., 1990; Maloney and Smith, 1995]. Most of the osteolysis is located in the distal and proximal end of the prosthesis. The osteolysis can be accelerated by the aseptic loosening of the implants, which in turn causes the accelerated wear of the UHMWPE cup. The initial mode of failure of the femoral prosthesis is the debonding of the stem/bone cement interface followed by osteolysis at the distal end of the stem where the particulate may accumulate by gravity [Jasty et al., 1991]. It is interesting to note that the firm fixation of the porous and hydroxyapatite coated implants cause a great deal of difficulties during revision surgery, sometimes necessitating cutting of the femora [Bhamra et al., 1996]. This result creates the Catch-22 dilemma for the fixation of implants—to have a stable implant, it should be firmly embedded into the intramedullary cavity, but when it fails one should be able to remove it easily.

Possible Solutions for Cemented Fixation There are many studies concerning the loosening of the femoral stem of the hip joint replacement but very few facts have been elucidated with respect to the loosening [Amstutz et al., 1976; Carlsson et al., 1983; Crowninshield et al., 1980; Fowler et al., 1988; Harris, 1992; Hedley et al., 1979; Jasty et al., 1991]. For instance, some believe that the strengthening of bone cement may not be beneficial to the enhancement of fixation due to the unfavorable stress distribution [Crowninshield et al., 1980; Crugnola et al., 1979]. On the other hand, the bone cement is the weakest link between bone and prosthesis. Failure of the bone cement itself and interface loosening between cement and stem may be large contributing factors toward the failure of the fixation of the prosthesis [Amstutz et al., 1976; Carlsson et al., 1983; Harris, 1992; Lu et al., 1993; Wykman et al., 1991]. Recent reports on the failure rate of the Boneloc® bone cement implanted prostheses alarmed the European orthopedic community and it was abandoned [Riegels-Nielsen et al., 1995]. The Boneloc® bone cement was designed to give less exothermic heat of polymerization but its mechanical properties were compromised [Thanner et al., 1995; Wykman et al., 1991]. Cement/Prosthesis Interface Bone cement fixation creates two interfaces: (1) cement/bone and (2) cement-implant. According to an earlier report [Amstutz et al., 1976] the incidence of loosening for the femoral prostheses were evenly divided at about 10 and 11% for cement/bone and cement-implant interfaces, respectively. Jasty et al. reported that the initiation of loosening of the cemented femoral components originates from the prosthesis-cement interface, especially the pores trapped between them [Jasty et al., 1991]. The cementimplant interface loosening can be minimized by “pre-coating” with bone cement or polymethylmethacrylate polymer [Ahmed et al., 1984; Barb et al., 1982; Park, 1983; Park et al., 1982; Park et al., 1978; Raab et al., 1982]. Pre-coating can achieve a good bonding between the “cement” and prosthesis during the manufacturing process. During surgery, the fresh cement adheres well to the pre-coated cement [Park et al., 1978; Park et al., 1979]. Mohler et al. reported early loosening of the femoral component at the cement-prosthesis interface after 1941 THRs performed between 1980 and 1990 [Mohler et al., 1995]. They analyzed 29 failed hips (27 patients). The femoral stems were matte finished; 20 were PMMA-precoated and 9 were not precoated. Initiation of debonding took place in zone 1 of Gruen between cement and prosthesis followed by progressive loosening with maintenance of the bone-cement interface. Extensive osteolysis took place after debonding in many hips (20/29). The average time for revision between the onset of the symptoms after surgery was 9 months. The authors attributed the progressive loosening to the stem geometry (a cylindrical shape, distal to the proximal cobra shape) and matted surface finish. The cylindrical shape


allowed the rotation of the prosthesis, permitting the matte-finished surface to grind the bone cement. This results in loosening and osteolysis. It is also not likely that the PMMA-coating would help to avoid rotation of the prosthesis since it is only applied in the proximal region and the resistance to rotation is greatly compromised by the cylindrical shaped stem. Shear strength of the PMMA bone cement is about 1/3 of the tensile strength. Therefore, if the bone cement is subjected to shear, such as rotation of the stem, the cement will not be able to resist the load. It has been suggested that the stem should be more square with a smooth surface finish. Malchau and Herberts reported inconclusive results on the question of the matte or polished finish on the survival of the implants [Malchau and Herberts, 1998]. It would be reasonable to assume that the matte finish may hold the implants better initially due to the larger surface area but may act to grind the bone cement when the implant becomes loose. The high strains or stresses on the cement at the tip of the prosthesis are strong indications that this location is a likely site for the initial event of failure [O’Connor et al., 1991]. Others reported that high strains at the proximal and distal tip of the prosthesis appeared to cause debonding between the stem and cement and was associated with radial crack initiation at the debonded surface of the pores in the cement surrounding the areas of debonding [Harrigan et al., 1992]. These radial cracks seemed to be due to stress changes secondary to debonding. There has been some discussion on the merits of smooth versus rough finish on the femoral stem surface with regards to the finish’s effect on the longevity of the implant [Harris, 1992; Ling, 1992]. Recent studies show that the effect of the surface roughness may depend on the effectiveness of the load transfer from cement to the stem as shown in Fig. 46.3. The smooth surface will generate local stress peaks in the cement

FIGURE 46.3 Illustration of the Von Mises stress patterns in the cement mantle from smooth to the rough surface finish (right) and corresponding axial load transfer effectiveness [Verdonschot and Huiskes, 1998].


FIGURE 46.4 Illustration of the reinforcement of the bone cement around the prosthesis to resist radial- and hoopstress on the cement mantle by incorporating a wire coil conforming to the contours of the femoral stem which can be placed during surgery or prefabricated during manufacturing of the implant [Kim et al., 1994].

mantle as well as ineffective load transfer. The ineffective (rough) irregular surface which has shallow profiles will be ineffective in axial load transfer and will act as a stress inducer. The effective load transfer profile will have deeper surface irregularity as shown in Fig. 46.3c [Verdonschot and Huiskes, 1998]. The cement/prosthesis interface strength could be enhanced by placing reinforcing material around the stem to resist the radial- and hoop-stress caused by body loading [Park, 1993]. It is also critical that the cement is reinforced to counteract the hoop-stress and to some extent the radial stress created by the prosthesis when it is loaded [Ahmed et al., 1984; Mann et al., 1991]. Therefore, a coil or mesh of wire conforming to the contours of the prosthesis could be placed prior to bone cement injection during surgery as shown in Fig. 46.4. It could also be pre-fabricated around the prosthesis. Also, some advocate the use of a mesh reinforcement around the prosthesis which can be fixed with fresh bone cement at the time of surgery. This mesh reinforcement is similar to the coil but is more difficult to incorporate [Davidson, 1988; Willert et al., 1991]. The simple wire coil would resist the hoop-stress developed by the prosthesis, which in turn will decrease the stress on the bone cement beyond the wire coil toward the bone. Changes to the mechanical property of the bone cement by reinforcing the cement with a wire coil in a simulated stem loading condition has been tried [Kim and Park, 1994] instead of the traditional reinforcement of adding wires, fibers, etc. to the bone cement [Saha and Pal, 1984]. The coils were made from 20 gage 302/304 stainless steel wire and were very effective counteracting the static radial- and hoop-stress as given in Table 46.6. The magnitude of the ultimate load of the control specimens (3.70 ± 1.13 kN) is similar to the ones obtained by Mann et al. (~32 kN), whose specimen was 8 times longer [Mann et al., 1991]. If one extrapolates the ultimate load of the wire reinforced specimens to that of Mann’s work (which is closer to the in vivo condition) the load to fracture would be about 72 kN; well above the normal loading on the prosthesis even in dynamic loading (6~8 times body weight; 4 kN) [Paul, 1976]. The percentage increases of improvements in mechanical properties are similar or higher TABLE 46.6

Summary of the Static Mechanical Property Measurements

Specimen Type

Maximum Strain (%)

Maximum Load (kN)

Stiffness (GNm/m)

Toughness (Nm/m)

Control Wire reinforced % Change

2.6 ± 0.5 3.2 ± 0.9 +22.7

3.70 ± 1.13 7.68 ± 2.33 +107.6

1.73 ± 0.39 4.17 ± 1.16 +141.0

49.33 ± 18.63 129.48 ± 65.3 +162.5

Source: Kim and Park, 1994.


than other studies although the increases are not as high as predicted by the composite theory [Taitsman and Saha, 1977; Topoleski et al., 1990]. The fatigue test showed a substantial increase in its cyclic load to failure by the wire coil specimens [Kim and Park, 1996]. It is believed that the beneficial effects of the wire (fiber) layer lies in its ability to inhibit the propagation of fatigue cracks [Manley and Stern, 1985; Manley et al., 1987]. At high stress levels in the cement mantle, fatigue cracks could initiate from defects such as air bubbles and surface discontinuities, and propagate gradually over a large number of load cycles. The presence of the fiber or wire coil increases the energy necessary for crack propagation by crack tip blunting, crack deflection, and the bridging of cracks by intact fibers [Taylor, 1987; Taylor et al., 1989]. Another observation gained from fatigue testing was a subsidence of the stem. The subsidence is due to the creep of the viscoelastic materials such as bone and bone cement [Park and Lakes, 1992]. The net effect from the creep would be a continued subsidence of the implant shortening the leg [Kim and Park, 1996; Manley et al., 1987]. All measured mechanical properties show that the wire coil reinforcement significantly enhanced the strength, fracture strain, stiffness, and toughness over the control [Kim and Park, 1996]. The most significant increases were in the toughness, indicating that the coil reinforced cement mantle will resist the load much more than the non-reinforced control. This may result in a prolonged fixation life of the prosthesis, which is the ultimate goal of all arthroplasties. Cement-Bone Interface The problems at the bone-cement interface cannot be easily overcome since these problems arise from the intrinsic properties of the bone cement as well as extrinsic factors such as cementing technique, condition of the bone surface, etc. The toxicity of the monomer, inherent weakness of the cement as a material due to the unavoidable inclusion of the pores (Table 46.7), and blood and tissue debris mixed during surgery can contribute to the problem of loosening at the bone-cement interface [Park, 1983]. The bone-cement interface strength may be enhanced by growing bone into the cement after fixation. Bone cement can be used for immediate fixation yet provides tissue ingrowth space later by incorporating resorbable particles (such as inorganic bone particles) as shown conceptually in Fig. 46.5. Some studies [Dai et al., 1991; Henrich et al., 1993; Liu et al., 1987; Park et al., 1986] indicate that the concept can be used effectively, at least in rabbit and canine models [Kwon et al., 1997]. In one study the bone-particle impregnated bone cement was used to fix femoral stem prostheses in the femora of dogs (one side was experimental, the other was control) and after a pre-determined time period, the femora were harvested and sectioned into disks for the push-out test to measure the interfacial strength between the bone and cement interface [Dai et al., 1991]. The results as shown in Fig. 46.6 indicate that the experimental side increased in strength for up to five months while the control side decreased slightly. The histology also showed the integration of tissues into the spaces of the dissolved particles. Of course, care should be taken to control the amount of particles to balance out the increased viscosity and decreased strength. Table 46.7 illustrates the relationship between the amount of resorbable particle (inorganic bone particles, TABLE 46.7 Effect of Inorganic Bone Particle Mixed with Acrylic Bone Cement on the Pore Size, Porosity, and Tensile Strength Particle Amount (%)

Pore Size (µm)

Porosity (%)

Tensile Strength (MPa)

0 10 20 30

154.7 ± 72 160.6 ± 68 172.9 ± 52* 218.0 ± 92*

7.2 ± 2.5 5.0 ± 1.7* 4.9 ± 1.5* 2.4 ± 0.7*

23.2 ± 2.3 22.6 ± 2.0* 20.1 ± 1.1* 19.7 ± 1.1*

Note: 15 specimens for each group. * Statistically significant vs. control (p < 0.05). Source: Kim et al., 1994.


FIGURE 46.5 Basic concept of the bone cement with resorbable particle fixation. Immediate fixation is achieved as in the ordinary bone cement with the enhanced biological fixation to be achieved later for a prolonged stabilization of the implant [Park, 1992 and 1995].

FIGURE 46.6 Maximum interfacial shear strength between bone and bone cement versus implant period when the femoral stems were implanted with ordinary bone cement and cement with bone particles. In both cases the interfacial strength was stabilized after 5 months for this canine model [Dai et al., 1991].

100–300 µm) and the porosity and pore size [Kim et al., 1994]. As more particles are incorporated onto the bone cement, a lesser amount of porosity resulted, however, the average pore size increased. The tensile strength decreased with an increased amount of bone particles. Fatigue properties improved with higher particle inclusion, however, it was found that about 30% (by weight) of bone particles can provide sufficient interconnected porosity for bony ingrowth and yet give reasonable compromises to other parameters [Liu et al., 1987]. The bone-particle impregnated bone cement has been used clinically but its long-term success remains to be proven since it has been only five years without any failures [Dai, 1991 and 1996].


Uncemented Fixation Efforts to develop a viable interface between the tissue and implants have been made ever since Moore designed a femoral prosthesis which had a fenestrated large hole in the proximal region [Moore, 1952]. Ironically, the fixation itself is by the passive mechanical fixation technique as discussed earlier. Smith [Smith, 1963] tried to develop a bone substitute with porous aluminate ceramic impregnated with an epoxy resin called Cerocium®. Although the material demonstrated a good adherence to the tissues, the pore size (average 18 µm diameter) was too small to allow any bony tissue ingrowth. Later, ceramics [Klawitter and Hulbert, 1972; Oonishi et al., 1989; Predecki et al., 1972], metals [Hirshhorn et al., 1972; Niles and Lapitsky, 1975], and polymers [Sauer et al., 1974] were used to test the ingrowth idea. Basically, any biocompatible material will allow bony tissues to grow into any space large enough to accommodate osteons. However, to be continuously viable the space must be large enough (more than 75 µm in diameter for the bony tissues) and contiguous to the surface of bone matrix [Heimke, 1990]. In addition, Wolff ’s law dictates that the ingrown tissues should be subjected to bodily loading in order to prevent resorption, even after the initial ingrowth has taken place. The same principle also makes it difficult to have a uniform tissue ingrowth throughout the implant surface. This is the reason why the tissue ingrowth takes place where the stress transfer occurs, e.g., in the distal lateral region of the femoral stem of the artificial hip joint. Figure 46.7 shows the general trends of the fixation (interfacial) strength variation with implantation period in animals to be up to six months for metallic implants (Co–Cr alloys, Ti and its alloys, and stainless steel). A few observations can be made from this data. First, the maximum interfacial shear strength between bone and porous implant peaked at about 12 to 13 MPa in about 6 to 8 weeks regardless of the implant material, animal type, or location of the implants (i.e., cortical or cancellous bone and transcortical or intramedullary canal). Second, the data are widely scattered. (The original data showed wide variations, therefore, they are not distinguished among the variables). The decrease in interfacial strength with time after the peak for the metallic implants is somewhat distressing since this may also be true with human patients. This may be caused by two factors. The first is that of bone resorption from the initially ingrown area due to the stress shielding. The second is the release of a large number of metal ions due to the increased surface area by making the implant porous. Porous polymers did not

FIGURE 46.7 Maximum interfacial shear strength between bone and porous metal implants versus implant of various metals and animals, data from [Spector, 1982]. © 2000 by CRC Press LLC

FIGURE 46.8 Maximum interfacial shear strength between bone and porous polymeric implants versus implant. [Data from Spector, 1982].

have the decreasing trends as shown in Fig. 46.8 [Spector, 1982]. This may be interpreted as the lack of stress shielding (after bone is ingrown into the pores) due to the low modulus of the polymers. However, there are not enough data available to come to a definitive conclusion [Keet and Runne, 1989; Sadr and Arden, 1987; Spector, 1982; Tullos et al., 1984]. Further problems with biological fixation are: (1) the unforgiving nature of the surgery, (2) the long immobilization time for tissue ingrowth, (3) the unpredictable time to ambulate, (4) the difficulty in eradicating infection, and (5) once the interface is destroyed by an accidental overloading, it cannot be regrown with certainty. Moreover, the porous coating may weaken the underlying prosthesis itself and, in case of metals, there is an increased danger of corrosion fatigue due to the increased surface area [Brand, 1987; Morscher, 1984]. Due to these problems and clinical results of porous fixation some investigators have insisted that the bone cement fixation technique is still the gold standard at this time [Johnston, 1987; Wroblewski, 1993]. In order to alleviate these problems several modifications have been tried. Pre-Coating the Metallic Porous Surface with Ceramics or Carbons This method has been tried with limited success [Cranin et al., 1972; Thomas et al., 1985]. The problem of coating deep in the pores and the thermal expansion difference between the metal and ceramic materials, both make a uniform and good adherent coating very difficult. An attractive material for coating is hydroxyapatite [Ca10(PO4)6(OH)2] ceramic which is similar to bone mineral. It is not yet conclusive that this material is more beneficial than other ceramics. Some preliminary studies in our laboratory indicate that during the early period of fixation, the bioactive hydroxyapatite coating may be more beneficial, but the effect may diminish after three weeks as shown Fig. 46.9. The result is for the simple cortical bone plug experiment performed on canine femora [Park, 1988]. Others have shown promising results with the same material system used on rabbits instead of canines as in our study [Oonishi et al., 1989]. Pre-Coating with Porous Polymeric Materials on Metal Stem Theoretically, this method has two distinct advantages over the ceramics or carbon coating methods discussed above [Keet and Runne, 1989; Sadr and Arden, 1987; Spector, 1982; Tullos et al., 1984; Homsy et al., 1972]. First, the low modulus polymeric material could transfer the load from the implant to the


FIGURE 46.9 Maximum interfacial shear strength between bone and bioactive ceramic coated porous plug implants versus implant period when the plugs with and without (control) coating were implanted in the cortices of canine femora [Park, 1988].

bone more gradually and evenly and thus, prevent stress-shielding effects on the ingrown tissues. Second, this method would prevent the metallic-surface ion release, i.e., less corrosion of the metal. One major problem with this technique is the weak interfacial strength between the polymer and metallic stem, especially in dynamic loading conditions in vivo. Enhancement of Porous Ingrowth with Electrical or Electromagnetic Stimulation This technique combines the porous ingrowth with the stimulating effects of electrical and/or electromagnetic stimulation [Park and Kenner, 1975; Weinstein et al., 1976]. Direct current stimulation can indeed accelerate tissue ingrowth in the early stages of healing but its effect diminishes over time [Weinstein et al., 1976]. Direct current stimulation has one distinct problem: the invasive nature of the power source. The pulsed electromagnetic field stimulation is a better method since the stimulation can be carried out extracorporeally. Porous Ingrowth with the Use of Filling Materials There has been some effort to use filling materials around the porous implant since it is very difficult to prepare the tissue bed with the exact shape of the prosthesis to eliminate the micro-motion of the prosthesis after implantation. Bone matrix proteins (BMP) or demineralized bone and hydroxyapatite crystals can be used for this purpose [Parsons et al., 1987]. The success of this technique has not been fully documented. Some glass-ceramics tend to achieve direct bonding with the bone through a selective dissolution of the surface film [Blencke et al., 1978; Hench and Paschall, 1973; Yamamuro et al., 1990]. Some researchers have reported a direct bonding of tissues to hydroxyapatites [Geesink et al., 1987; Kay, 1988]. It has been difficult to coat a metallic surface with these materials due to the large difference in the thermal expansion coefficient between the coating and base materials although dip-coating can be utilized [West et al., 1990]. The glass-ceramics have not been used to make load bearing implants due to their brittleness.

46.3 Articulating Surface of the Acetabular Cup and Femoral Head Any articulating joint will involve the friction and wear of the two opposing members. Ever since the total joint replacement has been practiced this subject gave the orthopedic community pause and posed a considerable problem to the biomedical engineers due to the fact that any material or design can avoid this subject. The lowest friction coefficient material such as polytetrafluoroethylene (PTFE), best known


as Teflon®, is not sufficient as a member of a joint as it was proved to be disastrous by Dr. J. Charnley. Earlier than Charnley, the all metal-metal combinations were used but abandoned. Some researchers advocate it again [Bohler, 1995; Muller, 1995] claiming that smaller head design resulted in drastically reduced wear after 10 years of implantation (60, 1330, and 2100 µm for CoCrMo/CoCrMo, Al2O3/UHMWPE, and CoCrNo/UHMWPE, respectively). Saikko reported wear of the UHMWPE cup and five different Al2O3 ceramic head combinations and found a vastly different wear volume with a different femoral head design and stem fit [Saikko, 1995]. The head on these implants was attached to the Ti-alloy stem by taper-fit. The wear particles from the ceramic and head resulted in severe third-body abrasive wear (3170 mg) compared to minimal wear of others (5.3 to 124 mg) after 26 million cycles. Obviously there was a loosening between the head and the stem collar of the femoral implant for the worst case (Protek®) since others had similar taper-fit. In this respect, the trunnion design of the head and stem should be avoided at all possible costs to eliminate the possibility of third-body wear. Modularity of the head of the femoral stem is sound in principle since it could reduce inventory while providing the surgeons with versatility. The full benefits of such flexibility have not been realized in actual usage. Third body wear of the acetabular cup, increased chance of corrosion, and less than optimal design (although the optimal design remains elusive) contribute to the problem of the modularity [Chmell et al., 1995; Jacobs et al., 1995; Salvati et al., 1995]. Similar problems of third-body wear exist for the cemented polyethylene acetabular cup and metalbaked (porous coated) cup. PMMA bone cement particles were found in autopsies for the cemented cups. Ritter [Ritter, 1995] analyzed uncemented metal-backed vs. a cemented polyethylene cups and found statistically significant better clinical performance of the cemented cup in terms of the linear wear (0.05 vs. 0.11 mm/yr). There were also differences in the wear between machined and molded cups, indicating that the molding resulted in better polyethylene powder consolidation than the raw material used in the machine from rod stocks. It is also now widely recognized that gamma radiation causes the polyethylene to lose its original chemical and mechanical properties due to the oxidation about 1 mm below the surface after about three years of implantation [Li et al., 1995; Sutula et al., 1995]. This increases the density by crystallization but causes brittleness of the material. Increased localized wear could result. Inert or vacuum sterilization may minimize the problem unless the sterilization method is changed.

Conclusions In summary, the wear of the articulating surface in a complex phenomenon like THR could be minimized by having the most consolidated polyethylene (molded) as a cup material with a small femoral head (either ceramic or metal) cemented with bone cement. Although some advocate the use of “hybrid” fixation techniques, i.e., cemented stem and uncemented cup, the final verdict is not yet in. A metalbacked polyethylene lined cup would cause more problems, either cemented or uncemented due to the increased complexity and decreased shock absorbing capacity of the cup during walking. The same is true with modularity of the femoral stem and head. It is strongly recommended that all nations follow the example of the Norwegian Arthroplasty Register in obtaining meaningful outcome data of all implantations. Thus, the clinical performance of a particular implant could be determined and disasters like the Boneloc® bone cement and Christianson prosthesis could be avoided more quickly [Linder, 1995]. Orthopedic surgeons often suggest that the fixation of implants should not be too strong since it would be difficult to remove, should it become necessary to have a revision arthroplasty to restore its function. The two requirements could be mutually exclusive since the interface must be strong enough to hold the prosthesis in vivo but not so strong that it may impair any remedial measures. The assessment of success or failure of the prosthesis fixation could be quite difficult clinically as demonstrated by some researchers. For example, some [Ling, 1992] contend that the relative motion between the implant and bone cement and subsidence of the prosthesis are not a direct cause of failure of arthroplasties while others consider it a failure [Harris, 1992].


Another problem is related to the more frequent use of the implants for younger patients, which requires firmer fixation and longer implant life [Park et al., 1995]. The original expectation of uncemented porous ingrowth fixation for the young is tempered somewhat and we need to explore a longer lasting method of fixation.

Acknowledgments This work is partially funded by the Biomaterials Research Fund, and the University of Iowa Research Foundation. A major part of the paper was presented at the International Symposium on Biomaterials and Drug Delivery Systems held at the Korea Research Institute of Chemical Technology in Taejon, Korea, July 4–5, 1996.

References Ahmed, A. M., Raab, S., and Miller, J. E. (1984). Metal/cement interface strength in cemented stem fixation. J. Orthop. Res. 2, 105–118. Amstutz, H. C., Markolf, K. L., McNeices, G. M., and Gruen, T. A. W. (1976). Loosening of total hip components: Cause and prevention. In: The Hip (St. Louis: Mosby), pp. 102–116. Anthony, P. P., Gie, G. A., Howie, C. R., and Ling, R. S. M. (1990). Localised endosteal bone lysis in relation to the femoral components of cemented total hip arthroplasties. J. Bone Joint Surg. 72B, 971–979. Barb, W., Park, J. B., von Recum, A. F., and Kenner, G. H. (1982). Intramedullary fixation of artificial hip joints with bone cement precoated implants: I. Interfacial strengths. J. Biomed. Mater. Res. 16, 447–458. Bhamra, M., Rao, G. S., and Robson, M. (1996). Hydroxyapatite-coated hip prostheses: Difficulties with revision in four cases. Acta Orthop. Scand. 67, 49–52. Blencke, B. A., Bromer, H., and Deutscher, K. K. (1978). Compatibility and long-term stability of glass ceramic implants. J. Biomed. Mater. Res. 12, 307–316. Bohler, N. (1995). Metal/metal articulating interfaces. Orthoped. 18, 879–880. Brand, R. A. (1987). The Hip: Non-Cemented Hip Implants (St. Louis: C.V. Mosby Co). Carlsson, A. S., Gentz, C. F., and Lindberg, H. O. (1983). Thirty-two noninfected total hip arthroplasties revised due to stem loosening. Clin. Orthop. Rel. Res. 181, 196–203. Charnley, J. (1970). Acrylic Cement in Orthopaedic Surgery (Edinburgh: Churchill/Livingstone). Charnley, J. (1972). The long-term results of low-friction arthroplasty of the hip, performed as a primary intervention. J. Bone Joint Surg. 54B, 61–76. Charnley, J. and Cupic, Z. (1973). The nine and ten year results of the low-friction arthroplasty of the hip. Clin. Orthop. Rel. Res. 95, 9–25. Chmell, M. J., Rispler, D., and Poss, R. (1995). The impact of modularity in total hip arthroplasty. Clin. Orthop. Rel. Res. 319, 77–84. Cranin, A. N., Schnitman, P. A., Rabkin, M., Dennison, T., and Onesto, E. J. (1972). Alumina and zirconia coated Vitallium oral endosteal implants. J. Biomed. Mater. Res. Symposium No. 6, 257–262. Crowninshield, R. (1988). An overview of prosthetic materials for fixation. Clin. Orthop. Rel. Res. 235, 166–172. Crowninshield, R. D., Brand, R. A., Johnston, R. C., and Milroy, J. C. (1980). An analysis of femoral component stem design in total hip arthroplasty. J. Bone Joint Surg. 62A, 68–78. Crugnola, A., Ellis, E. J., Radin, E. L., and Rose, R. M. (1979). A second generation of acrylic cements. Trans. Soc. Biomat. 3, 91. Dai, K. R. (1991 and 1996). Personal communications. Dai, K. R., Liu, Y. K., Park, J. B., and Zhang, Z. K. (1991). Bone particle impregnated bone cement: An in vivo weight-bearing study. J. Biomed. Mater. Res. 25, 141–156. Dall, D. M., Learmonth, I. D., Solomon, M., and Davenport, M. (1993). 811 Charnley hips followed for 3–17 years. Acta Orthop. Scand. 64, 252–256. © 2000 by CRC Press LLC

Davidson, J. A. (1988). Bone cement reinforcement and method. In Patent # 4,735,625 (USA). de Groot, K. (1983). Bioceramics of Calcium Phosphate (Boca Raton, FL: CRC Press). Dowd, J. E., Schwendeman, L., Macaulay, W., Doyle, J. S., Shanbhag, A. S., Wilson, S., Herndon, J. H., and Rubash, H. E. (1995). Aseptic loosening in uncemented total hip arthroplasty in a canine model. Clin. Orthop. Rel. Res. 319, 106–121. Ducheyne, P. (1988). Prosthesis fixation for orthopedics. In: Encyclopedia of Medical Devices and Instrumentation, J. G. Webster, Ed. (N.Y.: Wiley-Interscience), 2146–2154. Eftekhar, N. S. (1978). Principles of Total Hip Arthroplasty (St. Louis: Mosby). Espehaug, B., Havelin, L. I., Engesaeter, L. B., Vollset, S. E., and Langeland, N. (1995). Early revision among 12,179 prostheses: A comparison of ten different brands reported to the Norwegian arthroplasty register, 1987–1993. Acta Orthop. Scand. 66, 487–493. Foerch, R., Kill, G., and Walzak, M. (1994). Plasma Surface Modification of Polymers. In: Plasma Surface Modification of Polyethylene: Short-Term vs. Long-Term Plasma Treatment, M. Strobel, C. Lyons and K. L. Mittal, Eds. (Utrecht, Netherlands: VSP Press), 99–111. Fowler, J. L., Gie, G. A., Lee, A. J. C., and Ling, R. S. M. (1988). Experience with Exeter total hip replacement since 1970. Orthop. Clin. N. Am. 19, 477–489. Geesink, R. G., de Groot, K., and Klein, C. (1987). Chemical implant fixation using hydroxyaptite coatings. Clinic. Orthop. Rel. Res. 226, 147–170. Gruen, T. S., McNeice, G. M., and Amstutz, H. A. (1979). “Modes of failure” of cemented stem-type femoral components. Clin. Orthop. Rel. Res. 141, 17–27. Gul, R. M., Muratoglu, O. K., McGarry, F. J., Bragdon, C. R., Jasty, M., and Harris, W. H. (1998). The effect of the peroxide content on the cross-link density, mechanical properties, and wear behavior of UHMWPE. In: Orthop Res. Soc. (New Orleans, LA), 99. Harrigan, T. P., Kareh, J. A., O’Connor, D. O., Burke, D. W., and Harris, W. H. (1992). A finite element study of the initiation of failure of fixation in cemented femoral total hip components. J. Orthop. Res. 10, 134–144. Harris, W. H. (1992). Is it advantageous to strengthen the cement-metal interface and use a collar for cemented femoral components of total hip replacements? Clin. Orthop. Rel. Res. 285, 67–72. Harris, W. H. (1995). The problem is osteolysis. Clin. Orthop. Rel. Res. 311, 46–53. Harris, W. H. and McGann, W. A. (1984). Loosening of the femoral component after use of the medullaryplug cementing technique. J. Bone Joint Surg. 67B, 222–224. Harris, W. H. and Penenberg, B. L. (1987). Further follow-up on socket fixation using a metal-backed acetabular component for total hip replacement. A minimum ten year follow-up study. J. Bone Joint Surg. 69A, 1140–1143. Havelin, L. I., Espehaug, B., Vollset, S. E., and Engesaeter, L. B. (1994). Early failures among 14,009 cemented and 1,326 uncemented prostheses for primary coarthrosis: The Norwegian Arthroplasty Register, 1987–1992. Acta Orthop Scand 65, 1–6. Havelin, L. I., Vollset, S. E., and Engesaeter, L. B. (1995). Revision for aseptic loosening of uncemented cups in 4,352 primary total hip prostheses: A report from the Norwegian arthroplasty register, 1987–1993. Acta Orthop. Scand. 66, 494–500. Hedley, A. K., Moreland, J., Bloebaum, R. D., Coster, I., Gruen, T., and Clarke, I. (1979). Press-fit, cemented, and bone ingrowth surface replacement-canine fixation model. Trans. Orthop. Res. Soc. 4, 163. Heimke, G. E. (1990). Osseo-Integrated Implants, vol. I & II (Boca Raton, FL: CRC Press). Hench, L. L. and Paschall, H. A. (1973). Direct chemical bond of bioactive glass-ceramic materials to bone and muscle. J. Biomed. Mater. Res. Symposium No. 4, 25–43. Henrich, D. E., Cram, A. E., Park, J. B., Liu, Y. K., and Reddi, H. (1993). Inorganic bone and bone morphogenetic protein impregnated bone cement: A preliminary in vivo study. J. Biomed. Mater. Res. 27, 277–280. Hirshhorn, J. S., McBeath, A. A., and Dustoor, M. R. (1972). Porous titanium surgical implant materials. J. Biomed. Mater. Res. Symposium No. 2, 49–69.


Homsy, C. A., Cain, T. E., Kessler, F. B., Anderson, M. S., and King, J. W. (1972). Porous implant systems for prosthetic stabilization. Clinic. Orthop. Rel. Res. 89, 220–231. Jacobs, J. J., Urban, R. M., Gilbert, J. L., Skipor, A. K., Jasty, M., Black, J., and Galante, J. O. (1995). Local and distant products from modularity. Clin. Orthop. Rel. Res. 319, 94–105. Jasty, M., Maloney, W. J., Bragdon, C. R., O’Connor, D. O., Haire, T., and Harris, W. H. (1991). The initiation of failure in cemented femoral components of hip arthroplasties. J. Bone Joint Surg. 73B, 551–558. Johnston, R. C. (1987). The case for cemented hips. In The Hip, R. A. Brand, Ed. (St. Louis: C.V. Mosby Co.), 351–358. Kang, Y. H. (1998). Pre-coating of Ultra-High Molecular Weight Polyethylene (UHMWPE) with Polymethylmethacrylate (PMMA). In: Biomedical Engineering (Iowa City: The University of Iowa), 1–160. Kang, Y. H. and Park, J. B. (1998). Interfacial Tensile Strength of PMMA Pre-coated UHMWPE and Bone Cement. J. Biomed. Mat. Res. (Appl. Biomat.) 43, 261–269. Kay, J. F. (1988). Bioactive surface coatings: Cause for encouragement and caution. J. Oral Implant 16, 43–54. Keet, G. G. M., and Runne, W. C. (1989). The anaform endoprosthesis: A proplast-coated femoral endoprosthesis. Orthop. 12, 1185–1190. Khang, G., Kang, Y., Lee, H., and Park, J. (1996). Improved bonding strength of polyethylene/polymethylmethacrylate bone cement: A preliminary study. Biomed. Mater. Eng. 6, 335–344. Kim, J. K. and Park, J. B. (1996). Fatigue properties and stem subsidence of wire coil reinforced PMMA bone cement: An in vitro study. Biomed. Mater. Eng. 6, 453–462. Kim, J. K. and Park, J. B. (1994). Reinforcement of bone cement around prostheses by precoated wire coil: A preliminary study. Biomed. Mater. Eng. 4, 369–380. Kim, J. K. and Park, J. B. (1996). Reinforcement of bone cement around the femoral prosthesis tip by pre-coated wire coil: A human cadaver bone study. Biomed. Mater. Eng. 6, 159–164. Kim, Y. S., Kang, Y. H., Kim, J. K., and Park, J. B. (1994). Effect of bone mineral particles on the porosity of bone cement. Biomed. Mater. Eng. 4, 1–10. Klawitter, J. J., and Hulbert, S. F. (1972). Application of porous ceramics for the attachment of load bearing internal orthopedic applications. J. Biomed. Mater. Res. Symp. No. 2, 161–229. Kwon, S. Y., Kim, Y. S., Woo, Y. K., Kim, S. S., and Park, J. B. (1997). Hydroxyapatite impregnated bone cement: In vitro and in vivo studies. Biomed. Mar. Eng. 7, 129–140. Li, S., Chang, J. D., Barrena, E. G., Furman, B. D., Wright, T. M., and Salvati, E. (1995). Nonconsolidated polyethylene particles and oxidation in Charnley acetabular cups. Clin. Orthop. Rel. Res. 319, 54–63. Linder, L. (1995). Boneloc®—the Christiansen experience revisited, Editorial. Acta Orthop. Scand. 66, 205–206. Ling, R. S. M. (1992). The use of a collar and precoating on cemented femoral stems is unnecessary and detrimental. Clin. Orthop. Rel. Res. 285, 73–83. Liu, Y. K., Njus, G. O., Park, J. B., and Stienstra, D. (1987). Bone particle impregnated bone cement I. In vitro study. J. Biomed. Mater. Res. 21, 247–261. Lombardi, A. V., Mallory, T. H., Eberle, R. W., Mitchell, M. B., Lefkowitt, M. S., and Williams, J. R. (1995). Failure of intraoperatively customized non-porous femoral components inserted without cement in total hip arthroplasty. J. Bone Joint Surg. 77A, 1836–1844. Lu, Z., Ebramzadeh, E., and Sarmiento, A. (1993). The effect of failure of the cement interfaces on gross loosening of cemented total hip femoral components. Trans. Orthop. Res. Soc. 18, 519. Malchau, H. and Herberts, P. (1998). Prognosis of total hip replacement: Revision and re-revision rate in IHR: A revision-risk study of 148,359 primary operations. In 65th Annual Ame. Acad. Orthop. Surg. (New Orleans, LA). Maloney, W. J., Jasty, M., Rosenberg, A., and Harris, W. H. (1990). Bone lysis in well-fixed cemented femoral components. J. Bone Joint Surg. 72B, 966–970. Maloney, W. J. and Smith, R. L. (1995). Periprosthetic osteolysis in total hip arthroplasty: The role of particulate wear debris. J. Bone Joint Surg. 77A, 1448–1460. © 2000 by CRC Press LLC

Manley, M. T. and Stern, L. S. (1985). The load carrying and fatigue properties of the stem-cement interface with smooth and porous coated femoral components. J. Biomed. Mater. Res. 19, 563–575. Manley, M. T., Stern, L. S., Kotzar, G., and Stulberg, B. N. (1987). Femoral component loosening in hip arthroplasty. Acta Orthop. Scand. 58, 485–490. Mann, K. A., Bartel, D. L., Wright, T. M., and Ingraffea, A. R. (1991). Mechanical characteristics of stemcement interface. J. Orthop. Res., 798–808. Marrs, H., Barton, D. C., Ward, I. M., Doyle, C., Fisher, J., and Jones, R. A. (1998). Comparative wear under three different tribological conditions of acetylene crosslinked ultra high molecular weight polyethylene. In: Orthop Res. Soc. (New Orleans, LA), 100. McKellop, H. A., Shen, F. W., and Salovey, R. (1998). Extremely low wear of gamma crosslinked/remelted UHMW polyethylene acetabular cups. In: Orthop Res. Soc. (New Orleans, LA), 98. Mittlemeier, H. (1976). Anchoring hip endoprosthesis without bone cement. In: Advances in Artificial Hip and Knee Joint Technology, M. Schaldach and D. Hohmann, Eds. (Berlin: Springer-Verlag), 387–402. Mittlemeier, H. (1975). New development of wear-resistant ceramic and metal composite prostheses with ribbed support shafts for cement-free implantation. Hefte zur Unfallheilkunde: Beihefe zue Monatsschrift fur Unfallheilkunde. Verischerings, Versorgungs und Verkehrsmedizin 126, 333–336. Mohler, C. G., Callaghan, J. J., Collis, D. K., and Johnston, R. C. (1995). Early loosening of the femoral component at cement-prosthesis interface after total hip replacement. J. Bone Joint Surg. 77A, 1315–1322. Moore, A. T. (1952). Metal hip joint: A new self-locking Vitallium prosthesis. South. Med. J. 45, 10–15. Morscher, E. (1984). The Cementless Fixation of Hip Endoprosthesis (Heidelberg: Springer-Verlag). Morscher, E. W. (1995). The clinical significance of implant stiffness. Orthoped. 18, 795–800. Muller, M. E. (1995). The benefits of metal-on-metal total hip replacements. Clin. Orthop. Rel. Res. 311, 54–59. Mulroy, R. D. and Harris, W. H. (1990). The effect of improved cementing techniques on component loosening in total hip replacement: An eleven year radiographic review. J. Bone Joint Surg. 72B, 757–760. Niles, J. L. and Lapitsky, M. (1975). Biomechanical investigations of bone-porous carbon and metal interfaces. Biomed. Mater. Res. Symposium 4, 63–84. Nilsen, A. R. and Wiig, M. (1996). Total hip arthroplasty with Boneloc®: Loosening in 102/157 cases after 0.5–3 years. Acta Orthop. Scand. 67, 57–59. O’Connor, D. O., Burke, D. W., Sedlacek, R. C., and Harris, W. H. (1991). Peak cement strains in cemented femoral total hip. Trans. Orthop. Res. Soc. 16, 220. Onsten, I., Sanzen, L., Carlsson, A., and Besjakov, J. (1995). Migration of uncemented, long-term femoral components in revision hip arthroplasty: A 2–8 year clinical follow-up of 45 cases and radiostereometric analysis of 13 cases. Acta Orthop. Scand. 66, 220–224. Oonishi, H., Saito, M., and Kadoya, Y. (1998). Wear of high-dose gamma irradiated polyethylene in total joint replacement. Long-term radiological evaluation. In: Orthop Res. Soc. (New Orleans, LA), 97. Oonishi, H., Yamamoto, M., Ishimaru, H., Tsuji, E., Kushitani, S., Aono, M., and Ukon, Y. (1989). Comparisons of bone ingrowth into porous Ti-6Al-4V beads uncoated and coated with hydroxyapatite. In: Bioceramics, H. Ooonoishi, H. Aoki and K. Sawai, Eds. (Tokyo and St. Louis: Ishiyaku EuroAmerica, Inc.), 400–405. Park, J. B. (1983). Acrylic bone cement: In vitro and in vivo property-structure relationship—A selective review. Annals Biomed. Eng. 11, 297–312. Park, J. B. (1983). Implant fixation by pulsed electromagnetic field stimulation. Unpublished study (Iowa City, IA: The University of Iowa). Park, J. B. (1988). In vivo evaluation of hydroxyapatite coated implant. Unpublished study (Iowa City: The University of Iowa). Park, J. B. (1992). Orthopedic Prosthesis Fixation. Annals Biomed. Eng. 20, 583–594. Park, J. B. (1995). Orthopedic Prosthesis Fixation. In: Biomedical Engineering Handbook, J. D. Bronzino, Ed. (Boca Raton, FL: CRC Press), 704–723.


Park, J. B. (1993). Reinforcement of bone cement around prosthesis by pre-coated wire coil/wire mesh. In Patent Disclosure. Park, J. B. (1998). Sintering of pure and cross-linked UHMWPE, MMA, and PMMA treated powders to create graded total joint implants for better fixation and wear. In Patent Disclosure (USA). Park, J. B., Barb, W., Kenner, G. H., and von Recum, A. F. (1982). Intramedullary fixation of artificial hip joints with bone cement precoated implants. II. Density and histological study. J. Biomed. Mater. Res. 16, 459–469. Park, J. B., Choi, W. W., Liu, Y. K., and Haugen, T. W. (1986). Bone particle impregnated polymethylmethacrylate: In vitro and in vivo study. In: Tissue Integration in Oral and Facial Reconstruction, D. Van Steenberghe, Ed. (Amsterdam: Excerptu Medica), 118–124. Park, J. B. and Kenner, G. H. (1975). Effect of electrical stimulation on the tensile strength of the porous implant and bone interface. Biomater. Med. Dev. Artif. Org. 3, 233–243. Park, J. B. and Lakes, R. S. (1992). Biomaterials: An Introduction: 2nd Edition (N.Y.: Plenum Pub.). Park, J. B., Malstrom, C. S., and von Recum, A. F. (1978). Intramedullary fixation of implants pre-coated with bone cement: A preliminary study. Biomater. Med. Dev. Artif. Org. 6, 361–373. Park, J. B., von Recum, A. F., and Gratzick, G. E. (1979). Pre-coated orthopedic implants with bone cement. Med. Dev. Artif. Org. 7, 41–53. Park, K. D. and Park, J. B. (1998). Pre-coating UHMWPE powers with PMMA and sintering (Iowa City: The University of Iowa). Park, S.-H., Llinas, A., and Goel, V. K. (1995). Hard tissue replacement. In: Biomedical Engineering Handbook, J. D. Bronzino, Ed. (Boca Raton, FL: CRC Press), 672–691. Parsons, J. R., Ricci, J. L., Liebrecht, P., Salsbury, R. L., Patras, A. S., and Alexander, H. (1987). Enhanced stabilization of orthopedic implants with spherical hydroxylapatite particulate (Dublin, CA: OrthoMatrix, Inc.). Paul, J. P. (1976). Loading on normal hip and knee joints and joint replacements. In: Advances in Hip and Knee Joint Technology, M. Schaldach and D. Hohmann, Eds. (Berlin: Springer-Verlag), 53–70. Predecki, P., Stephan, J. E., Auslander, B. E., Mooney, V. L., and Kirkland, K. (1972). Kinetics of bone growth into cylindrical channels in aluminum oxide and titanium. J. Biomed. Mater. Res. 6, 375–400. Raab, S., Ahmed, A. M., and Provan, J. W. (1982). Thin film PMMA precoating for improved implant bone-cement fixation. J. Biomed. Mater. Res. 16, 679–704. Riegels-Nielsen, P., Sorensen, L., Andersen, H. M., and Lindequist, S. (1995). Boneloc® cemented total hip prostheses: Loosening in 28/43 cases after 3–38 months. Acta Orthop. Scand. 66, 215–217. Ritter, M. A. (1995). All polyethylene versus metal backing. Clin. Orthop. Rel. Res. 311, 69–75. Ritter, M. A., Keating, E. M., Faris, P. M., and Brugo, G. (1990). Metal-backed acetabular cups in total hip arthroplasty. J. Bone Joint Surg. 72A, 672–677. Sadr, B., and Arden, G. P. (1987). A comparison of the stability of proplast-coated and cemented Thompson prosthesis in the treatment of subcapital femoral fractures. Injury 8, 234–237. Saha, S., and Pal, S. (1984). Mechanical properties of bone cement: A review. J. Biomed. Mater. Res. 18, 435–462. Saikko, V. (1995). Wear of the polyethylene acetabular cup: The effect of head material, head diameter, and cup thickness studied with a hip simulator. Acta Orthop. Scand. 66, 501–506. Salvati, E. A., Lieberman, J. R., Huk, O. L., and Evans, B. G. (1995). Complications of femoral and acetabular modularity. Clin. Orthop. Rel. Res. 319, 85–93. Sauer, B. W., Weinstein, A. M., Klawitter, J. J., Hulbert, S. F., Leonard, R. B., and Bagwell, J. G. (1974). The role of porous polymeric materials in prosthesis attachment. J. Biomed. Mater. Res. Symposium 5, 145–156. Schmalzried, T. P., Kwong, L. M., Sedlacek, R. C., Jasty, M., Haire, T. C., O’Connor, D. O., Bragdon, C. R., Kabo, J. M., Malcolm, A. J., and Harris, W. H. (1992). The mechanism of loosening of cemented acetabular components in total hip arthroplasty: Analyis of specimens retrieved at autopsy. Clin. Orthop. Rel. Res. 274, 60–78. © 2000 by CRC Press LLC

Smith, L. (1963). Ceramic-plastic material as a bone substitute. Arch. Surg. 87, 653–661. Spector, M. (1982). Bone Ingrowth into porous metals. In: Biocompatibility of Orthopedic Implants, D. F. Williams, Ed. (Boca Raton, FL: CRC Press), 55–88. Spector, M. (1982). Bone Ingrowth into porous polymers. In: Biocompatibility of Orthopedic Implants, D. F. Williams, Ed. (Boca Raton, FL: CRC Press), 89–128. Stromberg, C. N., Herberts, P., Palmertz, B., and Garellick, G. (1996). Radiographic risk signs for loosening after cemented THA: 61 loose stems and 23 loose sockets compared with 42 controls. Acta Orthop. Scand. 67, 43–48. Sutula, L. C., Collier, J. P., Saum, K. A., Currier, B. H., Currier, J. H., Sanford, W. M., Mayor, M. B., Wooding, R. E., Spering, D. K., Williams, I. R., Kasprzak, D. J., and Sorprenant, V. A. (1995). Impact of gamma sterilization on clinical performance of polyethylene in the hip. Clin. Orthop. Rel. Res. 319, 28–40. Taitsman, J. P. and Saha, S. (1977). Tensile strength of wire-reinforced bone cement and twisted stainlesssteel wire. J. Bone Joint Surg. 59A, 419–425. Taylor, D. (1987). Fatigue failure in bone cements for hip joint implants. Fatigue 87, 1353. Taylor, D., Clarke, F., McCormack, B., and Sheehan, J. (1989). Reinforcement of bone cement using metal meshes. Engng. Med. 203, 49–53. Thanner, J., Freij-Larsson, C., Karrholm, J., Malchau, H., and Wesslen, B. (1995). Evaluation of Boneloc®: Chemical and mechanical properties, and a randomized clinical study of 30 total hip arthroplasties. Acta Orthop. Scand. 66, 207–214. Thomas, K. A., Cook, S. D., Renz, E. A., Anderson, R. C., Haddad, R. J. J., Haubold, A. D., and Yapp, R. (1985). The effect of surface treatments on the interface mechanics of LTI pyrolytic carbon implants. J. Biomed. Mater. Res. 19, 145–160. Topoleski, L. D. T., Ducheyne, P., and Cuckler, J. M. (1990). The fracture toughness of short-titaniumfiber reinforced bone cement. Trans. Soc. Biomat. 13, 107. Tullos, H. S., McCaskill, B. L., Dickey, R., and Davidson, J. (1984). Total hip arthroplasty with a lowmodulus porous-coated femoral component. J. Bone Joint Surg. 66A, 888–898. Verdonschot, N., and Huiskes, E. T. (1998). Effects of prosthesis surface roughness on the failure process of cemented hip implants after stem-cement debonding. J. Biom. Mater. Res. 42, 554–559. Weinstein, A. M., Klawitter, J. J., Cleveland, T. W., and Amoss, D. C. (1976). Electrical stimulation of bone growth into porous Al2O3. J. Biomed. Mater. Res. 10, 231–247. West, J. K., Clark, A. E., Hall, M. B., and Turner, G. F. (1990). In vivo bone-bonding study of Bioglass®coated titanium alloy. In: CRC Handbook of Bioactive Ceramics vol I: Bioactive Glasses and GlassCeramics, T. Yamamuro, L. L. Hench and J. Wilson, Eds. (Boca Raton, FL: CRC Press), 161–166. Willert, H. G., Koch, R., and Burgi, M. (1991). Reinforcement for a bone cement bed. In Patent # 5,035,714 (USA). Williams, D. F. and Roaf, R. (1973). Implants in Surgery (London: W.B. Saunders). Wroblewski, B. M. (1986). 15–21 Year results of the charnley low-friction arthroplasty. Clin. Orthop. Rel. Res. 211, 30–35. Wroblewski, B. M. (1993). Cementless versus cemented total hip arthroplasty: A scientific controversy? Orthop. Clin. North Am. 24, 591–597. Wykman, A. G. M., Olsson, A. E., Axdorph, G., and Goldie, I. (1991). Total hip arthroplasty. A comparison between cemented and press-fit noncemented fixation. J. Arthoplasty 6, 19–29. Yamamuro, T., Hench, L. L., and Wilson, J. E. (1990). CRC Handbook of Bioactive Ceramics vol I: Bioactive Glasses and Glass-Ceramics and vol. II Calcium Phosphate and Hydroxylapatite Ceramics (Boca Raton, FL: CRC Press).


Neuman, M.R. “Biomedical Sensors.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

V Biomedical Sensors Michael R. Neuman Joint Program in Biomedical Engineering, The University of Memphis and University of Tennessee, Memphis 47 Physical Measurements Michael R. Neuman Description of Sensors • Biomedical Applications of Physical Sensors

48 Biopotential Electrodes Michael R. Neuman Sensing Bioelectric Signals • Electrical Characteristics • Practical Electrodes for Biomedical Measurements • Biomedical Applications

49 Electrochemical Sensors Chung-Chiun Liu Conductivity/Capacitance Electrochemical Sensors • Potentiometric Sensors • Voltammetric Sensors • Reference Electrodes • Summary

50 Optical Sensors Yitzhak Mendelson Instrumentation • Optical Fibers • General Principles of Optical Sensing • Applications

51 Bioanalytic Sensors Richard P. Buck Classification of Biochemical Reactions in the Context of Sensor Design and Development • Classification of Transduction Processes—Detection Methods • Tables of Sensors from the Literature • Application of Microelectronics in Sensor Fabrication

S

ENSORS CONVERT SIGNALS OF ONE type of quantity such as hydrostatic fluid pressure into an equivalent signal of another type of quantity, for example, an electrical signal. Biomedical sensors take signals representing biomedical variables and convert them into what is usually an electrical signal. As such, the biomedical sensor serves as the interface between a biologic and an electronic system and must function in such a way as to not adversely affect either of these systems. In considering biomedical sensors, it is necessary to consider both sides of the interface: the biologic and the electronic, since both biologic and electronic factors play an important role in sensor performance. Many different types of sensors can be used in biomedical applications. Table V.1 gives a general classification of these sensors. It is possible to categorize all sensors as being either physical or chemical. In the case of physical sensors, quantities such as geometric, mechanical, thermal, and hydraulic variables are measured. In biomedical applications these can include things such as muscle displacement,


blood pressure, core body temperature, blood flow, TABLE V.1 Classifications of Biomedical Sensors cerebrospinal fluid pressure, and bone growth. Two Physical sensors types of physical sensors deserve special mention Geometric with regard to their biomedical application: Sensors Mechanical of electrical phenomena in the body, usually known Thermal Hydraulic as electrodes, play a special role as a result of their Electric diagnostic and therapeutic applications. The most Optical familiar of these are sensors used to pick up the Chemical sensors electrocardiogram, an electrical signal produced by Gas the heart. The other type of physical sensor that finds Electrochemical Photometric many applications in biology and medicine is the Other physical chemical methods optical sensor. These sensors can use light to collect Bioanalytic information, and, in the case of fiber optic sensors, light is the signal transmission medium as well. The second major classification of sensing devices is chemical sensors. In this case the sensors are concerned with measuring chemical quantities such as identifying the presence of particular chemical compounds, detecting the concentrations of various chemical species, and monitoring chemical activities in the body for diagnostic and therapeutic applications. A wide variety of chemical sensors can be classified in many ways. One such classification scheme is illustrated in Table V.1 and is based upon the methods used to detect the chemical components being measured. Chemical composition can be measured in the gas phase using several techniques, and these methods are especially useful in biomedical measurements associated with the pulmonary system. Electrochemical sensors measure chemical concentrations or, more precisely, activities based on chemical reactions that interact with electrical systems. Photometric chemical sensors are optical devices that detect chemical concentrations based upon changes in light transmission, reflection, or color. The familiar litmus test is an example of an optical change that can be used to measure the acidity or alkalinity of a solution. Other types of physical chemical sensors such as the mass spectrometer use various physical methods to detect and quantify chemicals associated with biologic systems. Although they are essentially chemical sensors, bioanalytic sensors are often classified as a separate major sensor category. These devices incorporate biologic recognition reactions such as enzyme-substrate, antigen-antibody, or ligand-receptor to identify complex biochemical molecules. The use of biologic reactions gives bioanalytic sensors high sensitivity and specificity in identifying and quantifying biochemical substances. One can also look at biomedical sensors from the standpoint of their applications. These can be generally divided according to whether a sensor is used for diagnostic or therapeutic purposes in clinical medicine and for data collection in biomedical research. Sensors for clinical studies such as those carried out in the clinical chemistry laboratory must be standardized in such a way that errors that could result in an incorrect diagnosis or inappropriate therapy are kept to an absolute minimum. Thus these sensors must not only be reliable themselves, but appropriate methods must exist for testing the sensors that are a part of the routine use of the sensors for making biomedical measurements. One can also look at biomedical sensors from the standpoint of how they are applied to the patient or research subject. Table V.2 shows the range of general approaches to attaching biomedical sensors. At the top of the list we have the method that involves the least interaction with the biologic object being studied; the bottom of the list includes sensors that interact to the greatest extent. Clearly if a measurement can be made equally well by a sensor that does not contact the subject being measured or by one that must be surgically implanted, the former is by far the TABLE V.2 Types of Sensor-Subject Interfaces most desirable. However, a sensor that is used to provide information to help control a device surgically Noncontacting (noninvasive) placed in the body to replace or assist a failing organ Skin surface (contacting) should be implanted, since this is the best way to comIndwelling (minimally invasive) Implantable (invasive) municate with the internal device. © 2000 by CRC Press LLC

You will notice in reading this section that the majority of biomedical sensors are essentially the same as sensors used in other applications. The unique part about biomedical sensors is their application. There are, however, special problems that are encountered by biomedical sensors that are unique to them. These problems relate to the interface between the sensor and the biologic system being measured. The presence of foreign materials, especially implanted materials, can affect the biologic environment in which they are located. Many biologic systems are designed to deal with foreign materials by making a major effort to eliminate them. The rejection reaction that is often discussed with regard to implanted materials or transplanted tissues is an example of this. Thus, in considering biomedical sensors, one must worry about this rejection phenomenon and how it will affect the performance of the sensor. If the rejection phenomenon changes the local biology around the sensor, this can result in the sensor measuring phenomena associated with the reaction that it has produced as opposed to phenomena characteristic of the biologic system being studied. Biologic systems can also affect sensor performance. This is especially true for indwelling and implanted sensors. Biologic tissue represents a hostile environment which can degrade sensor performance. In addition to many corrosive ions, body fluids contain enzymes that break down complex molecules as a part of the body’s effort to rid itself of foreign and toxic materials. These can attack the materials that make up the sensor and its package, causing the sensor to lose calibration or fail. Sensor packaging is an especially important problem. The package must not only protect the sensor from the corrosive environment of the body, but it must allow that portion of the sensor that performs the actual measurement to communicate with the biologic system. Furthermore, because it is frequently desirable to have sensors be as small as possible, especially those that are implanted and indwelling, it is important that the packaging function be carried out without significantly increasing the size of the sensor structure. Although there have been many improvements in sensor packaging, this remains a major problem in biomedical sensor research. High-quality packaging materials that do not elicit major foreign body responses from the biologic system are still being sought. Another problem that is associated with implanted sensors is that once they are implanted, access to them is very limited. This requires that these sensors be highly reliable so that there is no need to repair or replace them. It is also important that these sensors be highly stable, since in most applications it is not possible to calibrate the sensor in vivo. Thus, sensors must maintain their calibration once they are implanted, and for applications such as organ replacement, this can represent a potentially long time, the remainder of the patient’s life. In the following sections we will look at some of the sensors described above in more detail. We will consider physical sensors with special sections on biopotential electrodes and optical sensors. We will also look at chemical sensors, including bioanalytic sensing systems. We would also like to refer you to Appendix A “Basics of Blood Gas Instrumentation”, where applications of sensors for measurements of oxygen and carbon dioxide in blood samples and blood acid–base balance are discussed. Although it is not possible to cover the field in extensive detail in a handbook such as this, it is hoped that these sections can serve as an introduction to this important aspect of biomedical engineering and instrumentation.


Generic bioanalytical analyzer.

Neuman, M.R. “Physical Measurements.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

47 Physical Measurements 47.1

Linear and Angular Displacement Sensors • Inductance Sensors • Capacitive Sensors • Sonic and Ultrasonic Sensors • Velocity Measurement • Accelerometers • Force • Measurement of Fluid Dynamic Variables • Measurement • Temperature

Michael R. Neuman Joint Program in Biomedical Engineering, The University of Memphis and University of Tennessee, Memphis

Description of Sensors

47.2

Biomedical Applications of Physical Sensors

Physical variables associated with biomedical systems are measured by a group of sensors known as physical sensors. Although many specific physical variables can be measured in biomedical systems, these can be categorized into a simple list as shown in Table 47.1. Sensors for these variables, whether they are measuring biomedical systems or other systems, are essentially the same. Thus, sensors of linear displacement can frequently be used equally well for measuring the displacement of the heart muscle during the cardiac cycle or the movement of a robot arm. There is, however, one notable exception regarding the similarity of these sensors: the packaging of the sensor and attachment to the system being measured. Although physical sensors used in nonbiomedical applications need to be packaged so as to be protected from their environment, few of these sensors have to deal with the harsh environment of biologic tissue, especially with the mechanisms inherent in this tissue for trying to eliminate the sensor as a foreign body. Another notable exception to this similarity of sensors for measuring physical quantities in biologic and nonbiologic systems are the sensors used for fluidic measurements such as pressure and flow. Special needs for these measurements in biologic systems have resulted in special sensors and instrumentation systems for these measurements that can be quite different from systems for measuing pressure and flow in nonbiologic environments. In this chapter, we will attempt to review various examples of sensors used for physical measurement in biologic systems. Although it would be beyond the scope of this chapter to cover all these in detail, the principal sensors applied for biologic measurements will be described. Each section will include a brief description of the principle of operation of the sensor and the underlying physical principles, examples of some of the more common forms of these sensors for application in biologic systems, methods of signal processing for these sensors where appropriate, and important considerations for when the sensor is applied.

47.1 Description of Sensors Linear and Angular Displacement Sensors A comparison of various characteristics of displacement sensors described in detail below is outlined in Table 47.2. Variable Resistance Sensor One of the simplest sensors for measuring displacement is a variable resistor similar to the volume control on an audio electronic device [1]. The resistance between two terminals on this device is related to the


TABLE 47.1

Physical Variables and Sensors

Physical Quantity Geometric

Kinematic Force-Torque Fluidic Thermal

TABLE 47.2

Sensor

Variable Sensed

Strain gauge LVDT Ultrasonic transit time Velocimeter Accelerometer Load cell Pressure transducer Flow meter Thermometer Thermal flux sensor

Strain Displacement Displacement Velocity Acceleration Applied force or torque Pressure Flow Temperature Heat flux

Comparison of Displacement Sensors

Sensor

Electrical Variable

Variable resistor

Resistance

Foil strain gauge Liquid metal strain gauge Silicon strain gauge Mutual inductance coils

Resistance Resistance Resistance Inductance

Variable reluctance

Inductance

LVDT Parallel plate capacitor

Inductance Capacitance

Sonic/ultrasonic

Time

Measurement Circuit Voltage divider, ohmmeter, bridge, current source Bridge Ohmmeter, bridge Bridge Impedance bridge, inductance meter Impedance bridge, inductance meter Voltmeter Impedance bridge, capacitance meter Timer circuit

Sensitivity

Precision

High

Moderate

Large

Low Moderate High Moderate to high High

Moderate Moderate Moderate Moderate to low Moderate

Small Large Small Moderate to large Large

High Moderate to high High

High Moderate

High Moderate to large Large

High

Range

linear or angular displacement of a sliding tap along a resistance element. Precision devices are available that have a reproducible, linear relationship between resistance and displacement. These devices can be connected in circuits that measure such resistance as an ohmmeter or bridge, or they can be used as a part of a circuit that provides a voltage that is proportional to the displacement. Such circuits include the voltage divider (as illustrated in Fig. 47.1a) or driving a known constant current through the resistance and measuring the resulting voltage across it. This sensor is simple and inexpensive and can be used for measuring relatively large displacements. Strain Gauge Another displacement sensor based on an electrical resistance change is the strain gauge [2]. If a long narrow electrical conductor such as a piece of metal foil or a fine gauge wire is stretched within its elastic limit, it will increase in length and decrease in cross-sectional area. Because the electric resistance between both ends of this foil or wire can be given by

R=ρ

l A

(47.1)

where ρ is the electrical resistivity of the foil or wire material, l is its length, and A is its cross-sectional area, this stretching will result in an increase in resistance. The change in length can only be very small for the foil to remain within its elastic limit, so the change in electric resistance will also be small. The relative sensitivity of this device is given by its gauge factor, γ, which is defined as © 2000 by CRC Press LLC

FIGURE 47.1 Examples of displacement sensors: (a) variable resistance sensor, (b) foil strain gauge, (c) linear variable differential transformer (LVDT), (d) parallel plate capacitive sensor, and (e) ultrasonic transit time displacement sensor.

γ=

∆R R ∆l l

(47.2)

where ∆R is the change in resistance when the structure is stretched by an amount ∆ l. Foil strain gauges are the most frequently applied and consist of a structure such as shown in Fig. 47.1b. A piece of metal foil that is attached to an insulating polymeric film such as polyimide that has a much greater compliance than the foil itself is chemically etched into the pattern shown in Fig. 47.1b. When a strain is applied in the sensitive direction, the direction of the individual elements of the strain gauge, the length of the gauge will be slightly increased, and this will result in an increase in the electrical resistance seen between the terminals. Since the displacement or strain that this structure can measure is quite small for it to remain within its elastic limit, it can only be used to measure small displacements such as occur as loads are applied to structural beams. If one wants to increase the range of a foil strain gauge, one has to attach it to some sort of a mechanical impedance converter such as a cantilever beam. If the strain gauge is attached to one surface of the beam as shown in Fig. 47.2, a fairly large displacement at the unsupported end of the beam can be translated to a relatively small displacement on the beam’s surface. It would be possible for this structure to be used to measure larger displacements at the cantilever beam tip using a strain gauge on the beam. Because the electric resistance changes for a strain gauge are quite small, the measurement of this resistance change can be challenging. Generally, Wheatstone bridge circuits are used. It is important to note, however, that changes in temperature can also result in electric resistance changes that are of the same order of magnitude or even larger than the electric resistance changes due to the strain. Thus, it is important to temperature-compensate strain gauges in most applications. A simple method of temperature compensation is to use a double or quadruple strain gauge and a bridge circuit for measuring the resistance change. This is illustrated in Fig. 47.2. If one can use the strain gauge in an application such


FIGURE 47.2 Strain gauges on a cantilever structure to provide temperature compensation: (a) cross-sectional view of the cantilever and (b) placement of the strain gauges in a half bridge or full bridge for temperature compensation and enhanced sensitivity.

as the cantilever beam application described above, one can place one or two of the strain gauge structures on the concave side of the beam and the other one or two on the convex side of the beam. Thus, as the beam deflects, the strain gauge on the convex side will experience tension, and that on the concave side will experience compression. By putting these gauges in adjacent arms of the Wheatstone bridge, their effects can double the sensitivity of the circuit in the case of the double strain gauge and quadruple it in the case where the entire bridge is made up of strain gauges on a cantilever. In some applications it is not possible to place strain gauges so that one gauge is undergoing tension while the other is undergoing compression. In this case, the second strain gauge used for temperature compensation can be oriented such that its sensitive axis is in a direction where strain is minimal. Thus, it is still possible to have the temperature compensation by having two identical strain gauges at the same temperature in adjacent arms of the bridge circuit, but the sensitivity improvement described in the previous paragraph is not seen. Another constraint imposed by temperature is that the material to which the strain gauge is attached and the strain gauge both have temperature coefficients of expansion. Thus, even if a gauge is attached to a structure under conditions of no strain, if the temperature is changed, the strain gauge could experience some strain due to the different expansion that it will have compared to the structure to which it is attached. To avoid this problem, strain gauges have been developed that have identical temperature coefficients of expansion to various common materials. In selecting a strain gauge, one should choose a device with thermal expansion characteristics as close as possible to those of the object upon which the strain is to be measured.


A more compliant structure that has found applications in biomedical instrumentation is the liquid metal strain gauge [3]. Instead of using a solid electric conductor such as the wire or metal foil, mercury confined to a compliant, thin wall, narrow bore elastomeric tube is used. The compliance of this strain gauge is determined by the elastic properties of the tube. Since only the elastic limit of the tube is of concern, this sensor can be used to detect much larger displacements than conventional strain gauges. Its sensitivity is roughly the same as a foil or wire strain gauge, but it is not as reliable. The mercury can easily become oxidized or small air gaps can occur in the mercury column. These effects make the sensor’s characteristics noisy and sometimes results in complete failure. Another variation on the strain gauge is the semiconductor strain gauge. These devices are frequently made out of pieces of silicon with strain gauge patterns formed using semiconductor microelectronic technology. The principal advantage of these devices is that their gauge factors can be more than 50 times greater than that of the solid and liquid metal devices. They are available commercially, but they are a bit more difficult to handle and attach to structures being measured due to their small size and brittleness.

Inductance Sensors Mutual Inductance The mutual inductance between two coils is related to many geometric factors, one of which is the separation of the coils. Thus, one can create a very simple displacement sensor by having two coils that are coaxial but with different separation. By driving one coil with an ac signal and measuring the voltage signal induced in the second coil, this voltage will be related to how far apart the coils are from one another. When the coils are close together, the mutual inductance will be high, and so a higher voltage will be induced in the second coil; when the coils are more widely separated, the mutual inductance will be lower as will the induced voltage. The relationship between voltage and separation will be determined by the specific geometry of the coils and in general will not be a linear relationship with separation unless the change of displacement is relatively small. Nevertheless, this is a simple method of measuring separation that works reasonably well provided the coils remain coaxial. If there is movement of the coils transverse to their axes, it is difficult to separate the effects of transverse displacement from those of displacement along the axis. Variable Reluctance A variation on this sensor is the variable reluctance sensor wherein a single coil or two coils remain fixed on a form which allows a high reluctance slug to move into or out of the coil or coils along their axis. Since the position of this core material determines the number of flux linkages through the coil or coils, this can affect the self-inductance or mutual inductance of the coils. In the case of the mutual inductance, this can be measured using the technique described in the previous paragraph, whereas self-inductance changes can be measured using various instrumentation circuits used for measuring inductance. This method is also a simple method for measuring displacements, but the characteristics are generally nonlinear, and the sensor generally has only moderate precision. Linear Variable Differential Transformer By far the most frequently applied displacement transducer based upon inductance is the linear variable differential transformer (LVDT) [4]. This device is illustrated in Fig. 47.1c and is essentially a three-coil variable reluctance transducer. The two secondary coils are situated symmetrically about the primary coil and connected such that the induced voltages in each secondary oppose each other. When the core is located in the center of the structure equidistant from each secondary coil, the voltage induced in each secondary will be the same. Since these voltages oppose one another, the output voltage from the device will be zero. As the core is moved closer to one or the other secondary coils, the voltages in each coil will no longer be equal, and there will be an output voltage proportional to the displacement of the core from the central, zero-voltage position. Because of the symmetry of the structure, this voltage is linearly related to the core displacement. When the core passes through the central, zero point, the phase of the


output voltage from the sensor changes by 180 degrees. Thus, by measuring the phase angle as well as the voltage, one can determine the position of the core. The circuit associated with the LVDT not only measures the voltage but often measures the phase angle as well. LVDTs are available commercially in many sizes and shapes. Depending on the configuration of the coils, they can measure displacements ranging from tens of micrometers through centimeters.

Capacitive Sensors Displacement sensors can be based upon measurements of capacitance as well as inductance. The fundamental principle of operation is the capacitance of a parallel plate capacitor as given by

C=

A d

(47.3)

where is the dielectric constant of the medium between the plates, d is the separation between the plates, and A is the cross-sectional area of the plates. Each of the quantities in Eq. (47.3) can be varied to form a displacement transducer as shown in Fig. 47.1c. By moving one of the plates with respect to the other, Eq. (47.3) shows us that the capacitance will vary inversely with respect to the plate separation. This will give a hyperbolic capacitance-displacement characteristic. However, if the plate separation is maintained at a constant value and the plates are displaced laterally with respect to one another so that the area of overlap changes, this can produce a capacitance-displacement characteristic that can be linear, depending on the shape of the actual plates. The third way that a variable capacitance transducer can measure displacement is by having a fixed parallel plate capacitor with a slab of dielectric material having a dielectric constant different from that of air that can slide between the plates. The effective dielectric constant for the capacitor will depend on how much of the slab is between the plates and how much of the region between the plates is occupied only by air. This, also, can yield a transducer with linear characteristics. The electronic circuitry used with variable capacitance transducers, is essentially the same as any other circuitry used to measure capacitance. As with the inductance transducers, this circuit can take the form of a bridge circuit or specific circuits that measure capacitive reactance.

Sonic and Ultrasonic Sensors If the velocity of sound in a medium is constant, the time it takes a short burst of that sound energy to propagate from a source to a receiver will be proportional to the displacement between the two transducers. This is given by

d = cT

(47.4)

where c is the velocity of sound in the medium, T is the transit time, and d is the displacement. A simple system for making such a measurement is shown in Fig. 47.1e [5]. A brief sonic or ultrasonic pulse is generated at the transmitting transducer and propagates through the medium. It is detected by the receiving transducer at time T after the burst was initiated. The displacement can then be determined by applying Eq. (47.4). In practice, this method is best used with ultrasound, since the wavelength is shorter, and the device will neither produce annoying sounds nor respond to extraneous sounds in the environment. Small piezoelectric transducers to generate and receive ultrasonic pulses are readily available. The electronic circuit used with this instrument carries out three functions: (1) generation of the sonic or ultrasonic burst, (2) detection of the received burst, and (3) measurement of the time of propagation of the ultrasound. An advantage of this system is that the two transducers are coupled to one another only sonically. There is no physical connection as was the case for the other sensors described in this section. © 2000 by CRC Press LLC

Velocity Measurement Velocity is the time derivative of displacement, and so all the displacement transducers mentioned above can be used to measure velocity if their signals are processed by passing them through a differentiator circuit. There are, however, two additional methods that can be applied to measure velocity directly. Magnetic Induction If a magnetic field that passes through a conducting coil varies with time, a voltage is induced in that coil that is proportional to the time-varying magnetic field. This relationship is given by

v=N

dφ dt

(47.5)

where v is the voltage induced in the coil, N is the number of turns in the coil, and φ is the total magnetic flux passing through the coil (the product of the flux density and area within the coil). Thus a simple way to apply this principle is to attach a small permanent magnet to an object whose velocity is to be determined, and attach a coil to a nearby structure that will serve as the reference against which the velocity is to be measured. A voltage will be induced in the coil whenever the structure containing the permanent magnet moves, and this voltage will be related to the velocity of that movement. The exact relationship will be determined by the field distribution for the particular magnet and the orientation of the magnet with respect to the coil. Doppler Ultrasound When the receiver of a signal in the form of a wave such as electromagnetic radiation or sound is moving at a nonzero velocity with respect to the emitter of that wave, the frequency of the wave perceived by the receiver will be different than the frequency of the transmitter. This frequency difference, known as the Doppler shift, is determined by the relative velocity of the receiver with respect to the emitter and is given by

fd =

fou c

(47.6)

where fd is the Doppler frequency shift, fo is the frequency of the transmitted wave, u is the relative velocity between the transmitter and receiver, and c is the velocity of sound in the medium. This principle can be applied in biomedical applications as a Doppler velocimeter. A piezoelectric transducer can be used as the ultrasound source with a similar transducer as the receiver. When there is no relative movement between the two transducers, the frequency of the signal at the receiver will be the same as that at the emitter, but when there is relative motion, the frequency at the receiver will be shifted according to Eq. (47.6). The ultrasonic velocimeter can be applied in the same way that the ultrasonic displacement sensor is used. In this case the electronic circuit produces a continuous ultrasonic wave and, instead of detecting the transit time of the signal, now detects the frequency difference between the transmitted and received signals. This frequency difference can then be converted into a signal proportional to the relative velocity between the two transducers.

Accelerometers Acceleration is the time derivative of velocity and the second derivative with respect to time of displacement. Thus, sensors of displacement and velocity can be used to determine acceleration when their signals are appropriately processed through differentiator circuits. In addition, there are direct sensors of acceleration based upon Newton’s second law and Hooke’s law. The fundamental structure of an accelerometer


FIGURE 47.3

Fundamental structure of an accelerometer.

is shown in Fig. 47.3. A known seismic mass is attached to the housing by an elastic element. As the structure is accelerated in the sensitive direction of the elastic element, a force is applied to that element according to Newton’s second law. This force causes the elastic element to be distorted according to Hooke’s law, which results in a displacement of the mass with respect to the accelerometer housing. This displacement is measured by a displacement sensor. The relationship between the displacement and the acceleration is found by combining Newton’s second law and Hooke’s law

a=

k x m

(47.7)

where x is the measured displacement, m is the known mass, k is the spring constant of the elastic element, and a is the acceleration. Any of the displacement sensors described above can be used in an accelerometer. The most frequently used displacement sensors are strain gauges or the LVDT. One type of accelerometer uses a piezoelectric sensor as both the displacement sensor and the elastic element. A piezoelectric sensor generates an electric signal that is related to the dynamic change in shape of the piezoelectric material as the force is applied. Thus, piezoelectric materials can only directly measure time varying forces. A piezoelectric accelerometer is, therefore, better for measuring changes in acceleration than for measuring constant accelerations. A principal advantage of piezoelectric accelerometers is that they can be made very small, which is useful in many biomedical applications.

Force Force is measured by converting the force to a displacement and measuring the displacement with a displacement sensor. The conversion takes place as a result of the elastic properties of a material. Applying a force to the material distorts the material’s shape, and this distortion can be determined by a displacement sensor. For example, the cantilever structure shown in Fig. 47.2a could be a force sensor. Applying a vertical force at the tip of the beam will cause the beam to deflect according to its elastic properties. This deflection can be detected using a displacement sensor such as a strain gauge as described previously. A common form of force sensor is the load cell. This consists of a block of material with known elastic properties that has strain gauges attached to it. Applying a force to the load cell stresses the material, resulting in a strain that can be measured by the strain gauge. Applying Hooke’s law, one finds that the strain is proportional to the applied force. The strain gauges on a load cell are usually in a half-bridge © 2000 by CRC Press LLC

FIGURE 47.4 Structure of an unbonded strain gauge pressure sensor. Reproduced with permission from Neuman MR. 1993. Biomedical sensors. In RC Dorf (ed), The Electrical Engineering Handbook, Boca Raton, Fla, CRC Press.

or full-bridge configuration to minimize the temperature sensitivity of the device. Load cells come in various sizes and configurations, and they can measure a wide range of forces.

Measurement of Fluid Dynamic Variables The measurement of the fluid pressure and flow in both liquids and gases is important in many biomedical applications. These two variables, however, often are the most difficult variables to measure in biologic applications because of interactions with the biologic system and stability problems. Some of the most frequently applied sensors for these measurements are described in the following paragraphs. Pressure Measurement Sensors of pressure for biomedical measurements such as blood pressure [7] consist of a structure such as shown in Fig. 47.4. In this case a fluid coupled to the fluid to be measured is housed in a chamber with a flexible diaphragm making up a portion of the wall, with the other side of the diaphragm at atmospheric pressure. When a pressure exists across the diaphragm, it will cause the diaphragm to deflect. This deflection is then measured by a displacement sensor. In the example in Fig. 47.4, the displacement transducer consists of four fine-gauge wires drawn between a structure attached to the diaphragm and the housing of the sensor so that these wires serve as strain gauges. When pressure causes the diaphragm to deflect, two of the fine-wire strain gauges will be extended by a small amount, and the other two will contract by the same amount. By connecting these wires into a bridge circuit, a voltage proportional to the deflection of the diaphragm and hence the pressure can be obtained. Semiconductor technology has been applied to the design of pressure transducers such that the entire structure can be fabricated from silicon. A portion of a silicon chip can be formed into a diaphragm and semiconductor strain gauges incorporated directly into that diaphragm to produce a small, inexpensive, and sensitive pressure sensor. Such sensors can be used as disposable, single-use devices for measuring blood pressure without the need for additional sterilization before being used on the next patient. This minimizes the risk of transmitting blood-borne infections in the cases where the transducer is coupled directly to the patient’s blood for direct blood pressure measurement. In using this type of sensor to measure blood pressure, it is necessary to couple the chamber containing the diaphragm to the blood or other fluids being measured. This is usually done using a small, flexible plastic tube known as a catheter, that can have one end placed in an artery of the subject while the other is connected to the pressure sensor. This catheter is filled with a physiologic saline solution so that the arterial blood pressure is coupled to the diaphragm. This external blood-pressure-measurement method is used quite frequently in the clinic and research laboratory, but it has the limitation that the properties of the fluid in the catheter and the catheter itself can affect the measurement. For example, both ends


of the catheter must be at the same vertical level to avoid a pressure offset due to hydrostatic effects. Also, the compliance of the tube will affect the frequency response of the pressure measurement. Air bubbles in the catheter or obstructions due to clotted blood or other materials can introduce distortion of the waveform due to resonance and damping. These problems can be minimized by utilizing a miniature semiconductor pressure transducer that is located at the tip of a catheter and can be placed in the blood vessel rather than being positioned external to the body. Such internal pressure sensors are available commercially and have the advantages of a much broader frequency response, no hydrostatic pressure error, and generally clearer signals than the external system. Although it is possible to measure blood pressure using the techniques described above, this remains one of the major problems in biomedical sensor technology. Long-term stability of pressure transducers is not very good. This is especially true for pressure measurements of venous blood, cerebrospinal fluid, or fluids in the gastrointestinal tract, where pressures are relatively low. Long-term changes in baseline pressure for most pressure sensors require that they be frequently adjusted to be certain of zero pressure. Although this can be done relatively easily when the pressure transducer is located external to the body, this can be a major problem for indwelling pressure transducers. Thus, these transducers must be extremely stable and have low baseline drift to be useful in long-term applications. The packaging of the pressure transducer is also a problem that needs to be addressed, especially when the transducer is in contact with blood for long periods. Not only must the package be biocompatible, but it also must allow the appropriate pressure to be transmitted from the biologic fluid to the diaphragm. Thus, a material that is mechanically stable under corrosive and aqueous environments in the body is needed.

Measurement of Flow The measurement of true volummetric flow in the body represents one of the most difficult problems in biomedical sensing [8]. The sensors that have been developed measure velocity rather than volume flow, and they can only be used to measure flow if the velocity is measured for a tube of known crosssection. Thus, most flow sensors constrain the vessel to have a specific cross-sectional area. The most frequently used flow sensor in biomedical systems is the electromagnetic flow meter illustrated in Fig. 47.5. This device consists of a means of generating a magnetic field transverse to the flow vector in a vessel. A pair of very small biopotential electrodes are attached to the wall of the vessel such that the vessel diameter between them is at right angles to the direction of the magnetic field. As the

FIGURE 47.5 Fundamental structure of an electromagnetic flowmeter. Reproduced with permission from Neuman MR. 1986. Biosensors: Transducers, electrodes, and physiologic systems. IN JD Bronzino (ed), Biomedical Engineering and Instrumentation: Basic Concepts and Applications, Boston, PWS Publishers. © 2000 by CRC Press LLC

FIGURE 47.6 Structure of an ultrasonic Doppler flowmeter with the major blocks of the electronic signal processing system. The oscillator generates a signal that, after amplification, drives the transmitting transducer. The oscillator frequency is usually in the range of 1–10 MHz. The reflected ultrasound from the blood is sensed by the receiving transducer and amplified before being processed by a detector circuit. This block generates the frequency difference between the transmitted and received ultrasonic signals. This difference frequency can be converted into a voltage proportional to frequency, and hence flow velocity, by the frequency to voltage converter circuit.

blood flows in the structure, ions in the blood deflect in the direction of one or the other electrodes due to the magnetic field, and the voltage across the electrodes is given by

v = Blu

(47.8)

where B is the magnetic field, l is the distance between the electrodes, and u is the average instantaneous velocity of the fluid across the vessel. If the sensor constrains the blood vessel to have a specific diameter, then its cross-sectional area will be known, and multiplying this area by the velocity will give the volume flow. Although dc flow sensors have been developed and are available commercially, the most desirable method is to use ac excitation of the magnetic field so that offset potential effects from the biopotential electrodes do not generate errors in this measurement. Small ultrasonic transducers can also be attached to a blood vessel to measure flow as illustrated in Fig. 47.6. In this case the transducers are oriented such that one transmits a continuous ultrasound signal that illuminates the blood. Cells within the blood diffusely reflect this signal in the direction of the second sensor so that the received signal undergoes a Doppler shift in frequency that is proportional to the velocity of the blood. By measuring the frequency shift and knowing the cross-sectional area of the vessel, it is possible to determine the flow. Another method of measuring flow that has had biomedical application is the measurement of cooling of a heated object by convection. The object is usually a thermistor (see section on temperature measurement, below) placed either in a blood vessel or in tissue, and the thermistor serves as both the heating element and the temperature sensor. In one mode of operation, the amount of power required to maintain the thermistor at a temperature slightly above that of the blood upstream is measured. As the flow around the thermistor increases, more heat is removed from the thermistor by convection, and so more power is required to keep it at a constant temperature. Relative flow is then measured by determining the amount of power supplied to the thermistor. In a second approach the thermistor is heated by applying a current pulse and then measuring the cooling curve of the thermistor as the blood flows across it. The thermistor will cool more quickly as the blood flow increases. Both these methods are relatively simple to achieve electronically, but both also


TABLE 47.3

Properties of Temperature Sensors

Sensor Metal resistance thermometer Thermistor Thermocouple Mercury in glass thermometer Silicon p-n diode

Form

Sensitivity

Stability

Range

Coil of fine platinum wire Bead, disk, or rod Pair of wires Column of Hg in glass capillary Electronic component

Low High Low Moderate Moderate

High Moderate High High High

–100–700°C –50–100°C –100–>1000°C –50–400°C –50–150°C

have severe limitations. They are essentially qualitative measures and strongly depend on how the thermistor probe is positioned in the vessel being measured. If the probe is closer to the periphery or even in contact with the vessel wall, the measured flow will be different than if the sensor is in the center of the vessel.

Temperature There are many different sensors of temperature [9], but three find particularly wide application to biomedical problems. Table 47.3 summarizes the properties of various temperature sensors, and these three, including metallic resistance thermometers, thermistors, and thermocouples, are described in the following paragraphs. Metallic Resistance Thermometers The electric resistance of a piece of metal or wire generally increases as the temperature of that electric conductor increases. A linear approximation to this relationship is given by

[ (

R = R0 1 + α T − T0

)]

(47.9)

where R0 is the resistance at temperature T0 , α is the temperature coefficient of resistance, and T is the temperature at which the resistance is being measured. Most metals have temperature coefficients of resistance of the order of 0.1–0.4%/°C, as indicated in Table 47.4. The noble metals are preferred for resistance thermometers, since they do not corrode easily and, when drawn into fine wires, their crosssection will remain constant, thus avoiding drift in the resistance over time which could result in an unstable sensor. It is also seen from Table 47.4 that the noble metals of gold and platinum have some of the highest temperature coefficients of resistance of the common metals. Metal resistance thermometers are often fabricated from fine-gauge insulated wire that is wound into a small coil. It is important in doing so to make certain that there are not other sources of resistance change that could affect the sensor. For example, the structure should be utilized in such a way that no external strains are applied to the wire, since the wire could also behave as a strain gauge. Metallic films and foils can also be used as temperature sensors, and commercial products are available in the wire, TABLE 47.4

Temperature Coefficient of Resistance for Common Metals and Alloys

Metal or Alloy Platinum Gold Silver Copper Constantan (60% Cu, 40% Ni) Nichrome (80% Ni, 20% Cr)

Resistivity at 20°C microhm-cm

Temperature Coefficient of Resistance, %/°C

9.83 2.22 1.629 1.724 49.0 108.0

0.3 0.368 0.38 0.393 0.0002 0.013

Source: Pender H, McIlwain K. 1957. Electrical Engineers’ Handbook, 4th ed, New York, Wiley.


foil, or film forms. The electric circuits used to measure resistance, and hence the temperature, are similar to those used with the wire or foil strain gauges. A bridge circuit is the most desirable, although ohmmeter circuits can also be used. It is important to make sure that the electronic circuit does not pass a large current through the resistance thermometer to provide self-heating due to the Joule conversion of electric energy to heat. Thermistors Unlike metals, semiconductor materials have an inverse relationship between resistance and temperature. This characteristic is very nonlinear and cannot be characterized by a linear equation such as for the metals. The thermistor is a semiconductor temperature sensor. Its resistance as a function of temperature is given by

R = R0e

β  1 − 1   T T0 

(47.10)

where β is a constant determined by the materials that make up the thermistor. Thermistors can take a variety of forms and cover a large range of resistances. The most common forms used in biomedical applications are the bead, disk, or rod forms of the sensor as illustrated in Fig. 47.7. These structures can be formed from a variety of different semiconductors ranging from elements such as silicon and germanium to mixtures of various semiconducting metallic oxides. Most commercially available thermistors are manufactured from the latter materials, and the specific materials as well as the process for fabricating them are closely held industrial secrets. These materials are chosen not only to have high sensitivity but also to have the greatest stability, since thermistors are generally not as stable as the metallic resistance thermometers. However, thermistors are close to an order of magnitude more sensitive. Thermocouples When different regions of an electric conductor or semiconductor are at different temperatures, there is an electric potential between these regions that is directly related to the temperature differences. This phenomenon, known as the Seebeck effect, can be used to produce a temperature sensor known as a thermocouple by taking a wire of metal or alloy A and another wire of metal or alloy B and connecting them as shown in Fig. 47.8. One of the junctions is known as the sensing junction, and the other is the reference junction. When these junctions are at different temperatures, a voltage proportional to the temperature difference will be seen at the voltmeter when metals A and B have different Seebeck coefficients. This voltage is roughly proportional to the temperature difference and can be represented over the relatively small temperature differences encountered in biomedical applications by the linear equation

V = SAB ( Ts – Tr)

(47.11)

where SAB is the Seebeck coefficient for the thermocouple made up of metals A and B. Although this equation is a reasonable approximation, more accurate data are usually found in tables of actual voltages as a function of temperature difference. In some applications the voltmeter is


FIGURE 47.7

Common forms of thermistors.

FIGURE 47.8 Circuit arrangement for a thermocouple showing the voltage-measuring device, the voltmeter, interrupting one of the thermocouple wires (a) and at the cold junction (b).

located at the reference junction, and one uses some independent means such as a mercury in glass thermometer to measure the reference junction temperature. Where precision measurements are made, the reference junction is often placed in an environment of known temperature such as an ice bath. Electronic measurement of reference junction temperature can also be carried out and used to compensate for the reference junction temperature so that the voltmeter reads a signal equivalent to what would be seen if the reference junction were at 0°C. This electronic reference junction compensation is usually carried out using a metal resistance temperature sensor to determine reference junction temperature. The voltages generated by thermocouples used for temperature measurement are generally quite small being on the order of tens of microvolts per degree C. Thus, for most biomedical measurements where there is only a small difference in temperature between the sensing and reference junction, very sensitive amplifiers must be used to measure these potentials. Thermocouples have been used in industry for temperature measurement for many years. Several standard alloys to provide optimal sensitivity and stability of these sensors have evolved. Table 47.5 lists these common alloys, the Seebeck coefficient for thermocouples of these materials at room temperature, and the full range of temperatures over which these thermocouples can be used. TABLE 47.5

Common Thermocouples

Type

Materials

Seebeck Coefficient, µV/°C

Temperature Range

S T K J E

Platinum/platinum 10% rhodium Copper/constantan Chromel/alumel Iron/constantan Chromel/constantan

6 50 41 53 78

0–1700°C –190–400°C –200–1370°C –200–760°C –200–970°C


Thermocouples can be fabricated in many different ways depending on their applications. They are especially suitable for measuring temperature differences between two structures, since the sensing junction can be placed on one while the other has the reference junction. Higher-output thermocouples or thermopiles can be produced by connecting several thermocouples in series. Thermocouples can be made from very fine wires that can be implanted in biologic tissues for temperature measurements, and it is also possible to place these fine-wire thermocouples within the lumen of a hypodermic needle to make short-term temperature measurements in tissue.

47.2 Biomedical Applications of Physical Sensors Just as it is not possible to cover the full range of physical sensors in this chapter, it is also impossible to consider the many biomedical applications that have been reported for these sensors. Instead, some representative examples will be given. These are summarized in Table 47.6 and will be briefly described in the following paragraphs. Liquid metal strain gauges are especially useful in biomedical applications, because they are mechanically compliant and provide a better mechanical impedance match to most biomedical tissues than other types of strain gauges. By wrapping one of these strain gauges around a circumference of the abdomen, it will stretch and contract with the abdominal breathing movements. The signal from the strain gauge can then be used to monitor breathing in patients or experimental animals. The advantage of this sensor is its compliance so that it does not interfere with the breathing movements or substantially increase the required breathing effort. One of the original applications of the liquid metal strain gauge was in limb plethysmography [3]. One or more of these sensors are wrapped around an arm or leg at various points and can be used to measure changes in circumference that are related to the cross-sectional area and hence the volume of the limb at those points. If the venous drainage from the limb is occluded, the limb volume will increase as it fills with blood. Releasing the occlusion allows the volume to return to normal. The rate of this decrease in volume can be monitored using the liquid metal strain gauges, and this can be used to identify venous blockage when the return to baseline volume is too slow. Breathing movements, although not volume, can be seen using a simple magnetic velocity detector. By placing a small permanent magnet on the anterior side of the chest or abdomen and a flat, large-area coil on the posterior side opposite from the magnet, voltages are induced in the coil as the chest of abdomen moves during breathing. The voltage itself can be used to detect the presence of breathing movements, or it can be electronically integrated to give a signal related to displacement. The LVDT is a displacement sensor that can be used for more precise applications. For example, it can be used in studies of muscle physiology where one wants to measure the displacement of a muscle or where one is measuring the isometric force generated by the muscle (using a load cell) and must TABLE 47.6

Examples of Biomedical Applications of Physical Sensors

Sensor Liquid metal strain gauge Magnetic displacement sensor LVDT Load cell Accelerometer Miniature silicon pressure sensor

Electromagnetic flow sensor


Application

Signal Range

Breathing movement Limb plethysmography Breathing movement Muscle contraction Uterine contraction sensor Electronic scale Subject activity Intra-arterial blood pressure Urinary bladder pressure Intrauterine pressure Cardiac output (with integrator) Organ blood flow

0–0.05 0–0.02 0–10 mm 0–20 mm 0–5 mm 0–440 lbs (0–200 kg) 0–20 m/s2 0–50 Pa (0–350 mm Hg) 0–10 Pa (0–70 mm Hg) 0–15 Pa (0–100 mm Hg) 0–500 ml/min 0–100 ml/min

Reference 3 10 11 12 13

14 15

ensure that there is no muscle movement. It can also be incorporated into other physical sensor such as a pressure sensor or a tocodynamometer, a sensor used to electronically “feel” uterine contractions of patients in labor or those at risk of premature labor and delivery. In addition to studying muscle forces, load cells can be used in various types of electronic scales for weighing patients or study animals. The simplest electronic scale consists of a platform placed on top of a load cell. The weight of any object placed on the platform will produce a force that can be sensed by the load cell. In some critical care situations in the hospital, it is important to carefully monitor the weight of a patient. For example, this is important in watching water balance in patients receiving fluid therapy. The electronic scale concept can be extended by placing a load cell under each leg of the patient’s bed and summing the forces seen by each load cell to get the total weight of the patient and the bed. Since the bed weight remains fixed, weight changes seen will reflect changes in patient weight. Accelerometers can be used to measure patient or research subject activity. By attaching a small accelerometer to the individual being studied, any movements can be detected. This can be useful in sleep studies where movement can help to determine the sleep state. Miniature accelerometers and recording devices can also be worn by patients to study activity patterns and determine effects of disease or treatments on patient activity [13]. Miniature silicon pressure sensors are used for the indwelling measurement of fluid pressure in most body cavities. The measurement of intra-arterial blood pressure is the most frequent application, but pressures in other cavities such as the urinary bladder and the uterus are also measured. The small size of these sensors and the resulting ease of introduction of the sensor into the cavity make these sensors important for these applications. The electromagnetic flow sensor has been a standard method in use in the physiology laboratory for many years. Its primary application has been for measurement of cardiac output and blood flow to specific organs in research animals. New miniature inverted electromagnetic flow sensors make it possible to temporarily introduce a flow probe into an artery through its lumen to make clinical measurements. The measurement of body temperature using instruments employing thermistors as the sensor has greatly increased in recent years. Rapid response times of these low-mass sensors make it possible to quickly assess patients’ body temperatures so that more patients can be evaluated in a given period. This can then help to reduce health care costs. The rapid response time of low-mass thermistors makes them a simple sensor to be used for sensing breathing. By placing small thermistors near the nose and mouth, the elevated temperature of exhaled air can be sensed to document a breath. The potential applications of physical sensors in medicine and biology are almost limitless. To be able to use these devices, however, scientists must first be familiar with the underlying sensing principles. It is then possible to apply these in a form that addresses the problems at hand.

References 1. Doebelin EO. 1990. Measurement Systems: Applications and Design, New York, McGraw-Hill. 2. Dechow PC. 1988. Strain Gauges. In J Webster (ed), Encyclopedia of Medical Devices and Instrumentation, pp 2715–2721, New York, Wiley. 3. Whitney RJ. 1949. The measurement of changes in human limb-volume by means of a mercuryin-rubber strain gauge. J Physiol 109:5p. 4. Schaevitz H. 1947. The linear variable differential transformer. Proc Soc Stress Anal 4:79. 5. Stegall HF, Kardon MB, Stone HL, et al. 1967. A portable simple sonomicrometer. J Appl Physiol 23:289. 6. Angelsen BA, Brubakk AO. 1976. Transcutaneous measurement of blood flow velocity in the human aorta. Cardiovasc Res 10:368. 7. Geddes LA. 1970. The Direct and Indirect Measurement of Blood Pressure, Chicago, Year Book. 8. Roberts VC. 1972. Blood Flow Measurements, Baltimore, Williams & Wilkins. 9. Herzfeld CM (ed). 1962. Temperature: Its Measurement and Control in Science and Industry, New York, Reinhold. © 2000 by CRC Press LLC

10. Rolfe P. 1971. A magnetometer respiration monitor for use with premature babies. Biomed Eng 6:402. 11. Reddy NP, Kesavan SK. 1988. Linear variable differential transformers. In J Webster (ed), Encyclopedia of Medical Devices and Instrumentation, pp 1800–1806, New York, Wiley. 12. Roe FC. 1966. New equipment for metabolic studies. Nurs Clin N Am 1:621. 13. Patterson SM, Krantz DS, Montgomery LC, et al. 1993. Automated physical activity monitoring: Validation and comparison with physiological and self-report measures. Psychophysiology 30:296. 14. Fleming DG, Ko WH, Neuman MR (eds). 1977. Indwelling and Implantable Pressure Transducers, Cleveland, CRC Press. 15. Wyatt DG. 1971. Electromagnetic blood flow measurements. In BW Watson (ed), IEE Medical Electronics Monographs, London, Peregrinus.

Further Information Good overviews of physical sensors are found in these books: Doebelin EO. 1990. Measurement Systems: Application and Design, 4th ed, New York, McGraw-Hill; Harvey, GF (ed). 1969. Transducer Compendium, 2d ed, New York, Plenum. One can also find good descriptions of physical sensors in chapters of two works edited by John Webster. Chapters 2, 7, and 8 of his textbook (1992) Medical Instrumentation: Application and Design, 2d ed, Boston, Houghton Mifflin, and several articles in his Encyclopedia on Medical Devices and Instrumentation, published by Wiley in 1988, cover topics on physical sensors. Although a bit old, the text Transducers for Biomedical Measurements (New York, J. Wiley, 1974) by Richard S.C. Cobbold, remains one of the best descriptions of biomedical sensors available. By supplementing the material in this book with recent manufacturers’ literature, the reader can obtain a wealth of information on physical (and for that matter chemical) sensors for biomedical application. The journals IEEE Transactions on Biomedical Engineering and Medical and Biological Engineering and Computing are good sources of recent research on biomedical applications of physical sensors. The journals Physiological Measurement and Sensors and Actuators are also good sources for this material.


Neuman, M. R. “Biopotential Electrodes.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

48 Biopotential Electrodes

Michael R. Neuman Joint Program in Biomedical Engineering, The University of Memphis and University of Tennessee, Memphis

48.1 48.2 48.3

Sensing Bioelectric Signals Electrical Characteristics Practical Electrodes for Biomedical Measurements Body-Surface Biopotential Electrodes • Intracavity and Intratissue Electrodes • Microelectrodes • Electrodes Fabricated Using Microelectronic Technology

48.4

Biomedical Applications

Biologic systems frequently have electric activity associated with them. This activity can be a constant dc electric field, a constant flux of charge-carrying particles or current, or a time-varying electric field or current associated with some time-dependent biologic or biochemical phenomenon. Bioelectric phenomena are associated with the distribution of ions or charged molecules in a biologic structure and the changes in this distribution resulting from specific processes. These changes can occur as a result of biochemical reactions, or they can emanate from phenomena that alter local anatomy. One can find bioelectric phenomena associated with just about every organ system in the body. Nevertheless, a large proportion of these signals are associated with phenomena that are at the present time not especially useful in clinical medicine and represent time-invariant, low-level signals that are not easily measured in practice. There are, however, several signals that are of diagnostic significance or that provide a means of electronic assessment to aid in understanding biologic systems. These signals, their usual abbreviations, and the systems they measure are listed in Table 48.1. Of these, the most familiar is the electrocardiogram, a signal derived from the electric activity of the heart. This signal is widely used in diagnosing disturbances in cardiac rhythm, signal conduction through the heart, and damage due to cardiac ischemia and infarction. The electromyogram is used for diagnosing neuromuscular diseases, and the electroencephalogram is important in identifying brain dysfunction and evaluating sleep. The other signals listed in Table 48.1 are currently of lesser diagnostic significance but are, nevertheless, used for studies of the associated organ systems. Although Table 48.1 and the above discussion are concerned with bioelectric phenomena in animals and these techniques are used primarily in studying mammals, bioelectric signals also arise from plants [1]. These signals are generally steady-state or slowly changing, as opposed to the time-varying signals listed in Table 48.1. An extensive literature exists on the origins of bioelectric signals, and the interested reviewer is referred to the text by Plonsey and Barr for a general overview of this area [2].

48.1 Sensing Bioelectric Signals The mechanism of electric conductivity in the body involves ions as charge carriers. Thus, picking up bioelectric signals involves interacting with these ionic charge carriers and transducing ionic currents into electric currents required by wires and electronic instrumentation. This transducing function is carried out by electrodes that consist of electrical conductors in contact with the aqueous ionic solutions


TABLE 48.1 Bioelectric Signals Sensed by Biopotential Electrodes and Their Sources Bioelectric Signal

Abbreviation

Biologic Source

ECG — EMG EEG EOG ERG — EGG GSR

Heart—as seen from body surface Heart—as seen from within Muscle Brain Eye dipole field Eye retina Nerve or muscle Stomach Skin

Electrocardiogram Cardiac electrogram Electromyogram Electroencephalogram Electrooptigram Electroretinogram Action potential Electrogastrogram Galvanic skin reflex

of the body. The interaction between electrons in the electrodes and ions in the body can greatly affect the performance of these sensors and requires that specific considerations be made in their application. At the interface between an electrode and an ionic solution redox (oxidation-reduction), reactions need to occur for a charge to be transferred between the electrode and the solution. These reactions can be represented in general by the following equations:

C 1C n+ + ne −

(48.1)

Am− 1A + me −

(48.2)

where n is the valence of cation material C, and m is the valence of anion material, A. For most electrode systems, the cations in solution and the metal of the electrodes are the same, so the atoms C are oxidized when they give up electrons and go into solution as positively charged ions. These ions are reduced when the process occurs in the reverse direction. In the case of the anion reaction, Eq. (48.2), the directions for oxidation and reduction are reversed. For best operation of the electrodes, these two reactions should be reversible, that is, it should be just as easy for them to occur in one direction as the other. The interaction between a metal in contact with a solution of its ions produces a local change in the concentration of the ions in solution near the metal surface. This causes charge neutrality not to be maintained in this region, causing the electrolyte surrounding the metal to be at a different electrical potential from the rest of the solution. Thus, a potential difference known as the half-cell potential is established between the metal and the bulk of the electrolyte. It is found that different characteristic potentials occur for different materials, and some of these potentials are summarized in Table 48.2. These half-cell potentials can be important when using electrodes for low frequency or dc measurements. The relationship between electric potential and ionic concentrations or, more precisely, ionic activities is frequently considered in electrochemistry. Most commonly two ionic solutions of different activity are separated by an ion-selective semipermeable membrane that allows one type of ion to pass freely through TABLE 48.2 Half-cell Potentials for Materials and Reactions Encountered in Biopotential Measurement


Metal and Reaction

Half-cell Potential, V

Al → Al3+ + 3e– Ni → Ni2+ + 2e– H2 → 2H+ + 2e– Ag + Cl – → AgCl + e– Ag → Ag+ + e– Au → Au+ + e–

–1.706 –0.230 0.000 (by definition) +0.223 +0.799 +1.680

the membrane. It can be shown that an electric potential E will exist between the solutions on either side of the membrane, based upon the relative activity of the permeable ions in each of these solutions. This relationship is known as the Nernst equation

E=−

RT  a1  ln   nF  a2 

(48.3)

where a1 and a2 are the activities of the ions on either side of the membrane, R is the universal gas constant, T is the absolute temperature, n is the valence of the ions, and F is the Faraday constant. More detail on this relationship can be found in Chapter 49. When no electric current flows between an electrode and the solution of its ions or across an ionpermeable membrane, the potential observed should be the half-cell potential or the Nernst potential, respectively. If, however, there is a current, these potentials can be altered. The difference between the potential at zero current and the measured potentials while current is passing is known as the over voltage and is the result of an alteration in the charge distribution in the solution in contact with the electrodes or the ion-selective membrane. This effect is known as polarization and can result in diminished electrode performance, especially under conditions of motion. There are three basic components to the polarization over potential: the ohmic, the concentration, and the activation over potentials. Of these, the activation over potential is of greatest concern in bioelectric measurements. More details on these over potentials can be found in electrochemistry or biomedical instrumentation texts [4]. Perfectly polarizable electrodes pass a current between the electrode and the electrolytic solution by changing the charge distribution within the solution near the electrode. Thus, no actual current crosses the electrode-electrolyte interface. Nonpolarized electrodes, however, allow the current to pass freely across the electrode-electrolyte interface without changing the charge distribution in the electrolytic solution adjacent to the electrode. Although these types of electrodes can be described theoretically, neither can be fabricated in practice. It is possible, however, to come up with electrode structures that closely approximate their characteristics. Electrodes made from noble metals such as platinum are often highly polarizable. A charge distribution different from that of the bulk electrolytic solution is found in the solution close to the electrode surface. Such a distribution can create serious limitations when movement is present and the measurement involves low frequency or even dc signals. If the electrode moves with respect to the electrolytic solution, the charge distribution in the solution adjacent to the electrode surface will change, and this will induce a voltage change in the electrode that will appear as motion artifact in the measurement. Thus, for most biomedical measurements, nonpolarizable electrodes are preferred to those that are polarizable. The silver–silver chloride electrode is one that has characteristics similar to a perfectly nonpolarizable electrode and is practical for use in many biomedical applications. The electrode (Fig. 48.1a) consists of a silver base structure that is coated with a layer of the ionic compound silver chloride. Some of the silver chloride when exposed to light is reduced to metallic silver, so a typical silver–silver chloride electrode has finely divided metallic silver within a matrix of silver chloride on its surface. Since the silver chloride is relatively insoluble in aqueous solutions, this surface remains stable. Because there is minimal polarization associated with this electrode, motion artifact is reduced compared to polarizable electrodes such as the platinum electrode. Furthermore, due to the reduction in polarization, there is also a smaller effect of frequency on electrode impedance, especially at low frequencies. Silver–silver chloride electrodes of this type can be fabricated by starting with a silver base and electrolytically growing the silver chloride layer on its surface [4]. Although an electrode produced in this way can be used for most biomedical measurements, it is not a robust structure, and pieces of the silver chloride film can be chipped away after repeated use of the structure. A structure with greater mechanical stability is the sintered silver–silver chloride electrode in Fig. 48.1b. This electrode consists of a silver lead wire surrounded by a sintered cylinder made up of finely divided silver and silver-chloride powder pressed together. © 2000 by CRC Press LLC

FIGURE 48.1 Silver–silver electrodes for biopotential measurements: (a) metallic silver with a silver chloride surface layer and (b) sintered electrode structure. The lower views show the electrodes in cross-section.

In addition to its nonpolarizable behavior, the silver–silver chloride electrode exhibits less electrical noise than the equivalent polarizable electrodes. This is especially true at low frequencies, and so silver–silver chloride electrodes are recommended for measurements involving very low voltages for signals that are made up primarily of low frequencies. A more detailed description of silver–silver chloride electrodes and methods to fabricate these devices can be found in Janz and Ives [5] and biomedical instrumentation textbooks [4].

48.2 Electric Characteristics The electric characteristics of biopotential electrodes are generally nonlinear and a function of the current density at their surface. Thus, having the devices represented by linear models requires that they be operated at low potentials and currents. Under these idealized conditions, electrodes can be represented by an equivalent circuit of the form shown in Fig. 48.2. In this circuit Rd and Cd are components that represent the impedance associated with the electrode-electrolyte interface and polarization at this interface. Rs is the series resistance associated with interfacial effects and the resistance of the electrode materials themselves.

FIGURE 48.2


The equivalent circuit for a biopotential electrode.

FIGURE 48.3 An example of biopotential electrode impedance as a function of frequency. Characteristic frequencies will be somewhat different for electrode different geometries and materials. TABLE 48.3

The Effect of Electrode Properties on Electrode Impedance

Property Surface area Polarization Surface roughness Radius of curvature Surface contamination

Change in Property

Changes in Electrode Impedance

↑ ↑ ↑ ↑ ↑

↓ ↑ At low frequencies ↓ ↓ ↑

The battery Ehc represents the half-cell potential described above. It is seen that the impedance of this electrode will be frequency dependent, as illustrated in Fig. 48.3. At low frequencies the impedance is dominated by the series combination of Rs and Rd , whereas at higher frequencies Cd bypasses the effect of Rd so that the impedance is now close to Rs . Thus, by measuring the impedance of an electrode at high and low frequencies, it is possible to determine the component values for the equivalent circuit for that electrode. The electrical characteristics of electrodes are affected by many physical properties of these electrodes. Table 48.3 lists some of the more common physical properties of electrodes and qualitatively indicates how these can affect electrode impedance.

48.3 Practical Electrodes for Biomedical Measurements Many different forms of electrodes have been developed for different types of biomedical measurements. To describe each of these would go beyond the constraints of this article, but some of the more commonly used electrodes are presented in this section. The reader is referred to the monograph by Geddes for more details and a wider selection of practical electrodes [6].

Body-Surface Biopotential Electrodes This category includes electrodes that can be placed on the body surface for recording bioelectric signals. The integrity of the skin is not compromised when these electrodes are applied, and they can be used for short-term diagnostic recording such as taking a clinical electrocardiogram or long-term chronic recording such as occurs in cardiac monitoring.


Metal Plate Electrodes The basic metal plate electrode consists of a metallic conductor in contact with the skin with a thin layer of an electrolyte gel between the metal and the skin to establish this contact. Examples of metal plate electrodes are seen in Fig. 48.4a. Metals commonly used for this type of electrode include German silver (a nickel-silver alloy), silver, gold, and platinum. Sometimes these electrodes are made of a foil of the metal so as to be flexible, and sometimes they are produced in the form of a suction electrode (Fig. 48.4b) to make it easier to attach the electrode to the skin to make a measurement and then move it to another point to repeat the measurement. These types of electrodes are used primarily for diagnostic recordings of biopotentials such as the electrocardiogram or the electroencephalogram. Metal disk electrodes with a gold surface in a conical shape such as shown in Fig. 48.4c are frequently used for EEG recordings. The apex of the cone is open so that electrolyte gel or paste can be introduced to both make good contact between the electrode and the head and to allow this contact medium to be replaced should it dry out during its use.

FIGURE 48.4 Examples of different skin electrodes: (a) metal plate electrodes, (b) suction electrode for ECG, (c) metal cup EEG electrode, (d) recessed electrode, (e) disposable electrode with electrolyte-impregnated sponge (shown in cross-section), ( f ) disposable hydrogel electrode (shown in cross-section), (g) thin-film electrode for use with neonates (shown in cross-section), (h) carbon-filled elastomer dry electrode.




Electrodes for Chronic Patient Monitoring Long-term monitoring of biopotentials such as the electrocardiogram as performed by cardiac monitors places special constraints on the electrodes used to pick up the signals. These electrodes must have a stable interface between them and the body, and frequently nonpolarizable electrodes are, therefore, the best for this application. Mechanical stability of the interface between the electrode and the skin can help to reduce motion artifact, and so there are various approaches to reduce interfacial motion between the electrode and the coupling electrolyte or the skin. Figure 48.4d is an example of one approach to reduce motion artifact by recessing the electrode in a cup of electrolytic fluid or gel. The cup is then securely fastened to the skin surface using a double-sided adhesive ring. Movement of the skin with respect to the electrode may affect the electrolyte near the skin-electrolyte interface, but the electrode-electrolyte interface can be several millimeters away from this location, since it is recessed in the cup. The fluid movement is unlikely to affect the recessed electrode-electrolyte interface as compared to what would happen if the electrode was separated from the skin by just a thin layer of electrolyte. The advantages of the recessed electrode can be realized in a simpler design that lends itself to mass production through automation. This results in low per-unit cost so that these electrodes can be considered disposable. Figure 48.4e illustrates such an electrode in cross section. The electrolyte layer now consists of an open-celled sponge saturated with a thickened (high-viscosity) electrolytic solution. The sponge serves the same function as the recess in the cup electrodes and is coupled directly to a silver–silver chloride electrode. Frequently, the electrode itself is attached to a clothing snap through an insulatingadhesive disk that holds the structure against the skin. This snap serves as the point of connection to a lead wire. Many commercial versions of these electrodes in various sizes are available, including electrodes with a silver–silver chloride interface or ones that use metallic silver as the electrode material. A recently developed modification of this basic monitoring electrode structure is shown in Fig. 48.4f. In this case the metal electrode is a silver foil with a surface coating of silver chloride. The foil gives the electrode increased flexibility to fit more closely over body contours. Instead of using the sponge, a hydrogel film (really a sponge on a microscopic level) saturated with an electrolytic solution and formed from materials that are very sticky is placed over the electrode surface. The opposite surface of the hydrogel layer can be attached directly to the skin, and since it is very sticky, no additional adhesive is needed. The mobility and concentration of ions in the hydrogel layer is generally lower than for the electrolytic solution used in the sponge or the cup. This results in an electrode that has a higher source impedance as compared to these other structures. An important advantage of this structure is its ability to have the electrolyte stick directly on the skin. This greatly reduces interfacial motion between the skin surface and the electrolyte, and hence there is a smaller amount of motion artifact in the signal. This type of hydrogel electrode is, therefore, especially valuable in monitoring patients who move a great deal or during exercise. Thin-film flexible electrodes such as shown in Fig. 48.4g have been used for monitoring neonates. They are basically the same as the metal plate electrodes; only the thickness of the metal in this case is less than a micrometer. These metal films need to be supported on a flexible plastic substrate such as polyester or polyimide. The advantage of using only a thin metal layer for the electrode lies in the fact that these electrodes will then become x-ray transparent. This is especially important in infants where repeated placement and removal of electrodes, so that x-rays may be taken, can cause substantial skin irritation. Electrodes that do not use artificially applied electrolyte solutions or gels and, therefore, are often referred to as dry electrodes have been used in some monitoring applications. These sensors as illustrated in Fig. 48.4h can be placed on the skin and held in position by an elastic band or tape. They are made up of a graphite or metal-filled polymer such as silicone. The conducting particles are ground into a fine powder, and this is added to the silicone elastomer before it cures so to produce a conductive material with physical properties similar to that of the elastomer. When held against the skin surface, these electrodes establish contact with the skin without the need for an electrolytic fluid or gel. In actuality such a layer is formed by sweat under the electrode surface. For this reason these electrodes tend to perform better after they have been left in place for an hour or two so that this layer forms. Some investigators have found that placing a drop of physiologic saline solution on the skin before applying


the electrode accelerates this process. This type of electrode has found wide application in home infant cardiorespiratory monitoring because of the ease with which it can be applied by untrained caregivers.

Intracavitary and Intratissue Electrodes Electrodes can be placed within the body for biopotential measurements. These electrodes are generally smaller than skin surface electrodes and do not require special electrolytic coupling fluid, since natural body fluids serve this function. There are many different designs for these internal electrodes, and only a few examples are given in the following paragraphs. Basically these electrodes can be classified as needle electrodes, which can be used to penetrate the skin and tissue to reach the point where the measurement is to be made, or they are electrodes that can be placed in a natural cavity or surgically produced cavity in tissue. Figure 48.5 illustrates some of these internal electrodes. A catheter tip or probe electrode is placed in a naturally occurring cavity in the body such as in the gastrointestinal system. A metal tip or segment on a catheter makes up the electrode. The catheter or, in the case where there is no hollow lumen, probe, is inserted into the cavity so that the metal electrode

FIGURE 48.5 Examples of different internal electrodes: (a) catheter or probe electrode, (b) needle electrode, (c) coaxial needle electrode, (d) coiled wire electrode. (Reprinted with permission from Webster JG (ed). 1992. Medical Instrumentation: Application and Design, Houghton Mifflin, Boston.)


makes contact with the tissue. A lead wire down the lumen of the catheter or down the center of the probe connects the electrode to the external circuitry. The basic needle electrode shown in Fig. 48.5b consists of a solid needle, usually made of stainless steel, with a sharp point. An insulating material coats the shank of the needle up to a millimeter or two of the tip so that the very tip of the needle remains exposed. When this structure is placed in tissue such as skeletal muscle, electrical signals can be picked up by the exposed tip. One can also make needle electrodes by running one or more insulated wires down the lumen of a standard hypodermic needle. The electrode as shown in Fig. 48.5c is shielded by the metal of the needle and can be used to pick up very localized signals in tissue. Fine wires can also be introduced into tissue using a hypodermic needle, which is then withdrawn. This wire can remain in tissue for acute or chronic measurements. Caldwell and Reswick have used fine coiled wire electrodes in skeletal muscle for several years without adverse effects [7].

Microelectrodes The electrodes described in the previous paragraphs have been applied to studying bioelectric signals at the organism, organ, or tissue level but not at the cellular level. To study the electric behavior of cells, electrodes that are themselves smaller than the cells being studied need to be used. Three types of electrodes have been described for this purpose: etched metal electrodes, micropipette electrodes, and metal-film-coated micropipette electrodes. The metal microelectrode is essentially a subminiature version of the needle electrode described in the previous section (Fig. 48.6a). In this case, a strong metal such as tungsten is used. One end of this wire is etched electrolytically to give tip diameters on the order of a few micrometers. The structure is insulated up to its tip, and it can be passed through the membrane of a cell to contact the cytosol. The advantage of these electrodes is that they are both small and robust and

FIGURE 48.6 Microelectrodes: (a) metal, (b) micropipette, (c) thin metal film on micropipette. (Reprinted with permission from Webster JC (ed). 1992. Medical Instrumentation: Application and Design, Houghton Mifflin, Boston.) © 2000 by CRC Press LLC

can be used for neurophysiologic studies. Their principal disadvantage is the difficulty encountered in their fabrication and their high source impedance. The second and most frequently used type of microelectrode is the glass micropipette. This structure, as illustrated in Fig. 48.6b consists of a fine glass capillary drawn to a very narrow point and filled with an electrolytic solution. The point can be as narrow as a fraction of a micrometer, and the dimensions of this electrode are strongly dependent on the skill of the individual drawing the tip. The electrolytic solution in the lumen serves as the contact between the interior of the cell through which the tip has been impaled and a larger conventional electrode located in the shank of the pipette. These electrodes also suffer from high source impedances and fabrication difficulty. A combined form of these two types of electrodes can be achieved by depositing a metal film over the outside surface of a glass micropipette as shown in Fig. 48.6c. In this case, the strength and smaller dimensions of the micropipette can be used to support films of various metals that are insulated by an additional film up to a point very close to the actual tip of the electrode structure. These electrodes have been manufactured in quantity and made available as commercial products. Since they combine the features of both the metal and the micropipette electrodes, they also suffer from many of the same limitations. They do, however, have the advantage of flexibility due to the capability of being able to make films of different metals on the micropipette surface without having to worry about the strength of the metal, as would be the case if the metal were used alone.

Electrodes Fabricated Using Microelectronic Technology Modern microelectronic technology can be used to fabricate many different types of electrodes for specific biomedical applications. For example, dry electrodes with high source resistances or microelectrodes with similar characteristics can be improved by incorporating a microelectronic amplifier for impedance conversion right on the electrode itself. In the case of the conventional-sized electrodes, a metal disk 5–10 mm in diameter can have a high input impedance microelectronic amplifier configured as a follower integrated into the back of the electrode so that localized processing of the high source impedance signal can produce one of lower, more practical impedance for signal transmission [8]. Single- and multiple-element electrodes can be made from thin-film or silicon technology. Mastrototaro and colleagues have demonstrated probes for measuring intramyocardial potentials using thin, patterned gold films on polyimide or oxidised molybdenum substrates [9]. When electrodes are made from pieces of micromachined silicon, it is possible to integrate an amplifier directly into the electrode [10]. Multichannel amplifiers or multiplexers can be used with multiple electrodes on the same probe. Electrodes for contact with individual nerve fibers can be fabricated using micromachined holes in a silicon chip that are just big enough to pass a single growing axon. Electrical contacts on the sides of these holes can then be used to pick up electrical activity from these nerves [11]. These examples are just a few of the many possibilities that can be realized using microelectronics and three-dimensional micromachining technology to fabricate specialized electrodes.

48.4 Biomedical Applications Electrodes can be used to perform a wide variety of measurements of bioelectric signals. An extensive review of this would be beyond the scope of this chapter, but some typical examples of applications are highlighted in Table 48.4. The most popular application for biopotential electrodes is in obtaining the electrocardiogram for diagnostic and patient-monitoring applications. A substantial commercial market exists for various types of electrocardiographic electrodes, and many of the forms described in the previous section are available commercially. Other electrodes for measuring bioelectric potentials for application in diagnostic medicine are indicated in Table 48.4. Research applications of biopotential electrodes are highly varied and specific for individual studies. Although a few examples are given in Table 48.4, the field is far too broad to be completely covered here. Biopotential electrodes are one of the most common biomedical sensors used in clinical medicine. Although their basic principle of operation is the same for most applications, they take on many forms © 2000 by CRC Press LLC

TABLE 48.4

Examples of Applications of Biopotential Electrodes

Application

Biopotential

Cardiac monitoring

ECG

Infant cardiopulmonary monitoring

ECG impedance

Sleep encephalography

EEG

Diagnostic muscle activity Cardiac electrograms Implanted telemetry of biopotentials

EMG Electrogram ECG EMG EOG

Eye movement

Type of Electrode Ag/AgCl with sponge Ag/AgCl with hydrogel Ag/AgCl with sponge Ag/AgCl with hydrogel Thin-film Filled elastomer dry Gold cups Ag/AgCl cups Active electrodes Needle Intracardiac probe Stainless steel wire loops Platinum disks Ag/AgCl with hydrogel

and are used in the measurement of many types of bioelectric phenomena. They will continue to play an important role in biomedical instrumentation systems.

References 1. Yoshida T, Hayashi K, Toko K. 1988. The effect of anoxia on the spatial pattern of electric potential formed along the root. Ann Bot 62(5):497. 2. Plonsey R, Barr RC. 1988. Bioelectricity, New York, Plenum. 3. Weast RC (ed). 1974. Handbook of Chemistry and Physics, 55th ed, Boca Raton, Fla, CRC Press. 4. Webster JG (ed). 1992. Medical Instrumentation: Application and Design, Boston, Houghton Mifflin. 5. Janz GI, Ives DJG. 1968. Silver–silver chloride electrodes. Ann N Y Acad Sci 148:210. 6. Geddes LA. 1972. Electrodes and the Measurement of Bioelectric Events, New York, Wiley. 7. Caldwell CW, Reswick JB. 1975. A percutaneous wire electrode for chronic research use. IEEE Trans Biomed Eng 22:429. 8. Ko WH, Hynecek J. 1974. Dry electrodes and electrode amplifiers. In HA Miller, Harrison DC (eds), Biomedical Electrode Technology, pp 169–181, New York, Academic Press. 9. Mastrototaro JJ, Massoud HZ, Pilkington TC, et al., 1992. Rigid and flexible thin-film microelectrode arrays for transmural cardiac recording. IEEE Trans Biomed Eng 39(3):271. 10. Wise KD, Najafi K, Ji J, et al. 1990. Micromachined silicon microprobes for CNS recording and stimulation. Proc Ann Conf IEEE Eng Med Biol Soc 12:2334. 11. Edell DJ. 1986. A peripheral nerve information transducer for amputees: Long-term multichannel recordings from rabbit peripheral nerves. IEEE Trans Biomed Eng 33:203.

Further Information Good overviews of biopotential electrodes are found in Geddes LA. 1972. Electrodes and the Measurement of Bioelectric Events, New York, Wiley; and Ferris CD. 1974. Introduction to Bioelectrodes, New York, Plenum. Even though these references are more than 20 years old, they clearly cover the field, and little has changed since these books were written. Overviews of biopotential electrodes are found in chapters of two works edited by John Webster. Chapter 5 of his textbook, Medical Instrumentation: Application and Design, covers the material of this chapter in more detail, and there is a section on “Bioelectrodes” in his Encyclopedia on Medical Devices and Instrumentation, published by Wiley in 1988. The journals IEEE Transactions on Biomedical Engineering and Medical and Biological Engineering and Computing are good sources of recent research on biopotential electrodes.


Liu, C. C. “Electrochemical Sensors.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

49 Electrochemical Sensors Chung-Chiun Liu Electronics Design Center and Edison Sensor Technology Center, Case Western Reserve University

49.1 49.2 49.3 49.4 49.5

Conductivity/Capacitance Electrochemical Sensors Potentiometric Sensors Voltammetric Sensors Reference Electrodes Summary

Electrochemical sensors have been used extensively either as a whole or an integral part of a chemical and biomedical sensing element. For instance, blood gas (PO2, PCO2, and pH) sensing can be accomplished entirely by electrochemical means. Many important biomedical enzymatic sensors, including glucose sensors, incorporate an enzymatic catalyst and an electrochemical sensing element. The Clark type of oxygen sensor [Clark, 1956] is a well-known practical biomedical sensor based on electrochemical principles, an amperometric device. Electrochemical sensors generally can be categorized as conductivity/capacitance, potentiometric, amperometric, and voltammetric sensors. The amperometric and voltammetric sensors are characterized by their current-potential relationship with the electrochemical system and are less well-defined. Amperometric sensors can also be viewed as a subclass of voltammetric sensors. Electrochemical sensors are essentially an electrochemical cell which employs a two or three-electrode arrangement. Electrochemical sensor measurement can be made at steady-state or transient. The applied current or potential for electrochemical sensors may vary according to the mode of operation, and the selection of the mode is often intended to enhance the sensitivity and selectivity of a particular sensor. The general principles of electrochemical sensors have been extensively discussed in many electroanalytic references. However, many electroanalytic methods are not practical in biomedical sensing applications. For instance, dropping mercury electrode polarography is a well-established electroanalytic method, yet its usefulness in biomedical sensor development, particularly for potential in vivo sensing, is rather limited. In this chapter, we shall focus on the electrochemical methodologies which are useful in biomedical sensor development.

49.1 Conductivity/Capacitance Electrochemical Sensors Measurement of the conductivity of an electrochemical cell can be the basis for an electrochemical sensor. This differs from an electrical (physical) measurement, for the electrochemical sensor measures the conductivity change of the system in the presence of a given solute concentration. This solute is often the sensing species of interest. Electrochemical sensors may also involve a capacitative impedance resulting from the polarization of the electrodes and the faradaic or charge transfer processes. It has been established that the conductance of a homogeneous solution is directly proportional to the cross-sectional area perpendicular to the electrical field and inversely proportional to the segment of


solution along the electrical field. Thus, the conductance of this solution (electrolyte), G (Ω–1), can be expressed as

G=σ A L

(49.1)

where A is the cross-sectional area (in cm2), L is the segment of the solution along the electrical field (in cm), and σ (in Ω–1 cm–1) is the specific conductivity of the electrolyte and is related quantitatively to the concentration and the magnitude of the charges of the ionic species. For a practical conductivity sensor, A is the surface of the electrode, and L is the distance between the two electrodes. Equivalent and molar conductivities are commonly used to express the conductivity of the electrolyte. Equivalent conductance depends on the concentration of the solution. If the solution is a strong electrolyte, it will completely dissociate the components in the solution to ionic forms. Kohlrauch [MacInnes, 1939] found that the equivalent conductance of a strong electrolyte was proportional to the square root of its concentration. However, if the solution is a weak electrolyte which does not completely dissociate the components in the solution to respective ions, the above observation by Kohlrauch is not applicable. The formation of ions leads to consideration of their contribution to the overall conductance of the electrolyte. The equivalent conductance of a strong electrolyte approaches a constant limiting value at infinite dilution, namely,

Λ o = Λ lim→o = λ+o + λ−o

(49.2)

where Λo is the equivalent conductance of the electrolyte at infinite dilution and λ+o and λ–o are the ionic equivalent conductance of cations and anions at infinite dilution, respectively. Kohlrauch also established the law of independent mobilities of ions at infinite dilution. This implies that Λo at infinite dilution is a constant at a given temperature and will not be affected by the presence of other ions in the electrolytes. This provides a practical estimation of the value of Λo from the values of λ+o and λ–o. As mentioned, the conductance of an electrolyte is influenced by its concentration. Kohlrausch stated that the equivalent conductance of the electrolyte at any concentration C in mol/l or any other convenient units can be expressed as

Λ = Λ o − βC 0.5

(49.3)

where β is a constant depending on the electrolyte. In general, electrolytes can be classified as weak electrolytes, strong electrolytes, and ion-pair electrolytes. Weak electrolytes only dissociate to their component ions to a limited extent, and the degree of the dissociation is temperature dependent. However, strong electrolytes dissociate completely, and Eq. (49.3) is applicable to evaluate its equivalent conductance. Ion-pair electrolytes can by characterized by their tendency to form ion pairs. The dissociation of ion pairs is similar to that of a weak electrolyte and is affected by ionic activities. The conductivity of ion-pair electrolytes is often nonlinear related to its concentration. The electrolyte conductance measurement technique, in principle, is relatively straightforward. However, the conductivity measurement of an electrolyte is often complicated by the polarization of the electrodes at the operating potential. Faradaic or charge transfer processes occur at the electrode surface, complicating the conductance measurement of the system. Thus, if possible, the conductivity electrochemical sensor should operate at a potential where no faradaic processes occur. Also, another important consideration is the formation of the double layer adjacent to each electrode surface when a potential is imposed on the electrochemical sensor. The effect of the double layer complicates the interpretation of the conductivity measurement and is usually described by the Warburg impedance. Thus, even in the absence of faradaic processes, the potential effect of the double layer on the conductance of the electrolyte must be carefully assessed. The influence of a faradaic process can be minimized by maintaining a high center constant, L/A, of the electrochemical conductivity sensor, so that the cell resistance lies in the © 2000 by CRC Press LLC

region of 1–50 kΩ . This implies the desirable feature of a small electrode surface area and a relatively large distance between the two electrodes. Yet, a large electrode surface area enhances the accuracy of the measurement, since a large deviation from the null point facilitates the balance of the Wheatstone bridge, resulting in improvement of sensor sensitivity. These opposing features can be resolved by using a multiple-sensing electrode configuration in which the surface area of each electrode element is small compared to the distance between the electrodes. The multiple electrodes are connected in parallel, and the output of the sensor represents the total sum of the current through each pair of electrodes. In this mode of measurement, the effect of the double layer is included in the conductance measurement. The effects of both the double layers and the faradaic processes can be minimized by using a high-frequency, low-amplitude alternating current. The higher the frequency and the lower the amplitude of the imposed alternating current, the closer the measured value is to the true conductance of the electrolyte.

49.2 Potentiometric Sensors When a redox reaction, Ox + Ze = Red, takes place at an electrode surface in an electrochemical cell, a potential may develop at the electrode-electrolyte interface. This potential may then be used to quantify the activity (on concentration) of the species involved in the reaction forming the fundamental of potentiometric sensors. The above reduction reaction occurs at the surface of the cathode and is defined as a half-cell reaction. At thermodynamic equilibrium, the Nernst equation is applicable and can be expressed as:

E = Eo +

RT  aox  ln   ZF  a red 

(49.4)

where E and Eo are the measured electrode potential and the electrode potential at standard state, respectively; aox and ared are the activities of Ox (reactant in this case) and Red (product in this case), respectively; Z is the number of electrons transferred, F the Faraday constant, R the gas constant, and T the operating temperature in the absolute scale. In the electrochemical cell, two half-cell reactions will take place simultaneously. However, for sensing purposes, only one of the two half-cell reactions should involve the species of interest, and the other half-cell reaction is preferably reversible and noninterfering. As indicated in Eq. (49.4), a linear relation exists between the measured potential E and the natural logarithm of the ratio of the activities of the reactant and product. If the number of electrons transferred, Z, is one, at ambient temperature (25°C or 298°K) the slope is approximately 60 mV/decade. This slope value governs the sensitivity of the potentiometric sensor. Potentiometric sensors can be classified based on whether the electrode is inert or active. An inert electrode does not participate in the half-cell reaction and merely provides the surface for the electron transfer or provides a catalytic surface for the reaction. However, an active electrode is either an ion donor or acceptor in the reaction. In general, there are three types of active electrodes: the metal/metal ion, the metal/insoluble salt or oxide, and metal/metal chelate electrodes. Noble metals such as platinum and gold, graphite, and glassy carbon are commonly used as inert electrodes on which the half-cell reaction of interest takes place. To complete the circuitry for the potentiometric sensor, the other electrode is usually a reference electrode on which a noninterference half-cell reaction occurs. Silver–silver chloride and calomel electrodes are the most commonly used reference electrodes. Calomel consists of Hg/HgCl2 and is less desirable for biomedical systems in terms of toxicity. An active electrode may incorporate chemical or biocatalysts and is involved as either an ion donor or acceptor in the half-cell reaction. The other half-cell reaction takes place on the reference electrode and should also be noninterfering. If more than a single type of ion contributes to the measured potential in Eq. (49.4), the potential can no longer be used to quantify the ions of interest. This is the interference in a potentiometric sensor. © 2000 by CRC Press LLC

Thus, in many cases, the surface of the active electrode often incorporates a specific functional membrane which may be ion-selective, ion-permeable, or have ion-exchange properties. These membranes tend to selectivity permit the ions of interest to diffuse or migrate through. This minimizes the ionic interference. Potentiometric sensors operate at thermodynamic equilibrium conditions. Thus, in practical potentiometric sensing, the potential measurement needs to be made under zero-current conditions. Consequently, a high-input impedance electrometer is often used for measurements. Also, the response time for a potentiometric sensor to reach equilibrium conditions in order to obtain a meaningful reading can be quite long. These considerations are essential in the design and selection of potentiometric sensors for biomedical applications.

49.3 Voltammetric Sensors The current-potential relationship of an electrochemical cell provides the basis for voltammetric sensors. Amperometric sensors, that are also based on the current-potential relationship of the electrochemical cell, can be considered a subclass of voltammetric sensors. In amperometric sensors, a fixed potential is applied to the electrochemical cell, and a corresponding current, due to a reduction or oxidation reaction, is then obtained. This current can be used to quantify the species involved in the reaction. The key consideration of an amperometric sensor is that it operates at a fixed potential. However, a voltammetric sensor can operate in other modes such as linear cyclic voltammetric modes. Consequently, the respective current potential response for each mode will be different. In general, voltammetric sensors examine the concentration effect of the detecting species on the current-potential characteristics of the reduction or oxidation reaction involved. The mass transfer rate of the detecting species in the reaction onto the electrode surface and the kinetics of the faradaic or charge transfer reaction at the electrode surface directly affect the currentpotential characteristics. This mass transfer can be accomplished through (a) an ionic migration as a result of an electric potential gradient, (b) a diffusion under a chemical potential difference or concentration gradient, and (c) a bulk transfer by natural or forced convection. The electrode reaction kinetics and the mass transfer processes contribute to the rate of the faradaic process in an electrochemical cell. This provides the basis for the operation of the voltammetric sensor. However, assessment of the simultaneous mass transfer and kinetic mechanism is rather complicated. Thus, the system is usually operated under definitive hydrodynamic conditions. Various techniques to control either the potential or current are used to simplify the analysis of the voltammetric measurement. A description of these techniques and their corresponding mathematical analyses are well documented in many texts on electrochemistry or electroanalysis [Adams, 1969; Bard & Faulkner, 1980; Lingane, 1958; Macdonald, 1977; Murray & Reilley, 1966]. A preferred mass transfer condition is total diffusion, which can be described by Fick’s law of diffusion. Under this condition, the cell current, a measure of the rate of the faradaic process at an electrode, usually increases with increases in the electrode potential. This current approaches a limiting value when the rate of the faradaic process at the electrode surface reaches its maximum mass transfer rate. Under this condition, the concentration of the detecting species at the electrode surface is considered as zero and is diffusional mass transfer. Consequently, the limiting current and the bulk concentration of the detecting species can be related by

i = ZFkmC *

(49.5)

where km is the mass transfer coefficient and C* is the bulk concentration of the detecting species. At the other extreme, when the electrode kinetics are slow compared with the mass transfer rate, the electrochemical system is operated in the reaction kinetic control regime. This usually corresponds to a small overpotential. The limiting current and the bulk concentration of the detecting species can be related as


i = ZFkcC *

(49.6)

where kc is the kinetic rate constant for the electrode process. Both Eqs. (49.5) and (49.6) show the linear relationship between the limiting current and the bulk concentration of the detecting species. In many cases, the current does not tend to a limiting value with an increase in the electrode potential. This is because other faradaic or nonfaradaic processes become active, and the cell current represents the cumulative rates of all active electrode processes. The relative rates of these processes, expressing current efficiency, depend on the current density of the electrode. Assessment of such a system is rather complicated, and the limiting current technique may become ineffective. When a voltammetric sensor operates with a small overpotential, the rate of faradaic reaction is also small; consequently, a high-precision instrument for the measurement is needed. An amperometric sensor is usually operated under limiting current or relatively small overpotential conditions. Amperometric sensors operate under an imposed fixed electrode potential. Under this condition, the cell current can be correlated with the bulk concentration of the detecting species (the solute). This operating mode is commonly classified as amperometric in most sensor work, but it also is referred to as the chronosuperometric method, since time is involved. Voltammetric sensors can be operated in a linear or cyclic sweep mode. Linear sweep voltammetry involves an increase in the imposed potential linearly at a constant scanning rate from an initial potential to a defined upper potential limit. This is the so-called potential window. The current-potential curve usually shows a peak at a potential where the oxidation or reduction reaction occurs. The height of the peak current can be used for the quantification of the concentration of the oxidation or reduction species. Cyclic voltammetry is similar to the linear sweep voltammetry except that the electrode potential returns to its initial value at a fixed scanning rate. The cyclic sweep normally generates the current peaks corresponding to the oxidation and reduction reactions. Under these circumstances, the peak current value can relate to the corresponding oxidation or reduction reaction. However, the voltammogram can be very complicated for a system involving adsorption (nonfaradaic processes) and charge processes (faradaic processes). The potential scanning rate, diffusivity of the reactant, and operating temperature are essential parameters for sensor operation, similar to the effects of these parameters for linear sweep voltammograms. The peak current may be used to quantify the concentration of the reactant of interest, provided that the effect of concentration on the diffusivity is negligible. The potential at which the peak current occurs can be used in some cases to identify the reaction, or the reactant. This identification is based on the half-cell potential of the electrochemical reactions, either oxidation or reduction. The values of these half-cell reactions are listed extensively in handbooks and references. The described voltammetric and amperometric sensors can be used very effectively to carry out qualitative and quantitative analyses of chemical and biochemical species. The fundamentals of this sensing technique are well established, and the critical issue is the applicability of the technique to a complex, practical environment, such as in whole blood or other biologic fluids. This is also the exciting challenge of designing a biosensor using voltammetric and amperometric principles.

49.4 Reference Electrodes Potentiometric, voltammetric, and amperometric sensors employ a reference electrode. The reference electrode in the case of potentiometric and amperometric sensors serves as a counter electrode to complete the circuitry. In either case, the reaction of interest takes place at the surface of the working electrode, and this reaction is either an oxidation or reduction reaction. Consequently, the reaction at the counter electrode, i.e., the reference electrode, is a separate reduction or oxidation reaction, respectively. It is necessary that the reaction occurring at the reference electrode does not interfere with the reaction at the working electrode. For practical applications, the reaction occurring at the reference electrode should be highly reversible and, as stated, does not contribute to the reaction at the working electrode. In electrochemistry, the hydrogen electrode is universally accepted as the primary standard with which other


electrodes are compared. Consequently, the hydrogen electrode serves extensively as a standard reference. A hydrogen reference electrode is relatively simple to prepare. However, for practical applications hydrogen reference electrodes are too cumbersome to be useful in practice. A class of electrode called the electrode of the second kind, which forms from a metal and its sparingly soluble metal salt, finds use as the reference electrode. The most common electrode of this type includes the calomel electrode, Hg/HgCl2 and the silver–silver chloride electrode, Ag/AgCl. In biomedical applications, particularly in in vivo applications, Ag/AgCl is more suitable as a reference electrode. An Ag/AgCl electrode can be small, compact, and relatively simple to fabricate. As a reference electrode, the stability and reproducibility of an Ag/AgCl electrode is very important. Contributing factors to instability and poor reproducibility of Ag/AgCl electrodes include the purity of the materials used, the aging effect of the electrode, the light effect, and so on. When in use, the electrode and the electrolyte interface contribute to the stability of the reference electrode. It is necessary that a sufficient quantity of Cl– ions exists in the electrolyte when the Ag/AgCl electrode serves as a reference. Therefore, other silver–silver halides such as Ag/AgBr or Ag/AgI electrodes are used in cases where these other halide ions are present in the electrolyte. In a voltammetric sensor, the reference electrode serves as a true reference for the working electrode, and no current flows between the working and reference electrodes. Nevertheless, the stability of the reference electrode remains essential for a voltammetric sensor.

49.5 Summary Electrochemical sensors are used extensively in many biomedical applications including blood chemistry sensors1, PO2, PCO2, and pH electrodes. Many practical enzymatic sensors, including glucose and lactate sensors, also employ electrochemical sensors as sensing elements. Electrochemically based biomedical sensors have found in vivo and in vitro applications. We believe that electrochemical sensors will continue to be an important aspect of biomedical sensor development.

References Adams RN. 1969. Electrochemistry at Solid Electrodes, New York, Marcel Dekker. Bard A, Faulkner LR. 1980. Electrochemical Methods, New York, Wiley. Clark LC Jr. 1956. Monitor and control of blood and tissue oxygen tissues. Trans. Am. Soc. Artif. Organs 2:41. Lingane JJ. 1958. Electroanalytical Chemistry, New York, London, Interscience. Macdonald DD. 1977. Transient Techniques in Electrochemistry, New York, Plenum. MacInnes DA. 1939. The Principles of Electrochemistry, New York, Reinhold. Murray RW, Reilley CN. 1996. Electroanalytical Principles, New York-London, Interscience.

1

Further information on blood-gas sensors can be found in Appendix A.


Mendelson, Y. “Optical Sensors.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

50 Optical Sensors 50.1

Instrumentation Light Source • Optical Element • Photodetectors • Signal Processing

50.2

Optical Fibers Probe Configurations • Optical Fiber Sensors • IndicatorMediated Transducers

50.3

General Principles of Optical Sensing Evanescent Wave Spectroscopy • Surface Plasmon Resonance

Yitzhak Mendelson Worcester Polytechnic Institute

50.4

Applications Oximetry • Blood Gases • Glucose Sensors • Immunosensors

Optical methods are among the oldest and best-established techniques for sensing biochemical analytes. Instrumentation for optical measurements generally consists of a light source, a number of optical components to generate a light beam with specific characteristics and to direct this light to some modulating agent, and a photodetector for processing the optical signal. The central part of an optical sensor is the modulating component, and a major part of this chapter will focus on how to exploit the interaction of an analyte with optical radiation in order to obtain essential biochemical information. The number of publications in the field of optical sensors for biomedical applications has grown significantly during the past two decades. Numerous scientific reviews and historical perspectives have been published, and the reader interested in this rapidly growing field is advised to consult these sources for additional details. This chapter will emphasize the basic concept of typical optical sensors intended for continuous in vivo monitoring of biochemical variables, concentrating on those sensors which have generally progressed beyond the initial feasibility stage and reached the promising stage of practical development or commercialization. Optical sensors are usually based on optical fibers or on planar waveguides. Generally, there are three distinctive methods for quantitative optical sensing at surfaces: 1. The analyte directly affects the optical properties of a waveguide, such as evanescent waves (electromagnetic waves generated in the medium outside the optical waveguide when light is reflected from within) or surface plasmons (resonances induced by an evanescent wave in a thin film deposited on a waveguide surface). 2. An optical fiber is used as a plain transducer to guide light to a remote sample and return light from the sample to the detection system. Changes in the intrinsic optical properties of the medium itself are sensed by an external spectrophotometer. 3. An indicator or chemical reagent placed inside, or on, a polymeric support near the tip of the optical fiber is used as a mediator to produce an observable optical signal. Typically, conventional techniques, such as absorption spectroscopy and fluorimetry, are employed to measure changes in the optical signal.


FIGURE 50.1 applications.

General diagram representing the basic building blocks of an optical instrument for optical sensor

50.1 Instrumentation The actual implementation of instrumentation designed to interface with optical sensors will vary greatly depending on the type of optical sensor used and its intended application. A block diagram of a generic instrument is illustrated in Fig. 50.1. The basic building blocks of such an instrument are the light source, various optical elements, and photodetectors.

Light Source A wide selection of light sources are available for optical sensor applications. These include: highly coherent gas and semiconductor diode lasers, broad spectral band incandescent lamps, and narrow-band, solid-state, light-emitting diodes (LEDs). The important requirement of a light source is obviously good stability. In certain applications, for example in portable instrumentation, LEDs have significant advantages over other light sources because they are small and inexpensive, consume lower power, produce selective wavelengths, and are easy to work with. In contrast, tungsten lamps provide a broader range of wavelengths, higher intensity, and better stability but require a sizable power supply and can cause heating problems inside the apparatus.

Optical Elements Various optical elements are used routinely to manipulate light in optical instrumentation. These include lenses, mirrors, light choppers, beam splitters, and couplers for directing the light from the light source into the small aperture of a fiber optic sensor or a specific area on a waveguide surface and collecting the light from the sensor before it is processed by the photodetector. For wavelength selection, optical filters, prisms, and diffraction gratings are the most common components used to provide a narrow bandwidth of excitation when a broadwidth light source is utilized.

Photodetectors In choosing photodetectors for optical sensors, a number of factors must be considered. These include sensitivity, detectivity, noise, spectral response, and response time. Photomultipliers and semiconductor © 2000 by CRC Press LLC

quantum photodetectors, such as photoconductors and photodiodes, are both suitable. The choice, however, is somewhat dependent on the wavelength region of interest. Generally, both give adequate performance. Photodiodes are usually more attractive because of the compactness and simplicity of the circuitry involved. Typically, two photodetectors are used in optical instrumentation because it is often necessary to include a separate reference detector to track fluctuations in source intensity and temperature. By taking a ratio between the two detector readings, whereby a part of the light that is not affected by the measurement variable is used for correcting any optical variations in the measurement system, a more accurate and stable measurement can be obtained.

Signal Processing Typically, the signal obtained from a photodetector provides a voltage or a current proportional to the measured light intensity. Therefore, either simple analog computing circuitry (e.g., a current-to-voltage converter) or direct connection to a programmable gain voltage stage is appropriate. Usually, the output from a photodetector is connected directly to a preamplifier before it is applied to sampling and analogto-digital conversion circuitry residing inside a computer. Quite often two different wavelengths of light are utilized to perform a specific measurement. One wavelength is usually sensitive to changes in the species being measured, and the other wavelength is unaffected by changes in the analyte concentration. In this manner, the unaffected wavelength is used as a reference to compensate for fluctuation in instrumentation over time. In other applications, additional discriminations, such as pulse excitation or electronic background subtraction utilizing synchronized lock-in amplifier detection, are useful, allowing improved selectivity and enhanced signal-to-noise ratio.

50.2 Optical Fibers Several types of biomedical measurements can be made by using either plain optical fibers as a remote device for detecting changes in the spectral properties of tissue and blood or optical fibers tightly coupled to various indicator-mediated transducers. The measurement relies either on direct illumination of a sample through the endface of the fiber or by excitation of a coating on the side wall surface through evanescent wave coupling. In both cases, sensing takes place in a region outside the optical fiber itself. Light emanating from the fiber end is scattered or fluoresced back into the fiber, allowing measurement of the returning light as an indication of the optical absorption or fluorescence of the sample at the fiber optic tip. Optical fibers are based on the principle of total internal reflection. Incident light is transmitted through the fiber if it strikes the cladding at an angle greater than the so-called critical angle, so that it is totally internally reflected at the core/cladding interface. A typical instrument for performing fiber optic sensing consists of a light source, an optical coupling arrangement, the fiber optic light guide with or without the necessary sensing medium incorporated at the distal tip, and a light detector. A variety of high-quality optical fibers are available commercially for biomedical sensor applications, depending on the analytic wavelength desired. These include plastic, glass, and quartz fibers which cover the optical spectrum from the UV through the visible to the near IR region. On one hand, plastic optical fibers have a larger aperture and are strong, inexpensive, flexible, and easy to work with but have poor UV transmission below 400 nm. On the other hand, glass and quartz fibers have low attenuation and better transmission in the UV but have small apertures, are fragile, and present a potential risk in in vivo applications.

Probe Configurations There are many different ways to implement fiber optic sensors. Most fiber optic chemical sensors employ either a single-fiber configuration, where light travels to and from the sensing tip in one fiber, or a doublefiber configuration, where separate optical fibers are used for illumination and detection. A single fiber optic configuration offers the most compact and potentially least expensive implementation. However,


additional challenges in instrumentation are involved in separating the illuminating signal from the composite signal returning for processing. The design of intravascular catheters requires special considerations related to the sterility and biocompatibility of the sensor. For example, intravascular fiberoptic sensors must be sterilizable and their material nonthrombogenic and resistant to platelet and protein deposition. Therefore, these catheters are typically made of materials covalently bound with heparin or antiplatelet agents. The catheter is normally introduced into the jugular vein via a peripheral cut-down and a slow heparin flush is maintained until it is removed from the blood.

Optical Fiber Sensors Advantages cited for fiber sensors include their small size and low cost. In contrast to electrical measurements, where the difference of two absolute potentials must be measured, fiber optics are self-contained and do not require an external reference signal. Because the signal is optical, there is no electrical risk to the patient, and there is no direct interference from surrounding electric or magnetic fields. Chemical analysis can be performed in real-time with almost an instantaneous response. Furthermore, versatile sensors can be developed that respond to multiple analytes by utilizing multiwavelength measurements. Despite these advantages, optical fiber sensors exhibit several shortcomings. Sensors with immobilized dyes and other indicators have limited long-term stability, and their shelf life degrades over time. Moreover, ambient light can interfere with the optical measurement unless optical shielding or special timesynchronous gating is performed.

Indicator-Mediated Transducers Only a limited number of biochemical analytes have an intrinsic optical absorption that can be measured with sufficient selectivity directly by spectroscopic methods. Other species, particularly hydrogen, oxygen, carbon dioxide, and glucose, which are of primary interest in diagnostic applications, are not susceptible to direct photometry. Therefore, indicator-mediated sensors have been developed using specific reagents that are properly immobilized on the surface of an optical sensor. The most difficult aspect of developing an optical biosensor is the coupling of light to the specific recognition element so that the sensor can respond selectively and reversibly to a change in the concentration of a particular analyte. In fiber-optic-based sensors, light travels efficiently to the end of the fiber where it exists and interacts with a specific chemical or biologic recognition element that is immobilized at the tip of the fiber optic. These transducers may include indicators and ionophores (i.e., ion-binding compounds) as well as a wide variety of selective polymeric materials. After the light interacts with the sample, the light returns through the same or a different optical fiber to a detector which correlates the degree of change with the analyte concentration. Typical indicator-mediated fiber-optic-sensor configurations are shown schematically in Fig. 50.2. In (a) the indicator is immobilized directly on a membrane positioned at the end of a fiber. An indicator in the form of a powder can be either glued directly onto a membrane, as shown in (b), or physically retained in position at the end of the fiber by a special permeable membrane (c), a tubular capillary/membrane (d), or a hollow capillary tube (e).

50.3 General Principles of Optical Sensing Two major optical techniques are commonly available to sense optical changes at sensor interfaces. These are usually based on evanescent wave and surface plasmon resonance principles.

Evanescent Wave Spectroscopy When light propagates along an optical fiber, it is not confined to the core region but penetrates to some extent into the surrounding cladding region. In this case, an electromagnetic component of the light © 2000 by CRC Press LLC

FIGURE 50.2 Typical configuration of different indicator-mediated fiber optic sensor tips (from Otto S. Wolfbeis, Fiber Optic Chemical Sensors and Biosensors, vol. 1, CRC Press, Boca Raton, 1990).

penetrates a characteristic distance (on the order of one wavelength) beyond the reflecting surface into the less optically dense medium where it is attenuated exponentially according to Beer-Lambert’s law (Fig. 50.3). The evanescent wave depends on the angle of incidence and the incident wavelength. This phenomenon has been widely exploited to construct different types of optical sensors for biomedical applications. Because of the short penetration depth and the exponential decay of the intensity, the evanescent wave is absorbed mainly by absorbing compounds very close to the surface. In the case of particularly weak absorbing analytes, sensitivity can be enhanced by combining the evanescent wave principle with multiple internal reflections along the sides of an unclad portion of a fiber optic tip. Instead of an absorbing species, a fluorophore can also be used. Light is absorbed by the fluorophore emitting detectable fluorescent light at a higher wavelength, thus providing improved sensitivity. Evanescent wave sensors have been applied successfully to measure the fluorescence of indicators in solution, for pH measurement, and in immunodiagnostics.

Surface Plasmon Resonance Instead of the dielectric/dielectric interface used in evanescent wave sensors, it is possible to arrange a dielectric/metal/dielectric sandwich layer such that when monochromatic polarized light (e.g., from a laser source) impinges on a transparent medium having a metallized (e.g., Ag or Au) surface, light is absorbed within the plasma formed by the conduction electrons of the metal. This results in a phenomenon known as surface plasmon resonance (SPR). When SPR is induced, the effect is observed as a minimum in the intensity of the light reflected off the metal surface. As is the case with the evanescent wave, an SPR is exponentially decaying into solution © 2000 by CRC Press LLC

FIGURE 50.3 Schematic diagram of the path of a light ray at the interface of two different optical materials with index of refraction n1 and n2. The ray penetrates a fraction of a wavelength (dp) beyond the interface into the medium with the smaller refractive index.

with a penetration depth of about 20 nm. The resonance between the incident light and the plasma wave depends on the angle, wavelength, and polarization state of the incident light and the refractive indices of the metal film and the materials on either side of the metal film. A change in the dielectric constant or the refractive index at the surface causes the resonance angle to shift, thus providing a highly sensitive means of monitoring surface reactions. The method of SPR is generally used for sensitive measurement of variations in the refractive index of the medium immediately surrounding the metal film. For example, if an antibody is bound to or absorbed into the metal surface, a noticeable change in the resonance angle can be readily observed because of the change of the refraction index at the surface, assuming all other parameters are kept constant (Fig. 50.4). The advantage of this concept is the improved ability to detect the direct interaction between antibody and antigen as an interfacial measurement. SPR has been used to analyze immunochemicals and to detect gases. The main limitation of SPR, however, is that the sensitivity depends on the optical thickness of the adsorbed layer, and, therefore, small molecules cannot be measured in very low concentrations.

50.4 Applications Oximetry Oximetry refers to the colorimetric measurement of the degree of oxygen saturation, that is, the relative amount of oxygen carried by the hemoglobin in the erythrocytes, by recording the variation in the color of deoxyhemoglobin (Hb) and oxyhemoglobin (HbO2). A quantitative method for measuring blood oxygenation is of great importance in assessing the circulatory and respiratory status of a patient. Various optical methods for measuring the oxygen saturation of arterial (SaO2) and mixed venous (SvO2) blood have been developed, all based on light transmission through, or reflecting from, tissue and blood. The measurement is performed at two specific wavelengths: λ1, where there is a large difference in light absorbance between Hb and HbO2 (e.g., 660 nm red light), and λ2, which can be an isobestic wavelength (e.g., 805 nm infrared light), where the absorbance of light is independent of blood oxygenation, or a different wavelength in the infrared region (>805 nm), where the absorbance of Hb is slightly smaller than that of HbO2. Assuming for simplicity that a hemolyzed blood sample consists of a two-component homogeneous mixture of Hb and HbO2, and that light absorbance by the mixture of these two components is additive, a simple quantitative relationship can be derived for computing the oxygen saturation of blood:

( )  ( ) 

 OD λ 1 Oxygen saturation = A − B   OD λ 2 

where A and B are coefficients which are functions of the specific absorptivities of Hb and HbO2, and OD is the corresponding absorbance (optical density) of the blood. Since the original discovery of this phenomenon over 50 years ago, there has been progressive development in instrumentation to measure oxygen saturation along three different paths: bench-top oximeters for clinical laboratories,


FIGURE 50.4 Surface plasmon resonance at the interface between a thin metallic surface and a liquid (A). A sharp decrease in the reflected light intensity can be observed in (B). The location of the resonance angle is dependent on the refractive index of the material present at the interface.

FIGURE 50.5

Principle of a three-fiber optical catheter for SvO2/HCT measurement [2].

fiber optic catheters for invasive intravascular monitoring, and transcutaneous sensors, which are noninvasive devices placed against the skin. Intravascular Fiber Optic SvO2 Catheters In vivo fiberoptic oximeters were first described in the early 1960s by Polanyi and Heir [1]. They demonstrated that in a highly scattering medium such as blood, where a very short path length is required for a transmittance measurement, a reflectance measurement was practical. Accordingly, they showed that a linear relationship exists between oxygen saturation and the ratio of the infrared-to-red (IR/R) light backscattered from the blood

(

oxygen saturation = a − b IR R

)

where a and b are catheter-specific calibration coefficients. Fiber optic SvO2 catheters consist of two separate optical fibers. One fiber is used for transmitting the light to the flowing blood, and a second fiber directs the backscattered light to a photodetector. In some commercial instruments (e.g., Oximetrix), automatic compensation for hematocrit is employed utilizing three, rather than two, infrared reference wavelengths. Bornzin and coworkers [2] and Mendelson and coworkers [3] described a 5-lumen, 7.5F thermodilution catheter that is comprised of three unequally spaced optical fibers, each fiber 250 µm in diameter, and provides continuous SvO2 reading with automatic corrections for hematocrit variations (Fig. 50.5). Intravenous fiberoptic catheters are utilized in monitoring SvO2 in the pulmonary artery and can be used to indicate the effectiveness of the cardiopulmonary system during cardiac surgery and in the ICU. Several problems limit the wide clinical application of intravascular fiberoptic oximeters. These include the dependence of the individual red and infrared backscattered light intensities and their ratio on hematocrit (especially for SvO2 below 80%), blood flow, motion artifacts due to catheter tip “whipping” against the blood vessel wall, blood temperature, and pH. Noninvasive Pulse Oximetry Noninvasive monitoring of SaO2 by pulse oximetry is a rapidly growing practice in many fields of clinical medicine [4]. The most important advantage of this technique is the capability to provide continuous, safe, and effective monitoring of blood oxygenation at the patient’s bedside without the need to calibrate the instrument before each use.


FIGURE 50.6

Disposable finger probe of a noninvasive pulse oximeter.

Pulse oximetry, which was first suggested by Aoyagi and colleagues [5] and Yoshiya and colleagues [6], relies on the detection of the time-variant photoplethysmographic signal, caused by changes in arterial blood volume associated with cardiac contraction. SaO2 is derived by analyzing only the time-variant changes in absorbance caused by the pulsating arterial blood at the same red and infrared wavelengths used in conventional invasive type oximeters. A normalization process is commonly performed by which the pulsatile (ac) component at each wavelength, which results from the expansion and relaxation of the arterial bed, is divided by the corresponding nonpulsatile (dc) component of the photoplethysmogram, which is composed of the light absorbed by the blood-less tissue and the nonpulsatile portion of the blood compartment. This effective scaling process results in a normalized red/infrared ratio which is dependent on SaO2 but is largely independent of the incident light intensity, skin pigmentation, skin thickness, and tissue vasculature. Pulse oximeter sensors consist of a pair of small and inexpensive red and infrared LEDs and a single, highly sensitive, silicon photodetector. These components are mounted inside a reusable rigid springloaded clip, a flexible probe, or a disposable adhesive wrap (Fig. 50.6). The majority of the commercially available sensors are of the transmittance type in which the pulsatile arterial bed, e.g., ear lobe, fingertip, or toe, is positioned between the LEDs and the photodetector. Other probes are available for reflectance (backscatter) measurement where both the LEDs and photodetectors are mounted side-by-side facing the skin [7, 8]. Noninvasive Cerebral Oximetry Another substance whose optical absorption in the near infrared changes corresponding to its reduced and oxidized state is cytochrome aa3, the terminal member of the respiratory chain. Although the concentration of cytochrome aa3 is considerably lower than that of hemoglobin, advanced instrumentation including time-resolved spectroscopy and differential measurements is being used successfully to obtain noninvasive measurements of hemoglobin saturation and cytochrome aa3 by transilluminating areas of the neonatal brain [9-11].

Blood Gases1 Frequent measurement of blood gases, i.e., oxygen partial pressure (PO2), carbon dioxide partial pressure (PCO2), and pH, is essential to clinical diagnosis and management of respiratory and metabolic problems 1

Additional information on blood gas sensors can be found in Appendix A.


in the operating room and the ICU. Considerable effort has been devoted over the last two decades to developing disposable extracorporeal and in particular intravascular fiber optic sensors that can be used to provide continuous information on the acid-base status of a patient. In the early 1970s, Lübbers and Opitz [12] originated what they called optodes (from the Greek, optical path) for measurements of important physiologic gases in fluids and in gases. The principle upon which these sensors was designed was a closed cell containing a fluorescent indicator in solution, with a membrane permeable to the analyte of interest (either ions or gases) constituting one of the cell walls. The cell was coupled by optical fibers to a system that measured the fluorescence in the cell. The cell solution would equilibrate with the PO2 or PCO2 of the medium placed against it, and the fluorescence of an indicator reagent in the solution would correspond to the partial pressure of the measured gas. Extracorporeal Measurement Following the initial feasibility studies of Lübbers and Opitz, Cardiovascular Devices (CDI, USA) developed a GasStat™ extracorporeal system suitable for continuous online monitoring of blood gases ex vivo during cardiopulmonary bypass operations. The system consists of a disposable plastic sensor connected inline with a blood loop through a fiber optic cable. Permeable membranes separate the flowing blood from the system chemistry. The CO2-sensitive indicator consists of a fine emulsion of a bicarbonate buffer in a two-component silicone. The pH-sensitive indicator is a cellulose material to which hydroxypyrene trisulfonate (HPTS) is bonded covalently. The O2-sensitive chemistry is composed of a solution of oxygenquenching decacyclene in a one-component silicone covered with a thin layer of black PTFE for optical isolation and to render the measurement insensitive to the halothane anesthetic. The extracorporeal device has two channels, one for arterial blood and the other for venous blood, and is capable of recording the temperature of the blood for correcting the measurements to 37°C. Several studies have been conducted comparing the specifications of the GasStat™ with that of intermittent blood samples analyzed on bench-top blood gas analyzers [13-15]. Intravascular Catheters During the past decade, numerous efforts have been made to develop integrated fiber optic sensors for intravascular monitoring of blood gases. A few commercial systems for monitoring blood gases and pH are currently undergoing extensive clinical testing. Recent literature reports of sensor performance show that considerable progress has been made mainly in improving the accuracy and reliability of these intravascular blood gas sensors [16-19]. Most fiber optic intravascular blood gas sensors employ either a single- or a double-fiber configuration. Typically, the matrix containing the indicator is attached to the end of the optical fiber as illustrated in Fig. 50.7. Since the solubility of O2 and CO2 gases, as well as the optical properties of the sensing chemistry itself, are affected by temperature variations, fiber optic intravascular sensors include a thermocouple or thermistor wire running alongside the fiber optic cable to monitor and correct for temperature fluctuations near the sensor tip. A nonlinear response is characteristic of most chemical indicator sensors, so

FIGURE 50.7 Principle diagram of an integrated fiber optic blood gas catheter (from Otto S. Wolfbeis, Fiber Optic Chemical Sensors and Biosensors, vol. 2, CRC Press, Boca Raton, 1990).


they are designed to match the concentration region of the intended application. Also, the response time of the optode is somewhat slower compared to electrochemical sensors. Intravascular fiber optic blood gas sensors are normally placed inside a standard 20-gauge catheter, which is sufficiently small to allow adequate spacing between the sensor and the catheter wall. The resulting lumen is large enough to permit the withdrawal of blood samples, introduction of a continuous heparin flush, and the recording of a blood pressure waveform. In addition, the optical fibers are encased in a protective tubing to contain any fiber fragments in case they break off. pH Sensors In 1976, Peterson and coworkers [20] originated the development of the first fiber optic chemical sensor for physiological pH measurement. The basic idea was to contain a reversible color-changing indicator at the end of a pair of optical fibers. The indicator, phenol red, was covalently bound to a hydrophilic polymer in the form of water-permeable microbeads. This technique stabilized the indicator concentration. The indicator beads were contained in a sealed hydrogen-ion-permeable envelope made out of a hollow cellulose tubing. In effect, this formed a miniature spectrophotometric cell at the end of the fibers and represented an early prototype of a fiber optic chemical sensor. The phenol red dye indicator is a weak organic acid, and the acid form (un-ionized) and base form (ionized) are present in a concentration ratio determined by the ionization constant of the acid and the pH of the medium according to the familiar Henderson-Hasselbalch equation. The two forms of the dye have different optical absorption spectra, so the relative concentration of one of the forms, which varies as a function of pH, can be measured optically and related to variations in pH. In the pH sensor, green (560 nm) and red (longer than 600 nm) light emerging from the end of one fiber passes through the dye and is reflected back into the other fiber by light-scattering particles. The green light is absorbed by the base form of the indicator. The red light is not absorbed by the indicator and is used as an optical reference. The ratio of green to red light is measured and is related to pH by an S-shaped curve with an approximate high-sensitivity linear region where the equilibrium constant (pK) of the indicator matches the pH of the solution. The same principle can also be used with a reversible fluorescent indicator, in which case the concentration of one of the indicator forms is measured by its fluorescence rather than absorbance intensity. Light in the blue or UV wavelength region excites the fluorescent dye to emit longer wavelength light, and the two forms of the dye may have different excitation or emission spectra to allow their distinction. The original instrument design for a pH measurement was very simple and consisted of a tungsten lamp for fiber illumination, a rotating filter wheel to select the green and red light returning from the fiber optic sensor, and signal processing instrumentation to give a pH output based on the green-to-red ratio. This system was capable of measuring pH in the physiologic range between 7.0–7.4 with an accuracy and precision of 0.01 pH units. The sensor was susceptible to ionic strength variation in the order of 0.01 pH unit per 11% change in ionic strength. Further development of the pH probe for practical use was continued by Markle and colleagues [21]. They designed the fiber optic probe in the form of a 25-gauge (0.5 mm o.d.) hypodermic needle, with an ion-permeable side window, using 75-mm-diameter plastic optical fibers. The sensor had a 90% response time of 30 s. With improved instrumentation and computerized signal processing and with a three-point calibration, the range was extended to ±3 pH units, and a precision of 0.001 pH units was achieved. Several reports have appeared suggesting other dye indicator systems that can be used for fiber optic pH sensing [22]. A classic problem with dye indicators is the sensitivity of their equilibrium constant to ionic strength. To circumvent this problem, Wolfbeis and Offenbacher [23] and Opitz and Lübbers [24] demonstrated a system in which a dual sensor arrangement can measure ionic strength and pH and simultaneously can correct the pH measurement for variations in ionic strength. PCO2 Sensors The PCO2 of a sample is typically determined by measuring changes in the pH of a bicarbonate solution that is isolated from the sample by a CO2-permeable membrane but remains in equilibrium with the


CO2. The bicarbonate and CO2, as carbonic acid, form a pH buffer system, and, by the HendersonHasselbalch equation, hydrogen ion concentration is proportional to the pCO2 in the sample. This measurement is done with either a pH electrode or a dye indicator in solution. Vurek [25] demonstrated that the same techniques can also be used with a fiber optic sensor. In his design, one plastic fiber carries light to the transducer, which is made of a silicone rubber tubing about 0.6 mm in diameter and 1.0 mm long, filled with a phenol red solution in a 35-mM bicarbonate. Ambient PCO2 controls the pH of the solution which changes the optical absorption of the phenol red dye. The CO2 permeates through the rubber to equilibrate with the indicator solution. A second optical fiber carries the transmitted signal to a photodetector for analysis. The design by Zhujun and Seitz [26] uses a PCO2 sensor based on a pair of membranes separated from a bifurcated optical fiber by a cavity filled with bicarbonate buffer. The external membrane is made of silicone, and the internal membrane is HPTS immobilized on an ion-exchange membrane. PO2 Sensors The development of an indicator system for fiber optic PO2 sensing is challenging because there are very few known ways to measure PO2 optically. Although a color-changing indicator would have been desirable, the development of a sufficiently stable indicator has been difficult. The only principle applicable to fiber optics appears to be the quenching effect of oxygen on fluorescence. Fluorescence quenching is a general property of aromatic molecules, dyes containing them, and some other substances. In brief, when light is absorbed by a molecule, the absorbed energy is held as an excited electronic state of the molecule. It is then lost by coupling to the mechanical movement of the molecule (heat), reradiated from the molecule in a mean time of about 10 ns (fluorescence), or converted into another excited state with much longer mean lifetime and then reradiated (phosphorescence). Quenching reduces the intensity of fluorescence and is related to the concentration of the quenching molecules, such as O2. A fiber optic sensor for measuring PO2 using the principle of fluorescence quenching was developed by Peterson and colleagues [27]. The dye is excited at around 470 nm (blue) and fluoresces at about 515 nm (green) with an intensity that depends on the PO2. The optical information is derived from the ratio of green fluorescence to the blue excitation light, which serves as an internal reference signal. The system was chosen for visible light excitation, because plastic optical fibers block light transmission at wavelengths shorter than 450 nm, and glass fibers were not considered acceptable for biomedical use. The sensor was similar in design to the pH probe continuing the basic idea of an indicator packing in a permeable container at the end of a pair of optical fibers. A dye perylene dibutyrate, absorbed on a macroreticular polystyrene adsorbent, is contained in a oxygen-permeable porous polystyrene envelope. The ratio of green to blue intensity was processed according to the Stren-Volmer equation

I0 = 1 + K PO2 I where I and I0 are the fluorescence emission intensities in the presence and absence of a quencher, respectively, and I is the Stern-Volmer quenching coefficient. This provides a nearly linear readout of PO2 over the range of 0–150 mmHg (0–20 kPa), with a precision of 1 mm Hg (0.13 kPa). The original sensor was 0.5 mm in diameter, but it can be made much smaller. Although its response time in a gas mixture is a fraction of a second, it is slower in an aqueous system, about 1.5 min for 90% response. Wolfbeis and coworkers [28] designed a system for measuring the widely used halothane anesthetic which interferes with the measurement of oxygen. This dual-sensor combination had two semipermeable membranes (one of which blocked halothane) so that the probe could measure both oxygen and halothane simultaneously. The response time of their sensor, 15–20 s for halothane and 10–15 s for oxygen, is considered short enough to allow gas analysis in the breathing circuit. Potential applications of this device include the continuous monitoring of halothane in breathing circuits and in the blood.


FIGURE 50.8 [29].

Schematic diagram of a competitive binding fluorescence affinity sensor for glucose measurement

Glucose Sensors Another important principle that can be used in fiber optic sensors for measurements of high sensitivity and specificity is the concept of competitive binding. This was first described by Schultz, Mansouri, and Goldstein [29] to construct a glucose sensor. In their unique sensor, the analyte (glucose) competes for binding sites on a substrate (the lectin concanavalin A) with a fluorescent indicator-tagged polymer [fluorescein isothiocyanate (FITC)-dextran]. The sensor, which is illustrated in Fig. 50.8, is arranged so that the substrate is fixed in a position out of the optical path of the fiber end. The substrate is bound to the inner wall of a glucose-permeable hollow fiber tubing (300 µ O.D. × 200 µ I.D.) and fastened to the end of an optical fiber. The hollow fiber acts as the container and is impermeable to the large molecules of the fluorescent indicator. The light beam that extends from the fiber “sees” only the unbound indictor in solution inside the hollow fiber but not the indicator bound on the container wall. Excitation light passes through the fiber and into the solution, fluorescing the unbound indicator, and the fluorescent light passes back along the same fiber to a measuring system. The fluorescent indicator and the glucose are in competitive binding equilibrium with the substrate. The interior glucose concentration equilibrates with its concentration exterior to the probe. If the glucose concentration increases, the indicator is driven off the substrate to increase the concentration of the indicator. Thus, fluorescence intensity as seen by the optical fiber follows the glucose concentration. The response time of the sensor was found to be about 5 min. In vivo studies demonstrated fairly close correspondence between the sensor output and actual blood glucose levels. A time lag of about 5 min was found and is believed to be due to the diffusion of glucose across the hollow fiber membrane and the diffusion of FTIC-dextran within the tubing. In principle, the concept of competitive binding can be applied to any analysis for which a specific reaction can be devised. However, long-term stability of these sensors remains the major limiting factor that needs to be solved.

Immunosensors Immunologic techniques offer outstanding selectivity and sensitivity through the process of antibodyantigen interaction. This is the primary recognition mechanism by which the immune system detects and fights foreign matter and has therefore allowed the measurement of many important compounds at trace levels in complex biologic samples. In principle, it is possible to design competitive binding optical sensors utilizing immobilized antibodies as selective reagents and detecting the displacement of a labeled antigen by the analyte. Therefore,


FIGURE 50.9

Basic principle of a fiber optic antigen-antibody sensor [33].

antibody-based immunologic optical systems have been the subject of considerable research in the past few years [30-34]. In practice, however, the strong binding of antigens to antibodies and vice versa causes difficulties in constructing reversible sensors with fast dynamic responses. Several immunologic sensors based on fiber optic waveguides have been demonstrated for monitoring antibody-antigen reactions. Typically, several centimeters of cladding are removed along the fiber’s distal end, and the recognition antibodies are immobilized on the exposed core surface. These antibodies bind fluorophore-antigen complexes within the evanescent wave as illustrated in Fig. 50.9. The fluorescent signal excited within the evanescent wave is then transmitted through the cladded fiber to a fluorimeter for processing. Experimental studies have indicated that immunologic optical sensors can generally detect micromolar and even picomolar concentrations. However, the major obstacle that must be overcome to achieve high sensitivity in immunologic optical sensors is the nonspecific binding of immobilized antibodies.

References 1. Polanyi ML, Heir RM. 1962. In vivo oximeter with fast dynamic response. Rev Sci Instrum 33:1050. 2. Bornzin GA, Mendelson Y, Moran BL, et al. 1987. Measuring oxygen saturation and hematocrit using a fiberoptic catheter. Proc 9th Ann Conf Eng Med Bio Soc 807–809. 3. Mendelson Y, Galvin JJ, Wang Y. 1990. In vitro evaluation of a dual oxygen saturation/hematocrit intravascular fiberoptic catheter. Biomed Instrum Tech 24:199. 4. Mendelson Y. 1992. Pulse oximetry: Theory and application for noninvasive monitoring. Clin Chem 28(9):1601. 5. Aoyagi T, Kishi M, Yamaguchi K, et al. 1974. Improvement of the earpiece oximeter. Jpn Soc Med Electron Biomed Eng 90–91. 6. Yoshiya I, Shimada Y, Tanaka K. 1980. Spectrophotometric monitoring of arterial oxygen saturation in the fingertip. Med Biol Eng Comput 18:27. 7. Mendelson Y, Solomita MV. 1992. The feasibility of spectrophotometric measurements of arterial oxygen saturation from the scalp utilizing noninvasive skin reflectance pulse oximetry. Biomed Instrum Technol 26:215. 8. Mendelson Y, McGinn MJ. 1991. Skin reflectance pulse oximetry: In vivo measurements from the forearm and calf. J Clin Monit 7:7. 9. Chance B, Leigh H, Miyake H, et al. 1988. Comparison of time resolved and un-resolved measurements of deoxyhemoglobin in brain. Proc Nat Acad Sci 85:4971. 10. Jobsis FF, Keizer JH, LaManna JC, et al. 1977. Reflection spectrophotometry of cytochrome aa3 in vivo. Appl Physiol: Respirat Environ Excerc Physiol 43(5):858.


11. Kurth CD, Steven IM, Benaron D, et al. 1993. Near-infrared monitoring of the cerebral circulation. J Clin Monit 9:163. 12. Lübbers DW, Opitz N. 1975. The pCO2/pO2-optode: A new probe for measurement of pCO2 or pO2 in fluids and gases. Z Naturforsch C: Biosci 30C:532. 13. Clark CL, O’Brien J, McCulloch J, et al. 1986. Early clinical experience with GasStat. J Extra Corporeal Technol 18:185. 14. Hill AG, Groom RC, Vinansky RP, et al. 1985. On-line or off-line blood gas analysis: Cost vs. time vs. accuracy. Proc Am Acad Cardiovasc Perfusion 6:148. 15. Siggaard-Andersen O, Gothgen IH, Wimberley, et al. 1988. Evaluation of the GasStat fluorescence sensors for continuous measurement of pH, pCO2 and pO3 during CPB and hypothermia. Scand J Clin Lab Invest 48 (Suppl. 189):77. 16. Zimmerman JL, Dellinger RP. 1993. Initial evaluation of a new intra-arterial blood gas system in humans. Crit Care Med 21(4):495. 17. Gottlieb A. 1992. The optical measurement of blood gases—approaches, problems and trends: Fiber optic medical and fluorescent sensors and applications. Proc SPIE 1648:4. 18. Barker SL, Hyatt J. 1991. Continuous measurement of intraarterial pHa, PaCO2, and PaO2 in the operation room. Anesth Analg 73:43. 19. Larson CP, Divers GA, Riccitelli SD. 1991. Continuous monitoring of PaO2 and PaCO2 in surgical patients. Abstr Crit Care Med 19:525. 20. Peterson JI, Goldstein SR, Fitzgerald RV. 1980. Fiber optic pH probe for physiological use. Anal Chem 52:864. 21. Markle DR, McGuire DA, Goldstein SR, et al: 1981. A pH measurement system for use in tissue and blood, employing miniature fiber optic probes, In DC Viano (ed), Advances in Bioengineering, p 123, New York, American Society of Mechanical Engineers. 22. Wolfbeis OS, Furlinger E, Kroneis H, et al. 1983. Fluorimeter analysis: 1. A study on fluorescent indicators for measuring near neutral (physiological) pH values. Fresenius’ Z Anal Chem 314:119. 23. Wolfbeis OS, Offenbacher H. 1986. Fluorescence sensor for monitoring ionic strength and physiological pH values. Sens Actuators 9:85. 24. Opitz N, Lübbers DW. 1983. New fluorescence photomatrical techniques for simultaneous and continuous measurements of ionic strength and hydrogen ion activities. Sens Actuators 4:473. 25. Vurek GG, Feustel PJ, Severinghaus JW. 1983. A fiber optic pCO2 sensor. Ann Biomed Eng 11:499. 26. Zhujun Z, Seitz WR. 1984. A carbon dioxide sensor based on fluorescence. Anal Chim Acta 160:305. 27. Peterson JI, Fitzgerald RV, Buckhold DK. 1984. Fiber-optic probe for in vivo measurements of oxygen partial pressure. Anal Chem 56:62. 28. Wolfbeis OS, Posch HE, Kroneis HW. 1985. Fiber optical fluorosensor for determination of halothane and/or oxygen. Anal Chem 57:2556. 29. Schultz JS, Mansouri S, Goldstein IJ. 1982. Affinity sensor: A new technique for developing implantable sensors for glucose and other metabolites. Diabetes Care 5:245. 30. Andrade JD, Vanwagenen RA, Gregonis DE, et al. 1985. Remote fiber optic biosensors based on evanescent-excited fluoro-immunoassay: Concept and progress. IEEE Trans Electron Devices ED32:1175. 31. Sutherland RM, Daehne C, Place JF, et al. 1984. Optical detection of antibody-antigen reactions at a glass-liquid interface. Clin Chem 30:1533. 32. Hirschfeld TE, Block MJ. 1984. Fluorescent immunoassay employing optical fiber in a capillary tube. US Patent No. 4,447,546. 33. Anderson GP, Golden JP, Ligler FS. 1993. An evanescent wave biosensor: Part I. Fluorescent signal acquisition from step-etched fiber optic probes. IEEE Trans Biomed Eng 41(6):578. 34. Golden JP, Anderson GP, Rabbany SY, et al. 1994. An evanescent wave biosensor: Part II. Fluorescent signal acquisition from tapered fiber optic probes. IEEE Trans Biomed Eng 41(6):585.


Richard P. Buck. “Bioanalytic Sensors.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

51 Bioanalytic Sensors 51.1

Classification of Biochemical Reactions in the Context of Sensor Design and Development Introduction and Definitions • Classification of Recognition Reactions and Receptor Processes

51.2

Classification of Transduction Processes—Detection Methods Calorimetric, Thermometric, and Pyroelectric Transducers • Optical, Optoelectronic Transducers • Piezoelectric Transducers • Electrochemical Transducers

Richard P. Buck

51.3 51.4

University of North Carolina

Tables of Sensors from the Literature Application of Microelectronics in Sensor Fabrication

51.1 Classification of Biochemical Reactions in the Context of Sensor Design and Development Introduction and Definitions Since sensors generate a measurable material property, they belong in some grouping of transducer devices. Sensors specifically contain a recognition process that is characteristic of a material sample at the molecular-chemical level, and a sensor incorporates a transduction process (step) to create a useful signal. Biomedical sensors include a whole range of devices that may be chemical sensors, physical sensors, or some kind of mixed sensor. Chemical sensors use chemical processes in the recognition and transduction steps. Biosensors are also chemical sensors, but they use particular classes of biological recognition/transduction processes. A pure physical sensor generates and transduces a parameter that does not depend on the chemistry per se, but is a result of the sensor responding as an aggregate of point masses or charges. All these when used in a biologic system (biomatrix) may be considered bioanalytic sensors without regard to the chemical, biochemical, or physical distinctions. They provide an “analytic signal of the biologic system” for some further use. The chemical recognition process focuses on some molecular-level chemical entity, usually a kind of chemical structure. In classical analysis this structure may be a simple functional group: SiO– in a glass electrode surface, a chromophore in an indicator dye, or a metallic surface structure, such as silver metal that recognizes Ag+ in solution. In recent times, the biologic recognition processes have been better understood, and the general concept of recognition by receptor or chemoreceptor has come into fashion. Although these are often large molecules bound to cell membranes, they contain specific structures that permit a wide variety of different molecular recognition steps including recognition of large and small species and of charged and uncharged species. Thus, chemoreceptor appears in the sensor literature as a generic term for the principal entity doing the recognition. For a history and examples, see references [1–6].


FIGURE 51.1

Generic bioanalytic sensor.

Biorecognition in biosensors has especially stressed “receptors” and their categories. Historically, application of receptors has not necessarily meant measurement directly of the receptor. Usually there are coupled chemical reactions, and the transduction has used measurement of the subsidiary products: change of pH, change of dissolved O2,1 generation of H2O2, changes of conductivity, changes of optical adsorption, and changes of temperature. Principal receptors are enzymes because of their extraordinary selectivity. Other receptors can be the more subtle species of biochemistry: antibodies, organelles, microbes, and tissue slices, not to mention the trace level “receptors” that guide ants, such as pheromones, and other unusual species. A sketch of a generic bioanalytic sensor is shown in Fig. 51.1

Classification of Recognition Reactions and Receptor Processes The concept of recognition in chemistry is universal. It almost goes without saying that all chemical reactions involved recognition and selection on the basis of size, shape, and charge. For the purpose of constructing sensors, general recognition based on these factors is not usually enough. Frequently in inorganic chemistry a given ion will react indiscriminantly with similar ions of the same size and charge. Changes in charge from unity to two, for example, do change the driving forces of some ionic reactions. By control of dielectric constant of phases, heterogeneous reactions can often be “tailored” to select divalent ions over monovalent ions and to select small versus large ions or vice versa. Shape, however, has more special possibilities, and natural synthetic methods permit product control. Nature manages to use shape together with charge to build organic molecules, called enzymes, that have acquired remarkable selectivity. It is in the realm of biochemistry that these natural constructions are investigated and catalogued. Biochemistry books list large numbers of enzymes and other selective materials that direct chemical reactions. Many of these have been tried as the basis of selective sensors for bioanalytic and biomedical purposes. The list in Table 51.1 shows how some of the materials can be grouped into lists according to function and to analytic substrate, both organic and inorganic. The principles seem general, so there is no reason to discriminate against the inorganic substrates in favor or the organic substrates. All can be used in biomedical analysis.

51.2 Classification of Transduction Processes—Detection Methods Some years ago, the engineering community addressed the topic of sensor classification—Richard M. White in IEEE Trans. Ultra., Ferro., Freq. Control (UFFC), UFFC-34 (1987) 124, and Wen H. Ko in 1

Additional information on these types of sensors can be found in Chapter 50 and Appendix A.


TABLE 51.1

Recognition Reactions and Receptor Processes

1. Insoluble salt-based sensors a. S+ + R– 1(insoluble salt) Ion exchange with crystalline SR (homogeneous or heterogeneous crytals) chemical signal S+n receptor R–n inorganic cations inorganic anions examples: Ag+, Hg22+, Pb2+, Cd2+, Cu2+ S=, Se2=, SCN–, I–, Br–, Cl– b. S–n + R+n 1SR (insoluble salt) Ion exchange with crystalline SR (homogeneous or heterogeneous crystals) chemical signal S–n receptor R+n inorganic anions inorganic cations examples: F–, S=, Se2=, SCN–, I–, Br–, Cl– LaF2+, Ag+, Hg22+, Pb2+, Cd2+, Cu2+ 2. Solid ion exchanges a. S+n + R–n (sites) 1S+nR–n = SR (in ion exchanger phase) Ion exchange with synthetic ion exchangers containing negative fixed sites (homogeneous or heterogeneous, inorganic or organic materials) chemical signal S+n receptor R–n inorganic and organic ions examples: H+, Na+, K+ H+, Na+, K+, other M+n

inorganic and organic ion sites silicate glass Si-0– synthetic sulfonated, phosphorylated, EDTA-substituted polystyrenes b. S–n + R+n (sites) 1S–nR+n = SR (in ion exchanger phase) Ion exchange with synthetic ion exchangers containing positive fixed sites (homogeneous or heterogeneous, inorganic or organic materials) receptor R+n chemical signal S–n organic and inorganic ions organic and inorganic ion sites examples: hydrophobic anions quaternized polystyrene 3. Liquid ion exchanger sensors with electrostatic selection a. S–n + R–n (sites) 1S+nR–n = SR (in ion exchanger phase) Plasticized, passive membranes containing mobile trapped negative fixed sites (homogeneous or heterogeneous, inorganic or organic materials) chemical signal S+n receptor R–n inorganic and organic ions examples: Ca2+ M+n

inorganic and organic ion sites diester of phosphoric acid or monoester of a phosphonic acid dinonylnaphthalene sulfonate and other organic, hydrophobic anions tetraphenylborate anion or substituted derivatives

R1R2R3R4N+ and bis-Quaternary Cations cationic drugs tetrasubstituted arsonium+ b. S–n + R+n (sites) 1S–nR+n = SR (in ion exchanger phase) Plasticized, passive membranes containing mobile, trapped negative fixed sites (homogeneous or heterogeneous, inorganic or organic materials) chemical signal S–n receptor R+n inorganic and organic ions examples: anions, simple Cl–, Br–, ClO4–

inorganic and organic sites quaternary ammonium cations: e.g. tridodecylmethylammonium anions, complex, drugs quaternary ammonium cations: e.g. tridodecylmethylammonium 4. Liquid ion exchanger sensors with neutral (or charged) carrier selection a. S+n + X and R–n (sites) 1S+nX R–n = SXR (in ion exchanger phase) Plasticized, passive membranes containing mobile, trapped negative fixed sites (homogeneous or heterogeneous, inorganic or organic materials) chemical signal S+n receptor R–n inorganic and organic ions examples: Ca2+


inorganic and organic ion sites X = synthetic ionophore complexing agent selective to Ca2+ R–n usually a substituted tetra phenylborate salt

TABLE 51.1 (continued) +

+

Recognition Reactions and Receptor Processes

+

Na , K , H X = selective ionophore complexing agent b. S–n + X and R+n (sites) 1S–nX R+n = SXR (in ion exchanger phase) Plasticized, passive membranes containing mobile, trapped negative fixed sites (homogeneous or heterogeneous, inorganic or organic materials) receptor R+n chemical signal S–n inorganic and organic ions examples: HPO42=

inorganic and organic ion sites R+n = quaternary ammonium salt X = synthetic ionophore complexing agent; aryl organotin compound or suggested cyclic polyamido-polyamines HCO3– X = synthetic ionophore: trifluoro acetophenone Cl– X = aliphatic organotin compound 5. Bioaffinity sensors based on change of local electron densities S + R 1SR chemical signal S receptor R protein saccharide glycoprotein substrate inhibitor

dyes lectin enzyme

Transferases Hydrolases (peptidases, esterases, etc.) Lyases Isomerases Ligases prosthetic group apoenzyme antigen antibody hormone “receptor” substrate analogue transport system 6. Metabolism sensors based on substrate consumption and product formation S + R 1SR → P + R chemical signal S receptor R substrate examples: lactate (SH2)

enzyme hydrogenases catalyze hydrogen transfer from S to acceptor A (not molecular oxygen!) reversibly pyruvate + NADH + H+ using lactate dehydrogenase

SH2 + A 1S + AH2 lactate + NAD+ glucose (SH ) 1 2 oxidases catalyze hydrogen transfer to molecular oxygen SH2 + 2--- O2 1S + H2O or SH2 + O2 1S + H2O2 using glucose oxidase glucose + O2 1gluconolactone + H2O2 reducing agents (S) peroxidases catalyze oxidation of a substrate by H2O2 using 2S + 2H+ + H2O2 12S+ + 2H2O horseradish peroxidase Fe2+ + H2O2 + 2H+ 1Fe3+ + 2H2O reducing agents oxygenates catalyze substrate oxidations by molecular O2 L-lactate + O2 1acetate + CO2 + H2O cofactor organelle inhibitor microbe activator tissue slice enzyme activity 7. Coupled and hybrid systems using sequences, competition, anti-interference and amplification concepts and reactions. 8. Biomimetic sensors chemical signal S receptor R sound stress light Source: Adapted from [2, 6].


carrier-enzyme

IEEE/EMBS Symposium Abstract T.1.1 84CH2068-5 (1984). It is interesting because the physical and chemical properties are given equal weight. There are many ideas given here that remain without embodiment. This list is reproduced as Table 51.2. Of particular interest in this section are “detection means used in sensors” and “sensor conversion phenomena.” At present the principle transduction schemes use electrochemical, optical, and thermal detection effects and principles.

Calorimetric, Thermometric, and Pyroelectric Transducers

TABLE 51.2 Detection Means and Conversion Phenomena Used in Sensors Detection means Biologic Chemical Electric, magnetic, or electromagnetic wave Heat, temperature Mechanical displacement of wave Radioactivity, radiation Other Conversion phenomena Biologic Biochemical transformation Physical transformation Effect on test organism Spectroscopy Other Chemical Chemical transformation Physical transformation Electrochemical process Spectroscopy Other Physical Thermoelectric Photoelectric Photomagnetic Magnetoelectric Elastomagnetic Thermoelastic Elastoelectric Thermomagnetic Thermooptic Photoelastic Other

Especially useful for enzymatic reactions, the generation of heat (enthalpy change) can be used easily and generally. The enzyme provides the selectivity and the reaction enthalpy cannot be confused with other reactions from species in a typical biologic mixture. The ideal aim is to measure total evolved heat, i.e., to perform a calorimetric measurement. In real systems there is always heat loss, i.e., heat is conducted away by the sample and sample container so that the process cannot be adiabatic as required for a total heat evolution measurement. As a result, temperature difference before and after evolution is measured most often. It has to be assumed that the heat capacity of the specimen and container is constant over the small temperature range usually measured. The simplest transducer is a thermometer coated with the enzyme that permits the selected reaction to proceed. Thermistors are used rather than thermometers or thermocouples. The change of resistance of certain oxides is much greater than the change of length of a mercury column or the microvolt changes of thermocouple junctions. Pyroelectric heat flow transducers are relatively new. Heat flows from a heated region to a lower temperature region, controlled to occur in one dimension. The lower temperature side can be coated with an enzyme. When the substrate is converted, the lower temperature side is warmed. The pyroelectric material is from a category of materials that develops a spontaneous voltage difference in a thermal gradient. If the gradient is disturbed by evolution or adsorption of heat, the voltage temporarily changes. In biomedical sensing, some of the solid-state devices based on thermal sensing cannot be used effectively. The reason is that the sensor itself has to be heated or is heated quite hot by catalytic surface reactions. Thus pellistors (oxides with catalytic surfaces and embedded platinum wire thermometer), chemiresistors, and “Figaro” sensor “smoke” detectors have not found many biologic applications.

Optical, Optoelectronic Transducers Most optical detection systems for sensors are small, i.e., they occupy a small region of space because the sample size and volume are themselves small. This means that common absorption spectrophotometers and photofluorometers are not used with their conventional sample-containing cells, or with their conventional beam-handling systems. Instead light-conducting optical fibers are used to connect the sample with the more remote monochromator and optical readout system. The techniques still remain absorption spectrophotometry, fluorimetry including fluorescence quenching, and reflectometry. © 2000 by CRC Press LLC

The most widely published optical sensors use a miniature reagent contained or immobilized at the tip of an optical fiber. In most systems a permselective membrane coating allows the detected species to penetrate the dye region. The corresponding absorption change, usually at a sensitive externally preset wavelength, is changed and correlated with the sample concentration. Similarly, fluorescence can be stimulated by the higher-frequency external light source and the lower-frequency emission detected. Some configurations are illustrated in references [1, 2]. Fluorimetric detection of coenzyme A, NAD+/NADH, is involved in many so-called pyridine-linked enzyme systems. The fluorescence of NADH contained or immobilized can be a convenient way to follow these reactions. Optodes, miniature encapsulated dyes, can be placed in vivo. Their fluorescence can be enhanced or quenched and used to detect acidity, oxygen, and other species. A subtle form of optical transduction uses the “peeled” optical fiber as a multiple reflectance cell. The normal fiber core glass has a refractive index greater than that of the exterior coating; there is a range of angles of entry to the fiber so that all the light beam remains inside the core. If the coating is removed and materials of lower index of refraction are coated on the exterior surface, there can be absorption by multiple reflections, since the evanescent wave can penetrate the coating. Chemical reagent can be added externally to create selective layers on the optical fiber. Ellipsometry is a reflectance technique that depends on the optical constants and thickness of surface layer. For colorless layers, a polarized light beam will change its plane of polarization upon reflection by the surface film. The thickness can sometimes be determined when optical constants are known or approximated by constants of the bulk material. Antibody-antigen surface reaction can be detected this way.

Piezoelectric Transducers Cut quartz crystals have characteristic modes of vibration that can be induced by painting electrodes on the opposite surfaces and applying a megaHertz ac voltage. The frequency is searched until the crystal goes into a resonance. The resonant frequency is very stable. It is a property of the material and maintains a value to a few parts per hundred million. When the surface is coated with a stiff mass, the frequency is altered. The shift in frequency is directly related to the surface mass for thin, stiff layers. The reaction of a substrate with this layer changes the constants of the film and further shifts the resonant frequency. These devices can be used in air, in vacuum, or in electrolyte solutions.

Electrochemical Transducers Electrochemical transducers are commonly used in the sensor field. The main forms of electrochemistry used are potentiometry [zero-current cell voltage (potential difference measurements)], amperometry (current measurement at constant applied voltage at the working electrode), and ac conductivity of a cell. Potentiometric Transduction The classical generation of an activity-sensitive voltage is spontaneous in a solution containing both nonredox ions and redox ions. Classical electrodes of types 1, 2, and 3 respond by ion exchange directly or indirectly to ions of the same material as the electrode. Inert metal electrodes (sometimes called type 0)—Pt, Ir, Rh, and occasionally carbon C—respond by electrons exchange from redox pairs in solution. Potential differences are interfacial and reflect ratios of activities of oxidized to reduced forms. Amperometric Transduction For dissolved species that can exchange electrons with an inert electrode, it is possible to force the transfer in one direction by applying a voltage very oxidizing (anodic) or reducing (cathodic). When the voltage is fixed, the species will be, by definition, out of equilibrium with the electrode at its present applied voltage. Locally, the species (regardless of charge) will oxidize or reduce by moving from bulk solution to the electrode surface where they react. Ions do not move like electrons. Rather they diffuse from high to low concentration and do not usually move by drift or migration. The reason is that the electrolytes © 2000 by CRC Press LLC

in solutions are at high concentrations, and the electric field is virtually eliminated from the bulk. The field drops through the first 1000 Angstroms at the electrode surface. The concentration of the moving species is from high concentration in bulk to zero at the electrode surface where it reacts. This process is called concentration polarization. The current flowing is limited by mass transport and so is proportional to the bulk concentration. Conductometric Transducers Ac conductivity (impedance) can be purely resistive when the frequency is picked to be about 1000 to 10,000 Hz. In this range the transport of ions is sufficiently slow that they never lose their uniform concentration. They simply quiver in space and carry current forward and backward each half cycle. In the lower and higher frequencies, the cell capacitance can become involved, but this effect is to be avoided.

51.3 Tables of Sensors from the Literature The longest and most consistently complete references to the chemical sensor field is the review issue of Analytical Chemistry Journal. In the 1970s and 1980s these appeared in the April issue, but more recently they appear in the June issue. The editors are Jiri Janata and various colleagues [7-10]. Note all possible or imaginable sensors have been made according to the list in Table 51.2. A more realistic table can be constructed from the existing literature that describes actual devices. This list is Table 51.3. Book references are listed in Table 51.4 in reverse time order to about 1986. This list covers most of the major source books and many of the symposium proceedings volumes. The reviews [7-10] are a principal source of references to the published research literature.

TABLE 51.3

Chemical Sensors and Properties Documented in the Literature

1. General topics including items II-V; selectivity, fabrication, data processing II. Thermal sensors III. Mass sensors Gas sensors Liquid sensors IV. Electrochemical sensors Potentiometric sensors Reference electrodes Biomedical electrodes Applications to cations, anions Coated wire/hybrids ISFETs and related Biosensors Gas sensors Amperometric sensors Modified electrodes Gas sensors Biosensors Direct electron transfer Mediated electron transfer Biomedical Conductimetric sensors Semiconducting oxide sensors Zinc oxide-based Chemiresistors Dielectrometers V. Optical sensors Liquid sensors Biosensors Gas sensors


TABLE 51.4

Books and Long Reviews Keyed to Items in Table 51.3

(Reviewed since 1988 in reverse time sequence) 1. Yamauchi S (ed). 1992. Chemical Sensor Technology, vol 4, Tokyo, Kodansha Ltd. Flores JR, Lorenzo E. 1992. Amperometric Biosensors, In MR Smyth , JG Vos (eds), Comprehensive Analytical Chemistry Amsterdam, Elsevier Vaihinger S, Goepel W. 1991. Multicomponent analysis in chemical sensing. In W Goepel, J Hesse, J Zemel (eds), Sensors vol 2 Part 1, pp 191–237, Weinheim, Germany, VCH Publishers Wise DL (ed). 1991. Bioinstrumentation and Biosensors, New York, Marcel Dekker Scheller F, Schubert F. 1989. Biosensors, Basel, Switzerland, Birkhauser Verlag, see also [2]. Madou M, Morrison SR. 1989. Chemical Sensing with Solid State Devices, New York, Academic Press. Janata J. 1989. Principles of Chemical Sensors, New York, Plenum Press. Edmonds TE (ed). 1988. Chemical Sensors, Glasgow, Blackie. Yoda K. 1988. Immobilized enzyme cells. Methods Enzymology, 137:61. Turner APF, Karube I, Wilson GS (eds). 1987. Biosensors: Fundamentals and Applications, Oxford, Oxford University Press. Seiyama T (ed). 1986. Chemical Sensor Technology, Tokyo, Kodansha Ltd. II. Thermal Sensor There are extensive research and application papers and these are mentioned in books listed under I. However, the upto-date lists of papers are given in references 7–10. III. Mass Sensors There are extensive research and application papers and these are mentioned in books listed under I. However, the upto-date lists of papers are given in references 7–10. Fundamentals of this rapidly expanding field are recently reviewed: Buttry DA, Ward MD. 1992. Measurement of Interfacial processes at Electrode Surfaces with the Electrochemical Quartz Crystal Microbalance, Chemical Reviews 92:1355. Grate JW, Martin SJ, White RM. 1993. Acoustic Wave Microsensors, Part 1, Analyt Chem 65:940A; part 2, Analyt Chem 65:987A. Ricco AT. 1994. SAW Chemical Sensors, The Electrochemical Society Interface Winter: 38–44. IVA. Electrochemical Sensors—Liquid Samples Scheller F, Schmid RD (eds). 1992. Biosensors: Fundamentals, Technologies and Applications, GBF Monograph Series, New York, VCH Publishers. Erbach R, Vogel A, Hoffmann B. 1992. Ion-sensitive field-effect structures with Langmuir-Blodgett membranes. In F Scheller, RD Schmid (eds). Biosensors: Fundamentals, Technologies, and Applications, GBF Monograph 17, pp 353–357, New York, VCH Publishers. Ho May YK, Rechnitz GA. 1992. An Introduction to Biosensors, In RM Nakamura, Y Kasahara, GA Rechnitz (eds), Immunochemical Assays and Biosensors Technology, pp 275–291, Washington, DC, American Society Microbiology. Mattiasson B, Haakanson H. Immunochemically-based assays for process control, 1992. Advances in Biochemical Engineering and Biotechnology 46:81. Maas AH, Sprokholt R. 1990. Proposed IFCC Recommendations for electrolyte measurements with ISEs in clinical chemistry, In A Ivaska, A Lewenstam, R Sara (eds), Contemporary Electroanalytical Chemistry, Proceedings of the ElectroFinnAnalysis International Conference on Electroanalytical Chemistry, pp 311–315, New York, Plenum. Vanrolleghem P, Dries D, Verstreate W. RODTOX: Biosensor for rapid determination of the biochemical oxygen demand, 1990. In C Christiansen, L Munck, J Villadsen (eds), Proceedings of the 5th European Congress Biotechnology, vol 1, pp 161–164, Copenhagen, Denmark, Munksgaard. Cronenberg C, Van den Heuvel H, Van den Hauw M, Van Groen B. Development of glucose microelectrodes for measurements in biofilms, 1990. In C Christiansen, L Munck, J Villadsen J (eds), Proceedings of the 5th European Congress Biotechnology, vol 1, pp 548–551, Copenhagen, Denmark, Munksgaard. Wise DL (ed). 1989. Bioinstrumentation Research, Development and Applications, Boston, MA, ButterworthHeinemann. Pungor E (ed). 1989. Ion-Selective Electrodes—Proceedings of the 5th Symposium (Matrafured, Hungary 1988), Oxford, Pergamon. Wang J (ed). 1988. Electrochemical Techniques in Clinical Chemistry and Laboratory Medicine, New York, VCH Publishers. Evans A. 1987. Potentiometry and Ion-selective Electrodes, New York, Wiley. Ngo TT(ed). 1987. Electrochemical Sensors in Immunological Analysis, New York, Plenum. IVB. Electrochemical Sensors—Gas Samples Sberveglieri G (ed). 1992. Gas Sensors, Dordrecht, The Netherlands, Kluwer. Moseley PT, Norris JOW, Williams DE. 1991. Technology and Mechanisms of Gas Sensors, Bristol, U.K., Hilger. Moseley PT, Tofield BD (eds). 1989. Solid State Gas Sensors, Philadelphia, Taylor and Francis, Publishers.


TABLE 51.4 (continued)

Books and Long Reviews Keyed to Items in Table 51.3

(Reviewed since 1988 in reverse time sequence) V. Optical Sensors Coulet PR, Blum LJ. Luminescence in Biosensor Design, 1991. In DL Wise, LB Wingard, Jr (eds). Biosensors with Fiberoptics, pp 293–324, Clifton, N.J., Humana. Wolfbeis OS. 1991. Spectroscopic Techniques, In OS Wolfbeis (ed). Fiber Optic Chemical Sensors and Biosensors, vol 1, pp 25–60. Boca Raton, Fla, CRC Press. Wolfbeis OS. 1987. Fibre-optic sensors for chemical parameters of interest in biotechnology, In RD Schmidt (ed). GBF (Gesellschaft fur Biotechnologische Forschung) Monogr. Series, vol 10, pp 197–206, New York, VCH Publishers.

51.4 Applications of Microelectronics in Sensor Fabrication The reviews of sensors since 1988 cover fabrication papers and microfabrication methods and examples [7-10]. A recent review by two of the few chemical sensor scientists (chemical engineers) who also operate a microfabrication laboratory is C. C. Liu, Z.-R. Zhang. 1992. Research and development of chemical sensors using microfabrication techniques. Selective Electrode 14:147.

References 1. Janata J. 1989. Principles of Chemical Sensors, New York, Plenum. 2. Scheller F, Schubert F. 1989. Biosensors, #18 in Advances in Research Technologies (Beitrage zur Forschungstechnologie), Berlin, Akademie-Verlag, Amsterdam, Elsevier (English translation) 3. Turner APF, Karube I, Wilson GS. 1987. Biosensors: Fundamentals and Applications, Oxford, Oxford University Press. 4. Hall EAH. 1990. Biosensors, Milton Keynes, England, Open University Press. 5. Eddoes MJ. 1990. Theoretical methods for Analyzing Biosensor Performance. In AEG Cass (ed), Biosensor—A Practical Approach, Oxford, IRL Press at Oxford University Ch. 9 pp 211–262. 6. Cosofret VV, Buck RP. 1992. Pharmaceutical Applications of Membrane Sensors, Boca Raton, Fla, CRC Press. 7. Janata J, Bezegh A. 1988. Chemical sensors, Analyt Chem 60:62R. 8. Janata J. 1990. Chemical sensors, Analyt Chem 62:33R. 9. Janata J. 1992. Chemical sensors, Analyt Chem 66:196R. 10. Janata J, Josowicz M, DeVaney M. 1994. Chemical Sensors, Analyt Chem 66:207R.


Geddes, L. A. “The Electrocardiograph.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

Historical Perspectives 2

The Electrocardiograph

Leslie A. Geddes Purdue University

The First Electrocardiogram Capillary Electrometer Record Rheotome Record Mammalian Electrocardiograms Corrected Capillary Electrometer Records Clinical Electrocardiography The String Galvanometer • Vacuum Tube Electrocardiograph • HotStylus Record

Recording and display of bioelectric events occupied a long time and required the development and adaptation of a variety of primitive instruments, not all of which were electronic. The first bioelectric recorder was the rheoscopic frog, consisting of a sciatic nerve and its innervated gastrocnemius muscle. So sensitive was the nerve that it could be stimulated by the beating heart or a contracting muscle; both events contracted the gastrocnemius muscle. However, such a response provided no information on the time course of these bioelectric events. Sensitive and rapidly responding indicators were essential for this purpose. As Etienne Jules Marey [1885], champion of the graphic method, stated: In effect, in the field of rigorous experimentation all the sciences give a hand. Whatever is the object of these studies, that which measures a force or movement, an electrical state or a temperature, whether he be a physician, chemist or physiologist, he has recourse to the same method and employs the same instruments. Development of the galvanometer and the electric telegraph provided design concepts and instruments that could be adapted to the measurement of bioelectric events. For example, Thomson’s reflecting telegraphic galvanometer was used by Caton [1875] to display the first electroencephalogram. Ader’s telegraphic string galvanometer [1897] was modified by Einthoven [1903] to create the instrument that introduced clinical electrocardiography. Gasser and Erlanger [1922] adapted the Braun cathode-ray tube to enable recording of short-duration nerve action potentials. Garceau [1935] used the Western Union telegraphic recorder, called the Undulator to create the first direct-inking electroencephalograph. However, in the early days of bioelectricity, ingenious electrophysiologists appropriated many devices from physics and engineering to establish the existence of bioelectric phenomena.

The First Electrocardiogram The electric activity accompanying the heartbeat was discovered with the rheoscopic frog by Kolliker and Mueller [1856]. When these investigations laid the nerve over the beating ventricle of a frog heart, the muscle twitched once and sometimes twice. Stimulation of the nerve obviously occurred with depolarization and repolarization of the ventricles. Because at that time there were no rapidly responding galvanometers, Donders [1872] recorded the twitches of the rheoscope to provide a graphic demonstration of the existence of an electrocardiographic signal.


FIGURE HP2.1 The capillary electrometer (a) used by Marey and Lippmann in 1876 and a tortoise ventricular electrogram (b) made with it. This is the first cardiac electrogram from a spontaneously beating heart.

Capillary Electrometer Record The capillary electrometer was created especially for recording the electrocardiogram. The principle underlying its operation was being investigated by Lippmann, a colleague of Marey in France. The phenomenon of electrocapillarity is the change in contour of a drop of mercury in dilute sulfuric acid when a current is passed through the mercury—sulfuric acid interface. This phenomenon was put to practical use by Marey [1876], who placed the interface in a capillary tube, transilluminated it, and recorded the contour change on a moving (falling) photographic plate. The two wires from the electrometer were connected to electrodes placed against the exposed tortoise ventricle. Figure HP2.1a illustrates the capillary electrometer, and Fig. HP2.1b is a reproduction of the tortoise ventricular electrogram showing what we now call the R and T waves.

Rheotome Record Probably unaware that Marey had recorded the cardiac electrogram with the capillary electrometer, Burdon-Sanderson [1879] in England used a slow-speed, d’Arsonval-type galvanometer, the rheotome [see Hoff and Geddes, 1957], and induction-coil stimulator [see Geddes et al., 1989] to reconstruct the ventricular electrogram of the frog heart; Fig. HP2.2 illustrates his reconstruction, showing the R and T waves; note their similarity with those obtained by Marey in Fig. HP2.1b.


FIGURE HP2.2 Burdon-Sanderson’s plot of the frog cardiac electrogram. Thirty-five points were determined to make the reconstruction. [Burdon-Sanderson and Page, 1879.]

Mammalian Electrocardiograms When news of the capillary electrometer reached the United Kingdom, many investigators fabricated their own instruments. One of these was Waller, who used it to record the electrocardiogram of a patient whom he called Jimmy. In 1910, Waller revealed the identity of Jimmy, his pet bulldog, shown in Fig. HP2.3 having his ECG recorded with a forepaw and hindpaw in glass vessels containing saline and metal electrodes.

FIGURE HP2.3 Waller’s patient Jimmy having his ECG recorded with the capillary electrometer. (From Waller AD, Hitchcock Lectures, University of London, 1910.)


FIGURE HP2.4 First human apex cardiogram and capillary electrometer ECG (a) and the dipole map for the heart (b), (a from Waller [1887]; b from Waller [1889].)

Waller [1887] obtained the first ECGs from human subjects; Fig. HP2.4a is one of his records which displays the apex cardiogram and the capillary electrometer record, showing the R and T waves. At that time there were no standard sites for electrode placement. Using the extremities, Waller experimented with different sites, discovering that there were favorable and unfavorable sites, i.e., sites where the amplitude was large or small; Table HP2.1 summarizes his findings. From recordings made with these electrodes, Waller [1887] proposed that the heart could be represented as a dipole, as shown in Fig. HP2.4b.

Corrected Capillary Electrometer Records By the 1890s, it was known that the response of the capillary electrometer was slow, and methods were developed to correct recordings to obtain a true voltage-time record. Burch [1892] in the United Kingdom developed a geometric method that used the tangent at each point along the recording. Figure HP2.5 is an illustration showing a capillary electrometer record (dark hump) and the true voltage-time record (biphasic waveform).


TABLE HP2.1

Waller’s Electrode Locations

The unfavorable combinations were: Left hand and left foot Left hand and right foot Right foot and left foot Mouth and right hand The favorable combinations were: Front of chest and back of chest Left hand and right hand Right hand and right foot Right hand and left foot Mouth and left hand Mouth and right foot Mouth and left foot

One who was very dissatisfied with capillary-electrometer records was Einthoven in the Netherlands. He obtained a capillary electrometer record from a subject (Fig. HP2.6a) and applied the correct method to create the ECG shown in Fig. HP2.6b, revealing the intimate details of what he called the Q and S waves, not visible in the capillary electrometer record, in which he used A, B, and C to designate what he later called the P, R, and T waves.

Clinical Electrocardiography The String Galvanometer Einthoven [1903] set himself the task of creating a high-fidelity recorder by improving Ader’s string telegraphic galvanometer. Einthoven used a silvered quarter filament as the conductor and increased the field strength surrounding the filament by using a strong electromagnet. He added a lens to focus the image of the filament onto a moving photographic surface. Thus the thick baseline for the string galvanometer recording was the image of the “string” (quartz filament). Electrocardiac current caused the silvered filament to be deflected, and its excursions, when magnified optically and recorded photographically, constituted the electrocardiogram. Figure HP2.7a illustrates Einthoven’s string galvanometer, and Fig. HP2.7b shows a patient in hospital clothing having his ECG recorded. Measurement of calibration records published by Einthoven reveals a response time (10% to 90%) of 20 ms. The corresponding sinusoidal frequency response is 0 to 25 Hz (30% attenuation). Figure HP2.8 illustrates one of Einthoven’s electromagnets (lower left) and one of his early cameras (center) and string galvanometers (right). Above the bench supporting these are mounted a bow and arrow, which were used by Einthoven to provide the tension on the quartz rod while it was being heated to the melting point, after which the arrow left the bow and extruded the rod to a long slender filament, which was later silvered to make it conducting.


FIGURE HP2.5 Capillary electrometer record (dark hump) and the corrected voltage-time record (biphasic wave).

FIGURE HP2.6 A capillary electrometer record (a) and its corrected version (b) presented by Einthoven to show the inadequacy of the capillary electrometer in displaying rapidly changing waveforms. (From FA Willius, 1941. TE Keys, Cardiac Classics, St. Louis, Mosby.)

Einthoven borrowed handsomely from previous work; he used Marey’s recording chart speed (25 mm/s) and Waller’s bucket electrodes, as well as some of his leads. With the string galvanometer, Einthoven ushered in clinical electrocardiography in the early 1900s; soon the string galvanometer was used worldwide. The first string galvanometer appeared in the United States in 1912, when it was used at the Rockefeller Institute Hospital until 1959. This instrument, which is now in the Smithsonian Institute, was designed by Horatio B. Williams, professor of physiology at the College of Physicians and Surgeons at Columbia University. Williams had spent some time with Einthoven in Leiden in 1910 and 1911. On his return to New York, Williams had Charles F. Hindle, a machinist at Columbia, construct the first American string galvanometer. Soon thereafter, the Cambridge Instrument Company took over manufacture of the Hindle instrument and made them available for sale in the United States. Although it is clear that the concept of an electrical axis, i.e., a cardiac vector, was demonstrated by Waller’s studies, it remained for Einthoven [1913] to make practical use of the concept. Einthoven postulated that the heart was at the center of an equilateral triangle, the apices of which were the right and left shoulders and the point where both legs joined the trunk. In his early studies, Einthoven used the right and left arms and both feet in saline-filled buckets as the three electrodes. Soon he found that the electrocardiogram was negligibly altered if the right foot was removed from the bucket electrode. Thus he adopted three standard leads: right and left arms and left leg (foot). He postulated that if the amplitudes of the electrocardiographic waves are plotted on this triaxial reference frame, it is possible to calculate the magnitude and direction of an electric vector that produces these same voltages in leads I, II, and III, corresponding to the limb electrodes. He further stated that the arithmetic sum of the amplitudes in lead I plus III equals the amplitude in lead II. This is Einthoven’s law, and the relationship is true only for an equilateral triangle reference frame [Valentinuzzi et al., 1970].

Vacuum-Tube Electrocardiograph Not long after Einthoven described his string galvanometer, efforts were begun in the United States to create an electrocardiograph that used vacuum tubes. At that time, there were rapidly responding mirror © 2000 by CRC Press LLC

(b)

FIGURE HP2.7


Einthoven’s string galvanometer (a) and a patient having his ECG (lead 1) recorded (b).

FIGURE HP2.8 Some of Einthoven’s early equipment. On the bench (left) is an electromagnet. In the center is a camera, and on the right is a string galvanometer. On the wall are a bow and arrow; the latter was used to apply force to a quartz rod which was heated, and when pulled by the arrow, created the quartz filament which was later silvered.

galvanometers, as well as a limited number of vacuum tubes, despite the fact that they has been patented only a few years earlier [1907] by DeForest. According to Marvin [1954], the first discussions relative to such an instrument were held in 1917 between Steinmetz, Neuman, and Robinson of the General Electric Engineering Laboratory. The task of establishing feasibility fell to W. R. G. Baker, who assembled a unit and demonstrated its operation to those just identified. However, because of urgent wartime priorities, the project was shelved. In 1921, General Electric reopened the issue of a vacuum-tube ECG. A second prototype was built and demonstrated to the Schenectady County Medical Association some time in 1924 by Robinson and Marvin. The instrument was used by Drs. Newman, Pardee, Mann, and Oppenheim, all physicians in New York City. Subsequently, six commercial models were made. One instrument was sent to each of the four physicians just identified; the fifth was sent to the General Electric Company Hospital; and the sixth was sent to the AMA Convention in Atlantic City in 1925. This latter instrument became a prototype for future models provided by the General Electric X-Ray Division. A U.S. patent application was filed on January 15, 1925, and the instrument was described by Mann [1930]. On December 22, 1931, a patent on the General Electric vacuum-tube ECG was granted to Marvin and Leibing; Fig. HP2.9 shows the circuit diagram of the instrument, including a specimen record (lower left, Fig. HP2.9). The instrument used three triode vacuum tubes in a single-sided, resistance-capacitancecoupled amplifier. It was battery operated, and a unique feature was a viewing screen that allowed the operator to see the motion of the galvanometer beam as it was being recorded by the camera. Between introduction of the string galvanometer and the hot-stylus recorder for ECG, attempts were made to create direct-inking ECG recorders. In a review of scientific instruments, Brian Matthews [1935] reported © 2000 by CRC Press LLC

FIGURE HP2.9

Circuit diagram of the first vacuum-tube ECG, patented on December 12, 1931.

There are two ink-writing electrocardiographs available, the author’s and that of Drs. Duschel and Luthi. Both utilize a moving-iron driving unit with oil damping, a tubular pen writing on moving paper. The former has a battery-coupled amplifier. The latter gives, in effect, D.C. amplification; the potentials to be recorded are interrupted about 500 times per second by a special type of buzzer, after © 2000 by CRC Press LLC

amplification by resistance capacity coupled valves the interrupted output is reflected by the output valve; the amplifier achieves in effect what can be done with a battery-coupled amplifier and obviates the coupling batteries. At present the speed of these direct-recording instruments is barely adequate to show the finer details of the electrocardiogram, but they enable its main features to be recorded instantly. Despite the instant availability of inked recordings of the ECG, those produced by the string galvanometer were superior, and it took some time for a competitor to appear. Such an instrument did appear in the form of the hot-stylus recorder.

Hot-Stylus Recorder The final step toward modern electrocardiography was the introduction of the hot-stylus recorder by Haynes [1936] of the Bell Telephone Laboratories (New York). Prior to that time, there was colored, waxcoated recording paper, the wax being scraped off by a recording stylus, exposing the colored paper. Referring to the scraping method, Haynes wrote However, the method is not adaptable to many types of recording instruments because of the large amount of friction between the recording stylus and paper arising from the pressure necessary to engrave the wax. This pressure can be removed and the friction largely eliminated by the use of a special stylus consisting essentially of a small electric heating coil situated close to the end of a pointed rod in such a way that the temperature of the point may be raised to about 80°C. The point is then capable of melting wax and engraving its surface with only a very small fraction of the pressure before necessary. The described stylus, when used with waxed recording paper, will provide a means of obtaining a permanent record without the use of pen and ink which is adaptable to the most sensitive recording instruments. Following the end of World War II, vacuum-tube electrocardiographs with heated-stylus recorders became very popular; they are still in use today. However, the heritage of a thick baseline, derived from the string-galvanometer days, had to be preserved for some time because clinicians objected to a thin baseline. It took many years for the hot-stylus baseline to be narrowed without protest.

References Ader M. 1987. Sur un nouvel appareil enregistreur pour cables sousmarins. C R Acad Sci 124:1440. Burch GJ. 1892. On the time relations of the excursions of the capillary electrometer. Philos Trans R Soc (Lond) 83A:81. Burch GJ. 1892. On a method of determining the value of rapid variations of potential by means of the capillary electrometer communicated by J. B. Sanderson. Proc R Soc (Lond) 48:89. Burdon-Sanderson JS, Page FJM. 1879. On the time relations of the excitatory process of the ventricle of the heart of the frog. J Physiol (Lond) 2:384. Caton R. 1875. The electric currents of the brain. Br Med 2:278. DeForest L. U.S. patents 841,387 (1907) and 879,532 (1908). Donders FC. 1872. De secondaire contracties order den involed der systolen van het hart, met en zonder vagus-prikkung. Oncerszoek ged in physiol Lab d Utrecht Hoogesch, Bd. 1, S. 256, Derde reeks TI p 246 bis 255. Einthoven W. 1903. Ein neues Galvanometer. Ann Phys 12 (suppl 4):1059. Einthoven W, Fahr G, de Waart A. 1913. Uber die Richtung und die manifeste Grosse der Potential schwankunzen in menschlichen Herzen. Pflugers Arch 150:275. Garceau EL, Davis H. 1935. An Ink-writing electroencephalograph. Arch Neurol Psychiatry 34:1292.


Gasser HS, Erlanger J. 1922. A study of the action currents of nerve with the cathode ray oscillograph. Am Physiol 62:496. Gasser HS, Newcomer HS. 1921. Physiological action currents in the phrenic nerve. Am Physiol 57(1):1. Geddes LA, Foster KS, Senior J, Kahfeld A. 1989. The Inductorium: The stimulator associated with discovery. Med Instrum 23(4):308. Haynes JR. 1936. Heated stylus for use with waxed recording paper. Rev Sci Instrum 7:108. Hoff HE, Geddes LA. 1957. The rheotome and its prehistory: A study in the historical interrelation of electrophysiology and electromechanics. Bull Hist Med 31(3):212. Kolliker RA, Muller J. 1856. Nachweiss der negativen Schwankung des Muskelstromsnaturlich sic contrakinenden Muskel: Verhandl. Phys Med Ges Wurzburg 6:528. Mann H. 1930–31. A light weight portable EKG. Am Heart J 7:796. Marey EJ. 1885. Methode Graphique, 2d ed. Paris, Masosn. Marey EJ. 1876. Des variations electriques des muscles du coeur en particulier etudiee au moyen de l’ectrometre d M. Lippmann. C R Acad Sci 82:975. Marvin HB, et al. 1925. U.S. patent 1,817,913. Matthews BHC. 1935. Recent developments in electrical instruments for biological and medical purposes. J Sci Instrum 12(7):209. Valentinuzzi ME, Geddes LA, Hoff HE, Bourland JD. 1970. Properties of the 30° hexaxial (EinthovenGoldberger) system of vectorcardiography. Cardiovasc Res Cent Bull 9(2):64. Waller AD. 1887. A demonstration on man of electromotive changes accompanying the heart’s beat. J Physiol (Lond) 8:229.


Onaral, B. “Biomedical Signal Analysis.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

VI Biomedical Signal Analysis Banu Onaral Drexel University 52 Biomedical Signals: Origin and Dynamic Characteristics; Frequency-Domain Analysis Arnon Cohen Or ig in of Biomedical Sig nals • Classifi cat ion of Biosig nals • Sto chast ic Signals • Frequency-Domain Analysis • Discrete Signals • Data Windows • Short-Time Fourier Transform (STFT) • Spectral Estimation • Signal Enhancement • Optimal Filtering • Adaptive Filtering • Segmentation of Nonstationary Signals

53 Digital Biomedical Signal Acquisition and Processing Luca T. Mainardi, Anna M. Bianchi, Sergio Cerutti Acquisition • Signal Processing • Conclusion

54 Compression of Digital Biomedical Signals A. Enis Cetin, ¸ Hayrettin Köymen Time-Domain Coding of Biomedical Signals • Frequency-Domain Data Compression Methods • Wavelet or Subband Coding • Hybrid Multichannel ECG Coding

55 Time-Frequency Signal Representations for Biomedical Signals G. Faye Boudreaux-Bartels, Robin Murray One-Dimensional Signal Representations • Desirable Properties of Time-Frequency Representations • TFR Classes • Common TFRs and Their Use in Biomedical Applications

56 Wavelet (Time-Scale) Analysis in Biomedical Signal Processing Nitish V. Thakor, Boris Gramatikov, David Sherman The Wavelet Transform: Variable Time and Frequency Resolution • A Multiresolution Theory: Decomposition of Signals Using Orthogonal Wavelets • Further Developments of the Wavelet Transform • Applications

57 Higher-Order Spectral Analysis Athina P. Petropulu Definitions and Properties of HOS • HOS Computation from Real Data • Linear Processes • Nonlinear Processes • HOS in Biomedical Signal Processing

58 Neural Networks in Biomedical Signal Processing Evangelia Micheli-Tzanakou Neural Networks in Sensory Waveform Analysis • Neural Networks in Speech Recognition • Neural Networks in Cardiology • Neural Networks in Neurology • Discussion

59 Complexity, Scaling, and Fractals in Biomedical Signals Banu Onaral, Joseph P. Cammarota Complex Dynamics • Introduction to Scaling Theories • An Example of the Use of Complexity Theory in the Development of a Model of the Central Nervous System


60 Future Directions: Biomedical Signal Processing and Networked Multimedia Communications Banu Onaral Public Switched Network and ATM • Wireless Communication • Photonics • Virtual Reality

B

IOMEDICAL SIGNAL ANALYSIS CENTERS ON the acquisition and processing of informationbearing signals that emanate from living systems. These vital signals permit us to probe the state of the underlying biologic and physiologic structures and dynamics. Therefore, their interpretation has significant diagnostic value for clinicians and researchers. The detected signals are commonly corrupted with noise. Often, the information cannot be readily extracted from the raw signal, which must be processed in order to yield useful results. Signals and systems engineering knowledge and, in particular, signal-processing expertise are therefore critical in all phases of signal collection and analysis. Biomedical engineers are called on to conceive and implement processing schemes suitable for biomedical signals. They also play a key role in the design and development of biomedical monitoring devices and systems that match advances in signal processing and instrumentation technologies with biomedical needs and requirements. This section is organized in two main parts. In the first part, contributing authors review contemporary methods in biomedical signal processing. The second part is devoted to emerging methods that hold the promise for major enhancements in our ability to extract information from vital signals. The success of signal-processing applications strongly depends on the knowledge about the origin and the nature of the signal. Biomedical signals possess many special properties and hence require special treatment. Also, the need for noninvasive measurements presents unique challenges that demand a clear understanding of biomedical signal characteristics. In the lead chapter, entitled, “Biomedical Signals: Origin and Dynamic Characteristics; Frequency-Domain Analysis,” Arnon Cohen provides a general classification of biomedical signals and discusses basics of frequency domain methods. The advent of digital computing coupled with fast progress in discrete-time signal processing has led to efficient and flexible methods to acquire and treat biomedical data in digital form. The chapter entitled, “Digital Biomedical Signal Acquisition and Processing,” by Luca T. Mainardi, Anna M. Bianchi, and Sergio Cerutti, presents basic elements of signal acquisition and processing in the special context of biomedical signals. Especially in the case of long-term monitoring, digital biomedical signal-processing applications generate vast amounts of data that strain transmission and storage resources. The creation of multipatient reference signal bases also places severe demands on storage. Data compression methods overcome these obstacles by eliminating signal redundancies while retaining clinically significant information. A. Enis Cetin and Hayrettin Köymen provide a comparative overview of a range of approaches from conventional to modern compression techniques suitable for biomedical signals. Futuristic applications involving longterm and ambulatory recording systems, and remote diagnosis opportunities will be made possible by breakthroughs in biomedical data compression. This chapter serves well as a point of departure. Constraints such as stationarity (and time invariance), gaussianity (and minimum phaseness), and the assumption of a characteristic scale in time and space have constituted the basic, and by now implicit, assumptions upon which the conventional signals and systems theories have been founded. However, investigators engaged in the study of biomedical processes have long known that they did not hold under most realistic situations and hence could not sustain the test of practice. Rejecting or at least relaxing restrictive assumptions always opens new avenues for research and yields fruitful results. Liberating forces in signals and systems theories have conspired in recent years to create research fronts that target long-standing constraints in the established wisdom (dogma?) of classic signal processing and system analysis. The emergence of new fields in signals and system theories that address these shortcomings and aim to relax these restrictions has been motivated by scientists who, rather than mold natural behavior into artificial models, seek methods inherently suited to represent reality. Bio-


medical scientists and engineers are inspired by insights gained from a deeper appreciation for the dynamic richness displayed by biomedical phenomena; hence, more than their counterparts in other disciplines, they more forcefully embrace innovations in signal processing. One of these novel directions is concerned with time-frequency representations tailored for nonstationary and transient signals. Faye Boudreaux-Bartels and Robin Murray address this issue, provide an introduction to concepts and tools of time-frequency analysis, and point out candidate applications. Many physiologic structures and dynamics defy the concept of a characteristic spatial and temporal scale and must be dealt with employing methods compatible with their multiscale nature. Judging from the recent success of biomedical signal-processing applications based on time-scale analysis and wavelet transforms, the resolution of many outstanding processing issues may be at hand. The chapter entitled, “Time-Scale Analysis and Wavelets in Biomedical Signals,” by Nitish V. Thakor, familiarizes the reader with fundamental concepts and methods of wavelet analysis and suggests fruitful directions in biomedical signal processing. The presence of nonlinearities and statistics that do not comply with the gaussianity assumption and the desire for phase reconstruction have been the moving forces behind investigations of higher-order statistics and polyspectra in signal-processing and system-identification fields. An introduction to the topic and potential uses in biomedical signal-processing applications are presented by Athina Petropulu in the chapter entitled, “Higher-Order Spectra in Biomedical Signal Processing.” Neural networks derive their cue from biologic systems and, in turn, mimic many of the functions of the nervous system. Simple networks can filter, recall, switch, amplify, and recognize patterns and hence serve well many signal-processing purposes. In the chapter entitled, “Neural Networks in Biomedical Signal Processing,” Evangelia Tzanakou helps the reader explore the power of the approach while stressing how biomedical signal-processing applications benefit from incorporating neural-network principles. The dichotomy between order and disorder is now perceived as a ubiquitous property inherent in the unfolding of many natural complex phenomena. In the last decade, it has become clear that the common threads shared by natural forms and functions are the “physics of disorder” and the “scaling order,” the hallmark of broad classes of fractal entities. Biomedical signals are the global observables of underlying complex physical and physiologic processes. “Complexity” theories therefore hold the potential to provide mathematical tools that describe and possibly shed light on the internal workings of physiologic systems. In the next to last chapter in this section, Banu Onaral and Joseph P. Cammarota introduce the reader to basic tenets of complexity theories and the attendant scaling concepts with hopes to facilitate their integration into the biomedical engineering practice. The section concludes with a brief chapter on the visions of the future when biomedical signal processing will merge with the rising technologies in telecommunication and multimedia computing, and eventually with virtual reality, to enable remote monitoring, diagnosis, and intervention. The impact of this development on the delivery of health care and the quality of life will no doubt be profound. The promise of biomedical signal analysis will then be fulfilled.


Cross-biospectral plots of hippocampal EEG.

Cohen, A. “Biomedical Signals: Origin and Dynamic Characteristics; Frequency-Domain Analysis” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

52 Biomedical Signals: Origin and Dynamic Characteristics; Frequency-Domain Analysis 52.1 52.2 52.3 52.4 52.5 52.6 52.7 52.8

Origin of Biomedical Signals Classification of Biosignals Stochastic Signals Frequency-Domain Analysis Discrete Signals Data Windows Short-Time Fourier Transform (STFT) Spectral Estimation The Blackman-Tukey Method • The Periodogram • TimeSeries Analysis Methods

52.9 52.10

Signal Enhancement Optimal Filtering Minimization of Mean Squared Error: The Wiener Filter • Maximization of the Signal-to-Noise Ratio: The Matched Filter

Arnon Cohen Ben-Gurion University

52.11 52.12

Adaptive Filtering Segmentation of Nonstationary Signals

A signal is a phenomenon that conveys information. Biomedical signals are signals, used in biomedical fields, mainly for extracting information on a biologic system under investigation. The complete process of information extraction may be as simple as a physician estimating the patient’s mean heart rate by feeling, with the fingertips, the blood pressure pulse or as complex as analyzing the structure of internal soft tissues by means of a complex CT machine. Most often in biomedical applications (as in many other applications), the acquisition of the signal is not sufficient. It is required to process the acquired signal to get the relevant information “buried” in it. This may be due to the fact that the signal is noisy and thus must be “cleaned” (or in more professional terminology, the signal has to be enhanced) or due to the fact that the relevant information is not “visible” in the signal. In the latter case, we usually apply some transformation to enhance the required information.


The processing of biomedical signals poses some unique problems. The reason for this is mainly the complexity of the underlying system and the need to perform indirect, noninvasive measurements. A large number of processing methods and algorithms is available. In order to apply the best method, the user must know the goal of the processing, the test conditions, and the characteristics of the underlying signal. In this chapter, the characteristics of biomedical signals will be discussed [Cohen, 1986]. Biomedical signals will be divided into characteristic classes, requiring different classes of processing methods. Also in this chapter, the basics of frequency-domain processing methods will be presented.

52.1 Origin of Biomedical Signals From the broad definition of the biomedical signal presented in the preceding section, it is clear that biomedical signals differ from other signals only in terms of the application—signals that are used in the biomedical field. As such, biomedical signals originate from a variety of sources. The following is a brief description of these sources: • Bioelectric signals. The bioelectric signal is unique to biomedical systems. It is generated by nerve cells and muscle cells. Its source is the membrane potential, which under certain conditions may be excited to generate an action potential. In single cell measurements, where specific microelectrodes are used as sensors, the action potential itself is the biomedical signal. In more gross measurements, where, for example, surface electrodes are used as sensors, the electric field generated by the action of many cells, distributed in the electrode’s vicinity, constitutes the bioelectric signal. Bioelectric signals are probably the most important biosignals. The fact that most important biosystems use excitable cells makes it possible to use biosignals to study and monitor the main functions of the systems. The electric field propagates through the biologic medium, and thus the potential may be acquired at relatively convenient locations on the surface, eliminating the need to invade the system. The bioelectric signal requires a relatively simple transducer for its acquisition. A transducer is needed because the electric conduction in the biomedical medium is done by means of ions, while the conduction in the measurement system is by electrons. All these lead to the fact that the bioelectric signal is widely used in most fields of biomedicine. • Bioimpedance signals. The impedance of the tissue contains important information concerning its composition, blood volume, blood distribution, endocrine activity, automatic nervous system activity, and more. The bioimpedance signal is usually generated by injecting into the tissue under test sinusoidal currents (frequency range of 50 kHz to 1 MHz, with low current densities of the order of 20 µA to 20 mA). The frequency range is chosen to minimize electrode polarization problems, and the low current densities are chosen to avoid tissue damage mainly due to heating effects. Bioimpedance measurements are usually performed with four electrodes. Two source electrodes are connected to a current source and are used to inject the current into the tissue. The two measurement electrodes are placed on the tissue under investigation and are used to measure the voltage drop generated by the current and the tissue impedance. • Bioacoustic signals. Many biomedical phenomena create acoustic noise. The measurement of this acoustic noise provides information about the underlying phenomenon. The flow of blood in the heart, through the heart’s valves, or through blood vessels generates typical acoustic noise. The flow of air through the upper and lower airways and in the lungs creates acoustic sounds. These sounds, known as coughs, snores, and chest and lung sounds, are used extensively in medicine. Sounds are also generated in the digestive tract and in the joints. It also has been observed that the contracting muscle produces an acoustic noise (muscle noise). Since the acoustic energy propagates through the biologic medium, the bioacoustic signal may be conveniently acquired on the surface, using acoustic transducers (microphones or accelerometers). • Biomagnetic signals. Various organs, such as the brain, heart, and lungs, produce extremely weak magnetic fields. The measurements of these fields provides information not included in other


biosignals (such as bioelectric signals). Due to the low level of the magnetic fields to be measured, biomagnetic signals are usually of very low signal-to-noise ratio. Extreme caution must be taken in designing the acquisition system of these signals. • Biomechanical signals. The term biomechanical signals includes all signals used in the biomedicine fields that originate from some mechanical function of the biologic system. These signals include motion and displacement signals, pressure and tension and flow signals, and others. The measurement of biomechanical signals requires a variety of transducers, not always simple and inexpensive. The mechanical phenomenon does not propagate, as do the electric, magnetic, and acoustic fields. The measurement therefore usually has to be performed at the exact site. This very often complicates the measurement and forces it to be an invasive one. • Biochemical signals. Biochemical signals are the result of chemical measurements from the living tissue or from samples analyzed in the clinical laboratory. Measuring the concentration of various ions inside and in the vicinity of a cell by means of specific ion electrodes is an example of such a signal. Partial pressures of oxygen (pO2) and of carbon dioxide (pCO2) in the blood or respiratory system are other examples. Biochemical signals are most often very low frequency signals. Most biochemical signals are actually dc signals. • Biooptical signals. Biooptical signals are the result of optical functions of the biologic system, occurring naturally or induced by the measurement. Blood oxygenation may be estimated by measuring the transmitted and backscattered light from a tissue (in vivo and in vitro) in several wavelengths. Important information about the fetus may be acquired by measuring fluorescence characteristics of the amniotic fluid. Estimation of the heart output may be performed by the dye dilution method, which requires the monitoring of the appearance of recirculated dye in the bloodstream. The development of fiberoptic technology has opened vast applications of biooptical signals. Table 52.1 lists some of the more common biomedical signals with some of their characteristics.

52.2 Classification of Biosignals Biosignals may be classified in many ways. The following is a brief discussion of some of the most important classifications. • Classification according to source. Biosignals may be classified according to their source or physical nature. This classification was described in the preceding section. This classification may be used when the basic physical characteristics of the underlying process is of interest, e.g., when a model for the signal is desired. • Classification according to biomedical application. The biomedical signal is acquired and processed with some diagnostic, monitoring, or other goal in mind. Classification may be constructed according to the field of application, e.g., cardiology or neurology. Such classification may be of interest when the goal is, for example, the study of physiologic systems. • Classification according to signal characteristics. From point of view of signal analysis, this is the most relevant classification method. When the main goal is processing, it is not relevant what is the source of the signal or to which biomedical system it belongs; what matters are the signal characteristics. We recognize two broad classes of signals: continuous signals and discrete signals. Continuous signals are described by a continuous function s(t) which provides information about the signal at any given time. Discrete signals are described by a sequence s(m) which provides information at a given discrete point on the time axis. Most of the biomedical signals are continuous. Since current technology provides powerful tools for discrete signal processing, we most often transform a continuous signal into a discrete one by a process known as sampling. A given signal s(t) is sampled into the sequence s(m) by


TABLE 52.1

Biomedical Signals

Classification Bioelectric Action potential Electroneurogram (ENG) Electroretinogram (ERG) Electro-oculogram (EOG) Electroencephalogram (EEG) Surface

Acquisition

Frequency Range

Comments Invasive measurement of cell membrane potential Potential of a nerve bundle Evoked flash potential Steady-corneal-retinal potential

Microelectrodes

100 Hz–2 kHz

10 µV–100 mV

Needle electrode Microelectrode Surface electrodes

100 Hz–1 kHz 0.2–200 Hz dc–100 Hz

5 µV–10 mV 0.5 µV–1 mV 10 µV–5 mV

Surface electrodes

0.5–100 Hz

2–100 µV

Multichannel (6–32) scalp potential Young children, deep sleep and pathologies Temporal and central areas during alert states Awake, relaxed, closed eyes

50–100 µV 100–200 µV

Bursts of about 0.2 to 0.6 s Bursts during moderate and deep sleep Response of brain potential to stimulus Occipital lobe recordings, 200-ms duration Sensory cortex Vertex recordings Recordings from exposed surface of brain

Delta range

0.5–4 Hz

Theta range

4–8 Hz

Alpha range Beta range Sleep spindles K-complexes Evoked potentials (EP)

Dynamic Range

8–13 Hz 13–22 Hz 6–15 Hz 12–14 Hz

0.1–20 µV

Surface electrodes

Visual (VEP)

1–300 Hz

1–20 µV

Somatosensory (SEP) Auditory (AEP) Electrocorticogram

Needle electrodes

2 Hz–3 kHz 100 Hz–3 kHz 100 Hz–5 kHz

Electromyography (EMG) Single-fiber (SFEMG)

Needle electrode

500 Hz–10 kHz

1–10 µV

Needle electrode

5 Hz–10 kHz

100 µV–2 mV

2–500 Hz 0.01–1 Hz 0.05–100 Hz 100 Hz–1 kHz

50 µV–5 mV

Motor unit action potential (MUAP) Surface EMG (SEMG) Skeletal muscle Smooth muscle Electrocardiogram (ECG) High-Frequency ECG

0.5–10 µV

Action potentials from single muscle fiber

Surface electrodes

Surface electrodes Surface electrodes

( ) ()

s m =s t

t =mTs

1–10 mV 100 µV–2 mV

m = …, − 1, 0, 1,…

Notchs and slus waveforms superimposed on the ECG.

(52.1)

where Ts is the sampling interval and fs = (2π/Ts ) is the sampling frequency. Further characteristic classification, which applies to continuous as well as discrete signals, is described in Fig. 52.1. We divide signals into two main groups: deterministic and stochastic signals. Deterministic signals are signals that can be exactly described mathematically or graphically. If a signal is deterministic and its mathematical description is given, it conveys no information. Real-world signals are never deterministic. There is always some unknown and unpredictable noise added, some unpredictable change in the parameters, and the underlying characteristics of the signal that render it nondeterministic. It is, however, very often convenient to approximate or model the signal by means of a deterministic function. An important family of deterministic signals is the periodic family. A periodic signal is a deterministic signal that may be expressed by © 2000 by CRC Press LLC

FIGURE 52.1

Classification of signals according to characteristics.

() (

s t = s t + nT

)

(52.2)

where n is an integer, and T is the period. The periodic signal consists of a basic wave shape with a duration of T seconds. The basic wave shape repeats itself an infinite number of times on the time axis. The simplest periodic signal is the sinusoidal signal. Complex periodic signals have more elaborate wave shapes. Under some conditions, the blood pressure signal may be modeled by a complex periodic signal, with the heart rate as its period and the blood pressure wave shape as its basic wave shape. This is, of course, a very rough and inaccurate model. Most deterministic functions are nonperiodic. It is sometimes worthwhile to consider an “almost periodic” type of signal. The ECG signal can sometimes be considered “almost periodic.” The ECG’s RR interval is never constant; in addition, the PQRST complex of one heartbeat is never exactly the same as that of another beat. The signal is definitely nonperiodic. Under certain conditions, however, the RR interval is almost constant, and one PQRST is almost the same as the other. The ECG may thus sometimes be modeled as “almost periodic.”

52.3 Stochastic Signals The most important class of signals is the stochastic class. A stochastic signal is a sample function of a stochastic process. The process produces sample functions, the infinite collection of which is called the ensemble. Each sample function differs from the other in it fine details; however, they all share the same © 2000 by CRC Press LLC

FIGURE 52.2

The ensemble of the stochastic process s(t).

distribution probabilities. Figure 52.2 depicts three sample functions of an ensemble. Note that at any given time, the values of the sample functions are different. Stochastic signals cannot be expressed exactly; they can be described only in terms of probabilities which may be calculated over the ensemble. Assuming a signal s(t), the Nth-order joint probability function

[( )

( ) ] (

()

P s t1 ≤ s1, s t 2 ≤ s2 , …, s t N ≤ s N = P s1, s2 ,…, s N

)

(52.3)

is the joint probability that the signal at time ti will be less than or equal to Si and at time tj will be less than or equal to Sj , etc. This joint probability describes the statistical behavior and intradependence of the process. It is very often useful to work with the derivative of the joint probability function; this derivative is known as the joint probability density function (PDF):

(

)

p s1, s2 ,…, s N =

[(

∂N P s1, s2 ,…, s N ∂s1 ∂s2 L∂s N

)]

(52.4)

Of particular interest are the first- and second-order PDFs. The expectation of the process s(t), denoted by E{s(t)} or by ms , is a statistical operator defined as


{ ( )} = ∫ sp(s)ds = m, ∞

E st

(52.5)

−∞

The expectation of the function sn(t) is known as the nth-order moment. The first-order moment is thus the expectation of the process. The nth-order moment is given by

{ ( )} = ∫ s p(s)ds ∞

E sn t

n

(52.6)

−∞

Another important statistical operator is the nth central moment:

(

) ∫ (s − m ) p(s)ds

n  µ n = E  s − ms  =  

∞

n

(52.7)

s

−∞

The second central moment is known as the variance (the square root of which is the standard deviation). The variance is denoted by σ 2 :

(

) ∫ (s − m ) p(s)ds ∞

2  σ 2 = µ 2 = E  s − ms  =  

2

(52.8)

s

−∞

The second-order joint moment is defined by the joint PDF. Of particular interest is the autocorrelation function rss :

( ) { ( ) ( )} = ∫ ∫ s(t )s(t ) p(s , s )ds ds

rss t1, t 2 = E s t1 s t 2

∞

∞

−∞ −∞

1

2

1

2

1

(52.9)

2

The cross-correlation function is defined as the second joint moment of the signal s at time t1, s(t1), and the signal y at time t2, y(t2):

( ) { ( ) ( )} = ∫ ∫ s(t ) y(t ) p(s , y )ds dy

rsy t1, t 2 = E s t1 y t 2

∞

∞

−∞ −∞

1

2

1

2

1

2

(52.10)

Stationary stochastic processes are processes whose statistics do not change in time. The expectation and the variance (as with any other statistical mean) of a stationary process will be time-independent. The autocorrelation function, for example, of a stationary process will thus be a function of the time difference τ = t2 – t1 (one-dimensional function) rather than a function of t2 and t1 (two-dimensional function). Ergodic stationary processes possess an important characteristic: Their statistical probability distributions (along the ensemble) equal those of their time distributions (along the time axis of any one of its sample functions). For example, the correlation function of an ergodic process may be calculated by its definition (along the ensemble) or along the time axis of any one of its sample functions:

( ) { ( ) ( )}

1 T→ ∞ 2T

rss τ = E s t s t − τ = lim

∫ s(t )s(t − τ)dt T

−T

(52.11)

The right side of Eq. (52.11) is the time autocorrelation function. Ergodic processes are nice because one does not need the ensemble for calculating the distributions; a single sample function is sufficient. From the point of view of processing, it is desirable to model the


signal as an ergodic one. Unfortunately, almost all signals are nonstationary (and hence nonergodic). One must therefore use nonstationary processing methods (such as, for example, wavelet transformation) which are relatively complex or cut the signals into short-duration segments in such a way that each may be considered stationary. The sleep EEG signal, for example, is a nonstationary signal. We may consider segments of the signal, in which the subject was at a given sleep state, as stationary. In order to describe the signal, we need to estimate its probability distributions. However, the ensemble is unavailable. If we further assume that the process is ergodic, the distributions may be estimated along the time axis of the given sample function. Most of the standard processing techniques assume the signal to be stationary and ergodic.

52.4 Frequency-Domain Analysis Until now we have dealt with signals represented in the time domain, that is to say, we have described the signal by means of its value on the time axis. It is possible to use another representation for the same signal: that of the frequency domain. Any signal may be described as a continuum of sine waves having different amplitudes and phases. The frequency representation describes the signals by means of the amplitudes and phases of the sine waves. The transformation between the two representations is given by the Fourier transform (FT):

( ) ∫ s(t )e

Sω =

∞

− jωt

−∞

{ ( )}

dt = F s t

(52.12)

where ω = 2πf is the angular frequency, and F{*} is the Fourier operator. The inverse Fourier transform (IFT) is the operator that transforms a signal from the frequency domain into the time domain:

()

st =

1 2π

∫ S(ω)e ∞

−∞

jωt

{ ( )}

dw = F −1 S ω

(52.13)

The frequency domain representation S(ω) is complex; hence

() ()

Sω =Sω e

( )

jθ ω

(52.14)

where S(ω), the absolute value of the complex function, is the amplitude spectrum, and θ(ω), the phase of the complex function, is the phase spectrum. The square of the absolute value, S(ω) 2, is termed the power spectrum. The power spectrum of a signal describes the distribution of the signal’s power on the frequency axis. A signal in which the power is limited to a finite range of the frequency axis is called a band-limited signal. Figure 52.3 depicts an example of such a signal. The signal in Fig. 52.3 is a band-limited signal; its power spectrum is limited to the frequency range –ωmax ≤ ω ≤ ωmax . It is easy to show that if s(t) is real (which is the case in almost all applications), the amplitude spectrum is an even function and the phase spectrum is an odd function. Special attention must be given to stochastic signals. Applying the FT to a sample function would provide a sample function on the frequency axis. The process may be described by the ensemble of spectra. Another alternative to the frequency representation is to consider the correlation function of the process. This function is deterministic. The FT may be applied to it, yielding a deterministic frequency function. The FT of the correlation function is defined as the power spectral density function (PSD):

[ ( )] ( ) { ( )} ∫ r (τ)e

PSD s t = Sss ω = F rss τ =


∞

−∞

ss

− jωτ

dτ

(52.15)

FIGURE 52.3

Example of a signal described in the time and frequency domains.

The PSD is used to describe stochastic signals; it describes the density of power on the frequency axis. Note that since the autocorrelation function is an even function, the PSD is real; hence no phase spectrum is required. The EEG signal may serve as an example of the importance of the PSD in signal processing. When processing the EEG, it is very helpful to use the PSD. It turns out that the power distribution of the EEG changes according to the physiologic and psychological states of the subject. The PSD may thus serve as a tool for the analysis and recognition of such states. Very often we are interested in the relationship between two processes. This may be the case, for example, when two sides of the brain are investigated by means of EEG signals. The time-domain expression of such relationships is given by the cross-correlation function (Eq. 52.10). The frequencydomain representation of this is given by the FT of the cross-correlation function, which is called the cross-power spectral density function (C-PSD) or the cross-spectrum:

( ) { ( )}

()

Ssy ω = F rsy τ = Ssy ω e

( )

jθsy ω

(52.16)

Note that we have assumed the signals s(t) and y(t) are stationary; hence the cross-correlation function is not a function of time but of the time difference τ. Note also that unlike the autocorrelation function, rsy(τ) is not even; hence its FT is not real. Both absolute value and phase are required. It can be shown that the absolute value of the C-PSD is bounded:

()

Ssy ω

2

() ()

≤ Sss ω S yy ω

(52.17)

The absolute value information of the C-PSD may thus be normalized to provide the coherence function:

γ sy2

( ) ≤1 = S (ω )S (ω ) Ssy ω

ss

2

(52.18)

yy

The coherence function is used in a variety of biomedical applications. It has been used, for example, in EEG analysis to investigate brain asymmetry.


FIGURE 52.4 Amplitude spectrum of a sampled signal with sampling frequency above the Nyquist frequency (upper trace) and below the Nyquist frequency (lower trace).

52.5 Discrete Signals Assume now that the signal s(t) of Fig. 52.3 was sampled using a sampling frequency of fs = ωs /(2π) = (2π)/Ts . The sampled signal is the sequence s(m). The representation of the sampled signal in the frequency domain is given by applying the Fourier operator:

( ) { ( )} ( )

Ss ω = F s m = Ss ω e

( )

jθs ω

(52.19)

The amplitude spectrum of the sampled signal is depicted in Fig. 52.4. It can easily be proven that the spectrum of the sampled signal is the spectrum of the original signal repeated infinite times at frequencies of nωs . The spectrum of a sampled signal is thus a periodic signal in the frequency domain. It can be observed, in Fig. 52.4, that provided the sampling frequency is large enough, the wave shapes of the spectrum do not overlap. In such a case, the original (continuous) signal may be extracted from the sampled signal by low-pass filtering. A low-pass filter with a cutoff frequency of ωmax will yield at its output only the first period of the spectrum, which is exactly the continuous signal. If, however, the sampling frequency is low, the wave shapes overlap, and it will be impossible to regain the continuous signal. The sampling frequency must obey the inequality

ω s ≥ 2ω max

(52.20)

Equation (52.20) is known as the sampling theorem, and the lowest allowable sampling frequency is called the Nyquist frequency. When overlapping does occur, there are errors between the sampled and original © 2000 by CRC Press LLC

signals. These errors are known as aliasing errors. In practical applications, the signal does not possess a finite bandwidth; we therefore limit its bandwidth by an antialiasing filter prior to sampling. The discrete Fourier transform (DFT) [Proakis & Manolakis, 1988] is an important operator that maps a finite sequence s(m), m = 0, 1, … , N – 1, into another finite sequence S(k), k = 0, 1, … , N – 1. The DFT is defined as

{ ( )} ∑ s(m)e N −1

()

S k = DFT s m =

− jkm

(52.21)

m =0

An inverse operator, the inverse discrete Fourier transform (IDFT), is an operator that transforms the sequence S(k) back into the sequence s(m). It is given by

{ ( )} ∑S(k)e N −1

( )

s m = IDFT S k = −

jkm

(52.22)

k =0

It can be shown that if the sequence s(m) represents the samples of the band-limited signal s(t), sampled under Nyquist conditions with sampling interval of Ts , the DFT sequence S(k) (neglecting windowing effects) represents the samples of the FT of the original signal:

() ( )

S k = Ss ω

ω ω =k s N

k = 0, 1, …, N − 1

(52.23)

Figure 52.5 depicts the DFT and its relations to the FT. Note that the N samples of the DFT span the frequency range one period. Since the amplitude spectrum is even, only half the DFT samples carry the information; the other half is composed of the complex conjugates of the first half.

FIGURE 52.5 © 2000 by CRC Press LLC

The sampled signal s(m) and its DFT.

The DFT may be calculated very efficiently by means of the fast (discrete) Fourier transform (FFT) algorithm. It is this fact that makes the DFT an attractive means for FT estimation. The DFT provides an estimate for the FT with frequency resolution of

∆f =

2πf s 2π = N T

(52.24)

where T is the duration of the data window. The resolution may be improved by using a longer window. In cases where it is not possible to have a longer data window, e.g., because the signal is not stationary, zero padding may be used. The sequence may be augmented with zeroes:

( ) {( ) ()

(

)

}

s A m = s 0 , s 1 , …, s N − 1 , 0, …, 0

(52.25)

The zero padded sequence sA(m), m = 0, 1, … , L – 1, contains N elements of the original sequence and L – N zeroes. It can be shown that its DFT represents the samples of the FT with an increased resolution of ∆f = 2 πfsL–1.

52.6 Data Windows Calculation of the various functions previously defined, such as the correlation function, requires knowledge of the signal from minus infinity to infinity. This is, of course, impractical because the signal is not available for long durations and the results of the calculations are expected at a reasonable time. We therefore do not use the signal itself but the windowed signal. A window w(t) is defined as a real and even function that is also time-limited:

()

w t =0

∀t > T 2

The FT of a window W(ω) is thus real and even and is not band-limited. Multiplying a signal by a window will zero the signal outside the window duration (the observation period) and will create a windowed, time-limited signal sw(t):

() () ()

sw t = s t w t

(52.26)

In the frequency domain, the windowed signal will be

() () ()

Sw ω = S ω ∗ W ω

(52.27)

where (*) is the convolution operator. The effect of windowing on the spectrum of the signal is thus the convolution with the FT of the window. A window with very narrow spectrum will cause low distortions. A practical window has an FT with a main lobe, where most of its energy is located, and sidelobes, which cover the frequency axis. The convolution of the sidelobes with the FT of the signal causes distortions known as spectral leakage. Many windows have been suggested for a variety of applications. The simplest window is the rectangular (Dirichlet) window; in its discrete form it is given by w(m) = 1, m = 0, 1, … , N – 1. A more useful window is the Hamming window, given by

 2π  w m = 0.54 − 0.46 cos  m ; m = 0, 1, …, N − 1 N 

( )

The Hamming window was designed to minimize the effects of the first sidelobe. © 2000 by CRC Press LLC

(52.28)

52.7 Short-Time Fourier Transform (STFT) The Fourier analysis discussed in preceding sections assumed that the signal is stationary. Unfortunately, most signals are nonstationary. A relatively simple way to deal with the problem is to divide the signal into short segments. The segments are chosen such that each one by itself can be considered a windowed sample of a stationary process. The duration of the segments has to be determined either by having some a priori information about the signal or by examining its local characteristics. Depending on the signal and the application, the segments may be of equal or different duration. We want to represent such a segmented signal in the frequency domain. We define the short-time Fourier transform (STFT):

( ) { ( ) ( )} ∫ s(t )w(t − τ)e

STFTs ω, τ = F s t w t − τ =

∞

−∞

− jωt

dt

(52.29)

The window is shifted on the time axis to t = τ so that the FT is performed on a windowed segment in the range τ – (T/2) ≤ t ≤ τ + (T/2). The STFT describes the amplitude and phase-frequency distributions of the signal in the vicinity of t = τ. In general, the STFT is a two-dimensional, time-frequency function. The resolution of the STFT on the time axis depends on the duration T of the window. The narrower the window, the better the time resolution. Unfortunately, choosing a short-duration window means a wider-band window. The wider the window in the frequency domain, the larger the spectral leakage and hence the deterioration of the frequency resolution. One of the main drawbacks of the STFT method is the fact that the time and frequency resolutions are linked together. Other methods, such as the wavelet transform, are able to better deal with the problem. In highly nonstationary signals, such as speech signals, equal-duration windows are used. Window duration is on the order of 10 to 20 ms. In other signals, such as the EEG, variable-duration windows are used. In the EEG, windows on the order of 5 to 30 s are often used. A common way for representing the two-dimensional STFT function is by means of the spectrogram. In the spectrogram, the time and frequency axes are plotted, and the STFT PSD value is given by the gray-scale code or by a color code. Figure 52.6 depicts a simple spectrogram. The time axis is quantized to the window duration T. The gray scale codes the PSD such that black denotes maximum power and white denotes zero power. In Figure 52.6, the PSD is quantized into only four levels of gray. The spectrogram shows a signal that is nonstationary in the time range 0 to 8T. In this time range, the PSD possesses a peak that is shifted from about 0.6fs to about 0.1fs at time 0.7T. From time 0.8T, the signal becomes stationary with a PSD peak power in the low-frequency range and the high-frequency range.

FIGURE 52.6


A spectrogram.

52.8 Spectral Estimation The PSD is a very useful tool in biomedical signal processing. It is, however, impossible to calculate, since it requires infinite integration time. Estimation methods must be used to acquire an estimate of the PSD from a given finite sample of the process under investigation. Many algorithms for spectral estimation are available in the literature [Kay, 1988], each with its advantages and drawbacks. One method may be suitable for processes with sharp spectral peaks, while another will perform best for broad, smoothed spectra. An a priori knowledge on the type of PSD one is investigating helps in choosing the proper spectral estimation method. Some of the PSD estimation methods will be discussed here.

The Blackman-Tukey Method This method estimates the PSD directly from its definition (Eq. 52.15) but uses finite integration time and an estimate rather than the true correlation function. In its discrete form, the PSD estimation is

()

( )

∑ rˆ (m)e M

Sˆ xx ω = Ts

xx

− jωmTs

m = −M

1 rˆxx m = N

(52.30)

N − i −1

∑ x(m + i) x(i) i=0

where N is the number of samples used for the estimation of the correlation coefficients, and M is the number of correlation coefficients used for estimation of the PSD. Note that a biased estimation of the correlation is employed. Note also that once the correlations have been estimated, the PSD may be calculated by applying the FFT to the correlation sequence.

The Periodogram The periodogram estimates the PSD directly from the signal without the need to first estimate the correlation. It can be shown that

S xx

 1 ω = lim E  T→ ∞  2T

()

∫ () T

−T

xte

  dt   2

− jωt

(52.31)

The PSD presented in Eq. (52.31) requires infinite integration time. The periodogram estimates the PSD from a finite observation time by dropping the lim operator. It can be shown that in its discrete form, the periodogram estimator is given by

()

{ ( )}

T Sˆ xx ω = s DFT x m N

2

(52.32)

The great advantage of the periodogram is that the DFT operator can very efficiently be calculated by the FFT algorithm. A modification to the periodogram is weighted overlapped segment averaging (WOSA). Rather than using one segment of N samples, we divide the observation segment into shorter subsegments, perform a periodogram for each one, and then average all periodograms. The WOSA method provides a smoother estimate of the PSD.


FIGURE 52.7

Time-series model for the signal s(m).

Time-Series Analysis Methods Time-series analysis methods model the signal as an output of a linear system driven by a white source. Figure 52.7 depicts this model in its discrete form. Since the input is a white noise process (with zero mean and unity variance), the PSD of the signal is given by

() ()

Sss ω = H ω

2

(52.33)

The PSD of the signal may thus be represented by the system’s transfer function. Consider a general pole-zero system with p poles and q zeros [ARMA(p, q)]: q

()

∑b z

−i

i

i=0

H z =

(52.34)

p

1+

∑a z

−i

i

i =1

Its absolute value evaluated on the frequency axis is 2

q

()

H ω

∑b z

−i

i

2

=

i=0

1+

(52.35)

2

p

∑a z

−i

i

i =1

z =e − jωTs

Several algorithms are available for the estimation of the model’s coefficients. The estimation of the ARMA model parameters requires the solution of a nonlinear set of equations. The special case of q = 0, namely, an all-pole model [AR(p)], may be estimated by means of linear equations. Efficient AR estimation algorithms are available, making it a popular means for PSD estimation. Figure 52.8 shows the estimation of EMG PSD using several estimation methods.

52.9 Signal Enhancement The biomedical signal is very often a weak signal contaminated by noise. Consider, for example, the problem of monitoring the ECG signal. The signal is acquired by surface electrodes that pick up the electric potential generated by the heart muscle. In addition, the electrodes pick up potentials from other active muscles. When the subject is at rest, this type of noise may be very small, but when the subject is an athlete


FIGURE 52.8 PSD of surface EMG. (Upper trace) Blackman-Tukey (256 correlation coefficients and 256 padding zeroes). (Middle trace) Periodogram (512 samples and 512 padding zeroes). (Lower trace) AR model (p = 40).

performing some exercise, the muscle noise may become dominant. Additional noise may enter the system from electrodes motion, from the power lines, and from other sources. The first task of processing is usually to enhance the signal by “cleaning” the noise without (if possible) distorting the signal. Assume a simple case where the measured signal x(t) is given by

() () ()

x t = s t +n t

() () ()

X ω =S ω +N ω

(52.36)

where s(t) is the desired signal and n(t) is the additive noise. For simplicity, we assume that both the signal and noise are band-limited, namely, for the signal, S(ω) = 0, for ωmax ≤ ω, ωmin ≥ ω. Figure 52.9


FIGURE 52.9

Noisy signal in the frequency domain.

depicts the PSD of the signal in two cases, the first where the PSD of the signal and noise do not overlap and the second where they do overlap (for the sake of simplicity, only the positive frequency axis was plotted). We want to enhance the signal by means of linear filtering. The problem is to design the linear filter that will provide best enhancement. Assuming we have the filter, its output, the enhanced signal, is given by

( ) ( ) ( ) Y (ω ) = X (ω ) H (ω )

y t = x t ∗h t

(52.37)

where y(t) = ˆs(t) + no(t) is the enhanced output, and h(t) is the impulse response of the filter. The solution for the first case is trivial; we need an ideal bandpass filter whose transfer function H(ω) is

1 H ω = 0

()

ω min < ω < ω max otherwise

(52.38)

Such a filter and its output are depicted in Fig. 52.10. As is clearly seen in Fig. 52.10, the desired signal s(t) was completely recovered from the given noisy signal x(t). Practically, we do not have ideal filters, so some distortions and some noise contamination will always appear at the output. With the correct design, we can approximate the ideal filter so that the distortions and noise may be as small as we desire. The enhancement of overlapping noisy signals is far from being trivial.

52.10

Optimal Filtering

When the PSD of signal and noise overlap, complete, undistorted recovery of the signal is impossible. Optimal processing is required, with the first task being definition of the optimality criterion. Different criteria will result in different solutions to the problem. Two approaches will be presented here: the Wiener filter and the matched filter.


FIGURE 52.10 filter.

(a) An ideal bandpass filter. (b) Enhancement of a nonoverlapping noisy signal by an ideal bandpass

Minimization of Mean Squared Error: The Wiener Filter Assume that our goal is to estimate, at time t + ξ, the value of the signal s(t + ξ), based on the observations x(t). The case ξ = 0 is known as smoothing, while the case ξ > 0 is called prediction. We define an output error ε(t) as the error between the filter’s output and the desired output. The expectation of the square of the error is given by

E

{ (t )} = E [s(t + ξ) − y(t + ξ)] 

2

2

  = E  s t + ξ − 

 h τ x t − τ dτ  −∞ 

( ) ∫ ()( ) ∞

2

   

(52.39)

The integral term on the right side of Eq. (52.39) is the convolution integral expressing the output of the filter. The minimization of Eq. (52.39) with respect of h(t) yields the optimal filter (in the sense of minimum squared error). The minimization yields the Wiener-Hopf equation:


( ) ∫ h(η)r (τ ⋅ −η)dη

rsx τ + ξ =

∞

(52.40)

xx

−∞

In the frequency domain, this equation becomes

()

() ()

Ssx ω e jωξ = H opt ω S xx ω

(52.41)

from which the optimal filter Hopt(ω) can be calculated:

()

H opt ω =

( )e (ω )

Ssx ω S xx

jωξ

=

()

Ssx ω

()

()

Sss ω + Snn ω

e jωξ

(52.42)

If the signal and noise are uncorrelated and either the signal or the noise has zero mean, the last equation becomes

()

H opt ω =

()

Sss ω

()

()

Sss ω + Snn ω

e jωξ

(52.43)

The optimal filter requires a priori knowledge of the PSD of noise and signal. These are very often not available and must be estimated from the available signal. The optimal filter given in Eqs. (52.42) and (52.43) is not necessarily realizable. In performing the minimization, we have not introduced a constraint that will ensure that the filter is causal. This can be done, yielding the realizable optimal filter.

Maximization of the Signal-to-Noise Ratio: The Matched Filter The Wiener filter was optimally designed to yield an output as close as possible to the signal. In many cases we are not interested in the fine details of the signal but only in the question whether the signal exists at a particular observation or not. Consider, for example, the case of determining the heart rate of a subject under noisy conditions. We need to detect the presence of the R wave in the ECG. The exact shape of the wave is not important. For this case, the optimality criterion used in the last section is not suitable. To find a more suitable criterion, we define the output signal-to-noise ratio: Let us assume that the signal s(t) is a deterministic function. The response of the filter, ˆs(t) = s(t) * h(t), to the signal is also deterministic. We shall define the output signal-to-noise ratio

()

SNR o t =

() E {n (t )} sˆ t

(52.44)

2 o

as the optimality criterion. The optimal filter will be the filter that maximizes the output SNR at a certain given time t = to . The maximization yields the following integral equation:

∫ h(ξ)r (τ − ξ)dξ = αs(t − τ) T

0

nn

o

o ≤ τ ≤T

(52.45)

where T is the observation time and α is any constant. This equation has to be solved for any given noise and signal.


A special important case is the case where the noise is a white noise so that its autocorrelation function is a delta function. In this case, the solution of Eq. (52.45) is

()

hτ =

(

1 s to − τ N

)

(52.46)

where N is the noise power. For this special case, the impulse response of the optimal filter has the form of the signal run backward, shifted to the time to . This type of filter is called a matched filter.

52.11

Adaptive Filtering

The optimal filters discussed in the preceding section assumed the signals to be stationary with known PSD. Both assumptions rarely occur in reality. In most biomedical applications, the signals are nonstationary with unknown PSD. To enhance such signals, we require a filter that will continuously adjust itself to perform optimally under the changing circumstances. Such a filter is called an adaptive filter [Widrow & Stearns, 1985]. The general description of an adaptive filter is depicted in Fig. 52.11. The signal s(t) is to be corrected according to the specific application. The correction may be enhancement or some reshaping. The signal is given in terms of the noisy observation signal x(t). The main part of the system is a filter, and the parameters (gain, poles, and zeroes) are controllable by the adaptive algorithm. The adaptive algorithm has some a priori information on the signal and the noise (the amount and type of information depend on the application). It also has a correction criterion, according to which the signal is operating. The adaptive algorithm also gets the input and output signals of the filter so that its performance can be analyzed continuously. The adaptive filter requires a correction algorithm. This can best be implemented digitally. Most adaptive filters therefore are implemented by means of computers or special digital processing chips. An important class of adaptive filters requires a reference signal. The knowledge of the noise required by this type of adaptive filter is a reference signal that is correlated with the noise. The filter thus has two inputs: the noisy signal x(t) = s(t) + n(t) and the reference signal nR(t). The adaptive filter, functioning

FIGURE 52.11


Adaptive filter, general scheme.

as a noise canceler, estimates the noise n(t) and, by subtracting it from the given noisy input, gets an estimate for the signal. Hence

( ) ( ) ( ) ( ) [ ( ) ( )] ( )

y t = x t − nˆ t = s t + n t − nˆ t = sˆ t

(52.47)

The output of the filter is the enhanced signal. Since the reference signal is correlated with the noise, the following relationship exists:

() () ()

NR ω = G ω N ω

(52.48)

which means that the reference noise may be represented as the output of an unknown filter G(ω). The adaptive filter estimates the inverse of this unknown noise filter and from its estimates the noise:

{ ( ) ( )}

()

nˆ t = F −1 Gˆ −1 ω N R ω

(52.49)

The estimation of the inverse filter is done by the minimization of some performance criterion. There are two dominant algorithms for the optimization: the recursive least squares (RLS) and the least mean squares (LMS). The LMS algorithm will be discussed here. Consider the mean square error

  { ( )} = E{y (t )} = E  s(t ) + [n(t ) − nˆ(t )]  2

E ε2 t

2

{ ( )} [ ( ) ( )]

 = E s 2 t + E  n t − nˆ t 

2

  

(52.50)

The right side of Eq. (52.50) is correct, assuming that the signal and noise are uncorrelated. We are searching for the estimate Gˆ –1(ω) that will minimize the mean square error: E{[n(t) – nˆ (t)]2}. Since the estimated filter affects only the estimated noise, the minimization of the noise error is equivalent to the minimization of Eq. (52.50). The implementation of the LMS filter will be presented in its discrete form (see Fig. 52.12). The estimated noise is p

( ) ∑v w = v

nˆ R m =

i

i

T m

w

(52.51)

i=0

where

[ ] [ ( ) = [w , w , …, w ]

(

v Tm = v0 , v1, …, v p = 1, n R m − 1 , …, n R m − p w

T

0

)]

(52.52)

p

The vector w represents the filter. The steepest descent minimization of Eq. (52.50) with respect to the filter’s coefficients w yields the iterative algorithm

w j +1 = w j + 2µ j v j


(52.53)

FIGURE 52.12

Block diagram of LMS adaptive noise canceler.

where µ is a scalar that controls the stability and convergence of the algorithm. In the evaluation of Eq. (52.53), the assumption

∂E

{ }≅ ∂ j

2

j

∂wk

∂wk

2

(52.54)

was made. This is indeed a drastic approximation; the results, however, are very satisfactory. Figure 54.12 depicts the block diagram of the LMS adaptive noise canceler. The LMS adaptive noise canceler has been applied to many biomedical problems, among them cancellation of power-line interferences, elimination of electrosurgical interferences, enhancement of fetal ECG, noise reduction for the hearing impaired, and enhancement of evoked potentials.

52.12

Segmentation of Nonstationary Signals

Most biomedical signals are nonstationary, yet the common processing techniques (such as the FT) deal with stationary signals. The STFT is one method of processing nonstationary signals, but it does require, however, the segmentation of the signal into “almost” stationary segments. The signal is thus represented as a piecewise-stationary signal. An important problem in biomedical signal processing is efficient segmentation. In very highly nonstationary signals, such as the speech signal, short, constant-duration (of the order of 15 ms) segments are used. The segmentation processing in such a case is simple and inexpensive. In other cases such as the monitoring of nocturnal EEG, a more elaborate segmentation procedure is called for because the signal may consist of “stationary” segments with very wide duration range. Segmentation into a priori fixed-duration segments will be very inefficient in such cases.


FIGURE 52.13 Adaptive segmentation of simulated EEG. First 2.5 seconds and last 2.5 seconds were simulated by means of different AR models. (Lower trace) SEM calculated with fixed reference window. A new segment has been detected at t – 2.5. [From Cohen, 1986, with permission.]

Several adaptive segmentation algorithms have been suggested. Figure 52.13 demonstrates the basic idea of these algorithms. A fixed reference window is used to define an initial segment of the signal. The duration of the reference window is determined such that it is long enough to allow a reliable PSD estimate yet short enough so that the segment may still be considered stationary. Some a priori information about the signal will help in determining the reference window duration. A second, sliding window is shifted along the signal. The PSD of the segment defined by the sliding window is estimated at each window position. The two spectra are compared using some spectral distance measure. As long as this distance measure remains below a certain decision threshold, the reference segment and the sliding segment are considered close enough and are related to the same stationary segment. Once the distance


measure exceeds the decision threshold, a new segment is defined. The process continues by defining the last sliding window as the reference window of the new segment. Let us define a relative spectral distance measure

() ∫

Dt ω =

ωM

ωM

() () ()

2

S ω −S ω  R t   dω   SR ω  

(52.55)

where SR(ω) and St(ω) are the PSD estimates of the reference and sliding segments, respectively, and ωM is the bandwidth of the signal. A normalized spectral measure was chosen, since we are interested in differences in the shape of the PSD and not in the gain. Some of the segmentation algorithms use growing reference windows rather than fixed ones. This is depicted in the upper part of Fig. 52.13. The various segmentation methods differ in the way the PSDs are estimated. Two of the more well-known segmentation methods are the auto-correlation measure method (ACM) and the spectral error measure (SEM).

References Cohen A. 1986. Biomedical Signal Processing. Boca Raton, Fla, CRC Press. Kay SM. 1988. Modern Spectral Estimation: Theory and Application. Englewood Cliffs, NJ, Prentice-Hall. Proakis JG, Manolakis DG. 1988. Introduction to Digital Signal Processing. New York, Macmillan. Weitkunat R (ed). 1991. Digital Biosignal Processing. Amsterdam, Elsevier. Widrow B, Stearns SD. 1985. Adaptive Signal Processing. Englewood Cliffs, NJ, Prentice-Hall.


Mainardi, L.T.,Bianchi, A.M., Cerutti, S. “Digital Biomedical Signal Acquisition and Processing.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

53 Digital Biomedical Signal Acquisition and Processing Luca T. Mainardi Polytechnic University, Milan

Anna M. Bianchi St. Raffaele Hospital, Milan

53.1 53.2

Sergio Cerutti Polytechnic University, Milan

Acquisition The Sampling Theorem • The Quantization Effects

Signal Processing Digital Filters • Signal Averaging • Spectral Analysis

53.3

Conclusion

Biologic signals carry information that is useful for comprehension of the complex pathophysiologic mechanisms underlying the behavior of living systems. Nevertheless, such information cannot be available directly from the raw recorded signals; it can be masked by other biologic signals contemporaneously detected (endogenous effects) or buried in some additive noise (exogenous effects). For such reasons, some additional processing is usually required to enhance the relevant information and to extract from it parameters that quantify the behavior of the system under study, mainly for physiologic studies, or that define the degree of pathology for routine clinical procedures (diagnosis, therapy, or rehabilitation). Several processing techniques can be used for such purposes (they are also called preprocessing techniques); time- or frequency-domain methods including filtering, averaging, spectral estimation, and others. Even if it is possible to deal with continuous time waveforms, it is usually convenient to convert them into a numerical form before processing. The recent progress of digital technology, in terms of both hardware and software, makes digital rather than analog processing more efficient and flexible. Digital techniques have several advantages: Their performance is generally powerful, being able to easily implement even complex algorithms, and accuracy depends only on the truncation and round-off errors, whose effects can be predicted and controlled by the designer and are largely unaffected by other unpredictable variables such as component aging and temperature, which can degrade the performances of analog devices. Moreover, design parameters can be more easily changed because they involve software rather than hardware modifications. A few basic elements of signal acquisition and processing will be presented in the following; our aim is to stress mainly the aspects connected with acquisition and analysis of biologic signals, leaving to the cited literature a deeper insight into the various subjects for both the fundamentals of digital signal processing and the applications.

53.1 Acquisition A schematic representation of a general acquisition system is shown in Fig. 53.1. Several physical magnitudes are usually measured from biologic systems. They include electromagnetic quantities (currents,


FIGURE 53.1

General block diagram of the acquisition procedure of a digital signal.

potential differences, field strengths, etc.), as well as mechanical, chemical, or generally nonelectrical variables (pressure, temperature, movements, etc.). Electric signals are detected by sensors (mainly electrodes), while nonelectric magnitudes are first converted by transducers into electric signals that can be easily treated, transmitted, and stored. Several books of biomedical instrumentation give detailed descriptions of the various transducers and the hardware requirements associated with the acquisition of the different biologic signals [Cobbold, 1988; Tompkins & Webster, 1981; Webster, 1992]. An analog preprocessing block is usually required to amplify and filter the signal (in order to make it satisfy the requirements of the hardware such as the dynamic of the analog-to-digital converter), to compensate for some unwanted sensor characteristics, or to reduce the portion of undesired noise. Moreover, the continuous-time signal should be bandlimited before analog-to-digital (A/D) conversion. Such an operation is needed to reduce the effect of aliasing induced by sampling, as will be described in the next section. Here it is important to remember that the acquisition procedure should preserve the information contained in the original signal waveform. This is a crucial point when recording biologic signals, whose characteristics often may be considered by physicians as indices of some underlying pathologies (i.e., the ST-segment displacement on an ECG signal can be considered a marker of ischemia, the peak-and-wave pattern on an EEG tracing can be a sign of epilepsy, and so on). Thus the acquisition system should not introduce any form of distortion that can be misleading or can destroy real pathologic alterations. For this reason, the analog prefiltering block should be designed with constant modulus and linear phase (or zero-phase) frequency response, at least in the passband, over the frequencies of interest. Such requirements make the signal arrive undistorted up to the A/D converter. The analog waveform is then A/D converted into a digital signal; i.e., it is transformed into a series of numbers, discretized both in time and amplitude, that can be easily managed by digital processors. The A/D conversion ideally can be divided in two steps, as shown in Fig. 53.1: the sampling process, which converts the continuous signal in a discrete-time series and whose elements are named samples, and a quantization procedure, which assigns the amplitude value of each sample within a set of determined discrete values. Both processes modify the characteristics of the signal, and their effects will be discussed in the following sections.

The Sampling Theorem The advantages of processing a digital series instead of an analog signal have been reported previously. Furthermore, the basic property when using a sampled series instead of its continuous waveform lies in the fact that the former, under certain hypotheses, is completely representative of the latter. When this happens, the continuous waveform can be perfectly reconstructed just from the series of sampled values. This is known as the sampling theorem (or Shannon theorem) [Shannon, 1949]. It states that a continuoustime signal can be completely recovered from its samples if, and only if, the sampling rate is greater than twice the signal bandwidth. In order to understand the assumptions of the theorem, let us consider a continuous band-limited signal x(t) (up to fb) whose Fourier transform X( f ) is shown in Fig. 53.2a and suppose to uniformly sample it. The sampling procedure can be modeled by the multiplication of x(t) with an impulse train


FIGURE 53.2 Effect of sampling frequency (fs) on a band-limited signal (up to frequency fb). Fourier transform of the original time signal (a), of the sampled signal when fs < 2fb (b), and when fs > 2fb (c). The dark areas in part b indicate the aliased frequencies.

( ) ∑ δ(t − kT )

it =

(53.1)

s

k = − ∞, ∞

where δ(t) is the delta (Dirac) function, k is an integer, and Ts is the sampling interval. The sampled signal becomes

( ) ( ) ( ) ∑ x(t ) ⋅ δ(t − kT )

xs t = x t ⋅ i t =

(53.2)

s

k = − ∞, ∞

Taking into account that multiplication in time domain implies convolution in frequency domain, we obtain

() () () ()

Xs f = X f ∗ I f = X f ∗

1 Ts

∑ δ( f − kf ) = T1 ∑ X ( f − kf ) s

k = − ∞, ∞

s

(53.3)

s k = − ∞, ∞

where fs = 1/Ts is the sampling frequency. Thus Xs( f ), i.e., the Fourier transform of the sampled signal, is periodic and consists of a series of identical repeats of X( f ) centered around multiples of the sampling frequency, as depicted in Fig. 53.2b,c. It is worth noting in Fig. 53.2b that the frequency components of X( f ) placed above fs /2 appears, when fs < 2fb , as folded back, summing up to the lower-frequency components. This phenomenon is known as aliasing (higher component look “alias” lower components). When aliasing occurs, the original information (Fig. 53.2a) cannot be recovered because the frequency components of the original signal are irreversibly corrupted by the overlaps of the shifted versions of X( f ).


FIGURE 53.3 Power spectrum of an EEG signal (originally bandlimited up to 40 Hz). The presence of 50-Hz mains noise (a) causes aliasing error in the 30-Hz component (i.e., in the β diagnostic band) in the sampled signal (b) if fs = 80 Hz.

A visual inspection of Fig. 53.2 allows one to observe that such frequency contamination can be avoided when the original signal is bandlimited [X( f ) = 0 for f > fb] and sampled at a frequency fs ≥ 2fb . In this case, shown in Fig. 53.2c, no overlaps exist between adjacent reply of X( f ), and the original waveform can be retrieved by low-pass filtering the sampled signal [Oppenheim & Schafer, 1975]. Such observations are the basis of the sampling theorem previously reported. The hypothesis of a bandlimited signal is hardly verified in practice, due to the signal characteristics or to the effect of superimposed wideband noise. It is worth noting that filtering before sampling is always needed even if we assume the incoming signal to be bandlimited. Let us consider the following example of an EEG signal whose frequency content of interest ranges between 0 and 40 Hz (the usual diagnostic bands are δ, 0 to 3.5 Hz; ϑ, 4 to 7 Hz, α, 8 to 13 Hz; β, 14 to 40 Hz). We may decide to sample it at 80 Hz, thus literally respecting the Shannon theorem. If we do it without prefiltering, we could find some unpleasant results. Typically, the 50-Hz mains noise will replicate itself in the signal band (30 Hz, i.e., the β band), thus corrupting irreversibly the information, which is of great interest from a physiologic and clinical point of view. The effect is shown in Fig. 53.3a (before sampling) and Fig. 53.3b (after sampling). Generally, it is advisable to sample at a frequency greater than 2fb [Gardenhire, 1964] in order to take into account the nonideal behavior of the filter or the other preprocessing devices. Therefore, the prefiltering block of Fig. 53.1 is always required to bandlimit the signal before sampling and to avoid aliasing errors.

The Quantization Effects The quantization produces a discrete signal, whose samples can assume only certain values according to the way they are coded. Typical step functions for a uniform quantizer are reported in Fig. 53.4a,b, where the quantization interval ∆ between two quantization levels is evidenced in two cases: rounding and truncation, respectively. Quantization is a heavily nonlinear procedure, but fortunately, its effects can be statistically modeled. Figure 53.4c,d shows it; the nonlinear quantization block is substituted by a statistical model in which the error induced by quantization is treated as an additive noise e(n) (quantization error) to the signal x(n). The following hypotheses are considered in order to deal with a simple mathematical problem: 1. e(n) is supposed to be a white noise with uniform distribution. 2. e(n) and x(n) are uncorrelated. First of all, it should be noted that the probability density of e(n) changes according to the adopted coding procedure. If we decide to round the real sample to the nearest quantization level, we have –∆/2 ≤ © 2000 by CRC Press LLC

FIGURE 53.4 Nonlinear relationships for rounding (a) and truncation (b) quantization procedures. Description of quantization block (c) by a statistical model (d) and probability densities for the quantization noise e(n) for rounding (e) and truncation (f ). ∆ is the quantization interval.

e(n) < ∆/2, while if we decide to truncate the sample amplitude, we have –∆ ≤ e(n) < 0. The two probability densities are plotted in Fig. 53.4e,f. The two ways of coding yield processes with different statistical properties. In the first case the mean and variance value of e(n) are

me = 0

σ e2 = ∆2 12

while in the second case me = –∆/2, and the variance is still the same. Variance reduces in the presence of a reduced quantization interval as expected. Finally, it is possible to evaluate the signal-to-noise ratio (SNR) for the quantization process:

 σ2  σ2  SNR = 10 log10  x2  = 10 log10  −2b x  = 6.02b + 10.79 + 10 log10 σ x2  σe   2 12 

( )

(53.4)

having set ∆ = 2–2b and where σ 2x is the variance of the signal and b is the number of bits used for coding. It should be noted that the SNR increases by almost 6 dB for each added bit of coding. Several forms of quantization are usually employed: uniform, nonuniform (preceding the uniform sampler with a nonlinear block), or roughly (small number of quantization levels and high quantization step). Details can be found in Carassa [1983], Jaeger [1982], and Widrow [1956].

53.2 Signal Processing A brief review of different signal-processing techniques will be given in this section. They include traditional filtering, averaging techniques, and spectral estimators. © 2000 by CRC Press LLC

FIGURE 53.5 General block diagram of a digital filter. The output digital signal y(n) is obtained from the input x(n) by means of a transformation T[·] which identifies the filter.

Only the main concepts of analysis and design of digital filters are presented, and a few examples are illustrated in the processing of the ECG signal. Averaging techniques will then be described briefly and their usefulness evidenced when noise and signal have similar frequency contents but different statistical properties; an example for evoked potentials enhancement from EEG background noise is illustrated. Finally, different spectral estimators will be considered and some applications shown in the analysis of RR fluctuations [i.e., the heart rate variability (HRV) signal].

Digital Filters A digital filter is a discrete-time system that operates some transformation on a digital input signal x(n) generating an output sequence y(n), as schematically shown by the block diagram in Fig. 53.5. The characteristics of transformation T[·] identify the filter. The filter will be time-variant if T[·] is a function of time or time-invariant otherwise, while is said to be linear if, and only if, having x1(n) and x2(n) as inputs producing y1(n) and y2(n), respectively, we have

[

] [ ] [ ]

T ax1 + bx2 = aT x1 + bT x2 = ay1 + by2 .

(53.5)

In the following, only linear, time-invariant filters will be considered, even if several interesting applications of nonlinear [Glaser & Ruchkin, 1976; Tompkins, 1993] or time-variant [Cohen, 1983; Huta & Webster, 1973; Thakor, 1987; Widrow et al., 1975] filters have been proposed in the literature for the analysis of biologic signals. The behavior of a filter is usually described in terms of input-output relationships. They are usually assessed by exciting the filter with different inputs and evaluating which is the response (output) of the system. In particular, if the input is the impulse sequence δ(n), the resulting output, the impulse response, has a relevant role in describing the characteristic of the filter. Such a response can be used to determine the response to more complicated input sequences. In fact, let us consider a generic input sequence x(n) as a sum of weighted and delayed impulses.

( ) ∑ x(k) ⋅ δ(n − k)

xn =

(53.6)

k = − ∞, ∞

and let us identify the response to δ(n – k) as h(n – k). If the filter is time-invariant, each delayed impulse will produce the same response, but time-shifted; due to the linearity property, such responses will be summed at the output:

( ) ∑ x(k) ⋅ h(n − k)

yn =

(53.7)

k = − ∞, ∞

This convolution product links input and output and defines the property of the filter. Two of them should be recalled: stability and causality. The former ensures that bounded (finite) inputs will produce


bounded outputs. Such a property can be deduced by the impulse response; it can be proved that the filter is stable if and only if

∑ h(k) < ∞

(53.8)

k = − ∞, ∞

Causality means that the filter will not respond to an input before the input is applied. This is in agreement with our physical concept of a system, but it is not strictly required for a digital filter that can be implemented in a noncausal form. A filter is causal if and only if

()

h k =0

for k < 0

Even if relation (53.7) completely describes the properties of the filter, most often it is necessary to express the input-output relationships of linear discrete-time systems under the form of the z-transform operator, which allows one to express relation (53.7) in a more useful, operative, and simpler form. The z-Transform The z-transform of a sequence x(n) is defined by [Rainer et al., 1972]

( ) ∑ x(k) ⋅ z

X z =

−k

(53.9)

k = − ∞, ∞

where z is a complex variable. This series will converge or diverge for different z values. The set of z values which makes Eq. (53.9) converge is the region of convergence, and it depends on the series x(n) considered. Among the properties of the z-transform, we recall • The delay (shift) property:

If

() (

w n = x n −T

)

then

() ()

W z = X z ⋅ z −T

• The product of convolution:

If

( ) ∑ x(k) ⋅ y(n − k)

wn =

then

() () ()

W z = X z ⋅Y z

k = − ∞, ∞

The Transfer Function in the z-Domain Thanks to the previous property, we can express Eq. (53.7) in the z-domain as a simple multiplication:

() () ()

Y z =H z ⋅X z

(53.10)

where H(z), known as transfer function of the filter, is the z-transform of the impulse response. H(z) plays a relevant role in the analysis and design of digital filters. The response to input sinusoids can be evaluated as follows: Assume a complex sinusoid x(n) = e jωnTS as input, the correspondent filter output will be

( ) ∑ h(k)e

yn =

k = 0, ∞


( ) = e − jωnTS

jωT2 n −k

∑ h(k)e

k = 0, ∞

− jωkTS

() ()

=x n ⋅H z

z = e jωTs

(53.11)

Then a sinusoid in input is still the same sinusoid at the output, but multiplied by a complex quantity H(ω). Such complex function defines the response of the filter for each sinusoid of ω pulse in input, and it is known as the frequency response of the filter. It is evaluated in the complex z plane by computing H(z) for z = e jωnTS, namely, on the point locus that describes the unitary circle on the z plane (e jωnTS = 1). As a complex function, H(ω) will be defined by its module H(ω) and by its phase ∠H(ω) functions, as shown in Fig. 53.6 for a moving average filter of order 5. The figure indicates that the lower-frequency components will come through the filter almost unaffected, while the higher-frequency components will be drastically reduced. It is usual to express the horizontal axis of frequency response from 0 to π. This is obtained because only pulse frequencies up to ωs /2 are reconstructable (due to the Shannon theorem), and therefore, in the horizontal axis, the value of ωTs is reported which goes from 0 to π. Furthermore, Fig. 53.6b demonstrates that the phase is piecewise linear, and in correspondence with the zeros of H(ω), there is a change in phase of π value. According to their frequency response, the filters are usually classified as (1) low-pass, (2) high-pass, (3) bandpass, or (4) bandstop filters. Figure 53.7 shows the ideal frequency response for such filters with the proper lowand high-frequency cutoffs. For a large class of linear, time-invariant systems, H(z) can be expressed in the following general form:

∑b z H (z ) = 1 + ∑a z

−m

m

m = 0, M

FIGURE 53.6 Modulus (a) and phase (b) diagrams of the frequency response of a moving average filter of order 5. Note that the frequency plots are depicted up to π. In fact, taking into account that we are dealing with a sampled signal whose frequency information is up to fs/2, we have ωmax = 2πfs/2 = πfs or ωmax = π if normalized with respect to the sampling rate.

(53.12)

−k

k

k =1, N

which describes in the z domain the following difference equation in this discrete time domain:

( ) ∑ a y(n − k) + ∑ b x(n − m)

y n =−

k

k =1, N

m

(53.13)

m = 0, M

When at least one of the ak coefficients is different from zero, some output values contribute to the current output. The filter contains some feedback, and it is said to be implemented in a recursive form. On the other hand, when the ak values are all zero, the filter output is obtained only from the current or previous inputs, and the filter is said to be implemented in a nonrecursive form. The transfer function can be expressed in a more useful form by finding the roots of both numerator and denominator:


FIGURE 53.7

Ideal frequency-response moduli for low-pass (a), high-pass (b), bandpass (c), and bandstop filters (d).

(

b0 z N − M Πm =1, M z − z m

()

H z =

(

Πk =1, N z − pk

)

)

(53.14)

where zm are the zeroes and pk are the poles. It is worth noting that H(z) presents N – M zeros in correspondence with the origin of the z plane and M zeroes elsewhere (N zeroes totally) and N poles. The pole-zero form of H(z) is of great interest because several properties of the filter are immediately available from the geometry of poles and zeroes in the complex z plane. In fact, it is possible to easily assess stability and by visual inspection to roughly estimate the frequency response without making any calculations. Stability is verified when all poles lie inside the unitary circle, as can be proved by considering the relationships between the z-transform and the Laplace s-transform and by observing that the left side of the s plane is mapped inside the unitary circle [Jackson, 1986; Oppenheim & Schafer, 1975]. The frequency response can be estimated by noting that (z – zm)z=ejωnTs is a vector joining the mth zero with the point on the unitary circle identified by the angle ωTs . Defining

( ) r A = (z − p )

r Bm = z − z m k

k

z = e jωTs

(53.15)

z = e jωTs

we obtain

r b0Πm =1, M Bm H ω = r Πk =1, N Ak

()

() ∑

∠H ω =

r ∠Bm −

m =1, M


∑

(

)

r ∠Ak + N − M ωTs .

k =1, N

(53.16)

FIGURE 53.8 Poles and zeroes geometry (a) and relative frequency response modulus (b) and phase (c) characteristics. Moving around the unitary circle a rough estimation of H(ω) and ∠H(ω) can be obtained. Note the zeroes’ effects at π and π/2 and modulus rising in proximity of the poles. Phase shifts are clearly evident in part c closer to zeros and poles.

Thus the modulus of H(ω) can be evaluated at any frequency ω° by computing the distances between poles and zeroes and the point on the unitary circle corresponding to ω = ω°, as evidenced by Fig. 53.8, where a filter with two pairs of complex poles and three zeros is considered. To obtain the estimate of H(ω), we move around the unitary circle and roughly evaluate the effect of poles and zeroes by keeping in mind a few rules [Challis & Kitney, 1982]: (1) when we are close to a © 2000 by CRC Press LLC

zero, H(ω) will approach zero, and a positive phase shift will appear in ∠H(ω) as the vector from the zero reverses its angle; (2) when we are close to a pole, H(ω) will tend to peak, and a negative phase change is found in ∠H(ω) (the closer the pole to unitary circle, the sharper is the peak until it reaches infinite and the filter becomes unstable); and (3) near a closer pole-zero pair, the response modulus will tend to zero or infinity if the zero or the pole is closer, while far from this pair, the modulus can be considered unitary. As an example, it is possible to compare the modulus and phase diagram of Fig. 53.8b,c with the relative geometry of the poles and zeros of Fig. 53.8a. FIR and IIR Filters A common way of classifying digital filters is based on the characteristics of their impulse response. For finite impulse response (FIR) filters, h(n) is composed of a finite number of nonzero values, while for infinite response (IIR) filters, h(n) oscillates up to infinity with nonzero values. It is clearly evident that in order to obtain an infinite response to an impulse in input, the IIR filter must contain some feedback that sustains the output as the input vanishes. The presence of feedback paths requires putting particular attention to the filter stability. Even if FIR filters are usually implemented in a nonrecursive form and IIR filters in a recursive form, the two ways of classification are not coincident. In fact, as shown by the following example, a FIR filter can be expressed in a recursive form

( ) ∑z

H z =

k = 0 , N −1

−k

=

∑

k = 0 , N −1

(1 − z ) = 1 − z (1 − z ) 1 − z −1

z

−k

−1

−N −1

(53.17)

for a more convenient computational implementation. As shown previously, two important requirements for filters are stability and linear phase response. FIR filters can be easily designed to fulfill such requirements; they are always stable (having no poles outside the origin), and the linear phase response is obtained by constraining the impulse response coefficients to have symmetry around their midpoint. Such constrain implies • bm = ±bM −m

(53.18)

where the bm are the M coefficients of an FIR filter. The sign + or – stays in accordance with the symmetry (even or odd) and M value (even or odd). This is a necessary and sufficient condition for FIR filters to have linear phase response. Two cases of impulse response that yield a linear phase filter are shown in Fig. 53.9. It should be noted that condition (53.18) imposes geometric constrains to the zero locus of H(z). Taking into account Eq. (53.12), we have

 1 z MH z = H   z *

()

(53.19)

Thus, both zm and 1/z*m must be zeros of H(z). Then the zeroes of linear phase FIR filters must lie on the unitary circle, or they must appear in pairs and with inverse moduli. Design Criteria In many cases, the filter is designed in order to satisfy some requirements, usually on the frequency response, which depend on the characteristic of the particular application the filter is intended for. It is known that ideal filters, like those reported in Fig. 53.7, are not physically realizable (they would require an infinite number of coefficients of impulse response); thus we can design FIR or IIR filters that can


FIGURE 53.9

Examples of impulse response for linear phase FIR filters: odd (a) and even (b) number of coefficients.

only mimic, with an acceptable error, the ideal response. Figure 53.10 shows a frequency response of a not ideal low-pass filter. Here, there are ripples in passband and in stopband, and there is a transition band from passband to stopband, defined by the interval ωs – ωp . Several design techniques are available, and some of them require heavy computational tasks, which are capable of developing filters with defined specific requirements. They include window technique, frequency-sampling method, or equiripple design for FIR filters. Butterworth, Chebychev, elliptical design, and impulse-invariant or bilinear transformation are instead employed for IIR filters. For detailed analysis of digital filter techniques, see Antoniou [1979], Cerutti [1983], and Oppenheim and Schafer [1975]. Examples A few examples of different kinds of filters will be presented in the following, showing some applications on ECG signal processing. It is shown that the ECG contains relevant information over a wide range of © 2000 by CRC Press LLC

FIGURE 53.10 Amplitude response for a real low-pass filter. Ripples are admitted in both passband and stopband, but they are constrained into restricted areas. Limitations are also imposed to the width of the transition band.

frequencies; the lower-frequency contents should be preserved for correct measurement of the slow ST displacements, while higher-frequency contents are needed to correctly estimate amplitude and duration of the faster contributions, mainly at the level of the QRS complex. Unfortunately, several sources of noise are present in the same frequency band, such as, higher-frequency noise due to muscle contraction (EMG noise), the lower-frequency noise due to motion artifacts (baseline wandering), the effect of respiration or the low-frequency noise in the skin-electrode interface, and others. In the first example, the effect of two different low-pass filters will be considered. An ECG signal corrupted by an EMG noise (Fig. 53.11a) is low-pass filtered by two different low-pass filters whose frequency responses are shown in Fig. 53.11b,c. The two FIR filters have cutoff frequencies at 40 and 20 Hz, respectively, and were designed through window techniques (Weber-Cappellini window, filter length = 256 points) [Cappellini et al., 1978]. The output signals are shown in Fig. 53.11d,e. Filtering drastically reduces the superimposed noise but at the same time alters the original ECG waveform. In particular, the R wave amplitude is progressively reduced by decreasing the cutoff frequency, and the QRS width is progressively increased as well. On the other hand, P waves appear almost unaffected, having frequency components generally lower than 20 to 30 Hz. At this point, it is worth noting that an increase in QRS duration is generally associated with various pathologies, such as ventricular hypertrophy or bundle-branch block. It is therefore necessary to check that an excessive band limitation does not introduce a false-positive indication in the diagnosis of the ECG signal. An example of an application for stopband filters (notch filters) is presented in Fig. 53.12. it is used to reduce the 50-Hz mains noise on the ECG signal, and it was designated by placing a zero in correspondence of the frequency we want to suppress. Finally, an example of a high-pass filter is shown for the detection of the QRS complex. Detecting the time occurrence of a fiducial point in the QRS complex is indeed the first task usually performed in ECG signal analysis. The QRS complex usually contains the higher-frequency components with respect to the other ECG waves, and thus such components will be enhanced by a high-pass filter. Figure 53.13 shows how QRS complexes (Fig. 53.13a) can be identified by a derivative high-pass filter with a cutoff frequency to decrease the effect of the noise contributions at high frequencies (Fig. 53.13b). The filtered signal (Fig. 53.13c) presents sharp and well-defined peaks that are easily recognized by a threshold value.

Signal Averaging Traditional filtering performs very well when the frequency content of signal and noise do not overlap. When the noise bandwidth is completely separated from the signal bandwidth, the noise can be decreased easily by means of a linear filter according to the procedures described earlier. On the other hand, when the signal and noise bandwidth overlap and the noise amplitude is enough to seriously corrupt the signal, a traditional filter, designed to cancel the noise, also will introduce signal cancellation or, at least, distortion. As an example, let us consider the brain potentials evoked by a sensory stimulation (visual, acoustic, or somatosensory) generally called evoked potentials (EP). Such a response is very difficult to © 2000 by CRC Press LLC


FIGURE 53.11 Effects of two different low-pass filters (b) and (c) on an ECG trace (a) corrupted by EMG noise. Both amplitude reduction and variation in the QRS with induced by too drastic lowpass filtering are evidenced.


FIGURE 53.12

A 50-Hz noisy ECG signal (a); a 50-Hz rejection filter (b); a filtered signal (c).


FIGURE 53.13

Effect of a derivative high-pass filter (b) on an ECG lead (a). (c) The output of the filter.

determine because its amplitude is generally much lower than the background EEG activity. Both EP and EEG signals contain information in the same frequency range; thus the problem of separating the desired response cannot be approached via traditional digital filtering [Aunon et al., 1981]. Another typical example is in the detection of ventricular late potentials (VLP) in the ECG signal. These potentials are very small in amplitude and are comparable with the noise superimposed on the signal and also for what concerns the frequency content [Simson, 1981]. In such cases, an increase in the SNR may be achieved on the basis of different statistical properties of signal and noise. When the desired signal repeats identically at each iteration (i.e., the EP at each sensory stimulus, the VLP at each cardiac cycle), the averaging technique can satisfactorily solve the problem of separating signal from noise. This technique sums a set of temporal epochs of the signal together with the superimposed noise. If the time epochs are properly aligned, through efficient trigger-point recognition, the signal waveforms directly sum together. If the signal and the noise are characterized by the following statistical properties: 1. All the signal epochs contain a deterministic signal component x(n) that does not vary for all the epochs. 2. The superimposed noise w(n) is a broadband stationary process with zero mean and variance σ2 so that

[ ( )] E[w (n)] = σ E w n =0 2

(53.20) 2

3. Signal x(n) and noise w(n) are uncorrelated so that the recorded signal y(n) at the ith iteration can be expressed as

() () ()

y n = x n + wi n ,

(53.21)

i

Then the averaging process yields yt :

()

yt n =

1 N

N

∑ i =1

( ) ∑ w (n) N

yi = x n +

i

(53.22)

i =1

The noise term is an estimate of the mean by taking the average of N realizations. Such an average is a new random variable that has the same mean of the sum terms (zero in this case) and which has variance of σ 2/N. The effect of the coherent averaging procedure is then to maintain the amplitude of the signal and reduce the variance of the noise by a factor of N. In order to evaluate the improvement in the SNR (in rms value) in respect to the SNR (at the generic ith sweep):

SNR = SNR i ⋅ N

(53.23)

Thus signal averaging improves the SNR by a factor of N in rms value. A coherent averaging procedure can be viewed as a digital filtering process, and its frequency characteristics can be investigated. From expression (53.17) through the z-transform, the transfer function of the filtering operation results in

()

H z =


1 + z − h + z −2h + L + z N

(

)

− N −1 h

(53.24)

where N is the number of elements in the average, and h is the number of samples in each response. An alternative expression for H(z) is

()

H z =

1 1 − z − Nh N 1− zh

(53.25)

This is a moving average low-pass filter as discussed earlier, where the output is a function of the preceding value with a lag of h samples; in practice, the filter operates not on the time sequence but in the sweep sequence on corresponding samples. The frequency response of the filter is shown in Fig. 53.14 for different values of the parameter N. In this case, the sampling frequency fs is the repetition frequency of the sweeps, and we may assume it to be 1 without loss of generality. The frequency response is characterized by a main lobe with the first zero corresponding to f = 1/N and by successive secondary lobes separated by zeroes at intervals 1/N. The width of each tooth decreases as well as the amplitude of the secondary lobes when increasing the number N of sweeps. The desired signal is sweep-invariant, and it will be unaffected by the filter, while the broadband noise will be decreased. Some leakage of noise energy takes place in the center of the sidelobes and, of course, at zero frequency. Under the hypothesis of zero mean noise, the dc component has no effect, and the

FIGURE 53.14

Equivalent frequency response for the signal-averaging procedure for different values of N (see text).


diminishing sidelobe amplitude implies the leakage to be not relevant for high frequencies. It is important to recall that the average filtering is based on the hypothesis of broadband distribution of the noise and lack of correlation between signal and noise. Unfortunately, these assumptions are not always verified in biologic signals. For example, the assumptions of independence of the background EEG and the evoked potential may be not completely realistic [Gevins & Remond, 1987]. In addition, much attention must be paid to the alignment of the sweeps; in fact, slight misalignments (fiducial point jitter) will lead to a low-pass filtering effect of the final result. Example As mentioned previously, one of the fields in which signal-averaging technique is employed extensively is in the evaluation of cerebral evoked response after a sensory stimulation. Figure 53.15a shows the EEG recorded from the scalp of a normal subject after a somatosensory stimulation released at time t = 0. The evoked potential (N = 1) is not visible because it is buried in the background EEG (upper panel). In the successive panels there is the same evoked potential after averaging different numbers of sweeps corresponding to the frequency responses shown in Fig. 53.14. As N increases, the SNR is improved by a factor N (in rms value), and the morphology of the evoked potential becomes more recognizable while the EEG contribution is markedly diminished. In this way it is easy to evaluate the quantitative indices of clinical interest, such as the amplitude and the latency of the relevant waves.

Spectral Analysis The various methods to estimate the power spectrum density (PSD) of a signal may be classified as nonparametric and parametric. Nonparametric Estimators of PSD This is a traditional method of frequency analysis based on the Fourier transform that can be evaluated easily through the fast Fourier transform (FFT) algorithm [Marple, 1987]. The expression of the PSD as a function of the frequency P(f ) can be obtained directly from the time series y(n) by using the periodogram expression

()

1 P f = Ts Ts

2

N −1

∑ ()

y k e − j 2 πfkTs =

k =0

()

1 Y f NTs

2

(53.26)

where Ts is the sampling period, N is the number of samples, and Y(f ) is the discrete time Fourier transform of y(n). On the basis of the Wiener-Khintchin theorem, PSD is also obtainable in two steps from the FFT of the autocorrelation function Ryy(k) of the signal, where Ryy(k) is estimated by means of the following expression:

()

1 Rˆ yy k = N

N −k −1

∑ y(i) y * (i + k)

(53.27)

i=0

where * denotes the complex conjugate. Thus the PSD is expressed as

()

P f = Ts ⋅

∑ Rˆ (k)e N

yy

− j 2 πfkTs

k = −N

based on the available lag estimates Rˆ yy(k), where –(1/2Ts) ≤ f ≤ (1/2Ts). © 2000 by CRC Press LLC

(53.28)

FIGURE 53.15 Enhancement of evoked potential (EP) by means of averaging technique. The EEG noise is progressively reduced, and the EP morphology becomes more recognizable as the number of averaged sweeps (N) is increased.

FFT-based methods are widely diffused, for their easy applicability, computational speed, and direct interpretation of the results. Quantitative parameters are obtained by evaluating the power contribution at different frequency bands. This is achieved by dividing the frequency axis in ranges of interest and by integrating the PSD on such intervals. The area under this portion of the spectrum is the fraction of the total signal variance due to the specific frequencies. However, autocorrelation function and Fourier transform are theoretically defined on infinite data sequences. Thus errors are introduced by the need to operate on finite data records in order to obtain estimators of the true functions. In addition, for the finite data set it is necessary to make assumptions, sometimes not realistic, about the data outside the recording window; commonly they are considered to be zero. This implicit rectangular windowing of © 2000 by CRC Press LLC

the data results in a special leakage in the PSD. Different windows that smoothly connect the side samples to zero are most often used in order to solve this problem, even if they may introduce a reduction in the frequency resolution [Harris, 1978]. Furthermore, the estimators of the signal PSD are not statistically consistent, and various techniques are needed to improve their statistical performances. Various methods are mentioned in the literature; the methods of Dariell [1946], Bartlett [1948], and Welch [1970] are the most diffused ones. Of course, all these procedures cause a further reduction in frequency resolution. Parametric Estimators Parametric approaches assume the time series under analysis to be the output of a given mathematical model, and no drastic assumptions are made about the data outside the recording window. The PSD is calculated as a function of the model parameters according to appropriate expressions. A critical point in this approach is the choice of an adequate model to represent the data sequence. The model is completely independent of the physiologic, anatomic, and physical characteristics of the biologic system but provides simply the input-output relationships of the process in the so-called black-box approach. Among the numerous possibilities of modeling, linear models, characterized by a rational transfer function, are able to describe a wide number of different processes. In the most general case, they are represented by the following linear equation that relates the input-driving signal w(k) and the output of an autoregressive moving average (ARMA) process: p

q

( ) ∑ a y(k − i) + ∑ b w(k − j) + w(k)

y k =−

i

(53.29)

j

i =1

j =1

where w(k) is the input white noise with zero mean value and variance λ2, p and q are the orders of AR and MA parts, respectively, and ai and bj are the proper coefficients. The ARMA model may be reformulated as an AR or an MA if the coefficients bj or ai are, respectively, set to zero. Since the estimation of the AR parameters results in liner equations, AR models are usually employed in place of ARMA or MA models, also on the basis of the Wold decomposition theorem [Marple, 1987] that establishes that any stationary ARMA or MA process of finite variance can be represented as a unique AR model of appropriate order, even infinite; likewise, any ARMA or AR process can be represented by an MA model of sufficiently high order. The AR PPSD is then obtained from the following expression:

()

λ2Ts

P f =

2

p

1+

∑a z

∏ (z − z ) p

−i

(53.30)

2

l

i

i =1

λ2Ts

=

(

z = exp j 2 πfTs

)

l =1

(

z = exp j 2 πfTs

)

The right side of the relation puts into evidence the poles of the transfer function that can be plotted in the z-transform plane. Figure 53.16b shows the PSD function of the HRV signal depicted in Fig. 53.16a, while Fig. 53.16c displays the corresponding pole diagram obtained according to the procedure described in the preceding section. Parametric methods are methodologically and computationally more complex than the nonparametric ones, since they require an a priori choice of the structure and of the order of the model of the signalgeneration mechanism. Some tests are required a posteriori to verify the whiteness of the prediction error, such as the Anderson test (autocorrelation test) [Box & Jenkins, 1976] in order to test the reliability of the estimation. Postprocessing of the spectra can be performed as well as for nonparametric approaches by integrating the P(f ) function in predefined frequency ranges; however, the AR modeling has the advantage of allowing a spectral decomposition for a direct and automatic calculation of the power and


FIGURE 53.16 (a) Interval tachogram obtained from an ECG recording as the sequence of the RR time intervals expressed in seconds as a function of the beat number. (b) PSD of the signal (a) evaluated by means of an AR model (see text). (c) Pole diagram of the PSD shown in (b).

frequency of each spectral component. In the z-transform domain, the autocorrelation function (ACF) R(k) and the P(z) of the signal are related by the following expression:

()

Rk =

1 2πj

∫ P ( z )z z =1

k −1

dz

(53.31)

If the integral is calculated by means of the residual method, the ACF is decomposed into a sum of dumped sinusoids, each one related to a pair of complex conjugate poles, and of dumped exponential functions, related to the real poles [Zetterberg, 1969]. The Fourier transform of each one of these terms gives the expression of each spectral component that fits the component related to the relevant pole or pole pair. The argument of the pole gives the central frequency of the component, while the ith spectral component power is the residual γi in the case of real poles and 2Re(γi ) in case of conjugate pole pairs. γi is computed from the following expression:

(

)()

γ i = z −1 z − z i P z

z = zi

(53.32)

It is advisable to point out the basic characteristics of the two approaches that have been described above: the nonparametric and the parametric. The latter (parametric) has evident advantages with respect to the former, which can be summarized in the following:


• It has a more statistical consistency even on short segments of data; i.e., under certain assumptions, a spectrum estimated through autoregressive modeling is a maximum entropy spectrum (MES). • The spectrum is more easily interpretable with an “implicit” filtering of what is considered random noise. • An easy and more reliable calculation of the spectral parameters (postprocessing of the spectrum), through the spectral decomposition procedure, is possible. Such parameters are directly interpretable from a physiologic point of view. • There is no need to window the data in order to decrease the spectral leakage. • The frequency resolution does not depend on the number of data. On the other hand, the parametric approach • Is more complex from a methodologic and computational point of view. • Requires an a priori definition of the kind of model (AR, MA, ARMA, or other) to be fitted and mainly its complexity defined (i.e., the number of parameters). Some figures of merit introduced in the literature may be of help in determining their value [Akaike, 1974]. Still, this procedure may be difficult in some cases. Example As an example, let us consider the frequency analysis of the heart rate variability (HRV) signal. In Fig. 53.16a, the time sequence of the RR intervals obtained from an ECG recording is shown. The RR intervals are expressed in seconds as a function of the beat number in the so-called interval tachogram. It is worth noting that the RR series is not constant but is characterized by oscillations of up to the 10% of its mean value. These oscillations are not causal but are the effect of the action of the autonomic nervous system in controlling heart rate. In particular, the frequency analysis of such a signal (Fig. 53.16b shows the PSD obtained by mean of an AR model) has evidenced three principal contributions in the overall variability of the HRV signal. A very low frequency (VLF) component is due to the long-term regulation mechanisms that cannot be resolved by analyzing a few minutes of signal (3 to 5 minutes are generally studied in the traditional spectral analysis of the HRV signal). Other techniques are needed for a complete understanding of such mechanisms. The low-frequency (LF) component is centered around 0.1 Hz, in a range between 0.03 and 0.15 Hz. An increase in its power has always been observed in relation to sympathetic activations. Finally, the high-frequency (HF) component, in synchrony with the respiration rate, is due to the respiration activity mediated by the vagus nerve; thus it can be a marker of vagal activity. In particular, LF and HF power, both in absolute and in normalized units (i.e., as percentage value on the total power without the VLF contribution), and their ratio LF/HF are quantitative indices widely employed for the quantification of the sympathovagal balance in controlling heart rate [Malliani et al., 1991].

53.3 Conclusion The basic aspects of signal acquisition and processing have been illustrated, intended as fundamental tools for the treatment of biologic signals. A few examples also were reported relative to the ECG signal, as well as EEG signals and EPs. Particular processing algorithms have been described that use digital filtering techniques, coherent averaging, and power spectrum analysis as reference examples on how traditional or innovative techniques of digital signal processing may impact the phase of informative parameter extraction from biologic signals. They may improve the knowledge of many physiologic systems as well as help clinicians in dealing with new quantitative parameters that could better discriminate between normal and pathologic cases.


Defining Terms Aliasing: Phenomenon that takes place when, in A/D conversion, the sampling frequency fs is lower than twice the frequency content fb of the signal; frequency components above fs /2 are folded back and are summed to the lower-frequency components, distorting the signal. Averaging: Filtering technique based on the summation of N stationary waveforms buried in casual broadband noise. The SNR is improved by a factor of N . Frequency response: A complex quantity that, multiplied by a sinusoid input of a linear filter, gives the output sinusoid. It completely characterizes the filter and is the Fourier transform of the impulse response. Impulse reaction: Output of a digital filter when the input is the impulse sequence δ(n). It completely characterizes linear filters and is used for evaluating the output corresponding to different kinds of inputs. Notch filter: A stopband filter whose stopped band is very sharp and narrow. Parametric methods: Spectral estimation methods based on the identification of a signal generating model. The power spectral density is a function of the model parameters. Quantization error: Error added to the signal, during the A/D procedure, due to the fact that the analog signal is represented by a digital signal that can assume only a limited and predefined set of values. Region of convergence: In the z-transform plane, the ensemble containing the z-complex points that makes a series converge to a finite value.

References Akaike H. 1974. A new look at the statistical model identification. IEEE Trans Autom Contr (AC-19):716. Antoniou A. 1979. Digital Filters: Analysis and Design. New York: McGraw-Hill. Aunon JL, McGillim CD, Childers DG: 1981. Signal processing in evoked potential research: Averaging and modeling. CRC Crit Rev Bioing 5:323. Bartlett MS. 1948. Smoothing priodograms from time series with continuous spectra. Nature 61:686. Box GEP, Jenkins GM. 1976. Time Series Analysis: Forecasting and Control. San Francisco, Holden-Day. Cappellini V, Constantinides AG, Emiliani P. 1978. Digital Filters and Their Applications. London, Academic Press. Carassa F. 1983. Comunicazioni Elettriche. Torino, Boringhieri. Cerutti S. 1983. Filtri numerici per l’eleborazione di segnali biologici. Milano, CLUP. Challis RE, Kitney RI. 1982. The design of digital filters for biomedical signal processing: 1. Basic concepts. J Biomed Eng 5:267. Cobbold RSC. 1988. Transducers for Biomedical Measurements. New York, Wiley. Cohen A. 1983. Biomedical Signal Processing: Time and Frequency Domains Analysis. Boca Raton, Fla, CRC Press. Dariell PJ. 1946. On the theoretical specification and sampling properties of autocorrelated time-series (discussion). JR Stat Soc 8:88. Gardenhire LW. 1964. Selecting sample rate. ISA J 4:59. Gevins AS, Remond A (eds). 1987. Handbook of Electrophysiology and Clinical Neurophysiology. Amsterdam, Elsevier. Glaser EM, Ruchkin DS. 1976. Principles of Neurophysiological Signal Processing. New York, Academic Press. Harris FJ. 1978. On the use of windows for harmonic analysis with the discrete Fourier transform. Proc IEEE 64(1):51. Huta K, Webster JG. 1973. 60-Hz interference in electrocardiography. IEEE Trans Biomed Eng 20(2):91. Jackson LB. 1986. Digital Signal Processing. Hingham, Mass, Kluer Academic.


Jaeger RC. 1982. Tutorial: Analog data acquisition technology: II. Analog to digital conversion. IEEE Micro 8:46. Malliani A, Pagani M, Lombardi F, Cerutti S. 1991. Cardiovascular neural regulation explored in the frequency domain. Circulation 84:482. Marple SL. 1987. Digital Spectral Analysis with Applications. Englewood Cliffs, NJ, Prentice-Hall. Oppenheim AV, Schafer RW. 1975. Digital Signal Processing. Englewood Cliffs, NJ, Prentice-Hall. Rainer LR, Cooley JW, Helms HD, et al: 1972. Terminology in digital signal processing. IEEE Trans Audio Electroac AU-20:322. Shannon CE. 1949. Communication in presence of noise. Proc IRE 37:10. Simson MB. 1981. Use of signals in the terminal QRS complex to identify patients with ventricular tachycardia after myocardial infarction. Circulation 64:235. Thakor NV. 1987. Adaptive filtering of evoked potential. IEEE Trans Biomed Eng 34:1706. Tompkins WJ (ed). 1993. Biomedical Digital Signal Processing. Englewood Cliffs, NJ, Prentice-Hall. Tompkins WJ, Webster JG (eds). 1981. Design of Microcomputer-Based Medical Instrumentation. Englewood Cliffs, NJ, Prentice-Hall. Webster JG (ed). 1992. Medical Instrumentation, 2d ed. Boston, Houghton-Mufflin. Welch DP. 1970. The use of fast Fourier transform for the estimation of power spectra: A method based on time averaging over short modified periodograms. IEEE Trans Acoust AU-15:70. Widrow B. 1956. A study of rough amplitude quantization by means of Nyquist sampling theory. IRE Trans Cric Theory 3:266. Widrow B, Glover JRJ, Kaunitz J, et al; 1975. Adaptive noise cancelling: Principles and applications. Proc IEEE 63(12):1692. Zetterberg LH. 1969. Estimation of parameters for a linear difference equation with application to EEG analysis. Math Biosci 5:227.

Further Information A book that provides a general overview of basic concepts in biomedical signal processing is Digital Biosignal Processing, by Rolf Weitkunat (ed) (Elsevier Science Publishers, Amsterdam, 1991). Contributions by different authors provide descriptions of several processing techniques and many applicative examples on biologic signal analysis. A deeper and more specific insight of actual knowledge and future perspectives on ECG analysis can be found in Electrocardiography: Past and Future, by Philippe Coumel and Oscar B. Garfein (eds) (Annals of the New York Academy Press, vol 601, 1990). Advances in signal processing are monthly published in the journal IEEE Transactions on Signal Processing, while the IEEE Transaction on Biomedical Engineering provides examples of applications in biomedical engineering fields.


Cetin, A. E., Köymen, H. “Compression of Digital Biomedical Signals.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

54 Compression of Digital Biomedical Signals 54.1 54.2

A. Enis Cetin ¸ Bilkent University

Hayrettin Köymen Bilkent University

Introduction Time-Domain Coding of Biomedical Signals Data Compression by DPCM • AZTEC ECG Compression Method • Turning Point ECG Compression Method • ECG Compression via Parameter Extraction

54.3 54.4 54.5

Frequency-Domain Data Compression Methods Wavelet or Subband Coding Hybrid Multichannel ECG Coding Preprocessor • Linear Transformer

54.1 Introduction Computerized electrocardiogram (ECG), electroencephalogram (EEG), and magnetoencephalogram (MEG) processing systems have been widely used in clinical practice [1] and they are capable of recording and processing long records of biomedical signals. The use of such systems (1) enables the construction of large signal databases for subsequent evaluation and comparison, (2) makes the transmission of biomedical information feasible over telecommunication networks in real time or off line, and (3) increases the capabilities of ambulatory recording systems such as the Holter recorders for ECG signals. In spite of the great advances in VLSI memory technology, the amount of data generated by digital systems may become excessive quickly. For example, a Holter recorder needs more than 200 Mbits/day of memory space to store a dual-channel ECG signal sampled at a rate of 200 samples/sec. with 10 bit/sample resolution. Since the recorded data samples are correlated with each other, there is an inherent redundancy in most biomedical signals. This can be exploited by the use of data compression techniques which have been successfully utilized in speech, image, and video signals [2] as well. The aim of any biomedical signal compression scheme is to minimize the storage space without losing any clinically significant information, which can be achieved by eliminating redundancies in the signal, in a reasonable manner. Data compression methods can be classified into two categories: (1) lossless and (2) lossy coding methods. In lossless data compression, the signal samples are considered to be realizations of a random variable or a random process and the entropy of the source signal determines the lowest compression ratio that can be achieved. In lossless coding the original signal can be perfectly reconstructed. For typical biomedical signals lossless (reversible) compression methods can only achieve Compression Ratios1 (CR) in the order of 2 to 1. On the other hand lossy (irreversible) techniques may produce CR results in the order of 10 to 1. In lossy methods, there is some kind of quantization of the input data which leads to 1CR is defined as the ration of the total number of bits used to represent the digital signal before and after compression.


higher CR results at the expense of reversibility. But this may be acceptable as long as no clinically significant degradation is introduced to the encoded signal. The CR levels of 2 to 1 are too low for most practical applications. Therefore, lossy coding methods which introduce small reconstruction errors are preferred in practice. In this section we review the lossy biomedical data compression methods.

54.2 Time-Domain Coding of Biomedical Signals Biomedical signals can be compressed in time domain, frequency domain, or time-frequency domain. In this subsection the time domain techniques, which are the earlier approaches to biomedical signal compression, are reviewed.

Data Compression by DPCM Differential pulse code modulation (DPCM) is a well-known data coding technique in which the main idea is to decorrelate the input signal samples by linear prediction. The current signal sample, x(n), is estimated from the past samples by using either a fixed or adaptive linear predictor

( ) ∑ a x(n − k) N

xˆ n =

(54.1)

k

k =1

where xˆ (n) is the estimate of x(n) at discrete time instant n, and {ak} is the predictor weight.2 The samples of the estimation error sequence, e(n) = x(n) – xˆ (n) are less correlated with each other compared to the original signal, x(n) as the predictor removes the unnecessary information which is the predictable portion of the sample x(n). In a typical DPCM encoder, the error sequence is quantized by using a nonuniform quantizer and quantizer outputs are entropy coded by assigning variable-length codewords to the quantized error sequence according to the frequency of occurrence. The variable length codebook is constructed by Huffman coding which assigns shorter (longer) codewords to values occurring with higher (lower) probabilities. Huffman coding produces compression results that are arbitrarily close to the entropy of the quantized error sequence. A CR of 7.8 was reported for an ECG signal recorded at a rate of 500 Hz with 8 bit/sample resolution [6]. This means that about 1 bit/sample is used to represent the ECG signal. The corresponding Percent Root Mean Square Difference (PRD) was 3.5. The PRD is a measure of reconstruction error and it is defined as follows:

∑ [x(n) − x (n)] ∑ x (n) N −1

PRD =

rec

n=0

N −1

2

∗100

(54.2)

2

n=0

where N is the total number of samples in the ECG signal, x(n), and xrec(n) is the reconstructed ECG signal. A typical DPCM encoder may become a lossless coder if the quantization step is skipped. In this case CR drops drastically to low values.

AZTEC ECG Compression Method The Amplitude Zone Time Epoch Coding (AZTEC) is one of the earliest ECG coding methods. It was developed by Cox et al., [12] as a preprocessing software for real-time monitoring of ECGs. It was 2In practical DPCM implementation, instead of x(n – k), quantized past samples of the signal are used in the linear predictor [2].


FIGURE 54.1

AZTEC representation of an ECG waveform.

observed to be useful for automatic analysis of ECGs such as QRS detection, but it is inadequate for visual presentation of the ECG signal as the reconstructed signal has a staircase appearance. In this method the ECG signal is considered to consist of flat regions and “slopes”. If the signal value stays within a predetermined constant for more than three consecutive samples then that region is assumed to be constant and stored by its duration (number of samples) and a fixed amplitude value. Otherwise the signal is assumed to have a slope. A slope is stored by its duration and the amplitude of the last sample point. Linear interpolation is used to reconstruct the ECG signal. As a result, the resulting signal has a discontinuous nature as shown in Fig. 54.1. Even through AZTEC produces a high compression ratio (CR = 10 to 1, for 500 Hz sampled data with 12 bit/sample resolution) the quality of the reconstructed signal is very low and it is not acceptable to the cardiologists. Various modified of AZTEC are proposed [3]. One notable example is the CORTES technique [13]. The CORTES technique is a hybrid of the AZTEC and the Turning Point method which is described in the next subsection.

Turning Point ECG Compression Method The Turning Point data reduction method [14] is basically an adaptive downsampling method developed especially for ECGs. It reduces the sampling frequency of an ECG signal by a factor of two. The method is based on the trends of the signal samples. Three input samples are processed at a time. Let x(n) be the current sample at discrete-time instant n. Among the two consecutive input samples, x(n + 1) and x(n + 2), the one producing the highest slope (in magnitude) is retained and the other sample is dropped. In this way the overall sampling rate is reduced to one-half of the original sampling rate. no other coding is carried out. Therefore, the resulting CR is 2 to 1. A PRD of 5.3% is reported for an ECG signal sampled at 200 Hz with 12 bit/sample resolution [14]. In practice the CR value actually may be lower than 2 as the retained samples may not be equally spaced and some extra bits may be needed for timing determination.

ECG Compression via Parameter Extraction In these methods, the signal is analyzed and some important features such as typical cycles, extreme locations, etc. are determined. These features are properly stored. Reconstruction is carried out by using appropriate interpolation schemes. The location of the extrema or peaks in an ECG signal is important in diagnosis because they basically determine the shape of each ECG period. The ECG compression techniques described in [15] take advantage of this feature and record only the maxima, minima, slope changes, zero-crossing intervals, etc. of the signal. During reconstruction various interpolation schemes such as polynomial fitting and © 2000 by CRC Press LLC

spline functions are used. The performance of the extrema based methods is compared to the AZTEC method and it was reported that for a given CR the RMS error is half of that of AZTEC method [15]. Other parameter extraction methods include [4, 11] where ECG signals are analyzed in a cyclesynchronous manner. An AR model is fitted to the MEG signal and parameters of the AR model are stored [16]. An ECG signal is modeled by splines and the data is compressed by storing the spline parameters [19]. Review of some other time-domain ECG data compression methods can be found in [3].

54.3 Frequency-Domain Data Compression Methods Transform Coding (TC) is the most important frequency-domain digital waveform compression method [2]. The key is to divide the signal into frequency components and judiciously allocate bits in the frequency domain. In most TC methods, the input signal is first divided into blocks of data and each block is linearly transformed into the “frequency” domain. Let x = [x0 x1…xN–1]T be a vector obtained from a block of N input samples. The transform domain coefficients, vi , i = 0, 1, …, N – 1, are given by

v = Ax

(54.3)

where v = [v0v1… vN – 1]T, and A is the N × N transform matrix representing the linear transform. A variety of transform matrices including the discrete Fourier transform matrix is used in digital waveform coding [2]. In this subsection discrete Karhunen-Loeve Transform (KLT) and Discrete Cosine Transform, which are the most used ones, are reviewed. If the input signal is a wide sense stationary random process then the so-called optimum linear transform, KLT is well defined and it decorrelates the entries of the input vector. This is equivalent to removing all the unnecessary information contained in the vector x. Therefore, by coding the entries of the v vector, only the useful information is retained. The entries of the v vector are quantized and stored. Usually, different quantizers are used to quantize the entries of the v vector. In general, more bits are allocated to those entries which have high energies compared to the ones with low energies. The KLT is constructed from the eigenvectors of the autocovariance matrix of the input vector x. In most practical waveforms the statistics of the input vector change from block to block. Thus for each block a new KLT matrix must be constructed. This is computationally very expensive (in some practical cases a fixed KLT matrix is estimated and it is assumed to be constant for a reasonable amount of duration). Furthermore, there is no fast algorithm similar to the Fast Fourier Transform (FFT) to compute the KLT. The Discrete Cosine Transform (DCT) [2] was developed by Ahmed et. al., to approximate KLT when there is high correlation among the input samples, which is the case in many digital waveforms including speech, music, and biomedical signals. The DCT v = [v0 v1 …vN–1]T of the vector x is defined as follows

v0 =

vk =

2 N

N −1

∑ x cos n

n=0

1 N

N −1

∑x ,

(2n + 1)kπ 2N

(54.4)

n

n=0

(

)

k = 1, 2, …, N − 1

(54.5)

where vk is the kth DCT coefficient. The inverse discrete cosine transform (IDCT) of v is given as follows © 2000 by CRC Press LLC

xn =

1 N

v0 +

2 N

N −1

∑ v cos k

k =1

(2n + 1)kπ 2N

(

)

n = 0, 1, 2, …, N − 1

(54.6)

There exist fast algorithms, Order (N log N), to compute the DCT [2[. Thus, DCT can be implemented in a computationally efficient manner. Two recent image and video coding standards, JPEG and MPEG, use DCT as the main building block. A CR of 3 to 1 for single channel ECG is reported by using DCT and KLT based coders [8]. For multilead systems, two-dimensional (2-D) transform based methods can be used. A CR of 12 to 1 was reported [17] for a three-lead system by using a 2-D KLT. Recently, Philips [18] developed a new transform by using time-warped polynomials and obtained a CR of 26 to 9. In this case a DCT based coding procedure produced a CR of 24 to 3. Since signal recording conditions and noise levels vary from study to study, a thorough comparison of the coding methods is very difficult to make. But, frequency-domain coding methods produce higher coding results than time-domain coding methods.

54.4 Wavelet or Subband Coding Wavelet Transform (WT) is a powerful time-frequency signal analysis tool and it is used in a wide variety of applications including signal and image coding [20]. WT and Subband Coding (SBC) are closely related to each other. In fact the fast implementation of WTs are carried out using Subband (SB) filter banks. Due to this reason WT based waveform coding methods are essentially similar to the SBC based methods. In an SBC structure the basic building block is a digital filter bank which consists of a lowpass and a highpass filter. In the ideal case, the passband (stopband) of the lowpass filter is [0, π/2] [π/2, π]. Let Hl (ej ω)Hu (ej ω) be the frequency response of the lowpass (highpass) filter. In SBC the input signal which is sampled at the rate of fs is filtered by Hl (ej ω) and Hu (ej ω) and the filter outputs are downsampled by a factor of two (every other sample is dropped) in order to maintain the overall sampling rate to be equal to the original rate. In this way two subsignals, xι(n) and xu(n) which both have a sampling rate of fs /2 are obtained. The block diagram of a two-level SBC structure is shown in Fig. 54.1. The subsignal, xι(n) (xu(n)), contains the lowpass (highpass) frequency domain information of the original signal. The subsignals, xι(n) and xu(n), can be further divided into their frequency components in a similar manner. This SB division process can be repeated until the frequency domain is divided in a sufficient manner. In [7] the two-level subband structure of Fig. 54.2 which divides the frequency , is used to code a single channel ECG signal. The resulting domain into four regions, k4π , (k +41)π k = 0 , 1, 2 , 3 subsignals can be encoded by various coding methods including DPCM, Entropy coding, and transform coding which should be designed to exploit the special nature of the subsignals to achieve the best possible CR for each band. Recently, very successful special encoders which take advantage of the tree structure of the decomposition were developed for compressing the subsignals [26]. The signal reconstruction from the coded subsignals is carried out by using a filter bank consisting of two filters, Gl(e jω) and Gu(e jω). In this case the subsignal, xι(n) [xu(n)] is first upsampled by inserting a zero between every other sample and filtered by Gl(e jω) [Gu(e jω)]. The reconstructed signal, y(n), is the sum of the outputs of these two filters. By properly selecting the filters, Hi (z), Hu(z), Gi (z) and Gu(z), perfect reconstruction is possible, i.e.,

(

)

() (

y n = x n−K

)

(54.7)

in the absence of quantization errors [20]. In other words, filtering and downsampling operations in the SB decomposition structure do not introduce any loss.


FIGURE 54.2 four regions.

Two-level (four branch) subband coder structure. This structure divides the frequency domain into

A CR 5.22 corresponding to a PRD of 5.94% is reported for an ECG signal sampled at 500 Hz with 12 bit/sample resolution [7]. Other WT based ECG coding methods include [20–23].

54.5 Hybrid Multichannel ECG Coding In this subsection, a hybrid frequency domain multichannel compression method [24] for the so-called standard lead [27] ECG recording system is described. The block diagram of this system is shown in Fig. 54.3. The main idea is to exploit not only the correlation between the consecutive ECG samples, but also the inherent correlation among the ECG channels. The standard lead system has 12 ECG channels. In this method, the recorded digital signals are first passed through a preprocessor. The function of the preprocessor is to prepare raw ECG data for further processing. After preprocessing the input signals, the resulting discrete-time sequences are linearly transformed into another set of sequences. The aim of this linear transformation is to decorrelate the highly

FIGURE 54.3


Block Diagram of the Multichannel ECG data Compression Scheme.

correlated ECG lead signals. In a way, this idea is similar to representing the RGB color components of an image in terms of luminance and chrominance components. The transformation matrix, A, can be the matrix of the optimum transform, KLT, or the DCT matrix. Another approach is to use a nonlinear transform such as an Independent Component Analyzer (ICA) [25]. Lastly, to compress the transform domain signals, various coding schemes which exploit their special nature are utilized. In the following subsections, detailed descriptions of the sub-blocks of the multichannel ECG compression method are given.

Preprocessor The standard lead ECG recording configuration consists of 12 ECG leads, I, II, III, AVR, AVL, AVF, V1, V2, …, and V6. The leads, III, AVR, AVL, and AVF, are linearly related to I and II. Therefore, eight channels are enough to represent a standard 12-channel ECG recording system. The preprocessor discards the redundant channels, III, AVR, AVL, and AVF, and rearranges the order of the ECG channels. The six precordial (chest) leads, V1, …, V6, represent variations of the electrical heart vector amplitude with respect to time from six different narrow angles. During a cardiac cycle it is natural to expect high correlation among precordial leads so the channels V1, …, V6 are selected as the first 6 signals, i.e., xi–1 = Vi, i = 1, 2, …, 6. The two horizontal lead waveforms (I and II) which have relatively less energy contents with respect to precordial ECG lead waveforms are chosen as seventh, x6 = I, and eighth channels, x 7 = II. A typical set of standard ECG lead waveforms, xi , i = 0, 1, …, 7, are shown in Fig. 54.4. The aim of the reordering the ECG channels is to increase the efficiency of the linear transformation operation which is described in the next subsection.

Linear Transformer The outputs of the preprocessor block, xi , i = 0, 1, …, 7, are fed to the linear transformer. In this block, the ECG channels are linearly transformed to another domain, and eight new transform domain signals yi , i = 0, 1, …, 7, are obtained which are significantly less correlated (ideally uncorrelated) than the ECG signal set, xi , i = 0, 1, …, 7. The transform domain samples at discrete-time instant m are given as follows:

Ym = A ⋅ X m

(54.8)

where Ym = [yo(m),…,yN–1(m)]T, Xm = [xo(m),…, xn–1(m)]T, and A is the N × N transform matrix. The optimum linear transform, discrete KLT, can be properly defined for stationary random processes and the entries of the tranform matrix, AKLT depending on the statistics of the random processes. For slowly varying, unstationary signals, an approximate KLT matrix can also be defined. Although ECG signals cannot be considered to be wide-sense stationary-random processes, a covariance matrix, Cˆ x , of the ECG channels is estimated as follows:

1 Cˆ x = M

M −1

∑ i=0

( ) 

 x i  0  M   x N −1 i 

   

()

[x (i)Lx (i)] 0

N −1

(54.9)

where N is the number of the ECG channels and M is the number of ECG samples per channel used. The N × N ECG channel covariance matrix, Cˆ x, is used in the construction of an approximate KLT matrix. Rows of the approximate KLT matrix are the eigenvectors of Cˆ x . Typical approximate KLT matrices can be found in [24].


FIGURE 54.4

A typical set of standard ECG lead waveforms xi , i = 0. 1,…, 7.

Although there is no fast algorithm to compute the KL transform, the computational burden is not high because N = 8. The DCT can also be used as a linear transformer because it approximates the KLT. Compression of the Transform Domain Signals In this subsection the compression of the uncorrelated transform domain signals yk , k = 0, 1, …, 7, are described. In Fig. 54.5 a typical set of uncorrelated signals, yk, k = 0, 1, …, 7, are shown. The signals in Fig. 54.5 are obtained by KL transforming the ECG signals, xk , k = 0, 1, …, 7, shown in Fig. 54.4. Transform domain signals, yk, k = 0, 1, …, 7, are divided into two classes according to their energy contents. The first class of signals, yo , y1 , …, y4 , have higher energy than the second class of signals, y5 , y6 , and y7 . More bits are allocated to the high energy signals, y0, y1, …, y4, compared to the low energy signals, y5, y6, and y7 in coding. Subband Coder (SBC) Higher energy signals, yo(n), y1(n),…, y4(n), contain more information than the low energy signals, y5(n), y6(n), and y7(n). Therefore, the high energy signals, yo(n), …, y4(n) should be compressed very accurately. The signals, yo(n), …, y4(n), are compressed using the SBC [7] because this coding scheme does not introduce any visual degradation as pointed out in the previous subsection (4). © 2000 by CRC Press LLC

FIGURE 54.5

Uncorrelated signals yi , i = 0, 1, 2,…, 7, corresponding to the ECG signals shown in Fig. 54.4.

Each signal yi is decomposed into four subsignals by using a filter bank in a tree-structured manner. In this way the signal yi is divided into four consecutive bands, [lπ/4, (l + 1)π/4], l = 0, 1, 2, 3, in the frequency domain [2]. For example, yi00 (the subsignal at branch A of Fig. 54.2) comes from the lowpass frequency band, [0, π/4], of the signal yi . In the coding of the subband signals, yijk , j = 0, 1; k = 0, 1, the advantage is taken of the nonuniform distribution of energy in the frequency domain to judiciously allocate the bits. The number of bits used to encode each frequency band can be different, so the encoding accuracy is always maintained at the required frequency bands [2]. It is observed that the energy of the signal yi is mainly concentrated in the lowest frequency band [0, π/4]. Because of this, the lowband subsignal yi,0,0 has to be carefully coded. High correlation among neighboring samples of yi,0,0 makes this signal a good candidate for efficient predictive or transform coding. The subsignals yi,0,0 are compressed using a DCT based scheme. After the application of DCT with a block size of 64 samples to the lowband subsignal, yi,0,0, the transform domain coefficients, gi,0,0(k), k = 0, 1, …, 63 are obtained and they are thresholded and quantized. The coefficients whose magnitudes are above a preselected threshold, β, are retained and the other coefficients are discarded. Thresholded DCT coefficients are quantized for bit level representation. Thresholded and quantized nonzero coefficients are variable-length coded by using an amplitude table and the zero values are runlength coded. The amplitude and runlength lookup tables are Huffman coding tables which are obtained according to the histograms of the DCT coefficients [24]. In practice, ECG recording levels do not change from one recording to another. If drastic variations occur, first scale the input by an appropriate factor, then apply DCT. Bandpass and highpass subsignals, yi,0,1, yi,1,0 yi,1,1 (branches B, C, and D), are coded using non-uniform quantizers. After quantization, a code assignment procedure is realized using variable length amplitude and runlength lookup tables for zero values. The look up tables are obtained according to the histograms of quantized subband signals.


FIGURE 54.6

The original and reconstructed ECG lead signals I, II, V1, and V2 (CR = 6.17, APRD = 6).

The bit streams which are obtained from coding of four subband signals are multiplexed and stored. Appropriate decoders are assigned to each branch to convert the bit streams into time domain samples and the four-branch synthesis filter bank performs the reconstruction [7]. The low energy signals, y5(n), y6(n), y7(n), are also coded by using the SB coder previously explained. However, it is observed that all the subsignals except the subsignal at branch A of Fig. 54.2 contain very little information when subband decomposition is applied to the signals y5(n), y6(n), y7(n). Due to this fact, only the subsignals at branch A (lowband signals) are processed. The other branches which have very little energy are discarded. Original and reconstructed ECG waveforms are shown in Fig. 54.6 for CR = 6.17 (CR = 7.98) with PRD = 6.19% when DCT (KLT) is used as the Linear Transformer. Recorded ECG signals are sampled at 500 Hz with 12 bit/sample resolution. Also, the raw ECG signals are filtered to attenuate the high frequency noise with a 33-tap equiripple Parks-McClellan FIR filter whose cutoff frequency is equal to 125 Hz. In this case a CR = 9.41 is obtained for a PRD = 5.94%.


The effect of compressing the data on diagnostic computer analysis results is tested on Cardionics program, which is derived from the Mount Sinai program developed by Pordy et al., in conjunction with CRO-MED Bionics Company [10]. Morphological measurements include (1) intervals—PR, QRS, QT, (2) “width”s—Q, R, S, R′, S′, T, P, (3) amplitudes—P, Q, R, R′, S, S′, T, JJ, and (4) areas of QRS and QRST. There was no difference in the measurement results of both the compressed and the original data. It is experimentally observed that the multichannel technique produces better compression results than single channel schemes. Also, the computational complexity of the multichannel scheme is comparable to single channel ECG coding schemes and the algorithm can be implemented by using digital signal processor for real-time applications, such as transmission of ECG signals over telephone lines.

Conclusion In this section biomedical signal compression methods are reviewed. Most of the biomedical data compression methods have been developed for ECG signals. However, these methods can be applied to other biomedical signals with some modifications. It is difficult to compare the efficiency of biomedical data compression schemes because coding results of various data compression schemes are obtained under different recording conditions such as (1) sampling frequency, (2) bandwidth, (3) sample precision, and (4) noise level, which may drastically affect the currently used performance measures.

References [1] J. Willems, “Common Standards for quantitative electrocardiography,” J. Med. Eng. Techn., 9, 209–217, 1985. [2] N. S. Jayant, P. Noll, Digital Coding of Waveforms, Englewood Cliffs, NJ, Prentice-Hall, 1984. [3] S. M. S. Jalaleddine, C. G. Hutchens, R. D. Strattan, and W. A. Coberly, “ECG Data Compression Techniques: A Unified Approach,” IEEE Trans. Biomed. Eng., BME-37, 329–343, 1990. [4] S. M. S. Jalaleddine, C. G. Hutchens, R. D. Strattan, and W. A. Coberly, “Compression of Holter ECG Data,” Biomed. Sci Instrument, 24(4) 35–45, 1988. [5] M. Bertrand, R. Guardo, F. A. Roberge, and P. Blondeau, “Microprocessor Application for Numerical ECG Encoding and Transmission,” Proc. IEEE, 65(5) 714–722, 1977. [6] U.E. Ruttiman and H. V. Pipberger, “Compression of the ECG by Prediction or Interpolation and Entropy Coding,” IEEE Trans. Biomed. Eng., 26(4) 613–623, 1979. [7] M. C. Aydin, A. E. Cetin, ¸ H. Köymen, “ECG Data Compression by Sub-Band Coding,” Electron. Lett., 27(4) 359–360, 1991. [8] N. Ahmed, P. J. Milne, and S. G. Harris, “Electrocardiographic Data Compression via Orthogonal Transforms,” IEEE Trans. Biomed. Eng., 22(6) 484–487, 1975. [9] Y. Z. Ider, H. Köymen, “A new technique for line interference monitoring and reduction in biopotential amplifiers,” IEEE Trans. Biomed. Eng., 37, 624–631, 1990. [10] L. Pordy, H. Jaffe, K. Chelsky, C. K. Freiedberg, L. Fullowes, R. E. Bonner, “Computer Diagnosis of Electrodcardiograms: A Computer Program for Contour Analysis with Classical Results of Rhythm and Contour Interpretation,” Comput. Biomed. Res., 1, 408–433, 1968. [11] P. S. Hamilton, and W. J. Tompkins, “Compression of the Ambulatory ECG by Average Beat Subtraction and Residual Differencing,” IEEE Trans. Biomed. Eng., 38, 253–260, 1991. [12] J. R. Cox et al., “AZTEC: A Preprocessing Program for Real-Time ECG Rhythm Analysis,” IEEE Trans. Biomed. Eng., 15, 128–129, 1968. [13] J. P. Abenstein and W. J. Tompkins, “New Data Reduction Algorithm for Real-Time ECG Analysis,” IEEE Trans. Biomed. Eng., 29, 43–48, 1982. [14] W. C. Mueller, “Arrhythmia Detection Program for an Ambulatory ECG Monitor,” Biomed. Sci. Instrument, 14, 81–85, 1978.


[15] H. Imai, N. Kimura, and Y. Yoshida, “An Efficient Encoding Method for ECG Using Spline Functions,” Syst. Comput, Japan, 16, 85–94, 1985. [16] A. Angelidou et al., “On AR Modeling for MEG Spectral Estimation, Data Compression, and Classification,” Comput. Biol Med., 22, 379–387, 1992. [17] M. E. Womble et al., “Data Compression for Storing and Transmitting ECGs/VCGs,” Proc. IEEE, 65, 702–706, 1977. [18] W. Philips, “ECG Data Compression with Time-Warped Polynomials,” IEEE Trans. Biomed. Eng., 40, 1095–1101, 1993. [19] M. Karczewicz, and M. Gabbouj, “ECG Data Compression by Spline Approximation,” Signal Processing, 59, 43–59, 1997. [20] G. Strang and T. Nguyen, Wavelets and Filter Banks, Wellesley-Cambridge Press, MA, 1996. [21] J. A. Crowe et al., “Wavelet Transform as a Potential Tool for ECG Analysis and Compression,” J. Biomed. Eng., 14, 268–272, 1992. [22] S. C. Tai, “6-band Subband Coder on ECG Waveforms,” Med. and Biolog. Eng. and Comput., 30, 187–192, 1992. [23] A. E. Cetin, ¸ A. H. Tewfik, and Y. Yardimci, “ECG Coding by Wavelet Transform Extrema,” IEEE Symp. Time-Freq. and Time-Scale, Philadelphia, Oct. 1994. [24] A. E. Cetin, ¸ H. Köymen, and M. C. Aydn, “Multichannel ECG Data Compression by Multirate Signal Processing and Transform Coding Techniques,” IEEE Trans. Biomed. Eng., 40, 495–499, 1993. [25] R. Vigario and E. Oja, “ICA Fixed Point Algorithm in Extraction of Artifacts from EEG,” NORSIG’96, Finland, 383–386, 1996. [26] J. M. Shapiro, “Embedded Image Coding Using Zerotrees of Wavelet Coefficients,” IEEE Trans. SP, 41, 3445–3462, 1993. [26] J. B. Wyngaarden and L. H. Smith, Textbook of Medicine, W. B. Saunders, Toronto, 1985.

Further Information Biomedical signal compression is a current research area and most of the research articles describing the advances in biomedical data compression appear in the following journals; IEEE Transactions on Biomedical Engineering and Signal Processing, Journal of Biomedical Engineering, and Medical and Biological Engineering and Computing.


Murray, R., Boudreaux-Bartels, G. F. “Time-Frequency Signal Representations for Biomedical Signals.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

55 Time-Frequency Signal Representations for Biomedical Signals 55.1 55.2 55.3

One-Dimensional Signal Representations Desirable Properties of Time-Frequency Representations TFR Classes Cohen’s Class of TFRs • Affine Class of TFRs • Hyperbolic Class of TFRs • κth Power Class

G. Faye BoudreauxBartels

55.4

University of Rhode Island

Robin Murray University of Rhode Island

Common TFRs and Their Use in Biomedical Applications Wigner Distribution • Smoothed Wigner Distributions • Spectrogram • Choi-Williams Exponential and Reduced Interference Distributions • Scalogram or Wavelet Transform Squared Magnitude • Biomedical Applications

The Fourier transform of a signal x(t)

( ) ∫ ()

X f = x t e − j 2 πft dt

(55.1)

is a useful tool for analyzing the spectral content of a stationary signal and for transforming difficult operations such as convolution, differentiation, and integration into very simple algebraic operations in the Fourier dual domain. The inverse Fourier transform equation

() ∫ ( )

x t = X f e j 2 πft df

(55.2)

is a linear combination of complex sinusoids of infinite duration which implicitly assumes that each sinusoidal component is present at all times and, hence, that the spectral content of the signal does not change with time. However, many signals in bioengineering are produced by biologic systems whose spectral characteristics can change rapidly with time. To analyze these rapid spectral changes one needs a two-dimensional, mixed time-frequency signal representation (TFR) that is analogous to a musical score with time represented along one axis and frequency along the other, indicating which frequency components (notes) are present at each time instant. The first purpose of this chapter is to review several TFRs defined in Table 55.1 and to describe many of the desirable properties listed in Table 55.2 that an ideal TFR should satisfy. TFRs will be grouped into


TABLE 55.1

List of Time-Frequency Representations (TFR) of a Signal x(t) Whose Fourier Transform is X( f )

{ () ( ) }

( )

Ackroyd distribution:

ACK x t , f = Re x * t X f e j2πft

Affine Wigner distribution:

  v  v  v    v  v AWDX t , f ; G , F = G   X  f F   +  X *  f F   −  e j2πtvdv  f    f  2   f  2

Altes-Marinovic distribution:

AM X t , f = f

(Narrowband) ambiguity function:

 τ  τ   v v AFx τ, v = x  t +  x *  t −  e − j2πvt dt = X  f +  X *  f −  e j2πτf df 2 2  2  2  

Andrieux et al. distribution:

(

) ∫

( ) ∫ X ( fe )X * ( fe )e −u 2

u 2

j2 πtfu

du = f

( ) ∫

∫∫ W AF (τ,e )e β

(

j2 π tfβ − fτ

X

)dτ dβ

∫

2      t − t ′ 2  f − b t −t ′ t − t ′    ( )   1 exp − +    2σ 2 σ (t −t ′ )σ ( f − f ′ ) 2σ 2f − f ′  ( )   (t −t ′ )     

(

( ) ∫∫

ANDx t , f =

(

)

)

( )

WDx t ′, f ′ dt ′ df ′

()

where for x t = e

()

()

( ) , then b = d 2 ϕ t , σ = d3 ϕ t t t 2 3

j2 πϕ t

dt

dt

−1 3

, and 2πσt σ f = 12 .

Autocorrelation function:

( ) ∫ ( ) ( ) ∫ X( f ) e 2

Temporal:

act x τ = x * t x t + τ dt =

Spectral:

ACFX v = X * f X f + v df =

() ∫ () (

Bertrand Pk distribution:

)

j2 πfτ

∫ x(t ) e 2

df − j2 πtv

dt

) ∫ X ( f λ (u))X * ( f λ (−u))µ (u)e

(

BkDx t , f ; µ k = f

k

k

( ( ) ( )) du

j2 πtf λk u − λk − u

k

1

()

with λ 0 u =

 1 + ue − u   e u − 1  k −1 , λ1 u = exp  − u  , λ k u = k − ku  sinh u 2  e − 1   e − 1 

( ) ()

u 2e u 2

Born-Jordon distribution:

()



( ) ∫ 1τ ∫ 

BJDx t , f =

t +|τ| 2

t −|τ| 2

()

( )

k ≠ 0, 1, µ k u = µ k −u > 0

 τ  τ  x  t ′ +  x *  t ′ −  dt ′  e − j2πfτdτ 2  2    −1

( ) ∫∫

Butterworth distribution:

BUDx t , f =

Choi-Williams exponential distribution:

CWDx t , f =

   2M   2N  v τ j2 π tv − fτ 1 +      AFx τ, v e ( ) dτdv   τ 0   v0    

( )

2  σ 1 σ  t −t′   τ  τ exp −  x t ′ +  x *  t ′ −  e − j2πfτ dt ′ dτ   4  τ    4π τ 2  2  

( ) ∫∫

( ) X( f ) 2

Cohen’s nonnegative distribution:

( )

CNDx t , f =

()

with ξ x t =


2

xt

1 Ex

( ( ) ( ))

1 + cρ ξ t , η f  x x  

Ex

1 ∫ x(τ) dτ, η ( f ) = E ∫ X ( f ′) df ′, E = ∫ x(t ) dt t

−∞

f

2

x

x

−∞

2

2

x



Cone-Kernel distribution:

Cumulative attack spectrum:

Cumulative decay spectrum:

Flandrin D-distribution:

 CKDx t , f = g τ  

( ) ∫ ()∫

t −|τ| 2

( ) ∫ x(τ)e t

CASx t , f =

t +|τ| 2

2 − j2 πfτ

dτ

− j2 πfτ

dτ

−∞

( ) ∫ x(τ)e ∞

CDSx t , f =

 τ  τ  x  t ′ +  x *  t ′ −  dt ′  e − j2πfτ dτ 2  2   

2

t

 

2  

 

2  

2   





( ) ∫ X  f 1 + u4   X *  f 1 − u4   1 −  u4   e    

FC X t , f = f

()

(

)(

j2 π k∆F t x t = Σ n Σ k GEx n k ; g g t − n∆T e ( )

Generalized Altes distribution:

GAM X t , f ; α = f

Generalized exponential distribution:

GEDx t , f =

) ∫e

− αu

  1 −α u   − 1 +α u  X  fe  2   X *  fe  2   e j2πtfu du    

 



( ) ∫∫ exp − ττ   

2M

0

 v v   0

2N

  AF τ, v e j2π(tv − fτ ) dτdv  x 

( )

Generalized Wigner distribution:

 1    1   GWDx t , f ; α˜ = x  t +  + α˜  τ x *  t −  − α˜  τ e − j2πfτ dτ    2    2

Hyperbolic ambiguity function:

HAFX ζ, β =

Hyperbolic wavelet transform:

Hyperbologram:

Levin distribution:

(

du

)

Gabor expansion, Gex(n, k; g):

(

j2 πtfu

) ∫

( ) ∫ X ( fe )X * ( fe )e ∞

β 2

(

)

fr f

HWTx t , f ; Γ =

(

)

fr f

HYPx t , f ; Γ =

( )

−β 2

(

j2 πζ ln f fr

) df

0

LDx t , f = −

d dt

( )



∞

r

(

j2 πtf ln ξ fr

) dξ

0

2

 f  j2πtf ln( ξ fr ) dξ = HWTx t , f ; Γ X ξ Γ *  r ξ e  f 

∫ () ∞

0

(

t

{ () ( ) }

MH x t , f = Re x t X * f e − j2πft

Multiform, Tiltable distributions:

γ  2  2 α  2 α 2 β   Let µ˜ τ˜ , v˜; α , r , β, γ =  τ˜  v˜  +  τ˜  v˜ + 2r  τ˜˜v           

(

) () ()

( ) ∫∫

MTBDx t , f =

Exponential:

MTEDx t , f =


2

x τ e − j2πfτ dτ

Margineau-Hill distribution:

Butterworth:

)

2

∫ () ∞



f ∫ X (ξ)Γ * f ξ e

() ()

−1

( )

  τ v  j2 π tv − fτ 1 − µ˜ 2λ  , ; α , r , β, γ   AFx τ, v e ( )dτdv  τ 0 v0    

( ) ∫∫ exp −πµ˜ 

2λ

( )

 τ v   j2 π (tv − fτ ) dτdv  τ , v ; α , r , β, γ   AFx τ, v e  0 0  

( )


List of Time-Frequency Representations (TFR) of a Signal x(t) Whose Fourier Transform is X( f ) −1

( ) ( ) ∫∫

MT I C x t , f =

(Inverse) Chebyshev:

±1     1 + 2C ± 2  µ˜  τ , v ; α , r , β, γ    AF τ, v e j2π(tv − fτ ) dτdv λ x      τ 0 v0     

( )

()

where C λ a is a Chebyshev polynomial of order λ 









( ) ∫∫ exp−π  ττ  

NDx t , f =

Nutall-Griffin distribution:

( )

PDx t , f =

Page distribution:

d dt

2   v  τv    j2 π (tv − fτ ) dτdv +   + 2r   AFx τ, v e   v v τ  0  0 0   

2

( )

0

2

∫ () t

−∞

x τ e − j2πfτ dτ

κth power ζ  κ BX( ) ζ, β = AFWκ x  , f r β , Wκ X f =  fr 

( )

Ambiguity function:

(

)( )

  f  X  f r ξ κ−1     fr   

1

( ( f f )) −1 r κ

fr τ κ f ξ

r

  f   f  d t κ AM X( ) t , f = WDWκ x  , fr ξκ    , τ κ f = ξκ    fr τ κ f  f df  fr   r  

( )

Central member:

()

()

() ()

κ 1κ   sgn b b , b ∈ for κ ≠ 0 sgn b b , b ∈ , κ ≠ 0 where ξ κ b =  and ξ κ−1 b =  b > 0 for κ = 0 ln b , e b , b > 0, κ = 0 

()

()

( ) ( ) = ∫ act (τ)e 2

Power spectral density:

PSDx f = X f

Pseudo-Altes distribution:

PADX t , f ; Γ = f r

(

− j2 πfτ

x

) ∫

0

∞

()

dτ

  tf  df ′ f  AM Γ  0, f r AM X  , f ′ f ′    f′  f′

 τ  τ  τ  τ PWDx t , f ; η = x  t +  x *  t −  η  η *  −  e − j2πfτ dτ  2  2  2  2

(

Pseudo-Wigner distribution:

) ∫

( ) ∫∫

RAGD x t, f =

Radially adaptive Gaussian distribution:

(

) ( ) ()

2   τ τ0 + v v0 exp − 2σ x2 θ  

2

  j 2 π (tv − fτ ) dτdv  AFx τ, v e  

( )

 v v0  where θ = arctan    τ τ 0 

Real generalized Wigner distribution:

x    1    1     RGWDx t , f ; α˜ = Re x  t +  + α˜  τ  t −  − α˜  τ e − j2πfτ dτ     2     2   

Reduced interference distribution:

RIDx t , f ; s RID =

(

(

d  with SRID β , SRID 0 = 1,  SRID β  dβ

()


()

()

)

∫

) ∫∫ 1τ s

RID

 t −t′  τ  τ  − j2πfτ dt ′dτ  τ  x t ′ + 2  x *  t ′ − 2  e      

  1  = 0 , s RID α = 0 for α >  2   β =0 

()



( ) () ( )

Rihaczek distribution:

RDx t , f = x t X * f e − j2πtf

Running spectrum, past and future:

RSPx t , f =

( ) ∫ x(u)e t

− j2 πfu

−∞

( ) ∫ x(u)e

du, RSFx t , f =

∞

− j2 πfu

du

t

2

 f  f γ * τ − t dτ = WTx t , f ; γ fr  fr 

(

) ∫ x(τ)

( )

(

) ∫ () ( )

(

Scalogram:

SCALx t , f ; γ =

Short-time Fourier transform:

STFTx t , f ; γ = x τ γ * τ − t e − j2πfτ dτ = e − j2πtf

Smoothed Pseudo-Altes distribution:

c  SPADX t , f ; Γ , g = g tf − c PADX  , f ; Γ dc f 

Smoothed Pseudo-Wigner distribution:

SPWDx t , f ; γ , η =

Spectrogram:

SPEC x t , f ; γ =

Unterberger active distribution:

UADX t , f = f

Unterberger passive distribution:

∫ X ( f ′)Γ * ( f ′ − f )e

) ∫(

(

) ∫∫ γ (t − t ′)η 2τ  η *  − 2τ  x t ′ + 2τ  x *  t ′ − 2τ  e

j2 πtf ′

df ′

)

 

) ∫ x(τ) γ * (τ − t )e



− j2 πfτ

 



(

2



dτ = STFTx t , f ; γ

( ) ∫ X ( fu)X * ( f u)[1 + u ]e ∞

−2

(

j2 πtf u −1 u

)



− j2 πfτ

dt ′dτ

2

)du

0

( )

UPDX t , f = 2 f

∫ X ( fu)X * ( f u)[u ]e ∞

) ∫ ()

WTx t , f ; γ = x τ

Wideband ambiguity function:

WAFX τ, α =

( ) ∫

0

−1

(

j2 πtf u −1 u

)du

0

Wavelet transform:

Wigner distribution:

2

(

(

(

)

∞

 f f γ * τ −t fr  fr



( ) dτ = ∫ X ( f ′)

f  fr Γ *  r f ′ e j2πtf ′ df ′ f  f 

X  f α  X *  f α  e j2πτf df    

 τ  τ   v v WDx t , f = x  t +  x *  t −  e − j2πfτdτ = X  f +  X *  f −  e j2πtvdv 2 2  2  2  

( ) ∫

∫

classes satisfying similar properties to provide a more intuitive understanding of their similarities, advantages, and disadvantages. Further, each TFR within a given class is completely characterized by a unique set of kernels that provide valuable insight into whether or not a given TFR (1) satisfies other ideal TFR properties, (2) is easy to compute, and (3) reduces nonlinear cross-terms. The second goal of this chapter is to discuss applications of TFRs to signal analysis and detection problems in bioengineering. Unfortunately, none of the current TFRs is ideal; some give erroneous information when the signal’s spectra are rapidly time-varying. Researchers often analyze several TFRs side by side, keeping in mind the relative strengths and weaknesses of each TFR before drawing any conclusions.

55.1 One-Dimensional Signal Representations The instantaneous frequency and the group delay of a signal are one-dimensional representations that attempt to represent temporal and spectral signal characteristics simultaneously. The instantaneous frequency of the signal


)

( ) ( )

0

(

)

()

( ) ∫ ( ) ( ) ( ∫ () ) ( ) T (t , f ) = T (t , f − f ′)T (t , f ′)df ′ for y(t ) = h(t )x( t) ∫

)

h

∫ T (t , f )dt = X ( f )

P11: Energy distribution


x

2

( )

n

n

∫∫ f T (t , f )dt df = ∫ f X ( f ) df

2

P13: Frequency moments

n

( )

n

x

∫∫ t T (t , f )dt df = ∫ t x(t ) dt

ΨC τ, 0 = 1

ΨC 0, v = 1

ΨH 0, 0 = 1

P12: Time moments

ΨC 0, 0 = 1

( )

H

Γ

( ) ( ) Ψ (ζ, β) = HAF ( −ζ, − β) ΨH* −ζ, − β = ΨH ζ, β

( )

x

( )

sinh β 2

( )

β 2

( ) ( )

∫∫ T (t , f )dt df = ∫ X ( f ) df

x

()

− j2 πζ lnG β

( )(

Φ H b1 , β Φ H b2 , β = e b1 Φ H b1 , β δ b1 − b2

Always satisfied

Always satisfied

( )

2

( )

ΨC 0, v = 1

C

( ) ( ) Ψ ( τ, v ) = AFγ ( − τ, − v )

()

vpC τ

()

τPC v

ΨC* − τ, − v = ΨC τ, v

ΨC τ, v = e

( )

( )

ΨC τ, v = e

()

()

with G β =

( )

ΨH ζ, β = BH β e

Kernel Constraints for Hyperbolic Class

ΨH ζ, 0 = 1

2

x

( )

ΨC τ, v = SC τv

( )

Always satisfied

Always satisfied

Kernel Constraints for Cohen’s Class

ΨC τ, 0 = 1

x

P10: Frequency marginal

2

∫ T (t , f )df = x(t )

x

Tx* t , f = Tx t , f

( ) ( ) T (t , f ) ≥ 0

y

Ty t , f = Th t − τ, f Tx τ, f dτ for y t = h t − τ x τ dτ

P9: Time marginal

P8: Positivity

P7: Real-valued

P6: Modulation covariant

P5: Convolution covariant

( ) (

 f  Ty t , f = Tx t − c f , f if Y f = exp − j2πc ln  X f fr  

0

P4: Hyperbolic time shift

( ) (

x

Ty t , f = Tx at , f a for y t = a x at

y

P3: Scale covariant

P2: Time-shift covariant

Ty t , f = Tx t , f − f 0 for y t = x t e j2πf0t

( ) ( ) ( () ) T (t , f ) = T (t − t , f ) for y t( = x)(t − t )

TFR Property

List of Desirable Properties for Time-Frequency Representations and Their Corresponding Kernel Constraints

P1: Frequency-shift covariant

Property Name

TABLE 55.2

)

(

( )

P15: Finite frequency support

) ( )

)

( ) ()

( )

( ) (

∫∫ T (t , f )T (t , f )dt df = ∫ x(t )y *( )t dt

P25: Moyal’s formula


Ty t , f = Tx t , f − ct for y t = x t e j πct

( ∫ )( )

P24: Chirp multiplication

)

x

* y

)

( )( ) 2

2

Ty t , f = Tx t − f c , f for y t = x t − τ

( ) (

()

c 1  1 δ t −  , f > 0 if X c f = e f  f f

− j πcf

2

0

P23: Chirp convolution

Tx t , f =

x

P21: Linear chirp localization

0

Tx t , f = δ f − f 0 for X f = δ f − f 0

( ) ( ) ( )( ) T (t , f ) = δ(t − t ) for x( t =) δ(t − t ) T (t , f ) = δ(t − cf ) for X( f = )e

Ty t , f = Tx − f , t for y t = X t

( ) (

x

x

x

x

∫ fT (t , f )df = 1 d arg{x(t )} ∫ T (t , f )df 2π dt ∫ tT (t , f )dt = − 1 d arg{x( f )} ∫ T (t , f )dt 2π df

x

P22: Hyperbolic localization

( )

(

)

2

,f >0

c e j πcτ dτ

f − j2 πc ln fr

Tx t , f = 0 for f ∉ f1 , f 2 if X f = 0 for f ∉ f1 , f 2

P20: Time localization

P19: Frequency localization

P18: Fourier transform

P17: Group delay

P16: Instantaneous frequency

( ) ( )

( )

Tx t , f = 0 for t ∉ t 1 , t 2 if x t = 0 for t ∉ t 1 , t 2

P14: Finite time support

)

( )

( )

)

( ) ΨC τ, v = 1

( )

ΨC τ, v − cτ = ΨC τ, v

(

( )

∂ ΨC τ, v ∂v

 v  ΨC  τ − , v = ΨC τ, v c  

C

C

( ) Ψ (0, v ) = 1 Ψ ( τ, v ) = 1 ΨC τ, 0 = 1

( )

∂ ΨC τ, v ∂τ

ΨC −v , τ = ΨC τ, v

(

( )

ΨC τ, 0 = 1 and

( )

f 1 > v 2

ΨC 0, v = 1 and

( )

t 1 > τ 2

Φ C f , v = 0,

( )

ϕ C t , τ = 0,

v =0

τ =0

=0

=0

( )

ΨH ζ, β = 1

( )

ΨH 0, β = 1

( )

ΨH ζ, 0 = 1

( )

( ) ∂ ΨH ζ, β ∂β

c 1 > ζ 2

ΨH ζ, 0 = 1 and

( )

Φ H c , ζ = 0,

β =0

=0

()

fx t =

{ ( )}

1 d arg x t 2π dt

(55.3)

has been used in communication theory to characterize the time-varying frequency content of narrowband, frequency-modulated signals. It is a generalization of the fact that the frequency f0 of a complex sinusoidal signal x(t) = exp(j2πf0t) is proportional to the derivative of the signal’s phase. A dual concept used in filter analysis is the group delay

()

τH f = −

{ ( )}

1 d arg H f 2π df

(55.4)

which can be interpreted as the time delay or distortion introduced by the filter’s frequency response H(f ) at each frequency. Group delay is a generalization of the fact that time translations are coded in the derivative of the phase of the Fourier transform. Unfortunately, if the signal contains several signal components that overlap in time or frequency, then fx(t) or τH(f ) only provides average spectral characteristics, which are not very useful.

55.2 Desirable Properties of Time-Frequency Representations Mixed time-frequency representations (TFRs) map a one-dimensional signal into a two-dimensional function of time and frequency in order to analyze the time-varying spectral content of the signal. Before discussing any particular TFR in Table 55.1, it is helpful to first investigate what types of properties an “ideal” time-frequency representation should satisfy. The list of desirable TFR properties in Table 55.2 can be broken up conceptually into the following categories: covariance, statistical, signal analysis, localization, and inner products [Boashash, 1991; Claasen & Mecklenbräuker, 1980; Cohen, 1989; Flandrin, 1993; Hlawatsch & Boudreaux-Bartels, 1992]. The covariance properties P1 to P6 basically state that certain operations on the signal, such as translations, dilations, or convolution, should be preserved, i.e., produce exactly the same operation on the signal’s TFR. The second category of properties originates from the desire to generalize the concepts of the one-dimensional instantaneous signal energy x(t)2 and power spectral density x(f )2 into a two-dimensional statistical energy distribution Tx(t0, f0) that provides a measure of the local signal energy or the probability that a signal contains a sinusoidal component of frequency f0 at time t0. Properties P7 to P13 state that such an energy-distribution TFR should be real and nonnegative, have its marginal distributions equal to the signal’s temporal and spectral energy densities x(t) 2 and X(f ) 2, respectively, and preserve the signal energy, mean, variance, and other higher-order moments of the instantaneous signal energy and power spectral density. The next category of properties, P14 to P18 , arises from signal-processing considerations. A TFR should have the same duration and bandwidth as the signal under analysis. At any given time t, the average frequency should equal the instantaneous frequency of the signal, while the average or center of gravity in the time direction should equal the group delay of the signal. These two properties have been used to analyze the distortion of audio systems and the complex FM sonar signals used by bats and whales for echolocation. Property P18 is the TFR equivalent of the duality property of Fourier transforms. The group of properties P19 to P24 constitutes ideal TFR localization properties that are desirable for high resolution capabilities. Here, δ(a) is the Dirac function. These properties state that if a signal is perfectly concentrated in time or frequency, i.e., an impulse or a sinusoid, then its TFR also should be perfectly concentrated at the same time or frequency. Properties P21 and P22 state that the TFRs of linear or hyperbolic spectral FM chirp signals should be perfectly concentrated along the chirp signal’s group delay. Property P24 states that a signal modulated by a linear FM chirp should have a TFR whose instantaneous frequency has been sheared by an amount equal to the linear instantaneous frequency of the chirp. The last property, known as Moyal’s formula or the unitarity property, states that TFRs should preserve the signal projections, inner products,


and orthonormal signal basis functions that are used frequently in signal detection, synthesis, and approximation theory. Table 55.3 indicates which properties are satisfied by the TFRs listed in Table 55.1. A TFR should be relatively easy to compute and interpret. Interpretation is greatly simplified if the TFR is linear, i.e.,

( ) ∑T (t , f )

( ) ∑ x (t )

N

Ty t , f =

N

for y t =

xn

n =1

(55.5)

n

n =1

However, energy is a quadratic function of the signal, and hence so too are many of the TFRs in Table 55.1. The nonlinear nature of TFRs gives rise to troublesome cross-terms. If y(t) contains N signal components or auto-terms xn(t) in Eq. (55.5), then a quadratic TFR of y(t) can have as many nonzero cross-terms as there are unique pairs of autoterms, i.e., N(N – 1)/2. For many TFRs, these cross-terms are oscillatory and overlap with autoterms, obscuring visual analysis of TFRs. Two common methods used to reduce the number of cross-terms are to reduce any redundancy in the signal representation and to use local smoothing or averaging to reduce oscillating cross-terms. TFR analysis of real, bandpass signals should be carried out using the analytic signal representation, i.e., the signal added to – 1 times its Hilbert transform, in order to remove cross-terms between the positiveand negative-frequency axis components of the signal’s Fourier transform. As we will see in upcoming sections, cross-term reduction by smoothing is often achieved at the expense of significant autoterm distortion and loss of desirable TFR properties.

55.3 TFR Classes This section will briefly review Cohen’s class of shift covariant TFRs, the Affine class of affine covariant TFRs, the Hyperbolic class (developed for signals with hyperbolic group delay), and the Power class (which is useful for signals with polynomial group delay). Each class is formed by grouping TFRs that satisfy two properties. They provide very helpful insight as to which types of TFRs will work best in different situations. Within a class, each TFR is completely characterized by a unique set of TFR-dependent kernels which can be compared against a class-dependent list of kernel constraints in Tables 55.2 and 55.6 to quickly determine which properties the TFR satisfies.

Cohen’s Class of TFRs Cohen’s class consists of all quadratic TFRs that satisfy the frequency-shift and time-shift covariance properties, i.e., those TFRs with a check in the first two property rows in Table 55.3 [Claasen & Mecklenbräuker, 1980; Cohen, 1989; Flandrin, 1993; Hlawatsch & Boudreaux-Bartels, 1992]. Time- and frequency-shift covariances are very useful properties in the analysis of speech, narrowband Doppler systems, and multipath environments. Any TFR in Cohen’s class can be written in one of the four equivalent “normal forms”:

(

=










) ∫∫ ϕ (t − t ′, τ) x t ′ + 2τ  x *  t ′ − 2τ  e

C x t , f ; ΨC =

C

− j 2 πfτ

dt ′ dτ

  v v ΦC f − f ′, v X  f ′ +  X *  f ′ −  e j 2 πtvdf ′ dv 2 2  

∫∫ (

)

(55.6)

(55.7)

List of Desirable Properties Satisfied by Time-Frequency Representations


TABLE 55.3

=

∫∫ ψ (t − t ′, f − f ′) WD (t ′, f ′) dt ′ df ′

(55.8)

=

∫∫ Ψ (τ, v ) AF (τ, v )e

(55.9)

C

x

C

x

(

j 2 π tv − fτ

)dτ dv .

Each normal form is characterized by one of the four kernels ϕC(t, τ), ΦC(f, ν), ψC(t, f), and ΨC(τ, ν) which are interrelated by the following Fourier transforms:

( ) ∫∫ Φ ( f , v )e

ϕC t , τ =

C

( ) ∫∫ Ψ (τ, v )e

ψC t, f =

C

(

)df dv = Ψ τ, v e j 2 πvt dv C

∫ ( )

(55.10)

(

)dτ dv = Φ f , v e j 2 πvt dv . C

∫ ( )

(55.11)

j 2 π fτ + vt

j 2 π vt − f τ

The kernels for the TFRs in Cohen’s class are given in Table 55.4. The four normal forms offers various computational and analysis advantages. For example, the first two normal forms can be computed directly from the signal x(t) or its Fourier transform X(f ) via a onedimensional convolution with ϕC(t, τ) or ΦC(f, ν). If ϕC(t, τ) is of fairly short duration, then it may be possible to implement Eq. (55.6) on a digital computer in real time using only a small number of signal samples. The third normal form indicates that any TFR in Cohen’s shift covariant class can be computed by convolving the TFR-dependent kernel ψC(t, f) with the Wigner distribution (WD) of the signal, defined in Table 55.1. Hence the WD is one of the key members of Cohen’s class, and many TFRs correspond to smoothed WDs, as can be seen in the top of Table 55.5. Equation (55.11) and the fourth normal form in Eq. (55.9) indicate that the two-dimensional convolution in Eq. (55.8) transforms to multiplication of the Fourier transform of the kernel ψC(t, f) with the Fourier transform of the WD, which is the ambiguity function (AF) in Table 55.1. This last normal form provides an intuitive interpretation that the “AF domain” kernel ΨC(τ, ν) can be thought of as the frequency response of a two-dimensional filter. The kernels in Eqs. (55.6) to (55.11) are signal-independent and provide valuable insight into the performance of each Cohen class TFR, regardless of the input signal. For good cross-term reduction and little autoterm distortion, each TFR kernel ΨC(τ, ν) given in Table 55.4 should be as close as possible to an ideal low-pass filter. If these kernels satisfy the constraints in the third column of Table 55.2, then the TFR properties in the first column are guaranteed to always hold [Claasen & Mecklenbräuker, 1980; Hlawatsch & Boudreaux-Bartels, 1992]. For example, the last row of Table 55.2 indicates that Moyal’s formula is satisfied by any TFR whose AF domain kernel, listed in the third column of Table 55.4, has unit modulus, e.g., the Rihaczek distribution. Since the AF domain kernel of the WD is equal to 1, i.e., ΨWD(τ, ν) = 1, then the WD automatically satisfies the kernel constraints in Table 55.2 for properties P9 to P13 and P16 to P21 as well as Moyal’s formula. However, it also acts as an all-pass filter, passing all crossterms. The Choi-Williams Gaussian kernel in Table 55.4 was formulated to satisfy the marginal property constraints of having an AF domain kernel equal to 1 along the axes and to be a low-pass filter that reduces cross-terms.

Affine Class of TFRs TFRs that are covariant to scale changes and time translations, i.e., properties P2 and P3 in Tables 55.2 and 55.3, are members of the Affine class [Bertrand chapter in Boashash, 1991; Flandrin, 1993]. The scale covariance property P3 is useful when analyzing wideband Doppler systems, signals with fractal structure, octave-band systems such as the cochlea of the inner ear, and detecting short-duration “transients.” Any Affine class TFR can be written in four “normal form” equations similar to those of Cohen’s class:


TABLE 55.4 TFR ACK

Kernels of Cohen’s Shift-Invariant Class of Time-Frequency Representations (TFR) ψC(t, f)

( )

2 cos 4πtf

ΨC(τ, ν)

ϕC(t, τ)

(

( )

πτv

   2M   2N  τ v 1 +        τ 0   v0    

CWD

− 2 πτv e ( )

CKD

g τ τ

CAS

1 1  − jπ τ v  δ v + e jv  2

CDS

1 1  jπ τ v  δ −v −  e jv  2

GED2

   2M   2N  v τ exp −       τ 0   v0    

)

1  , f v < 1 2 v 0, f v > 1 2 

2  σ 1 σt exp−     4  τ  4π τ  

σ

2  σ 1 σ f  exp−     4 v  4π v  

()

( )

g τ , t τ < 1 2   t τ >1 2 0,

sin πτv πτv

()

( )

τ0 2 π τ v0

M

 −v 2 τ 2M t 2  exp  0 20M   4τ 

N = 1 only

(

MIN  v r 1 

GRD

1 j2πtf e α˜

˜ α

(

(

e jπ τ v

δt+τ 2

MH

cos πτv

(

)

(

) ( 2

−1

  τ v  1 + µ˜ 2λ  , ; α , r , β, γ    τ 0 v0   

MTC

     1 + 2pC λ2  µ˜  τ , v ; α , ρ, β, γ          τ 0 v0 

MTED

  τ v  exp −πµ˜ 2λ  , ; α , r , β, γ   τ v  0 0   


δ f − α˜ v

δ t +τ 2 +δ t −τ 2

MTBD

−1

v0 2 π v

N

 − τ 2v 2 N f 2  exp  0 02N   4v 

)

)

LD

( )

MIN

πt

˜ δ t + ατ

˜

τ0

M = 1 only

sin 2π r t τ

e j2πατv

GWD

) (

−1

BUD

()

(

δ f −v 2 +δ f +v 2 2

1  , t τ < 1 2 τ 0, t τ > 1 2 

( )

2

)

2

sin πτv

BJD

) (

δ t +τ 2 +δ t −τ 2

cos πτv

ΦC(f, ν)

)

(

) ) (

δ f −v 2 +δ f +v 2 2

)


Kernels of Cohen’s Shift-Invariant Class of Time-Frequency Representations (TFR)

ψC(t, f)

TFR

ΨC(τ, ν)

ϕC(t, τ)

MTIC

     1 + 2s C λ−2  µ˜ −1  τ , v ; α , r , β, γ          τ 0 v0 

ND

  τ v  exp −πµ˜  , ; 0, r , 1, 1  v τ  0 0   

PD

e − jπ τ v

()

( )

PWD

δ t WDη 0, f

RGWD

1 cos 2π tf α˜ α˜

RID

(

∫τs 1

RID

t  τ ,  

(

˜ cos 2πατv

RD

2e − j 4πtf

e − jπτv

SPWD

γ t WDη 0, f

SPEC

WDγ −t , − f

WD

δt δ f

(

() ( )

)

() ( ) ( ) ˜ ) + δ(t − ατ ˜ ) δ(t + ατ

)

1 τ

()

(

δ t −τ 2

()

2

t s RID   ,  τ

s RID α = 0, α >

 τ  τ η  η *  −  Γ v  2  2

(

( ) δ( f − α˜ v ) + δ( f + α˜ v )

WDη 0, f

2

RID

AFγ − τ, − v

)

δ t η τ 2 η * −τ 2

( ) (β)Real, S (0) = 1

SRID

( )

)

SRID τv ,

e − j2πfτdτ

()

(

( ) (

)

−1

δt−τ 2

η τ 2 η * −τ 2

ΦC(f, ν)

1 v

()

1 2

s RID α = 0, α >

)

)

()

( )

1 2

 τ  τ γ t η  η *  −   2  2

Γ v WDη 0, f

 τ  τ γ  −t −  γ *  −t +  2  2 

  v v Γ f −  Γ *−f +  2 2  

δ(t)

1

(

δ f +v 2

()

)

 f s RID  −  ,  v

δ(f)

˜ + 2r((˜τ ˜ν)β)γ) and C (a) is a Chebyshev polynomial of order λ. Functions Here, ˜µ(˜τ, ˜ν; α, r, β, γ) = ((˜τ)2 ((˜ν)2)α + ((˜τ)2)α(ν) λ with lowercase and uppercase letters, e.g. γ(t) and Γ(f ), indicate Fourier transform pairs.

(

) ∫∫ ϕ ( f (t ′ − t ), f τ)x(t + τ 2)x * (t − τ 2)dt ′ dτ

Ax t , f ; ΨA = f

A

∫∫

 f′ v Φ A  ,  X f ′ + v 2 X * f ′ − v 2 e j 2 πtvdf ′ dv  f f

(

) (

=

1

=

 f ′ ∫∫ ψ  f (t − t ′), f  W D (t ′, f ′)dt ′ df ′

=

f

A

∫∫

x

 v ΨA  f τ,  AFx τ, v e j 2 πtv dτ dv . f 

( )

)

(55.12)

(55.13)

(55.14)

(55.15)

The Affine class kernels are interrelated by the same Fourier transforms given in Eqs. (55.10) and (55.11). Note that the third normal form of the Affine class involves an Affine smoothing of the WD. Well-known


TABLE 55.5

Many TFRs Are Equivalent to Smoothed or Warped Wigner Distributions

TFR Name

TFR formulation

(

) ∫∫ ψ (t − t ′, f − f ′)WD (t ′, f ′)dt ′ df ′

Cohen’s class TFR

Cx t , f ; ψC =

Pseudo-Wigner distribution

PWDx t , f ; η = WDη 0, f − f ′ WDx t , f ′ df ′

Scalogram

SCALx t , f ; γ =

C

x

(

) ∫

(

)

( )

(

) ∫∫ WD  ff (t ′ − t ), f



γ

r

(

r

f ′ WDx t ′, f ′ dt ′ df ′ f 

(

)

) ∫∫ γ (t − t ′)WD (0, f − f ′)WD (t ′, f ′)dt ′ df ′

Smoothed Pseudo-Wigner distribution

SPWDx t , f ; γ , η =

Spectrogram

SPEC x t , f ; γ =

Altes distribution

 tf f  AM X t , f = WDX  , f r ln  f f  r r

κth Power Altes distribution

 t κ AM X( ) t , f = WD κ X    κ f fr

Hyperbologram

HYPX t , f ; Γ =

Pseudo-Altes distribution

PADX t , f ; Γ = f r

(

η

x

) ∫∫ WD (t ′ − t , f ′ − f )WD (t ′, f ′)dt ′ df ′ γ

x

( )

( )

∞

(

) ∫

 tf  f ′  df ′ f  WDΓ  0, f r ln  WDX  , f r ln  f ′ f fr  f ′   r

∞

−∞ 0 ∞

0

(

) ∫ ∫ g (tf − c )WD

)( )

Where fr > 0 is a positive reference frequency, H f = e f

()

 tf f  WDΓ  t ′ − , f ′ − f r ln  WDX t ′, f ′ dt ′ df ′ f f  r r

) ∫ ∫

(

(

()

(

SPADX t , f ; Γ , g = f r

Smoothed Pseudo-Altes distribution

 1, κ ≠ 0, and sgn f =  −1,

 κ  , sgn f f f f r r ,κ ≠0 κ −1  

fr

∞

∞

−∞ 0

(

H fre f

f >0 f ζ 2

( )

b −1 1 > β 2

( ) ( )

( )

∂ Φ A b, β ∂β

=0 β =0

( ) ( )

P19: Frequency localization

Φ A b, 0 = δ b − 1

P25: Moyal’s formula

˜ ) Φ (β, ηβ ˜ ) dβ = δ(b − 1), ∀ η ˜ ∫ Φ (bβ, ηβ


* A

)) ( (

fr Γ fr b − β 2 Γ * fr b + β 2

A

−1 2

δ b − 1 + β2 4   

))

∞

∞

 f  ψ H  tf − t ′f ′, ln  AM X t ′, f ′ dt ′ df ′ f ′ 

(

=

∫ ∫

=

∫∫ ( )

−∞ 0

( )

ΨH ζ, β HAFX ζ, β e

)

[ ( )]

j 2 π  tf β − ln f f r ζ 

(55.18)

dζ dβ

(55.19)

where AMX(t, f) is the Altes distribution and HAFX(ζ, β) is the hyperbolic ambiguity function defined in Table 55.1, νX(c, ζ) is defined in Table 55.7, ( X)(f ) = e f ⁄ fr X(fre f/fr) is a unitary warping on the frequency axis of the signal, and the kernels are interrelated via the Fourier transforms in Eqs. (55.10) and (55.11). Tables 55.3 reveals that the Altes-Marinovic, the Bertrands’ P0 , and the hyperbologram distributions are members of the Hyperbolic class. Their kernels are given in Table 55.7, and kernel property constraints are given in Table 55.2. The hyperbolic TFRs give highly concentrated TFR representations for signals with hyperbolic group delay. Each Hyperbolic class TFR, kernel, and property corresponds to a warped version of a Cohen’s class TFR, kernel, and property, respectively. For example, Table 55.5 shows that the Altes distribution is equal to the WD after both the signal and the time-frequency axes are warped appropriately. The WD’s perfect location of linear FM chirps (P21) corresponds to the Altes distribution’s perfect localization for hyperbolic FM chirps (P22). This one-to-one correspondence between the Cohen and Hyperbolic classes greatly facilitates their analysis and gives alternative methods for calculating various TFRs.

κth Power Class The Power class of TFRs consists of all TFRs that are scale covariant and power time-shift covariant, i.e.,

 d κ κ  PCY( ) t , f = PC X( )  t − c ξ f fr , f  df  

( )

TABLE 55.7

( )

ψH(c, b)

ψH(ζ, β)

AM

δ(c)δ(b)

B0D

∫ δ(b + ln λ(β))e

GAM

1 j2πcb α˜ e α˜

e j2παζβ

HYP

 −c  AM Γ  − b , f r e − b  f e  r 

HAFΓ −ζ, − β

PAD

f r δ c AM Γ 0, f r e b

j2 πcβ

dβ

()

∞

0


()

)X f

(55.20)

e

()

∫e

− j2 πζ ln λ β

(

( )

)

( )

() ( )

f r G β υ Γ 0, ζ

r

(

) (

(

˜ δ c + αζ

f r υ Γ 0, ζ

b

Γ

δ(b)

( )) dβ

j2π cβ −ζ ln λ β

(

(

ΦH(b, β)

)

υ Γ −c, − ζ

(

( ))

δ b + ln λ β

(

˜ δ b − αβ

)

)

(

VΓ −b, − β

)

(

)

() ( )

f r AM Γ 0, f r e b

() ( )

f r G β AM Γ 0, f r e b

f r δ c υ Γ 0, ζ

f r g c υ Γ 0, ζ

) ( ) (

(

()

) (

)

β 2 , V b, β = f r e bΓ f r e b+β 2 Γ * f r e b−β 2 , υ Γ c , ζ = ρΓ c + ζ 2 ρ*Γ c − ζ 2 and sinh β 2 Γ  f  Γ f    fr 

() ∫ ( )

ρΓ c =

δ(c)

˜

( ) f g (c ) AM (0, f e ) r

ϕH(c, ζ)

1

()

Here λ β =

(

− j 2 πcξκ f fr

Kernels of the Hyperbolic Class of Time-Frequency Representations

TFR

SPAD

()

for Y f = e

j2 πc

df f

.

)

where ζ κ(f ) = sgn(f ) f κ, for κ ≠ 0 [Hlawatsch et al., 1993]. Consequently, the κth Power class perfectly represents group delay changes in the signal that are powers of frequency. When κ = 1, the Power class is equivalent to the Affine class. The central member, AM(κ) in Table 55.1, is the Power class equivalent to the Altes-Marinovic distribution.

55.4 Common TFRs and Their Use in Biomedical Applications This section will briefly review some of the TFRs commonly used in biomedical analysis and summarize their relative advantages and disadvantages.

Wigner Distribution One of the oldest TFRs in Table 55.1 is the Wigner distribution (WD), which Wigner proposed in quantum mechanics as a two-dimensional statistical distribution relating the Fourier transform pairs of position and momentum of a particle. Table 55.3 reveals that the WD satisfies a large number of desirable TFR properties, P1 to P3, P5 to P7, P9 to P21, and P23 and P25. It is a member of both the Cohen and the Affine classes. The WD is a high-resolution TFR for linear FM chirps, sinusoids, and impulses. Since the WD satisfies Moyal’s formula, it has been used to design optimal signal-detection and synthesis algorithms. The drawbacks of the WD are that it can be negative, it requires the signal to be known for all time, and it is a quadratic TFR with no implicit smoothing to remove cross-terms.

Smoothed Wigner Distributions Many TFRs are related to the WD by either smoothing or a warping, e.g., see Eqs. (55.8) and (55.18) and Table 55.5. An intuitive understanding of the effects of cross-terms on quadratic TFRs can be obtained by analyzing the WD of a multicomponent signal y(t) in Eq. (55.5) under the assumption that each signal component is a shifted version of a basic envelope, i.e., xn(t) = x(t – tn)e j2πfnt:

( ) ∑WD (t − t , f − f ) N

WD y t , f =

x

n

n

n =1

N −1

+2

∑ ∑ WD (t − t N

x

k =1 q = k +1

(

)

k ,q

)

[ (

, f − fk ,q cos  2π ∆fk ,q t − ∆t k ,q

− ∆t k ,q f − ∆fk ,q + ∆fk ,q ∆t k ,q

)

(55.21)

]) –

where ∆fk,q = fk – fq is the difference or “beat” frequency and fk,q = (fk + fq)/2 is the average frequency – between the kth and qth signal components. Similarly, ∆tk,q is the difference time and tk,q is the average time. The auto-WD terms in the first summation properly reflect the fact that the WD is a member of Cohen’s shift covariant class. Unfortunately, the cross-WD terms in the second summation occur midway in the time- frequency plane between each pair of signal components and oscillate with a spatial frequency proportional to the distance between them. The Pseudo-WD (PWD) and the smoothed Pseudo-WD (SPWD) defined in Table 55.1 use low-pass smoothing windows η(τ) and γ(t) to reduce oscillatory crosscomponents. However, Table 55.5 reveals that the Pseudo-WD performs smoothly only in the frequency direction. Short smoothing windows greatly reduce the limit of integration in the Pseudo-WD and SPWD formulations and hence reduce computation time. However, Table 55.3 reveals that smoothing the WD reduces the number of desirable properties it satisfies from 18 to 7 for the Pseudo-WD and to only 3 for the SPWD.


Spectrogram One of the most commonly used TFRs for slowly time-varying or quasi-stationary signals is the spectrogram, defined in Tables 55.1 and 55.5 [Rabiner & Schafer, 1978]. It is equal to the squared magnitude of the short-time Fourier transform, performing a local or “short-time” Fourier analysis by using a sliding analysis window γ(t) to segment the signal into short sections centered near the output time t before computing a Fourier transformation. The spectrogram is easy to compute, using either FFTs or a parallel bank of filters, and it is often easy to interpret. The quadratic spectrogram smooths away all cross-terms except those which occur when two signal components overlap. This smoothing also distorts auto-terms. The spectrogram does a poor job representing rapidly changing spectral characteristics or resolving two closely spaced components because there is an inherent tradeoff between good time resolution, which requires a short analysis window, and good frequency resolution, which requires a long analysis window. The spectrogram satisfies only three TFR properties listed in Tables 55.2 and 55.3; i.e., it is a nonnegative number of Cohen’s shift invariant class.

Choi-Williams Exponential and Reduced Interference Distributions The Choi-Williams exponential distribution (CWD) and the reduced interference distribution (RID) in Table 55.1 are often used as a compromise between the high-resolution but cluttered WD versus the smeared but easy to interpret spectrogram [Jeong & Williams, 1992; Williams & Jeong, 1991]. Since they are members of both Cohen’s class and the Affine class, their AF domain kernels in Table 55.4 have a very special form, i.e., ΨC(τ, ν) = Sc(τν), called a product kernel, which is a one-dimensional kernel evaluated at the product of its time-frequency variables [Hlawatsch & Boudreaux-Bartels, 1992]. The CWD uses a Gaussian product kernel in the AF plane to reduce cross-terms, while the RID typically uses a classic window function that is time-limited and normalized to automatically satisfy many desirable TFR properties (see Table 55.3). The CWD has one scaling factor σ that allows the user to select either good cross-term reduction or good auto-term preservation but, unfortunately, not always both. The generalized exponential distribution in Table 55.1 is an extension of the CWD that permits both [Hlawatsch & Boudreaux-Bartels, 1992]. Because the CWD and RID product kernels have hyperbolic isocontours in the AF plane, they always pass cross-terms between signal components that occur at either the same time or frequency, and they can distort auto-terms of linear FM chirp signals whose instantaneous frequency has a slope close to 1. The multiform tiltable exponential distribution (MTED) [Costa and Boudreaux-Bartels, 1994], another extension of the CWD, works well for any linear FM chirp.

Scalogram or Wavelet Transform Squared Magnitude The scalogram [Flandrin, 1993], defined in Tables 55.1 and 55.5, is the squared magnitude of the recently introduced wavelet transform (WT) [Daubechies, 1992; Meyer, 1993] and is a member of the Affine class. It uses a special sliding analysis window γ(t), called the mother wavelet, to analyze local spectral information of the signal x(t). The mother wavelet is either compressed or dilated to give a multiresolution signal representation. The scalogram can be thought of as the multiresolution output of a parallel bank of octave band-filters. High-frequency regions of the WT domain have very good time resolution, whereas low-frequency regions of the WT domain have very good spectral resolution. The WT has been used to model the middle- to high-frequency range operation of the cochlea, to track transients such as speech pitch and the onset of the QRS complex in ECG signals, and to analyze fractal and chaotic signals. One drawback of the scalogram is its poor temporal resolution at low-frequency regions of the time-frequency plane and poor spectral resolution at high frequencies. Moreover, many “classic” windows do not satisfy the conditions needed for a mother wavelet. The scalogram cannot remove cross-terms when signal components overlap. Further, many discrete WT implementations do not preserve the important timeshift covariance property.


Biomedical Applications The electrocardiogram (ECG) signal is a recording of the time-varying electric rhythm of the heart. The short-duration QRS complex is the most predominant feature of the normal ECG signal. Abnormal heart rhythms can be identified on the ECG by detecting the QRS complex from one cycle to the next. The transient detection capability of the wavelet transform (WT) has been exploited for detection of the QRS complex by Kadambe et al. [1992] and Li and Zheng [1993]. The WT exhibits local maxima which align across successive (dyadic) scales at the location of transient components, such as QRS complexes. The advantage of using the WT is that it is robust both to noise and to nonstationarities in the QRS complex. Other pathologic features in the heart’s electrical rhythm that appear only in high-resolution signalaveraged ECG signals are ventricular late potentials (VLPs). VLPs are small-amplitude, short-duration components that occur after the QRS complex and are precursors of dangerous, life-threatening cardiac arrhythmias. Tuteur [1989] used the peak of the WT at a fixed scale to identify simulated VLPs. More recently, Jones et al. [1992] compared different time-frequency techniques, such as the spectrogram, short-time spectral estimators, the smoothed WD, and the WT, in their ability to discriminate between normal patients and patients susceptible to dangerous arrhythmias. Morlet et al. [1993] used the transient detection capability of the WT to identify VLPs. The WT has also been applied to the ECG signal in the context of ECG analysis and compression by Crowe et al. [1992]. Furthermore, Crowe et al. [1992] exploited the capability of the WT to analyze fractal-like signals to study heart rate variability (HRV) data, which have been described as having fractallike properties. The recording of heart sounds, or phonocardiogram (PCG) signal, has been analyzed using many timefrequency techniques. Bulgrin et al. [1993] compared the short-time Fourier transform and the WT for the analysis of abnormal PCGs. Picard et al. [1991] analyzed the sounds produced by different prosthetic valves using the spectrogram. The binomial RID, which is a fast approximation to the CWD, was used to analyze the short-time, narrow-bandwidth features of first heart sound in mongrel dogs by Wood et al. [1992]. TFRs have also been applied to nonstationary brain wave signals, including the electrocardiogram (EEG), the electrocorticogram (ECoG), and evoked potentials (EPs). Zaveri et al. [1992] used the spectrogram, the WD, and the CWD to characterize the nonstationary behavior of the ECoG of epileptic patients. Of the three techniques, the CWD exhibited superior results. The WT was used to identify the onset of epileptic seizures in the EEG by Schiff and Milton [1993], to extract a single EP by Bartnik et al. [1992], and to characterize changes in somatosensory EPs due to brain injury caused by oxygen deprivation by Thakor et al. [1993]. Crackles are lung sounds indicative of pathologic conditions. Verreault [1989] used AR models of slices of the WD to discriminate crackles from normal lung sounds. The electrogastrogram (EGG) is a noninvasive measure of the time-varying electrical activity of the stomach. Promising results regarding abnormal EGG rhythms and the frequency of the EGG slow wave were obtained using the CWD by Lin and Chen [1994]. Widmalm et al. [1991] analyzed temporomandibular joint (TMI) clicking using the spectrogram, the WD, and the RID. The RID allowed for better time-frequency resolution of the TMJ sounds than the spectrogram while reducing the cross-terms associated with the WD. TMJ signals were also modeled using nonorthogonal Gabor logons by Brown et al. [1994]. The primary advantage of this technique, which optimizes the location and support of each Gabor log-on, is that only a few such logons were needed to represent the TMJ clicks. Auditory applications of TFRs are intuitively appealing because the cochlea exhibits constant-bandwidth behavior at low frequencies and constant-Q behavior at middle to high frequencies. Applications include a wavelet-based model of the early stages of acoustic signal processing in the auditory system [Yang et al., 1992], a comparison of the WD and Rihaczek distribution on the response of auditory neurons to wideband noise stimulation [Eggermont & Smith, 1990], and spectrotemporal analysis of dorsal cochlear neurons in the guinea pig [Backoff & Clopton, 1992].


The importance of mammography, x-ray examination of the breast, lies in the early identification of tumors. Kaewlium and Longbotham [1993] used the spectrogram with a Gabor window as a texture discriminator to identify breast masses. Recently, a mammographic feature-enhancement technique using the WT was proposed by Laine et al. [1993]. The wavelet coefficients of the image are modified and then reconstructed to the desired resolution. This technique enhanced the visualization of mammographic features of interest without additional cost or radiation. Magnetic resonance imaging (MRI) allows for the imaging of the soft tissues in the body. Weaver et al. [1992] reduced the long processing time of traditional phase encoding of MRI images by WT encoding. Moreover, unlike phase-encoded images, Gibb’s ringing phenomena and motion artifacts are localized in WT encoded images. The Doppler ultrasound signal is the reflection of an ultrasonic beam due to moving red blood cells and provides information regarding blood vessels and heart chambers. Doppler ultrasound signals in patients with narrowing of the aortic valve were analyzed using the spectrogram by Cloutier et al. [1991]. Guo and colleagues [1994] examined and compared the application of five different time-frequency representations (the spectrogram, short-time AR model, CWD, RID, and Bessel distributions) with simulated Doppler ultrasound signals of the femoral artery. Promising results were obtained from the Bessel distribution, the CWD, and the short-time AR model. Another focus of bioengineering applications of TFRs has concerned the analysis of biologic signals of interest, including the sounds generated by marine mammals, such as dolphins and whales, and the sonar echolocation systems used by bats to locate and identify their prey. The RID was applied to sperm whale acoustic signals by Williams and Jeong [1991] and revealed an intricate time-frequency structure that was not apparent in the original time-series data. The complicated time-frequency characteristics of dolphin whistles were analyzed by Tyack et al. [1992] using the spectrogram, the WD, and the RID, with the RID giving the best results. Flandrin [1988] analyzed the time-frequency structure of the different signals emitted by bats during hunting, navigation, and identifying prey using the smoothed PseudoWD. In addition, the instantaneous frequency of the various signals was estimated using time-frequency representations. Saillant et al. [1993] proposed a model of the bat’s echolocation system using the spectrogram. The most common application of the spectrogram is the analysis and modification of quasistationary speech signals [Rabiner & Schafer, 1978].

Acknowledgments The authors would like to acknowledge the use of the personal notes of Franz Hlawatsch and Antonia Papandreou on TFR kernel constraints as well as the help given by Antonia Papandreou in critiquing the article and its tables.

References Backoff PM, Clopton BM. 1991. A spectrotemporal analysis of DCN single unit responses to wideband noise in guinea pig. Hear Res 53:28. Bartnik EA, Blinowska KJ, Durka PJ. 1992. Single evoked potential reconstruction by means of a wavelet transform. Biol Cybernet 67:175. Boashash B (ed). 1991. Time-Frequency Signal Analysis-Methods and Applications. Melbourne, Australia, Longman-Chesire. Brown ML, Williams WJ, Hero AO. 1994. Non-orthogonal Gabor representation for biological signals. In Proc Intl Conf ASSP, Australia, pp 305–308. Bulgrin JR, Rubal BJ, Thompson CR, Moody JM. 1993. Comparison of short-time Fourier transform, wavelet and time-domain analyses of intracardiac sounds. Biol Sci Instrum 29:465. Cloutier G, Lemire F, Durand L, et al. 1991. Change in amplitude distributions of Doppler spectrograms recorded below the aortic valve in patients with a valvular aortic stenosis. IEEE Trans Biomed Eng 39:502.


Cohen L. 1989. Time-frequency distributions—A review. Proc IEEE 77:941. Claasen TACM, Mecklenbräuker WFG. 1980. The Wigner distribution: A tool for time-frequency signal analysis, parts I–III. Philips J Res 35:217, 35:276, 35:372. Costa A, Boudreaux-Bartels GF. 1994. Design of time-frequency representations using multiform, tiltable kernels. In Proc IEEE-SP Intl Symp T-F and T-S Anal (Pacific Grove, CA). Crowe JA, Gibson NM, Woolfson MS, Somekh MG. 1992. Wavelet transform as a potential tool for ECG analysis and compression. J Biomed Eng 14:268. Daubechies I. 1992. Ten Lectures on Wavelets. Montpelier, Vt, Capital City Press. Eggermont JJ, Smith GM. 1990. Characterizing auditory neurons using the Wigner and Rihacek distributions: A comparison. JASA 87:246. Flandrin P. 1993. Temps-Fréquence. Hermes, Paris, France. Flandrin P. 1988. Time-frequency processing of bat sonar signals. In Nachtigall PE, Moore PWB, (eds), Animal Sonar: Processes and Performance, pp 797–802. New York, Plenum Press. Guo Z, Durand LG, Lee HC. 1994. Comparison of time-frequency distribution techniques for analysis of simulated Doppler ultrasound signals of the femoral artery. IEEE Trans Biomed Eng 41:332. Hlawatsch F, Boudreaux-Bartels GF. 1992. Linear and quadratic time-frequency signal representations. IEEE Sig Proc Mag March:21. Hlawatsch F, Papandreou A, Boudreaux-Bartels GF. 1993. Time-frequency representations: A generalization of the Affine and Hyperbolic classes. In Proc 26th Ann Asil Conf Sig Syst Comput (Pacific Grove, CA). Jeong J, Williams WJ. 1992. Kernel design for reduced interference distributions. IEEE Trans SP 40:402. Jones DL, Tovannas JS, Lander P, Albert DE. 1992. Advanced time-frequency methods for signal averaged ECG analysis. J Electrocardiol 25 (suppl):188. Kadambe S, Murray R, Boudreaux-Bartels GF. 1992. The dyadic wavelet transform based QRS detector. In Proc 26th Ann Asil Conf Sig Syst Comput (Pacific Grove, CA). Kaewlium A, Longbotham H. 1993. Application of Gabor transform as texture discriminator of masses in digital mammograms. Biol Sci Instrum 29:183. Laine A, Schuler S, Fan J. 1993. Mammographic feature enhancement by multiscale analysis. Submitted to IEEE Trans Med Imaging. Li C, Zheng C. 1993. QRS detection by wavelet transform. In Proc Ann Intl Conf IEEE EMBS, pp 330–331. Lin ZY, Chen JDZ. 1994. Time-frequency representation of the electrogastrogram: Application of the exponential distribution. IEEE Trans Biomed Eng 41:267. Morlet D, Peyrin F, Desseigne P, et al. 1993. Wavelet analysis of high resolution signal averaged ECGs in postinfarction patients. J Electrocardiol 26:311. Murray R. 1994. Summary of biomedical applications of time-frequency representations. Technical report no. 0195-0001, Univ. of Rhode Island. Papandreou A, Hlawatsch F, Boudreaux-Bartels GF. 1993. The Hyperbolic class of quadratic timefrequency representations: I. Constant-Q warping, the hyperbolic paradigm, properties, and members. IEEE Trans SP 41:3425. Picard D, Charara J, Guidoin F, et al. 1991. Phonocardiogram spectral analysis simulator of mitral valve prostheses. J Med Eng Technol 15:222. Porat B. 1994. Digital Processing of Random Signals: Theory and Methods. Englewood Cliffs, NJ, Prentice-Hall. Rabiner LR, Schafer RW. 1978. Digital Processing of Speech Signals. Englewood Cliffs, NJ, Prentice-Hall. Rioul O, Vetterli M. 1991. Wavelets and signal processing. IEEE Sig Proc Mag October:14. Saillant PA, Simmons JA, Dear SP. 1993. A computational model of echo processing and acoustic imaging in frequency modulated echo-locating bats: The spectrogram correlation and transformation receiver. JASA 94:2691. Schiff SJ, Milton JG. 1993. Wavelet transforms for electroencephalographic spike and seizure detection. In Proc SPIE—Intl Soc Opt Eng, pp 50–56.


Tuteur FB. 1989. Wavelet transformations in signal detection. In Proc Intl Conf ASSP, pp 1435–1438. Tyack PL, Williams WJ, Cunningham G. 1992. Time-frequency fine structure of dolphin whistles. In Proc IEEE—SP Intl Symp T-F and T-S Anal (Victoria, BC, Canada), pp 17–20. Verreault E. 1989. Détection et Caractérisation des Rales Crépitants (French). PhD thesis, l’Université Laval, Faculte des Sciences et de Genie. Weaver JB, Xu Y, Healy DM, Driscoll JR. 1992. Wavelet encoded MR imaging. Magnet Reson Med 24:275. Widmalm WE, Williams WJ, Zheng C. 1991. Reduced interference time-frequency distributions. In Boashash B (ed), Time frequency distributions of TMJ sounds. J Oral Rehabil 18:403. Williams WJ, Jeong J. 1991. Time-Frequency Signal Analysis—Methods and Applications, pp 878–881. Chesire, England, Longman. Yang Z, Wang K, Shamma S. 1992. Auditory representations of acoustic signals. IEEE Trans Info Theory 38:824. Zaveri HP, Williams WJ, Iasemidis LD, Sackellares JC. 1992. Time-frequency representations of electrocorticograms in temporal lobe epilepsy. IEEE Trans Biomed Eng 39:502.

Further Information Several TFR tutorials exist on the Cohen class [Boashash, 1991; Cohen, 1989; Flandrin, 1993; Hlawatsch and Boudreaux-Bartels, 1992], Affine class [Bertrand chapter in Boashash, 1991; Flandrin, 1993], hyperbolic class [Papandreou et al., 1993], and power class [Hlawatsch et al., 1993]. Several special conferences or issues of IEEE journals devoted to TFRs and the WT include Proc. of the IEEE-SP Time Frequency and Time-Scale Workshop, 1992, IEEE Sig. Proc. Soc.; Special issue on wavelet transforms and multiresolution signal analysis, IEEE Trans. Info. Theory, 1992; and Special issue on wavelets and signal processing, IEEE Trans. SP, 1993. An extended list of references on the application of TFRs to problems in biomedical or bio-engineering can be found in Murray [1994].


Thakor, N.V., Gramatikov, B., Sherman, D. “Wavelet (Time-Scale) Analysis in Biomedical Signal Processing.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

56 Wavelet (Time-Scale) Analysis in Biomedical Signal Processing 56.1 56.2

Introduction The Wavelet Transform: Variable Time and Frequency Resolution The Continuous Wavelet Transform • The Discrete Wavelet Transform

Nitish V. Thakor

56.3

Johns Hopkins School of Medicine

Boris Gramatikov Johns Hopkins School of Medicine

David Sherman Johns Hopkins School of Medicine

A Multiresolution Theory: Decomposition of Signals Using Orthogonal Wavelets Implementation of the Multiresolution Wavelet Transorm: Analysis and Synthesis of Algorithms

56.4 56.5

Further Developments of the Wavelet Transform Applications Cardiac Signal Processing • Neurological Signal Processing • Other Applications

Signals recorded from the human body provide information pertaining to its organs. Their characteristic shape, or temporal and spectral properties, can be correlated with normal or pathological functions. In response to dynamical changes in the function of these organs, the signals may exhibit time-varying as well as nonstationary responses. Time-frequency and time-scale analysis techniques are well suited for such biological signals. Joint time-frequency signal analysis techniques include short-term Fourier transform and Wigner-Ville distribution, and related reduced interference distribution. Joint time-scale analysis includes continuous and discrete, orthonormal and non-orthonormal wavelets. These techniques find applications in the analysis of transient and time-varying events in biological signals. Examples of applications include cardiac signals (for detection of ischemia and reperfusion-related changes in QRS complex, and late potentials in ECG) and neurological signals (evoked potentials and seizure spikes).

56.1 Introduction Digital signal processing uses sophisticated mathematical analysis and algorithms to extract information hidden in signals derived from sensors. In biomedical applications these sensors such as electrodes, accelerometers, optical imagers, etc., record signals from biological tissue with the goal of revealing their health and well-being in clinical and research settings. Refining those signal processing algorithms for biological applications requires building suitable signal models to capture signal features and components that are of diagnostic importance. As most signals of a biological origin are time-varying, there is a special need for capturing transient phenomena in both healthy and chronically ill states.


A critical feature of many biological signals is frequency domain parameters. Time localization of these changes is an issue for biomedical researchers who need to understand subtle frequency content changes over time. Certainly, signals marking the transition from severe normative to diseased states of an organism sometimes undergo severe changes which can easily be detected using methods such as the Short Time Fourier Transform (STFT) for deterministic signals and its companion, the spectrogram, for power signals. The basis function for the STFT is a complex sinusoid, e j2πft, which is suitable for stationary analyses of narrowband signals. For signals of a biological origin, the sinusoid may not be a suitable analysis signal. Biological signals are often spread out over wide areas of the frequency spectrum. Also as Rioul and Vetterli [1] point out, when the frequency content of a signal changes in a rapid fashion, the frequency content becomes smeared over the entire frequency spectrum as it does in the case of the onset of seizure spikes in epilepsy or a fibrillating heartbeat as revealed on an ECG. The use of a narrowband basis function does not accurately represent wideband signals. It is preferred that the basis functions be similar to the signal under study. In fact, for a compact representation using as few basis functions as possible, it is desirable to use basis functions that have a wider frequency spread as most biological signals do. Wavelet theory, which provides for wideband representation of signals [2-4], is therefore a natural choice for biomedical engineers involved in signal processing and currently under intense study [5-9].

56.2 The Wavelet Transform: Variable Time and Frequency Resolution. Continuous Wavelet Transform (CWT) A decomposition of a signal, based on a wider frequency mapping and consequently better time resolution is possible with the wavelet transform. The Continuous Wavelet Transform (CWT) [3] is defined thusly for a continuous signal, x(t),

 t − τ dt a 

(56.1a)

 τ CWTx τ, a = a x at g *  t −  dt  a

(56.1b)

( )

CWTx τ, a =

∫ x(at )g *  a

1

or with change of variable as

( )

∫( )

where g(t) is the mother or basic wavelet, * denotes a complex conjugate, a is the scale factor, and τ—a time shift. Typically, g(t) is a bandpass function centered around some center frequency, fo . Scale a allows the compression or expansion of g(t) [1, 3, 10]. A larger scale factor generates the same function compressed in time whereas a smaller scale factor generates the opposite. When the analyzing signal is contracted in time, similar signal features or changes that occur over a smaller time window can be studied. For the wavelet transform, the same basic wavelet is employed with only alterations in this signal arising from scale changes. Likewise, a smaller scale function enables larger time translations or delays in the basic signal. The notion of scale is a critical feature of the wavelet transform because of time and frequency domain reciprocity. When the scale factor, a, is enlarged, the effect on frequency is compression as the analysis window in the frequency domain is contracted by the amount 1/a [10]. This equal and opposite frequency domain scaling effect can be put to advantageous use for frequency localization. Since we are using bandpass filter functions, a center frequency change at a given scale yields wider or narrower frequency response changes depending on the size of the center frequency. This is the same in the analog or digital filtering theories as “constant-Q or quality factor” analysis [1, 10, 11]. At a given Q or scale factor,


frequency translates are accompanied by proportional bandwidth or resolution changes. In this regard, wavelet transforms are often written with the scale factor rendered as

a=

f f0

(56.2)

or

 f CWTx  τ, a =  = fo  

1 f fo

 t −τ dt . fo 

∫ x(t )g *  f

(56.3)

This is the equivalent to logarithmic scaling of the filter bandwidth or octave scaling of the filter bandwidth for power-of-two growth in center frequencies. Larger center frequency entails a larger bandwidth and vice versa. The analyzing wavelet, g(t), should satisfy the following conditions: (1) belong to L2 (R), i.e., be square integrable (be of finite energy) [2]; (2) be analytic [G(ω) = 0 for ω < 0] and thus be complex-valued. In fact many wavelets are realvalued; however, analytic wavelets often provide valuable phase information [3], indicative of changes of state, particularly in acoustics, speech, and biomedical signal processing [8]; and (3) be admissible. This condition was shown to enable invertibility of the transform [2, 6, 12]:

()

st =

1 cg

∞

∞





1 t −τ 1 ∫ ∫ ( ) a g  a  a da dτ W τ, a

2

− ∞ a>0

where cg is a constant that depends only on g(t) and a is positive. For an analytic wavelet the constant should be positive and convergent: ∞

cg =

∫ 0

()

Gω ω

2

dω < ∞

which in turn imposes an admissibility condition on g(t). For a real-valued wavelet, the integrals from both – ∞ to 0 and 0 to + ∞ should exist and be greater than zero. The admissibility condition along with the issue of reversibility of the transformation is not so critical for applications where the emphasis is on signal analysis and feature extraction. Instead, it is often more important to use a fine sampling of both the translation and scale parameters. This introduces redundancy which is typical for the CWT, unlike for the discrete wavelet transform, which is used in its dyadic, orthogonal, and invertible form. All admissible wavelets with g ∈ L1(R) have no zero-frequency contribution. That is, they are of zero mean, +∞

∫ g (t )dt = 0

−∞

or equivalently G(ω) = 0 for ω = 0, meaning that g(t) should not have non-zero DC [6, 12]. This condition is often being applied also to nonadmissible wavelets.


The complex-valued Morlet’s wavelet is often selected as the choice for signal analysis using the CWT. Morlet’s wavelet [3] is defined as

()

g t = e j 2 πf ot e

−t

2

(56.4a)

2

with its scaled version written as 2 πf o t t j t − g   = e a e 2a 2  a 2

(56.4b)

Morlet’s wavelet insures that the time-scale representation can be viewed as a time-frequency distribution as in Eq. (56.3). This wavelet has the best representation in both time and frequency because it is based on the Gaussian window. The Gaussian function guarantees a minimum time-bandwidth product, providing for maximum concentration in both time and frequency domains [1]. This is the best compromise for a simultaneous localization in both time and frequency as the Gaussian function’s Fourier transform is simply a scaled version of its time domain function. Also the Morlet wavelet is defined by an explicit function and leads to a quasi continuous discrete version [11]. A modified version of Morlet’s wavelet leads to fixed center frequency, fo , with width parameter, σ,

( )

g σ, t = e j 2 πf ot e

2 − t

2σ2

(56.4c)

Once again time-frequency (TF) reciprocity determines the degree of resolution available in time and frequency domains. Choosing a small window size, σ, in the time domain, yields poor frequency resolution while offering excellent time resolution and vice versa [11, 13]. To satisfy the requirement for admissibility and G(0) = 0, a correction term must be added. For ω > 5, this correction term becomes negligibly small and can be omitted. The requirements for the wavelet to be analytic and of zero mean is best satisfied for ω0 = 5.3 [3]. Following the definition in Eq. (56.1a,b) the discrete implementation of the CWT in the time-domain is a set of bandpass filters with complex-valued coefficients, derived by dilating the basic wavelet by the scale factor, a, for each analyzing frequency. The discrete form of the filters for each a is the convolution: k+ n

 i −k 1 S k, a = s i gm ∗ =   a  a i=k − n a

( )

1

∑ () 2

2

n 2

∑ s(k − i) g

i=− n

m

i ∗   a

2

with k = τ/Ts , where Ts is the sampling interval. The summation is over a number of terms, n. Because of the scaling factor a in the denominator of the argument of the wavelet, the wavelet has to be resampled at a sampling interval Ts /a for each scale a. Should the CWT cover a wide frequency range, a computational problem would arise. For example, if we wish to display the CWT over 10 octaves (a change by one octave corresponds to changing the frequency by a factor of 2), the computational complexity (size of the summation) increases by a factor of 210 = 1024. The algorithm by Holschneider et al. [14] solves this problem for certain classes of wavelets by replacing the need to resample the wavelet with a recursive application of an interpolating filter. Since scale is a multiplicative rather than an additive parameter, another way of reducing computational complexity would be by introducing levels between octaves (voices) [15]. Voices are defined to be the scale levels between successive octaves, uniformly distributed in a multiplicative sense [13, 16]. Thus, the ratio between two successive voices is constant. For example,


if one wishes to have ten voices per octave, then the ratio between successive voices is 21/10. The distance between two levels, ten voices apart is an octave. The CWT can also be implemented in the frequency domain. Equation (56.1) may be formulated in the frequency domain as:

( )

∫() ( )

CWT τ, a = a S ω G ∗ aω e j τω dω

(56.5)

where S(ω) and G(ω) denote the Fourier transformed s(t) and g(t), and j = (–1)1/2. The analyzing wavelet g(t) generally has the following Fourier transform:

()

( )

Gτ , a ω = a G aω e jωτ

(56.6a)

The Morlet wavelet [Eq. (56.4a,b)] in frequency domain is a Gaussian function:

()

Gm ω =

1 2ω

e

(

− ω −ω 0

)

2

2

(56.6b)

From Eq. (56.6a,b) it can be seen that for low frequencies, ω, (larger scales a) the width, ∆ω, of the Gaussian is smaller and vice versa. In fact, the ratio ∆ω/ω is constant [1], i.e., Morlet wavelets may be considered filter banks of the constant-Q factor. Based on Eqs. (56.5 and 56.6a,b) the wavelet transform can be implemented in the frequency domain. At each scale, the Fourier image of the signal can be computed as

( ) () ( )

Y ω, a = S ω • Gm ω, a

with S(ω) being the Fourier transform of the signal, Gm(ω, a) being the scaled Fourier image of the Morlet wavelet at scale a, and the operation ● standing for element-by-element multiplication (windowing in frequency domain). The signal at each scale a will finally be obtained by applying the inverse Fourier transform:

( ) ( ) Y (ω , a )

CWT τ, a = FFT

−1

This approach has the advantage of avoiding computationally intensive convolution of time-domain signals by using multiplication in the frequency domain, as well as the need of resampling the mother wavelet in the time domain [17, 18]. Note that the CWT is, in the general case, a complex-valued transformation. In addition to its magnitude, its phase often contains valuable information pertinent to the signal being analyzed, particularly in instants of transients [3]. Sometimes the TF distribution of the nonstationary signal is much more important. This may be obtained by means of real-valued wavelets. Alternatives to the complexvalued Morlet wavelet are simpler, real-valued wavelets that may be utilized for the purpose of the CWT. For example, the early Morlet wavelet, as used for seismic signal analysis [19], had the following realvalued form:

()

( )

g t = cos 5t e − t

2

2

It had a few cycles of a sine wave tapered by a Gaussian envelope. Though computationally attractive, this idea contradicts the requirement for an analytic wavelet, i.e., its Fourier transform G(ω) = 0 for


ω < 0. An analytic function is generally complex-valued in the time domain and has its real and imaginary parts as Hilbert transforms of each other [2, 20]. This guarantees only positive-frequency components of the analyzing signal. A variety of analyzing wavelets have been proposed in recent years for time-scale analysis of the ECG. For example, Senhadji et al. [21] applied a pseudo-Morlet’s wavelet to bandpass filtering to find out whether some abnormal ECG events like extrasystoles and ischemia are mapped on specific decomposition levels:

() (

(

))

g t = C 1 + cos 2πf0t e −2iπkf 0t , t ≤ 1 2 f0

and

{

}

k integer ∉ −1, 0,1

with the product kf0 defining the number of oscillations of the complex part, and C representing a normalizing constant such that ||g|| = 1. The above function is a modulated complex sine wave that would yield complex-valued CWT including phase information. However its envelope is a cosine, rather than a Gaussian, as in the case of the complex Morlet wavelet. It is well known that strictly the Gaussian function (both in time and frequency domain) guarantees the smallest possible time-bandwidth product which means maximum concentration in the time and frequency domains [22]. The STFT has the same time-frequency resolution regardless of frequency translations. The STFT can be written as ∞

( ) ∫ x(t ) g ∗ (t − τ)e

STFT τ, f =

−2 π jft

dt

(56.7)

−∞

where g(t) is the time window that selects the time interval for analysis or otherwise known as the spectrum localized in time. Figure 56.1 shows comparative frequency resolution of both the STFT as well as the wavelet transform. The STFT is often thought to be analogous to a bank of bandpass filters, each shifted by a certain modulation frequency, fo . In fact, the Fourier transform of a signal can be interpreted as passing the signal through multiple bandpass filters with impulse response, g(t)ej2πft, and then using complex demodulation to downshift the filter output. Ultimately, the STFT as a bandpass filter rendition simply translates the same low pass filter function through the operation of modulation. The characteristics of the filter stay the same though the frequency is shifted. Unlike the STFT, the wavelet transform implementation is not frequency independent so that higher frequencies are studied with analysis filters with wider bandwidth. Scale changes are not equivalent to

FIGURE 56.1 Comparative frequency resolution for Short Time Fourier Transform (STFT) and Wavelet Transform (WT). Note that frequency resolution of STFT is constant across frequency spectrum. The WT has a frequency resolution that is proportional to the center frequency of the bandpass filter.


varying modulation frequencies that the STFT uses. The dilations and contractions of the basis function allow for variation of time and frequency resolution instead of uniform resolution of the Fourier transform [15]. Both the wavelet and Fourier transform are linear Time-Frequency Representations (TFRs) for which the rules of superposition or linearity apply [10]. This is advantageous in cases of two or more separate signal constituents. Linearity means that cross-terms are not generated in applying either the linear TF or time-scale operations. Aside from linear TFRs, there are quadratic TF representations which are quite useful in displaying energy and correlation domain information. These techniques, also described elsewhere in this volume include the Wigner-Ville distribution (WVD), smoothed WVD, the reduced inference distribution (RID), etc. One example of the smoothed Wigner-Ville distribution is

 1   1   τ W t , f = s *  t − τ e − jτ 2 πf s *  t + τ h   dτ  2   2   2

( ) ∫

(56.8)

where h(t), is a smoothing function. In this case the smoothing kernel for the generalized or Cohen’s class of TFRs is

 τ φ t , τ = h  δ t .  2

( )

()

These methods display joint TF information in such a fashion as to display rapid changes of energy over the entire frequency spectrum. They are not subject to variations due to window selection as in the case of the STFT. A problematic area for these cases is the elimination of those cross-terms that are the result of the embedded correlation. It is to be noted that the scalogram or scaled energy representation for wavelets can be represented as a Wigner-Ville distribution as [1]





( ) = ∫∫ W (u, n)W ∗  u a− t , an dudn

CWTx τ, a

2

x

g

(56.9)

where

 1   1  W x t , f = x ∗  t − τ e − jτ 2 πf x *  t + τ dτ  2   2 

( ) ∫

The Discrete Wavelet Transform In the discrete TFRs both time and scale changes are discrete. Scaling for the discrete wavelet transform involves sampling rate changes. A larger scale corresponds to subsampling the signal. For a given number of samples a larger time swath is covered for a larger scale. This is the basis of signal compression schemes as well [23]. Typically, a dyadic or binary scaling system is employed so that given a discrete wavelet function, ψ(x), is scaled by values that are binary. Thus

()

( )

ψ2j t = 2 j ψ 2 j t

(56.10)

where j is the scaling index and j = 0, –1, –2, …. In a dyadic scheme, subsampling is always decimationin-time by a power of 2. Translations in time will be proportionally larger as well as for a more sizable scale.


It is for discrete time signals that scale and resolution are related. When the scale is increased, resolution is lowered. Resolution is strongly related to frequency. Subsampling means lowered frequency content. Rioul and Vetterli [1] use the microscope analogy to point out that smaller scale (higher resolution) helps to explore fine details of a signal. This higher resolution is apparent with samples taken at smaller time intervals.

56.3 A Multiresolution Theory: Decomposition of Signals Using Orthogonal Wavelets One key result of the wavelet theory is that signals can be decomposed in a series of orthogonal wavelets. This is similar to the notion of decomposing a signal in terms of discrete Fourier transform components or Walsh or Haar functions. Orthogonality insures a unique and complete representation of the signal. Likewise the orthogonal complement provides some measure of the error in the representation. The difference in terms of wavelets is that each of the orthogonal vector spaces offers component signals with varying levels of resolution and scale. This is why Mallat [24] named his algorithm the multiresolution signal decomposition. Each stage of the algorithm generates wavelets with sequentially finer representations of signal content. To achieve an orthogonal wavelet representation, a given wavelet function, φ(t), at a scaling index level equal to zero, is first dilated by the scale coefficient 2j, then translating it by 2–jn and normalizing by 2 – j gives:

(

2− j φ2 j t − 2− j n

)

The algorithm begins with an operator A2j for discrete signals that takes the projections of a signal, f(t) onto the orthonormal basis, V2j :

()

A2 j f t = 2

∞

−j

∑ f (u),φ (u − 2 n) φ (t − 2 n) −j

−j

2j

2j

(56.11)

n=−∞

where 2j defines the level of resolution. A2j is defined as the multi-resolution operator that approximates a signal at a resolution 2j. Signals at successively lower resolutions can be obtained by repeated application of the operator A2 j (–J ≤ j ≤ –1), where J specifies the maximum resolution, such that A2j f(x) is the closest approximation of function f(x) at resolution 2j. Here we note that < > is simply a convolution defined thusly,

() (

) ∫ f (u)φ(u − 2 n)du

f u ,φ2 j u − 2− j n =

∞

−j

−∞

(56.12)

Here φ(x) is the impulse response of the scaling function. The Fourier transforms of these functions are lowpass filter functions with successively smaller halfband lowpass filters. This convolution synthesizes the coarse signal at a resolution/scaling level j:

() (

C2 j f = f t , φ2 j t − 2− j n

)

(56.13)

Each level j generates new basis functions of the particular orthonormal basis with a given discrete approximation. In this case, larger j provides for decreasing resolution and increasing the scale in proportional fashion for each level of the orthonormal basis. Likewise, each sequentially larger j provides for time shift in accordance with scale changes, as mentioned above, and the convolution or inner product operation generates the set of coefficients for the particular basis function. A set of scaling functions at decreasing levels of resolution, j = 0, –1, –2, …, –6 is given in [25].


The next step in the algorithm is the expression of basis function of one level of resolution, φ2 j by at a higher resolution, φ2 j+1 . In the same fashion as above, an orthogonal representation of the basis V2 j in terms of V2 j+1 is possible, or

(

∞

)

φ2 j t − 2− j n = 2− j −1

∑ φ (u − 2 n), φ (u − 2 k) φ (t − 2 k) −j

2 j +1

2j

− j −1

2 j +1

− j −1

(56.14)

k=−∞

Here the coefficients are once again the inner products between the two basis functions. A means of translation is possible for converting the coefficients of one basis function to the coefficients of the basis function at a higher resolution:

() (

)

C2 j f = f u , φ2 j t − 2− j n = ∞

2

− j −1

∑ φ (u − 2 n), φ (u − 2 k) f (u), φ (t − 2 k) −j

− j −1

2 j +1

2j

2 j +1

(56.15)

− j −1

k=−∞

Mallat [24] also conceives of the filter function, h(n), whose impulse response provides this conversion, namely,

() (

∞

)

C2 j f = f u , φ2 j t − 2− j n = 2− j −1

∑ h˜(2n − k) f (u), φ (t − 2 k) 2 j +1

− j −1

(56.16)

k=−∞

˜ where h(n) = 2–j–1 〈φ2 j (u – 2–j n), φ2 j+1 (u – 2–j–1 k)〉 and h(n) = h(–n) is the impulse response of the appropriate mirror filter. Using the tools already described, Mallat [24] then proceeds to define the orthogonal complement, O2j to the vector space V2 j at resolution level j. This orthogonal complement to V2 j is the error in the approximation of the signal in V2 j+1 by use of basis function belonging to the orthogonal complement. The basis functions of the orthogonal complement are called orthogonal wavelets, ψ(x), or simply wavelet functions. To analyze finer details of the signal, a wavelet function derived from the scaling function is selected. The Fourier transform of this wavelet function has the shape of a bandpass filter in frequency domain. A basic property of the function ψ is that it can be scaled according to

()

( )

ψ2j t = 2 j ψ 2 j t

An orthonormal basis set of wavelet functions is formed by dilating the function ψ(x) with a coefficient 2j and then translating it by 2–jn, and normalizing by 2 – j. They are formed by the operation of convolving the scale function with the quadrature mirror filter

 ω  ω ψ ω = G  φ    2  2

()

—

where G(ω) = e–jω H (ω + π) is the quadrature mirror filter transfer response and g(n) = (–1)1–n h(1 – n) is the corresponding impulse response function. The set of scaling and wavelet functions presented here form a duality, together resolving the temporal signal into coarse and fine details, respectively. For a given level j then, this detail signal can once again be represented as a set of inner products: © 2000 by CRC Press LLC

() (

D2 j f = f t , ψ 2 j x − 2− j n

)

(56.17)

For a specific signal, f(x), we can employ the projection operator as before to generate the approximation to this signal on the orthogonal complement. As before, the detail signal can be decomposed using the higher resolution basis function:

() (

)

D2 j f = f u , ψ 2 j t − 2− j n = ∞

2

− j −1

∑ ψ (t − 2 n), φ (u − 2 k) f (u), φ (t − 2 ) −j

2j

− j −1

2 j +1

2 j +1

− j −1

k Y′

(56.18)

k=−∞

or in terms of the synthesis filter response for the orthogonal wavelet

() (

)

D2 j f = f u , ψ 2 j t − 2− j n = 2− j −1

∞

∑ g˜(2n − k) f (u), φ (t − 2 k) − j −1

k=−∞

2 j +1

(56.19)

At this point, the necessary tools are here for a decomposition of a signal in terms of wavelet components, coarse and detail signals. Multiresolution wavelet description provides for the analysis of a signal into lowpass components at each level of resolution called coarse signals through the C operators. At the same time the detail components through the D operator provide information regarding bandpass components. With each decreasing resolution level, different signal approximations are made to capture unique signal features. Procedural details for realizing this algorithm follow.

Implementation of the Multiresolution Wavelet Transform: Analysis and Synthesis of Algorithms A diagram of the algorithm for the multiresolution wavelet decomposition algorithm is shown in Fig. 56.2. A step-by-step rendition of the analysis is as follows: 1. Start with N samples of the original signal, x(t), at resolution level j = 0. 2. Convolve signal with original scaling function, φ(t), to find C1 f as in Eq. (56.13) with j = 0. 3. Find coarse signal at successive resolution levels, j = –1, –2, …, –J through Eq. (56.16); Keep every other sample of the output.

FIGURE 56.2 Flow chart of multiresolution algorithm showing how successive coarse and detail components of resolution level, j, are generated from higher resolution level, j + 1.


4. Find detail signals at successive resolution levels, j = –1, –2, …, –J through Eq. (56.19); Keep every other sample of the output. 5. Decrease j and repeat steps 3 through 5 until j = –J Signal reconstruction details are presented in [24, 26].

56.4 Further Developments of the Wavelet Transform Of particular interest are Daubechies’ orthonormal wavelets of compact support with the maximum number of vanishing moments [27, 28]. In these works, wavelets of very short length are obtained (number of coefficients 4, 6, 8, etc.) thus allowing for very efficient data compression. Another important development of the multiresolution decomposition is the wavelet packet approach [29]. Instead of the simple multiresolution three-stage filter bank (low-pass/highpass/downsampling) described previously in this chapter, wavelet packets analyse all subsystems of coefficients at all scale levels, thus yielding the full binary tree for the orthogonal wavelet transform, resulting in a completely evenly spaced frequency resolution. The wavelet packet system gives a rich orthogonal structure that allows adaptation to particular signals or signal classes. An optimal basis may be found, minimizing some function on the signal sequences, on the manifold of orthonormal bases. Wavelet packets allow for a much more efficient adaptive data compression [30]. Karrakchou and Kunt report an interesting method of interference cancelling using a wavelet packet approach [8]. They apply ‘local’ subband adaptive filtering by using a non-uniform filter bank, having a different filter for each subband (scale). Furthermore, the orthonormal basis has “optimally” been adapted to the spectral contents of the signal. This yields very efficient noise suppression without significantly changing signal morphology.

56.5 Applications Cardiac Signal Processing Signals from the heart, especially the ECG, are well suited for analysis by joint time-frequency and timescale distributions. That is because the ECG signal has a very characteristic time-varying morphology, identified as the P-QRS-T complex. Throughout this complex, the signal frequencies are distributed: (1) low frequency—P and T waves, (2) mid to high frequency—QRS complex [31, 32]. Of particular diagnostic importance are changes in the depolarization (activation phase) of the myocardium, represented by the QRS complex. These changes cause alterations in the propagation of the depolarization wave detectable by time-frequency analysis of one-dimensional electrocardiographic signals recorded from the body surface. Two examples are presented below. Analysis of the QRS Complex Under Myocardial Ischemia Ischemia-related intra-QRS changes. When the heart muscle becomes ischemic or infarcted, characteristic changes are seen in the form of elevation or depression of the ST-segment. Detection of these changes requires an extension of the signal bandwidth to frequencies down to 0.05 Hz and less, making the measurements susceptible to motion artifact errors. Ischemia also causes changes in conduction velocity and action potential duration, which results in fragmentation in the depolarization front (Fig. 56.3) and appearance of low-amplitude notches and slurs in the body surface ECG signals. These signal changes are detectable with various signal processing methods [33, 34]. Depolarization abnormalities due to ischemia may also cause arrhythmogenic reentry [35], which is one more reason to detect intra-QRS changes precisely. Identification of ischemia-related changes in the QRS complex is not as well known, and interpretation of the QRS complex would be less susceptible to artifactual errors, as compared to the ST analysis. Thus, time-frequency or time-scale analysis would serve a useful function in localizing the ischemia-related changes within the QRS complex, but would be somewhat independent of the artifactual errors.


FIGURE 56.3 Idealized example of (a) normal propagation and (b) wavefront fragmentation due to an ischemic zone of slow conduction. The superimposed lines are isochrones, connecting points at which the depolarization arrives at the same time.

Experimental findings. In experimental animals, the response of the heart to coronary artery occlusion and then reperfusion was studied. The Left Anterior Descending (LAD) branch of the coronary artery was temporarily occluded for 20 min. Subsequent to that, the occlusion was removed, and resulting reperfusion more or less restored the ECG signal after 20 min. The coronary artery was occluded a second time for 60 min, and once again occlusion was removed and blood flow was restored. Single ECG cycles were analyzed using the continuous wavelet transform [33]. Figure 56.4 shows the time-scale plots for the ECG cycles for each of the five stages of this experiment. The three-dimensional plots give time in the P-QRS-T complex on one axis, the scale (or equivalent frequency) on another axis, and the normalized magnitude on the third axis. First occlusion results in a localized alteration around 100 ms and the midscale, which shows up as a bump in the three-dimensional plot or a broadening in the contour plot. Upon reperfusion, the time-scale plot returns to the pre-occlusion state. The second occlusion brings about a far more significant change in the time-scale plot, with increased response in the 0–200 ms and mid-scale ranges. This change is reversible. We were thus able to show, using time-scale technique, ischemia related changes in the QRS complex, and the effects of occlusion as well as reperfusion. Potential role of ischemia-related intra-QRS changes in coronary angioplasty. The above results are also applicable to human ECGs and clinical cardiology. For example, a fairly common disorder is the occlusion of coronary vessels, causing cardiac ischemia and eventually infarction. An effective approach to the treatment of the occlusion injury is to open the coronary blood vessels using a procedure called coronary angioplasty (also known as percutaneous transluminal coronary angioplasty or PTCA). Vessels may be opened using a balloon-type or a laser-based catheter. When reperfusion occurs following the restoration of the blood flow, initially a reperfusion injury is known to occur (which sometimes leads to arrhythmias) [35]. The ST level changes as well, but its detection is not easy due to artifacts, common in a PTCA setting. In a clinical study, we analyzed ischemia and reperfusion changes before and after the PTCA procedure. Short-term occlusion and ischemia followed by reperfusion were carried out in a cardiac catheterization laboratory at the Johns Hopkins Hospital in connection with PTCA) [36]. Figure 56.5 shows time-scale plots of a patient derived from continuous wavelet transform. Characteristic midscale hump in the early stages of the QRS cycle is seen in the three-dimensional time-scale plot. Then, 60 min after angioplasty, the normal looking time-scale plot of the QRS complex is restored in this patient. This study suggests that time-scale analysis and resulting three-dimensional or contour plots may be usable in monitoring the effects of ischemia and reperfusion in experimental or clinical studies. In another study (4 patients, LAD) we monitored for intra-QRS changes during PTCA. Despite signal noise and availability



FIGURE 56.4 Time-frequency distributions of the vector magnitude of two ECG leads during five stages of a controlled animal experiment. The frequency scale is logarithmic, 16 to 200 Hz. The z-axis represents the modulus (normalized) of the complex wavelet-transformed signal.

FIGURE 56.5 Time-frequency distributions of human ECG study using WT. Pre-angioplasty plot (a) shows a characteristic hump at about 35 Hz, which disappears as indicated by the second plot (b) taken one hour after angioplasty treatment.

of recordings only from limb leads, superimposed mid-frequency components during ischemic states of the heart were observed, which disappeared when perfusion was restored. There was at least one lead that responded to changes in coronary perfusion. Figure 56.6 shows five different stages of a PTCA procedure as time plots (lead I) and CWT TFDs (topo-plots). Despite the presence of noise, the WT was able to unveil elevation of intra-QRS time-frequency components around 20 Hz during balloon inflation (ischemia), and a drop in the same components with reperfusion after balloon deflation. Frequency components 20 to 25 Hz during inflation. No substantial ST changes can be observed in the time-domain plot. The arrows show the zones of change in TFDs with ischemia and reperfusion. Note the representation of power line interference (50 Hz) as “clouds” in (b) and (d) topo plots—far from the region of interest. Another study analyzed ECG waveforms from patients undergoing the PTCA procedure by the multiresolution wavelet method, decomposing the whole P-QRS-T intervals into coarse and detail components [26, 37], as can be seen from the analysis of one pre-angioplasty ECG cycle in Fig. 56.7 [26]. The PTCA procedure results in significant morphological and spectral changes within the QRS complex. It was found that certain detail components are more sensitive than others: in this study, the detail components d6 and d5 corresponding to frequency band of 2.2 to 8.3 Hz are most sensitive to ECG changes following a successful PTCA procedure. From this study it was concluded that monitoring the energy of ECG signals at different detail levels may be useful in assessing the efficacy of angioplasty procedures [37]. A benefit of this approach is that a real-time monitoring instrument for the cardiac catheterization laboratory can be envisioned (whereas currently X-ray fluroscopy is needed). Detection of reperfusion during thrombolytic therapy. Detecting reperfusion-related intra-QRS changes, along with ST changes, in the time-frequency domain would possibly find application in thrombolysis monitoring after myocardial infarction. At present, the ST elevation and its recovery are the main electrocardiographic indicators of acute coronary ischemia and reperfusion. Reports using continuous ST-segment monitoring have indicated that 25 to 50% of patients treated with intravenous thrombolytic therapy display unstable ST recovery [38], and additional reperfusion indicators are necessary. Signal averaged ECG and highest frequency ECG components (150 to 250 Hz) have been utilized as a reperfusion marker during thrombolysis and after angioplasty, but their utility is uncertain, since the degree of change of the energy values chosen does not appear to be satisfactory [39, 40]. We have analyzed the QRS of the vector magnitude V = X 2 + Y 2 + Z 2 of body surface orthogonal ECG leads X, Y and Z during thrombolytic



FIGURE 56.6 Time domain signals (lead I) during PTCA on the LAD (upper plots) and time-frequency distributions TFD (lower plots) during (a) baseline, (b) at the end of a 3 min inflation (7 at), (c) 10 min. after first inflation, reducing stenosis from 95 to 60%, (d) at the end of a 20 min inflation (6 at), (e) 10 min after first inflation, reducing stenosis from 95 to 60%. The arrows show the zones of change in TFDs with ischemia/reperfusion.

FIGURE 56.7 Detail and coarse components from one ECG cycle. The coarse components represent the lowpass filtered versions of the signal at successive scales. Detail components, d1 and d2, consist mainly of electrical interference.

therapy of two patients with myocardial infarction. Figure 56.8 shows how TFDs may be affected by reperfusion during thrombolysis. Two interesting trends may be observed on this figure: (1) a midfrequency peak present during initial ischemia (a) disappears two hours after start of thrombolytic therapy (b) due to smoother depolarization front, and (2) high-frequency components appear with reestablished perfusion, possibly due to faster propagation velocity of the depolarization front. Analysis of Late Potentials in the ECG “Late potentials” are caused by fractionation of the depolarization front after myocardial infarction [35, 41]. They have been shown to be predictive of life threatening reentrant arrhythmias. Late potentials occur in the terminal portion of the QRS complex and are characterized by small amplitude and higher frequencies than in the normal QRS complex. The presence of late potentials may indicate underlying dispersion of electrical activity of the cells in the heart, and therefore may provide a substrate for production of arrhythmias. The conventional Fourier transform does not readily localize these features in time and frequency [15, 42]. STFT is more useful because the concentration of signal energy at various times in the cardiac cycle is more readily identified. The STFT techniques suffer from the problem of selecting a proper window function; for example, window width can affect whether high temporal or high spectral resolution is achieved [15]. Another approach sometimes considered is the Wigner-Ville distribution, which also produces a composite time-frequency distribution. However, the Wigner-Ville distribution suffers from the problem of interference from cross-terms. Comparative representations by smoothed Wigner-Ville, wavelet transform (scalogram), and traditional spectrogram are illustrated in Fig. 56.9. This problem causes high levels of signal power to be seen at frequencies not representing the original signal; for example, signal energy contributed at certain frequencies by the QRS complex may mask the signal energy contributed by the late potentials. In this regard, wavelet analysis methods provide a more accurate picture of the localized time-scale features indicative of the late potentials [11, 15, 43–45]. Figure 56.10 shows that the signal energies at 60 Hz and beyond are localized in the late stage of the QRS © 2000 by CRC Press LLC

FIGURE 56.8 Change of the time-frequency distributions (QRS complex) of the vector magnitude of orthogonal leads X, Y and Z during thrombolysis: (a) 5 minutes after start, and (b) 2 hr after initiation of therapy. A mid-frequency peak has disappeared due to smoother depolarization, and high-frequency components appear due to faster propagation.

and into the ST-segment. For comparison, the scalogram from a healthy person is illustrated in Fig. 56.11. This spreading of the high frequencies into the late cycle stages of the QRS complex is a hallmark of the late potentials. Time-scale analysis of late potentials may therefore serve as a noninvasive diagnostic tool for predicting the likelihood of life threatening arrhythmias in the heart. Role of Wavelets in Arrhythmia Analysis Detection of altered QRS morphology. Further applications to the generalized field of arrhythmia classification can be envisaged [46]. When arrhythmias, such as premature ventricular contractions (PVCs) and tachycardia do occur, the P-QRS-T complex undergoes a significant morphological change. Abnormal beats, such as PVCs, have different time-scale signatures than normal beats. Often the QRS complex may widen and sometimes invert with PVCs. As the QRS complex widens, its power spectrum shows diminished contribution at higher frequencies and these are spread out over a wider body of the signal [47, 48]. This empirical description of the time-domain features of the ECG signal lends itself particularly well to analysis by time-frequency and time-scale techniques. A more challenging problem is to distinguish multiform PVCs. Use of the orthogonal wavelet decomposition to separate dynamical activities embedded in a time series. Another interesting application of the discrete wavelet transform to arrhythmia research is the dynamical analysis of the ECG before and during ventricular tachycardia and ventricular fibrillation. Earlier works on heart rate variability have shown that heart rhythm becomes rigidly constant prior to the onset of life threatening arrhythmias, whereas the correlation dimension as a measure of randomness increases to values above 1 during disorganized rhythms like fast ventricular tachycardia, ventricular flutter, and ventricular fibrillation. This fact was used to identify patients at risk, and is promising with regard to prediction of arrhythmic episodes. Encouraging results were recently obtained by combining multiresolution wavelet analysis and dynamical analysis, in an attempt to find a decorrelated scale best projecting changes in low-dimensional dynamics [49]. The authors used records 115 and 207 from the MIT © 2000 by CRC Press LLC

FIGURE 56.9 Comparison of time-frequency representations of sinusoids with specific on-off times: (a) shows true time-frequency representation of 40 and 60 Hz sinusoids, (b) shows representation of smoothed Wigner-Ville transform, (c) spectrogram representation, (d) shows wavelet transform version of signal.

Arrhythmia Database to study nonlinear dynamics preceding ventricular flutter. Distinct changes in the correlation dimension (D2 ≈ 3) in frequency band 45 to 90 Hz were observed before the onset of arrythmia, indicating the presence of underlying low-dimensional activity.

Neurological Signal Processing Evoked Potentials Evoked potentials are the signals recorded from the brain in response to external stimulation. Evoked responses can be elicited by electrical stimulation (somatosensory evoked response), visual stimulation (visual evoked response), or auditory stimulation (brainstem auditory evoked response). Usually the signals are small, while the background noise, mostly the background EEG activity, is quite large. The low signal-to-noise ratio (SNR) necessitates use of ensemble averaging, sometimes signal averaging as many as a thousand responses [50]. After enhancing the SNR, one obtains a characteristic wave pattern that includes the stimulus artifact and an undulating pattern characterized by one or more peaks at specific latencies beyond the stimulus. Conventionally, the amplitude and the latency of the signal peaks is used in arriving at a clinical diagnosis. However, when the signals have a complex morphology, simple amplitude and latency analysis does not adequately describe all the complex changes that may occur as a result of brain injury or disease. Time-frequency and wavelet analysis have been shown to be useful in identifying the features localized within the waveform that are most indicative of the brain’s response [51]. In one recent experimental study, we evaluated the somatosensory evoked response from experimental animals in whom injury was caused by oxygen deprivation. The evoked response signal was decomposed © 2000 by CRC Press LLC

FIGURE 56.10 Healthy person: (a) first recorded beat, (b) 3-D representation of the modified WT for the first beat, (c) contour plot of the modified WT for the first beat, and (d) contour plot of the modified WT for the second beat.

into its coarse and detail components with the aid of the multiresolution wavelet analysis technique (Fig. 56.12) [25]. The magnitude of the detail components was observed to be sensitive to the cerebral hypoxia during its early stages. Figures 56.13a and 56.13b show a time trend of the magnitude of the detail components along with the trend of the amplitude and the latency of the primary peak of the somatosensory evoked response. The experimental animal was initially challenged by nitrous gas mixture with 100% oxygen (a non-injury causing event), and late by inspired air with 7 to 8% oxygen. As expected, the amplitude trend shows an initial rise because of the 100% oxygen, and later a gradual decline in response to hypoxia. The magnitude of the detail component shows a trend more responsive to injury: while there is not a significant change in response to the non-injury causing event, the magnitude of the detail component d4 drops quite rapidly when the brain becomes hypoxic. These data suggest that detail components of the evoked response may serve as indicators of early stages of brain injury. Evoked response monitoring can be useful in patient monitoring during surgery and in neurological critical care [52, 53]. Other applications include study of cognitive or event-related potentials in human patients for normal cognitive function evaluation or for assessment of clinical situations, such as a response in Alzheimer’s disease [54]. Proper characterization of evoked responses from multiple channel recordings facilitates localization of the source using the dipole localization theory [55]. EEG and Seizures Electroencephalographic signals are usually analyzed by spectrum analysis techniques, dividing the EEG signal into various arbitrary frequency bands (α, β, θ, δ). Conventional spectrum analysis is useful when these events are slowly unfolding, as when a person goes to sleep, the power in the EEG shifts from higher to lower frequency bands. However, when transient events such as epileptic seizures occur, there are often sharp spikes or a bursting series of events in the recorded waveform. This form of the signal, that is temporally well localized and has a spectrum that is distinctive from normal or ongoing events, lends


FIGURE 56.11 Patient with ventricular tachycardia diagnosis: (a) first beat, (b) 3-D representation of the modified WT for the first beat, (c) contour plot of the modified WT for the first beat, and (d) contour plot of the modified WT for the second beat.

itself to wavelet analysis. A patient’s EEG recorded over an extended period, preceding and following the seizure, was recorded and analyzed. Figure 56.14 shows a short segment of the EEG signal with a seizure burst. The multiresolution wavelet analysis technique was employed to identify the initiation of the seizure burst. Figure 56.15 shows a sudden burst onset in the magnitude of the detail components when the seizure event starts. The bursting subsides at the end of the seizure, as seen by a significant drop in the magnitude of the detail components. Wavelet analysis, thus, may be employed for the detection of onset and termination of seizures. Further possibilities exist in the use of this technique for discriminating interictal spikes and classifying them [56]. Certain seizures, like the petit mal and the grand mal epilepsy seizures, have very characteristic morphologies (e.g., spike and dome pattern). These waveforms would be expected to lend themselves very well to wavelet analysis.

Other Applications Wavelet, or time-scale analysis is applicable to problems in which signals have characteristic morphologies or equivalently differing spectral signature attributed to different parts of the waveform, and the events of diagnostic interest are well localized in time and scale. The examples of such situations and applications are many. In cardiac signal processing, there are several potential applications. Well localized features of ECG signals, such as the P-QRS-T lend themselves well to wavelet analysis [57]. The application of wavelet analysis to ischemic-reperfusion injury changes and the late potentials has been illustrated above. This idea has been extended to the study of body surface maps recorded using numerous electrodes placed on the chest. In a preliminary study [58], spatio-temporal maps can been constructed and interpreted using time-scale analysis techniques.


FIGURE 56.12 Coarse (a) and detail (b) components from somatosensory evoked potentials during normal, hypoxic, and reoxygenation phases of experiment.

In many situations noise and artifact result in inaccurate detection of the QRS complex. Wavelet analysis may prove to be helpful in removal of electrical interference from ECG [59]. Wavelet techniques have successfully been used in removing the noise from functional MRI data [8]. A more challenging application would be in distinguishing artifact from signal. Since wavelet analysis naturally decomposes the signals at different scales at well localized times, the artifactual events can be localized and eliminated. Fast computational methods may prove to be useful in real-time monitoring of ECG signal at a bedside or in analysis of signals recorded by Holter monitors. Other cardiovascular signals, such as heart sounds may be analyzed by time-frequency or time-scale analysis techniques. Characteristic responses to various normal and abnormal conditions along with sounds that are well localized in time and scale make these signals good candidates for wavelet analysis [60]. Normal patterns may be discriminated from pathological sound patterns, or sounds from various blood vessels can be identified [61]. Blood pressure waveform similarly has a characteristic pattern amenable to time-scale analysis. The dicrotic notch of the pressure waveform results from blood flow through the valves whose opening and closing affects the pressure signal pattern. The dicrotic notch can be detected by wavelet analysis [62]. Sounds from the chest, indicative of respiratory patterns are being investigated using wavelet techniques. Applications include analysis of respiratory patterns of infants [63] and respiration during sleep [64]. Two applications in neurological signal processing, evoked response and seizure detection, are described above. Other potential applications include detection and interpretation of signals from multiple neurons obtained using microelectrodes [65]. Since waveforms (called spikes) from individual


FIGURE 56.13 (a) Amplitude and latency of major evoked potential peak during control, hypoxic, and reoxygenation portions of experiment; (b) mean amplitude of respective detail components during phases of experiment.

neurons (called units) have different patterns because of their separation and orientation with respect to the recording microelectrode, multiunit spike analysis becomes a challenging problem. Time-scale analysis techniques may be employed to localize and analyze the responses of individual units and from that derive the overall activity and interrelation among these units so as to understand the behavior of neural networks. An analogous problem is that of detecting and discriminating signals from “motor units,” that is, the muscle cells and fibers [66]. Signals from motor units are obtained by using small, micro or needle electrodes, and characteristic spike trains are obtained, which can be further analyzed by time-scale analysis techniques to discriminate normal and abnormal motor unit activity.

Discussion and Conclusions Biological signals with their time-varying nature and characteristic morphologies and spectral signatures are particularly well suited for analysis and interpretation using time-frequency and time-scale analysis techniques. For example, the P-QRS-T complex of the ECG signal shows localized low frequencies in the P- and the ST-segments and high frequencies in the QRS complex. In time-scale frame, the ischemia related changes are seen in certain detail components of the QRS complex. The late segment of the QRS cycle exhibits the so-called late potentials more easily localized by means of time-scale analysis. Other cardiovascular signals, such as pressure waves, heart sounds, and blood flow are being analyzed by the newly developed wavelet analysis algorithms. Other examples of time-scale analysis include neurological


FIGURE 56.14

FIGURE 56.15


Example of epilepsy burst.

Localization of burst example with wavelet detail components.

signals with potential applications in the analysis of single and multiunit recordings from neurons, evoked response, EEG, and epileptic spikes and seizures. The desirable requirements for a successful application of time-scale analysis to biomedical signals is that events are well localized in time and exhibit morphological and spectral variations within the localized events. Objectively viewing the signal at different scales should provide meaningful new information. For example, are there fine features of signal that are observable only at scales that pick out the detail components? Are there features of the signal that span a significant portion of the waveform so that they are best studied at a coarse scale. The signal analysis should be able to optimize the trade-off between time and scale, i.e., distinguish short lasting events and long lasting events. For these reasons, the signals described in this article have been found to be particularly useful models for data analysis. However, one needs to be cautious in using any newly developed tool or technology. Most important questions to be addressed before proceeding with a new application are: Is the signal well suited to the tool, and in applying the tool, are any errors inadvertently introduced? Does the analysis provide a new and more useful interpretation of the data and assist in the discovery of new diagnostic information? Wavelet analysis techniques appear to have robust theoretical properties allowing novel interpretation of biomedical data. As new algorithms emerge, they are likely to find application in the analysis of more diverse biomedical signals. Analogously, the problems faced in the biomedical signal acquisition and processing world will hopefully stimulate development of new algorithms.

References 1. Rioul O and Vetterli M. Wavelets and signal processing, IEEE Signal Proc. Mag., (October), 14–38, 1991. 2. Grossmann A and Morlet J. Decomposition of Hardy functions into square integrable wavelets of constant shape, SIAM J. Math. Anal., 15, 723–736, 1984. 3. Kronland-Martinet R, Morlet J, and Grossmann A. Analysis of sound patterns through wavelet transforms, Intern. J. Pattern Rec. Artificial Intell., 1, 273–302, 1987. 4. Daubechies I. The wavelet transform, time-frequency localization, and signal analysis, IEEE Trans. Info. Theory, 36, 961–1005, 1990. 5. Raghuveer M, Samar V, Swartz KP, Rosenberg S, and Chaiyaboonthanit T. Wavelet Decomposition of Event Related Potentials: Toward the Definition of Biologically Natural Components, In: Proc. Sixth SSAP Workshop on Statistical Signal and Array Processing, Victoria, BC, CA, 1992, 38–41. 6. Holschneider M. Wavelets. An Analysis Tool. Clarendon Press, Oxford, 1995. 7. Strang G and Nguyen T. Wavelets and Filter Banks. Wellesley-Cambridge Press, Wellesley, MA, 1996. 8. Aldroubi A and Unser M. Wavelets in Medicine and Biology. CRC Press, Boca Raton, FL, 1996. 9. Ho KC. Fast CWT computation at integer scales by the generalized MRA structure, IEEE Transactions on Signal Processing, 46(2), 501–6, 1998. 10. Hlawatsch F and Bourdeaux-Bartels GF. Linear and Quadratic Time-Frequency Signal Representations, IEEE Signal Processing Magazine, (April), 21–67, 1992. 11. Meste O, Rix H, Jane P, Caminal P, and Thakor NV. Detection of late potentials by means of wavelet transform, IEEE Trans. Biomed. Eng., 41, 625–634, 1994. 12. Chan YT. Wavelet Basics. Kluwer Academic Publishers, Boston, 1995. 13. Najmi A-H and Sadowsky J. The continuous wavelet transform and variable resolution timefrequency analysis, The Johns Hopkins APL Technical Digest, 18(1), 134–140, 1997. 14. Holschneider M, Kronland-Martinet R, and Tchamitchian P. A real-time algorithm for signal analysis with the help of the wavelet transform. In: Wavelets: Time-Frequency Methods and Phase Space. Combes J, Grossmann A, Tchamitchian P, Eds. Springer Verlag, New York, 1989, 286–297. 15. Gramatikov B and Georgiev I. Wavelets as an alternative to STFT in signal-averaged electrocardiography, Med. Biolog. Eng. Comp., 33(3), 482–487, 1995. 16. Sadowsky J. The continuous wavelet transform: A tool for signal investigation and understanding, Johns Hopkins APL Tech. Dig., 15(4), 306–318, 1994.


17. Jones DL and Baraniuk RG. Efficient approximation of continuous wavelet transforms, Electron. Lett., 27(9), 748–750, 1991. 18. Vetterli M and Kovacevic J. Wavelets and Subband Coding. Prentice Hall, Englewood Cliffs, NJ, 1995. 19. Goupillaud P, Grossmann A, and Morlet, J. Cycle-octave and related transforms in seismic signal analysis, Geoexploration, 23, 85–102, 1984. 20. Oppenheim AV and Schaffer RW. Discrete-Time Signal Processing. Prentice-Hall, Englewood Cliffs, NJ, 1989. 21. Senhadji L, Carrault G, Bellanger JJ, and Passariello G. Some new applications of the wavelet transforms, In: Proc. 14th Ann. Internat. Conf. of the IEEE EMBS, Paris, 1992, 2592–2593. 22. Tuteur FB. Wavelet transformation in signal detection, Proc. IEEE Int. Conf. ASSP, 1435–1438, 1988. 23. Thakor NV, Sun YC, Rix H, and Caminal P. Mulitwave: A wavelet-based ECG data compression algorithm, IEICE Trans. Inf. Systems, E76-D, 1462–1469, 1993. 24. Mallat S. A theory for multiresolution signal decomposition: The wavelet representation, IEEE Trans. Pattern Ana. Machine Intell., 11, 674–693, 1989. 25. Thakor NV, Xin-Rong G, Yi-Chun S, and Hanley DF. Multiresolution wavelet analysis of evoked potentials, IEEE Trans. Biomed. Eng., 40, 1085–1094, 1993. 26. Thakor NV, Gramatikov B, and Mita M. Multiresolution Wavelet Analysis of ECG During Ischemia and Reperfusion, In: Proc. Comput. Cardiol., London, 1993, 895–898. 27. Daubechies I. Orthogonal bases of compactly supported wavelets, Commun. Pure Appl. Math., 41, 909–996, 1988. 28. Daubechies I. Orthonormal bases of compactly supported wavelets II. Variations on a Theme, SIAM J. Mathem. Anal., 24, 499–519, 1993. 29. Burrus CS, Gopinath RA, and Guo H. Introduction to Wavelets and Wavelet Transforms. Prentice Hall, Upper Saddle River, NJ, 1998. 30. Hilton ML. Wavelet and wavelet packet compression of electrocardiograms, IEEE Trans. Biom. Eng., 44(5), 394–402, 1997. 31. Thakor NV, Webster JG, and Tompkins WJ. Estimation of QRS complex power spectra for design of QRS filter, IEEE Trans. Biomed. Eng., 31, 702–706, 1984. 32. Gramatikov B. Digital filters for the detection of late potentials, Med. Biol. Eng. Comp., 31(4), 416–420, 1993. 33. Gramatikov B and Thakor N. Wavelet analysis of coronary artery occlusion related changes in ECG, In: Proc. 15th Ann. Int. Conf. IEEE Eng. Med. Biol. Soc., San Diego, 1993, 731. 34. Pettersson J, Warren S, Mehta N, Lander P, Berbari EJ, Gates K, Sornmo L, Pahlm O, Selvester RH, and Wagner G S. Changes in high-frequency QRS components during prolonged coronary artery occlusion in humans, J. Electrocardiol, 28 (Suppl), 225–7, 1995. 35. Wit AL and Janse MJ. The Ventricular Arrhythmias of Ischemia and Infarction: Electrophysiological Mechanisms. Futura Publishing Company, New York, 1993. 36. Thakor N, Yi-Chun S, Gramatikov B, Rix H, and Caminal P. Multiresolution wavelet analysis of ECG: Detection of ischemia and reperfusion in angioplasty, In: Proc. World Congress Med. Physics Biomed. Eng., Rio de Janeiro, 1994, 392. 37. Gramatikov B, Yi-Chun S, Rix H, Caminal P, and Thakor N. Multiresolution wavelet analysis of the body surface ECG before and after angioplasty, Ann. Biomed. Eng., 23, 553–561, 1995. 38. Kwon K, Freedman B, and Wilcox I. The unstable ST segment early after thrombolysis for acute infarction and its usefulness as a marker of coronary occlusion, Am. J. Cardiol., 67, 109, 1991. 39. Abboud S, Leor J, and Eldar M. High frequency ECG during reperfusion therapy of acute myocardial infarction, In: Proc. Comput. Cardiol., 351–353, 1990. 40. Xue Q, Reddy S, and Aversano T. Analysis of high-frequency signal-averaged ECG measurements, J. Electrocardiol., 28 (Supplement), 239–245, . 41. Berbari EJ. Critical review of late potential recordings, J. Electrocardiol., 20, 125–127, 1987.


42. Gramatikov B. Detection of late potentials in the signal-averaged ECG—combining time and frequency domain analysis, Med. Biol. Eng. Comp., 31(4), 333–339, 1993. 43. Nikolov Z, Georgiev I, Gramatikov B, and Daskalov I. Use of the Wavelet Transform for TimeFrequency Localization of Late Potentials, In: Proc. Congress ‘93 of the German, Austrian and Swiss Society for Biomedical Engineering, Graz, Austria, Biomedizinische Technik, Suppl. 38, 87–89, 1993. 44. Morlet D, Peyrin F, Desseigne P, Touboul P, and Roubel P. Wavelet analysis of high-resolution signal-averaged ECGs in postinfarction patients, J. Electrocardiol., 26, 311–320, 1993. 45. Dickhaus H, Khadral L, and Brachmann J. Quantification of ECG late potentials by wavelet transformation, Comput. Meth. Programs Biomedicine, 43, 185–192, 1994. 46. Jouney I, Hamilton P, and Kanapathipillai M. Adaptive wavelet representation and classification of ECG signals, In: Proc. IEEE Int.’l Conf. of the Eng. Med. Biol. Soc., Baltimore, MD, 1994. 47. Thakor NV, Baykal A, and Casaleggio A. Fundamental analyses of ventricular fibrillation signals by parametric, nonparametric, and dynamical methods. In: Inbar IGaGF, ed. Advances in Processing and Pattern Analysis of Biological Signals. Plenum Press, New York, 273–295, 1996. 48. Baykal A, Ranjan R, and Thakor NV. Estimation of the ventricular fibrillation duration by autoregressive modeling., IEEE Trans. on Biomed. Eng., 44(5), 349–356, 1997. 49. Casaleggio A, Gramatikov B, and Thakor NV. On the use of wavelets to separate dynamical activities embedded in a time series, In: Proc. Computers in Cardiology, Indianapolis, 181–184, 1996. 50. Aunon JI, McGillem CD, and Childers DG. Signal processing in evoked potential research: averaging, principal components, and modeling. Crit. Rev. Biomed. Eng., 5, 323–367, 1981. 51. Raz J. Wavelet models of event-related potentials, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., Baltimore, MD, 1994. 52. Grundy BL, Heros RC, Tung AS, and Doyle E. Intraoperative hypoxia detected by evoked potential monitoring, Anesth. Analg., 60, 437–439, 1981. 53. McPherson RW. Intraoperative monitoring of evoked potentials, Prog. Neurol. Surg., 12, 146–163, 1987. 54. Ademoglu A, Micheli-Tzanakou E, and Istefanopulos Y. Analysis of pattern reversal visual evoked potentials (PRVEP) in Alzheimer’s disease by spline wavelets, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., 1993, 320–321. 55. Sun M, Tsui F, and Sclabassi RJ. Partially reconstructible wavelet decomposition of evoked potentials for dipole source localization, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., 1993, 332–333. 56. Schiff SJ. Wavelet transforms for epileptic spike and seizure detection, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., Baltimore, MD, 1214–1215, 1994. 57. Li C and Zheng C. QRS detection by wavelet transform, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., 1993, 330–331. 58. Brooks DH, On H, MacLeond RS, and Krim H. Spatio-temporal wavelet analysis of body surface maps during PTCA-induced ischemia, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., Baltimore, MD, 1994. 59. Karrakchou M. New structures for multirate adaptive filtering: Application to intereference canceling in biomedical engineering, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., Baltimore, MD, 1994. 60. Bentley PM and McDonnel JTE. Analysis of heart sounds using the wavelet transform, in Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., Baltimore, MD, 1994. 61. Akay M, Akay YM, Welkowitz W, and Lewkowicz S. Investigating the effects of vasodilator drugs on the turbulent sound caused by femoral artery using short term Fourier and wavelet transform methods, IEEE Trans. Biomed. Eng., 41(10), 921–928, 1994. 62. Antonelli L. Dicrotic notch detection using wavelet transform analysis, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., Baltimore, MD, 1994. 63. Ademovic E, Charbonneau G, and Pesquet J-C. Segmentation of infant respiratory sounds with Mallat’s wavelets, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., Baltimore, MD, 1994.


64. Sartene R, Wallet JC, Allione P, Palka S, Poupard L, and Bernard JL. Using wavelet transform to analyse cardiorespiratory and electroencephalographic signals during sleep, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., Baltimore, MD, 1994. 65. Akay YM and Micheli-Tzanakou E. Wavelet analysis of the multiple single unit recordings in the optic tectum of the frog, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., 1993, 334–335. 66. Pattichis M and Pattichis CS. Fast wavelet transform in motor unit action potential analysis, In: Proc. IEEE Int. Conf. Eng. Med. Biol. Soc., 1993, 1225–1226.


Petropulu, A. P. “ Higher-Order Spectral Analysis.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

57 Higher-Order Spectral Analysis 57.1 57.2 57.3

Introduction Definitions and Properties of HOS HOS Computation from Real Data Indirect Method • Direct Method

57.4

Linear Processes Non-Parametric Methods • Parametric Methods

Athina P. Petropulu Drexel University

57.5 57.6

Nonlinear Processes HOS in Biomedical Signal Processing

57.1 Introduction The past 20 years witnessed an expansion of power spectrum estimation techniques, which have proved essential in many applications, such as communications, sonar, radar, speech/image processing, geophysics, and biomedical signal processing [32], [27], [20]. In power spectrum estimation, the process under consideration is treated as a superposition of statistically uncorrelated harmonic components. The distribution of power among these frequency components is the power spectrum. As such phase relations between frequency components are suppressed. The information in the power spectrum is essentially present in the autocorrelation sequence, which would suffice for the complete statistical description of a Gaussian process of known mean. However, there are applications where one would need to obtain information regarding deviations from the Gaussianity assumption and presence of nonlinearities. In these cases power spectrum is of little help, and a look beyond the power spectrum or autocorrelation domain is needed. Higher-Order Spectra (HOS) (of order greater than 2), which are defined in terms of higher-order cumulants of the data, do contain such information [35]. The third order spectrum is commonly referred to as bispectrum and the fourth-order as trispectrum. The power spectrum is also a member of the higher-order spectral class; it is the second-order spectrum. HOS consist of higher-order moment spectra, which are defined for deterministic signals, and cumulant spectra, which are defined for random processes. In general there are three motivations behind the use of HOS in signal processing: (1) to suppress Gaussian noise of unknown mean and variance, (2) to reconstruct the phase as well as the magnitude response of signals or systems, and (3) to detect and characterize nonlinearities in the data. The first motivation stems from the property of Gaussian processes to have zero higher-order spectra of order greater than 2). Due to this property, HOS are high signal-to-noise ratio domains in which one can perform detection, parameter estimation, or even signal reconstruction even if the time domain noise is spatially correlated. The same property of cumulant spectra can provide a means of detecting and characterizing deviations of the data from the Gaussian model.


The second motivation is based on the ability of cumulant spectra to preserve the Fourier phase of signals. In the modeling of time series, second-order statistics (autocorrelation) have been heavily used because they are the result of least-squares optimization criteria. However, an accurate phase reconstruction in the autocorrelation domain can be achieved only if the signal is at minimum phase. Nonminimum phase signal reconstruction can be achieved only in the HOS domain, due to the HOS ability to preserve the phase. Being nonlinear functions of the data, HOS are quite natural tools in the analysis of nonlinear systems operating under a random input. General relations for stationary random data passing through an arbitrary linear system exist and have been studied extensively. Such expressions, however, are not available for nonlinear systems, where each type of nonlinearity must be studied separately. Higher-order correlations between input and output can detect and characterize certain nonlinearities [56], and for this purpose several higher-order spectra based methods have been developed. This chapter is organized as follows. In Section 57.2 definitions and properties of cumulants and higher-order spectra are introduced. In section 57.3 methods for estimation of HOS from finite length data are presented and the asymptotic statistics of the corresponding estimates are provided. In Section 57.4 methods for identification of a linear time-invariant system from HOS of the system output are outlined, while in Section 57.6, non-linear system estimation is considered. Section 57.6 summarizes the research activity related to HOS as applied to biomedical signal analysis and provides details on a particular application, namely the improvement of resolution of medical ultrasound images via HOS is discussed.

57.2

Definitions and Properties of HOS

In this chapter only random one-dimensional processes are considered. The definitions can be easily extended to the two-dimensional case [34]. The join moments of order r of the random variables x1, …, xn are given by [42]:

[

] {

M om x1k1 , …, xnkn = E x1k1 , …, xnkn

( )

= −j

r

(

}

∂r Φ ω1, …, ω n

)

∂ω1k1 …∂ω knn

(57.1) ω1 = L = ω n = 0 ,

where k1 +…+ kn = r, and Φ(·) is their joint characteristic function. The corresponding joint cumulants are defined as:

[

] ( )

Cum x1k1 , …, xnkn = − j

r

(

)

∂r ln Φ ω1, …, ω n   ω1 =L= ωn = 0 . ∂ω1k1 …∂ω knn 

(57.2)

For a stationary discrete time random process x(k), (k denotes discrete time), the moments of order n are given by:

(

) {()(

) (

)}

mnx τ1, τ2 , …, τn−1 = E x k x k + τ1 L x k + τn−1 ,

(57.3)

where E{.} denotes expectation. The nth order cumulants are functions of the moments of order up to n, i.e., 1st order cumulants:

{ ( )} (mean)

c1x = m1x = E x k © 2000 by CRC Press LLC

(57.4)

2nd order cumulants:

( ) ( ) (covariance)

( )

c2x τ1 = m2x τ1 − m1x

2

(57.5)

3rd order cumulants:

(

)

(

)

(

) ( )[ ( )

( )

)] ( )

(

c 3x τ1, τ2 = m3x τ1, τ2 − m1x m2x τ1 + m2x τ2 + m2x τ2 − τ1 + 2 m1x

3

(57.6)

4th order cumulants:

(

)

( ) ( ) ( ) ( ) − m ( τ )m ( τ − τ ) − m [m ( τ − τ , τ − τ ) + m ( τ , τ ) + m ( τ , τ ) + m ( τ , τ )] + (m ) [m ( τ ) + m ( τ ) + m ( τ ) + m ( τ − τ ) + m ( τ − τ ) +m ( τ − τ )] − 6(m )

c 4x τ1, τ2 , τ 3 = m4x τ1, τ2 , τ 3 − m2x τ1 m2x τ 3 − τ2 − m2x τ2 m2x τ 3 − τ1 x 2

x 1

x 1

x 1

x 2

3

x 3

2

2

2

x 2

2

1

1

1

1

3

1

x 3

2

x 2

x 2

x 1

2

3

x 3

3

x 2

3

2

x 3

4

1

x 2

1

3

2

(57.7)

2

4

The general relationship between cumulants and moments can be found in [35]. Some important properties of moments and cumulants are summarized next. [P1] If x(k) is a jointly Gaussian process, then c xn (τ1, τ2, …, τn–1) = 0 for n > 2. In other words, all the information about a Gaussian process is contained in its first and second-order cumulants. This property can be used to suppress Gaussian noise, or as a measure for non-Gaussianity in time series. [P2] If x(k) is symmetrically distributed, then c x3 (τ1, τ2) = 0. Third-order cumulants suppress not only Gaussian processes, but also all symmetrically distributed processes, such as uniform, Laplace, and Bernoulli-Gaussian. [P3] Additivity holds for cumulants. If x(k) = s(k) + w(k), where s(k), w(k) are stationary and statistically independent random processes, then c xn (τ1, τ2, …, τn–1) = c ns (τ1, τ2, …, τn–1) + c wn (τ1, τ2, …, τn–1). It is important to note that additivity does not hold for moments. If w(k) is a Gaussian representing noise which corrupts the signal of interest, s(k), then by means of (P2) and (P3), we conclude that c xn (τ1, τ2, …, τn–1) = c ns (τ1, τ2 , …, τn–1), for n > 2. In other words, in higher-order cumulant domains the signal of interest propagates noise free. Property (P3) can also provide a measure of the statistical dependence of two processes. [P4] if x(k) has zero mean, then c xn (τ1, …, τn–1) = m xn (τ1, …, τn–1), for n ≤ 3. Higher-order spectra are defined in terms of either cumulants (e.g., cumulant spectra) or moments (e.g., moment spectra). Assuming that the nth order cumulant sequence is absolutely summable, the nth order cumulant spectrum of x(k), C xn (ω1 , ω2 , …, ωn–1), exists, and is defined to be the (n–1)-dimensional Fourier transform of the nth order cumulant sequence. In general, C xn (ω1, ω2 , …, ωn–1) is complex, i.e., it has magnitude and phase. In an analogous manner, moment spectrum is the multidimensional Fourier transform of the moment sequence. If v(k) is a stationary non-Gaussian process with zero mean and nth order cumulant sequence

(

)

(

)

cnv τ1, …, τn−1 = γ nvδ τ1, …, τn−1 ,


(57.8)

where δ(.) is the delta function, v(k) is said to be nth order white. Its nth order cumulant spectrum is then flat and equal to γ xn . Cumulant spectra are more useful in processing random signals than moment spectra since they posses properties that the moment spectra do not share, i.e., properties P1, P3, and flatness of spectra corresponding to higher-order white noise.

57.3 HOS Computation from Real Data The definitions of cumulants presented in the previous section are based on expectation operations, and they assume infinite length data. In practice we always deal with data of finite length, therefore, the cumulants can only be estimated. Two methods for cumulants and spectra estimation are presented next for the third-order case.

Indirect Method: Let x(k), k = 0,…, N – 1 be the available data. 1. Segment the data into K records of M samples each. Let x i(k), k = 0, …, M – 1, represent the ith record. 2. Subtract the mean of each record. 3. Estimate the moments of each segment x i(k) as follows:

(

)

m3x i τ1, τ 2 =

(

)

1 M

(

l2

∑ x (l)x (l + τ )x (l + τ ), i

i

i

1

2

l = l1

(57.9)

)

l1 = max 0, − τ1, − τ 2 , l2 = min M − 1 − τ1, M − 1 − τ 2 , M − 1 , τ1 < L, τ 2 < L, i = 1, 2, …, K . Since each segment has zero mean, its third-order moments and cumulants are identical, i.e., c x3i (τ 1, τ2) = mx3i (τ 1, τ2). 4. Compute the average cumulants as:

(

)

cˆ3x τ1, τ2 =

1 K

∑ m (τ , τ ) K

xi 3

1

(57.10)

2

i =1

5. Obtain the third-order spectrum (bispectrum) estimate as the two-dimensional Discrete Fourier Transform of size M × M of the windowed cx3i (τ1, τ2), i.e.,

( ) ∑ ∑ cˆ (τ , τ )e L

L

Cˆ 3x k1, k2 =

x 3

1

 2π  2π − j k1τ1 + k2 τ 2  M M 

2

τ1 = − L τ 2 = − L

(

)

w τ1, τ2 , k1, k2 = 0, …, M − 1

(57.11)

where L < M –1, and w(τ1, τ2) is a two-dimensional window of bounded support, introduced to smooth out edge effects. The bandwidth of the final bispectrum estimate is ∆ = 1/L. A complete description of appropriate windows that can be used in Eq. 57.11 and their properties can be found in [35]. A good choice of cumulant window is:

(

) ( )( )(

)

w τ1, τ2 = d τ1 d τ2 d τ1 − τ2 ,


(57.12)

where

( )cos

 1 sin πτ + 1 −  L d τ = π 0  

()

τ L

τ ≤L

πτ L

(57.13)

τ >L

which is known as the minimum bispectrum bias supremum [36].

Direct Method Let x(k), k = 1, … N be the available data. 1. Segment the data into K records of M samples each. Let x i(k), k = 0, …, M – 1, represent the ith record. 2. Subtract the mean of each record. 3. Compute the Discrete Fourier Transform X i(k) of each segment, based on M points, i.e., M −1

( ) ∑ x (n)e

X k = i

i

− j 2 π nk M

, k = 0, 1, …, M − 1, i = 1, 2, …, K .

(57.14)

n=0

4. The third-order spectrum of each segment is obtained as

(

)

C3xi k1, k2 =

() ( ) (

)

1 i X k1 X i k2 X i* k1 + k2 , i = 1, …, K . M

(57.15)

Due to the bispectrum symmetry properties, C x3i (k 1,k 2) needs to be computed only in the triangular region 0 ≤ k2 ≤ k1, k1 +k2 < M / 2. 5. In order to reduce the variance of the estimate, additional smoothing over a rectangular window of size (M3 × M3) can be performed around each frequency, assuming that the third order spectrum is smooth enough, i.e.,

( )

1 C˜ 3xi k1, k2 = 2 M3

M 3 2 −1

M 3 2 −1

∑ ∑C (k + n , k + n ). xi 3

1

1

2

2

(57.16)

n1 = − M 3 2 n2 = − M 3 2

6. Finally, the third-order spectrum is given as the average over all third-order spectra, i.e.,

( )

1 Cˆ 3x k1, k2 = K

∑C˜ (k , k ) K

xi 3

1

2

(57.17)

i =1

The final bandwidth of this bispectrum estimate is ∆ = M3 /M, which is the spacing between frequency samples in the bispectrum domain. For large N, and as long as

∆ → 0, and ∆2 N → ∞


(57.18)

both the direct and the indirect methods produce asymptotically unbiased and consistent bispectrum estimates, with real and imaginary part variances [50]:

[ ( )]

[ ( )]

var Re Cˆ 3x k1, k2  = var Im Cˆ 3x k1, k2 

( ) ( ) ( ( ) ( ) (

) )

 VL2 C x k C x k C x k + k  MK 2 1 2 2 2 1 2 = 2 C k1 C k2 C k1 + k2 =  M x x x ∆N  KM32 C 2 k1 C 2 k2 C 2 k1 + k2 1

x 2

( ) ( ) ( x 2

)

x 2

indirect

(57.19)

direct ,

where V is the energy of the bispectrum window. From the above expressions, it becomes apparent that the bispectrum estimate variance can be reduced by increasing the number of records, or reducing the size of the region of support of the window in the cumulant domain (L), or increasing the size of the frequency smoothing window (M3), etc. The relation between the parameters M, K, L, M3 should be such that 57.18 is satisfied.

57.4 Linear Processes Let x(k) be generated by exiting a linear, time-invariant (LTI), exponentially stable, mixed-phase system with frequency response H(ω) (or impulse response h(k)), with an nth order white, zero-mean, nonGaussian stationary process v(k), i.e.,

() () ()

x k = v k ∗h k

(57.20)

where ∗ denotes convolution. The output nth order cumulant equals [10]:

(

∞

)

cnx τ1, …, τn−1 = γ nv

∑ h(k)h(k + τ )Lh(k + τ ),

n≥3

n −1

1

(57.21)

k =0

where γ vn is a scalar constant and equals the nth order spectrum of v(k). The output nth order cumulant spectrum is given by

(

)

( ) ( ) (

)

Cnx ω1, ω 2 , …, ω n−1 = γ nv H ω1 L H ω n−1 H * ω1 + L + ω n−1 .

(57.22)

For a linear non-Gaussian random process x(k), the nth order spectrum can be factorized as in Eq. 57.22 for every order n, while for a nonlinear process such a factorization might be valid for some orders only (it is always valid for n = 2). It can be easily shown that the cumulant spectra of successive orders are related as follows:

(

)

(

) ( )γ

Cnx ω1, ω 2 , …, 0 = Cnx−1 ω1, ω 2 , …, ω n−2 H 0

γ nv v n −1

.

(57.23)

For example, the power spectrum of a non-Gaussian linear process can be reconstructed from the bispectrum up to a scalar factor, i.e.,

( )

( )γ

C3x ω, 0 = C2x ω


γ v3 v 2

.

(57.24)

While for n = 2, Eq. 57.22 provides the magnitude only of H(ω), as a function of C xn(ω 1, ω2, …, ωn–1), for n > 2 both the magnitude and phase of H(ω) can be obtained. A significant amount of research has been devoted to estimating the system phase from the phase of the bispectrum or trispectrum [33], [35], [45], and [47]. Let

() () ()

y k = x k +w k

(57.25)

where x(k) was defined in Eq. 57.20, and w(k) is zero-mean Gaussian noise, uncorrelated to x(k). Methods for reconstructing h(k) from y(k) can be divided into two main categories: non-parametric and parametric.

Non-Parametric Methods Non-parametric methods reconstruct the system by recovering its Fourier phase and magnitude. Some utilize the whole bispectrum (or polyspectrum) information [33], [7], [43], [53], [41], [30], and others use fixed, one-dimensional polyspectrum slices [31], [13]. In [45] and [49] methods for system recovery from arbitrarily selected HOS slices have been proposed. A system method based on preselected polyspectrum slices, allows regions where polyspectrum estimates exhibit high variance, or regions where the ideal polyspectrum is expected to be zero, such as in the case of bandlimited systems to be avoided. Assuming that there existed a mechanism to select “good” slices, it can also lead to computational savings. Let the bispectrum of y(k) in Eq. 57.25, i.e., C x3(ω 1, ω2), be discretized on a square grid of size 2π/N × 2π/N, and let {C y3(m, l ), m = 0, …, N–1}, {C y3(m, l + r), m = 0,…, N–1} be two bispectral slices. In [49] it was shown that unique identification of the system impulse response can be performed base on two horizontal slices of the discretized third-order spectrum of the system output, as long as the distance between slices, i.e., r and N are co-prime numbers. Let

[

()

(

)] (( N − 1) × 1)

hˆ = log H 1 , …, log H N − 1

T

(57.26)

where H(k) is the Discrete Fourier Transform of h(n). Then hˆ can be obtained as the solution of the system:

Ahˆ = b ,

(57.27)

where

()

(

)

(

)

b k = log C3h −k − r − l, l − log C3h −k − r − l, l + r + cl ,r k = 0, …, N − 1 1 c l ,r = N

(57.28)

N −1

∑[log C (k, l + r ) − log C (k, l)] y 3

y 3

k =0

and A is a [(N – 1) × (N – 1)] sparse matrix; its first row contains a 1 at the rth column and 0s elsewhere. The kth row contains a –1 at column (k – 1)moduloN, and a 1 at column (k + r – 1)moduloN. Although b(k) requires the computation of the logarithm of a complex quantity, it was shown in [49], that the logbispectrum can be computed based on the principal argument of the bispectral phase, thus bypassing the need for two-dimensional phase unwrapping. The bispectrum needed in this method can be substituted with a bispectral estimate, obtained with any of the methods outlined in section 57.3. Reduction of the HOS estimate variance can be achieved via a low-rank approach [4], [9], and [8]. © 2000 by CRC Press LLC

The above described reconstruction method can be easily extended to any order spectrum [49]. Matlab code for this method can be found at http://www.ece.drexel.edu/CSPL.

Parametric Methods These methods fit a parametric model to the data x(k). For example, let us fit a real autoregressive moving average (ARMA) model to x(k), i.e., p

q

∑ ( ) ( ) ∑ b( j)v(k − j) a i x k −i =

i=0

(57.29)

j =0

where v(k) is a stationary zero-mean nth order white non-Gaussian process, and a(i) and b(j) represent the AR and MA parameters. Let y(k) be as defined in Eq. 57.25. For estimation of the AR parameters from y(k), equations analogous to the Yule-Walker equations can be derived based on third-order cumulants of y(k), i.e., p

∑ a(i)c (τ − i, j) = 0, τ > q,

(57.30)

∑ a(i)c (τ − i, j) = −c (τ, j), τ > q,

(57.31)

y 3

i=0

or p

y 3

y 3

i =1

where it was assumed a(0) = 1. Concatenating Eq. 57.31 for τ = q + 1, …, q + M, M ≥ 0, and j = q – p, …, q, the matrix equation

Ca = c ,

(57.32)

can be formed, where C and c are a matrix and a vector, respectively, formed by third-order cumulants of the process according to Eq. 57.31, and the vector a contains the AR parameters. If the AR order p is unknown and Eq. 57.32 is formed based on an overestimate of p, the resulting matrix C always has rank p. In this case the AR parameters can be obtained using a low-rank approximation of C [16]. Using the estimated AR parameters, â(i), i = 1,…, p, a, pth order filter with transfer function Â(z) = 1 + Σ pi=1 â(i)z –i can be constructed. Based on the filtered-through Â(z) process y(k), i.e., ˜y(k), or otherwise known as the residual time series [16], the MA parameters can be estimated via any MA method [34], for example:

()

bk =

( ) , k = 0, 1, …, q c (q, 0)

c 3y˜ q, k y˜ 3

(57.33)

known as the c(q, k) formula [15]. Practical problems associated with the described approach are sensitivity to model order mismatch, and AR estimation errors that propagate in the estimation of the MA parameters. Additional parametric methods can be found in [5], [14], [34], [35], [44]. [63], [64].


FIGURE 57.1

Second-order Volterra system. Linear and quadratic parts are connected in parallel.

57.5 Nonlinear Processes Despite the fact that progress has been established in developing the theoretical properties of nonlinear models, only a few statistical methods exist for detection and characterization of nonlinearities from a finite set of observations. In this section we will consider nonlinear Volterra systems excited by Gaussian stationary inputs. Let y(k) be the response of a discrete time invariant, pth order, Volterra filter whose input is x(k). Then,

()

y k = h0 +

∑ ∑ h (τ , …, τ ) x(k − τ )Lx(k − τ ), i

1

i

1

(57.34)

i

τ1, …, τi

i

where hi(τ1,…, τi) are the Volterra kernels of the system, which are symmetric functions of their arguments; for causal systems hi(τ1,… τi) = 0 for any τi < 0. The output of a second-order Volterra system when the input is zero-mean stationary is

()

y k = h0 +

∑ h (τ ) x(k − τ ) + ∑ ∑ h (τ , τ ) x(k − τ ) x(k − τ ). 1

1

1

τ1

2

τ1

1

2

1

2

(57.35)

τ2

Equation 57.35 can be viewed as a parallel connection of a linear system h1(τ1) and a quadratic system h2(τ1, τ2) as illustrated in Fig. 57.1. Let

()

{ ( )[ ( ) ]}

c2xy τ = E x k + τ y k − m1y

(57.36)

be the cross-covariance of input and output, and

(

)

{ ( ) ( )[ ( ) ]}

c 3xxy τ1, τ2 = E x k + τ1 x k + τ2 y k − m1y

(57.37)

be the third-order cross-cumulant sequence of input and output. The system’s linear part can be obtained by [61]

( )

H1 −ω =


( ), (ω )

C2xy ω x 2

C

(57.38)

and the quadratic part by

(

)

H 2 −ω1, − ω 2 =

(

C3xxy ω1, ω 2

)

( ) ( )

2C ω1 C ω 2 x 2

x 2

,

(57.39)

xy xxy where C xy2 (ω) and C xxy 3 (ω 1, ω2) are the Fourier transforms of c 2 (τ) and c 2 (τ1, τ2), respectively. It should be noted that the above equations are valid only for Gaussian input signals. More general results assuming non-Gaussian input have been obtained in [22] and [48]. Additional results on particular nonlinear systems have been reported in [11] and [54]. An interesting phenomenon, caused by a second-order nonlinearity, is the quadratic-phase coupling. There are situations where nonlinear interaction between two harmonic components of a process contribute to the power of the sum and/or difference frequencies. The signal

()

(

)

(

x k = A cos λ1k + θ1 + B cos λ 2k + θ2

)

(57.40)

after passing through the quadratic system:

() ()

()

z k = x k + x 2 k , ≠ 0 ,

(57.41)

contains cosinusoidal terms in (λ1, θ1), (λ2,θ2), (2λ1, 2θ1), (2λ2, 2θ2), (λ1 + λ2, θ1 + θ2), (λ1 – λ2, θ1 – θ2). Such a phenomenon that results in phase relations that are the same as the frequency relations is called quadratic phase coupling [28]. Quadratic phase coupling can arise only among harmonically related components. Three frequencies are harmonically related when one of them is the sum or difference of the other two. Sometimes it is important to find out if peaks at harmonically related positions in the power spectrum are in fact phase coupled. Due to phase suppression, the power spectrum is unable to provide an answer to this problem. As an example, consider the process [51]

( ) ∑ cos(λ k + φ ) 6

xk =

i

(57.42)

i

i =1

where λ1 > λ2 > 0, λ4 + λ5 > 0, λ3 = λ1 + λ2, λ6 = λ4 + λ5, φ1,… φ5 are all independent, uniformly distributed random variables over (0, 2π), and φ6 = φ4 + φ5 . Among the six frequencies (λ1, λ2, λ3 ) and (λ4, λ5, λ6) are harmonically related, however, only λ6 is the result of phase coupling between λ4 and λ5 . The power spectrum of this process consists of six impulses at λi , i, …, 6 (Fig. 57.2), offering no indication whether each frequency component is independent or a result of frequency coupling. On the other hand, the bispectrum of x(k), (evaluated in its principal region) is zero everywhere, except at point (λ4, λ5) of the (ω1, ω2) plane, where it exhibits an impulse (Fig. 57.2b). The peak indicates that only λ4, λ5 are phase coupled. The biocherence index, defined as

(

)

P3x ω1, ω 2 =

( ) , ω C ω C ω + ω ) ( ) ( ) ( C3x ω1, ω 2

x 2

C

1

x 2

2

x 2

1

(57.43)

2

has been extensively used in practical situations for the detection and quantification of quadratic phase coupling [28]. The value of the bicoherence index at each frequency pair indicates the degree of coupling


FIGURE 57.2 Quadratic phase coupling. (a) The power spectrum of the process described in Eq. 57.42 cannot determine what frequencies are coupled. (b) The corresponding magnitude bispectrum is zero everywhere in the principle region, except at points corresponding to phase coupled frequencies.

among the frequencies of that pair. Almost all bispectral estimators can be used in Eq. 57.43. However, estimates obtained based on parametric modeling of the bispectrum have been shown to yield resolution superior [51 and 52] to the ones obtained with conventional methods.

57.6 HOS in Biomedical Signal Processing The applications of HOS on biomedical signals are clustered according to the HOS property they most rely on, i.e., (1) the ability to describe non-Gaussian processes and preserve phase, (2) the Gaussian noise immunity, and (3) the ability to characterize nonlinearities. In the first class are the works of [1], [3], and [60], where ultrasound imaging distortions are estimated from the ultrasound echo and subsequently compensated for, to improve the diagnostic quality of the image. HOS have been used in modeling the ultrasound radio-frequency (RF) echo [2], where schemes for estimation of resolvable periodicity as well as correlations among nonresolvable scatters have been proposed. The “tissue color,” a quantity that describes the scatterer spatial correlations, and which can be obtained from the HOS of the RF echo, has been proposed [2] as a tissue characterization feature.


The skewness and kurtosis of mammogram images have been proposed in [18] as a tool for detecting microcalcifications in mammograms. In the second class are methods that process multicomponent biomedical signals, treating one component as the signal of interest and the rest as noise. HOS have been used in [19] to process lung sounds in order to suppress sound originating from the heart, and in [65] to detect human afferent nerve signals, which are usually observed in very poor signal-to-noise ratio conditions. Most of the HOS applications in the biomedical area belong to the third class, and usually investigate the formation and the nature of frequencies present in signals through the presence of quadratic phase coupling (QPC). The bispectrum has been applied in EEG signals of the rat during various vigilance states [37], where QPC between specific frequencies was observed. QPC changes in auditory evoked potentials of healthy subjects and subjects with Alzheimer’s dementia has been reported [55]. Visual evoked potentials have been analyzed via the bispectrum in [59], [24], and [21]. Bispectral analysis of interactions between electrocerebral activity resulting from stimulation of the left and right visual fields revealed nonlinear interactions between visual fields [57]. The bispectrum has been used in the analysis of electromyogram recruitment [66], and in defining the pattern of summation of muscle fiber twitches in the surface mechanomyogram generation [40]. QPC was also observed in the EEG of humans [6] and [23]. In the sequel, presented in some detail are the application of HOS on improving the resolution of ultrasound images, a topic that has recently attracted a lot of attention. Ultrasonic imaging is a valuable tool in the diagnosis of tumors of soft tissues. Some of the distinctive features of ultrasound are its safe, non-ionizing acoustic radiation, and its wide availability as a low cost, portable equipment. The major drawback that limits the use of ultrasound images in certain cases, (e.g., breast imaging) is poor resolution. In B-Scan images, the resolution is compromised due to: (1) the finite bandwidth of the ultrasonic transducer, (2) the non-negligible beam width, and (3) phase aberrations and velocity variations arising from acoustic inhomogeneity of tissues themselves. The observed ultrasound image can be considered as a distorted version of the true tissue information. Along the axial direction the distortion is dominated by the pulse-echo wavelet of the imaging system, while along the lateral direction the distortion is mainly due to finite-width lateral beam profile. Provided that these distortions are known in advance, or non-invasively measurable, their effects can be compensated for in the observed image. With propagation in tissue, however, both axial and lateral distortions change due to the inhomogeneities of the media and the geometry of the imaging system. They also change among different tissue types and individuals. Distortions measure in a simplified setting, e.g. in a water tank, are rarely applicable to clinical images, due to the effects of tissue-dependent components. Therefore, distortions must be estimated based on the backscattered RF data that lead to the B-mode image. Assuming a narrow ultrasound beam, linear propagation and weak scattering, the ultrasonic RF echo, yi(n), corresponding to the ith axial line in the B-mode image is modeled as [1]:

() () () ()

yi n = hi n ∗ fi n + wi n , i = 1, 2, …,

(57.44)

where n is the discrete time; wi(n) is observation noise; fi(n) represents the underlying tissue structure and is referred to as tissue response; and hi(n) represents the axial distortion kernel. Let us assume that: (A1) yi(n) is non-Gaussian; (A2) fi(n) is white, non-Gaussian random process; (A3) wi(n) is Gaussian noise uncorrelated with fi(n); and (A4) yi(n) is a short enough segment, so that the attenuation effects stay constant over its duration. (Long segments of data can be analyzed by breaking them into a sequence of short segments.) A similar model can be assumed to hold in the lateral direction of the RF image. Even though the distortion kernels were known to be non-minimum phase [26], their estimation was mostly carried out using second-order statistics (SOS), such as autocorrelation or power spectrum, thereby neglecting Fourier phase information. Phase is important in preserving edges and boundaries in images. It is particularly important in medical ultrasound images where the nature of edges of a tumor provide important diagnostic information about the malignancy. Complex cepstrum-based operations


that take phase information into account have been used [60] to estimate distortions. HOS retain phase and in addition, are not as sensitive to singularities as the cepstrum. HOS were used [1] for the first time to estimate imaging distortions from B-scan images. It was demonstrated that the HOS-based distortion estimation and subsequent image deconvolution significantly improved resolution. For the case of breast data, in was demonstrated [3] that deconvolution via SOS-based distortion estimates was not as good as its HOS counterpart. In the following we present some results of distortion estimation followed by deconvolution of clinical B-scan images of human breast data. The data were obtained using a flat linear array transducer with a nominal center frequency of 7.5 MHz on a clinical imaging system UltraMark-9 Advanced Technology Laboratories. Data were sampled at a rate of 20 MHz. Figure 57.3(a) shows parts of the original image, where the the logarithm of the envelope has been used for display purposes. Axial and lateral distortion kernels were estimated from the RF data via the HOS-based non-parametric method outlined in Section 57.4 of this chapter [49], and also via an SOS power cepstrum based method [3]. Each kernel estimate was obtained from a rectangular block of RF data described by (x, y, Nx , Ny) where (x, y) are the co-ordinates of the upper left corner of the block, Nx is its lateral width, and Ny is the axial height. Note that y corresponds to the depth of the upper left corner of the image block from the surface of the transducer. In the following, all dimensions are specified in sample numbers. The size of the images used was 192 × 1024. Assuming that within the same block the axial distortion kernel does not significantly depend on the lateral location of the RF data, all axial RF data in the block can be concatenated to form a longer one-dimensional vector. In both HOS and SOS based axial kernel estimations, it was assumed that Nx = 10, Ny = 128, N = 128. Note that the Ny = 128 samples correspond to 5mm in actual tissue space, as is reasonable to assume that attenuation over such a small distance may be assumed constant. Fifty kernels were estimated from the blocks (x, 400, 10, 128), the x taking values in the range 1…50. Lateral kernels were estimated from the same images with parameters Nx = 192, Ny =10, and N = 64. All the kernels were estimated from the blocks (1, y, 192, 10), the y taking values in the range 400…600 increasing each time by 5. Data from all adjacent lateral image lines in the block were concatenated to make a long one-dimensional vector. Note that Ny = 10 corresponds to a very small depth, 0.4mm in the real tissue space, hence he underlying assumption that the lateral distortion kernel does not vary much over this depth, should be reasonable. Figure 57.3(b-c) show that the result of lateral followed by axial deconvolution of the image of Fig. 57.3(a), using respectively HOS- and SOS-based distortion kernel estimates. The deconvolution was performed via the constrained Wiener Filter technique [62]. According to Fig. 57.3, the deconvolution in the RF-domain resulted in a significant reduction in speckle size. The speckle size appears smaller in the HOS-deconvolved image than in SOS-deconvolved ones, which is evidence of higher resolution. The resolution improvement can be quantified based on the width of the main lobe of the autocovariance of the image. The axial resolution gain for the HOS-based approach was 1.8 times that of the SOS-based approach. The lateral resolution gain for the HOS-based approach was 1.73 times as much. From a radiologist’s point of view, overall, the HOS image appears to have better spatial resolution than the original as well as the SOS deconvolved images. Conversely, the SOS image seems to have incurred a loss of both axial and lateral resolution.

Acknowledgements Parts of this chapter have been based on the book: Nikias, C. L. and Petropulu, A. P., Higher Order Spectra Analysis: A Nonlinear Signal Processing Framework, Prentice Hall, Englewood Cliffs, NJ, 1993. Major support for this work came from the National Science Foundation under grant MIP–9553227, the Whitaker Foundation, the National Institute of Health under grant 2P01CA52823–07A1, Drexel University and the Laboratoire des Signaux et Systems, CNRS, Universite Paris Sud, Ecole Superieure d’Electric, France. The author would like to thank Drs. F. Forsberg and E. Conant for providing the ultrasound image and for evaluating the processed image. © 2000 by CRC Press LLC

FIGURE 57.3 (a) Original image, (b) HOS-based deconvolution, (c) SOS-based deconvolution. In all cases the logarithm of the envelope is displayed. Autocovariance plots for (d) original (e) HOS deconvolved and (f) SOS deconvolved RF data.

References [1] Abeyratne, U., Petropulu, A. P., and Reid, J. M. “Higher-Order Spectra Based Deconvolution of Ultrasound Images,” IEEE Trans. Ultrason. Ferroelec., and Freq. Con., vol. 42(6): 1064–1075, November 1995. [2] Abeyratne, U. R., Petropulu, A. P., and Reid, J. M. “On Modeling the Tissue Response from Ultrasound Images,” IEEE Trans. Med. Imag., vol. 14(4): 479–490, August 1996. © 2000 by CRC Press LLC

[3] Abeyratne, U. R., Petropulu, A. P., Golas, T., Reid, J. M., Forsberg, F., and Consant, E. “Blind deconvolution of ultrasound breast images: A comparative study of autocorrelation based versus higher-order statistics based methods,” IEEE Trans. Ultrason. Ferroelec., and Freq. Con., vol. 44(6): 1409–1416, November 1997. [4] Andre, T. F., Nowak, R. D., Van Veen, B. D. “Low Rank Estimation of Higher Order Statistics,” IEEE Trans. on Sig. Proc., vol. 45(3): 673–685, March 1997. [5] Alshbelli, S. A., Venetsanopoulos, A. N., and Cetin, A. E. “Cumulant Based Identification Approaches for Nonminimum Phase Systems,” IEEE Trans. Sig. Proc., vol. 41(4): 1576–1588, April 1993. [6] Barnett, T. P., Johnson, L.C. et al., “Bispectrum analysis of electroencephalogram signals during waking and sleeping,” Science, 172: 401–402, 1971. [7] Bartlet, H., Lohmann, A. W., and Wirnitzer, B. “Phase and amplitude recovery from bispectra,” Applied Optics, 23(18): 3121–3129, Sept. 1984. [8] Bradaric, I., and Petropulu, A. P. “Subspace Design of Low Rank Estimators for Higher Order Statistics,” IEEE Trans. Sig. Proc., submitted in 1999. [9] Bradaric, I., and Petropulu, A. P. “Low Rank Approach in System Identification Using Higher Order Statistics,” 9th IEEE Sig. Processing Workshop on Statistical Signal and Array Processing - SSAP’98, Portland, Oregon, September 1998. [10] Brillinger, D. R. and Rosenblatt, M. “Computation and Interpretation of kth-order Spectra,” In: Spectral Analysis of Time Series, B Harris, Ed., John Wiley and Sons, New York, NY, 189–232, 1967. [11] Brillinger, D. R., “The Identification of a Particular Nonlinear Time Series System,” Biometrika, 64(3): 509–515, 1977. [12] Cohen, F. N. “Modeling of Ultrasound Speckle with Applications in Flaw Detection in Metals,” IEEE Trans. on Signal Processing, 40(3): 624–632, 1992. [13] Dianat, S. A. and Raghuveer, M. R. “Fast Algorithms for Phase and Magnitude Reconstruction from Bispectra,” Optical Engineering, 29: 504–512, 1990. [14] Fonollosa, J. A. R. and Vidal, J. “System Identification Using a Linear Combination of Cumulant Slices,” IEEE Trans. Sig. Proc., 41: 2405–2411, 1993. [15] Giannakis, G. B. “Cumulants: A Powerful Tool in Signal Processing,” Proc. IEEE, 75, 1987. [16] Giannakis, G. B. and Mendel, J. M., “Cumulant-Based Order Determination of Non-Gaussian ARMA Models,” IEEE Trans. on Acoustics, Speech, and Signal Processing, 38: 1411–1423, 1990. [17] Giannakis, G. B. and Swami, A. “On Estimating Noncausal Nonminimum Phase ARMA Models of Non-Gaussian Processes,” IEEE Transactions Acoustics, Speech, and Signal Processing, 38(3): 478–495, 1990. [18] Gurcan, M. N., Yardimci, Y., Cetin, A. E., Ansari, R. “Detection of Microcalcifications in Mammograms Using Higher-Order Statistics,” IEEE Sig. Proc. Letters., 4(8), 1997. [19] Hadjileontiadis, L. and Panas, S. M., “Adaptive Reduction of Heart Sounds from Lung Sounds Using Fourth-Order Statistics,” IEEE Trans. Biomed. Eng., 44(7), 1997. [20] Haykin, S. Nonlinear Methods of Spectral Analysis, 2nd ed., Berlin, Germany, Springer-Verlag, 1983. [21] Henning, G. and Husar, P. “Statistical Detection of Visually Evoked Potentials,” IEEE Engineering in Medicine and Biology, 14(4), 386–390, 1995. [22] Hinich, M. J., “Identification of the Coefficients in a Nonlinear Time Series of the Quadratic Type,” J. of Economics, 30: 269–288, 1985. [23] Husar, P. J., Leiner, B., et al., “Statistical Methods for Investigating Phase Relations in Stochastic Processes,” IEEE Trans. on Audio and Electroacoustics, 19(1): 78–86, 1971. [24] Husar, P. and Henning, G. “Bispectrum Analysis of Visually Evoked Potentials,” IEEE Engineering in Medicine and Biology, 16(1), 1997. [25] Jensen, J. A., “Deconvolution of Ultrasound Images,” Ultrasonic Imaging, 14:1–15, 1992. [26] Jensen, J. A. and Leeman, S. “Nonparametric Estimation of Ultrasound Pulses,” IEEE Trans. Biomed. Eng., 41(10): 929–936, 1994. [27] Kay, S.M. Modern Spectral Estimation, Prentice-Hall, Inc., Englewood Cliffs, NJ, 1988. © 2000 by CRC Press LLC

[28] Kim, Y. C. and Powers, E. J. “Digital Bispectral Analysis of Self-Excited Fluctuation Spectra,” Phys. Fluids, 21(8): 1452–1453, 1978. [29] Le Roux, J, Coroyer, C., and Rossille, D. “Illustration of the Effects of Sampling on Higher-Order Spectra,” Signal Processing, 36: 375–390, 1994. [30] Le Roux, J. and Sole, P. “Least-Squared Error Reconstruction of a Deterministic Sampled Signal Fourier Transform Logarithm from its Nth Order Polyspectrum Logarithm,” Signal Processing, 35: 75–81, 1994. [31] Lii, K. S. and Rosenblatt, M. “Deconvolution and Estimation of Transfer Function Phase and Coefficients for Nongaussian Linear Processes,” The Annals of Statistics, 10: 1195–1208, 1982. [32] Marple, Jr., S. L. Digital Spectral Analysis with Applications, Prentice-Hall, Inc., Englewood Cliffs, NJ, 1987. [33] Matsuoka, T. and Ulrych, T. J. “Phase Estimation Using Bispectrum,” Proc. of IEEE, 72: 1403–1411, Oct., 1984. [34] Mendel, J. M. “Tutorial on Higher-Order Statistics (Spectra) in Signal Processing and System Theory: Theoretical Results and Some Applications,” IEEE Proc., 79: 278–305, 1991. [35] Nikias, C. L. and Petropulu, A P. Higher-Order Spectra Analysis: A Nonlinear Signal Processing Framework, Prentice Hall Inc., Englewood Cliffs, NJ, 1993. [36] Nikias, C. L. and Raghuveer, M. R. “Bispectrum Estimation: A Digital Signal Processing Framework,” Proceedings IEEE, 75(7): 869–891, 1987. [37] Ning, T. and Bronzino, J. D. “Bispectral Analysis of the EEG During Various Vigilance States,” IEEE Transactions on Biomedical Engineering, 36(4): 497–499, 1989. [38] Ning, T. and Bronzino, J. D. “Autogressive and Bispectral Analysis Techniques: EEG Applications,” IEEE Engineering in Medicine and Biology, March 1990. [39] Oppenheim, A. V. and Schafer, R. W. “Discrete-Time Signal Processing, Prentice-Hall, Englewood Cliffs, NJ, 1989. [40] Orizio, C., Liberati, D., Locatelli, C., De Grandis, D., Veicsteinas, A., “Surface Mechanomyogram Reflects Muscle Fibres Twitches Summation,” J. of Biomech. 29(4), 1996. [41] Pan, R. and Nikias, C. L. “The Complex Cepstrum of Higher-Order Cumulants and Nonminimum Phase System Identification,” IEEE Trans. Acoust., Speech, Signal Processing, 36: 186–205, 1988. [42] Papoulis, A. Probability Random Variables and Stochastic Processes, New York: McGraw-Hill, 1984. [43] Petropulu, A. P. and Nikias, C. L. “Signal Reconstruction from the Phase of the Bispectrum,” IEEE Trans. Acoust., Speech, Signal Processing, 40: 601–610, 1992. [44] Petropulu, A. P. “Noncausal Nonminimum Phase ARMA Modeling of Non-Gaussian Processes,” IEEE Trans. on Signal Processing, 43(8): 1946–1954, 1995. [45] Petropulu, A P. and Abeyratne, U. R. “System Reconstruction from Higher-Order Spectra Slices,” IEEE Trans. Sig. Proc. , 45(9): 2241–2251, 1997. [46] Petropulu, A. P. “Higher-Order Spectra in Signal Processing,” in Signal Processing Handbook, CRC Press, Boca Raton, FL, 1998. [47] Petropulu, A. P. and Pozidiz, H. “Phase Estimation from Bispectrum Slices,” IEEE Trans. Sig., Proc. 46(2): 527–531, 1998. [48] Powers, E. J., Ritz, C. K., et al. “Applications of Digital Polyspectral Analysis to Nonlinear Systems Modeling and Nonlinear Wave Phenomena,” Workshop on Higher-Order Spectral Analysis, 73–77, Vail, CO, 1989. [49] Pozidis, H. and Petropulu, A. P. “System Reconstruction from Selected Regions of the Discretized Higher-Order Spectrum,” IEEE Trans. Sig. Proc., 46(12): 3360–3377, 1998. [50] Subba Rao, T. and Gabr, M. M. “An Introduction to Bispectral Analysis and Bilinear Time Series Models,” Lecture Notes in Statistics, 24, New York: Springer-Verlag, 1984. [51] Raghuveer, M. R. and Nikias, C. L. “Bispectrum Estimation: A Parametric Approach,” IEEE Trans. on Acous., Speech, and Sig. Proc., ASSP 33(5): 1213–1230, 1985. [52] Raghuveer, M. R. and Nikias, C. L. “Bispectrum Estimation via AR Modeling,” Signal Processing, 10:35–45, 1986. © 2000 by CRC Press LLC

[53] Rangoussi, M. and Giannakis, G. B. “FIR Modeling Using Log-Bispectra: Weighted Least-Squares Algorithms and Performance Analysis,” IEEE Trans. Circuits and Systems, 38(3): 281–296, 1991. [54] Rozzario, N. and Papoulis, A. “The Identification of Certain Nonlinear Systems by Only Observing the Output,” Workshop on Higher-Order Spectral Analysis, 73–77, Vail, CO, 1989. [55] Samar, V. J, Swartz, K. P., et al. “Quadratic Phase Coupling in Auditory Evoked Potentials from Healthy Old Subjects and Subjects with Alzheimer’s Dementia,” IEEE Signal Processing Workshop on Higher-Order Statistics, 361–365, Tahoe, CA, 1993. [56] Schetzen, M. The Volterra and Wiener Theories on Nonlinear Systems, updated edition, Krieger Publishing Company, Malabar, FL, 1989. [57] “Bispectral Analysis of Visual Field Interactions,” Proc. of International Conference of the IEEE Engineering in Medicine and Biology, vol. 3, Amsterdam, Netherlands, 1996. [58] Swami, A. and Mendel, J. M. “ARMA Parameter Estimation Using Only Output Cumulants,” IEEE Trans. Acoust., Speech and Sig. Proc., 38: 1257–1265, 1990. [59] Tang, Y. and Norcia, A. M. “Coherent Bispectral Analysis of the Steady-State VEP,” Proc. Of International Conference of the IEEE Engineering in Medicine and Biology, 1995. [60] Taxt, T. “Comparison of Cepstrum-Based Methods for Radial Blind Deconvolution of Ultrasound Images,” IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 44(3), 1997. [61] Tick, L. J., “The Estimation of Transfer Functions of Quadratic Systems,” Technometrics, 3(4): 562–567, 1961. [62] Treitel, S. and Lines L. R. “Linear Inverse Theory and Deconvolution,” Geophysics, 47(8): 1153–1159, 1982. [63] Tugnait, J. “Fitting Non-Causal AR Signal Plus Noise Models to Noisy Non-Gaussian Linear Processes,” IEEE Trans. Automat. Control, 32: 547–552, 1987. [64] Tugnait, J. “Identification of Linear Stochastic Systems via Second- and Fourth-Order Cumulant Matching,” IEEE Trans. Inform. Theory, 33: 393–407, 1987. [65] Upshaw, B. “SVD and Higher-Order Statistical Processing of Human Nerve Signals,” Proc. of Int. Conf. of the IEEE Engineering in Medicine and Biology, 1996. [66] Yana, K., Mizuta, H., and Kajiyama, R. “Surface Electromyogram Recruitment Analysis Using Higher-Order Spectrum,” Proc. of International Conference of the IEEE Engineering in Medicine and Biology, 1995.


Micheli-Tzanakou, E. “ Neural Networks in Biomedical Signal Processing.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

58 Neural Networks in Biomedical Signal Processing 58.1

Neural Networks in Sensory Waveform Analysis Multineuronal Activity Analysis • Visual Evoked Potential

Evangelia MicheliTzanakou Rutgers University

58.2 58.3 58.4 58.5

Neural Networks in Speech Recognition Neural Networks in Cardiology Neural Networks in Neurology Discussion

Computing with neural networks (NNs) is one of the faster growing fields in the history of artificial intelligence (AI), largely because NNs can be trained to identify nonlinear patterns between input and output values and can solve complex problems much faster than digital computers. Owing to their wide range of applicability and their ability to learn complex and nonlinear relationships—including noisy or less precise information—NNs are very well suited to solving problems in biomedical engineering and, in particular, in analyzing biomedical signals. NNs have made strong advances in continuous speech recognition and synthesis, pattern recognition, classification of noisy data, nonlinear feature detection, and other fields. By their nature, NNs are capable of high-speed parallel signal processing in real time. They have an advantage over conventional technologies because they can solve problems that are too complex—problems that do not have an algorithmic solution or for which an algorithmic solution is too complex to be found. NNs are trained by example instead of rules and are automated. When used in medical diagnosis, they are not affected by factors such as human fatigue, emotional states, and habituation. They are capable of rapid identification, analysis of conditions, and diagnosis in real time. The most widely used architecture of an NN is that of a multilayer perceptron (MLP) trained by an algorithm called backpropagation (BP). Backpropagation is a gradient-descent algorithm that tries to minimize the average squared error of the network. In real applications, the network is not a simple one-dimensional system, and the error curve is not a smooth, bowl-shaped curve. Instead, it is a highly complex, multidimensional curve with hills and valleys (for a mathematical description of the algorithm, see Chapter 182). BP was first developed by P. Werbos in 1974 [1], rediscovered by Parker in 1982 [2], and popularized later by Rummelhart et al. in 1986 [3]. There exist many variations of this algorithm, especially trying to improve its speed and performance in avoiding getting stuck into local minima—one of its main drawbacks. In my work, I use the ALOPEX algorithm developed by my colleagues and myself (see Chapter 182) [4–10], and my colleagues and I have applied it in a variety of world problems of considerable complexity. This chapter will examine several applications of NNs in biomedical signal processing. One- and twodimensional signals are examined.


58.1 Neural Networks in Sensory Waveform Analysis Mathematical analysis of the equations describing the processes in NNs can establish any dependencies between quantitative network characteristics, the information capacity of the network, and the probabilities of recognition and retention of information. It has been proposed that electromyographic (EMG) patterns can be analyzed and classified by NNs [11] where the standard BP algorithm is used for decomposing surface EMG signals into their constituent action potentials (APs) and their firing patterns [12]. A system such as this may help a physician in diagnosing time-behavior changes in the EMG. The need for a knowledge-based system using NNs for evoked potential recognition was described in a paper by Bruha and Madhavan [13]. In this paper, the authors used syntax pattern-recognition algorithms as a first step, while a second step included a two-layer perceptron to process the list of numerical features produced by the first step. Myoelectric signals (MES) also have been analyzed by NNs [14]. A discrete Hopfield network was used to calculate the time-series parameters for a moving-average MES. It was demonstrated that this network was capable of producing the same time-series parameters as those produced by a conventional sequential least-squares algorithm. In the same paper, a second implementation of a two-layered perceptron was used for comparison. The features used were a time-series parameter and the signal power in order to train the perceptron on four separate arm functions, and again, the network performed well. Moving averages have been simulated for nonlinear processes by the use of NNs [15]. The results obtained were comparable with those of linear adaptive techniques. Moody and colleagues [16] used an adaptive approach in analyzing visual evoked potentials. This method is based on spectral analysis that results in spectral peaks of uniform width in the frequency domain. Tunable data windows were used. Specifically, the modified Bessel functions Io – sin h, the gaussian, and the cosine-taper windows are compared. The modified Bessel function window proved to be superior in classifying normal and abnormal populations. Pulse-transmission NNs—networks that consist of neurons that communicate with other neurons via pulses rather than numbers—also have been modeled [7,17]. This kind of network is much more realistic, since, in biological systems, action potentials are the means of neuronal communication. Dayhoff [18] has developed a pulse-transmission network that can temporally integrate arriving signals and also display some resistance to temporal noise. Another method is optimal filtering, which is a variation of the traditional matched filter in noise [19]. This has the advantage of separating even overlapping waveforms. It also carries the disadvantage that the needed knowledge of the noise power spectral density and the Fourier transform of the spikes might not be always available. Principal-components analysis also has been used. Here the incoming spike waveforms are represented by templates given by eigenvectors from their average autocorrelation functions [20]. The authors found that two or three eigenvectors are enough to account for 90% of the total energy of the signal. This way each spike can be represented by the coordinates of its projection onto the eigenvector space. These coordinates are the only information needed for spike classification, which is further done by clustering techniques.

Multineuronal Activity Analysis When dealing with single- or multineuron activities, the practice is to determine how many neurons (or units) are involved in the “spike train” evoked by some sensory stimulation. Each spike in a spike train represents an action potential elicited by a neuron in close proximity to the recording electrode. These action potentials have different amplitudes, latencies, and shape configurations and, when superimposed one each other, create a complex waveform—a composite spike. The dilemma that many scientists face is how to decompose these composite potentials into their constituents and how to assess the question of how many neurons their electrode is recording from. One of the most widely used methods is window discrimination, in which different thresholds are set, above which the activity of any given neuron is assigned, according to amplitude. Peak detection techniques also have been used [21]. These methods perform well if the number of neurons is very small and the spikes are well separated. Statistical methods © 2000 by CRC Press LLC

of different complexity also have been used [22-24] involving the time intervals between spikes. Each spike is assigned a unique instant of time so that a spike train can be described by a process of time points corresponding to the times where the action potential had occurred. Processes such as these are called point processes, since they are characterized by only one number. Given this, a spike train can be treated as a stochastic point process that may or may not be stationary. In the former case, its statistics do not vary in the time of observation [25]. In the second case, when nonstationarity is assumed, any kind of statistical analysis becomes formidable. Correlations between spike trains of neurons can be found because of many factors, but mostly because of excitatory or inhibitory interactions between them or due to a common input to both. Simulations on each possibility have been conducted in the past [26]. In our research, when recording from the optic tectum of the frog, the problem is the reversed situation of the one given above. That is, we have the recorded signal with noise superimposed, and we have to decompose it to its constituents so that we can make inferences on the neuronal circuitry involved. What one might do would be to set the minimal requirements on a neural network, which could behave the same way as the vast number of neurons that could have resulted in a neural spike train similar to the one recorded. This is a very difficult problem to attack with no unique solution. A method that has attracted attention is the one developed by Gerstein et al. [22,27]. This technique detects various functional groups in the recorded data by the use of the so-called gravitational clustering method. Although promising, the analysis becomes cumbersome due to the many possible subgroups of neurons firing in synchrony. Temporal patterns of neuronal activity also have been studied with great interest. Some computational methods have been developed for the detection of favored temporal patterns [28-31]. My group also has been involved in the analysis of complex waveforms by the development of a novel method, the ST-scan method [32]. This method is based on well-known tomographic techniques, statistical analysis, and template matching. The method proved to be very sensitive to even small variations of the waveforms due to the fact that many orientations of them are considered, as it is done in tomographic imaging. Each histogram represents the number of times a stroke vector at a specific orientation is cutting the composite waveform positioned at the center of the window. These histograms were then fed to a NN for categorization. The histograms are statistical representations of individual action potentials. The NN therefore must be able to learn to recognize histograms by categorizing them with the action potential waveform that they represent. The NN also must be able to recognize any histogram as belonging to one of the “learned” patterns or not belonging to any of them [33]. In analyzing the ST-histograms, the NN must act as an “adaptive demultiplexer.” That is, given a set of inputs, the network must determine the correspondingly correct output. This is a categorization procedure performed by a perceptron, originally described by Rosenblatt [34]. In analyzing the ST-histograms, the preprocessing is done by a perceptron, and the error is found either by an LMS algorithm [35] or by an ALOPEX algorithm [36,39,40].

Visual Evoked Potentials Visual evoked potentials (VEPs) have been used in the clinical environment as a diagnostic tool for many decades. Stochastic analysis of experimental recordings of VEPs may yield useful information that is not well understood in its original form. Such information may provide a good diagnostic criterion in differentiating normal subjects from subjects with neurological diseases as well as provide an index of the progress of diseases. These potentials are embedded in noise. Averaging is then used in order to improve the signal-tonoise (S/N) ratio. When analyzing these potentials, several methods have been used, such as spectral analysis of their properties [42], adaptive filtering techniques [43,44], and some signal enhancers, again based on adaptive processes [45]. In this latter method, no a priori knowledge of the signal is needed. The adaptive signal enhancer consists of a bank of adaptive filters, the output of which is shown to be a minimum mean-square error estimate of the signal. If we assume that the VEP represents a composite of many action potentials and that each one of these action potentials propagates to the point of the VEP recording, the only differences between the various © 2000 by CRC Press LLC

action potentials are their amplitudes and time delays. The conformational changes observed in the VEP waveforms of normal individuals and defected subjects can then be attributed to an asynchrony in the arrival of these action potentials at a focal point (integration site) in the visual cortex [36]. One can simulate this process by simulating action potentials and trying to fit them to normal and abnormal VEPs with NNs. Action potentials were simulated using methods similar to those of Moore and Ramon [37] and Bell and Cook [38]. Briefly, preprocessing of the VEP waveforms is done first by smoothing a five-point filter that performs a weighted averaging over its neighboring points:

( ) [ ( ) ( ) ( ) ( ) ( )] 9

S n = F n − 2 + 2F n − 1 + 3F n + 2F n + 1 + F n + 2

(58.1)

The individual signals νj are modulated so that at the VEP recording site each νj has been changed in amplitude am(j) and in phase ph(j). The amplitude change represents the propagation decay, and the phases represent the propagation delays of signals according to the equation

() [

()

( )]

v j i = am j ⋅ AP i − ph j

(58.2)

For a specific choice of am(j) and ph(j), j = 1, 2, … , N, the simulated VEP can be found by N

VEP = b + k

∑v

α j

(58.3)

j =1

where k is a scaling factor, b is a dc component, and α is a constant [39,40]. The ALOPEX process was used again in order to adjust the parameters (amplitude and phase) so that the cost function reaches a minimum and therefore the calculated waveform coincides with the experimental one. The modified ALOPEX equation is given by

() ( )

() () ()

pi n = pi n − 1 + γ∆pi n ∆R n + µri n

(58.4)

where pi(n) are the parameters at iteration n, and γ and µ are scaling factors of the deterministic and random components, respectively, which are adjusted so that at the beginning γ is small and µ is large. As the number of iterations increases, γ increases while µ decreases. The cost function R is monitored until convergence has been achieved at least 80% or until a preset number of iterations has been covered. The results obtained show a good separation between normal and abnormal VEPs. This separation is based on an index λ, which is defined as the ratio of two summations, namely, the summation of amplitudes whose ph(i) is less than 256 ms and the summation of amplitudes whose ph(i) is greater than 256 ms. A large value of λ indicates an abnormal VEP, while a small λ indicates a normal VEP. The convergences of the process for a normal and an abnormal VEP are shown in Figs. 58.1 and 58.2, respectively, at different iteration numbers. The main assumption here is that in normal individuals, the action potentials all arrive at a focal point in the cortex in resonance, while in abnormal subjects there exists as asynchrony of the arrival times. Maier and colleagues [41] noted the importance of source localization of VEPs in humans in studying the perceptual behavior in humans. The optimization procedure used in this section can help that task, since individual neuronal responses are optimally isolated. One of the interesting points is that signals (action potentials) with different delay times result in composite signals of different forms. Thus the reverse solution of this problem, i.e., extracting individual signals from a composite measurement, can help resolve the problem of signal source localization. The multiunit


(a)

(b)

(c)

FIGURE 58.1 Normal VEP. (a) The fitting at the beginning of the process, (b) after 500 iterations, and (c) after 1000 iterations. Only one action potential is repeated 1000 times. The x axis is ×10 ms; the y axis is in millivolts.


(a)

(b)

(c)

FIGURE 58.2 Abnormal VEP. (a) Fitting at t = 3 iterations, (b) after 500 iterations, and (c) after 2000 iterations. One action potential was used 1000 times.


recordings presented in the early sections of this paper fall in the same category with that of decomposing VEPs. This analysis might provide insight as to how neurons communicate.

58.2 Neural Networks in Speech Recognition Another place where NNs find wide application is in speech recognition. Tebelski et al. [46] have studied the performance of linked predictive NNs (LPNNs) for large-vocabulary, continuous-speech recognition. The authors used a six-state phoneme topology, and without any other optimization, the LPNN achieved an average of 90%, 58%, and 39% accuracy on tasks with perplexity of 5, 11, and 402, respectively, which is better than the performance of several other simpler NNs tested. These results show that the main advantages of predictive networks are mainly that they produce nonbinary distortion measures in a simple way and that they can model dynamic properties such as those found in speech. Their weakness is that they show poor discrimination, which may be corrected by corrective training and function work modeling. Allen and Kanam [47] designed in NN architecture that locates word boundaries and identifies words from phoneme sequences. They tested the model in three different regimes with a highly redundant corpus and a restricted vocabulary, and the NN was trained with a limited number of phonemic variations for the words in the corpus. These tests yielded a very low error rate. In a second experiment, the network was trained to identify words from expert transcriptions of speech. The error rate for correct simultaneous identification of words and word boundaries was 18%. Finally, they tested the use of the output of a phoneme classifier as the input to the word and word boundary identification network. The error rate increased almost exponentially to 49%. The best discrimination came from hybrid systems such as the use of an MLP and a hidden Markov model (HMM). These systems incorporate multiple sources of evidence such as features and temporal context without any restrictive assumptions of distributions or statistical independence. Bourland et al. [48] used MLPs as linear predictions for autoregressive HMMs. This approach, although more compatible with the HMM formalism, still suffers from several limitations. Although these authors generalized their approach to take into account time correlations between successive observations without any restrictive assumptions about the driving noise, the reputed results show that many of the tricks used to improve standard HMMs are also valid for this hybrid approach. In another study, Intrator [49] used an unsupervised NN for dimensionality reduction that seeks directions emphasizing multimodality and derived a new statistical insight to the synaptic modification equations learning as in the Bienenstock, Cooper, and Munro (BCM) neurons [50]. The speech data consisted of 20 consecutive time windows of 32 ms with 30-ms overlap aligned at the beginning of the time. For each window, a set of 22 energy levels was computed corresponding to the Zwicker criticalband filters [51]. The classification results were compared with those of a backpropagation network. These results showed that the backpropagation network does well in finding structures useful for classification of the trained data, but these structures are more sensitive to voicing. Classification results using a BCM network, on the other hand, suggest that for the specified task, structures that are more sensitive to voicing can be extracted, even though voicing imposes several effects on the speech signal. These features are more speaker-invariant. Phan et al. [52] have attempted to solve the “cocktail party effect,” which describes phenomena in which humans can selectively focus attention on one sound source among competing sound sources. This is an ability that is hampered for the hearing-impaired. A system was developed that successfully identifies a speaker in the presence of competing speakers for short utterances. Features used for identification are monaural, whose feature space represent a 90% data reduction from the original data. This system is presently used off-line and also has been applied successfully to intraspeaker speech recognition. The features used in the preprocessing were obtained by wavelet analysis. This multiresolution analysis decomposes a signal into a hierarchical system of subspaces that are one-dimensional and are square integrable. Each subspace is spanned by basis functions that have scaling characteristics of either dilation


or compression depending on the resolution. The implementation of these basis functions is incorporated in a recursive pyramidal algorithm in which the discrete approximation of a current resolution is convolved with quadrature mirror filters in the subsequent resolution [53]. After preprocessing, speech waveform is analyzed by the wavelet transform. Analysis is limited to four octaves. For pattern-recognition input configuration, the wavelet coefficients are mapped to a vector and used as inputs to a neural network trained by the ALOPEX algorithm. White noise was superimposed on the features as well in order to establish a threshold of robustness of the algorithm.

58.3 Neural Networks in Cardiology Two-dimensional echocardiography is an important noninvasive clinical tool for cardiac imaging. The endocardial and epicardial boundaries of the left ventricle (LV) are very useful quantitative measures of various functions of the heart. Cardiac structures detection in ECG images are also very important in recognizing the image parts. A lot of research in the last couple of years has taken place around these issues and the application of neural networks in solving them. Hunter et al. [54] have used NNs in detecting echocardiographic LV boundaries and the center of the LV cavity. The points detected are then linked by a “snake” energyminimizing function to give the epicardial and endocardial boundaries. A snake is an energy-minimizing deformable curve [55] fined by an internal energy that is the controller of the differential properties in terms of its curvature and its metrics. The most robust results were obtained by a 9 × 9 square input vector with a resolution reduction of 32:1. Energy minimization is carried out by simulated annealing. The minimum energy solution was obtained after 1000 iterations over the entire snake. The use of NNs as edge detectors allows the classification of points to be done by their probability of being edges rather than by their edge strength, a very important factor for echocardiographic images due to their wide variations in edge strength. A complex area in electrocardiography is the differentiation of wide QRS tachycardias in ventricular (VT) and supraventricular tachycardia (SVT). A set of criteria for this differentiation has been proposed recently by Brugada et al. [56]. One important aspect of applying NNs in interpreting ECGs is to use parameters that not only make sense but are also meaningful for diagnosis. Dassen et al. [57] developed an induction algorithm that further improved the criteria set by Brugada et al., also using NNs. Nadal and deBossan [58] used principal-components analysis (PCA) and the relative R-wave to R-wave intervals of P-QRS complexes to evaluate arrhythmias. Arrhythmias are one of the risks of sudden cardiac arrest in coronary care units. In this study, the authors used the first 10 PCA coefficients (PCCs) to reduce the data of each P-QRS complex and a feedforward NN classifier that splits the vector space generated by the PCCs into regions that separate the different classes of beats in an efficient way. They obtained better classification than using other methods, with correct classification of 98.5% for normal beats, 98.5% for ventricular premature beats, and 93.5% for fusion beats, using only the first two PCCs. When 4 PCCs were used, these classifications improved to 99.2%, 99.1%, and 94.1%, respectively. These results are better than those obtained by logistic regression when the input space is composed of more than two classes of beats. The difficulties encountered include the elimination of redundant data in order not to overestimate the importance of normal beat detection by the classifier. In another work, Silipo et al. [59] used an NN as an autoassociator. In previous work they had proved that NNs had better performances than the traditional clustering and statistical methods. They considered beat features derived from both morphologic and prematurity information. Their classification is adequate for ventricular ectopic beats, but the criteria used were not reliable enough to characterize the supraventricular ectopic beat. A lot of studies also have used NNs for characterization of myocardial infarction (MI). Myocardial infarction is one of the leading causes of death in the United States. The currently available techniques for diagnosis are accurate enough, but they suffer from certain drawbacks, such as accurate quantitative


FIGURE 58.3

Selection of region of interest. The intensity pattern in the box is used as input to the NN.

measure of severity, extent, and precise location of the infarction. Since the acoustic properties in the involved region are mostly changing, one can study them by the use of ultrasound. Baxt [60] used an NN to identify MI in patients presented to an emergency department with anterior chest pain. An NN was trained on clinical pattern sets retrospectively derived from patients hospitalized with a high likelihood of having MI, and the ability of the NN was compared with that of physicians caring for the same patients. The network performed with a sensitivity of 92% and a specificity of 96%. These figures were better than the physicians, with a sensitivity of 88% and a specificity of 71%, or any other computer technology (88% sensitivity, 74% specificity) [61]. Diagnosis of inferior MI with NNs was studied by Hedén et al. [62] with sensitivity of 84% and a specificity of 97%, findings that are similar to those of Pahlm et al. [63]. Yi et al. [64] used intensity changes in an echocardiogram to detect MI. Once an echocardiogram is obtained, it is digitized and saved in a file as a gray-scale image (512 × 400) (Fig. 58.3). A window of the region of interest is then selected between the systole and diastole of the cardiac cycle and saved in a different file. The window can be either of a constant size or it can be adaptive (varying) in size. This new file can be enlarged (zoom) for examination of finer details in the image. All image artifacts are filtered out, and contrast enhancement is performed. This new image is then saved and serves as input to the NN. A traditional three-layer NN with 300 input nodes (one node for each pixel intensity in the input file), a varying number of hidden nodes, and two output nodes were used. The output node indicates the status of the patient under testing. A “one” indicates normal and a “zero” an abnormal case. The weights of the connections were calculated using the optimization algorithms of ALOPEX. One sub-ALOPEX was used for the weights between hidden nodes and input nodes, and a second sub-ALOPEX was used for those from the output to the hidden layer. The network was trained with a population of 256 patients, some with scars and some normal. These patients were used to obtain “templates” for each category. These templates were then used for comparison with the test images. None of these testing images was included in the training set. The cost function used for the process was the least-squares rule, which, although slow, produces reliable results. A similar process is used for the output layer. The noise was made adaptable. The intensities of the images are normalized before being submitted to the NN. A cutoff of 0.2 was used. Therefore, anything above 0.8 is normal, below 0.2 is scar, and all the in-between values are classified as unknown. Due to the fact that the scar training set was very small compared with the normals, a better classification of normals than scars was observed. A study was also made as to how the number of hidden nodes influences the results for the same standard deviation of noise. © 2000 by CRC Press LLC

In another study, Kostis et al. [65] used NNs in estimating the prognosis of acute MI. Patients who survive the acute phase of an MI have an increased risk of mortality persisting up to 10 years or more. Estimation of the probability of being alive at a given time in the future is important to the patients and their physicians and is usually ascertained by the use of statistical methods. The purpose of the investigation was to use an NN to estimate future mortality of patients with acute MI. The existence of a large database (Myocardial Infarction Data Acquisition Systems, or MIDAS) that includes MI occurring in the state of New Jersey and has long-term follow-up allows the development and testing of such a computer algorithm. Since the information included in the database does not allow the exact prediction of vital status (dead or alive) in all patients with 100% accuracy, the NN should be able to categorize patients according to the probability of dying within a given period of time. Since information included in the database is not sufficient to allow the exact prediction of vital status (dead or alive) in all patients with 100% accuracy, we developed an NN able to categorize patients according to the probability of dying within a given period of time rather than predicting categorically whether a patient will be dead or alive at a given time in the future. It was observed that there were many instances where two or more patients had identical input characteristics while some were dead and some alive at the end of the study period. For this reason, it is difficult to train a standard NN. Since there is no unique output value for all input cases, the network had difficulty converging to a unique set of solutions. To alleviate this problem, a conflict-resolution algorithm was developed. The algorithm takes templates with identical input vectors and averages each of their input characteristics to produce a single case. Their output values of vital status are averaged, producing, in effect, a percentage probability of mortality for the particular set of input characteristics. As each new subject template is read into the program, its input characteristics are compared with those of all previous templates. If no match is found, its input values and corresponding output value are accepted. If a match is found, the output value is brought together with the stored output value (percentage probability of mortality and the number of templates on which it is based), and a new output value is calculated, representing the percentage average mortality of the entire characteristic group. Since each member of the group is an identical input case, no input characteristic averaging is necessary, thus preserving the statistical significance of the average mortality with respect to that exact case. This new algorithm, using the two-hidden-layer perceptron optimized by ALOPEX, has converged to greater than 98% using several thousand input cases. In addition, 10 output nodes were used in the final layer, each corresponding to a range of percent chance or mortality (e.g., node 1: 0% to 10%; node 2: 10% to 20%; etc.). The outputs of the network are designed to maximize one of the 10 potential output “binds,” each corresponding to a decile of mortality between 0% and 100%. The output node containing the correct probability value was set to a value of 1.0; the others to 0.0. In this manner, the network should be able to provide percentage probability of mortality and also to resolve input-case conflicts. An SAS program was written to present the predicted probability of mortality separately in patients who are dead or alive at the end of the follow-up period. The NNs constructed as described above were able to be trained and evaluated according to several definitions of network response: matching the output value at every output node, identifying the correct location of which node was to contain the peak output value, and matching the output value at the peak output location. The network was tested on known and then on unknown cases. A correspondence of the observed to the predicted probability of being alive at a given time was observed. The categorical classifications (dead or alive) yielded an overall accuracy of 74%. A reciprocal relationship between sensitivity and specificity of the rules for determination of vital status was observed.

58.4 Neural Networks in Neurology NNs found applications in neurology as well and, in particular, in characterizing memory defects, as are apparent in diseases such as Parkinson’s and Alzheimer’s. Both diseases exhibit devastating effects and disrupt the lives of those affected.


For several decades, in an attempt to further understand brain functions, computational neuroscientists have addressed the issue by modeling biologic neural network structure with computer simulations. A recent article by Stern et al. [66] reports on an important relationship of Alzheimer’s disease expression and levels of education. A similar inverse relationship was earlier reported by Zhang et al. [67] in a Chinese population. This inverse relationship was attributed to a protective role of the brain reserve capacity [68]. Such a capacity becomes important, since in an educated person’s brain more synapses exist, which might protect the individual in the expression of symptoms of the disease. It is not argued, however, that the disease will not be acquired; rather, that it will be delayed. Zahner et al. [69] have employed a three-layer feedforward NN trained with ALOPEX to simulate the effects of education in dementia, age-related changes, and in general, brain damage. Our results show that the higher the level of training of the NN, 50%, 60%, 70%, 80%, etc., the slower is the damage on the “brain.” Damage was simulated by systematically adding noise on the weights of the network. Noise had a gaussian distribution with varying standard deviations and mean of zero. Figure 58.4 shows the results of these simulations as recognition rate versus standard deviation of noise added to the inputs for increasingly damaged weights at various levels of training. Each point corresponds to an average of 100 samples. Notice how much slower the drop in recognition is for the 70% and 80% curves as compared with the 50% and 60% training. Also notice the distances of the starting points of the curves as we follow the progress of dementia at different stages (Fig. 58.4, a versus b versus c). All curves demonstrate impairment with damage, but they also show some threshold in the training level, after which the system becomes more robust.

58.5 Discussion Neural networks provide a powerful tool for analysis of biosignals. This chapter has reviewed applications of NNs in cardiology, neurology, speech processing, and brain waveforms. The literature is vast, and this chapter is by no means exhaustive. In the last 10 years, NNs have been used more and more extensively in a variety of fields, and biomedical engineering is not short of it. Besides the applications, a lot of research is still going on in order to find optimal algorithms and optimal values for the parameters used in these algorithms. In industry, an explosion of VLSI chip designs on NNs has been observed. The parallel character of NNs makes them very desirable solutions to computational bottlenecks.


FIGURE 58.4 Recognition rate versus standardization of noise added to the inputs. Different curves correspond to various learning levels with damaged weights. (a) Noise added only to the inputs. (b) Noise added to the weights to mimic “brain damage,” σ = 0.05. (c) Noise on the weights with σ = 0.1. Notice how much more robust the “brain” is to noise with higher levels of education. © 2000 by CRC Press LLC

References 1. Werbos P. 1974. Beyond Regression: New Tools for Prediction and Analysis in the Behavioral Sciences. Ph.D. thesis, Harvard University, Cambridge, Mass. 2. Parker DB. 1985. Learning logic, S-81-64, file 1, Office of Technology Licensing, Stanford University, Stanford, Calif. 3. Rumelhart DE, Hinton GE, Williams RJ. 1986. Learning internal representations by error propagation. In DE Rumelhart, JL McClelland (eds), Parallel Distributed Processing, vol 2: Foundations. Cambridge, Mass, MIT Press. 4. Harth E, Tzanakou E. 1974. A stochastic method for determining visual receptive fields. Vision Res 14:1475. 5. Tzanakou E, Michalak R, Harth E. 1984. The ALOPEX process: Visual receptive fields with response feedback. Biol Cybernet 51:53. 6. Micheli-Tzanakou E. 1984. Nonlinear characteristics in the frog’s visual system. Biol Cybernet 51:53. 7. Deutsch S, Micheli-Tzanakou E. 1987. Neuroelectric Systems. New York, NYU Press. 8. Marsic I, Micheli-Tzanakou E. 1990. Distributed optimization with the ALOPEX algorithms. In Proceedings of the 12th Annual International Conference of the IEEE/EMBS 12:1415. 9. Dasey TJ, Micheli-Tzanakou E. 1989. A pattern recognition application of the ALOPEX process with hexagonal arrays. In International Joint Conference on Neural Networks 12:119. 10. Xiao L-T., Micheli-Tzanakou E, Dasey TJ. 1990. Analysis of composite neuronal waveforms into their constituents. In Proceedings of the 12th Annual International Conference of the IEEE/EMBS 12(3):1433. 11. Hiraiwa A, Shimohara K, Tokunaga Y. 1989. EMG pattern analysis and classification by Neural Networks. In IEEE International Conference on Systems, Man and Cybernetics, part 3, pp 1113–1115. 12. Huang Q, Graupe D, Huang Y-F, Liu RW. 1989. Identification of firing patterns of neuronal signals. In Proceedings of the 28th IEEE Conference on Decision and Control, vol 1, pp 266–271. 13. Bruha I, Madhavan GP. 1989. Need for a knowledge-based subsystem in evoked potential neuralnet recognition system. In Proceedings of the 11th Annual International Conference of the IEEE/EMBS, vol 11, part 6, pp 2042–2043. 14. Kelly MF, Parker PA, Scott RN. 1990. Applications of neural networks to myoelectric signal analysis: A preliminary study. IEEE Trans Biomed Eng 37(3):221. 15. Ramamoorthy PA, Govid G, Iyer VK. 1988. Signal modeling and prediction using neural networks. Neural Networks 1(1):461. 16. Moody EB Jr, Micheli-Tzanakou E, Chokroverty S. 1989. An adaptive approach to spectral analysis of pattern-reversal visual evoked potentials. IEEE Trans Biomed Eng 36(4):439. 17. Dayhoff JE. 1990. Regularity properties in pulse transmission networks. Proc IJCNN 3:621. 18. Dayhoff JE. 1990. A pulse transmission (PT) neural network architecture that recognizes patterns and temporally integrates. Proc IJCNN 2:A-979. 19. Roberts WM, Hartile DK. 1975. Separation of multi-unit nerve impulse trains by the multichannel linear filter algorithm. Brain Res 94:141. 20. Abeles M, Goldstein MH. 1977. Multiple spike train analysis. Proc IEEE 65:762. 21. Wiemer W, Kaack D, Kezdi P. 1975. Comparative evaluation of methods for quantification of neural activity. Med Biol Eng 358. 22. Perkel DH, Gerstein GL, Moore GP. 1967. Neuronal spike trains and stochastic point processes: I. The single spike train. Biophys J 7:391. 23. Perkel DH, Gerstein GL, Moore GP: 1967. Neuronal spike trains and stochastic point processes: II. Simultaneous spike trains. Biophys J 7:419. 24. Gerstein GL, Perkel DH, Subramanian KN. 1978. Identification of functionally related neuronal assemblies. Brain Res 140:43.


25. Papoulis A. 1984. Probability, Random Variables and Stochastic Processes, 2d ed. New York, McGraw-Hill. 26. Moore GP, Segundo JP, Perkel DH, Levitan H. 1970. Statistical signs of synaptic interactions in neurons. Biophys J 10:876. 27. Gerstein GL, Perkel DH, Dayhoff JE. 1985. Cooperative firing activity in simultaneously recorded populations of neurons: detection and measurement. J Neurosci 5:881. 28. Dayhoff JE, Gerstein GL. 1983. Favored patterns in nerve spike trains: I. Detection. J Neurophysiol 49(6):1334. 29. Dayhoff JE, Gerstein GL. 1983. Favored patterns in nerve spike trains: II. Application. J Neurophysiol 49(6):1349. 30. Abeles M, Gerstein GL. 1988. Detecting spatiotemporal firing patterns among simultaneously recorded single neurons. J Neurophysiol 60(3):909. 31. Frostig RD, Gerstein GL. 1990. Recurring discharge patterns in multispike trains. Biol Cybernet 62:487. 32. Micheli-Tzanakou E, Iezzi R. 1985. Spike recognition by stroke density function calculation. In Proceedings of the 11th Northeast Bioengineering Conference, pp 309–312. 33. Iezzi R, Micheli-Tzanakou E. 1990. Neural network analysis of neuronal spike-trains. In Annual International Conference of the IEEE/EMBS, vol 12, no 3, pp 1435–1436. 34. Rosenblatt F. 1962. Principles of Neurodynamics. Washington, Spartan Books. 35. Davilla CE, Welch AJ, Rylander HG. 1986. Adaptive estimation of single evoked potentials. In Proceedings of the 8th IEEE EMBS Annual Conference, pp 406–409. 36. Wang J-Z, Micheli-Tzanakou E. 1990. The use of the ALOPEX process in extracting normal and abnormal visual evoked potentials. IEEE/EMBS Mag 9(1):44. 37. Moore J, Ramon F. 1974. On numerical integration of the Hodgkin and Huxley equations for a membrane action potential. J Theor Biol 45:249. 38. Bell J, Cook LP. 1979. A model of the nerve action potential. Math Biosci 46:11. 39. Micheli-Tzanakou E, O’Malley KG. 1985. Harmonic context of patterns and their correlations to VEP waveforms. In Proceedings of IEEE, 9th Annual Conference EMBS, pp 426–430. 40. Micheli-Tzanakou E. 1990. A neural network approach of decomposing brain waveforms to their constituents. In Proceedings of the IASTED International Symposium on Computers and Advanced Technology in Medicine, Healthcare and Bioengineering, pp 56–60. 41. Maier J, Dagnelie G, Spekreijse H, Van Duk W. 1987. Principal component analysis for source localization of VEPs in man. Vis Res 27:165. 42. Nahamura M, Nishida S, Shibasaki H. 1989. Spectral properties of signal averaging and a novel technique for improving the signal to noise ration. J Biomed Eng 2(1):72. 43. Orfanidis S, Aafif F, Micheli-Tzanakou E. 1987. Visual evoked potentials extraction by adaptive filtering. In Proceedings of the IEEE/EMBS International Conference, vol 2, pp 968–969. 44. Doncarli C, Goerig I. 1988. Adaptive smoothing of evoked potentials: A new approach. In Proceedings of the Annual International Conference on the IEEE/EMBS, part 3(of 4), pp 1152–1153. 45. Davilla E, Welch AJ, Rylander HG. 1986. Adaptive estimation of single evoked potentials. In Proceedings of the 8th IEEE/EMBS Annual International Conference, pp 406–409. 46. Tebelski J, Waibel A, Bojan P, Schmidbauer O. 1991. Continuous speech recognition by linked predictive neural networks. In PR Lippman, JE Moody, DS Touretzky (eds), Advances in Neural Information Processing Systems 3, pp 199–205. San Mateo, Calif, Morgan Kauffman. 47. Allen RB, Kanam CA. 1991. A recurrent neural network for word identification from continuous phonemena strings. In PR Lippman, JE Moody, DS Touretzky (eds), Advances in Neural Information Processing Systems 3, pp 206–212. San Mateo, Calif, Morgan Kauffman. 48. Bourland H, Morgan N, Wooters C. 1991. Connectionist approaches to the use of Markov models speech recognition. In PR Lippman, JE Moody, DS Touretzky (eds), Advances in Neural Information Processing systems 3, pp 213–219. San Mateo, Calif, Morgan Kauffman.


49. Intrator N. 1991. Exploratory feature extraction in speech signals. In PR Lippman, JE Moody, DS Touretzky (eds), Advances in Neural Information Processing Systems 3, pp 241–247. San Mateo, Calif, Morgan Kauffman. 50. Bienenstock EL, Cooper LN, Munro PW. 1992. Theory for the development of neuron selectivity: Orientation specificity and binocular interaction in visual cortex. J Neurosci 2:32. 51. Zwicker E. 1961. Subdivision of the audible frequency range into critical bands (frequenagruppen). J Acoust Soc Am 33:248. 52. Phan F, Zahner D, Micheli-Tzanakou E, Sideman S. 1994. Speaker identification through wavelet multiresolution decomposition and Alopex. In Proceedings of the 1994 Long Island Conference on Artificial Intelligence and Computer Graphics, pp 53–68. 53. Mallat SG. 1989. A theory of multiresolution signal decomposition: The wavelet representation. IEEE Trans Pattern Anal Mach Int 11:674. 54. Hunter IA, Soraghan JJ, Christie J, Durani TS. 1993. Detection of echocardiographic left ventricle boundaries using neural networks. Comput Cardiol 201. 55. Cohen LD. 1991. On active contour models and balloons. CVGIP Image Understanding 53(92):211. 56. Brugada P, Brugada T, Mont L, et al. 1991. A new approach to the differential diagnosis of a regular tachycardia with a wide QRS complex. Circulation 83(5):1649. 57. Dassen WRM, Mulleneers RGA, Den Dulk K, et al: 1993. Further improvement of classical criteria for differentiation of wide-QRS tachycardia in SUT and VT using artificial neural network techniques. Comput Cardiol 337. 58. Nadal J, deBossan MC. 1993. Classification of cardiac arrhythmias based on principal components analysis and feedforward neural networks. Comput Cardiol 341. 59. Silipo R, Gori M, Marchesi C. 1993. Autoassociator structured neural network for rhythm classification of long-term electrocardiogram. Comput Cardiol 349. 60. Baxt WB. 1991. Use of an artificial neural network for the diagnosis of myocardial infarction. Ann Intern Med 115(II):843. 61. Goldman L, Cook SF, Brand DA, et al. 1988. A computer protocol to predict myocardial infarction in emergency department patients with chest pain. N Engl J Med 18:797. 62. Hedén B, Edenbrandt L, Haisty WK Jr, Pahlm O. 1993. Neural networks for ECG diagnosis of inferior myocardial infarction. Comput Cardiol 345. 63. Pahlm O, Case D, Howard G, et al. 1990. Decision rules for the ECK diagnosis of inferior myocardial infarction. Comput Biomed Res 23:332. 64. Yi C, Micheli-Tzanakou E, Shindler D, Kostis JB. 1993. A new neural network algorithm to study myocardial ultrasound for tissue characterization. In 19th Northeastern Bioengineering Conference, NJIT, pp 109–110. 65. Kostis WJ, Yi C, Micheli-Tzanakou E. 1993. Estimation of long-term mortality of myocardial infarction using a neural network based on the ALOPEX algorithm. In ME Cohen and DL Hudson (eds), Comparative Approaches in Medical Reasoning. 66. Stern Y, Gurland B, Tatemichi TK, et al. 1994. Influence of education and occupation on the incidence of Alzheimer’s disease. JAMA 271:1004. 67. Zhang M, Katzman R, Salmon D, et al. 1990. The prevalence of dementia and Alzheimer’s disease in Shanghai, China: Impact of age, gender and education. Ann Neurol 27:428., 68. Satz P. 1993. Brain reserve capacity on symptom onset after brain injury: A formulation and review of evidence for threshold theory. Neurophysiology 7(3):723. 69. Zahner DA, Micheli-Tzanakou E, Powell A, et al. 1994. Protective effects of learning on a progressively impaired neural network model of memory. In Proceedings of the 16th IEEE/EMBS Annual International Conference, Baltimore, Md. vol 2, pp 1065–1066.


Onaral, B., Cammarota, J. P. “ Complexity, Scaling, and Fractals in Biomedical Signals.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

59 Complexity, Scaling, and Fractals in Biomedical Signals 59.1

Complex Dynamics. Overcoming the Limits of Newtonian Mathematics • Critical Phenomena: Phase Transitions • An Illustration of Critical Phenomena: Magnetism • A Model for Phase Transitions: Percolation • Self-Organized Criticality • Dynamics at the Edge of Chaos

59.2 Drexel University

Joseph P. Cammarota Naval Air Warfare Center, Aircraft Division

Introduction to Scaling Theories Fractal Preliminaries • Mathematical and Natural Fractals • Fractal Measures • Power Law and 1/f Processes • Distributed Relaxation Processes • Multifractals

Banu Onaral 59.3

An Example of the Use of Complexity Theory in the Development of a Model of the Central Nervous System

Complexity, a contemporary theme embraced by physical as well as social sciences, is concerned with the collective behavior observed in composite systems in which long-range order is induced by short-range interactions of the constituent parts. Complex forms and functions abound in nature. Particularly in biology and physiology, branched, nested, granular, or otherwise richly packed, irregular, or disordered objects are the rule rather than the exception. Similarly ubiquitous are distributed, broad-band phenomena that appear to fluctuate randomly. The rising science of complexity holds the promise to lead to powerful tools to analyze, model, process, and control the global behavior of complex biomedical systems. The basic tenets of the complexity theory rest on the revelation that large classes of complex systems (composed of a multitude of richly interacting components) are reducible to simple rules. In particular, the structure and dynamics of complex systems invariably exist or evolve over a multitude of spatial and temporal scales. Moreover, they exhibit a systematic relationship between scales. From the biomedical engineering standpoint, the worthwhile outcome is the ability to characterize these intricate objects and processes in terms of straightforward scaling and fractal concepts and measures that often can be translated into simple iterative rules. In this sense, the set of concepts and tools, emerging under the rubric of complexity, complements the prediction made by the chaos theory that simple (low-order deterministic) systems may generate complex behavior. In their many incarnations, the concepts of complexity and scaling are playing a refreshingly unifying role among diverse scientific pursuits; therein lie compelling opportunities for scientific discoveries and technical innovations. Since these advances span a host of disciplines, hence different scientific languages, cultures, and dissemination media, finding one’s path has become confusing. One of the aims of this


presentation is to serve as a resource for key literature. We hope to guide the reader toward substantial contributions and away from figments of fascination in the popular press that have tended to stretch emerging concepts ahead of the rigorous examination of evidence and the scientific verification of facts. This chapter is organized in three mains parts. The first part is intended to serve as a primer for the fundamental aspects of the complexity theory. An overview of the attendant notions of scaling theories constitutes the core of the second part. In the third part, we illustrate the potential of the complexity approach by presenting an application to predict acceleration-induced loss of consciousness in pilots.

59.1 Complex Dynamics There exists a class of systems in which very complex spatial and temporal behavior is produced through the rich interactions among a large number of local subsystems. Complexity theory is concerned with systems that have many degrees of freedom (composite systems), are spatially extended (systems with both spatial and temporal degrees of freedom), and are dissipative as well as nonlinear due to the interplay among local components (agents). In general, such systems exhibit emergent global behavior. This means that macroscopic characteristics cannot be deduced from the microscopic characteristics of the elementary components considered in isolation. The global behavior emerges from the interactions of the local dynamics. Complexity theories draw their power from recognition that the behavior of a complex dynamic system does not, in general, depend on the physical particulars of the local elements but rather on how they interact to collectively (cooperatively or competitively) produce the globally observable behavior. The local agents of a complex dynamic system interact with their neighbors through a set of usually (very) simple rules. The emergent global organization that occurs through the interplay of the local agents arises without the intervention of a central controller. That is, there is self-organization, a spontaneous emergence of global order. Long-range correlations between local elements are not explicitly defined in such models, but they are induced through local interactions. The global organization also may exert a top-down influence on the local elements, providing feedback between the macroscopic and microscopic structures [Forrest, 1990] (Fig. 59.1).

Overcoming the Limits of Newtonian Mathematics Linearity, as well as the inherent predictive ability, was an important factor in the success of Newtonian mechanics. If a linear system is perturbed by a small amount, then the system response will change by a proportionally small amount. In nonlinear systems, however, if the system is perturbed by a small amount, the response could be no change, a small change, a large change, oscillations (limit cycle), or

FIGURE 59.1


A complex dynamic system.

chaotic behavior. The response depends on the state of the system at the time it was perturbed. Since most of nature is nonlinear, the key to success in understanding nature lies in embracing this nonlinearity. Another feature found in linear systems is the property of superposition. Superposition means that the whole is equal to the sum of the parts. All the properties of a linear system can be understood through the analysis of each of its parts. This is not the case for complex systems, where the interaction among simple local elements can produce complex emergent global behavior. Complexity theory stands in stark contrast to a purely reductionist approach that would seek to explain global behavior by breaking down the system into its most elementary components. The reductionist approach is not guaranteed to generate knowledge about the behavior of a complex system, since it is likely that the information about the local interactions (which determine the global behavior) will not be revealed in such an analysis. For example, knowing everything there is to know about a single ant will reveal nothing about why an ant colony is capable of such complex behaviors as waging war, farming, husbandry, and the ability to quickly adapt to changing environmental conditions. The approach that complexity theory proposes is to look at the system as a whole and not merely as a collection of irreducible parts. Complexity research depends on digital computers for simulation of interactions. Cellular automata (one of the principal tools of complexity) have been constructed to model sand piles, earthquakes, traffic patterns, satellite communication networks, evolution, molecular autocatalysis, forest fires, and species interactions (among others) [Toffoli & Margoulis, 1987]. We note here that complexity is building on, and in some cases unifying, developments made in the fields of chaotic dynamics [Devaney, 1992], critical phenomena, phase transitions, renormalization [Wilson, 1983], percolation [Stauffer & Aharony, 1992], neural networks [Harvey, 1994; Simpson, 1990], genetic algorithms [Goldberg, 1989] and artificial life [Langton, 1989; Langton et al., 1992].

Critical Phenomena: Phase Transitions For the purpose of this discussion, a phase transition can be defined as any abrupt change between the physical and/or dynamic states of a system. The most familiar examples of phase transitions are between the fundamental stages of matter: solid, liquid, gas, and plasma. Phase transitions are also used to define other changes in matter, such as changes in the crystalline structure or state of magnetism. There are also phase transitions in the dynamics of systems from ordered (fixed-point and limit-cycle stability) to disordered (chaos). Determining the state of matter is not always straightforward. Sometimes the apparent state of matter changes when the scale of the observation (macroscopic versus microscopic) is changed. A critical point is a special case of phase transitions where order and disorder are intermixed at all scales [Wilson, 1983]. At criticality, all spatial and temporal features become scale invariant or self-similar. Magnetism is a good example of this phenomenon.

All Illustration of Critical Phenomena: Magnetism The atoms of a ferromagnetic substance have more electrons with spins in one direction than in the other, resulting in a net magnetic field for the atom as a whole. The individual magnetic fields of the atoms tend to line up in one direction, with the result that there is a measurable level of magnetism in the material. At a temperature of absolute zero, all the atomic dipoles are perfectly aligned. At normal room temperature, however, some of the atoms are not aligned to the global magnetic field due to thermal fluctuations. This creates small regions that are nonmagnetic, although the substance is still magnetic. Spatial renormalization, or coarse graining, is the process of averaging the microscopic properties of the substance over a specified range in order to replace the multiple elements with a single equivalent element. If measurements of the magnetic property were taken at a very fine resolution (without renormalization), there would be some measurements that detect small pockets of nonmagnetism, although most measurements would indicate that the substance was magnetic. As the scale of the measurements is increased, i.e., spatially renormalized, the small pockets of nonmagnetism would be averaged out and


would not be measurable. Therefore, measurements at the larger scale would indicate that the substance is magnetic, thereby decreasing its apparent temperature and making the apparent magnetic state dependent on the resolution of the measurements. The situation is similar (but reversed) at high temperatures. That is, spatial renormalization results in apparently higher temperatures, since microscopic islands of magnetism are missed because of the large areas of disorder in the material. At the Curie temperature there is long-range correlation in both the magnetic and nonmagnetic regions. The distribution of magnetic and nonmagnetic regions is invariant under the spatial renormalization transform. These results are independent of the scale at which the measure is taken, and the apparent temperature does not change under the renormalization transform. This scale invariance (selfsimilarity) occurs at only three temperatures: absolute zero, infinity, and the Curie temperature. The Curie temperature represents a critical point (criticality) in the tuning parameter (temperature) that governs the phase transition from a magnetic to a nonmagnetic state [Pietgen & Richter, 1986].

A Model for Phase Transitions: Percolation A percolation model is created by using a simple regular geometric framework and by establishing simple interaction rules among the elements on the grid. Yet these models give rise to very complex structures and relationships that can be described by using scaling concepts such as fractals and power laws. A percolation model can be constructed on any regular infinite n-dimensional lattice [Stauffer & Aharony, 1992]. For simplicity, the example discussed here will use a two-dimensional finite square grid. In site percolation, each node in the grid has only two states, occupied or vacant. The nodes in the lattice are populated-based on a uniform probability distribution, independent of the sate of any other node. The probability of a node being occupied is p (and thus the probability of a node being vacant is 1 – p). Nodes that are neighbors on the grid link together to form clusters (Fig. 59.2).

FIGURE 59.2


A percolation network.

Clusters represent connections between nodes in the lattice. Anything associated with the cluster can therefore travel (flow) to any node that belongs to the cluster. Percolation can describe the ability of water to flow through a porous medium such as igneous rock, oil fields, or finely ground Colombian coffee. As the occupation probability increases, the clusters of the percolation network grow from local connectedness to global connectedness [Feder, 1988]. At the critical occupation probability, a cluster that spans the entire lattice emerges. It is easy to see how percolation could be used to describe such phenomena as phase transitions by viewing occupied nodes as ordered matter, with vacant nodes representing disordered matter. Percolation networks have been used to model magnetism, forest fires, and the permeability of ion channels in cell membranes.

Self-Organized Criticality The concept of self-organized criticality has been introduced as a possible underlying principle of complexity [Bak et al., 1988; Bak & Chen, 1991]. The class of self-organized critical systems is spatially extended, composite, and dissipative with many locally interacting degrees of freedom. These systems have the capability to naturally evolve (i.e., there is no explicit tuning parameter such as temperature or pressure) toward a critical state. Self-organized criticality is best illustrated by a sand pile. Start with a flat plate. Begin to add sand one grain at a time. The mound will continue to grow until criticality is reached. This criticality is dependent only on the local interactions among the grains of sand. The local slope determines what will happen if another grain of sand is added. If the local slope is below the criticality (i.e., flat) the new grain of sand will stay put and increase the local slope. If the local slope is at the criticality, then adding the new grain of sand will increase the slope beyond the criticality, causing it to collapse. The collapsing grains of sand spread to adjoining areas. If those areas are at the criticality, then the avalanche will continue until local areas with slopes below the criticality are reached. Long-range correlations (up to the length of the sand pile) may emerge from the interactions of the local elements. Small avalanches are very common, while large avalanches are rare. The size (and duration) of the avalanche plotted against the frequency of occurrence of the avalanche can be described by a power law [Bak et al., 1988]. The sand pile seeks the criticality on its own. The slope in the sand pile will remain constant regardless of even the largest avalanches. These same power laws are observed in traffic patterns, earthquakes, and many other complex phenomena.

Dynamics at the Edge of Chaos The dynamics of systems can be divided into several categories. Dynamic systems that exhibit a fixedpoint stability will return to their initial state after being perturbed. A periodic evolution of states will result from a system that exhibits a limit-cycle stability. Either of these systems may display a transient evolution of states before the stable regions are reached. Dynamic systems also may exhibit chaotic behavior. The evolution of states associated with chaotic behavior is aperiodic, well-bounded, and very sensitive to initial conditions and resembles noise but is completely deterministic [Tsonis & Tsonis, 1989]. The criticality that lies between highly ordered and highly disordered dynamics has been referred to as the edge of chaos [Langton, 1990] and is analogous to a phase transition between states of matter, where the highly ordered system can be thought of as a solid and the highly disordered system a liquid. The edge of chaos is the critical boundary between order and chaos. If the system dynamics are stagnant (fixed-point stability, highly ordered system), then there is no mechanism for change. The system cannot adapt and evolve because new states cannot be encoded into the system. If the system dynamics are chaotic (highly disordered), then the system is in a constant state of flux, and there is no memory, no learning, and no adaptation (some of the main qualities associated with life). Systems may exhibit transients in the evolution of states before settling down into either fixed-point or limit-cycle behavior. As the dynamics of a complex system enter the edge of chaos region, the length of these transients quickly grows. The chaotic region is where the length of the “transient” is infinite. At the edge of chaos (the


dynamic phase transition) there is no characteristic scale due to the emergence of arbitrarily long correlation lengths in space and time [Langton, 1990]. The self-organized criticality in the sand piles of Per Bak is an example of a system that exists at the edge of chaos. It is in this region that there is no characteristic space or time scale. A single grain of sand added to the pile could cause an avalanche that consists of two grains of sand, or it could cause an avalanche that spreads over the entire surface of the sand pile.

59.2 Introduction to Scaling Theories Prior to the rise of complexity theories, the existence of a systematic relationship between scales eluded the mainstream sciences. As a consequence, natural structures and dynamics have been commonly dismissed as too irregular and complex and often rejected as monstrous formations, intractable noise, or artifacts. The advent of scaling concepts [Mandelbrot, 1983] has uncovered a remarkable hierarchical order that persists over a significant number of spatial or temporal scales. Scaling theories capitalize on scale-invariant symmetries exhibited by many natural broadband (i.e., multiscale) phenomena. According to the theory of self-organized criticality (see Section 59.1), this scaling order is a manifestation of dilation (compression) symmetries that define the organization inherent to complex systems which naturally evolve toward a critical state while dissipating energies on broad ranges of space and time scales. Long overlooked, this symmetry is now added to the repertoire of mathematical modeling concepts, which had included approaches based largely on displacement invariances under translation and/or rotation. Many natural forms and functions maintain some form of exact or statistical invariance under transformations of scale and thus belong in the scaling category. Objects and processes that remain invariant under ordinary geometric similarity constitute the self-similar subset in this class. Methods to capture scaling information in the form of simple rules that relate features on different scales are actively developed in many scientific fields [Barnsley, 1993]. Engineers are coping with scaling nature of forms and functions by investigating multiscale system theory [Basseville et al., 1992], multiresolution and multirate signal processing [Akansu & Hadad, 1992; Vaidyanathan, 1993], subband coding, wavelets, and filter banks [Meyer, 1993], and fractal compression [Barnsley & Hurd, 1993]. These emerging tools empower engineers to reexamine old data and to re-formulate the question at the root of many unresolved inverse problems—What can small patterns say about large patterns, and vice versa? They also offer the possibility to establish cause-effect relationships between a given physical (spatial) medium and the monitored dynamic (temporal) behavior that constitutes the primary preoccupation of diagnostic scientists.

Fractal Preliminaries In the broadest sense, the noun or adjective fractal refers to physical objects or dynamic processes that reveal new details on space or time magnification. A staple of a truly fractal object or process is therefore the lack of characteristic scale in time or space. Most structures in nature are broadband over a finite range, covering at least a number of frequency decades in space or time. Scaling fractals often consist of a hierarchy or heterarchy of spatial or temporal structures in cascade and are often accomplished through recursive replication of patterns at finer scales. If the replication rule preserves scale invariance throughout the entity, such fractals are recognized as self-similar in either an exact or a statistical sense. A prominent feature of fractals is their ability to pack structure with economy of resources, whether energy, space, or whatever other real estate. Fitting nearly infinite networks into finite spaces is just one such achievement. These types of fractals are pervasive in physiology, i.e., the branching patterns of the bronchi, the cardiovascular tree, and the nervous tissue [West and Goldberger, 1987], which have the additional feature of being “fault tolerant” [West, 1990]. Despite expectations heightened by the colorful publicity campaign mounted by promoters of fractal concepts, it is advisable to view fractals only as a starting approximation in analyzing scaling shapes and © 2000 by CRC Press LLC

fluctuations in nature. Fractal concepts are usually descriptive at a phenomenologic level without pretense to reveal the exact nature of the underlying elementary processes. They do not offer, for that matter, conclusive evidence of whatever particular collective, coupled, or isolated repetitive mechanism that created the fractal object. In many situations, the power of invoking fractal concepts resides in the fact that they bring the logic of constraints, whether in the form of asymmetry of motion caused by defects, traps, energy barriers, residual memories, irreversibility, or any other appropriate interaction or coupling mechanisms that hinder free random behavior. As discussed earlier, the spontaneous or forced organization and the ensuing divergence in correlations and coherences that emerges out of random behavior are presumably responsible for the irregular structures pervasive throughout the physical world. More important, the versatility of fractal concepts as a magnifying tool is rooted in the facility to account for scale hierarchies and/or scale invariances in an exact or statistical sense. In the role of a scale microscope, they suggest a fresh look, with due respect to all scales of significance, at many structural and dynamic problems deemed thus far anomalous or insoluble.

Mathematical and Natural Fractals The history of mathematics is rife with “pathologic” constructions of the iterated kind which defy the euclidian dimension concepts. The collection once included an assortment of anomalous dust sets, lines, surfaces, volumes, and other mathematical miscellenia mostly born out of the continuous yet nondifferentiable category of functions such as the Weierstrass series. The feature unifying these mathematical creations with natural fractals is the fractional or integer dimensions distinct from the euclidian definition. Simply stated, a fractional dimension positions an object in between two integer dimensions in the euclidian sense, best articulated by the critical dimension in the Hausdorff-Besicovith derivation [Feder, 1988]. When this notion of dimension is pursued to the extreme and the dimension reaches an integer value, one is confronted with the counterintuitive reality of space-filling curves, volume-filling planes, etc. These objects can be seen readily to share intrinsic scaling properties with the nearly infinite networks accomplished by the branching patterns of bronchi and blood vessels and the intricate folding of the cortex. A rewarding outcome afforded by the advent of scaling concepts is the ability to characterize such structures in terms of straightforward “scaling” or “dimension” measures. From these, simple iterative rules may be deduced to yield models with maximum economy (or minimum number) or parameters [Barnsley, 1993]. This principle is suspected to underlie the succinct, coding adopted by nature in order to store extensive information needed to create complex shapes and forms.

Fractal Measures The measure most often used in the diagnosis of a fractal is the basic fractal dimension, which, in the true spirit of fractals, has eluded a rigorous definition embracing the entire family of fractal objects. The guiding factor in the choice of the appropriate measures is the recognition that most fractal objects scale self-similarly; in other words, they can be characterized by a measure expressed in the form of a power factor, or scaling exponent ∂, that links the change in the observed dependent quantity V to the independent variable x as V(x) ≈ x ∂ [Falconer, 1990, p 36]. Clearly, ∂ is proportional to the ratio of the logarithm of V(x) and x, i.e., ∂ = log V(x)/log x. In the case of fractal objects, ∂ is the scaling exponent in the fractal sense and may have a fractional value. In the final analysis, most scaling relationships can be cast into some form of a logarithmic dependence on the independent variable with respect to which a scaling property is analyzed, the latter also expressed on the logarithmic scale. A number of dimension formulas have been developed based on this observation, and comprehensive compilations are now available [Falconer, 1990; Feder, 1988]. One approach to formalize the concept of scale invariance utilizes the homogeneity or the renormalization principle given by f (µ) = f (aµ)/b, where a and b are constants and µ is the independent variable


[West & Goldberger, 1987]. The function f that satisfies this relationship is referred as a scaling function. The power-law function f (µ) ≈ µβ is a prominent example in this category provided β = log b/log a. The usefulness of this particular scaling function has been proven many times over in many areas of science, including the thermodynamics of phase transitions and the threshold behavior of percolation networks [Schroeder, 1991; Stauffer & Aharony, 1992; Wilson, 1983].

Power Law and 1/f Processes The revived interest in power-law behavior largely stems from the recognition that a large class of noisy signals exhibits spectra that attenuate with a fractional power dependence on frequency [West & Shlesinger, 1989; Wornell, 1993]. Such behavior is often viewed as a manifestation of the interplay of a multitude of local processes evolving on a spectrum of time scales that collectively give rise to the socalled 1/f β or, more generally, the 1/f-type behavior. As in the case of spatial fractals that lack a characteristic length scale, 1/f processes such as the fractional brownian motion cannot be described adequately within the confines of a “characteristic” time scale and hence exhibit the “fractal time” property [Mandelbrot, 1967].

Distributed Relaxation Processes Since the later part of the nineteenth century, the fractional power function dependence of the frequency spectrum also has been recognized as a macroscopic dynamic property manifested by strongly interacting dielectric, viscoelastic, and magnetic materials and interfaces between different conducting materials [Daniel, 1967]. More recently, the 1/f-type dynamic behavior has been observed in percolating networks composed of random mixtures of conductors and insulators and layered wave propagation in heterogeneous media [Orbach, 1986]. In immittance (impedance or admittance) studies, this frequency dispersion has been analyzed conventionally to distinguish a broad class of the so-called anomalous, i.e., nonexponential, relaxation/dispersion systems from those which can be described by the “ideal” single exponential form due to Debye [Daniel, 1967]. The fractal time or the multiplicity of times scales prevalent in distributed relaxation systems necessarily translates into fractional constitutive models amenable to analysis by fractional calculus [Ross, 1977] and fractional state-space methods [Bagley & Calico, 1991]. This corresponds to logarithmic distribution functions ranging in symmetry from the log-normal with even center symmetry at one extreme to singlesided hyperbolic distributions with diverging moments at the other. The realization that systems that do not possess a characteristic time can be described in terms of distributions renewed the interest in the field of dispersion/relaxation analysis. Logarithmic distribution functions have been used conventionally as means to characterize such complexity [West, 1994].

Multifractals Fractal objects and processes in nature are rarely strictly homogeneous in their scaling properties and often display a distribution of scaling exponents that echos the structural heterogeneities occurring at a myriad of length or time scales. In systems with spectra that attenuate following a pure power law over extended frequency scales, as in the case of Davidson-Cole dispersion [Daniel, 1967], the corresponding distribution of relaxation times is logarithmic and single-tailed. In many natural relaxation systems, however, the spectral dimension exhibits a gradual dependence on frequency, as in phenomena conventionally modeled by the Cole-Cole type dispersion. The equivalent distribution functions exhibit doublesided symmetries on the logarithmic relaxation time scale ranging from the even symmetry of the lognormal through intermediate symmetries down to strictly one-sided functions. The concept that a fractal structure can be composed of fractal subsets with uniform scaling property within the subset has gained popularity in recent years [Feder, 1988]. From this perspective, one may view a complicated fractal object, say, the strange attractor of a chaotic process, as a superposition of simple fractal subsystems. The idea has been formalized under the term multifractal. It follows that


each individual member contributes to the overall scaling behavior according to a spectrum of scaling exponents or dimensions. The latter function is called the multifractal spectrum and summarizes the global scaling information of the complete set.

59.3 An Example of the Use of Complexity Theory in the Development of a Model of the Central Nervous System Consciousness can be viewed as an emergent behavior arising from the interactions among a very large number of local agents, which, in this case, range from electrons through neurons and glial cells to networks of neurons. The hierarchical organization of the brain [Churchland & Sejnowski, 1992; Newell, 1990], which exists and evolves on a multitude of spatial and temporal scales, is a good example of the scaling characteristics found in many complex dynamic systems. There is no master controller for this emergent behavior, which results from the intricate interactions among a very large number of local agents. A model that duplicates the global dynamics of the induction of unconsciousness in humans due to cerebral ischemia produced by linear acceleration stress (G-LOC) was constructed using some of the tenets of complexity [Cammarota, 1994]. It was an attempt to provide a theory that could both replicate historical human acceleration tolerance data and present a possible underlying mechanism. The model coupled the realization that an abrupt loss of consciousness could be thought of as a phase transition from consciousness to unconsciousness with the proposed neurophysiologic theory of G-LOC [Whinnery, 1989]. This phase transition was modeled using a percolation network to evaluate the connectivity of neural pathways within the central nervous system. In order to construct the model, several hypotheses had to be formulated to account for the unobservable interplay among the local elements of the central nervous system. The inspiration for the characteristics of the locally interacting elements (the nodes of the percolation lattice) was provided by the physiologic mechanism of arousal (the all-or-nothing aspect of consciousness), the utilization of oxygen in neural tissue during ischemia, and the response of neural cells to metabolic threats. The neurophysiologic theory of acceleration tolerance views unconsciousness as an active protective mechanism that is triggered by a metabolic threat which in this case is acceleration-induced ischemia. The interplay among the local systems is determined by using a percolation network that models the connectivity of the arousal mechanism (the reticular activating system). When normal neuronal function is suppressed due to local cerebral ischemia, the corresponding node is removed from the percolation network. The configuration of the percolation network varies as a function of time. When the network is no longer able to support arousal, unconsciousness results. The model simulated a wide range of human data with a high degree of fidelity. It duplicated the population response (measured as the time it took to lose consciousness) over a range of stresses that varied from a simulation of the acute arrest of cerebral circulation to a gradual application of acceleration stress. Moreover, the model was able to offer a possible unified explanation for apparently contradictory historical data. An analysis of the parameters responsible for the determination of the time of LOC indicated that there is a phase transition in the dynamics that was not explicitly incorporated into the construction of the model. The model spontaneously captured an interplay of the cardiovascular and neurologic systems that could not have been predicted based on existing data. The keys to the model’s success are the reasonable assumptions that were made about the characteristics and interaction of the local dynamic subsystems through the integration of a wide range of human and animal physiologic data in the design of the model. None of the local parameters was explicitly tuned to produce the global (input-output) behavior. By successfully duplicating the observed global behavior of humans under acceleration stress, however, this model provided insight into some (currently) unobservable inner dynamics of the central nervous system. Furthermore, the model suggests new experimental protocols specifically aimed at exploring further the microscopic interplay responsible for the macroscopic (observable) behavior.


Defining Terms 1/f process: Signals or systems that exhibit spectra which attenuate following a fractional power dependence on frequency. Cellular automata: Composite discrete-time and discrete space dynamic systems defined on a regular lattice. Neighborhood rules determine the state transitions of the individual local elements (cells). Chaos: A state the produces a signal that resembles noise and is aperiodic, well-bounded, and very sensitive to initial conditions but is governed by a low-order deterministic differential or difference equation. Complexity: Complexity theory is concerned with systems that have many degrees of freedom (composite systems), are spatially extended (systems with both spatial and temporal degrees of freedom), and are dissipative as well as nonlinear due to the rich interactions among the local components (agents). Some of the terms associated with such systems are emergent global behavior, collective behavior, cooperative behavior, self-organization, critical phenomena, and scale invariance. Criticality: A state of a system where spatial and/or temporal characteristics are scale invariant. Emergent global behavior: The observable behavior of a system that cannot be deduced from the properties of constituent components considered in isolation and results from the collective (cooperative or competitive) evolution of local events. Fractal: Refers to physical objects or dynamic processes that reveal new details on space or time magnification. Fractals lack a characteristic scale. Fractional brownian motion: A generalization of the random function created by the record of the motion of a “brownian particle” executing random walk. Brownian motion is commonly used to model diffusion in constraint-free media. Fractional brownian motion is often used to model diffusion of particles in constrained environments or anomalous diffusion. Percolation: A simple mathematical construct commonly used to measure the extent of connectedness in a partially occupied (site percolation) or connected (bond percolation) lattice structure. Phase transition: Any abrupt change between the physical and/or the dynamic states of a system, usually between ordered and disordered organization or behavior. Renormalization: Changing the characteristic scale of a measurement though a process of systematic averaging applied to the microscopic elements of a system (also referred to as coarse graining). Scaling: Structures or dynamics that maintain some form of exact or statistical invariance under transformations of scale. Self-organization: The spontaneous emergence of order. This occurs without the direction of a global controller. Self-similarity: A subset of objects and processes in the scaling category that remain invariant under ordinary geometric similarity.

References Akansu AN, Haddad RA. 1992. Multiresolution Signal Decomposition: Transforms, Subbands, and Wavelets. New York, Academic Press. Bagley R, Calico R. 1991. Fractional order state equations for the control of viscoelastic damped structures. J Guidance 14(2):304. Bak P, Tang C, Wiesenfeld K. 1988. Self-organized criticality. Phys Rev A 38(1):364. Bak P, Chen K. 1991. Self-organized. Sci Am Jan:45. Barnsley MF. 1993. Fractals Everywhere, 2d ed. New York, Academic Press. Barnsley MF, Hurd LP. 1993. Fractal Image Compression. Wellesley, AK Peters. Basseville M, Benveniste A, Chou KC, et al. 1992. Modeling and estimation of multiresolution stochastic processes. IEEE Trans Information Theory 38(2):766. Cammarota JP. 1994. A Dynamic Percolation Model of the Central Nervous System under Acceleration (+Gz) Induced/Hypoxic Stress. Ph.D. thesis, Drexel University, Philadelphia.


Churchland PS, Sejnowski TJ. 1992. The Computational Brain. Cambridge, Mass, MIT Press. Daniel V. 1967. Dielectric Relaxation. New York, Academic Press. Devaney RL. 1992. A First Course in Chaotic Dynamical Systems: Theory and Experiment. Reading, Mass, Addison-Wesley. Falconer K. 1990. Fractal Geometry: Mathematical Foundations and Applications. New York, Wiley. Feder J. 1988. Fractals. New York, Plenum Press. Forrest S. 1990. Emergent computation: Self-organization, collective, and cooperative phenomena in natural and artificial computing networks. Physica D 42:1. Goldberg DE. 1989. Genetic Algorithms in Search, Optimization, and Machine Learning. Reading, Mass, Addison-Wesley. Harvey RL. 1994. Neural Network Principles. Englewood Cliffs, NJ, Prentice-Hall. Langton CG. 1989. Artificial Life: Proceedings of an Interdisciplinary Workshop on the Synthesis and Simulation of Living Systems, September 1987, Los Alamos, New Mexico, Redwood City, Calif, Addison-Wesley. Langton CG. 1990. Computation at the edge of the chaos: Phase transitions and emergent computation. Physica D 42:12. Langton CG, Taylor C, Farmer JD, Rasmussen S. 1992. Artificial Life II: Proceedings of the Workshop on Artificial Life, February 1990, Sante Fe, New Mexico. Redwood City, Calif, Addison-Wesley. Mandelbrot B. 1967. Some noises with 1/f spectrum, a bridge between direct current and white noise. IEEE Trans Information Theory IT-13(2):289. Mandelbrot B. 1983. The Fractal Geometry of Nature. New York, WH Freeman. Meyer Y. 1993. Wavelets: Algorithms and Applications. Philadelphia, SIAM. Newell A. 1990. Unified Theories of Cognition. Cambridge, Mass, Harvard University Press. Orbach R. 1986. Dynamics of fractal networks. Science 231:814. Peitgen HO, Richter PH. 1986. The Beauty of Fractals. New York, Springer-Verlag. Ross B. 1977. Fractional calculus. Math Mag 50(3):115. Shroeder M. 1990. Fractals, Chaos, Power Laws. New York, WH Freeman. Simpson PK. 1990. Artificial Neural Systems: Foundations, Paradigms, Applications, and Implementations. New York, Pergamon Press. Stauffer D, Aharony A. 1992. Introduction to Percolation, 2d ed. London, Taylor & Francis. Toffoli T, Margolus N. 1987. Cellular Automata Machines: A New Environment for Modeling. Cambridge, Mass, MIT Press. Tsonis PA, Tsonis AA. 1989. Chaos: Principles and implications in biology. Comput Appl Biosci 5(1):27. Vaidyanathan PP. 1993. Multi-rate Systems and Filter Banks. Englewood Cliffs, NJ, Prentice-Hall. West BJ. 1990. Physiology in fractal dimensions: Error tolerance. Ann Biomed Eng 18:135. West BJ. 1994. Scaling statistics in biomedical phenomena. In Proceedings of the IFAC Conference on Modeling and Control in Biomedical Systems, Galveston, Texas. West BJ, Goldberger A. 1987. Physiology in fractal dimensions. Am Scientist 75:354. West BJ, Shlesinger M. 1989. On the ubiquity of 1/f noise. Int J Mod Phys 3(6):795. Whinnery JE. 1989. Observations on the neurophysiologic theory of acceleration (+Gz) induced loss of consciousness. Aviat Space Environ Med 60:589. Wilson KG. 1983. The renormalization group and critical phenomena. Rev Mod Phys 55(3):583. Wornell GW. 1993. Wavelet-based representations for the 1/f family of fractal processes. IEEE Proc 81(10):1428.


Onaral, B. “Future Directions: Biomedical Signal Processing and Networked Multimedia Communications.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

60 Future Directions: Biomedical Signal Processing and Networked Multimedia Communications

Banu Onaral Drexel University

60.1 60.2 60.3 60.4

Public Switched Network and ATM Wireless Communication Photonics Virtual Reality

The long anticipated “information age” is taking shape at the cross-section of multimedia signal processing and telecommunications-based networking. By defying the traditional concepts of space and time, these emerging technologies promise to affect all facets of our lives in a pervasive and profound manner [Mayo, 1992]. The physical constraints of location have naturally led, over the centuries, to the creation of the conventional patient care services and facilities. As the “information superhighway” is laid down with branches spanning the nation, and eventually the world via wired and wireless communication channels, we will come closer to a bold new era in health care delivery, namely, the era of remote monitoring, diagnosis, and intervention. Forward-looking medical industries are engaging in research and development efforts to capitalize on the emerging technologies. Medical institutions in particular recognize the transforming power of the impending revolution. A number of hospitals are undertaking pilot projects to experiment with the potentials of the new communication and interaction media that will constitute the foundations of futuristic health care systems. There is consensus among health care administrators that the agility and effectiveness with which an institution positions itself to fully embrace the new medical lifestyle will decide its viability in the next millennium. Although multimedia communications is yet in its infancy, recent developments foretell a bright future. Many agree that multimedia networking is becoming a reality thanks to advances in digital signalprocessing research and development. Trends toward implementation of algorithms by fewer components are leading to decreasing hardware complexity while increasing processing functionality [Andrews, 1994]. The vast and vibrant industry producing multimedia hardware and software ranging from applicationspecific digital signal processors and video chip sets to videophones and multimedia terminals heavily relies on digital signal processing know-how.


As in the case of generic digital signal processing, biomedical signal processing is expected to play a key role in mainstreaming patient care at a distance. Earlier in this section, emerging methods in biomedical signal analysis that promise major enhancements in our ability to extract information from vital signals were introduced. This chapter provides a glimpse of the future—when biomedical signals will be integrated with other patient information and transmitted via networked multimedia—by examining trends in key communications technologies, namely, public switched-network protocols, wireless communications, photonics, and virtual reality.

60.1 Public Switched Network and ATM The public switched network already can accommodate a wide array of networked multimedia communications. The introduction of new standards such as the asynchronous transfer mode (ATM) is a strong sign that the network will evolve to handle an array of novel communication services. ATM is a technology based on a switched network that uses dedicated media connections [ATM Networking, 1994]. Each connection between users is physically established by setting up a path or virtual channel through a series of integrated circuits in the switch. In conventional shared media networks, connections are made by breaking the information into packets labeled with their destination; these packets then share bandwidth until they reach their destination. In switched networks, instead of sharing bandwidth, connections can be run in parallel, since each connection has its own pathway. This approach prevents degradation of the response time despite an increased number of users who are running intensive applications on the network, such as videoconferencing. Therefore, ATM offers consistently high performance to all users on the network, particularly in real-time networked multimedia applications which often encounter severe response time degradation on shared media networks. Also, the ATM standards for both local area networks (LANs) and wide area networks (WANs) are the same; this allows for seamless integration of LANs and WANs. Small-scale experiments based on ATM-based multimedia communications have already been launched in a number of medical centers. Early examples involve departments where doctors and staff can remotely work on chest x-rays, CT exams, and MRI images around the hospital over an ATM switched network. Since ATM integrates LANs and WANs, collaborating institutions will be able to access the same information in the future. Similar patient multimedia information-sharing efforts are underway which integrate all vital information including physiologic signals and sounds, images, and video and patient data and make them available remotely. In some recent experiments, the access is accomplished in real time such that medical conferencing, and hence remote diagnosis, becomes a possibility.

60.2 Wireless Communication Wireless communication is the fastest growing sector of the telecommunications industry [Wittman, 1994]. Wireless networking technology is rapidly coming of age with the recent passage of initial standards, widespread performance improvements, and the introduction of personal communications networks (PCNs). Pocket-sized portable “smart” terminals are combined with wireless communication to free users from the constraints of wired connection to networks. The progress in this direction is closely monitored by the health care community because the technology holds the potential to liberate ambulatory patients who require long-term monitoring and processing of biomedical signals for timely intervention. Wireless and interactive access by medical personnel to physiologic multimedia information will no doubt be a staple of the future distributed health care delivery systems.

60.3 Photonics Photonics, or lightwave, is an evolving technology with the capacity to support a wide range of highbandwidth multimedia applications. In current practice, photonics plays a complementary role to


electronics in the hybrid technology referred to as electro-optics. Optical fibers are widely used for transmission of signals, from long-distance links to undersea cables to local loops linking customers with the central switching office. The trend in photonics is to move beyond functions limited to transmission toward logic operations. Recent advances in photonic logic devices suggest that optical computers may present characteristics more desirable than electronics in many biomedical processing applications requiring parallel tasks. A case in point is online pattern recognition, which may bring a new dimension to remote diagnosis.

60.4 Virtual Reality Virtual reality—the ultimate networked multimedia service—is a simulated environment that enables one to remotely experience an event or a place in all dimensions. Virtual reality makes telepresence possible. The nascent technology builds on the capabilities of interactive telecommunications and is expected to become a consumer reality early in the next century. Applications in the field of endoscopic surgery are being developed. Demonstration projects in remote surgery are underway, paving the way for remote medical intervention.

Acknowledgment Contributions by Prabhakar R. Chitrapu, Dialogic Corporation, to material on networked multimedia communications in this chapter are gratefully acknowledged.

Defining Terms ATM (asynchronous transfer mode): Technology based on a switched network that uses dedicated media connections. CT: Computer tomography. MRI: Magnetic resonance imaging. Multimedia: Technology to integrate sights, sounds, and data. The media may include audio signals such as speech, music, biomedical sounds, images, animation, and video signals, as well as text, graphics, and fax. One key feature is the common linkage, synchronization, and control of the various media that contribute to the overall multimedia signal. In general, multimedia services and products are interactive, hence real-time, as in the case of collaborative computing and videoconferencing. Photonics: Switching, computing, and communications technologies based on lightwave. Virtual reality: Technology that creates a simulated environment enabling one to remotely experience an event or a place in all dimensions. Wireless network: Technology based on communication between nodes such as stationary desktops, laptops, or personal digital assistants (PDAs) and the LAN hub or access point using a wireless adapter with radio circuitry and an antenna. The LAN hub has a LAN attachment on one interface and one or more antennas on another.

References Andrews, D. 1994. Digital signal processing: The engine to make multimedia mainstream. Byte 22. ATM Networking: Increasing performance on the network. HEPC Syllabus 3(7):12, 1994. Mayo JS. 1992. The promise of networked multimedia communications. Bear Stearns Sixth Annual Media and Communications Conference, Coronado, Calif. Wittmann A. 1994. Will wireless win the war? Network Computing 58.


Mudry, K. M. “Imaging.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

VII Imaging Karen M. Mudry The Whitaker Foundation 61 X-Ray Robert E. Shroy, Jr., Michael S. Van Lysel, Martin J. Yaffe X-Ray Equipment • X-Ray Projection Angiography • Mammography

62 Computed Tomography Ian A. Cunningham, Philip F. Judy Instrumentation • Reconstruction Principles

63 Magnetic Resonance Imaging Steven Conolly, Albert Macovski, John Pauly, John Schenck, Kenneth K. Kwong, David A. Chesler, Xiaoping Hu, Wei Chen, Maqbool Patel, Kamil Ugurbil Acquisition and Processing • Hardware/Instrumentation • Functional MRI • Chemical-Shift Imaging: An Introduction to Its Theory and Practice

64 Nuclear Medicine Barbara Y. Croft, Benjamin M. W. Tsui Instrumentation • SPECT (Single-Photon Emission Computed Tomography)

65 Ultrasound Richard L. Goldberg, Stephen W. Smith, Jack G. Mottley, K. Whittaker Ferrara Transducer • Ultrasonic Imaging • Blood Flow Measurement Using Ultrasound

66 Magnetic Resonance Microscopy Xiaohong Zhou, G. Allan Johnson Basic Principles • Resolution Limits • Instrumentation • Applications

67 Positron-Emission Tomography (PET) Thomas E. Budinger, Henry F. VanBrocklin Radiopharmaceuticals • Instrumentation

68 Electrical Impedance Tomography D. C. Barber The Electrical Impedance of Tissue • Conduction in Human Tissues • Determination of the Impedance Distribution • Areas of Clinical Application • Summary and Future Developments

69 Medical Applications of Virtual Reality Technology Tom Piantanida

Walter Greenleaf,

Overview of Virtual Reality Technology • VR Application Examples • Current Status of Virtual Reality Technology • Overview of Medical Applications of Virtual Reality Technology • Summary

T

HE FIELD OF MEDICAL IMAGING has experienced phenomenal growth within the last century. Whereas imaging was the prerogative of the defense and the space science communities in the past, with the advent of powerful, less-expansive computers, new and expanded imaging systems have found their way into the medical field. Systems range from those devoted to planar imaging using


x-rays to technologies that are just emerging, such as virtual reality. Some of the systems, such as ultrasound, are relatively inexpensive, while others, such as positron emission tomography (PET) facilities, cost millions of dollars for the hardware and the employment of Ph.D.-level personnel to operate them. Systems that make use of x-rays have been designed to image anatomic structures, while others that make use of radioisotopes provide functional information. The fields of view that can be imaged range from the whole body obtained with nuclear medicine bone scans to images of cellular components using magnetic resonance (MR) microscopy. The design of transducers for the imaging devices to the postprocessing of the data to allow easier interpretation of the images by medical personnel are all aspects of the medical imaging devices field. Even with the sophisticated systems now available, challenges remain in the medical imaging field. With the increasing emphasis on health care costs, and with medical imaging systems often cited as not only an example of the investment that health care providers must make and consequently recover, but also as a factor contributing to escalating costs, there is increasing emphasis on lowering the costs of new systems. Researchers, for example, are trying to find alternatives to the high-cost superconducting magnets used for magnetic resonance systems. With the decreasing cost of the powerful computers that are currently contained within most imaging systems and with the intense competition among imaging companies, prices for these systems are bound to fall. Other challenges entail presentation of the imaging data. Often multiple modalities are used during a clinical evaluation. If both anatomic and functional information are required, methods to combine and present these data for medical interpretation need to be achieved. The use of medical image data to more effectively execute surgery is a field that is only starting to be explored. How can anatomic data obtained with a tomographic scan be correlated with the surgical field, given that movement of tissues and organs occurs during surgery? Virtual reality is likely to play an important role in this integration of imaging information with surgery. There also are imaging modalities that are just beginning to be intensively explored, such as the detection of impedance and magnetic field data or the use of optical sources and detectors. Engineers and physical scientists are involved throughout the medical imaging field. They are employed by both large and small companies. The names of the medical imaging giants, such as General Electric, Siemens, Picker, and Acuson are familiar to most. Small startup companies are prevalent. In addition to the medical imaging companies, Ph.D.-trained engineers and scientists are employed by departments of engineering, physics, and chemistry in universities and more and more by radiology departments of research-oriented medical centers. Whereas only a few years ago researchers working in the medical imaging field would submit papers to general scientific journals, such as the IEEE Transactions on Biomedical Engineering, now there is a journal, IEEE Transactions on Medical Imaging, devoted to the medical imaging field and journals dedicated to certain modalities, such as Magnetic Resonance in Medicine and Journal of Computer Assisted Tomography. Large scientific meetings for medical imaging, such as the Radiological Society of North America’s annual meeting with over 20,000 attendees, are held each year. Modality-specific meetings, such as that of the Society for Magnetic Resonance in Imaging, have thousands of attendees. Although entire books have been written on each of the medical imaging modalities, this section will provide an overview of the main medical imaging devices and also highlight a few emerging systems. Chapter 61 describes x-ray systems, the backbone of the medical imaging field. X-ray systems are still quite important because of their relatively low system acquisition cost, the low cost of the diagnostic procedures, and the speed with which results are obtained. Chapter 62 describes computed tomographic (CT) systems. This technology became available in the 1970s, with current improvements focused on acquisition speed and data presentations. Chapter 63 highlights magnetic resonance imaging (MRI), a technology that first became available in the 1980s. The technology is rapidly evolving, with major advances recently in the areas of functional and spectroscopic MRI. Nuclear medicine, the subject of Chapter 64, covers both planar and single-photon emission computed tomography (SPECT) systems. Chapter 65 covers ultrasound, the technology that is widely used for obstetrical imaging and vascular


flow evaluation. The latest research on linear and two-dimensional transducers, which will be able to provide real-time three-dimensional ultrasound images, is also covered in this chapter. Less prevalent technologies are presented in Chapters 66 to 69. These include the field of MR microscopy, which requires high-field-strength magnets; position emission tomography (PET), which has a tremendous potential for functional imaging; impedance tomography, which is aimed at constructing images based on difference in conductivity between different body tissues; and virtual reality, which provides an overview of the high-tech field of interactive imaging that is bound to become increasingly important as computing power increases.

VR system used by the disabled.


Shroy, R. E. Jr., Van Lysel, M.S., Yaffe, M. J. “X-Ray.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

61 X-Ray 61.1

Robert E. Shroy, Jr. Picker International

61.2

X-Ray Projection Angiography X-Ray Generation • Image Formation • Digital Angiography • Cine • Summary

Michael S. Van Lysel University of Wisconsin

X-Ray Equipment Production of X-rays • Image Detection: Screen Film Combinations • Image Detection: X-Ray Image Intensifiers with Televisions • Image Detection: Digital Systems

61.3

Martin J. Yaffe University of Toronto

Mammography Principles of Mammography • Physics of Image Formation • Equipment • Quality Control • Stereotactic Biopsy Devices • Digital Mammography • Summary

61.1 X-Ray Equipment Robert E. Shroy, Jr. Conventional x-ray radiography produces images of anatomy that are shadowgrams based on x-ray absorption. The x-rays are produced in a region that is nearly a point source and then are directed on the anatomy to be imaged. The x-rays emerging from the anatomy are detected to form a two-dimensional image, where each point in the image has a brightness related to the intensity of the x-rays at that point. Image production relies on the fact that significant numbers of x-rays penetrate through the anatomy and that different parts of the anatomy absorb different amounts of x-rays. In cases where the anatomy of interest does not absorb x-rays differently from surrounding regions, contrast may be increased by introducing strong x-ray absorbers. For example, barium is often used to image the gastrointestinal tract. X-rays are electromagnetic waves (like light) having an energy in the general range of approximately 1 to several hundred kiloelectronvolts (keV). In medical x-ray imaging, the x-ray energy typically lies between 5 and 150 keV, with the energy adjusted to the anatomic thickness and the type of study being performed. X-rays striking an object may either pass through unaffected or may undergo an interaction. These interactions usually involve either the photoelectric effect (where the x-ray is absorbed) or scattering (where the x-ray is deflected to the side with a loss of some energy). X-rays that have been scattered may undergo deflection through a small angle and still reach the image detector; in this case they reduce image contrast and thus degrade the image. This degradation can be reduced by the use of an air gap between the anatomy and the image receptor or by use of an antiscatter grid. Because of health effects, the doses in radiography are kept as low as possible. However, x-ray quantum noise becomes more apparent in the image as the dose is lowered. This noise is due to the fact that there is an unavoidable random variation in the number of x-rays reaching a point on an image detector. The quantum noise depends on the average number of x-rays striking the image detector and is a fundamental limit to radiographic image quality.


FIGURE 61.1

X-ray tube.

The equipment of conventional x-ray radiography mostly deals with the creation of a desirable beam of x-rays and with the detection of a high-quality image of the transmitted x-rays. These are discussed in the following sections.

Production of X-Rays X-Ray Tube The standard device for production of x-rays is the rotating anode x-ray tube, as illustrated in Fig. 61.1. The x-rays are produced from electrons that have been accelerated in vacuum from the cathode to the anode. The electrons are emitted from a filament mounted within a groove in the cathode. Emission occurs when the filament is heated by passing a current through it. When the filament is hot enough, some electrons obtain a thermal energy sufficient to overcome the energy binding the electron to the metal of the filament. Once the electrons have “boiled off ” from the filament, they are accelerated by a voltage difference applied from the cathode to the anode. This voltage is supplied by a generator (see below). After the electrons have been accelerated to the anode, they will be stopped in a short distance. Most of the electrons’ energy is converted into heating of the anode, but a small percentage is converted to x-rays by two main methods. One method of x-ray production relies on the fact that deceleration of a charged particle results in emission of electromagnetic radiation, called bremmstrahlung radiation. These x-rays will have a wide, continuous distribution of energies, with the maximum being the total energy the electron had when reaching the anode. The number of x-rays is relatively small at higher energies and increases for lower energies. A second method of x-ray production occurs when an accelerated electron strikes an atom in the anode and removes an inner electron from this atom. The vacant electron orbital will be filled by a neighboring electron, and an x-ray may be emitted whose energy matches the energy change of the electron. The result is production of large numbers of x-rays at a few discrete energies. Since the energy of these characteristic x-rays depends on the material on the surface of the anode, materials are chosen


partially to produce x-rays with desired energies. For example, molybdenum is frequently used in anodes of mammography x-ray tubes because of its 20-keV characteristic x-rays. Low-energy x-rays are undesirable because they increase dose to the patient but do not contribute to the final image because they are almost totally absorbed. Therefore, the number of low-energy x-rays is usually reduced by use of a layer of absorber that preferentially absorbs them. The extent to which lowenergy x-rays have been removed can be quantified by the half-value layer of the x-ray beam. It is ideal to create x-rays from a point source because any increase in source size will result in blurring of the final image. Quantitatively, the effects of the blurring are described by the focal spot’s contribution to the system modulation transfer function (MTF). The blurring has its main effect on edges and small objects, which correspond to the higher frequencies. The effect of this blurring depends on the geometry of the imaging and is worse for larger distances between the object and the image receptor (which corresponds to larger geometric magnifications). To avoid this blurring, the electrons must be focused to strike a small spot of the anode. The focusing is achieved by electric fields determined by the exact shape of the cathode. However, there is a limit to the size of this focal spot because the anode material will melt if too much power is deposited into too small an area. This limit is improved by use of a rotating anode, where the anode target material is rotated about a central axis and new (cooler) anode material is constantly being rotated into place at the focal spot. To further increase the power limit, the anode is made with an angle surface. This allows the heat to be deposited in a relatively large spot while the apparent spot size at the detector will be smaller by a factor of the sine of the anode angle. Unfortunately, this angle cannot be made too small because it limits the area that can be covered with x-rays. In practice, tubes are usually supplied with two (or more) focal spots of differing sizes, allowing choice of a smaller (sharper, lower-power) spot or a larger (more blurry, higher-power) spot. The x-ray tube also limits the total number of x-rays that can be used in an exposure because the anode will melt if too much total energy is deposited in it. This limit can be increased by using a more massive anode. Generator The voltages and currents in an x-ray tube are supplied by an x-ray generator. This controls the cathodeanode voltage, which partially defines the number of x-rays made because the number of x-rays produced increases with voltage. The voltage is also chosen to produce x-rays with desired energies: Higher voltages makes x-rays that generally are more penetrating but give a lower contrast image. The generator also determines the number of x-rays created by controlling the amount of current flowing from the cathode to anode and by controlling the length of time this current flows. This points out the two major parameters that describe an x-ray exposure: the peak kilovolts (peak kilovolts from the anode to the cathode during the exposure) and the milliampere-seconds (the product of the current in milliamperes and the exposure time in seconds). The peak kilovolts and milliampere-seconds for an exposure may be set manually by an operator based on estimates of the anatomy. Some generators use manual entry of kilovolts and milliamperes but determine the exposure time automatically. This involves sampling the radiation either before or after the image sensor and is referred to as phototiming. The anode-cathode voltage (often 15 to 150 kV) can be produced by a transformer that converts 120 or 220 V ac to higher voltages. This output is then rectified and filtered. Use of three-phase transformers gives voltages that are nearly constant versus those from single-phase transformers, thus avoiding low kilovoltages that produce undesired low-energy x-rays. In a variation of this method, the transformer output can be controlled at a constant voltage by electron tubes. This gives practically constant voltages and, further, allows the voltage to be turned on and off so quickly that millisecond exposure times can be achieved. In a third approach, an ac input can be rectified and filtered to produce a nearly dc voltage, which is then sent to a solid-state inverter that can turn on and off thousands of times a second. This higher-frequency ac voltage can be converted more easily to a high voltage by a transformer. Equipment operating on this principle is referred to as midfrequency or high-frequency generators.


Image Detection: Screen Film Combinations Special properties are needed for image detection in radiographic applications, where a few high-quality images are made in a study. Because decisions are not immediately made from the images, it is not necessary to display them instantly (although it may be desirable). The most commonly used method of detecting such a radiographic x-ray image uses light-sensitive negative film as a medium. Because high-quality film has a poor response to x-rays, it must be used together with x-ray–sensitive screens. Such screens are usually made with CaWo2 or phosphors using rare earth elements such as doped Gd2O2S or LaOBr. The film is enclosed in a light-tight cassette in contact with an x-ray screen or in between two x-ray screens. When a x-ray image strikes the cassette, the x-rays are absorbed by the screens with high efficiency, and their energy is converted to visible light. The light then exposes a negative image on the film, which is in close contact with the screen. Several properties have been found to be important in describing the relative performance of different films. One critical property is the contrast, which describes the amount of additional darkening caused by an additional amount of light when working near the center of a film’s exposure range. Another property, the latitude of a film, describes the film’s ability to create a usable image with a wide range in input light levels. Generally, latitude and contrast are competing properties, and a film with a large latitude will have a low contrast. Additionally, the modulation transfer function (MTF) of a film is an important property. MTF is most degraded at higher frequencies; this high-frequency MTF is also described by the film’s resolution, its ability to image small objects. X-ray screens also have several key performance parameters. It is essential that screens detect and use a large percentage of the x-rays striking them, which is measured as the screen’s quantum detection efficiency. Currently used screens may detect 30% of x-rays for images at higher peak kilovolts and as much 60% for lower peak kilovolt images. Such efficiencies lead to the use of two screens (one on each side of the film) for improved x-ray utilization. As with films, a good high-frequency MTF is needed to give good visibility of small structures and edges. Some MTF degradation is associated with blurring that occurs when light spreads as it travels through the screen and to the film. This leads to a compromise on thickness; screens must be thick enough for good quantum detection efficiency but thin enough to avoid excess blurring. For a film/screen system, a certain amount of radiation will be required to produce a usable amount of film darkening. The ability of the film/screen system to make an image with a small amount of radiation is referred to as its speed. The speed depends on a number of parameters: the quantum detection efficiency of the screen, the efficiency with which the screen converts x-ray energy to light, the match between the color emitted by the screen and the colors to which the film is sensitive, and the amount of film darkening for a given amount of light. The number of x-rays used in producing a radiographic image will be chosen to give a viewable amount of exposure to the film. Therefore, patient dose will be reduced by the use of a high-speed screen/film system. However, high-speed film/screen combinations gives a “noisier” image because of the smaller number of x-rays detected in its creation.

Image Detection: X-Ray Image Intensifiers with Televisions Although screen-film systems are excellent for radiography, they are not usable for fluoroscopy, where lower x-ray levels are produced continuously and many images must be presented almost immediately. Fluoroscopic images are not used for diagnosis but rather as an aid in performing tasks such as placement of catheters in blood vessels during angiography. For fluoroscopy, x-ray image intensifiers are used in conjunction with television cameras. An x-ray image intensifier detects the x-ray image and converts it to a small, bright image of visible light. This visible image is then transferred by lenses to a television camera for final display on a monitor. The basic structure of an x-ray image intensifier is shown in Fig. 61.2. The components are held in a vacuum by an enclosure made of glass and/or metal. The x-rays enter through a low-absorption window and then strike an input phosphor usually made of doped CsI. As in the x-ray screens described above, the x-rays are converted to light in the CsI. On top of the CsI layer is a photoemitter, which absorbs the © 2000 by CRC Press LLC

FIGURE 61.2

X-ray image intensifier.

light and emits a number of low-energy electrons that initially spread in various directions. The photoelectrons are accelerated and steered by a set of grids that have voltages applied to them. The electrons strike an output phosphor structure that converts their energy to the final output image made of light. This light then travels through an output window to a lens system. The grid voltages serve to add energy to the electrons so that the output image is brighter. Grid voltages and shapes are also chosen so that the x-ray image is converted to a light image with minimal distortion. Further, the grids must be designed to take photoelectrons that are spreading from a point on the photoemitter and focus them back together at a point on the output phosphor. It is possible to adjust grid voltages on an image intensifier so that it has different fields of coverage. Either the whole input area can be imaged on the output phosphor, or smaller parts of the input can be imaged on the whole output. Use of smaller parts of the input is advantageous when only smaller parts of anatomy need to be imaged with maximum resolution and a large display. For example, an image intensifier that could cover a 12-in-diameter input also might be operated so that a 9-in-diameter or 6-in-diameter input covers all the output phosphor. X-ray image intensifiers can be described by a set of performance parameters not unlike those of screen/film combinations. It is important that x-rays be detected and used with a high efficiency; current image intensifiers have quantum detection efficiencies of 60% to 70% for 59-keV x-rays. As with film/screens, a good high-frequency MTF is needed to image small objects and sharp edges without blurring. However, low-frequency MTF also must be controlled carefully in image intensifiers, since it can be degraded by internal scattering of x-rays, photoelectrons, and light over relatively large distances. The amount of intensification depends on brightness and size of the output image for a given x-ray input. This is described either by the gain (specified relative to a standard x-ray screen) or by conversion efficiency [a light output per radiation input measured in (cd/m2)/(mR/min)]. Note that producing a smaller output image is as important as making a light image with more photons because the small image can be handled more efficiently by the lenses that follow. Especially when imaging the full input area, image intensifiers introduce a pincushion distortion into the output image. Thus a square object placed off-center will produce an image that is stretched in the direction away from the center. © 2000 by CRC Press LLC

Although an image intensifier output could be viewed directly with a lens system, there is more flexibility when the image intensifier is viewed with a television camera and the output is displayed on a monitor. Televisions are currently used with pickup tubes and with CCD sensors. When a television tube is used, the image is focused on a charged photoconducting material at the tube’s input. A number of materials are used, including SbS3, PbO, and SeTeAs. The light image discharges regions of the photoconductor, converting the image to a charge distribution on the back of the photoconducting layer. Next, the charge distribution is read by scanning a small beam of electrons across the surface, which recharges the photoconductor. The recharging current is proportional to the light intensity at the point being scanned; this current is amplified and then used to produce an image on a monitor. The tube target is generally scanned in an interlaced mode in order to be consistent with broadcast television and allow use of standard equipment. In fluoroscopy, it is desirable to use the same detected dose for all studies so that the image noise is approximately constant. This is usually achieved by monitoring the image brightness in a central part of the image intensifier’s output, since brightness generally increases with dose. The brightness may be monitored by a photomultiplier tube that samples it directly or by analyzing signal levels in the television. However, maintaining a constant detected dose would lead to high patient doses in the case of very absorptive anatomy. To avoid problems here, systems are generally required by federal regulations to have a limit on the maximum patient dose. In those cases where the dose limit prevents the image intensifier from receiving the usual dose, the output image becomes darker. To compensate for this, television systems are often operated with automatic gain control that gives an image on the monitor of a constant brightness no matter what the brightness from the image intensifier.

Image Detection: Digital Systems In both radiography and fluoroscopy, there are advantages to the use of digital images. This allows image processing for better displayed images, use of lower doses in some cases, and opens the possibility for digital storage with a PACS system or remote image viewing via teleradiology. Additionally, some digital systems provide better image quality because of fewer processing steps, lack of distortion, or improved uniformity. A common method of digitizing medical x-ray images uses the voltage output from an image-intensifier/TV system. This voltage can be digitized by an analog-to-digital converter at rates fast enough to be used with fluoroscopy as well as radiography. Another technology for obtaining digital radiographs involves use of photostimulable phosphors. Here the x-rays strike an enclosed sheet of phosphor that stores the x-ray energy. This phorphor can then be taken to a read-out unit, where the phosphor surface is scanned by a small light beam of proper wavelength. As a point on the surface is read, the stored energy is emitted as visible light, which is then detected, amplified, and digitized. Such systems have the advantage that they can be used with existing systems designed for screen-film detection because the phosphor sheet package is the same size as that for screen films. A new method for digital detection involves use of active-matrix thin-film-transistor technology, in which an array of small sensors is grown in hydrogenated amorphous silicon. Each sensor element includes an electrode for storing charge that is proportional to its x-ray signal. Each electrode is coupled to a transistor that either isolates it during acquisition or couples it to digitization circuitry during readout. There are two common methods for introducing the charge signal on each electrode. In one method, a layer of x-ray absorber (typically selenium) is deposited on the array of sensors; when this layer is biased and x-rays are absorbed there, their energy is converted to electron-hole pairs and the resulting charge is collected on the electrode. In the second method, each electrode is part of the photodiode that makes electron-hole pairs when exposed to light; the light is produced from x-rays by a layer of scintillator (such as CsI) that is deposited on the array. Use of a digital system provides several advantages in fluoroscopy. The digital image can be processed in real time with edge enhancement, smoothing, or application of a median filter. Also, frame-to-frame


averaging can be used to decrease image noise, at the expense of blurring the image of moving objects. Further, digital fluoroscopy with TV system allows the TV tube to be scanned in formats that are optimized for read-out; the image can still be shown in a different format that is optimized for display. Another advantage is that the displayed image is not allowed to go blank when x-ray exposure is ended, but a repeated display of the last image is shown. This last-image-hold significantly reduces doses in those cases where the radiologist needs to see an image for evaluation, but does not necessarily need a continuously updated image. The processing of some digital systems also allows the use of pulsed fluoroscopy, where the x-rays are produced in a short, intense burst instead of continuously. In this method the pulses of x-rays are made either by biasing the x-ray tube filament or by quickly turning on and off the anode-cathode voltage. This has the advantage of making sharper images of objects that are moving. Often one x-ray pulse is produced for every display frame, but there is also the ability to obtain dose reduction by leaving the x-rays off for some frames. With such a reduced exposure rate, doses can be reduced by a factor of two or four by only making x-rays every second or fourth frame. For those frames with no x-ray pulse, the system repeats a display of the last frame with x-rays.

Defining Terms Antiscatter grid: A thin structure made of alternating strips of lead and material transmissive to x-rays. Strips are oriented so that most scattered x-rays go through lead sections and are preferentially absorbed, while unscattered x-rays go through transmissive sections. Focal spot: The small area on the anode of an x-ray tube from where x-rays are emitted. It is the place where the accelerated electron beam is focused. Half-value layer (HVL): The thickness of a material (often aluminum) needed to absorb half the x-ray in a beam. keV: A unit of energy useful with x-rays. It is equal to the energy supplied to an electron when accelerated through 1 kilovolt. Modulation transfer function (MTF): The ratio of the contrast in the output image of a system to the contrast in the object, specified for sine waves of various frequencies. Describes blurring (loss of contrast) in an imaging system for different-sized objects. Quantum detection efficiency: The percentage of incident x-rays effectively used to create an image.

References Bushberg JT, Seibert JA, Leidholdt EM, Boone JM. 1994. The Essential Physics of Medical Imaging. Baltimore, Williams & Wilkins. Curry TS, Dowdey JE, Murry RC. 1984. Christensen’s Introduction to the Physics of Diagnostic Radiology. Philadelphia, Lea & Febiger. Hendee WR, Ritenour R. 1992. Medical Imaging Physics: St. Louis, Mosby-Year Book. Ter-Pogossian MM. 1969. The Physical Aspects of Diagnostic Radiology. New York, Harper & Row.

Further Information Medical Physics is a monthly scientific and informational journal published for the American Association of Physicists in Medicine. Papers here generally cover evaluation of existing medical equipment and development of new systems. For more information, contact the American Association of Physicists in Medicine, One Physics Ellipse, College Park, MD 20740-3846. The Society of Photo-Optical Instrumentation Engineers (SPIE) sponsors numerous conferences and publishes their proceedings. Especially relevant is the annual conference on Medical Imaging. Contact SPIE, P.O. Box 10, Bellham, WA 98277-0010.


Several corporations active in medical imaging work together under the National Electrical Manufacturers Association to develop definitions and testing standards for equipment used in the field. Information can be obtained from NEMA, 2101 L Street, N.W., Washington, DC 20037. Critical aspects of medical x-ray imaging are covered by rules of the Food and Drug Administration, part of the Department of Health and Human Services. These are listed in the Code of Federal Regulations, Title 21. Copies are for sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402.

61.2 X-Ray Projection Angiography Michael S. Van Lysel Angiography is a diagnostic and, increasingly, therapeutic modality concerned with diseases of the circulatory system. While many imaging modalities (ultrasound, computed tomography, magnetic resonance imaging, angioscopy) are now available, either clinically or for research, to study vascular structures, this section will focus on projection radiography. In this method, the vessel of interest is opacified by injection of a radiopaque contrast agent. Serial radiographs of the contrast material flowing through the vessel are then acquired. This examination is performed in an angiographic suite, a special procedures laboratory, or a cardiac catheterization laboratory. Contrast material is needed to opacify vascular structures because the radiographic contrast of blood is essentially the same as that of soft tissue. Contrast material consists of an iodine-containing (Z = 53) compound, with maximum iodine concentrations of about 350 mg/cm3. Contrast material is injected through a catheter ranging in diameter roughly from 1 to 3 mm, depending on the injection flow rate to be used. Radiographic images of the contrast-filled vessels are recorded using either film or video. Digital imaging technology has become instrumental in the acquisition and storage of angiographic images. The most important application of digital imaging is digital subtraction angiography (DSA). Temporal subtraction is a DSA model in which a preinjection image (the mask) is acquired, the injection of contrast agent is then performed, and the sequential images of the opacified vessel(s) are acquired and subtracted from the mask. The result, ideally, is that the fixed anatomy is canceled, allowing contrast enhancement (similar to computed tomographic windowing and leveling) to provide increased contrast sensitivity. An increasingly important facet of the angiographic procedure is the use of transluminal interventional techniques to effect a therapeutic result. These techniques, including angioplasty, atherectomy, laser ablation, and intraluminal stents, rely on digital angiographic imaging technology to facilitate the precise catheter manipulations necessary for a successful result. In fact, digital enhancement, storage, and retrieval of fluoroscopic images have become mandatory capabilities for digital angiographic systems. Figure 61.3 is a schematic representation of an angiographic imaging system. The basic components include an x-ray tube and generator, image intensifier, video camera, cine camera (optional for cardiac imaging), and digital image processor.

X-ray Generation Angiographic systems require a high-power, sophisticated x-ray generation system in order to produce the short, intense x-ray pulses needed to produce clear images of moving vessels. Required exposure times range from 100 to 200 ms for cerebral studies to 1 to 10 ms for cardiac studies. Angiographic systems use either a constant potential generator, or, increasingly, a medium/high-frequency inverter generator. Power ratings for angiographic generators are generally greater than or equal to 80 kW at 100 kW and must be capable of producing reasonably square x-ray pulses. In most cases (pediatric cardiology being an exception), pulse widths of 5 ms or greater are necessary to keep the x-ray tube potential in the desirable, high-contrast range of 70 to 90 kVp.


FIGURE 61.3 Schematic diagram of an image intensifier-based digital angiographic and cine imaging system. Solid arrows indicate image signals, and dotted arrows indicate control signals.

The x-ray tube is of the rotating-anode variety. Serial runs of high-intensity x-ray pulses result in high heat loads. Anode heat storage capacities of 1 mega-heat units (MHU) or greater are required, especially in the case of cine angiography. Electronic “heat computers,” which calculate and display the current heat load of the x-ray tube, are very useful in preventing damage to the x-ray tube. In a high-throughput angiographic suite, forced liquid cooling of the x-ray tube housing is essential. Angiographic x-ray tubes are of multifocal design, with the focal spot sizes tailored for the intended use. A 0.6-mm (50-kW loading), 1.2-mm (100-kW loading) bifocal insert is common. The specified heat loads for a given focal spot are for single exposures. When serial angiographic exposures are performed, the focal spot load limit must be derated (reduced) to account for the accumulation of heat in the focal spot target during the run. Larger focal spots (e.g., 1.8 mm) can be obtained for high-load procedures. A smaller focal spot [e.g., 0.3 and 0.1 mm (bias)] is needed for magnification studies. Small focal spots have become increasingly desirable for high-resolution fluoroscopic interventional studies performed with 1000-line video systems.

Image Formation Film From the 1950s through to the 1980s, the dominant detector system used for recording the angiographic procedure was film/screen angiography using a rapid film changer [Amplatz, 1997]. However, digital angiography, described below, has almost completely replaced this technology. Film provides a higher spatial resolution than digital. However, as the spatial resolution of digital continues to improve, the overwhelming advantages of ease of use and image processing algorithms have prevailed. While film changers can still be found in use, few new film/screen systems are being purchased. One application where film use is still common is cardiac imaging. Cine angiography is performed with a 35-mm motion picture camera optically coupled to the image intensifier output phosphor


FIGURE 61.4 Schematic diagram of the optical distributor used to couple the image intensifier (I.I.) output phosphor to a video and cine camera. (From Van Lysel MS. Digital angiography. In S Baum (ed), Abrams’ Angiography, 4th ed. Boston, Little, Brown, with permission.)

(Fig. 61.4). The primary clinical application is coronary angiography and ventriculography. During cine angiography, x-ray pulses are synchronized both with the cine camera shutter and the vertical retrace of the system video camera. To limit motion blurring, it is desirable to keep the x-ray pulses as short as possible, but no longer than 10 msec. Imaging runs generally last from 5 to 10 sec in duration at a frame rate of 30 to 60 frames/sec (fps). Sixty fps is generally used for pediatric studies, where higher heart rates are encountered, while 30 fps is more typical for adult studies (“digital cine,” discussed below, generally is performed at 15 fps). Some cine angiographic installations provide biplane imaging in which two independent imaging chains can acquire orthogonal images of the injection sequence. The eccentricity of coronary lesions and the asymmetric nature of cardiac contraction abnormalities require that multiple x-ray projections be acquired. Biplane systems allow this to be done with a smaller patient contrast load. For this reason, biplane systems are considered a requirement for pediatric cath labs. They are relatively uncommon for adult cath labs, however, where the less complicated positioning of a single plane system is often valued over the reduced number of injections possible with a biplane system. The AP and lateral planes of a biplane system are energized out of phase, which results in the potential for image degradation due to detection of the radiation scattered from the opposite plane. Image intensifier blanking, which shuts down the accelerating voltage of the non-imaging plane, is used to eliminate this problem. While film remains in common use for cardiac imaging, here too the medium is being replaced by digital. Digital angiographic systems (described below) were integrated into the cardiac catheterization


imaging system beginning in the 1980s, to provide short term image storage and replay. However, the high resolution and frame rate required for cardiac imaging precluded the use of digital hardware for long-term (greater than one day) storage. Film and cine are acquired simultaneously using the beam splitting mirror in the optical distributor (Fig. 61.4). This situation is rapidly changing, as high-speed video servers and the recordable compact disc (CD-R) now provide acceptable performance for this application. Most new cardiac imaging systems are purchased without a cine film camera. Image Intensifier/Video Camera The image intensifier (II) is fundamental to the modern angiographic procedure. The purpose of the image intensifier is (1) to produce a light image with sufficient brightness to allow the use of video and film cameras and (2) to produce an output image of small enough size to allow convenient coupling to video and film cameras. The image intensifier provides both real-time imaging capability (fluoroscopy), which allows patient positioning and catheter manipulation, and recording of the angiographic injection (digital angiography, analog video recording, photospot, cine). Image intensifier output phosphors are approximately 25 mm in diameter, although large (e.g., 50 to 60 mm) output phosphor image intensifiers have been developed to increase spatial resolution. The modulation transfer function (MTF) of the image intensifier is determined primarily by the input and output phosphor stages, so mapping a given input image to a larger output phosphor will improve the MTF of the system. The input phosphor of a modern image intensifier is cesium iodide (CsI). The largest currently available image intensifier input phosphors are approximately 16 in. The effective input phosphor diameter is selectable by the user. For example, an image intensifier designated 9/7/5 allows the user to select input phosphor diameters of 9, 7, or 5 in. These selections are referred to as image intensifier modes. The purpose of providing an adjustable input phosphor is to allow the user to trade off between spatial resolution and field of view. A smaller mode provides higher spatial resolution both because the MTF of the image intensifier improves and because it maps a smaller field of view to the fixed size of the video camera target. Generally speaking, angiographic suites designed exclusively for cardiac catheterization use 9-in. image intensifiers, neuroangiography suites use 12-in. intensifiers, while suites that must handle pulmonary, renal, or peripheral angiography require the larger (i.e., 14 to 16 in.) intensifiers. The brightness gain of the image intensifier derives from two sources: (1) the increase in electron energy produced by the accelerating potential (the flux gain) and (2) the decrease in size of the image as it is transferred from the input to the output phosphor (the minification gain). The product of these two factors can exceed 5000. However, since the minification gain is a function of the area of the input phosphor exposed to the radiation beam (i.e., the image intensifier mode), the brightness gain drops as smaller image intensifier modes are selected. This is compensated for by a combination of increasing x-ray exposure to maintain the image intensifier light output and opening the video camera aperture. Image intensifier brightness gain declines with age and must be monitored to allow timely replacement. The specification used for this purpose is the image intensifier conversion factor, defined as the light output of the image intensifier per unit x-ray exposure input. Modern image intensifiers have a conversion factor of 100 cd/m2/mR/s or more for the 9-in. mode. With the increasing emergence of digital angiography as the primary angiographic imaging modality, image intensifier performance has become increasingly important. In the field, the high-spatial-frequency of an image intensifier is assessed by determining the limiting resolution [in the neighborhood of 4 to 5 line-pairs/mm (lp/mm) in the 9-in mode], while the low-spatial-frequency response is assessed using the contrast ratio (in the neighborhood of 15:1 to 30:1). The National Electrical Manufacturers Association (NEMA) has defined test procedures for measuring the contrast ratio [NEMA, 1992]. The detective quantum efficiency (DQE), which is a measure of the efficiency with which the image intensifier utilizes the x-ray energy incident on it, is in the neighborhood of 65% (400 µm phosphor thickness, 60 keV). A tabulation of the specifications of several commercially available image intensifiers has been published by Siedband [1994].


Optical Distributor The image present at the image intensifier output phosphor is coupled to the video camera, and any film camera present (e.g., cine), by the optical distributor. The components of the distributor are shown in Fig. 61.4. There is an aperture for each camera to allow the light intensity presented to each camera to be adjusted independently. The video camera aperture is usually a motor-driven variable iris, while the film camera aperture is usually fixed. It is important to realize that while the aperture does ensure that the proper light level is presented to the camera, more fundamentally, the aperture determines the x-ray exposure input to the image intensifier. As a result, both the patient exposure and the level of quantum noise in the image are set by the aperture. The noise amplitude in a fluoroscopic or digital angiographic image is inversely proportional to the f-number of the optical system. Because the quantum sink of a properly adjusted fluorographic system is at the input of the image intensifier, the aperture diameter is set, for a given type of examination, to provide the desired level of quantum mottle present in the image. The x-ray exposure factors are then adjusted for each patient, by an automatic exposure control (AEC) system, to produce the proper postaperture light level. However, some video systems do provide for increasing the video camera aperture during fluoroscopy when the maximum entrance exposure does not provide adequate light levels on a large patient. The beam-splitting mirror was originally meant to provide a moderate-quality video image simultaneous with cine recording in order to monitor the contrast injection during cardiac studies. More recently, as the importance of digital angiography has mushroomed, precision-quality mirrors with higher transmission have been used in order to provide simultaneous diagnostic-quality cine and video. The latest development has been the introduction of cine-less digital cardiac systems, in which the video image is the sole recording means. The introduction of these systems has sometimes been accompanied by the claim that a cine-less system requires less patient exposure due to the fact that light does not have to be provided to the cine camera. However, a conventional cine system operates with an excess of light (i.e., the cine camera aperture is stopped down). Because the image intensifier input is the quantum sink of the system, exposure is determined by the need to limit quantum mottle, not to maintain a given light level at the image intensifier output. Therefore, the validity of this claim is dubious. It should be noted, however, that because of the difference in spatial resolution capabilities of cine and video, the noise power spectrum of images acquired with equal exposure will be different. It is possible that observers accept a lower exposure in a video image than in the higher-resolution film image. Video System The video system in an angiographic suite consists of several components, including the camera head, camera control unit (CCU), video monitors, and video recording devices. In addition, a digital image processor is integrated with the video system. The video camera is responsible for signal generation. Traditionally, pickup-tube based cameras (discussed below) have been used. Recently, high resolution (10242), high frame rate (30 frame/sec) chargecoupled device (CCD) video cameras have become available and are replacing tube-based cameras in some angiographic installations. This trend will probably continue. The image quality ramifications are a matter of current research [Blume, 1998]. The advantages of CCD cameras include low-voltage operation, reliability, little required setup tuning, and freedom from geometric distortions. Frame-charge transfer is the preferred CCD read-out scheme, due to the higher optical fill factor of this configuration. Video camera pickup tubes used for angiography are of the photoconductive vidicon-style of construction. This type of tube uses a low-velocity scanning beam and generates a signal from the recharge of the target by the scanning electron beam. There are several types of vidicon-style tubes in use (Plumbicon, Primicon, Saticon, Newvicon) which differ from each other in the material and configuration used for target construction. There is an unfortunate confusion in terminology because the original vidicon-style tube is referred to simply as a vidicon. The original vidicon has an antimony trisulfide target (Sb2S3) and exhibits such a high degree of lag (image retention) that it is not useful for angiographic work. Even with low-lag angiographic cameras, the residual lag can result in artifacts in subtraction images. Light bias is


often used to further reduce lag by ensuring that the target surface is not driven to a negative potential by energetic electrons in the scanning beam [Sandrik, 1984]. Image noise in a well-designed system is due to x-ray quantum fluctuations and noise related to signal generation in the video camera. When used for digital angiographic purposes, it is important that the video camera exhibit a high signal-to-noise ratio (at least 60 dB) so that video camera noise does not dominate the low-signal (dark) portions of the image. In order to achieve this, the pickup tube must be run at high beam currents (2 to 3 µA). Because long-term operation of the tube at these beam currents can result in damage to the target, beam current is usually either blanked when imaging is not being performed, or the current is held at a low level (e.g., 400 nA) and boosted only when high-quality angiographic images are required. All pickup tubes currently in use for angiographic imaging exhibit a linear response to light input (i.e., γ = 1, where the relationship between signal current I and image brightness B is described by a relationship of the form I/Io = (B/Bo)γ). This has the disadvantage that the range in image brightness presented to the camera often exceeds the camera’s dynamic range when a highly transmissive portion of the patient’s anatomy (e.g., lung) or unattenuated radiation is included in the image field, either saturating the highlights, forcing the rest of the image to a low signal level, or both. To deal with this problem, it is desirable for the operator to mechanically bolus the bright area with metal filters, saline bags, etc. Specially constructed filters are available for commonly encountered problems, such as the transmissive region between the patient’s legs during runoff studies of the legs, and most vendors provide a controllable metal filter in the x-ray collimator that the operator can position over bright spots with a joystick. In addition to mechanical bolusing performed by the laboratory staff, most system vendors have incorporated some form of gamma curve modification into the systems. Usually performed using analog processing in the CCU, the technique applies a nonlinear transfer curve to the originally linear data. There are two advantages to this technique. First, the CRT of the display monitor has reduced gain at low signal, so imposing a transfer function with γ ≈ 0.5 via gamma-curve modification provides a better match between the video signal and the display monitor. Second, if the modification is performed prior to digitization, a γ < 1 results in a more constant ratio between the ADC step size and the image noise amplitude across the full range of the video signal. This results in less contouring in the dark portions of the digital image, especially in images that have been spatially filtered. It is important to note, however, that gamma-curve modification does not eliminate the desirable effects of mechanically bolusing the image field prior to image acquisition. This is so because bolusing allows more photon flux to be selectively applied to the more attenuating regions of the patient, which decreases both quantum and video noise in those regions. Bolusing is especially important for subtraction imaging. The video system characteristic most apparent to the user is the method employed in scanning the image. Prior to the advent of the digital angiography, EIA RS-170 video (525-line, 30 Hz frames, 2:1 interlace) was the predominant standard used for fluoroscopic systems in the United States. However, this method of scanning has definite disadvantages for angiography, including low resolution and image artifacts related to the interlaced format. The inclusion of a digital image processor in the imaging chain, functioning as an image buffer, allows the scanning mode to be tailored to the angiographic procedure. Many of the important video scanning modes used for angiographic work are dependent on the ability of image processors to perform scan conversion operations. Two typical scan conversion operations are progressive-to-interlaced conversion and upscanning. Progressive-to-interlaced scan conversion allows progressive scanning (also referred to as sequential scanning) to be used for image acquisition and interlaced scanning for image display. Progressive scanning is a noninterlaced scan mode in which all the lines are read out in a single vertical scan. Progressive scanning is especially necessary when imaging moving arteries, such as during coronary angiography [Seibert et al., 1984]. In noncardiac work, progressive scan acquisition is usually combined with beam blanking. Beam blanking refers to the condition in which the pickup tube beam current is blanked (turned off) for one or more integer number of frames. This mode is used in order to allow the image to integrate on the camera target prior to readout (Fig. 61.5). In this way, x-ray pulses shorter than one frame period can be acquired without scanning artifacts, and x-ray pulses longer than one frame period can be used in order to increase the x-ray quantum statistics © 2000 by CRC Press LLC

FIGURE 61.5 Timing diagram for image acquisition using the pulsed-progressive mode. (From Van Lysel MS. Digital angiography. In S Baum (ed), Abrams’ Angiography, 4th ed. Boston, Little, Brown, with permission.)

of an image. Upscanning refers to the acquisition of data at a low line rate (e.g., 525 lines) and the display of that data at a higher line rate (e.g., 1023 lines) [Holmes et al., 1989]. The extra lines are produced by either replication of the actual data or, more commonly, by interpolation (either linear or spline). Upscanning also can be performed in the horizontal direction as well, but this is less typical. Upscanning is used to decrease the demands on the video camera, system bandwidth, and digital storage requirements while improving display contrast and decreasing interfield flicker. The image buffering capability of a digital system can provide several operations aimed at reducing patient exposure. If fluoroscopy is performed with pulsed rather than continuous x-rays, then the operator has the freedom to choose the frame rate. During pulsed-progressive fluoroscopy, the digital system provides for display refresh without flicker. Frame rates of less than 30 frames per second can result in lower patient exposure, though; because of the phenomenon of eye integration, the x-ray exposure per pulse must be increased as the frame rate drops in order to maintain low contrast detectability [Aufrichtig et al., 1994]. Last image hold, which stores and displays the last acquired fluoroscopic frame, also can result in an exposure reduction. Combining last image hold with graphic overlays allows collimator shutters and bolus filters to be positioned with the x-rays turned off.

Digital Angiography Digital imaging technology has quickly replaced film-based recording for most angiographic procedures. Digital image processing provides the ability to manipulate the contrast and spatial-frequency characteristics of the angiographic image, as well as providing immediate access to the image data during the procedure. The rapid application of digital imaging technology to the angiographic procedure was facilitated by the fact that the image intensifier and video camera imaging chain was already in use when digital imaging appeared. It is a relatively simple matter to digitize the video camera output. Theoretically, additional noise is added to the image due to quantitization errors associated with the digitization process. This additional noise can be kept insignificantly small by using a sufficient number of digital levels so that © 2000 by CRC Press LLC

FIGURE 61.6 Detector modulation transfer function, including limits due to the image intensifier (II), video camera (TV), and sampling, including the antialiasing filter (pixels). Figures are for the 6.5-in image intensifier mode for 512- and 1024-pixel matrices. In both cases, the total detector MTF is given by the solid line. Data used for this figure from Verhoeven [1985].

the amplitude of one digital level is approximately equal to the amplitude of the standard deviation in image values associated with the noise (x-ray quantum and electronic noise) in the image prior to digitization [Kruger et al., 1981]. To meet this condition, most digital angiographic systems employ a 10-bit (1024-level) analog-to-digital convertor (ADC). Those systems which are designed to digitize highnoise (i.e., low x-ray exposure) images exclusively can employ an 8-bit (256-level) ADC. Such systems include those designed for cardiac and fluoroscopic applications. The spatial resolution of a digital angiographic image is determined by several factors, including the size of the x-ray tube focal spot, the modulation transfer function of the image intensifier-video camera chain, and the size of the pixel matrix. The typical image matrix dimensions for non-cardiac digital angiographic images is 1024 × 1024. Cardiac systems use 512 × 512, 1024(H) × 512(V), or 1024 × 1024. Sampling rates required to digitize 5122 and 10242 matrices in the 33-ms frame period of conventional video are 10 and 40 MHz, respectively. Figure 61.6 shows an example of the detector MTF of a digital angiographic system. The product of the image intensifier and video system MTF constitute the presampling MTF [Fujita et al., 1985]. It is seen that the effect of sampling with a 512-pixel matrix is to truncate the high-frequency tail of the presampling MTF, while a 1024-pixel matrix imposes few additional limitations beyond that associated with the analog components (especially for small image intensifier modes) [Verhoeven, 1985]. However, because of blurring associated with the x-ray tube focal spot, the full advantage of the higher density pixel matrix can rarely be realized, clinically (Fig. 61.7). Digital Image Processor The digital image processor found in a modern angiographic suite is a dedicated device designed specially to meet the demands of the angiographic procedure. Hardware is structured as a pipeline processor to perform real-time processing at video rates. Image-subtraction, integration, spatial-filtration, and temporal-filtration algorithms are hardwired to meet this requirement. Lookup tables (LUTs) are used to perform intensity transformations (e.g., contrast enhancement and logarithmic transformation). A more general-purpose host computer is used to control the pipeline processor and x-ray generator, respond to user input, and perform non-real-time image manipulations. The most clinically important image-processing algorithm is temporal subtraction (DSA). Subtraction imaging is used for most vascular studies. Subtraction allows approximately a factor of 2 reduction in the amount of injected contrast material. As a result, DSA studies can be performed with less contrast load and with smaller catheters than film/screen angiography. The primary limitation of DSA is a susceptibility to misregistration artifacts resulting from patient motion. Some procedures that are particularly susceptible to motion artifacts are routinely performed in a nonsubtracted mode. Unsubtracted digital angiographic studies are usually performed with the same amount of contrast material as © 2000 by CRC Press LLC

FIGURE 61.7 Experimental determination of the limiting resolution (high-contrast object and high x-ray exposure) for 512- and 1024-pixel matrices, focal spots of actual dimensions 1.0 and 1.5 mm, and a 6.5-in image intensifier mode, as a function of geometric magnification. (From Mistretta CA, Peppler WW. 1988. Digital cardiac x-ray imaging: Fundamental principles. Am J Cardiac Imaging 2:26, with permission.)

film/screen studies. Cardiac angiography is one procedure that is generally performed without subtraction, although in any particular study, if patient and respiratory motion are absent, it is possible to obtain high-quality time subtractions using phase-matched mask subtractions in which the preinjection mask and postinjection contrast images are matched with respect to cardiac phase. In order to do this efficiently, it is necessary to digitize the ECG signal along with the image data. Additional examples of unsubtracted digital angiographic studies are those in uncooperative patients (e.g., trauma) and digital runoff studies of the vessels in the leg. In a digital runoff study it is necessary to follow the bolus down the legs by moving the patient (or the image gantry). This motion causes difficulties with both mask registration and uniform exposure intensities between the mask and contrast images. The high contrast sensitivity of DSA is valuable for runoff studies, however, because the small vessels and slow flow in the lower extremities can make vessel visualization difficult. Recently, making use of programmed table (or gantry) motion and pixel-shifting strategies, x-ray vendors have begun to offer a viable digital subtraction runoff mode. In addition to subtraction angiography, two filtration algorithms have become clinically important. The first is high-pass spatial filtration for the purposes of providing edge enhancement. Real-time edge enhancement of fluoroscopy is especially important for interventional procedures, such as angioplasty, where the task requires visualization of high-frequency objects such as vessel edges and guidewires. Because increasing the degree of edge enhancement also increases image noise, operator control of the degree of enhancement (accomplished by adjusting the size and weighing of the convolution kernel) is an important feature.


The second filtration algorithm often available from a digital angiographic system is low-pass temporal filtration (recursive filtering) [Rowlands, 1992]. Temporal filtration is used to reduce quantum noise levels in fluoroscopic images without increasing patient exposure. Recursive filtering is a more desirable method than simple image integration because it requires only two memory planes and because it is a simple matter to turn the filtering algorithm off, on a pixel-by-pixel basis, when motion is detected. Motion-detection circuits monitor the frame-to-frame change in pixel values and assume that an object has moved into or out of the pixel if the change exceeds a preset threshold. Image Storage Storage needs during the angiographic procedure are easily met by modern hard drives. These often are configured to provide real-time (i.e., video rate) storage. The immediate access to images provided by real-time disk technology is one of the major advantages of digital angiography over film. Not only is it unnecessary to wait for film to be developed, but review of the image data after the patient’s procedure is completed is facilitated by directories and specialized review software. For example, a popular image menu feature is the presentation to users of a low-resolution collage of available images from which they may select a single image or entire run for full-resolution display. While online storage of recent studies is a strength of digital angiography, long-term (archival) storage is a weakness. Archival devices provided by vendors are generally proprietary devices that make use of various recording media. There is an established communications protocol (ACR-NEMA) [NEMA, 1993] for network transfer of images. For the time being, while large institutions and teaching hospitals are investing in sophisticated digital picture archiving and communications systems (PACS), archival needs at most institutions are met by storage of hardcopy films generated from the digital images. Hardcopy devices include multiformat cameras (laser or video) and video thermal printers. Laser cameras, using either an analog or a digital interface to the digital angiographic unit, can provide diagnostic-quality, large-format, high-contrast, high-resolution films. Multiformat video cameras, which expose the film with a CRT, are a less expensive method of generating diagnostic-quality images but are also more susceptible to drift and geometric distortions. Thermal printer images are generally used as convenient method to generate temporary hardcopy images. Cardiac angiography labs have long employed a similar method, referred to as parallel cine, in which both digital and cine images are recorded simultaneously by use of the semitransparent mirror in the image intensifier optical distributor. Immediate diagnosis would be performed off the digital monitor while the 35-mm film would provide archival storage and the ability to share the image data with other institutions. However, due to the promulgation of an image transfer standard using the recordable CD (CD-R), many laboratories are abandoning cine film. While CD-R cannot replay images at full real-time rates, the read-rate (>1 MB/sec) is sufficient to load old data quickly from CD-R to a workstation hard drive for review. Low volume laboratories can use CD-R as an archival storage method. Higher volume labs are installing networked video file servers to provide online or near-line access to archived patient studies.

Summary For decades, x-ray projection film/screen angiography was the only invasive modality available for the diagnosis of vascular disease. Now several imaging modalities are available to study the cardiovascular system, most of which are less invasive than x-ray projection angiography. However, conventional angiography also has changed dramatically during the last decade. Digital angiography has replaced film/screen angiography in most applications. In addition, the use and capabilities of transluminal interventional techniques have mushroomed, and digital angiographic processor modes have expanded significantly in support of these interventional procedures. As a consequence, while less invasive technologies, such as MR angiography, make inroads into conventional angiography’s diagnostic applications, it is likely that x-ray projection angiography will remain an important clinical modality for many years to come.


Defining Terms Bolus: This term has two, independent definitions: (1) material placed in a portion of the x-ray beam to reduce the sense dynamic range and (2) the injection contrast material. Digital subtraction angiography (DSA): Methodology in which digitized angiographic images are subtracted in order to provide contrast enhancement of the opacified vasculature. Clinically, temporal subtraction is the algorithm used, though energy subtraction methods also fall under the generic term DSA. Parallel cine: The simultaneous recording of digital and cine-film images during cardiac angiography. In this mode, digital image acquisition provides diagnostic-quality images. Picture archiving and communications systems (PACS): Digital system or network for the electronic storage and retrieval of patient images and data. Progressive scanning: Video raster scan method in which all horizontal lines are read out in a single vertical scan of the video camera target. Pulsed-progressive fluoroscopy: Method of acquiring fluoroscopic images in which x-rays are produced in discrete pulses coincident with the vertical retrace period of the video camera. The video camera is then read out using progressive scanning. This compares with the older fluoroscopic method of producing x-rays continuously, coupled with interlaced video camera scanning. Temporal subtraction: Also known as time subtraction or mask-mode subtraction. A subtraction mode in which an unopacified image (the mask, usually acquired prior to injection) is subtracted from an opacified image. Upscanning: Scan conversion method in which the number of pixels or video lines displayed is higher (usually by a factor of 2) than those actually acquired from the video camera. Extra display data are produced by either replication or interpolation of the acquired data.

References Amplatz K. 1997. Rapid film changers. In S Baum (ed), Abrams’ Angiography, Chap 6. Boston, Little, Brown and Company. Aufrichtig R, Xue P, Thomas CW, et al. 1994. Perceptual comparison of pulsed and continuous fluoroscopy. Med Phys 21(2):245. Blume H. 1998. The imaging chain. In Nickoloff EL, Strauss KJ (eds): Categorical Course in Diagnostic Radiology Physics: Cardiac Catheterization Imaging, pp 83–103. Oak Brook, IL. Radiological Society of North America. Fujita H, Doi K, Lissak Giger M. 1985. Investigation of basic imaging properties in digital radiography: 6. MTFs of II-TV digital imaging systems. Med Phys 12(6):713. Holmes DR Jr, Wondrow MA, Reeder GS, et al. 1989. Optimal display of the coronary arterial tree with an upscan 1023-line video display system. Cathet Cardiovasc Diagn 18(3):175. Kruger RA, Mistretta CA, Riederer SJ. 1981. Physical and technical considerations of computerized fluoroscopy difference imaging. IEEE Trans Nucl Sci 28:205. National Electrical Manufacturers Association. 1992. Test Standard for the Determination of the System Contrast Ratio and System Veiling Glare Index of an X-ray Image Intensifier System, NEMA Standards Publication No. XR 16. Washington, National Electrical Manufacturers Association. National Electric Manufacturers Association. 1993. Digital Imaging and Communications in Medicine (DICOM), NEMA Standards Publication PS3.0(1993). Washington, National Electrical Manufacturers Association. Rowlands JA. 1992. Real-time digital processing of video image sequences for videofluoroscopy. SPIE Proc 1652:294. Sandrik JM. 1984. The video camera for medical imaging. In GD Fullerton, WR Hendee, JC Lasher, et al. (eds): Electronic Imaging in Medicine, pp 145–183. New York, American Institute of Physics.


Seibert JA, Barr DH, Borger DJ, et al. 1984. Interlaced versus progressive readout of television cameras for digital radiographic acquisitions. Med Phys 11:703. Siedband MP. 1994. Image intensification and television. In JM Taveras, JT Ferrucci (eds), Radiology: Diagnosis-Imaging-Intervention, Chap 10. Philadelphia, JB Lippincott. Verhoeven LAJ. 1985. DSA imaging: Some physical and technical aspects. Medicamundi 30:46.

Further Information Balter S, Shope TB (eds). 1995. A Categorical Course in Physics: Physical and Technical Aspects of Angiography and Interventional Radiology. Oak Brook, IL. Radiological Society of North America. Baum S (ed). 1997. Abrams’ Angiography, 4th ed. Boston. Little, Brown, and Company. Kennedy TE, Nissen SE, Simon R, Thomas JD, Tilkemeier PL. 1997. Digital Cardiac Imaging in the 21st Century: A Primer. Bethesda, MD. The Cardiac and Vascular Information Working Group (American College of Cardiology). Moore RJ. 1990. Imaging Principles of Cardiac Angiography. Rockville, MD. Aspen Publishers. Nickoloff EL, Strauss KJ (eds). 1998. Categorical Course in Diagnostic Radiology Physics: Cardiac Catheterization Imaging. Oak Brook, IL. Radiological Society of North America. Seibert JA, Barnes GT, Gould RG (eds). 1994. Medical Physics Monograph No. 20: Specification, Acceptance Testing and Quality Control of Diagnostic X-Ray Imaging Equipment. Woodbury, NY. American Institute of Physics.

61.3 Mammography Martin J. Yaffe Mammography is an x-ray imaging procedure for examination of the breast. It is used primarily for the detection and diagnosis of breast cancer, but also for pre-surgical localization of suspicious areas and in the guidance of needle biopsies. Breast cancer is a major killer of women. Approximately 179,000 women were diagnosed with breast cancer in the U.S. in 1998 and 43,500 women died of this disease [Landis, 1998]. Its cause is not currently known; however, it has been demonstrated that survival is greatly improved if disease is detected at an early stage [Tabar, 1993; Smart, 1993]. Mammography is at present the most effective means of detecting early stage breast cancer. It is used both for investigating symptomatic patients (diagnostic mammography) and for screening of asymptomatic women in selected age groups. Breast cancer is detected on the basis of four types of signs on the mammogram: 1. 2. 3. 4.

The characteristic morphology of a tumor mass. Certain presentations of mineral deposits as specks called microcalcifications. Architectural distortion of normal tissue patterns caused by the disease. Asymmetry between corresponding regions of images of the left and right breast.

Principles of Mammography The mammogram is an x-ray shadowgram formed when x-rays from a quasi-point source irradiate the breast and the transmitted x-rays are recorded by an image receptor. Because of the spreading of the x-rays from the source, structures are magnified as they are projected onto the image receptor. The signal is a result of differential attenuation of x-rays along paths passing through the structures of the breast. The essential features of image quality are summarized in Fig. 61.8. This is a one-dimensional profile of x-ray transmission through a simplified computer model of the breast [Fahrig, 1992], illustrated in Fig. 61.9. A region of reduced transmission corresponding to a structure of interest such as a tumor, a calcification, or normal fibroglandular tissue is shown. The imaging system must have sufficient spatial


FIGURE 61.8 Profile of a simple x-ray projection image, illustrating the role of contrast, spatial resolution, and noise in mammographic image quality.

FIGURE 61.9

Simplified computer model of the mammographic image acquisition process.

resolution to delineate the edges of fine structures in the breast. Structural detail as small as 50 µm must be adequately resolved. Variation in x-ray attenuation among tissue structures in the breast gives rise to contrast. The detectability of structures providing subtle contrast is impaired, however, by an overall random fluctuation in the profile, referred to as mottle or noise. Because the breast is sensitive to ionizing radiation, which at least for high doses is known to cause breast cancer, it is desirable to use the lowest radiation dose compatible with excellent image quality. The components of the imaging system will be described and their design will be related to the imaging performance factors discussed in this section.

Physics of Image Formation In the model of Fig. 61.9, an “average” breast composed of 50% adipose tissue and 50% fibroglandular tissue is considered. For the simplified case of monoenergetic x-rays of energy, E, the number of x-rays recorded in a fixed area of the image is proportional to


()

N B = N 0 E e −µT

(61.1)

in the “background” and

()

N L = N0 E e

[( ) ]

− µ T −t + µ ′t

(61.2)

in the shadow of the lesion or other structure of interest. In Eqs. (61.1) and (61.2), N0(E) is the number of x-rays that would be recorded in the absence of tissue in the beam, µ and µ´ are the attenuation coefficients of the breast tissue and the lesion, respectively, T is the thickness of the breast, and t is the thickness of the lesion. The difference in x-ray transmission gives rise to subject contrast which can be defined as:

C0 =

NB − NL NB + NL

(61.3)

For the case of monoenergetic x-rays and temporarily ignoring scattered radiation,

[ [

1− e

− µ ′ −µt

1+ e

− µ ′ −µt

] , ]

(61.4)

i.e., contrast would depend only on the thickness of the lesion and the difference between its attenuation coefficient and that of the background material. These are not valid assumptions and in actuality contrast also depends on µ and T. Shown in Fig. 61.10 are x-ray attenuation coefficients measured vs. energy on samples of three types of materials found in the breast: adipose tissue, normal fibroglandular breast tissue, and infiltrating ductal carcinoma (one type of breast tumor) [Johns, 1987]. Both the attenuation coefficients themselves and

FIGURE 61.10 Measured x-ray linear attenuation coefficients of breast fibroglandular tissue, breast fat, and infiltrating ductal carcinoma plotted vs. x-ray energy.


FIGURE 61.11

Dependence of mammographic subject contrast on x-ray energy.

their difference (µ′ – µ) decrease with increasing E. As shown in Fig. 61.11, which is based on Eq. (61.4), this causes Cs to fall as x-ray energy increases. Note that the subject contrast of even small calcifications in the breast is greater than that for a tumor because of the greater difference in attenuation coefficient between calcium and breast tissue. For a given image recording system (image receptor), a proper exposure requires a specific value of x-ray energy transmitted by the breast and incident on the receptor, i.e., a specific value of NB. The breast entrance skin exposure1 (ESE) required to produce an image is, therefore, proportional to

()

N 0 = NB E e +µT

(61.5)

Because µ decreases with energy, the required exposure for constant signal at the image receptor, NB, will increase if E is reduced to improve image contrast. A better measure of the risk of radiation-induced breast cancer than ESE is the mean glandular dose (MGD) [BEIR V, 1990]. MGD is calculated as the product of the ESE and a factor, obtained experimentally or by Monte Carlo radiation transport calculations, which converts from incident exposure to dose [Wu, 1991, 1994]. The conversion factor increases with E so that MGD does not fall as quickly with energy as does entrance exposure. The trade-off between image contrast and radiation dose necessitates important compromises in establishing mammographic operating conditions.

Equipment The mammography unit consists of an x-ray tube and an image receptor mounted on opposite sides of a mechanical assembly or gantry. Because the breast must be imaged from different aspects and to accommodate patients of different height, the assembly can be adjusted in a vertical axis and rotated about a horizontal axis as shown in Fig. 61.12. Most general radiography equipment is designed such that the image field is centered below the x-ray source. In mammography, the system’s geometry is arranged as in Fig. 61.13a where a vertical line from the x-ray source grazes the chest wall of the patient and intersects orthogonally with the edge of the image receptor closest to the patient. If the x-ray beam were centered over the breast as in Fig. 61.13b, 1Exposure is expressed in Roentgens (R) (which is not an SI unit) or in Coulombs of ionization collected per kilogram of air.


FIGURE 61.12

Schematic diagram of a dedicated mammography machine.

some of the tissue near the chest wall would be projected inside of the patient where it could not be recorded. Radiation leaving the x-ray tube passes through a metallic spectral-shaping filter, a beam-defining aperture, and a plate which compresses the breast. Those rays transmitted through the breast are incident on an anti-scatter “grid” and then strike the image receptor where they interact and deposit most of their energy locally. A fraction of the x-rays pass through the receptor without interaction and impinge upon a sensor which is used to activate the automatic exposure control mechanism of the unit. X-Ray Source Practical monoenergetic x-ray sources are not available and the x-rays used in mammography arise from bombardment of a metal target by electrons in a hot-cathode vacuum tube. The x-rays are emitted from the target over a spectrum of energies, ranging up to the peak kilovoltage applied to the x-ray tube. Typically, the x-ray tube employs a rotating anode design in which electrons from the cathode strike the anode target material at a small angle (0 to 16°) from normal incidence (Fig. 61.14). Over 99% of the energy from the electrons is dissipated as heat in the anode. The angled surface and the distribution of the electron bombardment along the circumference of the rotating anode disk allows the energy to be spread over a larger area of target material while presenting a much smaller effective focal spot as viewed from the imaging plane. On modern equipment, the typical “nominal” focal spot size for normal contact mammography is 0.3 mm, while the smaller spot used primarily for magnification is 0.1 mm. The specifications for x-ray focal spot size tolerance, established by NEMA (National Electrical Manufacturers Association) or the IEC (International Electrotechnical Commission) allow the effective focal spot size to be considerably larger than these nominal sizes. For example, the NEMA specification allows the effective focal spot size to be 0.45 mm in width and 0.65 mm in length for a nominal 0.3 mm spot and 0.15 mm in each dimension for a nominal 0.1 mm spot. The nominal focal spot size is defined relative to the effective spot size at a “reference axis”. As shown in Fig. 61.14, this reference axis, which may vary from manufacturer to manufacturer, is normally specified at some mid-point in the image. The effective size of the focal spot will monotonically increase © 2000 by CRC Press LLC

FIGURE 61.13 Geometric arrangement of system components in mammography. (a) Correct alignment provides good tissue coverage, (b) incorrect alignment causes tissue near the chest wall not to be imaged.

from the anode side to the cathode side of the imaging field. Normally, x-ray tubes are arranged such that the cathode side of the tube is adjacent to the patient’s chest wall, since the highest intensity of x-rays is available at the cathode side, and the attenuation of x-rays by the patient is generally greater near the chest wall of the image. The spatial resolution capability of the imaging system is partially determined by the effective size of the focal spot and by the degree of magnification of the anatomy at any plane in the breast. This is illustrated in Fig. 61.15 where, by similar triangles, the unsharpness region due to the finite size of the focal spot is linearly related to the effective size of the spot and to the ratio of OID to SOD, where SOD is the sourceobject distance and OID is the object-image receptor distance. Because the breast is a three-dimensional structure, this ratio and, therefore, the unsharpness will vary for different planes within the breast. The size of the focal spot determines the heat loading capability of the x-ray tube target. For smaller focal spots, the current through the x-ray tube must be reduced, necessitating increased exposure times and the possibility of loss of resolution due to motion of anatomical structures. Loss of geometric resolution can be controlled in part by minimizing OID/SOD, i.e., by designing the equipment with greater source-breast distances, by minimizing space between the breast and the image receptor, and by compressing the breast to reduce its overall thickness. Magnification is often used intentionally to improve the signal-to-noise ratio of the image. This is accomplished by elevating the breast above the image receptor, in effect reducing SOD and increasing OID. Under these conditions, resolution is invariably limited by focal spot size and use of a small spot for magnification imaging (typically a nominal size of 0.1 mm) is critical. © 2000 by CRC Press LLC

FIGURE 61.14 Angled-target x-ray source provides improved heat loading but causes effective focal spot size to vary across the image.

Since monoenergetic x-rays are not available, one attempts to define a spectrum providing energies which give a reasonable compromise between radiation dose and image contrast. The spectral shape can be controlled by adjustment of the kilovoltage, choice of the target material, and the type and thickness of metallic filter placed between the x-ray tube and the breast. Based on models of the imaging problem in mammography, it has been suggested that the optimum energy for imaging lies between 18 and 23 keV, depending on the thickness and composition of the breast [Beaman, 1982]. It has been found that for the breast of typical thickness and composition, the characteristic x-rays from molybdenum at 17.4 and 19.6 keV provide good imaging performance. For this reason, molybdenum target x-ray tubes are used on the vast majority of mammography machines. Most mammography tubes use beryllium exit windows between the evacuated tube and the outside world since glass or other metals used in general purpose tubes would provide excessive attenuation of the useful energies for mammography. Figure 61.16 compares tungsten target and molybdenum target spectra for beryllium window x-ray tubes. Under some conditions, tungsten may provide appropriate image quality for mammography; however, it is essential that the intense emission of L radiation from tungsten be filtered from the beam before it is incident upon the breast, since extremely high doses to the skin would result from this radiation without useful contribution to the mammogram. Filtration of the X-Ray Beam In conventional radiology, filters made of aluminum or copper are used to provide selective removal of low x-ray energies from the beam before it is incident upon the patient. In mammography, particularly when a molybdenum anode x-ray tube is employed, a molybdenum filter 20 to 35 µm thick is generally used. This filter attenuates x-rays both at low energies and those above its own K-absorption edge allowing the molybdenum characteristic x-rays from the target to pass through the filter with relatively high © 2000 by CRC Press LLC

FIGURE 61.15

Dependence of focal spot unsharpness on focal spot size and magnification factor.

efficiency. As illustrated in Fig. 61.17, this K edge filtration results in a spectrum enriched with x-ray energies in the range of 17 to 20 keV. Although this spectrum is relatively well suited for imaging the breast of average attenuation, slightly higher energies are desirable for imaging dense thicker breasts. Because the molybdenum target spectrum is so heavily influenced by the characteristic x-rays, an increase in the kilovoltage alone does not substantially change the shape of the spectrum. The beam can be “hardened”, however, by employing filters of higher atomic number than molybdenum. For example, rhodium (atomic no. 45) has a K absorption edge at 23 keV, providing strong attenuation both for x-rays above this energy and for those at substantially lower energies. Used with a molybdenum target x-ray tube and slightly increased kV, it provides a spectrum with increased penetration (reduced dose) compared to the Mo/Mo combination. It is possible to go further in optimizing imaging performance, by “tuning” the effective spectral energy by using other target materials in combination with appropriate K-edge filters [Jennings, 1993]. One manufacturer employs an x-ray tube incorporating both molybdenum and rhodium targets, where the electron beam can be directed toward one or the other of these materials [Heidsieck, 1991]. With this system, the filter material (rhodium, molybdenum, etc.) can be varied to suit the target that has been selected. Similarly, work has been reported on K-edge filtration of tungsten spectra [Desponds, 1991], where the lack of pronounced K characteristic peaks provides more flexibility in spectral shaping with filters. Compression Device There are several reasons for applying firm (but not necessarily painful) compression to the breast during the examination. Compression causes the different tissues to be spread out, minimizing superposition from different planes and thereby improving a conspicuity of structures. As will be discussed later, scattered radiation can degrade contrast in the mammogram. The use of compression decreases the ratio © 2000 by CRC Press LLC

FIGURE 61.16


Comparison of tungsten and molybdenum target x-ray spectra.

Molybdenum target spectrum filtered by 0.03 mm Mo foil.

of scattered to directly transmitted radiation reaching the image receptor. Compression also decreases the distance from any plane within the breast to the image receptor (i.e., OID) and in this way reduces geometric unsharpness. The compressed breast provides lower overall attenuation to the incident x-ray beam, allowing the radiation dose to be reduced. The compressed breast also provides more uniform attenuation over the image. This reduces the exposure range which must be recorded by the imaging system, allowing more flexibility in choice of films to be used. Finally, compression provides a clamping action which reduces anatomical motion during the exposure reducing this source of image unsharpness. It is important that the compression plate allows the breast to be compressed parallel to the image receptor, and that the edge of the plate at the chest wall be straight and aligned with both the focal spot and image receptor to maximize the amount of breast tissue which is included in the image (see Fig. 61.13). Anti-Scatter Grid Lower x-ray energies are used for mammography than for other radiological examinations. At these energies, the probability of photoelectric interactions within the breast is significant. Nevertheless, the probability of Compton scattering of x-rays within the breast is still quite high. Scattered radiation recorded by the image receptor has the effect of creating a quasi-uniform haze on the image and causes the subject contrast to be reduced to

CS =

C0 1 + SPR

(61.6)

where C0 is the contrast in the absence of scattered radiation, given by Eq. 61.4 and SPR is the scatterto-primary (directly transmitted) x-ray ratio at the location of interest in the image. In the absence of an anti-scatter device, 37 to 50% of the total radiation incident on the image receptor would have experienced a scattering interaction within the breast, i.e., the scatter-to-primary ratio would be 0.6:1.0. In addition to contrast reduction, the recording of scattered radiation uses up part of the dynamic range of the image receptor and adds statistical noise to the image. Anti-scatter grids have been designed for mammography. These are composed of linear lead (Pb) septa separated by a rigid interspace material. Generally, the grid septa are not strictly parallel but focused (toward the x-ray source). Because the primary x-rays all travel along direct lines from the x-ray source to the image receptor, while the scatter diverges from points within the breast, the grid presents a smaller acceptance aperture to scattered radiation than to primary radiation and thereby discriminates against scattered radiation. Grids are characterized by their grid ratio (ratio of the path length through the interspace material to the interseptal width) which typically ranges from 3.5:1 to 5:1. When a grid is used, the SPR is reduced typically by a factor of about 5, leading in most cases to a substantial improvement in image contrast [Wagner, 1991]. On modern mammography equipment, the grid is an integral part of the system, and during x-ray exposure is moved to blur the image of the grid septa to avoid a distracting pattern in the mammogram. It is important that this motion be uniform and of sufficient amplitude to avoid non-uniformities in the image, particularly for short exposures that occur when the breast is relatively lucent. Because of absorption of primary radiation by the septa and by the interspace material, part of the primary radiation transmitted by the patient does not arrive at the image receptor. In addition, by removing some of the scattered radiation, the grid causes the overall radiation fluence to be reduced from that which would be obtained in its absence. To obtain a radiograph of proper optical density, the entrance exposure to the patient must be increased by a factor known as the Bucky factor to compensate for these losses. Typical Bucky factors are in the range of 2 to 3. A linear grid does not provide scatter rejection for those quanta traveling in planes parallel to the septa. Recently a crossed grid that consists of septa that run in orthogonal directions has been introduced for this purpose. The improved scatter rejection is accomplished at doses comparable to those required


FIGURE 61.18

Design of a screen-film image receptor for mammography.

with a linear grid, because the interspace material of the crossed grid is air rather than solid. To avoid artifacts, the grid is moved in a very precise way during the exposure to ensure a uniform blurring of the image of the grid itself. Image Receptor Fluorescent Screens When first introduced, mammography was carried out using direct exposure radiographic film in order to obtain the high spatial resolution required. Since the mid-1970s, high resolution fluorescent screens have been used to convert the x-ray pattern from the breast into an optical image. These screens are used in conjunction with single-coated radiographic film, and the configuration is shown in Fig. 61.18. With this arrangement, the x-rays pass through the cover of a light-tight cassette and the film to impinge upon the screen. Absorption is exponential, so that a large fraction of the x-rays are absorbed near the entrance surface of the screen. The phosphor crystals which absorb the energy produce light in an isotropic distribution. Because the film emulsion is pressed tightly against the entrance surface of the screen, the majority of the light quanta have only a short distance to travel to reach the film. Light quanta traveling longer distances have an opportunity to spread laterally (see Fig. 61.18), and in this way degrade the spatial resolution. To discriminate against light quanta which travel along these longer oblique paths, the phosphor material of the screen is generally treated with a dye which absorbs much of this light, giving rise to a sharper image. A typical phosphor used for mammography is gadolinium oxysulphide (Gd2O2S). Although the K-absorption edge of gadolinium occurs at too high an energy to be useful in mammography, the phosphor material is dense (7.44 g/cm3) so that the quantum efficiency (the fraction of incident x-rays which interact with the screen), is good (about 60%). Also, the conversion efficiency of this phosphor (fraction of the absorbed x-ray energy converted to light) is relatively high. The light emitted from the fluorescent screen is essentially linearly dependent upon the total amount of energy deposited by x-rays within the screen.


FIGURE 61.19

Characteristic curve of a mammographic screen-film image receptor.

Film The photographic film emulsion for mammography is designed with a characteristic curve such as that shown in Fig. 61.19, which is a plot of the optical density (blackness) provided by the processed film vs. the logarithm of the x-ray exposure to the screen. Film provides non-linear input-output transfer characteristics. The local gradient of this curve controls the display contrast presented to the radiologist. Where the curve is of shallow gradient, a given increment of radiation exposure provides little change in optical density, rendering structures imaged in this part of the curve difficult to visualize. Where the curve is steep, the film provides excellent image contrast. The range of exposures over which contrast is appreciable is referred to as the latitude of the film. Because the film is constrained between two optical density values—the base + fog density of the film, where no intentional x-ray exposure has resulted, and the maximum density provided by the emulsion—there is a compromise between maximum gradient of the film and the latitude that it provides. For this reason, some regions of the mammogram will generally be underexposed or overexposed, i.e., rendered with sub-optimal contrast. Film Processing Mammography film is processed in an automatic processor similar to that used for general radiographic films. It is important that the development temperature, time, and rate of replenishment of the developer chemistry be compatible with the type of film emulsion used and be designed to maintain good contrast of the film. Noise and Dose Noise in mammography results primarily from two sources—the random absorption of x-rays in the detector and the granularity associated with the screen and the film emulsion. The first, commonly known as quantum noise, is governed by Poisson statistics so that for a given area of image the standard deviation in the number of x-rays recorded is equal to the square root of the mean number recorded. In other words, the noise in the image is dependent on both the amount of radiation which strikes the imaging


system per unit area, and the quantum efficiency of the imaging system. The quantum efficiency is related to the attenuation coefficient of the phosphor material and the thickness of the screen. In order to maintain high spatial resolution, the screen must be made relatively thin to avoid lateral diffusion of light. The desirability of maintaining a relatively low quantum noise level in the image mandates that the conversion efficiency of the screen material and the sensitivity of the film not be excessively high. With very high conversion efficiency, the image could be produced at low dose but with an inadequate number of quanta contributing to the image. Similarly, film granularity increases as more sensitive films are used, so that again film speed must be limited to maintain high image quality. For current high quality mammographic imaging employing an anti-scatter grid, with films exposed to a mean optical density of at least 1.6, the mean glandular dose to a 5-cm thick compressed breast consisting of 50% fibroglandular and 50% adipose tissue is in the range of 1 to 2 milligray [Conway, 1992]. Automatic Exposure Control It is difficult for the technologist to estimate the attenuation of the breast by inspection and, therefore, modern mammography units are equipped with automatic exposure control (AEC). The AEC radiation sensors are located behind the image receptor so that they do not cast a shadow on the image. The sensors measure the x-ray fluence transmitted through both the breast and the receptor and provide a signal which can be used to discontinue the exposure when a certain preset amount of radiation has been received by the image receptor. The location of the sensor must be adjustable so that it can be placed behind the appropriate region of the breast in order to obtain proper image density. AEC devices must be calibrated so that constant image optical density results are independent of variations in breast attenuation, kilovoltage setting, or field size. With modern equipment, automatic exposure control is generally microprocessor-based so that relatively sophisticated corrections can be made during the exposure for the above effects and for reciprocity law failure of the film. Automatic Kilovoltage Control Many modern mammography units also incorporate automatic control of the kilovoltage or target/filter/kilovoltage combination. Penetration through the breast depends on both breast thickness and composition. For a breast that is dense, it is possible that a very long exposure time would be required to achieve adequate film blackening. This results in high dose to the breast and possibly blur due to anatomical motion. It is possible to sense the compressed breast thickness and the transmitted exposure rate and to employ an algorithm to automatically choose the x-ray target and/or beam filter as well as the kilovoltage.

Quality Control Mammography is one of the most technically demanding radiographic procedures, and in order to obtain optimal results, all components of the system must be operating properly. Recognizing this, the American College of Radiology implemented and administers a Mammography Accreditation Program [McClelland, 1991], which evaluates both technical and personnel-related factors in facilities applying for accreditation. In order to verify proper operation, a rigorous quality control program should be in effect. In fact, the U.S. Mammography Quality Standards Act stipulates that a quality control program must be in place in all facilities performing mammography. A program of tests (summarized in Table 61.1) and methods for performing them are contained in the quality control manuals for mammography published by the American College of Radiology [Hendrick, 1999].

Stereotactic Biopsy Devices Stereoscopic x-ray imaging techniques are currently used for the guidance of needle “core” biopsies. These procedures can be used to investigate suspicious mammographic or clinical findings without the need for surgical excisional biopsies, resulting in reduced patient risk, discomfort, and cost. In these stereotactic biopsies, the gantry of a mammography machine is modified to allow angulated views of the breast


TABLE 61.1

Mammographic Quality Control Minimum Test Frequencies

Test Darkroom cleanliness Processor quality control Screen cleanliness Viewboxes and viewing conditions Phantom images Visual check list Repeat analysis Analysis of fixer retention in film Darkroom fog Screen-film contact Compression Mammographic unit assembly evaluation Collimation assessment Focal spot size performance kVp Accuracy/reproducibility Beam quality assessment (half-value-layer) Automatic exposure control (AEC) system performance assessment Uniformity of screen speed Breast entrance exposure and mean glandular dose Image quality—Phantom evaluation Artifact assessment Radiation output rate Viewbox luminance and room illuminance Compression release mechanism

Performed By:

Minimum Frequency

Radiologic technologist

Daily Daily Weekly Weekly Weekly Monthly Quarterly Quarterly Semi-annually Semi-annually Semi-annually Annually Annually Annually Annually Annually Annually Annually Annually Annually Annually Annually Annually Annually

Medical physicist

Source: Hendrick, R.E. et al., 1999. Mammography Quality Control Manuals (radiologist, radiologic technologist, medical physicist). American College of Radiology. Reston, VA. With permission.

(typically ±15° from normal incidence) to be achieved. From measurements obtained from these images, the three-dimensional location of a suspicious lesion is determined and a needle equipped with a springloaded cutting device can be accurately placed in the breast to obtain tissue samples. While this procedure can be performed on an upright mammography unit, special dedicated systems have recently been introduced to allow its performance with the patient lying prone on a table. The accuracy of sampling the appropriate tissue depends critically on the alignment of the system components and the quality of the images produced. A thorough review of stereotactic imaging, including recommended quality control procedures is given by Hendrick and Parker [Hendrick, 1994].

Digital Mammography There are several technical factors associated with screen-film mammography which limit the ability to display the finest or most subtle details, and produce images with the most efficient use of radiation dose to the patient. In screen-film mammography, the film must act as an image acquisition detector as well as a storage and display device. Because of its sigmoidal shape, the range of x-ray exposures over which the film display gradient is significant, i.e., the image latitude, is limited. If a tumor is located in either a relatively lucent or more opaque region of the breast, then the contrast displayed to the radiologist may be inadequate because of the limited gradient of the film. This is particularly a concern in patients whose breasts contain large amounts of fibroglandular tissue, the so-called dense breast. Another limitation of film mammography is the effect of structural noise due to the granularity of the film emulsion used to record the image. This impairs the detectibility of microcalcifications and other fine structures within the breast. While Poisson quantum noise is unavoidable, it should be possible to virtually eliminate structural noise by technical improvements. Existing screen-film mammography also suffers because of the inefficiency of grids in removing the effects of scattered radiation and with compromises in spatial resolution vs. quantum efficiency inherent in the screen-film image receptor. © 2000 by CRC Press LLC

FIGURE 61.20

Schematic representation of a digital mammography system.

Many of the limitations of conventional mammography can be effectively overcome with a digital mammography imaging system (Fig. 61.20), in which image acquisition, display, and storage are performed independently, allowing optimization of each. For example, acquisition can be performed with low noise, highly linear x-ray detectors, while since the image is stored digitally, it can be displayed with contrast independent of the detector properties and defined by the needs of the radiologist. Whatever image processing techniques are found useful, ranging from simple contrast enhancement to histogram modification and spatial frequency filtering, could conveniently be applied. The challenges in creating a digital mammography system with improved performance are mainly related to the x-ray detector and the display device. There is active development of high resolution display monitors and hard copy devices to meet the demanding requirements (number of pixels, luminance, speed, multi-image capacity) of displaying digital mammography images, and suitable systems for this purpose should be available in the near future. The detector should have the following characteristics: 1. 2. 3. 4. 5. 6.

Efficient absorption of the incident radiation beam. Linear response over a wide range of incident radiation intensity. Low intrinsic noise. Spatial resolution on the order of 10 cycles/mm (50 µm sampling). Can accommodate at least an 18 × 24 cm and preferably a 24 × 30 cm field size. Acceptable imaging time and heat loading of the x-ray tube.

Two main approaches have been taken in detector development—area detectors and slot detectors. In the former, the entire image is acquired simultaneously, while in the latter only a portion of the image is acquired at one time and the full image is obtained by scanning the x-ray beam and detector across the breast. Area detectors offer convenient fast image acquisition and could be used with conventional x-ray machines, but may still require a grid, while slot systems are slower and require a scanning x-ray beam, but use relatively simple detectors and have excellent intrinsic efficiency at scatter rejection.


FIGURE 61.21 (a) Small-format detector system for biopsy imaging, (b) full-breast detector incorporating 12 detector modules, (c) slot detector for a full-breast scanning digital mammography system.

At the time of writing, small-format (5 × 5 cm) digital systems for guidance of stereotactic breast biopsy procedures are in widespread use. These use a lens or a fiberoptic taper to couple a phosphor to a CCD whose format is approximately square and typically provides 1 × 1K images with 50 µm pixels (Fig. 61.21a). Adjustment of display contrast enhances the localization of the lesion while the immediate display of images (no film processing is required) greatly accelerates the clinical procedure. Four designs of full breast digital mammography systems are undergoing clinical evaluation. Various detector technologies are being developed and evaluated for use in digital mammography. In three of the systems, x-rays are absorbed by a cesium iodide (CsI) phosphor layer and produce light. In one system, the phosphor is deposited directly on a matrix of about 20002 photodiodes with thin film transistor switches fabricated on a large area amorphous silicon plate. The electronic signal is read out on a series of data lines as the switches in each row of the array are activated. In another system, the light is coupled through demagnifying fiberoptic tapers to a CCD readout. A mosaic of 3 × 4 detector modules is formed (Fig. 61.21b) to obtain a detector large enough to cover the breast [Cheung, 1998]. A third system also uses fiberoptic coupling of CsI to CCDs; however, the detector is in a slot configuration and is scanned beneath the breast in synchrony with a fan beam of x-rays to acquire the transmitted signal (Fig. 61.21c). The fourth system employs a plate formed of a photostimulable phosphor material. When exposed to x-rays, traps in the phosphor are filled with electrons, the number being related to x-ray intensity. The plate is placed in a reader device and scanned with a red HeNe laser beam which stimulates the traps to release the electrons. The transition of these electrons through energy levels in the phosphor crystal result in the formation of blue light, which is measured as a function of the laser position on the plate to form the image signal. Other materials in which x-ray energy is directly converted to charge are under development for digital mammography. These materials include lead iodide, amorphous selenium, zinc cadmium telluride, and thallium bromide. A review of the current status of digital mammography is given in [Yaffe, 1994; Pisano, 1998] and of detectors for digital x-ray imaging in [Yaffe, 1997]. © 2000 by CRC Press LLC

Summary Mammography is a technically demanding imaging procedure which can help reduce mortality from breast cancer. To be successful at this purpose, both the technology and technique used for imaging must be optimized. This requires careful system design and attention to quality control procedures. Imaging systems for mammography are still evolving and, in the future, are likely to make greater use of digital acquisition and display methods.

Defining Terms Conversion efficiency: The efficiency of converting the energy from x-rays absorbed in a phosphor material into that of emitted light quanta. Fibroglandular tissue: A mixture of tissues within the breast composed of the functional glandular tissue and the fibrous supporting structures. Focal spot: The area of the anode of an x-ray tube from which the useful beam of x-rays is emitted. Also known as the target. Grid: A device consisting of evenly spaced lead strips which functions like a venetian blind in preferentially allowing x-rays traveling directly from the focal spot without interaction in the patient to pass through, while those whose direction has been diverted by scattering in the patient strike the slats of the grid and are rejected. Grids improve the contrast of radiographic images at the price of increased dose to the patient. Image receptor: A device that records the distribution of x-rays to form an image. In mammography, the image receptor is generally composed of a light-tight cassette containing a fluorescent screen, which absorbs x-rays and produces light, coupled to a sheet of photographic film. Quantum efficiency: The fraction of incident x-rays which interact with a detector or image receptor. Screening: Examination of asymptomatic individuals to detect disease. Survival: An epidemiological term giving the fraction of individuals diagnosed with a given disease alive at a specified time after diagnosis, e.g., “10-year survival”.

References Beaman, S.A. and Lillicrap, S.C. 1982 Optimum x-ray spectra for mammography. Phys. Med. Biol. 27, 1209-1220. Cheung L., Bird R., Ashish C., Rego A., Rodriguez C., Yuen J. 1998 Initial operating and clinical results of a full-field mammography system. Health Effects of Exposure to Low Levels of Ionizing Radiation (BEIR V) 1990 National Academy Press, Washington, D.C. 163-170. Landis SH, Murray T, Bolden S, Wingo P 1998 Cancer Statistics, 1998 CA Cancer J Clin 48, 6-29. Conway, B.J., Suleiman, O.H., Rueter, F.G., Antonsen, R.G., Slayton, R.J. and McCrohan, J.L. 1992. Does credentialing make a difference in mammography? Radiology 185(P): 250. Desponds, L., Depeursinge, C., Grecescu, M., Hessler, C., Samiri, A. and Valley, J.F. 1991. Image of anode and filter material on image quality and glandular dose for screen-film mammography. Phys. Med. Biol. 36:1165-1182. Fahrig, R., Maidment, A.D.A., Yaffe, M.J. 1992. Optimization of peak kilovoltage and spectral shape for digital mammography. Proc SPIE 1651:74-83. Heidsieck, R., Laurencin, G., Ponchin, A., Gabbay, E. and Klausz, R. 1991. Dual target x-ray tubes for mammographic examinations: Dose reduction with image quality equivalent to that with standard mammographic tubes. Radiology, 181 (P):311. Hendrick, R.E. et al., 1994. Mammography Quality Control Manuals (radiologist, radiologic technologist, medical physicist). American College of Radiology. Reston, VA. Hendrick, R.E. and Parker, S.H. 1994. Stereotaxic Imaging. In A Categorical Course in Physics: Technical Aspects of Breast Imaging. 3rd ed. A.G. Haus and M.J.Yaffe, Eds, p 263-274. RSNA Publications, Oak Brook, IL. © 2000 by CRC Press LLC

Jennings, R.J., Quinn, P.W., Gagne, R.M. and Fewell, T.R. 1993. Evaluation of x-ray sources for mammography. Proc SPIE 1896:259-268. Johns, P.C., Yaffe, M.J. 1987. X-ray characterization of normal and neoplastic breast tissues. Phys. Med. Biol. 32:675-695. Karellas, A., Harris, L.J. and D’Orsi, C.J., 1990. Small field digital mammography with a 2048x2048 pixel charge-coupled device. Radiology. 177:288. McClelland, R., Hendrick, R.E., Zinninger, M.D. and Wilcox, P.W. 1991. The American College of Radiology Mammographic Accreditation Program. Am. J. Roentgenol. 157:473-479. Nishikawa, R.M. and Yaffe, M.J. 1985. Signal-to-noise properties of mammography film-screen systems. Medical Physics 12:32-39. Pisano ED, Yaffe MJ, 1998 Digital Mammography. Contemporary Diagnostic Radiology 21:1-6. Smart, C.R., Hartmann, W.H., Beahrs, O.H. et al. 1993. Insights into breast cancer screening of younger women: Evidence from the 14-year follow-up of the Breast Cancer Detection Demonstration Project. Cancer. 72:1449-1456. Tabar L., Duffy S.W., Burhenne L.W. 1993. New Swedish breast cancer detection results for women aged 40-49. Cancer (suppl). 72:1437-1448. Wagner A.J. 1991. Contrast and grid performance in mammography. In Screen Film Mammography: Imaging Considerations and Medical Physics Responsibilities. ed. G.T Barnes and G.D. Frey, p 115-134. Medical Physics Publishing, Madison, WI. Wu X, Barnes GT, Tucker DM. 1991 Spectral dependence of glandular tissue dose in screen-film mammography, Radiology 179:143-148. Wu X, Gingold EL, Barnes GT, Tucker DM. 1994 Normalized average glandular dose in molybdenum target-rhodium filter and rhodium target-rhodium filter mammography, Radiology 193:83-89. Yaffe M.J. 1994 Digital Mammography. In A Categorical Course in Physics: Technical Aspects of Breast Imaging, 3rd ed. A.G. Haus and M.J.Yaffe, Eds., p 275-286. RSNA Publications, Oak Brook, IL. Yaffe MJ, Rowlands JA. 1997 X-ray detectors for digital radiography. Phys Med Biol 42 1-39.

Further Information Yaffe, M.J. et al., 1993. Recommended Specifications for New Mammography Equipment: ACR-CDC Cooperative Agreement for Quality Assurance Activities in Mammography, ACR Publications, Reston, VA. Haus, A.G. and Yaffe, M.J. 1994. A Categorical Course in Physics: Technical Aspects of Breast Imaging. RSNA Publications, Oak Brook, IL. In this syllabus to a course presented at the Radiological Society of North America all technical aspects of mammography are addressed by experts and a clinical overview is presented in language understandable by the physicist or biomedical engineer. Screen Film Mammography: Imaging Considerations and Medical Physics Responsibilities. Eds., G.T Barnes and G.D. Frey. Medical Physics Publishing, Madison, WI. Considerable practical information related to obtaining and maintaining high quality mammography is provided here. Film Processing in Medical Imaging. A.G. Haus Ed., Medical Physics Publishing. Madison, WI. This book deals with all aspects of medical film processing with particular emphasis on mammography.


Cunningham, I. A., Judy, P. F. “ Computed Tomography.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

62 Computed Tomography Ian A. Cunningham Victoria Hospital, the John P. Robarts Research Institute, and the University of Western Ontario

Philip F. Judy

62.1

Instrumentation Data-Acquisition Geometries • X-Ray System • Patient Dose Considerations • Summary

62.2

Brigham and Women’s Hospital and Harvard Medical School

Reconstruction Principles Image Processing: Artifact and Reconstruction Error • Projection Data to Image: Calibrations • Projection Data to Image: Reconstruction

62.1 Instrumentation Ian A. Cunningham The development of computed tomography (CT) in the early 1970s revolutionized medical radiology. For the first time, physicians were able to obtain high-quality tomographic (cross-sectional) images of internal structures of the body. Over the next 10 years, 18 manufacturers competed for the exploding world CT market. Technical sophistication increased dramatically, and even today, CT continues to mature, with new capabilities being researched and developed. Computed tomographic images are reconstructed from a large number of measurements of x-ray transmission through the patient (called projection data). The resulting images are tomographic “maps” of the x-ray linear attenuation coefficient. The mathematical methods used to reconstruct CT images from projection data are discussed in the next section. In this section, the hardware and instrumentation in a modern scanner are described. The first practical CT instrument was developed in 1971 by DR. G. N. Hounsfield in England and was used to image the brain [Hounsfield, 1980]. The projection data were acquired in approximately 5 minutes, and the tomographic image was reconstructed in approximately 20 minutes. Since then, CT technology has developed dramatically, and CT has become a standard imaging procedure for virtually all parts of the body in thousands of facilities throughout the world. Projection data are typically acquired in approximately 1 second, and the image is reconstructed in 3 to 5 seconds. One special-purpose scanner described below acquires the projection data for one tomographic image in 50 ms. A typical modern CT scanner is shown in Fig. 62.1, and typical CT images are shown in Fig. 62.2. The fundamental task of CT systems is to make an extremely large number (approximately 500,000) of highly accurate measurements of x-ray transmission through the patient in a precisely controlled geometry. A basic system generally consists of a gantry, a patient table, a control console, and a computer. The gantry contains the x-ray source, x-ray detectors, and the data-acquisition system (DAS).

Data-Acquisition Geometries Projection data may be acquired in one of several possible geometries described below, based on the scanning configuration, scanning motions, and detector arrangement. The evolution of these geometries

0-8493-????-?/98/$0.00+$.50 © 2000 1998 by CRC Press LLC

FIGURE 62.1 Schematic drawing of a typical CT scanner installation, consisting of (1) control console, (2) gantry stand, (3) patient table, (4) head holder, and (5) laser imager. (Courtesy of Picker International, Inc.)

is descried in terms of “generations,” as illustrated in Fig. 62.3, and reflects the historical development [Newton and Potts, 1981; Seeram, 1994]. Current CT scanners use either third-, fourth-, or fifthgeneration geometries, each having their own pros and cons. First Generation: Parallel-Beam Geometry Parallel-beam geometry is the simplest technically and the easiest with which to understand the important CT principles. Multiple measurements of x-ray transmission are obtained using a single highly collimated x-ray pencil beam and detector. The beam is translated in a linear motion across the patient to obtain a projection profile. The source and detector are then rotated about the patient isocenter by approximately 1 degree, and another projection profile is obtained. This translate-rotate scanning motion is repeated until the source and detector have been rotated by 180 degrees. The highly collimated beam provides excellent rejection of radiation scattered in the patient; however, the complex scanning motion results in long (approximately 5-minute) scan times. This geometry was used by Hounsfield in his original experiments [Hounsfield, 1980] but is not used in modern scanners. Second Generation: Fan Beam, Multiple Detectors Scan times were reduced to approximately 30 s with the use of a fan beam of x-rays and a linear detector array. A translate-rotate scanning motion was still employed; however, a larger rotate increment could be used, which resulted in shorter scan times. The reconstruction algorithms are slightly more complicated than those for first-generation algorithms because they must handle fan-beam projection data. Third Generation: Fan Beam, Rotating Detectors Third-generation scanners were introduced in 1976. A fan beam of x-rays is rotated 360 degrees around the isocenter. No translation motion is used; however, the fan beam must be wide enough to completely contain the patient. A curved detector array consisting of several hundred independent detectors is mechanically coupled to the x-ray source, and both rotate together. As a result, these rotate-only motions acquire projection data for a single image in as little as 1 s. Third-generation designs have the advantage that thin tungsten septa can be placed between each detector in the array and focused on the x-ray source to reject scattered radiation.


FIGURE 62.2

Typical CT images of (a) brain, (b) head showing orbits, (c) chest showing lungs, and (d) abdomen.

Fourth Generation: Fan Beam, Fixed Detectors In a fourth-generation scanner, the x-ray source and fan beam rotate about the isocenter, while the detector array remains stationary. The detector array consists of 600 to 4800 (depending on the manufacturer) independent detectors in a circle that completely surrounds the patient. Scan times are similar to those of third-generation scanners. The detectors are no longer coupled to the x-ray source and hence cannot make use of focused septa to reject scattered radiation. However, detectors are calibrated twice during each rotation of the x-ray source, providing a self-calibrating system. Third-generation systems are calibrated only once every few hours. Two detector geometries are currently used for fourth-generation systems: (1) a rotating x-ray source inside a fixed detector array and (2) a rotating x-ray source outside a nutating detector array. Figure 62.4 shows the major components in the gantry of a typical fourth-generation system using a fixed-detector array. Both third- and fourth-generation systems are commercially available, and both have been highly successful clinically. Neither can be considered an overall superior design. Fifth Generation: Scanning Electron Beam Fifth-generation scanners are unique in that the x-ray source becomes an integral part of the system design. The detector array remains stationary, while a high-energy electron beams is electronically swept along a semicircular tungsten strip anode, as illustrated in Fig. 62.5. X-rays are produced at the point


FIGURE 62.3

Four generations of CT scanners illustrating the parallel- and fan-beam geometries [Robb, 1982].

where the electron beam hits the anode, resulting in a source of x-rays that rotates about the patient with no moving parts [Boyd et al., 1979]. Projection data can be acquired in approximately 50 ms, which is fast enough to image the beating heart without significant motion artifacts [Boyd and Lipton, 1983]. An alternative fifth-generation design, called the dynamic spatial reconstructor (DSR) scanner, is in use at the Mayo Clinic [Ritman, 1980, 1990]. This machine is a research prototype and is not available commercially. It consists of 14 x-ray tubes, scintillation screens, and video cameras. Volume CT images can be produced in as little as 10 ms. Spiral/Helical Scanning The requirement for faster scan times, and in particular for fast multiple scans for three-dimensional imaging, has resulted in the development of spiral (helical) scanning systems [Kalendar et al., 1990]. Both third- and fourth-generation systems achieve this using self-lubricating slip-ring technology (Fig. 62.6) to make the electrical connections with rotating components. This removes the need for power and signal cables which would otherwise have to be rewound between scans and allows for a continuous rotating motion of the x-ray fan beam. Multiple images are acquired while the patient is translated through the gantry in a smooth continuous motion rather than stopping for each image. Projection data for multiple images covering a volume of the patient can be acquired in a single breath hold at rates of approximately one slice per second. The reconstruction algorithms are more sophisticated because they must accommodate the spiral or helical path traced by the x-ray source around the patient, as illustrated in Fig. 62.7.

X-Ray System The x-ray system consists of the x-ray source, detectors, and a data-acquisition system. X-Ray Source With the exception of one fifth-generation system described above, all CT scanners use bremsstrahlung x-ray tubes as the source of radiation. These tubes are typical of those used in diagnostic imaging and


FIGURE 62.4 The major internal components of a fourth-generation CT gantry are shown in a photograph with the gantry cover removed (upper) and identified in the line drawing (lower), (Courtesy of Picker International, Inc.)

produce x-rays by accelerating a beam of electrons onto a target anode. The anode area from which x-rays are emitted, projected along the direction of the beam, is called the focal spot. Most systems have two possible focal spot sizes, approximately 0.5 × 1.5 mm and 1.0 × 2.5 mm. A collimator assembly is used to control the width of the fan beam between 1.0 and 10 mm, which in turn controls the width of the imaged slice. The power requirements of these tubes are typically 120 kV at 200 to 500 mA, producing x-rays with an energy spectrum ranging between approximately 30 and 120 keV. All modern systems use highfrequency generators, typically operating between 5 and 50 kHz [Brunnett et al., 1990]. Some spiral systems use a stationary generator in the gantry, requiring high-voltage (120-kV) slip rings, while others use a rotating generator with lower-voltage (480-V) slip rings. Production of x-rays in bremsstrahlung tubes is an inefficient process, and hence most of the power delivered to the tubes results in heating of


FIGURE 62.5 Schematic illustration of a fifth-generation ultrafast CT system. Image data are acquired in as little as 50 ms, as an electron beam is swept over the strip anode electronically. (Courtesy of Imatron, Inc.)

the anode. A heat exchanger on the rotating gantry is used to cool the tube. Spiral scanning, in particular, places heavy demands on the heat-storage capacity and cooling rate of the x-ray tube. The intensity of the x-ray beam is attenuated by absorption and scattering processes as it passes through the patient. The degree of attenuation depends on the energy spectrum of the x-rays as well as on the average atomic number and mass density of the patient tissues. The transmitted intensity is given by − µ ( x ) dx I t = I oe ∫0 L

(62.1)

where Io and II are the incident and transmitted beam intensities, respectively; L is the length of the x-ray path; and µ(x) is the x-ray linear attenuation coefficient, which varies with tissue type and hence is a function of the distance x through the patient. The integral of the attenuation coefficient is therefore given by

1 ∫ µ(x )dx = − L ln(I I ) L

0

t

o

(62.2)

The reconstruction algorithm requires measurements of this integral along many paths in the fan beam at each of many angles about the isocenter. The value of L is known, and Io is determined by a system calibration. Hence values of the integral along each path can be determined from measurements of It . X-Ray Detectors X-ray detectors used in CT systems must (a) have a high overall efficiency to minimize the patient radiation dose, have a large dynamic range, (b) be very stable with time, and (c) be insensitive to temperature variations within the gantry. Three important factors contributing to the detector efficiency are geometric efficiency, quantum (also called capture) efficiency, and conversion efficiency [Villafanaet et al., 1987]. Geometric efficiency refers to the area of the detectors sensitive to radiation as a fraction of the total exposed area. Thin septa between detector elements to remove scattered radiation, or other


FIGURE 62.6 Photograph of the slip rings used to pass power and control signals to the rotating gantry. (Courtesy of Picker International, Inc.)

insensitive regions, will degrade this value. Quantum efficiency refers to the fraction of incident x-rays on the detector that are absorbed and contribute to the measured signal. Conversion efficiency refers to the ability to accurately convert the absorbed x-ray signal into an electrical signal (but is not the same as the energy conversion efficiency). Overall efficiency is the product of the three, and it generally lies between 0.45 and 0.85. A value of less than 1 indicates a nonideal detector system and results in a required increase in patient radiation dose if image quality is to be maintained. The term dose efficiency sometimes has been used to indicate overall efficiency. Modern commercial systems use one of two detector types: solid-state or gas ionization detectors. Solid-State Detectors. Solid-state detectors consist of an array of scintillating crystals and photodiodes, as illustrated in Fig. 62.8. The scintillators generally are either cadmium tungstate (CdWO4) or a ceramic material made of rare earth oxides, although previous scanners have used bismuth germanate crystals with photomultiplier tubes. Solid-state detectors generally have very high quantum and conversion efficiencies and a large dynamic range. Gas Ionization Detectors. Gas ionization detectors, as illustrated in Fig. 62.9, consist of an array of chambers containing compressed gas (usually xenon at up to 30 atm pressure). A high voltage is applied to tungsten septa between chambers to collect ions produced by the radiation. These detectors have excellent stability and a large dynamic range; however, they generally have a lower quantum efficiency than solid-state detectors.


FIGURE 62.7 Spiral scanning causes the focal spot to follow a spiral path around the patient as indicated. (Courtesy of Picker International, Inc.)

FIGURE 62.8 (a) A solid-state detector consists of a scintillating crystal and photodiode combination. (b) Many such detectors are placed side by side to form a detector array that may contain up to 4800 detectors.

Data-Acquisition System The transmitted fraction It/Io in Eq. (62.2) through an obese patient can be less than 10–4. Thus it is the task of the data-acquisition system (DAS) to accurately measure It over a dynamic range of more than 104, encode the results into digital values, and transmit the values to the system computer for reconstruction. Some manufacturers use the approach illustrated in Fig. 62.10, consisting of precision preamplifiers, current-to-voltage converters, analog integrators, multiplexers, and analog-to-digital converters. Alternatively, some manufacturers use the preamplifier to control a synchronous voltage-to-frequency converter (SVFC), replacing the need for the integrators, multiplexers, and analog-to-digital converters [Brunnett, et al., 1990]. The logarithmic conversion required in Eq. (62.2) is performed with either an analog logarithmic amplifier or a digital lookup table, depending on the manufacturer.


FIGURE 62.9 Gas ionization detector arrays consist of high-pressure gas in multiple chambers separated by thin septa. A voltage is applied between alternating septa. The septa also act as electrodes and collect the ions created by the radiation, converting them into an electrical signal.

Sustained data transfer rates to the computer are as high as 10 Mbytes/s for some scanners. This can be accomplished with a direct connection for systems having a fixed detector array. However, third-generation slip-ring systems must use more sophisticated techniques. At least one manufacturer uses optical transmitters on the rotating gantry to send data to fixed optical receivers [Siemens, 1989]. Computer System

FIGURE 62.10

The data-acquisition system con-

Various computer systems are used by manufacturers verts the electrical signal produced by each detector to a digital value for the computer. to control system hardware, acquire the projection data, and reconstruct, display, and manipulate the tomographic images. A typical system is illustrated in Fig. 62.11, which uses 12 independent processors connected by a 40-Mbyte/s multibus. Multiple custom array processors are used to achieve a combined computational speed of 200 MFLOPS (million floating-point operations per second) and a reconstruction time of approximately 5 s to produce an image on a 1024 × 1024 pixel display. A simplified UNIX operating system is used to provide a multitasking, multiuser environment to coordinate tasks.

Patient Dose Considerations The patient dose resulting from CT examinations is generally specified in terms of the CT dose index (CTDI) [Felmlee et al., 1989; Rothenberg and Pentlow, 1992], which includes the dose contribution from radiation scattered from nearby slices. A summary of CTDI values, as specified by four manufacturers, is given in Table 62.1.

Summary Computed tomography revolutionized medical radiology in the early 1970s. Since that time, CT technology has developed dramatically, taking advantage of developments in computer hardware and detector technology. Modern systems acquire the projection data required for one tomographic image in approximately


FIGURE 62.11 The computer system controls the gantry motions, acquires the x-ray transmission measurements, and reconstructs the final image. The system shown here uses 12 68000-family CPUs. (Courtesy of Picker International, Inc.)

TABLE 62.1 Summary of the CT Dose Index (CTDI) Values at Two Positions (Center of the Patient and Near the Skin) as Specified by Four CT Manufacturers for Standard Head and Body Scans. Manufacturer

Detector

kVp

mA

Scan Time (s)

CTDI, center (mGy)

CTDI, skin (mGy)

A, head A, body A, head A, body B, head B, body C, head C, body D, head D, body

Xenon Xenon Solid state Solid state Solid state Solid state Solid state Solid state Solid state Solid state

120 120 120 120 130 130 120 120 120 120

170 170 170 170 80 80 500 290 200 200

2 2 2 2 2 2 2 1 2 2

50 14 40 11 37 15 39 12 78 9

48 25 40 20 41 34 50 28 78 16

1 s and present the reconstructed image on a 1024 × 1024 matrix display within a few seconds. The images are high-quality tomographic “maps” of the x-ray linear attenuation coefficient of the patient tissues.

Defining Terms Absorption: Some of the incident x-ray energy is absorbed in patient tissues and hence does not contribute to the transmitted beam. Anode: A tungsten bombarded by a beam of electrons to produce x-rays. In all but one fifth-generation system, the anode rotates to distribute the resulting heat around the perimeter. The anode heatstorage capacity and maximum cooling rate often limit the maximum scanning rates of CT systems.


Attenuation: The total decrease in the intensity of the primary x-ray beam as it passes through the patient, resulting from both scatter and absorption processes. It is characterized by the linear attenuation coefficient. Computed tomography (CT): A computerized method of producing x-ray tomographic images. Previous names for the same thing include computerized tomographic imaging, computerized axial tomography (CAT), computer-assisted tomography (CAT), and reconstructive tomography (RT). Control console: The control console is used by the CT operator to control the scanning operations, image reconstruction, and image display. Cormack, Dr. Allan MacLeod: A physicist who developed mathematical techniques required in the reconstruction of tomographic images. Dr. Cormack shared the Nobel Prize in Medicine and Physiology with Dr. G. N. Hounsfield in 1979 [Cormack, 1980]. Data-acquisition system (DAS): Interfaces the x-ray detectors to the system computer and may consist of a preamplifier, integrator, multiplexer, logarithmic amplifier, and analog-to-digital converter. Detector array: An array of individual detector elements. The number of detector elements varies between a few hundred and 4800, depending on the acquisition geometry and manufacturer. Each detector element functions independently of the others. Fan beam: The x-ray beam is generated at the focal spot and so diverges as it passes through the patient to the detector array. The thickness of the beam is generally selectable between 1.0 and 10 mm and defines the slice thickness. Focal spot: The region of the anode where x-rays are generated. Focused septa: Thin metal plates between detector elements which are aligned with the focal spot so that the primary beam passes unattenuated to the detector elements, while scattered x-rays which normally travel in an altered direction are blocked. Gantry: The largest component of the CT installation, containing the x-ray tube, collimators, detector array, DAS, other control electronics, and the mechanical components required for the scanning motions. Helical scanning: The scanning motions in which the x-ray tube rotates continuously around the patient while the patient is continuously translated through the fan beam. The focal spot therefore traces a helix around the patient. Projection data are obtained which allow the reconstruction of multiple contiguous images. This operation is sometimes called spiral, volume, or three-dimensional CT scanning. Hounsfield, Dr. Godfrey Newbold: An engineer who developed the first practical CT instrument in 1971. Dr. Hounsfield received the McRobert Award in 1972 and shared the Nobel Prize in Medicine and Physiology with Dr. A. M. Cormack in 1979 for this invention [Hounsfield, 1980]. Image plane: The plane through the patient that is imaged. In practice, this plane (also called a slice) has a selectable thickness between 1.0 and 10 mm centered on the image plane. Pencil beam: A narrow, well-collimated beam of x-rays. Projection data: The set of transmission measurements used to reconstruct the image. Reconstruct: The mathematical operation of generating the tomographic image from the projection data. Scan time: The time required to acquire the projection data for one image, typically 1.0 s. Scattered radiation: Radiation that is removed from the primary beam by a scattering process. This radiation is not absorbed but continues along a path in an altered direction. Slice: See Image plane. Spiral scanning: See Helical scanning. Three-dimensional imaging: See Helical scanning. Tomography: A technique of imaging a cross-sectional slice. Volume CT: See Helical scanning. X-ray detector: A device that absorbs radiation and converts some or all of the absorbed energy into a small electrical signal.


X-ray linear attenuation coefficient µ: Expresses the relative rate of attenuation of a radiation beam as it passes through a material. The value of µ depends on the density and atomic number of the material and on the x-ray energy. The units of µ are cm–1. X-ray source: The device that generates the x-ray beam. All CT scanners are rotating-anode bremsstrahlung x-ray tubes except one-fifth generation system, which uses a unique scanned electron beam and a strip anode. X-ray transmission: The fraction of the x-ray beam intensity that is transmitted through the patient without being scattered or absorbed. It is equal to It/Io in Eq. (62.2), can be determined by measuring the beam intensity both with (It) and without (Io) the patient present, and is expressed as a fraction. As a rule of thumb, n2 independent transmission measurements are required to reconstruct an image with an n × n sized pixel matrix.

References Body DP, et al. 1979. A proposed dynamic cardiac 3D densitometer for early detection and evaluation of heart disease. IEEE Trans Nucl Sci 2724. Boyd DP, Lipton MJ. 1983. Cardiac computed tomography. Proc IEEE 198. Brunnett CJ, Heuscher DJ, Mattson RA, Vrettos CJ. 1990. CT Design Considerations and Specifications. Picker International, CT Engineering Department, Ohio. Cormack AM. 1980. Nobel Award Address: Early two-dimensional reconstruction and recent topics stemming from it. Med Phys 7(4):277. Felmlee JP, Gray JE, Leetzow ML, Price JC. 1989. Estimated fetal radiation dose from multislice CT studies. Am Roent Ray Soc 154:185. Hounsfield GN. 1980. Nobel Award Address: Computed medical imaging. Med Phys 7(4):283. Kalendar WA, Seissler W, Klotz E, et al. 1990. Spiral volumetric CT with single-breath-hold technique, continuous transport, and continuous scanner rotation. Radiology 176:181. Newton TH, Potts DG (eds). 1981. Radiology of the Skull and Brain: Technical Aspects of Computed Tomography. St. Louis, Mosby. Picker. 1990. Computed Dose Index PQ2000 CT Scanner. Picker International, Ohio. Ritman EL. 1980. Physical and technical considerations in the design of the DSR, and high temporal resolution volume scanner. AJR 134:369. Ritman EL. 1990. Fast computed tomography for quantitative cardiac analysis—State of the art and future perspectives. Mayo Clin Proc 65:1336. Robb RA. 1982. X-ray computed tomography: An engineering synthesis of multiscientific principles. CRC Crit Rev Biomed Eng 7:265. Rothenberg LN, Pentlow KS. 1992. Radiation dose in CT. RadioGraphics 12:1225. Seeram E. 1994. Computed Tomography: Physical Principles, Clinical Applications and Quality Control. Philadelphia, Saunders. Siemens. 1989. The Technology and Performance of the Somatom Plus. Siemens Aktiengesellschaft, Medical Engineering Group, Erlangen, Germany. Villafana T, Lee SH, Rao KCVG (eds). 1987. Cranial Computed Tomography. New York, McGraw-Hill.

Further Information A recent summary of CT instrumentation and concepts is given by E. Seeram in Computed Tomography: Physical Principles, Clinical Applications and Quality Control. The author summarizes CT from the perspective of the nonmedical, nonspecialist user. A summary of average CT patient doses is described by Rothenberg and Pentlow [1992] in Radiation Dose in CT. Research papers on both fundamental and practical aspects of CT physics and instrumentation are published in numerous journals, including Medical Physics, Physics in Medicine and Biology, Journal of Computer Assisted Tomography, Radiology, British Journal of Radiology, and the IEEE Press. A comparison of technical specifications of CT systems


provided by the manufacturers is available from ECRI to help orient the new purchaser in a selection process. Their Product Comparison System includes a table of basic specifications for all the major international manufactures.

62.2 Reconstruction Principles Philip F. Judy Computed tomography (CT) is a two-step process: (1) the transmission of an x-ray beam is measured through all possible straight-line paths as in a plane of an object, and (2) the attenuation of an x-ray beam is estimated at points in the object. Initially, the transmission measurements will be assumed to be the results of an experiment performed with a narrow monoenergetic beam of x-rays that are confined to a plane. The designs of devices that attempt to realize these measurements are described in the preceding section. One formal consequence of these assumptions is that the logarithmic transformation of the measured x-ray intensity is proportional to the line integral of attenuation coefficients. In order to satisfy this assumption, computer processing procedures on the measurements of x-ray intensity are necessary even before image reconstruction is performed. These linearization procedures will reviewed after background. Both analytical and iterative estimations of linear x-ray attenuation have been used for transmission CT reconstruction. Iterative procedures are of historic interest because an iterative reconstruction procedure was used in the first commercially successful CT scanner [EMI, Mark I, Hounsfield, 1973]. They also permit easy incorporation of physical processes that cause deviations from the linearity. Their practical usefulness is limited. The first EMI scanner required 20 minutes to finish its reconstruction. Using the identical hardware and employing an analytical calculation, the estimation of attenuation values was performed during the 4.5-minute data acquisition and was made on a 160 × 160 matrix. The original iterative procedure reconstructed the attenuation values on an 80 × 80 matrix and consequently failed to exploit all the spatial information inherent in transmission data. Analytical estimation, or direct reconstruction, uses a numerical approximation of the inverse Radon transform [Radon, 1917]. The direct reconstruction technique (convolution-backprojection) presently used in x-ray CT was initially applied in other areas such as radio astronomy [Bracewell and Riddle, 1967] and electron microscopy [Crowther et al., 1970; Ramachandran and Lakshminarayana, 1971]. These investigations demonstrated that the reconstructions from the discrete spatial sampling of bandlimited data led to full recovery of the cross-sectional attenuation. The random variation (noise) in x-ray transmission measurements may not be bandlimited. Subsequent investigators [e.g., Chesler and Riederer, 1975; Herman and Roland, 1973; Shepp and Logan, 1974] have suggested various bandlimiting windows that reduce the propagation and amplification of noise by the reconstruction. These issues have been investigated by simulation, and investigators continue to pursue these issues using a computer phantom [e.g., Guedon and Bizais, 1994, and references therein] described by Shepp and Logan. The subsequent investigations of the details of choice of reconstruction parameters has had limited practical impact because real variation of transmission data is bandlimited by the finite size of the focal spot and radiation detector, a straightforward design question, and because random variation of the transmission tends to be uncorrelated. Consequently, the classic precedures suffice.

Image Processing: Artifact and Reconstruction Error An artifact is a reconstruction defect that is obviously visible in the image. The classification of an image feature as an artifact involves some visual criterion. The effect must produce an image feature that is greater than the random variation in image caused by the intrinsic variation in transmission measurements. An artifact not recognized by the physician observer as an artifact may be reported as a lesion. Such false-positive reports could lead to an unnecessary medical procedure, e.g., surgery to remove an imaginary tumor. A reconstruction error is a deviation of the reconstruction value from its expected value.


Reconstruction errors are significant if the application involves a quantitative measurement, not a common medical application. The reconstruction errors are characterized by identical material at different points in the object leading to different reconstructed attenuation values in the image which are not visible in the medical image. Investigators have used computer simulation to investigate artifact [Herman, 1980] because image noise limits the visibility of their visibility. One important issue investigated was required spatial sampling of transmission slice plane [Crawford and Kak, 1979; Parker et al., 1982]. These simulations provided a useful guideline in design. In practice, these aliasing artifacts are overwhelmed by random noise, and designers tend to oversample in the slice plane. A second issue that was understood by computer simulation was the partial volume artifact [Glover and Pelc, 1980]. This artifact would occur even for mononergetic beams and finite beam size, particularly in the axial dimension. The axial dimension of the beams tend to be greater (about 10 mm) than their dimensions in the slice plane (about 1 mm). The artifact is created when the variation of transmission within the beam varies considerably, and the exponential variation within the beam is summed by the radiation detector. The logarithm transformation of the detected signal produces a nonlinear effect that is propagated throughout the image by the reconstruction process. Simulation was useful in demonstrating that isolated features in the same crosssection act together to produce streak artifacts. Simulations have been useful to illustrate the effects of patient motion during the data-acquisition streaks off high-contrast objects.

Projection Data to Image: Calibrations Processing of transmission data is necessary to obtain high-quality images. In general, optimization of the projection data will optimize the reconstructed image. Reconstruction is a process that removes the spatial correlation of attenuation effects in the transmitted image by taking advantage of completely sampling the possible transmissions. Two distinct calibrations are required: registration of beams with the reconstruction matrix and linearization of the measured signal. Without loss of generalization, a projection will be considered a set of transmissions made along parallel lines in the slice plane of the CT scanner. Without loss of generalization means that essential aspects of all calibration and reconstruction procedures required for fan-beam geometries are captured by the calibration and reconstruction procedures described for parallel projections. One line of each projection is assumed to pass through the center of rotation of data collection. Shepp et al. [1979] showed that errors in the assignment of that center-of-rotation point in the projections could lead to considerable distinctive artifacts and that small errors (0.05 mm) would produce these effects. The consequences of these errors have been generalized to fan-beam collection schemes, and images reconstructed from 180degree projection sets were compared with images reconstructed from 360-degree data sets [Kijewski and Judy, 1983]. A simple misregistration of the center of rotation was found to produce blurring of image without the artifact. These differences may explain the empirical observation that most commercial CT scanners collect a full 360-degree data set even though 180 degrees of data will suffice. The data-acquisition scheme that was designed to overcome the limited sampling inherent in thirdgeneration fan-beam systems by shifting detectors a quarter sampling distance while opposite 180-degree projection is measured, has particularly stringent registration requirements. Also, the fourth-generation scanner does not link the motion of the x-ray tube and the detector; consequently, the center of rotation is determined as part of a calibration procedure, and unsystematic effects lead to artifacts that mimic noise besides blurring the image. Misregistration artifacts also can be mitigated by feathering. This procedure requires collection of redundant projection data at the end of the scan. A single data set is produced by linearly weighting the redundant data at the beginning and end of the data collection [Parker et al., 1982]. These procedures have be useful in reducing artifacts from gated data collections [Moore et al., 1987]. The other processing necessary before reconstruction of project data is linearization. The formal requirement for reconstruction is that the line integrals of some variable be available; this is the variable that ultimately is reconstructed. The logarithm of x-ray transmission approximates this requirement. There © 2000 by CRC Press LLC

are physical effects in real x-ray transmissions that cause deviations from this assumption. X-ray beams of sufficient intensity are composed of photons of different energies. Some photons in the beam interact with objects and are scattered rather than absorbed. The spectrum of x-ray photons of different attenuation coefficients means the logarithm of the transmission measurement will not be proportional to the line integral of the attenuation coefficient along that path, because an attenuation coefficient cannot even be defined. An effective attenuation coefficient can only be defined uniquely for a spectrum for a small mass of material that alters that intensity. It has to be small enough not to alter the spectrum [McCullough, 1979]. A straightforward approach to this nonunique attenuation coefficient error, called hardening, is to assume that the energy dependence of the attenuation coefficient is constant and that differences in attenuation are related to a generalized density factor that multiplies the spectral dependence of attenuation. The transmission of an x-ray beam then can be estimated for a standard material, typically water, as a function of thickness. This assumption is that attenuations of materials in the object, the human body, differ because specific gravities of the materials differ. Direct measurements of the transmission of an actual x-ray beam may provide initial estimates that can be parameterized. The inverse of this function provides the projection variable that is reconstructed. The parameters of the function are usually modified as part of a calibration to make the CT image of a uniform water phantom flat. Such a calibration procedure does not deal completely with the hardening effects. The spectral dependence of bone differs considerably from that of water. This is particularly critical in imaging of the brain, which is contained within the skull. Without additional correction, the attenuation values of brain are lower in the center than near the skull. The detection of scattered energy means that the reconstructed attenuation coefficient will differ from the attenuation coefficient estimated with careful narrow-beam measurements. The x-rays appear more penetrating because scattered x-rays are detected. The zero-ordered scatter, a decrease in the attenuation coefficient by some constant amount, is dealt with automatically by the calibration that treats hardening. First-order scattering leads to a widening of the x-ray beam and can be dealt with by a modification of the reconstruction kernel.

Projection Data to Image: Reconstruction The impact of CT created considerable interest in the formal aspects of reconstruction. There are many detailed descriptions of direct reconstruction procedures. Some are presented in textbooks used in graduate courses for medical imaging [Barrett and Swindell, 1981; Cho et al., 1993]. Herman (1980) published a textbook that was based on a two-semester course that dealt exclusively with reconstruction principles, demonstrating the reconstruction principles with simulation. The standard reconstruction method is called convolution-backprojection. The first step in the procedure is to convolve the projection, a set of transmissions made along parallel lines in the slice plane, with a reconstruction kernel derived from the inverse Radon transform. The choice of kernel is dictated by bandlimiting issues [Chesler and Riederer, 1975; Herman and Roland, 1973; Shepp and Logan, 1974]. It can be modified to deal with the physical aperture of the CT system [Bracewell, 1977], which might include the effects of scatter. The convolved projection is then backprojected onto a two-dimensional image matrix. Backprojection is the opposite of projection; the value of the projection is added to each point along the line of the projection. This procedure makes sense in the continuous description, but in the discrete world of the computer, the summation is done over the image matrix. Consider a point of the image matrix; very few, possibly no lines of the discrete projection data intersect the point. Consequently, to estimate the projection value to be added to that point, the procedure must interpolate between two values of sampled convolve projection. The linear interpolation scheme is a significant improvement over nearest project nearest to the point. More complex schemes get confounded with choices of reconstruction kernel, which are designed to accomplish standard image processing in the image, e.g., edge enhancement. Scanners have been developed to acquire a three-dimensional set of projection data [Kalender et al., 1990]. The motion of the source defines a spiral motion relative to the patient. The spiral motion defines © 2000 by CRC Press LLC

an axis. Consequently, only one projection is available for reconstruction of the attenuation values in the plane. This is the back-projection problem just discussed; no correct projection value is available from the discrete projection data set. The solution is identical: a projection value is interpolated from the existing projection values to estimate the necessary projections for each plane to be reconstructed. This procedure has the advantage that overlapping slices can be reconstructed without additional exposure, and this eliminates the risk that a small lesion will be missed because it straddles adjacent slices. This data-collection scheme is possible because systems that continuously rotate have been developed. The spiral scan motion is realized by moving the patient through the gantry. Spiral CT scanners have made possible the acquisition of an entire data set in a single breath hold.

References Barrett HH, Swindell W. 1981. Radiological Imaging: The Theory and Image Formation, Detection, and Processing, vol 2. New York, Academic Press. Bracewell RN, Riddle AC. 1976. Inversion of fan-beam scans in radio astronomy. The Astrophysical Journal 150:427-434. Chesler DA, Riederer SJ. 1975. Ripple suppression during reconstruction in transverse tomography. Phys Med Biol 20(4):632-636. Cho Z, Jones JP, Singh M. 1993. Foundations of medical imaging. New York, Wiley & Sons, Inc. Crawford CR, Kak AC. 1979. Aliasing artifacts in computerized tomography. Applied Optics 18:3704-3711. Glover GH, Pelc NJ. 1980. Nonlinear partial volume artifacts in x-ray computed tomography. Med Phys 7:238-248. Guedon J-P, Bizais. 1994. Bandlimited and harr filtered back-projection reconstruction. IEEE Trans Medical Imaging 13(3):430-440. Herman GT, Rowland SW. 1973. Three methods for reconstruction objects for x-rays—a comparative study. Comp Graph Imag Process 2:151–178. Herman GT. 1980. Image Reconstruction from Projection: The Fundamentals of Computerized Tomography. New York, New York, Academic Press. Hounsfield, GN. 1973. Computerized transverse axial scanning (tomography): Part I. Brit J Radiol 46:1016–1022. Kalender WA, Weissler, Klotz E, et al. 1990. Spiral volumetric CT with single-breath-hold technique, continuous transport, and continuous scanner rotation. Radiology 176:181–183. Kijewski MF, Judy PF. 1983. The effect of misregistration of the projections on spatial resolution of CT scanners. Med Phys 10:169–175. McCullough EC. 1979. Specifying and evaluating the performance of computed tomographic (CT) scanners. Med Phys 7:291–296. Moore SC, Judy PF, Garnic JD, et al. 1983. The effect of misregistration of the projections on spatial resolution of CT scanners. Med Phys 10:169–175.


Macovski, A., Pauly, J., Schenck, J., Kwong, K. K., Chesler, D. A., Hu. X., Chen, W., Patel, M., Ugurbil, K. Conolly, S. "Magnetic Resonance Imaging" The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

63 Magnetic Resonance Imaging Steven Conolly

63.1

Acquisition and Processing

63.2

Hardware / Instrumentation

Fundamentals of MRI • Contrast Mechanisms

Stanford University

Albert Macovski

Fundamentals of MRI Instrumentation • Static Field Magnets • Gradient Coils • Digital Data Processing • Current Trends in MRI

Stanford University

John Pauly Stanford University

63.3

John Schenck General Electric Corporate Research and Development Center

Kenneth K. Kwong

63.4

Massachusetts General Hospital and Harvard University Medical School

David A. Chesler Massachusetts General Hospital and Harvard University Medical School

Xiaoping Hu Center for Magnetic Resonance Research and the University of Minnesota Medical School

Wei Chen Center for Magnetic Resonance Research and the University of Minnesota Medical School

Maqbool Patel Center for Magnetic Resonance Research and the University of Minnesota Medical School

Kamil Ugurbil Center for Magnetic Resonance Research and the University of Minnesota Medical School


Functional MRI Advanced in Functional Brain Mapping • Mechanism • Problem and Artifacts in fMRI: The Brain-Vein Problem? The Brain-Inflow Problem? • Techniques to Reduce the Large Vessel Problems

Chemical-Shift Imaging: An Introduction to Its Theory and Practice General Methodology • Practical Examples

63.1

Acquisition and Processing

Steven Conolly, Albert Macovski, and John Pauly Magnetic resonance imaging (MRI) is a clinically important medical imaging modality due to its exceptional soft-tissue contrast. MRI was invented in the early 1970s [1]. The first commercial scanners appeared about 10 years later. Noninvasive MRI studies are now supplanting many conventional invasive procedures. A 1990 study [2] found that the principal applications for MRI are examinations of the head (40%), spine (33%), bone and joints (17%), and the body (10%). The percentage of bone and joint studies was growing in 1990. Although typical imaging studies range from 1 to 10 minutes, new fast imaging techniques acquire images in less than 50 ms. MRI research involves fundamental tradeoffs between resolution, imaging time, and signal-to-noise ratio (SNR). It also depends heavily on both gradient and receiver coil hardware innovations. In this section we provide a brief synopsis of basic nuclear magnetic resonance (NMR) physics. We then derive the k-space analysis of MRI, which interprets the received signal as a scan of the Fourier

transform of the image. This powerful formalism is used to analyze the most important imaging sequences. Finally, we discuss the fundamental contrast mechanisms for MRI.

Fundamentals of MRI Magnetic resonance imaging exploits the existence of induced nuclear magnetism in the patient. Materials with an odd number of protons or neutrons possess a weak but observable nuclear magnetic moment. Most commonly protons (1H) are imaged, although carbon (13C), phosphorous (31P), sodium (23Na), and fluorine (19F) are also of significant interest. The nuclear moments are normally randomly oriented, but they align when placed in a strong magnetic field. Typical field strengths for imaging range between 0.2 and 1.5 T, although spectroscopic and functional imaging work is often performed with higher field strengths. The nuclear magnetization is very weak; the ratio of the induced magnetization to the applied fields is only 4 × 10–9. The collection of nuclear moments is often referred to as magnetization or spins. The static nuclear moment is far too weak to be measured when it is aligned with the strong static magnetic field. Physicists in the 1940s developed resonance techniques that permit this weak moment to be measured. The key idea is to measure the moment while it oscillates in a plane perpendicular to the static field [3,4]. First one must tip the moment away from the static field. When perpendicular to the static field, the moment feels a torque proportional to the strength of the static magnetic field. The torque always points perpendicular to the magnetization and causes the spins to oscillate or precess in a plane perpendicular to the static field. The frequency of the rotation ω0 is proportional to the field:

ω 0 = − γB0 where γ, the gyromagnetic ratio, is a constant specific to the nucleus, and B0 is the magnetic field strength. The direction of B0 defines the z axis. The precession frequency is called the Larmor frequency. The negative sign indicates the direction of the precession. Since the precessing moments constitute a time-varying flax, they produce a measurable voltage in a loop antenna arranged to receive the x and y components of induction. It is remarkable that in MRI we are able to directly measure induction from the precessing nuclear moments of water protons. Recall that to observe this precession, we first need to tip the magnetization away from the static field. This is accomplished with a weak rotating radiofrequency (RF) field. It can be shown that a rotating RF field introduces a fictitious field in the z direction of strength ω/γ. By tuning the frequency of the RF field to ω0 , we effectively delete the B0 field. The RF slowly nutates the magnetization away from the z axis. The Larmor relation still holds in this “rotating frame,” so the frequency of the nutation is γB1, where B1 is the amplitude of the RF field. Since the coils receive x and y (transverse) components of induction, the signal is maximized by tipping the spins completely into the transverse plane. This is accomplished by a π/2 RF pulse, which requires γB1τ = π/2, where τ is the duration of the RF pulse. Another useful RF pulse rotates spins by π radians. This can be used to invert spins. It also can be used to refocus transverse spins that have dephased due to B0 field inhomogeneity. This is called a spin echo and is widely used in imaging. NMR has been used for decades in chemistry. A complex molecule is placed in a strong, highly uniform magnetic field. Electronic shielding produces microscopic field variations within the molecule so that geometrically isolated nuclei rotate about distinct fields. Each distinct magnetic environment produces a peak in the spectra of the received signal. The relative size of the spectral peaks gives the ratio of nuclei in each magnetic environment. Hence the NMR spectrum is extremely useful for elucidating molecular structure. The NMR signal from a human is due predominantly to water protons. Since these protons exist in identical magnetic environments, they all resonate at the same frequency. Hence the NMR signal is simply proportional to the volume of the water. They key innovation for MRI is to impose spatial variations on the magnetic field to distinguish spins by their location. Applying a magnetic field gradient causes each region of the volume to oscillate at a distinct frequency. The most effective nonuniform field is a linear © 2000 by CRC Press LLC

gradient where the field and the resulting frequencies vary linearly with distance along the object being studied. Fourier analysis of the signal obtains a map of the spatial distribution of spins. This argument is formalized below, where we derive the powerful k-space analysis of MRI [5,6]. k-Space Analysis of Data Acquisition In MRI, we receive a volume integral from an array of oscillators. By ensuring that the phase, “signature” of each oscillator is unique, one can assign a unique location to each spin and thereby reconstruct an image. During signal reception, the applied magnetic field points in the z direction. Spins precess in the xy plane at the Larmor frequency. Hence a spin at position r = (x,y,z) has a unique phase θ that describes its angle relative to the y axis in the xy plane:

( )

θ r, t = − γ

∫ B (r, τ)dτ t

z

0

(63.1)

where Bz(r,t) is the z component of the instantaneous, local magnetic flux density. This formula assumes there are no x and y field components. A coil large enough to receive a time-varying flux uniformly from the entire volume produces an EMF proportional to

()

st ∝

d dt

∫ M (r)e

( )dr

− iθ r ,t

(63.2)

V

where M(r) represents the equilibrium moment density at each point r. The key idea for imaging is to superimpose a linear field gradient on the static field B0. This field points in the direction z, and its magnitude varies linearly with a coordinate direction. For example, an x gradient points in the z direction and varies along the coordinate x. This is described by the vector field xGx zˆ , where zˆ is the unit vector in the z direction. In general, the gradient is (xGx + yGy + zGz) zˆ, which can be written compactly as the dot product G·r zˆ . These gradient field components can vary with time, so the total z field is

( )

()

Bz r, t = B0 + G t ⋅ r

(63.3)

In the presence of this general time-varying gradient, the received signal is

()

st ∝

d dt

∫

− iγ G ( τ ) ⋅ r dτ e − iγB0t M r e ∫0 dr

()

V

t

(63.4)

The center frequency γB0 is always much larger than the bandwidth of the signal. Hence the derivative operation is approximately equivalent to multiplication by –i ω0 . The signal is demodulated by the waveform ei γB0t to obtain the “baseband” signal:

()

s t ∝ −iω 0

− iγ G ( τ ) ⋅ r dτ M r e ∫0 dr

∫ ()

t

(63.5)

V

It will be helpful to define the term k(t):

( ) ∫ G(τ) dτ t

k t =γ

0


(63.6)

FIGURE 63.1 The drawing on the left illustrates the scanning pattern of the 2D Fourier transform imaging sequence. On the right is a plot of the gradient and RF waveforms that produce this pattern. Only four of the Ny horizontal k-space lines are shown. The phase-encode period initiates each acquisition at a different ky and at –kx(max). Data are collected during the horizontal traversals. After all Ny k-space lines have been acquired, a 2D FFT reconstructs the image. Usually 128 or 256 ky lines are collected, each with 256 samples along kx . The RF pulse and the z gradient waveform together restrict the image to a particular slice through the subject.

Then we can rewrite the received baseband signal as

( ) ∫ M (r)e

St ∝

( ) dr

− ik t ⋅ r

(63.7)

V

which we can now identify as the spatial Fourier transform of M(r) evaluated at k(t). That is, S(t) scans the spatial frequencies of the function M(r). This can be written explicitly as

()

( ( ))

S t ∝M k t

(63.8)

where M(k) is the three-dimensional Fourier transform of the object distribution M(r). Thus we can view MRI with linear gradients as a “scan” of k-space or the spatial Fourier transform of the image. After the desired portion of k-space is scanned, the image M(r) is reconstructed using an inverse Fourier transform. 2D Imaging. Many different gradient waveforms can be used to scan k-space and to obtain a desired image. The most common approach, called two-dimensional Fourier transform imaging (2D FT), is to scan through k-space along several horizontal lines covering a rectilinear grid in 2D k-space. See Fig. 63.1 for a schematic of the k-space traversal. The horizontal grid lines are acquired using 128 to 256 excitations separated by a time TR, which is determined by the desired contrast, RF flip angle, and the T1 of the desired components of the image. The horizontal-line scans through k-space are offset in ky by a variable area y-gradient pulse, which happens before data acquisition starts. These variable offsets in ky are called phase encodes because they affect the phase of the signal rather than the frequency. Then for each ky phase encode, signal is acquired while scanning horizontally with a constant x gradient. Resolution and Field of View. The fundamental image characteristics of resolution and field of view (FOV) are completely determined by the characteristics of the k-space scan. The extent of the coverage


of k-space determines the resolution of the reconstructed image. The resolution is inversely proportional to the highest spatial frequency acquired:

( )

γG T 1 k x max = = x ∆x 2π π

( )

γG T 1 k y max = = y phase π π ∆y

(63.9)

(63.10)

where Gx is the readout gradient amplitude and T is the readout duration. The time Tphase is the duration of the phase-encode gradient Gy . For proton imaging on a 1.5-T imaging system, a typical gradient strength is Gx = 1 G/cm. The signal is usually read for about 8 ms. For water protons, γ = 26,751 rad/s/G, so the maximum excursion in kx is about 21 rad/mm. Hence we cannot resolve an object smaller than 0.3 mm in width. From this one might be tempted to improve the resolution dramatically using very strong gradients or very long readouts. But there are severe practical obstacles, since higher resolution increases the scan time and also degrades the image SNR. In the phase-encode direction, the k-space data are sampled discretely. This discrete sampling in k-space introduces replication in the image domain [7]. If the sampling in k-space is finer than 1/FOV, then the image of the object will not fold back on itself. When the k-space sampling is coarser than 1/FOV, the image of the object does fold back over itself. This is termed aliasing. Aliasing is prevented in the readout direction by the sampling filter. Perspective. For most imaging systems, diffraction limits the resolution. That is, the resolution is limited to the wavelength divided by the angle subtended by the receiver aperture, which means that the ultimate resolution is approximately the wavelength itself. This is true for imaging systems based on optics, ultrasound, and x-rays (although there are other important factors, such as quantum noise, in x-ray). MRI is the only imaging system for which the resolution is independent of the wavelength. In MRI, the wavelength is often many meters, yet submillimeter resolution is routinely achieved. The basic reason is that no attempt is made to focus the radiation pattern to the individual pixel or voxel (volume element), as is done in all other imaging modalities. Instead, the gradients create spatially varying magnetic fields so that individual pixels emit unique waveform signatures. These signals are decoded and assigned to unique positions. An analogous problem is isolating the signals from two transmitting antenna towers separated by much less than a wavelength. Directive antenna arrays would fail because of diffraction spreading. However, we can distinguish the two signals if we use the a priori knowledge that the two antennas transmit at different frequencies. We can receive both signals with a wide-angle antenna and then distinguish the signals through frequency-selective filtering. SNR Considerations. The signal strength is determined by the EMF induced from each voxel due to the processing moments. The magnetic moment density is proportional to the polarizing field B0. Recall that the EMF is proportional to the rate of change of the coil flux. The derivative operation multiples the signal by the Larmor frequency, which is proportional to B0, so the received signal is proportional to B02 times the volume of the voxel Vv . In a well-designed MRI system, the dominant noise source is due to thermally generated currents within the conductive tissues of the body. These currents create a time-varying flux which induces noise voltages in the receiver coil. Other noise sources include the thermal noise from the antenna and from the first amplifier. These subsystems are designed so that the noise is negligible compared with the noise from the patient. The noise received is determined by the total volume seen by the antenna pattern Vn and the effective resistivity and temperature of the conductive tissue. One can show [8] that the standard deviation of the noise from conductive tissue varies linearly with B0. The noise is filtered by an integration


over the total acquisition time Tacq , which effectively attenuates the noise standard deviation by T acq . Therefore, the SNR varies as

SNR ∝

B02Vv B0Vn

Tacq

(

= B0 Tacq Vv Vn

)

(63.11)

The noise volume Vn is the effective volume based on the distribution of thermally generated currents. For example, when imaging a spherical object of radius r, the noise standard deviation varies as r 5/2 [9]. The effective resistance depends strongly on the radius because currents near the outer radius contribute more to the noise flux seen by the receiver coil. To significantly improve the SNR, most systems use surface coils, which are simply small coils that are just big enough to see the desired region of the body. Such a coil effectively maximizes the voxel-volume to noise-volume ratio. The noise is significantly reduced because these coils are sensitive to currents from a smaller part of the body. However, the field of view is somewhat limited, so “phased arrays” of small coils are now being offered by the major manufacturers [10]. In the phased array, each coil sees a small noise volume, while the combined responses provide the wide coverage at a greatly improved SNR. Fast Imaging. The 2D FT scan of k-space has the disadvantage that the scan time is proportional to the number of phase encodes. It is often advantageous to trade off SNR for a shorter scan time. This is especially true when motion artifacts dominate thermal noise. To allow for a flexible tradeoff of SNR for imaging time, more than a single line in k-space must be covered in a single excitation. The most popular approach, called echo-planar imaging (EPI), traverses k-space back and forth on a single excitation pulse. The k-space trajectory is drawn in Fig. 63.2. It is important that the tradeoff be flexible so that you can maximize the imaging time given the motion constraints. For example, patients can hold their breath for about 12 seconds. So a scan of 12 seconds’ duration gives the best SNR given the breath-hold constraint. The EPI trajectory can be interleaved to take full advantage of the breath-hold interval. If each acquisition takes about a second, 12 interleaves can be collected. Each interleaf acquires every twelfth line in k-space.

FIGURE 63.2 Alternative methods for the rapid traversal of k space. On the left is the echo planar trajectory. Data are collected during the horizontal traversals. When all Ny horizontal lines in k space have been acquired, the data are sent to a 2D FFT to reconstruct the image. On the right is an interleaved spiral trajectory. The data are interpolated to a 2D rectilinear grid and then Fourier transformed to reconstruct the image. These scanning techniques allow for imaging within a breathhold.


Another trajectory that allows for a flexible tradeoff between scan time and SNR is the spiral trajectory. Here the trajectory starts at the origin in k-space and spirals outward. Interleaving is accomplished by rotating the spirals. Figure 63.2 shows two interleaves in a spiral format. Interleaving is very helpful for reducing the hardware requirements (peak amplitude, peak slew rate, average dissipation, etc.) for the gradients amplifiers. For reconstruction, the data are interpolated to a 2D rectilinear grid and then Fourier-transformed. Our group has found spiral imaging to be very useful for imaging coronary arteries within a breath-hold scan [11]. The spiral trajectory is relatively immune to artifacts due to the motion of blood.

Contrast Mechanisms The tremendous clinical utility of MRI is due to the great variety of mechanisms that can be exploited to create image contrast. If magnetic resonance images were restricted to water density, MRI would be considerably less useful, since most tissues would appear identical. Fortunately, many different MRI contrast mechanisms can be employed to distinguish different tissues and disease processes. The primary contrast mechanisms exploit relaxation of the magnetization. The two types of relaxations are termed spin-lattice relaxation, characterized by a relaxation time T1, and spin-spin relaxation, characterized by a relaxation time T2 . Spin-lattice relaxation describes the rate of recovery of the z component of magnetization toward equilibrium after it has been disturbed by RF pulses. The recovery is given by

()

(

)

()

M z t = M 0 1 − e − t T1 + M z 0 e − t T1

(63.12)

where M0 is the equilibrium magnetization. Differences in the T1 time constant can be used to produce image contrast by exciting all magnetization and then imaging before full recovery has been achieved. This is illustrated on the left in Fig. 63.3. An initial π/2 RF pulse destroys all the longitudinal magnetization. The plots show the recovery of two different T1 components. The short T1 component recovers faster and produces more signal. This gives a T1-weighted image.

FIGURE 63.3 The two primary MRI contrast mechanisms, T1 and T2 . T1, illustrated on the left, describes the rate at which the equilibrium Mzk magnetization is restored after it has been disturbed. T1 contrast is produced by imaging before full recovery has been obtained. T2, illustrated on the right, describes the rate at which the MRI signal decays after it has been created. T2 contrast is produced by delaying data acquisition, so shorter T2 components produce less signal.


FIGURE 63.4 right.

Example images of a normal volunteer demonstrating T1 contrast on the left and T2 contrast on the

Spin-spin relaxation describes the rate at which the NMR signal decays after it has been created. The signal is proportional to the transverse magnetization and is given by

()

()

M xy t = M xy 0 e − t T2

(63.13)

Image contrast is produced by delaying the data acquisition. The decay of two different T2 species is plotted on the right in Fig. 63.3. The signal from the shorter T2 component decays more rapidly, while that of the longer T2 component persists. At the time of data collection, the longer T2 component produces more signal. This produces a T2-weighted image. Figure 63.4 shows examples of these two basic types of contrast. These images are of identical axial sections through the brain of a normal volunteer. The image of the left was acquired with an imaging method that produces T1 contrast. The very bright ring of subcutaneous fat is due to its relatively short T1 . White matter has a shorter T1 than gray matter, so it shows up brighter in this image. The image on the right was acquired with an imaging method that produces T2 contrast. Here the cerebrospinal fluid in the ventricles is bright due to its long T2 . White matter has a shorter T2 than gray matter, so it is darker in this image. There are many other contrast mechanisms that are important in MRI. Different chemical species have slightly different resonant frequencies, and this can be used to image one particular component. It is possible to design RF and gradient waveforms so that the image shows only moving spins. This is of great utility in MR angiography, allowing the noninvasive depiction of the vascular system. Another contrast mechanism is called T 2• . This relaxation parameter is useful for functional imaging. It occurs when there is a significant spread of Larmor frequencies within a voxel. The superposition signal is attenuated faster than T2 due to destructive interference between the different frequencies. In addition to the intrinsic tissue contrast, artificial MRI contrast agents also can be introduced. These are usually administered intravenously or orally. Many different contrast mechanisms can be exploited, but the most popular agents decrease both T1 and T2. One agent approved for clinical use is gadolinium DPTA. Decreasing T1 causes faster signal recovery and a higher signal on a T1-weighted image. The contrast-enhanced regions then show up bright relative to the rest of the image.


Defining Terms Gyromagnetic ratio γ: An intrinsic property of a nucleus. It determines the Larmor frequency through the relation ω0 = –γB0 . k-space: The reciprocal of object space, k-space describes MRI data acquisition in the spatial Fourier transform coordinate system. Larmor frequency ω0: The frequency of precession of the spins. It depends on the product of the applied flux density B0 and on the gyromagnetic ratio γ. The Larmor frequency is ω0 = –γB0 . Magnetization M: The macroscopic ensemble of nuclear moments. The moments are induced by an applied magnetic field. At body temperatures, the amount of magnetization is linearly proportional (M0 = 4 × 10–9H0) to the applied magnetic field. Precession: The term used to describe the motion of the magnetization about an applied magnetic field. The vector locus traverses a cone. The precession frequency is the frequency of the magnetization components perpendicular to the applied field. The precession frequency is also called the Larmor frequency ω0 . Spin echo: The transverse magnetization response to a π RF pulse. The effects of field inhomogeneity are refocused at the middle of the spin echo. Spin-lattice relaxation T1 : The exponential rate constant describing the decay of the z component of magnetization toward the equilibrium magnetization. Typical values in the body are between 300 and 3000 ms. Spin-Spin relaxation T2 : The exponential rate constant describing the decay of the transverse components of magnetization (Mx and My). Spins M: Another name for magnetization. 2D FT: A rectilinear trajectory through k-space. This popular acquisition scheme requires several (usually 128 to 256) excitations separated by a time TR, which is determined by the desired contrast, RF flip angle, and the T1 of the desired components of the image.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

Lauterbur PC. 1973. Nature 242:190. Evens RG, Evens JRG. 1991. AJR 157:603. Bloch F, Hansen WW, Packard ME. 1946. Phys Rev 70:474. Bloch F. 1946. Phys Rev 70:460. Twieg DB. 1983. Med Phys 10:610. Ljunggren S. 1983. J Magnet Reson 54:338. Bracewell RN. 1978. The Fourier Transform and Its Applications. New York, McGraw-Hill. Hoult DI, Lauterbur PC. 1979. J Magnet Reson 34:425. Chen CN, Hoult D. 1989. Biomedical Magnetic Resonance Technology. New York, Adam Hilger. Roemer PB, Edelstein WA, Hayes CE, et al. 1990. Magn Reson Med 16:192. Meyer CH, Hu BS, Nishimura DG, Macovski A. 1992. Magn Reson Med 28(2):202.

63.2 Hardware/Instrumentation John Schenck This section describes the basic components and the operating principles of MRI scanners. Although scanners capable of making diagnostic images of the human internal anatomy through the use of magnetic resonance imaging (MRI) are now ubiquitous devices in the radiology departments of hospitals in the U.S. and around the world, as recently as 1980 such scanners were available only in a handful of research institutions. Whole-body superconducting magnets became available in the early 1980s and greatly increased the clinical acceptance of this new imaging modality. Market research data indicate that between


1980 and 1996 more than 100 million clinical MRI scans were performed worldwide. By 1996 more than 20 million MRI scans were being performed each year. MRI scanners use the technique of nuclear magnetic resonance (NMR) to induce and detect a very weak radio frequency signal that is a manifestation of nuclear magnetism. The term nuclear magnetism refers to weak magnetic properties that are exhibited by some materials as a consequence of the nuclear spin that is associated with their atomic nuclei. In particular, the proton, which is the nucleus of the hydrogen atom, possesses a nonzero nuclear spin and is an excellent source of NMR signals. The human body contains enormous numbers of hydrogen atoms—especially in water (H2O) and lipid molecules. Although biologically significant NMR signals can be obtained from other chemical elements in the body, such as phosphorous and sodium, the great majority of clinical MRI studies utilize signals originating from protons that are present in the lipid and water molecules within the patient's body. The patient to be imaged must be placed in an environment in which several different magnetic fields can be simultaneously or sequentially applied to elicit the desired NMR signal. Every MRI scanner utilizes a strong static field magnet in conjunction with a sophisticated set of gradient coils and radiofrequency coils. The gradients and the radiofrequency components are switched on and off in a precisely timed pattern, or pulse sequence. Different pulse sequences are used to extract different types of data from the patient. MR images are characterized by excellent contrast between the various forms of soft tissues within the body. For patients who have no ferromagnetic foreign bodies within them, MRI scanning appears to be perfectly safe and can be repeated as often as necessary without danger [Shellock and Kanal, 1998]. This provides one of the major advantages of MRI over conventional X-ray and computed tomographic (CT) scanners. The NMR signal is not blocked at all by regions of air or bone within the body, which provides a significant advantage over ultrasound imaging. Also, unlike the case of nuclear medicine scanning, it is not necessary to add radioactive tracer materials to the patient.

Fundamentals of MRI Instrumentation Three types of magnetic fields—main fields or static fields (B2), gradient fields, an radiofrequency (RF) fields (B1)—are required in MRI scanners. In practice, it is also usually necessary to use coils or magnets that produce shimming fields to enhance the spatial uniformity of the static field Bo. Most MRI hardware engineering is concerned with producing and controlling these various forms of magnetic fields. The ability to construct NMR instruments capable of examining test tube-sized samples has been available since shortly after World War II. The special challenge associated with the design and construction of medical scanners was to develop a practical means of scaling these devices up to sizes capable of safely and comfortably accommodating an entire human patient. Instruments capable of human scanning first became available in the late 1970s. The successful implementation of MRI requires a two-way flow of information between analog and digital formats (Fig. 63.5). The main magnet, the gradient and RF coils, and the gradient and RF power supplies operate in the analog domain. The digital domain is centered on a general-purpose computer (Fig. 63.6) that is used to provide control information (signal timing and amplitude) to the gradient and RF amplifiers, to process time-domain MRI signal data returning from the receiver, and to drive image display and storage systems. The computer also provides miscellaneous control functions, such as permitting the operator to control the position of the patient table.

Static Field Magnets The main field magnet [Thomas, 1993] is required to produce an intense and highly uniform, static magnetic field over the entire region to be imaged. To be useful for imaging purposes, this field must be extremely uniform in space and constant in time. In practice, the spatial variation of he main field of a whole-body scanner must be less than about 1 to 10 parts per million (ppm) over a region approximately 40 cm in diameter. To achieve these high levels of homogeneity requires careful attention to magnet design and to manufacturing tolerances. The temporal drift of the field strength is normally required to be less than 0.1 ppm/h.


FIGURE 63.5 Digital and analog domains for MRI imaging. MRI involves the flow of data and system commands between these two domains (Courtesy of WM Leue. Reprinted with permission from Schenck and Leue, 1991).

FIGURE 63.6 Block diagram for an MRI scanner. A general-purpose computer is used to generate the commands that control the pulse sequence and to process data during MR scanning (Courtesy of WM Leue. Reprinted with permission from Schenck and Leue, 1991.)

Two units of magnetic field strength are now in common use. The gauss (G) has a long historical usage and is firmly embedded in the older scientific literature. The tesla (T) is a more recently adopted unit, but is a part of the SI system of units and, for this reason, is generally preferred. The tesla is a much larger unit than the gauss—1 T corresponds to 10,000 G. The magnitude of the earth's magnetic field is about .05 mT (5000 G). The static magnetic fields of modern MRI scanners arc most commonly in the range of from 0.5 to 1.5 T; useful scanners, however, have been built using the entire range from 0.02 to 8 T. The signal-to-noise ration (SNR) is the ratio of the NMR signal voltage to the ever-present noise voltages that arise within the patient and within the electronic components of the receiving system. The


SNR is one of the key parameters that determine the performance capabilities of a scanner. The maximum available SNR increases linearly with field strength. The improvement in SNR as the field strength is increased is the major reason that so much effort has gone into producing high-field magnets for MRI systems. Magnetic fields can be produced by using either electric currents or permanently magnetized materials as sources. In either case, the field strength falls off rapidly away from the source, and it is not possible to create a highly uniform magnetic field on the outside of a set of sources. Consequently, to produce the highly uniform field required for MRI, it is necessary to more or less surround the patient with a magnet. The main field magnet must be large enough, therefore, to effectively surround the patient; in addition, it must meet other stringent performance requirements. For these reasons, the main field magnet is the most important determinant of the cost, performance, and appearance of an MRI scanner. Four different classes of main magnets—(1) permanent magnets, (2) electromagnets, (3) resistive magnets, and (4) superconducting magnets—have been used in MRI scanners. Permanent Magnets and Electromagnets Both these magnet types use magnetized materials to produce the field that is applied to the patient. In a permanent magnet, the patient is placed in the gap between a pair of permanently magnetized pole faces. Electromagnets use a similar configuration, but the pole faces are made of soft magnetic materials, which become magnetized only when subjected to the influence of electric current coils that are wound around them. Electromagnets, but not permanent magnets, require the use of an external power supply. For both types of magnets, the magnetic circuit is completed by use of a soft iron yoke connecting the pole faces to one another (Fig. 63.7). The gap between the pole faces must be large enough to contain the patient as well as the gradient and RF coils. The permanent magnet materials available for use in MRI scanners include high-carbon iron, alloys such as Alnico, ceramics such as barium ferrite, and rare earth alloys such as samarium cobalt. Permanent magnet scanners have some advantages: They produce a relatively small fringing field and do not require power supplies. However, they tend to be very heavy (up to 100 tons) can produce only

FIGURE 63.7 Permanent magnet. The figure shows a schematic cross-section of a typical permanent magnet cinfiguration. Electromagnets have a similar construction but are energized by current-carrying coils wound around the iron yoke. Soft magnetic shims are used to enhance the homogeneity of the field. (Reprinted with permission from Schenck and Leue, 1991.)


relatively low fields—on the order of 0.3 T or less. They are also subject to temporal field drift caused by temperature changes. If the pole faces are made from an electrically conducting material, eddy currents induced in the pole faces by the pulsed gradient fields can limit performance as well. A recently introduced alloy of neodymium, boron, and iron (usually referred to as neodymium iron) has been used to make lighter-weight permanent magnet scanners. Resistive Magnets The first whole-body scanners, manufactured in the late 1970s and early 1980s, used four to six large coils of copper or aluminum wire surrounding the patient. These coils are energized by powerful (40 to 100 kW) direct-current (dc) power supplies. The electrical resistance of the coils leads to substantial joule heating, and the use of cooling water flowing through the coils is necessary to prevent overheating. The heat dissipation increases rapidly with field strength, and it is not feasible to build resistive magnets operating at fields much higher than 0.15 to 0.3 T. At present, resistive magnets are seldom used except for very low field strength (0.02 to 0.06 T) applications. Superconducting Magnets Since the early 1980s, the use of cryogenically cooled superconducting magnets [Wilson, 1983] has been the most satisfactory solution to the problem of producing the static magnet field for MRI scanners. The property of exhibiting absolutely no electrical resistance near absolute zero has been known as an exotic property of some materials since 1911. Unfortunately, the most common of these materials, such as lead, tin, and mercury, exhibit a phase change back to the normal state at relatively low magnetic field strengths and cannot be used to produce powerful magnetic fields. In the 1950s, a new class of materials (type II superconductors) was discovered. These materials retain the ability to carry loss-free electric currents in very high fields. One such material, an alloy of niobium and titanium, has been used in most of the thousands of superconducting whole-body magnets that have been constructed for use in MRI scanners (Fig. 63.8). The widely publicized discovery in 1986 of another class of materials which remain superconducting at much higher temperatures than any previously known material has not yet lead to any material capable of carrying sufficient current to be useful in MRI scanners.

FIGURE 63.8 Superconducting magnet. This figure shows a 1.5-T whole-body superconducting magnet. The nominal warm bore diameter is 1 m. The patient to be imaged, as well as the RF and gradient coils, are located within this bore. (Courtesy of General Electric Medical Systems. Reprinted with permission from Schenck and Leue, 1991.)


FIGURE 63.9 Schematic drawing of a superconducting magnet. The main magnet coils and the superconducting shim coils are maintained at liquid helium temperature. A computer-controlled table is used to advance the patient into the region of imaging. (Reprinted with permission from Schenck and Leue, 1991.)

Figure 63.9 illustrates the construction of a typical superconducting whole-body magnet. In this case, six coils of superconducting wire are connected in a series and carry an intense current—on the order of 200 A—to produce the 1.5-T magnetic field at the magnet's center. The diameter of the coils is about 1.3 m, and the total length of wire is about 65 km (40 miles). The entire length of this wire must be without any flaws—such as imperfect welds—that would interrupt the superconducting properties. If the magnet wire has no such flaws, the magnet can be operated in the persistent mode—that is, once the current is established, the terminals may be connected together, and a constant persistent current flow indefinitely so long as the temperature of the coils is maintained below the superconducting transition temperature. This temperature is about 10 K for niobium–titanium wire. The coils are kept at this low temperature by encasing them in a double-walled cryostat (analogous to a Thermos bottle) that permits them to be immersed in liquid helium at a temperature of 4.2 K. The gradual boiling of liquid helium caused by inevitable heat leaks into the cryostat requires that the helium be replaced on a regular schedule. Many magnets now make use of cryogenic refrigerators that reduce or eliminate the need for refilling the liquid helium reservoir. The temporal stability of superconducting magnets operating in the persistent mode is truly remarkable—magnets have operated for years completely disconnected from power supplies and maintained their magnetic field constant to within a few parts per million. Because of their ability to achieve very strong and stable magnetic field strengths without undue power consumption, superconducting magnets have become the most widely used source of the main magnetic fields for MRI scanners. Magnetic Field Homogeneity The necessary degree of spatial uniformity of the field can be achieved only by carefully placing the coils at specific spatial locations. It is well known that a single loop of wire will produce, on its axis, a field that is directed along the coil axis and that can be expressed as a sum of spherical harmonic fields. The first term in this sum is constant in space and represents the desired field that is completely independent


of position. The higher-order terms represent contaminating field inhomogeneities that spoil the field uniformity. More than a century ago, a two-coil magnet system—known as the Helmholtz pair—was developed which produced a much more homogeneous field at its center than is produced by a single current loop. This design is based on the mathematical finding that when two coaxial coils of the same radius are separated by a distance equal to their radius, the first nonzero contaminating term in the harmonic expansion is of the fourth order. This results in an increased region of the field homogeneity, which, although it is useful in many applications, is far too small to be useful in MRI scanners. However, the principle of eliminating low-order harmonic fields can be extended by using additional coils. This is the method now used to increase the volume of field homogeneity to values that are useful for MRI. For example, in the commonly used six-coil system, it is possible to eliminate all the error fields through the twelfth order. In practice, manufacturing tolerances and field perturbations caused b extraneous magnetic field sources—such as steel girders in the building surrounding the magnet—produce additional inhomogeneity in the imaging region. These field imperfections are reduced by the use of shimming fields. One approach—active shimming—uses additional coils (either resistive coils, superconducting coils, or some of each) which are designed to produce a magnetic field corresponding to a particular term in the spherical harmonic expansion. When the magnet is installed, the magnetic field is carefully mapped, and the currents in the shim coils are adjusted to cancel out the terms in the harmonic expansion to some prescribed high order. The alternative approach—passive shimming—utilizes small permanent magnets that are placed at the proper locations along the inner walls of the magnet bore to cancel out contaminating fields. If a large object containing magnetic materials—such as a power supply—is moved in the vicinity of superconducting magnets, it may be necessary to reset the shimming currents or magnet locations to account for the changed pattern of field inhomogeneity. Fringing Fields A large, powerful magnet produces a strong magnetic field in the region surrounding it as well as in its interior. This fringing field can produce undesirable effects such as erasing magnetic tapes (and credit cards). It is also a potential hazard to people with implanted medical devices such as cardiac pacemakers. For safety purposes, it is general practice to limit access to the region where the fringing field becomes intense. A conventional boundary for this region is the "5-gaussline," which is about 10 to 12 m from the center of an unshielded 1.5-T magnet. Magnetic shielding—in the form of iron plates (passive shielding) or external coils carrying current in the direction opposite to the main coil current (active shielding)—is frequently used to restrict the region in which the fringing field is significant.

Gradient Coils Three gradient fields, one each for the x, y, and z directions of a Cartesian coordinate system, are used to code position information into the MRI signal and to permit the imaging of thin anatomic slices [Thomas, 1993]. Along with their larger size, it is the use of these gradient coils that distinguishes MRI scanners from the conventional NMR systems such as those used in analytical chemistry. The direction of the static field, along the axis of the scanner, is conventionally taken as the z direction, and it is only the Cartesian component of the gradient field in this direction that produces a significant contribution to the resonant behavior of the nuclei. Thus, the three relevant gradient fields are Bz = GxX, Bz = Gyy, and Bz = GzZ. MRI scans are carried out by subjecting the spin system to a sequence of pulsed gradient and RF fields. Therefore, it is necessary to have three separate coils—one for each of the relevant gradient fields—each with its own power supply and under independent computer control. Ordinarily, the most practical method for constructing the gradient coils is to wind them on a cylindrical coil form that surrounds the patient and is located inside the warm bore of the magnet. The z gradient field can be produced by sets of circular coils wound around the cylinder with the current direction reversed for coils on the opposite sides of the magnet center (z = 0). To reduce deviations from a perfectly linear Bz gradient field, a spiral winding can be used with the direction of the turns reversed at z = 0 and the spacing


FIGURE 63.10 Z-gradient coil. The photograph shows a spiral coil wound on a cylindrical surface with an overwiding near the end of the coil (Courtesy of R. J. Dobberstein, General Electric Medical Systems. Reprinted with permission from Schenck and Leue, 1991.)

FIGURE 63.11 Transverse gradient coil. The photograph shows the outer coil pattern of an actively shielded transverse gradient coil. (Courtesy of R. J. Dobberstien, General Electric Medical Systems. Reprinted with permission from Schenck and Leue, 1991).

between windings decreasing away from the coil center (Fig. 63.10). A more complex current pattern is required to produce the transverse (x and y) gradients. As indicated in Fig. 63.11, transverse gradient fields are produced by windings which utilize a four-quadrant current pattern. The generation of MR images requires that a rapid sequence of time-dependent gradient fields (on all three axes) be applied to the patient. For example, the commonly used technique of spin-warp imaging [Edelstein et al., 1980] utilizes a slice-selection gradient pulse to select the spins in a thin (3 to10 mm)


slice of the patient and then applies readout and phase-encoding gradients in the two orthogonal directions to encode two-dimensional spatial information into the NMR signal. This, in turn, requires that the currents in the three gradient coils be rapidly switched by computer-controlled power supplies. The rate at which gradient currents can be switched is an important determinant of the imaging capabilities of a scanner. In typical scanners, the gradient coils have an electrical resistance of about 1Ω and an inductance of about 1 mH, and the gradient field can be switched from 0 to 10 mT/m (1 G/cm) in about 0.5 ms. The current must be switched from 0 to about 100 A in this interval, and the instantaneous voltage on the coils, L di/dt, is on the order of 200 V. The power dissipation during the switching interval is about 20 kW. In more demanding applications, such as are met in cardiac MRI, the gradient field may be as high as 4 to 5 mT/m and switched in 0.2 ms or less. In this case, the voltage required during gradient switching is more than 1 kV. In many pulse sequences, the switching duty cycle is relatively low, and coil heating is not significant. However, fast-scanning protocols use very rapidly switched gradients at a high duty cycle. This places very strong demands on the power supplies, and it is often necessary to use water cooling to prevent overheating the gradient coils.

Radiofrequency Coils Radiofrequency (RF) coils are components of every scanner and are used for two essential purposes—transmitting and receiving signals at the resonant frequency of the protons within the patient [Schenck, 1993]. The precession occurs at the Larmor frequency of the protons, which is proportional to the static magnetic field. At IT this frequency is 42.58 MHz. Thus in the range of field strengths currently used in whole-body scanners, 0.02 to 4 T, the operating frequency ranges from 0.85 to 170.3 MHz. For the commonly used 1.5-T scanners, the operating frequency is 63.86 MHz. The frequency of MRI scanners overlaps the spectral region used for radio and television broadcasting. As an example, the frequency of a 1.5-T scanner is within the frequency hand 60 to 66 MHz, which is allocated to television channel 3. Therefore, it is not surprising that the electronic components in MRI transmitter and receiver chains closely resemble corresponding components in radio and television circuitry. An important difference between MRI scanners and broadcasting systems is that the transmitting and receiving antennas of broadcast systems operate in the far field of the electromagnetic wave. These antennas are separated by many wavelengths. On the other hand, MRI systems operate in the near field, and the spatial separation of the sources and receivers is much less than a wavelength. In far-field systems, the electromagnetic energy is shared equally between the electric and magnetic components of the wave. However, in the near field of magnetic dipole sources, the field energy is almost entirely in the magnetic component of the electromagnetic wave. This difference accounts for the differing geometries that are most cost effective for broadcast and MRI antenna structures. Ideally, the RF field is perpendicular to the static field, which is in the z direction. Therefore, the RF field can be linearly polarized in either the x or y direction. However, the most efficient RF field results from quadrature excitation, which requires a coil that is capable of producing simultaneous x and y fields with a 90-degree phase shift between them. Three classes of RF coils—body coils, head coils, and surface coils—are commonly used in MRI scanners. These coils are located in the space between the patient and the gradient coils. Conducting shields just inside the gradient coils are used to prevent electromagnetic coupling between the RF coils and the rest of the scanner. Head and body coils are large enough to surround the legion being imaged and are designed to produce an RF magnetic field that is uniform across the region to be imaged. Body coils are usually constructed on cylindrical coil forms and have a large enough diameter (50 to 60 cm) to entirely surround the patient's body. Coils are designed only for head imaging (Fig. 63.12) have a smaller diameter (typically 28 cm). Surface coils are smaller coils designed to image a restricted region of the patient's anatomy. They come in a wide variety of shapes and sizes. Because they can be closely applied to the region of interest, surface coils can provide SNR advantages over head and body coils for localized regions, but because of their asymmetric design, they do not have uniform sensitivity.


FIGURE 63.12 Birdcage resonator. This is a head coil designed to perate in a 4-T scanner at 170 MHz. Quadrature excitation and receiver performance are achieved by using two adjacent ports with a 90-degree phase shift between them. (Reprinted with permission from Schenck and Leue, 1991.)

A common practice is to use separate coils for the transmitter and receiver functions. This permits the use of a large coil—such as the body coil—with a uniform excitation pattern as the transmitter and a small surface coil optimized to the anatomic region—such as the spine—being imaged. When this two-coil approach is used, it is important to provide for electronically decoupling of the two coils because they are tuned at the same frequency and will tend to have harmful mutual interactions.

Digital Data Processing A typical scan protocol calls for a sequence of tailored RF and gradient pulses with duration controlled in steps of 0.1 ms. To achieve sufficient dynamic range in control of pulse amplitudes, 12- to 16-bit digital-to-analog converters are used. The RF signal at the Larmor frequency (usually in the range from 1 to 200 MHz) is mixed with a local oscillator to produce a baseband signal which typically has a bandwidth of 16 to 32 kHz. The data-acquisition system must digitize the baseband signal at the Nyquist rate, which requires sampling the detected RF signal at a rate one digital data point every 5 to 20 ms. Again, it is necessary to provide sufficient dynamic range. Analog-to-digital converters with 16 to18 bits are used to produce the desired digitized signal data. During the data acquisition, information is acquired at a rate on the order of 800 kilobytes per second, and each image can contain up to a megabyte of digital data. The array processor (AP) is a specialized computer that is designed for the rapid performance of specific algorithms, such as the fast Fourier transform (FFT), which are used to convert the digitized time-domain data to image data. Two-dimensional images are typically displayed as 256 × 128, 256 × 256, or 512 × 512 pixel arrays. The images can be made available for viewing within about 1 sec after data acquisition. Three-dimensional imagining data, however, require more computer processing, and this results in longer delays between acquisition and display.


A brightness number, typically containing 16 bits of gray-scale information, is calculated for each pixel element of the image, and this corresponds to the signal intensity originating in each voxel of the object. To make the most effective use of the imaging information, sophisticated display techniques, such as multi-image displays, rapid sequential displays (cine loop), and three-dimensional renderings of anatomic surfaces, are frequently used. These techniques are often computationally intensive and require the use of specialized computer hardware. Interfaces to microprocessor-based workstations are frequently used to provide such additional display and analysis capabilities. MRI images are available as digital data; therefore, there is considerable utilization of local arena networks (LANs) to distribute information throughout the hospital, and long-distance digital transmission of the images can be used for purposes of teleradiology.

Current Trends in MRI At present, there is a substantial effort directed at enhancing the capabilities and cost-effectiveness of MR imagers. The activities include efforts to reduce the cost of these scanners, improve image quality, reduce scan times, and increase the number of useful clinical applications. Examples of these efforts include the development of high-field scanners, the advent of MRI-guided therapy, and the development of niche scanners that are designed for specific anatomical and clinical applications. Scanners have been developed that are dedicated to low-field imaging of the breast and other designs are dedicated to orthopedic applications such as the knees, wrists, and elbows. Perhaps the most promising incipient application of MRI is to cardiology. Scanners are now being developed to permit studies of cardiac wall motion, cardiac perfusion, and the coronary arteries in conjunction with cardiac stress testing. These scanners emphasize short magnet configurations to permit close monitoring and interaction with the patient, and high strength rapidly switched gradient fields. Conventional spin-warp images typically require several minutes to acquire. The fast spin echo (FSE) technique can reduce this to the order of 20 s, and gradient-echo techniques can reduce this time to a few seconds. The echo-planar technique (EPI) [Cohen and Weisskoff, 1991; Wehrli, 1990] requires substantially increased gradient power and receiver bandwidth but can produce images in 40 to 60 ms. Scanners with improved gradient hardware that are capable of handling higher data-acquisition rates are now available. For most of the 1980s and 1990s, the highest field strength commonly used in MRI scanners was 1.5 T. To achieve better SNRs, higher-field scanners, operating at fields up to 4 T, were studied experimentally. The need for very high-field scanners has been enhanced by the development of functional brain MRI. This technique utilizes magnetization differences between oxygenated and deoxygenated hemoglobin, and this difference is enhanced at higher field strengths. It has now become possible to construct 3- and 4-T and even 8-T [Robitaille et al., 1998], whole-body scanners of essentially the same physical size (or footprint) as conventional 1.5-T systems. Along with the rapidly increasing clinical interest in functional MRI, this is resulting in a considerable increase in the use of high-field systems. For the first decade or so after their introduction, MRI scanners were used almost entirely to provide diagnostic information. However, there is now considerable interest in systems capable of performing image-guided, invasive surgical procedures. Because MRI is capable of providing excellent soft-tissue contrast and has the potential for providing excellent positional information with submillimeter accuracy, it can be used for guiding biopsies and stereotactic surgery. The full capabilities of MRI-guided procedures can only be achieved if it is possible of provide surgical access to the patient simultaneously with the MRI scanning. This has lead to the development of new system designs, including the introduction of a scanner with a radically modified superconducting magnet system that permits the surgeon to operate at the patient's side within the scanner (Fig. 63.13) [Schenck et al., 1995; Black et al., 1997]. These systems have lead to the introduction of magnetic field-compatible surgical instruments, anesthesia stations, and patient monitoring equipment [Schenck, 1996].


FIGURE 63.13 Open magnet for MEI-guided therapy. This open-geometry superconducting magnet provides a surgion with direct patient access and the ability to interactively control the MRI scanner. This permits imaging to be performed simultaneously with surgical interventions.

Defining Terms Bandwidth: The narrow frequency range, approximately 32 kHz, over which the MRI signal is transmitted. The bandwidth is proportional to the strength of the readout gradient field. Echo-planar imaging (EPI): A pulse sequence used to produce very fast MRI scans. EPI times can be as short as 50 ms. Fast Fourier transform (FFT): A mathematical technique used to convert data sampled from the MRI signal into image data. This version of the Fourier transform can be performed with particular efficiency on modern array processors. Gradient coil: A coil designed to produce a magnetic field for which the field component B: varies linearly with position. Three gradient coils, one each for the x, y, and z directions, are required MRI. These coils are used to permit slice selection and to encode position information into the MRI signal. Larmor frequency: The rate at which the magnetic dipole moment of a particle precesses in an applied magnetic field. It is proportional to the field strength and is 42.58 MHz for protons in a 1-T magnetic field. Magnetic resonance imaging (MRI): A technique for obtaining images of the internal anatomy based on the use of nuclear magnetic resonance signals. During the 1980s, it became a major modality for medical diagnostic imaging. Nuclear magnetic resonance (NMR): A technique for observing and studying nuclear magnetism. It is based on partially aligning the nuclear spins by use of a strong, static magnetic field, stimulating these spins with a radiofrequency field oscillating at the Larmor frequency, and detecting the signal that is induced at this frequency. Nuclear Magnetism: The magnetic properties arising from the small magnetic dipole moments possessed by the atomic nuclei of some materials. This form of magnetism is much weaker than the more familiar form that originates from the magnetic dipole moments of the atomic electrons. Pixel: A single element or a two-dimensional array of image data.


Pulse sequence: A series of gradient and radiofrequency pulses used to organize the nuclear spins into a pattern that encodes desired imaging information into the NMR signal. Quadrature excitation and detection: The use of circularly polarized, rather than linearly polarized, radio frequency fields to excite an detect the NMR signal. It provides a means of reducing the required excitation power by 1/2 and increasing the signal-to-noise ration by 2. Radiofrequency (RF) coil: A coil designed to excite and/or detect NMR signals. These coils can usually be tuned to resonate at the Larmor frequency of the nucleus being studied. Spin: The property of a particle, such as an electron or nucleus, that leads to the presence of an intrinsic angular momentum and magnetic moment. Spin-warp imagining: The pulse sequence used in the most common method of MRI imaging. It uses a sequence of gradient field pulses to encode position information into the NMR signal and applies Fourier transform mathematics to this signal to calculate the image intensity value for each pixel. Static magnetic field: The field of the main magnet that is used to magnetize the spins and to drive their Larmor precession. Voxel: The volume element associated with a pixel. The voxel volume is equal to the pixel area multiplied by the slice thickness.

References Black, PMcL, Moriarty T, Alexander, E III., et al. 1997. Development and implementation of intraoperative magnetic resonance imaging and its neurosurgical applications. Neurosurgery 41:831. Cohen MS, Weisskoff RM. 1991. Ultra-fast imaging. Magn Reson Imaging 9:1. Edelstein WA, Hutchinson JMS, Johnson G, Redpath TW. 1980. Spin-warp NMR imaging and applications to human whole-body imaging. Phys Med Biol 25:751. Robitaille P-ML, Abdujalil AM, Kangarlu A, et al. 1998. Human magnetic resonance imaging at 8 T. NMR Biomed 11:263. Schenck JF, Leue WM. 1996. Instrumentation: Magnets coils and hardware. In SW Atlas (ed), Magnetic Resonance Imaging of the Brain and Spine, 2nd ed. pp 1–27. Philadelphia, Lippincott-Raven. Schenck JF. 1993. Radiofrequency coils: Types and characteristics. In MI Bronskill, P Sprawls (eds), The Physics of MRI, Medical Physics Monograph No. 21, pp 98–134. Woodbury, NY, American Institute of Physics. Schenck JF, Jolesz A, Roemer PB, et al. 1995. Superconducting open-configuration MR imaging system for image-guided therapy. Radiology 195:805. Schenck JF. 1996. The role of magnetic susceptibility in magnetic resonance imaging: magnetic field compatibility of the first and second kinds. Med Phys 23:815. Shellock FG, Kanal E. 1998. Magnetic Resonance: Bioeffects, Safety and Patient Management, 2nd ed. Philadelphia, Saunders. Thomas SR. 1993. Magnet and gradient coils: Types and characteristics. In MJ Bronskill, P Sprawls (eds), The Physics of MRI, Medical Physics Monograph No. 21, pp 56–97. Woodbury, NY, American Institute of Physics. Wehrli FW. 1990. Fast scan magnetic resonance: principles and applications. Magn Reson Q 6:165. Wilson MN. 1983. Superconducting Magnets. Oxford, Clarendon Press.

Further Information There are several journals devoted entirely to MR imaging. These include Magnetic Resonance in Medicine, JMRI—Journal of Magnetic Resonance Imaging, and NMR in Biomedicine, all three of which are published by Wiley-Liss, 605 Third Avenue, New York, NY 10158. Another journal dedicated to this field is Magnetic Resonance Imaging (Elsevier Publishing, 655 Avenue of the Americas, New York, NY 10010). The clinical aspects of MRI are covered extensively in Radiology (Radiological Society of North America, 2021 Spring


Road, Suite 600, Oak Brook, IL 60521), The American Journal of Radiology (American Roentgen Ray Society, 1891 Preston White Drive, Reston, VA 20191), as well as in several other journals devoted to the practice of radiology. There is a professional society, now known as the International Society for Magnetic Resonance in Medicine (ISMRM), devoted to the medical aspects of magnetic resonance. The main offices of this society are at 2118 Milvia, Suite 201, Berkeley, CA 94704. This society holds an annual meeting that includes displays of equipment and the presentation of approximately 2800 technical papers on new developments in the field. The annual Book of Abstracts of this meeting provides an excellent summary of current activities in the field. Similarly, the annual meeting of the Radiological Society of North America (RSNA) provides extensive coverage of MRI that is particularly strong on the clinical applications. The RSNA is located at 2021 Spring Road, Suite 600, Oak Brook, IL 60521. Several book-length accounts of MRI instrumentation and techniques are available. Biomedical Magnetic Resonance Technology (Adam Higler, Bristol, 1989) by Chen and D. I. Hoult, The Physics of MRI (Medical Physics Monograph 21, American Institute of Physics, Woodbury, NY, 1993), edited by M. J. Bronskill and P. Sprawls, and Electromagnetic Analysis and Design in Magnetic Resonance Imaging (CRC Press, Boca Raton, FL, 1998) by J.M. Jin each contain thorough accounts of instrumentation and the physical aspects of MRI. There are many books that cover the clinical aspects of MRI. Of particular interest are Magnetic Resonance Imaging, 3rd edition (Mosby, St. Louis, 1999), edited by D. D. Stark and W. G. Bradley, Jr., and Magnetic Resonance Imaging of the Brain and Spine, 2nd ed. (Lipincott-Raven, Philadelphia, 1996), edited by S. W. Atlas.

63.3 Functional MRI Kenneth K. Kwong and David A. Chesler Functional magnetic resonance imaging (fMRI), a technique that images intrinsic blood signal change with magnetic resonance (MR) imagers, has in the last 3 years become one of the most successful tools used to study blood flow and perfusion in the brain. Since changes in neuronal activity are accompanied by focal changes in cerebral blood flow (CBF), blood volume (CBV), blood oxygenation, and metabolism, these physiologic changes can be used to produce functional maps of mental operations. There are two basic but completely different techniques used in fMRI to measure CBF. The first one is a classic steady-state perfusion technique first proposed by Detre et al. [1], who suggested the use of saturation or inversion of incoming blood signal to quantify absolute blood flow [1–5]. By focusing on blood flow change and not just steady-state blood flow, Kwong et al. [6] were successful in imaging brain visual functions associated with quantitative perfusion change. There are many advantages in studying blood flow change because many common baseline artifacts associated with MRI absolute flow techniques can be subtracted out when we are interested only in changes. And one obtains adequate information in most functional neuroimaging studies with information of flow change alone. The second technique also looks at change of a blood parameter—blood oxygenation change during neuronal activity. The utility of the change of blood oxygenation characteristics was strongly evident in Turner’s work [7] with cats with induced hypoxia. Turner et al. found that with hypoxia, the MRI signal from the cats’ brains went down as the level of deoxyhemoglobin rose, a result that was an extension of an earlier study by Ogawa et al. [8,9] of the effect of deoxyhemoglobin on MRI signals in animals’ veins. Turner’s new observation was that when oxygen was restored, the cats’ brain signals climbed up and went above their baseline levels. This was the suggestion that the vascular system overcompensated by bringing more oxygen, and with more oxygen in the blood, the MRI signal would rise beyond the baseline. Based on Turner’s observation and the perfusion method suggested by Detre et al., movies of human visual cortex activation utilizing both the perfusion and blood oxygenation techniques were successfully acquired in May of 1991 (Fig. 63.14) at the Massachusetts General Hospital with a specially equipped superfast 1.5-T system known as an echo-planar imaging (EPI) MRI system [10]. fMRI results using intrinsic blood contrast were first presented in public at the Tenth Annual Meeting of the Society of Magnetic Resonance in Medicine in August of 1991 [6,11]. The visual cortex activation work was carried


FIGURE 63.14 Functional MR image demonstrating activation of the primary visual cortex (V1). Image acquired on May 9, 1991 with a blood oxygenation–sensitive MRI gradient-echo (GE) technique.

out with flickering goggles, a photic stimulation protocol employed by Belliveau et al. [12] earlier to acquire the MRI functional imaging of the visual cortex with the injection of the contrast agent gadolinium-DTPA. The use of an external contrast agent allows the study of change in blood volume. The intrinsic blood contrast technique, sensitive to blood flow and blood oxygenation, uses no external contrast material. Early model calculation showed that signal due to blood perfusion change would only be around 1% above baseline, and the signal due to blood oxygenation change also was quite small. It was quite a pleasant surprise that fMRI results turned out to be so robust and easily detectable. The blood oxygenation–sensitive MRI signal change, coined blood oxygenation level dependent (BOLD) by Ogawa et al. [8,9,13], is in general much larger than the MRI perfusion signal change during brain activation. Also, while the first intrinsic blood contrast fMRI technique was demonstrated with a superfast EPI MRI system, most centers doing fMRI today are only equipped with conventional MRI systems, which are really not capable of applying Detre’s perfusion method. Instead, the explosive growth of MR functional neuroimaging [14–33] in the last three years relies mainly on the measurement of blood oxygenation change, utilizing a MR parameter called T 2• . Both high speed echo planar (EPI) and conventional MR have now been successfully employed for functional imaging in MRI systems with magnet field strength ranging from 1.5 to 4.0 T.

Advances in Functional Brain Mapping The popularity of fMRI is based on many factors. It is safe and totally noninvasive. It can be acquired in single subjects for a scanning duration of several minutes, and it can be repeated on the same subjects as many times as necessary. The implementation of the blood oxygenation sensitive MR technique is universally available. Early neuroimaging, work focused on time-resolved MR topographic mapping of human primary visual (VI) (Figs. 63.15, 63.16), motor (MI), somatosensory (S1), and auditory (A1) cortices during task activation. Today, with BOLD technique combined with EPI, one can acquire 20 or more contiguous brain slices covering the whole head (3 × 3 mm in plane and 5 mm slice thickness) every 3 seconds for a total duration of several minutes. Conventional scanners can only acquire a couple of slices at a time. The benefits of whole-head imaging are many. Not only can researchers identify and test their hypotheses on known brain activation centers, they can also search for previous unknown or unsuspected sites. High resolution work done with EPI has a resolution of 1.5 × 1.5 mm in plane and a slice thickness of 3 mm. Higher spatial resolution has been reported in conventional 1.5-T MR systems [34]. © 2000 by CRC Press LLC

FIGURE 63.15 Movie of fMRI mapping of primary visual cortex (V1) activation during visual stimulation. Images are obliquely aligned along the calcarie fissures with the occipital pole at the bottom. Images were acquired at 3-s intervals using a blood oxygenation–sensitive MRI sequence (80 images total). A baseline image acquired during darkness (upper left) was subtracted from subsequent images. Eight of these subtraction images are displayed, chosen when the image intensities reached a steady-state signal level during darkness (OFF) and during 8-Hz photic stimulation (ON). During stimulation, local increases in signal intensity are detected in the posteromedial regions of the occipital lobes along the calcarine fissures.

Of note with Fig. 63.16 is that with blood oxygenation–sensitive MR technique, one observers an undershoot [6,15,35] in signal in V1 when the light stimulus is turned off. The physiologic mechanism underlying the undershoot is still not well understood. The data collected in the last 3 years have demonstrated that fMRI maps of the visual cortex correlate well with known retinotopic organization [24,36]. Higher visual regions such as V5/MT [37] and motorcortex organization [6,14,27,38] have been explored successfully. Preoperative planning work (Fig. 63.17) using motor stimulation [21,39,40] has helped neurosurgeons who attempt to preserve primary areas from tumors to be resected. For higher cognitive functions, several fMRI language studies have already demonstrated known language-associated regions [25,26,41,42] (Fig. 63.18). There is more detailed modeling work on the mechanism of functional brain mapping by blood-oxygenation change [43–46]. Postprocessing techniques that would help to alleviate the serious problem of motion/displacement artifacts are available [47].

Mechanism Flow-sensitive images show increased perfusion with stimulation, while blood oxygenation–sensitive images show changes consistent with an increase in venous blood oxygenation. Although the precise biophysical mechanisms responsible for the signal changes have yet to be determined, good hypotheses exist to account for our observations. Two fundamental MRI relaxation rates, T1 and T 2• , are used to describe the fMRI signal. T1 is the rate at which the nuclei approach thermal equilibrium, and perfusion change can be considered as an additional T1 change. T 2• represents the rate of the decay of MRI signal due to magnetic field inhomogeneities, and the change of T 2• is used to measure blood-oxygenation change.


FIGURE 63.16 Signal intensity changes for a region of interest (~60 mm2) within the visual cortex during darkness and during 8-Hz photic stimulation. Results using oxygenation-sensitive (top graph) and flow-sensitive (bottom graph) techniques are shown. The flow-sensitive data were collected once every 3.5 s, and the oxygenation-sensitive data were collected once every 3 s. Upon termination of photic stimulation, an undershoot in the oxygenation-sensitive signal intensity is observed.

T 2• changes reflect the interplay between changes in cerebral blood flow, volume, and oxygenation. As hemoglobin becomes deoxygenated, it becomes more paramagnetic than the surrounding tissue [48] and thus creates a magnetically inhomogeneous environment. The observed increased signal on T 2• -weighted images during activation reflects a decrease in deoxyhemoglobin content, i.e., an increase in venous blood oxygenation. Oxygen delivery, cerebral blood flow, and cerebral blood volume all increase with neuronal activation. Because CBF (and hence oxygen-delivery) changes exceed CBV changes by 2 to 4 times [49], while blood-oxygen extraction increases only slightly [50,51], the total paramagnetic blood deoxyhemoglobin content within brain tissue voxels will decrease with brain activation. The resulting decrease in the tissue-blood magnetic susceptibility difference leads to less intravoxel dephasing within brain tissue voxels and hence increased signal on T 2• -weighted images [6,14,15,17]. These results independently confirm PET observations that activation-induced changes in blood flow and volume are accompanied by little or no increases in tissue oxygen consumption [50,51,52]. Since the effect of volume susceptibility difference ∆χ is more pronounced at high field strength [53], higher-field imaging magnets [17] will increase the observed T 2• changes. Signal changes can also be observed on T1-weighted MR images. The relationship between T1 and regional blood flow was characterized by Detre et al. [1]:


FIGURE 63.17 Functional MRI mapping of motor cortex for preoperative planning. This three-dimensional rendering of the brain represents fusion of functional and structural anatomy. Brain is viewed from the top. A tumor is shown in the left hemisphere, near the midline. The other areas depict sites of functional activation during movement of the right hand, right foot, and left foot. The right foot cortical representation is displaced by tumor mass effect from its usual location. (Courtesy of Dr. Brad Buchbinder.)

dM M 0 − M f = + fM b − M dt T1 λ

(63.14)

where M is tissue magnetization and Mb is incoming blood signal. M0 is proton density, f is the flow in ml/gm/unit time, and λ is the brain-blood partition coefficient of water (~0.95 ml/gm). From this equation, the brain tissue magnetization M relaxes with an apparent T1 time constant T1app given by

f 1 1 = − λ T1 app T1

(63.15)

where the T1app is the observed (apparent) longitudinal relaxation time with flow effects included. T1 is the true tissue longitudinal relaxation time in the absence of flow. If we assume that the true tissue T1 remains constant with stimulation, a change in blood flow ∆f will lead to a change in the observed T1app :

∆

1 f =∆ λ T1 app

Thus the MRI signal change can be used to estimate the change in blood flow. © 2000 by CRC Press LLC

(63.16)

FIGURE 63.18 Left hemisphere surface rendering of functional data (EPI, gradient-echo, 10 oblique coronal slices extending to posterior sylvian fissure) and high-resolution anatomic image obtained on a subject (age 33 years) during performance of a same-different (visual matching) task of pairs of words or nonwords (false font strings). Foci of greatest activation for this study are located in dominant perisylvian cortex, i.e., inferior frontal gyrus (Broca’s area), superior temporal gyrus (Wernicke’s area), and inferior parietal lobule (angular gyrus). Also active in this task are sensorimotor cortex and prefrontal cortex. The perisylvian sites of activation are known to be key nodes in a left hemisphere language network. Prefrontal cortex probably plays a more general, modulatory role in attentional aspects of the task. Sensorimotor activation is observed in most language studies despite the absence of overt vocalization. (Courtesy of Dr. Randall Benson.)

From Eq. (63.14), if the magnetization of blood and tissue always undergoes a similar T1 relaxation, the flow effect would be minimized. This is a condition that can be approximated by using a flownonsensitive T1 technique inverting all the blood coming into the imaged slice of interest. This flownonsensitive sequence can be subtracted from a flow-sensitive T1 technique to provide an index of CBF without the need of external stimulation [54,55] (Fig. 63.19). Initial results with tumor patients show that such flow-mapping techniques are useful for mapping out blood flow of tumor regions [55]. Other flow techniques under investigation include the continuous inversion of incoming blood at the carotid level [1] or the use of a single inversion pulse at the carotid level (EPIstar) inverting the incoming blood [56,57]. Compared with the flow-nonsensitive and flow-sensitive methods, the blood-tagging techniques at the carotid level are basically similar concepts except that the MRI signal of tagged blood is expected to be smaller by a factor that depends on the time it takes blood to travel from the tagged site to the imaged slice of interest [55]. The continuous-inversion technique also has a significant problem of magnetization transfer [1] that contaminates the flow signal with a magnetization transfer signal that is several times larger. On the other hand, the advantage of the continuous inversion is that it can under optimal conditions provide a flow contrast larger than all the other methods by a factor of e [55].

Problem and Artifacts in fMRI: The Brain-Vein Problem? The Brain-Inflow Problem? The artifacts arising from large vessels pose serious problems to the interpretation of oxygenation sensitive fMRI data. It is generally believed that microvascular changes are specific to the underlying region of © 2000 by CRC Press LLC

FIGURE 63.19 Functional MRI cerebral blood flow (CBF) index (right) of a low-flow brain tumor (dark region right of the midline) generated by the subtraction of a flow-nonsensitive image from a flow-sensitive image. This low-flow region matches well with a cerebral blood volume (CBV) map (left) of the tumor region generated by the injection of a bolus of MRI contrast agent Gd-DTPA, a completely different and established method to measure hemodynamics with MRI.

neuronal activation. However, MRI gradient echo (GE) is sensitive to vessels of all dimensions [46,58], and there is concern that macrovascular changes distal to the site of neuronal activity can be induced [20]. This has been known as the brain-vein problem. For laboratories not equipped with EPI, gradient echo (GE) sensitive to variations in T 2• and magnetic susceptibility are the only realistic sequences available for fMRI acquisition, so the problem is particularly acute. In addition, there is a non-deoxyhemoglobin-related problem, especially acute in conventional MRI. This is the inflow problem of fresh blood that can be time-locked to stimulation [28,29,59]. Such nonparenchymal and macrovascular responses can introduce error in the estimate of activated volumes.

Techniques to Reduce the Large Vessel Problems In dealing with the inflow problems, EPI has special advantages over conventional scanners. The use of long repetition times (2 to 3 s) in EPI significantly reduces the brain-inflow problem. Small-flip-angle methods in conventional MRI scanners can be used to reduce inflow effect [59]. Based on inflow modeling, one observes that at an angle smaller than the Ernst angle [60], the inflow effect drops much faster than the tissue signal response to activation. Thus one can effectively remove the inflow artifacts with small-flip-angle techniques. A new exciting possibility is to add small additional velocity-dephasing gradients to suppress slow inplane vessel flow [60,61]. Basically, moving spins lose signals, while stationary spins are unaffected. The addition of these velocity-dephasing gradients drops the overall MRI signal (Fig. 63.20). The hypothesis that large vessel signals are suppressed while tissue signals remain intact is a subject of ongoing research. Another advantage with EPI is that another oxygenation-sensitive method such as the EPI T2-weighted spin-echo (T2SE) is also available. T2SE methods are sensitive to the MRI parameter T2, which is affected by microscopic susceptibility and hence blood oxygenation. Theoretically, T2SE methods are far less sensitive to large vessel signals [1,6,46,58]. For conventional scanners, T2SE methods take too long to perform and therefore are not practical options. The flow model [1] based on T1-weighted sequences and independent of deoxyhemoglobin is also not so prone to large vessel artifacts, since the T1 model is a model of perfusion at the tissue level. Based on the study of volunteers, the average T 2• -weighted GE signal percentage change at V1 was 2.5 ± 0.8%. The average oxygenation-weighted T2SE signal percentage change was 0.7 ± 0.3%. The © 2000 by CRC Press LLC

FIGURE 63.20 The curves represent time courses of MRI response to photic stimulation (off-on-off-on L) with different levels of velocity-dephasing gradients turned on to remove MRI signals coming from the flowing blood of large vessels. The top curve had no velocity-dephasing gradients turned on. The bottom curve was obtained with such strong velocity-dephasing gradients turned on that all large vessel signals were supposed to have been eliminated. The middle curve represents a moderate amount of velocity-dephasing gradients, a tradeoff between removing large vessel signals and retaining a reasonable amount of MRI signal to noise.

average perfusion-weighted and T1-weighted MRI signal percentage change was 1.5 ± 0.5%. These results demonstrate that T2SE and T1 methods, despite their ability to suppress large vessels, are not competitive with T 2• effect at 1.5 T. However, since the microscopic effect detected by T2SE scales up with field strength [62], we expect the T2SE to be a useful sequence at high field strength such as 3 or 4 T. Advancing field strength also should benefit T1 studies due to better signal-to-noise and to the fact that T1 gets longer at higher field strength. While gradient-echo sequence has a certain ambiguity when it comes to tissue versus vessels, its sensitivity at current clinical field strength makes it an extremely attractive technique to identify activation sites. By using careful paradigms that rule out possible links between the primary activation site and secondary sites, one can circumvent many of the worries of “signal from the primary site draining down to secondary sites.” A good example is as follows: Photic stimulation activates both the primary visual cortex and the extrastriates. To show that the extrastriates are not just a drainage from the primary cortex, one can utilize paradigms that activate the primary visual cortex but not the extrastriate, and vice versa. There are many permutations of this [37]. This allows us to study the higher-order functions umambiguously even if we are using gradient-echo sequences. The continuous advance of MRI mapping techniques utilizing intrinsic blood-tissue contrast promises the development of a functional human neuroanatomy of unprecedented spatial and temporal resolution.

References 1. 2. 3. 4. 5.

Detre J, Leigh J, Williams D, Koretsky A. 1992. Magn Reson Med 23:37. Williams DS, Detre JA, Leigh JS, Koretsky AP. 1992. Proc Natl Acad Sci USA 89:212. Zhang W, Williams DS, Detre JA. 1992. Magn Reson Med 25:362. Zhang W, Williams DS, Koretsky AP. 1993. Magn Reson Med 29:416. Dixon WT, Du LN, Faul D, et al. 1986. Magn Reson Med 3:454.


6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50.

Kwong KK, Belliveau JW, Chesler DA, et al. 1992. Proc Natl Acad Sci USA 89:5675. Turner R, Le Bihan D, Moonen CT, et al. 1991. Magn Reson Med 22:159. Ogawa S, Lee TM, Kay AR, Tank DW. 1990. Proc Natl Acad Sci USA 87:9868. Ogawa S, Lee TM. 1990. Magn Reson Med 16:9. Cohen MS, Weisskoff RM. 1991. Magn Reson Imaging 9:1. Brady TJ, Society of Magnetic Resonance in Medicine, San Francisco, CA 2, 1991. Belliveau JW, Kennedy DN Jr, McKinstry RC, et al. 1991. Science 254:716. Ogawa S, Lee TM, Nayak AS, Glynn P. 1990. Magn Reson Med 14:68. Bandettini PA, Wong EC, Hinks RS, et al. 1992. Magn Reson Med 25:390. Ogawa S, Tank DW, Menon R, et al. 1992. Proc Natl Acad Sci USA 89:5951. Frahm J, Bruhn H, Merboldt K, Hanicke W. 1992. J Magn Reson Imaging 2:501. Turner R, Jezzard P, Wen H, et al. 1992. Society of Magnetic Resonance in Medicine Eleventh Annual Meeting, Berlin. Blamire A, Ogawa S, Ugurbil K, et al. 1992. Proc Natl Acad Sci USA 89:11069. Menon R, Ogawa S, Tank D, Ugurbil K. 1993. Magn Reson Med 30:380. Lai S, Hopkins A, Haacke E, et al. 1993. Magn Reson Med 30:387. Cao Y, Towle VL, Levin DN, et al. 1993. Society of Magnetic Resonance in Medicine Meeting. Connelly A, Jackson GD, Frackowiak RSJ, et al. 1993. Radiology 125. Kim SG, Ashe J, Georgopouplos AP, et al. 1993. J Neurophys 69:297. Schneider W, Noll DC, Cohen JD. 1993. Nature 365:150. Hinke RM, Hu X, Stillman AE, et al. 1993. Neurol Rep 4:675. Binder JR, Rao SM, Hammeke TA, et al. 1993. Neurology (suppl 2):189. Rao SM, Binder JR, Bandettini PA, et al. 1993. Neurology 43:2311. Gomiscek G, Beisteiner R, Hittmair K, et al. 1993. MAGMA 1:109. Duyn J, Moonen C, de Boer R, et al. 1993. Society of Magnetic Resonance in Medicine, 12th Annual Meeting, New York, New York. Hajnal JV, Collins AG, White SJ, et al. 1993. Magn Reson Med 30:650. Hennig J, Ernst T, Speck O, et al. 1994. Magn Reson Med 31:85. Constable RT, Kennan RP, Puce A, et al. 1994. Magn Res Med 31:686. Binder JR, Rao SM, Hammeke TA, et al. 1994. Ann Neurol 35:662. Frahm J, Merboldt K, Hänicke W. 1993. Magn Reson Med 29:139. Stern CE, Kwong KK, Belliveau JW, et al. 1992. Society of Magnetic Resonance in Medicine Annual Meeting, Berlin, Germany. Belliveau JW, Kwong KK, Baker JR, et al. 1992. Society of Magnetic Resonance in Medicine Annual Meeting, Berlin, Germany. Tootell RBH, Kwong KK, Belliveau JW, et al. 1993. Investigative Ophthalmology and Visual Science, p 813. Kim S-G, Ashe J, Hendrich K, et al. 1993. Science 261:615. Buchbinder BR, Jiang HJ, Cosgrove GR, et al. 1994. ASNR 162. Jack CR, Thompson RM, Butts RK, et al. 1994. Radiology 190:85. Benson RR, Kwong KK, Belliveau JW, et al. 1993. Soc Neurosci. Benson RR, Kwong KK, Buchbinder BR, et al. 1994. Society of Magnetic Resonance, San Francisco. Ogawa S, Menon R, Tank D, et al. 1993. Biophys J 64:803. Ogawa S, Lee TM, Barrere B. 1993. Magn Reson Med 29:205. Kennan RP, Zhong J, Gore JC. 1994. Magn Reson Med 31:9. Weisskoff RM, Zuo CS, Boxerman JL, Rosen BR. 1994. Magn Res Med 31:601. Bandettini PA, Jesmanowicz A, Wong EC, Hyde JS. 1993. Magn Reson Med 30:161. Thulborn KR, Waterton JC, Matthews PM, Radda GK. 1982. Biochim Biophys Acta 714:265. Grubb RL, Raichle ME, Eichling JO, Ter-Pogossian MM. 1974. Stroke 5:630. Fox PT, Raichle ME, 1986. Proc Natl Acad Sci USA 83:1140.


51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62.

Fox PT, Raichle ME, Mintun MA, Dence C. 1988. Science 241:462. Prichard J, Rothman D, Novotny E, et al. 1991. Proc Natl Acad Sci USA 88:5829. Brooks RA, Di Chiro G. 1987. Med Phys 14:903. Kwong K, Chesler D, Zuo C, et al. 1993. Society of Magnetic Resonance in Medicine, 12th Annual Meeting, New York, New York, p 172. Kwong KK, Chesler DA, Weisskoff RM, Rosen BR. 1994. Society of Magnetic Resonance, San Francisco. Edelman R, Sievert B, Wielopolski P, et al. 1994. JMRI 4(P). Warach S, Sievert B, Darby D, et al. 1994. JMRI 4(P):S8. Fisel CR, Ackerman JL, Buxton RB, et al. 1991. Magn Reson Med 17:336. Frahm J, Merboldt K, Hanicke W. 1993. Society of Magnetic Resonance in Medicine, 12th Annual Meeting, New York, New York, p 1427. Kwong KK, Chesler DA, Boxerman JL, et al. 1994. Society of Magnetic Resonance, San Francisco. Song W, Bandettini P, Wong E, Hyde J. 1994. Personal communication. Zuo C, Boxerman J, Weisskoff R. 1992. Society of Magnetic Resonance in Medicine, 11th Annual Meeting, Berlin, p 866.

63.4 Chemical-Shift Imaging: An Introduction to Its Theory and Practice Xiaoping Hu, Wei Chen, Maqbool Patel, and Kamil Ugurbil Over the past two decades, there has been a great deal of development in the application of nuclear magnetic resonance (NMR) to biomedical research and clinical medicine. Along with the development of magnetic resonance imaging [1], in vivo magnetic resonance spectroscopy (MRS) is becoming a research tool for biochemical studies of humans as well as a potentially more specific diagnostic tool, since it provides specific information on individual chemical species in living systems. Experimental studies in animals and humans have demonstrated that MRS can be used to study the biochemical basis of disease and to follow the treatment of disease. Since biologic subjects (e.g., humans) are heterogeneous, it is necessary to spatially localize the spectroscopic signals to a well-defined volume or region of interest (VOI or ROI, respectively) in the intact body. Toward this goal, various localization techniques have been developed (see ref. [2] for a recent review). Among these techniques, chemical-shift imaging (CSI) or spectroscopic imaging [3–6] is an attractive technique, since it is capable of producing images reflecting the spatial distribution of various chemical species of interest. Since the initial development of CSI in 1982 [3], further developments have been made to provide better spatial localization and sensitivity, and the technique has been applied to numerous biomedical problems. In this section we will first present a qualitative description of the basic principles of chemical-shift imaging and subsequently present some practical examples to illustrate the technique. Finally, a summary is provided in the last subsection.

General Methodology In an NMR experiment, the subject is placed in a static magnetic field B0. Under the equilibrium condition, nuclear spins with nonzero magnetic moment are aligned along B0, giving rise to an induced bulk magnetization. To observe the bulk magnetization, it is tipped to a direction perpendicular to B0 (transverse plane) with a radiofrequency (RF) pulse that has a frequency corresponding to the resonance frequency of the nuclei. The resonance frequency is determined by the product of the gyromagnetic ratio of the nucleus γ and the strength of the static field, i.e., γB0 , and is called the Larmor frequency. The Larmor frequency also depends on the chemical environment of the nuclei, and this dependency gives


rise to chemical shifts that allow one to identify different chemical species in an NMR spectrum. Upon excitation, the magnetization in the transverse plane (perpendicular to the main B0 field direction) oscillates with the Larmor frequencies of all the different chemical species and induces a signal in a receiving RF coil; the signal is also termed the free induction decay (FID). The FID can be Fourier transformed with respect to time to produce a spectrum in frequency domain. In order to localize an NMR signal from an intact subject, spatially selective excitation and/or spatial encoding are usually utilized. Selective excitation is achieved as follows: In the excitation, an RF pulse with a finite bandwidth is applied in the presence of a linear static magnetic field gradient. With the application of the gradient, the Larmor frequency of spins depends linearly on the spatial location along the direction of the gradient. Consequently, only the spins in a slice whose resonance frequency falls into the bandwidth of the RF pulse are excited. The RF excitation rotates all or a portion of the magnetization to the transverse plane, which can be detected by a receiving RF coil. Without spatial encoding, the signal detected is the integral of the signals over the entire excited volume. In CSI based on Fourier imaging, spatial discrimination is achieved by phase encoding. Phase encoding is accomplished by applying a gradient pulse after the excitation and before the data acquisition. During the gradient pulse, spins precess at Larmor frequencies that vary linearly along the direction of the gradient and accrue a phase proportional to the position along the phase-encoding gradient as well as the strength and the duration of the gradient pulse. This acquired → → → → → spatially encoded phase is typically expressed as k · r = ∫ γ g(t)· r dt, where γ is the gyromagnetic ratio; r → is the vector designating spatial location; g(t) defines the magnitude, the direction, and the time dependence of the magnetic field gradient applied during the phase-encoding; and the integration is performed over time when the phase-encoding gradient is on. Thus, in one-dimensional phase encoding, if the phase encoding is along, for example, the y axis, the phase acquired becomes k × y = ∫ γgy(t) × y dt. The acquired signal S(t) is the integral of the spatially distributed signals modulated by a spatially dependent phase, given by the equation

() ∫ ( )

r r ik ⋅ r S t = ρ r , t e ( )d 3r r

(63.17)

where ρ is a function that describes the spatial density and the time evolution of the transverse magnetization of all the chemical species in the sample. This signal mathematically corresponds to a sample of the Fourier transform along the direction of the gradient. The excitation and detection process is repeated with various phase-encoding gradients to obtain many phase-encoded signals that can be inversely Fourier-transformed to resolve an equal number of pixels along this direction. Taking the example of one-dimensional phase-encoding along the y axis to obtain a one-dimensional image along this direction of n pixels, the phase encoding gradient is incremented n times so that n FIDs are acquired, each of which is described as

( ) ∫ ( )

S t , n = ρ * y, t e(

ink0 y

) dy

(63.18)

where ρ* is already integrated over the x and z directions, and k0 is the phase-encoding increment; the latter is decided on using the criteria that the full field of view undergo a 360-degree phase difference when n = 1, as dictated by the sampling theorem. The time required for each repetition (TR), which is dictated by the longitudinal relaxation time, is usually on the order of seconds. In CSI, phase encoding is applied in one, two, or three dimensions to provide spatial localization. Meanwhile, selective excitation also can be utilized in one or more dimensions to restrict the volume to be resolved with the phase encodings. For example, with selective excitation in two dimensions, CSI in one spatial dimension can resolve voxels within the selected column. In multidimensional CSI, all the phase-encoding steps along one dimension need to be repeated for all the steps along the others. Thus,


for three dimensions with M, N, and L number of phase encoding steps, one must acquire M × N × L number of FIDS:

(

) ∫( )

r i mkx x +nky0 y +lkz0 z ) 3 r S t , m, n, l = ρ r , t e ( 0 dr

(63.19)

where m, n, and l must step through M, N, and L in integer steps, respectively. As a result, the time needed for acquiring a chemical-shift image is proportional to the number of pixels desired and may be very long. In practice, due to the time limitation as well as the signal-to-noise ratio (SNR) limitation, chemical-shift imaging is usually performed with relatively few spatial encoding steps, such as 16 × 16 or 32 × 32 in a two-dimensional experiment. The data acquired with the CSI sequence need to be properly processed before the metabolite information can be visualized and quantitated. The processing consists of spatial reconstruction and spectral processing. Spatial reconstruction is achieved by performing discrete inverse Fourier transformation, for each of the spatial dimensions, with respect to the phase-encoding steps. The spatial Fourier transform is applied for all the points of the acquired FID. For example, for a data set from a CSI in two spatial dimensions with 32 × 32 phase-encoding steps and 1024 sampled data points for each FID, a 32 × 32 two-dimensional inverse Fourier transform is applied to each of the 1024 data points. Although the nominal spatial resolution achieved by the spatial reconstruction is determined by the number of phaseencoding steps and the field of view (FOV), it is important to note that due to the limited number of phase-encoding steps used in most CSI experiments, the spatial resolution is severely degraded by the truncation artifacts, which results in signal “bleeding” between pixels. Various methods have been developed to reduce this problem [7–14]. The localized FIDs derived from the spatial reconstruction are to be further processed by spectral analysis. Standard procedures include Fourier transformation, filtering, zero-filling, and phasing. The localized spectra can be subsequently presented for visualization or further processed to produce quantitative metabolite information. The presentation of the localized spectra in CSI is not a straightforward task because there can be thousands of spectra. In one-dimensional experiments, localized spectra are usually presented in a stack plot. In two-dimensional experiments, localized spectra are plotted in small boxes representing the extent of the pixels, and the plots can be overlaid on corresponding anatomic image for reference. Spectra from three-dimensional CSI experiments are usually presented slice by slice, each displaying the spectra as in the two-dimensional case. To derive metabolite maps, peaks corresponding to the metabolites of interest need to be quantified. In principle, the peaks can be quantified using the standard methods developed for spectral quantification [15–17]. The most straightforward technique is to calculate the peak areas by integrating the spectra over the peak of interest if it does not overlap with other peaks significantly. In integrating all the localized spectra, spectral shift due to B0 inhomogeneity should be taken into account. A more robust approach is to apply spectral fitting programs to each spectrum to obtain various parameters of each peak. The fitted area for the peak of interest can then be used to represent the metabolite signal. The peak areas are then used to generate metabolite maps, which are images with intensities proportional to the localized peak area. The metabolite map can be displayed by itself as a gray-scale image or color-coded image or overlaid on a reference anatomic image.

Practical Examples To illustrate the practical utility of CSI, we present two representative CSI studies in this section. The sequence for the first study is shown in Fig. 63.21. This is a three-dimensional sequence in which phase encoding is applied in all three directions and no slice selection is used. Such a sequence is usually used with a surface RF coil whose spatial extent of sensitivity defines the field of view. In this sequence, the FID is acquired immediately after the application of the phase-encoding gradient to minimize the decay


FIGURE 63.21 RF pulse.

Sequence diagram for a three-dimensional chemical shift imaging sequence using a nonselective

of the transverse magnetization, and the sequence is suitable for imaging metabolites with short transverse relaxation time (e.g., ATP). With the sequence shown in Fig. 63.21, a phosphorus-31 CSI study of the human brain was conducted using a quadrature surface coil. A nonselective RF pulse with an Ernest angle (40 degrees) optimized for the repetition time was used for the excitation. Phase-encoding gradients were applied for a duration of 500 µs; the phase-encoding gradients were incremented according to a FOV of 25 × 25 × 20 cm3. Phaseencoded FIDs were acquired with 1024 complex data points over a sampling window of 204.8 ms; the corresponding spectral width was 5000 Hz. To reduce intervoxel signal contamination, a technique that utilizes variable data averaging to introduce spatial filtering during the data acquisition for optimal signalto-noise ratio is employed [7–10], resulting in spherical voxels with diameter of 3 cm (15 cc volume). The data were acquired with a TR of 1 s, and the total acquisition time was approximately 28 minutes. The acquired data were processed to generate three-dimensional voxels, each containing a localized phosphorus spectrum, in a 17 × 13 × 17 matrix. In Fig. 63.22a–c, spectra in three slices of the threedimensional CSI are presented; these spectra are overlaid on the corresponding anatomic images obtained with a T1-weighted imaging sequence. One representative spectrum of the brain is illustrated in Fig. 63.22d, where the peaks corresponding to various metabolites are labeled. It is evident that the localized phosphorus spectra contain a wealth of information about several metabolites of interest, including adenosine triphosphate (ATP), phosphocreatine (PCr), phosphomonoester (PME), inorganic phosphate (Pi), and phosphodiester (PDE). In pathologic cases, focal abnormalities in phosphorus metabolites have been detected in patients with tumor, epilepsy, and other diseases [18–25]. The second study described below is performed with the sequence depicted in Fig. 63.23. This is a two-dimensional spin-echo sequence in which a slice is selectively excited by a 90-degree excitation pulse. The 180-degree refocusing pulse is selective with a slightly broader slice profile. Here the phase-encoding gradients are applied before the refocusing pulse; they also can be placed after the 180-degree pulse or split to both sides of the 180-degree pulse. This sequence was used for a proton CSI experiment. In proton CSI, a major problem arises from the strong water signal that overwhelms that of the metabolites. In order to suppress the water signal, many techniques have been devised [26–29]. In this study, a three-pulse


(A)

(B)

FIGURE 63.22 (A–C) Boxed plot of spectra in three slices from the three-dimensional 31P CSI experiment overlaid on corresponding anatomic images. The spectral extent displayed is from 10 to –20 ppm. A 20-Hz line broadening is applied to all the spectra. (D) Representative spectrum from the three-dimensional 31P CSI shown in (B). Metabolite peaks are labeled.

CHESS [26] technique was applied before the application of the excitation pulse as shown in Fig. 63.23. The CSI experiment was performed on a 1.4-cm slice with 32 × 32 phase encodings over a 22 × 22 cm2 FOV. The second half of the spin-echo was acquired with 512 complex data points over a sampling window of 256 ms, corresponding to a spectral width of 2000 Hz. Each phase-encoding FID was acquired twice for data averaging. The repetition time (TR) and the echo time (TE) used were 1.2 s and 136 ms, respectively. The total acquisition time was approximately 40 minutes.



Another major problem in proton CSI study of the brain is that the signal from the subcutaneous lipid usually is much stronger than those of the metabolites, and this strong signal leaks into pixels within the brain due to truncation artifacts. To avoid lipid signal contamination, many proton CSI studies of the brain are performed within a selected region of interest excluding the subcutaneous fat [30–34]. Recently, several techniques have been proposed to suppress the lipid signal and consequently suppress the lipid signal contamination. These include the use of WEFT [27] and the use of outer-volume signal suppression [34]. In the example described below, we used a technique that utilizes the spatial location of the lipid to extrapolate data in the k-space to reduce the signal contamination due to truncation [35]. In Fig. 63.24, the results from the proton CSI study are presented. In panel (a), the localized spectra are displayed. Note that the spectra in the subcutaneous lipid are ignored because they are all off the scale. The nominal spatial resolution is approximately 0.66 cc. A spectrum from an individual pixel in this study is presented in Fig. 63.24b with metabolite peaks indicated. Several metabolite peaks, such as those corresponding to the N-acetyl aspartate (NAA), creatine/phosphocreatine (Cr/PCr), and choline


FIGURE 63.23 proton study.

A two-dimensional spin-echo CSI sequence with chemical selective water suppression (CHESS) for

(Cho), are readily identified. In addition, there is still a noticeable amount of residual lipid signal contamination despite the use of the data extrapolation technique. Without the lipid suppression technique, the brain spectra would be severely contaminated by the lipid signal, making the detection of the metabolite peaks formidable. The peak of NAA in these spectra is fitted to generate the metabolite map shown in panel (c). Although the metabolite map is not corrected for coil sensitivity and other factors and only provides a relative measure of the metabolite concentration in the brain, it is a reasonable measure of the NAA distribution in the brain slice. The spatial resolution of the CSI study can be appreciated from the brain structure present in the map. In biomedical research, proton CSI is potentially the most promising technique, since it provides best sensitivity and spatial resolution. Various in vivo applications of proton spectroscopy can be found in the literature [36].

Summary CSI is a technique for generating localized spectra that provide a wealth of biochemical information that can be used to study the metabolic activity of living system and to detect disease associated biochemical changes. This section provides an introduction to the technique and illustrates it by two representative examples. More specific topics concerning various aspects of CSI can be found in the literature.

Acknowledgments The authors would like to thank Dr. Xiao-Hong Zhu for assisting data acquisition and Mr. Gregory Adriany for hardware support. The studies presented here are supported by the National Institutes of Health (RR08079).


(a)

FIGURE 63.24 (a) Boxed plot of spectra for the two-dimensional proton study overlaid on the anatomic image. A spectral range of 1.7 to 3.5 ppm is used in the plot to show Cho, PCr/Cr, and NAA. A 5-Hz line broadening is applied in the spectral processing. (b) A representative spectrum from the two-dimensional CSI in panel (a). Peaks corresponding to Cho, PCr/Cr, and NAA are indicated. (c) A map of the area under the NAA peak obtained by spectral fitting. The anatomic image is presented along with the metabolite map for reference. The spatial resolution of the metabolite image can be appreciated from the similarities between the two images. The lipid suppression technique has successfully eliminated the signal contamination from the lipid in the skull.


(c) FIGURE 63.24 (continued)

References 1. Lauterbur PC. 1973. Image formation by induced local interactions: Examples employing nuclear magnetic resonance. Nature 242:190. 2. Alger JR. 1994. Spatial localization for in vivo magnetic resonance spectroscopy: Concepts and commentary. In RJ Gillies (ed), NMR in Physiology and Biomedicine, pp 151–168. San Diego, Calif, Academic Press. 3. Brown TR, Kincaid MB, Ugurbil K. 1982. NMR chemical shift imaging in three dimensions. Proc Natl Acad Sci USA 79:3523. 4. Maudsley AA, Hilal SK, Simon HE, Perman WH. 1983. Spatially resolved high resolution spectroscopy by “four dimensional” NMR. J Magn Reson 51:147. 5. Haselgrove JC, Subramanian VH, Leigh JS Jr, et al. 1983. In vivo one-dimensional imaging of phosphorous metabolites by phosphorus-31 nuclear magnetic resonance. Science 220:1170. 6. Maudsley AA, Hilal SK, Simon HE, Wittekoek S. 1984. In vivo MR spectroscopic imaging with P-31. Radiology 153:745. 7. Garwood M, Schleich T, Ross BD, et al. 1985. A modified rotating frame experiment based on a Fourier window function: Application to in vivo spatially localized NMR spectroscopy. J Mag Reson 65:239. 8. Garwood M, Robitalle PM, Ugurbil K. 1987. Fourier series windows on and off resonance using multiple coils and longitudinal modulation. J Magn Reson 75:244. 9. Mareci TH, Brooker HR. 1984. High-resolution magnetic resonance spectra from a sensitive region defined with pulsed gradients. J Magn Reson 57:157. 10. Brooker HR, Mareci TH, Mao JT. 1987. Selective Fourier transform localization. Magn Reson Med 5:417. 11. Hu X, Levin DN, Lauterbur PC. Spraggins TA. 1988. SLIM: Spectral localization by imaging. Magn Reson Med 8:314. 12. Liang ZP, Lauterbur PC. 1991. A generalized series approach to MR spectroscopic imaging. IEEE Trans Med Imag MI-10:132. 13. Hu X, Stillman AE. 1991. Technique for reduction of truncation artifact in chemical shift images. IEEE Trans Med Imag MI-10(3):290. 14. Hu X, Patel MS, Ugurbil K. 1993. A new strategy for chemical shift imaging. J Magn Reson B103:30.


15. van den Boogaart A, Ala-Korpela M, Jokisaari J, Griffiths JR. 1994. Time and frequency domain analysis of NMR data compared: An application to 1D 1H spectra of lipoproteins. Magn Reson Med 31:347. 16. Ernst T, Kreis R, Ross B. 1993. Absolute quantification of water and metabolites in human brain: I. Compartments and water. J Magn Reson 102:1. 17. Kreis R, Ernst T, Ross B. 1993. Absolute quantification of water and metabolites in human brain. II. Metabolite concentration. J Magn Reson 102:9. 18. Lenkinski RE, Holland GA, Allman T, et al. 1988. Integrated MR imaging and spectroscopy with chemical shift imaging of P-31 at 1.5 T: Initial clinical experience. Radiology 169:201. 19. Hugg JW, Matson GB, Twieg DB, et al. 31P MR spectroscopic imaging of normal and pathological human brains. Magn Reson Imaging 10:227. 20. Vigneron DB, Nelson SJ, Murphy-Boesch J, et al. 1990. Chemical shift imaging of human brain: Axial, sagittal, and coronal 31P metabolite images. Radiology 177:643. 21. Hugg JW, Laxer KD, Matson GB, et al. 1992. Lateralization of human focal epilepsy by 31P magnetic resonance spectroscopic imaging. Neurology 42:2011. 22. Meyerhoff DJ, Maudsley AA, Schafer S, Weiner MW. 1992. Phosphorous-31 magnetic resonance metabolite imaging in the human body. Magn Reson Imaging 10:245. 23. Bottomley PA, Hardy C, Boemer P. 1990. Phosphate metabolite imaging and concentration measurements in human heart by nuclear magnetic resonance. Magn Reson Med 14:425. 24. Robitaille PM, Lew B, Merkle H, et al. 1990. Transmural high energy phosphate distribution and response to alterations in workload in the normal canine myocardium as studied with spatially localized 31P NMR spectroscopy. Magn Reson Med 16:91. 25. Ugurbil K, Garwood M, Merkle H, et al. 1989. Metabolic consequences of coronary stenosis: Transmurally heterogeneous myocardial ischemia studied by spatially localized 31P NMR spectroscopy. NMR Biomed 2:317. 26. Hasse A, Frahm J, Hanicker H, Mataei D. 1985. 1H NMR chemical shift selective (CHESS) imaging. Phys Med Biol 30(4):341. 27. Patt SL, Sykes BD. 1972. T1 water eliminated Fourier transform NMR spectroscopy. Chem Phys 56:3182. 28. Moonen CTW, van Zijl PCM. 1990. Highly effective water suppression for in vivo proton NMR spectroscopy (DRYSTEAM). J Magn Reson 88:28. 29. Ogg R, Kingsley P, Taylor JS. 1994. WET: A T1 and B1 insensitive water suppression method for in vivo localized 1H NMR spectroscopy. B104:1. 30. Lampman DA, Murdoch JB, Paley M. 1991. In vivo proton metabolite maps using MESA 3D technique. Magn Reson Med 18:169. 31. Luyten PR, Marien AJH, Heindel W, et al. 1990. Metabolic imaging of patients with intracranial tumors: 1H MR spectroscopic imaging and PET. Radiology 176:791. 32. Arnold DL, Matthews PM, Francis GF, et al. 1992. Proton magnetic resonance spectroscopic imaging for metabolite characterization of demyelinating plaque. Ann Neurol 31:319. 33. Duijin JH, Matson GB, Maudsley AA, et al. 1992. Human brain infarction: Proton MR spectroscopy. Radiology 183:711. 34. Duyn JH, Gillen J, Sobering G, et al. 1993. Multisection proton MR spectroscopic imaging of the brain. Radiology 188:277. 35. Patel MS, Hu X. 1994. Selective data extrapolation for chemical shift imaging. Soc Magn Reson Abstr 3:1168. 36. Rothman DL. 1994 1H NMR studies of human brain metabolism and physiology. In RJ Gillies (ed), NMR in Physiology and Biomedicine, pp 353–372. San Diego, Calif, Academic Press.


Croft, B. Y., Tsui, B. M. W. “ Nuclear Medicine.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

64 Nuclear Medicine 64.1

Barbara Y. Croft National Institutes of Health

Benjamin M.W. Tsui University of North Carolina at Chapel Hill

Instrumentation Parameters for Choices in Nuclear Medicine • Detection of Photon Radiation • Various Detector Configurations • Ancillary Electronic Equipment for Detection • Place of Planar Imaging in Nuclear Medicine Today: Applications and Economics

64.2

SPECT (Single-Photon Emission Computed Tomography) Basic Principles of SPECT • SPECT Instrumentation • Reconstruction Methods • Discussion

64.1 Instrumentation Barbara Y. Croft Nuclear medicine can be defined as the practice of making patients radioactive for diagnostic and therapeutic purposes. The radioactivity is injected intravenously, rebreathed, or ingested. It is the internal circulation of radioactive material that distinguishes nuclear medicine from diagnostic radiology and radiation oncology in most of its forms. This section will examine only the diagnostic use and will concentrate on methods for detecting the radioactivity from outside the body without trauma to the patient. Diagnostic nuclear medicine is successful for two main reasons: (1) It can rely on the use of very small amounts of materials (picomolar concentrations in chemical terms) thus usually not having any effect on the processes being studied, and (2) The radionuclides being used can penetrate tissue and be detected outside the patient. Thus the materials can trace processes or “opacify” organs without affecting their function.

Parameters for Choices in Nuclear Medicine Of the various kinds of emanations from radioactive materials, photons alone have a range in tissue great enough to escape so that they can be detected externally. Electrons or beta-minus particles of high energy can create bremsstrahlung in interactions with tissue, but the radiation emanates from the site of the interaction, not the site of the beta ray’s production. Positrons or beta-plus particles annihilate with electrons to create gamma rays so that they can be detected (see Chap. 67). For certain radionuclides, the emanation being detected is x-rays, in the 50- to 100-KeV energy range. The half-lives of materials in use in nuclear medicine range from a few minutes to weeks. The halflife must be chosen with two major points in mind: the time course of the process being studied and the radiation dose to the target organ, i.e., that organ with the highest concentration over the longest time (the cumulated activity or area underneath the activity versus time curve). In general, it is desired to stay under 5 rad to the target organ. The choice of the best energy range to use is also based on two major criteria: the energy that will penetrate tissue but can be channeled by heavy metal shielding and collimation and that which will


TABLE 64.1

Gamma Ray Detection

Type of Sample

Activity

Energy

Type of Instrument

Patient samples, e.g., blood, urine

0.001 µCi

20–5000 keV

Small organ function 10λ). For particles that are small compared to the wavelength, the scattering is equal in all directions. However, as the particle size becomes larger than the wavelength of light, it becomes preferentially scattered in the forward direction. Lightscattering techniques are widely used to detect the formation of antigen-antibody complexes in immunoassays. When light scattering is measured by the attenuation of a beam of light through a solution, it is called turbidimetry. This is essentially the same as absorption measurements with a photometer except that a large pass-band is acceptable. When maximum sensitivity is required a different method is used—direct measurement of the scattered light with a detector placed at an angle to the central beam. This method is called nephelometry. A typical nephelometer will have a light source, filter, sample cuvette, and detector set at an angle to the incident beam (Fig. 75.5).

Defining Terms Accuracy: The degree to which the average value of repeated measurements approximate the true value being measured. Fluorescence: Emission of light by an atom or molecule following absorption of a photon by greater energy. Emission normally occurs within 10–8 of absorption. Nephelometry: Measurement of the amount of light scattered by particles suspended in a fluid. Precision: A measure of test reproducibility. Sensitivity: A measure of how small an amount or concentration of an analyte can be detected. Specificity: A measure of how well a test detects the intended analyte without being “fooled” by other substances in the sample. Turbidimetry: Measurement of the attenuation of a light beam due to light lost to scattering by particles suspended in a fluid.


References Burtis CA, Ashwood ER (eds). 1994. Tietz Textbook of Clinical Chemistry, 2d ed, Philadelphia, Saunders. Hicks MR, Haven MC, Schenken JR, et al. (eds). 1987. Laboratory Instrumentation, 3d ed, Philadelphia: Lippincott. Kaplan LA, Pesce AJ (eds). 1989. Clinical Chemistry: Theory, Analysis, and Correlation, 2d ed, St. Louis, Mosby. Tietz NW (ed). 1987. Fundamentals of Clinical Chemistry, 3d ed, Philadelphia, Saunders. Ward JM, Lehmann CA, Leiken AM. 1994. Clinical Laboratory Instrumentation and Automation: Principles, Applications, and Selection, Philadelphia, Saunders.


Roa, R. L. “Clinical Laboratory: Nonspectral Methods and Automation.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

76 Clinical Laboratory: Nonspectral Methods and Automation

Richard L. Roa Baylor University Medical Center

76.1 76.2 76.3 76.4 76.5 76.6 76.7 76.8

Particle Counting and Identification Electrochemical Methods Ion-Specific Electrodes Radioactive Methods Coagulation Timers Osmometers Automation Trends in Laboratory Instrumentation

76.1 Particle Counting and Identification The Coulter principle was the first major advance in automating blood cell counts. The cells to be counted are drawn through a small aperture between two fluid compartments, and the electric impedance between the two compartments is monitored (see Fig. 76.1). As cells pass through the aperture, the impedance increases in proportion to the volume of the cell, allowing large numbers of cells to be counted and sized rapidly. Red cells are counted by pulling diluted blood through the aperture. Since red cells greatly outnumber white cells, the contribution of white cells to the red cell count is usually neglected. White cells are counted by first destroying the red cells and using a more concentrated sample. Modern cell counters using the Coulter principle often use hydrodynamic focusing to improve the performance of the instrument. A sheath fluid is introduced which flows along the outside of a channel with the sample stream inside it. By maintaining laminar flow conditions and narrowing the channel, the sample stream is focused into a very thin column with the cells in single file. This eliminates problems with cells flowing along the side of the aperture or sticking to it and minimizes problems with having more than one cell in the aperture at a time. Flow cytometry is a method for characterizing, counting, and separating cells which are suspended in a fluid. The basic flow cytometer uses hydrodynamic focusing to produce a very thin stream of fluid containing cells moving in single file through a quartz flow chamber (Fig. 76.2). The cells are characterized on the basis of their scattering and fluorescent properties. This simultaneous measurement of scattering and fluorescence is accomplished with a sophisticated optical system that detects light from the sample both at the wavelength of the excitation source (scattering) as well as at longer wavelengths (fluorescence) at more than one angle. Analysis of these measurements produces parameters related to the cells’ size, granularity, and natural or tagged fluorescence. High-pressure mercury or xenon arc lamps can be used as light sources, but the argon laser (488 nm) is the preferred source for high-performance instruments. One of the more interesting features of this technology is that particular cells may be selected at rates that allow collection of quantities of particular cell types adequate for further chemical testing. This is


FIGURE 76.1 Coulter method. Blood cells are surrounded by an insulating membrane, which makes them nonconductive. The resistance of electrolyte-filled channel will increase slightly as cells flow through it. This resistance variation yields both the total number of cells which flow through the channel and the volume of each cell.

FIGURE 76.2 Flow cytometer. By combining hydrodynamic focusing, state-of-the-art optics, fluorescent labels, and high-speed computing, large numbers of cells can be characterized and sorted automatically.

accomplished by breaking the outgoing stream into a series of tiny droplets using piezoelectric vibration. By charging the stream of droplets and then using deflection plates controlled by the cell analyzer, the cells of interest can be diverted into collection vessels. The development of monoclonal antibodies coupled with flow cytometry allows for quantitation of T and B cells to assess the status of the immune system as well as characterization of leukemias, lymphomas, and other disorders. © 2000 by CRC Press LLC

FIGURE 76.3

Electrochemical cell.

76.2 Electrochemical Methods Electrochemical methods are increasingly popular in the clinical laboratory, for measurement not only of electrolytes, blood gases, and pH but also of simple compounds such as glucose. Potentiometry is a method in which a voltage is developed across electrochemical cells as shown in Fig. 76.3. This voltage is measured with little or no current flow. Ideally, one would like to measure all potentials between the reference solution in the indicator electrode and the test solution. Unfortunately there is no way to do that. Interface potentials develop across any metal-liquid boundary, across liquid junctions, and across the ion-selective membrane. The key to making potentiometric measurements is to ensure that all the potentials are constant and do not vary with the composition of the test solution except for the potential of interest across the ion-selective membrane. By maintaining the solutions within the electrodes constant, the potential between these solutions and the metal electrodes immersed in them is constant. The liquid junction is a structure which severely limits bulk flow of the solution but allows free passage of all ions between the solutions. The reference electrode commonly is filled with saturated KCl, which produces a small, constant liquidjunction potential. Thus, any change in the measured voltage (V) is due to a change in the ion concentration in the test solution for which the membrane is selective. The potential which develops across an ion-selective membrane is given by the Nernst equation:

 RT  a2 V =  ln  zF  a1 where R T z F an

(76.1)

= gas constant = 8.314 J/K·mol = temperature in K = ionization number = Faraday constant = 9.649 × 104 C/Mol = activity of ion in solution n

When one of the solutions is a reference solution, this equation can be rewritten in a convenient form as

V = V0 +

N log10 a z

where V0 = a constant voltage due to reference solution N = Nernst slope ≈ 59 mV/decade at room temperature © 2000 by CRC Press LLC

(76.2)

The actual Nernst slope is usually slightly less than the theoretical value. Thus, the typical pH meter has two calibration controls. One adjusts the offset to account for the value of V0 , and the other adjusts the range to account for both temperature effects and deviations from the theoretical Nernst slope.

76.3 Ion-Specific Electrodes Ion-selective electrodes use membranes which are permeable only to the ion being measured. To the extent that this can be done, the specificity of the electrode can be very high. One way of overcoming a lack of specificity for certain electrodes is to make multiple simultaneous measurement of several ions which include the most important interfering ones. A simple algorithm can then make corrections for the interfering effects. This technique is used in some commercial electrolyte analyzers. A partial list of the ions that can be measured with ion-selective electrodes includes H+ (pH), Na+, K+, Li+, Ca++, Cl–, F–, NH+4, and CO2 . NH+4, and CO2 are both measured with a modified ion-selective electrode. They use a pH electrode modified with a thin layer of a solution (sodium bicarbonate for CO2 and ammonium chloride for NH+4 ) whose pH varies depending on the concentration of ammonium ions or CO2 it is equilibrated with. A thin membrane holds the solution against the pH glass electrode and provides for equilibration with the sample solution. Note that the CO2 electrode in Fig. 76.4 is a combination electrode. This means that both the reference and indicating electrodes have been combined into one unit. Most pH electrodes are made as combination electrodes. The Clark electrode measures pO2 by measuring the current developed by an electrode with an applied voltage rather than a voltage measurement. This is an example of amperometry. In this electrode a voltage of approximately –0.65 V is applied to a platinum electrode relative to a Ag/AgCl electrode in an electrolyte solution. The reaction

O2 + 2H+ + 2e − → H2O2 proceeds at a rate proportional to the partial pressure of oxygen in the solution. The electrons involved in this reaction form a current which is proportional to the rate of the reaction and thus to the pO2 in the solution.


Clark electrode.

76.4 Radioactive Methods Isotopes are atoms which have identical atomic number (number of protons) but different atomic mass numbers (protons + neutrons). Since they have the same number of electrons in the neutral atom, they have identical chemical properties. This provides an ideal method for labeling molecules in a way that allows for detection at extremely low concentrations. Labeling with radioactive isotopes is extensively used in radioimmunoassays where the amount of antigen bound to specific antibodies is measured. The details of radioactive decay are complex, but for our purposes there are three types of emission from decaying nuclei: alpha, beta, and gamma radiation. Alpha particles are made up of two neutrons and two protons (helium nucleus). Alpha emitters are rarely used in the clinical laboratory. Beta emission consists of electrons or positrons emitted from the nucleus. They have a continuous range of energies up to a maximum value characteristic of the isotope. Beta radiation is highly interactive with matter and cannot penetrate very far in most materials. Gamma radiation is a high-energy form of electromagnetic radiation. This type of radiation may be continuous, discrete, or mixed depending on the details of the decay process. It has greater penetrating ability than beta radiation. (See Fig. 76.5.) The kinetic energy spectrum of emitted radiation is characteristic of the isotope. The energy is commonly measured in electron volts (eV). One electron volt is the energy acquired by an electron falling through a potential of 1 volt. The isotopes commonly used in the clinical laboratory have energy spectra which range from 18 keV–3.6 MeV. The activity of a quantity of radioactive isotope is defined as the number of disintegrations per second which occur. The usual units are the curie (Ci), which is defined as 3.7 × 1010 dps, and the becquerel (Bq), defined as 1 dps. Specific activity for a given isotope is defined as activity per unit mass of the isotope. The rate of decay for a given isotope is characterized by the decay constant λ, which is the proportion of the isotope which decays in unit time. Thus, the rate of loss of radioactive isotope is governed by the equation

dN = −λN dt

(76.3)

where N is the amount of radioactive isotope present at time t. The solution to this differential equation is:

N = N 0e − λt

(76.4)

FIGURE 76.5 Gamma counted. The intensity of the light flash produced when a gamma photon interacts with a scintillator is proportional to the energy of the photon. The photomultiplier tube converts these light flashes into electric pulses which can be selected according to size (gamma energy) and counted. © 2000 by CRC Press LLC

It can easily be shown that the amount of radioactive isotope present will be reduced by half after time

t1 2 =

0.693 λ

(76.5)

This is known as the half-life for the isotope and can vary widely; for example, carbon-14 has a half-life of 5760 years, and iodine-131 has a half-life of 8.1 days. The most common method for detection of radiation in the clinical laboratory is by scintillation. This is the conversion of radiation energy into photons in the visible or near-UV range. These are detected with photomultiplier tubes. For gamma radiation, the scintillating crystal is made of sodium iodide doped with about 1% thallium, producing 20 to 30 photons for each electron-volt of energy absorbed. The photomultiplier tube and amplifier circuit produce voltage pulses proportional to the energy of the absorbed radiation. These voltage pulses are usually passed through a pulse-height analyzer which eliminates pulses outside a preset energy range (window). Multichannel analyzers can discriminate between two or more isotopes if they have well-separated energy maxima. There generally will be some spill down of counts from the higherenergy isotope into the lower-energy isotope’s window, but this effect can be corrected with a simple algorithm. Multiple well detectors with up to 64 detectors in an array are available which increase the throughput for counting systems greatly. Counters using the sodium iodide crystal scintillator are referred to as gamma counters or well counters. The lower energy and short penetration ability of beta particles requires a scintillator in direct contact with the decaying isotope. This is accomplished by dissolving or suspending the sample in a liquid fluor. Counters which use this technique are called beta counters or liquid scintillation counters. Liquid scintillation counters use two photomultiplier tubes with a coincidence circuit that prevents counting of events seen by only one of the tubes. In this way, false counts due to chemiluminescence and noise in the phototube are greatly reduced. Quenching is a problem in all liquid scintillation counters. Quenching is any process which reduces the efficiency of the scintillation counting process, where efficiency is defined as

Efficiency = counts per minute decays per minute

(76.6)

A number of techniques have been developed that automatically correct for quenching effects to produce estimates of true decays per minute from the raw counts. Currently there is a trend away from betaemitting isotopic labels, but these assays are still used in many laboratories.

76.5 Coagulation Timers Screening for and diagnosis of coagulation disorders is accomplished by assays that determine how long it takes for blood to clot following initiation of the clotting cascade by various reagents. A variety of instruments have been designed to automate this procedure. In addition to increasing the speed and throughput of such testing, these instruments improve the reproducibility of such tests. All the instruments provide precise introduction of reagents, accurate timing circuits, and temperature control. They differ in the method for detecting clot formation. One of the older methods still in use is to dip a small metal hook into the blood sample repeatedly and lift it a few millimeters above the surface. The electric resistance between the hook and the sample is measured, and when fibrin filaments form, they produce a conductive pathway which is detected as clot formation. Other systems detect the increase in viscosity due to fibrin formation or the scattering due to the large polymerized molecules formed. Absorption and fluorescence spectroscopy can also be used for clot detection.


76.6 Osmometers The colligative properties of a solution are a function of the number of solute particles present regardless of size or identity. Increased solute concentration causes an increase in osmotic pressure and boiling point and a decrease in vapor pressure and freezing point. Measuring these changes provides information on the total solute concentration regardless of type. The most accurate and popular method used in clinical laboratories is the measurement of freezing point depression. With this method, the sample is supercooled to a few degrees below 0°C while being stirred gently. Freezing is then initiated by vigorous stirring. The heat of fusion quickly brings the solution to a slushy state where an equilibrium exists between ice and liquid, ensuring that the temperature is at the freezing point. This temperature is measured. A solute concentration of 1 osmol/kg water produces a freezing point depression of 1.858°C. The measured temperature depression is easily calibrated in units of milliosmols/kg water. The vapor pressure depression method has the advantage of smaller sample size. However, it is not as precise as the freezing point method and cannot measure the contribution of volatile solutes such as ethanol. This method is not used as widely as the freezing point depression method in clinical laboratories. Osmolality of blood is primarily due to electrolytes such as Na+ and Cl–. Proteins with molecular weights of 30,000 or more atomic mass units (amu) contribute very little to total osmolality due to their smaller numbers (a single Na+ ion contributes just as much to osmotic pressure as a large protein molecule). However, the contribution to osmolality made by proteins is of great interest when monitoring conditions leading to pulmonary edema. This value is known as colloid osmotic pressure, or oncotic pressure, and is measured with a membrane permeable to water and all molecules smaller than about 30,000 amu. By placing a reference saline solution on one side and the unknown sample on the other, an osmotic pressure is developed across the membrane. This pressure is measured with a pressure transducer and can be related to the true colloid osmotic pressure through a calibration procedure using known standards.

76.7 Automation Improvements in technology coupled with increased demand for laboratory tests as well as pressures to reduce costs have led to the rapid development of highly automated laboratory instruments. Typical automated instruments contain mechanisms for measuring, mixing, and transport of samples and reagents, measurement systems, and one or more microprocessors to control the entire system. In addition to system control, the computer systems store calibration curves, match test results to specimen IDs, and generate reports. Automated instruments are dedicated to complete blood counts, coagulation studies, microbiology assays, and immunochemistry, as well as high-volume instruments used in clinical chemistry laboratories. The chemistry analyzers tend to fall into one of four classes: continuous flow, centrifugal, pack-based, and dry-slide-based systems. The continuous flow systems pass successive samples and reagents through a single set of tubing, where they are directed to appropriate mixing, dialyzing, and measuring stations. Carry-over from one sample to the next is minimized by the introduction of air bubbles and wash solution between samples. Centrifugal analyzers use plastic rotors which serve as reservoirs for samples and reagents and also as cuvettes for optical measurements. Spinning the plastic rotor mixes, incubates, and transports the test solution into the cuvette portion of the rotor, where the optical measurements are made while the rotor is spinning. Pack-based systems are those in which each test uses a special pack with the proper reagents and sample preservation devices built-in. The sample is automatically introduced into as many packs as tests required. The packs are then processed sequentially. Dry chemistry analyzers use no liquid reagents. The reagents and other sample preparation methods are layered onto a slide. The liquid sample is placed on the slide, and after a period of time the color developed is read by reflectance photometry. Ion-selective electrodes have been incorporated into the same slide format.


There are a number of technological innovations found in many of the automated instruments. One innovation is the use of fiberoptic bundles to channel excitation energy toward the sample as well as transmitted, reflected, or emitted light away from the sample to the detectors. This provides a great deal of flexibility in instrument layout. Multiwavelength analysis using a spinning filter wheel or diode array detectors is commonly found. The computers associated with these instruments allow for innovative improvements in the assays. For instance, when many analytes are being analyzed from one sample, the interference effects of one analyte on the measurement of another can be predicted and corrected before the final report is printed.

76.8 Trends in Laboratory Instrumentation Predicting the future direction of laboratory instrumentation is difficult, but there seem to be some clear trends. Decentralization of the laboratory functions will continue with more instruments being located in or around ICUs, operating rooms, emergency rooms, and physician offices. More electrochemistrybased tests will be developed. The flame photometer is already being replaced with ion-selective electrode methods. Instruments which analyze whole blood rather than plasma or serum will reduce the amount of time required for sample preparation and will further encourage testing away from the central laboratory. Dry reagent methods increasingly will replace wet chemistry methods. Radioimmunoassays will continue to decline with the increasing use of methods for performing immunoassays that do not rely upon radioisotopes such as enzyme-linked fluorescent assays.

Defining Terms Alpha radiation: Particulate radiation consisting of a helium nucleus emitted from a decaying anucleus. Amperometry: Measurements based on current flow produced in an electrochemical cell by an applied voltage. Beta radiation: Particulate radiation consisting of an electron or positron emitted from a decaying nucleus. Colligative properties: Physical properties that depend on the number of molecules present rather than on their individual properties. Gamma radiation: Electromagnetic radiation emitted from an atom undergoing nuclear decay. Hydrodynamic focusing: A process in which a fluid stream is first surrounded by a second fluid and then narrowed to a thin stream by a narrowing of the channel. Isotopes: Atoms with the same number of protons but differing numbers of neutrons. Plasma: The liquid portion of blood. Potentiometry: Measurement of the potential produced by electrochemical cells under equilibrium conditions with no current flow. Scintillation: The conversion of the kinetic energy of a charged particle or photon to a flash of light. Serum: The liquid portion of blood remaining after clotting has occurred.

References Burtis CA, Ashwood ER (eds). 1994. Tietz Textbook of Clinical Chemistry, 2d ed, Philadelphia, Saunders Company. Hicks MR, Haven MC, Schenken JR, et al. (eds). 1987. Laboratory Instrumentation, 3d ed, Philadelphia, Lippincott Company, 1987. Kaplan LA, Pesce AJ (eds). 1989. Clinical Chemistry: Theory, Analysis, and Correlation, 2d ed, St. Louis, Mosby. Tietz NW (ed). 1987. Fundamentals of Clinical Chemistry, 3d ed, Philadelphia, Saunders. Ward JM, Lehmann CA, Leiken AM. 1994. Clinical Laboratory Instrumentation and Automation: Principles, Applications, and Selection, Philadelphia, Saunders. © 2000 by CRC Press LLC

Forde, M., Ridgely, P. “Implantable Cardiac Pacemakers.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

77 Implantable Cardiac Pacemakers 77.1 77.2

Indications Pulse Generators Sensing Circuit • Output Circuit • Timing Circuit • Telemetry Circuit • Power Source

Michael Forde Medtronic, Inc.

Pat Ridgely Medtronic, Inc.

77.3 77.4 77.5 77.6 77.7

Leads Programmers System Operation Clinical Outcomes and Cost Implications Conclusion

The practical use of an implantable device for delivering a controlled, rhythmic electric stimulus to maintain the heartbeat is relatively recent: Cardiac pacemakers have been in clinical use only slightly more than 30 years. Although devices have gotten steadily smaller over this period (from 250 grams in 1960 to 25 grams today), the technological evolution goes far beyond size alone. Early devices provided only single-chamber, asynchronous, nonprogrammable pacing coupled with questionable reliability and longevity. Today, advanced electronics afford dual-chamber multiprogrammability, diagnostic functions, rate response, data collection, and exceptional reliability, and lithium-iodine power sources extend longevity to upward of 10 years. Continual advances in a number of clinical, scientific, and engineering disciplines have so expanded the use of pacing that it now provides cost-effective benefits to an estimated 350,000 patients worldwide each year. The modern pacing system is comprised of three distinct components: pulse generator, lead, and programmer (Fig. 77.1). The pulse generator houses the battery and the circuitry which generates the stimulus and senses electrical activity. The lead is an insulated wire that carries the stimulus from the generator to the heart and relays intrinsic cardiac signals back to the generator. The programmer is a telemetry device used to provide two-way communications between the generator and the clinician. It can alter the therapy delivered by the pacemaker and retrieve diagnostic data that are essential for optimally titrating that therapy. Ultimately, the therapeutic success of the pacing prescription rests on the clinician’s choice of an appropriate system, use of sound implant technique, and programming focused on patient outcomes. This chapter discusses in further detail the components of the modern pacing system and the significant evolution that has occurred since its inception. Our focus is on system design and operations, but we also briefly overview issues critical to successful clinical performance.

77.1 Indications The decision to implant a permanent pacemaker for bradyarrhythmias usually is based on the major goals of symptom relief (at rest and with physical activity), restoration of functional capacity and quality


FIGURE 77.1 The pacing systems comprise a programmer, pulse generator, and lead. There are two programmers pictured above; one is portable, and the other is an office-based unit.

of life, and reduced mortality. As with other healthcare technologies, appropriate use of pacing is the intent of indications guidelines established by Medicare and other third-party payors. In 1984 and again in 1991, a joint commission of the American College of Cardiology and the American Heart Association established guidelines for pacemaker implantation [Committee on Pacemaker Implantation, 1991]. In general, pacing is indicated when there is a dramatic slowing of the heart rate or a failure in the connection between the atria and ventricles resulting in decreased cardiac output manifested by such symptoms as syncope, light-headedness, fatigue, and exercise intolerance. Failure of impulse formation and/or conduction is the overriding theme of all pacemaker indications. There are four categories of pacing indications: 1. 2. 3. 4.

Heart block (e.g., complete heart block, symptomatic 2° AV block) Sick sinus syndrome (e.g., symptomatic bradycardia, sinus arrest, sinus exit block) Myocardial infarction (e.g., conduction disturbance related to the site of infarction) Hypersensitive carotid sinus syndrome (e.g., recurrent syncope)

Within each of these four categories the ACC/AHA provided criteria for classifying a condition as group I (pacing is considered necessary), group II (pacing may be necessary), or group III (pacing is considered inappropriate). New indications for cardiac pacing are being evaluated under the jurisdiction of the Food and Drug Administration. For example, hypertrophic obstructive cardiomyopathy (HOCM) is one of these new potential indications, with researchers looking at dual-chamber pacing as a means of reducing left ventricular outflow obstruction. Though efforts in these areas are ongoing and expanding, for now they remain unapproved as standard indications for pacing.

77.2 Pulse Generators The pulse generator contains a power source, output circuit, sensing circuit, and a timing circuit (Fig. 77.2). A telemetry coil is used to send and receive information between the generator and the programmer. Rate-adaptive pulse generators include the sensor components along with the circuit to process the information measured by the sensor. Modern pacemakers use CMOS circuit technology. One to 2 kilobytes of read-only memory (ROM) are used to direct the output and sensing circuits; 16–512 bytes of random-access memory (RAM) are © 2000 by CRC Press LLC

used to store diagnostic data. Some manufacturers offer fully RAM-based pulse generators, providing greater storage of diagnostic data and the flexibility for changing feature sets after implantation. All components of the pulse generator are housed in a hermetically sealed titanium case with a connector block that accepts the lead(s). Because pacing leads are available with a variety of different connector sites and configurations, the pulse generator is available with an equal variety of connectors. The outer casing is laseretched with the manufacturer, name, type (e.g., singleversus dual-chamber), model number, serial number, and the lead connection diagram for each identification. Once implanted, it may be necessary to use an x-ray to reveal the identity of the generator. Some manufacturers use radiopaque symbols and ID codes for this purpose, whereas others give their generators characteristic shapes.

FIGURE 77.2

Internal view of pulse generator.

Sensing Circuit Pulse generators have two basic functions, pacing and sensing. Sensing refers to the recognition of an appropriate signal by the pulse generator. This signal is the intrinsic cardiac depolarization from the chamber or chambers in which the leads are placed. It is imperative for the sensing circuit to discriminate between these intracardiac signals and unwanted electrical interference such as far-field cardiac events, diastolic potentials, skeletal muscle contraction, and pacing stimuli. An intracardiac electrogram (Fig. 77.3) shows the waveform as seen by the pacemaker; it is typically quite different from the corresponding event as shown on the surface ECG. Sensing (and pacing) is accomplished with one of two configurations, bipolar and unipolar. In bipolar, the anode and cathode are close together, with the anode at the tip of the lead and the cathode a ring electrode about 2 cm proximal to the tip. In unipolar, the anode and cathode may be 5–10 cm apart. The anode is at the lead tip and the cathode is the pulse generator itself (usually located in the pectoral region).

FIGURE 77.3 The surface ECG (ECG LEAD II) represents the sum total of the electrical potentials of all depolarizing tissue. The intracardiac electrogram (V EGM) shows only the potentials measured between the lead electrodes. This allows the evaluation of signals that may be hidden within the surface ECG.


FIGURE 77.4 This is a conceptual depiction of the bandpass filter demonstrating the typical filtering of unwanted signals by discriminating between those with slew rates that are too low and/or too high.

In general, bipolar and unipolar sensing configurations have equal performance. A drawback of the unipolar approach is the increased possibility of sensing noncardiac signals: The large electrode separation may, for example, sense myopotentials from skeletal muscle movement, leading to inappropriate inhibition of pacing. Many newer pacemakers can be programmed to sense or pace in either configuration. Once the electrogram enters the sensing circuit, it is scrutinized by a bandpass filter (Fig. 77.4). The frequency of an R-wave is 10 to 30 Hz. The center frequency of most sensing amplifiers is 30 Hz. T-waves are slower, broad signals that are composed of lower frequencies (approximately 5 Hz or less). Far-field signals are also lower-frequency signals, whereas skeletal muscle falls in the range of 10–200 Hz. At the implant, the voltage amplitude of the R-wave (and the P-wave, in the case of dual-chamber pacing) is measured to ensure the availability on an adequate signal. R-wave amplitudes are typically 5–25 mV, and P-wave amplitudes are 2–6 mV. The signals passing through the sense amplifier are compared to an adjustable reference voltage called the sensitivity. Any signal below the reference voltage is not sensed, and those above it are sensed. Higher-sensitivity settings (high-reference voltage) may lead to substandard sensing, and a lower reference voltage may result in oversensing. A minimum 2:1 safety margin should be maintained between the sensitivity setting and the amplitude of the intracardiac signal. The circuit is protected from extremely high voltages by a Zener diode. The slope of the signal is also surveyed by the sensing circuit and is determined by the slew rate (the time rate of change in voltage). A slew rate that is too flat or too steep may be eliminated by the bandpass filter. On the average, the slew rate measured at implant should be between 0.75 and 2.50 V/s. The last line of defense in an effort to remove undesirable signals is to “blind” the circuit at specific times during the cardiac cycle. This is accomplished with blanking and refractory periods. Some of these periods are programmable. During the blanking period the sensing circuit is turned off, and during the refractory period the circuit can see the signal but does not initiate any of the basic timing intervals. Virtually all paced and sensed events begin concurrent blanking and refractory periods, typically ranging from 10–400 ms. These are especially helpful in dual-chamber pacemakers where there exists the potential


for the pacing output of the atrial side to inhibit the ventricular pacing output, with dangerous consequences for patients in complete heart block. Probably the most common question asked by the general public about pacing systems is the effect of electromagnetic interference (EMI) on their operation. EMI outside of the hospital is an infrequent problem, though patients are advised to avoid such sources of strong electromagnetic fields as arc welders, high-voltage generators, and radar antennae. Some clinicians suggest that patients avoid standing near antitheft devices used in retail stores. Airport screening devices are generally safe, though they may detect a pacemaker’s metal case. Microwave ovens, ham radio equipment, video games, computers, and office equipment rarely interfere with the operation of modern pacemakers. A number of medical devices and procedures may on occasion do so, however; electrocautery, cardioversion and defibrillation, MRI, lithotripsy, diathermy, TENS units, and radiation therapy. Pacemakers affected by interference typically respond with temporary loss of output or temporary reversion to asynchronous pacing (pacing at a fixed rate, with no inhibition from intrinsic cardiac events). The usual consequence for the patient is a return of the symptoms that originally led to the pacemaker implant.

Output Circuit Pacing is the most significant drain on the pulse generator power source. Therefore, current drain must be minimized while maintaining an adequate safety margin between the stimulation threshold and the programmed output stimulus. Modern permanent pulse generators use constant voltage. The voltage remains at the programmed value while current fluctuates in relation to the source impedance. Output energy is controlled by two programmable parameters, pulse amplitude and pulse duration. Pulse amplitudes range from 0.8–5 V and, in some generators, can be as high as 10 V (used for troubleshooting or for pediatric patients). Pulse duration can range from 0.05–1.5 ms. The prudent selection of these parameters will greatly influence the longevity of the pulse generator. The output pulse is generated from the discharge of a capacitor charged by the battery. Most modern pulse generators contain a 2.8 V battery. The higher voltages are achieved using voltage multipliers (smaller capacitors used to charge the large capacitor). The voltage can be doubled by charging two smaller capacitors in parallel, with the discharge delivered to the output capacitor in series. Output pulses are emitted at a rate controlled by the timing circuit; output is commonly inhibited by sensed cardiac signals.

Timing Circuit The timing circuit regulates such parameters as the pacing cycle length, refractory and blanking periods, pulse duration, and specific timing intervals between atrial and ventricular events. A crystal oscillator generating frequencies in the kHz range sends a signal to a digital timing and logic control circuit, which in turn operates internally generated clocks at divisions of the oscillatory frequency. A rate-limiting circuit is incorporated into the timing circuit to prevent the pacing rate from exceeding an upper limit should a random component failure occur (an extremely rare event). This is also referred to as “runaway” protection and is typically 180–200 ppm.

Telemetry Circuit Today’s pulse generators are capable of both transmitting information from an RF antenna and receiving information with an RF decoder. This two-way communication occurs between the pulse generator and the programmer at approximately 300 Hz. Real-time telemetry is the term used to describe the ability of the pulse generator to provide information such as pulse amplitude, pulse duration, lead impedance, battery impedance, lead current, charge, and energy. The programmer, in turn, delivers coded messages to the pulse generator to alter any of the programmable features and to retrieve diagnostic data. Coding


requirements reduce the likelihood of inappropriate programming alterations by environmental sources of radiofrequency and magnetic fields. It is also prevents the improper use of programmers from other manufacturers.

Power Source Over the years, a number of different battery technologies have been tried, including mercury-zinc, rechargeable silver-modified-mercuric-oxide-zinc, rechargeable nickel-cadmium, radioactive plutonium or promethium, and lithium with a variety of different cathodes. Lithium-cupric-sulfide and mercuryzinc batteries were associated with corrosion and early failure. Mercury-zinc produced hydrogen gas as a by-product of the battery reaction; the venting required made it impossible to hermetically seal the generator. This led to fluid infiltration followed by the risk of sudden failure. The longevity of very early pulse generators was measured in hours. With the lithium-iodide technology now used, longevity has been reported as high as 15 years. The clinical desire to have a generator that is small and full-featured yet also long-lasting poses a formidable challenge to battery designers. One response by manufacturers has been to offer different models of generators, each offering a different balance between therapy, size, and longevity. Typical battery capacity is in the range of 0.8–3.0 amp-hours. Many factors affect longevity, including pulse amplitude and duration, pacing rate, singleversus dualchamber pacing, degree to which the patient uses the pacemaker, lead design, and static current drain from the sensing circuits. Improvements in lead design are often overlooked as a factor in improving longevity, but electrodes used in 1960 required a pulse generator output of 675 µJ for effective stimulation, whereas the electrodes of the 1990s need only 3–6 µJ. Another important factor in battery design lies in the electrolyte that separates the anode and the cathode. The semisolid layer of lithium iodide that is used gradually thickens over the life of the cell, increasing the internal resistance of the battery. The voltage produced by lithium-iodine batteries is inversely related to this resistance and is linear from 2.8 V to approximately 2.4 V, representing about 90% of the usable battery life. It then declines exponentially to 1.8 V as the internal battery resistance increases from 10,000 Ω to 40,000 Ω (Fig. 77.5). When the battery reaches between 2.0 and 2.4 V (depending on the manufacturer), certain functions of the pulse generator are altered so as to alert the clinician. These alterations are called the electivereplacement indicators (ERI). They vary from one pulse generator to another and include signature decreases in rate, a change to a specific pacing mode, pulse duration stretching, and the telemetered battery voltage. When the battery voltage reaches 1.8 V, the pulse generator may operate erratically or cease to function and is said to have reached “end of life.” The time period between appearance of the ERI and end-of-life status averages about 3 to 4 months.

FIGURE 77.5 The initial decline in battery voltage is slow and then more rapid after the battery reaches the ERI voltage. An important aspect of battery design is the predictability of this decline so that timely generator replacement is anticipated. © 2000 by CRC Press LLC

FIGURE 77.6

The four major lead components.

77.3 Leads Implantable pacing leads must be designed not only for consistent performance within the hostile environment of the body but also for easy handling by the implanting physician. Every lead has four major components (Fig. 77.6): the electrode, the conductor, the insulation, and the connector pin(s). The electrode is located at the tip of the lead and is in direct contact with the myocardium. Bipolar leads have a tip electrode and a ring electrode (located about 2 cm proximal to the tip); unipolar leads have tip electrodes only. A small-radius electrode provides increased current density resulting in lower stimulation thresholds. The electrode also increases resistance at the electrode-myocardial interface, thus lowering the current drain further and improving battery longevity. The radius of most electrodes is 6–8 mm2, though there are clinical trials underway using a “high-impedance” lead with a tip radius as low as 1.5 mm2. Small electrodes, however, historically have been associated with inferior sensing performance. Lead designers were able to achieve both good pacing and good sensing by creating porous-tip electrodes containing thousands of pores in the 20–100 µm range. The pores allow the ingrowth of tissue, resulting in the necessary increase in effective sensing area while maintaining a small pacing area. Some commonly used electrode materials include platinum-iridium. Elgiloy (an alloy of cobalt, iron, chromium, molybdenum, nickel, and manganese), platinum coated with platinized titanium, and vitreous or pyrolytic carbon coating a titanium or graphite core. Another major breakthrough in lead design is the steroid-eluting electrode. About 1 mg of a corticosteroid (dexamethasone sodium phosphate) is contained in a silicone core that is surrounded by the electrode material (Fig. 77.7). The “leaking” of the steroid into the myocardium occurs slowly over several

FIGURE 77.7


The steroid elution electrode.

years and reduces the inflammation that results from the lead placement. It also retards the growth of the fibrous sack that forms around the electrode which separates it from viable myocardium. As a result, the dramatic rise in acute thresholds that is seen with nonsteroid leads over the 8 to 16 weeks postimplant is nearly eliminated. This makes it possible to program a lower pacing output, further extending longevity. Once a lead has been implanted, it must remain stable (or fixated). The fixation device is either active or passive. The active fixation leads incorporate corkscrew mechanisms, barbs, or hooks to attach themselves to the myocardium. The passive fixation leads are held into place with tines that become entangled into the netlike lining (trabeculae) of the heart. Passive leads generally have better acute pacing and sensing performance but are difficult to remove chronically. Active leads are easier to remove chronically and have the advantage of unlimited placement sites. Some implanters prefer to use active-fixation leads in the atrium and passive-fixation leads in the ventricle. The conductor carries electric signals to the pulse generator and delivers the pacing pulses to the heart. It must be strong and flexible to withstand the repeated flexing stress placed on it by the beating heart. The early conductors were a single, straight wire that was vulnerable to fracturing. They have evolved into coiled (for increased flexibility) multifilar (to prevent complete failure with partial fractures) conductors. The conductor material is a nickel alloy called MP35N. Because of the need for two conductors, bipolar leads are usually larger in diameter than unipolar leads. Current bipolar leads have a coaxial design that has significantly reduced the diameter of bipolar leads. Insulation materials (typically silicone and polyurethane) are used to isolate the conductor. Silicone has a longer history and the exclusive advantage of being repairable. Because of low tear strength, however, silicone leads tend to be thicker than polyurethane leads. Another relative disadvantage of silicone is its high coefficient of friction in blood, which makes it difficult for two leads to pass through the same vein. A coating applied to silicone leads during manufacturing has diminished this problem. A variety of generator-lead connector configurations and adapters are available. Because incompatibility can result in disturbed (or even lost) pacing and sensing, an international standards (IS-1) has been developed in an attempt to minimize incompatibility. Leads can be implanted epicardially and endocardially. Epicardial leads are placed on the outer surface of the heart and require the surgical exposure of a small portion of the heart. They are used when venous occlusion makes it impossible to pass a lead transvenously, when abdominal placement of the pulse generator is needed (as in the case of radiation therapy to the pectoral area), or in children (to allow for growth). Endocardial leads are more common and perform better in the long term. These leads are passed through the venous system and into the right side of the heart. The subclavian or cephalic veins in the pectoral region are common entry sites. Positioning is facilitated by a thin, firm wire stylet that passes through the central lumen of the lead, stiffening it. Fluoroscopy is used to visualize lead positioning and to confirm the desired location. Manufacturers are very sensitive to the performance reliability of the leads. Steady improvements in materials, design, manufacturing, and implant technique have led to reliability well in excess of 99% over 3-year periods.

77.4 Programmers Noninvasive reversible alteration of the functional parameters of the pacemaker is critical to ongoing clinical management. For a pacing system to remain effective throughout its lifetime, it must be able to adjust to the patient’s changing needs. The programmer is the primary clinical tool for changing settings, for retrieving diagnostic data, and for conducting noninvasive tests. The pacing rate for programmable pacemakers of the early 1960s was adjusted via a Keith needle manipulated percutaneously into a knob on the side of the pacemaker; rotating the needle changed the pacing rate. Through the late 1960s and early 1970s, magnetically attuned reed switches in the pulse generator made it possible to noninvasively change certain parameters such as rate, output, sensitivity, and polarity. The application of a magnet could alter the parameters which were usually limited to only one of two choices. It wasn’t until the late 1970s, when radiofrequency energy was incorporated as the © 2000 by CRC Press LLC

transmitter of information, that programmability began to realize its full potential. Radiofrequency transmission is faster, provides bidirectional telemetry, and decreases the possibility of unintended programming from inappropriate sources. Most manufacturers today are moving away from a dedicated proprietary instrument and toward a PC-based design. The newer designs are generally more flexible, more intuitive to use, and more easily updated when new devices are released. Manufacturers and clinicians alike are becoming more sensitive to the role that time-efficient programming can play in the productivity of pacing clinics, which may provide follow-up for as many as 500–1000 patients a year.

77.5 System Operation Pacemakers have gotten steadily more powerful over the last three decades, but at the cost of steadily greater complexity. Manufacturers have come to realize the challenge that this poses for busy clinicians and have responded with a variety of interpretive aids (Fig. 77.8). Much of the apparent complexity of the timing rules that determine pacemaker operation is due to a design goal of mimicking normal cardiac function without interfering with it. One example is the dualchamber feature that provides sequential stimulation of the atrium before the ventricle. Another example is rate response, designed for patients who lack the normal ability to increase their heart rate in response to a variety of physical conditions (e.g., exercise). Introduced in the mid-1980s, rate-responsive systems use some sort of sensor to measure the change in a physical variable correlated to heart rate. The sensor output is signal-processed and then used by the output circuit to specify a target pacing rate. The clinician controls the aggressiveness of the rate increase through a variety of parameters (including a choice of transfer function); pacemaker-resident diagnostics provide data helpful in titrating the rate-response therapy. The most common sensor is the activity sensor, which uses piezoelectric materials to detect vibrations caused by body movement. Systems using a transthoracic-impedance sensor to estimate pulmonary minute ventilation are also commercially available. Numerous other sensors (e.g., stroke volume, blood

FIGURE 77.8 The Marker Channel Diagram is just one tool that makes interpretation of the ECG strip faster and more reliable for the clinician. It allows quick checking of the timing operations of the system.


TABLE 77.1

The NASPE/NPEG Code

Position

I

II

III

Category

Chamber(s) paced

Chamber(s) sensed O = None A = Atrium V = Ventricle D = Dual (A+V)

Response to sensing O = None T = Triggered I = Inhibited D = Dual (T+I)

O = None A = Atrium V = Ventricle D = Dual (A+V) Manufacturers’ designation only

S = Single (A or V)

IV

V

Programmability rate Antitachyarrhythmia modulation function(s) O = None O = None P = Simple programmable P = Packing M = Multiprogrammable S = Shock C = Communicating D = Dual (P+S) R = Rate modulation

S = Single (A or V)

Note: Positions I through III are used exclusively for antibradyarrhythmia function. (From Bernstein AD, et al., PACE, Vol. 10, July–Aug. 1987.)

temperature or pH, oxygen saturation, preejection interval, right ventricular pressure) are in various stages of clinical research or have been market released outside the United States. Some of these systems are dual-sensor, combining the best features of each sensor in a single pacing system. To make it easier to understand the gross-level system operation of modern pacemakers, a five-letter code has been developed by the North American Society of Pacing and Electrophysiology and the British Pacing and Electrophysiology Group [Bernstein et al., 1987]. The first letter indicates the chamber (or chambers) that are paced. The second letter reveals those chambers in which sensing takes place, and the third letter describes how the pacemaker will respond to a sensed event. The pacemaker will “inhibit” the pacing output when intrinsic activity is sensed or will “trigger” a pacing output based on a specific previously sensed event. For example, in DDD mode: D: Pacing takes place in the atrium and the ventricle. D: Sensing takes place in the atrium and the ventricle. D: Both inhibition and triggering are the response to a sensed event. An atrial output is inhibited with an atrial-sensed event, whereas a ventricular output is inhibited with a ventricular-sensed event; a ventricular pacing output is triggered by an atrial-sensed event (assuming no ventricular event occurs during the A-V interval). The fourth letter in the code is intended to reflect the degree of programmability of the pacemaker but is typically used to indicate that the device can provide rate response. For example, a DDDR device is one that is programmed to pace and sense in both chambers and is capable of sensor-driven rate variability. The fifth letter is reserved specifically for antitachycardia functions (Table 77.1). Readers interested in the intriguing details of pacemaker timing operations are referred to the works listed at the end of this chapter.

77.6 Clinical Outcomes and Cost Implications The demonstrable hemodynamic and symptomatic benefits provided by rate-responsive and dual-chamber pacing have led U.S. physicians to include at least one of these features in over three-fourths of implants in recent years. Also, new prospective data [Andersen et al., 1993] support a hypothesis investigated retrospectively since the mid-1980s: namely, that pacing the atrium in patients with sinus node dysfunction can dramatically reduce the incidence of such life-threatening complications as congestive heart failure and stroke associated with chronic atrial fibrillation. Preliminary analysis of the cost implications suggest that dual-chamber pacing is significantly cheaper to the U.S. healthcare system than is single-chamber pacing over the full course of therapy, despite the somewhat higher initial cost of implanting the dual-chamber system.


77.7

Conclusion

Permanent cardiac pacing is the beneficiary of three decades of advances in a variety of key technologies: biomaterials, electrical stimulation, sensing of bioelectrical events, power sources, microelectronics, transducers, signal analysis, and software development. These advances, informed and guided by a wealth of clinical experience acquired during that time, have made pacing a cost-effective cornerstone of cardiac arrhythmia management.

Defining Terms Atrial fibrillation: An atrial arrhythmia resulting in chaotic current flow within the atria. The effective contraction of the atria is lost, allowing blood to pool and clot, leading to stroke if untreated. Battery capacity: Given by the voltage and the current delivery. The voltage is a result of the battery chemistry, and current delivery (current × time) is measured in ampere hours and is related to battery size. CMOS circuit: Abbreviation for complementary metallic oxide semiconductor, which is a form of semiconductor often used in pacemaker technology. Congestive heart failure: The pathophysiologic state in which an abnormality of cardiac function is responsible for the failure of the heart to pump blood at a rate commensurate with the requirements of the body. Endocardium: The inner lining of the heart. Epicardium: The outer lining of the heart. Hermeticity: The term, as used in the pacemaker industry, refers to a very low rate of helium gas leakage from the sealed pacemaker container. This reduces the chance of fluid intruding into the pacemaker generator and causing damage. Hypertrophic obstructive cardiomyopathy: A disease of the myocardium characterized by thickening (hypertrophy) of the interventricular septum, resulting in the partial obstruction of blood from the left ventricle. Minute ventilation: Respiratory rate × tidal volume (the amount of air taken in with each breath) = minute ventilation. This parameter is used as a biologic indicator for rate-adaptive pacing. Mode: The type of pacemaker response to the patient’s intrinsic heartbeat. The three commonly used modes are asynchronous, demand, and triggered. Programmable: The ability to alter the pacemaker settings noninvasively. A variety of selections exist, each with its own designation. Rate-adaptive: The ability to change the pacemaker stimulation interval caused by sensing a physiologic function other than the intrinsic atrial rhythm. Sensitivity: A programmable setting that adjusts the reference voltage to which signals entering the sensing circuit are compared for filtering. Stimulation threshold: The minimum output energy required to consistently “capture” (cause depolarization) of the heart.

References Andersen HR, Thuesen L, Bagger JP, et al. 1993. Atrial versus ventricular pacing in sick sinus syndrome: A prospective randomized trial in 225 consecutive patients. Eur Heart J 14(abstr suppl):252. Bernstein AD, Camm AJ, Fletcher RD, et al. 1987. The NASPE/BPEG generic pacemaker code for antibradyarrhythmia and adaptive-rate pacing and antitachyarrhythmia devices. PACE 10:794. Committee on Pacemaker Implantation. 1991. Guidelines for implantation of cardiac pacemakers and antiarrhythmic devices. J Am Coll Cardiol 18(1):1.


Further Information A good basic introduction to pacing from a clinical perspective is the third edition of A Practical Guide to Cardiac Pacing by H. Weston Moses, Joel Schneider, Brain Miller, and George Taylor (Little, Brown, 1991). Cardiac Pacing (Blackwell Scientific, 1992), edited by Kenneth Ellenbogen, is an excellent intermediate treatment of pacing. The treatments of timing cycles and troubleshooting are especially good. In-depth discussion of a wide range of pacing topics is provided by the third edition of A Practice of Cardiac Pacing by Seymour Furman, David Hayes, and David Holmes (Futura, 1993), and by New Perspectives in Cardiac Pacing 3, edited by Serge Barold and Jacques Mugica (Futura, 1993). Detailed treatment of rate-responsive pacing is given in Rate-Adaptive Cardiac Pacing: Single and Dual Chamber by Chu-Pak Lau (Futura, 1993), and in Rate-Adaptive Pacing, edited by David Benditt (Blackwell Scientific, 1993). The Foundations of Cardiac Pacing, Part I by Richard Sutton and Ivan Bourgeois (Futura, 1991) contains excellent illustrations of implantation techniques. Readers seeking a historical perspective may wish to consult “Pacemakers, Pastmakers, and the Paced: An Informal History from A to Z,” by Dwight Harken in the July/August 1991 issue of Biomedical Instrumentation and Technology. PACE is the official journal of the North American Society of Pacing and Electrophysiology (NASPE) and of the International Cardiac Pacing and Electrophysiology Society. It is published monthly by Futura Publishing (135 Bedford Road, PO Box 418, Armonk, NY 10504 USA).


Strojnik, P.,Peckham, P. H. “ Implantable Stimulators for Neuromuscular Control.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

78 Implantable Stimulators for Neuromuscular Control 78.1 78.2 78.3 78.4

Functional Electrical Stimulation Technology for Delivering Stimulation Pulses to Excitable Tissue Stimulation Parameters Implantable Neuromuscular Stimulators Receiving Circuit • Power Supply • Data Retrieval • Data Processing • Output Stage

78.5 78.6 78.7 78.8

Peripheral Nerve Stimulators • Stimulators of Central Nervous System

Primozˇ Strojnik Case Western Reserve University

78.9

P. Hunter Peckham Case Western Reserve University and Veterans Affairs Medical Center

Packaging of Implantable Electronics Leads and Electrodes Safety Issues of Implantable Stimulators Implantable Stimulators in Clinical Use Future of Implantable Electrical Stimulators Distributed Stimulators • Sensing of Implantable Transducer-Generated and Physiological Signals

78.10

Summary

78.1 Functional Electrical Stimulation Implantable stimulators for neuromuscular control are the technologically most advanced versions of functional electrical stimulators. Their function is to generate contraction of muscles, which cannot be controlled volitionally because of the damage or dysfunction in the neural paths of the central nervous system (CNS). Their operation is based on the electrical nature of conducting information within nerve fibers, from the neuron cell body (soma), along the axon, where a travelling action potential is the carrier of excitation. While the action potential is naturally generated chemically in the head of the axon, it may also be generated artificially by depolarizing the neuron membrane with an electrical pulse. A train of electrical impulses with certain amplitude, width, and repetition rate, applied to a muscle innervating nerve (a motor neuron) will cause the muscle to contract, very much like in natural excitation. Similarly, a train of electrical pulses applied to the muscular tissue close to the motor point will cause muscle contraction by stimulating the muscle through the neural structures at the motor point.

78.2 Technology for Delivering Stimulation Pulses to Excitable Tissue A practical system used to stimulate a nerve consists of three components: (1) a pulse generator to generate a train of pulses capable of depolarizing the nerve, (2) a lead wire, the function of which is to deliver the


pulses to the stimulation site, and (3) an electrode, which delivers the stimulation pulses to the excitable tissue in a safe and efficient manner. In terms of location of the above three components of an electrical stimulator, stimulation technology can be described in the following terms: Surface or transcutaneous stimulation, where all three components are outside the body and the electrodes are placed on the skin above or near the motor point of the muscle to be stimulated. This method has been used extensively in medical rehabilitation of nerve and muscle. Therapeutically, it has been used to prevent atrophy of paralyzed muscles, to condition paralyzed muscles before the application of functional stimulation, and to generally increase the muscle bulk. As a functional tool, it has been used in rehabilitation of plegic and paretic patients. Surface systems for functional stimulation have been developed to correct drop-foot condition in hemiplegic individuals [Liberson, 1961], for hand control [Rebersek, 1973], and for standing and stepping in individuals with paralysis of the lower extremities [Kralj and Bajd, 1989]. This fundamental technology was commercialized by Sigmedics, Inc. [Graupe, 1998]. The inability of surface stimulation to reliably excite the underlying tissue in a repeatable manner and to selectively stimulate deep muscles has limited the clinical applicability of surface stimulation. Percutaneous stimulation employs electrodes which are positioned inside the body close to the structures to be stimulated. Their lead wires permanently penetrate the skin to be connected to the external pulse generator. State of the art embodiments of percutaneous electrodes utilize a smalldiameter insulated stainless steel lead that is passed through the skin. The electrode structure is formed by removal of the insulation from the lead and subsequent modification to ensure stability within the tissue. This modification includes forming barbs or similar anchoring mechanisms. The percutaneous electrode is implanted using a hypodermic needle as a trochar for introduction. As the needle is withdrawn, the anchor at the electrode tip is engaged into the surrounding tissue and remains in the tissue. A connector at the skin surface, next to the skin penetration point, joins the percutaneous electrode lead to the hardwired external stimulator. The penetration site has to be maintained and care must be taken to avoid physical damage of the lead wires. In the past, this technology has helped develop the existing implantable systems, and it may be used for short and long term, albeit not permanent, stimulation applications [Memberg, 1993; Marsolais, 1986]. The term implantable stimulation refers to stimulation systems in which all three components, pulse generator, lead wires, and electrodes, are permanently surgically implanted into the body and the skin is solidly closed after the implantation procedure. Any interaction between the implantable part and the outside world is performed using telemetry principles in a contact-less fashion. This chapter is focused on implantable neuromuscular stimulators, which will be discussed in more detail.

78.3 Stimulation Parameters In functional electrical stimulation, the typical stimulation waveform is a train of rectangular pulses. This shape is used because of its effectiveness as well as relative ease of generation. All three parameters of a stimulation train, i.e., frequency, amplitude, and pulse-width, have effect on muscle contraction. Generally, the stimulation frequency is kept as low as possible, to prevent muscle fatigue and to conserve stimulation energy. The determining factor is the muscle fusion frequency at which a smooth muscle response is obtained. This frequency varies; however, it can be as low as 12 to 14 Hz and as high as 50 Hz. In most cases, the stimulation frequency is kept constant for a certain application. This is true both for surface as well as implanted electrodes. With surface electrodes, the common way of modulating muscle force is by varying the stimulation pulse amplitude at a constant frequency and pulse width. The stimulation amplitudes may be as low as 25 V at 200 µs for the stimulation of the peroneal nerve and as high as 120 V or more at 300 µs for activation of large muscles such as the gluteus maximus.


In implantable stimulators and electrodes, the stimulation parameters greatly depend on the implantation site. When the electrodes are positioned on or around the target nerve, the stimulation amplitudes are on the order of a few milliamperes or less. Electrodes positioned on the muscle surface (epimysial electrodes) or in the muscle itself (intramuscular electrodes), employ up to ten times higher amplitudes. For muscle force control, implantable stimulators rely either on pulse-width modulation or amplitude modulation. For example, in upper extremity applications, the current amplitude is usually a fixed paramter set to 16 or 20 mA, while the muscle force is modulated with pulse-widths within 0 to 200 µs.

78.4 Implantable Neuromuscular Stimulators Implantable stimulation systems use an encapsulated pulse generator that is surgically implanted and has subcutaneous leads that terminate at electrodes on or near the desired nerves. In low power consumption applications such as the cardiac pacemaker, a primary battery power source is included in the pulse generator case. When the battery is close to depletion, the pulse generator has to be surgically replaced. Most implantable systems for neuromuscular application consist of an external and an implanted component. Between the two, an inductive radio-frequency link is established, consisting of two tightly coupled resonant coils. The link allows transmission of power and information, through the skin, from the external device to the implanted pulse generator. In more advanced systems, a back-telemetry link is also established, allowing transmission of data outwards, from the implanted to the external component. Ideally, implantable stimulators for neuromuscular control would be stand alone, totally implanted devices with an internal power source and integrated sensors detecting desired movements from the motor cortex and delivering stimulation sequences to appropriate muscles, thus bypassing the neural damage. At the present developmental stage, they still need a control source and an external controller to provide power and stimulation information. The control source may be either operator driven, controlled by the user, or triggered by an event such as the heel-strike phase of the gait cycle. Figure 78.1 depicts a neuromuscular prosthesis developed at the Case Western Reserve University (CWRU) and Cleveland Veterans Affairs Medical Center for the restoration of hand functions using an implantable neuromuscular stimulator. In this application, the patient uses the shoulder motion to control opening and closing of the hand. The internal electronic structure of an implantable neuromuscular stimulator is shown in Fig. 78.2. It consists of receiving and data retrieval circuits, power supply, data processing circuits, and output stages.

FIGURE 78.1


Implanted FES hand grasp system.

FIGURE 78.2

Block diagram of an implantable neuromuscular stimulator.

Receiving Circuit The stimulator’s receiving circuit is an LC circuit tuned to the resonating frequency of the external transmitter, followed by a rectifier. Its task is to provide the raw DC power from the received rf signal and at the same time allow extraction of stimulation information embedded in the rf carrier. There are various encoding schemes allowing simultaneous transmission of power and information into an implantable electronic device. They include amplitude and frequency modulation with different modulation indexes as well as different versions of digital encoding such as Manchester encoding where the information is hidden in a logic value transition position rather than the logic value itself. Synchronous and asynchronous clock signals may be extracted from the modulated carrier to drive the implant’s logic circuits. The use of radiofrequency transmission for medical devices is regulated and in most countries limited to certain frequencies and radiation powers. (In the U.S., the use of the rf space is regulated by the Federal Communication Commission [FCC]). Limited rf transmission powers as well as conservation of power in battery operated external controllers dictate high coupling efficiencies between the transmitting and receiving antennas. Optimal coupling parameters cannot be uniformly defined; they depend on application particularities and design strategies.

Power Supply The amount of power delivered into an implanted electronic package depends on the coupling between the transmitting and the receiving coil. The coupling is dependent on the distance as well as the alignment between the coils. The power supply circuits must compensate for the variations in distance for different users as well as for the alignment variations due to skin movements and consequent changes in relative coil-to-coil position during daily usage. The power dissipated on power supply circuits must not raise the overall implant case temperature. In implantable stimulators that require stimulation voltages in excess of the electronics power supply voltages (20 to 30 V), the stimulation voltage can be provided directly through the receiving coil. In that case, voltage regulators must be used to provide the electronics supply voltage (usually 5 V), which heavily taxes the external power transmitter and increases the implant internal power dissipation.

Data Retrieval Data retrieval technique depends on the data-encoding scheme and is closely related to power supply circuits and implant power consumption. Most commonly, amplitude modulation is used to encode the in-going data stream. As the high quality factor of resonant LC circuits increases the efficiency of power transmission, it also effectively reduces the transmission bandwidth and therefore the transmission data © 2000 by CRC Press LLC

rate. Also, high quality circuits are difficult to amplitude modulate since they tend to continue oscillating even with power removed. This has to be taken into account when designing the communication link in particular for the start-up situation when the implanted device does not use the power for stimulation and therefore loads the transmitter side less heavily, resulting in narrower and higher resonant curves. The load on the receiving coil may also affect the low pass filtering of the received rf signal. Modulation index (m) or depth of modulation affects the overall energy transfer into the implant. At a given rf signal amplitude, less energy is transferred into the implanted device when 100% modulation is used (m = 1) as compared to 10% modulation (m = 0.053). However, retrieval of 100% modulated signal is much easier than retrieval of a 10% modulated signal.

Data Processing Once the information signal has been satisfactorily retrieved and reconstructed into logic voltage levels, it is ready for logic processing. For synchronous data processing a clock signal is required. It can be generated locally within the implant device, reconstructed from the incoming data stream, or can be derived from the rf carrier. A crystal has to be used with a local oscillator to assure stable clock frequency. Local oscillator allows for asynchronous data transmission. Synchronous transmission is best achieved using Manchester data encoding. Decoding of Manchester encoded data recovers the original clock signal, which was used during data encoding. Another method is using the downscaled rf carrier signal as the clock source. In this case, the information signal has to be synchronized with the rf carrier. Of course, 100% modulation scheme cannot be used with carrier-based clock signal. Complex command structure used in multichannel stimulators requires intensive data decoding and processing and consequently extensive electronic circuitry. Custom-made, application specific circuits (ASIC) are commonly used to minimize the space requirements and optimize the circuit performance.

Output Stage The output stage forms stimulation pulses and defines their electrical characteristics. Even though a mere rectangular pulse can depolarize a nervous membrane, such pulses are not used in clinical practice due to their noxious effect on the tissue and stimulating electrodes. These effects can be significantly reduced by charge balanced stimulating pulses where the cathodic stimulation pulse is followed by an anodic pulse containing the same electrical charge, which reverses the electrochemical effects of the cathodic pulse. Charge balanced waveforms can be assured by capacitive coupling between the pulse generator and stimulation electrodes. Charge balanced stimulation pulses include symmetrical and asymmetrical waveforms with anodic phase immediately following the cathodic pulse or being delayed by a short, 20 to 60 µs interval. The output stages of most implantable neuromuscular stimulators have constant current characteristics, meaning that the output current is independent on the electrode or tissue impedance. Practically, the constant current characteristics ensure that the same current flows through the excitable tissues regardless of the changes that may occur on the electrode-tissue interface, such as the growth of fibrous tissue around the electrodes. Constant current output stage can deliver constant current only within the supply voltage—compliance voltage. In neuromuscular stimulation, with the electrode impedance being on the order of 1 kΩ , and the stimulating currents in the order of 20 mA, the compliance voltage must be above 20 V. Considering the voltage drops and losses across electronic components, the compliance voltage of the output stage may have to be as high as 33 V. The stimulus may be applied through either monopolar or bipolar electrodes. The monopolar electrode is one in which a single active electrode is placed near the excitable nerve and the return electrode is placed remotely, generally at the implantable unit itself. Bipolar electrodes are placed at the stimulation site, thus limiting the current paths to the area between the electrodes. Generally, in monopolar stimulation the active electrode is much smaller than the return electrode, while bipolar electrodes are the same size.


78.5 Packaging of Implantable Electronics Electronic circuits must be protected from the harsh environment of the human body. The packaging of implantable electronics uses various materials, including polymers, metals, and ceramics. The encapsulation method depends somewhat on the electronic circuit technology. Older devices may still use discrete components in a classical form, such as leaded transistors and resistors. The newer designs, depending on the sophistication of the implanted device, may employ application-specific integrated circuits (ASICs) and thick film hybrid circuitry for their implementation. Such circuits place considerable requirements for hermeticity and protection on the implanted circuit packaging. Epoxy encapsulation was the original choice of designers of implantable neuromuscular stimulators. It has been successfully used with relatively simple circuits using discrete, low impedance components. With epoxy encapsulation, the receiving coil is placed around the circuitry to be “potted” in a mold, which gives the implant the final shape. Additionally, the epoxy body is coated with silicone rubber that improves the biocompatibility of the package. Polymers do not provide an impermeable barrier and therefore cannot be used for encapsulation of high density, high impedance electronic circuits. The moisture ingress ultimately will reach the electronic components, and surface ions can allow electric shorting and degradation of leakage-sensitive circuitry and subsequent failure. Hermetic packaging provides the implant electronic circuitry with a long-term protection from the ingress of body fluids. Materials that provide hermetic barriers are metals, ceramics, and glasses. Metallic packaging generally uses a titanium capsule machined from a solid piece of metal or deep-drawn from a piece of sheet metal. Electrical signals, such as power and stimulation, enter and exit the package through hermetic feedthroughs, which are hermetically welded onto the package walls. The feedthrough assembly utilizes a ceramic or glass insulator to allow one or more wires to exit the package without contact with the package itself. During the assembly procedures, the electronic circuitry is placed in the package and connected internally to the feedthroughs, and the package is then welded closed. Tungsten Inert Gas (TIG), electron beam, or laser welding equipment is used for the final closure. Assuming integrity of all components, hermeticity with this package is ensured. This integrity can be checked by detecting gas leakage from the capsule. Metallic packaging requires that the receiving coil be placed outside the package to avoid significant loss of rf signal or power, thus requiring additional space within the body to accommodate the volume of the entire implant. Generally, the hermetic package and the receiving antenna are jointly imbedded in an epoxy encapsulant, which provides electric isolation for the metallic antenna and stabilizes the entire implant assembly. Figure 78.3 shows such an implantable

FIGURE 78.3 Photograph of a multichannel implantable stimulator telemeter. Hybrid circuit in titanium package is shown exposed. Receiving coil (left) is imbedded in epoxy resin together with titanium case. Double feedthroughs are seen penetrating titanium capsule wall on the right.


stimulator designed and made by the CWRU/Veterans Administration Program. The hermetic package is open, displaying the electronic hybrid circuit. More recently, alumina-based ceramic packages have been developed that allow hermetic sealing of the electronic circuitry together with enclosure of the receiving coil [Strojnik, 1994]. This is possible due to the rf transparency of ceramics. The impact of this type of enclosure is still not fully investigated. The advantage of this approach is that the volume of the implant can be reduced, thus minimizing the biologic response, which is a function of volume. Yet, an unexplored issue of this packaging method is the effect of powerful electromagnetic fields on the implant circuits, lacking the protection of the metal enclosure. This is a particular concern with high gain (EMG, ENG, or EKG sensing) amplifiers, which in the future may be included in the implant package as part of back-telemetry circuits. Physical strength of ceramic packages and their resistance to impact will also require future investigation.

78.6 Leads and Electrodes Leads connect the pulse generator to the electrodes. They must be sufficiently flexible to move across the joints while at the same time sufficiently sturdy to last for the decades of the intended life of the device. They must also be stretchable to allow change of distance between the pulse generator and the electrodes, associated with body movements. Ability to flex and to stretch is achieved by coiling the lead conductor into a helix and inserting the helix into a small-diameter silicone tubing. This way, both flexing movements and stretching forces exerted on the lead are attenuated, while translated into torsion movements and forces exerted on the coiled conductor. Using multi-strand rather than solid conductors further enhances the longevity. Several individually insulated multi-strand conductors can be coiled together, thus forming a multiple conductor lead wire. Most lead configurations include a connector at some point between the implant and the terminal electrode, allowing for replacement of the implanted receiver or leads in the event of failure. The connectors used have been either single pin in-line connectors located somewhere along the lead length or a multiport/multilead connector at the implant itself. Materials used for lead wires are stainless steels, MP35N (Co, Cr, Ni alloy), and noble metals and their alloys. Electrodes deliver electrical charge to the stimulated tissues. Those placed on the muscle surface are called epimysial, while those inserted into the muscles are called intramuscular. Nerve stimulating electrodes are called epineural when placed against the nerve, or cuff electrodes when they encircle the nerve. Nerve electrodes may embrace the nerve in a spiral manner individually, or in an array configuration. Some implantable stimulation systems merely use exposed lead-wire conductor sutured to the epineurium as the electrode. Generally, nerve electrodes require approximately one-tenth of the energy for muscle activation as compared to muscle electrodes. However, they require more extensive surgery and may be less selective, but the potential for neural damage is greater than, for example, nerve encircling electrodes. Electrodes are made of corrosion resistant materials, such as noble metals (platinum or iridium) and their alloys. For example, a platinum–iridium alloy consisting of 10% iridium and 90% platinum is commonly used as an electrode material. Epimysial electrodes developed at CWRU use Ø4 mm Pt90Ir10 discs placed on Dacron reinforced silicone backing. CWRU intramuscular electrodes employ a stainless steel lead-wire with the distal end de-insulated and configured into an electrode tip. A small, umbrellalike anchoring barb is attached to it. With this arrangement, the diameter of the electrode tip does not differ much from the lead wire diameter and this electrode can be introduced into a deep muscle with a trochar-like insertion tool. Figure 78.4 shows enlarged views of these electrodes.

78.7 Safety Issues of Implantable Stimulators The targeted lifetime of implantable stimulators for neuromuscular control is the lifetime of their users, which is measured in tens of years. Resistance to premature failure must be assured by manufacturing processes and testing procedures. Appropriate materials must be selected that will withstand the working


FIGURE 78.4 Implantable electrodes with attached lead wires. Intramuscular electrode (top) has stainless steel tip and anchoring barbs. Epimysial electrode has PtIr disk in the center and is backed by silicone-impregnated Dacron mesh.

environment. Protection against mechanical and electrical hazards that may be encountered during the device lifetime must be incorporated in the design. Various procedures are followed and rigorous tests must be performed during and after its manufacturing to assure the quality and reliability of the device. • Manufacturing and testing—Production of implantable electronic circuits and their encapsulation in many instances falls under the standards governing production and encapsulation of integrated circuits. To minimize the possibility of failure, the implantable electronic devices are manufactured in controlled clean-room environments, using high quality components and strictly defined manufacturing procedures. Finished devices are submitted to rigorous testing before being released for implantation. Also, many tests are carried out during the manufacturing process itself. To assure maximum reliability and product confidence, methods, tests, and procedures defined by military standards, such as MIL-STD-883, are followed. • Bio-compatibility—Since the implantable stimulators operate surgically implanted in living tissue, an important part of their design has to be dedicated to biocompatibility, i.e., their ability to dwell in living tissue without disrupting the tissue in its functions, creating adverse tissue response, or changing its own properties due to the tissue environment. Elements of biocompatibility include tissue reaction to materials, shape, and size, as well as electrochemical reactions on stimulation electrodes. There are known biomaterials used in the making of implantable stimulators. They include stainless steels, titanium and tantalum, noble metals such as platinum and iridium, as well as implantable grades of selected epoxy and silicone-based materials. • Susceptibility to electromagnetic interference (EMI) and electrostatic discharge (ESD)—Electromagnetic fields can disrupt the operation of electronic devices, which may be lethal in situations with life support systems, but they may also impose risk and danger to users of neuromuscular stimulators. Emissions of EMI may come from outside sources; however, the external control unit is also a source of electromagnetic radiation. Electrostatic discharge shocks are not uncommon during the dry winter season. These shocks may reach voltages as high as 15 kV and more. Sensitive © 2000 by CRC Press LLC

electronic components can easily be damaged by these shocks unless protective design measures are taken. The electronic circuitry in implantable stimulators is generally protected by the metal case. However, the circuitry can be damaged through the feedthroughs either by handling or during the implantation procedure by the electrocautery equipment. ESD damage may happen even after implantation when long lead-wires are utilized. There are no standards directed specifically towards implantable electronic devices. The general standards put in place for electromedical equipment by the International Electrotechnical Commission provide guidance. The specifications require survival after 3 kV and 8 kV ESD discharges on all conductive and nonconductive accessible parts, respectively.

78.8 Implantable Stimulators in Clinical Use Peripheral Nerve Stimulators • Manipulation—Control of complex functions for movement, such as hand control, requires the use of many channels of stimulation. At the Case Western Reserve University and Cleveland VAMC, an eight-channel stimulator has been developed for grasp and release [Smith, 1987]. This system uses eight channels of stimulation and a titanium-packaged, thick-film hybrid circuit as the pulse generator. The implant is distributed by the Neurocontrol Corporation (Cleveland, OH) under the name of Freehand®. It has been implanted in approximately 150 patients in the U.S., Europe, Asia, and Australia. The implant is controlled by a dual-microprocessor external unit carried by the patient with an input control signal provided by the user’s remaining volitional movement. Activation of the muscles provides two primary grasp patterns and allows the person to achieve functional performance that exceeds his or her capabilities without the use of the implanted system. This system received pre-market approval from the FDA in 1998. • Locomotion—The first implantable stimulators were designed and implanted for the correction of the foot drop condition in hemiplegic patients. Medtronic’s Neuromuscular Assist (NMA) device consisted of an rf receiver implanted in the inner thigh and connected to a cuff electrode embracing the peroneal nerve just beneath the head of fibula at the knee [McNeal, 1977; Waters 1984]. The Ljubljana peroneal implant had two versions [Vavken, 1976; Strojnik,1987] with the common feature that the implant–rf receiver was small enough to be implanted next to the peroneal nerve in the fossa poplitea region. Epineural stimulating electrodes were an integral part of the implant. This feature and the comparatively small size make the Ljubljana implant a precursor of the micro-stimulators described in Section 78.9. Both NMA and the Ljubljana implants were triggered and synchronized with gait by a heel switch. The same implant used for hand control and developed by the CWRU has also been implanted in the lower extremity musculature to assist incomplete quadriplegics in standing and transfer operations [Triolo, 1996]. Since the design of the implant is completely transparent, it can generate any stimulation sequence requested by the external controller. For locomotion and transfer-related tasks, stimulation sequences are preprogrammed for individual users and activated by the user by means of pushbuttons. The implant (two in some applications) is surgically positioned in the lower abdominal region. Locomotion application uses the same electrodes as the manipulation system; however, the lead wires have to be somewhat longer. • Respiration—Respiratory control systems involve a two-channel implantable stimulator with electrodes applied bilaterally to the phrenic nerve. Most of the devices in clinical use were developed by Avery Laboratories (Dobelle Institute) and employed discrete circuitry with epoxy encapsulation of the implant and a nerve cuff electrode. Approximately 1000 of these devices have been implanted in patients with respiratory disorders such as high-level tetraplegia [Glenn, 1986]. Activation of the phrenic nerve results in contraction of each hemidiaphragm in response to electrical stimulation. In order to minimize damage to the diaphragms during chronic use, alternation of the diaphragms has been employed, in which one hemidiaphragm will be activated for © 2000 by CRC Press LLC

several hours followed by the second. A review of existing systems was given by Creasy et al. [1996]. Astrotech of Finland also recently introduced a phrenic stimulator. More recently, DiMarco [1997] has investigated use of CNS activation of a respiratory center to provide augmented breathing. • Urinary control—Urinary control systems have been developed for persons with spinal cord injury. The most successful of these devices has been developed by Brindley [1982] and is manufactured by Finetech, Ltd. (England). The implanted receiver consists of three separate stimulator devices, each with its own coil and circuitry, encapsulated within a single package. The sacral roots (S2, S3, and S4) are placed within a type of encircling electrode, and stimulation of the proper roots will generate contraction of both the bladder and the external sphincter. Cessation of stimulation results in faster relaxation of the external sphincter than of the bladder wall, which then results in voiding. Repeated trains of pulses applied in this manner will eliminate most urine, with only small residual amounts remaining. Approximately 1500 of these devices have been implanted around the world. This technology also has received FDA pre-market approval and is currently distributed by NeuroControl Corporation. • Scoliosis treatment—Progressive lateral curvature of the adolescent vertebral column with simultaneous rotation is known as idiopathic scoliosis. Electrical stimulation applied to the convex side of the curvature has been used to stop or reduce its progression. Initially rf powered stimulators have been replaced by battery powered totally implanted devices [Bobechko, 1979; Herbert, 1989]. Stimulation is applied intermittently, stimulation amplitudes are under 10.5 V (510 Ω), and frequency and pulsewidth are within usual FES parameter values.

Stimulators of Central Nervous System Some stimulation systems have electrodes implanted on the surface of the central nervous system or in its deep areas. They do not produce functional movements; however, they “modulate” a pathological motor brain behavior and by that stop unwanted motor activity or abnormality. Therefore, they can be regarded as stimulators for neuromuscular control. • Cerebellar stimulation—Among the earliest stimulators from this category are cerebellar stimulators for control of reduction of effects of cerebral palsy in children. Electrodes are placed on the cerebellar surface with the leads penetrating cranium and dura. The pulse generator is located subcutaneously in the chest area and produces intermittent stimulation bursts. There are about 600 patients using these devices [Davis, 1997]. • Vagal stimulation—Intermittent stimulation of the vagus nerve with 30 sec on and five min off has been shown to reduce frequency of epileptic seizures. A pacemaker-like device, developed by Cyberonics, is implanted in the chest area with a bipolar helical electrode wrapped around the left vagus nerve in the neck. The stimulation sequence is programmed (most often parameter settings are 30 Hz, 500 µs, 1.75 mA); however, patients have some control over the device using a hand-held magnet [Terry, 1991]. More than 3000 patients have been implanted with this device, which received the pre-marketing approval (PMA) from the FDA in 1997. • Deep brain stimulation—Recently, in 1998, an implantable stimulation device (Activa by Medtronic) was approved by the FDA that can dramatically reduce uncontrollable tremor in patients with Parkinson’s disease or essential tremor [Koller, 1997]. With this device, an electrode array is placed stereotactically into the ventral intermediate nucleus of thalamic region of the brain. Lead wires again connect the electrodes to a programmable pulse generator implanted in the chest area. Application of high frequency stimulation (130 Hz, 60 to 210 µs, 0.25 to 2.75 V) can immediately suppress the patient’s tremor.


FIGURE 78.5 Microstimulator developed at A.E. Mann Foundation. Dimensions are roughly 2 × 16 mm. Electrodes at the ends are made of tantalum and iridium, respectively.

78.9 Future of Implantable Electrical Stimulators Distributed Stimulators One of the major concerns with multichannel implantable neuromuscular stimulators is the multitude of leads that exit the pulse generator and their management during surgical implantation. Routing of multiple leads virtually increases the implant size and by that the burden that an implant imposes on the tissue. A solution to that may be distributed stimulation systems with a single outside controller and multiple single-channel implantable devices implanted throughout the structures to be stimulated. This concept has been pursued both by the Alfred E. Mann Foundation [Strojnik, 1992; Cameron, 1997] and the University of Michigan [Ziaie, 1997]. Micro-injectable stimulator modules have been developed that can be injected into the tissue, into a muscle, or close to a nerve through a lumen of a hypodermic needle. A single external coil can address and activate a number of these devices located within its field, on a pulse-to-pulse basis. A glass-encapsulated microstimulator developed at the AEMF is shown in Fig. 78.5.

Sensing of Implantable Transducer–Generated and Physiological Signals External command sources such as the shoulder-controlled joystick utilized by the Freehand® system impose additional constraints on the implantable stimulator users, since they have to be donned by an attendant. Permanently implanted control sources make neuro-prosthetic devices much more attractive and easier to use. An implantable joint angle transducer (IJAT) has been developed at the CWRU that consists of a magnet and an array of magnetic sensors implanted in the distal and the proximal end of a joint, respectively [Smith, 1998]. The sensor is connected to the implantable stimulator package, which provides the power and also transmits the sensor data to the external controller, using a back-telemetry link. Figure 78.6 shows a radiograph of the IJAT implanted in a patient’s wrist. Myoelectric signals (MES) from muscles not affected by paralysis are another attractive control source for implantable neuromuscular stimulators. Amplified and bin-integrated EMG signal from uninvolved muscles, such as the sternocleido-mastoid muscle, has been shown to contain enough information to control an upper extremity neuroprosthesis [Scott, 1996]. EMG signal is being utilized by a multichannel stimulator-telemeter developed at the CWRU, containing 12 stimulator channels and 2 MES channels integrated into the same platform [Strojnik, 1998].


FIGURE 78.6 Radiograph of the joint angle transducer (IJAT) implanted in the wrist. The magnet is implanted in the lunate bone (left) while the magnetic sensor array is implanted in the radius. Leads going to the implant case can be seen as well as intramuscular and epimysial electrodes with their individual lead wires.

78.10

Summary

Implantable stimulators for neuromuscular control are an important tool in rehabilitation of paralyzed individuals with preserved neuro-muscular apparatus, as well as in the treatment of some neurological disorders that result in involuntary motor activity. Their impact on rehabilitation is still in its infancy; however, it is expected to increase with further progress in microelectronics technology, development of smaller and better sensors, and with improvements of advanced materials. Advancements in neurophysiological science are also expected to bring forward wider utilization of possibilities offered by implantable neuromuscular stimulators. © 2000 by CRC Press LLC

Defining Terms Biocompatibility: Ability of a foreign object to coexist in a living tissue. Electrical stimulation: Diagnostic, therapeutic, and rehabilitational method used to excite motor nerves with the aim of contracting the appropriate muscles and obtain limb movement. EMG activity: Muscular electrical activity associated with muscle contraction and production of force. Feedthrough: Device that allows passage of a conductor through a hermetic barrier. Hybrid circuit: Electronic circuit combining miniature active and passive components on a single ceramic substrate. Implantable stimulator: Biocompatible electronic stimulator designed for surgical implantation and operation in a living tissue. Lead wire: Flexible and strong insulated conductor connecting pulse generator to stimulating electrodes. Paralysis: Loss of power of voluntary movement in a muscle through injury to or disease to its nerve supply. Stimulating electrode: Conductive device that transfers stimulating current to a living tissue. On its surface, the electric charge carriers change from electrons to ions or vice versa. rf-radiofrequency: Pertaining to electromagnetic propagation of power and signal in frequencies above those used in electrical power distribution.

References Bobechko WP, Herbert MA, Friedman HG. 1979. Electrospinal instrumentation for scoliosis: Current status. Orthop Clin North Am 10(4):927. Brindley GS, Polkey CE, Rushton DN. 1982. Sacral anterior root stimulators for bladder control in paraplegia. Paraplegia 20(6):365. Cameron T, Loeb GE, Peck RA, Schulman JH, Strojnik P, Troyk PR. 1997. Micromodular implants to provide electrical stimulation of paralyzed muscles and limbs. IEEE Trans Biomed Eng 44(9):781. Creasey G, Elefteriades J, DiMarco A, Talonen P, Bijak M, Girsch W, Kantor C. 1996. Electrical stimulation to restore respiration. J Rehab Res Dev 33(2):123. Davis R.: Cerebellar stimulation for movement disorders. 1997. In PL Gildenberg and RR Tasker (eds), Textbook of Stereotactic and Functional Neurosurgery, McGraw-Hill, New York. DiMarco AF, Romaniuk JR, Kowalski KE, Supinski GS. 1997. Efficacy of combined inspiratory intercostal and expiratory muscle pacing to maintain artificial ventilation. Am J Respir Crit Care Med 156(1):122. Glenn WW, Phelps ML, Elefteriades JA, Dentz B, Hogan JF. 1986. Twenty years of experience in phrenic nerve stimulation to pace the diaphragm pacing. Clin Electrophysiol 9(6 Pt 1):780. Graupe D, Kohn KH. 1998. Functional neuromuscular stimulator for short-distance ambulation by certain thoracic-level spinal-cord-injured paraplegics. Surg Neurol 50(3):202. Herbert MA, Bobechko WP. 1989. Scoliosis treatment in children using a programmable, totally implantable muscle stimulator (ESI). IEEE Trans Biomed Eng 36(7):801. Koller W, Pahwa R, Busenbark K, Hubble J, Wilkinson S, Lang A, Tuite P, Sime E, Lazano A, Hauser R, Malapira T, Smith D, Tarsy D, Miyawaki E, Norregaard T, Kormos T, Olanow CW. 1997. Highfrequency unilateral thalamic stimulation in the treatment of essential and parkinsonian tremor. Ann Neurol 42(3):292. Kralj A, Bajd T. 1989. Functional electrical stimulation: Standing and walking after spinal cord injury, CRC Press, Inc., Boca Raton, FL. Liberson WT, Holmquest, HJ Scot D, Dow M. 1961. Functional electrotherapy: Stimulation of the peroneal nerve synchronized with the swing phase of the gait of hemiplegic patients. Arch Phys Med & Reb 42:101. Marsolais EB, Kobetic R. 1986. Implantation techniques and experience with percutaneous intramuscular electrodes in lower extremities. J Rehab Res Dev 23(3):1. McNeal DR, Waters R, Reswick J. 1977. Experience with implanted electrodes, Neurosurgery 1(2):228. © 2000 by CRC Press LLC

Memberg, W, Peckham PH, Thorpe GB, Keith, MW, Kicher TP. 1993. An analysis of the reliability of percutaneous intramuscular electrodes in upper extremity FNS applications. IEEE Trans Biomed Eng, 1(2):126. Rebersek S, Vodovnik L. 1973. Proportionally controlled functional electrical stimulation of hand. Arch Phys Med Rehab 54:378. Scott TRD., Peckham PH., Kilgore KL. 1996. Tri-State myoelectric control of bilateral upper extremity neuroprosthesies for tetraplegic individuals. IEEE Trans Rehab Eng 2:251. Smith B, Peckham PH, Keith, MW, Roscoe DD. 1987. An externally powered, multichannel, implantable stimulator for versatile control of paralyzed muscle. IEEE Trans Biomed Engr 34:499. Smith B, Tang, Johnson MW, Pourmehdi S, Gazdik MM, Buckett JR, Peckham PH. 1998. An externally powered, multichannel, implantable stimulator-telemeter for control of paralyzed muscle. IEEE Trans Biomed Eng 45(4):463. Strojnik P, Pourmehdi S, Peckham P. 1998. Incorporating FES control sources into implanatable stimulators. Proc 6th Vienna International Workshop on Functional Electrostimulation, Vienna, Austria. Strojnik P, Meadows P, Schulman JH, Whitmoyer D. 1994. Modification of a cochlear stimulation system for FES applications. Basic & Applied Myology, BAM 4(2): 129. Strojnik P, Acimovic R, Vavken E, Simic V, Stanic U. 1987. Treatment of drop foot using an implantable peroneal underknee stimulator. Scand J Rehab Med 19:37. Strojnik P, Schulman J, Loeb G, Troyk P. 1992. Multichannel FES system with distributed microstimulators. Proc 14th Ann Int Conf IEEE, MBS, Paris, pp 1352. Terry RS, Tarver WB, Zabara J. 1991. The implantable neurocybernetic prosthesis system. Pacing Clin Electrophysiol 14(1):86. Triolo RJ, Bieri C, Uhlir J, Kobetic R, Scheiner A, Marsolais EB. 1996. Implanted functional neuromuscular stimulation systems for individuals with cervical spinal cord injuries: Clinical case reports. Arch Phys Med Rehabil 77(11):1119. Vavken E, Jeglic A. 1976. Application of an implantable stimulator in the rehabilitation of paraplegic patients. Int Surg 61(6-7):335-9. Waters RL, McNeal DR, Clifford B. 1984. Correction of footdrop in stroke patients via surgically implanted peroneal nerve stimulator. Acta Orthop Belg 50(2):285. Ziaie B, Nardin MD, Coghlan AR, Najafi K. 1997. A single-channel implantable microstimulator for functional neuromuscular stimulation. IEEE Trans Biomed Eng 44(10):909.

Further Information Additional references on early work in FES which augment peer review publications can be found in Proceedings from Conferences in Dubrovnik and Vienna. These are the External Control of Human Extremities and the Vienna International Workshop on Electrostimulation, respectively.


Tacker, W. A. “ External Defibrillators.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

79 External Defibrillators

Willis A. Tacker Purdue University

79.1 79.2 79.3 79.4 79.5 79.6 79.7

Mechanism of Fibrillation Mechanism of Defibrillation Clinical Defibrillators Electrodes. Synchronization Automatic External Defibrillators Defibrillator Safety

Defibrillators are devices used to supply a strong electric shock (often referred to as a countershock) to a patient in an effort to convert excessively fast and ineffective heart rhythm disorders to slower rhythms that allow the heart to pump more blood. External defibrillators have been in common use for many decades for emergency treatment of life-threatening cardiac rhythms as well as for elective treatment of less threatening rapid rhythms. Figure 79.1 shows an external defibrillator. Cardiac arrest occurs in more than 500,000 people annually in the United States, and more than 70% of the out-of-hospitals are due to cardiac arrhythmia treatable with defibrillators. The most serious arrhythmia treated by a defibrillator is ventricular fibrillation. Without rapid treatment using a defibrillator, ventricular fibrillation causes complete loss of cardiac function and death within minutes. Atrial fibrillation and the more organized rhythms of atrial flutter and ventricular tachycardia can be treated on a less emergent basis. Although they do not cause immediate death, their shortening of the interval between contractions can impair filling of the heart chambers and thus decrease cardiac output. Conventionally, treatment of ventricular fibrillation is called defibrillation, whereas treatment of the other tachycardias is called cardioversion.

79.1 Mechanism of Fibrillation Fibrillation is chaotic electric excitation of the myocardium and results in loss of coordinated mechanical contraction characteristic of normal heart beats. Description of mechanisms leading to, and maintaining, fibrillation and other rhythm disorders are reviewed elsewhere [1] and are beyond the scope of this chapter. In summary, however, these rhythm disorders are commonly held to be a result of reentrant excitation pathways within the heart. The underlying abnormality that leads to the mechanism is the combination of conduction block of cardiac excitation plus rapidly recurring depolarization of the membranes of the cardiac cells. This leads to rapid repetitive propagation of a single excitation wave or of multiple excitatory waves throughout the heart. If the waves are multiple, the rhythm may degrade into total loss of synchronization of cardiac fiber contraction. Without synchronized contraction, the chamber affected will not contract, and this is fatal in the case of ventricular fibrillation. The most common cause of these conditions, and therefore of these rhythm disorders, is cardiac ischemia or infarction as a complication of atherosclerosis. Additional relatively common causes include other cardiac


FIGURE 79.1

Photograph of a trans-chest defibrillator (provided by Physio-Control Corporation with permission).

disorders, drug toxicity, electrolyte imbalances in the blood, hypothermia, and electric shocks (especially from alternating current).

79.2 Mechanism of Defibrillation The corrective measure is to extinguish the rapidly occurring waves of excitation by simultaneously depolarizing most of the cardiac cells with a strong electric shock. The cells then can simultaneously repolarize themselves, and thus they will be back in phase with each other. Despite years of intensive research, there is still no single theory for the mechanism of defibrillation that explains all the phenomena observed. However, it is generally held that the defibrillating shock must be adequately strong and have adequate duration to affect most of the heart cells. In general, longer duration shocks require less current than shorter duration shocks. This relationship is called the strengthduration relationship and is demonstrated by the curve shown in Fig. 79.2. Shocks of strength and duration above and to the right of the current curve (or above the energy curve) have adequate strength to defibrillate, whereas shocks below and to the left do not. From the exponentially decaying current curve an energy curve can also be determined (also shown in Fig. 79.2), which is high at very short durations due to high current requirements at short durations, but which is also high at longer durations due to additional energy being delivered as the pulse duration is lengthened at nearly constant current. Thus, for most electrical waveforms there is a minimum energy for defibrillation at approximate pulse durations of 3–8 ms. A strength-duration charge curve can also be determined as shown in Fig. 79.2, which demonstrates that the minimum charge for defibrillation occurs at the shortest pulse duration


FIGURE 79.2 Strength-duration curves for current, energy, and charge. Adequate current shocks are above and to the right of the current curve. (Modified from Tacker WA, Geddes LA, 1980. Electrical Defibrillation, Boca Raton, Fla, CRC Press, with permission.)

tested. Very-short-duration pulses are not used, however, since the high current and voltage required is damaging to the myocardium. It is also important to note that excessively strong or long shocks may cause immediate refibrillation, thus failing to restore the heart function. In practice, for a shock applied to electrodes on the skin surface of the patient’s chest, durations are on the order of 3–10 milliseconds and have an intensity of a few thousand volts and tens of amperes. The energy delivered to the subject by these shocks is selectable by the operator and is on the order of 50–360 joules for most defibrillators. The exact shock intensity required at a given duration of electric pulse depends on several variables, including the intrinsic characteristics of the patient (such as the underlying disease problem or presence of certain drugs and the length of time the arrhythmia has been present), the techniques for electrode application, and the particular rhythm disorder being treated (more organized rhythms require less energy than disorganized rhythms).

79.3 Clinical Defibrillators Defibrillator design has resulted from medical and physiologic research and advances in hardware technology. It is estimated that for each minute that elapses between onset of ventricular fibrillation and the first shock application, survival to leave hospital decreases by about 10%. The importance of rapid response led to development of portable, battery-operated defibrillators and more recently to automatic external defibrillators (AEDs) that enable emergency responders to defibrillate with minimal training. All clinical defibrillators used today store energy in capacitors. Desirable capacitor specifications include small size, light weight, and capability to sustain several thousands of volts and many chargedischarge cycles. Energy storage capacitors account for at least one pound and usually several pounds of defibrillator weight. Energy stored by the capacitor is calculated from

Ws =


1 CE 2 2

(79.1)

FIGURE 79.3 Block diagram of a typical defibrillator. (From Feinberg B. 1980. Handbook Series in Clinical Laboratory Science, vol 2, Boca Raton, Fla, CRC Press, with permission.)

where WS = stored energy in joules, C = capacitance in farads, and E = voltage applied to the capacitor. Delivered energy is expressed as

 R  Wd = WS ×    Ri + R 

(79.2)

where Wd = delivered energy, WS = stored energy, R = subject resistance, and Ri = device resistance. Figure 79.3 shows a block diagram for defibrillators. Most have a built-in monitor and synchronizer (dashed lines in Fig. 79.3). Built-in monitoring speeds up diagnosis of potentially fatal arrhythmias, especially when the ECG is monitored through the same electrodes that are used to apply the defibrillating shock. The great preponderance of defibrillators for trans-chest defibrillation deliver shocks with either a damped sinusoidal waveform produced by discharge of an RCL circuit or a truncated exponential decay waveform (sometimes called trapezoidal). Basic components of exemplary circuits for damped sine waveform and trapezoidal waveform defibrillators are shown in Figs. 79.4 and 79.5. The shape of the waveforms generated by RCL defibrillators depend on the resistance of the patient as well as the energy storage capacitance and resistance and inductance of the inductor. When discharged into a 50-Ω load (to stimulate the patient’s resistance), these defibrillators produce either a critically damped sine waveform or a slightly underdamped sine waveform (i.e., having a slight reversal of waveform polarity following the main waveform) into the 50-Ω load.

FIGURE 79.4 Resister-capacitor-inductor defibrillator. The patient is represented by R. (Modified from Feinberg B. 1980. Handbook Series in Clinical Laboratory Science, vol 2, Boca Raton, Fla, CRC Press, with permission.)


FIGURE 79.5 Trapezoidal wave defibrillator. The patient is represented by R. (Modified from Feinberg B. 1980. Handbook Series in Clinical Laboratory Science, vol 2, Boca Raton, Fla, CRC Press, with permission.)

The exact waveform can be determined by application of Kirkchoff ’s voltage law to the circuit

L

(

)

1 di + Ri + R i + dt C

∫ idt = 0

(79.3)

where L = inductance in H, i = instantaneous current in amperes, t = time in seconds, Ri = device resistance, R = subject resistance, and C = capacitance. From this, the second-order differential equation describes the RCL defibrillator.

L

(

)

d 2i di 1 + Ri + R + i =0 2 dt C dt

(79.4)

Trapezoidal waveform (actually, these are truncated exponential decay waveform) defibrillators are also used clinically. The circuit diagram in Fig. 79.4 is exemplary of one design for producing such a waveform. Delivered energy calculation for this waveform is expressed as

 d   Ii Wd = 0.5 I R   log e  I  f  2 i

 2    If     1−       I i   

(79.5)

where Wd = delivered energy, Ii = initial current in amperes, If = final current, R = resistance of the patient, and d = pulse duration in seconds. Both RCL and trapezoidal waveforms defibrillate effectively. Implantable defibrillators now use alternative waveforms such as a biphasic exponential decay waveform, in which the polarity of the electrodes is reversed part way through the shock. Use of the biphasic waveform has reduced the shock intensity required for implantable defibrillators but has not yet been extended to trans-chest use except on an experimental basis. RCL defibrillators are the most widely available. They store up to about 440 joules and deliver up to about 360 joules into a patient with 50-ohm impedance. Several selectable energy intensities are available, typically from 5–360 J, so that pediatric patients, very small patients, or patients with easily converted arrhythmias can be treated with low-intensity shocks. The pulse duration ranges from 3–6 ms. Because the resistance (R) varies between patients (25–150 ohms) and is part of the RCL discharge circuit, the duration and damping of the pulse also varies; increasing patient impedance lengthens and dampens the pulse. Figure 79.6 shows waveforms from RCL defibrillators with critically damped and with underdamped pulses.

79.4 Electrodes Electrodes for external defibrillation are metal and from 70–100 cm2 in surface area. They must be coupled to the skin with an electrically conductive material to achieve low impedance across the electrode-patient © 2000 by CRC Press LLC

FIGURE 79.6 The damped sine wave. The interval O–D represents a duration for the critically and overdamped since waves. By time D, more than 99% of the energy has been delivered. O–U is taken as the duration for an underdamped sine wave. (Modified from Tacker WA, Geddes LA. 1980. Electrical Defibrillation, Boca Raton, Fla, CRC Press, with permission.)

interface. There are two types of electrodes: hand-held (to which a conductive liquid or solid gel is applied) and adhesive, for which an adhesive conducting material holds the electrode in place. Hand-held electrodes are reusable and are pressed against the patient’s chest by the operator during shock delivery. Adhesive electrodes are disposable and are applied to the chest before the shock delivery and left in place for reuse if subsequent shocks are needed. Electrodes are usually applied with both electrodes on the anterior chest as shown in Fig. 79.7 or in anterior-to-posterior (front-to-back) position, as shown in Fig. 79.8.

79.5 Synchronization Most defibrillators for trans-chest use have the feature of synchronization, which is an electronic sensing and triggering mechanism for application of the shock during the QRS complex of the ECG. This is required when treating arrhythmias other than ventricular fibrillation, because inadvertent application of a shock during the T wave of the ECG often produces ventricular fibrillation. Selection by the operator of the synchronized mode of defibrillator operation will cause the defibrillator to automatically sense the QRS complex and apply the shock during the QRS complex. Furthermore, on the ECG display, the timing of the shock on the QRS is graphically displayed so the operator can be certain that the shock will not fall during the T wave (see Fig. 79.9).

79.6 Automatic External Defibrillators Automatic external defibrillators (AEDs) are defibrillators that automatically or semiautomatically recognize and treat rapid arrhythmias, usually under emergency conditions. Their operation requires less © 2000 by CRC Press LLC

FIGURE 79.7 Cross-sectional view of the chest showing position for standard anterior wall (precordial) electrode placement. Lines of presumed current flow are shown between the electrodes on the skin surface. (Modified from Tacker WA (ed). 1994. Defibrillation of the Heart: ICDs, AEDs and Manual, St. Louis, Mosby-Year Book, with permission.)

FIGURE 79.8 Cross-sectional view of the chest showing position for front-to-back electrode placement. Lines of presumed current flow are shown between the electrodes on the skin surface. (Modified from Tacker WA (ed). 1994. Defibrillation of the Heart: ICDs, AEDs and Manual, St. Louis, Mosby-Year Book, with permission.)

training than operation of manual defibrillators because the operator need not know which ECG waveforms indicate rhythms requiring a shock. The operator applies adhesive electrodes from the AED to the patient and turns on the AED, which monitors the ECG and determines by built-in signal processing © 2000 by CRC Press LLC

FIGURE 79.9 Timing mark (M) as shown on a synchronized defibrillator monitor. The M designates when in the cardiac cycle a shock will be applied. The T wave must be avoided, since a shock during the vulnerable period (V.P.) may fibrillate the ventricles. This tracing shows atrial fibrillation as identified by the irregular wavy baseline of the ECG. (Modified from Feinberg B. 1980. Handbook Series in Clinical Laboratory Science, vol 2, Boca Raton, Fla, CRC Press, with permission.)

whether or not and when to shock the patient. In a completely automatic mode, the AED does not have a manual control as shown in Fig. 79.3 but instead has an automatic control. In semiautomatic mode, the operator must confirm the shock advisory from the AED to deliver the shock. AEDs have substantial potential for improving the chances of survival from cardiac arrest because they enable emergency personnel, who typically reach the patient before paramedics do, to deliver defibrillating shocks. Furthermore, the reduced training requirements make feasible the operation of AEDs in the home by a family member of a patient at high risk of ventricular fibrillation.

79.7 Defibrillator Safety Defibrillators are potentially dangerous devices because of their high electrical output characteristics. The danger to the patient of unsynchronized shocks has already been presented, as has the synchronization design to prevent inadvertent precipitation of fibrillation by a cardioversion shock applied during the T wave. There are other safety issues. Improper technique may result in accidental shocking of the operator or other personnel in the vicinity, if someone is in contact with the electric discharge pathway. This may occur if the operator is careless in holding the discharge electrodes or if someone is in contact with the patient or with a metal bed occupied by the subject when the shock is applied. Proper training and technique is necessary to avoid this risk. Another safety issue is that of producing damage to the patient by application of excessively strong or excessively numerous shocks. Although cardiac damage has been reported after high-intensity and repetitive shocks to experimental animals and human patients, it is generally held that significant cardiac damage is unlikely if proper clinical procedures and guidelines are followed. Failure of a defibrillator to operate correctly may also be considered a safety issue, since inability of a defibrillator to deliver a shock in the absence of a replacement unit means loss of the opportunity to resuscitate the patient. A recent review of defibrillator failures found that operator errors, inadequate defibrillator care and maintenance, and, to a lesser extent, component failure accounted for the majority of defibrillator failures [7].


References 1. Tacker WA Jr (ed). 1994. Defibrillation of the Heart: ICDs, AEDs, and Manual. St. Louis, MosbyYear Book. 2. Tacker WA Jr, Geddes LA. 1980. Electrical Defibrillation. Boca Raton, Fla, CRC Press. 3. Emergency Cardiac Care Committees, American Heart Association. 1992. Guidelines for cardiopulmonary resuscitation and emergency cardiac care. JAMA 268:2199. 4. American National Standard ANSI/AAMI DF2. 1989 (second edition, revision of ANSI/AAMI DF2-1981). Safety and performance standard: Cardiac defibrillator devices. 5. Canadian National Standard CAN/CSA C22.2 No. 601.2.4-M90. 1990. Medical electrical equipment, part 2: Particular requirements for the safety of cardiac defibrillators and cardiac defibrillator/monitors. 6. International Standard IEC 601-2-4. 1983. Medical electrical equipment, part 2: Particular requirements for the safety of cardiac defibrillators and cardiac defibrillator/monitors. 7. Cummins RO, Chesemore K, White RD, and the Defibrillator Working Group. 1990. Defibrillator failures: Causes of problems and recommendations for improvement. JAMA 264:1019.

Further Information Detailed presentation of material on defibrillator waveforms, algorithms for ECG analysis, and automatic defibrillation using AED’s, electrodes, design, clinical use, effects of drugs on shock strength required to defibrillate, damage due to defibrillator shocks, and use of defibrillators during open-thorax surgical procedures or trans-esophageal defibrillation are beyond the scope of this chapter. Also, the historical aspects of defibrillation are not presented here. For more information, the reader is referred to the publications at the end of this chapter [1–3]. For detailed description of specific defibrillators with comparisons of features, the reader is referred to articles from Health Devices, a monthly publication of ECRI, 5200 Butler Pike, Plymouth Meeting, Pa USA. For American, Canadian, and European defibrillator standards, the reader is referred to published standards [3–6] and Charbonnier’s discussion of standards [1].


Duffin, E. G. “ Implantable Defibrillators.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

80 Implantable Defibrillators

Edwin G. Duffin Medtronic, Inc.

80.1 80.2 80.3 80.4 80.5 80.6 80.7 80.8

Pulse Generators Electrode Systems (“LEADS”) Arrhythmia Detection Arrhythmia Therapy Implantable Monitoring Follow-up Economics Conclusion

The implantable cardioverter defibrillator (ICD) is a therapeutic device that can detect ventricular tachycardia or fibrillation and automatically deliver high-voltage (750 V) shocks that will restore normal sinus rhythm. Advanced versions also provide low-voltage (5–10 V) pacing stimuli for painless termination of ventricular tachycardia and for management of bradyarrhythmias. The proven efficacy of the automatic implantable defibrillator has placed it in the mainstream of therapies for the prevention of sudden arrhythmic cardiac death. The implantable defibrillator has evolved significantly since first appearing in 1980. The newest devices can be implanted in the patient’s pectoral region and use electrodes that can be inserted transvenously, eliminating the traumatic thoracotomy required for placement of the earlier epicardial electrode systems. Transvenous systems provide rapid, minimally invasive implants with high assurance of success and greater patient comfort. Advanced arrhythmia detection algorithms offer a high degree of sensitivity with reasonable specificity, and extensive monitoring is provided to document performance and to facilitate appropriate programming of arrhythmia detection and therapy parameters. Generator longevity can now exceed 4 years, and the cost of providing this therapy is declining.

80.1 Pulse Generators The implantable defibrillator consists of a primary battery, high-voltage capacitor bank, and sensing and control circuitry housed in a hermetically sealed titanium case. Commercially available devices weigh between 197 and 237 grams and range in volume from 113 to 145 cm3. Clinical trials are in progress on devices with volumes ranging from 178 cm3 to 60 cm3 and weights between 275 and 104 grams. Further size reductions will be achieved with the introduction of improved capacitor and integrated circuit technologies and lead systems offering lower pacing and defibrillation thresholds. Progress should parallel that made with antibradycardia pacemakers that have evolved from 250-gram, nonprogrammable, VOO units with 600-µJ pacing outputs to 26-gram, multiprogrammable, DDDR units with dual 25-µJ outputs. Implantable defibrillator circuitry must include an amplifier, to allow detection of the millivolt-range cardiac electrogram signals; noninvasively programmable processing and control functions, to evaluate


the sensed cardiac activity and to direct generation and delivery of the therapeutic energy; high-voltage switching capability; dc-dc conversion functions to step up the low battery voltages; random access memories, to store appropriate patient and device data; and radiofrequency telemetry systems, to allow communication to and from the implanted device. Monolithic integrated circuits on hybridized substrates have made it possible to accomplish these diverse functions in a commercially acceptable and highly reliable form. Defibrillators must convert battery voltages of approximately 6.5 V to the 600–750 V needed to defibrillate the heart. Since the conversion process cannot directly supply this high voltage at current strengths needed for defibrillation, charge is accumulated in relatively large (≈85–120µF effective capacitance) aluminum electrolytic capacitors that account for 20–30% of the volume of a typical defibrillator. These capacitors must be charged periodically to prevent their dielectric from deteriorating. If this is not done, the capacitors become electrically leaky, yielding excessively long charge times and delay of therapy. Early defibrillators required that the patient return to the clinic periodically to have the capacitors reformed, whereas newer devices do this automatically at preset or programmable times. Improved capacitor technology, perhaps ceramic or thin-film, will eventually offer higher storage densities, greater shape variability for denser component packaging, and freedom from the need to waste battery capacity performing periodic reforming charges. Packaging density has already improved from 0.03 J/cm3 for devices such as the early cardioverter to 0.43 J/cm3 with some investigational ICDs. Capacitors that allow conformal shaping could readily increase this density to more than 0.6 J/cm3. Power sources used in defibrillators must have sufficient capacity to provide 50–400 full energy charges (≈34 J) and 3 to 5 years of bradycardia pacing and background circuit operation. They must have a very low internal resistance in order to supply the relatively high currents needed to charge the defibrillation capacitors in 5–15 s. This generally requires that the batteries have large surface area electrodes and use chemistries that exhibit higher rates of internal discharge than those seen with the lithium iodide batteries used in pacemakers. The most commonly used defibrillator battery chemistry is lithium silver vanadium oxide.

80.2 Electrode Systems (“Leads”) Early implantable defibrillators utilized patch electrodes (typically a titanium mesh electrode) placed on the surface of the heart, requiring entry through the chest (Fig. 80.1). This procedure is associated with approximately 3–4% perioperative mortality, significant hospitalization time and complications, patient discomfort, and high costs. Although subcostal, subxiphoid, and thoracoscopic techniques can minimize the surgical procedure, the ultimate solution has been development of fully transvenous lead systems with acceptable defibrillation thresholds. Currently available transvenous leads are constructed much like pacemaker leads, using polyurethane or silicone insulation and platinum-iridium electrode materials. Acceptable thresholds are obtained in 67–95% of patients, with mean defibrillation thresholds ranging from 10.9–18.1 J. These lead systems use a combination of two or more electrodes located in the right ventricular apex, the superior vena cava, the coronary sinus, and sometimes, a subcutaneous patch electrode is placed in the chest region. These leads offer advantages beyond the avoidance of major surgery. They are easier to remove should there be infections or a need for lead system revision. The pacing thresholds of current transvenous defibrillation electrodes are typically © 2000 by CRC Press LLC

FIGURE 80.1 Epicardial ICD systems typically use two or three large defibrillating patch electrodes placed on the epicardium of the left and right ventricles and a pair of myocardial electrodes for detection and pacing. The generator is usually placed in the abdomen. (Copyright Medtronic, Inc. Used with permission.)

FIGURE 80.2 The latest transvenous fibrillation systems employ a single catheter placed in the right ventricular apex. In panel a, a single transvenous catheter provides defibrillation electrodes in the superior vena cava and in the right ventricle. This catheter provides a single pace/sense electrode which is used in conjunction with the right ventricular high-voltage defibrillation electrode for arrhythmia detection and antibradycardia/antitachycardia pacing (a configuration that is sometimes referred to as integrated bipolar). With pulse generators small enough to be placed in the pectoral region, defibrillation can be achieved by delivering energy between the generator housing and one high-voltage electrode in the right ventricle (analogous to unipolar pacing) as is shown in panel b. This catheter provided bipolar pace/sense electrodes for arrhythmia detection and antibradycardia/antitachycardia pacing. (Copyright Medtronic, Inc. Used with permission.)

0.96 ± 0.39 V, and the electrogram amplitudes are on the order of 16.4 ± 6.4 mV. The eventual application of steroid-eluting materials in the leads should provide increased pacing efficiency with transvenous lead systems, thereby reducing the current drain associated with pacing and extending pulse generator longevity. Lead systems are being refined to simplify the implant procedures. One approach is the use of a single catheter having a single right ventricular low-voltage electrode for pacing and detection, and a pair of high-voltage defibrillation electrodes spaced for replacement in the right ventricle and in the superior vena cava (Fig. 80.2a). A more recent approach parallels that used for unipolar pacemakers. A single right-ventricular catheter having bipolar pace/sense electrodes and one right ventricular high-voltage electrode is used in conjunction with a defibrillator housing that serves as the second high-voltage electrode (Fig. 80.2b). Mean biphasic pulse defibrillation thresholds with the generator-electrode placed in the patient’s left pectoral region are reported to be 9.8 ± 6.6 J (n = 102). This approach appears to be practicable only with generators suitable for pectoral placement, but such devices will become increasingly available.

80.3 Arrhythmia Detection Most defibrillator detection algorithms rely primarily on heart rate to indicate the presence of a treatable rhythm. Additional refinements sometimes include simple morphology assessments, as with the probability density function, and analysis of rhythm stability and rate of change in rate. The probability density function evaluates the percentage of time that the filtered ventricular electrogram spends in a window centered on the baseline. The rate-of-change-in-rate or onset evaluation discriminates sinus tachycardia from ventricular tachycardia on the basis of the typically gradual acceleration of sinus rhythms versus the relatively abrupt acceleration of many pathologic tachycardias. The rate stability © 2000 by CRC Press LLC

function is designed to bar detection of tachyarrhythmias as long as the variation in ventricular rate exceeds a physician-programmed tolerance, thereby reducing the likelihood of inappropriate therapy delivery in response to atrial fibrillation. This concept appears to be one of the more successful detection algorithm enhancements. Because these additions to the detection algorithm reduce sensitivity, some defibrillator designs offer a supplementary detection mode that will trigger therapy in response to any elevated ventricular rate of prolonged duration. These extended-high-rate algorithms bypass all or portions of the normal detection screening, resulting in low specificity for rhythms with prolonged elevated rates such as exercise-induced sinus tachycardia. Consequently, use of such algorithms generally increases the incidence of inappropriate therapies. Improvements in arrhythmia detection specificity are desirable, but they must not decrease the excellent sensitivity offered by current algorithms. The anticipated introduction of defibrillators incorporating dual-chamber pacemaker capability will certainly help in this quest, since it will then be possible to use atrial electrograms in the rhythm classification process. It would also be desirable to have a means of evaluating the patient’s hemodynamic tolerance of the rhythm, so that the more comfortable pacing sequences could be used as long as the patient was not syncopal yet branch quickly to a definitive shock should the patient begin to lose consciousness. Although various enhanced detection processes have been proposed, many have not been tested clinically, in some cases because sufficient processing power was not available in implantable systems, and in some cases because sensor technology was not yet ready for chronic implantation. Advances in technology may eventually make some of these very elegant proposals practicable. Examples of proposed detection enhancements include extended analyses of cardiac event timing (PR and RR stability, AV interval variation, temporal distribution of atrial electrogram intervals and of ventricular electrogram intervals, timing differences and/or coherency of multiple ventricular electrograms, ventricular response to a provocative atrial extrastimuli), electrogram waveform analyses (paced depolarization integral, morphology analyses of right ventricular or atrial electrograms), analyses of hemodynamic parameters (right-ventricular pulsatile pressure, mean right atrial and mean right ventricular pressures, wedge coronary sinus pressure, static right ventricular pressure, right atrial pressure, right ventricular stroke volume, mixed venous oxygen saturation and mixed venous blood temperature, left ventricular impedance, intramyocardial pressure gradient, aortic and pulmonary artery flow), and detection of physical motion. Because defibrillator designs are intentionally biased to overtreat in preference to the life-threatening consequences associated with failure to treat, there is some incidence of inappropriate therapy delivery. Unwarranted therapies are usually triggered by supraventricular tachyarrhythmias, especially atrial fibrillation, or sinus tachycardia associated with rates faster than the ventricular tachycardia detection rate threshold. Additional causes include nonsustained ventricular tachycardia, oversensing of T waves, double counting of R waves and pacing stimuli from brady pacemakers, and technical faults such as loose leadgenerator connections or lead fractures. Despite the bias for high detection sensitivity, undersensing does occur. It has been shown to result from inappropriate detection algorithm programming, such as an excessively high tachycardia detection rate; inappropriate amplifier gain characteristics; and electrode designs that place the sensing terminals too close to the high-voltage electrodes with a consequent reduction in electrogram amplitude following shocks. Undersensing can also result in the induction of tachycardia should the amplifier gain control algorithm result in undersensing of sinus rhythms.

80.4 Arrhythmia Therapy Pioneering implantable defibrillators were capable only of defibrillation shocks. Subsequently, synchronized cardioversion capability was added. Antibradycardia pacing had to be provided by implantation of a standard pacemaker in addition to the defibrillator, and, if antitachycardia pacing was prescribed, it


was necessary to use an antitachycardia pacemaker. Several currently marketed implantable defibrillators offer integrated ventricular demand pacemaker function and tiered antiarrhythmia therapy (pacing/cardioversion/defibrillation). Various burst and ramp antitachycardia pacing algorithms are offered, and they all seem to offer comparably high success rates. These expanded therapeutic capabilities improve patient comfort by reducing the incidence of shocks in conscious patients, eliminate the problems and discomfort associated with implantation of multiple devices, and contribute to a greater degree of success, since the prescribed regimens can be carefully tailored to specific patient needs. Availability of devices with antitachy pacing capability significantly increases the acceptability of the implantable defibrillator for patients with ventricular tachycardia. Human clinical trials have shown that biphasic defibrillation waveforms are more effective than monophasic waveforms, and newer devices now incorporate this characteristic. Speculative explanations for biphasic superiority include the large voltage change at the transition from the first to the second phase or hyperpolarization of tissue and reactivation of sodium channels during the initial phase, with resultant tissue conditioning that allows the second phase to more readily excite the myocardium. Antitachycardia pacing and cardioversion are not uniformly successful. There is some incidence of ventricular arrhythmia acceleration with antitachycardia pacing and cardioversion, and it is also not unusual for cardioversion to induce atrial fibrillation that in turn triggers unwarranted therapies. An ideal therapeutic solution would be one capable of preventing the occurrence of tachycardia altogether. Prevention techniques have been investigated, among them the use of precisely timed subthreshold stimuli, simultaneous stimulation at multiple sites, and pacing with elevated energies at the site of the tachycardia, but none has yet proven practical. The rudimentary VVI antibradycardia pacing provided by current defibrillators lacks rate responsiveness and atrial pacing capability. Consequently, some defibrillator patients require implantation of a separate dual-chamber pacemaker for hemodynamic support. It is inevitable that future generations of defibrillators will offer dual-chamber pacing capabilities. Atrial fibrillation, occurring either as a consequence of defibrillator operation or as a natural progression in many defibrillator patients, is a major therapeutic challenge. It is certainly possible to adapt implantable defibrillator technology to treat atrial fibrillation, but the challenge is to do so without causing the patient undue discomfort. Biphasic waveform defibrillation of acutely induced atrial fibrillation has been demonstrated in humans with an 80% success rate at 0.4 J using epicardial electrodes. Stand-alone atrial defibrillators are in development, and, if they are successful, it is likely that this capability would be integrated into the mainstream ventricular defibrillators as well. However, most conscious patients find shocks above 0.5 J to be very unpleasant, and it remains to be demonstrated that a clinically acceptable energy level will be efficacious when applied with transvenous electrode systems to spontaneously occurring atrial fibrillation. Moreover, a stand-alone atrial defibrillator either must deliver an atrial shock with complete assurance of appropriate synchronization to ventricular activity or must restrict the therapeutic energy delivery to atrial structures in order to prevent inadvertent induction of a malignant ventricular arrhythmia.

80.5 Implantable Monitoring Until recently, defibrillator data recording capabilities were quite limited, making it difficult to verify the adequacy of arrhythmia detection and therapy settings. The latest devices record electrograms and diagnostic channel data showing device behavior during multiple tachyarrhythmia episodes. These devices also include counters (number of events detected, success and failure of each programmed therapy, and so on) that present a broad, though less specific, overview of device behavior (Fig. 80.3). Monitoring capability in some of the newest devices appears to be the equivalent of 32 Kbytes of random access memory, allowing electrogram waveform records of approximately 2-min duration, with some opportunity for later expansion by judicious selection of sampling rates and data compression techniques. Electrogram storage has proven useful for documenting false therapy delivery due to atrial fibrillation,


FIGURE 80.3 Typical data recorded by an implantable defibrillator include stored intracardiac electrograms with annotated markers indicating cardiac intervals, paced and sensed events, and device classification of events (TF = fast tachycardia; TP = antitachy pacing stimulus; VS = sensed nontachy ventricular event). In the example, five rapid pacing pulses convert a ventricular tachycardia with a cycle length of 340 ms into sinus rhythm with a cycle length of 830 ms. In the lower portion of the figure is an example of the summary data collected by the ICD, showing detailed counts of the performance of the various therapies (Rx) for ventricular tachycardia (VT), fast ventricular (VTF), and ventricular (VF). (Copyright Medtronic, Inc. Used with permission.)

lead fractures, and sinus tachycardia, determining the triggers of arrhythmias; documenting rhythm accelerations in response to therapies; and demonstrating appropriate device behavior when treating asymptomatic rhythms. Electrograms provide useful information by themselves, yet they cannot indicate how the device interpreted cardiac activity. Increasingly, electrogram records are being supplemented with event markers that indicate how the device is responding on a beat-by-beat basis. These records can include measurements of the sensed and paced intervals, indication as to the specific detection zone an event falls in, indication of charge initiation, and other device performance data.

80.6 Follow-up Defibrillator patients and their devices require careful follow-up. In one study of 241 ICD patients with epicardial lead systems, 53% of the patients experienced one or more complications during an average exposure of 24 months. These complications included infection requiring device removal in 5%, postoperative respiratory complications in 11%, postoperative bleeding and/or thrombosis in 4%, lead system migration or disruption in 8%, and documented inappropriate therapy delivery, most commonly due to atrial fibrillation, in 22%. A shorter study of eighty patients with transvenous defibrillator systems reported no postoperative pulmonary complications, transient nerve injury (1%), asymptomatic subclavian vein occlusion (2.5%), pericardial effusion (1%), subcutaneous patch pocket hematoma (5%), pulse © 2000 by CRC Press LLC

generator pocket infection (1%), lead fracture (1%), and lead system dislodgement (10%). During a mean follow-up period of 11 months, 7.5% of the patients in this series experienced inappropriate therapy delivery, half for atrial fibrillation and the rest for sinus tachycardia. Although routine follow-up can be accomplished in the clinic, detection and analysis of transient events depends on the recording capabilities available in the devices or on the use of various external monitoring equipment.

80.7 Economics The annual cost of ICD therapy is dropping as a consequence of better longevity and simpler implantation techniques. Early generators that lacked programmability, antibradycardia pacing capability, and event recording had 62% survival at 18 months and 2% at 30 months. Some recent programmable designs that include VVI pacing capability and considerable event storage exhibit 96.8% survival at 48 months. It has been estimated that an increase in generator longevity from 2–5 years would lower the cost per life-year saved by 55% in a hypothetical patient population with a 3-year sudden mortality of 28%. More efficient energy conversion circuits and finer line-width integrated circuit technology with smaller, more highly integrated circuits and reduced current drains will yield longer-lasting defibrillators while continuing the evolution to smaller volumes. Cost of the implantation procedure is clearly declining as transvenous lead systems become commonplace. Total hospitalization duration, complication rates, and use of costly hospital operating rooms and intensive care facilities all are reduced, providing significant financial benefits. One study reported requiring half the intensive care unit time and a reduction in total hospitalization from 26 to 15 days when comparing transvenous to epicardial approaches. Another center reported a mean hospitalization stay of 6 days for patients receiving transvenous defibrillation systems. Increasing sophistication of the implantable defibrillators paradoxically contributes to cost efficacy. Incorporation of single-chamber brady pacing capability eliminates the cost of a separate pacemaker and lead for those patients who need one. Eventually even dual-chamber pacing capability will be available. Programmable detection and therapy features obviate the need for device replacement that was required when fixed parameter devices proved to be inappropriately specified or too inflexible to adapt to a patient’s physiologic changes. Significant cost savings may be obtained by better patient selection criteria and processes, obviating the need for extensive hospitalization and costly electrophysiologic studies prior to device implantation in some patient groups. One frequently discussed issue is the prophylactic role that implantable defibrillators will or should play. Unless a means is found to build far less expensive devices that can be placed with minimal time and facilities, the life-saving yield for prophylactic defibrillators will have to be high if they are to be cost-effective. This remains an open issue.

80.8 Conclusion The implantable defibrillator is now an established and powerful therapeutic tool. The transition to pectoral implants with biphasic waveforms and efficient yet simple transvenous lead systems is simplifying the implant procedure and drastically reducing the number of unpleasant VF inductions required to demonstrate adequate system performance These advances are making the implantable defibrillator easier to use, less costly, and more acceptable to patients and their physicians.

Acknowledgment Portions of this text are derived from Duffin EG, Barold SS. 1994. Implantable cardioverter-defibrillators: An overview and future directions, Chapter 28 of I Singer (ed), Implantable Cardioverter-Defibrillator, and are used with permission of Futura Publishing Company, Inc.


References Josephson M, Wellens H (eds). 1992. Tachycardias: Mechanisms and Management. Mount Kisco, NY, Futura Publishing. Kappenberger L, Lindemans F (eds). 1992. Practical Aspects of Staged Therapy Defibrillators. Mount Kisco, NY, Futura Publishing. Singer I (ed). 1994. Implantable Cardioverter-Defibrillator. Mount Kisco, NY, Futura Publishing. Tacker W (ed). 1994. Defibrillation of the Heart: ICD’s, AED’s, and Manual. St. Louis, Mosby. PACE, 14:865. (Memorial issue on implantable defibrillators honoring Michel Mirowski.)


Eggleston, J. L., Von Maltzahn, W.W. “Electrosurgical Devices.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

81 Electrosurgical Devices

Jeffrey L. Eggleston Valleylab, Inc.

Wolf W. von Maltzahn Whitaker Foundation

81.1 81.2 81.3 81.4 81.5 81.6 81.7 81.8

Theory of Operation Monopolar Mode Bipolar Mode ESU Design Active Electrodes Dispersive Electrodes ESU Hazards Recent Developments

An electrosurgical unit (ESU) passes high-frequency electric currents through biologic tissues to achieve specific surgical effects such as cutting, coagulation, or desiccation. Although it is not completely understood how electrosurgery works, it has been used since the 1920s to cut tissue effectively while at the same time controlling the amount of bleeding. Cutting is achieved primarily with a continuous sinusoidal waveform, whereas coagulation is achieved primarily with a series of sinusoidal wave packets. The surgeon selects either one of these waveforms or a blend of them to suit the surgical needs. An electrosurgical unit can be operated in two modes, the monopolar mode and the bipolar mode. The most noticeable difference between these two modes is the method in which the electric current enters and leaves the tissue. In the monopolar mode, the current flows from a small active electrode into the surgical site, spreads through the body, and returns to a large dispersive electrode on the skin. The high current density in the vicinity of the active electrode achieves tissue cutting or coagulation, whereas the low current density under the dispersive electrode causes no tissue damage. In the bipolar mode, the current flows only through the tissue held between two forceps electrodes. The monopolar mode is used for both cutting and coagulation. The bipolar mode is used primarily for coagulation. This chapter beings with the theory of operation for electrosurgical units, outlines various modes of operation, and gives basic design details for electronic circuits and electrodes. It then describes how improper application of electrosurgical units can lead to hazardous situations for both the operator and the patient and how such hazardous situations can be avoided or reduced through proper monitoring methods. Finally, the chapter gives an update on current and future developments and applications.

81.1 Theory of Operation In principle, electrosurgery is based on the rapid heating of tissue. To better understand the thermodynamic events during electrosurgery, it helps to know the general effects of heat on biologic tissue. Consider a tissue volume that experiences a temperature increase from normal body temperature to 45°C within a few seconds. Although the cells in this tissue volume show neither microscopic nor macroscopic changes, some cytochemical changes do in fact occur. However, these changes are reversible, and the cells return to their normal function when the temperature returns to normal values. Above 45°C, irreversible changes


take place that inhibit normal cell functions and lead to cell death. First, between 45°C and 60°C, the proteins in the cell lose their quaternary configuration and solidify into a glutinous substance that resembles the white of a hard-boiled egg. This process, termed coagulation, is accompanied by tissue blanching. Further increasing the temperature up to 100°C leads to tissue drying; that is, the aqueous cell contents evaporate. This process is called desiccation. If the temperature is increased beyond 100°C, the solid contents of the tissue reduce to carbon, a process referred to as carbonization. Tissue damage depends not only on temperature, however, but also on the length of exposure to heat. Thus, the overall temperature-induced tissue damage is an integrative effect between temperature and time that is expressed mathematically by the Arrhenius relationship, where an exponential function of temperature is integrated over time [1]. In the monopolar mode, the active electrode either touches the tissue directly or is held a few millimeters above the tissue. When the electrode is held above the tissue, the electric current bridges the air gap by creating an electric discharge arc. A visible arc forms when the electric field strength exceeds 1 kV/mm in the gap and disappears when the field strength drops below a certain threshold level. When the active electrode touches the tissue and the current flows directly from the electrode into the tissue without forming an arc, the rise in tissue temperature follows the bioheat equation

T − To =

1 2 Jt σρc

(81.1)

where T and To are the final and initial temperatures (K), σ is the electrical conductivity (S/m), ρ is the tissue density (kg/m3), c is the specific heat of the tissue (Jkg–1K–1), J is the current density (A/m2), and t is the duration of heat applications [1]. The bioheat equation is valid for short application times where secondary effects such as heat transfer to surrounding tissues, blood perfusion, and metabolic heat can be neglected. According to Eq. (81.1), the surgeon has primarily three means of controlling the cutting or coagulation effect during electrosurgery: the contact area between active electrode and tissue, the electrical current density, and the activation time. In most commercially available electrosurgical generators, the output variable that can be adjusted is power. This power setting, in conjunction with the output power vs. tissue impedance characteristics of the generator, allow the surgeon some control over current. Table 81.1 lists typical output power and mode settings for various surgical procedures. Table 81.2 lists some typical impedance ranges seen during use of an ESU in surgery. The values are shown as ranges because the impedance increases as the tissue dries out, and at the same time, the output power of the ESU decreases. The surgeon may control current density by selection of the active electrode type and size.

81.2 Monopolar Mode A continuous sinusoidal waveform cuts tissue with very little hemostasis. This waveform is simply called cut or pure cut. During each positive and negative swing of the sinusoidal waveform, a new discharge arc forms and disappears at essentially the same tissue location. The electric current concentrates at this tissue location, causing a sudden increase in temperature due to resistive heating. The rapid rise in temperature then vaporizes intracellular fluids, increases cell pressure, and ruptures the cell membrane, thereby parting the tissue. This chain of events is confined to the vicinity of the arc, because from there the electric current spreads to a much larger tissue volume, and the current density is no longer high enough to cause resistive heating damage. Typical output values for ESUs, in cut and other modes, are shown in Table 81.3. Experimental observations have shown that more hemostasis is achieved when cutting with an interrupted sinusoidal waveform or amplitude modulated continuous waveform. These waveforms are typically called blend or blended cut. Some ESUs offer a choice of blend waveforms to allow the surgeon to select the degree of hemostasis desired.


Table 81.1

Typical ESU Power Settings for Various Surgical Procedures

Power-Level Range Low power 70 W coag

Procedures Neurosurgery Dermatology Plastic surgery Oral surgery Laparoscopic sterilization Vasectomy General surgery Laparotomies Head and neck surgery (ENT) Major orthopedic surgery Major vascular surgery Routine thoracic surgery Polypectomy Transurethral resection procedures (TURPs) Thoracotomies Ablative cancer surgery Mastectomies

Note: Ranges assume the use of a standard blade electrode. Use of a needle electrode, or other small current-concentrating electrode, allows lower settings to be used; users are urged to use the lowest setting that provides the desired clinical results.

TABLE 81.2 Typical Impedance Ranges Seen During Use of an ESU in Surgery Cut Mode Application Prostate tissue Oral cavity Liver tissue Muscle tissue Gall bladder Skin tissue Bowel tissue Periosteum Mesentery Omentum Adipose tissue Scar tissue Adhesions

Impedance Range (Ω) 400–1700 1000–2000

1500–2400 1700–2500 2500–3000 3000–4200 3500–4500

Coag Mode Application Contact coagulation to stop bleeding

100–1000

When a continuous or interrupted waveform is used in contact with the tissue and the output voltage current density is too low to sustain arcing, desiccation of the tissue will occur. Some ESUs have a distinct mode for this purpose called desiccation or contact coagulation. In noncontact coagulation, the duty cycle of an interrupted waveform and the crest factor (ratio of peak voltage to rms voltage) influence the degree of hemostasis. While a continuous waveform reestablishes the arc at essentially the same tissue location concentrating the heat there, an interrupted waveform


TABLE 81.3

Typical Output Characteristics of ESUs

Monopolar modes Cut Blend Desiccate Fulgurate/spray Bipolar mode Coagulate/desiccate

Frequency, kHz

Crest Factor  Vpeak   Vrms 

Duty Cycle

1–400 1–300 1–200 1–200

300–1750 300–1750 240–800 300–800

1.4–2.1 2.1–6.0 3.5–6.0 6.0–20.0

100% 25–80% 50–100% 10–70%

1–70

300–1050

1.6–12.0

25–100%

Output Voltage Range Open Circuit, Vpeak-peak, V

Output Power Range, W

200–5000 1500–5800 400–6500 6000–12000 200–1000

causes the arc to reestablish itself at different tissue locations. The arc seems to dance from one location to the other raising the temperature of the top tissue layer to coagulation levels. These waveforms are called fulguration or spray. Since the current inside the tissue spreads very quickly from the point where the arc strikes, the heat concentrates in the top layer, primarily desiccating tissue and causing some carbonization. During surgery, a surgeon can easily choose between cutting, coagulation, or a combination of the two by activating a switch on the grip of the active electrode or by use of a footswitch.

81.3 Bipolar Mode The bipolar mode concentrates the current flow between the two electrodes, requiring considerably less power for achieving the same coagulation effect than the monopolar mode. For example, consider coagulating a small blood vessel with 3-mm external diameter and 2-mm internal diameter, a tissue resistivity of 360 Ωcm, a contract area of 2 × 4 mm, and a distance between the forceps tips of 1 mm. The tissue resistance between the forceps is 450 Ω as calculated from R = ρL/A, where ρ is the resistivity, L is the distance between the forceps, and A is the contact area. Assuming a typical current density of 200 mA/cm2, then a small current of 16 mA, a voltage of 7.2 V, and a power level of 0.12 W suffice to coagulate this small blood vessel. In contrast, during monopolar coagulation, current levels of 200 mA and power levels of 100 W or more are not uncommon to achieve the same surgical effect. The temperature increase in the vessel tissue follows the bioheat equation, Eq. (81.1). If the specific heat of the vessel tissue is 4.2 Jg –1κ –1 and the tissue density is 1 g/cm3, then the temperature of the tissue between the forceps increases from 37°C to 57°C in 5.83 s. When the active electrode touches the tissue, less tissue damage occurs during coagulation, because the charring and carbonization that accompanies fulguration is avoided.

81.4 ESU Design Modern ESUs contain building blocks that are also found in other medical devices, such as microprocessors, power supplies, enclosures, cables, indicators, displays, and alarms. The main building blocks unique to ESUs are control input switches, the high-frequency power amplifier, and the safety monitor. The first two will be discussed briefly here, and the latter will be discussed later. Control input switches include front panel controls, footswitch controls, and handswitch controls. In order to make operating an ESU more uniform between models and manufacturers, and to reduce the possibility of operator error, the ANSI/AAMI HF-18 standard [5] makes specific recommendations concerning the physical construction and location of these switches and prescribes mechanical and electrical performance standards. For instance, front panel controls need to have their function identified by a permanent label and their output indicated on alphanumeric displays or on graduated scales; the pedals of foot switches need to be labeled and respond to a specified activation force; and if the active


electrode handle incorporates two finger switches, their position has to correspond to a specific function. Additional recommendations can be found in Reference [5]. Four basic high-frequency power amplifiers are in use currently; the somewhat dated vacuum tube/spark gap configuration, the parallel connection of a bank if bipolar power transistors, the hybrid connection of parallel bipolar power transistors cascaded with metal oxide silicon field effect transistors (MOSFETs), and the bridge connection of MOSFETs. Each has unique properties and represents a stage in the evolution of ESUs. In a vacuum tube/spark gap device, a tuned-plate, tuned-grid vacuum tube oscillator is used to generate a continuous waveform for use in cutting. This signal is introduced to the patient by an adjustable isolation transformer. To generate a waveform for fulguration, the power supply voltage is elevated by a step-up transformer to about 1600 V rms which then connects to a series of spark gaps. The voltage across the spark gaps is capacitively coupled to the primary of an isolation transformer. The RLC circuit created by this arrangement generates a high crest factor, damped sinusoidal, interrupted waveform. One can adjust the output power and characteristics by changing the turns ratio or tap on the primary and/or secondary side of the isolation transformer, or by changing the spark gap distance. In those devices that use a parallel bank of bipolar power transistors, the transistors are arranged in a Class A configuration. The bases, collectors, and emitters are all connected in parallel, and the collective base node is driven through a current-limiting resistor. A feedback RC network between the base node and the collector node stabilizes the circuit. The collectors are usually fused individually before the common node connects them to one side of the primary of the step-up transformer. The other side of the primary is connected to the high-voltage power supply. A capacitor and resistor in parallel to the primary create a resonance tank circuit that generates the output waveform at a specific frequency. Additional elements may be switched in and out of the primary parallel RLC to alter the output power and waveform for various electrosurgical modes. Small-value resistors between the emitters and ground improve the current sharing between transistors. This configuration sometimes requires the use of matched sets of high-voltage power transistors. A similar arrangement exists in amplifiers using parallel bipolar transistors cascaded with a power MOSFET. This arrangement is called a hybrid cascode amplifier. In this type of amplifier, the collectors of a group if bipolar transistors are connected, via protection diodes, to one side of the primary of the step-up output transformer. The other side of the primary is connected to the high-voltage power supply. The emitters of two or three bipolar transistors are connected, via current limiting resistors, to the drain of an enhancement mode MOSFET. The source of the MOSFET is connected to ground, and the gate of the MOSFET is connected to a voltage-snubbing network driven by a fixed amplitude pulse created by a high-speed MOS driver circuit. The bases of the bipolar transistors are connected, via current control RC networks, to a common variable base voltage source. Each collector and base is separately fused. In cut modes, the gate drive pulse is a fixed frequency, and the base voltage is varied according to the power setting. In the coagulation modes, the base voltage is fixed and the width of the pulses driving the MOSFET is varied. This changes the conduction time of the amplifier and controls the amount of energy imparted to the output transformer and its load. In the coagulation modes and in high-power cut modes, the bipolar power transistors are saturated, and the voltage across the bipolar/MOSFET combination is low. This translates to high efficiency and low power dissipation. The most common high-frequency power amplifier in use is a bridge connection of MOSFETs. In this configuration, the drains of a series of power MOSFETs are connected, via protection diodes, to one side of the primary of the step-up output transformer. The drain protection diodes protect the MOSFETs against the negative voltage swings of the transformer primary. The other side of the transformer primary is connected to the high-voltage power supply. The sources of the MOSFETs are connected to ground. The gate of each MOSFET has a resistor connected to ground and one to its driver circuitry. The resistor to ground speeds up the discharge of the gate capacitance when the MOSFET is turned on while the gate series resistor eliminates turn-off oscillations. Various combinations of capacitors and/or LC networks can be switched across the primary of the step-up output transformer to obtain different waveforms. In


the cut mode, the output power is controlled by varying the high-voltage power supply voltage. In the coagulation mode, the output power is controlled by varying the on time of the gate drive pulse.

81.5 Active Electrodes The monopolar active electrode is typically a small flat blade with symmetric leading and trailing edges that is embedded at the tip of an insulated handle. The edges of the blade are shaped to easily initiate discharge arcs and to help the surgeon manipulate the incision; the edges cannot mechanically cut tissue. Since the surgeon holds the handle like a pencil, it is often referred to as the “pencil.” Many pencils contain in their handle one or more switches to control the electrosurgical waveform, primarily to switch between cutting and coagulation. Other active electrodes include needle electrodes, loop electrodes, and ball electrodes. Needle electrodes are used for coagulating small tissue volumes like in neurosurgery or plastic surgery. Loop electrodes are used to resect nodular structures such as polyps or to excise tissue samples for pathologic analysis. An example would be the LLETZ procedure where the transition zone of the cervix is excised. Electrosurgery at the tip of an endoscope or laparoscope requires yet another set of active electrodes and specialized training of the surgeon.

81.6 Dispersive Electrodes The main purpose of the dispersive electrode is to return the high-frequency current to the electrosurgical unit without causing harm to the patient. This is usually achieved by attaching a large electrode to the patient’s skin away from the surgical site. The large electrode area and a small contact impedance reduce the current density to levels where tissue heating is minimal. Since the ability of a dispersive electrode to avoid tissue heating and burns is of primary importance, dispersive electrodes are often characterized by their heating factor. The heating factor describes the energy dissipated under the dispersive electrode per Ω of impedance and is equal to I2t, where I is the rms current and t is the time of exposure. During surgery a typical value for the heating factor is 3 A2s, but factors of up to 9 A2s may occur during some procedures [2]. Two types of dispersive electrodes are in common use today, the resistive type and the capacitive type. In disposable form, both electrodes have a similar structure and appearance. A thin, rectangular metallic foil has an insulating layer on the outside, connects to a gel-like material on the inside, and may be surrounded by an adhesive foam. In the resistive type, the gel-like material is made of an adhesive conductive gel, whereas in the capacitive type, the gel is an adhesive dielectric nonconductive gel. The adhesive foam and adhesive gel layer ensure that both electrodes maintain good skin contact to the patient, even if the electrode gets stressed mechanically from pulls on the electrode cable. Both types have specific advantages and disadvantages. Electrode failures and subsequent patient injury can be attributed mostly to improper application, electrode dislodgment, and electrode defects rather than to electrode design.

81.7 ESU Hazards Improper use of electrosurgery may expose both the patient and the surgical staff to a number of hazards. By far the most frequent hazards are electric shock and undesired burns. Less frequent are undesired neuromuscular stimulation, interference with pacemakers or other devices, electrochemical effects from direct currents, implant heating, and gas explosions [1,3]. Current returns to the ESU through the dispersive electrode. If the contact area of the dispersive electrode is large and the current exposure time short, then the skin temperature under the electrode does not rise above 45°C, which has been shown to be the maximum safe temperature [4]. However, to include a safety margin, the skin temperature should not rise more than 6°C above the normal surface


temperature of 29–33°C. The current density at any point under the dispersive electrode has to be significantly below the recognized burn threshold of 100 mA/cm2 for 10 seconds. To avoid electric shock and burns, the American National Standard for Electrosurgical Devices [5] requires that “any electrosurgical generator that provides for a dispersive electrode and that has a rated output power of greater than 50 W shall have at least one patient circuit safety monitor.” The most common safety monitors are the contact quality monitor for the dispersive electrode and the patient circuit monitor. A contact quality monitor consists of a circuit to measure the impedance between the two sides of a split dispersive electrode and the skin. A small high-frequency current flows from one section of the dispersive electrode through the skin to the second section of the dispersive electrode. If the impedance between these two sections exceeds a certain threshold, or changes by a certain percentage, an audible alarm sounds, and the ESU output is disabled. Patient circuit monitors range from simple to complex. The simple ones monitor electrode cable integrity while the complex ones detect any abnormal condition that could result in electrosurgical current flowing in other than normal pathways. Although the output isolation transformer present in most modern ESUs usually provides adequate patient protection, some potentially hazardous conditions may still arise. If a conductor to the dispersive electrode is broken, undesired arcing between the broken conductor ends may occur, causing fire in the operating room and serious patient injury. Abnormal current pathways may also arise from capacitive coupling between cables, the patient, operators, enclosures, beds, or any other conductive surface or from direct connections to other electrodes connected to the patient. The patient circuit monitoring device should be operated from an isolated power source having a maximum voltage of 12 V rms. The most common device is a cable continuity monitor. Unlike the contact quality monitor, this monitor only checks the continuity of the cable between the ESU and the dispersive electrode and sounds an alarm if the resistance in that conductor is greater than 1 kΩ . Another implementation of a patient circuit monitor measures the voltage between the dispersive electrode connection and ground. A third implementation functions similarly to a ground fault circuit interrupter (GFCI) in that the current in the wire to the active electrode and the current in the wire to the dispersive electrode are measured and compared with each other. If the difference between these currents is greater than a preset threshold, the alarm sounds and the ESU is disconnected. There are other sources of undesired burns. Active electrodes get hot when they are used. After use, the active electrode should be placed in a protective holster, if available, or on a suitable surface to isolate it from the patient and surgical staff. The correct placement of an active electrode will also prevent the patient and/or surgeon from being burned if an inadvertent activation of the ESU occurs (e.g., someone accidentally stepping on a foot pedal). Some surgeons use a practice called buzzing the hemostat in which a small bleeding vessel is grasped with a clamp or hemostat and the active electrode touched to the clamp while activating. Because of the high voltages involved and the stray capacitance to ground, the surgeon’s glove may be compromised. If the surgical staff cannot be convinced to eliminate the practice of buzzing hemostats, the probability of burns can be reduced by use of a cut waveform instead of a coagulation waveform (lower voltage), by maximizing contact between the surgeon’s hand and the clamp, and by not activating until the active electrode is firmly touching the clamp. Although it is commonly assumed that neuromuscular stimulation ceases or is insignificant at frequencies above 10 kHz, such stimulation has been observed in anesthetized patients undergoing certain electrosurgical procedures. This undesirable side effect of electrosurgery is generally attributed to nonlinear events during the electric arcing between the active electrode and tissue. These events rectify the high-frequency current leading to both dc and low-frequency current components. These current components can reach magnitudes that stimulate nerve and muscle cells. To minimize the probability of unwanted neuromuscular stimulation, most ESUs incorporate in their output circuit a high-pass filter that suppresses dc and low-frequency current components. The use of electrosurgery means the presence of electric discharge arcs. This presents a potential fire hazard in an operating room where oxygen and flammable gases may be present. These flammable gases may be introduced by the surgical staff (anesthetics or flammable cleaning solutions), or may be generated


within the patients themselves (bowel gases). The use of disposable paper drapes and dry surgical gauze also provides a flammable material that may be ignited by sparking or by contact with a hot active electrode. Therefore, prevention of fires and explosions depends primarily on the prudence and judgment of the ESU operator.

81.8 Recent Developments Electrosurgery is being enhanced by the addition of a controlled column of argon gas in the path between the active electrode and the tissue. The flow of argon gas assists in clearing the surgical site of fluid and improves visibility. When used in the coagulation mode, the argon gas is turned into a plasma allowing tissue damage and smoke to be reduced, and producing a thinner, more flexible eschar. When used with the cut mode, lower power levels may be used. Many manufacturers have begun to include sophisticated computer-based systems in their ESUs that not only simplify the use of the device but also increase the safety of patient and operator [7]. For instance, in a so-called soft coagulation mode, a special circuit continuously monitors the current between the active electrode and the tissue and turns the ESU output on only after the active electrode has contacted the tissue. Furthermore, the ESU output is turned off automatically, once the current has reached a certain threshold level that is typical for coagulated and desiccated tissue. This feature is also used in a bipolar mode termed autobipolar. Not only does this feature prevent arcing at the beginning of the procedure, but it also keeps the tissue from being heated beyond 70°C. Some devices offer a so-called power-peak-system that delivers a very short power peak at the beginning of electrosurgical cutting to start the cutting arc. Other modern devices use continuous monitoring of current and voltage levels to make automatic power adjustments in order to provide for a smooth cutting action from the beginning of the incision to its end. Some manufacturers are developing waveforms and instruments designed to achieve specific clinical results such as bipolar cutting tissue lesioning, and vessel sealing. With the growth and popularity of laparoscopic procedures, additional electrosurgical instruments and waveforms tailored to this surgical specialty should also be expected. Increased computing power, more sophisticated evaluation of voltage and current waveforms, and the addition of miniaturized sensors will continue to make ESUs more user-friendly and safer.

Defining Terms Active electrode: Electrode used for achieving desired surgical effect. Coagulation: Solidification of proteins accompanied by tissue whitening. Desiccation: Drying of tissue due to the evaporation of intracellular fluids. Dispersive electrode: Return electrode at which no electrosurgical effect is intended. Fulguration: Random discharge of sparks between active electrode and tissue surface in order to achieve coagulation and/or desiccation. Spray: Another term for fulguration. Sometimes this waveform has a higher crest factor than that used for fulguration.

References 1. Pearce John A. 1986. Electrosurgery, New York, John Wiley. 2. Gerhard Glen C. 1988. Electrosurgical unit. In JG Webster (ed), Encyclopedia of Medical Devices and Instrumentation, vol 2, pp 1180–1203, New York, John Wiley. 3. Gendron Francis G. 1988. Unexplained Patient Burns: Investigating Latrogenic Injuries, Brea, Calif, Quest Publishing. 4. Pearce JA, Geddes LA, Van Vleet JF, et al. 1983. Skin burns from electrosurgical current. Med Instrum 17(3):225.


5. American National Standard for Electrosurgical Devices. 1994. HF18, American National Standards Institute. 6. LaCourse JR, Miller WT III, Vogt M, et al. 1985. Effect of high frequency current on nerve and muscle tissue. IEEE Trans Biomed Eng 32:83. 7. Haag R, Cuschieri A. 1993. Recent advances in high-frequency electrosurgery: Development of automated systems. J R Coll Surg Ednb 38:354.

Further Information American National Standards Institute, 1988. International Standard, Medical Electrical Equipment, Part 1: General Requirements for Safety, IEC 601-1, 2d ed, New York. American National Standards Institute. 1991. International Standard, Medical Electrical Equipment, Part 2: Particular Requirements for the Safety of High Frequency Surgical Equipment, IEC 601-2-2, 2d ed, New York. National Fire Protection Association. 1993. Standard for Health Care Facilities, NFPA 99.


Behbehani, K. “Mechanical Ventilation.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

82 Mechanical Ventilation 82.1 82.2 82.3 82.4

Introduction Negative-Pressure Ventilators Positive-Pressure Ventilators Ventilation Modes Mandatory Ventilation • Spontaneous Ventilation

82.5 The University of Texas at Arlington and The University of Texas Southwestern Medical Center at Dallas

Breath Delivery Control Mandatory Volume Controlled Inspiratory Flow Delivery • Pressure Controlled Inspiratory Flow Delivery • Expiratory Pressure Control in Mandatory Mode • Spontaneous Breath Delivery Control

Khosrow Behbehani

82.6

Summary

82.1 Introduction This chapter presents an overview of the structure and function of mechanical ventilators. Mechanical ventilators, which are often also called respirators, are used to artificially ventilate the lungs of patients who are unable to naturally breathe from the atmosphere. In almost 100 years of development, many mechanical ventilators with different designs have been developed [Mushin et al., 1980Philbeam, 1998]. The very early devices used bellows that were manually operated to inflate the lungs. Today’s respirators employ an array of sophisticated components such as microprocessors, fast response servo valves, and precision transducers to perform the task of ventilating the lungs. The changes in the design of ventilators have come about as the result of improvements in engineering the ventilator components and the advent of new therapy modes by clinicians. A large variety of ventilators are now available for short-term treatment of acute respiratory problems as well as long-term therapy for chronic respiratory conditions. It is reasonable to broadly classify today’s ventilators into two groups. The first and indeed the largest group encompasses the intensive care respirators used primarily in hospitals to support patients following certain surgical procedures or assist patients with acute respiratory disorders. The second group includes less complicated machines that are primarily used at home to treat patients with chronic respiratory disorders. The level of engineering design and sophistication for the intensive care ventilators is higher than the ventilators used for chronic treatment. However, many of the engineering concepts employed in designing intensive care ventilators can also be applied in the simpler chronic care units. Therefore, this presentation focuses on the design of intensive care ventilators; the terms respirator, mechanical ventilator, or ventilator that will be used from this point on refer to the intensive care unit respirators. At the beginning, the designers of mechanical ventilators realized that the main task of a respirator was to ventilate the lungs in a manner as close to natural respiration as possible. Since natural inspiration is a result of negative pressure in the pleural cavity generated by distention of the diaphragm, designers initially developed ventilators that created the same effect. These ventilators are called negative-pressure


ventilators. However, more modern ventilators use pressures greater than atmospheric pressures to ventilate the lungs; they are known as positive-pressure ventilators.

82.2 Negative-Pressure Ventilators The principle of operation of a negative-pressure respirator is shown in Fig. 82.1. In this design, the flow of air to the lungs is created by generating a negative pressure around the patient’s thoracic cage. The negative pressure moves the thoracic walls outward expanding the intra-thoracic volume and dropping the pressure inside the lungs. The pressure gradient between the atmosphere and the lungs causes the flow of atmospheric air into the lungs. The inspiratory and expiratory phases of the respiration are controlled by cycling the pressure inside the body chamber between a sub-atmospheric level (inspiration) and the atmospheric level (exhalation). Flow of the breath out of the lungs during exhalation is caused by the recoil FIGURE 82.1 A simplified illustration of a negative-presof thoracic muscles. sure ventilator. Although it may appear that the negativepressure respirator incorporates the same principles as natural respiration, the engineering implementation of this concept has not been very successful. A major difficulty has been in the design of a chamber for creating negative pressure around the thoracic walls. One approach has been to make the chamber large enough to house the entire body with the exception of the head and neck. Using foam rubber around the patient’s neck, one can seal the chamber and generate a negative pressure inside the chamber. This design configuration, commonly known as the iron lung, was tried back in the 1920s and proved to be deficient in several aspects. The main drawback was that the negative pressure generated inside the chamber was applied to the chest as well as the abdominal wall, thus creating a venous blood pool in the abdomen and reducing cardiac output. More recent designs have tried to restrict the application of the negative pressure to the chest walls by designing a chamber that goes only around the chest. However, this has not been successful because obtaining a seal around the chest wall (Fig. 82.1) is difficult. Negative-pressure ventilators also made the patient less accessible for patient care and monitoring. Further, synchronization of the machine cycle with the patient’s effort was has been difficult and they are also typically noisy and bulky [McPherson, & Spearman 1990]. These deficiencies of the negativepressure ventilators have led to the development of the positive-pressure ventilators.

82.3 Positive-Pressure Ventilators Positive-pressure ventilators generate the inspiratory flow by applying a positive pressure (greater than the atmospheric pressure) to the airways. Figure 82.2 shows a simplified block diagram of a positivepressure ventilator. During inspiration, the inspiratory flow delivery system creates a positive pressure in the tubes connected to the patient airway, called patient circuit, and the exhalation control system closes a valve at the outlet of the tubing to the atmosphere. When the ventilator switches to exhalation, the inspiratory flow delivery system stops the positive pressure and the exhalation system opens the valve to allow the patient’s exhaled breath to flow to the atmosphere. The use of a positive pressure gradient


FIGURE 82.2

A simplified diagram of the functional blocks of a positive-pressure ventilator.

in creating the flow allows treatment of patients with high lung resistance and low compliance. As a result, positive-pressure ventilators have been very successful in treating a variety of breathing disorders and have become more popular than negative-pressure ventilators. Positive-pressure ventilators have been employed to treat patients ranging from neonates to adults. Due to anatomical differences between various patient populations, the ventilators and their modes of treating infants are different than those for adults. Nonetheless, their fundamental design principles are similar and adult ventilators comprise a larger percentage of ventilators manufactured and used in clinics. Therefore, the emphasis here is on the description of adult positive-pressure ventilators. Also, the concepts presented will be illustrated using a microprocessor-based design example, as almost all modern ventilators use microprocessor instrumentation.

82.4 Ventilation Modes Since the advent of respirators, clinicians have devised a variety of strategies to ventilate the lungs based on patient conditions. For instance, some patients need the respirator to completely take over the task of ventilating their lungs. In this case, the ventilator operates in mandatory mode and delivers mandatory breaths. On the other hand, some patients are able to initiate a breath and breathe on their own, but may need oxygen-enriched air flow or slightly elevated airway pressure. When a ventilator assists a patient who is capable of demanding a breath, the ventilator delivers spontaneous breaths and operates in spontaneous mode. In many cases, it is first necessary to treat the patient with mandatory ventilation and as the patient’s condition improves spontaneous ventilation is introduced; it is used primarily to wean the patient from mandatory breathing.

Mandatory Ventilation Designers of adult ventilators have employed two rather distinct approaches for delivering mandatory breaths: volume controlled ventilation and pressure controlled ventilation. Volume controlled ventilation, which presently is more popular, refers to delivering a specified tidal volume to the patient during the inspiratory phase. Pressure controlled ventilation, however, refers to raising the airway pressure to a level, set by the therapist, during the inspiratory phase of each breath. Regardless of the type, a ventilator operating in mandatory mode must control all aspects of breathing such as tidal volume, respiration rate, inspiratory flow pattern, and oxygen concentration of the breath. This is often labeled as controlled mandatory ventilation (CMV). Figure 82.3 shows the flow and pressure waveforms for a volume controlled ventilation (CMV). In this illustration, the inspiratory flow waveform is chosen to be a half sinewave. In Fig. 82.3a, ti is the inspiration duration, te is the exhalation period, and Qi is the amplitude of inspiratory flow. The ventilator


FIGURE 82.3 (a) Inspiratory flow for a controlled mandatory volume controlled ventilation breath, (b) airway pressure resulting from the breath delivery with a non-zero PEEP.

delivers a tidal volume equal to the area under the flow waveform in Fig. 82.3a at regular intervals (ti + te) set by the therapist. The resulting pressure waveform is shown in Fig. 82.3b. It is noted that during volume controlled ventilation, the ventilator delivers the same volume irrespective of the patient’s respiratory mechanics. However, the resulting pressure waveform such as the one shown in Fig. 82.3b, will be different among patients. Of course, for safety purposes, the ventilator limits the maximum applied airway pressure according to the therapist’s setting. As can be seen in Fig. 82.3b, the airway pressure at the end of exhalation may not end at atmospheric pressure (zero gauge). The positive end expiratory pressure (PEEP) is sometimes used to keep the alveoli from collapsing during expiration [Norwood, 1990]. In other cases, the expiration pressure is allowed to return to the atmospheric level. Figure 82.4 shows a plot of the pressure and flow during a mandatory pressure controlled ventilation. In this case, the respirator raises and maintains the airway pressure at the desired level independent of

FIGURE 82.4 (a) Inspiratory pressure pattern for a controlled mandatory pressure controlled ventilation breath, (b) airway flow pattern resulting from the breath delivery. Note that PEEP is zero.


patient airway compliance and resistance. The level of pressure during inspiration, Pi, is set by the therapist. While the ventilator maintains the same pressure trajectory for patients with different respiratory resistance and compliance, the resulting flow trajectory, shown in Fig. 82.4b, will depend on the respiratory mechanics of each patient. In the following, the presentation will focus on volume ventilators, as they are more common. Further, in a microprocessor-based ventilator, the mechanism for delivering mandatory volume and pressure controled ventilation have many similar main components. The primary difference lies in the control algorithms governing the delivery of breaths to the patient.

Spontaneous Ventilation An important phase in providing respiratory therapy to a recovering pulmonary patient is weaning the patient from the respirator. As the patient recovers and gains the ability to breathe independently, the ventilator must allow the patient to initiate a breath and control the breath rate, flow rate, and the tidal volume. Ideally, when a respirator is functioning in the spontaneous mode, it should let the patient take breaths with the same ease as breathing from the atmosphere. This, however, is difficult to achieve because the respirator does not have an infinite gas supply or an instantaneous response. In practice, the patient generally has to exert more effort to breathe spontaneously on a respirator than from the atmosphere. However, patient effort is reduced as the ventilator response speed increases [McPherson, 1990]. Spontaneous ventilation is often used in conjunction with mandatory ventilation since the patient may still need breaths that are delivered entirely by the ventilator. Alternatively, when a patient can breathe completely on his own but needs oxygen-enriched breath or elevated airway pressure, spontaneous ventilation alone may be used. As in the case of mandatory ventilation, several modes of spontaneous ventilation have been devised by therapists. Two of the most important and popular spontaneous breath delivery modes are described below. Continuous Positive Airway Pressure (CPAP) in Spontaneous Mode In this mode, the ventilator maintains a positive pressure at the airway as the patient attempts to inspire. Figure 82.5 illustrates a typical airway pressure waveform during CPAP breath delivery. The therapist sets the sensitivity level lower than PEEP. When the patient attempts to breathe, the pressure drops below the sensitivity level and the ventilator responds by supplying breathable gases to raise the pressure back to the PEEP level. Typically, the PEEP and sensitivity levels are selected such that the patient will be impelled to exert effort to breathe independently. As in the case of the mandatory mode, when the patient exhales the ventilator shuts off the flow of gas and opens the exhalation valve to allow the exhaled gases to flow into the atmosphere. Pressure Support in Spontaneous Mode This mode is similar to the CPAP mode with the exception that during the inspiration the ventilator attempts to maintain the patient airway pressure at a level above PEEP. In fact, CPAP may be considered a special case of pressure support ventilation in which the support level is fixed at the atmospheric level.

FIGURE 82.5


Airway pressure during a CPAP spontaneous breath delivery.

FIGURE 82.6

Airway pressure during a pressure support spontaneous breath delivery.

Figure 82.6 shows a typical airway pressure waveform during the delivery of a pressure support breath. In this mode, when the patient’s airway pressure drops below the therapist-set sensitivity line, the ventilator inspiratory breath delivery system raises the airway pressure to the pressure support level (>PEEP), selected by the therapist. The ventilator stops the flow of breathable gases when the patient starts to exhale and controls the exhalation valve to achieve the set PEEP level.

82.5 Breath Delivery Control Figure 82.7 shows a simplified block diagram for delivering mandatory or spontaneous ventilation. Compressed air and oxygen are normally stored in high pressure tanks (≅1400 kPa) that are attached to the inlets of the ventilator. In some ventilators, an air compressor is used in place of a compressed air tank. Manufacturers of mechanical respirators have designed a variety of blending and metering devices [McPherson, 1990]. The primary mission of the device is to enrich the inspiratory air flow with the proper level of oxygen and to deliver a tidal volume according to the therapist’s specifications. With the

FIGURE 82.7

A simplified block diagram of a control structure for mandatory and spontaneous breath delivery.


introduction of microprocessors for control of metering devices, electromechanical valves have gained popularity [Puritan-Bennett, 1987]. In Fig. 82.7, the air and oxygen valves are placed in closed feedback loops with the air and oxygen flow sensors. The microprocessor controls each the valves to deliver the desired inspiratory air and oxygen flows for mandatory and spontaneous ventilation. During inhalation, the exhalation valve is closed to direct all the delivered flows to the lungs. When exhalation starts, the microprocessor actuates the exhalation valve to achieve the desired PEEP level. The airway pressure sensor, shown on the right side of Fig. 82.7, generates the feedback signal necessary for maintaining the desired PEEP (in both mandatory and spontaneous modes) and airway pressure support level during spontaneous breath delivery.

Mandatory Volume Controlled Inspiratory Flow Delivery In a microprocessor-controlled ventilator (Fig. 82.7), the electronically actuated valves open from a closed position to allow the flow of blended gases to the patient. The control of flow through each valve depends on the therapist’s specification for the mandatory breath. That is, the clinician must specify the following parameters for delivery of CMV breaths: (1) respiration rate; (2) flow waveform; (3) tidal volume; (4) oxygen concentration (of the delivered breath); (5) peak flow; and (6) PEEP, as shown in the lower left side of Fig. 82.7. It is noted that the PEEP selected by the therapist in the mandatory mode is only used for control of exhalation flow; that will be described in the following section. The microprocessor utilizes the first five of the above parameters to compute the total desired inspiratory flow trajectory. To illustrate this point, consider the delivery of a tidal volume using a half sinewave as shown in Fig. 82.3. If the therapist selects a tidal volume of Vt (L), a respiration rate of n breaths per minute (bpm), the amplitude of the respirator flow, Qi (L/s), then the total desired inspiratory flow, Q d (t), for a single breath, can be computed from the following equation:

 πt Qi sin Qd t =  ti 0 

()

o ≤ t < ti

(82.1)

ti < t ≤ te

where ti signifies the duration of inspiration and is computed from the following relationship:

ti =

Vt . 2Qi

(82.2)

The duration of expiration in seconds is obtained from

te =

60 − ti . n

(82.3)

The ratio of inspiratory to expiratory periods of a mandatory breath is often used for adjusting the respiration rate. This ratio is represented by I:E (ratio) and is computed as follows. First, the inspiratory and expiratory periods are normalized with respect to ti . Hence, the normalized inspiratory period becomes unity and the normalized expiratory period is given by R = te /ti . Then, the I:E ratio is simply expressed as 1:R. To obtain the desired oxygen concentration in the delivered breath, the microprocessor computes the discrete form of Qd (t) as Qd (k) where k signifies the kth sample interval. Then, the total desired flow, Qd (k), is partitioned using the following relationships:


1− m Q k ( ) ( (1 −) c ) ( )

(82.4)

m−c Q k ( ) ( (1 −) c ) ( )

(82.5)

d

Qda k =

and

d

Qdx k =

where k signifies the sample interval, Qda(k)is the desired air flow (the subscript da stands for desired air), Q dx(k) is the desired oxygen flow (the subscript dx stands for desired oxygen), m is the desired oxygen concentration, and c is the oxygen concentration of the ventilator air supply. A number of control design strategies may be appropriate for the control of the air and oxygen flow delivery valves. A simple controller is the proportional plus integral controller that can be readily implemented in a microprocessor. For example, the controller for the air valve has the following form:

()

()

()

()

()

()

(82.7)

() ( ) ()

(82.8)

I k = K pE k + K i A k

(82.6)

where E(k) and A(k) are given by

E k = Qda k − Qsa k

A k = A k −1 + E k

where I(k) is the input (voltage or current) to the air valve at the kth sampling interval, E(k) is the error in the delivered flow, Qda(k) is the desired air flow, Qsa(k) is the sensed or actual air flow (the subscript sa stands for sensed air flow), A(k) is the integral (rectangular integration) part of the controller, and Kp and Ki are the controller proportionality constants. It is noted that the above equations are applicable to the control of either the air or oxygen valve. For control of the oxygen flow valve, Qdx(k) replaces Qda(k) and Qsx(k) replaces Qsa(k) where Qsx(k) represents the sensed oxygen flow (the subscript sx stands for sensed oxygen flow). The control structure shown in Fig. 82.7 provides the flexibility of quickly adjusting the percentage of oxygen in the enriched breath gases. That is, the controller can regulate both the total flow and the percent oxygen delivered to the patient. Since the internal volume of the flow control valve is usually small (< 50 ml), the desired change in the oxygen concentration of the delivered flow can be achieved within one inspiratory period. In actual clinical applications, rapid change of percent oxygen from one breath to another is often desirable, as it reduces the waiting time for the delivery of the desired oxygen concentration. A design similar to the one shown in Fig. 82.7 has been successfully implemented in a microprocessor-based ventilator [Behbehani, 1984] and is deployed in hospitals around the world.

Pressure Controlled Inspiratory Flow Delivery The therapist entry for pressure-controlled ventilation is shown in Fig. 82.7 (lower left-hand side). In contrast to the volume-controlled ventilation where Qd(t) was computed directly from the operator’s entry [Eq. (82.1) through Eq. (82.23)], the total desired flow is generated by a closed loop controller labeled as Airway Pressure Controller in Fig. 82.7. This controller uses the therapist-selected inspiratory pressure, respiration rate, and the I:E ratio to compute the desired inspiratory pressure trajectory. The


trajectory serves as the controller reference input. The controller then computes the flow necessary to make the actual airway pressure track the reference input. Assuming a proportional-plus-integral controller, the governing equations are

()

()

()

Qd k = C pE p k + Ci Ap k

(82.9)

where Qd is the computed desired flow, Cp and Ci are the proportionality constants, k represents the sample interval, and Ep(k) and Ap(k) are computed using the following equations:

() () ()

E p k = Pd k − Ps k

(82.10)

()

(82.11)

( ) ()

Ap k = Ap k − 1 + E p k

where Ep(k) is the difference between the desired pressure trajectory, Pd(k), and the sensed airway pressure, Ps(k), the parameter Ap(k) represents the integral portion of the controller. Using Q d from Eq. (82.9), the control of air and O2 valves is accomplished in the same manner as in the case of volume-controlled ventilation described earlier [Eq. (82.4) through Eq. (82.8)].

Expiratory Pressure Control in Mandatory Mode It is often desirable to keep the patient’s lungs inflated at the end of expiration at a pressure greater than atmospheric level [Norwood, 1990]. That is, rather than allowing the lungs to deflate during the exhalation, the controller closes the exhalation valve when the airway pressure reaches the PEEP level. When expiration starts, the ventilator terminates flow to the lungs; hence, the regulation of the airway pressure is achieved by controlling the flow of patient exhaled gases through the exhalation valve. In a microprocessor-based ventilator, an electronically actuated valve can be employed that has adequate dynamic response (20 ms rise time) to regulate PEEP. For this purpose, the pressure in the patient breath delivery circuit is measured using a pressure transducer (Fig. 82.7). The microprocessor will initially open the exhalation valve completely to minimize resistance to expiratory flow. At the same time, it will sample the pressure transducer’s output and start to close the exhalation valve as the pressure begins to approach the desired PEEP level. Since the patient’s exhaled flow is the only source of pressure, if the airway pressure drops below PEEP, it cannot be brought back up until the next inspiratory period. Hence, an overrun (i.e., a drop to below PEEP) in the closed-loop control of PEEP cannot be tolerated.

Spontaneous Breath Delivery Control The small diameter (≅5 mm) pressure sensing tube, shown on the right side of Fig. 82.7, pneumatically transmits the pneumatic pressure signal from the patient airway to a pressure transducer placed in the ventilator. The output of the pressure transducer is amplified, filtered, and then sampled by the microprocessor. The controller receives the therapist’s inputs regarding the spontaneous breath characteristics such as the PEEP, sensitivity, and oxygen concentration, as shown on the lower right-hand side of Fig. 82.7. The desired airway pressure is computed from the therapist entries of PEEP, pressure support level, and sensitivity. The multiple-loop control structure shown in Fig. 82.7 is used to deliver a CPAP or a pressure support breath. The sensed proximal airway pressure is compared with the desired airway pressure. The airway pressure controller computes the total inspiratory flow level required to raise the airway pressure to the desired level. This flow level serves as the reference input or total desired flow for the flow control loop. Hence, in general, the desired total flow trajectory for the spontaneous breath delivery may be different for each inspiratory cycle. If the operator has specified oxygen concentration greater than 21.6% (the atmospheric air oxygen concentration of the ventilator air supply), the controller will partition the


total required flow into the air and oxygen flow rates using Eqs. (82.4) and (82.5). The flow controller then uses the feedback signals from air and oxygen flow sensors and actuates the air and oxygen valves to deliver the desired flows. For a microprocessor-based ventilator, the control algorithm for regulating the airway pressure can also be a proportional plus integral controller [Behbehani, 1984; Behbehani and Watanabe, 1986]. In this case, the governing equations are identical to Eqs. (82.9) through (82.11). If a non-zero PEEP level is specified, the same control strategy as the one described for mandatory breath delivery can be used to achieve the desired PEEP.

82.6 Summary Today’s mechanical ventilators can be broadly classified into negative-pressure and positive-pressure ventilators. Negative-pressure ventilators do not offer the flexibility and convenience that positive-pressure ventilators provide; hence, they have not been very popular in clinical use. Positive-pressure ventilators have been quite successful in treating patients with pulmonary disorders. These ventilators operate in either mandatory or spontaneous mode. When delivering mandatory breaths, the ventilator controls all parameters of the breath such as tidal volume, inspiratory flow waveform, respiration rate, and oxygen content of the breath. Mandatory breaths are normally delivered to the patients that are incapable of breathing on their own. In contrast, spontaneous breath delivery refers to the case where the ventilator responds to the patient’s effort to breathe independently. Therefore, the patient can control the volume and the rate of the respiration. The therapist selects the oxygen content and the pressure at which the breath is delivered. Spontaneous breath delivery is typically used for patients who are on their way to full recovery, but are not completely ready to breathe from the atmosphere without mechanical assistance.

Defining Terms Continuous positive airway pressure (CPAP): A spontaneous ventilation mode in which the ventilator maintains a constant positive pressure, near or below PEEP level, in the patient’s airway while the patient breathes at will. I:E ratio: The ratio of normalized inspiratory interval to normalized expiratory interval of a mandatory breath. Both intervals are normalized with respect to the inspiratory period. Hence, the normalized inspiratory period is always unity. Mandatory mode: A mode of mechanically ventilating the lungs where the ventilator controls all breath delivery parameters such as tidal volume, respiration rate, flow waveform, etc. Patient circuit: A set of tubes connecting the patient airway to the outlet of a respirator. Positive end expiratory pressure (PEEP): A therapist-selected pressure level for the patient airway at the end of expiration in either mandatory or spontaneous breathing. Pressure controlled ventilation: A mandatory mode of ventilation where during the inspiration phase of each breath, a constant pressure is applied to the patient’s airway independent of the patient’s airway resistance and/or compliance.respiratory mechanics. Pressure support: A spontaneous breath delivery mode during which the ventilator applies a positive pressure greater than PEEP to the patient’s airway during inspiration. Pressure support level: Refers to the pressure level, above PEEP, that the ventilator maintains during the spontaneous inspiration. Spontaneous mode: A ventilation mode in which the patient initiates and breathes from the ventilatorsupplied gas at will. Volume controlled ventilation: A mandatory mode of ventilation where the volume of each breath is set by the therapist and the ventilator delivers that volume to the patient independent of the patient’s airway resistance and/or compliance.respiratory mechanics.


References Puritan-Bennett 7200 Ventilator System Series, “Ventilator, Options and Accessories”, Part. No. 22300A, Carlsbad, CA, September 1990. Pilbeam, S.P., 1998. Mechanical Ventilation: Physiological and Clinical Applications, 3rd ed, Mosby, St. Louis, MO. McPherson, S.P. and Spearman, C.B. 1990. Respiratory Therapy Equipment, 4th ed. C.V. Mosby Co., St. Louis, MO. Norwood, S. 1990. Physiological Principles of Conventional Mechanical Ventilation. In Clinical Application of Ventilatory Support, Kirby, R.R., Banner, M.J., and Downs, J.B., Eds., Churchill Livingstone, New York, 145-172. Behbehani, K. 1984. PLM-Implementation of a Multiple Closed-Loop Control Strategy for a Microprocessor-Controlled Respirator. Proc. ACC Conf., p. 574-576. Behbehani, K. and Watanabe, N.T. 1986. A New Application of Digital Computer Simulation in the Design of a Microprocessor-Based Respirator. Summer Simulation Conf., p. 415-420.


Voss, G. I., Butterfield, R. D. “Parenteral Infusion Devices.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

83 Parenteral Infusion Devices 83.1

Gregory I. Voss

83.2 83.3

Gravity Flow/Resistance Regulation • Volumetric Infusion Pumps • Controllers • Syringe Pumps

IVAC Corporation

Robert D. Butterfield IVAC Corporation

Performance Criteria for Intravenous Infusion Devices Flow Through an IV Delivery System Intravenous Infusion Devices

83.4 83.5

Managing Occlusions of the Delivery System Summary

The circulatory system is the body’s primary pathway for both the distribution of oxygen and other nutrients and the removal of carbon dioxide and other waste products. Since the entire blood supply in a healthy adult completely circulates within 60 seconds, substances introduced into the circulatory system are distributed rapidly. Thus intravenous (IV) and intraarterial access routes provide an effective pathway for the delivery of fluid, blood, and medicants to a patient’s vital organs. Consequently, about 80% of hospitalized patients receive infusion therapy. Peripheral and central veins are used for the majority of infusions. Umbilical artery delivery (in neonates), enteral delivery of nutrients, and epidural delivery of anesthetics and analgesics comprise smaller patient populations. A variety of devices can be used to provide flow through an intravenous catheter. An intravenous delivery system typically consists of three major components: (1) fluid or drug reservoir, (2) catheter system for transferring the fluid or drug from the reservoir into the vasculature through a venipuncture, and (3) device for regulation and/or generating flow (see Fig. 83.1). This chapter is separated into five sections. The first describes the clinical needs associated with intravenous drug delivery that determine device performance criteria. The second section reviews the principles of flow through a tube; the third section introduces the underlying electromechanical principles for flow regulation and/or generation and their ability to meet the clinical performance criteria. The fourth section reviews complications associated with intravenous therapy, and the fifth section concludes with a short list of articles providing more detailed information.

83.1 Performance Criteria for Intravenous Infusion Devices The intravenous pathway provides an excellent route for continuous drug therapy. The ideal delivery system regulates drug concentration in the body to achieve and maintain a desired result. When the drug’s effect cannot be monitored directly, it is frequently assumed that a specific blood concentration or infusion rate will achieve the therapeutic objective. Although underinfusion may not provide sufficient therapy, overinfusion can produce even more serious toxic side effects.


FIGURE 83.1

A typical IV infusion system.

The therapeutic range and risks associated with under- and overinfusion are highly drug and patient dependent. Intravenous delivery of fluids and electrolytes often does not require very accurate regulation. Low-risk patients can generally tolerate well infusion rate variability of ±30% for fluids. In some situations, however, specifically for fluid-restricted patients, prolonged under- or overinfusion of fluids can compromise the patient’s cardiovascular and renal systems. The infusion of many drugs, especially potent cardioactive agents, requires high accuracy. For example, post-coronary-artery-bypass-graft patients commonly receive sodium nitroprusside to lower arterial blood pressure. Hypertension, associated with underinfusion, subjects the graft sutures to higher stress with an increased risk for internal bleeding. Hypotension associated with overinfusion can compromise the cardiovascular state of the patient. Nitroprusside’s potency, short onset delay, and short half-life (30–180 s) provide for very tight control, enabling the clinician to quickly respond to the many events that alter the patient’s arterial pressure. The fast response of drugs such as nitroprusside creates a need for short-term flow uniformity as well as long-term accuracy. The British Department of Health employs Trumpet curves in their Health Equipment Information reports to compare flow uniformity of infusion pumps. For a prescribed flow rate, the trumpet curve is the plot of the maximum and minimum measured percentage flow rate error as a function of the accumulation interval (Fig. 83.2). Flow is measured gravimetrically in 30-s blocks for 1 hour. These blocks are summed to produce 120-s, 300-s, and other longer total accumulation intervals. Though the 120-s window may not detect flow variations important in delivery of the fastest acting agents, the trumpet curve provides a helpful means for performance comparison among infusion devices. Additional statistical information such as standard deviations may be derived from the basic trumpet flow measurements. The short half-life of certain pharmacologic agents and the clotting reaction time of blood during periods of stagnant flow require that fluid flow be maintained without significant interruption. Specifically, concern has been expressed in the literature that the infusion of sodium nitroprusside and other short half-life drugs occur without interruption exceeding 20 s. Thus, minimization of false alarms and rapid detection of occlusions are important aspects of maintaining a constant vascular concentration. © 2000 by CRC Press LLC

FIGURE 83.2 Trumpet curve for several representative large volume infusion pumps operated a 5 mL/hr. Note that peristaltic pumps were designed for low risk patients.

Accidental occlusions of the IV line due to improper positioning of stopcocks or clamps, kinked tubing, and clotted catheters are common. Occlusions between pump and patient present a secondary complication in maintaining serum drug concentration. Until detected, the pump will infuse, storing fluid in the delivery set. When the occlusion is eliminated, the stored volume is delivered to the patient in a bolus. With concentrated pharmaceutic agents, this bolus can produce a large perturbation in the patient’s status. Occlusions of the pump intake also interrupt delivery. If detection is delayed, inadequate flow can result. During an intake occlusion, in some pump designs removal of the delivery set can produce abrupt aspiration of blood. This event may precipitate clotting and cause injury to the infusion site. The common practice of delivering multiple drugs through a single venous access port produces an additional challenge to maintaining uniform drug concentration. Although some mixing will occur in the venous access catheter, fluid in the catheter more closely resembles a first-in/first-out digital queue: During delivery, drugs from the various infusion devices mix at the catheter input, an equivalent fluid volume discharges from the outlet. Rate changes and flow nonuniformity cause the mass flow of drugs at the outlet to differ from those at the input. Consider a venous access catheter with a volume of 2 mL and a total flow of 10 mL/hr. Due to the digital queue phenomenon, an incremental change in the intake flow rate of an individual drug will not appear at the output for 12 min. In addition, changing flow rates for one drug will cause short-term marked swings in the delivery rate of drugs using the same access catheter. When the delay becomes significantly larger than the time constant for a drug that is titrated to a measurable patient response, titration becomes extremely difficult leading to large oscillations. As discussed, the performance requirements for drug delivery vary with multiple factors: drug, fluid restriction, and patient risk. Thus the delivery of potent agents to fluid-restricted patients at risk require the highest performance standards defined by flow rate accuracy, flow rate uniformity, and ability to minimize risk of IV-site complications. These performance requirements need to be appropriately balanced with the device cost and the impact on clinician productivity.

83.2 Flow Through an IV Delivery System The physical properties associated with the flow of fluids through cylindrical tubes provide the foundation for understanding flow through a catheter into the vasculature. Hagen-Poiseuille’s equation for laminar flow of a Newtonian fluid through a rigid tube states

Q = π ⋅ r4 ⋅

(P − P ) 1

2

8⋅ η⋅ L

where Q is the flow; P1 and P2 are the pressures at the inlet and outlet of the tube, respectively; L and r are the length and internal radius of the tube, respectively; and η is fluid viscosity. Although many drug delivery systems do not strictly meet the flow conditions for precise application of the laminar flow equation, it does provide insight into the relationship between flow and pressure in a catheter. The fluid analog of Ohms law describes the resistance to flow under constant flow conditions © 2000 by CRC Press LLC

TABLE 83.1 Infusion

Resistance Measurements for Catheter Components Used for

Component

Length, cm

Standard administration set Extension tube for CVP monitoring 19-gauge epidural catheter 18-gauge needle 23-gauge needle 25-gauge needle Vicra Quick-Cath Catheter 18-gauge Extension set with 0.22 micron air-eliminating filter 0.2 micron filter

91–213 15 91 6–9 2.5–9 1.5–4.0 5

Flow Resistance, Fluid Ohm, mmHg/(L/hr) 4.3–5.3 15.5 290.4–497.1 14.1–17.9 165.2–344.0 525.1–1412.0 12.9 623.0 555.0

Note: Mean values are presented over a range of infusions (100, 200, and 300 mL/hr) and sample size (n = 10).

R=

P1 − P2 Q

Thus, resistance to flow through a tube correlates directly with catheter length and fluid viscosity and inversely with the fourth power of catheter diameter. For steady flow, the delivery system can be modeled as a series of resistors representing each component, including administration set, access catheter, and circulatory system. When dynamic aspects of the delivery system are considered, a more detailed model including catheter and venous compliance, fluid inertia, and turbulent flow is required. Flow resistance may be defined with units of mmHg/(L/hr), so that 1 fluid ohm = 4.8 × 10–11 Pa s/m3. Studies determining flow resistance for several catheter components with distilled water for flow rates of 100, 200, and 300 mL/hr appear in Table 83.1.

83.3 Intravenous Infusion Devices From Hagen-Poiselluie’s equation, two general approaches to intravenous infusion become apparent. First, a hydrostatic pressure gradient can be used with adjustment of delivery system resistance controlling flow rate. Complications such as partial obstructions result in reduced flow which may be detected by an automatic flow monitor. Second, a constant displacement flow source can be used. Now complications may be detected by monitoring elevated fluid pressure and/or flow resistance. At the risk of overgeneralization, the relative strengths of each approach will be presented.

Gravity Flow/Resistance Regulation The simplest means for providing regulated flow employs gravity as the driving force with a roller clamp as controlled resistance. Placement of the fluid reservoir 60–100 cm above the patient’s right atrium provides a hydrostatic pressure gradient Ph equal to 1.34 mmHg per cm of elevation. The modest physiologic mean pressure in the veins, Pv, minimally reduces the net hydrostatic pressure gradient. The equation for flow becomes

Q=

Ph − Pv Rmfr + Rn

where Rmfr and Rn are the resistance to flow through the mechanical flow regulator and the remainder of the delivery system, respectively. Replacing the variables with representative values for an infusion of 5% saline solution into a healthy adult at 100 mL/hr yields © 2000 by CRC Press LLC

FIGURE 83.3 Drift in flow rate (mean ± standard deviation) over a four-hour period for three mechanical flow regulators at initial flow rates of 10, 60, and 120 mL/hr with distilled water at constant hydrostatic pressure gradient.

100 mL hr =

(68 − 8)mmHg (550+50) mmHg (L hr)

Gravity flow cannot be used for arterial infusions since the higher vascular pressure exceeds available hydrostatic pressure. Flow stability in a gravity infusion system is subject to variations in hydrostatic and venous pressure as well as catheter resistance. However, the most important factor is the change in flow regulator resistance caused by viscoelastic creep of the tubing wall (see Fig. 83.3). Caution must be used in assuming that a preset flow regulator setting will accurately provide a predetermined rate. The clinician typically estimates flow rate by counting the frequency of drops falling through an in-line drip-forming chamber, adjusting the clamp to obtain the desired drop rate. The cross-sectional area of the drip chamber orifice is the major determinant of drop volume. Various manufacturers provide minidrip sets designed for pediatric (e.g., 60 drops/mL) and regular sets designed for adult (10–20 drops/mL) patients. Tolerances on the drip chamber can cause a 3% error in minidrip sets and a 17% error in regular sets at 125 mL/hr flow rate with 5% dextrose in water. Mean drop size for rapid rates increased by as much as 25% over the size of drops which form slowly. In addition, variation in the specific gravity and surface tension of fluids can provide an additional large source of drop size variability. Some mechanical flow regulating devices incorporate the principle of a Starling resistor. In a Starling device, resistance is proportional to hydrostatic pressure gradient. Thus, the device provides a negative feedback mechanism to reduce flow variation as the available pressure gradient changes with time. Mechanical flow regulators comprise the largest segment of intravenous infusion systems, providing the simplest means of operation. Patient transport is simple, since these devices require no electric power. Mechanical flow regulators are most useful where the patient is not fluid restricted and the acceptable therapeutic rate range of the drug is relatively wide with minimal risk of serious adverse sequelae. The most common use for these systems is the administration of fluids and electrolytes.

Volumetric Infusion Pumps Active pumping infusion devices combine electronics with a mechanism to generate flow. These devices have higher performance standards than simple gravity flow regulators. The Association for the Advancement of Medical Instrumentation (AAMI) recommends that long-term rate accuracy for infusion pumps remain within ±10% of the set rate for general infusion and, for the more demanding applications, that long-term flow remain within ±5%. Such requirements typically extend to those agents with narrow therapeutic indices and/or low flow rates, such as the neonatal population or other fluid-restricted © 2000 by CRC Press LLC

patients. The British Department of Health has established three main categories for hospital-based infusion devices: neonatal infusions, high-risk infusions, and low-risk infusions. Infusion control for neonates requires the highest performance standards, because their size severely restricts fluid volume. A fourth category, ambulatory infusion, pertains to pumps worn by patients.

Controllers These devices automate the process of adjusting the mechanical flow regulator. The most common controllers utilize sensors to count the number of drops passing through the drip chamber to provide flow feedback for automatic rate adjustment. Flow rate accuracy remains limited by the rate and viscosity dependence of drop size. Delivery set motion associated with ambulation and improper angulation of the drip chamber can also hinder accurate rate detection. An alternative to the drop counter is a volumetric metering chamber. A McGaw Corporation controller delivery set uses a rigid chamber divided by a flexible membrane. Instrument-controlled valves allow fluid to fill one chamber from the fluid reservoir, displacing the membrane driving the fluid from the second chamber toward the patient. When inlet and outlet valves reverse state, the second chamber is filled while the first chamber delivers to the patient. The frequency of state change determines the average flow rate. Volumetric accuracy demands primarily on the dimensional tolerances of the chamber. Although volumetric controllers may provide greater accuracy than drop-counting controllers, their disposables are inherently more complex, and maximum flow is still limited by head height and system resistance. Beyond improvements in flow rate accuracy, controllers should provide an added level of patient safety by quickly detecting IV-site complications. The IVAC Corporation has developed a series of controllers employing pulsed modulated flow providing for monitoring of flow resistance as well as improved accuracy. The maximum flow rate achieved by gravimetric based infusion systems can become limited by Rn and by concurrent infusion from other sources through the same catheter. In drop-counting devices, flow rate uniformity suffers at low flow rates from the discrete nature of the drop detector. In contrast with infusion controllers, pumps generate flow by mechanized displacement of the contents of a volumetric chamber. Typical designs provide high flow rate accuracy and uniformity for a wide rate range (0.1–1000.0 mL/hr) of infusion rates. Rate error correlates directly with effective chamber volume, which, in turn, depends on both instrument and disposable repeatability, precision, and stability under varying load. Stepper or servo-controlled dc motors are typically used to provide the driving force for the fluid. At low flow rates, dc motors usually operate in a discrete stepping mode. On average, each step propels a small quanta of fluid toward the patient. Flow rate uniformity therefore is a function of both the average volume per quanta and the variation in volume. Mechanism factors influencing rate uniformity include: stepping resolution, gearing and activator geometries, volumetric chamber coupling geometry, and chamber elasticity. When the quanta volume is not inherently uniform over the mechanism’s cycle, software control has been used to compensate for the variation.

Syringe Pumps These pumps employ a syringe as both reservoir and volumetric pumping chamber. A precision leadscrew is used to produce constant linear advancement of the syringe plunger. Except for those ambulatory systems that utilize specific microsyringes, pumps generally accept syringes ranging in size from 5–100 mL. Flow rate accuracy and uniformity are determined by both mechanism displacement characteristics and tolerance on the internal syringe diameter. Since syringe mechanisms can generate a specified linear travel with less than 1% error, the manufacturing tolerance on the internal cross-sectional area of the syringe largely determines flow rate accuracy. Although syringes can be manufactured to tighter tolerances, standard plastic syringes provide long-term accuracy of ±5%. Flow rate uniformity, however, can benefit from the ability to select syringe size (see Fig. 83.4). Since many syringes have similar stroke length, diameter variation provides control of volume. Also the linear advancement per step is typically


FIGURE 83.4

Effect of syringe type on Trumpet curve of a syringe pump at 1 mL/hr.

fixed. Therefore selection of a lower-volume syringe provides smaller-volume quanta. This allows tradeoffs among drug concentration, flow rate, and duration of flow per syringe. Slack in the gear train and drive shaft coupling as well as plunger slip cause rate inaccuracies during the initial stages of delivery (see Fig. 83.5a). Since the syringe volumes are typically much smaller than reservoirs used with other infusion devices, syringe pumps generally deliver drugs in either fluid-restricted environments or for short duration. With high-quality syringes, flow rate uniformity in syringe pumps is generally superior to that accomplished by other infusion pumps. With the drug reservoir enclosed within the device, syringe pumps manage patient transport well, including the operating room environment. Cassette pumps conceptually mimic the piston type action of the syringe pump but provide an automated means of repeatedly emptying and refilling the cassette. The process of refilling the cassette in single piston devices requires an interruption in flow (see Fig. 83.5b). The length of interruption relative to the drug’s half-life determines the impact of the refill period on hemodynamic stability. To eliminate the interruption caused by refill, dual piston devices alternate refill and delivery states, providing nearly continuous output. Others implement cassettes with very small volumes which can refill in less than a second (see Fig. 83.2). Tight control of the internal cross-sectional area of the pumping chamber provides exceptional flow rate accuracy. Manufacturers have recently developed remarkably small cassette pumps that can still generate the full spectrum of infusion rate (0.1–999.0 mL/hr). These systems combine pumping chamber, inlet and outlet valving, pressure sensing, and air detec- FIGURE 83.5 Continuous flow pattern for a representative, (a) tion into a single complex component. syringe, (b) cassette, and (c) linear peristaltic pump at 10 mL/hr. © 2000 by CRC Press LLC

FIGURE 83.6 Impact of 5 variables on flow rate accuracy in 4 different infusion pumps. Variables tested included Solution: Distilled water and 25% dextrose in water; Back Pressure: –100 mmHg and 300 mmHg; Pumping Segment Filling Pressure: –30 inches of water and +30 inches of water; Temperature: 10°C and 40°C; and Infusion Rate: 5 mL/hr and 500 mL/hr. Note: First and second peristaltic mechanism qualified for Low Risk patients, while the third peristaltic device qualified for high-risk patients.

Peristaltic pumps operate on a short segment of the IV tubing. Peristaltic pumps can be separated into two subtypes. Rotary peristaltic mechanisms operate by compressing the pumping segment against the rotor housing with rollers mounted on the housing. With rotation, the rollers push fluid from the container through the tubing toward the patient. At least one of the rollers completely occludes the tubing against the housing at all times precluding free flow from the reservoir to the patient. During a portion of the revolution, two rollers trap fluid in the intervening pumping segment. The captured volume between the rollers determines volumetric accuracy. Linear peristaltic pumps hold the pumping segment in a channel pressed against a rigid backing plate. An array of cam-driven actuators sequentially occlude the segment starting with the section nearest the reservoir forcing fluid toward the patient with a sinusoidal wave action. In a typical design using uniform motor step intervals, a characteristic flow wave resembling a positively biased sine wave is produced (see Fig. 83.5c). Infusion pumps provide significant advantages over both manual flow regulators and controllers in several categories. Infusion pumps can provide accurate delivery over a wide range of infusion rates (0.1–999.0 mL/hr). Neither elevated system resistance nor distal line pressure limit the maximum infusion rate. Infusion pumps can support a wider range of applications including arterial infusions, spinal and epidural infusions, and infusions into pulmonary artery or central venous catheters. Flow rate accuracy of infusion pumps is highly dependent on the segment employed as the pumping chamber (see Fig. 83.2). Incorporating special syringes or pumping segments can significantly improve flow rate accuracy (see Fig. 83.6). Both manufacturing tolerances and segment material composition significantly dictate flow rate accuracy. Time- and temperature-related properties of the pumping segment further impact longterm drift in flow rate.

83.4 Managing Occlusions of the Delivery System One of the most common problems in managing an IV delivery system is the rapid detection of occlusion in the delivery system. With a complete occlusion, the resistance to flow approaches infinity. In this condition, gravimetric-based devices cease to generate flow. Mechanical flow regulators have no mechanism for adverse event detection and thus must rely on the clinician to identify an occlusion as part of routine patient care. Electronic controllers sense the absence of flow and alarm in response to their inability to sustain the desired flow rate. The problem of rapidly detecting an occlusion in an infusion pump is more complex. Upstream occlusions that occur between the fluid reservoir and the pumping mechanism impact the system quite differently than downstream occlusions which occur between the pump and the patient. When an occlusion occurs downstream from an infusion pump, the pump continues to propel fluid into the section of tubing between the pump and the occlusion. The time rate of pressure rise in that section increases


in direct proportion to flow rate and inversely with tubing compliance (compliance, C, is the volume increase in a closed tube per mmHg pressure applied). The most common approach to detecting downstream occlusion requires a pressure transducer immediately below the pumping mechanism. These devices generate an alarm when either the mean pressure or rate of change in pressure exceeds a threshold. For pressure-limited designs, the time to downstream alarm (TTA) may be estimated as

TTA =

Palarm ⋅ Cdelivery− set flow rate

Using a representative tubing compliance of 1 µL/mmHg, flow rate of 1 mL/hr, and a fixed alarm threshold set of 500 mmHg, the time to alarm becomes

TTA =

500mmHg ⋅ 1000 ml mmHg 1mL hr

= 30 min

where TTA is the time from occlusion to alarm detection. Pressure-based detection algorithms depend on accuracy and stability of the sensing system. Lowering the threshold on absolute or relative pressure for occlusion alarm reduces the TTA, but at the cost of increasing the likelihood of false alarms. Patient movement, patient-to-pump height variations, and other clinical circumstances can cause wide perturbations in line pressure. To optimize the balance between fast TTA and minimal false alarms, some infusion pumps allow the alarm threshold to be set by the clinician or be automatically shifted upward in response to alarms; other pumps attempt to optimize performance by varying pressure alarm thresholds with flow rate. A second approach to detection of downstream occlusions uses motor torque as an indirect measure of the load seen by the pumping mechanism. Although this approach eliminates the need for a pressure sensor, it introduces additional sources for error including friction in the gear mechanism or pumping mechanism that requires additional safety margins to protect against false alarms. In syringe pumps, where the coefficient of static friction of the syringe bunge (rubber end of the syringe plunger) against the syringe wall can be substantial, occlusion detection can exceed 1 hr at low flow rates. Direct, continuous measurement of downstream flow resistance may provide a monitoring modality which overcomes the disadvantages of pressure-based alarm systems, especially at low infusion rates. Such a monitoring system would have the added advantage of performance unaffected by flow rate, hydrostatic pressure variations, and motion artifacts. Upstream occlusions can cause large negative pressures as the pumping mechanism generates a vacuum on the upstream tubing segment. The tube may collapse and the vacuum may pull air through the tubing walls or form cavitation bubbles. A pressure sensor situated above the mechanism or a pressure sensor below the mechanism synchronized with filling of the pumping chamber can detect the vacuum associated with an upstream occlusion. Optical or ultrasound transducers, situated below the mechanism, can detect air bubbles in the catheter, and air-eliminating filters can remove air, preventing large air emboli from being introduced into the patient. Some of the most serious complications of IV therapy occur at the venipuncture site; these include extravasation, postinfusion phlebitis (and thrombophlebitis), IV-related infections, ecchymosis, and hematomas. Other problems that do not occur as frequently include speed shock and allergic reactions. Extravasation (or infiltration) is the inadvertent perfusion of infusate into the interstitial tissue. Reported percentage of patients to whom extravasation has occurred ranges from 10% to over 25%. Tissue damage does not occur frequently, but the consequences can be severe, including skin necrosis requiring significant plastic and reconstructive surgery and amputation of limbs. The frequency of extravasation injury correlates with age, state of consciousness, and venous circulation of the patient as well as the type, location, and placement of the intravenous cannula. Drugs that have high osmolality,


vessicant properties, or the ability to induce ischemia correlate with frequency of extravasation injury. Neonatal and pediatric patients who possess limited communication skills, constantly move, and have small veins that are difficult to cannulate require superior vigilance to protect against extravasation. Since interstitial tissue provides a greater resistance to fluid flow than the venous pathway, infusion devices with accurate and precise pressure monitoring systems have been used to detect small pressure increases due to extravasation. To successfully implement this technique requires diligence by the clinician, since patient movement, flow rate, catheter resistance, and venous pressure variations can obscure the small pressure variations resulting from the extravasation. Others have investigated the ability of a pumping mechanism to withdraw blood as indicative of problems in a patent line. The catheter tip, however, may be partially in and out of the vein such that infiltration occurs yet blood can be withdrawn from the patient. A vein might also collapse under negative pressure in a patent line without successful blood withdrawal. Techniques currently being investigated which monitor infusion impedance (resistance and compliance) show promise for assisting in the detection of extravasation. When a catheter tip wedges into the internal lining of the vein wall, it is considered positional. With the fluid path restricted by the vein wall, increases in line resistance may indicate a positional catheter. With patient movement, for example wrist flexation, the catheter may move in and out of the positional state. Since a positional catheter is thought to be more prone toward extravasation than other catheters, early detection of a positional catheter and appropriate adjustment of catheter position may be helpful in reducing the frequency of extravasation. Postinfusion phlebitis is acute inflammation of a vein used for IV infusion. The chief characteristic is a reddened area or red streak that follows the course of the vein with tenderness, warmth, and edema at the venipuncture site. The vein, which normally is very compliant, also hardens. Phlebitis positively correlates with infusion rate and with the infusion of vesicants. Fluid overload and speed shock result from the accidental administration of a large fluid volume over a short interval. Speed shock associates more frequently with the delivery of potent medications, rather than fluids. These problems most commonly occur with manually regulated IV systems, which do not provide the safety features of instrumented lines. Many IV sets designed for instrumented operation will free-flow when the set is removed from the instrument without manual clamping. To protect against this possibility, some sets are automatically placed in the occluded state on disengagement. Although an apparent advantage, reliance on such automatic devices may create a false sense of security and lead to manual errors with sets not incorporating these features.

83.5 Summary Intravenous infusion has become the mode of choice for delivery of a large class of fluids and drugs both in hospital and alternative care settings. Modern infusion devices provide the clinician with a wide array of choices for performing intravenous therapy. Selection of the appropriate device for a specified application requires understanding of drug pharmacology and pharmacokinetics, fluid mechanics, and device design and performance characteristics. Continuing improvements in performance, safety, and cost of these systems will allow even broader utilization of intravenous delivery in a variety of settings.

References Association for the Advancement of Medical Instrumentation. 1992. Standard for Infusion Devices. Arlington, Virg. Bohony J. 1993. Nine common intravenous complications and what to do about them. Am J Nursing 10:45. British Department of Health. 1990. Evaluation of Infusion Pumps and Controllers. HEI Report #198. Glass PSA, Jacobs JR, Reves JG. 1991. Technology for continuous infusions in anesthesia. Continuous Infusions in Anesthesia. International Anesthesiology Clinics 29(4):39. MacCara M. 1983. Extravasation: A hazard of intravenous therapy. Drug Intelligence Clin Pharm 17:713. © 2000 by CRC Press LLC

Further Information Peter Glass provides a strong rationale for intravenous therapy including pharmacokinetic and pharmacodynamic bases for continuous delivery. Clinical complications around intravenous therapy are well summarized by MacCara [1983] and Bohony [1993]. The AAMI Standard for Infusion Devices provides a comprehensive means of evaluating infusion device technology, and the British Department of Health OHEI Report #198 provides a competitive analysis of pumps and controllers.


Paulsen, A. W. “Essentials of Anesthesia Delivery.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

84 Essentials of Anesthesia Delivery 84.1

Gases Used During Anesthesia and their Sources Oxygen • Air • Nitrous Oxide • Carbon Dioxide • Helium

84.2 84.3 84.4 84.5 84.6

A. William Paulsen Emory University

Gas Blending and Vaporization System Breathing Circuits Gas Scavenging Systems Monitoring the Function of the Anesthesia Delivery System Monitoring the Patient Control of Patient Temperature • Monitoring the Depth of Anesthesia • Anesthesia Computer-Aided Record Keeping • Alarms • Ergonomics • Simulation in Anesthesia • Reliability

The intent of this chapter is to provide an introduction to the practice of anesthesiology and to the technology currently employed. Limitations on the length of this work and the enormous size of the topic require that this chapter rely on other elements within this Handbook and other texts cited as general references for many of the details that inquisitive minds desire and deserve. The practice of anesthesia includes more than just providing relief from pain. In fact, pain relief can be considered a secondary facet of the specialty. In actuality, the modern concept of the safe and efficacious delivery of anesthesia requires consideration of three fundamental tenets, which are ordered here by relative importance: 1. maintenance of vital organ function 2. relief of pain 3. maintenance of the “internal milieu” The first, maintenance of vital organ function, is concerned with preventing damage to cells and organ systems that could result from inadequate supply of oxygen and other nutrients. The delivery of blood and cellular substrates is often referred to as perfusion of the cells or tissues. During the delivery of an anesthetic, the patient’s “vital signs” are monitored in an attempt to prevent inadequate tissue perfusion. However, the surgery itself, the patient’s existing pathophysiology, drugs given for the relief of pain, or even the management of blood pressure may compromise tissue perfusion. Why is adequate perfusion of tissues a higher priority than providing relief of pain for which anesthesia is named? A rather obvious extreme example is that without cerebral perfusion, or perfusion of the spinal cord, delivery of an anesthetic is not necessary. Damage to other organ systems may result in a range of complications from delaying the patient’s recovery to diminishing their quality of life to premature death.


In other words, the primary purpose of anesthesia care is to maintain adequate delivery of required substrates to each organ and cell, which will hopefully preserve cellular function. The second principle of anesthesia is to relieve the pain caused by surgery. Chronic pain and suffering caused by many disease states is now managed by a relatively new sub-specialty within anesthesia, called Pain Management. The third principle of anesthesia is the maintenance of the internal environment of the body, for example, the regulation of electrolytes (sodium, potassium, chloride, magnesium, calcium, etc.), acidbase balance, and a host of supporting functions on which cellular function and organ system communications rest. The person delivering anesthesia may be an Anesthesiologist (physician specializing in anesthesiology), an Anesthesiology Physician Assistant (a person trained in a medical school at the masters level to administer anesthesia as a member of the care team lead by an Anesthesiologist), or a nurse anesthetist (a nurse with Intensive Care Unit experience that has additional training in anesthesia provided by advanced practice nursing programs). There are three major categories of anesthesia provided to patients: (1) general anesthesia; (2) conduction anesthesia; and (3) monitored anesthesia care. General anesthesia typically includes the intravenous injection of anesthetic drugs that render the patient unconscious and paralyze their skeletal muscles. Immediately following drug administration a plastic tube is inserted into the trachea and the patient is connected to an electropneumatic system to maintain ventilation of the lungs. A liquid anesthetic agent is vaporized and administered by inhalation, sometimes along with nitrous oxide, to maintain anesthesia for the surgical procedure. Often, other intravenous agents are used in conjunction with the inhalation agents to provide what is called a balanced anesthetic. Conduction anesthesia refers to blocking the conduction of pain and possibly motor nerve impulses traveling along specific nerves or the spinal cord. Common forms of conduction anesthesia include spinal and epidural anesthesia, as well as specific nerve blocks, for example, axillary nerve blocks. In order to achieve a successful conduction anesthetic, local anesthetic agents such as lidocaine, are injected into the proximity of specific nerves to block the conduction of electrical impulses. In addition, sedation may be provided intravenously to keep the patient comfortable while he/she is lying still for the surgery. Monitored anesthesia care refers to monitoring the patient’s vital signs while administering sedatives and analgesics to keep the patient comfortable, and treating complications related to the surgical procedure. Typically, the surgeon administers topical or local anesthetics to alleviate the pain. In order to provide the range of support required, from the paralyzed mechanically ventilated patient to the patient receiving monitored anesthesia care, a versatile anesthesia delivery system must be available to the anesthesia care team. Today’s anesthesia delivery system is composed of six major elements: 1. The primary and secondary sources of gases (O2, air, N2O, vacuum, gas scavenging, and possibly CO2 and helium). 2. The gas blending and vaporization system. 3. The breathing circuit (including methods for manual and mechanical ventilation). 4. The excess gas scavenging system that minimizes potential pollution of the operating room by anesthetic gases. 5. Instruments and equipment to monitor the function of the anesthesia delivery system. 6. Patient monitoring instrumentation and equipment. The traditional anesthesia machine incorporated elements 1, 2, 3, and more recently 4. The evolution to the anesthesia delivery system adds elements 5 and 6. In the text that follows, references to the “anesthesia machine” refer to the basic gas delivery system and breathing circuit as contrasted with the “anesthesia delivery system” which includes the basic “anesthesia machine” and all monitoring instrumentation.

84.1 Gases Used During Anesthesia and their Sources Most inhaled anesthetic agents are liquids that are vaporized in a device within the anesthesia delivery system. The vaporized agents are then blended with other breathing gases before flowing into the


breathing circuit and being administered to the patient. The most commonly administered form of anesthesia is called a balanced general anesthetic, and is a combination of inhalation agent plus intravenous analgesic drugs. Intravenous drugs often require electromechanical devices to administer an appropriately controlled flow of drug to the patient. Gases needed for the delivery of anesthesia are generally limited to oxygen (O2), air, nitrous oxide (N2O), and possibly helium (He) and carbon dioxide (CO2). Vacuum and gas scavenging lines are also required. There needs to be secondary sources of these gases in the event of primary failure or questionable contamination. Typically, primary sources are those supplied from a hospital distribution system at 345 kPa (50 psig) through gas columns or wall outlets. The secondary sources of gas are cylinders hung on yokes on the anesthesia delivery system.

Oxygen Oxygen provides an essential metabolic substrate for all human cells, but it is not without dangerous side effects. Prolonged exposure to high concentrations of oxygen may result in toxic effects within the lungs that decrease diffusion of gas into and out of the blood, and the return to breathing air following prolonged exposure to elevated O2 may result in a debilitating explosive blood vessel growth in infants. Oxygen is usually supplied to the hospital in liquid form (boiling point of –183°C), stored in cryogenic tanks, and supplied to the hospital piping system as a gas. The efficiency of liquid storage is obvious since 1 liter of liquid becomes 860 liters of gas at standard temperature and pressure. The secondary source of oxygen within an anesthesia delivery system is usually one or more E cylinders filled with gaseous oxygen at a pressure of 15.2 MPa (2200 psig).

Air (78% N2, 21% O2, 0.9% Ar, 0.1% Other Gases) The primary use of air during anesthesia is as a diluent to decrease the inspired oxygen concentration. The typical primary source of medical air (there is an important distinction between “air” and “medical air” related to the quality and the requirements for periodic testing) is a special compressor that avoids hydrocarbon based lubricants for purposes of medical air purity. Dryers are employed to rid the compressed air of water prior to distribution throughout the hospital. Medical facilities with limited need for medical air may use banks of H cylinders of dry medical air. A secondary source of air may be available on the anesthesia machine as an E cylinder containing dry gas at 15.2 MPa.

Nitrous Oxide Nitrous oxide is a colorless, odorless, and non-irritating gas that does not support human life. Breathing more than 85% N2O may be fatal. N2O is not an anesthetic (except under hyperbaric conditions), rather it is an analgesic and an amnestic. There are many reasons for administering N2O during the course of an anesthetic including: enhancing the speed of induction and emergence from anesthesia; decreasing the concentration requirements of potent inhalation anesthetics (i.e., halothane, isoflurane, etc.); and as an essential adjunct to narcotic analgesics. N2O is supplied to anesthetizing locations from banks of H cylinders that are filled with 90% liquid at a pressure of 5.1 MPa (745 psig). Secondary supplies are available on the anesthesia machine in the form of E cylinders, again containing 90% liquid. Continual exposure to low levels of N2O in the workplace has been implicated in a number of medical problems including spontaneous abortion, infertility, birth defects, cancer, liver and kidney disease, and others. Although there is no conclusive evidence to support most of these implications, there is a recognized need to scavenge all waste anesthetic gases and periodically sample N2O levels in the workplace to maintain the lowest possible levels consistent with reasonable risk to the operating room personnel and cost to the institution [Dorsch and Dorsch, 1998]. Another gas with analgesic properties similar to N2O is xenon, but its use is experimental, and its cost is prohibitive at this time.


TABLE 84.1

Physical Properties of Gases Used During Anesthesia

GAS Oxygen Nitrogen Air Nitrous oxide Carbon dioxide Helium

TABLE 84.2 Agent Generic Name Halothane Enflurane Isoflurane Desflurane Sevoflurane

Molecular Wt.

Density (g/L)

Viscosity (cp)

Specific Heat (KJ/Kg°C)

31.999 28.013 28.975 44.013 44.01 4.003

1.326 1.161 1.200 1.836 1.835 0.1657

0.0203 0.0175 0.0181 0.0144 0.0148 0.0194

0.917 1.040 1.010 0.839 0.850 5.190

Physical Properties of Currently Available Volatile Anesthetic Agents Boiling Point (°C at 760 mmHg)

Vapor Pressure (mmHg at 20°C)

Liquid Density (g/ml)

MAC* (%)

50.2 56.5 48.5 23.5 58.5

243 175 238 664 160

1.86 1.517 1.496 1.45 1.51

0.75 1.68 1.15 6.0 2.0

* Minimum Alveolar Concentration is the percent of the agent required to provide surgical anesthesia to 50% of the population in terms of a cummulative dose response curve. The lower the MAC, the more potent the agent.

Carbon Dioxide Carbon dioxide is colorless and odorless, but very irritating to breathe in higher concentrations. CO2 is a byproduct of human cellular metabolism and is not a life-sustaining gas. CO2 influences many physiologic processes either directly or through the action of hydrogen ions by the reaction CO2 + H2O ↔ H2CO3 ↔ H+ + HCO3–. Although not very common in the U.S. today, in the past CO2 was administered during anesthesia to stimulate respiration that was depressed by anesthetic agents and to cause increased blood flow in otherwise compromised vasculature during some surgical procedures. Like N2O, CO2 is supplied as a liquid in H cylinders for distribution in pipeline systems or as a liquid in E cylinders that are located on the anesthesia machine.

Helium Helium is a colorless, odorless, and non-irritating gas that will not support life. The primary use of helium in anesthesia is to enhance gas flow through small orifices as in asthma, airway trauma, or tracheal stenosis. The viscosity of helium is not different from other anesthetic gases (refer to Table 84.1) and is therefore of no benefit when airway flow is laminar. However, in the event that ventilation must be performed through abnormally narrow orifices or tubes which create turbulent flow conditions, helium is the preferred carrier gas. Resistance to turbulent flow is proportional to the density rather than viscosity of the gas and helium is an order of magnitude less dense than other gases. A secondary advantage of helium is that it has a large specific heat relative to other anesthetic gases and therefore can carry the heat from laser surgery out of the airway more effectively than air, oxygen, or nitrous oxide.

84.2 Gas Blending and Vaporization System The basic anesthesia machine utilizes primary low pressure gas sources of 345 kPa (50 psig) available from wall or ceiling column outlets, and secondary high pressure gas sources located on the machine as pictured schematically in Fig. 84.1. Tracing the path of oxygen in the machine demonstrates that oxygen comes from either the low pressure source, or from the 15.2 Mpa (2200 psig) high pressure yokes via © 2000 by CRC Press LLC

FIGURE 84.1 machine.

Schematic diagram of gas piping within a simple two-gas (oxygen and nitrous oxide) anesthesia

cylinder pressure regulators and then branches to service several other functions. First and foremost, the second stage pressure regulator drops the O2 pressure to approximately 110 kPa (16 psig) before it enters the needle valve and the rotameter type flowmeter. From the flowmeter O2 mixes with gases from other flowmeters and passes through a calibrated agent vaporizer where specific inhalation anesthetic agents are vaporized and added to the breathing gas mixture. Oxygen is also used to supply a reservoir canister that sounds a reed alarm in the event that the oxygen pressure drops below 172 kPa (25 psig). When the oxygen pressure drops to 172 kPa or lower, then the nitrous oxide pressure sensor shutoff valve closes and N2O is prevented from entering its needle valve and flowmeter and is therefore eliminated from the breathing gas mixture. In fact, all machines built in the U.S. have pressure sensor shutoff valves installed in the lines to every flowmeter, except oxygen, to prevent the delivery of a hypoxic gas mixture in the event of an oxygen pressure failure. Oxygen may also be delivered to the common gas outlet or machine outlet via a momentary normally closed flush valve that typically provides a flow of 65 to 80 liters of O2 per minute directly into the breathing circuit. Newer machines are required to have a safety system for limiting the minimum concentration of oxygen that can be delivered to the patient to 25%. The flow paths for nitrous oxide and other gases are much simpler in the sense that after coming from the high pressure regulator or the low pressure hospital source, gas is immediately presented to the pressure sensor shutoff valve from where it travels to its specific needle valve and flowmeter to join the common gas line and enter the breathing circuit. © 2000 by CRC Press LLC

FIGURE 84.2 Schematic diagram of a calibrated in-line vaporizer that uses the flow-over technique for adding anesthetic vapor to the breathing gas mixture.

Currently all anesthesia machines manufactured in the U.S. use only calibrated flow-through vaporizers, meaning that all of the gases from the various flowmeters are mixed in the manifold prior to entering the vaporizer. Any given vaporizer has a calibrated control knob that, once set to the desired concentration for a specific agent, will deliver that concentration to the patient. Some form of interlock system must be provided such that only one vaporizer may be activated at any given time. Figure 84.2 schematically illustrates the operation of a purely mechanical vaporizer with temperature compensation. This simple flow-over design permits a fraction of the total gas flow to pass into the vaporizing chamber where it becomes saturated with vapor before being added back to the total gas flow. Mathematically this is approximated by:

FA =

(

QVC ∗ PA

)

PB ∗ QVC + QG − PA ∗ QG

where FA is the fractional concentration of agent at the outlet of the vaporizer, QG is the total flow of gas entering the vaporizer, QVC is the amount of QG that is diverted into the vaporization chamber, PA is the vapor pressure of the agent, and PB is the barometric pressure. From Fig. 84.2, the temperature compensator would decrease QVC as temperature increased because vapor pressure is proportional to temperature. The concentration accuracy over a range of clinically


expected gas flows and temperatures is approximately ± 15%. Since vaporization is an endothermic process, anesthetic vaporizers must have sufficient thermal mass and conductivity to permit the vaporization process to proceed independent of the rate at which the agent is being used.

84.3 Breathing Circuits The concept behind an effective breathing circuit is to provide an adequate volume of a controlled concentration of gas to the patient during inspiration, and to carry the exhaled gases away from the patient during exhalation. There are several forms of breathing circuits which can be classified into 2 basic types; (1) open circuit, meaning no rebreathing of any gases and no CO2 absorber present; and (2) closed circuit, indicating presence of CO2 absorber and some rebreathing of other gases. Figure 84.3 illustrates the Lack modification of a Mapleson open circuit breathing system. There are no valves and no CO2 absorber. There is a great potential for the patient to rebreath their own exhaled gases unless the fresh gas inflow is 2 to 3 times the patient’s minute volume. Figure 84.4 illustrates the most popular form of breathing circuit, the circle system, with oxygen monitor, circle pressure gage, volume monitor (spirometer), and airway pressure sensor. The circle is a closed system, or semi-closed when the fresh gas inflow exceeds the patient’s requirements. Excess gas evolves into the scavenging device, and some of the exhaled

FIGURE 84.3 An example of an open circuit breathing system that does not use unidirectional flow valves or contain a carbon dioxide absorbent.

FIGURE 84.4 A diagram of a closed circuit circle breathing system with unidirectional valves, inspired oxygen sensor, pressure sensor, and CO2 absorber. © 2000 by CRC Press LLC

gas is rebreathed after having the CO2 removed. The inspiratory and expiratory valves in the circle system guarantee that gas flows to the patient from the inspiratory limb and away from the patient through the exhalation limb. In the event of a failure of either or both of these valves, the patient will rebreath exhaled gas that contains CO2, which is a potentially dangerous situation. There are two forms of mechanical ventilation used during anesthesia: (1) volume ventilation, where the volume of gas delivered to the patient remains constant regardless of the pressure that is required; and (2) pressure ventilation, where the ventilator provides whatever volume to the patient that is required to produce some desired pressure in the breathing circuit. Volume ventilation is the most popular since the volume delivered remains theoretically constant despite changes in lung compliance. Pressure ventilation is useful when compliance losses in the breathing circuit are high relative to the volume delivered to the lungs. Humidification is an important adjunct to the breathing circuit because it maintains the integrity of the cilia that line the airways and promote the removal of mucus and particulate matter from the lungs. Humidification of dry breathing gases can be accomplished by simple passive heat and moisture exchangers inserted into the breathing circuit at the level of the endotracheal tube connectors, or by elegant dual servo electronic humidifiers that heat a reservoir filled with water and also heat a wire in the gas delivery tube to prevent rain-out of the water before it reaches the patient. Electronic safety measures must be included in these active devices due to the potential for burning the patient and the fire hazard.

84.4 Gas Scavenging Systems The purpose of scavenging exhaled and excess anesthetic agents is to reduce or eliminate the potential hazard to employees who work in the environment where anesthetics are administered, including operating rooms, obstetrical areas, special procedures areas, physician’s offices, dentist’s offices, and veterinarian’s surgical suites. Typically more gas is administered to the breathing circuit than is required by the patient, resulting in the necessity to remove excess gas from the circuit. The scavenging system must be capable of collecting gas from all components of the breathing circuit, including adjustable pressure level valves, ventilators, and sample withdrawal type gas monitors, without altering characteristics of the circuit such as pressure or gas flow to the patient. There are two broad types of scavenging systems as illustrated in Fig. 84.5: the open interface is a simple design that requires a large physical space for the reservoir volume, and the closed interface with an expandable reservoir bag and which must include relief valves for handling the cases of no scavenged flow and great excess of scavenged flow. Trace gas analysis must be performed to guarantee the efficacy of the scavenging system. The National Institutes of Occupational Safety and Health (NIOSH) recommends that trace levels of nitrous oxide be maintained at or below 25 parts per million (ppm) time weighted average and that halogenated anesthetic agents remain below 2 ppm.

84.5 Monitoring the Function of the Anesthesia Delivery System The anesthesia machine can produce a single or combination of catastrophic events, any one of which could be fatal to the patient: 1. delivery of a hypoxic gas mixture to the patient; 2. the inability to adequately ventilate the lungs by not producing positive pressure in the patient’s lungs, by not delivering an adequate volume of gas to the lungs, or by improper breathing circuit connections that permit the patient’s lungs to receive only rebreathed gases; 3. the delivery of an overdose of an inhalational anesthetic agent. The necessary monitoring equipment to guarantee proper function of the anesthesia delivery system includes at least:


FIGURE 84.5 Examples of open and closed gas scavenger interfaces. The closed interface requires relief valves in the event of scavenging flow failure.

• Inspired Oxygen Concentration monitor with absolute low level alarm of 19%. • Airway Pressure Monitor with alarms for: 1. low pressure indicative of inadequate breathing volume and possible leaks 2. sustained elevated pressures that could compromise cardiovascular function 3. high pressures that could cause pulmonary barotrauma 4. subatmospheric pressure that could cause collapse of the lungs • Exhaled Gas Volume Monitor. • Carbon Dioxide Monitor (capnography). • Inspired and Exhaled Concentration of anesthetic agents by any of the following: 1. mass spectrometer 2. Raman spectrometer 3. infrared or other optical spectrometer A mass spectrometer is a very useful cost-effective device since it alone can provide capnography, inspired and exhaled concentrations of all anesthetic agents, plus all breathing gases simultaneously (O2 , N2 , CO2 , N2O, Ar, He, halothane, enflurane, isoflurane, desflurane, and suprane). The mass spectrometer is unique in that it may be tuned to monitor an assortment of exhaled gases while the patient is asleep, including: (1) ketones for detection of diabetic ketoacidosis; (2) ethanol or other marker in the irrigation solution during transurethral resection of the prostate for early detection of the TURP syndrome, which


results in a severe dilution of blood electrolytes; and (3) pentanes during the evolution of a heart attack, to mention a few. Sound monitoring principles require: (1) earliest possible detection of untoward events (before they result in physiologic derangements); and (2) specificity that results in rapid identification and resolution of the problem. An extremely useful rule to always consider is “never monitor the anesthesia delivery system performance through the patient’s physiologic responses”. That is, never intentionally use a device like a pulse oximeter to detect a breathing circuit disconnection since the warning is very late and there is no specific information provided that leads to rapid resolution of the problem.

84.6 Monitoring the Patient The anesthetist’s responsibilities to the patient include: providing relief from pain and preserving all existing normal cellular function of all organ systems. Currently the latter obligation is fulfilled by monitoring essential physiologic parameters and correcting any substantial derangements that occur before they are translated into permanent cellular damage. The inadequacy of current monitoring methods can be appreciated by realizing that most monitoring modalities only indicate damage after an insult has occurred, at which point the hope is that it is reversible or that further damage can be prevented. Standards for basic intraoperative monitoring of patients undergoing anesthesia, that were developed and adopted by the American Society of Anesthesiologists, became effective in 1990. Standard I concerns the responsibilities of anesthesia personnel, while Standard II requires that the patient’s oxygenation, ventilation, circulation, and temperature be evaluated continually during all anesthetics. The following list indicates the instrumentation typically available during the administration of anesthetics. Electrocardiogram Non-Invasive or Invasive Blood Pressure Pulse Oximetry Temperature Urine Output Nerve Stimulators Cardiac Output Mixed Venous Oxygen Saturation Electroencephalogram (EEG) Transesophageal Echo Cardiography (TEE) Evoked Potentials Coagulation Status Blood gases and electrolytes (Po2, Pco2, pH, BE, Na+, K+, Cl–, Ca++, and glucose) Mass Spectrometry, Raman Spectrometry or Infrared Breathing Gas Analysis

Control of Patient Temperature Anesthesia alters the thresholds for temperature regulation and the patient becomes unable to maintain normal body temperature. As the patient’s temperature falls even a few degrees toward room temperature, several physiologic derangements occur: (1) drug action is prolonged; (2) blood coagulation is impaired; and (3) post-operative infection rate increases. On the positive side, cerebral protection from inadequate perfusion is enhanced by just a few degrees of cooling. Proper monitoring of core body temperature and forced hot air warming of the patient is essential.

Monitoring the Depth of Anesthesia There are two very unpleasant experiences that patients may have while undergoing an inadequate anesthetic: (1) the patient is paralyzed and unable to communicate their state of discomfort, and they are feeling the pain of surgery and are aware of their surroundings; (2) the patient may be paralyzed, unable to communicate, and is aware of their surroundings, but is not feeling any pain. The ability to monitor the depth of anesthesia would provide a safeguard against these unpleasant experiences. However, despite numerous instruments and approaches to the problem it remains elusive. Brain stem auditory evoked responses have come the closest to depth of anesthesia monitoring, but it is difficult to perform, is expensive, and is not possible to perform during many types of surgery. A promising new technology,


called bi-spectral index (BIS monitoring) is purported to measure the level of patient awareness through multivariate analysis of a single channel of the EEG.

Anesthesia Computer-Aided Record Keeping Conceptually, every anesthetist desires an automated anesthesia record keeping system. Anesthesia care can be improved through the feedback provided by correct record keeping, but today’s systems have an enormous overhead associated with their use when compared to standard paper record keeping. No doubt that automated anesthesia record keeping reduces the drudgery of routine recording of vital signs, but to enter drugs and drips and their dosages, fluids administered, urine output, blood loss, and other data requires much more time and machine interaction than the current paper system. Despite attempts to use every input/output device ever produced by the computer industry from keyboards to bar codes to voice and handwriting recognition, no solution has been found that meets wide acceptance. Tenants of a successful system must include: 1. The concept of a user transparent system, which is ideally defined as requiring no communication between the computer and the clinician (far beyond the concept of user friendly), and therefore that is intuitively obvious to use even to the most casual users. 2. Recognition of the fact that educational institutions have very different requirements from private practice institutions. 3. Real time hard copy of the record produced at the site of anesthetic administration that permits real time editing and notation. 4. Ability to interface with a great variety of patient and anesthesia delivery system monitors from various suppliers. 5. Ability to interface with a large number of hospital information systems. 6. Inexpensive to purchase and maintain.

Alarms Vigilance is the key to effective risk management, but maintaining a vigilant state is not easy. The practice of anesthesia has been described as moments of shear terror connected by times of intense boredom. Alarms can play a significant role in redirecting one’s attention during the boredom to the most important event regarding patient safety, but only if false alarms can be eliminated, alarms can be prioritized, and all alarms concerning anesthetic management can be displayed in a single clearly visible location.

Ergonomics The study of ergonomics attempts to improve performance by optimizing the relationship between people and their work environment. Ergonomics has been defined as a discipline which investigates and applies information about human requirements, characteristics, abilities, and limitations to the design, development, and testing of equipment, systems, and jobs [Loeb, 1993]. This field of study is only in its infancy and examples of poor ergonomic design abound in the anesthesia workplace.

Simulation in Anesthesia Complete patient simulators are hands-on realistic simulators that interface with physiologic monitoring equipment to simulate patient responses to equipment malfunctions, operator errors, and drug therapies. There are also crisis management simulators. Complex patient simulators, which are analogous to flight simulators, are currently being marketed for training anesthesia personnel. The intended use for these complex simulators is currently being debated in the sense that training people to respond in a preprogrammed way to a given event may not be adequate training.


Reliability The design of an anesthesia delivery system is unlike the design of most other medical devices because it is a life support system. As such, its core elements deserve all of the considerations of the latest failsafe technologies. Too often in today’s quest to apply microprocessor technology to everything, tradeoffs are made among reliability, cost, and engineering elegance. The most widely accepted anesthesia machine designs continue to be based upon simple ultra-reliable mechanical systems with an absolute minimum of catastrophic failure modes. The replacement of needle valves and rotameters, for example, with microprocessor controlled electromechanical valves can only introduce new catastrophic failure modes. However, the inclusion of microprocessors can enhance the safety of anesthesia delivery if they are implemented without adding catastrophic failure modes.

Further Information Blitt, C.D. and Hines R.L., Eds. 1995. Monitoring in Anesthesia and Critical Care Medicine, 3rd ed. Churchill Livingstone, New York. Dorsch, J.A. and Dorsch S.E. 1998. Understanding Anesthesia Equipment, 4th ed. Williams and Wilkins, Baltimore, MD. Ehrenwerth, J. and Eisenkraft, J.B. 1993. Anesthesia Equipment: Principles and Applications. Mosby, St. Louis, MO. Gravenstein N. and Kirby R.R., Eds. 1996. Complications in Anesthesiology, 2nd ed. Lippincott—Raven, Philadelphia, PA. Loeb, R. 1993. Ergonomics of the anesthesia workplace. STA Interface 4(3):18. Miller, R.D. Ed. 1999. Anesthesia, 5th ed. Churchill Livingstone, New York. Miller, R.D. Ed. 1998. Atlas of Anesthesia. Churchill Livingstone, New York. Saidman, L.J. and Smith, N.T., Eds. 1993. Monitoring in Anesthesia, 3rd ed. Butterworth-Heinemann, Stoneham, MA.


Judy, M. M. “Biomedical Lasers.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

85 Biomedical Lasers 85.1

Interaction and Effects of UV-IR Laser Radiation on Biologic Tissues

85.2

Penetration and Effects of UV-IR Laser Radiation into Biologic Tissues Effects of Mid-IR Laser Radiation Effects of Near-IR Laser Radiation Effects of Visible-Range Laser Radiation Effects of UV Laser Radiation Effects of Continuous and Pulsed IR-Visible Laser Radiation and Associated Temperature Rise General Description and Operation of Lasers Biomedical Laser Beam Delivery Systems

Scattering in Biologic Tissue • Absorption in Biologic Tissue

85.3 85.4 85.5 85.6 85.7 85.8 85.9

Millard M. Judy Baylor Research Institute

Optical Fiber Transmission Characteristics • Mirrored Articulated Arm Characteristics • Optics for Beam Shaping on Tissues • Features of Routinely Used Biomedical Lasers • Other Biomedical Lasers

Approximately 20 years ago the CO2 laser was introduced into surgical practice as a tool to photothermally ablate, and thus to incise and to debulk, soft tissues. Subsequently, three important factors have led to the expanding biomedical use of laser technology, particularly in surgery. These factors are: (1) the increasing understanding of the wave-length selective interaction and associated effects of ultravioletinfrared (UV-IR) radiation with biologic tissues, including those of acute damage and long-term healing, (2) the rapidly increasing availability of lasers emitting (essentially monochromatically) at those wavelengths that are strongly absorbed by molecular species within tissues, and (3) the availability of both optical fiber and lens technologies as well as of endoscopic technologies for delivery of the laser radiation to the often remote internal treatment site. Fusion of these factors has led to the development of currently available biomedical laser systems. This chapter briefly reviews the current status of each of these three factors. In doing so, each of the following topics will be briefly discussed: 1. The physics of the interaction and the associated effects (including clinical efforts) of UV-IR radiation on biologic tissues. 2. The fundamental principles that underlie the operations and construction of all lasers. 3. The physical properties of the optical delivery systems used with the different biomedical lasers for delivery of the laser beam to the treatment site. 4. The essential physical features of those biomedical lasers currently in routine use ranging over a number of clinical specialties, and brief descriptions of their use. 5. The biomedical uses of other lasers used surgically in limited scale or which are currently being researched for applications in surgical and diagnostic procedures and the photosensitized inactivation of cancer tumors.


In this review, effort is made in the text and in the last section to provide a number of key references and sources of information for each topic that will enable the reader’s more in-depth pursuit.

85.1 Interaction and Effects of UV-IR Laser Radiation on Biologic Tissues Electromagnetic radiation in the UV-IR spectral range propagates within biologic tissues until it is either scattered or absorbed.

Scattering in Biologic Tissue Scattering in matter occurs only at the boundaries between regions having different optical refractive indices and is a process in which the energy of the radiation is conserved [Van de Hulst, 1957]. Since biologic tissue is structurally inhomogeneous at the microscopic scale, e.g., both subcellular and cellular dimensions, and at the macroscopic scale, e.g., cellular assembly (tissue) dimensions, and predominantly contains water, proteins, and lipids, all different chemical species, it is generally regarded as a scatterer of UV-IR radiation. The general result of scattering is deviation of the direction of propagation of radiation. The deviation is strongest when wavelength and scatterer are comparable in dimension (Mie scattering) and when wavelength greatly exceeds particle size (Rayleigh scattering) [Van de Hulst, 1957]. This dimensional relationship results in the deeper penetration into biologic tissues of those longer wavelengths which are not absorbed appreciably by pigments in the tissues. This results in the relative transparency of nonpigmented tissues over the visible and near-IR wavelength ranges.

Absorption in Biologic Tissue Absorption of UV-IR radiation in matter arises from the wavelength-dependent resonant absorption of radiation by molecular electrons of optically absorbing molecular species [Grossweiner, 1989]. Because of the chemical inhomogeneity of biologic tissues, the degree of absorption of incident radiation strongly depends upon its wavelength. The most prevalent or concentrated UV-IR absorbing molecular species in biologic tissues are listed in Table 85.1 along with associated high-absorbance wavelengths. These species include the peptide bonds; the phenylalanine, tyrosine, and tryptophan residues of proteins, all of which absorb in the UV range; oxy- and deoxyhemoglobin of blood which absorb in the visible to near-IR range; melanin, which absorbs throughout the UV to near-IR range, which decreasing absorption occurring with increasing wavelength; and water, which absorbs maximally in the mid-IR range [Hale & Querry, 1973; Miller & Veitch, 1993; White et al., 1968]. Biomedical lasers and their emitted radiation wavelength values also are tabulated also in Table 85.1. The correlation between the wavelengths of clinically useful lasers and wavelength regions of absorption by constituents of biological tissues is evident. Additionally, exogenous light-absorbing chemical species may be intentionally present in tissues. These include: 1. Photosensitizers, such as porphyrins, which upon excitation with UV-visible light initiate photochemical reactions which are cytotoxic to the cells of the tissue, e.g., a cancer which concentrates the photosensitizer relative to surrounding tissues [Spikes, 1989]. 2. Dyes such as indocyanine green which, when dispersed in a concentrate fibrin protein gel can be used to localize 810 nm GaAlAs diode laser radiation and the associated heating to achieve localized thermal denaturation and bonding of collagen to effect joining or welding of tissue [Bass et al., 1992; Oz et al., 1989]. 3. Tattoo pigments including graphite (black) and black, blue, green, and red organic dyes [Fitzpatrick, 1994; McGillis et al., 1994].


TABLE 85.1

UV-IR-Radiation-Absorbing Constituent of Biological Tissues and Biomedical Laser Wavelengths Optical Absorption

Constituent

Tissue Type

Proteins Peptide bond Amino acid Residues Tryptophan Tyrosine Phenylalanine Pigments Oxyhemoglobin

All

Deoxyhemoglobin

Melanin Water

Wavelength*, nm 2.94 (r)

Relative† Strength +++++++

Laser Type

Wavelength, nm

ArF

193

488–514.5 532

+++ ++

Ar ion Frequency Doubled Nd:YAG Diode Nd:YAG Dye Nd:YAG

++++ +++ +++++++ ++++

Ruby Ho:YAG Er:YAG CO2

+ + + +++ ++ ++ +

810 1064 400–700 1064 693 2100 2940 10,640

* (p): Peak absorption wavelength; (r): wavelength range. † The number of + signs qualitatively ranks the magnitude of the optical absorbtion.

85.2 Penetration and Effects of UV-IR Laser Radiation into Biologic Tissue Both scattering and absorption processes affect the variations of the intensity of radiation with propagation into tissues. In the absence of scattering, absorption results in an exponential decrease of radiation intensity described simply by Beers law [Grossweiner, 1989]. With appreciable scattering present, the decrease in incident intensity from the surface is no longer monotonic. A maximum in local internal intensity is found to be present due to efficient back-scattering, which adds to the intensity of the incoming beam as shown, for example, by Miller and Veitch [1993] for visible light penetrating into the skin and by Rastegar and coworkers [1992] for 1.064 µm Nd:YAG laser radiation penetrating into the prostate gland. Thus, the relative contributions of absorption and scattering of incident laser radiation will stipulate the depth in a tissue at which the resulting tissue effects will be present. Since the absorbed energy can be released in a number of different ways including thermal vibrations, fluorescence, and resonant electronic energy transfer according to the identity of the absorber, the effects on tissue are in general different. Energy release from both hemoglobin and melanin pigments and from water is by molecular vibrations resulting in a local temperature rise. Sufficient continued energy absorption and release can result in local temperature increases which, as energy input increases, result in protein denaturation (41–65°C), water evaporation and boiling (up to 300°C under confining pressure of tissue), thermolysis of proteins, generation of gaseous decomposition products and of carbonaceous residue or char (≥ 300°C). The generation of residual char is minimized by sufficiently rapid energy input to support rapid gasification reactions. The clinical effect of this chain of thermal events is tissue ablation. Much smaller values of energy input result in coagulation of tissues due to protein denaturation. Energy release from excited exogenous photosensitizing dyes is via formation of free-radical species or energy exchange with itinerant dissolved molecular oxygen [Spikes, 1989]. Subsequent chemical


reactions following free-radical formation or formation of an activated or more reactive form of molecular oxygen following energy exchange can be toxic to cells with takeup of the photosensitizer. Energy release following absorption of visible (VIS) radiation by fluorescent molecular species, either endogenous to tissue or exogenous, is predominantly by emission of longer wavelength radiation [Lakowicz, 1983]. Endogenous fluorescent species include tryptophan, tyrosine, phenylalanine, flavins, and metal-free porphyrins. Comparison of measured values of the intensity of fluorescence emission from hyperplastic (transformed precancerous) cervical cells to cancerous cervical cells with normal cervical epithelial cells shows a strong potential for diagnostic use in the automated diagnosis and staging of cervical cancer [Mahadevan et al., 1993].

85.3 Effects of Mid-IR Laser Radiation Because of the very large absorption by water of radiation with wavelength in the IR range ≥ 2.0 µm, the radiation of Ho:YAG, Er: YAG, and CO2 lasers is absorbed within a very short distance of the tissue surface, and scattering is essentially unimportant. Using published values of the water absorption coefficient [Hale & Querry, 1973] and assuming an 80% water content and that the decrease in intensity is exponential with distance, the depth in the “average” soft tissue at which the intensity has decreased to 10% of the incident value (the optical penetration depth) is estimated to be 619, 13, and 170 micrometers, respectively, for Ho:YAG, Er:YAG, and CO2 laser radiation. Thus, the absorption of radiation from these laser sources and thermalization of this energy results essentially in the formation of a surface heat source. With sufficient energy input, tissue ablation through water boiling and tissue thermolysis occur at the surface. Penetration of heat to underlying tissues is by diffusion alone; thus, the depth of coagulation of tissue below the surface region of ablation is limited by competition between thermal diffusion and the rate of descent of the heated surface impacted by laser radiation during ablation of tissue. Because of this competition, coagulation depths obtained in soft biologic tissues with use of mid-IR laser radiation are typically ≤ 205–500 µm, and the ability to achieve sealing of blood vessels leading to hemostatic (“bloodless”) surgery is limited [Judy et al., 1992; Schroder et al., 1987].

85.4 Effects of Near-IR Laser Radiation The 810-nm and 1064-µm radiation, respectively, of the GaAlAs diode laser and Nd:YAG laser penetrate more deeply into biologic tissues than the radiation of longer-wavelength IR lasers. Thus, the resulting thermal effects arise from absorption at greater depth within tissues, and the depths of coagulation and degree of hemostasis achieved with these lasers tend to be greater than with the longer-wavelength IR lasers. For example, the optical penetration depths (10% incident intensity) for 810-nm and 1.024-µm radiation are estimated to be 4.6 and 8.6 mm respectively in canine prostate tissue [Rastegar et al., 1992]. Energy deposition of 3600 J from each laser onto the urethral surface of the canine prostate results in maximum coagulation depths of 8 and 12 mm respectively using diode and Nd:YAG lasers [Motamedi et al., 1993]. Depths of optical penetration and coagulation in porcine liver, a more vascular tissue than prostate gland, of 2.8 and 9.6 mm, respectively, were obtained with a Nd:YAG laser beam, and of 7 and 12 mm respectively with an 810-nm diode laser beam [Rastegar et al., 1992]. The smaller penetration depth obtained with 810-nm diode radiation in liver than in prostate gland reflects the effect of greater vascularity (blood content) on near-IR propagation.

85.5 Effects of Visible-Range Laser Radiation Blood and vascular tissues very efficiently absorb radiation in the visible wavelength range due to the strong absorption of hemoglobin. This absorption underlies, for example, the use of:


1. The argon ion laser (488–514.5 nm) in the localized heating and thermal coagulation of the vascular choroid layer and adjacent retina, resulting in the anchoring of the retina in treatment of retinal detachment [Katoh and Peyman, 1988]. 2. The argon ion laser (488–514.5 nm), frequency-doubled Nd:YAG laser (532 nm), and dye laser radiation (585 nm) in the coagulative treatment of cutaneous vascular lesions such as port wine stains [Mordon et al., 1993]. 3. The argon ion (488–514.5 nm) and frequency-doubled Nd:YAG lasers (532 nm) in the ablation of pelvic endometrial lesions which contain brown iron-containing pigments [Keye et al., 1983]. Because of the large absorption by hemoglobin and iron-containing pigments, the incident laser radiation is essentially absorbed at the surface of the blood vessel or lesion, and the resulting thermal effects are essentially local [Miller and Veitch, 1993].

85.6 Effects of UV Laser Radiation Whereas exposure of tissue to IR and visible-light-range laser energy result in removal of tissue by thermal ablation, exposure to argon fluoride (ArF) laser radiation of 193-nm wavelength results predominantly in ablation of tissue initiated by a photochemical process [Garrison and Srinivasan, 1985]. This ablation arises from repulsive forces between like-charged regions of ionized protein molecules that result from ejection of molecular electrons following UV photon absorption [Garrison and Srinivasan, 1985]. Because the ionization and repulsive processes are extremely efficient, little of the incident laser energy escapes as thermal vibrational energy, and the extent of thermal coagulation damage adjacent to the site of incidence is very limited [Garrison and Srinivasan, 1985]. This feature and the ability to tune very finely the fluence emitted by the ArF laser so that micrometer depths of tissue can be removed have led to ongoing clinical trials to investigate the efficiency of the use of the ArF laser to selectively remove tissue from the surface of the human cornea for correction of short-sighted vision to eliminate the need for corrective eyewear [Van Saarloos and Constable, 1993].

85.7 Effects of Continuous and Pulsed IR-Visible Laser Radiation and Association Temperature Rise Heating following absorption of IR-visible laser radiation arises from molecular vibration during loss of the excitation energy and initially is manifested locally within the exposed region of tissue. If incidence of the laser energy is maintained for a sufficiently long time, the temperature within adjacent regions of biologic tissue increases due to heat diffusion. The mean squared distance < X 2> over which appreciable heat diffusion and temperature rise occur during exposure time t can be described in terms of the thermal diffusion time τ by the equation:

X 2 = τt

(85.1)

where τ is defined as the ratio of the thermal conductivity to the product of the heat capacity and density. For soft biologic tissues τ is approximately 1 × 103 cm2 s–1 [Meijering et al., 1993]. Thus, with continued energy input, the distance over which thermal diffusion and temperature rise occurs increases. Conversely, with use of pulsed radiation, the distance of heat diffusion can be made very small; for example, with exposure to a 1-µs pulse, the mean thermal diffusion distance is found to be approximately 0.3 µm, or about 3–10% of a biologic cell diameter. If the laser radiation is strongly absorbed and the ablation of tissues is efficient, then little energy diffuses away from the site of incidence, and lateral thermally induced coagulation of tissue can be minimized with pulses of short duration. The effect of limiting lateral thermal


damage is desirable in the cutting of cornea [Hibst et al., 1992] and sclera of the eye [Hill et al., 1993], and joint cartilage [Maes and Sherk, 1994], all of which are avascular (or nearly so, with cartilage), and the hemostasis arising from lateral tissue coagulation is not required.

85.8 General Description and Operation of Lasers Lasers emit a beam of intense electromagnetic radiation that is essentially monochromatic or contains at most a few nearly monochromatic wavelengths and is typically only weakly divergent and easily focused into external optical systems. These attributes of laser radiation depend on the key phenomenon which underlies laser operation, that of light amplification by stimulated emission of radiation, which in turn gives rise to the acronym LASER. In practice, a laser is generally a generator of radiation. The generator is constructed by housing a light-emitting medium within a cavity defined by mirrors which provide feedback of emitted radiation through the medium. With sustained excitation of the ionic or molecular species of the medium to give a large density of excited energy states, the spontaneous and attendant stimulated emission of radiation from these states by photons of identical wavelength (a lossless process), which is amplified by feedback due to photon reflection by the cavity mirrors, leads to the generation of a very large photon density within the cavity. With one cavity mirror being partially transmissive, say 0.1 to 1%, a fraction of the cavity energy is emitted as an intense beam. With suitable selection of a laser medium, cavity geometry, and peak wavelengths of mirror reflection, the beam is also essentially monochromatic and very nearly collimated. Identity of the lasing molecular species or laser medium fixes the output wavelength of the laser. Laser media range from gases within a tubular cavity, organic dye molecules dissolved in a flowing inert liquid carrier and heat sink, to impurity-doped transparent crystalline rods (solid state lasers) and semiconducting diode junctions [Lengyel, 1971]. The different physical properties of these media in part determine the methods used to excite them into lasing states. Gas-filled, or gas lasers are typically excited by dc or rf electric current. The current either ionizes and excites the lasing gas, e.g., argon, to give the electronically excited and lasing Ar+ ion, or ionizes a gaseous species in a mixture also containing the lasing species, e.g., N2, which by efficient energy transfer excites the lasing molecular vibrational states of the CO2 molecule. Dye lasers and so-called solid-state lasers are typically excited by intense light from either another laser or from a flash lamp. The excitation light wavelength range is selected to ensure efficient excitation at the absorption wavelength of the lasing species. Both excitation and output can be continuous, or the use of a pulsed flashlamp or pulsed exciting laser to pump a solid-state or dye laser gives pulsed output with high peak power and short pulse duration of 1 µs to 1 ms. Repeated excitation gives a train of pulses. Additionally, pulses of higher peak power and shorter duration of approximately 10 ns can be obtained from solid lasers by intracavity Q-switching [Lengyel, 1971]. In this method, the density of excited states is transiently greatly increased by impeding the path between the totally reflecting and partially transmitting mirror of the cavity interrupting the stimulated emission process. Upon rapid removal of the impeding device (a beam-interrupting or -deflecting device), stimulated emission of the very large population of excited lasing states leads to emission of an intense laser pulse. The process can give single pulses or can be repeated to give a pulse train with repetition frequencies typically ranging from 1 Hz to 1kHz. Gallium aluminum (GaAlAs) lasers are, as are all semiconducting diode lasers, excited by electrical current which creates excited hole-electron pairs in the vicinity of the diode junction. Those carrier pairs are the lasing species which emit spontaneously and with photon stimulation. The beam emerges parallel to the function with the plane of the function forming the cavity and thin-layer surface mirrors providing reflection. Use of continuous or pulsed excitation current results in continuous or pulsed output.


85.9 Biomedical Laser Beam Delivery Systems Beam delivery systems for biomedical lasers guide the laser beam from the output mirror to the site of action on tissue. Beam powers of up to 100 W are transmitted routinely. All biomedical lasers incorporate a coaxial aiming beam, typically from a HeNe laser (632.8 nm) to illuminate the site of incidence on tissue. Usually, the systems incorporate two different beam-guiding methods, either (1) a flexible fused silica (SiO2) optical fiber or light guide, generally available currently for laser beam wavelengths between 400 nm and 2.1 µm, where SiO2 is essentially transparent and (2) an articulated arm having beamguiding mirrors for wavelengths greater than circa 2.1 µm (e.g., CO2 lasers), for the Er:YAG and for pulsed lasers having peak power outputs capable of causing damage to optical fiber surfaces due to ionization by the intense electric field (e.g., pulsed ruby). The arm comprises straight tubular sections articulated together with high-quality power-handling dielectric mirrors at each articulation junction to guide the beam through each of the sections. Fused silica optical fibers usually are limited to a length of 1–3 m and to wavelengths in the visible-to-low midrange IR ( θc where

n  sin θc =  1   n2 

(85.2)

n  cos α c =  1   n2 

(85.3)

or in terms of the complementary angle αc

For a focused input beam with apical angle αm incident upon the flat face of the fiber as shown in Fig. 85.1, total internal reflection and beam guidance within the fiber core will occur [Levi, 1980] for

(

) [

NA = sin α m 2 ≤ n22 − n12 where NA is the numerical aperture of the fiber.


]

0.5

(85.4)

FIGURE 85.1

Critical reflection and propagation within an optical fiber.

This relationship ensures that the critical angle of incidence of the interface is not exceeded and that total internal reflection occurs [Levi, 1980]. Typical values of NA for fused SiO2 fibers with polymer cladding are in the range of 0.36–0.40. The typical values of αm = 14 degrees used to insert the beam of the biomedical laser into the fiber is much smaller than those values (21–23 degrees) corresponding to typical NA values. The maximum value of the propagation angle α typically used in biomedical laser systems is 4.8 degrees. Leakage of radiation at the core-cladding interface of the fused SiO2 fiber is negligible, typically being 0.3 dB/m at 400 nm and 0.01 dB/m at 1.064 µm. Bends along the fiber length always decrease the angle of the incidence at the core cladding interface. Bends do not give appreciable losses for values of the bending radius sufficiently large that the angle of incidence θ of the propagating beam in the bent core does not becomes less than θc at the core-cladding interface [Levi, 1980]. The relationship given by Levi [1980] between the bending radius rb , the fiber core radius ro, the ratio (n2/n1) of fiber core to cladding refractive indices, and the propagation angle α in Fig. 85.1 which ensures that the beam does not escape is

n1 1 − ρ > cos α n2 1 + ρ

(85.5)

where ρ = (ro/rb). The inequality will hold for all α ≤ αc provided that

n1 1 − ρ ≤ n2 1 + ρ

(85.6)

Thus, the critical bending radius rbc is the value of rb such that Eq. (85.6) is an equality. Use of Eq. (85.6) predicts that bends with radii ≥ 12, 18, and 30 mm, respectively, will not result in appreciable beam leakage from fibers having 400-, 600-, and 1000-micron diameter cores, which are typical in biomedical use. Thus, use of fibers in flexible endoscopes usually does not compromise beam guidance. Because the integrity of the core-cladding interface is critical to beam guiding, the clad fiber is encased typically in a tough but flexible protective fluoropolymer buffer coat.

Mirrored Articulated Arm Characteristics Typically two or three relatively long tubular sections or arms of 50–80 cm length make up the portion of the articulated arm that extends from the laser output fixturing to the handpiece, endoscope, or operating microscope stage used to position the laser beam onto the tissue proper. Mirrors placed at the articulation of the arms and within the articulated handpiece, laparoscope, or operating microscope stage maintain the centration of the trajectory of the laser beam along the length of the delivery system. Dielectric multilayer mirrors [Levi, 1980] routinely are used in articulated devices. Their low high reflectivity ≤ 99.9 + % and power-handling capabilities ensure efficient power transmission down the © 2000 by CRC Press LLC

arm. Mirrors in articulated devices typically are held in kinetically adjustable mounts for rapid stable alignment to maintain beam concentration.

Optics for Beam Shaping on Tissues Since the rate of heating on tissue, and hence rates of ablation and coagulation, depends directly on energy input per unit volume of tissue, selection of ablation and coagulation rates of various tissues is achieved through control of the energy density (J/cm2 or W·s/cm2) of the laser beam. This parameter is readily achieved through use of optical elements such as discrete focusing lenses placed in the handpiece or rigid endoscope which control the spot size upon the tissue surface or by affixing a so-called contact tip to the end of an optical fiber. These are conical or spherical in shape with diameters ranging from 300–1200 µm and with very short focal lengths. The tip is placed in contact with the tissue and generates a submillimeter-sized focal spot in tissue very near the interface between the tip and tissue. One advantage of using the contact tip over a focused beam is that ablation proceeds with small lateral depth of attendant coagulation [Judy et al., 1993a]. This is because the energy of the tightly focused beam causes tissue thermolysis essentially at the tip surface and because the resulting tissue products strongly absorb the beam resulting in energy deposition and ablation essentially at the tip surface. This contrasts with the radiation penetrating deeply into tissue before thermolysis which occurs with a less tightly focused beam from a free lens or fiber. An additional advantage with the use of contact tips in the perception of the surgeon is that the kinesthetics of moving a contact tip along a resisting tissue surface more closely mimics the “touch” encountered in moving a scalpel across the tissue surface. Recently a class of optical fiber tips has been developed which laterally directs the beam energy from a silica fiber [Judy et al., 1993b]. These tips, either a gold reflective micromirror or an angled refractive prism, offer a lateral angle of deviation ranging from 35–105 degrees from the optical fiber axis (undeviated beam direction). The beam reflected from a plane micromirror is unfocused and circular in crosssection, whereas the beam from a concave mirror and refractive devices is typically elliptical in shape, fused with distal diverging rays. Fibers with these terminations are currently finding rapidly expanding, large-scale application in coagulation (with 1.064-µm Nd:YAG laser radiation) of excess tissue lining the urethra in treatment of benign prostatic hypertrophy [Costello et al., 1992]. The capability for lateral beam direction may offer additional utility of these terminated fibers in other clinical specialties.

Features of Routinely Used Biomedical Lasers Currently four lasers are in routine large-scale clinical biomedical use to ablate, dissect, and to coagulate soft tissue. Two, the carbon dioxide (CO2) and argon ion (Ar-ion) lasers, are gas-filled lasers. The other two employ solid-state lasing media. One is the Neodymium-yttrium-aluminum-garnet (Nd:YAG) laser, commonly referred to as a solid-state laser, and the other is the gallium-aluminum arsenide (GaAlAs) semiconductor diode laser. Salient features of the operating characteristics and biomedical applications of those lasers are listed in Tables 85.2 to 85.5. The operational descriptions are typical of the lasers currently available commercially and do not represent the product of any single manufacturer.

Other Biomedical Lasers Some important biomedical lasers have smaller-scale use or currently are being researched for biomedical application. The following four lasers have more limited scales of surgical use: The Ho:YAG (Holmium:YAG) laser, emitting pulses of 2.1 µm wavelength and up to 4 J in energy, used in soft tissue ablation in arthroscopic (joint) surgery (FDA approved). The Q-switched Ruby (Cr:Al203) laser, emitting pulses of 694-nm wavelength and up to 2 J in energy is used in dermatology to disperse black, blue, and green tattoo pigments and melanin in pigmented lesions (not melanoma) for subsequent removal by phagocytosis by macrophages (FDA approved).


TABLE 85.2

Operating Characteristics of Principal Biomedical Lasers

Characteristics

Ar Ion Laser

CO2 Laser

Argon gas, 133 Pa Ar+ ion Electric discharge, continuous 208 VAC, 60 A 0.06%

10% CO2 10% Ne, 80% He; 1330 Pa CO2 molecule Electric discharge, continuous, pulsed 110 VAC , 15 A 10%

Characteristics

Nd:YAG Laser

GaAlAs Diode Laser

Cavity medium Lasing species Excitation Electric input

Nd-doped YAG Nd3t in YAG lattice Flashlamp, continuous, pulsed 208/240 VAC , 30 A continuous 110 VAC , 10 A pulsed 1%

n-p junction, GaAlAs diode Hole-electron pairs at diode junction Electric current, continuous pulsed 110 VAC , 15 A

Cavity medium Lasing species Excitation Electric input Wall plug efficiency

Wall plug efficiency

TABLE 85.3

23%

Output Beam Characteristics of Ar-Ion and CO2 Biomedical Lasers

Output Characteristics

Argon Laser

Output power Wavelength(s) Electromagnetic wave propagation mode Beam guidance, shaping

2–8 W, continuous Multiple lines (454.6–528.7 nm), 488, 514.5 dominant TEM∞

1–100 W, continuous 10.6 µm TEM∞

Fused silica optical fiber with contact tip or flat-ended for beam emission, lensed handpiece. Slit lamp with ocular lens

Flexible articulated arm with mirrors; lensed handpiece or mirrored microscope platen

TABLE 85.4

Output Beam Characteristics of Nd:YAG and GaAlAs Diode Biomedical Lasers

Output Characteristics Output power

Wavelength(s) Electromagnetic wave propagation modes Beam guidance and shaping

TABLE 85.5

CO2 Laser

Nd:YAG Lasers

GaAlAs Diode Laser

1–100 W continuous at 1.064 millimicron 1–36 W continuous at 532 nm (frequency doubled with KTP) 1.064 µm/532 nm Mixed modes

1–25 W continuous

Fused SiO2 optical fiber with contact tip directing mirrored or refracture tip

Fused SiO2 optical fiber with contact tip or laterally directing mirrored or refracture tip

810 nm Mixed modes

Clinical Uses of Principal Biomedical Lasers

Ar-ion laser: Pigmented (vascular) soft-tissue ablation in gynecology; general and oral surgery; otolaryngology; vascular lesion coagulation in dermatology; retinal coagulation in ophthalmology Nd:YAG laser: Soft-tissue, particularly pigmented vascular tissue, ablation—dissection and bulk tissue removal—in dermatology; gastroenterology; gynecology; general, arthroscopic, neuro-, plastic, and thoracic surgery; urology; posterior capsulotomy (ophthalmology) with pulsed 1.064 millimicron and ocular lens


CO2 laser: Soft-tissue ablation—dissection and bulk tissue removal in dermatology; gynecology; general, oral, plastic, and neurosurgery; otolaryngology; podiatry; urology GaAlAs diode laser: Pigmented (vascular) softtissue ablation—dissection and bulk removal in gynecology; gastroenterology, general surgery, and urology; FDA approval for otolaryngology and thoracic surgery pending

The flashlamp pumped pulsed dye laser emitting 1- to 2-J pulses at either 577- or 585-nm wavelength (near the 537–577 absorption region of blood) is used for treatment of cutaneous vascular lesions and melanin pigmented lesions except melanoma. Use of pulsed radiation helps to localize the thermal damage to within the lesions to obtain low damage of adjacent tissue. The following lasers are being investigated for clinical uses. The Er:YAG laser, emitting at 2.94 µm near the major water absorption peak (OH stretch), is currently being investigated for ablation of tooth enamel and dentin [Li et al., 1992]. Dye lasers emitting at 630–690 nm are being investigated for application as light sources for exciting dihematoporphyrin ether or benzoporphyrin derivatives in investigation of the efficacy of these photosensitives in the treatment of esophageal, bronchial, and bladder carcinomas for the FDA approved process.

Defining Terms Biomedical Laser Radiation Ranges Infrared (IR) radiation: The portion of the electromagnetic spectrum within the wavelength range 760 nm–1 mm, with the regions 760 nm–1.400 µm and 1.400–10.00 µm, respectively, called the near- and mid-IR regions. Ultraviolet (UV) radiation: The portion of the electromagnetic spectrum within the wavelength range 100–400 nm. Visible (VIS) radiation: The portion of the electromagnetic spectrum within the wavelength range 400–760 nm. Laser Medium Nomenclature Argon fluoride (ArF): Argon fluoride eximer laser (an eximer is a diatomic molecule which can exist only in an excited state). Ar ion: Argon ion. CO2 : Carbon dioxide. Cr:Al203 : Ruby laser. Er:YAG: Erbium yttrium aluminum garnet. GaAlAs: Gallium aluminum laser. HeNe: Helium neon laser. Ho:YAG: Holmium yttrium aluminum garnet. Nd:YAG: Neodymium yttrium aluminum garnet. Optical Fiber Nomenclature Ag halide: Silver halide, halide ion, typically bromine (Br) and chlorine (Cl). Fused silica: Fused SiO2.

References Bass LS, Moazami N, Pocsidio J, et al. 1992. Change in type I collagen following laser welding. Lasers Surg Med 12(5):500. Costello AJ, Johnson DE, Bolton DM. 1992. Nd:YAG laser ablation of the prostate as a treatment for benign prostate hypertrophy. Lasers Surg Med 12(2):121. Fitzpatrick RE. 1993. Comparison of the Q-switched ruby, Nd:YAG, and alexandrite lasers in tattoo removal. Lasers Surg Med Suppl 6(266):52. Gannot I, Dror J, Calderon S, et al. 1994. Flexible waveguides for IR laser radiation and surgery applications. Lasers Surg Med 14(2):184.


Garrison BJ, Srinivasan R. 1985. Laser ablation of organic polymers: microscopic models for photochemical and thermal processes. J Appl Physiol 58(9):2909. Grossweiner LI. 1989. Photophysics. In KC Smith (ed), The Science of Photobiology, p 1–47. New York, Plenum. Hale GM, Querry MR. 1973. Optical constants of water in the 200 nm to 200 µm wavelength region. Appl Optics 12(12):555. Hibst R, Bende T, Schröder D. 1992. Wet corneal ablation by Er:YAG laser radiation. Lasers Surg Med Suppl 4(236):56. Hill RA, Le MT, Yashiro H, et al. 1993. Ab-interno erbium (Er:YAG) laser schlerostomy with iridotomy in dutch cross rabbits. Lasers Surg Med 13(5):559. Judy MM, Matthews JL, Aronoff BL, et al. 1993a. Soft tissue studies with 805 nm diode laser radiation: Thermal effects with contact tips and comparison with effects of 1064 nm. Nd:YAG laser radiation. Lasers Surg Med 13(5):528. Judy MM, Matthews JL, Gardetto WW, et al. 1993b. Side firing laser-fiber technology for minimally invasive transurethral treatment of benign prostate hyperplasia. Proc Soc Photo-Optical Instr Eng (SPIE) 1982:86. Judy MM, Matthews JL, Goodson JR, et al. 1992. Thermal effects in tissues from simultaneous coaxial CO2 and Nd:YAG laser beams. Lasers Surg Med 12(2):222. Katoh N, Peyman GA. 1988. Effects of laser wavelengths on experimental retinal detachments and retinal vessels. Jpn J Ophthalmol 32(2):196. Keye WR, Matson GA, Dixon J. 1983. The use of the argon laser in treatment of experimental endometriosis. Fertil Steril 39(1):26. Lakowicz JR. 1983. Principles of Fluorescence Spectroscopy. New York, Plenum. Lengyel BA. 1971. Lasers. New York, John Wiley. Levi L. 1980. Applied Optics, vol 2. New York, John Wiley. Li ZZ, Code JE, Van de Merve WP. 1992. Er:YAG laser ablation of enamel and dentin of human teeth: determination of ablation rates at various fluences and pulse repetition rates. Laser Surg Med 12(6):625. Maes KE, Sherk HH. 1994. Bone and meniscal ablation using the erbium YAG laser. Lasers Surg Med Suppl 6(166):31. Mahadevan A, Mitchel MF, Silva E, et al. 1993. Study of the fluorescence properties of normal and neoplastic human cervical tissue. Lasers Surg Med 13(6):647. McGillis ST, Bailin PL, Fitzpatrick RE, et al. 1994. Successful treatments of blue, green, brown and reddishbrown tattoos with the Q-switched alexandrite laser. Laser Surg Med Suppl 6(270):52. Meijering LJT, VanGermert MJC, Gijsbers GHM, et al. 1993. Limits of radial time constants to approximate thermal response of tissue. Lasers Surg Med 13(6):685. Merberg GN. 1993. Current status of infrared fiberoptics for medical laser power delivery. Lasers Surg Med 13(5):572. Miller ID, Veitch AR. 1993. Optical modeling of light distributions in skin tissue following laser irradiation. Lasers Surg Med 13(5):565. Mordon S, Beacco C, Rotteleur G, et al. 1993. Relation between skin surface temperature and minimal blanching during argon, Nd:YAG 532, and cw dye 585 laser therapy of port-wine stains. Lasers Surg Med 13(1):124. Motamedi M, Torres JH, Cammack T, et al. 1993. Thermodynamics of cw laser interaction with prostatic tissue: Effects of simultaneous cooling on lesion size. Lasers Surg Med Suppl 5(314):64. Oz MC, Chuck RS, Johnson JP, et al. 1989. Indocyanine green dye-enhanced welding with a diode laser. Surg Forum 40(4):316. Rastegar S, Jacques SC, Motamedi M, et al. 1992. Theoretical analysis of high-power diode laser (810 nm) and Nd:YAG laser (1064 nm) for coagulation of tissue: Predictions for prostate coagulation. Proc Soc Photo-Optical Instr Eng (SPIE) 1646:150.


Schroder T, Brackett K, Joffe S. 1987. An experimental study of effects of electrocautery and various lasers on gastrointestinal tissue. Surgery 101(6):691. Spikes JD. 1989. Photosensitization. In KC Smith (ed), The Science of Photobiology, 2d ed, pp 79–110. New York, Plenum. Van de Hulst HC. 1957. Light Scattering by Small Particles. New York, John Wiley. Van Saarloos PP, Constable IJ. 1993. Improved eximer laser photorefractive keratectomy system. Lasers Surg Med 13(2):189. White A, Handler P, Smith EL.1968. Principles of Biochemistry, 4th ed. New York, McGraw-Hill.

Further Information Current research on the optical, thermal, and photochemical interactions of radiation and their effect on biologic tissues, are published routinely in the journals: Laser in Medicine and Surgery, Lasers in the Life Sciences, and Photochemistry Photobiology and to a lesser extent in Applied Optics and Optical Engineering. Clinical evaluations of biomedical laser applications appear in Lasers and Medicine and Surgery and in journals devoted to clinical specialties such as Journal of General Surgery, Journal of Urology, Journal of Gastroenterological Surgery. The annual symposium proceedings of the biomedical section of the Society of Photo-Optical Instrumentation Engineers (SPIE) contain descriptions of new and current research on application of lasers and optics in biomedicine. The book Lasers (a second edition by Bela A. Lengyel), although published in 1971, remains a valuable resource on the fundamental physics of lasers—gas, dye solid-state, and semiconducting diode. A more recent book, The Laser Guidebook by Jeffrey Hecht, published in 1992, emphasizes the technical characteristics of the gas, diode, solid-state, and semiconducting diode lasers. The Journal of Applied Physics, Physical Review Letters, and Applied Physics Letters carry descriptions of the newest advances and experimental phenomena in lasers and optics. The book Safety with Lasers and Other Optical Sources by David Sliney and Myron Wolbarsht, published in 1980, remains a very valuable resource on matters of safety in laser use. Laser safety standards for the United States are given for all laser uses and types in the American National Standard (ANSI) Z136.1-1993, Safe Use of Lasers.


Flewelling, R. “Noninvasive Optical Monitoring.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

86 Noninvasive Optical Monitoring 86.1

Oximetry and Pulse Oximetry Background • Theory • Applications and Future Directions

86.2

Ross Flewelling Nellcor Incorporation

Nonpulsatile Spectroscopy Background • Cytochrome Spectroscopy • Near-Infrared Spectroscopy and Glucose Monitoring • Time-Resolved Spectroscopy

86.3

Conclusions

Optical measures of physiologic status are attractive because they can provide a simple, noninvasive, yet real-time assessment of medical condition. Noninvasive optical monitoring is taken here to mean the use of visible or near-infrared light to directly assess the internal physiologic status of a person without the need of extracting a blood of tissue sample or using a catheter. Liquid water strongly absorbs ultraviolet and infrared radiation, and thus these spectral regions are useful only for analyzing thin surface layers or respiratory gases, neither of which will be the subject of this review. Instead, it is the visible and near-infrared portions of the electromagnetic spectrum that provide a unique “optical window” into the human body, opening new vistas for noninvasive monitoring technologies. Various molecules in the human body possess distinctive spectral absorption characteristics in the visible or near-infrared spectral regions and therefore make optical monitoring possible. The most strongly absorbing molecules at physiologic concentrations are the hemoglobins, myoglobins, cytochromes, melanins, carotenes, and bilirubin (see Fig. 86.1 for some examples). Perhaps less appreciated are the less distinctive and weakly absorbing yet ubiquitous materials possessing spectral characteristics in the near-infrared: water, fat, proteins, and sugars. Simple optical methods are now available to quantitatively and noninvasively measure some of these compounds directly in intact tissue. The most successful methods to date have used hemoglobins to assess the oxygen content of blood, cytochromes to assess the respiratory status of cells, and possibly near-infrared to assess endogenous concentrations of metabolytes, including glucose.

86.1 Oximetry and Pulse Oximetry Failure to provide adequate oxygen to tissues—hypoxia—can in a matter of minutes result in reduced work capacity of muscles, depressed mental activity, and ultimately cell death. It is therefore of considerable interest to reliably and accurately determine the amount of oxygen in blood or tissues. Oximetry is the determination of the oxygen content of blood of tissues, normally by optical means. In the clinical laboratory the oxygen content of whole blood can be determined by a bench-top cooximeter or blood


FIGURE 86.1 Absorption spectra of some endogenous biologic materials (a) hemoglobins, (b) cytochrome aa3, (c) myoglobins, and (d) melanin.

gas analyzer. But the need for timely clinical information and the desire to minimize the inconvenience and cost of extracting a blood sample and later analyze it in the lab has led to the search for alternative noninvasive optical methods. Since the 1930s, attempts have been made to use multiple wavelengths of light to arrive at a complete spectral characterization of a tissue. These approaches, although somewhat successful, have remained of limited utility owing to the awkward instrumentation and unreliable results. It was not until the invention of pulse oximetry in the 1970s and its commercial development and application in the 1980s that noninvasive oximetry became practical. Pulse oximetry is an extremely easyto-use, noninvasive, and accurate measurement of real-time arterial oxygen saturation. Pulse oximetry is now used routinely in clinical practice, has become a standard of care in all U.S. operating rooms, and is increasingly used wherever critical patients are found. The explosive growth of this new technology and its considerable utility led John Severinghaus and Poul Astrup [1986] in an excellent historical review to conclude that pulse oximetry was “arguably the most significant technological advance ever made in monitoring the well-being and safety of patients during anesthesia, recovery and critical care.”


FIGURE 86.2 Hemoglobin oxygen dissociation curve showing the sigmoidal relationship between the partial pressure of oxygen and the oxygen saturation of blood. The curve is given approximately by %SaO2 = 100%/[1 + P50 /pO2n], with n = 2.8 and P50 = 26 mm Hg.

Background The partial pressure of oxygen (pO2) in tissues need only be about 3 mmHg to support basic metabolic demands. This tissue level, however, requires capillary pO2 to be near 40 mmHg, with a corresponding arterial pO2 of about 95 mmHg. Most of the oxygen carried by blood is stored in red blood cells reversibly bound to hemoglobin molecules. Oxygen saturation (SaO2) is defined as the percentage of hemoglobinbound oxygen compared to the total amount of hemoglobin available for reversible oxygen binding. The relationship between the oxygen partial pressure in blood and the oxygen saturation of blood is given by the hemoglobin oxygen dissociation curve as shown in Fig. 86.2. The higher the pO2 in blood, the higher the SaO2 . But due to the highly cooperative binding of four oxygen molecules to each hemoglobin molecule, the oxygen binding curve is sigmoidal, and consequently the SaO2 value is particularly sensitive to dangerously low pO2 levels. With a normal arterial blood pO2 above 90 mmHg, the oxygen saturation should be at least 95%, and a pulse oximeter can readily verify a safe oxygen level. If oxygen content falls, say to a pO2 below 40 mmHg, metabolic needs may not be met, and the corresponding oxygen saturation will drop below 80%. Pulse oximetry therefore provides a direct measure of oxygen sufficiency and will alert the clinician to any danger of imminent hypoxia in a patient. Although endogenous molecular oxygen is not optically observable, hemoglobin serves as an oxygensensitive “dye” such that when oxygen reversibly binds to the iron atom in the large heme prosthetic group, the electron distribution of the heme is shifted, producing a significant color change. The optical absorption of hemoglobin in its oxygenated and deoxygenated states is shown in Fig. 86.1. Fully oxygenated blood absorbs strongly in the blue and appears bright red; deoxygenated blood absorbs through the visible region and is very dark (appearing blue when observed through tissue due to light scattering effects). Thus the optical absorption spectra of oxyhemaglobin (O2Hb) and “reduced” deoxyhemoglobin (RHb) differ substantially, and this difference provides the basis for spectroscopic determinations of the proportion of the two hemoglobin states. In addition to these two normal functional hemoglobins, there are also dysfunctional hemoglobins—carboxyhemoglobin, methemoglobin, and sulhemoglobin—which are spectroscopically distinct but do not bind oxygen reversibly. Oxygen saturation is therefore defined in Eq. (86.1) only in terms of the functional saturation with respect to O2Hb and RHb:


SaO2 =

O2 Hb × 100% RHb + O2 Hb

(86.1)

Cooximeters are bench-top analyzers that accept whole blood samples and utilize four or more wavelengths of monochromatic light, typically between 500 and 650 nm, to spectroscopically determine the various individual hemoglobins in the sample. If a blood sample can be provided, this spectroscopic method is accurate and reliable. Attempts to make an equivalent quantitative analysis noninvasively through intact tissue have been fraught with difficulty. The problem has been to contend with the wide variation in scattering and nonspecific absorption properties of very complex heterogeneous tissue. One of the more successful approaches, marketed by Hewlett-Packard, used eight optical wavelengths transmitted through the pinna of the ear. In this approach a “bloodless” measurement is first obtained by squeezing as much blood as possible from an area of tissue; the arterial blood is then allowed to flow back, and the oxygen saturation is determined by analyzing the change in the spectral absorbance characteristics of the tissue. While this method works fairly well, it is cumbersome, operator dependent, and does not always work well on poorly perfused or highly pigmented subjects. In the early 1970s, Takuo Aoyagi recognized that most of the interfering nonspecific tissue effects could be eliminated by utilizing only the change in the signal during an arterial pulse. Although an early prototype was built in Japan, it was not until the refinements in implementation and application by Biox (now Ohmeda) and Nellcor Incorporated in the 1980s that the technology became widely adopted as a safety monitor for critical care use.

Theory Pulse oximetry is based on the fractional change in light transmission during an arterial pulse at two different wavelengths. In this method the fractional change in the signal is due only to the arterial blood itself, and therefore the complicated nonpulsatile and highly variable optical characteristics of tissue are eliminated. In a typical configuration, light at two different wavelengths illuminating one side of a finger will be detected on the other side, after having traversed the intervening vascular tissues (Fig. 86.3). The transmission of light at each wavelength is a function of the thickness, color, and structure of the skin, tissue, bone, blood, and other material through which the light passes. The absorbance of light by a sample is defined as the negative logarithm of the ratio of the light intensity in the presence of the sample (I) to that without (Io): A = –log(I/Io). According to the Beer-Lambert law, the absorbance of a sample

FIGURE 86.3 Typical pulse oximeter sensing configuration on a finger. Light at two different wavelengths is emitted by the source, diffusely scattered through the finger, and detected on the opposite side by a photodetector.


at a given wavelength with a molar absorptivity () is directly proportional to both the concentration (c) and pathlength (l) of the absorbing material: A = cl. (In actuality, biologic tissue is highly scattering, and the Beer-Lambert law is only approximately correct; see the references for further elaboration). Visible or near-infrared light passing through about one centimeter of tissue (e.g., a finger) will be attenuated by about one or two orders of magnitude for a typical emitter-detector geometry, corresponding to an effective optical density (OD) of 1–2 OD (the detected light intensity is decreased by one order of magnitude for each OD unit). Although hemoglobin in the blood is the single strongest absorbing molecule, most of the total attenuation is due to the scattering of light away from the detector by the highly heterogeneous tissue. Since human tissue contains about 7% blood, and since blood contains typically about 14 g/dL hemoglobin, the effective hemoglobin concentration in tissue is about 1 g/dL (~150 uM). At the wavelengths used for pulse oximetry (650–950 nm), the oxy- and deoxyhemoglobin molar absorptivities fall in the range of 100–1000 M–1cm–1, and consequently hemoglobin accounts for less than 0.2 OD of the total observed optical density. Of this amount, perhaps only 10% is pulsatile, and consequently pulse signals of only a few percent are ultimately measured, at times even one-tenth of this. A mathematical model for pulse oximetry begins by considering light at two wavelengths, λ 1 and λ 2, passing through tissue and being detected at a distant location as in Fig. 86.3. At each wavelength the total light attenuation is described by four different component absorbances: oxyhemoglobin in the blood (concentration co , molar absorptivity o , and effective pathlength lo), “reduced” deoxyhemoglobin in the blood (concentration cr , molar absorptivity r, and effective pathlength lr ), specific variable absorbances that are not from the arterial blood (concentration cx , molar absorptivity x , and effective pathlength lx), and all other non-specific sources of optical attenuation, combined as Ay , which can include light scattering, geometric factors, and characteristics of the emitter and detector elements. The total absorbance at the two wavelengths can then be written:

 Aλ = o colo + r cr lr + x c x l x + Ay 1 1 1 1 1  A = c l + c l + c l + A o2 o o r2 r r x2 x x y2  λ 2

(86.2)

The blood volume change due to the arterial pulse results in a modulation of the measured absorbances. By taking the time rate of change of the absorbances, the two last terms in each equation are effectively zero, since the concentration and effective pathlength of absorbing material outside the arterial blood do not change during a pulse [d(cx lx )/dt = 0], and all the nonspecific effects on light attenuation are also effectively invariant on the time scale of a cardiac cycle (dAy /dt = 0). Since the extinction coefficients are constant, and the blood concentrations are constant on the time scale of a pulse, the time-dependent changes in the absorbances at the two wavelengths can be assigned entirely to the change in the blood pathlength (dlo /dt and dlr /dt). With the additional assumption that these two blood pathlength changes are equivalent (or more generally, their ratio is a constant), the ratio R of the time rate of change of the absorbance at wavelength 1 to that at wavelength 2 reduces to the following:

R=

dAλ1 dt dAλ 2 dt

=

( ) = (∆I I ) = c + c −d log ( I I ) dt ( ∆I I ) c + c −d log I1 I o dt 2

o

1

2

1

o1 o

r1 r

2

o2 o

r2 r

(86.3)

Observing that functional oxygen saturation is given by S = co /(co + cr), and that (1–S) = cr /(co + cr), the oxygen saturation can then be written in terms of the ratio R as follows

S=


(

r1 − r 2 R

) (

)

r 1 − o1 − r 2 − o 2 R

(86.4)

FIGURE 86.4 Relationship between the measured ratio of fractional changes in light intensity at two wavelengths, R, and the oxygen saturation S. Beer-Lambert model is from Eq. (86.4) with o1 = 100, o2 = 300, r1 = 800, and r2 = 200. Empirical calibration is based on %S = 100% × (a – bR)/(c – dR) with a = 1000, b = 550, c = 900, and d = 350, with a linear extrapolation below 70%.

Equation (86.4) provides the desired relationship between the experimentally determined ratio R and the clinically desired oxygen saturation S. In actual use, commonly available LEDs are used as the light sources, typically a red LED near 660 nm and a near-infrared LED selected in the range 890–950 nm. Such LEDs are not monochromatic light sources, typically with bandwidths between 20 and 50 nm, and therefore standard molar absorptivities for hemoglobin cannot be used directly in Eq. (86.4). Further, the simple model presented above is only approximately true; for example, the two wavelengths do not necessarily have the exact same pathlength changes, and second-order scattering effects have been ignored. Consequently the relationship between S and R is instead determined empirically by fitting the clinical data to a generalized function of the form S = (a – bR)/(c – dR). The final empirical calibration will ultimately depend on the details of an individual sensor design, but these variations can be determined for each sensor and included in unique calibration parameters. A typical empirical calibration for R versus S is shown in Fig. 86.4, together with the curve that standard molar absorptivities would predict. In this way the measurement of the ratio of the fractional change in signal intensity of the two LEDs is used along with the empirically determined calibration equation to obtain a beat-by-beat measurement of the arterial oxygen saturation in a perfused tissue—continuously, noninvasively, and to an accuracy of a few percent.

Application and Future Directions Pulse oximetry is now routinely used in nearly all operating rooms and critical care areas in the United States and increasingly throughout the world. It has become so pervasive and useful that it is now being called the “fifth” vital sign (for an excellent review of practical aspects and clinical applications of the technology see Kelleher [1989]). The principal advantages of pulse oximetry are that it provides continuous, accurate, and reliable monitoring of arterial oxygen saturation on nearly all patients, utilizing a variety of convenient sensors, reusable as well as disposable. Single-patient-use adhesive sensors can easily be applied to fingers for adults and children and to arms for legs or neonates. Surface reflectance sensors have also been developed based on the same principles and offer a wider choice for sensor location, though they tend to be less accurate and prone to more types of interference.


Limitations of pulse oximetry include sensitivity to high levels of optical or electric interference, errors due to high concentrations of dysfunctional hemoglobins (methemoglobin or carboxyhemoglobin) or interference from physiologic dyes (such as methylene blue). Other important factors, such as total hemoglobin content, fetal hemoglobin, or sickle cell trait, have little or no effect on the measurement except under extreme conditions. Performance can also be compromised by poor signal quality, as may occur for poorly perfused tissues with weak pulse amplitudes or by motion artifact. Hardware and software advances continue to provide more sensitive signal detection and filtering capabilities, allowing pulse oximeters to work better on more ambulatory patients. Already some pulse oximeters incorporate ECG synchronization for improved signal processing. A pulse oximeter for use in labor and delivery is currently under active development by several research groups and companies. A likely implementation may include use of a reflectance surface sensor for the fetal head to monitor the adequacy of fetal oxygenation. This application is still in active development, and clinical utility remains to be demonstrated.

86.2 Nonpulsatile Spectroscopy Background Nonpulsatile optical spectroscopy has been used for more than half a century for noninvasive medical assessment, such as in the use of multiwavelength tissue analysis for oximetry and skin reflectance measurement for bilirubin assessment in jaundiced neonates. These early applications have found some limited use, but with modest impact. Recent investigations into new nonpulsatile spectroscopy methods for assessment of deep-tissue oxygenation (e.g., cerebral oxygen monitoring), for evaluation of respiratory status at the cellular level, and for the detection of other critical analytes, such as glucose, may yet prove more fruitful. The former applications have led to spectroscopic studies of cytochromes in tissues, and the latter has led to considerable work into new approaches in near-infrared analysis of intact tissues.

Cytochrome Spectroscopy Cytochromes are electron-transporting, heme-containing proteins found in the inner membranes of mitochondria and are required in the process of oxidative phosphorylation to convert metabolytes and oxygen into CO2 and high-energy phosphates. In this metabolic process the cytochromes are reversibly oxidized and reduced, and consequently the oxidation-reduction states of cytochromes c and aa3 in particular are direct measures of the respiratory condition of the cell. Changes in the absorption spectra of these molecules, particularly near 600 nm and 830 nm for cytochrome aa3, accompany this shift. By monitoring these spectral changes, the cytochrome oxidation state in the tissues can be determined (see, for example, Jöbsis [1977] and Jöbsis et al. [1977]). As with all nonpulsatile approaches, the difficulty is to remove the dependence of the measurements on the various nonspecific absorbing materials and highly variable scattering effects of the tissue. To date, instruments designed to measure cytochrome spectral changes can successfully track relative changes in brain oxygenation, but absolute quantitation has not yet been demonstrated.

Near-Infrared Spectroscopy and Glucose Monitoring Near-infrared (NIR), the spectral region between 780 nm and 3000 nm, is characterized by broad and overlapping spectral peaks produced by the overtones and combinations of infrared vibrational modes. Figure 86.5 shows typical NIR absorption spectra of fat, water, and starch. Exploitation of this spectral region for in vivo analysis has been hindered by the same complexities of nonpulsatile tissue spectroscopy described above and is further confounded by the very broad and indistinct spectral features characteristic of the NIR. Despite these difficulties, NIR spectroscopy has garnered considerable attention, since it may enable the analysis of common analytes.


FIGURE 86.5

Typical near-infrared absorption spectra of several biologic materials.

Karl Norris and coworkers pioneered the practical application of NIR spectroscopy, using it to evaluate water, fat, and sugar content of agricultural products (see Osborne et al. [1993] and Burns and Cuirczak [1992]). The further development of sophisticated multivariate analysis techniques, together with new scattering models (e.g., Kubelka-Munk theory) and high-performance instrumentation, further extended the application of NIR methods. Over the past decade, many research groups and companies have touted the use of NIR techniques for medical monitoring, such as for determining the relative fat, protein, and water content of tissue, and more recently for noninvasive glucose measurement. The body composition analyses are useful but crude and are mainly limited to applications in nutrition and sports medicine. Noninvasive glucose monitoring, however, is of considerable interest. More than 2 million diabetics in the United States lance their fingers three to six times a day to obtain a drop of blood for chemical glucose determination. The ability of these individuals to control their glucose levels, and the quality of their life generally, would dramatically improve if a simple, noninvasive method for determining blood glucose levels could be developed. Among the noninvasive optical methods proposed for this purpose are optical rotation, NIR analysis, and raman spectroscopy. The first two have received the most attention. Optical rotation methods aim to exploit the small optical rotation of polarized light by glucose. To measure physiologic glucose levels in a 1-cm thick sample to an accuracy of 25 mg/dL would require instrumentation that can reliably detect an optical rotation of at least 1 millidegree. Finding an appropriate in vivo optical path for such measurements has proved most difficult, with most approaches looking to use either the aqueous humor or the anterior chamber of the eye [Coté et al., 1992; Rabinovitch et al., 1982]. Although several groups have developed laboratory analyzers that can measure such a small effect, so far in vivo measurement has not been demonstrated, due both to unwanted scattering and optical activity of biomaterials in the optical path and to the inherent difficulty in developing a practical instrument with the required sensitivity. NIR methods for noninvasive glucose determination are particularly attractive, although the task is formidable. Glucose has spectral characteristics near 1500 nm and in the 2000–2500 nm band where many other compounds also absorb, and the magnitude of the glucose absorbance in biologic samples is typically two orders of magnitude lower than those of water, fat, or protein. The normal detection limit for NIR spectroscopy is on the order of one part in 103, whereas a change of 25 mg/dL in glucose concentration corresponds to an absorbance change of 10–4 to 10–5. In fact, the temperature dependence of the NIR absorption of water alone is at least an order of magnitude greater than the signal from glucose


in solution. Indeed, some have suggested that the apparent glucose signature in complex NIR spectra may actually be the secondary effect of glucose on the water. Sophisticated chemometric (particularly multivariate analysis) methods have been employed to try to extract the glucose signal out of the noise (for methods reviews see Martens and Næs [1989] and Haaland [1992]). Several groups have reported using multivariate techniques to quantitate glucose in whole blood samples, with encouraging results [Haaland et al., 1992]. And despite all theoretical disputations to the contrary, some groups claim the successful application of these multivariate analysis methods to noninvasive in vivo glucose determination in patients [Robinson et al., 1992]. Yet even with the many groups working in this area, much of the work remains unpublished, and few if any of the reports have been independently validated.

Time-Resolved Spectroscopy The fundamental problem in making quantitative optical measurements through intact tissue is dealing with the complex scattering phenomena. This scattering makes it difficult to determine the effective pathlength for the light, and therefore attempts to use the Beer-Lambert law, or even to determine a consistent empirical calibration, continue to be thwarted. Application of new techniques in time-resolved spectroscopy may be able to tackle this problem. Thinking of light as a packet of photons, if a single packet from a light source is sent through tissue, then a distant receiver will detected a photon distribution over time—the photons least scattered arriving first and the photons most scattered arriving later. In principle, the first photons arriving at the detector passed directly through the tissue. For these first photons the distance between the emitter and the detector is fixed and known, and the Beer-Lambert law should apply, permitting determination of an absolute concentration for an absorbing component. The difficulty in this is, first, that the measurement time scale must be on the order of the photon transit time (subnanosecond), and second, that the number of photons getting through without scattering will be extremely small, and therefore the detector must be exquisitely sensitive. Although these considerable technical problems have been overcome in the laboratory, their implementation in a practical instrument applied to a real subject remains to be demonstrated. This same approach is also being investigated for noninvasive optical imaging, since the unscattered photons should produce sharp images (see Chance et al., [1988], Chance [1991], and Yoo and Alfano [1989]).

86.3 Conclusions The remarkable success of pulse oximetry has established noninvasive optical monitoring of vital physiologic functions as a modality of considerable value. Hardware and algorithm advances in pulse oximetry are beginning to broaden its use outside the traditional operating room and critical care areas. Other promising applications of noninvasive optical monitoring are emerging, such as for measuring deep tissue oxygen levels, determining cellular metabolic status, or for quantitative determination of other important physiologic parameters such as blood glucose. Although these latter applications are not yet practical, they may ultimately impact noninvasive clinical monitoring just as dramatically as pulse oximetry.

Defining Terms Beer-Lambert law: Principle stating that the optical absorbance of a substance is proportional to both the concentration of the substance and the pathlength of the sample. Cytochromes: Heme-containing proteins found in the membranes of mitochondria and required for oxidative phosphorylation, with characteristic optical absorbance spectra. Dysfunctional hemoglobins: Those hemoglobin species that cannot reversibly bind oxygen (carboxyhemoglobin, methemoglobin, and sulfhemoglobin).


Functional saturation: The ratio of oxygenated hemoglobin to total nondysfunctional hemoglobins (oxyhemoglobin plus deoxyhemoglobin). Hypoxia: Inadequate oxygen supply to tissues necessary to maintain metabolic activity. Multivariate analysis: Empirical models developed to relate multiple spectral intensities from many calibration samples to known analyte concentrations, resulting in an optimal set of calibration parameters. Oximetry: The determination of blood or tissue oxygen content, generally by optical means. Pulse oximetry: The determination of functional oxygen saturation of pulsatile arterial blood by ratiometric measurement of tissue optical absorbance changes.

References Burns DA, Ciurczak EW (eds). 1992. Handbook of Near-Infrared Analysis. New York, Marcel Dekker. Chance B. 1991. Optical method. Annu Rev Biophys Biophys Chem 20:1. Chance B, Leigh JS, Miyake H, et al. 1988. Comparison of time-resolved and -unresolved measurements of deoxyhemoglobin in brain. Proc Natl Acad Sci USA 85(14):4971. Coté GL, Fox MD, Northrop RB. 1992. Noninvasive optical polarimetric glucose sensing using a true phase measurement technique. IEEE Trans Biomed Eng 39(7):752. Haaland DM. 1992. Multivariate calibration methods applied to the quantitative analysis of infrared spectra. In PC Jurs (ed), Computer-Enhanced Analytical Spectroscopy, vol 3, pp 1–30. New York, Plenum. Haaland DM, Robinson MR, Koepp GW, et al. 1992. Reagentless near-infrared determination of glucose in whole blood using multivariate calibration. Appl Spectros 46(10):1575. Jöbsis FF. 1977. Noninvasive, infrared monitoring of cerebral and myocardial oxygen sufficiency and circulatory parameters. Science 198(4323):1264. Jöbsis FF, Keizer LH, LaManna JC, et al. 1977. Reflectance spectrophotometry of cytochrome aa3 in vivo. J Appl Physiol 43(5):858. Kelleher JF. 1989. Pulse oximetry. J Clin Monit 5(1):37. Martens H, Næs T. 1989. Multivariate Calibration. New York, John Wiley. Osborne BG, Fearn T, Hindle PH. 1993. Practical NIR Spectroscopy with Applications in Food and Beverage Analysis. Essex, England, Longman Scientific & Technical. Payne JP, Severinghaus JW (eds). 1986. Pulse Oximetry. New York, Springer-Verlag. Rabinovitch B, March WF, Adams RL. 1982. Noninvasive glucose monitoring of the aqueous humor of the eye: Part I. Measurement of very small optical rotations. Diabetes Care 5(3):254. Robinson MR, Eaton RP, Haaland DM, et al. 1992. Noninvasive glucose monitoring in diabetic patients: A preliminary evaluation. Clin Chem 38(9):1618. Severinghaus JW, Astrup PB. 1986. History of blood gas analysis. VI. Oximetry. J Clin Monit 2(4):135. Severinghaus JW, Honda Y. 1987a. History of blood gas analysis. VII. Pulse oximetry. J Clin Monit 3(2):135. Severinghaus JW, Honda Y. 1987b. Pulse oximetry. Int Anesthesiol Clin 25(4):205. Severinghaus JW, Kelleher JF. 1992. Recent developments in pulse oximetry. Anesthesiology 76(6):1018. Tremper KK, Barker SJ. 1989. Pulse oximetry. Anesthesiology 70(1):98. Wukitsch MW, Petterson MT, Tobler DR, et al. 1988. Pulse oximetry: Analysis of theory, technology, and practice. J Clin Monit 4(4):290. Yoo KM, Alfano RR. 1989. Photon localization in a disordered multilayered system. Phys Rev B 39(9):5806.

Further Information Two collections of papers on pulse oximetry include a book edited by J. P. Payne and J. W. Severinghaus, Pulse Oximetry (New York, Springer-Verlag, 1986), and a journal collection—International Anesthesiology Clinics [25(4), 1987]. For technical reviews of pulse oximetry, see J. A. Pologe’s 1987 “Pulse Oximetry”


[Int Anesthesiol Clin 25(3):137], Kevin K. Tremper and Steven J. Barker’s 1989 “Pulse Oximetry” [Anesthesiology 70(1):98], and Michael W. Wukitsch, Michael T. Patterson, David R. Tobler, and coworkers’ 1988 “Pulse Oximetry: Analysis of Theory, Technology, and Practice” [J Clin Monit 4(4):290]. For a review of practical and clinical applications of pulse oximetry, see the excellent review by Joseph K. Kelleher [1989] and John Severinghaus and Joseph F. Kelleher [1992]. John Severinghaus and Yoshiyuki Honda have written several excellent histories of pulse oximetry [1987a, 1987b]. For an overview of applied near-infrared spectroscopy, see Donald A. Burns and Emil W. Ciurczak [1992] and B. G. Osborne, T. Fearn, and P. H. Hindle [1993]. For a good overview of multivariate methods, see Harald Martens and Tormod Næs [1989].


Bowman, B. R., Schuck, E. “Medical Instruments and Devices Used in the Home.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

87 Medical Instruments and Devices Used in the Home 87.1 87.2

The Device Must Provide a Positive Clinical Outcome • The Device Must be Safe to Use • The Device Must be Designed So That It Will be Used

Bruce R. Bowman EdenTec Corporation

Edward Schuck EdenTec Corporation

Scope of the Market for Home Medical Devices Unique Challenges to the Design and Implementation of High-Tech Homecare Devices

87.3 87.4

Infant Monitor Example Conclusions

87.1 Scope of the Market for Home Medical Devices The market for medical devices used in the home and alternative sites has increased dramatically in the last 10 years and has reached an overall estimated size of more than $1.6 billion [FIND/SVP, 1992]. In the past, hospitals have been thought of as the only places to treat sick patients. But with the major emphasis on reducing healthcare costs, increasing numbers of sicker patients move from hospitals to their homes. Treating sicker patients outside the hospital places additional challenges on medical device design and patient use. Equipment designed for hospital use can usually rely on trained clinical personnel to support the devices. Outside the hospital, the patient and/or family members must be able to use the equipment, requiring these devices to have a different set of design and safety features. This chapter will identify some of the major segments using medical devices in the home and discuss important design considerations associated with home use. Table 87.1 outlines market segments where devices and products are used to treat patients outside the hospital [FIND/SVP, 1992]. The durable medical equipment market is the most established market providing aids for patients to improve access and mobility. These devices are usually not life supporting or sustaining, but in many cases they can make the difference in allowing a patient to be able to function outside a hospital or nursing or skilled facility. Other market segments listed employ generally more sophisticated solutions to clinical problems. These will be discussed by category of use. The incontinence and ostomy area of products is one of the largest market segments and is growing in direct relationship to our aging society. Whereas sanitary pads and colostomy bags are not very “hightech,” well-designed aids can have a tremendous impact on the comfort and independence of these patients. Other solutions to incontinence are technically more sophisticated, such as use of electric stimulation of the sphincter muscles through an implanted device or a miniature stimulator inserted as an anal or vaginal plug to maintain continence [Wall et al., 1993].


TABLE 87.1

Major Market Segments Outside Hospitals Estimated Equipment Size 1991

Market Segment Durable medical equipment Incontinence and ostomy products Respiratory equipment

$373 M* $600 M* $180 M*

Drug infusion, drug measurement

$300 M

Pain control and functional stimulation

$140 M

Device Examples Specialty beds, wheelchairs, toilet aids, ambulatory aids Sanitary pads, electrical stimulators, colostomy bags Oxygen therapy, portable ventilators, nasal CPAP, monitors, apnea monitors Infusion pumps, access ports, patient-controlled analgesia (PCA), glucose measurement, implantable pumps Transcutaneous electrical nerve stimulation (TENS), functional electrical nerve stimulation (FES)

*Source: FIND/SVP [1992].

Many forms of equipment are included in the Respiratory segment. These devices include those that maintain life support as well as those that monitor patients’ respiratory function. These patients, with proper medical support, can function outside the hospital at a significant reduction in cost and increased patient comfort [Pierson, 1994]. One area of this segment, infant apnea monitors, provides parents or caregivers the cardio/respiratory status of an at-risk infant so that intervention (CPR etc.) can be initiated if the baby has a life-threatening event. The infant monitor shown in Fig. 87.1 is an example of a patient monitor designed for home use and will be discussed in more detail later in this chapter. Pulse oximetry monitors are also going home with patients. They are used to measure noninvasively the oxygen level of patients receiving supplemental oxygen or ventilator-dependent patients to determine if they are being properly ventilated. Portable infusion pumps are an integral part of providing antibiotics, pain management, chemotherapy, and parenteral and enteral nutrition. The pump shown in Fig. 87.2 is an example of technology that allows the patient to move about freely while receiving sometimes lengthy drug therapy. Implantable drug pumps are also available for special long-term therapy needs.

FIGURE 87.1

Infant apnea monitor used in a typical home setting (photo courtesy of EdenTec Corporation).


FIGURE 87.2

Portable drug pump used throughout the day (photo courtesy of Pharmacia Deltec Inc.).

Pain control using electric stimulation in place of drug therapy continues to be an increasing market. The delivery of small electric impulses to block pain is continuing to gain medical acceptance for treatment outside the hospital setting. A different form of electric stimulation called functional electric stimulation (FES) applies short pulses of electric current to the nerves that control weak or paralyzed muscles. This topic is covered as a separate chapter in this book. Growth of the homecare business has created problems in overall healthcare costs since a corresponding decrease in hospital utilization has not yet occurred. In the future, however, increased homecare will necessarily result in reassessment and downsizing in the corresponding hospital segment. There will be clear areas of growth and areas of consolidation in the new era of healthcare reform. It would appear, however, that homecare has a bright future of continued growth.

87.2 Unique Challenges to the Design and Implementation of High-Tech Homecare Devices What are some of the unique requirements of devices that could allow more sophisticated equipment to go home with ordinary people of varied educational levels without compromising their care? Even though each type of clinical problem has different requirements for the equipment that must go home with the patient, certain common qualities must be inherent in most devices used in the home. Three areas to consider when equipment is used outside of the hospital are that the device (1) must provide a positive clinical outcome, (2) must be safe and easy to use, and (3) must be user-friendly enough so that it will be used.

The Device Must Provide a Positive Clinical Outcome Devices cannot be developed any longer just because new technology becomes available. They must solve the problem for which they were intended and make a significant clinical difference in the outcome or management of the patient while saving money. These realities are being driven by those who reimburse for devices, as well as by the FDA as part of the submission for approval to market a new device.


The Device Must Be Safe to Use Homecare devices may need to be even more reliable and even safer than hospital devices. We often think of hospitals as having the best quality and most expensive devices that money can buy. In addition to having the best equipment to monitor patients, hospitals have nurses and aids that keep an eye on patients so that equipment problems may be quickly discovered by the staff. A failure in the home may go unnoticed until it is too late. Thus systems for home use really need extra reliability with automatic backup systems and/or early warning signals. Safety issues can take on a different significance depending on the intended use of the device. Certain safety issues are important regardless of whether the device is a critical device such as an implanted cardiac pacemaker or a noncritical device such as a bed-wetting alarm. No device should be able to cause harm to the patient regardless of how well or poorly it may be performing its intended clinical duties. Devices must be safe when exposed to all the typical environmental conditions to which the device could be exposed while being operated by the entire range of possible users of varied education and while exposed to siblings and other untrained friends or relatives. For instance, a bed-wetting alarm should not cause skin burns under the sensor if a glass of water spills on the control box. This type of safety issue must be addressed even when it significantly affects the final cost to the consumer. Other safety issues are not obviously differentiated as to being actual safety issues or simply nuisances or inconveniences to the user. It is very important for the designer to properly define these issues; although some safety features can be included with little or no extra cost, other safety features may be very costly to implement. It may be a nuisance for the patient using a TENS pain control stimulator to have the device inadvertently turned off when its on/off switch is bumped while watching TV. In this case, the patient only experiences a momentary cessation of pain control until the unit is turned back on. But it could mean injuries or death to the same patient driving an automobile who becomes startled when his TENS unit inadvertently turns on and he causes an accident. Reliability issues can also be mere inconveniences or major safety issues. Medical devices should be free of design and materials defects so that they can perform their intended functions reliably. Once again, reliability does not necessarily need to be expensive and often can be obtained with good design. Critical devices, i.e., devices that could cause death or serious injury if they stopped operating properly, may need to have redundant systems for backup, which likely will increase cost.

The Device Must be Designed So That It Will Be Used A great deal of money is being spent in healthcare on devices for patients that end up not being used. There are numerous reasons for this happening including that the wrong device was prescribed for the patient’s problem in the first place; the device works, but it has too many false alarms; the device often fails to operate properly; it is cumbersome to use or difficult to operate or too uncomfortable to wear. Ease of Use User-friendliness is one of the most important features in encouraging a device to be used. Technological sophistication may be just as necessary in areas that allow ease of use as in attaining accuracy and reliability in the device. The key is that the technologic sophistication be transparent to the user so that the device does not intimidate the user. Transparent features such as automatic calibration or automatic sensitivity adjustment may help allow successful use of a device that would otherwise be too complicated. Notions of what makes a device easy to use, however, need to be thoroughly tested with the patient population intended for the device. Caution needs to be taken in defining what “simple” means to different people. A VCR may be simple to the designer because all features can be programmed with one button, but it may not be simple to users if they have to remember that it takes two long pushes and one short to get into the clock-setting program. Convenience for the user is also extremely important in encouraging use of a device. Applications that require devices to be portable must certainly be light enough to be carried. Size is almost always important for anything that must fit within the average household. Either a device must be able to be left in place © 2000 by CRC Press LLC

in the home or it must be easy to set up, clean, and put away. Equipment design can make the difference between the patient appropriately using the equipment or deciding that it is just too much hassle to bother. Reliability Users must also have confidence in the reliability of the device being used and must have confidence that if it is not working properly, the device will tell them that something is wrong. Frequent breakdowns or false alarms will result in frustration and ultimately in reduced compliance. Eventually patients will stop using the device altogether. Most often, reliability can be designed into a product with little or no extra cost in manufacturing, and everything that can be done at no cost to enhance reliability should be done. It is very important, however, to understand what level of additional reliability involving extra cost is necessary for product acceptance. Reliability can always be added by duplicated backup systems, but the market or application may not warrant such an approach. Critical devices which are implanted, such as cardiac pacemakers, have much greater reliability requirements, since they involve not only patient frustration but also safety. Cost Reimbursement Devices must be paid for before the patient can realize the opportunity to use new, effective equipment. Devices are usually paid for by one of two means. First, they are covered on an American Medical Association Current Procedural Terminology Code (CPT-code) which covers the medical, surgical, and diagnostic services provided by physicians. The CPT-codes are usually priced out by Medicare to establish a baseline reimbursement level. Private carriers usually establish a similar or different level of reimbursement based on regional or other considerations. Gaining new CPT-codes for new devices can take a great deal of time and effort. The second method is to cover the procedure and device under a capitated fee where the hospital is reimbursed a lump sum for a procedure including the device, hospital, homecare, and physician fees. Every effort should be made to design devices to be low cost. Device cost is being scrutinized more and more by those who reimburse. It is easy to state, however, that a device needs to be inexpensive. Unfortunately the reality is that healthcare reforms and new regulations by the FDA are making medical devices more costly to develop, to obtain regulatory approvals for [FDA, 1993], and to manufacture. Professional Medical Service Support The more technically sophisticated a device is, the more crucial that homecare support and education be a part of a program. In fact, in many cases, such support and education are as important as the device itself. Medical service can be offered by numerous homecare service companies. Typically these companies purchase the equipment instead of the patient, and a monthly fee is charged for use of the equipment along with all the necessary service. The homecare company then must obtain reimbursement from third-party payers. Some of the services offered by the homecare company include training on how to use the equipment, CPR training, transporting the equipment to the home, servicing/repairing equipment, monthly visits, and providing on-call service 24 hours a day. The homecare provider must also be able to provide feedback to the treating physician on progress of the treatment. This feedback may include how well the equipment is working, the patient’s medical status, and compliance of the patient.

87.3 Infant Monitor Example Many infants are being monitored in the home using apnea monitors because they have been identified with breathing problems [Kelly, 1992]. These include newborn premature babies who have apnea of prematurity [Henderson-Smart, 1992; NIH, 1987], siblings of babies who have died of sudden infant death syndrome (SIDS) [Hunt, 1992; NIH, 1987], or infants who have had an apparent life-threatening episode (ALTE) related to lack of adequate respiration [Kahn et al., 1992; NIH, 1987]. Rather than keeping infants in the hospital for a problem that they may soon outgrow (1–6 months), doctors often discharge © 2000 by CRC Press LLC

them from the hospital with an infant apnea monitor that measures the duration of breathing pauses and heart rate and sounds an alarm if either parameter crosses limits prescribed by the doctor. Infant apnea monitors are among the most sophisticated devices used routinely in the home. These devices utilize microprocessor control, sophisticated breath-direction and artifact rejection firmware algorithms, and internal memory that keeps track of use of the device as well as recording occurrence of events and the physiologic waveforms associated with the events. The memory contents can be downloaded directly to computer or sent via modem remotely where a complete 45-day report can be provided to the referring physician (see Fig. 87.3). Most apnea monitors measure breathing effort through impedance pneumography. A small (100–200 uA) high-frequency (25–100 kHz) constant-current train of pulses is applied across the chest between a pair of electrodes. The voltage needed to drive the current is measured, and thereby the effective impedance between the electrodes can be calculated. Impedance across the chest increases as the chest expands and decreases as the chest contracts with each breath. The impedance change with each breath can be as low as 0.2 ohms on top of an electrode base impedance of 2000 ohms, creating some interesting signal-to-noise challenges. Furthermore, motion artifact and blood volume changes in the heart and chest can cause impedance changes of 0.6 ohms or more that can look just like breathing. Through the same pair of electrodes, heart rate is monitored by picking up the electrocardiogram (ECG) [AAMI, 1988]. Because the impedance technique basically measures the motion of the chest, this technique can only be used to monitor central apnea or lack of breathing effort. Another less common apnea in infants called obstructive apnea results when an obstruction of the airway blocks air from flowing in spite of breathing effort. Obstructive apnea can not be monitored using impedance pneumography [Kelly, 1992]. There is a very broad socioeconomic and educational spectrum of parents or caregivers who may be monitoring their infants with an apnea monitor. This creates an incredible challenge for the design of the device so that it is easy enough to be used by a variety of caregivers. It also puts special requirements on the homecare service company that must be able to respond to these patients within a matter of minutes, 24 hours a day. The user-friendly monitor shown in Fig. 87.1 uses a two-button operation, the on/off switch, and a reset switch. The visual alarm indicators are invisible behind a back-lit panel except when an actual alarm occurs. A word describing the alarm then appears. By not showing all nine possible alarm conditions unless an alarm occurs, parent confusion and anxiety is minimized. Numerous safety features are built into the unit, some of which are noticeable but many of which are internal to the operation of the monitor. One useful safety feature is the self-check. When the device is turned on, each alarm LED lights in sequence, and the unit beeps once indicating that the self-check was completed successfully. This gives users the opportunity to confirm that all the alarm visual indicators and the audible indicator are working and provides added confidence for users leaving their baby on the monitor. A dual-level battery alarm gives an early warning that the battery will soon need charging. The weak battery alarm allows users to reset the monitor and continue monitoring their babies for several more hours before depleting the battery to the charge battery level where the monitor must be attached to the ac battery charger/adapter. This allows parents the freedom to leave their homes for a few hours knowing that their child can continue to be monitored. A multistage alarm reduces the risk of parents sleeping through an alarm. Most parents are sleepdeprived with a new baby. Consequently, it can be easy for parents in a nearby room to sleep through a monitor alarm even when the monitor sounds at 85 dB. A three-stage alarm helps to reduce this risk. After 10 seconds of sounding at 1 beep per second, the alarm switches to 3 beeps per second for the next 10 seconds. Finally, if an alarm has not resolved itself after 20 seconds, the alarm switches to 6 beeps per second. Each stage of alarm sounds more intense than the previous one and offers the chance of jolting parents out of even the deepest sleep. The physician always prescribes what alarm settings should be used by the homecare service company when setting up the monitor. As a newborn matures, these settings may need to be adjusted. Sometimes the parents can be relied upon for making these setting changes. To allow both accessibility to these switches as well as to keep them safe from unauthorized tampering from a helping brother or sister, a © 2000 by CRC Press LLC

FIGURE 87.3 Infant apnea monitor with memory allows data to be sent by modem to generate physician report (drawing courtesy of EdenTec Corporation).

special tamper-resistant-adjustment procedure is utilized. Two simultaneous actions are required in order to adjust the alarm limit settings. The reset button must be continually pressed on the front of the unit while changing settings on the back of the unit. Heart rate levels are set in beats per minute, and apnea © 2000 by CRC Press LLC

duration is set in single-second increments. Rather than using easy-to-set push-button switches, “penset” switches are used which require a pen or other sharp implement to make the change. If the proper switch adjustment procedure is not followed, the monitor alarms continuously and displays a switch alarm until the settings are returned to their original settings. A similar technique is used for turning the monitor off. The reset button must first be pressed and then the on/off switch turned to the off position. Violation of this procedure will result in a switch alarm. Other safety features are internal to the monitor and are transparent to the user. The monitor’s alarm is designed to be normally on from the moment the device is turned on. Active circuitry controlled by the microprocessor turns the alarm off when there are no active alarm conditions. If anything hangs up the processor or if any of a number of components fail, the alarm will not turn off and will remain on in a fail-safe mode. This “alarm on unless turned off ” technique is also used in a remote alarm unit for parents with their baby in a distant room. If a wire breakage occurs between the monitor and the remote alarm unit, or a connector pulls loose, or a component fails, the remote alarm no longer is turned off by the monitor and it alarms in a fail-safe condition. Switches, connectors, and wires are prone to fail. One way to circumvent this potential safety issue is use of switches with a separate line for each possible setting. The monitor continuously polls every switch line of each switch element to check that “exactly” one switch position is making contact. This guards against misreading bad switch elements, a switch inadvertently being set between two positions, or a bad connector or cable. Violation of the “exactly one contact condition” results in a switch alarm. It is difficult to manage an apnea monitoring program in rural areas where the monitoring family may be a hundred miles or more away from the homecare service company. There are numerous ways to become frustrated with the equipment and stop using the monitor. Therefore, simplicity of use and reliability are important. Storing occurrence of alarms and documenting compliance in internal memory in the monitor help the homecare service company and the remote family cope with the situation. The monitor shown in Fig. 87.1 stores in digital memory the time, date, and duration of (1) each use of the monitor; (2) occurrence of all equipment alarms; and (3) all physiologic alarms including respiratory waveforms, heart rate, and ECG for up to a 45-day period. These data in the form of a report (see Fig. 87.3) can be downloaded to a laptop PC or sent via modem to the homecare service company or directly to the physician.

87.4 Conclusions Devices that can provide positive patient outcomes with reduced overall cost to the healthcare system while being safe, reliable, and user-friendly will succeed based on pending healthcare changes. Future technology in the areas of sensors, communications, and memory capabilities should continue to increase the potential effectiveness of homecare management programs by using increasingly sophisticated devices. The challenge for the medical device designer is to provide cost-effective, reliable, and easy-to-use solutions that can be readily adopted by the multidisciplinary aspects of homecare medicine while meeting FDA requirements.

Defining Terms Apnea: Cessation of breathing. Apnea can be classified as central, obstructive, or mixed, which is a combination. Apnea of prematurity: Apnea in which the incidence and severity increases with decreasing gestational age attributable to immaturity of the respiratory control system. The incidence has increased due to improved survival rates for very-low-birth-weight premature infants. Apparent life-threatening episode (ALTE): An episode characterized by a combination of apnea, color change, muscle tone change, choking, or gagging. To the observer it may appear the infant has died. Capitated fee: A fixed payment for total program services versus the more traditional fee for service in which each individual service is charged. © 2000 by CRC Press LLC

Cardiac pacemaker: A device that electrically stimulates the heart at a certain rate used in absence of normal function of the heart’s sino-atrial node. Central apnea: Apnea secondary to lack of respiratory or diaphragmatic effort. Chemotherapy: Treatment of disease by chemical agents. Term popularly used when fighting cancer chemically. Colostomy: The creation of a surgical hole as an alternative opening of the colon. CPR (cardiopulmonary resuscitation): Artificially replacing heart and respiration function through rhythmic pressure on the chest. CPT-code (current procedural terminology code): A code used to describe specific procedures/tests developed by the AMA. Electrocardiogram (ECG): The electric potential recorded across the chest due to depolarization of the heart muscle with each heartbeat. Enteral nutrition: Chemical nutrition injected intestinally. Food and Drug Administration (FDA): Federal agency that oversees and regulates foods, drugs, and medical devices. Functional electrical stimulation (FES): Electric stimulation of peripheral nerves or muscles to gain functional, purposeful control over partially or fully paralyzed muscles. Incontinence: Loss of voluntary control of the bowel or bladder. Obstructive apnea: Apnea in which the effort to breath continues but airflow ceases due to obstruction or collapse of the airway. Ostomy: Surgical procedure that alters the bladder or bowel to eliminate through an artificial passage. Parenteral nutrition: Chemical nutrition injected subcutaneously, intramuscular, intrasternally, or intravenously. Sphincter: A band of muscle fibers that constricts or closes an orifice. Sudden infant death syndrome (SIDS): The sudden death of an infant which is unexplained by history or postmortem exam. Transcutaneous electrical nerve stimulation (TENS): Electrical stimulation of sensory nerve fibers resulting in control of pain.

References AAMI. 1988. Association for the Advancement of Medical Instrumentation Technical Information Report. Apnea Monitoring by Means of Thoracic Impedance Pneumography, Arlington, Virg. FDA. November 1993. Reviewers Guidance for Premarket Notification Submissions (Draft), Anesthesiology and Respiratory Device Branch, Division of Cardiovascular, Respiratory, and Neurological Devices. Food and Drug Administration. Washington, DC. FIND/SVP. 1992. The Market for Home Care Products, a Market Intelligence Report. New York. Henderson-Smart DJ. 1992. Apnea of prematurity. In R Beckerman, R Brouillette, C Hunt (eds), Respiratory Control Disorders in Infants and Children, pp 161–177, Baltimore, Williams and Wilkins. Hunt CE. 1992. Sudden infant death syndrome. In R Beckerman, R Brouillette, C Hunt (eds), Respiratory Control Disorders in Infants and Children, pp 190–211, Baltimore, Williams and Wilkins. Kahn A, Rebuffat E, Franco P, et al. 1992. Apparent life-threatening events and apnea of infancy. In R Beckerman, R Brouillette, C Hunt (eds), Respiratory Control Disorders in Infants and Children, pp 178–189, Baltimore, Williams and Wilkins. Kelly DH. 1992. Home monitoring. In R Beckerman, R Brouillette, C Hunt (eds), Respiratory Control Disorders in Infants and Children, pp 400–412, Baltimore, Williams and Wilkins. NIH. 1987. Infantile Apnea and Home Monitoring Report of NIH Consensus Development Conference, US Department of Health and Human Services, NIH publication 87-2905. Pierson DJ. 1994. Controversies in home respiratory care: Conference summary. Respir Care 39(4):294. Wall LL, Norton PA, Dehancey JOL. 1993. Practical Urology, Baltimore, Williams and Wilkins.


Rosow, E., Adam . J. “Virtual Instrumentation: Applications in Biomedical Engineering.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

88 Virtual Instrumentation: Applications in Biomedical Engineering Eric Rosow

88.1

A Revolution—Graphical Programming and Virtual Instrumentation

Hartford Hospital and Premise Development Corporation

Joseph Adam Premise Development Corporation

Overview

88.2

Virtual Instrumentation and Biomedical Engineering Example 1 • Example 2

88.1 Overview A Revolution—Graphical Programming and Virtual Instrumentation Over the last decade, the graphical programming revolution has empowered engineers to develop customized systems, the same way the spreadsheet has empowered business managers to analyze financial data. This software technology has resulted in another type of revolution—the virtual instrumentation revolution, which is rapidly changing the instrumentation industry by driving down costs without sacrificing quality. Virtual Instrumentation can be defined as: A layer of software and/or hardware added to a general-purpose computer in such a fashion that users can interact with the computer as though it were their own custom-designed traditional electronic instrument. Today, computers can serve as the engine for instrumentation. Virtual instruments utilize the open architecture of industry-standard computers to provide the processing, memory, and display capabilities; while the off-the-shelf, inexpensive interface boards plugged into an open bus, standardized communications bus provides the vehicle for the instrument’s capabilities. As a result, the open architecture of PCs and workstations allow the functionality of virtual instruments to be user defined. In addition, the processing power of virtual instruments is much greater than stand-alone instruments. This advantage will continue to accelerate due to the rapid technology evolution of PCs and workstations that results from the huge investments made in this industry. The major benefits of virtual instrumentation include increased performance and reduced costs. In addition, because the user controls the technology through software, the flexibility of virtual instrumentation


is unmatched by traditional instrumentation. The modular, hierarchical programming environment of virtual instrumentation is inherently reusable and reconfigurable.

88.2 Virtual Instrumentation and Biomedical Engineering Virtual Instrumentation applications have encompassed nearly every industry including the telecommunications, automotive, semiconductor, and biomedical industries. In the fields of healthcare and biomedical engineering, virtual instrumentation has empowered developers and end-users to conceive of, develop, and implement a wide variety of research-based biomedical applications and executive information tools. These applications fall into several categories including: clinical research, equipment testing and quality assurance, data management, and performance improvement. In a collaborative approach, physicians, researchers, and biomedical and software engineers at Hartford Hospital (Hartford, CT) and Premise Development Corporation (Avon, CT) have developed various data acquisition and analysis systems that successfully integrate virtual instrumentation principles in a wide variety of environments. These include: • • • • •

“The EndoTester™”, a patented quality assurance system for fiberoptic endoscopes a Non-Invasive Pulmonary Diffusion and Cardiac Output Measurement System a Cardiovascular Pressure-Dimension Analysis System “BioBench™”, a powerful turnkey application for physiological data acquisition and analysis “PIVIT™”, a Performance Indicator Virtual Instrument Toolkit to manage and forecast financial data • a “Virtual Intelligence Program” to manage the discrete components within the continuum of care • “BabySave™”, an analysis and feedback system for apnea interruption via vibratory stimulation This chapter will describe several of these applications and describe how they have allowed clinicians and researchers to gain new insights, discover relationships that may not have been obvious, and test and model hypotheses based on acquired data sets. In some cases, these applications have been developed into commercial products to address test and measurement needs at other healthcare facilities throughout the world.

Example 1: BioBench™—A Virtual Instrument Application for Data Acquisition and Analysis of Physiological Signals The biomedical industry is an industry that relies heavily on the ability to acquire, analyze, and display large quantities of data. Whether researching disease mechanisms and treatments by monitoring and storing physiological signals, researching the effects of various drugs interactions, or teaching students in labs where students study physiological signs and symptoms, it was clear that there existed a strong demand for a flexible, easy-to-use, and cost-effective tool. In a collaborative approach, biomedical engineers, software engineers and clinicians, and researchers created a suite of virtual instruments called BioBench™. BioBench™ (National Instruments, Austin, TX) is a new software application designed for physiological data acquisition and analysis. It was built with LabVIEW™, the world’s leading software development environment for data acquisition, analysis, and presentation.1 Coupled with National Instruments data acquisition (DAQ) boards, BioBench integrates the PC with data acquisition for the life sciences market. Many biologists and physiologists have made major investments over time in data acquisition hardware built before the advent of modern PCs. While these scientists cannot afford to throw out their investment in this equipment, they recognize that computers and the concept of virtual instrumentation yield tremendous benefits in terms of data analysis, storage, and presentation. In many cases, traditional 1BioBench™ was developed for National Instruments (Austin, TX) by Premise Development Corporation (Avon, CT).


FIGURE 88.1

A typical biomedical application using BioBench (courtesy of National Instruments).

medical instrumentation may be too expensive to acquire and/or maintain. As a result, researchers and scientists are opting to create their own PC-based data monitoring systems in the form of virtual instruments. Other life scientists, who are just beginning to assemble laboratory equipment, face the daunting task of selecting hardware and software needed for their application. Many manufacturers for the life sciences field focus their efforts on the acquisition of raw signals and converting these signals into measurable linear voltages. They do not concentrate on digitizing signals or the analysis and display of data on the PC. BioBench™ is a low-cost turnkey package that requires no programming. BioBench is compatible with any isolation amplifier or monitoring instrument that provides an analog output signal. The user can acquire and analyze data immediately because BioBench automatically recognizes and controls the National Instruments DAQ hardware, minimizing configuration headaches. Some of the advantages of PC-Based Data Monitoring include: • • • • • •

Easy-to-use-software applications Large memory and the PCI bus Powerful processing capabilities Simplified customization and development More data storage and faster data transfer More efficient data analysis

Figure 88.1 illustrates a typical setup of a data acquisition experiment using BioBench. BioBench also features pull-down menus through which the user can configure devices. Therefore, those who have made large capital investments can easily migrate their existing equipment into the computer age. Integrating a combination of old and new physiological instruments from a variety of manufacturers is an important and straightforward procedure. In fact, within the clinical and research setting, it is a common requirement to be able to acquire multiple physiological signals from a variety of medical devices and instruments which do not necessarily communicate with each other. Often times, this situation is compounded by the fact that end-users would like to be able to view and analyze an entire waveform and not just an average value. In order to accomplish this, the end-user must acquire multiple channels of data at a relatively high sampling rate and have the ability to manage many large data files. BioBench can collect up to 16 channels simultaneously at a sampling rate of 1000 Hz per channel. Files are stored in an efficient binary format which significantly reduces the amount of hard disk and memory requirements of the PC. During data acquisition, a number of features are available to the end-user. These features include: © 2000 by CRC Press LLC

Data Logging: Logging can be enabled prior to or during an acquisition. The application will either prompt the user for a descriptive filename or it can be configured to automatically assign a filename for each acquisition. Turning the data logging option on and off creates a log data event record that can be inspected in any of the analysis views of BioBench. Event Logging: The capacity to associate and recognize user commands associated with a data file may be of significant value. BioBench has been designed to provide this capability by automatically logging user-defined events, stimulus events, and file logging events. With user-defined events, the user can easily enter and associate date and time-stamped notes with user actions or specific subsets of data. Stimulus events are also data and time-stamped and provide the user information about whether a stimulus has been turned on or off. File logging events note when data has been logged to disk. All of these types of events are stored with the raw data when logging data to file and they can be searched for when analyzing data. Alarming: To alert the user about specific data values and thresholds, BioBench incorporates userdefined alarms for each signal which is displayed. Alarms appear on the user interface during data acquisition and notify the user that an alarm condition has occurred. Figure 88.2 is an example of the Data Acquisition mode of BioBench. Once data has been acquired, BioBench can employ a wide array of easy-to-use analysis features. The user has the choice of importing recently acquired data or opening a data file that had been previously acquired for comparison or teaching purposes. Once a data set has been selected and opened, BioBench allows the user to simply select and highlight a region of interest and choose the analysis options to perform a specific routine. BioBench implements a wide array of scalar and array analyses. For example, scalar analysis tools will determine the minimum, maximum, mean, integral, and slope of a selected data set, while the array analysis tools can employ Fast Fourier Transforms (FFTs), peak detection, histograms, and X vs. Y plots. The ability to compare multiple data files is very important in analysis and BioBench allows the user to open an unlimited number of data files for simultaneous comparison and analysis. All data files can be scanned using BioBench’s search tools in which the user can search for particular events that are

FIGURE 88.2


BioBench acquisition mode with alarms enabled.

FIGURE 88.3

BioBench analysis mode.

associated with areas of interest. In addition, BioBench allows the user to employ filters and transformations to their data sets and all logged data can be easily exported to a spreadsheet or database for further analysis. Finally, any signal acquired with BioBench can be played back, thus taking lab experience into the classroom. Figure 88.3 illustrates the analysis features of BioBench.

Example 2: A Cardiovascular Pressure-Dimension Analysis System Introduction The intrinsic contractility of the heart muscle (myocardium) is the single most important determinant of prognosis in virtually all diseases affecting the heart (e.g., coronary artery disease, valvular heart disease, and cardiomyopathy). Furthermore, it is clinically important to be able to evaluate and track myocardial function in other situations, including chemotherapy (where cardiac dysfunction may be a side effect of treatment) and liver disease (where cardiac dysfunction may complicate the disease). The most commonly used measure of cardiac performance is the ejection fraction. Although it does provide some measure of intrinsic myocardial performance, it is also heavily influenced by other factors such as heart rate and loading conditions (i.e., the amount of blood returning to the heart and the pressure against which the heart ejects blood). Better indices of myocardial function based on the relationship between pressure and volume throughout the cardiac cycle (pressure–volume loops) exist. However, these methods have been limited because they require the ability to track ventricular volume continuously during rapidly changing loading conditions. While there are many techniques to measure volume under steady state situations, or at enddiastole and end-systole (the basis of ejection fraction determinations), few have the potential to record volume during changing loading conditions. Echocardiography can provide online images of the heart with high temporal resolution (typically 30 frames per second). Since echocardiography is radiation-free and has no identifiable toxicity, it is ideally suited to pressure–volume analyses. Until recently however, its use for this purpose has been limited by the need for manual tracing of the endocardial borders, an extremely tedious and time-consuming endeavor. © 2000 by CRC Press LLC

The System Biomedical and software engineers at Premise Development Corporation (Avon, CT), in collaboration with physicians and researchers at Hartford Hospital, have developed a sophisticated research application called the “Cardiovascular Pressure-Dimension Analysis (CPDA) System”. The CPDA system acquires echocardiographic volume and area information from the acoustic quantification (AQ) port, in conjunction with vetricular pressure(s) and ECG signals to rapidly perform pressure–volume and pressure-area analyses. This fully automated system allows cardiologists and researchers to perform online pressuredimension and stroke work analyses during routine cardiac catheterizations and open-heart surgery. The system has been designed to work with standard computer hardware. Analog signals for ECG, pressure, and area/volume (AQ) are connected to a standard BNC terminal board. Automated calibration routines ensure that each signal is properly scaled and allows the user to immediately collect and analyze pressuredimension relationships. The CPDA can acquire up to 16 channels of data simultaneously. Typically, only three physiological parameters, ECG, pressure, and the AQ signals are collected using standard data acquisition hardware. In addition, the software is capable of running on multiple operating systems including Macintosh, Windows 95/98/NT, and Solaris. The CPDA also takes advantage of the latest hardware developments and form-factors and can be used with either a desktop or a laptop computer. The development of an automated, online method of tracing endocardial borders (Hewlett Packard’s AQ Technology) (Hewlett-Packard Medical Products Group, Andover, MA) has provided a method for rapid online area and volume determinations. Figure 88.4 illustrates this AQ signal from a Hewlett Packard Sonos Ultrasound Machine. This signal is available as an analog voltage (–1 to +1 volts) through the Sonos Dataport option (BNC connector). Data Acquisition and Analysis Upon launching this application, the user is presented with a dialog box that reviews the license agreement and limited warranty. Next, the Main Menu is displayed, allowing the user to select from one of six options as shown in Fig. 88.5. Clinical Significance Several important relationships can be derived from this system. Specifically, a parameter called the EndSystolic Pressure-Volume Relationship (ESPVR) describes the line of best fit through the peak-ratio (maximum pressure with respect to minimum volume) coordinates from a series of pressure–volume loops


The Acoustic Quantification (AQ) signal (Hewlett Packard).

FIGURE 88.5

Cardiovascular pressure-dimension analysis main menu.

FIGURE 88.6

The data selection front panel.

generated under varying loading conditions. The slope of this line has been shown to be a sensitive index of myocardial contractility that is independent of loading conditions. In addition, several other analyses, including time varying elastance (Emax) and stroke work, are calculated. Time-varying elastance is measured by determining the maximum slope of a regression line through a series of isochronic pressure–volume coordinates. Stroke work is calculated by quantifying the area of each pressure–volume loop. Statistical parameters are also calculated and displayed for each set of data. Figure 88.7 illustrates the pressuredimension loops and each of the calculated parameters along with the various analysis options. Finally, the user has the ability to export data sets into spreadsheet and database files and export graphs and indicators into third-party presentation software packages such as Microsoft PowerPoint®. © 2000 by CRC Press LLC

FIGURE 88.7

The cardiac cycle analysis front panel.

Summary Virtual Instrumentation allows the development and implementation of innovative and cost-effective biomedical applications and information management solutions. As the healthcare industry continues to respond to the growing trends of managed care and capitation, it is imperative for clinically useful, cost-effective technologies to be developed and utilized. As application needs will surely continue to change, virtual instrumentation systems will continue to offer users flexible and powerful solutions without requiring new equipment or traditional instruments.

References 1. Fisher, J.P., Mikan, J.S., Rosow, E., Nagle, J., Fram, D.B., Maffucci, L.M., McKay, R.G., and Gillam, L.D., “Pressure-Dimension Analysis of Regional Left Ventricular Performance Using Echocardiographic Automatic Boundary Detection: Validation in an Animal Model of Inotropic Modulation,” J. Am. College of Cardiol., 19(3), 262A, 1992. 2. Fisher, J.P., McKay, R.G., Mikan, J.S., Rosow, E., Nagle, J., Mitchel, J.F., Kiernan, F.J., Hirst, J.A., Primiano, C.A., Fram, D.B., Gillam, L.D., Hartford Hospital and University of Connecticut, Hartford, CT, “Human Left Ventricular Pressure-Area and Pressure-Volume Analysis Using Echocardiographic Automatic Boundary Detection.” 65th Scientific Session of the American Heart Association (11/92). 3. Fisher, J.P., Mitchel, J.F., Rosow, E., Mikan, J.S., Nagle, J., Kiernan, F.J., Hirst, J.A., Primiano, Gillam, L.D., Hartford Hospital and University of Connecticut, Hartford, CT, “Evaluation of Left Ventricular Diastolic Pressure-Area Relations with Echocardiographic Automatic Boundary Detection,” 65th Scientific Session of the American Heart Association (11/92). 4. Fisher, J.P., McKay, R.G., Mikan, J.S., Rosow, E., Nagle, J., Mitchel, J.F., Fram, D.B., Gillam, L.D., Hartford Hospital, “A Comparison of Echocardiographic Methods of Evaluating Regional LV Systolic Function: Fractional Area Change vs. the End-Systolic Pressure-Area Relation.” 65th Scientific Session of the American Heart Association (11/92). © 2000 by CRC Press LLC

5. Fisher, J.P., McKay, R.G., Rosow, E. Mikan, J. Nagle, J. Hirst, J.A., Fram, D.B., Gillam, L.D., “OnLine Derivation of Human Left Ventricular Pressure-Volume Loops and Load Independent Indices of Contractility Using Echocardiography with Automatic Boundary Detection-A Clinical Reality.” Circulation 88, I-304, 1993. 6. Fisher, J.P., Chen, C., Krupowies, N., Li Li, Kiernan, F.J., Fram, D.B., Rosow, E., Adam, J., Gillam, L.D., “Comparison of Mechanical and Pharmacologic Methods of Altering Loading Conditions to Determine End-Systolic Indices of Left Ventricle Function.” Circulation 90 (II), 1-494, 1994. 7. Fisher, JP, Martin, J., Day FP, Rosow, E., Adam J, Chen C, Gillam LD. “Validation of a Less Invasive Method for Determining Preload Recruitable Stroke Work Derived with Echocardiographic Automatic Boundary Detection.” Circulation 92, 1-278, 1995. 8. Fontes ML. Adam J, Rosow E, Mathew J, DeGraff AC. “Non-Invasive Cardiopulmonary Function Assessment System,” J Clin Monit 13, 413, 1997. 9. Johnson, GW, LabVIEW Graphical Programming: Practical Applications in Instrumentation and Control, 2nd ed., McGraw-Hill, New York, 1997. 10. Mathew JP, Adam J, Rosow E, Fontes ML, Davis L, Barash PG, Gillam L., “Cardiovascular Pressure-Dimension Analysis System,” J Clin Monit 13, 423, 1997. 11. National Instruments 1999 Measurement and Automation Catalog, National Instruments, Austin, TX. 12. Rosow, E. “Technology and Sensors”, Presented at the United States Olympic Committee’s Sports Equipment and Technology Conference, Colorado Springs, CO; November 19-23, 1992. 13. Rosow, E. “Biomedical Applications using LabVIEW”, Presented at the New England Society of Clinical Engineering, Sturbridge, MA; November 14, 1993. 14. Rosow, E., Adam, J.S., Nagle, J., Fisher, J.P. and Gillam, L.D.; Premise Development Corporation, Avon, CT and Hartford Hospital, Hartford, CT, “A Cardiac Analysis System: LabVIEW and Acoustic Quantification (AQ)”, Presented at the Association for the Advancement of Medical Instrumentation in Washington, D.C. May 21-25, 1994.


Geddes, L. A. “Recording of Action Potentials.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

Historical Perspectives 3:

Recording of Action Potentials

Leslie A. Geddes Purdue University

Nerve Action Potential Action Potentials in Afferent Fibers Code of the Nervous System True Form of the Nerve Action Potential

Nerve Action Potential The question for the form and nature of the nerve action potential used instruments that we would now call primitive. Yet, in skilled hands, these instruments led directly to discovery of the code used by the nervous system to transmit information. In fact, the code was discovered before the true form of the nerve action potential was known. Using a slow-speed galvanometer ballistically to measure time [Hoff & Geddes, 1960], Helmholtz [1850, 1851, 1853] measured the velocity of the frog sciatic nerve action potential and determined it to be 30 m/s. This was far below the speed of electricity and caused much controversy among physiologists of the day. Using the rheotome, a slowly responding galvanometer, and an induction coil stimulator [Geddes et al., 1989], Bernstein [1868] reconstructed the action potential of the frog sciatic nerve [Hoff & Geddes, 1957]. The sampling time was 0.3 ms, and the action potential that he obtained is shown in Fig. HP3.1. Not only did Bernstein chart the time course of this 0.5626- to 0.8041-ms action potential, he measured its propagation velocity, obtaining an average of 28.718 m/s, agreeing with the speed obtained by Helmholtz. Thus, with primitive electromechanical instruments, the time course and velocity of the nerve action potential were determined accurately. When the capillary electrometer appeared in 1876, it was thought that it might be possible to use it to record the nerve action potential. Many attempts were made with variable results because the concept of response time was not fully appreciated. Gotch and Horsley [1888] in the United Kingdom stimulated peripheral nerves in the cat using induction-coil shocks, and with one recording electrode over intact nerve and one in an area of injury, they recorded the action potential. The type of record that they obtained is shown in Fig. HP3.2a. Note that the response (action potential) was in the same direction for the break (b) and make (m) shocks. Recall that the break shock from an induction coil is stronger than the make shock. They knew that the make and break shocks were of opposite polarity and proved it by recording them, as shown in Fig. HP3.2b. After a long discussion about the nerve response being always in the same direction, irrespective of the polarity of the stimulus, they concluded that they had recorded single action potentials in response to single-induction-coil stimuli. They wrote: There is thus no doubt that the movement [of the mercury contour] that we obtained and photographed was due to the electromotive change which was due to the electromotive change which accompanies the propagation of an excitatory state along the mammalian nerve when this state is evoked by the application of a single stimulus.


Like Gotch and Horsley [1888], others who attempted recording nerve action potentials with the capillary electrometer were rewarded with a small-amplitude record. When Einthoven’s string galvanometer appeared in 1903, it was also used to record nerve action potentials with the same result [Einthoven, 1903]. In 1907 and 1908, De Forest patented the triode (audion), and the vacuumtube amplifier could be constructed. Therefore it appeared logical to use a triode amplifier to enlarge nerve action potentials and display them with the string galvanometer and capillary electrometer. The first to put the vacuum-tube amplifier to work in electrophysiology were Forbes and Thacher [1920], who coupled a triode to a string galvanometer and recorded frog nerve and human muscle action potentials. Their paper is essentially a tutorial that FIGURE HP3.1 Bernstein’s recondescribes three ways of coupling the triode to the string galvanom- struction of the action potential of the eter so that the delicate string would not be damaged. The first frog sciatic nerve. From his experimethod placed the vacuum tube in one arm of a Wheatstone bridge. ments, the mean duration was 0.6833 The string galvanometer was in the detector position, and a resistor ms. This reconstruction was made constituted the arm adjacent to the triode. The second method with a slow-speed galvanometer and employed transformer coupling, and the third method employed the rheotome. (From Bernstein, capacitor (C) coupling, as shown in Fig. HP3.3a. Of the three meth- 1868.) ods, the capacitive-coupling method was preferred. Figure HP3.3b is a record obtained by Forbes and Thacher showing the amplified and directly recorded frog sciatic nerve action potential. The timing signal (bottom) is 100 Hz. Desirous of displaying repetitive nerve action potentials with the string galvanometer, Gasser and Newcomer [1921] employed a two-stage resistance-capacity-coupled amplifier connected to the string galvanometer. They chose the phrenic nerve as the object of their study because of its spontaneous periodic activity in causing the diaphragm to contract tetanically and produce inspiration. The frequency of the phrenic nerve action potentials ranged from 71 to 105 per second, noting that their amplitude appeared to be largest at the peak of inspiration. They also reported a one-to-one correspondence between phrenic nerve and diaphragm action potentials. Their important observations were later to become recognized as the two ways that intensity is signaled in the nervous system. The addition of amplification to the string galvanometer did not improve its ability to respond to rapidly rising, short-duration action potentials. It required the cathode-ray tube and a multistage amplifier to solve this problem. Gasser and Erlanger [1922] were the first to show the form of the nerve action potential recorded extracellularly using the cathoderay tube and triode amplifier that they had previously developed for use with the string galvanometer. Not only did they achieve their goal, but they made the fundamental discovery that nerve propagation velocity was proportional to nerve fiber diameter. Although rapidly responding, the cathode-ray FIGURE HP3.2 The first nerve action potentials in response to single stimuli, recorded with the capillary tube is quite insensitive, requiring about 20 to 50 V electrometer in response to a break (b) and make (m) for a 1-cm deflection of the spot made by the elec- induction-oil stimulus (top). (Bottom) The capillary tron beam on the face of the tube. Therefore, con- electrometer record of the break (b) and make (m) siderable amplification was needed to display the stimuli from the induction coil; note that they are of millivolt nerve action potentials detected with elec- opposite polarity and that of the nerve responses (top) trodes on the surface of a nerve trunk. Gasser and are of the same polarity. (From Gotch & Horsley, 1888).


FIGURE HP3.3 Use of the capacitively coupled (C) triode to enlarge action potentials from a nerve (N) and display them on the string galvanometer (G). (b) (upper) An amplified action potential and (lower) the unamplified action potential. At the bottom is a record from a 100-Hz tuning fork [Forbes & Thacher 1920].

Erlanger [1922] built a three-stage amplifier to display nerve action potentials on the cathode-ray tube screen. Figure HP3.4 is the circuit diagram of their equipment. The cathode-ray tube was of the lowvoltage (300) type and was provided by J. B. Johnson and E. B. Craft of the Western Electric Company. The tube contained a little argon gas. The fluorescent screen was green with a long persistence (5 to 10 s). The deflection sensitivity was 10 to 20 V/cm. Before describing the Gasser and Erlanger Nobel prize-winning research, it is of interest to examine how the nerve was stimulated and the action potentials were recorded. Figure HP3.4 shows the nerve (N) on the left with the stimulating electrodes (T) and the recording electrodes (E). The nerve was crushed under the electrode connected to the grid (G) of the first vacuum-tube amplifier. Therefore, the nerve action potential appeared under the electrode on the uninjured nerve. Three stages of amplification were used to enlarge the nerve action potential 7000 to 8000 times, which caused the beam of the cathoderay tube to be deflected vertically. On the right of Fig. HP3.4 is a rheotome that (1) delivered the stimulus to an induction coil (PS) connected to the nerve stimulating electrodes (T), (2) started the cathode-ray tube beam moving across the face of the tube to provide a time axis by starting the 1-mF (nowadays µF) capacitor to charge (the voltage on this capacitor was connected to the horizontal deflecting plates in the cathode-ray tube), and (3) later discharged the 1-µF capacitor so that the cycle could be restarted. In this way, the stimulus was delivered to the nerve at the instant when the spot started horizontally across the face of the cathoderay tube. Typically, 30 stimuli were delivered per second to produce a clear (standing wave) action potential on the cathode-ray tube. Commenting on the action potential recorded from frog sciatic nerve, Gasser and Erlanger [1922] wrote: The action current has a gradual ascent, a steep smooth anacrotic limb [rising phase] and a more gradual catacrotic limb [falling phase]. The latter like the former shows a period of great initial acceleration so that the crest is situated near the anacrotic side. In frog nerve and some mammalian nerves there are secondary waves on the catacrotic limb n. Suggestions are made as to the cause of these waves. Soon an explanation was given for the secondary waves; it came in 1924 when Erlanger and Gasser [1924] published their classic report which will now be described. Having no camera, Gasser and Erlanger either placed tissue paper on the cathode-ray tube face and traced the waveform with a pencil or pressed photographic paper against the face of the tube and obtained a contact print. Figure HP3.5 is such an illustration of the action potential of the bullfrog sciatic nerve.



FIGURE HP3.4

Circuit diagram of the three-stage amplifier, cathode-ray tube, and stimulator used by Gasser and Erlanger [1922] to record nerve action potentials.

FIGURE HP3.5 Contact print from the cathode-ray tube showing the action potential of the bullfrog sciatic nerve recorded by Erlanger & Gasser [1924].

Erlanger and Gasser knew that their oscilloscope time base was exponential and corrected the oscillograms accordingly. A favorite method employed transilluminating a contact print and tracing the action potential on the back with a pencil to obtain a positive image that was then replotted on semilogarthmic paper. This step was essential because they desired a true temporal display of the compound action potential from a nerve trunk. By varying the distance over which the action potential was propagated, they were able to reveal that the nerve trunk contained groups of fibers that propagated with different velocities. Commenting on their results, they stated: Each of the waves of these compound action currents as it progresses along the nerve changes its form just as does the simple action current in the phrenic nerve. It is suggested that these changes are due, in part, at least, to slight differences in the propagation rate in individual, or in many small groups of fibers, whose action currents therefore get slightly out of phase as they progress. Finally, Erlanger and Gasser [1937] summarized their work in a monograph entitled Electrical Signs of Nervous Activity. In it they not only showed how the compound action potential of nerve depends on the velocities of propagation of the different bundles of nerve fibers in a trunk but also provided histograms (fiber maps) of the diameters of different-sized fibers in a nerve trunk. Figure HP3.6 shows the action potentials of the A, B, and C fibers. For this pioneering work Erlanger and Gasser received the Nobel prize in physiology and medicine in 1944. Meanwhile, in the United Kingdom, Adrian was conducting experiments that showed that a nerve fiber responds in an all-or-none manner and that intensity is signaled in a single nerve fiber by the frequency of action potentials. In addition, he showed that intensity is also signaled by the number of nerve fibers carrying messages. These remarkable discoveries were made with the capillary electrometer, which even then was considered to be a primitive instrument. Adrian [1928] defended his use of the capillary electrometer in the following way: The ideal instrument for recording nerve action currents is undoubtedly the cathode-ray oscillograph devised by Erlanger and Gasser, for in this the moving system is a stream of cathode rays, the inertia of which is completely negligible. At present, however, the intensity of the illumination from the ray is far too small to allow photographs to be made from a single excursion, and similar excursions must be repeated many times over before the plate or the eye is affected. As a result, the cathode-ray oscillograph can only be used in experiments where the same sequence of action currents can be repeated over and over again, and it is not suitable for recording an irregular series of action currents such as are produced by the activity of the central nervous system. Another instrument in which the


FIGURE HP3.6 Action potential components of the A, B, and C fibers of nerve. (From Erlanger J, Gasser HS. Electrical Signs of Nervous Activity. Philadelphia, University of Pennsylvania Press, 1937. With permission.)

inertia factor is extremely small is the capillary electrometer. This has fallen out of favor with the majority of physiologists because its records need analysis and because of its low sensitivity compared with that of the string galvanometer. These objections have now become of little importance. With the advent of reliable valve [vacuum-tube] amplifiers a low sensitivity in the recording instrument is no drawback at all, and the analysis of capillary electrometer records can be made in a few moments by the machine designed by Keith Lucas. As will be seen, the combination of value amplifier and [capillary] electrometer gives us an instrument of such range and precision that it promises access to fields of investigation which are as yet almost unexplored. In Adrian’s first studies [1928] he used the three-valve (vacuum-tube) amplifier shown in Fig. HP3.7 to record action potentials in different nerve fibers. In his next studies with Bronk [Adrian & Bronk, 1928], he used a single-stage amplifier (Fig. HP3.8) that drove a capillary electrometer across which could be connected earphones or another amplifier (B) that drove a loudspeaker (LS) to enable listening to the nerve action potentials. Adrian stated:

FIGURE HP3.7 The three-stage valve (vacuum-tube) amplifier and capillary electrometer (CE) used by Adrian to record action potentials in afferent nerves. The dashed lines represent shielding. (Redrawn from [Adrian, 1926].)


FIGURE HP3.8 Amplifier and capillary electrometer used by Adrian to record nerve action potentials along with headphones, amplifier (B), and loudspeaker (LS) used to aurally monitor their frequency. (From Adrian & Bronk, 1928.)

The three-valve amplifier owes much to the great kindness of Prof. Gasser, who supplied me with details of the amplifier used by him in America, and to the staff of Messers W. G. Pye and Co. of Cambridge, who redesigned an instrument on the same general line and planned the very compact and well shielded lay-out of the apparatus. Adrian provided details on his capillary electrometer by stating, “The capillary tube at present has a diameter of 0.03 mm at its working part, and a pressure of 140 mmHg is needed to bring the mercury to this point.” The working part refers to the sulfuric acid–mercury interface, the contour of which changes when current traverses it. Adrian stated that he could detect a voltage as small as 0.01 mV when his electrometer was connected to the three-stage amplifier. Adrian knew that a photographic recording of the change in contour of the meniscus of the capillary electrometer was not a true representation of the applied voltage; this fact had been pointed out by Burch [1892]. Keith Lucas, a collaborator, devised a mechanical instrument for correcting capillary electrometer records. However, Adrian was less interested in waveform than in the presence or absence of action potentials.

Action Potentials in Afferent Fibers In the paper that described the three-valve (vacuum-tube) amplifier and capillary electrometer, Adrian [1928] presented numerous examples of the nature of the action potentials in afferent nerve fibers in the frog sciatic nerve when the gastrocnemius muscle was stretched with known weights. He found that the frequency of the action potentials was related to the weight. Fig. HP3.9 shows his corrected capillary electrometer records of this experiment. After applying strict criteria to test the validity of his results, Adrian then recorded trains of afferent impulses in the saphenous nerve of a decapitated cat when a forcep was used to pinch the skin of the foot. The same result was obtained with a pin prick. He then recorded afferent impulses in the vagus nerve in the spinal and decerebrate cat and in the anesthetized rabbit. He also recorded trains of action potentials in the vagus nerve that were synchronous with the heartbeat and with respiration. Commenting on the latter he wrote: “The striking result is the absence of any sign of renewed discharge of impulses at the moment when the lungs are most deflated.” In the summary to his paper, Adrian stated:


FIGURE HP3.9 Corrected capillary electrometer records of afferent action potentials in a frog sciatic nerve when weights were applied to the gastrocnemius muscle. (c) shows the action potentials from 10 g applied for 10 s; (d) shows the same weight applied for 24 s; and (e) shows the action potentials for 100 g weight applied for 10 s. Note the higher frequency of action potentials with 100-g weight. (From Adrian, 1928.)

It is probable that many of the oscillations represent action currents in a single nerve fibre, and these same general form and the same general time relations (allowing for temperature differences) in all sensory nerves in which they can be isolated sufficiently for measurement. There is no evidence that an increase in the stimulus increases the size of the action currents in single fibres, but the frequency of the impulses in the nerve trunk increases and leads to interference and overlapping of impulses in different fibres. He continued: More detailed analysis of the results is postponed until experiments have been made on preparations containing a known number of sensory endings, if possibly only one. It was not long before Adrian succeeded in recording action potentials in a single nerve fiber. Adrian and Bronk [1928] stated: To be sure of what is happening in the single nerve fibre we have to devise a method which will put out of action all the other fibres in the nerve. The recording of impulses in a single fibre presents no difficulty, but the problem of isolating the fibre seemed much more formidable. Fortunately this has turned out not nearly so difficult as we had imagined. Adrian and Bronk [1928] used the cervical rabbit phrenic nerve and dissected single fibers free with a needle and viewed the dissection with a binocular microscope. Teasing out a selected single fiber and sectioning it, electrodes were placed at the distal end. In this way, only afferent information was recorded. The equipment that they used in Fig. HP3.8. Commenting on use of the headphones or a loudspeaker, they stated: The amplified action currents can be photographed with the capillary electrometer, but until the final stages are reached, it is usually more convenient to lead them to a telephone or loudspeaker and estimate the character of the discharged by the ear instead of the eye. They continued by showing how aural monitoring aided in the experiment: When only a few fibres are in action the electrometer excursions may be too small to detect on a screen, but they produce a series of faint clicks in the loudspeaker, and it is thus possible to control


FIGURE HP3.10 Capillary electrometer records from the phrenic nerve of the decerebrate cat breathing spontaneously. (a) The airway is open. (b) The trachea was clamped. (From Adrian and Bronk, 1928.)

the dissection, to expose a plate at the moment when the discharge is at its height, etc., without the inconvenience of wearing telephones. Adrian and Bronk [1928] investigated the frequency of action potentials in an intact phrenic nerve of the decerebrate cat breathing spontaneously. The single-state amplifier (Fig. HP3.8) was used. Figure HP3.10a illustrates the phrenic nerve action potentials during spontaneous breathing, and Fig. HP3.10b shows the recordings with the trachea clamped, the breathing becoming labored. Having shown that the frequency of the action potentials in the phrenic nerve increased with depth of breathing, Adrian and Bronk [1928] set themselves the task of investigating how intrapulmonic pressure was related to the frequency of stimuli applied to the phrenic nerve. Using a rabbit in which the central circulation was occluded above C1, they connected a manometer to the trachea that could be clamped so that the manometer read the negative intrapulmonic pressure when the C4 phrenic nerve root was stimulated. They used a coreless induction coil connected to a rotary switch (rheotome) to enable delivery of stimuli at any desired frequency. By delivering bursts of stimuli of different frequencies, they plotted the negative intrapulmonic pressure as a function of frequency, clearly demonstrating the dependence of the amplitude of inspiration on the frequency of phrenic nerve stimuli. Adrian published two monographs on his work; one was entitled, The Mechanism of Nervous Action, and the other, The Basis of Sensation. Interestingly, neither publication contains illustrations obtained with the cathode-ray oscilloscope; however, there are string galvanometer and capillary electrometer recordings. Obviously, Adrian had brought the capillary electrometer to a high degree of perfection, but it is important to remember that Adrian’s interest lay in the frequency of the action potentials, not the intimate details of their waveforms. Perhaps the best summary of Adrian’s contributions appears in a single sentence in The Mechanism of Nervous Action. The section entitled, “Gradation of Activity,” states: There is certainly no evidence to suggest that the impulses are graded in size, for the fact that sensory messages may produce a small or large effect according to the intensity of the stimulus is naturally explained by the varying number of fibres in action and by the varying frequency of the discharge in each fibre. In 1932, Adrian shared the Nobel prize in physiology and medicine with Sherrington. The citation read, “For their discoveries regarding the function of the neuron.”


FIGURE HP3.11 The transmembrane potential of the giant axon of the squid at rest (left) and during activity. Time marks 2 ms. (From Hodgkin and Huxley, 1939. With permission.)

Code of the Nervous System From the impressive research performed by Gasser and Erlanger and by Adrian, the code of the nervous system was discovered. Stated simply, (1) intensity is signaled by the frequency of action potentials in a single axon, the action potentials all being the same, and (2) intensity is also signaled by the number of axons transmitting the information. In modern terms, the nervous system is a communications system that is binary and frequency modulated.

True Form of the Nerve Action Potential Although the studies by Bernstein, Gasser, Erlanger, and Adrian provided information on the nerve action potential, its true form could not be established until the micropipet electrode was used to measure the transmembrane potential. Hodgkin and Huxley [1939] obtained the true waveform of the nerve action potential and thereby ushered in the modern era of electrophysiology. In a paper published in Nature on October 21, 1939, they reported that J. Z. Young [1930] called their attention to the giant axon (500 µm in diameter) of the squid, which was ideal for electrophysiologic studies. The first micropipet electrode used by Hodgkin and Huxley consisted of a glass tube, 100 µm in diameter, that was slipped into the cut end of the giant axon. The electrode was mounted vertically and filled with seawater, and a silver–silver chloride electrode was inserted. The axon was then dipped into a container of seawater in which a second chlorided silver electrode was placed. The transmembrane potential was found to be –45 mV at 20°C. When the axon was stimulated, the action potential was 90 mV. Figure HP3.11 is a reproduction of this historic record. The Hodgkin and Huxley experiment clearly demonstrated two important phenomena: (1) there is a measurable resting transmembrane potential, and (2) the action potential is larger than the resting transmembrane potential, the latter indicating that activity represents more than a mere disappearance of the transmembrane potential. The same team later provided the explanation with their classic papers on ion fluxes. Today, the action of many drugs is explained on the basis of ion fluxes.


References Adrian ED, Bronk DW. 1928. The discharge of impulses in motor nerve fibres: 1. Impulses in single fibres of the phrenic nerve. J Physiol 66:81. Adrian ED. 1928. The Basis of Sensation. New York, WW Norton. Bernstein J. 1868. Über den zeitlichen verlauf der negativen Schwankung des nervenstroms. Arch Ges Physiol 1:73-207. Burch GJ. 1892a. On the time relations of the excursions of the capillary electrometer. Phil Trans R Soc (Lond) 83A:81. Burch GJ. 1892b. On a method of determining the value of rapid variations of potential by means of the capillary electrometer communicated by J.B. Sanderson. Proc R Soc (Lond) 48:89. DeForest L. 1907, 1908. U.S. patent nos. 841,387 (1907) and 879,532 (1908). Einthoven W. 1903. Ein neues Galvanometer. Ann Phys 12(suppl 4):1059. Erlanger J, Gasser HS. 1924. The compound nature of the action current of nerve as disclosed by the cathode ray oscillograph. Am J Physiol 70:624. Erlanger J, Gasser HS. 1937. Electrical Signs of Nervous Activity. Philadelphia, University of Pennsylvania Press. Forbes A, Thacher C. 1920. Amplification of action currents with the electron tube in recording with string galvanometer. Am J Physiol 56:409. Gasser HS, Erlanger J. 1922. A study of the action currents of nerve with the cathode ray oscillograph. Am J Physiol 62:496. Gasser HS, Newcomer HS. 1921. Physiological action currents in the phrenic nerve. Am J Physiol 57(1):1. Geddes LA, Foster KS, Senior J, Kahfeld A. 1989. The inductorium: The stimulator associated with discovery. Med Instrum 23(4):308. Gotch F, Horsley V. 1888. Observations upon the electromotive changes in the mammalian spinal cord following electrical excitation of the cortex cerebri. Proc R Soc (Lond) 45:18. Helmhotz H. 1850. Note sur la vitesse de propagation de l’agent nerveux dans les nerfs rachidiens. C R Acad Sci 30:204. Helmholtz H. 1951. Note sur la vitesse de propagation de l’agent nerveux. C R Acad Sci 32:262. Helmholtz H. 1853. On the methods of measuring very small portions of time and their application to physiological purposes. Phil Mag J Sci 6:313. Hodgkin AL, Huxley AF. 1939. Action potentials recorded from inside a nerve fiber. Nature 144:710. Hoff HE, Geddes LA. 1957. The rheotome and its prehistory: A study in the historical interrelation of electrophysiology and electromechanics. Bull Hist Med 31(3):212. Hoff HE, Geddes LA. 1960. Ballistics and the instrumentation of physiology: The velocity of the projectile and of the nerve impulse. J Hist Med All Sci 15(2):133. Young JZ. 1930. Structure of nerve fibers and synapses in some invertebrates. Cold Spring Harbor Symp Quant Biol 4:1.


Polk, C. “Biological Effects of Nonionizing Electromagnetic Fields .” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

Noninvasive inductive coupling device used to apply low field intensities.

IX Biological Effects of Nonionizing Electromagnetic Fields Charles Polk University of Rhode Island 89 Dielectric Properties of Tissues

Kenneth R. Foster

Definitions and Basic Phenomena • In Vivo Versus in Vitro Properties • Temperature Coefficients • Dielectric Data: Tabulated

90 Low-Frequency Magnetic Fields: Dosimetry, Cellular, and Animal Effects Maria A. Stuchly Physical Interactions — Dosimetry • Biological Effects • Concluding Remarks

91 Therapeutic Applications of Low-Frequency Sinusoidal and Pulsed Electric and Magnetic Fields Charles Polk Bone and Cartilage Repair with PEMF and Other Signals • Soft-Tissue Repair and Nerve Regeneration • Mechanisms of Field-Tissue Interaction

92 Biologic Effects of Radiofrequency and Microwave Fields: In Vivo and In Vitro Experimental Results Edward Elson Cellular and Molecular Biology • Reproduction, Growth, and Development • Effects on the Nervous System • Behavioral Effects • Effects on the Cardiovascular and Hematopoietic Systems • Auditory and Ocular Effects • Conclusion

93 Radio Frequency Hyperthermia in Cancer Therapy C. K. Chou, Rulong Ren Methods of EM Heating • Conclusion

94 Electroporation of Cells and Tissues

James C. Weaver

Background • Biological Barriers Based on Lipids • Decrease of the Membrane Barrier to Transport • Basis of Electroporation • Molecular Transport • In Vitro Electroporation Applications • In Vivo Electroporation Applications


T

HE PURPOSE OF BIOMEDICAL ENGINEERING is to develop and employ the best available technology for the benefit of human health. Implicitly, the discussion of biologic effects of nonionizing electromagnetic fields in a biomedical engineering handbook will then be concerned with beneficial medical applications of such fields. However, these applications must be based to as large an extent as possible on scientific knowledge. This section therefore begins with a description of experimentally established dielectric properties of biological materials (Chapter 89). This is followed by a general discussion of biologic effects—adverse as well as beneficial—of low-frequency magnetic fields (Chapter 90). This chapter is restricted to low-frequency magnetic—rather than electric and magnetic—fields because the succeeding chapter (Chapter 91), which deals with therapeutic applications, indicates that time-varying magnetic fields are of greater interest. At the present time, it is not known whether observed biologic effects of low-frequency magnetic fields (particularly of those having modest amplitude) are due to the electric fields and currents which they induce in tissue as a consequence of Faraday’s law or are due to an as yet unknown mechanism of direct magnetic field–tissue interaction. Since the fundamental equations that characterize field-tissue interaction are very different in the power to audiofrequency range from those which describe biologic effects at radio and microwave frequencies, separate chapters are devoted to the different frequency domains. Thus Chapter 92 is devoted to discussing biologic effects, again not necessarily beneficial, of RF fields, and Chapter 93 describes methods for their principal medical application, which is the production of desired hyperthermia. While Chapters 89 to 92 describe primarily measurements, effects, and applications of relatively lowintensity fields, Chapter 94 is devoted to the fundamentals of the increasingly important area of the interaction of short, high-intensity electric field pulses with biologic cells. It is shown that understanding of “electroporation” is not only important for understanding the nature of electric injury, but that this phenomenon is becoming a very important tool in biotechnology for transferring genetic material between different cells, as well as being applied for local drug delivery to tissue. The division of this section along the indicated lines is mandated by the observations that high- and low-intensity fields have different biologic effects—to consider extremes, electrocution versus healing of fractured bones—and that different frequencies of electromagnetic energy interact differently with materials of all types. Of particular importance is the relation between the wavelength λ of electromagnetic waves and the magnitudes L of field-generating and field-absorbing structures of their distances. Since

λ=

c f

(IX.1)

where f is the frequency and c is the propagation velocity (3.0 × 108 m/s in free space), it is obvious that L > E). Here E is the electric and H is the magnetic field intensity. AT RFs or microwaves, on the other hand, one is generally interested in the “far field” or “radiation field” region, where the ratio between E and H is equal to the “wave impedance” η. It is given in a loss-free medium by η = µ ⁄ , where µ and are, respectively, the magnetic permeability and dielectric permittivity of the medium in which the wave is propagating. (The magnetic flux density B is given by B = µH.)


TABLE IX.1

Biologic Effects of ELF and Pulsed Electric Fields V/m Inside Tissue or Fluid

Electroporation (transient or permanent, depending on amplitude and duration of pulse) Cell rotation in insulating fluid Nerve/muscle stimulation Initiation of firing Alteration of firing rate Subtle “long-term” (t > 10 min) effects (bone/soft tissue repair, Ca-efflux)

105 104 103 10 10–3 to 10–4

The only precaution that is needed in the application of quasi-statics at ELF is to remember the relation between E and B when the sources involved produce predominantly a magnetic field (such as the application coils in a bone therapy apparatus):

∇×E=−

∂B ≠0 ∂t

(IX.2)

Thus in the tissue, the curl of the electric field, i.e., the circulation of the electric field per unit area, is not zero, as it is when the applied quasi-static field comes directly from an electric field source. As a consequence, current distributions in tissue due to such different sources are likely to be very different even if current magnitudes are possibly, in a particular case, identical. As already mentioned, Chapters 89 to 92 deal primarily (although not exclusively) with low-intensity fields while Chapter 94 describes only high-field-intensity effects. It is of interest here to summarize the known biologic effects of various intensities of dc and low-frequency electric fields (Table IX.1). The basic laws that give forces (F) and torques (T) due to electric and magnetic fields are listed on Table IX.2. In this table, q = electric charge, v = charge velocity, p = electric dipole moment (magnitude of p = charge × separation distance between opposite charges), and m = magnetic moment (magnitude of m = magnitude of circulating current × area enclosed by current; direction of m is perpendicular to the plane of the circulating current). The derivatives with respect to the distance x are a one-dimensional representation of spatial field gradients. The equations listed in Table IX.2, as well as Maxwell’s equations (or the appropriate applicable approximations) and the known laws of classic and/or quantum mechanics, subject to the boundary conditions characteristic for living tissue, must provide a quantitative explanation for the observed biologic effects of electric and magnetic fields. Biologic tissue and cells are obviously extremely complex media; they are not only extremely inhomogeneous and anisotropic, but they are also not in thermodynamic equilibrium (unless dead). Thus the application of physical laws to the explanation of field-tissue interactions becomes a very complex problem, and the physicist and engineer must be careful not to provide “explanations” or to set limits on what should be “possible” or “impossible” based on physical models that are very far from even an approximate representation of biologic conditions. It is not surprising, therefore, that only relatively TABLE IX.2 Forces and Torques on Electric gross effects, such as cell damage in electroporation Charges and Electric and Magnetic Dipole Moments or heating by radiofrequency fields of sufficiently high intensity, are reasonably well (although not F = qE F = q (v × B) completely) understood. Some “nonthermal” dE F=P F = m dB dx microwave effects (Chapter 92) and several lowdx T=p×E T=m×B frequency effects of magnetically induced electric fields of less than 10–3 V/m (Table IX.1 and Chapters Noninvasive inductive coupling device used to apply low field intensities. 90 and 91) or alternating magnetic fields of less


than about 100 µT are presently not understood. The interested reader may want to consult references given in the following chapters and/or discussions by Adair [1991], Adey [1990], Blanchard and Blackman [1994], Douglass et al. [1993] Kaiser [1993], Kirschvink et al. [1992], Lednev [1991, 1993], Luben [1993], Polk [1992, 1994], Tenforde [1992b], Weaver and Astumian [1990], or additional references listed by these authors. There are numerous, experimentally confirmed in vitro biologic effects at low ELF field intensities, e.g. Harland and Liburdy [1997]. These effects may be beneficial, may be adverse, or may be of no consequence in entire animals or humans. Some epidemiologic data (Chapter 90) suggest the possible existence of adverse health effects in Portier and Wolfe [1998]. The experimental evidence existing at the present time is insufficient, however, to decide whether any of the more promising physical models that are discussed in the given references can provide an adequate explanation for any of the observed biologic effects. Two additional observations appear to be appropriate. The first is that the biologic environment is much more complex than the man-made inanimate world, where many different electric and magnetic field effects are used in devices as diverse as large electric motors, Xerox machines, feedback controllers, or television sets. It would therefore be surprising if identical electric or magnetic field “mechanisms” would be responsible for effects of different frequencies or field intensities on different biologic systems (e.g., immune cells versus bone tissue). The second observation concerns the existence and amplitude of the ambient, natural electric and magnetic fields. It is sometimes pointed out—correctly—that humans have been exposed for eons to the natural geomagnetic field of about 50 µT and to the natural dc electric field, which at the earth’s surface is about 100 V/m and can be several thousand volts per meter under a thunder cloud. Without conducting contact, dc electric fields of this magnitude have indeed not been implicated in any biologic effect, nor have dc magnetic fields of about 50 µT, without the simultaneous presence of an alternating field, been shown to have significant biologic consequences [Tenforde, 1992a; Frankel and Liburdy, 1995]. Virtually all the reported biologic effects of low-intensity electric and magnetic fields on living organisms, including cells, involve time variation, particularly at frequencies below 100 Hz (or microwaves amplitude modulated at these frequencies). The low-frequency effects on animals living outside water appear also to be primarily due to time-varying magnetic fields. Clearly, the natural ambient magnetic field of about 50 µT is a static field that exhibits diurnal slow variations (i.e., with frequency components well below 1 Hz) only of the order of 10–3 µT and with “large” variations (taking minutes to hours for one cycle) during magnetic storms occurring a few times each year, rarely as large as 0.5 µT [Matsushita and Campbell, 1967]. The natural background of magnetic fields with “higher” frequencies (10 to 100 Hz) is on the order of 10 pT [Polk, 1982]. Thus the geomagnetic field is clearly a static field that cannot induce an electric field into a stationary object (as a consequence of Faraday’s law) anywhere near the magnitude of that induced by a 60-Hz magnitude field even as low as 0.1 µT. Uniform linear motion of an object, such as that of a walking human, in a nearly uniform magnetic field of about 50 µT will produce an induced “Lorentz” electric field proportional to the product of velocity and flux density. However, that field cannot produce circulating electric currents as long as the total magnetic flux integrated over the cross-sectional area of the object (animal or human), ∫∫ B ds, does not change. Only tumbling motion, such as linear motion by means of somersaults, through the earth’s magnetic field could produce induced electric currents comparable in magnitude to those induced by a 1-µT, 60-Hz field. Thus there is no basis for statements, sometimes made, that biologic effects of weak ELF magnetic fields are impossible, because animals or humans have been exposed throughout their development to the geomagnetic field.

References Adair RK. 1991. Constraints on biological effects of weak extremely-low-frequency electromagnetic fields. Phys Rev A 43(2):1039. Adey WR. 1981. Tissue interaction with nonionizing electromagnetic fields. Physiol Rev 61:435.


Adey WR. 1990. Electromagnetic fields, cell membrane amplification and cancer promotion. In BW Wilson, RG Stevens, LE Anderson (eds), Extremely Low Frequency Electromagnetic Fields: The Question of Cancer, pp 211–249. Columbus, Ohio, Battelle Press. Blanchard JP, Blackman CF. 1994. See refs. to Chap. 92. Douglass JK, Wilkens L, Pantazelou E, Moss F. 1993. Noise enhancement of information transfer in crayfish mechanoreceptors by stochastic resonance. Nature 365:337. Frankel RB, Liburdy R. 1995. Biological effects of static magnetic fields. In CRC Handbook of Biological Effects of Electromagnetic Fields, 2d ed. Boca Raton, Fla, CRC Press. Harland JD, Liburdy R. 1997. Environmental magnetic fields inhibit the antiproliferative action of Tamoxifen and Melatonin in a human breast cancer cell line. Bioelectromagnetics 18:555–562. Kaiser F. 1993. Explanation of biological effects of low-intensity electric, magnetic and electromagnetic fields by nonlinear dynamics. In 9th Annual Review of Progress in Applied Computational Electromagnetics. Monterey, Calif. Kirschvink JL, Kobayashi-Kirschvink A, Woodford BJ. 1992. Magnetite biomineralization in the human brain. Proc Natl Acad Sci USA 89(16):7683. Lednev VV. 1991, 1993. See refs. to Chap. 92. Liboff AR, McLeod BR. 1988. See refs. to Chap. 92. Litovitz TA, Montrose CJ, Doinov P, et al. 1994. Superimposing spatially coherent electromagnetic noise inhibits field-induced abnormalities in developing chick embryos. Bioelectromagnetics 15:105. Luben RA. 1993. See refs. to Chap 92. Matsushita S. 1967. Geomagnetic disturbances and storms. In S Matsushita, WM Campbell (eds), Physics of Geomagnetic Phenomena, vol 2, pp 793–821. New York, Academic Press. Polk C. 1982. Schumann resonances. In H Volland (ed), CRC Handbook of Atmospherics, vol 1, pp 111–178. Boca Raton, Fla, CRC Press. Polk C. 1992. Dosimetry of extremely-low-frequency magnetic fields. Bioelectromagnetics 13(S1):209. Polk C. 1994. Physical/chemical mechanisms and signal-to-noise ratios. In T Tenforde (ed), Proceedings of the 1994 Annual Meeting of the National Council on Radiation Protection. Washington, NCRP. Portier CJ, Wolfe MS (eds). 1998. Assessment of Health Effects from Exposure to Power-Line Frequency Electric and Magnetic Fields. NIEHS, PO Box 12233, Research Triangle Park, NC 27709, NIH Publication No. 98-3981. Tenforde TS. 1992a. Interaction mechanisms and biological effects of static magnetic fields. Automedica 14:271. Tenforde TS. 1992b. Biological interactions and potential health effects of extremely-low-frequency magnetic fields from power lines and other common sources. Annu Rev Public Health 13:173. Weaver JC, Astumian RD. 1990. The response of cells to very weak electric fields: The thermal noise limit. Science 347:459.


Noninvasive inductive coupling device used to apply low field intensities.

Foster, K. R. “Dielectric Properties of Tissues.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

89 Dielectric Properties of Tissues 89.1 89.2 89.3

Kenneth R. Foster University of Pennsylvania

Definitions and Basic Phenomena In Vivo Versus in Vitro Properties Temperature Coefficients Reversible Changes • Irreversible Changes

89.4

Dielectric Data: Tabulated

The bulk electrical properties of tissues and cell suspensions have been of interest for many reasons for over a century. These properties determine the pathways of current flow through the body. This gives them fundamental importance in studies of biologic effects of electromagnetic fields, in measurements of physiologic parameters using impedance, and in basic and applied studies in electrocardiography, muscle contraction, nerve transmission, and numerous other fields. I will briefly define the quantities used to characterize the bulk electrical properties of tissues and give some of the background information needed to interpret the data. A more extensive review is presented elsewhere [1]. Other reviews of tissue properties are by Schwan [2], Pethig [3], Grant et al. [4], Schanne and P.-Ceretti [5], and Duck [6]. Other tabulations of tissue properties are by Schwan [2], Geddes and Baker [7], and Stuchly and Stuchly [8]; Schwan [9] has published an extensive review of practical measurement techniques.

89.1 Definitions and Basic Phenomena The dielectric permittivity 0 and conductivity σ of a material are, respectively, the dipole and current densities induced in response to an applied electric field of unit amplitude.1 The significance of these quantities can be illustrated by considering an ideal parallel-plate capacitor, whose plates have surface area A and separation d. The capacitance C and conductance G of the capacitor are then

C=

0 A d

σA G= d

(89.1)

(This neglects the effects of fringing fields and applies at low frequencies where propagation effects can be neglected.) At radian frequency ω, the admittance Y of the capacitor can be written 1

*In MKS units, the permittivity and conductivity have units of farads per meter and siemens per meter, respectively. For convenience, we write the permittivity as (the relative permittivity) times 0, the permittivity of vacuum. 0 = 8.85 (10–12 F/m). The resistivity ρ = 1/σ.


( ) A = (σ + jω ) d

Y = G + jωC

0

= σ*

(89.2)

A d

= jω * 0

A d

where σ * = σ + j ω0 is the complex conductivity, and * = – j σ /ω0 is the complex permittivity.2 In the usual notation, * = ′ – j″, where ″ is the loss and tan (″/′) is the loss tangent. Typically, for soft tissues at low frequencies,

σ >> ω o and the tissue can, for many purposes, be approximated adequately by considering it to be a pure conductor and neglecting the permittivity entirely. For tissues, both and σ are strong functions of frequency (Fig. 89.1). This frequency dependence (dispersion) arises from several mechanisms. These mechanisms are discussed, with reference to simple biophysical models, in Foster and Schwan [1]. For a typical soft tissue, different mechanisms dominate at different frequency ranges: • At low frequencies (typically below several hundred kilohertz), the conductivity of the tissue is dominated by conduction in the electrolytes in the extracellular space. The bulk conductivity of the tissue is then a sensitive function of the volume fraction of extracellular space and the conductivity of the extracellular medium. • At low frequencies, the tissue exhibits a dispersion (the alpha dispersion), centered in the lowkilohertz range, due to several physical processes. These include polarization of counterions near charged surfaces in the tissue and possibly the polarization of large membrane-bound structures in the tissue. At frequencies below the alpha dispersion, the relative permittivity of tissue reaches very high values, in the tens of millions. The alpha dispersion is very apparent in the permittivity but hardly noticeable in the conductivity of the tissue. • At radiofrequencies, the tissue exhibits a dispersion (the beta dispersion), centered in the range 0.1 to 10 MHz, due to the charging of cell membranes through the intracellular and extracellular media. Above the beta dispersion, the cell membranes have negligible impedance, and the current passes through both the extracellular and intracellular media. The beta dispersion is apparent in both the permittivity and conductivity of the tissue. • At microwave frequencies (above 1 GHz), the tissue exhibits a dispersion (the gamma dispersion) due to rotational relaxation of tissue water. This dispersion is centered at 20 GHz and is the same as that found in pure liquid water. In addition to these three major dispersions, other smaller dispersions occur due to rotational relaxation of bound water or tissue proteins, charging of membranes of intracellular organelles, and other effects. These dispersions overlap in frequency and lead to a broad and often featureless dielectric dispersion in tissue. 2The term dielectric constant is used, often in the chemical literature, to indicate the relative permittivity of pure liquids at low frequencies, where is essentially independent of frequency.


FIGURE 89.1 Data from liver tissue (the composite of several sets of data, from the Table 89.1). () Relative permittivity , () conductivity σ, () ω0. When σ >> ω0, the tissue may be regarded for many purposes as a pure conductor. The major dispersion regions (α,β,γ) are indicated on the figure. The lines are regression lines through the data.

These dispersions do not affect the permittivity and conductivity in the same way. For a single-timeconstant dispersion centered at frequency fc , the change in permittivity ∆ is related to the change in conductivity ∆σ :

∆σ = 2πfc ∆ 0 Thus the alpha dispersion (at kilohertz frequencies) is associated with a small (usually imperceptible) increase in tissue conductivity but a very large decrease in permittivity. By contrast, the beta dispersion represents a large decrease in the permittivity (from several thousand to less than 100) and a large increase in conductivity (by a factor of 10 or so).

89.2 In Vivo Versus In Vitro Properties The dielectric properties of the tissues that are summarized in the Table 89.1 pertain, for the most part, to excised tissues. The relation between these properties and the dielectric properties of tissue in vivo is a complicated matter. At low frequencies (below about 0.1 MHz), electric current largely passes through the extracellular space, and the tissue conductivity is a sensitive function of the extracellular volume fraction. Any changes in the fluid distribution between intracellular and extracellular compartments can lead to a pronounced change in the low-frequency conductivity of the tissue. For example, substantial (twofold) decrease in the conductivity of rat kidney [10] and sheep myocardium [11] have been reported within a few minutes after death of the animal or after experimentally



1  2  10 Hz  3 4  5  100 Hz  6 7  8  1 kHz  9 10   11  10 kHz  12  13   14  100 kHz  15  16   17  1 MHz  18  19   20  10 MHz  21   22  23  100 MHz  24  25   26  1 GHz  27  28   29  3 GHz  30  31   32  10 GHz  33 

Frequency

TABLE 89.1

0.076

0.08

0.085

0.40

0.52

0.52

0.55

0.65 0.56–0.59 0.38–0.44

2.7 ± 0.07 2.8 8.3 7.7 8.8

0.83–0.85 0.58–0.63 0.86–0.87 0.92–0.96 0.69–0.75 0.95–0.99 0.9 ± 0.08 0.75–0.82 1.38–1.45 1.3 1.5

0.076

B Skeletal Muscle Perpendicular (Nonoriented)

0.52

Skeletal Muscle Parallel

A

C

0.73

0.80 ± 0.02

0.94–1.05 0.73–0.76 0.48–0.51

1.05 0.53 0.75 ± 0.02 1.2 1.09–1.13 2.0 2.5 ± 0.03

2.3 ± 0.05

4.5–7.4

2.7 6.5 10.0

0.95–0.97 1.0

0.64–0.68 0.50–0.57

0.37–0.39

0.84 0.55–0.53

0.63

Kidney

F

0.24–0.25

Conductivity (S/m)

0.27 0.30 0.47 0.46 0.42–0.46 0.72 0.70 0.60–0.71 0.98 1.2 0.95–1.0 2.0 2.4 2.8 5.8–6.7 10.0

E

Spleen

0.62

0.11

0.096

0.092

0.089

Lung

D

0.15 0.16

0.15

0.13

0.13

0.12

Liver

Dielectric Properties of Selected Tissues

8

0.21–0.28 0.30 0.29–0.31 0.36–0.48 0.66–0.72 0.52–0.85 0.89–0.94 0.80 0.81–0.82 1.8–2.1 1.5

0.14–0.19

0.12–0.15

Brain White Matter

G

10

2.0

1.1 0.89–1.17

0.35 0.38 0.45–0.63 0.69 0.45

0.21

0.17

Brain Gray Matter

H

9.1 10.5

2.5–3.1

0.16

0.5–1.7

1.4–1.6 1.3

1.0

0.0237

0.0173

0.55 0.68

0.68

0.68

0.60

Whole Blood

J

0.05

0.0574 0.7

0.0144

0.0133

0.0129

0.0126

Bone

I

0.3–0.4

0.03–0.09

0.7–0.8

0.02–0.07

1.11

0.71

0.02–0.07

Fat

K


1  2  10 Hz  3 4  5  100 Hz  6 7  8  1 kHz  9 10   11  10 kHz  12  13   14  100 kHz  15  16   17  1 MHz  18  19   20  10 MHz  21  22   23  100 MHz  24  25   26  1 GHz  27  28   29  3 GHz  30  31   32  10 GHz  33  3.2 × 105

1.2 × 105

7 × 104

3 × 104

1.1 × 106

2.2 × 105

8 × 104

1.5 × 104 24,800–27,300 14,400–15,800

52.5 ± 0.7 46 40–42 37 35

2,460–2,530 1,900–2,150 170–190 187–204 162–181 67–72 68 ± 2 64–70 57–59 58 48

106

107

338 300 251–265 77 79 65–68 46 55 47–49 42 53 42–43 34–38 37 35

89–95 56–62 85 ± 1

83 71–76 81 ± 3 54 50–51 50 520.6 47.5 ± 1

30–37

46 42 38

43 46

431–499 190–204

2,390–2,690

321 352–410

1,450

1,970 1,970

10,900–12,500

Relative Permittivity

3,260

35

2.5 × 104

8.5 × 104

4.5 × 105

2.5 × 107

9760 1.4 × 104

5.5 × 104

1.3 × 105

8.5 × 105

5 × 107

25

163–209 200 190–191 57–66 65 58–64 40–44 35 38–39 35–41 33

543–827

1,960–3,400

40

44

45 47–51

352 380 237–289 90 90 65–80

1,250

3,800

8

7.5

8

23

37

87

280

640

1,000

3,800

50–52 45

67 72–74 58–62 63–67 63 55–56

200

2,040

4,000 2,740

2,810

2,900

3.5–4.0

4–7

4.3–7.5 3–6

4.5–7.5

2 × 104

5 × 104

1.5 × 105


Dog skeletal muscle, 37°C (av of 5 measurements, SD ~ 30%)

Nonoriented dog skeletal muscle, 37°C (range of 3 measurements)

Nonoriented rat skeletal muscle, 37°C (range of 2 measurements)

Dog liver, 37°C (single specimen)

Rabbit liver, 37°C (single specimen)

Dog spleen, 37°C (single specimen)

Dog kidney, 37°C (range of 2 measurements)

Dog brain, white and gray matter, 37°C (range of 2 measurements)

Dog spleen, 37°C (values at 1,3 GHz interpolated, single specimen)

19

20

20

20

20

20

20

20

21,22

1 A,B 4 A,B 7 A,B 10 A,B 13 A,B 15 A,B 18 A,B 21 A,B 24 A,B 14 A,B 17 A,B 20 A,B 13 C 16 C 19 C 22 C 14 C 17 C 20 C 23 C 13 E 16 E 19 E 22 E 13 F 16 F 19 F 22 F 13 G,H 16 G,H 19 G,H 22 G,H 27 E 30 E 33E

Tissues

Ref.

Coordinates 33 A,B 1C 4C 7C 10 C 1D 4D 7D 10 D 7K 10 K 19 A,B 22 A,B 25 A,B 20 E 23 E 26 E 20 F 23 F 26 F 21 G,H 24 G,H 27 G,H 21 C 24 C 27 C 28 C 23 D,J 22 K 26 K 26 D 27 J 23 A,B 26 A,B 29 A,B

Coordinates

References

27

Various cat tissues, in vivo (av ± SD, 55 measurements in 4 animals); value at 10 GHz extrapolated from 8.0 GHz; 1 GHz interpolated

Bovine liver, 37°C Beef blood Excised human tissues (deflated lung) 27°C (measurement frequencies 0.2–0.9 GHz)

Cat liver, in vivo, 35°C (range of 3 measurements)

24

25 26

Cat brain, in vivo, 33°C (range of 3 measurements)

Cat kidney, in vivo, 35°C (range of 3 measurements)

24

24

Cat spleen, in vivo, 35°C (range of 3 measurements)

Cat skeletal muscle, in vivo, 31°C (range of 3 measurements)

24

24

Dog fat, in situ

Dog lung, inflated, in situ (av of 20 measurements, SD ~ 25%)

Dog liver, in situ (av of 20 measurements, SD ~ 25%)

Tissues

23

23

23

Ref.


Dog skeletal muscle, 37°C (values at 1,3 GHz interpolated, single specimen)

Dog liver, 37°C (value at 1.3 GHz extrapolated)

Rat femur, 37°C, immersed in Hank’s buffered saline, radial direction (single sample)

Mixed brain tissue, mouse, 37°C (value at 3 GHz interpolated)

Rabbit blood, room temperature

21,22

21,22

28

29

15

Reprinted from Foster and Schwan, 1994.

27 A,B 30 A,B 25 E 28 E 31 E 24 F 27 F 30 F 33F 26 C 29 C 32 C 4I 7I 10 I 13 I 16 I 19 I 22 I 25 G 28 G 8J 11 J 14 J

32 A,B 24 E 17 J 20 J 13 J 25 C 30 C 25 K 28 K 25 J 28 J 25 I 28 I 31 A,B,I,J,K 32 J 20 G,H 23 G,H 26 G,H 29 G,H 32 G,H 4J 7J 29

17 31

30 16

Sheep blood, 18°C

Human (9.4 GHz), 37°C Dog brain, white and gray matter, 37°C

Normal human blood, hematocrit 40%, 21°C (50 kHz) Various tissues, dog, horse, 38°C (except 25 I, 28 I, 25°C); the measurements were made at 8.6 GHz and extrapolated to 10 GHz

induced ischemia. These changes are almost certainly a result of changes in fluid distribution within the tissue associated with cell swelling. These changes are likely to be much less pronounced in the permittivity and in the conductivity above about 0.1 MHz. Most of the data in Table 89.1 were taken from excised tissues within minutes to hours after the death of the animal and may well represent systematic under-estimates of the conductivity of tissues in vivo. This calls for caution in their use.

89.3 Temperature Coefficients The dielectric properties of tissues change with temperature. Below about 44 to 45°C these changes are generally reversible. They principally reflect the changes in conductivity of electrolyte with temperature, either directly (in the conductivity of tissue) or indirectly, as they affect the dielectric dispersion. At higher temperatures, thermal damage will result in irreversible changes in the dielectric properties of tissue. The extent of such changes depends on the tissue type, duration of heating, and other factors. For canine skeleton, such changes occur above 44.5°C. Figure 89.2 shows the temperature coefficient, defined in the figure, for the conductivity and permittivity of canine muscle and brain. These data pertain to reversible changes only.

89.4 Dielectric Data: Tabulated Table 89.1 presents selected permittivity and conductivity data from various tissues. Where available, the table presents up to three values for the tissues, including measurements performed on excised tissues of various species and in vivo. The aim is to present primarily new data; in some cases I have included earlier data. Sources are References 15 to 31. The large variability in reported properties of these tissues is illustrated in the table and is due in part to biological variability and in part to differences in condition of the tissues. All data pertain to tissues at body temperature (37 to 38°C). Another extensive tabulation of dielectric data has recently been prepared by Gabriel et al. [32]. These data are available (as of time of publication) on the Internet at http://www.brooks.af.mil/AFRL/HED/ hedr/reports/dielectric/home.html and several other sites.


FIGURE 89.2 Fractional change with reciprocal temperature of dielectric properties of (a) dog muscles and (b) brain near 37°C (from Foster and Schwan, ref. 1). In most cases, the temperature coefficients have been calculated from two measurements at 25 to 28 and 37°C.


References 1. Foster KR, Schwan HP. 1994. Dielectric properties of tissues. In C Polk, E Postow (eds), Handbook of Biological Effects of Electromagnetic Fields, 2d ed. Boca Raton, Fla, CRC Press. 2. Schwan HP. 1957. Electrical properties of tissue and cell suspensions. In Advances in Biological and Medical Physics, vol 5, p 47. New York, Academic Press. 3. Pethig R. 1979. Dielectric and Electronic Properties of Biological Materials. New York, Wiley. 4. Grant EH, Sheppard RJ, South GP. 1978. Dielectric Behavior of Biological Molecules in Solution. Oxford, Oxford University Press. 5. Schanne OF, P.-Ceretti ER. 1978. Impedance Measurements in Biological Cells. New York, Wiley. 6. Duck FA. 1990. Physical Properties of Tissue. New York, Academic Press. 7. Geddes LA, Baker LE. 1967. The specific resistance of biological material—A compendium of data for the biomedical engineer and physiologist. Med Biol Eng 5:271. 8. Stuchly MA, Stuchly SS. 1980. Dielectric properties of biological substances—Tabulated. J Microwave Power 15:19. 9. Schwan HP. 1963. Determination of biological impedances. In G Oster et al (eds), Physical Techniques in Biological Research, vol 6, p 323. New York, Academic Press. 10. Löfgren B. 1951. The electrical impedance of a complex tissue and its relation to changes in volume and fluid distribution. Acta Physiol Scand 23(suppl 81):1. 11. Fallert MA, Mirotznik MS, Bogen DK, et al. 1993. Myocardial electrical impedance mapping of ischemic sheep hearts and healing aneurysms. Circulation 87:188. 12. McRae DA, Esrick MA. 1992. The dielectric parameters of excised EMT-6 tumors and their change during hyperthermia. Phys Med Biol 37:2045. 13. McRae DA, Esrick MA. 1993. Changes in electrical impedance of skeletal muscle measured during hyperthermia. Int J Hyperthermia 9:247. 14. Esrick MA, McRae DA. 1994. The effect of hyperthermia-induced tissue conductivity changes on electrical-impedance temperature mapping so physics in medicine and biology. Phys Med Biol 39:133. 15. Fricke H, Curtis HJ. 1935. The electric impedance of hemolyzed suspensions of mammalian erythrocytes. J Gen Physiol 18:821. 16. Herrick JF, Jelatis DG, Lee GM. 1950. Dielectric properties of tissues important in microwave diathermy. Fed Proc Fed Am Soc Exp Biol 9:60 (abstract only; data summarized in ref. 2). 17. England TS, Sharples NA. 1949. Dielectric properties of the human body in the microwave region of the spectrum. Nature 163:487. 18. Schwan HP. 1941. Über die Niederfrequenzleitfahigkeit von Bluten und Blutserum bei verschiedenen Temperaturen. Z Ges Exp Med 109:531. 19. Epstein BR, Foster KR. 1983. Anisotropy in the dielectric properties of skeletal muscle. Med Biol Eng Comput 21:51. 20. Stoy RD, Foster KR, Schwan HP. 1982. Dielectric properties of mammalian tissues from 0.1 to 100 MHz: A summary of recent data. Phys Med Biol 27:501. 21. Schepps JL, Foster KR. 1980. The UHF and microwave dielectric properties of normal and tumor tissues: Variation in dielectric properties with tissue water content. Phys Med Biol 25:1149. 22. Schepps JL. 1980. The measurement and analysis of the dielectric properties of normal and tumor tissues at UHF and microwave frequencies, Ph.D. dissertation, University of Pennsylvania, Philadelphia. 23. Schwan HP, Kay CF. 1957. The conductivity of living tissues. Ann NY Acad Sci 65:1007. 24. Stuchly MA, Athey TW, Stuchly SS, et al. 1981. Dielectric properties of animal tissues in vivo at frequencies 10 MHz-1 GHz. Bioelectromagnetics 2:93. 25. Brady MM, Symonds SA, Stuchly SS. 1981. Dielectric behavior of selected animal tissues in vitro at frequencies from 2 to 4 GHz. IEEE Trans Biomed Eng BME-28:305.


26. Schwan HP, Li K. 1953. Capacity and conductivity of body tissues at ultrahigh frequencies. Proc IRE 41:1735. 27. Kraszewski A, Stuchly MA, Stuchly SS, Smith AM. 1982. In vivo and in vitro dielectric properties of animal tissues at radiofrequencies. Bioelectromagnetics 3:421. 28. Kosterich JD, Foster KR, Pollack SR. 1983. Dielectric permittivity and electrical conductivity of fluid saturated bone. IEEE Trans Biomed Eng BME-30:81. 29. Nightingale NRV, Goodridge VD, Sheppard RJ, Christie JL. 1983. The dielectric properties of cerebellum, cerebrum, and brain stem of mouse brain at radiowave and microwave frequencies. Phys Med Biol 28:897. 30. Pfutzner H. 1984. Dielectric analysis of blood by means of a raster-electrode technique. Med Biol Eng Computing 22:142. 31. Foster KR, Schepps JL, Stoy RD, Schwan HP. 1979. Dielectric properties of brain tissue between 0.01 and 10 GHz. Phys Med Biol 24:1177. 32. Gabriel S, Lau RW, Gabriel C. 1996. The dielectric properties of biological tissues. 2. Measurements in the frequency range 10 Hz to 20 GHz. Phys Med Biol 41:2251.


Stuchly, M. A. “Low-Frequency Magnetic Fields:Dosimetry, Cellular, and Animal Effects.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

90 Low-Frequency Magnetic Fields: Dosimetry, Cellular, and Animal Effects 901. 90.2

Introduction Physical Interactions—Dosimetry Basic Principles • People and Animals • Cells and Cell Assemblies

90.3

Maria A. Stuchly University of Victoria

Biological Effects Human Data • Laboratory Animals • Cellular Systems

90.4

Concluding Remarks

90.1 Introduction Low frequency magnetic fields are of interest to a biomedical engineer for at least two reasons. They have found applications in a few diagnostic procedures and their therapeutic applications have been growing. The diagnostic applications include the use of strong pulses of magnetic fields for neural stimulation. Another well-established diagnostic use is in magnetic resource imaging. Therapeutic applications are reviewed in Chapter 91. Information on other applications can be found in a review by Stuchly [1990]. The other reason for the importance of low frequency magnetic fields is their potential impact on human health. This is the area of considerable research activities, public concern, and a subject of this section. Low frequency magnetic fields, for the purpose of this review, mostly refer to extremely low frequency (ELF) fields, but the techniques and interactions extend to higher frequencies up to a few kilohertz. The impetus for the inquiry into interactions of ELF magnetic fields with living systems has been provided by epidemiological reports (for a summary of these reports see NAS [1996] and NIEHS [1998]). The epidemiological studies have not provided convincing evidence that ELF, or more specifically the power line frequency magnetic fields are associated with increases in cancer rates. They have shown some supportive evidence for possible increases in childhood leukemia and some adult leukemia in those exposed occupationally. In an overview that follows, attention is focused on physical interactions of ELF magnetic fields with living systems and reported effects in cellular and animal studies. One general conclusion that can be drawn is that low frequency magnetic fields at moderate to low levels are biologically active, i.e., interact with various cells or systems of the animal body. What, however, is not known are the biophysical mechanisms responsible for the interactions observed, and parameters of the field and biosystem that are essential in eliciting the interactions. It has also not been resolved what strengths of the field are required to elicit an effect. Several reviews and reports provide detailed information on the subject [NAS, 1996; NIEHS, 1998; Moulder, 1998].


90.2 Physical Interactions—Dosimetry Basic Principles ELF magnetic fields interact with conductive bodies such as tissues by inducing electric fields and currents inside them. While in general, all four Maxwell’s equations have to be satisfied, in the case of ELF fields, certain simplifications apply. As discussed in the preceding chapter, tissues are conductive and nonmagnetic (their bulk magnetic permeability is equal to that of free space). Consideration of the dimensions of biologic objects (e.g., people) in comparison with wavelengths at ELF, the tissue conductivity as compared with the dielectric constant, lead to the following conclusions. Provided that:

σ ωε >> 1

(90.1)

fµ 0σL2 500 W/m2) results in varying degrees of testicular damage, such as edema, necrosis of seminiferous tubules, and atrophy, at 2.45, 3, and 10 GHz. Lebovitz et al. [23] compared pulse-modulated 1.3-GHz microwaves with conventional heating in rats. The authors stated that all damage could be explained by heating. Although the general consensus is that damage is thermally mediated, Cleary et al. [24] reported decreases in sperm function in vitro utilizing suspension of mouse sperm exposed to SARs of 50 W/kg or greater but with stringent temperature control. Effects on embryonic development have been studied. Thermal stress appears to be the primary mechanism by which RF energy absorption exerts a teratogenic action. Chernovetz et al. [25] and others have pointed out evidence that indicates that increases in mortality and resorption are probably related to peak body temperature and its duration regardless of the method by which the temperature elevation is elicited. Rugh and McManaway [26] were able to prevent the increase in incidence of teratogenic


activity, which they had previously reported, by lowering the maternal body temperature through controlled use of pentobarbital anesthesia. The most common result from many studies [27] appears to be a reduced or retarded gain of body mass, thermally mediated.

92.3 Effects on the Nervous System Guy and Chou [28] studied response thresholds in rats for short, high-intensity microwave pulses. Rats were exposed to single 915-MHz RF pulses of 1-µs to 300-ms duration. No reaction occurred until the SAR rose to 29 kJ/kg, which correlated with a temperature rise in the brain of 8°C. Seizures occurred followed by unconsciousness for 4 to 5 minutes. Complete recovery ensued. Postmortem examination revealed some demyelination of neurons 1 day following exposure and focal gliosis 1 month after exposure. More intense exposures could produce a fatal outcome. Brown et al. [29] looked for response thresholds at lower energies, with particular attention to the onset of an involuntary, generalized muscular contraction that did not appear to be injurious. Studies of low-level chronic effects, mostly dating to the 1960s and 1970s, have yielded inconsistent results. Problems of dosimetric measurement and quantitation of reproducible biologic endpoints have left uncertain whether extended exposure to low-intensity or nonthermal levels of radiofrequency energy produces effects, adverse or not. Bawin et al. [30] reported that electromagnetic energy of 147 MHz, amplitude modulated, at brain wave frequencies (8 to 16 Hz), influenced spontaneous and conditioned EEG patterns in the cat at 10 W/m2. No effects were seen at other modulation frequencies. At high field intensities, when death is a result of hyperthermia, pathologic changes are identical to those of hyperthermia. At lower levels of exposure, there are no changes that are specific to RF radiation. From the work of Albert and DeSantis [31], there appears to be a threshold for permanent histologic damage to the brain for exposures lasting several hours per day for up to 3 weeks at between 100 and 250 W/m2 at 2.45 GHz CW. An historically controversial issue has been the effect of radiofrequency radiation on the blood-brain barrier (BBB). This poorly understood functional “barrier” provides resistance to movements of largemolecular-weight, fat-insoluble substances from the blood vessels of the brain into brain tissue, presumably to protect the brain from invasion by various blood-borne pathogens and toxic substances. Early reports asserted that RF radiation reduces the barrier, allowing many substances normally barred from brain tissue to enter including many drugs. It does appear that gross hyperthermia (brain tissue elevated to 43°C) compromises the barrier [32]. But there is evidence that the level of a drug entering the brain might actually be reduced at “moderate” levels of hyperthermia [32]. A number of studies on isolated nerve preparations have been performed. There is no direct production of a nerve impulse or action potential by CW or PW microwaves, but conduction velocity and amplitude can be changed, mediated by temperature elevations of at least 1°C in the solution. Similar changes can be produced by ambient changes in temperature [33].

92.4 Behavioral Effects Studies have been conducted on the effects of RFR on performance of trained tasks or operant behavior by rats and rhesus and squirrel monkeys [34,35]. All the studies indicated that the exposure would result in suppressed performance of the trained task and that an energy/dose threshold for achieving the suppression existed. Depending on duration and other parameters of exposure, the threshold power density for affecting trained behavior ranged between 50 and 500 W/m2. Lebovitz [36] and Akyel et al. [37] were unable to find any specific effect attributed to pulse power and noted that interference with behavior appeared to be of thermal origin. Raslear et al. [39] and Akyel et al. [38] later found performance deficits at very high peak, short-pulse powers at low repetition rates, such that no measurable temperature changes occurred in the brains of rats. Peak brain SAR was reported to be 7 MW/kg for 80-ns-wide pulses with an average brain SAR of 0.07 W/kg.


92.5 Effects on the Cardiovascular and Hematopoietic Systems At nonthermalizing levels of exposure, both bradycardia and tachycardia have been found in different studies of different animals and with inconsistent results in the same animal. The specific conditions of exposure, biologic variability, and other sources of error could account for the findings. Hyperthermia of RFR or non-RFR origin produces tachycardia and a decrease in total peripheral resistance caused by vasodilation, a heat dissipating response to the thermal burden. Variations from this general principle have been found in unusual circumstances, including the application of localized pulse power to the head, neck, and thoracic region [40]. Changes in the concentrations of circulating white blood cells have been observed in a number of animal species exposed to microwave energy. The changes were not consistent within or between species and depended on exposure conditions and thermal changes in tissue. A number of mechanisms have been proposed to explain changes in cellular dynamics, including stimulation of synthesis at thermal levels, recirculation of sequestered cells, and increased hypothalamic-hypophysial-adrenal function following thermal stress. Such changes may be related to changes in immune function noted by a number of investigators [41].

92.6 Auditory and Ocular Effects Microwave hearing, the perception of a clicking or buzzing in the presence of pulsed microwave energy (but not CW) at low power densities [42], has been attributed to thermoelastic transduction of pulsed microwaves in the head, with detection by the sensory epithelium of the cochlea. The process, familiar to occupationally exposed workers, appears not to be harmful at energies commonly encountered [43]. A threshold for cataract production in the lens of the rabbit eye has been found for 2.45-GHz microwaves at 1.5 kW/m2 for 100 minutes. An intraocular temperature elevated to at least 45°C appears to be required [44]. For exposures lower than this there does not appear to be a cumulative effect from microwave exposure, i.e., no pathologic damage following many exposures with time, for which each individual exposure produces no detectable damage.

92.7 Conclusion Elucidation of the biologic effects of RF exposure requires study of the available literature. Evaluating the research is a difficult task even for scientists specializing in the field. The possible sources of error of the biologic sciences are coupled with the sources of error associated with RF engineering and dosimetry. It is often difficult to make meaningful comparisons between studies. The issue of effects or even hazards at “nonthermal” levels of exposure stimulates continuing debate and research that may affect existing consensus safety standards. Whether existing standards are adequate frequently rests on a determination of whether documented effects constitute actual hazards at energy levels not producing morbidity. The research of the future will continue to affect the issue of safe levels of exposure.

Acknowledgments I wish to thank Mrs. Doris Michaelson for her support and permission to base this review on Dr. S. Michaelson’s monograph that appeared in the Handbook of Biological Effects of Electromagnetic Fields, published by CRC Press in 1986. The opinions or assertions contained are private views of the author and are not to be construed as reflecting the official views of the Department of the Army or the Department of Defense.


Defining Terms Blood-brain barrier: An anatomical and physiologic barrier to the movement of large-molecularweight, fat-insoluble substances from the blood vessels of the brain into brain tissue. Bradycardia: An abnormal slowness of the heartbeat, as evidenced by slowing of the pulse rate to 60 or less. Radiofrequency radiation (RFR): Defined by the Institute of Electrical and Electronics Engineers (IEEE) as that part of the electromagnetic spectrum extending from 3 kHz to 300 GHz. Relative absorption cross section (RAC): The ratio of the absorbed power to the power incident on the geometrical cross-sectional area of an animal. Tachycardia: An excessive rapidity in the action of the heart, usually applied to a pulse rate above 100 per minute.

References 1. Biological Effects of Radiofrequency Radiation, EPA-600/8-83-026F, Health Effects Research Laboratory, United States Environmental Protection Agency, Research Triangle Park, NC, 1984, pp 5–76. 2. Pere JR, Foster KR, Blick DW, Adair ER. 1997. A thermal model for human thresholds of microwave-evoked warmth sensations. Bioelectromagnetics 18:578–583. 3. Durney CH, Johnson CC, Barber PW, et al. 1978. Radiofrequency Radiation Dosimetry Handbook, 2d ed. USAF Report SAM-TR-78-22, Brooks Air Force Base, Texas. 4. Gandhi OP. 1975. Conditions of strongest electromagnetic power deposition in man and animals. IEEE Trans Microwave Theory Technol 23:1021. 5. Gandhi OP. 1980. State of knowledge for electromagnetic absorbed dose in man and animals. Proc IEEE 68:24. 6. Gandhi OP, Hunt EL, D’Andrea JA. 1977. Deposition of electromagnetic energy in animals and in models of man with and without grounding and reflector effects. Radio Sci 12(6S):39. 7. Guy AW. 1974. Quantitation of induced electromagnetic field patterns and associated biologic effects. In P Czerski et al (eds), Biologic Effects and Health Hazards of Microwave Radiation, p. 203, Warsaw, Polish Medical Publishers. 8. Johnson CC, 1975. Recommendations for specifying EM wave irradiation conditions in bioeffects research. J Microwave Power 10:249. 9. Anne A, Saito M, Salati OM, Schwan HP. 1962. Relative microwave absorption cross sections of biological significance. In MF Peyton (ed), Proceedings of the 4th Annual Tri-Service Conference on the Biological Effects of Microwave Radiating Equipment; Biological Effects of Microwave Radiation, New York, Plenum Press. 10. Gandhi OP. 1990. Electromagnetic energy absorption in humans and animals. In OP Gandhi (ed), Biological Effects and Medical Applications of Electromagnetic Energy, p 175. Englewood Cliffs, New Jersey, Prentice-Hall. 11. Anderstam B, Hamnerius Y, Hussain S, Ehrenberg L. 1983. Studies of possible genetic effects in bacteria of high frequency electromagnetic fields. Hereditas 98:11. 12. Varma MM, Traboulay EA. 1976. Evaluation of dominant lethal test and DNA studies in measuring mutagenicity caused by non-ionizing radiation. In CC Johnson, ML Shore (eds), Biological Effects of Electromagnetic Waves, vol 1, p 386. Publication (FDA) 77-8010, U.S. Department of Health Education, and Welfare, Rockville, MD. 13. Mittler S. 1976. Failure of 2 and 20 meter radio waves to induce genetic damage in Drosophila melanogaster. Environ Res 11:326.


14. Livingston GK, Johnson CC, Dethlefsen LA. 1977. Comparative effects of water bath and microwave-induced hyperthermia on cell survival and sister chromatid exchange in Chinese hamster ovary cells. In Abstracts of the International Symposium on the Biological Effects of Electromagnetic Waves, Airlie, VA, p 106. 15. McRee DI, MacNichols G, Livingston GD. 1981. Incidence of sister chromatid exchange in bone marrow cells of the mouse following microwave exposure. Radiat Res 85:340. 16. Stodolnik-Baranska W. 1974. The effects of microwaves on human lymphocyte cultures. In P Czerski et al (eds), Biologic Effects and Health Hazards of Microwave Radiation, p 189. Warsaw, Polish Medical Publishers. 17. Chen KM, Samuel A, Hoopingavner R. 1974. Chromosomal abberrations of living cells induced by microwave radiation. Environ Lett 6:37. 18. Balcer-Kubiczek EK, Harrison GH. 1985. Evidence for microwave carcinogenesis in vitro. Carcinogenesis 6:859. 19. Chou C-K, Guy AW, Kung LL, et al. 1992. Long-term, low-level microwave irradiation of rats. Bioelectromagnetics 13:469. 20. Ely TS, Goldman D, Hearon JZ, et al. 1964. Heating characteristics of laboratory animals exposed to ten centimeter microwaves. U.S. Naval Medical Research Institute (res. rep. proj. NM 001056.13.02), Bethesda, MD. IEEE Trans Biomed Eng 11:123. 21. Gorodetskaya SF. 1963. The effect of centimeter radio waves on mouse fertility. Fiziol Zh 9:394. 22. Imig CJ, Thomson JD, Hines HM. 1948. Testicular degeneration as a result of microwave irradiation. Proc Soc Exp Biol 69:382. 23. Lebovitz RM, Johnson L, Samson WK. 1987. Effects of pulse-modulated microwave radiation and conventional heating on sperm production. J Appl Physiol 62:245. 24. Cleary SF, Liu LM, Graham R, East J. 1989. In vitro fertilization of mouse ova by spermatazoa exposed isothermally to radio-frequency radiation. Bioelectromagnetics 10:361. 25. Chernovetz ME, Justesen DR, King NW, Wagner JE. 1975. Teratology, survival, and reversal learning after fetal irradiation of mice by 2450 MHz microwave energy. J Microwave Power 10:391. 26. Rugh R, McManaway M. 1976. Can electromagnetic waves cause congenital anomalies? In International IEEE/AP-S USN/URSI Symposium, Amherst, MA, p 143. 27. O’Connor ME. 1980. Mammalian teratogenesis and radiofrequency fields. Proc IEEE 68:56. 28. Guy AW, Chou CK. 1982. Effects of high-intensity microwave pulse exposure of rat brain. Radio Sci 17(5S):169. 29. Brown DO, Lu S-T, Elson EC. 1994. Characteristics of microwave evoked body movements in mice. Bioelectromagnetics 15:143. 30. Bawin SM, Adey WR. 1977. Calcium binding in cerebral tissue. In DG Hazzard (ed), Symposium on Biological Effects and Measurement of Radio Frequency/Microwaves. Publication (FDA) 778026, U.S. Department of Health, Education and Welfare, Rockville, MD. 31. Albert EN, DeSantis M. 1975. Do microwaves alter nervous system-structure? Ann NY Acad Sci 247:87. 32. Williams WM, Lu S-T, Del Cerro M, et al. 1984. Effects of 2450 MHz microwave energy on the blood-brain barrier: An overview and critique of past and present research. IEEE Trans Microwave Theory Technol 32:808. 33. Chou C-K, Guy AW. 1978. Effects of electromagnetic fields on isolated nerve and muscle preparations. IEEE Trans Microwave Theory Technol 26:141. 34. Gage MI, Guyer WM. 1982. Interaction of ambient temperature and controlled behavior in the rat. Radio Sci 17(5S):179. 35. deLorge JO. 1983. Operant behavior and colonic temperature of Rhesus monkeys, Macaca mulatta, exposed to microwaves at frequencies above and near whole-body resonance. Report no. NAMRL1289, Naval Aerospace Medical Research Lab., Pensacola, FL. 36. Lebovitz RM. 1983. Pulse modulated and continuous wave microwave radiation yield equivalent changes in operant behavior of rodents. Physiol Behav 30:391. © 2000 by CRC Press LLC

37. Akyel Y, hunt EL, Gambrill C, Vargas C Jr. 1991. Immediate post-exposure effects of high-peakpower microwave pulses on operant behavior of Wistar rats. Bioelectromagnetics 12:183. 38. Akyel Y, Belt M, Raslear TG, Hammer RM. 1993. The effects of high-peak power pulsed microwaves on treadmill performance. In M Blank (ed), Electricity and Magnetism in Biology and Medicine, p 668. San Francisco, San Francisco Press. 39. Raslear TG, Akyel Y, Bates F, et al. 1993. Temporal bisection in rats: The effects of high-peak power pulsed microwave irradiation. Bioelectromagnetics 14:459. 40. Lu S-T, Brown DO, Johnson CE, et al. 1992. Abnormal cardiovascular responses induced by localized high power microwave exposure. IEEE Trans Biomed Eng 39:484. 41. Budd RA, Czerski P. 1985. Modulation of mammalian immunity by electromagnetic radiation. J Microwave Power Electromagnet Energy 20:217. 42. Frey AH. 1962. Human auditory system response to modulated electromagnetic energy. J Appl Physiol 17:689. 43. Lin JC. 1991. Pulsed radiofrequency field effects in biological systems. In JC Lin (ed), Electromagnetic Interaction with Biological Systems, pp 165–177. New York, Plenum Press. 44. Kramar PO, Guy AW, Emergy AF, et al. 1976. Quantitation of microwave radiation effects on the eyes of rabbits and primates at 2450 MHz and 918 MHz. University of Washington, Bioelectromagnetics Research Laboratory Scientific report no. 6, Seattle, WA.

Further Information The reader is encouraged to consult the Handbook of Biological Effects of Electromagnetic Fields, edited by C. Polk and E. Postow, (CRC Press). Bioelectromagnetics, the journal of the Bioelectromagnetics Society and the Society for Physical Regulation in Biology and Medicine, is the leading publication of contemporary research on the biologic effects of electromagnetic radiation.


93 Radio Frequency Hyperthermia in Cancer Therapy C.K. Chou Motorola Inc.

Rulong Ren

93.1 93.2

Introduction Methods of EM Heating

City of Hope National Medical Center

93.3

Conclusion

Local Heating • Regional Heating • Whole Body Heating

93.1 Introduction Cancer hyperthermia is a treatment to raise the tissue temperature either locally or whole body to a therapeutic level to eradicate tumors. Over the last three decades, much has been learned about the effects of heat on cells and the interactions between heat and radiation and chemotherapeutic agents [1-4]. The scientific rationale for its use either alone or combined with other methods is multifactorial, and new justifications for its use are continuously being identified. For example, heat may be directly cytotoxic to tumor cells or inhibit repair of both sublethal and potentially lethal damage after radiation. Hyperthermia has been used in combination with chemotherapy because heating increases membrane permeability and the potency of some drugs. The synergism of radiation and hyperthermia is accomplished by the thermal killing of hypoxic and S phase (DNA syntheses) cells which are resistive to radiation alone. Although the biologic rationale is strong, and hyperthermia has been studied in phase I-III trials, the results of phase III clinical trials are controversial and the use of hyperthermia is still a developing project in the U.S. [4-8]. Numerous factors could affect the results of hyperthermia. The foremost problem in hyperthermia, however, is the generation and control of heat in tumors. Current heating methods include whole body heating, using hot wax, hot air, hot water suits, or infrared radiation, and partial body heating, utilizing ultrasound, heated blood, fluid perfusion, radio frequency (RF) fields, or microwaves. The effective temperature range of hyperthermia is very small: 42 to 45°C. At lower temperatures, the effect is minimal. At temperatures higher than 45°C, normal cells are damaged. Due to this narrow temperature range, the response rate of the tumor is highly dependent on how much of it is heated to a therapeutic level. The clinical use of hyperthermia has been hampered by a lack of adequate equipment to effectively deliver heat to deep-seated and even large superficial lesions and by a lack of thermometry techniques that provide reliable information on heat distribution in the target tissues. In RF hyperthermia, the final temperature of tumors is mainly dependent on energy deposition. When electromagnetic (EM) heating methods are used, the energy deposition is a complex function of the frequency, intensity, and polarization of the applied fields, the applicator’s size and geometry, as well as the size, depth, geometry, and dielectric property of the tumor [9,10]. The material, thickness, and construction of a cooling bolus also influence the amount of energy deposition. In this chapter, the main methods of RF hyperthermia will be reviewed.


93.2 Methods of EM Heating The EM energy used in hyperthermia is usually classified by frequency as either microwave energy or RF energy. Microwaves occupy the EM frequency band between 300 MHz and 300 GHz. Strictly speaking, RF is between 3 kHz and 300 GHz, but for hyperthermia it generally refers to frequencies below the microwave range. The most commonly used microwave frequencies in hyperthermia are 433, 915, and 2450 MHz, which are the designated ISM (industrial, scientific, and medical) frequencies in the U.S. and Europe. Common RF frequencies are 13.56 and 27.12 MHz, which have also been widely used in diathermy. Frequencies higher than 2450 MHz have no practical value due to their limited penetrations. At lower frequencies field penetration is deeper, but the applicator must be larger and focusing is difficult. Despite these limitations, EM heating methods have been developed for local, regional, and whole body hyperthermia.

Local Heating External The superficial cooling mechanism of tissue makes deep heating difficult by conductive methods. Two RF methods have been used to provide subcutaneous heating. For the first method, tissues are placed between two capacitor plates and heated by displacement currents. This method is simple, but overheating of the fat, caused by the perpendicular electric field, remains a major problem for obese patients. In planar tissue models, the rate of temperature rise is about 17 times greater in fat than in muscle, due to the large differences in their dielectric properties and specific heats [11,12]. Additionally, blood flow in fat is significantly less than that in muscle. Therefore, the final fat temperature is much higher than the muscle temperature, and a water bolus is necessary to minimize the fat heating. The second RF method uses solenoidal loops or “pancake” magnetic coils to generate a magnetic field. This field then produces heat in tissue by inducing eddy currents. Since the induced electric fields are parallel to the tissue interface, heating is maximized in muscle rather than in fat. However, the heating pattern is generally toroidal with a null at the center of the coil. Fujita et al. [13] described a pairedaperture type inductive applicator to produce deep heating in phantom. In the microwave frequency range, energy is coupled into tissues by waveguides, dipoles, microstrips, or other radiating devices. The shorter wavelengths of microwaves, as compared to RF wavelengths, provide the capability to direct and focus energy into tissues by direct radiation from a small applicator. Many applicators of various sizes operate over a frequency range of 300 to 1000 MHz [14-16]. Most of them are dielectrically loaded and have a water bolus for surface cooling. Low profile, lightweight microstrip applicators, which are easier to use clinically, have also been reported [17,18]. High permittivity dielectric materials, electric wall boundaries, and magnetic materials have been used to reduce applicator size and weight. For the most part, these applicators are used for treatment of tumors a few centimeters below the skin. Toxicities usually associated with treatment are pain and thermal blistering. Intracavitary Certain tumor sites at hollow visceras or cavities may be treated by intracavitary techniques. The advantages of intracavitary hyperthermia over external hyperthermia include better power deposition due to the proximity of the applicators to the tumors and the reduction of normal tissue exposure. There have been clinical and research studies on hyperthermia and radiation or chemotherapy of the esophagus, rectum, cervix, prostate, nasopharyngeal, and bladder cancers. Chou et al. [19] designed an esophageal applicator for intracavitary 915 MHz hyperthermia and intracavitary radiation. In a phantom test, the result showed that the effective heating length (>50% peak heating rate) is more than 12.0 cm. This applicator was used for a pilot study. The esophageal lesion was circumferentially and uniformly heated [20]. A microwave applicator with an integrated cooling system for intracavitary hyperthermia of vaginal carcinoma was recently developed in Sweden [21]. The advantage of this applicator is that it can lead to a highly targeted heating of tumors and a reduced risk of unwanted © 2000 by CRC Press LLC

heating of normal tissues. Li et al. [22] developed a 915 MHz applicator for cervical and upper vaginal cancers and a 2450 MHz applicator for uterine cancer. The average temperature distributions in the upper vagina, cervix, and uterus was 46 ± 2.1°C. In Italy, a computerized 915 MHz microwave system was designed to deliver simultaneously local bladder hyperthermia and intravesical chemotherapy for superficial bladder cancer [23]. The bladder walls could be directly heated by using a transurethral catheter. A cooled microwave transrectal applicator was used for prostate treatment [24]. The applicator was found capable of effectively heating a tissue volume extending radially 3 to 25 mm from the applicator surface. An intracavitary microwave antenna array system has also been developed for prostate cancer hyperthermia [25]. In addition, special applicators for nasopharyngeal cancer [26] and rectal cancer [27] were developed. Both microwave and RF energy have been used for intracavitary hyperthermia. The main problem, however, is that the tumor temperature is unknown. Most temperatures have been measured on the surface of the applicators, which can be very different from those in the tumor. Furthermore, many investigators have used thermocouples or thermistors to measure temperatures, not knowing the perturbation problem caused by the metallic sensors [28]. One solution to this problem is to measure tissue temperature in animals and then extrapolate to humans [19]. Recently, the MR thermography was developed for noninvasive monitoring of temperature distributions of deep-seated tumors [29]. Interstitial Interstitial techniques for radiation implants as primary or boost treatments have been practiced successfully by radiation oncologists for many years. When hyperthermia was learned to be cytotoxic and synergistic with radiation, it was natural to consider this combination with conventional interstitial radioactive implantation. The advantages of this technique over external hyperthermia include confined treatment volume, better sparing of normal tissue, accessibility of deeper tumors, more homogeneous therapeutic temperature distribution, and better control and evaluation of thermal parameters [30]. Interstitial hyperthermia has been used for various site tumors; however, a phase III study of interstitial thermoradiotherapy for recurrent or persistent human tumors did not show any additional beneficial effects over interstitial radiation alone. The delivery of hyperthermia remains a major obstacle [6, 31-34]. Methods such as resistive heating, the microwave technique, or ferromagnetic seed implants can be used for interstitial hyperthermia. With resistive heating, multi-electrodes inserted in plastic tubes were implanted in the treatment volume. The mean power deposition of the individual electrodes was controlled by varying the duty cycle of the RF signal applied to the electrodes [35]. To prevent excitation of nerve action potentials, an operating frequency greater than 100 kHz should be used. The microwave technique utilizes small microwave antennas inserted into hollow plastic tubing to produce interstitial heat. In the U.S., 915 MHz is a commonly used frequency for this technique. However, satisfactory heating patterns can be produced between 300 and 2450 MHz [36]. A small coaxial antenna can irradiate a volume of approximately 60 cc. It is necessary to use an array of microwave antennas to heat a large tumor. In a brain hyperthermia study, the array of four dipole antennas spaced 2.0 cm apart were capable of heating a volume of 5.9 cm × 2.8 cm × 2.8 cm [37]. As in RF resistive hyperthermia, the degree of control of microwave power radiating from these antennas is important in order to achieve homogeneous heating. Since the antennas couple to each other, the spacing, phasing, and insertion depth affect the heating patterns of array applicators [38,39]. Interstitial heat can also be produced by using ferromagnetic seed implants. This technique is applicable for delivering thermal energy to deep-seated tumors. The seed characteristics and implant geometry must be determined prior to the treatment [40]. When exposed to RF magnetic fields (~100 kHz), the implants absorb power and become heated until they reach the Curie point. Here, the implants become nonferromagnetic and no longer produce heat. The surrounding tissues are then heated by thermal conduction. The influence of blood flow and tissue inhomogeneities of the tumor, which may affect the temperature distribution, can be compensated by the self-regulation of the implants. It is therefore possible to maintain a temperature close to the Curie point [41,42]. Another method, which exposes magnetic fluid in a tumor to an RF magnetic field (0.3 to 80 MHz), has also been shown to be feasible for inducing selective heating [43]. Cetas et al. [44] developed an alterative form of ferromagnetic seed. A nearly © 2000 by CRC Press LLC

lossless ceramic ferrite core (FC) is surrounded by an electrically conductive metallic sheath (MS). The authors indicated that the FC/MS configuration solves many of the technical problems that have hindered the clinical implementation of thermally regulating ferromagnetic implants for thermal therapies.

Regional Heating Electric Field Heating deep-seated tumors is difficult. RF energy can be deposited into the center of the body, but a large region is affected. Differential increases of blood flow in normal and tumor tissues may result in higher temperatures in the tumor than in normal organs. However, this temperature differential cannot be assured. Strohbehn [45] used the term “dump and pray” to describe the situation of putting large amounts of EM energy into the region and hoping for satisfactory results. In Japan, the 8 MHz Thermotron system uses a capacitive electric field to heat deep tumors and a water-cooled bolus to minimize the heating of fat tissue. The sizes of two electrodes are adjusted to control the heating patterns in patients. Most other electric field heating systems generate electric fields parallel to the body surface. These include the annular phased array systems, the CDRH helix system, the coaxial TEM cell, the ring electrode, the segmented cylindrical array, the toroidal inductor, and the loosely coupled TEM system. The APAS, made by the BSD Company (Salt Lake City, Utah), radiates 16 RF fields in phase toward the patient. These systems, with variations in phase, frequency, amplitude, and orientation of the applied fields, can add more dimensions to the control of heating patterns during treatment. To determine the excitation phases of an array for heating an inhomogeneous medium, the retro-focusing technique was applied [46]. A small probe was first inserted into a tumor. A signal was radiated from the probe and received by the array of applicators outside the patient. By the reciprocity theory, conjugate fields were radiated from the applicators and focused at the tumor. The technique was demonstrated experimentally in a water tank. A significant power increase at the desired focus was observed. In general, superficial heating and hot spots in normal tissues are the limiting factors of treatment effectiveness in the existing systems. Invasive techniques using interstitial hyperthermia have been shown to solve some of the deep heating problems. However, no adequate deep EM heating system is available. Scanned ultrasound provides an alternative method [47]. Magnetic Field Magnetic fields heat tissue by induced eddy currents. The magnetic-loop applicators of the Magnetrode unit (Henry Radio, Los Angeles, CA) are self-resonant, non-contact cylindrical coils with built-in impedance matching circuitry; they operate at 13.56 MHz with a maximum power output of 1000 W. The RF current in the coil creates strong magnetic fields that are parallel to the center axis of the coil where the body or limb of a patient is located. Since the magnitude of the induced eddy current is a function of the radius of the exposed object, there is no energy deposition at the center of the exposed tissue. However, Storm et al. [48] showed that the heating of tumors deep in the body was possible as demonstrated in live dogs and humans, with no injury of surface tissue. This was apparently due to the redistribution of the thermal energy by blood flow. Nevertheless, the FDA has forbidden the use of the Magnetrode.

Whole Body Heating During the last 20 years, hyperthermia has been used primarily for treating localized tumors. However, tumors that are resistant to conventional therapy tend to be metastatic. For these patients, local and regional hyperthermia can only be palliative. For disseminated disease, whole body hyperthermia (WBH) in conjunction with chemotherapy and radiation has been studied by many groups [49,50]. Preclinical studies are consistent with the concept that 41.8°C whole body hyperthermia can enhance the therapeutic index of irradiation and specific chemotherapeutic agents without a commensurate increase in normal tissue toxicity, induce elevated plasma levels of granulcyte-colony stimulating tumor necrosis factoralpha, and other cytokine factors [51,52].


Methods of WBH include hot wax, water blankets, water suits, radiant heat, and extracorporeal blood heating. To heat the whole body uniformly with EM energy is impossible. However, it is possible to heat the body regionally so that the blood flow will redistribute the heat to the whole body. In the past, applications of 434 and 468 MHz microwaves have been explored in Europe for WBH, but the results did not produce any significant impact [53]. Several regional RF systems have been attempted for WBH. The BSD dipole system and the CDRH helix system require that the body be inserted into a tunnel applicator. The BSD system also requires a water bolus in contact with the patient to provide better energy coupling and skin cooling. For hour-long use, the bolus is very heavy and uncomfortable. The RF electric field system, designed by the UCLA group, uses three electrode plates to heat deep-seated tumors in the torso. The patient lies on a table and the top plate is swung over the abdomen to heat the thoracic region. No water bolus is needed and there is no contact with the patient. It is very simple to use. After two years of extensive phantom and animal studies, it was found to be a very promising system for regional and WBH [54].

93.3 Conclusion There has been much progress in the application of EM energy for clinical hyperthermia. Based on medical demands, many technological advances have been made. As a result, new forms of treatment equipment have been developed, and existing methods have been improved. It is impossible for a single piece of equipment to fulfill all of the clinical requirements for patient treatments. Depending on the location and vascularity of the tumor and adjacent tissues and the general physical condition of the patient, the hyperthermia practitioners should have the option of choosing the most appropriate equipment. Hyperthermia is a complicated technique and should be applied only by individuals well trained in its use. Due to the complexity of EM energy coupling to human tumors, careful heating pattern studies should be performed on all exposure geometries and contingencies prior to treatment to assure the best treatment conditions for the patient. Since hyperthermia in combination with high energy radiotherapy cannot be repeated after the tumor receives a maximal dosage of ionizing radiation, the physician must try to reach the critical tumor temperatures in optimal conjunction with radiotherapy. Accurate thermometry is particularly important in all phases of clinical hyperthermia, especially when the patient is anesthetized. The benefit of a good treatment outweighs minor risks. If there is no other choice, it would be more beneficial for the patient to have an effective treatment with a few blisters rather than a safe but ineffective treatment. It is easier to treat the burns than the cancer.

Acknowledgment Supported in part by NCI under Grants CA33572 and CA56116.

References 1. Field, S. B. and Hand, J. W., An Introduction to the Practical Aspects of Clinical Hyperthermia, Taylor & Francis, London, UK, 1990. 2. Kapp, D. S., From laboratory studies to clinical trials: Past benefits and future directions in hyperthermia, Int J Hyperthermia, 10, 355, 1994. 3. Engin, K., Biological rationale and clinical experience with hyperthermia, Control Clin Trials, 17, 316, 1996. 4. Dewhirst, M. W., Prosnitz, L., Thrall, D., Prescott, D., Clegg, S., Charles, C., MacFall, J., Rosner, G., Samulski, T., Gillette, E., LaRue, S., Hyperthermic treatment of malignant diseases: Current status and a view toward the future, Seminars in Oncology, 24, 616, 1997. 5. Perez, C. A., Pajak, T., Emami, B., Hornback, N. B., Tupchong, L., Rubin, P., Randomized phase III study comparing irradiation and hyperthermia with irradiation alone in superficial measurable tumors, Am J Clin Oncol, 14, 133, 1991.


6. Emami, B., Scott, C., Perez, C. A., Asbell, S., Swift, P., Grigsby, P., Montesano, A., Rubin, P., Curran, W., Delrowe, J., Arastu, H., Fu, K., Moros, E., Phase III study of interstitial thermoradiotherapy compared with interstitial radiotherapy alone in the treatment of recurrent or persistent human tumors. A prospectively controlled randomized study by the Radiation Therapy group., Int Radiat Oncol Biol Phys, 34, 1097, 1996. 7. Overgaard, J., Gonzalez, D. G., Hulshof, M. C. C. H., Arcangeli, G., Dahl, O., Mella, O., Bentzen, M., Hyperthermia as an adjuvant to radiation therapy of recurrent or metastatic malignant melonoma. A multicenter randomized trial by the European Society for Hyperthermic Oncology, Int J Hyperthermia, 12, 3, 1996. 8. Vernon, C. C., Hand, J. W., Field, S. B., Machin, D., Whaley, J. B., van der Zee, J., van Putten, W. L. J., van Rhoon, G. C., van Dijk, J. D. P., Gonzalez, D. G., Liu, F. F., Goodman, P., Sherar, M., Radiotherapy with or without hyperthermia in the treatment of superficial localized breast cancer: Results from five randomized controlled trials, Int J Radiat Oncol Biol Phys, 35, 731, 1996. 9. Chou, C. K. Safety considerations for clinical hyperthermia, in An Introduction to the Practical Aspects of Clinical Hyperthermia, Field, S. B. and Hand, J. W., Eds., Taylor & Francis Ltd., London, UK., 1990, chap. 18. 10. Chou, C. K., Evaluation of microwave hyperthermia applicators, Bioelectromag, 13, 581, 1992. 11. Guy, A. W. and Chou, C. K., Physical Aspects of Localized Heating by Radiowaves and Microwaves, in Hyperthermia in Cancer Therapy, Storm, F. K., Eds., GK Hall Medical Publishers, Boston, MA, 1983, 279. 12. Hand, J. W., Biophysics and technology of electromagnetic hyperthermia, in Methods of External Hyperthermic Heating, Gautherie, M., Eds., Springer-Verlag, Berlin, Germany, 1990, 1. 13. Fujita, Y., Kato, H., Ishida, T., An RF concentrating method using inductive aperture-type applicator, IEEE Trans Biomed Eng, 40, 110, 1993. 14. Johnson, R. H., Preece, A. W., Green, J. L., Theoretical and experimental comparison of three types of electromagnetic hyperthermia applicator, Phys Med Biol, 35, 761, 1990. 15. Nikawa, Y. and Okada, F., Dielectric loaded lens applicator for microwave hyperthermia, IEEE Trans Microwave Theory & Tech, 39, 1173, 1991. 16. Lee, E. R., Wilsey, T., Tarczys-Hornoch, P., Kapp, D. S., Fessenden, P., Lohrbach, A. W., Prionas, S. D., Body conformable 915 MHz microstrip array applicators for large surface area hyperthermia, IEEE Trans Biomed Eng, 39, 470, 1992. 17. Samulski, T. V., Fessenden, P., Lee, E. R., Kapp, D. S., Tanabe, E., McEuen, A., Spiral microstrip hyperthermia applicators: Technical design and clinical performance, Int J Radiat Oncol Biol Phys, 18, 233, 1990. 18. Cerri, G. and Marriani, V., Theoretical and experimental analysis of microstrip spiral antennas, in Italian Recent Advances in Applied Electromagnetics, Franceschetti, G., and Pierri, R., Eds., Liguori, Napoli, Italy, 1991, 195. 19. Chou, C. K., McDougall, J. A., Chan, K. W., Vora, N., Howard, H., Staud, C., Terr, L., Intracavitary hyperthermia and radiation of esophageal cancer, in Electricity and Magnetism in Biology and Medicine, Blank, M., Eds., San Francisco Press, San Francisco, CA, 1993, 793. 20. Ren, R. L., Chou, C. K., Vora, N., Luk, K., Vora, L., Ma, L., Ahn, C., Staud, C. L., Li, B., McDougall, J. A., Chan, K. W., Xiong, X. B., Li, D. J., A pilot study of intracavitary hyperthermia combined with radiation in the treatment of oesophageal carcinoma, Int J Hyperthermia, 14, 245, 1998. 21. Roos, D., Seegenschmiedt, M. H., Klautke, G., Erb, J., Sorbe, B., A new microwave applicator with integrated cooling system for intracavitary hyperthermia of vaginal carcinoma, Int J Hyperthermia, 12, 743, 1996. 22. Li, D. J., Chou, C. K., Luk, K. H., Wang, J. H., Xie, C. F., McDougall, J. A., Huang, G. Z., Design of intracavitary microwave applicators for the treatment of uterine cervix carcinoma, Int J Hyperthermia, 7, 693, 1991.


23. Colombo, R., Da Pozzo, L. F., Lev, A., Freschi, M., Gallus, G., Rigatti, P., Neoadjuvant combined microwave induced local hyperthermia and topical chemotherapy vs. chemotherapy alone for superficial bladder cancer, J Urol, 155, 1227, 1996. 24. Debicki, P. S., Okoniewski, M., Okoniewska, E., Shrivastava, P. N., Debicka, A. M., Baert, L. V., Petrovich, Z., Cooled microwave transrectal applicator with adjustable directional beam for prostate treatment, Int J Hyperthermia, 11, 95, 1995. 25. Yeh, M. M., Trembly, B. S., Douple, E. B., Ryan, T. P., Hoopes, P. J., Jonsson, E., Heaney, J. A., Theoretical and experimental analysis of air cooling for intracavitary microwave hyperthermia applicators, IEEE Trans Biomed Eng, 41, 874, 1994. 26. Zhong, Q. R., Chou, C. K., McDougall, J. A., Chan, K. W., Luk, K. H., Intracavitary hyperthermia applicators for treating nasopharyngeal and cervical cancers, Int J Hyperthermia, 6, 997, 1990. 27. Wust, P., Rau, B., Gellerman, J., Pegions, W., Loffel, J., Riess, H., Felix, R., Schlag, P.M., Radiochemotherapy and hyperthermia in the treatment of rectal cancer, Recent Results Cancer Research, 146, 175, 1998. 28. Cetas, T. C., Temperature, in Therapeutic Heat and Cold, Lehmann, J. F., Eds., Williams & Wilkins, Baltimore, MD, 1990, 1. 29. Wlodarczyk, W., Boroschewski, R., Hentschel, M., Wust, P., Monich, G., Felix, R., Three-dimensional monitoring of small temperature changes for therapeutic hyperthermia using MR, J Magn Reson Imaging, 8, 165, 1998. 30. Seegenschmiedt, M. H., Brady, L. W., Sauer, R., Interstitial thermoradiotherapy: Review on technical and clinical aspects, Am J Clin Oncol, 13, 352, 1990. 31. Kapp, K. S., Kapp, D. S., Stuecklschweiger, G., Berger, A., Geyer, E., Interstitial hyperthermia and high dose rate brachytherapy in the treatment of anal cancer: A phase I/II study, Int Radiat Oncol Biol Phys, 28, 189, 1994. 32. Stea, B., Rossman, K., Kittelson, J., Shetter, A., Hamiton, A., Cassady, J. R., Interstitial irradiation vs. interstitial thermoradiotherapy for supratentorial malignant gliomas: A comparative survival analysis, Int Radiat Oncol Biol Phys, 30, 591, 1994. 33. Peschke, P., Hahn, E. W., Wolber, G., Hildenbrand, D., Zuna, I., Interstitial radiation and hyperthermia in the Dunning R3327 prostate tumor model: Therapeutic efficacy depends on radiation dose-rate, sequence and frequency of heating, Int J Radiat Biol, 70, 609, 1996. 34. Djavan, B., Susani, M., Shariat, S., Zlotta, A. R., Silverman, D. E., Schulman, C. C., Marberger, M., Transperineal radiofrequency interstitial tumor ablation (RITA) of the prostate, Tech Urol, 4, 103, 1998. 35. van der Koijk, J. F., Crezee, J., Lagendijk, J. J., Thermal properties of capacitively coupled electrodes in iterstitial hyperthermia, Phys Med Biol, 43, 139, 1998. 36. Iskander, M. F. and Tumeh, A. M., Design optimization of interstitial antennas, IEEE Trans Biomed Eng, 36, 238, 1989. 37. Ryan, T. P., Trembly, B. S., Roberts, D. W., Strohbehn, J. W., Coughlin, C. T., Hoopes, P. J., Brain hyperthermia: I. Interstitial microwave antenna array techniques—the Dartmouth experience, Int J Radiat Oncol Biol Phys, 29, 1065, 1994. 38. Chan, K. W., Chou, C. K., McDougall, J. A., Luk, K. H., Changes in heating patterns of interstitial microwave antenna arrays at different insertion depths, Int J Hyperthermia, 5, 499, 1989. 39. Zhang, Y., Joines, W. T., Oleson, J. R., Prediction of heating patterns of a microwave interstitial antenna array at various insertion depths, Int J Hyperthermia, 7, 197, 1991. 40. Kotte, A. N., van Wieringen, N., Lagendijk, J. J., Modeling tissue heating with ferromagnetic seeds, Phys Med Biol, 43, 105, 1998. 41. Mack, C. F., Stea, B., Kittelson, J. M., Shimm, D. S., Sneed, P. K., Phillips, T. L., Swift, P. S., Luk, K., Stauffer, P. R., Chan, K. W., Steeves, R., Cassady, J. R., Cetas, T. C., Interstitial thermoradiotherapy with ferromagnetic implants for locally advanced and recurrent neoplasms, Int J Radiat Oncol Biol Phys, 27, 109, 1993.


42. van Wieringen, N., Kotte, A. N., van Leeuwen, G. M., Lagendijk, J. J., van Dijk, J. D., Nieuwenhuys, G. J., Dose uniformity of ferromagnetic seed implants in tissue with discrete vasculature: A numerical study on the impact of seed characteristics and implantation techniques, Phys Med Biol, 43, 121, 1998. 43. Jordan, A., Wust, P., Fahling, H., John, W., Hinz, A., Felix, R., Inductive heating of ferromagnetic particles and magnetic fluids; physical evaluation of their potential for hyperthermia, Int J Hyperthermia, 9, 51, 1993. 44. Cetas, T. C., Gross, E. J., Contractor, Y., A ferrite core/metallic sheath thermoseed for interstitial thermal therapies, IEEE Trans Biomed Eng, 45, 68, 1998. 45. Strohbehn, J.W., Evaluation of hyperthermia equipment, in Hyperthermia in Cancer Treatment, Anghileri, L.J. and Robert, J., Eds., CRC Press, Boca Raton, FL, 1986, 179. 46. Loane, J., Ling, H., Wang, B. F., Lee, S. W., Experimental investigation of a retrofocusing microwave hyperthermia applicator; conjugate-field matching scheme, IEEE Trans Microwave Theory and Tech, 34, 490, 1986. 47. Hynynen, K., Biophysics and technology of ultrasound hyperthermia, in Methods of External Hyperthermic Heating, Gautherie, M., Eds., Springer-Verlag, Berlin, Germany, 1990, 61. 48. Storm, F.K., Harrison, W.H., Elliott, R.S., Morton D.L., Physical Aspects of Localized Heating by Magnetic-Loop Induction, in Hyperthermia in Cancer Therapy, Storm, F.K., Ed., GK Hall Medical Publisher: Boston, MA., 1983, 305. 49. Anhalt, D., Hynynen, K., Deyoung, D., Shimm, D., Kundrat, M., Cetas, T., The CDRH helix: An in-vivo evaluation, Int J Hyperthermia, 6, 241, 1990. 50. Shen, R. N., Hornback, N. B., Shidnia, H., Wu, B., Lu, L., Broxmeyer, H. E., Whole body hyperthermia: A potent radioprotector in vivo, Int J Radiat Oncol Biol Phys, 20, 525, 1991. 51. Robins, H. I., Kutz, M., Wiedemann, G. J., Katschinski, D. M., Paul, D., Grosen, E., Tiggelaar, C. L., Spriggs, D., Gillis, W., d’Oleire, F., Cytokine induction by 41.8°C whole body hyperthermia, Cancer Letter, 97, 195, 1995. 52. Wiedemann, G. J., Robins, H. I., Katschinski, D. M., Mentzel, M., D’Oleire, F., Kutz, M., Wagner, T., Systemic hyperthermia and ICE chemotherapy for sarcoma patients: Rational and clinical status, Anticancer Res, 17, 2899, 1997. 53. van der Zee, J., van Rhoon, G. C., Faithful, N. S., van den Berg, A. P. Clinical hyperthermic practice: Whole-body hyperthermia, in An Introduction to the Practical Aspects of Clinical Hyperthermia, Field, S. B. and Hand, J. W., Eds., Taylor & Francis, London, UK, 1990, 185. 54. Chou, C. K., McDougall, J. A., Chan, K. W., Vora, N., Howard, H., Whole-body hyperthermia with an RF electric field system, in Proceedings 15th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Part 3, p. 1461- 1462. 1993.

Further Information 1. International Journal of Hyperthermia is published bimonthly by Taylor & Francis Ltd, Journals Customer Services, Rankine Road, Basingstoke, Hants, RG24 8PR UK. Tel: +44 (0) 1256 813000; Fax: +44 (0) 1256 2. North American Hyperthermia Society, 2021 Spring Road, Suite 600, Oak Brooks, IL 60521, Tel: (708) 571-2881, Fax: (708) 571-7837. 3. Asian Society of Hyperthermic Oncology, Kenshu Bldg. 501, 7-2-4, Hongo, Bunkyo-ku, Tokyo 113, Japan. Tel: 81-3-3811-3666; Fax: 81-3-3811-0676. 4. European Society for Hyperthermic Oncology, Contact J. Haveman. Radiotherapy Department BO, Academisch Medisch Centrum, Meibergdreef 9, 1105 AZ Amesterdam, the Netherlands. Tel: +31(20)566-4231; Fax: +31(20)566-4440.


Weaver, J. C. “Electroporation of Cells and Tissues.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

94 Electroporation of Cells and Tissues

James C. Weaver Massachusetts Institute of Technology

94.1 94.2 94.3 94.4 94.5 94.6 94.7

Background Biological Barriers Based on Lipids Decrease of the Membrane Barrier to Transport Basis of Electroporation Molecular Transport In Vitro Electroporation Applications In Vivo Electroporation Applications

94.1 Background Bioelectric phenomena are well-established topics in biomedical engineering. Electroporation is of interest because of its ability to rapidly and locally deliver molecules across biological barriers comprised of lipid bilayer membranes.1-6 Many applications use in vitro cellular conditions, and focus on DNA introduction. In vivo tissue electroporation can provide local “drug delivery” that can be electrically controlled. Electroporation allows reversible or irreversible alteration of cell membranes, and also other lipid-based barriers in tissues such as human skin. With these barriers, ionic and molecular transport is reduced within microseconds by several orders of magnitude. Simultaneously, the local electric field across the barrier drives molecular transport across the reduced barrier. The necessary transient voltage across a single bilayer membrane is about 0.5 to 1 V (depending on pulse duration) so that electroporation is mild at the molecular level, and can result in negligible damage.

94.2 Biological Barriers Based on Lipids A biological cell has intra- and extracellular compartments separated by a lipid-based membrane, which defines the spatial extent of the cell. The membrane presents a chemical boundary, which supports significant chemical concentration gradients. Although the membrane contains variable amounts and types of specific membrane proteins, the barrier-defining element is a thin (~6 nm) region of low dielectric constant (Km ≈ 2) lipid. The ability of a thin fluid sheet of low dielectric constant material to exclude ions and charged molecules is impressive, and can be understood in terms of a “Born energy” barrier ∆WBorn.7 Briefly, in the case of ion transport ∆WBorn is the electrostatic energy increase due to electric charge moving from a polarizable liquid medium (here water, dielectric constant Kw ≈ 80) to a relatively non-polarizable fluid lipid region within the membrane (dielectric constant Km ≈ 2). Formally, ∆ WBorn is the electrostatic energy associated with assembly of a particular configuration of charge, and is expressed in terms of the electric field, E, and the permittivity of the surrounding medium, = K0 (K is the dielectric constant and 0 = 8.85 × 10–12 F m–1,8)


WBorn =

∫

all space but ion

1 2 E dV . 2

(94.1)

The electrostatic exclusion of ions and charged molecules can be understood by considering the change, ∆WBorn , as an ion is moved into the membrane. The maximum value, ∆WBorn,max , is the electrostatic contribution to the barrier. Because WBorn increases rapidly as the ion enters the membrane, the membrane barrier size can be estimated by considering the simple problem of moving a charge from water into lipid. An ion of type “s” is treated as a charged sphere of radius r s and charge qs = zse with zs = ±1 and e = 1.6 × 10–19 C. Initially the sphere is surrounded by water, far from the membrane (WBorn,i) and is subsequently moved to the center of the membrane (WBorn,f ). The difference in these Born energies is the barrier height. A simple estimate of ∆WBorn is made by noting that a small ion diameter (2rs ≈ 0.4 nm) is significantly less than a typical membrane thickness (hm ≈ 3 to 6 nm; smaller values for some artificial planar bilayer membranes, larger for cell membranes). This allows ∆WBorn to be estimated by neglecting the membrane thickness, and instead considering bulk lipid. This is justified because the greatest contribution to the electric field is in the volume near the ion. This estimate yields

∆WBorn ≈

e2 8π 0rs

 1 1  −   ≈ 100 kT  K m K w 

(94.2)

where the relevant temperature is T = 37°C = 310 K. Numerical solutions to the electrostatic problem for a thin low dielectric constant sheet immersed in water yields ∆WBorn ≈ 65 kT. A barrier of this size is surmounted at a negligible rate by thermal fluctuations (spontaneous ion movement). Moreover, a transmembrane voltage, Udirect, which is much larger than physiological values, would be needed to provide this energy. Uncharged molecules that can partition into the membrane and then cross the membrane by diffusion are not significantly affected by ∆WBorn. Instead, their transport is governed by a passive permeability due to the combined effect of dissolution and diffusion. But large uncharged molecules will not readily cross the membrane either because their combined solubility and effective diffusion constant within the membrane are small.

94.3 Decrease of the Membrane Barrier to Transport Two types of membrane structural “defects” can greatly reduce ∆WBorn,max : (1) a mobile aqueous cavity (shuttling carrier), or (2) an aqueous perforation (a “pore”). Functionally, both structures provide aqueous transport pathways, so that charged molecules can cross the membrane. Indeed, biological systems have specific carriers and channels to regulate ionic and molecular transport. Electroporation based on hydrophilic pores (Fig. 94.1), and their constraints on membrane electrical behavior and molecular transport have been reviewed in detail.7 At the membrane level, structural re-arrangements such as those shown in Fig. 94.1 are hypothesized to generate a rapidly changing, heterogeneous population of pores in the membrane.7 These drawings are hypothetical constructs, because visualization of the “primary pores” of electroporation is likely to be impossible. As the transmembrane voltage, Um(t), increases, membrane conformational changes involving entry of water into the membrane increase nonlinearly in their frequency-of-occurrence. Moreover, pore population evolution is highly interactive. As pores appear and expand, the conductance of the membrane increases tremendously, and quickly becomes so large that Um(t) cannot rise further, and rapidly decays when the applied pulse ceases. Formation of pores in lipid bilayer membrane was first proposed as a purely spontaneous event, due solely to thermal fluctuations.7 An important concept is membrane rupture, a destructive mechanical


FIGURE 94.1 Illustrations of hypothetical structures of both transient and metastable membrane conformations that may be involved in electroporation. Hypothetical bilayer membrane structures related to electroporation. A: membrane free volume fluctuation. B: aqueous protrusion into the membrane (“dimple”). C: Hydrophobic pore proposed by Chizmadzhev and co-workers.9 D: Hydrophilic pore proposed by Litster and by Taupin and co-workers, usually regarded as the “primary pores” through which ion and molecules pass during electroporation. E: Composite pore with one or more proteins at the pore’s inner edge. F: Composite pore with “foot-in-the-door” charged macromolecule inserted into a hydrophilic pore. Although the actual transitions are not known, the transient aqueous pore model assumes that transitions from A → B → C or D occur with increasing frequency as Um increases. Type E may form by entry of a tethered macromolecule while Um is large, and then persist after Um has decayed by pore conduction. These hypothetical structures have not been directly observed, but are consistent with a variety of experimental observations and with theoretical models. Reproduced with permission from Weaver J, Cellular Biochem 51:426–435 (1993).

event caused by emergence of a very large pore. A “cookie cutter” model provides a simple way of imagining pore creation and motivates the mathematical form of the pore formation energy, Wp(r). Pore formation requires supplying an “edge energy,” 2πrpγ (γ is the edge energy per length) while simultaneously removing an area-related energy, πr 2p Γ, (Γ is the energy per area of a flat, pore-free membrane) arising from loss of a circular region of membrane. The idea is simple: a pore-free membrane is envisioned, then a circular region is removed, and the difference in energy between these two states is identified as Wp(r). Thus, the pore creation energy at Um = 0 is

()

∆Wp rp = 2πγrp − πΓrp2 at U m = 0

(94.3)

Equation (94.3) describes a parabolic barrier, with height Wp,max . For typical values γ ≈ 2 × 10–11 J m–1 and Γ ≈ 1 × 10–3 J m–2, and the “critical radius” rp,c at Wp,max are

rp,c =

γ πγ 2 ≈ 20 nm and Wp,max = ≈ 1.3 × 10−18 J ≈ 300 kT Γ Γ

(94.4)

Structural arguments and interpretation of experiments indicate that the minimum pore size is rp,min ≈ 1 nm. The large barrier height suggests that spontaneous rupture is negligible, consistent with normal cell stability.


94.4 Basis of Electroporation The creation of hydrophilic pores by increasing Um(t) was first suggested by Chizmadzhev and coworkers.7 It was hypothesized that the creation process involves a transition from spontaneously occurring hydrophobic pores to hydrophilic pores (Fig. 94.1), which is governed by a rate equation with a modified pore barrier function, Wp(r, Um). A hydrophilic pore (hereafter simply “pore”) is believed to dominate most electroporation phenomena. It can be modeled as a leaky, microscopic capacitor whose parallel resistance for small radii (rp,min ≈ 1 nm) is extremely large (Rp > 109 Ω). Many pores are created, which also increase in size, such that their contribution to the membrane conductance, Gm, results in a several order of magnitude drop in membrane resistance, Rm = 1/Gm . This dramatic electrical event occurs even though there is insufficient energy (qsUm,max ≈ 1 eV) for classic dielectric breakdown that involves ionization by monovalent ions. A quantitative model for electroporation of artificial planar bilayer membranes provides a test of this electroporation hypothesis. In the presence of a transmembrane voltage, the free energy of pore formation is

(

)

Wp r ,U m = 2πγrp − πΓrp2 − 0.5CpU m2 rp2 ,

(94.5)

where Um is the spatially averaged transmembrane voltage. The change in specific capacitance at the site of a pore due to water replacing lipids is

  Clw = C0  w − 1  m 

(94.6)

Here C0 = m /hm is the capacitance per area of a pore-free lipid bilayer of thickness hm, with w = Kw0 and m = K10 the permittivities of water and lipid, respectively. Using K1 ≈ 2 and Kw ≈ 80, the barrier ∆Wp(rp, Um) decreases as Um increases. The pore creation energy maximum, Wp,max(r, Um), is a fundamentally important quantity, along with the corresponding critical pore radius, rp,c . Both diminish with increasing Um . The aqueous interior of the pore has a much larger permittivity than the surrounding lipid. More specifically, as Um increases, it therefore becomes progressively more favorable for water to enter the pore. Thus, pore expansion is electrically driven. Qualitatively, this provides a readily visualized explanation of the rupture of planar bilayer membranes. As Um increases Wp,max , decreases, so that the probability of one or more “supracritical” pores [rp > rp,c(Um)] goes up. But even one supracritical pore can cause rupture. Once a pore expands beyond the barrier maximum, it can expand until reaches whatever physical boundary confines the membrane. In the case of artificial planar bilayer membranes, this boundary is a small but macroscopic aperture (e.g., raperture ≈ 10–3 m). In the case of a cell membrane, the cytoskeleton or other cellular structure may provide a boundary. In either case, expansion of a supracritical pore to the boundary is essentially irreversible. This is “irreversible breakdown.” One significant achievement of the transient aqueous pore model is a quantitative description of rupture, briefly presented here. The extension of Eq. (94.4) to Um > 0 yields,9

rp,c =

γ Γ + 0.5ClwU m2

and ∆Wp,max =

πγ 2 Γ + 0.5ClwU m2

(94.7)

Note that the critical pore radius decreases as Um increases, and more significantly, the corresponding pore energy, Wp,max (Um), decreases. Physical processes governed by activation energies (barriers) involve Boltzmann factors. In such cases, including that of pore creation, there is a nonlinear dependence on system parameters such as Um. This is the primary source of the nonlinear dependence of pore creation


on Um. As discussed below, only a moderate increase in Um is needed to significantly diminish the stability of a planar membrane, and can cause rupture. Self-consistent electroporation theories predict that pores are located randomly but widely spaced, on average separated by several tens of pore radii. The equipotentials outside the pore but near the membrane have a significant, nonlinear gradient. Thus, the current flowing through a pore results in a voltage drop both within the pore and within the electrolyte external to the pore but near the entrance of the pore. This potential drop near a pore’s mouth is due to a “spreading resistance,” Rspd, within the bathing electrolyte, which is

Rspd ≈

1 2σ erp

(94.8)

The voltage drop associated with Rspd is external to the pore, and therefore does not contribute to the driving force that expands pores. The electrical conductivity within the smallest pores is suppressed, because of Born energy exclusion due to the nearby low dielectric constant lipid. Ion entry is less likely due to the energy cost of bringing an ion into a small pore. This leads to a reduction in electrical conductivity within a pore, σp . Moreover, ion movement through a pore only slightly larger than the ion involves significant steric hindrance. The resistance within a pore can be estimated using a simple cylindrical geometry, but with a Born energyreduced conductivity, σp . This “interior” pore resistance is

Rint ≈

hm πrp2σ p

(94.9)

where σe is the bulk electrolyte conductivity. The two resistances Rspd and Rint are in series, creating a voltage divider effect.

 R  int  U m,p = U m   Rint + Rspd 

(94.10)

For this reason Um,p ≤ Um , and as a result the electrical expanding force associated with ∂Wp(r, Um)/∂r in pore radius space is diminished, and pores expand more slowly than expected from using Um in Eq. (94.5). Electrical measurements on artificial planar bilayer membranes show that some small pores remain after Um is decreased. Cellular experiments tested the response to dye uptake after electrical pulsing, and these revealed that some cells have a persisting capability to take up these molecules. One hypothesis is that some types of complex pores are created, which may involve a more permanent portion of the cell, e.g., the cytoskeleton or tethered cytoplasmic molecules (Fig. 94.1). Thus, some type of long lifetime metastable pores may occur. This idea is consistent with experiments involving DNA uptake by cells. DNA is a large, highly charged molecule, so that while temporarily occupying a pore, DNA charge groups should inhibit pore closure by coulombic repulsion.7 This is a “foot-in-the-door” hypothesis (Fig. 94.1), which is also consistent with skin electroporation experiments using heparin.6 However, most pores appear to be destroyed quickly, with pore destruction independent of other pores. This independence is plausible, as pores are predicted to be widely spaced even when the maximum number of pores is present. A heterogeneous pore size is fundamentally expected, because there are two contributions of energy that lead to pore formation: (1) thermal fluctuations (“kT energy”; stochastic), and (2) an increased transmembrane voltage (“electric field energy”; deterministic). The combination is therefore stochastic, and leads to the expectation that a distribution of pore sizes occurs. The flux of pores in pore radius space is described by


 ∂n n ∂∆Wp   J p = − Dp  +  ∂rp kT ∂rp 

(94.11)

Diffusion within “pore radius space” is contained in the term with Dp , the diffusion constant of the pores in pore radius space. For a statistical description, n = n(rp , t) is the pore distribution function, in which n = n(rp , t) is a probability density function. Within a radical increment, ∆rp there are n = n (rp, t) ∆rp pores instantaneously present with radii between rp and rp + ∆rp . This approach leads to a stochastic description of rupture, with average membrane lifetimes,

τ=

(kT )

32

(

4 πc p Am Dp γ Γ + 0.5ClwU m2

)

 πγ 2 exp   kT Γ + 0.5C U 2 m m 

(

)

   

(94.12)

Here, Am is the planar membrane area, and cp (the surface pore concentration) is the number of pores per area. Another simpler approach for estimating τ is based on an absolute rate estimate for critical pore appearance. In this case, τ depends on the reciprocal of a Boltzmann factor, in which Wp(Um) is an argument. By using an order of magnitude estimate for the prefactor, the mean membrane lifetime against rupture is

τ≈

[

1 exp + ∆Wp,c kT v0Vm

]

(94.13)

Here v0 is an attempt rate density based on the collision frequency and number density of phospholipid molecules within the membrane. The quantity Vm = hmAm is the membrane volume. By choosing a plausible mean lifetime, (e.g., τm ≈ 1 s), the magnitude of ∆Wp,c can be estimated, and from that, Um,c . Because of the strong non-linear behavior of Eqs. (94.12) and (94.13), an order of magnitude difference in the choice of mean lifetime, e.g., 0.1 or 10 s, results in only small differences in Um,c . Interaction of the membrane and its pore population with its environment should also be considered. Specifically, there is a current pathway through the bulk electrolyte by which ions are transported to the membrane in order to charge the membrane capacitance to Um . During electroporation, the membrane capacitance, Cm , changes by less than 2% because only a small fraction of the membrane is occupied by aqueous pores. In contrast, the membrane conductance, Gm(t) = 1/Rm(t), increases by orders of magnitude. Typically a short voltage pulse is applied, and current flows into and/or across the membrane. The membrane usually cannot discharge through the pulse generator. Instead, the discharge current passes through pores in the membrane. During a rectangular pulse, the circuit differential equation is

Cm

I R dU m = p N − Um dt RE + RN

 1   for 0 < t < t pulse Gm t + RE + RN  

()

(94.14)

Once the pulse stops (and the pulse generator is switched out of the circuit), the transmembrane voltage can change only by ions moving through the membrane, and the corresponding equation begins

Cm

dU m U = − m for t > t pulse dt Rm

(94.15)

Subject to the initial condition Um(t = 0) of zero, Eqs. (94.14) and (94.15) are solved numerically to provide a computer stimulation of planar membrane electroporation.


Experiments have shown that during rupture Um(t) exhibits a much longer discharge time than during reversible electrical breakdown (REB, see below), and that the rupture discharge results in Um(t) having a sigmoidal shape. A transient aqueous pore theory can account for general features of this behavior.7 The striking phenomenon of REB is observed for both artificial planar bilayer membranes and in cell membranes. However, the term “breakdown” is misleading, as REB is actually protective. During REB the membrane acquires a very large conductance by evolving a large pore population, and the resulting highly nonlinear increase in Gm prevents Um from exceeding Um ≈ 1.5 V. If some types of artificial planar bilayer membranes are challenged by short pulses, a characteristic feature of REB is observed: there is progressive shortening of the postpulse membrane discharge as larger and larger pulses are used. This behavior is consistent with a progressively smaller Rm being achieved. The transition from irreversible behavior (“rupture”) to reversible behavior (REB or incomplete reversible electrical breakdown) can be quantitatively described by the evolution of a dynamic pore population. If the pore population is small (moderate Um), then the membrane discharge can be so slow that one or more pores evolve to rupture the membrane. But if the pore population is large (high Um), then there are so many pores that the membrane discharges before any membrane-rupturing supracritical pores evolve. Irreversible electroporation of cell membranes is more complicated. A planar membrane is bounded by the aperture of the experimental apparatus, and for this reason has a total surface tension (both sides of the membrane), Γ, which favors pore expansion. But there may be no corresponding reservoir of membrane molecules in a cell, and almost certainly none in the case of a vesicle (approximately an empty spherical membrane). If the osmotic or hydrodynamic pressure difference across the cell membrane is zero, the cell membrane has Γ ≈ 0, and rupture is therefore not expected. If, however, a region of a cell membrane is bounded by the cytoskeleton (or other cellular structure) the membrane may behave like a microscopic planar membrane. One or more regions of a cell membrane may rupture, creating permanent openings and leading to cell death.

94.5 Molecular Transport From a biomedical engineering viewpoint, electrically controlled transport of molecules across cell membranes and tissues is likely to be the most important feature of electroporation. There is strong evidence for electrophoretic transport through pores as a major contribution for charged molecules. This is consistent with the observations that charged molecule transport due to a single exponential pulse exhibits a subequilibrium plateau, in that net transport (cell uptake) is independent of field pulse magnitude for large pulse magnitudes. A transient aqueous pore model of electroporation uses only an electrophoretic drift contribution through pores, and this is consistent with a subequilibrium plateau. Most potential biomedical applications seek controlled molecular transport, without significant side effects. However, assessment of cell viability/death following electroporation is nontrivial. Convenient, short term viability assays based on membrane-impermanent dyes assume that membrane openings are themselves evidence for cell death. The fact that electroporation by definition causes openings with unknown recovery kinetics means that vital stains and membrane exclusion probes cannot be used without validation. Partly for this reason, cell death by electroporation is not well understood. Chemical stress resulting from nonspecific molecular transport has been hypothesized to cause cell death by either reversible or irreversible cell membrane electroporation.4 The irreversible case clearly can lead to stress, but in principle membrane-reversible electroporation that generates large chemical exchange can also create stress. In either case, significant molecular exchange can occur. As already noted, in the case of irreversible electroporation, a region of the cell membrane may behave much like a microscopic planar membrane, and undergo rupture. In the case of reversible electroporation, significant molecular transport between the intra- and extracellular volumes could cause a significant chemical imbalance. If too large, the associated chemical stress (loss of essential compounds; entry of harmful compounds) may kill the cell. The cell’s local environment should be relevant, and the ratio


Rvolume =

Vextracellular Vintracellular

(94.16)

should govern the degree of chemical stress, and may correlate with cell survival or death. If Rvolume >> 1, (typical of in vitro cell suspensions and anchorage dependent cell culture) cell death should be favored, while for the other extreme Rvolume Tm : cout − cin = c0e − t TBC e TM

TBC

(96.6a)

]

−1

(96.6b)

CSTR: Here the internal concentration is uniform at the outlet level, and

[

]

 R  − τ − Rτ cout − cin =  − e − τt ; R = TBC TM  e −e  R − 1

(96.7)

These results are plotted in Fig. 96.1b for two time-constant ratios, TBC /TM , = 1 and TBC /TM = 10. The two flow conditions are seen to produce very different behavior for equal time constants, but these differences quickly become minor when Tm TBC . More specifically, the effect of our system flow conditions on the inlet concentration becomes insensitive to these conditions when two criteria are met:

Tm

TBC

t = TOBS ≥ Tm

(96.8a) (96.8b)

Here TOBS is the observer time, i.e., the time before which there is no interest in the system behavior. These are the conditions of time constant separation, of great practical importance in all engineering design calculations. Usually “much less than” can be taken to be less than a third, and vice versa, and


FIGURE 96.2

Modeling hemodialysis.

one tenth is almost always sufficient. Thus, using time constant separation to simplify process descriptions is usually referred to as an order-of-magnitude approximation. Returning to Fig. 96.1a we may now write a macroscopic mass balance [Bird et al., 1960, Ch. 22] of the form

(

)

Vdc dt ≈ Q cin − c + V R

(96.9)

where c is both average internal and exit concentration, and 〈R〉 is the average rate of solute formation by chemical reaction. Here we have used the CSTR approximation as the simplest to handle mathematically. This expression is valid only under the constraints of Eq. (96.8), but these are commonly met in conditions of medical interest. The major utility of Eq. (96.9) is in the description of networks such as the organ network of the human body, and applications include transient drug distribution, anaesthesia, and the study of metabolic processes [Welling, 1997; Bassingthwaighte et al., 1998]. Here an equation similar to Eq. (96.8) must be written for each element of the network, but individual organs can be combined, so long as the resulting element conforms to Eq. (96.8). These processes are often lumped under the heading of pharmacokinetics, and a large literature has developed [Welling, 1997]. An example of great economic importance is hemodialysis for the treatment of end stage kidney disease. Here the body can be approximated by the simple diagram of Fig. 96.2, and the defining equations reduce to

(

)

VT dcT dt = QB c B − cT + G

(96.9a)

(

(96.9b)

)

VBdc B dt = −QB c B − cT − Cl ⋅ c B

Here the subscript B and T refer to the blood and (whole body) tissue respectively; G is the rate of toxin generation, and Cl is the clearance of the dialyzer. Comparison of prediction with observation is shown for creatinine in Fig. 96.3. Here the parameters of Eq. (96.9) were determined for the first day’s treatment. It may be seen that the pharmacokinetic approximation permits accurate extrapolation for four additional days.


FIGURE 96.3

Actual creatinine dynamics compared to prediction.

96.3 Extracorporeal Systems Next we look at the problem of designing extracorporeal systems, and these can normally be classified into a small number of categories. We shall consider steady-state membrane separators, chromatographic devices, and flow reactors.

Membrane Separators The purpose of these devices is to transfer solute from one flowing steam to another, and there are two subcategories distinguished by the ratio of transmembrane flow induced (convective) and diffusional solute. This ratio in turn is defined by a Péclet number,

Pe ≡ v

P

(96.10)

P ≡ N i ∆ci

(96.11)

Here 〈v〉 is the observable transmembrane solvent velocity, and P is the membrane solute diffusional permeability. The permeability in turn is defined as the ratio of the molar flux of solute transport, moles/area-time, to the solute concentration difference causing this transport. The most familiar examples of low-Pe devices are blood oxygenators and hemodialyzers. High-Pe systems include micro-, ultra-, and nano-filtration and reverse osmosis. The design and operation of membrane separators is discussed in some detail in standard references [Noble and Stern, 1995; Ho and Sirkar, 1992], and a summary of useful predictions is provided in Section 96.4. Low-Pe devices are by far the simpler. Solute transport in dialyzers is essentially unaffected by the small amount of transmembrane water flow, and one may therefore use standard design techniques based on membrane permeabilities, usually supplied by the vendor, and mass transfer coefficients in the adjacent fluids. Local fluxes can be described by the simple expression

(

N i = K c cib − cie


)

(96.12)

TABLE 96.1 Asymptotic Nusselt Numbers for Laminar Duct Flow (With Fully Developed Velocity Profiles) Constant Wall Concentration

Constant Wall Mass Flux

Thermal Entrance Region Plug flow: Nuloc =

1  vD 2   —   π  Dz

12

Nuloc =

π 2

 vD 2   Dz   — 

12

Parabolic velocity profile: Nuloc =

 gD3   —  13 9 Γ 4 3  Dz 

13

1

( )

Nuloc =

( )

Γ 2 3  gD3   —  91 3  Dz 

13

Here z is distance in the flow direction and g is the rate of change of velocity with distance from the wall, evaluated at the wall. Nusselt numbers are local values, evaluated at z. Fully Developed Region Plug flow: Parabolic Flow:

Nuloc = 5.783 Nu = 3.656

Nuloc = 8 Nu = 48/11

where Ni is the molar flux of the solute “i” across the membrane, moles/area-time, Kc is the overall mass transfer coefficient, ci is molar solute concentration while the subscripts “b” and “e” refer to blood and the external fluid, respectively. The overall mass transfer coefficient must be calculated from the two fluid phase coefficients and membrane permeability. Hemodialysis solutes tend to distribute uniformly between blood (on a cell-free basis), and one may use the simple approximation

( ) ( ) ( )

1 K c = 1 kb + 1 P + 1 ke

(96.13)

Here kb and ke are the mass transfer coefficients for the blood and external fluid, here dialysate, respectively. These phase mass transfer coefficients can usually be estimated to an acceptable accuracy from simple asymptotic formulas [Lightfoot and Lightfoot, 1997]. Examples of the latter are given in Table 96.1. For unequally distributed solutes, Eq. (96.3) must be appropriately modified [Lightfoot and Lightfoot, 1997], and, for blood this can be something of a problem [Popel, 1989]. Equipment performance for dialyzers is normally expressed in terms of clearance

Cl = fci ,inQ

(96.14)

where f is the fraction of incoming solute “i” removed from the blood, ci,in is concentration of solute “i” in the entering blood, and Q is volumetric blood flow rate. Clearance is a convenient measure as it is independent of solute concentration, and it is easily determined experimentally. Prediction is useful for device design, but in operation, clearance is usually determined along with effective blood flow rates and tissue water volumes from numerical analysis of a test dialysis procedure. The efficiency of blood oxygenators can be dominated by either membrane permeability or mass transfer in the flowing blood, and it is complicated by the kinetics and thermodynamics of the oxygen/hemoglobin system. These aspects are discussed in detail by Popel [1989]. High-Pe devices are dominated by transmembrane water transport, and detailed discussion must be left to the above cited references. However, it is important to recognize their primary function is to remove water and undesired solutes while retaining one solute which is desired. Rejection of the desired product increases toward an asymptote as water flux increases, and one should operate near this asymptote if at all possible. The relation between water flux and rejection is perhaps best determined experimentally.


However, as water flux increases the rejected solute concentration at the interface between the feed stream and the membrane also increases. This process, usually known as concentration polarization, typically produces a significant increase in osmotic pressure which acts to reduce the flow. Polarization is a complex process, but to a good approximation the trans-membrane water velocity is given by

(

)(

)

v ≈ 1.4 ln c sδ c s 0 ⋅ ρδ ρ0 ⋅ Θ ⋅ kc

(96.15)

Here cs is concentration of the rejected solute, ρ is solution density, kc is the concentration-based mass transfer coefficient in the absence of water flux, and the subscripts δ and 0 refer to conditions at the membrane surface and bulk of the fed solution, respectively. The factor Θ is a correction for variable viscosity and diffusivity, approximated at least for some systems by

–* Θ≡ D

2 3

1 µ*

13

(96.16)

– *〉 and 〈1/µ*〉 are the averages of solute diffusivity and reciprocal solution viscosity at the memand 〈 D brane surface and bulk solution divided by these quantities in the bulk solution. These equations are reasonable once appreciable polarization has occurred, if rejection is high, and they are a modification of earlier boundary-layer analyses of Kozinski and Lightfoot [1972]. They are more accurate than the more recent results made using simple film theory discussed in [Noble and Stern, 1995; Ho and Sirkar, 1992], both in incorporating the coefficient 1.4 and in the corrections for variable diffusivity and viscosity. The coefficient of 1.4 is a correction to account for boundary-layer compression accompanying transmembrane flow, not allowed for in film theory. Typical geometry insensitive boundary layer behavior is assumed, and these equations are not restricted to any given rejected solute or equipment configuration. However, in order to calculate the pressure drop required to obtain this flow, one must use the expression ∆p = π + v kh

(96.17)

where ∆p is transmembrane pressure drop, π is solute osmotic pressure at the membrane surface, and kh is the hydraulic permeability of the membrane. The osmotic pressure in turn is a function of solute concentration at the membrane surface, and thus is different for each rejected solute.

Chromatographic Columns Chromatography is very widely used in biomedical analyses and to a significant extent for extracorporeal processing of blood and other body fluids. Recovery of proteins from blood is of particular importance, and these applications can be expected to grow. Good basic texts are available for underlying dynamic theory [Guiochon et al., 1994] and chemistry [Snyder et al., 1997], and a series of design papers is recommended [Lightfoot et al., 1997; Lode et al., 1998; Yuan et al., in press; Athalye et al., 1992]. Differential chromatography, in which small mixed-solute pulses are separated by selective migration along a packed column, is the simplest, and much of the chromatographic literature is based on concepts developed for this basic process. In differential chromatography, the individual solutes do not interact, and the effluent curve for each is usually close to a Gaussian distribution: 2   t  c f L, t = c0 exp − − 1   t    

( )

(2π) (σ t ) 12

(96.18) –

where cf (L,t) is the fluid phase concentration leaving a column of length L at time t, t is the mean solute residence time, and σ is the standard deviation of the distribution; c0 is the maximum effluent concentration. © 2000 by CRC Press LLC

FIGURE 96.4

Resolution in differential chromatography.

Degree of separation is defined in terms of the resolution, R12 defined as in Fig. 96.4 where separation with – – a resolution of unity is shown. Here the distance between the two mean residence times (t 1-t 2), 40 and 20 in Fig. 96.4, increase in direct proportion to column length, while the standard deviations increase with its square root. Thus resolution is proportional to the square root of column length. Column performance is normally rated in terms of the number of equivalent theoretical plates, N defined by

( )

2

N≡ t σ .

(96.19)

Methods of predicting N from column design and operating conditions are described in the above mentioned references (see for example [Athalye et al., 1992]).

Flow Reactors Development of small flow reactors for diagnostic purposes is a fast growing field. Patient monitoring of blood glucose is probably the largest current example, but new applications for a wide variety of purposes are rapidly being developed. These are essentially miniaturized variants of industrial reactors and normally consist of channels and reservoirs with overall dimensions of centimeters and channel diameters of millimeters or even tenths of millimeters. The key elements in their description are convective mass transfer, dispersion, and reaction, and they differ from larger systems only in parameter magnitudes. In particular, flow is almost invariably laminar and Schmidt numbers — Sc ≡ µ ρD

(96.20)

– is effective solute are always high, never less than 105. Here µ is solvent viscosity, ρ is its density, and D diffusivity through the solution. These aspects are briefly described in the next section.

96.4 Useful Correlations Convective Mass Transfer Mass transfer coefficients kc are best correlated in the dimensionless form of mass transfer Nusselt numbers defined as © 2000 by CRC Press LLC

— Nu ≡ kc D D

(96.21)

where D is any convenient characteristic length, typically diameter for tubes. For the high-Sc systems considered above, these Nusselt numbers are functions only of a dimensionless ratio L*, system geometry, and boundary conditions [Lightfoot and Lightfoot, 1997]. The scaled length L* in turn is just the ratio of mean solvent residence time to lateral diffusion time:

(

L* ≡ L v

) (D

2

— 24D

)

(96.22)

For all geometries these functions have the same three characteristics: an entrance region solution for L* “much” less than unity, a constant asymptotic value for L* “much” greater than unity (the “fully developed” region), and a relatively unimportant transition region for L* close to unity. Normally the entrance region is of most practical importance and can be used without great error up to L* of unity. Typical results are shown in Table 96.1. The entrance region results are valid for both tubes (circular cross-section) and between parallel plates; it may be noted that the reference length D cancels. It appears to the same power on both sides of the equations. The correlations for laminar flow are also valid for non-Newtonian fluids if the appropriate wall velocity gradients are used. The solutions for fully developed flow are, however, only useful for round tubes and either plug or Poiseuille flow as indicated. These are much less important as they result in very low transfer rates. Plug and Poiseuille (parabolic) flows are limiting situations, and most real situations will be between them. Plug flow is approximated in ultrafiltration where the high solute concentrations produce a locally high viscosity. Poiseuille flow is a good approximation for dialysis.

Convective Dispersion and the One-Dimensional Convective Diffusion Equation Non-uniform velocity profiles tend to reduce axial concentration gradients, a result which is qualitatively similar to a high axial diffusivity. As a result, it is common practice to approximate the diffusion equation for duct flow as — ∂2c ∂z 2 + R ∂c ∂t + v ∂c ∂z ≡ D eff i ,eff

(96.23)

Here the overline indicates an average over the flow cross-section, usually further approximated as the – eff is an cup-mixing or bulk concentration [Bird et al., 1960, p. 297), 〈v〉 is the flow average velocity, D effective diffusivity and 〈Ri,eff〉 is the rate of solute addition per unit volume by reaction plus mass transfer across the wall:

( ) (

Ri ,eff = Ri ,chem + C S K c ce − c

)

(96.24)

Here 〈Ri,chem〉 is the average volumetric rate of formation of species “i”, moles/volume-time, C is the duct circumference, S is its cross-sectional area, and ce is the solute concentration outside the duct wall, i.e., in the surrounding fluid. For Newtonian tube flow at L* greater than unity, it is reasonable to write

(

)(

— D eff D ≈ 1 + 1 192 D v

)

2

D

—

(96.25)

The restrictions on this result and some additional useful approximations are widely discussed in the mass transfer literature (see for example [Ananthkrishnan et al., 1965]). For flow in packed beds, one often sees


(

Deff D ≈ 0.4 Dv2 — D

—

)

(96.26)

but more accurate expressions are available [Athalye et al., 1992]. Convective dispersion is a very complex subject with a large literature, and successful application requires careful study [Brenner, 1962; Ananthkrishnan et al., 1965; Edwards et al., 1991].

References Anthakrishnan, V, WN Gill, and AJ Barduhn, 1965, Laminar Dispersion in Capillaries, AIChEJ, 11(6), 1063-1072. Athalye, AM, SJ Gibbs, and EN Lightfoot, 1992, Predictability of Chromatographic Separations: Study of Size Exclusion Media with Narrow Particle Size Distributions, J. Chromatogr A, 589, 71-85. Bassingthwaighte, JB, CA Goresky and JH Linehan, 1998, Whole Organ Approaches to Cellular Metabolism, Springer, New York. Bird, RB, WE Stewart and EN Lightfoot, 1960, Transport Phenomena, Wiley, New York. Brenner, H, 1962, The diffusion model of longitudinal mixing in beds of finite length, CES, 17, 229-243. Dedrick, RL, KB Bischoff, and DS Zaharko, 1970, Interspecies Correlation of Plasma Concentration History of Methotrexate (NSC-740), Cancer Ther Rep, Part I, 54, 95. Edwards, DA, M. Shapiro, H. Brenner, and M. Shapira, 1991, Transport in Porous Media. Guiochon, G, SG Shirazi, and AM Katti, 1994, Fundamentals of Preparative and Non-Linear Chromatography, Academic Press, New York. Ho, WSW, and KK Sirkar, 1992, Membrane Handbook, Van Nostrand Reinhold, New York. Kozinski, AA, and EN Lightfoot, 1972, Protein Ultrafiltration, AIChEJ, 18, 1030-1040. Lightfoot, EN, 1974, Transport Phenomena and Living Systems, Wiley-Interscience, New York. Lightfoot, EN and Lightfoot, EJ, 1997, Mass Transfer in Kirk-Othmer Encyclopedia of Separation Technology. Lightfoot, EN and KA Duca, 2000, The Roles of Mass Transfer in Tissue Function, in The Biomedical Engineering Handbook, 2nd ed., Bronzino, J., Ed., CRC Press, Boca Raton, FL, chap. 115. Lightfoot, EN, JL Coffman, F Lode, TW Perkins, and TW Root, 1997, Refining the Description of Protein Chromatography, J. Chromatgr A, 760, 130. Lode, F, A Rosenfeld, QS Yuan, TW Root and EN Lightfoot, 1998, Refining the Scale-up of Chromatographic Separations, J. Chromatogr A, 796, 3-14. Noble, RD, and SA Stern, 1995, Membrane Separations Technology, Elsevier, London. Popel, AS, 1989, Theory of oxygen transport to tissue, Clin Rev Biomed Eng 17(3), 257. Schmidt-Nielsen, 1997, Animal Physiology, 5th ed., Cambridge University Press, Cambridge. Snyder, LR, JJ Kirkland, and JL Glajch, 1997, Wiley, New York. Welling, PG, 1997, Pharmacokinetics, American Chemical Society. Yuan, QS, A Rosenfeld, TW Root, DJ Klingenberg, and EN Lightfoot, in press, Flow Distribution in Chromatographic Columns, J. Chromatogr A.


Shuler, M. L. “Animal Surrogate Systems.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

97 Animal Surrogate Systems 97.1

Background Limitations of Animal Studies • Alternatives to Animal Studies

97.2 97.3 97.4

Michael L. Shuler Cornell University

97.5

The Cell Culture Analog Concept Prototype CCA Use of Engineered Tissues or Cells for Toxicity/Pharmacology Future Prospects

97.1 Background Animal surrogate or cell culture analog (CCA) systems mimic the biochemical response of an animal or human when challenged with a chemical or drug. A true animal surrogate is a device that replicates the circulation, metabolism, or adsorption of a chemical and its metabolites using interconnected multiple compartments to represent key organs. These compartments make use of engineered tissues or cell cultures. Physiologically based pharmacokinetic models (PBPK) guide the design of the device. The animal surrogate, particularly a human surrogate, can provide important insights into toxicity and efficacy of a drug or chemical when it is impractical or imprudent to use living animals (or humans) for testing. The combination of a CCA and PBPK provides a rational basis to relate molecular mechanisms to whole animal response.

Limitations of Animal Studies The primary method used to test the potential toxicity of a chemical or action of a pharmaceutical is to use animal studies, predominantly with rodents. Animal studies are problematic. The primary difficulties are that the results may not be meaningful to assessment of human response [Gura, 1997]. Because of the intrinsic complexity of a living organism and the inherent variability within a species, animal studies are difficult to use to identify unambiguously the underlying molecular mechanism for action of a chemical. The lack of a clear relationship among all of the molecular mechanisms to whole animal response makes extrapolation across species difficult. This factor is particularly crucial when extrapolation of rodent data to humans is an objective. Further, without a good mechanistic model it is difficult to rationally extrapolate from high doses to low doses. However, this disadvantage due to complexity can be an advantage; the animal is a “black box” and provides response data even when the mechanism of action is unknown. Further disadvantages reside in the high cost of animal studies, the long period of time often necessary to secure results, and the potential ethical problems in animal studies.


Alternatives to Animal Studies In vitro methods using isolated cells (e.g., [Del Raso, 1993]) are inexpensive, quick, and have almost no ethical constraints. Because the culture environment can be specified and controlled, the use of isolated cells facilitates interpretation in terms of a biochemical mechanism. Since human cells can be used as well as animal cells, cross-species extrapolation is facilitated. However, these techniques are not fully representative of human or animal response. Typical in vitro experiments expose isolated cells to a static dose of a chemical or drug. It is difficult to relate this static exposure to specific doses in a whole animal. The time-dependent change in the concentration of a chemical in an animal’s organ cannot be replicated. If one organ modifies a chemical or prodrug which acts elsewhere, these situations would not be revealed by the normal in vitro test. A major limitation on the use of cell cultures is that isolated cells do not fully represent the full range of biochemical activity of the corresponding cell type when in a whole animal. Engineered tissues, especially co-cultures [Bhatia et al., 1998], can provide a more “natural” environment which can improve (i.e., make normal) cell function. Another alternative is use of tissue slices, typically from the liver [Olinga et al., 1997]. Tissue slices require the sacrifice of the animal, there is intrinsic variability, and biochemical activities can decay rapidly after harvest. The use of isolated tissue slices also does not reproduce interchange of metabolites among organs and the time-dependent exposure that occurs within an animal. An alternative to both animal and in vitro studies is the use of computer models based on PBPK models [Connolly and Anderson, 1991]. PBPK models can be applied to both humans and animals. Because PBPK models mimic the integrated, multicompartment nature of animals, they can predict the time-dependent changes in blood and tissue concentrations of a parent chemical or its metabolites. Although construction of a robust, comprehensive PBPK is time-consuming, once the PBPK is in place many scenarios concerning exposure to a chemical or treatment strategies with a drug can be run quickly and inexpensively. Since PBPKs can be constructed for both animals and humans, cross-species extrapolation is facilitated. There are, however, significant limitations in relying solely on PBPK models. The primary limitation is that a PBPK can only provide a response based on assumed mechanisms. Secondary and unexpected effects are not included. A further limitation is the difficulty in estimating parameters, particularly kinetic parameters. None of these alternatives satisfactorily predict human response to chemicals or drugs.

97.2 The Cell Culture Analog Concept A CCA is a physical replica of the structure of a PBPK where cells or engineered tissues are used in organ compartments to achieve the metabolic and biochemical characteristics of the animal. Cell culture medium circulates between compartments and acts as a “blood surrogate”. Small-scale bioreactors with the appropriate cell types in the physical device represent organs or tissues. The CCA concept combines attributes of a PBPK and other in vitro systems. Unlike other in vitro systems, the CCA is an integrated system that can mimic dose dynamics and allows for conversion of a parent compound into metabolites and the interchange of metabolites between compartments. A CCA system allows dose exposure scenarios that can replicate the exposure scenarios of animal studies. A CCA is intended to work in conjunction with a PBPK as a tool to test and refine mechanistic hypotheses. A molecular model can be embedded in a tissue model which is embedded in the PBPK. Thus, the molecular model is related to overall metabolic response. The PBPK can be made an exact replica of the CCA; the predicted response and measured CCA response should exactly match if the PBPK contains a complete and accurate description of the molecular mechanisms. In the CCA all flow rates, the number of cells in each compartment, and the levels of each enzyme can be measured independently, so no adjustable parameters are required. If the PBPK predictions and CCA results disagree, then the description of the molecular mechanisms is incomplete. The CCA and PBPK can be used in an iterative manner to test modifications in the proposed mechanism. When the PBPK is extended to describe the whole animal, failure to predict animal response would be due to inaccurate description of transport (particularly within an organ), inability to accurately measure kinetic parameters (e.g., in vivo enzyme


levels or activities), or the presence in vivo or metabolic activities not present in the cultured cells or tissues. Advances in tissue engineering will provide tissue constructs to use in a CCA that will display more authentic metabolism than isolated cell cultures. The goal is predicting human pharmacological response to drugs or assessing risk due to chemical exposure. A PBPK that can make an accurate prediction of both animal CCA and animal experiments would be “validated”. If we use the same approach to constructing a human PBPK and CCA for the same compound, then we would have a rational basis to extrapolate animal response to predict human response when human experiments would be inappropriate. Also, since the PBPK is mechanistically based, it would provide a basis for extrapolation to low doses. The CCA/PBPK approach complements animal studies by potentially providing an improved basis for extrapolation to humans. Further, PBPKs validated as described above provide a basis for prediction of human response to mixtures of drugs or chemicals. Drug and chemical interactions may be synergistic or antagonistic. If a PBPK for compound A and a PBPK for compound B are combined, then the response to any mixture of A and B should be predictable since the mechanisms for response to both A and B are included.

97.3 Prototype CCA A simple three-component CCA mimicking rodent response to a challenge by naphthalene has been tested by Sweeney et al. [1995]. While this prototype system did not fulfill the criteria for a CCA of physically realistic organ residence times or ratio of cell numbers in each organ, it did represent a multicompartment system with fluid recirculation. The three components were liver, lung, and other perfused tissues. These experiments used cultured rat hepatoma (H4IIE) cells for the liver and lung (L2) cells for the lung compartment. No cells were required in “other tissues” in this model since no metabolic reactions were postulated to occur elsewhere for naphthalene or its metabolites. The H4IIE cells contained enzyme systems for activation of naphthalene (cytochrome P450IA1) to the epoxide form and conversion to dihydriol (epoxide hydrolase) and conjugation with glutathione (glutathione-S-transferase). The L2 cells had no enzymes for naphthalene activation. Cells were cultured in glass vessels as monolayers. Experiments with this system using lactate dihydrogenase release (LDH) and glutathione levels as dependent parameters supported a hypothesis where naphthalene is activated in the “liver” and reactive metabolites circulate to the “lung” causing glutathione depletion and cell death as measured by LDH release. Increasing the level of cytochrome p450 activity in the “liver” by increasing cell numbers or by preinducing H4IIE cells led to increased death of L2 cells. Experiments with “liver”-blank; “lung”-“lung”, and “lung”-blank combinations all supported the hypothesis of a circulating reactive metabolite as the cause of L2 cell death. This prototype system [Sweeney et al., 1995] was difficult to operate, very non-physiologic, and made time course experiments very difficult. An alternative system using packed bed reactors for the “liver” and “lung” compartments has been tested [Shuler et al., 1996; Ghanem, 1998]. This system successfully allowed time course studies, was more compact and simpler to operate, and was physiological with respect to the ratio of “liver” to “lung” cells. While liquid residence times improved in this system, they still were not physiologic (i.e., 114s vs. an in vivo value of 21s in the liver and 6.6s vs. in vivo lung value of about 1.5s) due to physical limitations on flow through the packed beds. Unlike the prototype system, no response to naphthalene was observed. This difference in response of the two CCA designs was explained through the use of PBPK models of each CCA. In the prototype system, the large liquid residence times in the liver (6 min) and the lung (2.1 min) allowed formation of large amounts of naphthol from naphthalene oxide and presumably the conversion of napthol into quinones that were toxic. In the packed bed system, liquid residence times were sufficiently small so that the predicted naphthol level was negligible. Thus, the PBPK provided a mechanistic basis to explain the differences in response of the two experimental configurations. Using a very simple CCA, Mufti and Shuler [1998] demonstrated that response of human hepatoma (HepG2) to exposure to dioxin (2,3,7,8-tetrachlorodibenzo-p-dioxin) is dependent on how the dose is delivered. The induction of cytochrome p450IA1 activity was used as a model response for exposure to dioxin. Data were evaluated to estimate dioxin levels giving cytochrome P450IA1 activity 0.01% of


maximal induced activity. Such an analysis mimics the type of analysis used to estimate risk due to chemical exposure. The “allowable” dioxin concentration was 4 × 10–3 nM using a batch spinner flask, 4 × 10–4 nM using a one-compartment system with continuous feed, and 1 × 10–5 nM using a simple two-compartment CCA. Further, response could be correlated to an estimate of the amount of dioxin bound to the cytosolic Ah receptor with a simple model for two different human hepatoma cell lines. This work illustrates the potential usefulness of a CCA approach in risk assessment. Ma et al. [1997] have discussed an in vitro human placenta model for drug testing. This was a twocompartment perfusion system using human trophoblast cells attached to a chemically modified polyethylene terephatholate fibrous matrix as a cell culture scaffold. This system is a CCA in the same sense as the two-compartment system used to estimate response to dioxin. The above examples are the first that attempt to mimic circulation and metabolic response to model an animal as an integrated system. However, others have used engineered tissues as a basis for testing the efficacy of drugs or toxicity of chemicals. These tissues are important in themselves and could become elements in an integrated CCA.

97.4 Use of Engineered Tissues or Cells for Toxicity/Pharmacology The primary use of engineered tissues for toxicity testing has been with epithelial cells that mimic the barrier properties of the skin, gut, or blood brain barrier. The use of isolated cell cultures has been of modest utility due to artifacts introduced by dissolving test agents in medium and due to the extreme sensitivity of isolated cells to these agents compared to in vivo tissue. One of the first reports on the use of engineered cells is that by Gay et al. [1992] reporting on the use of a living skin equivalent as an in vitro dermatotoxicity model. The living skin equivalent consists of a co-culture of human dermal fibroblasts in a collagen-containing matrix overlaid with human keratinocytes that have formed a stratified epidermis. This in vitro model used a measurement of mitochondrial function (i.e., the colormetric thiazolyl blue assay) to determine toxicity. Eighteen chemicals were tested. Eleven compounds classified as nonirritating had minimal or no effect on mitochondrial activity. For seven known human skin irritants, the concentration that inhibited mitochondrial activity by 50% corresponded to the threshold value for each of these compounds to cause irritation on human skin. However, living skin equivalents did not fully mimic the barrier properties of human skin; water permeability was 30-fold greater in the living skin equivalent than in human skin. Kriwet and Parenteau [1996] report on permeabilities of 20 different compounds in in vitro skin models. Skin cultures are slightly more permeable (two- or threefold) for highly lipophilic substances and considerably more permeable (about tenfold) for polar substances than human-cadaver or freshly excised human skin. The above tests are static. Pasternak and Miller [1996] have tested a system combining perfusion and a tissue model consisting of MDCK (Madin-Darby canine kidney) epithelial cells cultured on a semiporous cellulose ester membrane filter. The system could be fully automated using measurement of transepithelial electrical resistance (TER) as an end point. A decrease in TER is an indicator of cell damage. The system was tested using nonionic surfactants and predicted the relative occular toxicity of these compounds. The perfusion system mimics some dose scenarios (e.g., tearing) more easily than a static system and provides a more consistent environment for the cultured cells. The authors cite as a major advantage that the TER can be measured throughout the entire exposure protocol without physically disturbing the tissue model and introducing artifacts in the response. Probably the most used cell-based assay is the Caco-2 model of the intestine to determine oral availability of a drug or chemical. The Caco-2 cell cultures are derived from a human colon adenocarcinoma cell line. The cell line, C2Bbel, is a clonal isolate of Caco-2 cells that is more homogeneous in apical brush border expression than the Caco-2 cell line. These cells form a polarized monolayer with an apical brush border morphologically comparable to the human colon. Tight junctions around the cells act to restrict passive diffusion by the paracellular route mimicking the transport resistance in the


intestine. Hydrophobic solutes pass primarily by the transcellular route and hydrophilic compounds by the paracellular route. Yu and Sinko [1997] have demonstrated that the substratum (e.g., membrane) properties upon which the monolayer forms can become important in estimating the barrier properties of such in vitro systems. The barrier effects of the substratum need to be separated from the intrinsic property of the monolayers. Further, Anderle et al. [1998] have shown that the chemical nature of substratum and other culture conditions can alter transport properties. Sattler et al. [1997] provide one example (with hypericin) of how this model system can be used to evaluate effects of formulation (e.g., use of cyclodextrin or liposomes) on oral bioavailability. Another example is the application of the Caco-2 system to transport of paclitaxel across the intestine [Walle and Walle, 1998]. Rapid passive transport was partially counter-balanced by an efflux pump (probably P-glycoprotein) limiting oral bioavailability. Another barrier of intense interest for drug delivery is the blood-brain barrier. The blood-brain barrier is formed by the endothelial cells of the brain capillaries. A primary characteristic is the high resistance of the capillary due to the presence of complex tight junctions inhibiting paracellular transport and the low endocytic activity of this tissue. Acceptable in vitro models have been more difficult to formulate but one commercial system marketed by Cellco, Inc. is available that uses a co-culture of endothelial cells and astrocytes to form a barrier. A recent example of a different in vitro system is described by Glynn and Yazdanian [1998] who used bovine brain microvessel endothelial cells grown on porous polycarbonate filters to compare the transport of nevirapine, a reverse transcriptase inhibitor to other HIV antiretroviral agents. Nevirapine was the most permeable antiretroviral agent and hence may have value in HIV treatment in reducing levels of HIV in the brain. These isolated cultures mimic an important aspect of cell physiology (oral uptake or transport into the brain). In principle, they could be combined with other tissue mimics to form a CCA that would be especially useful in testing pharmaceuticals.

97.5 Future Prospects Significant progress is being made in the construction of tissue engineered constructs [Baldwin and Saltzman, 1996]. These efforts include highly vascularized tissues, which are especially important for toxicological or pharmacological studies in the liver. Liver constructs, often based on co-cultures, have become increasingly more normal in behavior (see [Griffith et al., 1997]). For toxicological studies multiple test systems operated in parallel are desirable which suggests the need for miniaturization to conserve cells, reagents, and time. Bhatia et al. [1998] have described a microfabricated device for hepatocyte-fibroblast co-cultures. A CCA based on the concepts described here and incorporating these advanced-engineered tissues could become a powerful tool for preclinical testing of pharmaceuticals. While drug leads are expanding rapidly in number, the capacity to increase animal and human clinical studies is limited. It is imperative that preclinical testing and predictions for human response become more accurate. A CCA should become an important tool in preclinical testing.

Defining Terms Animal surrogate: A physiologically based cell or tissue multi-compartmented device with fluid circulation to mimic metabolism and fate of a drug or chemical. Engineered tissues: Cell culture mimic of a tissue or organ; often combines a polymer scaffold and one or more cell types. Physiologically based pharmacokinetic model (PBPK): A computer model that replicates animal physiology by subdividing the body into a number of anatomical compartments, each compartment interconnected through the body fluid systems; used to describe the time-dependent distribution and disposition of a substance. Tissue slice: A living organ is sliced into thin sections for use in toxicity studies; one primary organ can provide material for many tests.


References Anderle P, Niederer E., Werner R., Hilgendorf C, Spahn-Langguth H, Wunderu-Allenspach H, Merkle HP, Langguth P. 1998. P-Glycoprotein (P-gp) mediated efflux in Caco-2 cell monolayers: the influence of culturing conditions and drug exposure on P-gp expression levels. J. Pharm. Sci. 87:757 Bhatia SN, Balis UJ, Yarmush ML, Toner M. 1998. Microfabrication of hepatocyte/fibroblast co-cultures: role of homotypic cell interactions. Biotechnol. Prog. 14:378 Baldwin SP, Saltzman WM. 1996. Polymers for tissue engineering. Trends Polym. Sci. 4:177 Connolly RB, Andersen ME. 1991. Biologically based pharmacodynamic models: tool for toxicological research and risk assessment. Annu. Rev. Pharmacol. Toxicol. 31:503 Del Raso NJ. 1993. In vitro methodologies for enhanced toxicity testing. Toxicol. Lett. 68:91 Gay R, Swiderek M, Nelson D, Ernesti A. 1992. The living skin equivalent as a model in vitro for ranking the toxic potential of dermal irritants. Toxic. In Vitro 6:303. Ghanem A. 1998. Application of a novel packed bed cell culture analog bioreactor and corresponding pharmacokinetic model to naphthalene toxicology. Ph.D. Thesis. Cornell University, Ithaca, New York. Glynn SL, Mehran Y. 1998. In vitro blood-brain barrier permeability of nevirapine compared to other HIV antiretroviral agents. J. Pharm. Sci. 87:306. Griffith LG, Wu B, Cima MJ, Powers M, Chaignaud B, Vacanti JP. 1997. In vitro orgarogensis of vacularized liver tissue. Ann. N.Y. Acad. Sci. 831:382. Gura T. 1997. Systems for identifying new drugs are often faulty. Science 273:1041 Kirwet K, Parenteau NL. 1996. In vitro skin models. Cosmetics & Toiletries 111 (Feb):93 Ma T, Yang S-T, Kniss DA. 1997. Development of an in vitro human placenta model by the cultivation of human trophoblasts in a fiber-based bioreactor system. Am. Inst. Chem. Eng. Ann. Mtg., Los Angeles, Ca, Nov. 16-21. Mufti NA, Shuler ML. 1998. Different in vitro systems affect CYPIA1 activity in response to 2,3,7,8tetrachlorodibenzo-p-dioxin. Toxicol. in Vitro 12:259. Olinga, P, Meijer DKF, Slooff, MJH, Groothuis, GMM. 1997. Liver slices in in vitro pharmacotoxicology with special reference to the use of human liver tissue. Toxicol. In Vitro 12:77. Pasternak AS, Miller WM. 1996. Measurement of trans-epitheial electrical resistance in perfusion: potential application for in vitro ocular toxicity testing. Biotechnol. Bioeng. 50:568. Sattler S, Schaefer U, Schneider W, Hoelzl J, Lehr C-M. 1997. Binding, uptake, and transport of hypericin by Caco-2 cell monolayers. J. Pharm. Sci. 86:1120. Shuler ML, Ghanem A, Quick D, Wong MC, Miller P. 1996. A self-regulating cell culture analog device to mimic animal and human toxicological responses. Biotechnol. Bioeng. 52:45. Sweeney LM, Shuler ML, Babish JG, Ghanem A. 1995. A cell culture analog of rodent physiology: application to naphthalene toxicology. Toxicol. In Vitro 9:307. Walle UK, Walle T. Taxol transport by human intestinal epithelial Caco-2 cells. Drug Metabol. Disposit. 26:343. Yu H, Sinko PJ. 1997. Influence of the microporous substratum and hydrodynamics on resistances to drug transport in cell culture systems: calculation of intrinsic transport parameters. J. Pharm. Sci. 86:1448.


Baish, J. W. “ Microvascular Heat Transfer.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

98 Microvascular Heat Transfer 98.1 98.2 98.3

Introduction and Conceptual Challenges Basic Concepts Vascular Models

98.4 98.5

Heat Transfer Inside of a Blood Vessel Models of Perfused Tissues

Equilibration Lengths • Countercurrent Heat Exchange

Continuum Models • Multi-Equation Models • Vasculature Reconstruction Models

98.6

James W. Baish Bucknell University

Parameter Values Thermal Properties • Thermoregulation • Clinical Heat Generation

98.7

Solution of Models

98.1 Introduction and Conceptual Challenges Models of microvascular heat transfer are useful for optimizing thermal therapies such as hyperthermia treatment, for modeling thermoregulatory response at the tissue level, for assessing environmental hazards that involve tissue heating, for using thermal means of diagnosing vascular pathologies, and for relating blood flow to heat clearance in thermal methods of blood perfusion measurement. For example, the effect of local hyperthermia treatment is determined by the length of time that the tissue is held at an elevated temperature, nominally 43˚C or higher. Since the tissue temperature depends on the balance between the heat added by artificial means and the tissue’s ability to clear that heat, an understanding of the means by which the blood transports heat is essential. This chapter outlines the general problems associated with such processes, while more extensive reviews and tutorials on microvascular heat transfer may be found elsewhere [1-4]. The temperature range of interest for all of these applications is intermediate between freezing and boiling, making only sensible heat exchange by conduction and convection as important mechanisms of heat transfer. At high and low temperatures, such as those present during laser ablation or electrocautery and cryopreservation or cryosurgery, the change of phase and accompanying mass transport present problems beyond the scope of this section. See, for example, [5]. While the equations that govern heat transport are formally similar to those that govern passive mass transport, the processes involve the microvasculature in fundamentally different ways because the thermal diffusivity of most tissues is roughly two orders of magnitude greater than the diffusivity for mass transport of most mobile species (1.5 × 10–7 m2/s for heat vs. 1.5 × 10–9 m2/s for O2). Mass transport is largely restricted to the smallest blood vessels, the capillaries, arterioles, and venules, whereas heat transport occurs in somewhat larger, so-called thermally significant blood vessels with diameters in the range from 80 µm to 1 mm. The modes of heat transport differ from those of mass transport, not simply


because these vessels are larger, but because they have a different geometrical arrangement than the vessels primarily responsible for mass transport. Many capillary beds approximate a uniformly spaced array of parallel vessels that can be well modeled by the Krogh cylinder model. In contrast, the thermally significant vessels are in a tree-like arrangement that typically undergoes several generations of branching within the size range of interest and are often found as countercurrent pairs in which the artery and vein may be separated by one vessel diameter or less. Moreover, the vascular architecture of the thermally significant vessels is less well characterized than that of either the primary mass exchange vessels or the larger, less numerous supply vessels that carry blood over large distances in the body. There are too few supply vessels to contribute much to the overall energy balance in the tissue, but they are often far from thermal equilibrium with the surrounding tissue producing large local perturbations in the tissue temperature. Much of the microvascular heat exchange occurs as the blood branches from the larger supply vessels into the more numerous and densely spaced thermally significant blood vessels. While the details of the vascular architecture for particular organs have been well characterized in individual cases, variability among individuals makes the use of such published data valid only in a statistical sense. Imaging technology to map the thermally significant blood vessels in any given individual is not readily available. An additional challenge arises from the spatial and temporal variability of the blood flow in tissue. The thermoregulatory system and the metabolic needs of tissues can change the blood perfusion rates in some tissues by a factor as great as 15 to 25.

98.2 Basic Concepts For purposes of thermal analysis, vascular tissues are generally assumed to consist of two interacting subvolumes, a solid tissue subvolume and a blood subvolume which contains flowing blood. These subvolumes thermally interact through the walls of the blood vessels where heat, but little mass, is exchanged. Because the tissue subvolume can transport heat by conduction alone, it may be modeled by the standard heat diffusion equation [6]

( )

r ∂Tt r , t r r ˙ ∇⋅ kt ∇Tt r , t + qt′′′ r , t = ρt ct ∂t

( )

( )

(98.1)

where Tt is the local tissue temperature, kt is the thermal conductivity of the tissue, q· t is the rate of volumetric heat generation from metabolism or external source, ρt is the tissue density, and ct is the tissue specific heat. The properties used in Eq. (98.1) may be assumed to be bulk properties that average over the details of the interstitial fluid, extracellular matrix, and cellular content of the tissue. In the blood subvolume, heat may also be transported by advection which adds a blood velocity dependent term as given by [6]

( )

r ∂Tb r , t r r r r r ∇⋅ kb∇Tb r , t − ρbcbub r , t ⋅∇Tb r , t + q˙ b′′′ r , t = ρbcb ∂t

( )

→

( )

( )

( )

(98.2)

where ub is the local blood velocity and all other parameters pertain to the local properties of the blood. Potential energy, kinetic energy, pulsatility, and viscous dissipation effects are normally taken to be negligible. → → At the internal boundary on the vessel walls we expect a continuity of heat flux kb∇Tb(rw , t) = kt∇Tt(rw , t) → → → and temperature Tb(rw , t) = Tt(rw , t) where rw represents points on the vessel wall. Few attempts have been made to solve Eqs. (98.1) and (98.2) exactly, primarily due to the complexity of the vascular architecture and the paucity of data on the blood velocity field in any particular instance. The sections


that follow present approaches to the problem of microvascular heat transport that fall broadly into the categories of vascular models that consider the response of one or a few blood vessels to their immediate surroundings and continuum models that seek to average the effects of many blood vessels to obtain a single field equation that may be solved for a local average of the tissue temperature.

98.3 Vascular Models Most vascular models are based on the assumption that the behavior of blood flowing in a blood vessel is formally similar to that of a fluid flowing steadily in a roughly circular tube (See Fig. 98.1), that is [7]

πra2ρbcbu

( ) = q ′( s )

dTa s ds

(98.3)

—

where Ta is the mixed mean temperature of the blood for a given vessel cross section, ra is the vessel – radius, u is mean blood speed in the vessel, q′(s) is the rate at which heat conducts into the vessel per unit length, and s is the spatial coordinate along the vessel axis. For a vessel that interacts only with a cylinder of adjacent tissue, we have

( ( ) ( ))

()

q′ s = U ′2πra Tt s − Ta s

(98.4)

where U ′ is the overall heat transfer coefficient between the tissue and the blood. Typically, the thermal resistance inside the blood vessel is much smaller than that in the tissue cylinder so we may approximate — U ′2πra ≈ ktσ where the conduction shape factor σ relating local tissue temperature Tt(s) to the blood temperature is given by

σ≈

FIGURE 98.1 coordinates.

2π r  ln t   ra 

(98.5)

Representative tissue cylinder surrounding a blood vessel showing the radial and axial position


Equilibration Lengths One of the most useful concepts that arises from the simple vascular model presented above is the equilibration length Le, which may be defined as the characteristic length over which the blood changes — temperature from an inlet temperature Tao to eventually equilibrate with tissue at a constant temperature — Tt . The solution for Eqs. (98.3) and (98.4) under these conditions is given by

()

Ta s − Tt

 s = exp  −  Tao − Tt  Le 

(98.6)

where the equilibration length is given by

Le =

πra2ρbcbu kt σ

(98.7)

Chen and Holmes [8] found that vessels with diameters of about 175 µm have an anatomical length comparable to their thermal equilibration length, thus making vessels of this approximate size the dominant site of tissue-blood heat exchange. Accordingly, these vessels are known as the thermally significant blood vessels. Much smaller vessels, while more numerous, carry blood that has already equilibrated with the surrounding tissue. Much larger vessels, while not in equilibrium with the surrounding tissue, are too sparsely spaced to contribute significantly to the overall energy balance [7]. Even though the larger vessels do not exchange large quantities of heat with the tissue subvolume, they cannot be ignored because these vessels produce large local perturbations to the tissue temperature and form a source of blood for tissues that is at a much different temperature than the local tissue temperature.

Countercurrent Heat Exchange Thermally significant blood vessels are frequently found in closely spaced countercurrent pairs. Only a slight modification to the preceding formulas is needed for heat exchange between adjacent arteries and veins with countercurrent flow [9]

( ( ) ( ))

()

q′ s = kt σ ∆ Tv s − Ta s

(98.8)

—

where Tv(s) is the mixed mean temperature in the adjacent vein and the conduction shape factor is given approximately by [6]

σ∆ ≈

2π  w2 − r2 − r2  a v cosh   r r 2   a v

(98.9)

−1

where w is the distance between the vessel axes and rv is the radius of the vein. The blood temperatures in the artery and vein must be obtained simultaneously, but still yield an equilibration length of the form given in Eq. (98.7). Substitution of representative property values, blood speeds, and vessel dimensions reveals that countercurrent vessels have equilibration lengths that are about one third that of isolated vessels of similar size [9]. Based on this observation, the only vessels that participate significantly in the overall energy balance in the tissue are those larger than about 50 µm in diameter.


TABLE 98.1

Shape Factors for Various Vascular Geometries

Geometry

Ref.

Single vessel to skin surface Single vessel to tissue cylinder Countercurrent vessel to vessel Countercurrent vessels to tissue cylinder Countercurrent vessels with a thin tissue layer Multiple parallel vessels Vessels near a junction of vessels

[9] [8] [9, 40] [24] [41] [42] [43]

Typical dimension of blood vessels are available in Chapter 1, Tables 1.3 and 1.4 of this handbook.

The shape factors given above are only rough analytical approximations that do not include the effects of finite thermal resistance within the blood vessels and other geometrical effects. The reader is referred to Table 98.1 for references that address these issues.

98.4 Heat Transfer Inside of a Blood Vessel A detailed analysis of the heat transfer between the blood vessel wall and the mixed mean temperature of the blood can be done using standard heat transfer methods

()

( () )

q′ s = hπd Tw s − Tb

(98.10)

where d is the vessel diameter, Tw(s) is the vessel wall temperature, and the convective heat transfer coefficient h may be found from Victor and Shah’s [10] recommendation that

NuD =

(

hd = 4 + 0.155 exp 1.58 log10 Gz kb

)

Gz < 103

(98.11)

where Gz is the Graetz number defined as

Gz =

ρbcbud 2 kb L

(98.12)

where L is the vessel length. See also Barozzi and Dumas [11].

98.5 Models of Perfused Tissues Continuum Models Continuum models of microvascular heat transfer are intended to average over the effects of many vessels so that the blood velocity field need not be modeled in detail. Such models are usually in the form of a modified heat diffusion equation in which the effects of blood perfusion are accounted for by one or more additional terms. These equations then can be solved to yield a sort of moving average of the local temperature that does not give the details of the temperature field around every individual vessel, but provides information on the broad trends in the tissue temperature. The temperature they predict may be defined as


TABLE 98.2

Representative Thermal Property Values

Tissue

Thermal Conductivity (W/m-K)

Thermal Diffusivity (m2/s)

Aorta Fat of spleen Spleen Pancreas Cerebral cortex Renal cortex Myocardium Liver Lung Adenocarcinoma of breast Resting muscle bone Whole blood (21˚C) Plasma (21˚C) Water

0.461 [16] 0.3337 [44] 0.5394 [44] 0.5417 [44] 0.5153 [44] 0.5466 [44] 0.5367 [44] 0.5122 [44] 0.4506 [44] 0.5641 [44] 0.478 [50] 0.492 [50] 0.570 [50] 0.628 [6]

1.25 × 10–7 [16] 1.314 × 10–7 [44] 1.444 × 10–7 [44] 1.702 × 10–7 [44] 1.468 × 10–7 [44] 1.470 × 10–7 [44] 1.474 × 10–7 [44] 1.412 × 10–7 [44] 1.307 × 10–7 [44] 1.436 × 10–7 [44] 1.59 × 10–7 [50] 1.19 × 10–7 [50] 1.21 × 10–7 [50] 1.5136 × 10–7 [6]

Perfusion (m3/m3-sec)

0.023 [45] 0.0091 [45] 0.0067 [46] 0.077 [47] 0.0188 [48] 0.0233 [49]

0.0007 [48]

All conductivities and diffusivities are from humans at 37˚C except the value for skeletal muscle which is from sheep at 21˚C. Perfusion values are from various mammals as noted in the references. Significant digits do not imply accuracy. The temperature coefficient for thermal conductivity ranges from –0.000254 to 0.0039 W/mK-˚C with 0.001265 W/m-K-˚C typical of most tissues as compared to 0.001575 W/mK-˚C for water [44]. The temperature coefficient for thermal diffusivity ranges from –4.9 × 10–10 m2/s-˚C to 8.4 × 10–10 m2/s-˚C with 5.19 × 10–10 m2/s-˚C typical of most tissues as compared to 4.73 × 10–10 m2/s-˚C for water [44]. The values provided in this table are representative values presented for tutorial purposes. The reader is referred to the primary literature for values appropriate for specific design applications.

( )

r 1 Tt r , t = δV

r ∫ T (r ′, t )dV ′ δV

(98.13)

t

where δV is a volume that is assumed to be large enough to encompass a reasonable number of thermally significant blood vessels, but much smaller than the scale of the tissue as a whole. Much of the confusion concerning the proper form of the bioheat equation stems from the difficulty in precisely defining such a length scale. Unlike a typical porous medium, such as water percolating through sand where the grains of sand fall into a relatively narrow range of length scales, blood vessels form a branching structure with length scales spaning many orders of magnitude. Formulations Pennes Heat Sink Model In 1948, physiologist Harry Pennes modeled the temperature profile in the human forearm by introducing the assumptions that the primary site of equilibration was the capillary bed and that each volume of tissue has a supply of arterial blood that is at the core temperature of the body. The Pennes’ Bioheat equation has the form [12]

( )

r ∂Tt r , t r r r r ∇⋅ k∇Tt r , t + ω˙ b r , t ρbcb Ta − Tt r , t + q˙ ′′′ r , t = ρc ∂t

( ) ( ) (

( ))

( )

(98.14)

where ω· b is taken to be the blood perfusion rate in volume of blood per unit volume of tissue per unit time and Ta is an arterial supply temperature which is generally assumed to remain constant and equal


to the core temperature of the body, nominally 37˚C. The other thermal parameters are taken to be effective values that average over the blood and tissue subvolumes. Major advantages of this formulation are that it is readily solvable for constant parameter values, requires no anatomical data, and in the absence of independent measurement of the actual blood rate and heat generation rate gives two adjust· able parameters (ωb(→r,t) and Ta) that can be used to fit the majority of the experimental results available. On the downside, the model gives no prediction of the actual details of the vascular temperatures, the actual blood perfusion rate is usually unknown and not exactly equal to ω· b that best fits the thermal data, the assumption of constant arterial temperature is not generally valid, and, based on the equilibration length studies presented in the previous section, thermal equilibration occurs prior to the capillary bed. Despite these weaknesses, the Pennes formulation is the primary choice of modelers. Equilibration prior to the capillary bed does not invalidate the model as long as the averaging volume is large enough to encompass many vessels of the size in which equilibration does occur and as long as the venous return does not exchange significant quantities of heat after leaving the equilibrating vessels. As long as ω· b and Ta are taken as adjustable, curve-fitting parameters rather than literally as the perfusion rate and arterial blood temperature, the model may be used fruitfully, provided that the results are interpreted accordingly. Directed Perfusion Some of the shortcomings of the Pennes model were addressed by Wulff [13] in a formulation that is essentially the same as that used for common porous media

( )

r ∂Tt r , t r r r r r ˙ ∇⋅ k∇Tt r , t − ρcu r , t ⋅∇Tt r , t + q′′′ r , t = ρc ∂t

( )

( )

( )

( )

(98.15)

→

where u is a velocity averaged over both the tissue and blood subvolumes. Among the difficulties with this model are that it is valid only when the tissue and blood are in near-thermal equilibrium and when the averaging volume is small enough to prevent adjacent arteries and veins from canceling out their contributions to the average velocity, thus erroneously suggesting that the blood perfusion has no net effect on the tissue heat transfer. Eq. (98.15) is rarely applied in practical situations, but served as an important conceptual challenge to the Pennes formulation in the 1970s and 1980s. Effective Conductivity Model The oldest continuum formulation is the effective conductivity model

( )

r ∂Tt r , t r r ∇⋅ keff ∇Tt r , t + q˙ ′′′ r , t = ρt ct ∂t

( )

( )

(98.16)

where the effective conductivity is comprised of the intrinsic thermal conductivity of the tissue and a perfusion dependent increment. In principle, an effective conductivity can be defined from any known heat flow and temperature difference, that is

keff =

()

q ∆T

 L f   A

(98.17)

where f AL is a function of geometry with dimensions length–1 (for example, ∆x/A in a slab geometry). Originally introduced as an empirical quantity [14], the effective conductivity has been linked to the Pennes formulation in the measurement for blood perfusion rates via small, heated, implanted probes [15-17]. In 1985, Weinbaum and Jiji [18] theoretically related the effective conductivity to the blood flow and anatomy for a restricted class of tissues and heating conditions which are dominated by a closely spaced artery-vein architecture and which can satisfy the constraint [19]


(

dTt 1 d Ta + Tv ≈ ds 2 ds

)

(98.18)

Here the effective conductivity is a tensor quantity related to the flow and anatomy according to [18]

 π2ρ2c 2nr 4u 2 cos2 φ  b b a keff = kt 1 +  kt2 σ ∆  

(98.19)

where the enhancement is in the direction of the vessel axes and where n is the number of artery-vein pairs per unit area and φ is the angle of the vessel axes relative to the temperature gradient. The near equilibrium between the tissue and the blood vessels needed for this model to be valid is likely to occur only in the outer few millimeters near the skin and when the volumetric heat source is neither intense nor localized. Artery-vein pairs have been shown to act like a single highly conductive fiber even when the near equilibrium condition in Eq. (98.18) is violated [20]. The radius of the thermally equivalent fiber is given by

( )

rfiber = wra

12

(98.20)

and its conductivity is given by

(ρ c u ) r cosh (w r ) = 2

k fiber

b b

−1

3 a

a

(98.21)

wkt

Under these non-equilibrium conditions, the tissue-blood system acts like a fiber-composite material, but cannot be well modeled as a single homogeneous material with effective properties. Combination Recognizing that several mechanisms of heat transport may be at play in tissue, Chen and Holmes [8] suggested the following formulation which incorporates the effects discussed above

∇⋅ keff

( )

r ∂Tt r , t r r r r r r r r * ˙ ˙ r , t ∇Tt r , t + ω b r , t ρbcb Ta − Tt r , t − ρbcbub r , t ⋅∇Tt r , t + q′′′ r , t = ρt ct ∂t (98.22)

( ) ( ) ( ) (

( ))

( )

( )

( )

where Ta* is the temperature exiting the last artery that is individually modeled. The primary value of this formulation is its conceptual generality. In practice, this formulation is difficult to apply because it requires knowledge of a great many adjustable parameters, most of which have not been independently measured to date. Heat Sink Model with Effectiveness Using somewhat different approaches, Brinck and Werner [21] and Weinbaum et al. [22] have proposed that the shortcomings of the Pennes model can be overcome by introducing a heat transfer effectiveness factor ε to modify the heat sink term as follows:

( )

r ∂Tt r , t r r r r r ∇⋅ kt ∇Tt r , t + ε r , t ω˙ b r , t ρbcb T r , t − Ta + q˙ ′′′ r , t = ρt ct t ∂t

( ) ( ) ( ) (( )


) ( )

(98.23)

where 0 ≤ ε ≤ 1. In the Brinck and Werner formulation, ε is a curve-fitting parameter that allows the actual (rather than the thermally equivalent) perfusion rate to be used [21]. Weinbaum et al. provide an analytical result for ε that is valid for blood vessels smaller than 300 µm diameter in skeletal muscle [22]. In both formulations, ε < 1 arises from the countercurrent heat exchange mechanism that shunts heat directly between the artery and vein without requiring the heat-carrying blood to first pass through the smaller connecting vessels. Typical values for the effectiveness are in the range 0.6 ≤ ε ≤ 0.8 [22].

Multi-Equation Models The value of the continuum models is that they do not require a separate solution for the blood subvolume. In each continuum formulation, the behavior of the blood vessels is modeled by introducing assumptions that allow solution of only a single differential equation. But by solving only one equation, all detailed information on the temperature of the blood in individual blood vessels is lost. Several investigators have introduced multi-equation models that typically model the tissue, arteries, and veins as three separate, but interacting, subvolumes [9, 23-27]. As with the other non-Pennes formulations, these methods are difficult to apply to particular clinical applications, but provide theoretical insights into microvascular heat transfer.

Vascular Reconstruction Models As an alternative to the three equation models, a more complete reconstruction of the vasculature may be used along with a scheme for solving the resulting flow, conduction, and advection equations [28-33]. Since the reconstructed vasculature is similar to the actual vasculature only in a statistical sense, these models provide the mean temperature as predicted by the continuum models, as well as insight into the mechanisms of heat transport, the sites of thermal interaction, and the degree of thermal perturbations produced by vessels of a given size, but they cannot provide the actual details of the temperature field in a given living tissue. These models tend to be computationally intensive due to the high spatial resolution needed to account for all of the thermally significant blood vessels.

98.6 Parameter Values Thermal Properties The intrinsic thermal properties of tissues depend strongly on their composition. Cooper and Trezek [34] recommend the following correlations for thermal conductivity:

(

)

k = ρx10−3 0.628 f water + 0.117 f proteins + 0.231 f fats W m − K ;

(98.24)

c p = 4, 200 f water + 1, 090 f proteins + 2, 300 f fats J kg − K ;

(98.25)

specific heat:

and density:

ρ=

1 f water 1, 000 + f proteins 1, 540 + f fats 815

kg m 3

where fwater , fproteins , and ffats are the mass fractions of water, proteins, and fats, respectively.


(98.26)

Thermoregulation Humans maintain a nearly constant core temperature through a combination of physiological and behavioral responses to the environment. For example, heat loss or gain at the skin surface may be modified by changes in the skin blood flow, the rate of sweating, or clothing. In deeper tissues, the dependence of the blood perfusion rate, the metabolic heat generation rate, and vessel diameters depend on the environmental and physiological conditions in a complex, organ-specific manner. The blood perfusion varies widely among tissue types and for some tissues can change dramatically depending on the metabolic or thermoregulatory needs of the tissue. The situation is further complicated by the feedback control aspects of the thermoregulatory systems that utilize a combination of central and peripheral temperature sensors as well as local and more distributed actuators. The following examples are provided to illustrate some of the considerations, not to exhaustively explore this complicated issue. A model of the whole body is typically needed even for a relatively local stimulus, especially when the heat input represents a significant fraction of the whole body heat load. The reader is referred to extensive handbook entries on environmental response for more information [35, 36]. Whole body models of the thermoregulatory system are discussed in Wissler [37]. Chato [1] suggests that the temperature dependence of the blood perfusion effect can be approximated by a scalar effective conductivity

( Tt − 25 )   keff = 4.82 − 4.448331.00075−1.575 W m−K  

(98.27)

which is intended for use in Eq. (98.16). Under conditions of local hyperthermia, where the heated volume is small compared to the body as a whole, the blood perfusion rate in skin and muscle may increase by roughly nine- or tenfold as the tissue warms to about 44˚C and then drop again for higher temperatures. In contrast, tumors, the typical target of local hyperthermia, thermoregulate erratically and undergo a lesser change in blood flow up to 42˚C and then may drop to near zero for higher temperatures [38]. The metabolic rate may also undergo thermoregulatory changes. For example, the temperature dependence of the metabolism in the leg muscle and skin may be modeled with [39]

( )[(

q˙ m′′′ = 170 2

To −Tt

) 10]

W m3

(98.28)

The metabolic rate and blood flow may also be linked through processes that reflect the fact that sustained increased metabolic activity generally requires increased blood flow.

Clinical Heat Generation Thermal therapies such as hyperthermia treatment rely on local heat generation rates several orders of magnitude greater than that produced by metabolism. Under these circumstances, the metabolic heat generation is often neglected with little error.

98.7 Solutions of Models The steady-state solution with spatially and temporally constant parameter values including the rate of heat generation for a tissue half-space with a fixed temperature on the skin Tskin is given by 12  1 2     ˙   ˙  ω bρbcb  ω bρbcb  q˙ ′′′       Tt x = Tskin exp − x + Ta + x  1 − exp −     k     ˙ ω bρbcb    kt  t      

()


(98.29)

This solution reveals that perturbations to the tissue temperature decay exponentially with a characteristic length of

 kt  Lc =    ω˙ bρbcb 

12

(98.30)

which for typical values of the perfusion rate ω· b = 0.1 × 10–3 to 3.0 × 10–3 m3/m3-sec yields Lc = 6.5 × 10–3 to 36 × 10–3 m (Fig. 98.2). The transient solution of Pennes’ bioheat equation with constant perfusion rate for an initial uniform temperature of To , in the absence of any spatial dependence, is

  ω˙ ρ c      ω˙ ρ c    q˙ ′′′   b b b Tt t = To exp − b b b  t  + Ta + 1 − exp −   t  c ω˙ bρbcb   ρ       ρt ct     t t  

()

(98.31)

Here the solution reveals a characteristic time scale

tc =

ρt ct ω˙ bρbcb

(98.32)

that has typical values in the range of tc = 300 to 10,000 sec (Fig. 98.3). This solution is valid only when thermoregulatory changes in the perfusion rate are small or occur over a much longer time than the characteristic time scale tc . Numerical solution of the heat sink model is readily obtained by standard methods such as finite differences, finite element, boundary element, and Green’s functions provided that the parameter values and appropriate boundary conditions are known.

FIGURE 98.2

One-dimensional steady-state solution of Pennes bioheat equation for constant parameter values.


FIGURE 98.3 Transient solution of Pennes bioheat heat equation for constant parameter values in the absence of spatial effects.

Defining Terms Conduction shape factor: Dimensionless factor used to account for the geometrical effects in steadystate heat conduction between surfaces at different temperatures. Effective conductivity: A modified thermal conductivity that includes the intrinsic thermal conductivity of the tissue as well as a contribution from blood perfusion effects. Equilibration length: Characteristic length scale over which blood in a blood vessel will change temperature in response to surrounding tissue at a different temperature. Perfusion rate: Quantity of blood provided to a unit of tissue per unit time. Specific heat: Quantity of energy needed for a unit temperature increase for a unit of mass. Thermal conductivity: Rate of energy transfer by thermal conduction for a unit temperature gradient per unit of cross-sectional area. Thermally significant vessel: Blood vessels large enough and numerous enough to contribute significantly to overall heat transfer rates in tissue.

References 1. J. C. Chato, “Fundamentals of bioheat transfer,” in Thermal Dosimetry and Treatment Planning, M. Gautherie, Ed. New York: Springer-Verlag, 1990, pp. 1-56. 2. C. K. Charny, “Mathematical models of bioheat transfer,” in Bioengineering Heat Transfer: Advances in Heat Transfer, vol. 22, Y. I. Cho, Ed. Boston: Academic Press, 1992, pp. 19-155. 3. M. Arkin, L. X. Xu, and K. R. Holmes, “Recent developments in modeling heat transfer in blood perfused tissues,” IEEE Trans Biomed Eng, 41, 97-107, 1994. 4. T. K. Eto and B. Rubinsky, “Bioheat Transfer,” in Introduction to Bioengineering, S. A. Berger, W. Goldsmith, and E. R. Lewis, Eds. Oxford: Oxford University Press, 1996, pp. 203-227. 5. K. R. Diller, “Modeling of bioheat transfer processes at high and low temperatures,” in Bioengineering Heat Transfer: Advances in Heat Transfer, vol. 22, Y. I. Cho, Ed. Boston: Academic Press, 1992, pp. 157-357. 6. F. P. Incropera and D. P. DeWitt, Fundamentals of Heat and Mass Transfer, 4th ed. New York: John Wiley & Sons, 1996. 7. J. C. Chato, “Heat transfer to blood vessels,” J Biomechan Eng, 102, 110-118, 1980. © 2000 by CRC Press LLC

8. M. M. Chen and K. R. Holmes, “Microvascular contributions in tissue heat transfer,” Ann NY Acad Sci, 325, 137-50, 1980. 9. S. Weinbaum, L. M. Jiji, and D. E. Lemons, “Theory and experiment for the effect of vascular microstructure on surface tissue heat transfer-part I: anatomical foundation and model conceptualization,” J Biomech Eng, 106, 321-330, 1984. 10. S. A. Victor and V. L. Shah, “Steady state heat transfer to blood flowing in the entrance region of a tube,” Int J Heat Mass Trans, 19, 777-783, 1976. 11. G. S. Barozzi and A. Dumas, “Convective heat transfer coefficients in the circulation,” J Biomech Eng, 113, 308-313, 1991. 12. H. H. Pennes, “Analysis of tissue and arterial blood temperatures in the resting forearm,” J Appl Phys, 1, 93-122, 1948. 13. W. Wulff, “The energy conservation equation for living tissue,” IEEE Trans Biomed Eng, 21, 494-495, 1974. 14. H. C. Bazett and B. McGlone, “Temperature gradients in tissues in man,” Am J Physiol, 82, 415-428, 1927. 15. W. Perl, “Heat and matter distribution in body tissues and the determination of tissue bloodflow by local clearance methods,” J Theoret Biol, 2, 201-235, 1962. 16. J. W. Valvano, L. J. Hayes, A. J. Welch, and S. Bajekal, “Thermal conductivity and diffusivity of arterial walls,” 1984. 17. H. Arkin, K. R. Holmes, and M. M. Chen, “A technique for measuring the thermal conductivity and evaluating the “apparent conductivity” concept in biomaterials,” J Biomech Eng, 111, 276-282, 1989. 18. S. Weinbaum and L. M. Jiji, “A new simplified bioheat equation for the effect of blood flow on local average tissue temperature,” J Biomech Eng, 107, 131-139, 1985. 19. S. Weinbaum and L. M. Jiji, “The matching of thermal fields surrounding countercurrent microvessels and the closure approximation in the Weinbaum-Jiji equation,” J Biomech Eng, 111, 271-275, 1989. 20. J. W. Baish, “Heat Transport by countercurrent blood vessels in the presence of an arbitrary temperature gradient,” J Biomech Eng, 112, 207-211, 1990. 21. H. Brinck and J. Werner, “Efficiency function: improvement of classical bioheat approach,” J Appl Phys, 77, 1617-1622, 1994. 22. S. Weinbaum, L. X. Xu, L. Zhu, and A. Ekpene, “A new fundamental bioheat equation for muscle tissue: part I blood perfusion term,” J Biomech Eng, 119, 278-288, 1997. 23. L. M. Jiji, S. Weinbaum, and D. E. Lemons, “Theory and experiment for the effect of vascular microstructure on surface tissue heat transfer-part II: model formulation and solution,” J Biomech Eng, 106, 331-341, 1984. 24. J. W. Baish, P. S. Ayyaswamy, and K. R. Foster, “Small-scale termperature fluctuations in perfused tissue during local hyperthermia,” J Biomech Eng, 108, 246-250, 1986. 25. J. W. Baish, P. S. Ayyaswamy, and K. R. Foster, “Heat transport mechanisms in vascular tissues: a model comparison,” J Biomech Eng, 108, 324-331, 1986. 26. C. K. Charny and R. L. Levin, “Heat transfer normal to paired arterioles and venules embedded in perfused tissue during hyperthermia,” J Biomech Eng, 110, 277-282, 1988. 27. C. K. Charny and R. L. Levin, “Bioheat transfer in a branching countercurrent network during hyperthermia,” J Biomech Eng, 111, 263-270, 1989. 28. J. W. Baish, “Formulation of a statistical model of heat transfer in perfused tissue,” J Biomech Eng, 116, 521-527, 1994. 29. H. W. Huang, Z. P. Chen, and R. B. Roemer, “A counter current vascular network model of heat transfer in tissues,” J Biomech Eng, 118, 120-129, 1996. 30. J. F. Van der Koijk, J. J. W. Lagendijk, J. Crezee, J. De Bree, A. N. T. J. Kotte, G. M. J. Van Leeuwen, and J. J. Battermann, “The influence of vasculature on temperature distributions in MECS interstitial hyperthermia: importance of longitudinal control,” Int J Hyperthermia, 13, 365-386, 1997.


31. G. M. J. van Leeuwen, A. N. T. J. Kotte, J. Crezee, and J. J. W. Lagendijk, “Tests of the geometrical description of blood vessels in a thermal model using counter-current geometries,” Phys Med Biol, 42, 1515-1532, 1997. 32. G. M. J. van Leeuwen, A. N. T. J. Kotte, J. de Bree, J. F. van der Koijk, J. Crezee, and J. J. W. Lagendijk, “Accuracy of geometrical modeling of heat transfer from tissue to blood vessels,” Phys Med Biol, 42, 1451-1460, 1997. 33. G. M. J. Van Leeuwen, A. N. T. J. Kotte, and J. J. W. Lagendijk, “A flexible algorithm for construction of 3-D vessel networks for use in thermal modeling,” IEEE Trans Biomed Eng, 45, 596-605, 1998. 34. T. E. Cooper and G. J. Trezek, “Correlation of thermal properties of some human tissue with water content,” Aerospace Med, 42, 24-27, 1971. 35. ASHRAE, “Physiological Principles and Thermal Comfort,” in 1993 ASHRAE Handbook: Fundamentals. Atlanta, Georgia: ASHRAE Inc., 1993, 8.1-8.29. 36. M. J. Fregly and C. M. Blatteis, “Section 4: Environmental Physiology,” in Handbook of Physiology, vol. I. New York: American Physiological Society, 1996. 37. E. H. Wissler, “Mathematical simulation of human thermal behavior using whole body models,” in Heat Transfer in Medicine and Biology: Analysis and Applications, vol. 1, A. Shitzer and R. C. Eberhart, Eds. New York: Plenum Press, 1985, 325-373. 38. G. M. Hahn, “Blood Flow,” in Physics and Technology of Hyperthermia, S. B. Field and C. Franconi, Eds. Dordrecht: Martinus Nijhoff Publishers, 1987, 441-447. 39. J. W. Mitchell, T. L. Galvez, J. Hangle, G. E. Myers, and K. L. Siebecker, “Thermal response of human legs during cooling,” J Appl Phys, 29, 859-856, 1970. 40. E. H. Wissler, “An analytical solution of countercurrent heat transfer between parallel vessels with a linear axial temperature gradient,” J Biomech Eng, 110, 254-256, 1988. 41. L. Zhu and S. Weinbaum, “A model for heat transfer from embedded blood vessels in twodimensional tissue preparations,” J Biomech Eng, 117, 64-73, 1995. 42. Cousins, “On the Nusselt number in heat transfer between multiple parallel blood vessels,” J Biomech Eng, 119, 127-129, 1997. 43. J. W. Baish, J. K. Miller, and M. J. Zivitz, “Heat transfer in the vicinity of the junction of two blood vessels,” in Advances in Bioheat and Mass Transfer: Microscale Analysis of Thermal Injury Processes, Instrumentation, Modeling and Clinical Applications, vol. HTD-Vol. 268, R. B. Roemer, Ed. New York: ASME, 1993, 95-100. 44. J. W. Valvano, J. R. Cochran, and K. R. Diller, “Thermal conductivity and diffusivity of biomaterials measured with self-heated thermistors,” Int J Thermophys, 6, 301-311, 1985. 45. M. A. Kapin and J. L. Ferguson, “Hemodynamic and regional circulatory alterations in dog during anaphylactic challenge,” Am J Physiol, 249, H430-H437, 1985. 46. C. W. Haws and D. D. Heistad, “Effects of nimodipine on cerebral vasoconstrictor responses,” Am J Physiol, 247, H170-H176, 1984. 47. J. C. Passmore, R. E. Neiberger, and S. W. Eden, “Measurement of intrarenal anatomic distribution of krypton-85 in endotoxic shock in dogs,” Am J Physiol, 232, H54-H58, 1977. 48. R. C. Koehler, R. J. Traystman, and J. Jones, M. D., “Regional blood flow and O2 transport during hypoxic and CO hypoxia in neonatal and sheep,” Am J Physiol, 248, H118-H124, 1985. 49. W. C. Seyde, L. McGowan, N. Lund, B. Duling, and D. E. Longnecker, “Effects of anesthetics on regional hemodynamics in normovolemic and hemorrhaged rats,” Am J Physiol, 249, H164-H173, 1985. 50. T. A. Balasubramaniam and H. F. Bowman, “Thermal conductivity and thermal diffusivity of biomaterials: A simultaneous measurement technique,” Trans ASME, J Biomech Eng, 99, 148-154, 1977.


Saltzman. W. M. “Interstitial Transport in the Brain: Principles for Local Drug Delivery.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

99 Interstitial Transport in the Brain: Principles for Local Drug Delivery 99.1 99.2 99.3 99.4 99.5

Introduction Implantable Controlled Delivery Systems for Chemotherapy Drug Transport After Release from the Implant Application of Diffusion–Elimination Models to Intracranial BCNU Delivery Systems Limitations and Extensions of the Diffusion– Elimination Model Failure of the Model in Certain Situations • Effect of Drug Release Rate • Determinants of Tissue Penetration • Effect of Fluid Convection • Effect of Metabolism

99.6

W. Mark Saltzman Cornell University

99.7

New Approaches to Drug Delivery Suggested by Modeling Conclusion

99.1 Introduction Traditional methods for delivering drugs to the brain are inadequate. Many drugs, particularly watersoluble or high molecular weight compounds, do not enter the brain following systemic administration because they permeate through blood capillaries very slowly. This blood–brain barrier (BBB) severely limits the number of drugs that are candidates for treating brain disease. Several strategies for increasing the permeability of brain capillaries to drugs have been tested. Since the BBB is generally permeable to lipid soluble compounds which can dissolve and diffuse through endothelial cell membranes [1, 2], a common approach for enhancing brain delivery of compounds is chemical modification to enhance lipid solubility [3]. Unfortunately, lipidization approaches are not useful for drugs with molecular weight larger than 1,000. Another approach for increasing permeability is the entrapment of drugs in liposomes [4], but delivery may be limited by liposome stability in the plasma and uptake at other tissue sites. Specific nutrient transport systems in brain capillaries can be used to facilitate drug entry into the brain. L-dopa (L-3,4-dihydroxyphenylalanine), a metabolic precursor of dopamine, is transported across endothelial cells by the neutral amino acid transport system [5]. L-dopa permeates through capillaries into the striatal tissue, where it is decarboxylated to form dopamine. Therefore, systemic administration of L-dopa is often beneficial to patients with Parkinson’s disease. Certain protein modifications, such as cationization [6] and anionization [7], produce enhanced uptake in the brain. Modification of drugs [8, 9] by linkage to an anti-transferrin receptor antibody also appears to enhance transport into the brain.


This approach depends on receptor-mediated transcytosis of transferrin-receptor complexes by brain endothelial cells; substantial uptake also occurs in the liver. The permeability of brain capillaries can be transiently increased by intra-arterial injection of the hyperosmolar solutions, which disrupt interendothelial tight junctions [10]. But BBB disruption affects capillary permeability throughout the brain, enhancing permeability to all compounds in the blood, not just the agent of interest. Intraventricular therapy, where agents are administered directly into the CSF of the ventricles, results in high concentrations within the brain tissue, but only in regions immediately surrounding the ventricles [11, 12]. Because the agent must diffuse into the brain parenchyma from the ventricles, and because of the high rate of clearance of agents in the CNS into the peripheral circulation, this strategy cannot be used to deliver agents deep into the brain. Because of the difficulty in achieving therapeutic drug levels by systemic administration, methods for direct administration of drugs into the brain parenchyma have been tested. Drugs can be delivered directly into the brain tissue by infusion, implantation of a drug-releasing matrix, or transplantation of drugsecreting cells [13]. These approaches provide sustained drug delivery that can be limited to specific sites, localizing therapy to a brain region. Because these methods provide a localized and continuous source of active drug molecules, the total drug dose can be less than needed with systemic administration. With polymeric controlled release, the implants can also be designed to protect unreleased drug from degradation in the body and to permit localization of extremely high doses (up to the solubility of the drug) at precisely defined locations in the brain. Infusion systems require periodic refilling; the drug is usually stored in a liquid reservoir at body temperature and many drugs are not stable under these conditions. This chapter describes the transport of drug molecules that are directly delivered into the brain. For purposes of clarity, a specific example is considered: polymeric implants that provide controlled release of chemotherapy. The results can be extended to other modes of administration [13, 14] and other types of drug agents [15].

99.2 Implantable Controlled Delivery Systems for Chemotherapy The kinetics of drug release from a controlled release system are usually characterized in vitro, by measuring the amount of drug released from the matrix into a well-stirred reservoir of phosphate buffered water or saline at 37°C. Controlled release profiles for some representative anticancer agents are shown in Fig. 99.1; all of the agents selected for these studies—1,3-bis(2-chloroethyl)-1-nitrosourea (BCNU), 4-HC, cisplatin, and taxol—are used clinically for chemotherapy of brain tumors. The controlled release period can vary from several days to many months, depending on properties of the drug, the polymer, and the method of formulation. Therefore, the delivery system can be tailored to the therapeutic situation by manipulation of implant properties. The release of drug molecules from polymer matrices can be regulated by diffusion of drug through the polymer matrix or degradation of the polymer matrix. In many cases (including the release of BCNU, cisplatin, and 4HC from the degradable matrices shown in Fig. 99.1), drug release from biodegradable polymers appears to be diffusion-mediated, probably because the time for polymer degradation is longer than the time required for drug diffusion through the polymer. In certain cases linear release, which appears to correlate with the polymer degradation rate, can be achieved; this might be the case for taxol release from the biodegradable matrix (Fig. 99.1), although the exceedingly low solubility of taxol in water may also contribute substantially to the slowness of release. For diffusion-mediated release, the amount of drug released from the polymer is proportional to the concentration gradient of the drug in the polymer. By performing a mass balance on drug molecules within a differential volume element in the polymer matrix, a conservation equation for drug within the matrix is obtained:

∂C p ∂t © 2000 by CRC Press LLC

= D p∇2C p

(99.1)

FIGURE 99.1 Controlled release of anticancer compounds from polymeric matrices. (a) Release of cisplatin (circles) from a biodegradable copolymer of fatty acid dimers and sebacic acid, p(FAD:SA), initially containing 10% drug (see [32, 33] for details). (b) Release of BCNU from EVAc matrices (circles), polyanhydride matrices p(CPP:SA) (squares), and p(FAD:SA) (triangles) matrices initially containing 20% drug. (c) Release of BCNU (squares), 4HC (circles), and taxol (triangles) from p(CPP:SA) matrices initially containing 20% drug. Note that panel (c) has two time axes: the lower axis applies to the release of taxol and the upper axis applies to the release of BCNU and 4HC.

where Cp is the local concentration of drug in the polymer and Dp is the diffusion coefficient of the drug in the polymer matrix. This equation can be solved, with appropriate boundary and initial conditions, to obtain the cumulative mass of drug released as a function of time; the details of this procedure are described elsewhere [16]. A useful approximate solution, which is valid for the initial 60% of release, is:

Mt = 4Mo

Di : pt

(99.2)

πL2

where Mt is the cumulative mass of drug released from the matrix, Mo is the initial mass of drug in the matrix, and L is the thickness of the implant. By comparing Eq. (99.2) to the experimentally determined profiles, the rate of diffusion of the agent in the polymer matrix can be estimated (Table 99.1).

TABLE 99.1 Diffusion Coefficients for Chemotherapy Drug Release from Biocompatible Polymer Matricesa Drug

Polymer

Initial loading (%)

Dp (cm2/sec)

Cisplatin BCNU BCNU BCNU

P(FAD:SA) EVAc P(FAD:SA) P(CPP:SA)

10 20 20 20

4HC Taxol

P(CPP:SA) P(CPP:SA)

20 20

6.8 × 10–9 1.6 × 10–8 6.9 × 10–8 2.3 × 10–8 (panel b) 2.0 × 10–8 (panel c) 3.1 × 10–10 n.a.

a Diffusion coefficients were obtained by comparing the experimental data show in Fig. 99.1 to Eq. (99.2) and determining the best value of the diffusion coefficient to represent the data. Abbreviations: not applicable (n.a.).


99.3 Drug Transport After Release from the Implant Bypassing the BBB is necessary, but not sufficient for effective drug delivery. Consider the consequences of implanting a delivery system, such as one of the materials characterized above, within the brain. Molecules released into the interstitial fluid in the brain extracellular space must penetrate into the brain tissue to reach tumor cells distant from the implanted device. Before these drug molecules can reach the target site, however, they might be eliminated from the interstitium by partitioning into brain capillaries or cells, entering the cerebrospinal fluid, or being inactivated by extracellular enzymes. Elimination always accompanies dispersion; therefore, regardless of the design of the delivery system, one must understand the dynamics of both processes in order to predict the spatial pattern of drug distribution after delivery. The polymer implant is surrounded by biological tissue, composed of cells and an extracellular space (ECS) filled with extracellular fluid (ECF). Immediately following implantation, drug molecules escape from the polymer and penetrate the tissue. Once in the brain tissue, drug molecules (1) diffuse through the tortuous ECS in the tissue, (2) diffuse across semipermeable tissue capillaries to enter the systemic circulation and, therefore, are removed from the brain tissue, (3) diffuse across cell membranes by passive, active, or facilitated transport paths, to enter the intracellular space, (4) transform, spontaneously or by an enzyme-mediated pathway, into other compounds, and (5) bind to fixed elements in the tissue. Each of these events influence drug therapy: diffusion through the ECS is the primary mechanism of drug distribution in brain tissue; elimination of the drug occurs when it is removed from the ECF or transformed; and binding or internalization may slow the progress of the drug through the tissue. A mass balance on a differential volume element in the tissue [17] gives a general equation describing drug transport in the region near the polymer [18]:

( )

∂Ct ∂B + v • ∇Ct = Db∇2Ct + Re Ct − ∂t ∂t

(99.3)

where C is the concentration of the diffusible drug in the tissue surrounding the implant (g/cm3 tissue), v is the fluid velocity (cm/sec), Db is the diffusion coefficient of the drug in the tissue (cm2/sec), Re(C) is the rate of drug elimination from the ECF (g/sec-cm3 tissue), B is the concentration of drug bound or internalized in cells (g/cm3 tissue), and t is the time following implantation.1 In deriving this equation, the conventions developed by Nicholson [20], based on volume-averaging in a complex medium, and Patlak and Fenstermacher [18] were combined. In this version of the equation, the concentrations C and B and the elimination rate Re(C) are defined per unit volume of tissue. Db is an effective diffusion coefficient, which must be corrected from the diffusion coefficient for the drug in water to account for the tortuosity of the ECS. When the binding reactions are rapid, the amount of intracellular or bound drug can be assumed to be directly proportional, with an equilibrium coefficient Kbind, to the amount of drug available for internalization or binding:

B = K bind C

(99.4)

Substitution of Eq. (99.4) into Eq. (99.3) yields, with some simplification:

(

( )

∂Ct 1 = Db∇2Ct + Re Ct − v • ∇Ct ∂t 1 + K bind

)

(99.5)

1An alternate form of this equation, which accounts more rigorously for transfer of drug between different phases in the tissue, is also available [19].¸


The drug elimination rate, Re(C), can be expanded into the following terms:

C  V C Re Ct = kbbb  t − C plasma + max t + kneCt  ε ecs  K m + Ct

( )

(99.6)

where kbbb is the permeability of the BBB (defined based on concentration in the ECS), Cplasma is the concentration of drug in the blood plasma, Vmax and Km are Michaelis-Menton constants, and kne is a first order rate constant for drug elimination due to non-enzymatic reactions. For any particular drug, some of the rate constants may be very small, reflecting the relative importance of each mechanism of drug elimination. If it is assumed that the permeability of the BBB is low (Cpl C) and the concentration of drug in the brain is sufficiently low so that any enzymatic reactions are in the first order regime (C Km), Eq. (99.6) can be reduced to:

( )

− Re Ct =

kbbb V Ct + max Ct + kneCt = kappCt Km ε ecs

(99.7)

where kapp is a lumped first order rate constant. With these assumptions, Eq. (99.4) can be simplified by definition of an apparent diffusion coefficient, D*, and an apparent first order elimination constant, k *:

∂Ct v • ∇Ct = D * ∇2Ct + k * Ct − ∂t 1 + K bind where k* =

k app 1 + K bind

and D =

Db 1 + K bind

(99.8)

.

Boundary and initial conditions are required for solution of differential Eq. (99.8). If a spherical implant of radius R is implanted into a homogeneous region of the brain, at a site sufficiently far from anatomical boundaries, the following assumptions are reasonable:

Ct = 0 for t = 0; r > R

(99.9)

Ct = Ci for t > 0; r = R

(99.10)

Ct = 0 for t > 0; r → ∞

(99.11)

In many situations, drug transport due to bulk flow can be neglected. This assumption (v is zero) is common in previous studies of drug distribution in brain tissue [18]. For example, in a previous study of cisplatin distribution following continuous infusion into the brain, the effects of bulk flow were found to be small, except within 0.5 mm of the site of infusion [21]. In the cases considered here, since drug molecules enter the tissue by diffusion from the polymer implant, not by pressure-driven flow of a fluid, no flow should be introduced by the presence of the polymer. With fluid convection assumed negligible, the general governing equation in the tissue, Eq. (99.8), reduces to:

∂Ct = D * ∇2Ct + k * Ct ∂t

(99.12)

The no-flow assumption may be inappropriate in certain situations. In brain tumors, edema and fluid movement are significant components of the disease. In addition, some drugs can elicit cytotoxic edema.


Certain drug/polymer combinations can also release drugs in sufficient quantity to create density-induced fluid convection. Equation (99.12), with conditions (99.9) through (99.11), can be solved by Laplace transform techniques [13] to yield:

   ζ −1 ζ −1 Ct 1  − φ τ  + exp φ ζ − 1 erfc  + φ τ  = exp − φ ζ − 1 erfc  Ci 2ζ      2 τ  2 τ

[ ( )]

[ ( )]

(99.13)

where the dimensionless variables are defined as follows:

ζ=

r D*t k* ;τ= 2 ;φ=R R D* R

(99.14)

The differential equation also has a steady-state solution, which is obtained by solving Eq. (99.12) with the time derivative set equal to zero and subject to the boundary conditions (99.9) and (99.10):

[ ( )]

Ct 1 = exp − φ ζ − 1 Ci ζ

(99.15)

Figure 99.2 shows concentration profiles calculated using Eqs. (99.13) and (99.15). In this situation, which was obtained using reasonable values for all of the parameters, steady-state is reached approximately 1 h after implantation of the delivery device. The time required to achieve steady-state depends on the rate of diffusion and elimination, as previously described [22], but will be significantly less than 24 h for most drug molecules.

FIGURE 99.2 Concentration profiles after implantation of a spherical drug-releasing implant. (Panel a, Transient) Solid lines represent the transient solution to Eq. (99.12) [i.e., Eq. (99.13)] with the following parameter values: D* = 4 × 10–7 cm2/s; R = 0.032 cm; k* = 1.9 × 10–4 s–1 (t1/2 = 1 h). The dashed line represents the steady-state solution [i.e., Eq. (99.15)] for the same parameters. (Panel b, Steady-state) Solid lines in this plot represent Eq. (99.15) with the following parameters: D* = 4 × 10–7 cm2/s; R = 0.032 cm. Each curve represents the steady-state concentration profile for drugs with different elimination half-lives in the brain, corresponding to different dimensionless moduli, φ: t1/2 = 10 min (φ = 1.7); 1 h (0.7); 34 h (0.12); and 190 h (0.016).


99.4 Application of Diffusion-Elimination Models to Intracranial BCNU Delivery Systems The preceding mathematical analysis, which assumes diffusion and first-order elimination in the tissue, agrees well with experimental concentration profiles obtained after implantation of controlled release polymers (Fig. 99.3). At 3, 7, and 14 d after implantation of a BCNU-releasing implant, the concentration profile at the site of the implant was very similar. The parameter values (obtained by fitting Eq. (99.15) to the experimental data) were consistent with parameters obtained using other methods [19], suggesting that diffusion and first-order elimination were sufficient to account for the pattern of drug concentration observed during this period. Parameter values were similar at 3, 7, and 14 d, indicating that the rates of drug release, drug dispersion, and drug elimination did not change during this period. This equation has been compared to concentration profiles measured for a variety of molecules delivered by polymer implants to the brain—dexamethasone [22], molecular weight fractions of dextran [23], nerve growth factor in rats [24, 25], BCNU in rats [19], rabbits [26], and monkeys [27]. In each of these cases, the steady-state diffusion-elimination model appears to capture most of the important features of drug transport. This model can be used to develop guidelines for the design of intracranial delivery systems. Table 99.2 lists some of the important physical and biological characteristics of a few compounds that have been considered for interstitial delivery to treat brain tumors. When the implant is surrounded by tissue, the maximum rate of drug release is determined by the solubility of the drug, Cs, and the rate of diffusive transport through the tissue:

 dM t  ∂Ct 2  dt  = Maximum flux × Surface area = − D * ∂r 4 πR   max R

(

) (

)

(99.16)

FIGURE 99.3 Concentration profiles after implantation of a BCNU-releasing implant. Solutions to Eq. (99.15) were compared to experimental data obtained by quantitative autoradiographic techniques. The solid lines in the three panels labeled 3, 7, and 14 d were all obtained using the following parameters: R = 0.15 cm; φ = 2.1; and CI = 0.81 mM. The solid line in the panel labeled 1 d was obtained using the following parameters: R = 0.15 cm; φ = 0.7; and CI = 1.9 mM. Modified from [19].


TABLE 99.2

Implant Design Applied to Three Chemotherapy Compoundsa

Molecular weight (daltons) Solubility (mM) log10K k* (h) D* (10–7 cm2/sec) Toxic concentration in culture (µM) Maximum release rate (mg/d) Implant lifetime at max rate (d) Maximum concentration in tissue for 1-week-releasing implant (mM) RT (mm)

BCNU

4HC

Methotrexate

214 12 1.53 70 14 25 1.2 0.85 1.5 1.3

293 100 0.6 70 14 10 14 0.07 1.1 2.5

454 100 –1.85 1 5 0.04 17 0.06 1.8 5

a K is the octanol:water partition coefficient, k is the rate of elimination due to permeation through capillaries, Db is the diffusion coefficient of the drug in the brain. The following values are assumed, consistent with our results from polymer delivery to rats and rabbits: radius of spherical implant, R = 1.5 mm; mass of implant, M = 10 mg; drug loading in implant, Load = 10%.

Evaluating the derivative in Eq. (99.16) from the steady-state concentration profile [Eq. (99.15)] yields:

 dM t   dt  = 8πD * Cs R   max

(99.17)

Regardless of the properties of the implant, it is not possible to release drug into the tissue at a rate faster than determined by Eq. (99.17). If the release rate from the implant is less than this maximum rate, Ci (the concentration in the tissue immediately outside the implant) is less than the saturation concentration, Cs . The actual concentration Ci can be determined by balancing the release rate from the implant (dMt /dt, which can be determined from Eq. (99.2) provided that diffusion is the mechanism of release from the implant) with the rate of penetration into the tissue obtained by substituting CI for Cs in Eq. (99.17):

C* =

dM t  1  dt  8πD * R 

(99.18)

The effective region of therapy can be determined by calculating the distance from the surface of the implant to the point where the concentration drops below the cytotoxic level (estimated as the cytotoxic concentration determined from in vitro experiments). Using Eq. (99.15), and defining the radial distance for effective treatment as RT , yields:

Ccytotoxic Ci

=

 R k *  RT   − 1  exp− R RT D *  R   

(99.19)

Alternately, an effective penetration distance, dP , can be defined as the radial position at which the drug concentration has dropped to 10% of the peak concentration:

0.10 =

 R k *  d P   exp − R −1  dP D *  R    

(99.20)

These equations provide quantitative criteria for evaluating the suitability of chemotherapy agents for direct intracranial delivery (Table 99.2).


99.5 Limitations and Extensions of the Diffusion-Elimination Model Failure of the Model in Certain Situations The previous section outlined one method for analysis of drug transport after implantation of a drugreleasing device. A simple pseudo-steady-state equation [Eq. (99.15)] yielded simple guidelines [Eq. (99.16 through 99.19] for device design. Because the assumptions of the model were satisfied over a substantial fraction of the release period (days 3 to 14, based on the data shown in Fig. 99.3), this analysis may be useful for predicting the effects of BCNU release from biodegradable implants. Pseudosteady-state assumptions are reasonable during this period of drug release, presumably because the time required to achieve steady-state (which is on the order of minutes) is much less than the characteristic time associated with changes in the rate of BCNU release from the implant (which is on the order of days). But experimental concentration profiles measured 1 d after implantation were noticeably different: the peak concentration was substantially higher and the drug penetration into the surrounding tissue was deeper (see the left-hand panel of Fig. 99.3). This behavior cannot be easily explained by the pseudosteady-state models described above. For example, if the difference observed at 1 d represents transient behavior, the concentration observed at a fixed radial position should increase with time (Fig. 99.2); in contrast, the experimental concentration at any radial position on Day 1 is higher than the concentration measured at that same position on subsequent days.

Effect of Drug Release Rate Alternately, the observed difference at 1 d might be due to variability in the rate of BCNU release from the polymer implant over this period, with transport characteristics in the tissue remaining constant. When similar BCNU-releasing implants are tested in vitro, the rate of drug release did decrease over time (Fig. 99.1). Equation (99.18) predicts the variation in peak concentration with release rate; the twofold higher concentration observed at the interface on Day 1 (as compared to Days 3 through 14) could be explained by a twofold higher release rate on Day 1. But the effective penetration distance, dP , does not depend on release rate. Experimentally measured penetration distances are ~1.4 mm on Days 3, 7, and 14 and ~5 mm on Day 1. This difference in penetration is shown more clearly in the Day 1 panel of Fig. 99.3: the dashed line shows the predicted concentration profile if k* and D* were assumed equal to the values obtained for Days 3, 7, and 14. Changes in the rate of BCNU release are insufficient to explain the differences observed experimentally.

Determinants of Tissue Penetration Penetration of BCNU is enhanced at Day 1 relative to penetration at Days 3, 7, and 14. For an implant of fixed size, penetration depends only on the ratio of elimination rate to diffusion rate: k*/D*. Increased penetration results from a decrease in this ratio (Fig. 99.2), which could occur because of either a decreased rate of elimination (smaller k*) or an increased rate of diffusion (larger D*). But there are no good reasons to believe that BCNU diffusion or elimination are different on Day 1 than on Days 3 through 14. With its high lipid solubility, BCNU can diffuse readily through brain tissue. In addition, elimination of BCNU from the brain occurs predominantly by partitioning into the circulation; since BCNU can permeate the capillary wall by diffusion, elimination is not a saturable process. Perhaps the enhanced penetration of BCNU is due to the presence of another process for drug dispersion, such as bulk fluid flow, which was neglected in the previous analysis. The diffusion/elimination model compares favorably with available experimental data, but the assumptions used in predicting concentration profiles in the brain may not be appropriate in all cases. Deviations from the predicted concentration profiles may occur due to extracellular fluid flows in the brain, complicated patterns of drug binding, or multistep elimination pathways. The motion of interstitial fluid in


the vicinity of the polymer and the tumor periphery may not always be negligible, as mentioned above. The interstitial fluid velocity is proportional to the pressure gradient in the interstitium; higher interstitial pressure in tumors—due to tumor cell proliferation, high vascular permeability, and the absence of functioning lymphatic vessels—may lead to steep interstitial pressure gradients at the periphery of the tumor [28]. As a result, interstitial fluid flows within the tumor may influence drug transport. A drug at the periphery of the tumor must overcome outward convection to diffuse into the tumor. Similarly, local edema after surgical implantation of the polymer may cause significant fluid movement in the vicinity of the polymer. More complete mathematical models that include the convective contribution to drug transport are required.

Effect of Fluid Convection When bulk fluid flow is present (v ≠ 0), concentration profiles can be predicted from Eq. (99.8), subject to the same boundary and initial conditions [Eqs. (99.9) through (99.11)]. In addition to Eq. (99.8), continuity equations for water are needed to determine the variation of fluid velocity in the radial direction. This set of equations has been used to describe concentration profiles during microinfusion of drugs into the brain [14]. Relative concentrations were predicted by assuming that the brain behaves as a porous medium (i.e., velocity is related to pressure gradient by Darcy’s law). Water introduced into the brain can expand the interstitial space; this effect is balanced by the flow of water in the radial direction away from the infusion source and, to a lesser extent, by the movement of water across the capillary wall. In the presence of fluid flow, penetration of drug away from the source is enhanced (Fig. 99.4). The extent of penetration depends on the velocity of the flow and the rate of elimination of the drug. These calculations were performed for macromolecular drugs, which have limited permeability across the brain capillary wall. The curves indicate steady-state concentration profiles for three different proteins with

FIGURE 99.4 Concentration profiles predicted in the absence (solid lines) and presence (dashed lines) of interstitial fluid flow. Solid lines were obtained from Eq. (99.15) with the following parameter values: R = 0.032 cm; D = 4 × 10–7 cm2/sec; and k* = ln(2)/t1/2 where t1/2 is either 10 min, 1 h, or 33.5 h as indicated on the graph. Dashed lines were obtained from [14] using the same parameter values and an infusion rate of 3 µL/min. The dashed line indicating the interstitial flow calculation for the long-lived drug (t1/2 = 33.5 h) was not at steady-state, but at 12 h after initiation of the flow.


TABLE 99.3 Interstitial Fluid Velocity as a Function of Radial Position During Microinfusiona Radial Position (mm)

Interstitial Velocity (µm/sec)

2 5 10 20 a

5.0 0.8 0.2 0.05

Calculated by method reported in [14].

metabolic half-lives of 10 min, 1 h, or 33.5 h. In the absence of fluid flow, drugs with longer half-lives penetrate deeper into the tissue [solid lines in Fig. 99.4 were obtained from Eq. (99.15)]. This effect is amplified by the presence of flow (dashed lines in Fig. 99.4). During microinfusion, drug is introduced by pressure-driven fluid flow from a small catheter. Therefore, pressure gradients are produced in the brain interstitial space, which lead to fluid flow through the porous brain microenvironment. Volumetric infusion rates of 3 µL/min were assumed in the calculations reproduced in Fig. 99.4. Since loss of water through the brain vasculature is small, the velocity can be determined as a function of radial position:

vr =

q 4 πr 2 ε

(99.21)

where q is the volumetric infusion rate and ε is the volume fraction of the interstitial space in the brain (approximately 0.20). Fluid velocity decreases with radial distance from the implant (Table 99.3); but at all locations within the first 20 mm of the implant site, predicted velocity was much greater than the velocities reported previously during edema or tumor growth in the brain. The profiles predicted in Fig. 99.4 were associated with the introduction of substantial volumes of fluid at the delivery site. Flow-related phenomena are probably less important in drug delivery by polymer implants. Still, this model provides useful guidelines for predicting the influence of fluid flow on local rates of drug movement. Clearly, the effect of flow velocity on drug distribution is substantial (Fig. 99.4). Even relatively low flows, perhaps as small as 0.03 µm/sec, are large enough to account for the enhancement in BCNU penetration observed at Day 1 in Fig. 99.3.

Effect of Metabolism The metabolism, elimination, and binding of drug are assumed to be first order processes in our simple analysis. This assumption may not be realistic in all cases, especially for complex agents such as proteins. The metabolism of drugs in normal and tumor tissues is incompletely understood. Other cellular factors (e.g., the heterogeneity of tumor-associated antigen expression and multidrug resistance) that influence the uptake of therapeutic agents may not be accounted for by our simple first order elimination. Finally, changes in the brain that occur during the course of therapy are not properly considered in this model. Irradiation can be safely administered when a BCNU-loaded polymer has been implanted in monkey brains, suggesting the feasibility of adjuvant radiotherapy. However, irradiation also causes necrosis in the brain. The necrotic region has a lower perfusion rate and interstitial pressure than tumor tissue; thus, the convective interstitial flow due to fluid leakage is expected to be smaller. Interstitial diffusion of macromolecules is higher in tumor tissue than in normal tissue, as the tumor tissue has larger interstitial spaces [29]. The progressive changes in tissue properties—due to changes in tumor size, irradiation, and activity of chemotherapy agent—may be an important determinant of drug transport and effectiveness of therapy in the clinical situation.


99.6 New Approaches to Drug Delivery Suggested by Modeling Mathematical models, which describe the transport of drug following controlled delivery, can predict the penetration distance of drug and the local concentration of drug as a function of time and location. The calculations indicate that drugs with slow elimination will penetrate deeper into the tissue. The modulus φ, which represents the ratio of elimination to diffusion rates in the tissue, provides a quantitative criterion for selecting agents for interstitial delivery. For example, high molecular weight dextrans were retained longer in the brain space, and penetrated a larger region of the brain than low molecular weight molecules following release from an intracranial implant [23]. This suggests a strategy for modifying molecules to improve their tissue penetration by conjugating active drug molecules to inert polymeric carriers. For conjugated drugs, the extent of penetration should depend on the modulus φ for the conjugated compound as well as the degree of stability of the drug-carrier linkage. The effects of conjugation and stability of the linkage between drug and carrier on enhancing tissue penetration in the brain have been studied in a model system [30]. Methotrexate (MTX)-dextran conjugates with different dissociation rates were produced by linking MTX to dextran (molecular weight 70,000) through a short-lived ester bond (half-life ≈ 3 d) and a longer-lived amide bond (half-life > 20 d). The extent of penetration for MTX-dextran conjugates was studied in three-dimensional human brain tumor cell cultures; penetration was significantly enhanced for MTX-dextran conjugates and the increased penetration was correlated with the stability of the linkage. These results suggest that modification of existing drugs may increase their efficacy against brain tumors when delivered directly to the brain interstitium.

99.7 Conclusion Controlled release polymer implants are a useful new technology for delivering drugs directly to the brain interstitium. This approach is already in clinical use for treatment of tumors [31], and could soon impact treatment of other diseases. The mathematical models described in this paper provide a rational framework for analyzing drug distribution after delivery. These models describe the behavior of chemotherapy compounds very well and allow prediction of the effect of changing properties of the implant or the drug. More complex models are needed to describe the behavior of macromolecules, which encounter multiple modes of elimination and metabolism and are subject to the effects of fluid flow. In addition, variations on this approach may be useful for analyzing drug delivery in other situations.

References 1. Lieb, W. and W. Stein, Biological Membranes behave as Non-Porous Polymeric Sheets with Respect to the Diffusion of Non-Electrolytes. Nature, 1969. 224: 240-249. 2. Stein, W.D., The Movement of Molecules across Cell Membranes. 1967, New York: Academic Press. 3. Simpkins, J., N. Bodor, and A. Enz, Direct Evidence for Brain-Specific Release of Dopamine from a Redox Delivery System. Journal of Pharmaceutical Sciences, 1985. 74: 1033-1036. 4. Gregoriadis, G., The carrier potential of liposomes in biology and medicine. The New England Journal of Medicine, 1976. 295: 704-710. 5. Cotzias, C.G., M.H. Van Woert, and L.M. Schiffer, Aromatic amino acids and modificatino of parkinsonism. The New England Journal of Medicine, 1967. 276: 374-379. 6. Triguero, D., J.B. Buciak, J. Yang, and W.M. Pardridge, Blood-brain barrier transport of cationized immunoglobulin G: enhanced delivery compared to native protein. Proceedings of the National Academy of Sciences USA, 1989. 86: 4761-4765. 7. Tokuda, H., Y. Takakura, and M. Hashida, Targeted delivery of polyanions to the brain. Proceed. Intern. Symp. Control. Rel. Bioact. Mat., 1993. 20: 270-271.


8. Friden, P., L. Walus, G. Musso, M. Taylor, B. Malfroy, and R. Starzyk, Anti-transferrin receptor antibody and antibody-drug conjugates cross the blood-brain barrier. Proceedings of the National Academy of Sciences USA, 1991. 88: 4771-4775. 9. Friden, P.M., L.R. Walus, P. Watson, S.R. Doctrow, J.W. Kozarich, C. Backman, H. Bergman, B. Hoffer, F. Bloom, and A.-C. Granholm, Blood-brain barrier penetration and in vivo activity of an NGF conjugate. Science, 1993. 259: 373-377. 10. Neuwelt, E., P. Barnett, I. Hellstrom, K. Hellstrom, P. Beaumier, C. McCormick, and R. Weigel, Delivery of melanoma-associated immunoglobulin monoclonal antibody and Fab fragments to normal brain utilizing osmotic blood-brain barrier disruption. Cancer Research, 1988. 48: 4725-4729. 11. Blasberg, R., C. Patlak, and J. Fenstermacher, Intrathecal chemotherapy: brain tissue profiles after ventriculocisternal perfusion. The Journal of Pharmacology and Experimental Therapeutics, 1975. 195: 73-83. 12. Yan, Q., C. Matheson, J. Sun, M.J. Radeke, S.C. Feinstein, and J.A. Miller, Distribution of intracerebral ventricularly administered neurotrophins in rat brain and its correlation with Trk receptor expression. Experimental Neurology, 1994. 127: 23-36. 13. Mahoney, M.J. and W.M. Saltzman, Controlled release of proteins to tissue transplants for the treatment of neurodegenerative disorders. Journal of Pharmaceutical Sciences, 1996. 85(12): 1276-1281. 14. Morrison, P.F., D.W. Laske, H. Bobo, E.H. Oldfield, and R.L. Dedrick, High-flow microinfusion: tissue penetration and pharmacodynamics. American Journal of Physiology, 1994. 266: R292-R305. 15. Haller, M.F. and W.M. Saltzman, Localized delivery of proteins in the brain: Can transport be customized? Pharmaceutical Research, 1998. 15: 377-385. 16. Wyatt, T.L. and W.M. Saltzman, Protein delivery from non-degradable polymer matrices, in Protein Delivery-Physical Systems, L. Saunders and W. Hendren, Editors. 1997, Plenum Press: New York, NY. p. 119-137. 17. Bird, R.B., W.E. Stewart, and E.N. Lightfoot, Transport Phenomena. 1960, New York: John Wiley & Sons. 780. 18. Patlak, C. and J. Fenstermacher, Measurements of dog blood-brain transfer constants by ventriculocisternal perfusion. American Journal of Physiology, 1975. 229: 877-884. 19. Fung, L., M. Shin, B. Tyler, H. Brem, and W.M. Saltzman, Chemotherapeutic drugs released from polymers: distribution of 1,3-bis(2-chloroethyl)-1-nitrosourea in the rat brain. Pharmaceutical Research, 1996. 13: 671-682. 20. Nicholson, C., Diffusion from an injected volume of a substance in brain tissue with arbitrary volume fraction and tortuosity. Brain Research, 1985. 333: 325-329. 21. Morrison, P. and R.L. Dedrick, Transport of cisplatin in rat brain following microinfusion: an analysis. Journal of Pharmaceutical Sciences, 1986. 75: 120-128. 22. Saltzman, W.M. and M.L. Radomsky, Drugs released from polymers: diffusion and elimination in brain tissue. Chemical Engineering Science, 1991. 46: 2429-2444. 23. Dang, W. and W.M. Saltzman, Dextran retention in the rat brain following controlled release from a polymer. Biotechnology Progress, 1992. 8: 527-532. 24. Krewson, C.E., M. Klarman, and W.M. Saltzman, Distribution of nerve growth factor following direct delivery to brain interstitium. Brain Research, 1995. 680: 196-206. 25. Krewson, C.E. and W.M. Saltzman, Transport and elimination of recombinant human NGF during long-term delivery to the brain. Brain Research, 1996. 727: 169-181. 26. Strasser, J.F., L.K. Fung, S. Eller, S.A. Grossman, and W.M. Saltzman, Distribution of 1,3-bis(2chloroethyl)-1-nitrosourea (BCNU) and tracers in the rabbit brain following interstitial delivery by biodegradable polymer implants. The Journal of Pharmacology and Experimental Therapeutics, 1995. 275(3): 1647-1655. 27. Fung, L.K., M.G. Ewend, A. Sills, E.P. Sipos, R. Thompson, M. Watts, O.M. Colvin, H. Brem, and W.M. Saltzman, Pharmacokinetics of interstitial delivery of carmustine, 4-hydroperoxycyclophosphamide, and paclitaxel from a biodegradable polymer implant in the monkey brain. Cancer Research, 1998. 58: 672-684.


28. Jain, R.K., Barriers to drug delivery in solid tumors. Scientific American, 1994. 271(1): 58-65. 29. Clauss, M.A. and R.K. Jain, Interstitial transport of rabbit and sheep antibodies in normal and neoplastic tissues. Cancer Research, 1990. 50: 3487-3492. 30. Dang, W.B., O.M. Colvin, H. Brem, and W.M. Saltzman, Covalent coupling of methotrexate to dextran enhances the penetration of cytotoxicity into a tissue-like matrix. Cancer Research, 1994. 54: 1729-1735. 31. Brem, H., S. Piantadosi, P.C. Burger, M. Walker, R. Selker, N.A. Vick, K. Black, M. Sisti, S. Brem, G. Mohr, P. Muller, R. Morawetz, S.C. Schold, and P.-B.T.T. Group, Placebo-controlled trial of safety and efficacy of intraoperative controlled delivery by biodegradable polymers of chemotherapy for recurrent gliomas. Lancet, 1995. 345: 1008-1012. 32. Dang, W. and W.M. Saltzman, Controlled release of macromolecules from a biodegradable polyanhydride matrix. Journal of Biomaterials Science, Polymer Edition, 1994. 6(3): 291-311. 33. Dang, W., Engineering Drugs and Delivery Systems for Brain Tumor Therapy. 1993, The Johns Hopkins University:


Tarbell, J. M., Qiu, Y. “Arterial Wall Mass Transport: The Possible Role of Blood Phase Resistance in the Localization of Arterial Disease" The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

100 Arterial Wall Mass Transport: The Possible Role of Blood Phase Resistance in the Localization of Arterial Disease 100.1

Steady-State Transport Modeling

100.2

Damkhöler Numbers for Important Solutes

Reactive Surface • Permeable Surface • Reactive Wall Adenosine Triphosphate (ATP) • Albumin and LDL • Oxygen

100.3

Sherwood Numbers in the Circulation

100.4

Non-Uniform Geometries Associated with Atherogenesis

Straight Vessels

Sudden Expansion • Stenosis • Bifurcation • Curvature

John M. Tarbell The Pennsylvania State University

Yuchen Qiu The Pennsylvania State University

100.5 100.6

Discussion Possible Role of Blood Phase Transport in Atherogenesis Direct Mechanical Effects on Endothelial Cells • Hypoxic Effect on Endothelial Cells • Hypoxia Induces VEGF

Atherosclerosis is a disease of the large arteries which involves a characteristic accumulation of high molecular weight lipoprotein in the arterial wall [1]. The disease tends to be localized in regions of curvature and branching in arteries where fluid shear stress (shear rate) is altered from its normal patterns in straight vessels [2]. The possible role of fluid mechanics in the localization of atherosclerosis has been debated for many years [3,4]. One possibility considered early on was that the blood phase resistance to lipid transport, which could be affected by local fluid mechanics, played a role in the focal accumulation of lipid in arteries. Studies by Caro and Nerem [5], however, showed that the uptake of lipid in arteries could not be correlated with fluid phase mass transport, leading to the conclusion that the wall (endothelium) and not the blood, was the limiting resistance to lipid transport. This suggested that fluid mechanical effects on macromolecular transport were the result of direct mechanical influences on the transport characteristics of the endothelium.


FIGURE 100.1 Schematic diagram of arterial wall transport processes showing the concentration profile of a solute which is being transported from the blood, where its bulk concentration is Cb, to the surface of the endothelium, where its concentration is Cs , then across the endothelium, where the subendothelial concentration is Cw , and finally to a minimum value within the tissue, Cmin . Transport of the solute in the blood phase is characterized by the mass transport coefficient, kL ; consumption of the solute at the endothelial surface is described by a first-order reaction with rate constant, kr ; movement of the solute across the endothelium depends on the permeability coefficient, Pe; ° and reaction of the solute within the tissue volume is quantified by a zeroeth order consumption rate, Q.

While the transport of large molecules such as low density lipoprotein (LDL) and other high molecular weight materials, which are highly impeded by the endothelium, may be limited by the wall and not the fluid (blood), other low molecular weight species which undergo rapid reaction on the endothelial surface (e.g., adenosine triphosphate—ATP) or which are consumed rapidly by the underlying tissue (e.g, oxygen) may be limited by the fluid phase. With these possibilities in mind, the purpose of this short review is to compare the rates of transport in the blood phase to the rates of reaction on the endothelial surface, the rates of transport across the endothelium, and the rates of consumption within the wall of several important biomolecules. It will then be possible to assess quantitatively the importance of fluid phase transport; to determine which molecules are likely to be affected by local fluid mechanics; to determine where in blood vessels these influences are most likely to be manifest; and finally, to speculate about the role of fluid phase mass transport in the localization of atherosclerosis.

100.1 Steady-State Transport Modeling Reactive Surface Referring to Fig. 100.1, we will assume that the species of interest is transported from the blood vessel lumen, where its bulk concentration is Cb, to the blood vessel surface, where its concentration is Cs , by a convective-diffusive mechanism which depends on the local fluid mechanics and can be characterized by a fluid-phase mass transfer coefficient kL (see [6] for further background). The species flux in the blood phase is given by

(

Js = k L C b = C s

)

(100.1)

At the endothelial surface, the species may undergo an enzyme-catalyzed surface reaction (e.g., the hydrolysis of ATP to ADP) which can be modeled using classical Michaelis-Menten kinetics with a rate given by


V=

VmaxC s k m + Cs

(100.2)

where Vmax is the maximum rate (high Cs) and km is the Michaelis constant. When Cs km, as is often the case, then the reaction rate is pseudo-first order

V = k r Cs

(100.3)

with the rate constant for the surface reaction given by kr = Vmax /km. At steady state, the transport to the surface is balanced by the consumption at the surface so that

(

)

k L C b − C s = k rC s

(100.4)

It will be convenient to cast this equation into a dimensionless form by multiplying it by d/D, where d is the vessel diameter and D is the diffusion coefficient of the transported species in blood, or the media of interest. Equation (100.4) then becomes

(

)

(100.5)

kL d D

(100.6)

Sh C b − C s = Da rC s where

Sh ≡

is the Sherwood number (dimensionless mass transfer coefficient), and

Da r ≡

kr d D

(100.7)

is the Damkhöler number (dimensionless reaction rate coefficient). Solving Eq. (100.5) for the surface concentration one finds

Cs C b =

1 1 + Da r Sh

(100.8)

When Dar Sh,

Cs = C b

(100.9)

and the process is termed “wall-limited” or “reaction-limited”. On the other hand, when Dar Sh,

 Sh  Cs =   Cb  Da r 

(100.10)

and the process is termed “transport-limited”or “fluid phase-limited”. It is in this transport-limited case that the surface concentration, and in turn the surface reaction rate, depend on the fluid mechanics © 2000 by CRC Press LLC

which determines the Sherwood number. It will therefore be useful to compare the magnitudes of Dar and Sh to determine whether fluid mechanics plays a role in the overall transport process of a surface reactive species.

Permeable Surface Many species will permeate the endothelium without reacting at the luminal surface (e.g., albumin, LDL) and their rate of transport (flux) across the surface layer can be described by

(

Js = Pe C s − C w

)

(100.11)

where Pe is the endothelial permeability coefficient and Cw is the wall concentration beneath the endothelium. If the resistance to transport offered by the endothelium is significant, then it will be reasonable to assume

Cw

Cs

(100.12)

so that at steady state when the fluid and surface fluxes balance,

(

)

k L C b − C s = PeC s

(100.13)

Multiplying Eq. (100.13) by d/D to introduce dimensionless parameters and then solving for the surface concentration leads to

Cs C b =

1 1 + Da e Sh

(100.14)

where Sh was defined in Eq. (100.6) and

Da e ≡

Ped D

(100.15)

is a Damkhöler number based on endothelial permeability. Equation (100.14) shows that when Dae Sh, the transport process is again “wall-limited”. When Dae Sh, fluid mechanics again becomes important through the Sherwood number.

Reactive Wall Oxygen is transported readily across the endothelium (Hellums), but unlike most proteins, is rapidly consumed by the underlying tissue. In this case it is fair to neglect the endothelial transport resistance (assume Cw = Cs), and then by equating the rate of transport to the wall with the (zeroeth order) consumption rate within the wall we obtain

(

)

° T k L C b − Cs = Q

(100.16)

° is the tissue consumption rate and T is the tissue thickness (distance from the surface to the where Q minimum tissue concentration—see Fig. 100.1). For the specific case of O2 transport, it is conventional to replace concentration (C) with partial pressure (P) through the Henry’s law relationship C = KP, where


K is the Henry’s law constant. Invoking this relationship and rearranging Eq. (100.16) into a convenient dimensionless form, we obtain

Ps Da = 1− w Pb Sh

(100.17)

where Sh was defined in Eq. (100.6), and Daw is another Damkhöler number based on the wall consumption rate

Da w =

° QTd KD Pb

(100.18)

Clearly when Daw Sh, the process is wall limited. But, as Daw → Sh, the process becomes limited by transport in the fluid phase (Ps → 0), and fluid mechanics plays a role. Because we are treating the tissue consumption rate as a zeroeth order reaction, the case Daw > Sh is not meaningful (Ps < O). In reality, as Sh is reduced, the tissue consumption rate must be reduced due to the lack of oxygen supply from the blood.

100.2 Damkhöler Numbers for Important Solutes A wide range of Damkhöler numbers characterize the transport of biomolecular solutes in vessel walls of the cardiovascular system, and in this section we focus on four important species as examples of typical biotransport processes: adenosine triphosphate (ATP), a species that reacts vigorously on the endothelial surface, albumin, and low density lipoprotein (LDL), species which are transported across a permeable endothelial surface; and oxygen, which is rapidly consumed within the vessel wall. Since most vascular disease (atherosclerosis) occurs in vessels between 3 and 10 mm in diameter, we use a vessel of 5 mm diameter to provide estimates of typical Damkhöler numbers.

Adenosine Triphosphate (ATP) ATP is degraded at the endothelial surface by enzymes (ectonucleotidases) to form adenosine diphosphate (ADP). The Michaelis-Menten kinetics for this reaction have been determined by Gordon et al. [7] using cultured porcine aortic endothelial cells: km = 249 µM, Vmax = 22 nmol/min/106 cells. Vmax can be converted to a molar flux by using a typical endothelial cell surface density of 1.2 × 105 cells/cm2, with the result that the pseudo-first order rate constant (Eq. 100.3) is kr = 1.77 × 10–4 cm/sec. Assuming a diffusivity of 5.0 × 10–6 cm2/s for ATP [8], and a vessel diameter of 5 mm, we find

Dar = 17.7

Albumin and LDL These macromolecules are transported across the endothelium by a variety of mechanisms including non-specific and receptor-mediated trancytosis, and paracellular transport through normal or “leaky” inter-endothelial junctions [9, 10]. In rabbit aortas, Truskey et al. [11] measured endothelial permeability to LDL and observed values on the order of Pe = 1.0 × 10–8 cm/sec in uniformly permeable regions, but found that permeability increased significantly in punctate regions associated with cells in mitosis to a level of Pe = 5 × 10–7 cm/sec. Using this range of values for Pe, assuming a diffusivity of 2.5 × 10–7 cm2/sec for LDL, and a vessel diameter of 5 mm, we find

Da e = 0.02 – 1.0


(LDL)

For albumin, Truskey et al. [12] reported values of the order Pe = 4.0 × 10–8 cm/sec in the rabbit aorta. This presumably corresponded to regions of uniform permeability. They did not report values in punctate regions of elevated permeability. More recently, Lever et al. [13] reported Pe values of similar magnitude in the thoracic and abdominal aorta as well as the carotid and renal arteries of rabbits. In the ascending aorta and pulmonary artery, however, they observed elevated permeability to albumin on the order of Pe = 1.5 × 10–7 cm/sec. Assuming a diffusivity of 7.3 × 10–7 cm2/sec for albumin, a vessel diameter of 5mm, and the range of Pe values described above, we obtain

Da e = 0.027 – 0.10

(albumin)

Oxygen The first barrier encountered by oxygen after being transported from the blood is the endothelial layer. Although arterial endothelial cells consume oxygen [14], the pseudo-first order rate constant for this consumption is estimated to be an order of magnitude lower than that of ATP, and it is therefore reasonable to neglect the endothelial cell consumption relative to the much more significant consumption by the underlying tissue. Liu et al. [15] measured the oxygen permeability of cultured bovine aortic and human umbilical vein endothelial cells and obtained values of 1.42 × 10–2 cm/sec for bovine cell monolayers and 1.96 × 10–2 cm/sec for human cell monolayers. Because the endothelial permeability to oxygen is so high, it is fair to neglect the transport resistance of the endothelium and to direct attention to the oxygen consumption rate within the tissue. To evaluate the Damkhöler number based on the tissue consumption rate [Eq. (100.17)], we turn to ° measured both in vivo and in vitro in dog, rabbit, and pig data of Buerk and Goldstick [16] for Q/(KD) ° blood vessels. The values of Q/(KD) reported by Buerk and Goldstick are based on tissue properties for KD. To translate these tissue values into blood values, as required in our estimates [Eq. (100.17)], we use the relationship (KD)tissue = ⅓ (KD)water suggested by Paul et al. [17] and assume (KD)blood = (KD)water . ° ranged from 1.29 × 105 torr/cm2 to 5.88 × 105 torr/cm2 in the In the thoracic aorta of dogs, Q/(KD) tissue. The thickness (distance to the minimum tissue O2 concentration) of the thoracic aorta was 250 µm and the diameter is estimated to be 0.9 cm [18]. PO2 measured in the blood (Pb) was 90 torr. Introducing these values into Eq. (100.17) we find:

Daw = 10.8 – 49.0

(thoracic aorta)

° In the femoral artery of dogs, Q/(KD) ranged from 35.2 × 105 torr/cm2 to 46.9 × 105 torr/cm2 in the tissue. The thickness of the femoral artery was 50 µm and the estimated diameter is 0.4 cm [18]. PO2 measured in the blood was about 80 torr. These values lead to the following estimates:

Daw = 29.3 – 39.1

(femoral artery)

100.3 Sherwood Numbers in the Circulation Straight Vessels For smooth, cylindrical tubes (a model of straight blood vessels) with well-mixed entry flow, one can invoke the thin concentration boundary layer theory of Lévêque [6] to estimate the Sherwood number in the entry region of the vessel where the concentration boundary is developing. This leads to

Sh = 1.08x * −1 3 1.30 x * −1 3


(constant wall concentration) (constant wall flux)

(100.19a) (100.19b)

TABLE 100.1

Transport Characteristics in a Straight Aorta

Species

Sc

x*

2,900 7,000 48,000 140,000

4.1 × 10 1.7 × 10–5 2.5 × 10–6 8.6 × 10–7

O2 ATP Albumin LDL

–5

Sh

Da

31.1 41.8 79.2 114

10.8–49.0 17.7 0.027–0.100 0.02–1.00

Note: d = 1 cm, x = 60 cm, Re = 500, ν = .035 cm2/sec.

where

x* =

x d Re ⋅ Sc

(100.20)

is a dimensionless axial distance which accounts for differing rates of concentration boundary layer growth due to convection and diffusion. In Eq. (100.20), Re = vd/ν is the Reynolds number, Sc = ν/D is the Schmidt number, and their product is the Péclet number. Equation (100.19) is quite accurate for distances from the entrance satisfying x* < .001. Sh continues to drop with increasing axial distance as the concentration boundary layer grows, as described by the classical Graetz solution of the analogous heat transfer problem [19]. When the concentration boundary layer becomes fully developed, Sh approaches its asymptotic minimum value,

Sh = 3.66

(constant wall concentration)

(100.21a)

(constant wall flux)

(100.21b)

Sh = 4.36

For a straight vessel, Sh cannot drop below these asymptotic values. Equations (100.19) and (100.21) also indicate that the wall boundary condition has little effect on the Sherwood number. It is instructive to estimate Sh at the end of a straight tube having dimensions and flow rate characteristics of the human aorta (actually a tapered tube). Table 100.1 compares Sh and Da (for O2, ATP, albumin, and LDL) at the end of a 60-cm long model aorta having a diameter of 1 cm and a flow characterized by Re = 500. Table 100.1 clearly reveals that for a straight aorta, transport is in the entry or Lévêque regime (x* 5 h in the presence of purified nucleases [25]. Like PO oligonucleotides, PS oligonucleotides are believed to be internalized by receptor mediated endocytosis [27]. Their affinity for complementary RNA is not as high as that observed with PO oligonucleotides, and they are more likely to bind non-specifically to proteins. Nevertheless, PS oligonucleotides have been shown to efficiently inhibit gene expression [28]. One concern with PS oligonucleotides (and all chemically modified oligonucleotides) is their potential for toxicity. The metabolic byproducts of PS oligonucleotides are not native to the cell. If these were incorporated into cellular DNA, they could cause mutations [29]. Studies have also shown that PS oligonucleotides activate complement, resulting in immunological complications for in vivo applications [30]. These complications and issues of affinity for the RNA, RNase H activity, stability, and cellular uptake have led to investigation of other types of modifications (e.g., N3′-P5′ phosphoramidates) and chimeric oligonucleotides (see Fig. 103.3 and Table 103.1). One particularly promising modification focuses on the 2′ position of the sugar ring. RNA:RNA hybrids are more stable than RNA:DNA hybrids, presumably due to the formation of A-form helices rather than


FIGURE 103.3 Analogues of natural oligonucleotides. (A) Backbone modifications in which the phosphorus bridge atom is retained. (B) Backbone modifications in which the phosphorus bridge atom is replaced. (C) 2′ ribose modifications. (D) Peptide nucleic acids—the entire backbone is replaced with amino acids. See Table 103.1 for legend.

the A:B-form of the heteroduplex. However, with a hydroxyl at the 2′ position, RNA is exceptionally labile to nucleases. Thus, to maintain the A-form geometry (the preferred geometry for RNA helices) and prevent degradation, replacement of the 2′-hydroxyl is being investigated. A study by Monia, et al. investigated both 2′-halide and 2′-O-alkyl substitutions, finding the affinity for the target ranked 2′-fluoro > 2′-O-methyl > 2′-O-propyl > 2′-O-pentyl > 2′-O-deoxy [31]. However, fully modified oligonucleotides were ineffective in inhibiting Ha-ras gene expression in HeLa cells unless the interior of the molecule contained at least 5 2′-deoxy nucleotides to initiate RNase H cleavage. Hence, chimeric oligonucleotides have begun to receive more attention as the next generation of antisense therapeutics. Chimeric oligonucleotides take advantage of the properties of multiple modifications in the same molecule. Earlier chimeric oligonucleotides used end caps of nuclease resistant bases with unmodified central bases [32]. More recently, “mixed-backbone” oligonucleotides have been synthesized with both backbone end modifications (primarily PS) and central sugar modifications (primarily 2′-O-methoxy) [33]. These molecules have been shown to have improved target affinity over PS oligonucleotides while maintaining the ability to initiate RNase H cleavage of the RNA. Investigators have also replaced the phosphodiester backbone entirely, as with peptide nucleic acids (PNAs) in which the entire ribose-phosphodiester backbone is replaced with a polyamide (see Fig. 103.3C). PNAs, with a covalently bound terminal lysine residue to prevent self-aggregation, have been shown to preferentially bind to complementary sequences and inhibit gene expression in vitro [34, 35]. PNAs have an unusually high affinity for their complementary target sequences, possibly because they can form a triple helix with homopurine stretches of RNA (PNA:RNA-PNA) [36]. Duplex forming © 2000 by CRC Press LLC

TABLE 103.1

The Names and Key Characteristics of Several Oligonucleotide Analogues

Phosphorus Analogues (Fig. 103.3a) phosphodiester (PO) phosphorothioate (PS) methylphosphonate (MP) phosphoramidate phosphorodithioate phosphoethyltriester phosphoroselenoate

RNase H Nuclease Chiral activation resistance center yes yes no no yes no yes

Affinity

Charge

X (Fig. 103.3)

Y Z (Fig. 103.3) (Fig. 103.3)

no yes yes yes yes yes yes

no yes yes yes no yes yes

= PO < PO < PO < PO < PO > PO < PO

negative negative neutral neutral negative neutral negative

O– S– CH3 NH-R S– O-C2H5 Se–

O O O O S O O

P P P P P P P

yes yes yes yes yes yes yes

no no no no no no no

< PO > PO < or > PO < PO < PO < PO < PO

neutral neutral neutral neutral neutral neutral neutral

O S O O O CH2 CH2

O O NH CH2 NH CH2 CH2

CH2 CH2 CO SO2 SO2 SO S

yes yes yes yes yes yes yes yes

N/A N/A N/A N/A N/A N/A N/A1 no

>PO >PO >PO >PO >PO >PO < PO > PO

N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 positive3

F O-CH3 O-(CH2)2CH3 O-(CH)4CH3 O-CHCH2 O-(CH2)2-O-CH3 N/A2 N/A4

N/A2 N/A4

N/A2 N/A4

Non-Phosphorus Analogues (Fig. 103.3b) formacetal 3′ thioformacetal 5′-N-carbamate sulfonate sulfamate sulfoxide sulfide

? ? ? ? ? ? ?

2′ Modified Analogues (Fig. 103.3c) fluoro methoxy propoxy pentoxy O-allyl methoxyethoxy α-analogues Peptide Nucleic Acids (Fig. 103.3d)

no no no no no no no no

? = unknown. Chirality and charge depends on backbone structure used. 2 Structure not drawn; The bond between the sugar and base (an N-glycosidic bond) of α-analogues have the reverse orientation (α-configuration) from natural (β-configuration) oligonucleotides. 3 Typically, the C-terminus is covalently linked to a postively charged lysine residue, giving the PNA a positive charge. 4 See Fig. 103.3D for chemical structure. X, Y, Z: these columns reference Fig. 103.3; Replace the designated letter in Fig. 103.3 with the molecule indicated in the Table I to determine the chemical structure of the oligo. 1

PNAs have shown inhibition of translation when targeted to the mRNA start codon, while triple helix formation is required to inhibit protein elongation [37]. The ability of PNAs to invade the duplex of DNA and form a stable triple helix structure remains an issue for their in vivo application. Other modifications to the ribose sugars (e.g., α-anomeric) and modifications to the nucleoside bases (e.g., 5-methyl or 5-bromo-2′-deoxycytidine and 7′-deazaguanosine and 7′-deazaadenosine oligonucleotides) have also been examined [10]. Oligonucleotides covalently linked to active groups (e.g., intercalators such as acridine, photoactivated crosslinking agents such as psoralens and chelating agents such as EDTA-Fe) are also being actively investigated as potential antisense molecules [9]. Whether for clinical application or for research/diagnostic tools, studies with modified oligonucleotides will provide valuable information regarding the mechanistic steps of antisense oligonucleotide activity. These future generations of antisense “oligonucleotides” may bear little structural resemblance to natural oligonucleotides, but their inhibition of gene expression will still rely on sequence specific base pairing. The reader is directed to several reviews for a comprehensive treatment of all classes of chemically modified oligonucleotides [9, 38-40].


103.4 Oligonucleotide Synthesis and Purification The growth of the field of antisense therapeutics has also resulted in a need for increased production of oligonucleotides for in vitro and in vivo applications. As a result, new synthesis procedures are being explored to increase batch size and synthesis rate. Large-scale automated synthesizers have been developed, with which batches of 100 mmol have been produced [http://www.hybridon.com]. This is a significant improvement given that a large batch on a standard synthesizer is 1 µmol. Standard procedures begin with the 3′ base of the oligonucleotide attached to a solid controlled pore glass (CPG) support, although other supports have also been proposed [41]. The process then continues with deprotection, monomer introduction, activated coupling, oxidation, and capping in a cycle for the addition of each oligonucleotide. Depending on the type of monomer being used (e.g., β-cyanoethyl phosphoramidite or hydrogen phosphonate), sulfurization to generate PS oligonucleotides occurs either during the synthesis cycling or following the termination of synthesis. Given the imperfections in the synthesis procedures, the resulting oligonucleotide product must then be purified to remove failure products. Oligonucleotides are typically purified using reverse-phase high performance liquid chromatography (RP-HPLC). During synthesis, a hydrophobic dimethoxy-trityl (DMT) group is removed from the phosphoramidite to allow the next nucleoside to attach. Leaving this group on the final nucleoside of the synthesis increases the hydrophobicity of the full, correct oligo, increasing its retention time on an RPHPLC column. The eluent is then fractionated to collect only the full sequence product. However, PS oligonucleotides, unlike PO oligonucleotides, are chiral molecules, i.e., molecules that are not identical to their mirror image. Current synthesis techniques do not control the orientation of each internucleotide bond, and the resultant oligonucleotides are therefore a mixture of 2(n-1) diastereomers (where n = the number of bases in the complete oligo). The resulting sequences elute with similar retention times on RP-HPLC as all of the complete sequence diastereomers should have the hydrophobic DMT group attached. However, analysis following removal of the DMT group can be used to confirm or deny the presence of multiple diastereomers under higher resolution conditions. Capillary gel electrophoresis and anion-exchange HPLC are being examined as ways to enhance the selectivity of purification of PS oligonucleotides. Additional details on the synthesis and purification of oligonucleotides are available for the interested reader [42-44].

103.5 Specificity of Oligonucleotides The ability to block the expression of a single gene without undesired side effects (specificity) is the major advantage, in principle, of antisense based strategies. This specificity is primarily determined by the length (i.e., the number of bases) of the oligonucleotide. Experimental and theoretical data suggest that there is an optimum length at which specific inhibition of gene expression is maximized and non-specific effects are minimized [13, 14]. Affinity and specificity limit the minimum effective length of oligonucleotides. Oligonucleotides that are too short do not inhibit gene expression, because they do not bind with sufficient affinity to their substrates. The shortest oligonucleotide reported to affect gene expression in mammalian cells was 7 bases in length (7-mer) [45]. This oligonucleotide was highly modified, containing 5-(1-propynyl) cytosines and uracils as well as a phosphorothioate backbone. An oligonucleotide that is too short is less likely to represent a unique sequence in a given cell’s genome and thus more likely to bind to a non-targeted RNA and inhibit its expression. Woolf et al. estimated the minimum sequence length that will be statistically unique in a given pool of mRNAs [13]. Since each position in a given sequence can be occupied by any of four nucleotides (A, C, G or U), the total number of different possible sequences of length N bases is 4N. Letting R equal the total number of bases in a given mRNA pool and assuming that it is a random and equal mixture of the four nucleotides, then the frequency (F) of occurrence in that pool of a sequence of length N is given by:


F=

R 4N

(103.1)

For a typical human cell, which contains approximately 104 unique mRNA species whose average length is 2000 bases, R is approximately equal to 2 × 107. Therefore, for a sequence to be unique (F < 1), N must be greater than or equal to 13 bases [13]. The minimum oligonucleotide length will also be constrained by the minimum binding affinity to form a stable complex. However, oligonucleotides cannot be made arbitrarily long, because longer oligonucleotides are more likely to contain internal sequences complementary to non-targeted RNA molecules. This has also been expressed mathematically [13]. The expected number of complementary sites (S) of length L for an oligonucleotide of length N in an mRNA pool with R bases is given by:

S=

[(N − L + 1) × R] 4L

(103.2)

For example, an 18-mer (N = 18) has 6 internal 13-mers. Since a 13-mer is expected to occur 0.3 times in a mRNA pool containing 2 × 107 bases, the 18-mer is expected to match 1.8 (i.e., 6 × 0.3) 13-mers in the mRNA pool. Equation (103.2) also gives this result (N = 18, L = 13, and R = 2 × 107; therefore, S = 1.8). Woolf et al. have demonstrated that significant degradation of non-targeted mRNAs can occur [13]. They compared the effectiveness of three different 25-mers in suppressing the expression of fibronectin mRNA in Xenopus oocytes. Nearly 80% of the fibronectin mRNA was degraded after the oocytes were microinjected with a 25-mer in which all 25 of its bases were complementary to the mRNA. However, when the oocytes were microinjected with 25-mers that had only 17 or 14 complementary bases flanked by random sequences, greater than 30% of their fibronectin mRNA was still degraded. They also showed that a single mismatch in an oligonucleotide did not completely eliminate its antisense effect. Over 40% of the target mRNA was degraded when oocytes were treated with a 13-mer with one internal mismatch, though the mismatch left a 9 base consecutive complementary sequence that showed nearly the same activity as a 13-mer with 10 complementary bases in succession. Although these studies were conducted at lower temperatures and therefore under less stringent hybridization conditions than those found in mammalian cells, they clearly showed that complementary oligonucleotides flanked by unrelated sequences and even oligonucleotides with mismatched sequences can lead to significant degradation of RNA. The possibility of undesired inhibition of partially complementary sequences must be considered when designing and testing any antisense oligonucleotide. Hence, a useful heuristic is that an oligonucleotide should be long enough to be unique and possess high affinity for its target but short enough to minimize side effects due to degradation of non-targeted mRNAs. The efforts to sequence the human genome have begun to provide more information about the actual specificity of oligonucleotides. It is possible to scan for sequences against all of the genomic information available using resources such as BLAST at the National Center for Biotechnology Information [46; http://www.ncbi.nlm.nih.gov]. The results will determine the uniqueness of the oligonucleotide within the known database sequences. As the completeness of the databases grows, computational comparisons of target RNA sequences against the database will provide more reliable assessments of the uniqueness of these target sequences within the genome.

103.6 Oligonucleotide Delivery In many cases, the concentrations of antisense oligonucleotides required to achieve a biological effect are currently too large to be of therapeutic value. Achieving the necessary concentration in the vicinity of the target cells is thought to be limited at least in part by charge repulsion between the negatively-charged oligonucleotide and the negatively-charged cell surface. Four major methods are currently being investigated to overcome this and other extracellular transport barriers: (1) chemical modification of the © 2000 by CRC Press LLC

oligonucleotide to increase its hydrophobicity; (2) conjugation of the oligonucleotide to a polycation; (3) conjugation of the oligonucleotide to a ligand specific for a cellular receptor; and (4) encapsulation of the oligonucleotide in a liposome. Increasing the hydrophobicity of oligonucleotides was first accomplished by synthesis of methylphosphonate backbones. Despite their neutral backbone charge, cellular uptake of MP oligonucleotides was shown to be very inefficient, perhaps due in part to their inability to escape lysosomal degradation and gain entry into the cytoplasm. The hydrophobicity of an oligonucleotide can also be increased by conjugation with hydrophobic moieties such as cholesterol derivatives (chol-oligonucleotide). The increased hydrophobicity of chol-oligonucleotides reportedly improves their association with cell membranes and their internalization by cells [47]. Cellular accumulation of chol-oligonucleotides after 2 hours has been shown to occur when no accumulation was seen with unconjugated oligonucleotides [48]. The mechanism by which chol-oligonucleotides are more easily and rapidly taken up by cells has not been elucidated. Krieg [49] showed that chol-oligonucleotides with cholesterol at the 5′ end could be bound by low density lipoprotein (LDL) and that this markedly increased the association of the oligonucleotide with the cell membrane and its internalization in vitro. LDL-associated chol-oligonucleotides were effective at 8-fold lower concentration than PO oligonucleotide controls. Oligonucleotides conjugated with polycations such as poly-L-lysine (PLL) also have improved cellular uptake. Covalent conjugates can be constructed by coupling the 3′ end of the oligonucleotide to the epsilon amino groups of the lysine residues [50]. Oligo-PLL conjugates complementary to the translation initiation site of the tat gene (a key HIV regulatory protein) protected cells from HIV-1 infection with concentrations 100-fold lower than non-conjugated oligonucleotides [51]. Low concentrations of oligoPLL conjugates (100 nM), in which the 15-mer oligonucleotide was complementary to the initiation region of the viral N protein mRNA, also inhibited vesicular stomatitis virus (VSV) infection [50]. Internalization of oligonucleotides has also been improved by conjugation to ligands. One such ligand, transferrin, is internalized by receptor mediated endocytosis. Mammalian cells acquire iron carried by transferrin. Oligo-PLL complexes have been conjugated to transferrin to take advantage of this pathway. For example, an 18-mer complementary to c-myb mRNA (an oncogene that is responsible for the hyperproliferation of some leukemia cells) was complexed with a transferrin PLL conjugate and rapidly internalized by human leukemia (HL-60) cells [52]. The expression of c-myb was greatly reduced and the uncontrolled proliferation of these cells inhibited. Because an oligo-PLL conjugate (without transferrin) was not tested, it is not clear whether the improved antisense effects were due to the PLL moiety, the transferrin moiety, or a combination thereof. Oligo-ligand conjugates have also been used to target oligonucleotides to cells with specific cellular receptors, such as the asialoglycoprotein (ASGP) receptor, which is expressed uniquely by hepatocytes. These receptors bind and internalize ASGPs, serum glycoproteins that have exposed galactose residues at the termini of their glycosylation chains. Oligonucleotides complexed to ASGP are rapidly internalized by hepatocytes. A number of researchers have used ASGP conjugated to cationic PLL. When the ASGP:PLL is conjugated to oligonucleotides, their cellular uptake and antisense effectiveness in vitro is increased [53-55]. However, toxicity of the PLL moiety and instability of the noncovalent complexes between ASGPPLL and oligonucleotides have limited their in vivo applicability [56]. One possible solution is the direct conjugation of the oligonucleotides to ASGP via a cleavable disulfide linkage [57]. Another promising means to deliver oligonucleotides is lipofection. Oligonucleotides are mixed with cationic lipids that condense around the negatively charged oligonucleotide forming a lipid vesicle (liposome). The positively charged lipids reduce the electrostatic repulsion between the negativelycharged oligonucleotide and the similarly charged cell surface. Bennett et al. made liposomes using DOTMA {N-{1-(2,3-dioleyloxy)propyl)}-N,N,N-trimethylammonium chloride} as the cationic lipid and an oligonucleotide complementary to the translation initiation codon of human intracellular adhesion molecule 1 (ICAM-1), increasing the potency of the oligonucleotide by >1000-fold [12]. A recent study has shown that liposomally encapsulated oligonucleotides at 0.01 nM were as active as free oligonucleotides at 1.5 µM at inhibiting HIV-1 replication in chronically infected cells [58]. It was found that the oligonucleotides separate from cationic liposomes following cellular internalization [59]. Cationic lipids


subsequently accumulate in cytoplasmic structures and the plasma membrane, suggesting that they have fused with the endosomal vesicles and are being recirculated throughout the secretory pathways [59-61]. The increase in oligonucleotide activity suggests that cellular uptake may be a significant limitation in the effectiveness of antisense oligonucleotides. In vitro studies using electroporation, streptolysin O or α-toxin permeabilization, and particle bombardment to enhance cellular uptake of nucleic acids support this hypothesis [62-65]. Efforts have been made to use antibodies to target liposomes containing oligonucleotides to specific cell types. Protein A, which binds to many IgG antibodies, is covalently bound to the liposome to form a liposome-oligo-protein A complex. An antibody specific for the target cell’s surface is bound to the protein A, and these complexes are incubated with the target cells. Leonetti [66] synthesized such a complex to target the mouse major histocompatibility complex H-2K molecule on mouse L929 cells. The complex contained a 15-mer complementary to the 5′ end of mRNA encoding the N protein of VSV. The complexes inhibited viral replication by more than 95%. Unconjugated liposomes, and liposomes conjugated to antibodies specific for a nonexpressed antigen, had no effect.

103.7 Potential Applications of Antisense Oligonucleotides The potential applications for antisense oligonucleotides are limited only by the genetic information available. Antisense oligonucleotides can be developed against any target in which the inhibition of protein production or the inhibition of RNA processing yields the therapeutic result. Currently, clinical trials are underway using antisense oligonucleotides to treat rheumatoid arthritis, psoriasis, renal transplant rejection, and inflammatory bowel disease (Crohn’s disease) [http://www.phrma.org]. However, the primary targets remain refractory viral diseases and cancers for which the necessary target genetic information is typically available.

Viral Diseases One of the many potential therapeutic applications of antisense technology is the treatment of infectious viral diseases. The use of antisense oligonucleotides as anti-viral agents is particularly promising because viral nucleic acid sequences are unique to the infected cell and are not found in normal healthy cells. The goal of antiviral therapy is to block the expression of key viral proteins that are vital to the life cycle of the virus. This has been achieved in vitro with several viruses including HIV, HSV, influenza and human papilloma virus [10]. The in vitro work with HIV is representative of other viruses and will be highlighted here. Human Immunodeficiency Virus Retroviruses, and HIV in particular, have high rates of mutation and genetic recombination. The effectiveness of many anti-HIV drugs has been severely reduced because drug resistant viral strains often arise after prolonged drug treatment. This is especially relevant to strategies that use an antisense approach that relies on a specific nucleotide sequence. One strategy to inhibit HIV replication is to target conserved sequences in key regulatory proteins such as tat and rev. Part of the rev sequence is highly conserved, with the known 16 isolates of HIV differing by at most one base pair in this conserved region [67]. The importance of targeting a highly conserved region as is found in rev was demonstrated by Lisziewicz, who investigated the efficacy of antisense oligonucleotides directed against various viral proteins [68]. Although several 28-mers initially inhibited viral replication, resistant mutant viruses developed after 25 days to all of the oligonucleotides with the exception of those directed at the highly conserved rev regions. A second concern in treating HIV and other viruses is that viral replication can restart after antisense treatment is stopped [67]. Can oligonucleotides be continuously administered to prevent viral replication without unwanted side effects? These issues and others common to all antisense based therapies (oligonucleotide stability, specificity, affinity, and delivery) must be addressed prior to the successful implementation of antisense-based HIV therapies.


Cancer In principle, antisense technology can be used against cancer, but the target is more challenging. Oncogenes are typically genes that play a vital regulatory role in the growth of a cell, the mutation or inappropriate expression of which can result in the development of cancer. In the case of mutation, it is often difficult to distinguish an oncogene from its normal counterpart, because they may differ by as little as one base. Thus, attempts to inhibit oncogene expression might block the expression of the normal gene in non-cancerous cells and cause cytotoxic effects. Despite these challenges, steady progress has been made in the development of effective antisense oligonucleotides that have inhibited many types of oncogenes including bcr-abl, c-myc, c-myb, c-ras, Ha-ras, neu/erbB2 and bFGF [11, 25, 61, 69, 70]. Studies targeting ras oncogenes are encouraging and are representative of progress against other classes of oncogenes. Chang [71] targeted a ras p21 gene point mutation with an antisense oligonucleotide. Only the expression of the mutated genes was inhibited, suggesting that it is possible to selectively inhibit an oncogene that differs by only a single base from the normal gene. Using a 17-mer PS oligo, it was shown that ~98% inhibition of Ha-ras expression could be achieved at 1 µM concentration [25]. Successes in inhibiting oncogene expression in vitro are encouraging, but many problems remain to be solved before antisense oligonucleotides are therapeutically useful for cancer patients. For example, finding a suitable oncogene to target is difficult. Even if a genetic defect is common to all cells of a given cancer, there is no guarantee that inhibition of that oncogene will halt cancer cell proliferation [70]. Even if cell growth is inhibited, tumor growth may restart after antisense treatment is stopped. Oligonucleotides are needed that induce cancer cell terminal differentiation or death [70]. In order to avoid toxic side effects by inhibiting gene expression of regulated oncogenes in normal tissues, oligonucleotides that specifically target cancer cells are required. Even if appropriate antisense oligonucleotides are developed, they will still be ineffective against solid tumors unless they can reach the interior of the tumor at biologically effective concentrations.

103.8 In Vivo Pharmacology The first in vivo antisense studies were designed to test the biodistribution and toxicity of antisense oligonucleotides. These studies demonstrated that PO oligonucleotides are rapidly excreted from the body with PS oligonucleotides being retained longer. One study in mice noted that completely modified PS oligonucleotides were retained significantly longer than a chimeric 20-mer that contained 15 phosphodiester bonds and only 4 phosphorothioate bonds [72]. Only 30% of the PS oligonucleotide was excreted in the urine after 24 hours, whereas 75% of the chimeric oligonucleotide was excreted after only 12 hours. A similar study demonstrated that PS oligonucleotides were retained in body tissues of adult male rats with a half-life of 20 to 40 hours [73]. These studies suggested that antisense oligonucleotides may need to be administered repeatedly to maintain a therapeutic effect for chronic disorders. At least with PS oligonucleotides, this has proven to be problematic. Continuous or repeated administration of PS oligonucleotides results in immunologic complement activation. It has been shown that infusion of PS oligonucleotides to rhesus monkeys results in almost immediate (within 10 min.) decreases in white blood cell count and the fraction of neutrophils among the white cells [30]. At 40 min. post-infusion, the neutrophil counts had increased to higher than baseline levels. A recent study indicated that a plasma concentration of >50 µg/ml results in complement activation [74]. These difficulties have further hastened the development of novel oligonucleotide chemistries and chimeric oligonucleotides for which the immunologic properties are expected to be improved due to the possible reduction in the use of PS backbone linkages [33, 75]. The targeted delivery of oligonucleotides has been proposed as a way to improve the effectiveness of systemically administered antisense oligonucleotides by oligonucleotide concentration in the desired cells and minimizing non-specific side effects. Targeting oligonucleotides to a specific organ has been demonstrated. As previously discussed, a soluble DNA carrier system has been developed to specifically target hepatocytes [76]. This system was used to target a 67-mer PO oligonucleotide (complementary to the


5′ end of rat serum albumin mRNA) to the liver [56]. Following tail vein injection into rats, the complex rapidly and preferentially accumulated in the liver, but the efficiency of this targeting method was limited by the rapid dissociation of the oligonucleotide from the ASGP:PLL complex (30% dissociated within 7 minutes).

103.9 Animal Models The effectiveness of antisense oligonucleotides at inhibiting gene expression in vivo has been demonstrated in several animal models. Immunodeficient mice, bearing a human myeloid leukemia cell line, were continuously infused with 100 µg/day of a 24-mer PS oligonucleotide complementary to c-myb mRNA [77]. Mice that received the antisense oligonucleotide survived on average 8.5 times longer than control mice that received no treatment and 5.7 times longer than sense-treated controls. The treated animals also had a significantly lower tumor burden in the ovaries and brain. Antisense oligonucleotides have also inhibited the growth of solid tumors in vivo. Immuno-deficient mice bearing a fibrosarcoma or melanoma were injected subcutaneously twice weekly with 1.4 mg of PS oligonucleotides complementary to the 5′ end of the p65 subunit of the mRNA for the NF-κB transcription factor (NF-κB activates a wide variety of genes and is believed to be important in cell adhesion and tumor cell metastasis) [78]. Greater than 70% of antisense treated mice exhibited a marked reduction in their tumor size. Administration of control oligonucleotides complementary to GAPDH and jun-D had no effect on tumor size. The capability of antisense oligonucleotides to inhibit viral replication in vivo has also been studied. Fourteen one day old ducklings were infected with duck hepatitis B virus (DHBV), which is closely related to the strain of hepatitis B virus that is a major cause of chronic liver disease and cancer in humans [79]. Two weeks later, the ducks were injected daily, for 10 consecutive days, with 20 µg/gm body weight of an 18-mer PS oligonucleotide complementary to the start site of the Pre-S region of the DHBV genome. The antisense oligonucleotides blocked viral gene expression and eliminated the appearance of viral antigens in the serum and liver of all treated ducks. No toxic effects due to the oligonucleotides were noted. Unfortunately, residual amounts of DNA precursors of viral transcripts were detected in the nuclei of liver cells, which resulted in a slow restart of viral replication after antisense treatment was stopped. Further studies are needed to determine if prolonged treatment with antisense will eliminate this residual viral DNA. Though many of the oligonucleotides under investigation target viruses or cancer, other refractory targets are also being investigated. Antisense oligonucleotides have been evaluated in the treatment of restenosis, a narrowing of an artery following corrective surgery caused by mitogen-induced proliferation of vascular smooth muscle cells (SMC). Several studies have tested the ability of antisense oligonucleotides to inhibit genes whose expression is important in SMC proliferation, including c-myc, c-myb, cdc2 and cdk2 [80-82]. In one study with an antisense therapeutic complementary to c-myc mRNA, the formation of a thick intimal layer (neointima) in treated areas was significantly reduced in antisense treated rats when compared to control rats with maximum c-myc expression reduced by 75% relative to controls [82]. Dextran sodium sulfate-induced colitis in Swiss-Webster mice (as measured by disease activity index, DAI) was reduced by ~64% following 5 days of daily subcutaneous administration of an antisense oligonucleotide complementary to ICAM-1 at 0.3 mg/kg/day [83]. Interestingly, in this study, higher oligonucleotide concentrations (> 5 mg/kg/day) were ineffective in reducing inflammation.

103.10

Clinical Trials

The FDA recently approved the first antisense therapeutic, fomivirsen, a treatment for cytomegalovirus (CMV) retinitis in AIDS patients [http://www.isip.com]. CMV infection is fairly common (~50% of people over 35 are seropositive); however, it results in complications only in immunocompromised individuals. The oligonucleotide is complementary to the immediate-early viral RNA, a region encoding


TABLE 103.2 Compound

A Sample of Current Clinical Trials of Antisense Oligodeoxynucleotide Therapeutics Type of Oligo

Disease Target

Gene Target

Phase

ISIS13312 ISIS2302

2′-modified—2nd Gen. PS

G3139 ISIS2503 ISIS3521 ISIS5132 LR-3001 LR-3280

N/A PS PS PS PS PS

CMV Retinitis Crohn’s Disease Psoriasis Renal Transplant Rejection Rheumatoid Arthritis Ulcerative Colitis Cancer Cancer Cancer Cancer Chromic Myelogenous Leukemia Restenosis

N/A ICAM-1 ICAM-1 ICAM-1 ICAM-1 ICAM-1 N/A Ha-ras PKC-α c-raf bcr-abl myc

I/II II/Pivotal II II II II I I II II I II

N/A—Not available Potential applications of antisense technology include the treatment of cancer, infectious diseases, and inflammatory diseases.

proteins that regulate virus gene expression [84]. The activity of the oligonucleotide therapeutic is potentiated by traditional drugs such as ganciclovir. Additional studies suggested, however, that the oligonucleotide acts not only via an antisense mechanism, though complementarity to the target was required for maximal inhibitory capacity [85]. Clinical trials on many other antisense oligonucleotides have advanced to Phase II. However, complications associated with antisense oligonucleotides have also been found. An antisense compound targeted to the gag region of HIV was pulled from clinical trials due to the resulting decreases in platelet counts in 30% of the patients after 10 days of daily administration [http://www.hybridon.com]. The immunologic complications are likely related to the activation of complement discussed earlier. Perhaps, fomivirsen avoids these complications due to its local administration by intravitreal injection rather than systemic infusion. Table 103.2 lists clinical trials that are currently ongoing and the disease targets. Published results from these trials are not yet available.

103.11

Summary

Antisense oligonucleotides have the potential to selectively inhibit the expression of any gene with a known sequence. Although there have been some remarkable successes, realizing this potential is proving difficult because of problems with oligonucleotide stability, specificity, affinity, and delivery. Natural oligonucleotides contain a phosphodiester backbone that is highly susceptible to nuclease degradation. Chemically modified oligonucleotides (PS and chimeric oligonucleotides) have been synthesized that are more stable. However, the increased stability of PS oligonucleotides comes at a cost, in particular the activation of complement. Another issue for chemically modified oligonucleotides is their large scale production and purification. Alternative chemistries are needed that are more biologically active and more amenable to large scale production. Targeted delivery of antisense oligonucleotides is needed to minimize side effects and maximize oligonucleotide concentration in target cells. Attempts to improve delivery include chemical modification of the oligonucleotides to increase their cellular permeability, the conjugation of oligonucleotides to specific ligands to utilize more efficient receptor-mediated internalization pathways, and the use of antibody conjugated liposomes to deliver oligonucleotides to specific cells. The potential applications of antisense oligonucleotides to the treatment of disease are vast. Antisensebased therapies are under development for the treatment of infectious diseases such as CMV, HIV, herpes simplex virus, influenza, hepatitis and human papilloma virus as well as the treatment of complex genetic disorders like cancer. Animal models are being used to determine if (1) antisense oligonucleotides can


be delivered to target cells at high enough concentrations to be effective, (2) repeated treatments with oligonucleotides are toxic or elicit an immune response, and (3) antisense oligonucleotides directed against a single gene can be effective against complex genetic diseases such as cancer. Multiple clinical trials against viral, cancer, and other targets are now in progress with one drug having been approved for the clinic. Improvements in our understanding of the mechanisms of antisense inhibition, the pharmacology of antisense oligonucleotides in vivo and the development of chemically modified oligonucleotides with high affinity, specificity and stability are needed to realize the clinical potential of antisense-based strategies for the treatment of a wide variety of diseases.

Defining Terms Antisense: Any DNA or RNA molecule whose sequence is complementary to the RNA transcribed from a target gene. Chiral: A molecule whose configuration is not identical with its mirror image. Complementary: A nucleic acid sequence is complementary to another if it is able to form a perfectly hydrogen-bonded duplex with it, according to the Watson-Crick rules of base pairing (A opposite U or T, G opposite C). Diastereomer: Optically active isomers that are not enantiomorphs (mirror images). Exonuclease: An enzyme that catalyzes the release of one nucleotide at a time, serially, from one end of a polynucleotide. In vitro: In an artificial environment, referring to a process or reaction occurring therein, as in a test tube or culture dish. In vivo: In the living body, referring to a process or reaction occurring therein. Lipofection: Delivery of therapeutic drugs (antisense oligonucleotides) to cells using cationic liposomes. Lipophilic: Capable of being dissolved in lipids (organic molecules that are the major structural elements of biomembranes). Liposome: A spherical particle of lipid substance suspended in an aqueous medium. Plasmid: A small, circular extrachromosomal DNA molecule capable of independent replication in a host cell. Receptor mediated endocytosis: The selective uptake of extracellular proteins, oligonucleotides and small particles, usually into clathrin coated pits, following their binding to cell surface receptor proteins. Restenosis: A narrowing of an artery or heart valve following corrective surgery on it. RNase H: An enzyme that specifically recognizes RNA:DNA duplexes and cleaves the RNA portion, leaving the DNA portion intact.

References 1. Zamecnik, P. C. and Stephenson, M. L. 1978. Inhibition of Rous sarcoma virus replication and cell transformation by a specific oligodeoxynucleotide, Proc. Natl. Acad. Sci. USA, 75, 280. 2. Tomizawa, J. and Itoh, T. 1981. Plasmid ColE1 incompatibility determined by interaction of RNA I with primer transcript, Proc. Natl. Acad. Sci. USA, 78, 6096. 3. Tomizawa, J., Itoh, T., Selzer, G., and Som, T. 1981. Inhibition of ColE1 RNA primer formation by a plasmid-specified small RNA, Proc. Natl. Acad. Sci. USA, 78, 1421. 4. Izant, J. G. and Weintraub, H. 1985. Constitutive and conditional suppression of exogenous and endogenous genes by anti-sense RNA, Science, 229, 345. 5. Stull, R. A. and Szoka, F. C., Jr. 1995. Antigene, ribozyme and aptamer nucleic acid drugs: progress and prospects, Pharm Res, 12, 465. 6. Hélène, C., Giovannangeli, C., Guieysse-Peugeot, A. L., and Praseuth, D. 1997. Sequence-specific control of gene expression by antigene and clamp oligonucleotides, Ciba Found. Symp., 209, 94.


7. Gewirtz, A. M., Sokol, D. L., and Ratajczak, M. Z. 1998. Nucleic acid therapeutics: state of the art and future prospects, Blood, 92, 712. 8. Rossi, J. J. 1997. Therapeutic applications of catalytic antisense RNAs (ribozymes), Ciba Found. Symp., 209, 195. 9. Hélène, C. and Toulme, J. J. 1990. Specific regulation of gene expression by antisense, sense and antigene nucleic acids, Biochem. Biophys. Acta., 1049, 99. 10. Milligan, J. F., Matteucci, M. D., and Martin, J. C. 1993. Current concepts in antisense drug design, J. Med. Chem., 36, 1923. 11. Nagel, K. M., Holstad, S. G., and Isenberg, K. E. 1993. Oligonucleotide pharmacotherapy: an antigene strategy, Pharmacotherapy, 13, 177. 12. Bennett, C. F., Condon, T. P., Grimm, S., Chan, H., and Chiang, M. Y. 1994. Inhibition of endothelial cell adhesion molecule expression with antisense oligonucleotides, J. Immunol., 152, 3530. 13. Woolf, T. M., Melton, D. A., and Jennings, C. G., 1992. Specificity of antisense oligonucleotides in vivo, Proc. Natl. Acad. Sci. USA, 89, 7305. 14. Herschlag, D. 1991. Implications of ribozyme kinetics for targeting the cleavage of specific RNA molecules in vivo: more isn’t always better, Proc. Natl. Acad. Sci. USA, 88, 6921. 15. Stull, R. A., Zon, G., and Szoka, F. C., Jr. 1996. An in vitro messenger RNA binding assay as a tool for identifying hybridization-competent antisense oligonucleotides, Antisense Nucleic Acid Drug Dev., 6, 221. 16. Milner, N., Mir, K. U., and Southern, E. M. 1997. Selecting effective antisense reagents on combinatorial oligonucleotide arrays, Nat. Biotechnol., 15, 537. 17. Jaroszewski, J. W., Syi, J. L., Ghosh, M., Ghosh, K., and Cohen, J. S. 1993. Targeting of antisense DNA: comparison of activity of anti-rabbit beta- globin oligodeoxyribonucleoside phosphorothioates with computer predictions of mRNA folding, Antisense Res. Dev., 3, 339. 18. Sczakiel, G., Homann, M., and Rittner, K. 1993. Computer-aided search for effective antisense RNA target sequences of the human immunodeficiency virus type 1, Antisense Res. Dev., 3, 45. 19. Lima, W. F., Monia, B. P., Ecker, D. J., and Freier, S. M. 1992. Implication of RNA structure on antisense oligonucleotide hybridization kinetics, Biochemistry, 31, 12055. 20. Monia, B. P., Johnston, J. F., Geiger, T., Muller, M., and Fabbro, D. 1996. Antitumor activity of a phosphorothioate antisense oligodeoxynucleotide targeted against C-raf kinase, Nat. Med., 2, 668. 21. Goodchild, J., Carroll, E. D., and Greenberg, J. R. 1988. Inhibition of rabbit beta-globin synthesis by complementary oligonucleotides: identification of mRNA sites sensitive to inhibition, Arch. Biochem. Biophys., 263, 401. 22. Wakita, T. and Wands, J. R. 1994. Specific inhibition of hepatitis C virus expression by antisense oligodeoxynucleotides. In vitro model for selection of target sequence, J. Biol. Chem., 269, 14205. 23. Peyman, A., Helsberg, M., Kretzschmar, G., Mag, M., Grabley, S., and Uhlmann, E. 1995. Inhibition of viral growth by antisense oligonucleotides directed against the IE110 and the UL30 mRNA of herpes simplex virus type-1, Biol. Chem. Hoppe Seyler, 376, 195. 24. McKay, R. A., Cummins, L. L., Graham, M. J., Lesnik, E. A., Owens, S. R., Winniman, M., and Dean, N. M. 1996. Enhanced activity of an antisense oligonucleotide targeting murine protein kinase C-alpha by the incorporation of 2′-O-propyl modifications, Nucleic Acids Res., 24, 411. 25. Monia, B. P., Johnston, J. F., Sasmor, H., and Cummins, L. L. 1996. Nuclease resistance and antisense activity of modified oligonucleotides targeted to Ha-ras, J. Biol. Chem., 271, 14533. 26. Iribarren, A. M., Cicero, D. O., and Neuner, P. J. 1994. Resistance to degradation by nucleases of (2’S)2′-deoxy-2′-C-methyloligonucleotides, novel potential antisense probes, Antisense Res. Dev., 4, 95. 27. Loke, S. L., Stein, C. A., Zhang, X. H., Mori, K., Nakanishi, M., Subasinghe, C., Cohen, J. S., and Neckers, L. M. 1989. Characterization of oligonucleotide transport into living cells, Proc. Natl. Acad. Sci. USA, 86, 3474. 28. Stein, C. A., Subasinghe, C., Shinozuka, K., and Cohen, J. S. 1988. Physicochemical properties of phosphorothioate oligodeoxynucleotides, Nucleic Acids Res., 16, 3209.


29. Neckers, L. and Whitesell, L. 1993. Antisense technology: biological utility and practical considerations, Am. J. Physiol., 265, L1. 30. Galbraith, W. M., Hobson, W. C., Giclas, P. C., Schechter, P. J., and Agrawal, S. 1994. Complement activation and hemodynamic changes following intravenous administration of phosphorothioate oligonucleotides in the monkey, Antisense Res. Dev., 4, 201. 31. Monia, B. P., Lesnik, E. A., Gonzalez, C., Lima, W. F., McGee, D., Guinosso, C. J., Kawasaki, A. M., Cook, P. D., and Freier, S. M. 1993. Evaluation of 2′-modified oligonucleotides containing 2′-deoxy gaps as antisense inhibitors of gene expression, J. Biol. Chem., 268, 14514. 32. Pickering, J. G., Isner, J. M., Ford, C. M., Weir, L., Lazarovits, A., Rocnik, E. F., and Chow, L. H. 1996. Processing of chimeric antisense oligonucleotides by human vascular smooth muscle cells and human atherosclerotic plaque. Implications for antisense therapy of restenosis after angioplasty, Circulation, 93, 772. 33. Agrawal, S., Jiang, Z., Zhao, Q., Shaw, D., Cai, Q., Roskey, A., Channavajjala, L., Saxinger, C., and Zhang, R. 1997. Mixed-backbone oligonucleotides as second generation antisense oligonucleotides: in vitro and in vivo studies, Proc. Natl. Acad. Sci. USA, 94, 2620. 34. Hanvey, J. C., et al. 1992. Antisense and antigene properties of peptide nucleic acids, Science, 258, 1481. 35. Nielsen, P. E., Egholm, M., Berg, R. H., and Buchardt, O. 1993. Peptide nucleic acids (PNAs): potential antisense and anti-gene agents, Anticancer Drug Des., 8, 53. 36. Egholm, M., Nielsen, P. E., Buchardt, O., and Berg, R. H. 1992. Recognition of guanine and adenine in DNA by cytosine and thymine containing peptide nucleic acids (PNA), J. Am. Chem. Soc., 114, 9677. 37. Good, L. and Nielsen, P. E. 1997. Progress in developing PNA as a gene-targeted drug, Antisense Nucleic Acid Drug Dev., 7, 431. 38. Uhlmann, E. and Peyman, A. 1990. Antisense oligonucleotides: A new therapeutic principle, Chem. Rev., 90, 544. 39. Cook, P. D. 1991. Medicinal chemistry of antisense oligonucleotides—future opportunities, Anticancer Drug Des., 6, 585. 40. Matteucci, M. 1997. Oligonucleotide analogues: an overview, Ciba Found. Symp., 209, 5. 41. Fearon, K. L., Hirschbein, B. L., Chiu, C. Y., Quijano, M. R., and Zon, G. 1997. Phosphorothioate oligodeoxynucleotides: large-scale synthesis and analysis, impurity characterization, and the effects of phosphorus stereochemistry, Ciba Found. Symp., 209, 19. 42. Righetti, P. G. and Gelfi, C. 1997. Recent advances in capillary electrophoresis of DNA fragments and PCR products, Biochem. Soc. Trans., 25, 267. 43. Righetti, P. G. and Gelfi, C. 1997. Recent advances in capillary electrophoresis of DNA fragments and PCR products in poly(n-substituted acrylamides), Anal. Biochem., 244, 195. 44. Schlingensiepen, R., Brysch, W., and Schlingensiepen, K.-H. 1997. Antisense—from technology to therapy, Blackwell Wissenschaft Berlin, Vienna, Austria. 45. Wagner, R. W., Matteucci, M. D., Grant, D., Huang, T., and Froehler, B. C. 1996. Potent and selective inhibition of gene expression by an antisense heptanucleotide, Nat. Biotechnol., 14, 840. 46. Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Zhang, Z., Miller, W., and Lipman, D. J. 1997. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs, Nucleic Acids Res., 25, 3389. 47. Boutorine, A. S. and Kostina, E. V. 1993. Reversible covalent attachment of cholesterol to oligodeoxyribonucleotides for studies of the mechanisms of their penetration into eucaryotic cells, Biochimie, 75, 35. 48. Alahari, S. K., Dean, N. M., Fisher, M. H., Delong, R., Manoharan, M., Tivel, K. L., and Juliano, R. L. 1996. Inhibition of expression of the multidrug resistance-associated P- glycoprotein of by phosphorothioate and 5′ cholesterol-conjugated phosphorothioate antisense oligonucleotides, Mol. Pharmacol., 50, 808.


49. Krieg, A. M., Tonkinson, J., Matson, S., Zhao, Q., Saxon, M., Zhang, L. M., Bhanja, U., Yakubov, L., and Stein, C. A. 1993. Modification of antisense phosphodiester oligodeoxynucleotides by a 5′ cholesteryl moiety increases cellular association and improves efficacy, Proc. Natl. Acad. Sci. USA, 90, 1048, 1993. 50. Leonetti, J. P., Degols, G., Clarenc, J. P., Mechti, N., and Lebleu, B. 1993. Cell delivery and mechanisms of action of antisense oligonucleotides, Prog. Nucleic Acid Res. Mol. Biol., 44, 143. 51. Degols, G., Leonetti, J. P., Benkirane, M., Devaux, C., and Lebleu, B. 1992. Poly(L-lysine)-conjugated oligonucleotides promote sequence-specific inhibition of acute HIV-1 infection, Antisense Res. Dev., 2, 293. 52. Citro, G., Perrotti, D., Cucco, C., I, D. A., Sacchi, A., Zupi, G., and Calabretta, B. 1992. Inhibition of leukemia cell proliferation by receptor-mediated uptake of c-myb antisense oligodeoxynucleotides, Proc. Natl. Acad. Sci. USA, 89, 7031. 53. Bonfils, E., Depierreux, C., Midoux, P., Thuong, N. T., Monsigny, M., and Roche, A. C. 1992. Drug targeting: synthesis and endocytosis of oligonucleotide- neoglycoprotein conjugates, Nucleic Acids Res., 20, 4621. 54. Wu, G. Y. and Wu, C. H. 1992. Specific inhibition of hepatitis B viral gene expression in vitro by targeted antisense oligonucleotides, J. Biol. Chem., 267, 12436. 55. Roth, C. M., Reiken, S. R., Le Doux, J. M., Rajur, S. B., Lu, X.-M., Morgan, J. R., and Yarmush, M. L. 1997. Targeted antisense modulation of inflammatory cytokine receptors, Biotechnol. Bioeng., 55, 72. 56. Lu, X. M., Fischman, A. J., Jyawook, S. L., Hendricks, K., Tompkins, R. G., and Yarmush, M. L. 1994. Antisense DNA delivery in vivo: liver targeting by receptor-mediated uptake, J. Nucl. Med., 35, 269. 57. Rajur, S. B., Roth, C. M., Morgan, J. R., and Yarmush, M. L. 1997. Covalent protein-oligonucleotide conjugates for efficient delivery of antisense molecules, Bioconjugate Chem., 8, 935. 58. Lavigne, C. and Thierry, A. R. 1997. Enhanced antisense inhibition of human immunodeficiency virus type 1 in cell cultures by DLS delivery system, Biochem. Biophys. Res. Commun., 237, 566. 59. Zelphati, O. and Szoka, F. C., Jr. 1996. Mechanism of oligonucleotide release from cationic liposomes, Proc. Natl. Acad. Sci. USA, 93, 11493. 60. Marcusson, E. G., Bhat, B., Manoharan, M., Bennett, C. F., and Dean, N. M. 1998. Phosphorothioate oligodeoxyribonucleotides dissociate from cationic lipids before entering the nucleus, Nucleic Acids Res., 26, 2016. 61. Bhatia, R. and Verfaillie, C. M. 1998. Inhibition of BCR-ABL expression with antisense oligodeoxynucleotides restores beta1 integrin-mediated adhesion and proliferation inhibition in chronic myelogenous leukemia hematopoietic progenitors, Blood, 91, 3414. 62. Bergan, R., Connell, Y., Fahmy, B., and Neckers, L. 1993. Electroporation enhances c-myc antisense oligodeoxynucleotide efficacy, Nucleic Acids Res., 21, 3567. 63. Schiedlmeier, B., Schmitt, R., Muller, W., Kirk, M. M., Gruber, H., Mages, W., and Kirk, D. L. 1994. Nuclear transformation of Volvox carteri, Proc. Natl. Acad. Sci. USA, 91, 5080. 64. Lesh, R. E., Somlyo, A. P., Owens, G. K., and Somlyo, A. V. 1995. Reversible permeabilization. A novel technique for the intracellular introduction of antisense oligodeoxynucleotides into intact smooth muscle, Circ Res, 77, 220. 65. Spiller, D. G. and Tidd, D. M. 1995. Nuclear delivery of antisense oligodeoxynucleotides through reversible permeabilization of human leukemia cells with streptolysin O, Antisense Res. Dev., 5, 13. 66. Leonetti, J. P., Machy, P., Degols, G., Lebleu, B., and Leserman, L. 1990. Antibody-targeted liposomes containing oligodeoxyribonucleotides complementary to viral RNA selectively inhibit viral replication, Proc. Natl. Acad. Sci. USA, 87, 2448. 67. Stein, C. A. and Cheng, Y. C. 1993. Antisense oligonucleotides as therapeutic agents—is the bullet really magical?, Science, 261, 1004.


68. Lisziewicz, J., Sun, D., Klotman, M., Agrawal, S., Zamecnik, P., and Gallo, R. 1992. Specific inhibition of human immunodeficiency virus type 1 replication by antisense oligonucleotides: an in vitro model for treatment, Proc. Natl. Acad. Sci. USA, 89, 11209. 69. Monia, B. P., Johnston, J. F., Ecker, D. J., Zounes, M. A., Lima, W. F., and Freier, S. M. 1992. Selective inhibition of mutant Ha-ras mRNA expression by antisense oligonucleotides, J. Biol. Chem., 267, 19954. 70. Carter, G. and Lemoine, N. R. 1993. Antisense technology for cancer therapy: does it make sense?, Br. J. Cancer, 67, 869. 71. Chang, E. H., Miller, P. S., Cushman, C., Devadas, K., Pirollo, K. F., Ts’o, P. O., and Yu, Z. P. 1991. Antisense inhibition of ras p21 expression that is sensitive to a point mutation, Biochemistry, 30, 8283. 72. Agrawal, S., Temsamani, J., and Tang, J. Y. 1991. Pharmacokinetics, biodistribution, and stability of oligodeoxynucleotide phosphorothioates in mice, Proc. Natl. Acad. Sci. USA, 88, 7595. 73. Iversen, P. 1991. In vivo studies with phosphorothioate oligonucleotides: pharmacokinetics prologue, Anticancer Drug Des., 6, 531. 74. Henry, S. P., Giclas, P. C., Leeds, J., Pangburn, M., Auletta, C., Levin, A. A., and Kornbrust, D. J. 1997. Activation of the alternative pathway of complement by a phosphorothioate oligonucleotide: potential mechanism of action, J. Pharmacol. Exp. Ther., 281, 810. 75. Yu, D., Iyer, R. P., Shaw, D. R., Lisziewicz, J., Li, Y., Jiang, Z., Roskey, A., and Agrawal, S. 1996. Hybrid oligonucleotides: synthesis, biophysical properties, stability studies, and biological activity, Bioorg. Med. Chem., 4, 1685. 76. Wu, C. H., Wilson, J. M., and Wu, G. Y. 1989. Targeting genes: delivery and persistent expression of a foreign gene driven by mammalian regulatory elements in vivo, J. Biol. Chem., 264, 16985. 77. Ratajczak, M. Z., Kant, J. A., Luger, S. M., Hijiya, N., Zhang, J., Zon, G., and Gewirtz, A. M. 1992. In vivo treatment of human leukemia in a scid mouse model with c-myb antisense oligodeoxynucleotides, Proc. Natl. Acad. Sci. USA, 89, 11823. 78. Higgins, K. A., Perez, J. R., Coleman, T. A., Dorshkind, K., McComas, W. A., Sarmiento, U. M., Rosen, C. A., and Narayanan, R. 1993. Antisense inhibition of the p65 subunit of NF-kappa B blocks tumorigenicity and causes tumor regression, Proc. Natl. Acad. Sci. USA, 90, 9901. 79. Offensperger, W. B., Offensperger, S., Walter, E., Teubner, K., Igloi, G., Blum, H. E., and Gerok, W. 1993. In vivo inhibition of duck hepatitis B virus replication and gene expression by phosphorothioate modified antisense oligodeoxynucleotides, Embo. J., 12, 1257. 80. Simons, M., Edelman, E. R., DeKeyser, J. L., Langer, R., and Rosenberg, R. D. 1992. Antisense c-myb oligonucleotides inhibit intimal arterial smooth muscle cell accumulation in vivo, Nature, 359, 67. 81. Abe, J., Zhou, W., Taguchi, J., Takuwa, N., Miki, K., Okazaki, H., Kurokawa, K., Kumada, M., and Takuwa, Y. 1994. Suppression of neointimal smooth muscle cell accumulation in vivo by antisense cdc2 and cdk2 oligonucleotides in rat carotid artery, Biochem. Biophys. Res. Commun., 198, 16. 82. Bennett, M. R., Anglin, S., McEwan, J. R., Jagoe, R., Newby, A. C., and Evan, G. I. 1994. Inhibition of vascular smooth muscle cell proliferation in vitro and in vivo by c-myc antisense oligodeoxynucleotides, J. Clin. Invest., 93, 820. 83. Bennett, C. F., Kornbrust, D., Henry, S., Stecker, K., Howard, R., Cooper, S., Dutson, S., Hall, W., and Jacoby, H. I. 1997. An ICAM-1 antisense oligonucleotide prevents and reverses dextran sulfate sodium-induced colitis in mice, J. Pharmacol. Exp. Ther., 280, 988. 84. Azad, R. F., Brown-Driver, V., Buckheit, R. W., Jr., and Anderson, K. P. 1995. Antiviral activity of a phosphorothioate oligonucleotide complementary to human cytomegalovirus RNA when used in combination with antiviral nucleoside analogs, Antiviral Res., 28, 101. 85. Anderson, K. P., Fox, M. C., Brown-Driver, V., Martin, M. J., and Azad, R. F. 1996. Inhibition of human cytomegalovirus immediate-early gene expression by an antisense oligonucleotide complementary to immediate-early RNA, Antimicrob. Agents Chemother, 40, 2004.


Further Information The book Antisense Nucleic Acids and Proteins: Fundamentals and Applications, edited by Joseph N. M. Mol and Alexander R. van der Krol, is a collection of reviews in the use of antisense nucleic acids to modulate or downregulate gene expression. The book Antisense RNA and DNA, edited by James A. H. Murray, explores the use of antisense and catalytic nucleic acids for regulating gene expression. The journal Antisense and Nucleic Acid Drug Development, published by Mary Ann Liebert, Inc., presents original research on antisense technology. For subscription information, contact Antisense and Nucleic Acid Drug Development, Mary Ann Liebert, Inc., 2 Madison Avenue, Larchmont, NY 10538; (914) 834-3100, Fax (914) 834-3688, email: [email protected]. The biweekly journal Nucleic Acids Research publishes papers on physical, chemical, and biologic aspects of nucleic acids, their constituents, and proteins with which they interact. For subscription information, contact Journals Marketing; Oxford University Press, Inc., 2001 Evans Road, Cary, NC 27513; 1-800-852-7323, Fax, (919) 677-1714, email: [email protected].


Kaiser, R. “Tools for Genome Analysis.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

104 Tools for Genome Analysis

Robert Kaiser University of Washington

104.1 104.2 104.3 104.4

General Principles Enabling Technologies Tools for Genome Analysis Conclusions

The development of sophisticated and powerful recombinant techniques for manipulating and analyzing genetic material has led to the emergence of a new biologic discipline, often termed molecular biology. The tools of molecular biology have enabled scientists to begin to understand many of the fundamental processes of life, generally through the identification, isolation, and structural and functional analysis of individual or, at best, limited numbers of genes. Biology is now at a point where it is feasible to begin a more ambitious endeavor—the complete genetic analysis of entire genomes. Genome analysis aims not only to identify and molecularly characterize all the genes that orchestrate the development of an organism but also to understand the complex and interactive regulatory mechanisms of these genes, their organization in the genome, and the role of genetic variation in disease, adaptability, and individuality. Additionally, the study of homologous genetic regions across species can provide important insight into their evolutionary history. As can be seen in Table 104.1, the genome of even a small organism consists of a very large amount of information. Thus the analysis of a complete genome is not simply a matter of using conventional techniques that work well with individual genes (comprised of perhaps 1000 to 10,000 base pairs) a sufficient (very large) number of times to cover the genome. Such a brute-force approach would be too slow and too expensive, and conventional data-handling techniques would be inadequate for the task of cataloging, storing, retrieving, and analyzing such a large amount of information. The amount of manual labor and scientific expertise required would be prohibitive. New technology is needed to provide highthroughput, low-cost automation and reduced reliance on expert intervention at intermediate levels in the processes required for large-scale genetic analysis. Novel computational tools are required to deal with the large volumes of genetic information produced. Individual tools must be integrated smoothly to produce an analytical system in which samples are tracked through the entire analytical process, intermediate decisions and branch points are few, a stable, reliable and routine protocol or set of protocols is employed, and the resulting information is presented to the biologic scientist in a useful and meaningful format. It is important to realize that the development of these tools requires the interdisciplinary efforts of biologists, chemists, physicists, engineers, mathematicians, and computer scientists. Genome analysis is a complex and extended series of interrelated processes. The basic processes involved are diagrammed in Fig. 104.1. At each stage, new biologic, chemical, physical (mechanical, optical), and computational tools have been developed within the last 10 years that have begun to enable large-scale (megabase) genetic analysis. These developments have largely been spurred by the goals of


TABLE 104.1 DNA Content of Various Genomes in Monomer Units (Base Pairs) Organism Phage T4 Escherichia coli Saccharomyces Arabidopsis thaliana Caenorhabditis elegans Drosophila melanogaster Mouse Human

Type

Size

Bacteriophage (virus) Bacterium Yeast Plant Nematode Insect (fruit fly) Mammal Mammal

160,000 4,000,000 14,000,000 100,000,000 100,000,000 165,000,000 3,000,000,000 3,500,000,000

Source: Adapted from [1] and [2].

the Human Genome Project, a worldwide effort to decipher the entirety of human genetics. However, biologists are still a significant ways away from having a true genome analysis capability [3], and as such, new technologies are still emerging. This chapter cannot hope to describe in depth the entire suite of tools currently in use in genome analysis. Instead, it will attempt to present the basic principles involved and to highlight some of the recent enabling technological developments that are likely to remain in use in genome analysis for the foreseeable future. Some fundamental knowledge of biology is assumed; in this regard, an excellent beginning text for individuals with a minimal background in molecular biology is that by Watson et al. [1].

FIGURE 104.1


Basic steps in genome analysis.

TABLE 104.2

Enzymes Commonly Used in Genome Analysis

Enzyme Restriction endonuclease DNA polymerase Polynucleotide kinase Terminal transferase Reverse transcriptase DNA ligase

Function

Common Use

Cleave double-stranded DNA at specific sites Synthesize complementary DNA strand Adds phosphate to 5′ end of single-stranded DNA Adds nucleotides to the 3′ end of single-stranded DNA Makes DNA copy from RNA Covalently joins two DNA fragments

Mapping, cloning DNA sequencing, amplification Radiolabeling, cloning Labeling RNA sequencing, cDNA cloning Cloning

104.1 General Principles The fundamental blueprint for any cell or organism is encoded in its genetic material, its deoxyribonucleic acid (DNA). DNA is a linear polymer derived from a four-letter biochemical alphabet—A, C, G, and T. These four letters are often referred to as nucleotides or bases. The linear order of bases in a segment of DNA is termed its DNA sequence and determines its function. A gene is a segment of DNA whose sequence directly determines its translated protein product. Other DNA sequences are recognized by the cellular machinery as start and stop sites for protein synthesis, regulate the temporal or spatial expression of genes, or play a role in the organization of higher-order DNA structures such as chromosomes. Thus a thorough understanding of the DNA sequence of a cell or organism is fundamental to an understanding of its biology. Recombinant DNA technology affords biologists the capability to manipulate and analyze DNA sequences. Many of the techniques employed take advantage of a basic property of DNA, the molecular complementarity of the two strands of the double helix. This complementarity arises from the specific hydrogen-bonding interactions between pairs of DNA bases, A with T and C with G. Paired double strands of DNA can be denatured, or rendered into the component single strands, by any process that disrupts these hydrogens bonds—high temperature, chaotropic agents, or pH extremes. Complementary single strands also can be renatured into the duplex structure by reversing the disruptive element; this process is sometimes referred to as hybridization or annealing, particularly when one of the strands has been supplied from some exogenous source. Molecular biology makes extensive use of the DNA-modifying enzymes employed by cells during replication, translation, repair, and protection from foreign DNA. A list of commonly used enzymes, their functions, and some of the experimental techniques in which they are utilized is provided in Table 104.2.

104.2 Enabling Technologies The following are broadly applicable tools that have been developed in the context of molecular biology and are commonly used in genome analysis. Cloning. Cloning is a recombinant procedure that has two main purposes: First, it allows one to select a single DNA fragment from a complex mixture, and second, it provides a means to store, manipulate, propagate, and produce large numbers of identical molecules having this single ancestor. A cloning vector is a DNA fragment derived from a microorganism, such as a bacteriophage or yeast, into which a foreign DNA fragment may be inserted to produce a chimeric DNA species. The vector contains all the genetic information necessary to allow for the replication of the chimera in an appropriate host organism. A variety of cloning vectors have been developed which allow for the insertion and stable propagation of foreign DNA segments of various sizes; these are indicated in Table 104.3. Electrophoresis. Electrophoresis is a process whereby nucleic acids are separated by size in a sieving matrix under the influence of an electric field. In free solution, DNA, being highly negatively charged by virtue of its phosphodiester backbone, migrates rapidly toward the positive pole of an electric field. If © 2000 by CRC Press LLC

TABLE 104.3

Common Cloning Vectors

Vector Bacteriophage M13 Plasmid Bacteriophage lambda Cosmid Yeast artificial chromosome (YAC)

Approximate Insert Size Range (Base Pairs) 100–5000 100–10,000 10,000–15,000 25,000–50,000 100,000–1,000,000

the DNA is forced instead to travel through a molecularly porous substance, such as a gel, the smaller (shorter) fragments of DNA will travel through the pores more rapidly than the larger (longer) fragments, thus effecting separation. Agarose, a highly purified derivative of agar, is commonly used to separate relatively large fragments of DNA (100 to 50,000 base pairs) with modest resolution (50 to 100 base pairs), while cross-linked polyacrylamide is used to separate smaller fragments (10 to 1,000 base pairs) with single base-pair resolution. Fragment sizes are generally estimated by comparison with standards run in another lane of the same gel. Electrophoresis is used extensively as both an analytical and a preparative tool in molecular biology. Enzymatic DNA Sequencing. In the late 1970s, Sanger and coworkers [4] reported a procedure employing DNA polymerase to obtain DNA sequence information from unknown cloned fragments. While significant improvements and modifications have been made since that time, the basic technique remains the same: DNA polymerase is used to synthesize a complementary copy of an unknown single-stranded DNA (the template) in the presence of the four DNA monomers (deoxynucleotide triphosphates, or dNTPs). DNA polymerase requires a double-stranded starting point, so a single-stranded DNA (the primer) is hybridized at a unique site on the template (usually in the vector), and it is at this point that DNA synthesis is initiated. Key to the sequencing process is the use of a modified monomer, a dideoxynucleotide triphosphate (ddNTP), in each reaction. The ddNTP lacks the 3′-hydroxyl functionality (it has been replaced by a hydrogen) necessary for phosphodiester bond formation, and its incorporation thus blocks further elongation of the growing chain by polymerase. Four reactions are carried out, each containing all four dNTPs and one of the four ddNTPs. By using the proper ratios of dNTPs to ddNTP, each reaction generates a nested set of fragments, each fragment beginning at exactly the same point (the primer) and terminating with a particular ddNTP at each base complementary to that ddNTP in the template sequence. The products of the reactions are then separated by electrophoresis in four lanes of a polyacrylamide slab gel. Since conventional sequencing procedures utilize radiolabeling (incorporation of a small amount of 32P- or 35S-labeled dNTP by the polymerase), visualization of the gel is achieved by exposing it to film. The sequence can be obtained from the resulting autoradiogram, which appears as a series of bands (often termed a ladder) in each of the four lanes. Each band is composed of fragments of a single size, the shortest fragments being at the bottom of the gel and the longest at the top. Adjacent bands represent a single base pair difference, so the sequence is determined by reading up the ladders in the four lanes and noting which lane contains the band with the next largest sized fragments. The enzymatic sequencing process is diagrammed in Fig. 104.2. It should be noted that although other methods exist, the enzymatic sequencing technique is currently the most commonly used DNA sequencing procedure due to its simplicity and reliability. Polymerase Chain Reaction (PCR). PCR [5] is an in vitro procedure for amplifying particular DNA sequences up to 108-fold that is utilized in an ever-increasing variety of ways in genome analysis. The sequence to be amplified is defined by a pair of single-stranded primers designed to hybridize to unique sites flanking the target sequence on opposite strands. DNA polymerase in the presence of the four dNTPs is used to synthesize a complementary DNA copy across the target sequence starting at the two primer sites. The amplification procedure is performed by repeating the following cycle 25 to 50 times (see Fig. 104.3). First, the double-stranded target DNA is denatured at high temperature (94 to 96°C). Second, the mixture is cooled, allowing the primers to anneal to their complementary sites on the target single


FIGURE 104.2 Enzymatic DNA sequencing. A synthetic oligonucleotide primer is hybridized to its complementary site on the template DNA. DNA polymerase and dNTPs are then used to synthesize a complementary copy of the unknown portion of the template in the presence of a chain-terminating ddNTP (see text). A nested set of fragments beginning with the primer sequence and ending at every ddNTP position is produced in each reaction (the ddTTP reaction products are shown). Four reactions are carried out, one for each ddNTP. The products of each reaction are then separated by gel electrophoresis in individual lanes, and the resulting ladders are visualized. The DNA sequence is obtained by reading up the set of four ladders, one base at a time, from smallest to largest fragment.

strands. Third, the temperature is adjusted for optimal DNA polymerase activity, initiating synthesis. Since the primers are complementary to the newly synthesized strands as well as the original target, each cycle of denaturation/annealing/synthesis effectively doubles the amount of target sequence present in the reaction, resulting in a 2″ amplification (n = number of cycles). The initial implementation of PCR utilized a polymerase that was unstable at the high temperatures required for denaturation, thus requiring manual addition of polymerase prior to the synthesis step of every cycle. An important technological development was the isolation of DNA polymerase from a thermophilic bacterium, Thermus aquaticus (Taq), which can withstand the high denaturation temperatures [6]. Additionally, the high optimal synthesis temperature (70 to 72°C) of Taq polymerase improves the specificity of the amplification process by reducing spurious priming from annealing of the primers to nonspecific secondary sites in the target. While PCR can be performed successfully manually, it is a tedious process, and numerous thermal cycling instruments have become commercially available. Modern thermal cyclers are programmable and capable of processing many samples at once, using either small plastic tubes or microtiter plates, and are characterized by accurate and consistent temperature control at all sample positions, rapid temperature ramping, and minimal temperature over/undershoot. Temperature control is provided by a variety of means (Peltier elements, forced air, water) using metal blocks or water or air baths. Speed, precise temperature control, and high sample throughput are the watchwords of current thermal cycler design. © 2000 by CRC Press LLC

FIGURE 104.3 The first cycle in the polymerase chain reaction. In step 1, the double-stranded target DNA is thermally denatured to produce single-stranded species. A pair of synthetic primers, flanking the specific region of interest, are annealed to the single strands to form initiation sites for DNA synthesis by polymerase (step 2). Finally, complementary copies of each target single strand are synthesized by polymerase in the presence of dNTPs, thus doubling the amount of target DNA initially present (step 3). Repetition of this cycle effectively doubles the target population, affording one million-fold or greater amplification of the initial target sequence.

PCR technology is commonly used to provide sufficient material for cloning from genomic DNA sources, to identify and characterize particular DNA sequences in an unknown mixture, to rapidly produce templates for DNA sequencing from very small amounts of target DNA, and in cycle sequencing, a modification of the enzymatic sequencing procedure that utilizes Taq polymerase and thermal cycling to amplify the products of the sequencing reactions. Chemical Synthesis of Oligodeoxynucleotides. The widespread use of techniques based on DNA polymerase, such as enzymatic DNA sequencing and the PCR, as well as of numerous other techniques utilizing short, defined-sequence, single-stranded DNAs in genome analysis, is largely due to the ease with which small oligodeoxynucleotides can be obtained. The chemical synthesis of oligonucleotides has become a routine feature of both individual biology laboratories and core facilities. The most widely used chemistry for assembling short (90% pure product [Aboud et al., 1994]. The intracellular virus is released from the cells by Triton detergent lysis, followed by nuclease enzyme treatment for removal of RNA and DNA. The virus is concentrated and detergent removed by ion-exchange chromatography. PEG precipitation and chloroform solvent extraction purify away most of the cellular proteins, and final purification and polishing are achieved by ionexchange and size-exclusion chromatography. The virus particle is then inactivated with formaldehyde. In this case, inactivation comes last for two reasons. First, the virus is attenuated, so there is no risk to process personnel. Second, placing the inactivation after the size-exclusion step ensures that there are no contaminants or virus aggregates that may cause incomplete inactivation. The first hepatitis B virus vaccines were derived from human plasma [Hilleman, 1993]. The virus is a 22-nm-diameter particle, much larger than most biologic molecules. Isolation was achieved by ammonium sulfate or PEG precipitation, followed by rate zonal centrifugation and isopycnic banding to take advantage of the large particle size. The preparation was then treated with pepsin protease, urea, and formaldehyde or heat. The latter steps ensure inactivation of possible contaminant viruses from the blood serum. More recently, recombinant DNA–derived hepatitis B vaccines are expressed as an intracellular noninfectious particle in yeast and use a completely different purification process. Here, the emphasis is to remove the yeast host contaminants, particularly high levels of nucleic acids and polysaccharides. Details on the various manufacturing processes have been described by Sitrin et al. [1993]. Antibody Preparations Antibody preparation starts from the plasma pool prepared by removing the cellular components of blood. Cold ethanol is added in increments to precipitate fractions of the blood proteins, and the precipitate containing IgG antibodies is collected. This is further redissolved and purified by ultrafiltration, which also exchanges the buffer to the stabilizer formulation. Sometimes ion-exchange chromatography is used for further purification. Although the plasma is screened for viral contamination prior to pooling, all three purification techniques remove some virus.

105.3 Formulation and Delivery Successful vaccination requires both the development of a stable dosage form for in vitro storage and the proper delivery and presentation of the antigen to elicit a vigorous immune response in vivo. This is done by adjuvanting the vaccine and/or by formulating the adjuvanted antigen. An adjuvant is defined as an agent that enhances the immune response against an antigen. A formulation contains an antigen in a delivery vehicle designed to preserve the (adjuvenated) antigen and to deliver it to a specific target


organ or over a desired time period. Despite adjuvanting and formulation efforts, most current vaccines require multiple doses to create immune memory.

Live Organisms Live viruses and bacteria die relatively quickly in liquid solution (without an optimal environment) and are therefore usually stored in the frozen state. Preserving the infectivity of frozen live-organism vaccines is typically accomplished by lyophilization or freeze-drying. The freeze-drying process involves freezing the organism in the presence of stabilizers, followed by sublimation of both bulk water (primary drying) and more tightly bound water (secondary drying). The dehydration process reduces the conformational flexibility of the macromolecules, providing protection against thermal denaturation. Stabilizers also provide conformational stability and protect against other inactivating mechanisms such as amino acid deamidation, oxidation, and light-catalyzed reaction. Final water content of the freeze-dried product is the most important parameter for the drying process. Although low water content enhances storage stability, overdrying inactivates biologic molecules, since removal of tightly bound water disrupts antigen conformation. Influenza virus suspensions have been shown to be more stable at 1.7% (w/w) water than either 0.4% to 1% or 2.1% to 3.2% [Greiff and Rightsel, 1968]. Other lyophilization parameters that must be optimized pertain to heat and mass transfer, including (1) the rate of freezing and sublimation, (2) vial location in the freeze-drier, and (3) the type of vial and stopper used to cap the vial. Rates of freezing and drying affect phase transitions and compositions, changing the viable organism yield on lyophilization and the degradation rate of the remaining viable organisms on storage. Stabilizers are identified by trial-and-error screening and by examining the mechanisms of inactivation. They can be classified into four categories depending on their purpose: specific, nonspecific, competitive, and pharmaceutical. Specific stabilizers are ligands that naturally bind biologic macromolecules. For example, enzyme antigens are often stabilized by their natural substrates or closely related compounds. Antigen stabilizers for the liquid state also stabilize during freezing. Nonspecific stabilizers such as sugars, amino acids, and neutral salts stabilize proteins and virus structures via a variety of mechanisms. Sugars and polyols act as bound water substitutes, preserving conformational integrity without possessing the chemical reactivity of water. Buffer salts preserve optimal pH. Competitive inhibitors outcompete the organism or antigen for inactivating conditions, such as gas-liquid interfaces, oxygen, or trace-metal ions [Volkin and Klibanov, 1989]. Finally, pharmaceutical stabilizers may be added to preserve pharmaceutical elegance, i.e., to prevent collapse of the lyophilized powder during the drying cycle, which creates difficult redissolution. Large-molecular-weight polymers such as carbohydrates (dextrans or starch) or proteins such as albumin or gelatin are used for this purpose. For example, a buffered sorbitol-gelatin medium has been used successfully to preserve the infectivity of measles virus vaccine during lyophilized storage for several years at 2 to 8°C [Hilleman, 1989]. An example of live bacterium formulation to preserve activity on administration is typhoid fever vaccine, administered orally, S. typhi bacteria are lyophilized to a powder to preserve viability on the shelf, and the powder is encapsulated in gelatin to preserve bacterial viability when passing through the low-pH stomach. The gelatin capsule dissolves in the intestine to deliver the live bacteria. Oral polio vaccine is an exception to the general rule of lyophilization, since polio virus is inherently quite stable relative to other viruses. It is formulated as a frozen liquid and can be used for a limited time after thawing [Melnick, 1984]. In the presence of specific stabilizers such as MgCl2, extended 4°C stability can be obtained.

Subunit Antigens Inactivated and/or purified viral and bacterial antigens inherently offer enhanced stability because wholeorganism infectivity does not need to be preserved. However, these antigens are not as immunogenic as live organisms and thus are administered with an adjuvant. They are usually formulated as an aqueous liquid suspension or solution, although they can be lyophilized under the same principles as above. The major


adjuvant recognized as safe for human use is alum. Alum is a general term referring to various hydrated aluminum salts; a discussion of the different alums can be found in Shirodkar et al. [1990]. Vaccines can be formulated with alum adjuvants by two distinct methods: adsorption to performed aluminum precipitates or precipitation of aluminum salts in the presence of the antigen, thus adsorbing and entrapping the antigen. Alum’s adjuvant activity is classically believed to be a “depot” effect, slowly delivering antigen over time in vivo. In addition, alum particles are believed to be phagocytized by macrophages. Alum properties vary depending on the salt used. Adjuvants labeled aluminum hydroxide are actually aluminum oxyhydroxide, AlO(OH). This material is crystalline, has a fibrous morphology, and has a positive surface charge at neutral pH. In contrast, aluminum phosphate adjuvants are networks of platelike particles of amorphous aluminum hydroxyphosphate and possess a negative surface charge at neutral pH. Finally, alum coprecipitate vaccines are prepared by mixing an acidic alum solution of KAl(SO4)2 ·12H2O with an antigen solution buffered at neutral pH, sometimes actively pH-controlled with base. At neutral pH, the aluminum forms a precipitate, entrapping and adsorbing the antigen. The composition and physical properties of this alum vary with processing conditions and the buffer anions. In general, an amorphous aluminum hydroxy(buffer anion)sulfate material is formed. Process parameters must be optimized for each antigen to ensure proper adsorption and storage stability. First, since antigen adsorption isotherms are a function of the antigen’s isoelectric point and the type of alum used [Seeber et al., 1991], the proper alum and adsorption pH must be chosen. Second, the buffer ions in solution can affect the physical properties of alum over time, resulting in changes in solution pH and antigen adsorption. Finally, heat sterilization of alum solutions and precipitates prior to antigen adsorption can alter their properties. Alum is used to adjuvant virtually all the existing inactivated or formaldehyde-treated vaccines, as well as purified subunit vaccines such as HBsAg and Hib-conjugate vaccines. The exception is for some bacterial polysaccharide vaccines and for new vaccines under development (see below). An interesting vaccine development challenge was encountered with Hib-conjugate pediatric vaccines, which consist of purified capsular polysaccharides. Although purified, unadjuvanted polysaccharide is used in adults, it is not sufficiently immunogenic in children under age 2, the population is greatest risk [Ellis, 1992; Howard, 1992]. Chemical conjugation, or cross-linking, of the PRP polysaccharide to an antigenic protein adjuvant elicits T-helper cell activation, resulting in higher antibody production. Variations in conjugation chemistry and protein carriers have been developed; example proteins are the diphtheria toxoid (CRM 197), tetanus toxoid, and the outer membrane protein complex of N. meningitidis [Ellis, 1992; Howard, 1992]. The conjugated polysaccharide is sometimes adsorbed to alum for further adjuvant action.

105.4 Future Trends The reader will have noted that many production aspects for existing vaccines are quite archaic. This is so because most vaccines were developed before the biotechnology revolution, which is creating a generation of highly purified and better-characterized subunit vaccines. As such, for older vaccines “the process defines the product,” and process improvements cannot readily be incorporated into these poorly characterized vaccines without extensive new clinical trials. With improved scientific capabilities, we can understand the effects of process changes on the physicochemical properties of new vaccines and on their behavior in vivo.

Vaccine Cultivation Future cultivation methods will resemble existing methods of microbial and virus culture. Ill-defined medium components and cells will be replaced to enhance reproducibility in production. For bacterial and ex vivo cultivated virus, analytical advances will make monitoring the environment and nutritional status of the culture more ubiquitous. However, the major changes will be in novel product types—single-molecule subunit antigens, virus-like particles, monoclonal antibodies, and gene-therapy vaccines, each of which will incorporate novel processes.


Newer subunit vaccine antigens will be cultivated via recombinant DNA in microbial or animal cells. Several virus-like particle vaccines are under development using recombinant baculovirus (nuclear polyhedrosis virus) to infect insect cells (spodoptera frugipeeda or trichoplusia ni). Like the hepatitis B vaccine, the viral antigens spontaneously assemble into a noninfectious capsid within the cell. Although the metabolic pathways of insect cells differ from vertebrates, cultivation principles are similar. Insect cells do not require surface attachment and are grown much like bacteria. However, they also lack a cell wall and are larger and hence more fragile than vertebrate cells. Passive antibody vaccines have been prepared up to now from human blood serum. Consequently, there has been no need for cultivation methods beyond vaccination and conventional harvest of antibodycontaining blood from donors. Due to safety concerns over using human blood, passive vaccines will likely be monoclonal antibodies or cocktails thereof prepared in vitro by the cultivation of hybridoma or myeloma cell lines. This approach is under investigation for anti-HIV-1 antibodies [Emini et al., 1992]. Cultivation of these cell lines involves the same principles of animal cell cultivation as described above, with the exception that hybridomas can be less fastidious in nutritional requirements, and they do not require surface attachment for growth. These features will allow for defined serum-free media and simpler cultivation vessels and procedures. For the gene-therapy approach, the patient actually produces the antigen. A DNA polynucleotide encoding protein antigen(s) is injected intramuscularly into the patient. The muscle absorbs the DNA and produces the antigen, thereby eliciting an immune response [Ulmer et al., 1993]. For cultivation, production of the DNA plasmid is the objective, which can be done efficiently by bacteria such as Escherichia coli. Such vaccines are not sufficiently far along in development to generalize the factors that influence their production; however, it is expected that producer cells and process conditions that favor high cell mass, DNA replication, and DNA stability will be important. A potential beauty of this vaccination approach is that for cultivation, purification, and formulation, many vaccines can conceivably be made by identical processes, since the plasmids are inactive within the bacterium and possess roughly the same nucleotide composition.

Downstream Processing Future vaccines will be more highly purified in order to minimize side effects, and future improvements will be to assist this goal. The use of chemically defined culture media will impact favorably on downstream processing by providing a cleaner feedstock. Advances in filtration membranes and in chromatographic support binding capacity and throughput will improve ease of purification. Affinity purification methods that rely on specific “lock and key” interactions between a chromatographic support and the antigen will see greater use as well. Techniques amenable to larger scales will be more important to meet increased market demands and to reduce manufacturing costs. HPLC and other analytical techniques will provide greater process monitoring and control throughout purification. As seen during the evolution of diphtheria and tetanus toxoid vaccines, the trend will be to purify toxins prior to inactivation to reduce their cross-linking with other impurities. New inactivating agents such as hydrogen peroxide and ethyl dimethylaminopropyl carbodiimide have been investigated for pertussis toxin, which do not have problems of cross-linking or reversion of the toxoid to toxin status. Molecular biology is likely to have an even greater impact on purification. Molecular cloning of proteins allows the addition of amino acid sequences that can facilitate purification, e.g., polyhistidine or polyalanine tails for metal ion, or ion-exchange chromatography. Recent efforts also have employed genetic manipulation to inactivate toxins, eliminating the need for the chemical treatment step.

Vaccine Adjuvants and Formulation Many new subunit antigens lack the inherent immunogenicity found in the natural organism, thereby creating the need for better adjuvants. Concomitantly, the practical problem of enhancing worldwide immunization coverage has stimulated development of single-shot vaccine formulations in which


booster doses are unnecessary. Thus future vaccine delivery systems will aim at reducing the number of doses via controlled antigen release and will increase vaccine efficacy by improving the mechanism of antigen presentation (i.e., controlled release of antigen over time or directing of antigen to specific antigen-presenting cells). Major efforts are also being made to combine antigens into single-shot vaccines to improve immunization rates for infants, who currently receive up to 15 injections during the first 2 years of life. Coadministration of antigens presents unique challenges to formulation as well. Recent advances in the understanding of in vivo antigen presentation to the immune system has generated considerable interest in developing novel vaccine adjuvants. The efficacy of an adjuvant is judged by its ability to stimulate specific antibody production and killer cell proliferation. Developments in biology now allow analysis of activity by the particular immune cells that are responsible for these processes. Examples of adjuvants currently under development include saponin detergents, muramyl dipeptides, and lipopolysaccharides (endotoxin), including lipid A derivatives. As well, cytokine growth factors that stimulate immune cells directly are under investigation. Emulsion and liposome delivery vehicles are also being examined to enhance the presentation of antigen and adjuvant to the immune system [Edelman, 1992; Allison and Byars, 1992]. Controlled-release delivery systems are also being developed that encapsulate antigen inside a polymer-based solid microsphere. The size of the particles typically varies between 1 and 300 µm depending on the manufacturing process. Microspheres are prepared by first dissolving the biodegradable polymer in an organic solvent. The adjuvanted antigen, in aqueous solution or lyophilized powder form, is then emulsified into the solvent-polymer continuous phase. Microspheres are then formed by either solvent evaporation, phaseseparation, or spray-drying, resulting in entrapment of antigen [Morris et al., 1994]. The most frequently employed biodegradable controlled-released delivery systems use FDA-approved poly(lactide-co-glycolide) copolymers (PLGA), which hydrolyze in vivo to nontoxic lactic and glycolic acid monomers. Degradation rate can be optimized by varying the microsphere size and the monomer ratio. Antigen stability during encapsulation and during in vivo release from the microspheres remains a challenge. Other challenges to manufacturing include encapsulation process reproducibility, minimizing antigen exposure to denaturing organic solvents, and ensuring sterility. Methods are being developed to address these issues, including the addition of stabilizers for processing purposes only. It should be noted that microsphere technology may permit vaccines to be targeted to specific cells; they can potentially be delivered orally or nasally to produce a mucosal immune response. Other potential delivery technologies include liposomes and alginate polysaccharide and poly(dicarboxylatophenoxy)phosphazene polymers. The latter two form aqueous hydrogels in the presence of divalent cations [Khan et al., 1994]. Antigens can thus be entrapped under aqueous conditions with minimal processing by simply mixing antigen and soluble aqueous polymer and dripping the mixture into a solution of CaCl2. The particles erode by Ca2+ loss, mechanical and chemical degradation, and macrophage attack. For alginate polymers, monomer composition also determines the polymer’s immunogenicity, and thus the material can serve as both adjuvant and release vehicle. For combination vaccines, storage and administration compatibility of the different antigens must be demonstrated. Live-organism vaccines are probably not compatible with purified antigens, since the former usually require lyophilization and the latter are liquid formulas. Within a class of vaccines, formulation is challenging. Whereas it is relatively straightforward to adjuvant and formulate a single antigen, combining antigens is more difficult because each has its own unique alum species, pH, buffer ion, and preservative optimum. Nevertheless, several combination vaccines have reached the market, and others are undergoing clinical trials.

105.5 Conclusions Although vaccinology and manufacturing methods have come a considerable distance over the past 40 years, much more development will occur. There will be challenges for biotechnologists to arrive at safer, more effective vaccines for an ever-increasing number of antigen targets. If government interference


and legal liability questions do not hamper innovation, vaccines will remain one of the most cost-effective and logical biomedical technologies of the next century, as disease is prevented rather than treated. Challenges are also posed in bringing existing vaccines to technologically undeveloped nations, where they are needed most. This problem is almost exclusively dominated by the cost of vaccine manufacture and the reliability of distribution. Hence it is fertile ground for engineering improvements in vaccine production.

Defining Terms Adjuvant: A chemical or biologic substance that enhances immune response against an antigen. Used here as a verb, the action of combining an antigen and an adjuvant. Antigen: A macromolecule or assembly of macromolecules from a pathogenic organism that the immune system recognizes as foreign. Attenuation: The process of mutating an organism so that it no longer causes disease. Immunogen: A molecule or assembly of molecules with the ability to invoke an immune system response. Pathogen: A disease-causing organism, either a virus, mycobacterium, or bacterium.

References Aboud RA, Aunins JG, Buckland BC, et al. 1994. Hepatitis A Virus Vaccine. International patent application, publication number WO 94/03589, Feb. 17, 1994. Allison AC, Byars NE. 1992. Immunologic adjuvants and their mode of action. In RW Ellis (ed), Vaccines: New Approaches to Immunological Problems, p 431. Reading, Mass, Butterworth-Heinemann. Atkinson B, Mavituna F. 1991. Biochemical Engineering and Biotechnology Handbook, 2d ed. London, Macmillan. Aunins JG, Henzler H-J. 1993. Aeration in cell culture bioreactors. In H-J Rehm et al. (eds), Biotechnology, 2d ed., vol 3, p 219. Weinheim, Germany, VCH Verlag. Bachmayer H. 1976. Split and subunit vaccines. In P. Selby (ed), Influenza Virus, Vaccines, and Strategy, p 149. New York, Academic Press. Barbet AF. 1989. Vaccines for parasitic infections. Adv Vet Sci Comp Med 33:345. Cryz SJ, Reinhard G. 1990. Large-scale production of attenuated bacterial and viral vaccines. In GC Woodrow, MM Levine (eds), New Generation Vaccines, p 921. New York, Marcel Dekker. Datar RV, Rosen C-G. 1993. Cell and cell debris removal: Centrifugation and crossflow filtration. In H-J Rehm et al. (eds), Biotechnology, 2d ed., vol 3, p 469. Weinheim, Germany, VCH Verlag. Dobrota M, Hinton R. 1992. Conditions for density gradient separations. In D Rickwood (ed), Preparative Centrifugation: A Practical Approach, p 77. New York, Oxford U Press. Eagle H, Habel K. 1956. The nutritional requirements for the propagation of poliomyelitis virus by the HeLa cell. J Exp Med 104:271. Edelman R. 1992. An update on vaccine adjuvants in clinical trial. AIDS Res Hum Retrovir 8(8):1409. Ellis RW. 1992. Vaccine development: Progression from target antigen to product. In JE Ciardi et al. (eds), Genetically Engineered Vaccines, p 263. New York, Plenum Press. Emini EA, Schleif WA, Nunberg JH, et al. 1992. Prevention of HIV-1 infection in chimpanzees by gp120 V3 domain-specific monoclonal antibodies. Nature 355:728. Greiff D, Rightsel WA. 1968. Stability of suspensions of influenza virus dried to different contents of residual moisture by sublimation in vacuo. Appl Microbiol 16(6):835. Hanisch W. 1986. Cell harvesting. In WC McGregor (ed), Membrane Separations in Biotechnology, p 66. New York, Marcel Dekker. Hewlett EL, Cherry JD. 1990. New and improved vaccines against pertussis. In GC Woodrow, MM Levine (eds), New Generation Vaccines, p 231. New York, Marcel Dekker.


Hilleman MR. 1989. Improving the heat stability of vaccines: Problems, needs and approaches. Rev Infect Dis 11(suppl 3):S613. Hilleman MR. 1993. Plasma-derived hepatitus B vaccine: A breakthrough in preventive medicine. In R Ellis (ed), Hepatitus B Vaccines in Clinical Practice, p 17. New York, Marcel Dekker. Howard AJ. 1992. Haemophilus influenzae type-b vaccines. Br J Hosp Med 48(1):44. Ingraham JL, Maaløe O, Neidhardt FC. 1983. Growth of the Bacterial Cell. Sunderland, Mass, Sinauer. Jagicza A, Balla P, Lendvai N, et al. 1986. Additional information for the continuous cultivation of Bordetella pertussis for the vaccine production in bioreactor. Ann Immunol Hung 26:89. Janson J-C, Ryden L (eds). 1989. Protein Purification Principles, High Resolution Methods, and Applications. Weinheim, Germany, VCH Verlag. Kelley BD, Hatton TA. 1993. Protein purification by liquid-liquid extraction. In H-J Rehm et al. (eds), Biotechnology, 2d ed, vol 3, p 594. Weinheim, Germany, VCH Verlag. Khan MZI, Opdebeeck JP, Tucker IG. 1994. Immunopotentiation and delivery systems for antigens for single-step immunization: Recent trends and progress. Pharmacol Res 11(1):2. Melnick JL. 1984. Live attenuated oral poliovirus vaccine. Rev Infect Dis 6(suppl 2):S323. Metzgar DP, Newhart RH. 1977. U.S. patent no. 4,057,626, Nov. 78, 1977. Montagnon B, Vincent-Falquet JC, Fanget B. 1984. Thousand litre scale microcarrier culture of vero cells for killed polio virus vaccine: Promising results. Dev Biol Stand 55:37. Morris W, Steinhoff MC, Russell PK. 1994. Potential of polymer microencapsulation technology for vaccine innovation. Vaccine 12(1):5. Pappenheimer AM. 1984. Diphtheria. In R Germanier (ed), Bacterial Vaccines, p 1. New York, Academic Press. Polson A. 1993. Virus Separation and Preparation. New York, Marcel Dekker. Prokop A, Rosenberg MZ. 1989. Bioreactor for mammalian cell culture. In A Fiechter (ed), Advances in Biochemical Engineering, vol 39: Vertebrate Cell Culture II and Enzyme Technology, p 29. Berlin, Springer-Verlag. Rappuoli R. 1990. New and improved vaccines against diphtheria and tetanus. In GC Woodrow, MM Levine (eds), New Generation Vaccines, p 251. New York, Marcel Dekker. Relyveld EH. 1980. Current developments in production and testing of tetanus and diphtheria vaccines. In A Mizrahi et al. (eds), Progress in Clinical and Biological Research, vol 47: New Developments with Human and Veterinary Vaccines, p 51. New York, Alan R Liss. Relyveld EH, Ben-Efraim S. 1983. Preparation of vaccines by the action of glutaraldehyde on toxins, bacteria, viruses, allergens and cells. In SP Colowic, NO Kaplan (eds), Methods in Enzymology, vol 93, p 24. New York, Academic Press. Reuveny S. 1990. Microcarrier culture systems. In AS Lubiniecki (ed), In Large-Scale Mammalian Cell Culture Technology, p 271. New York, Marcel Dekker. Sato Y, Kimura M, Fukumi H. 1984. Development of a pertussis component vaccine in Japan. Lancet 1(8369):122. Seeber SJ, White JL, Helm SL. 1991. Predicting the adsorption of proteins by aluminum-containing adjuvants. Vaccine 9:201. Shirodkar S, Hutchinson RL, Perry DL, et al. 1990. Aluminum compounds used as adjuvants in vaccines. Pharmacol Res 7(12):1282. Sitrin RD, Wampler DE, Ellis R. 1993. Survey of licensed hepatitis B vaccines and their product processes. In R Ellis (ed), Hepatitus B Vaccines in Clinical Practice, p 83. New York, Marcel Dekker. Sureau P. 1987. Rabies vaccine production in animal cell cultures. In A Fiechter (ed), Advances in Biochemical Engineering and Biotechnology, vol 34, p 111. Berlin, Springer-Verlag. Tyrrell DAJ. 1976. Inactivated whole virus vaccine. In P Selby (ed), Influenza, Virus, Vaccines and Strategy, p 137. New York, Academic Press. Ulmer JB, Donnelly JJ, Parker SE, et al. 1993. Heterologous protection against influenza by injection of DNA encoding a viral protein. Science 259(5102):1745.


Volkin DB, Klibanov AM. 1989. Minimizing protein inactivation. In TE Creighton (ed), Protein Function: A Practical Approach, pp 1–12. Oxford, IRL Press.

Further Information A detailed description of all the aspects of traditional bacterial vaccine manufacture may be found in the World Health Organization technical report series for the production of whole-cell pertussis, diphtheria, and tetanus toxoid vaccines: World Health Organization. 1997a. BLG/UNDP/77.1 Rev. 1. Manual for the Production and Control of Vaccines: Diphtheria Toxoid. World Health Organization. 1997b. BLG/UNDP/77.2 Rev. 1. Manual for the Production and Control of Vaccines: Tetanus Toxoid. World Health Organization. 1997c. BLG/UNDP/77.3 Rev. 1. Manual for the Production and Control of Vaccines: Pertussis Vaccine. A description of all the aspects of cell culture and viral vaccine manufacture may be found in Spier RE, Griffiths JB. 1985. Animal Cell Biotechnology, vols 1 to 3. London, Academic Press. For a review of virology and virus characteristics, the reader is referred to Fields BN, Knipe DM (eds). 1990. Virology, 2d ed, vols 1 and 2. New York, Raven Press.


Le Doux, J. M., Morgan, J. R.,Yarmush, M. L. “ Gene Therapy.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

106

Joseph M. Le Doux The Center for Engineering in Medicine, and Surgical Services, Massachusetts General Hospital, Harvard Medical School, and the Shriners Burns Hospital

Gene Therapy

Jeffrey R. Morgan The Center for Engineering in Medicine, and Surgical Services, Massachusetts General Hospital, Harvard Medical School, and the Shriners Burns Hospital

Martin L. Yarmush The Center for Engineering in Medicine, and Surgical Services, Massachusetts General Hospital, Harvard Medical School, and the Shriners Burns Hospital

106.1 106.2 106.3 106.4 106.5 106.6 106.7 106.8 106.9

Background Recombinant Retroviruses Recombinant Adenoviruses Recombinant Adeno-Associated Viruses Direct Injection of Naked DNA Particle-Mediated Gene Transfer Liposome-Mediated Gene Delivery Other Gene Transfer Methods Summary and Conclusion

Gene therapy, the transfer of genes into cells for a therapeutic effect, is a new approach to the treatment of disease. The first clinically applicable system for efficiently delivering genes into mammalian cells was developed in the early 1980s and was based on a genetically engineered retrovirus, which, as part of its lifecycle, stably integrates its genome into the target cell’s chromosomal DNA. Using recombinant DNA technology perfected in the mid-1970s, investigators replaced the viral genes with therapeutic genes and the resulting recombinant retrovirus shuttled these genes into the target cells. The potential applications of gene therapy are far reaching (Table 106.1) since there are over 4000 known human genetic diseases (many of which have no viable treatment) and virtually every human disease is profoundly influenced by genetic factors [Anderson, 1992]. In one of the first clinical applications, gene therapy was used to treat patients with ADA (adenosine deaminase) deficiency, a genetic defect which causes severe combined immune deficiency (SCID) and death at an early age [Anderson, 1992]. The patient’s lymphocytes were isolated and transduced (insertion of a foreign gene into the genome of a cell) by a recombinant retrovirus encoding a functional ADA gene. These transduced cells were expanded in culture then reinfused back to the patient. These protocols were conducted after an exhaustive peer review process that laid the groundwork for future gene therapy protocols. Successful treatment of other genetic diseases is likely to be achieved in the future [Levine and Friedmann, 1993; Roemer and Friedmann, 1992]. In addition to inherited diseases, other viable targets for gene therapy include more prevalent disorders that show a complex genetic dependence (i.e., cancer and heart disease) as well as infectious diseases (i.e., human immunodeficiency virus (HIV) and applications in tissue engineering [Anderson, 1992; Morgan and Yarmush, 1998].

106.1 Background Gene therapy protocols conduct gene transfer in one of two settings; either ex vivo or in vivo [Mulligan, 1993]. For ex vivo gene therapy, target cells or tissue are removed from the patient, grown in culture, genetically modified and then reinfused or retransplanted into the patient [Ledley, 1993]. Ex-vivo gene therapy is limited to those tissues which can be removed, cultured in vitro and returned to the patient


TABLE 106.1

Target Diseases for Gene Therapy

Target Disease

Target Tissues

Corrective Gene

Gene Delivery Systems Used

Inherited ADA deficiency Alpha-1 antitrypsin deficiency

Alzheimer’s Disease Cystic fibrosis Diabetes Duchenne muscular dystrophy Familial hypercholesterolemia Gaucher disease Growth hormone deficiency

Hemoglobinopathies Hemophilia

Leukocyte adhesion deficiency Parkinson’s disease Phenylketonuria Purine nucleoside phosphorylase deficiency Urea cycle disorders

Hematopoietic cells Fibroblasts Hepatocytes Lung epithelia cells Peritoneal mesothelial cells Nervous system Lung epithelia cells Fibroblasts Hepatocytes Muscle cells Hepatocytes Hematopoietic cells Fibroblasts Endothelial cells Fibroblasts Keratinocytes Muscle cells Hematopoietic cells Fibroblasts Keratinocytes Hepatocytes Muscle cells Hematopoietic cells Nervous system Hepatocytes Fibroblasts

ADA Alpha-1 antitrypsin

Nerve growth factor CFTR Human insulin Dystrophin LDL receptor Glucocerebrosidase Human growth hormone

α or β-globin Factor VIII, IX

CD-18 Tyrosine hydroxylase Phenylalanine hydroxylase Purine nucleoside phosphorylase

RV, RM RV RV AV AV RV, AV, HSV AV,AAV, RM, L RV L AV, DI RV, AV, RM RV RV RV TR RV RV RV, AAV RV, DI, L RV RV, AV, RM, L RV RV RV, AV, HSV RV RV

Hepatocytes

Ornithine transcarbamylase or arginosuccinate synthetase

AV

Infectious diseases

Acute lymphoblastic leukemia Brain tumors Carcinoma Melanoma Retinoblastoma HIV

Cardiomyopathy Emphysema Local thrombosis Vaccines

Muscle cells Lung epithelial cells Endothelial cells Muscle cells

p53 HSV thymidine kinase γ-interferon Tumor necrosis factor Retinoblastoma gene Dominant negative Rev TAR decoy RRE decoy Diptheria toxin A (Used reporter gene) Alpha-1 antitrypsin Anti-clotting factors Various

RV RV TR RV RV RV RV TR RV DI AV RV DI

Acquired Cancer

Note: AAV = recombinant adeno-associated viruses AV = recombinant adenoviruses DI = direct injection HSV = recombinant herpes simplex virus L = lipofection RM = receptor mediated RV = recombinant retroviruses TR = transfection


TABLE 106.2

Physical Characteristics of Recombinant Virions

Characteristic Genome type Genome size Genome MW Particle diameter Particle mass Composition DNA/RNA Protein Lipid Density Enveloped? Shape Surface projections Number Length Max diameter Virus titer Integration?

Units Bases Daltons nm Grams % % % g/cm3 CsCl Yes/no Yes/no nm nm pfu/ml

Recombinant Retroviruses

Recombinant Adenoviruses

Recombinant AAV

ss RNA (2 per virion) 8300 3 × 106 90–147 3.6 × 10–16

ds DNA 36000 20–25 × 106 65–80 2.9 × 10–16

ss DNA 4700 1.2–1.8 × 106 20–24 1.0 × 10–17

2 62 36 1.15–1.16 Yes Spherical Yes ~60–200 5 8 106–107 Yes, random

13 87 0 1.33–1.35 No Icosahedral Yes 12 25–30 4 1010–1012 No, episomal

26 74 0 1.39–1.42 No Icosahedral No

105–107 Yes, chromosome 19

and cannot be applied to many important target tissues and organs such as the lungs, brain and heart. For in vivo gene therapy, the gene transfer agent is delivered directly to the target tissue/organ, and gene transfer occurs in the patient rather than in the tissue culture dish [Mulligan, 1993]. Both strategies have inherent advantages and disadvantages and current research is evaluating which approach can best meet the needs of a particular disease. Gene delivery systems can be classified as either viral or non-viral [Friedmann, 1989]. For viral gene transfer, one of several different types of viruses is engineered to deliver genes. Typically, viral genes are removed to prevent self-replication of the virus and to provide room for the insertion of one or more therapeutic genes that the recombinant virus will carry. To further ensure the safety of the recombinant viruses, specialized packaging cell lines have been developed to produce the recombinant viruses and minimize the production of infectious wild-type viruses. Some viruses are able to integrate the therapeutic genes into the target cell’s nuclear DNA (retroviruses, adeno associated viruses), whereas others are not (adeno viruses) (Table 106.2). Non-viral gene transfer systems are based on a variety of technologies that employ physical/chemical means to deliver genes [Felgner and Rhodes, 1991]. These technologies include direct plasmid injection, bombardment with DNA coated microprojectiles, and DNA complexed with liposomes. Some non-viral transfection techniques are too inefficient (e.g., coprecipitation of DNA with calcium phosphate [Chen and Okayama, 1987], DNA complexed with DEAE-dextran [Pagano et al., 1967], electroporation [Neumann et al., 1982]) or laborious (e.g., microinjection of DNA [Capecchi, 1980]) for clinical use. Only those gene delivery systems (viral and non-viral) with potential for clinical application will be discussed in this article. The main features of these technologies (Table 106.3) will be described and specific examples of their applications highlighted.

106.2 Recombinant Retroviruses Many of the approved clinical trials have utilized recombinant retroviruses for gene delivery. Retroviral particles contain two copies of identical RNA genomes that are wrapped in a protein coat and further encapsidated by a lipid bilayer membrane. The virus attaches to specific cell surface receptors via surface proteins that protrude from the viral membrane. The particle is then internalized and its genome is released into the cytoplasm, reverse transcribed from RNA to DNA, transported into the nucleus and


TABLE 106.3

Features of the Various Gene Transfer Systems

Features Maximum transgene size Maximum concentration (vectors/ml) Transfers genes to quiescent cells Integrates transgene into target cell genome Persistence of gene expression Immunological problems Pre-existing host immunity Stability Ease of large scale production Safety concerns

Retrovirus

AAV

Adenovirus

Non-viral

8 kb ~107

4.7 kb ~1012

36 kb ~1012

36 kb 36 kb

No/Yes* Yes

Yes Yes

Yes No

Yes No

wks–yrs Few No Poor Difficult Insertional mutagenesis

yrs None known Yes Good Difficult Inflammation toxicity

wks–mos Extensive Yes Good Easy Inflammation toxicity

days–wks None No Good Easy Toxicity

* Recombinant lentiviruses, such as human immunodeficiency virus, are capable of transducing quiescent cells.

then integrated into the cell’s chromosomal DNA. The integrated viral genome has LTRs (long terminal repeats) at both ends which encode the regulatory sequences that drive the expression of the viral genome [Weiss et al., 1982]. Retroviruses used for gene transfer are derived from wild-type murine retroviruses. The recombinant viral particles are structurally identical to the wild type virus but carry a genetically engineered genome (retroviral vector) which encodes the therapeutic gene of interest. These recombinant viruses are incapable of self-replication but can infect and insert their genomes into a target cell’s genome [Morgan et al., 1993]. Recombinant retroviruses, like all other recombinant viruses, are produced by a two part system composed of a packaging cell line and a recombinant vector (Fig. 106.1) [Anderson, 1992; Levine and Friedmann, 1993]. The packaging cell line has been engineered to express all the structural viral genes (gag, pol and env) necessary for the formation of an infectious virus particle. gag encodes the capsid proteins and is necessary for encapsidation of the vector. pol encodes the enzymatic activities of the virus including reverse transcriptase and integrase. env encodes the surface proteins on the virus particle which are necessary for attachment to the target cell’s receptors. The retroviral vector is essentially the wild-type genome with all the viral genes removed. This vector encodes the transgene(s) and the regulatory sequences necessary for their expression as well as a special packaging sequence, (ψ), that is required for encapsidation of the genome into an infectious viral particle [Morgan et al., 1993]. To produce recombinant retrovirus particles, the retroviral vector is transfected into the packaging cell line. The structural proteins expressed by the packaging cell line recognize the packaging sequence on RNAs transcribed from the transfected vector and encapsidate them into an infectious virus particle that is subsequently exocytosed by the cell and released into the culture medium. This medium containing infectious recombinant viruses is harvested and used to transduce target cells. Many different retroviral vector designs have been used (Fig. 106.2). A commonly used vector encodes two genes, one expressed from the LTR and the other from an internal promoter (Fig. 106.2c) [Miller, 1992]. Often, one gene expresses a therapeutic protein and the other a selectable marker that makes the modified cells resistant to selective media or a drug. This allows the investigator to establish a culture composed solely of transduced cells by growing them under the selective conditions. Several configurations are possible, but the optimum design is often dictated by the transgene(s) being expressed and the cell type to be transduced. Vector configuration is crucial for maximizing viral titer and transgene expression [Roemer and Friedmann, 1992]. As with all gene transfer technologies, there are advantages and disadvantages to the use of recombinant retroviruses. Retroviruses can only transduce dividing cells since integration requires passage of the target cells through mitosis [Roe et al., 1993]. This limits the use of recombinant retroviruses for in vivo gene therapy, since few normal cells are actively dividing. Recently, however, new retroviral vectors based on lentiviruses have been developed that are capable of transducing non-dividing cells [Naldini et al., 1996]. © 2000 by CRC Press LLC

FIGURE 106.1 Packaging cell line for retrovirus. A simple retroviral vector composed of 2 LTR regions which flank sequences encoding the packaging sequence (Ψ) and a therapeutic gene. A packaging cell line is transfected with this vector. The packaging cell line expresses the three structural proteins necessary for formation of a virus particle (gag, pol, and env). These proteins recognize the packaging sequence on the vector and form an infectious virion around it. Infectious virions bud from the cell surface into the culture medium. The virus-laden culture medium is filtered to remove cell debris and then either immediately used to transduce target cells or the virions are purified and/or concentrated and frozen for later use.

These HIV based vectors were able to achieve stable gene transfer after injection into a non-dividing tissue (brain) in vivo. Other drawbacks of retroviral vectors include: (a) a limitation to the size of inserted genes (9–10). To reduce hydrolysis, surfaces are commonly activated in nonaqueous solvents; ligands are then coupled using concentrated aqueous solutions (>1 mg/ml) for extended periods (>12 hr). Some ligands are soluble in polar solvents such as dimethylsulfoxide (DMSO), dimethylformamide (DMF), acetone, dioxane, or ethanol, or their aqueous cosolvents, and these combinations may be used to reduce hydrolysis during coupling.

FIGURE 110.1


General chemical scheme for grafting a ligand onto a surface.

110.2 Surface Bioconjugation The typical chemical groups involved in immobilizing ligands for cell-surface studies are hydroxyls, amines, carboxylic acids, and thiols. Other groups, such as amides, disulfides, phenols, guanidines, thioethers, indoles, and imidazoles, have also been modified, but these will not be described. Hydroxyls and carboxylic acids are commonly activated to produce more reactive agents for ligand acylation. Elimination is a possible competing reaction for secondary and aryl alcohols. Primary amines, present on many biologic ligands from N-amino acid termini and lysine, are good nucleophiles when unprotonated (pKa ~ 9); moderately basic pH (8–10) ensures their reactivity. Thiols, present on cysteine (pKa ~ 8.5), are stronger nucleophiles than amines, and as such they can be selectively coupled in the presence of amines at lower pH (5–7). Sulhydryl groups often exist as their disulfide form and may be reduced to free thiols using mild agents such as dithiothreitol. The following paragraphs review activation schemes for particular chemical groups present on the surface, giving brief coupling schemes. Due to the advantages of surface versus ligand activation, only general methods involving surface activation will be discussed.

Immobilization to Surface Alcohols A hydroxyl-bearing surface may be activated with numerous reagents to produce more reactive species for substitution, the most common example being the cyanogen bromide activation of cellulose and agarose derivatives [2–5, 11, 17]. Due to problems with high volatility of the cyanogen bromide, sensitivity of the activated species to hydrolysis, competing reactions during activation and coupling, and the desire to use culture substrates other than polysaccharides, other activation schemes have been developed for more defined activation and coupling. These schemes have been used to modify functionalized glasses and polymers for cell-mediated adhesion studies. Alcohols react with sulfonyl halides [18–25], carbonyldiimidazole [26, 27], succinimidyl chloroformate [28], epoxides [29, 30], isocyanates [31–33] and heterocyclic [34, 35] and alkyl halides [21]. Activation of surface alcohols, with these agents can be performed in organic solvents such as acetonitrile, methylene chloride, acetone, benzene, dioxane, diethyl ether, toluene, or DMF. Reactive Esters Alcohols can be activated to reactive sulfonic ester leaving groups by reaction with sulfonyl halides [18–25] (Fig. 110.2) such as p-toluenesulfonyl chloride (tosyl chloride), trifluoroethanesulfonyl chloride (tresyl chloride), methanesulfonyl chloride (mesyl chloride), or fluorobenzenesulfonyl chloride (fosyl chloride). The resulting sulfonic esters are readily displaced in mild aqueous conditions with amines or thiols to produce amino- or thioether-bound ligands. Aryl alcohols should not be activated in this manner, since the sulfonate group may irreversibly transfer to the aromatic nucleus [36]. Sulfonic esters differ in their ease of nucleophilic substitution and resistance to hydrolysis, tosyl esters being low in coupling potential [23] and fosyl esters being high in hydrolysis resistance [24]. Activation can be performed in many organic solvents that are properly dry: Trace water and other species such as ammonia will react. DMSO should not be used as a solvent, as it will react with sulfonyl halides. A tert-amino base such as pyridine, dimethylaminopyridine, triethylamine, diisopropylethyl amine, or ethylmorpholine can be added in equimolar amounts to serve as a nucleophilic catalyst and to combine with the liberated HCl. It has been suggested that hydroxyls be converted to alkoxides prior to sulfonic ester activation in the presence of ethers as they may be sensitive to the HCl generated during activation [25].

FIGURE 110.2 Alcohols can be activated with sulfonyl halides, which are readily displaced with amine- (shown) or thiol-containing (not shown) ligands.


An alcohol-containing surface is typically incubated with the sulfonyl halide for 0.5–6.0 hr. Any precipitated salts can be rinsed away with 1 mM HCl. The ligand is coupled to the surface for 12–24 hr in borate or carbonate buffer (pH 9–10) at a concentration of 1 mg/ml. Coupling can proceed at more mild pH (~7) with more concentrated solutions (>10 mg ligand/ml) or with thiol-bearing ligands. Excess sulfonic esters can be displaced with aqueous solutions (10–50 mM, pH = 8–9) of tris(hydroxymethyl)aminomethane, aminoethanol, glycine, or mercaptoethanol, or by hydrolysis. Other Acylating Agents Surface hydroxyls may be activated to groups other than sulfonic esters. Alcohols react readily with carbonyldiimidazole [26, 27] (Fig. 110.3) and succinimidyl chloroformate [28] (Fig. 110.4) to produce reactive imidazole-N-carboxylates and succinimidyl carbonates, respectively. These species acylate amines to urethane linkages. Activation typically proceeds in organic solvent for 2–6 hr. Dimethylaminopyridine catalyzes formation of the succinimidyl carbonate. Amine-bearing ligands (>10 mg/ml) are coupled to the surface in borate or phosphate buffer (pH 8–9) for 12–24 hr, or 4°C, 2–3 days. Thiol-bearing groups may also be immobilized onto these activated groups; however the thiocarbamate linkage can be sensitive to hydrolysis and may not be generally applicable to preparing well-defined substrates for long-term biologic investigations. Bifunctional Bridges In lieu of alcohol conversion to activated reaction groups, the hydroxyl may be added to homo- or heterobifunctional bridges, wherein alcholysis consumes one terminus to produce newly functionalized surfaces (Figure 110.5). Epoxide bridges can be added to surface-bound alcohols in aqueous base (to reduce hydrolysis or polymerization of the epoxide, 10–100 mM NaOH) or in organic solvent, 12 hr. Isocyanate bridges are added to surface alcohols in organic solvents with organotin catalysts (such as dibutyltin dilaurate), 12 hr. Halo alkylation of surface-bound alcohols or alkoxides proceeds in organic solvents, 2–3 days, or with heat (~80°C), 12 hr. The unconsumed free group (isocyanate, epoxide, or alkyl halide) is substituted with amine- or thiol-bearing ligands in buffered aqueous conditions, pH 8–10, 1–10 mg/ml, 12–24 hr. Hydroxyl-bearing ligands, such as carbohydrates, can be coupled in aqueous DMSO, dioxane, or DMF, 12–24 hr, 60–80°C. Hydroxyl coupling may also be performed in aqueous base (10–100 mM NaOH) for ligands that are resistant to ester- or amide hydrolysis, such as mono- or polysaccharides.

FIGURE 110.3 Alcohols react readily with carbonyldiimidazole to produce reactive imidazole-N-carboxylates for coupling to amine-containing ligands.


FIGURE 110.4 Alcohols react readily with succinimidyl chloroformate to produce a succinimidyl carbonate for coupling to amine-containing ligands.

Heterocyclic aryl halides, such as cyanuric chloride [34, 35], react with free hydroxyl groups in polar solvents containing sodium carbonate, 40–50°C, 30 min. Amine-bearing ligands are immobilized to the surface in borate buffer, pH 9, 1–10 mg/ml, 5–10°C, 12–24 hr.

Immobilization to Surface Carboxylic Acids Acids may be converted to activated leaving groups by reaction with carbodiimides [36]. The generated O-acylureas react with amines to produce amide linkages and can react with alcohols to produce ester linkages with acid or base catalysts. An undesirable competing reaction is urea rearrangement to nonreactive N-acylurea; this effect may be accelerated in aprotic polar solvents such as DMF but is reduced at low temperatures (0–10°C) or by the addition of agents such as hydroxybenztriazole to convert the O-acylurea to benztriazole derivatives. Acids may also be activated with carbonyldiimidizole to produce easily amino- and alcoholyzed imidazolide intermediates; however, imidazolides are highly susceptible to hydrolysis and necessitate anhydrous conditions for activation and coupling [37]. Water-soluble carbodiimides such as ethyl(dimethylaminopropyl)-carbodiimide (EDC) can be used in either aqueous or organic media for the immobilization of ligands to produce bioactive substrates [20, 38, 39]. To reduce hydrolysis of the O-acylurea, conversion to more resistant succinimidyl esters can be performed using EDC and N-hydroxysuccinimide or its water-soluble sulfonate derivative, and base catalysts (Fig. 110.6). Alternatively, the acid may be converted directly to a succinimidyl ester via reaction with tetramethyl(succinimido)uronium tetrafluoroborate [40, 41]. The activation of surface-bound carboxylic acids with EDC in organic media (ethanol, DMF, dioxane) is complete with 1–2 hr at 0–10°C;

FIGURE 110.5 Bifunctional coupling agents may be used, e.g., to couple an amine-containing ligand to a hydroxylcontaining surface via a spacer.


FIGURE 110.6 ligands.

Carboxylated surfaces can be achieved to succimydil esters to subsequently couple amine-containing

in aqueous media at pH 4–6, 0–5°C, 0.5–2 hr. Amine-bearing ligands may be coupled onto the surface in buffered media, pH 7.5–9, 1 mg/ml, 2–24 hr. Since the reaction of carbodiimides with amines is slow compared to reaction with acids, activation and coupling may proceed simultaneously [42] by including amine-bearing ligands (1–10 mg/ml) with the EDC and allowing coupling to proceed in buffered media, pH 5–7, 0–10°C, 12–24 hr, or 25–30°C, 4–6 hr. Acid dehydration directly to succinimidyl esters using tetramethyl(succinimido)uronium tetrafluoroborate is performed in anhydrous conditions with equimolar tert-amino base, 2–4 hr, followed by coupling in aqueous organic cosolvents (water/DMF or water/dioxane), 1–10 mg/ml ligand, pH 8–9, 12–24 hr.

Immobilization to Surface Amines Primary and secondary amine-containing surfaces may be reacted with homo- or heterobifunctional bridges (Fig. 110.7). Amines are more nucleophilic than alcohols, they generally do not require the addition of catalysts, and their addition is faster. These bridges may contain isocyanates [31, 33], isothiocyanates [43], cyclic anhydrides [44], succinimidyl esters [45–47], or epoxides [48, 49]. Isocyanates add to amines with good efficiency but are susceptible to hydrolysis; epoxides and cyclic anhydrides are somewhat less reactive yet are still sensitive to hydrolysis. Hydrolysis-resistant diisothiocyanates have been used for many years to label ligands with reporter molecules; however, the thiourea linkage may be hydrolytically labile (especially at lower pH) and may be unsuitable for investigations of cell-surface interactions. Succinimidyl esters, although not as resistant to hydrolysis as isothiocyanates, have very good reactivity to amines and form stable amide linkages. Hydrolysis is all these reagents is accelerated at higher pH (≥9–10). Coupling to surface-bound amines is performed in organic conditions (DMF, DMSO, acetone) for 1–3 hr. Coupling of amine-bearing ligands onto immobilized bridges is performed in buffered media, pH 8–10.5, 2–12, 1–10 mg/ml. Excess reagent can be displaced with buffered solutions of tris(hydroxymethyl)aminomethane, aminoethanol, glycine, or mercaptoethanol, or hydrolysis. Bifunctional aldehydes, such as glutaraldehyde and formaldehyde, have been used classically as crosslinkers for purposes of immunohistochemistry and ultrastructural investigations. They have been used also to couple ligands onto amine-bearing substrates [50–52]. Hydrolysis of aldehydes is usually not a concern, since the hydrolysis product, alkyl hydrate, is reversible back to the carbonyl. Amines add to aldehydes to produce imine linkages over a wide range of pH (6–10). These Schiff bases are potentially hydrolytically labile; reductive amination can be performed with mild reducing agents such as sodium

FIGURE 110.7 bridges.

Primary and secondary amine-containing surfaces may be reacted with homo- or heterobifunctional


FIGURE 110.8 Thiol-containing surfaces may be coupled to malcimide activated species, either heterobifunctional linkers or ligands.

cyanoborohydride, pH 8–9, without substantial losses in ligand bioactivity [53]. Acetalization of polyhydric alcohols may commence in the presence of Lewis acid catalysts followed by dehydrating the hemiacetal linkage, an acetal [54, 55]. The dehydration conditions (air-drying followed by 70–90°C, 2 hr) may damage many biologic ligands. Alkyl halide-bearing surfaces can be coupled to amine-bearing ligands [49, 56, 57]; their reaction is slower, but they are resistant to hydrolysis. Ligands can be immobilized in buffered medium (pH 9–10), 12–24 hr, 1–10 mg/ml, or in organic medium. Heat may be used to increase yields (60–80°C).

Immobilization to Surface Thiols Thiols are more nucleophilic than amines or alcohols and may be selectively coupled in their presence at more neutral pH. Thiols can be reacted with reagents that are not reactive toward amines, such as homo- or heterobifunctional maleimide bridges [32, 58–60]. Surface-bound thiols react with maleimides in acetone, methanol, DMF, etc., for 1–2 hr, to produce a thioether linkage (Fig. 110.8). Thiol-bearing ligands can be immobilized to surface maleimides by incubating in mild buffered media, pH 5–7, 12–24 hr, 1–10 mg/ml. More basic pH should be avoided as hydrolysis of the maleimide to nonreactive maleamic acid may occur.

Photoimmobilization Surfaces can be modified using photoactivated heterobifunctional bridges in one of three schemes: (1) the bridge may be immobilized onto a functionalized surface, followed by photocoupling of the ligand onto the photoactivated group [58, 61]; (2) the reagent may be photocoupled onto the surface, followed by immobilization onto the free terminus [58, 62]; or (3) the ligand may be coupled to the reagent, followed by photoimmobilization to the surface [63, 64]. The photoactivatible group may be light-sensitive azides, benzophenones, diazinines, or acrylates; polymerization initiator-transfer agent-terminators [65] and plasma-deposited free radicals [64] have also been used. Photoactivation produces highly reactive radical intermediates that immobilize onto the surface via a nonspecific insertion of the radical into a carboncarbon bond. Photolabile reagents may be coupled to amine- or thiol-bearing ligands using succinimidyl esters or maleimides. One utility of photoimmobilization is the ability to produce patterns of ligands on the surface by lithography.

110.3 Surface Preparation Polymeric materials can be functionalized using plasma deposition or wet chemical methods to produce surfaces containing reactive groups for ligand bioconjugation. Plasma polymerization has been widely used to deposit numerous functional groups and alter the surface chemistry of many medically relevant polymers [66, 67]. Polyesters such as polyethyleneterephthalate can be partially saponified with aqueous base (5% NaOH, 100°C, 30 min) to produce surface-bound carboxylic acids [39], partially aminolyzed with diamines such as ethylenediamine (50% aqueous, 12 hr) to produce surface-bound amines [68] or


electrophilically substituted with formaldehyde (20% in 1 M acetic acid) to produce surface-bound alcohols [19]. Polyurethanes can be carboxylated via a bimolecular nucleophilic substitution [69]. Carbamate anions are prepared by abstraction of hydrogen from the urethane nitrogen at low temperatures to prevent chain cleavage (–5°C), followed by coupling of cyclic lactones such as β-propiolactone. Ligands are then immobilized onto the grafted carboxylic acids. Polytetrafluoroethylene can be functionalized with a number of reactive groups using wet chemical or photochemical methods. Fluoropolymers are reduced with concentrated benzion dianion solutions to form carbonaceous surface layers which can then be halogenated, hydroxylated, carboxylated, or aminated [70]. Photochemical modification of fluoropolymers is possible by incubating the polymer in solutions of alkoxides or thiolate anions and exposing to UV light [71]. The patterning of surfaces is possible using this technique in conjunction with photolithography. Glasses and oxidized polymers can be functionalized with bifunctional silanating reagents to generate surfaces containing alkyl halides, epoxides, amines, thiols, or carboxylic acids [19, 49, 59, 72]. The substrate must be thoroughly clean and free of any contaminating agents. Glass is soaked in strong acid or base for 30–60 min. Polymers are cleaned with plasma etching or with strong oxidizers such as chromic acid. Clean substrates are immersed in silane solutions (5–10% in acetone, toluene, or ethanol/water 95%/5%) for 1 hr and cured 50–100°C, 2–4 hr, or at 25°C, 24 hr, for oxidation-sensitive silanes such as mercaptosilanes. Prepared surfaces stored under Ar are stable for several weeks. Metals such as gold can be functionalized via the chemisorption of self-assembled monolayers of alkanethiols [73–75]. Gold substrates, prepared by evaporating gold on chromium-primed silicon substrates, are immersed in organic solutions of alkanethiol (1–10 mM) for 5–60 min. The monolayer is adsorbed via a gold-sulfur bond; competitive displacement of the alkanethiol may occur [75], and it is unclear if these surfaces are applicable for use in reducing media. Similar substrates can be prepared from the chemisorption of carboxylic acids onto alumina [76]. Prepared substrates stored under Ar are stable for several months. Amine-bearing surfaces have been produced by the adsorption of biologically inert proteins such as albumin. Cells do not have adhesion receptors for albumin, and for this reason albumin has been used to passivate surfaces against cell adhesion. Bioactive ligands can be amino-immobilized onto the albumincoated surfaces [45, 46, 48]. Functional polymers and copolymers containing alcohols, amines, alkyl halides, carboxylic acids, or other groups, can be synthesized, coated onto a surface, and used as the substrate for immobilization of bioactive ligands [77, 78]. Examples include poly(vinyl alcohol) [31, 33], chloromethyl polystyrene [56], aminopolystyrene [79, 80], poly(acrylic acid) [20, 38], polyallylamine [80], poly(maleic acid anhydride) [12], poly(carbodiimide) [81], and poly(succinimide) [82].

110.4 Ligand/Polymer Hybrids Hybrid copolymers may be synthesized in which one of the components is the biologically active ligand. These copolymers may then be coated onto a substrate or crosslinked into a three-dimensional network. Since the ligand is a component of the copolymer, no additional ligand immobilization may be necessary to produce bioactive substrates. Examples are available for particular cases of hybrid copolymers (54, 55, 82–87], including gammairradiated crosslinked poly(peptide) [86, 88], dialdehyde crosslinked poly(vinyl alcohol)-glycosaminoglycan [54], poly(amino acid-etherurethane) [47, 87], poly(amino acid-lactic acid) [89], poly(amino acid-carbonate) [90], poly(peptide-styrene) [84], and linear [83] or crosslinked [82] poly(glycoside-acrylamide).

110.5 Determining Ligand Surface Densities The concentration of ligands immobilized upon a surface can be measured using radiolabeling, photochrome labeling, surface analysis, or gravimetry. Since most materials can support the nonspecific © 2000 by CRC Press LLC

adsorption of bioactive ligands, especially proteins, controls must be utilized to differentiate between covalent immobilization and physicochemical adsorption occurring during coupling. For example, ligands immobilized onto unactivated versus activated substrates give relative differences between nonspecific adsorption and specific bioconjugation. The surface immobilization of radiolabeled ligands with markers such as 3H, 35S, or 125I, can be followed to give information on the kinetics of coupling, the coupling capacity of the substrate, and the surface density of ligands. Densities on the order of pmol-fmol/cm2 are detectable using radiolabeled molecules. The coupling capacity and density of immobilized acids, amines, and thiols can be evaluated using colorimetric procedures. Substrates can be incubated in solutions of Ellman’s reagent [91, 92]; the absorbances of the reaction products give the surface density. Antigens which have been immobilized can be exposed to photochrome-labeled antibodies and surface concentrations calculated using standard enzyme immunosorbent assays [49, 93]. Verification of ligand immobilization may be performed using a number of surface analysis techniques. Mass spectroscopy, x-ray photoelectron spectroscopy, and dynamic contact angle analysis can give information on the chemical composition of the substance’s outmost layers. Changes in composition are indicative of modification. Ellipsometry can be used to gauge the thickness of overlapping surface layers; increases imply the presence of additional layers. Highly sensitive gravimetric balances, such as quartz crystal microbalances, can detect in situ changes in the mass of immobilized ligand in the nanogram range [65].

110.6 Applications of Immobilized Ligands Extracellular matrix proteins such as fibronectin, laminin, vitronectin and collagen, or adhesion molecules such as ICAM-1, VCAM-1, PCAM-1, and sialyl Lewis X, interact with cell surface receptors and mediate cell adhesion. The tripeptide adhesion sequence Arg-Gly-Asp (RGD) is a ubiquitous signal present in many cell adhesion proteins. It interacts with the integrin family of cell surface adhesion receptors, and comprises the best studied ligand-receptor pair [94–96]. In lieu of immobilizing complex multifunctional proteins for purposes of cell adhesion studies, synthetic RGD sequences have instead been immobilized onto many substrates as simplified models to understand various molecular aspects of cell adhesion phenomena. The following paragraphs cite examples of RGD-grafted substrates that have been used in biomedicine and bioengineering. The density of RGD necessary to mediate cell adhesion has been determined in a number of fashions. RGD-containing peptides and protein fragments have been physicochemically adsorbed onto tissue culture substrates [97, 98] or covalently bound to albumin-coated substrates [45, 46] to titrate the dependency of cell adhesion function upon RGD surface densities. To remove potential complications due to desorption of ligands or albumin, RGD has been covalently bound onto functionalized substrates. Immobilization also restricts the number of conformations the peptide may assume, helping to ensure that all the peptide is accessible to the cells. RGD has been immobilized onto silanated glasses by its amino [19, 99] and carboxyl [15] termini. The effects of RGD density on cell adhesion, spreading, and cytoskeletal organization was examined [99] using this well-defined system. Other peptides have been immobilized in identical fashion to determine if they influence cell physiology [100]. RGD peptides have been immobilized onto highly cell-resistant materials to ensure that the peptide is the only cell adhesion signal responsible for cell adhesion to diminish signals borne of nonspecifically adsorbed serum proteins. Hydrogels of polyacrylamide [101], poly(vinyl alcohol) [31] and poly(ethylene glycol) [85] and nonhydrogel networks of polyacrylate/poly(ethylene glycol) [102] have been grafted with RGD; these background materials were highly resistant to the adhesion of cells even in the presence of serum proteins, demonstrating that the RGD sequence was solely responsible for mediating cell adhesion. RGD-containing peptides have been immobilized onto medically relevant polymers in an effort to enhance their biocompatibilities by containing an adhered layer of viable cells. RGD-grafted surfaces can be more efficient in supporting the number and strength of cell adhesion by the peptide facilitating cell adhesion additionally to adsorbed adhesion proteins from the biological milieu. RGD has been conjugated © 2000 by CRC Press LLC

onto surfaces by means of photoimmobilization [103] and plasma glow discharge [64]. In an effort to promote cell adhesion onto biodegradable implants, RGD peptides have been covalently grafted onto poly(amino acid-lactic acid) copolymers [89]. In this manner, cells adherent on the degradable material can eventually obtain a completely natural environment. Self-assembled monolayers of biologic ligands have been immobilized onto gold substrates by adsorbing functionalized thiol-containing bridges followed by covalent grafting of the ligand [93] or by adsorbing alkanethiol-containing ligands [104]. It has been suggested [105] that RGD-containing peptides could be immobilized onto gold substrates in these manners to engineer highly defined surfaces for cell culture systems. These examples only partially illustrate the utility of ligand-grafted substrates for bioengineering and biomedicine. These substrates offer simplified models of basement membranes to elucidate mechanisms and requirements of cell adhesion. They have applications in biomedicine as biocompatible, cell-adhesive biomaterials for tissue engineering or for clinical implantation.

References 1. Brinkley M. 1992. A brief survey of methods for preparing protein conjugate with dyes, haptens, and cross-linking reagents. Bioconjugate Chem 3:2. 2. Means GE, Feeney RE. 1990. Chemical modification of proteins: History and applications. Bioconjugate Chem 1:2. 3. Wong SS. 1991. Chemistry of Protein Conjugation and Cross-Linking, Boca Raton, Fla, CRC Press. 4. Pharmacia Inc. 1988. Affinity chromatography principles and methods, Tech Bull, Uppsala, Sweden. 5. Trevan MD. 1980. Immobilized Enzymes: An Introduction and Applications in Biotechnology, New York, Wiley. 6. Wimalasena RL, Wilson GS. 1991. Factors affecting the specific activity of immobilized antibodies and their biologically active fragments. J Chromatogr 572:85. 7. Mason RS, Little MC. 1988. Strategy for the immobilization of monoclonal antibodies on solidphase supports. J Chromatogr 458:67. 8. Wingard LB Jr, Katchalski-Katzir E, Goldstein L. 1976. Applied Biochemistry and Bioengineering: Immobilized Enzyme Principles, vol 1, New York, Academic. 9. Smalla K, Turkova J, Coupek J, et al. 1988. Influence on the covalent immobilization of proteins to modified copolymers of 2-hydroxyethyl methacrylate with ethylene dimethacrylate. Biotech Appl Biochem 10:21. 10. Schneider C, Newmanm RA, Sutherland DR, et al. 1982. A one-step purification of membrane proteins using a high efficiency immunomatrix. J Biol Chem 257:10766. 11. Scouten W. 1987. A survey of enzyme coupling techniques. Methods Enzymol 135:30. 12. Maeda H, Seymour LW, Miyamoto Y. Conjugates of anticancer agents and polymers: advantages of macromolecular therapeutics in vivo. Bioconjugate Chem 3:351. 13. Nojiri C, Okano T, Park KD, et al. 1988. Suppression mechanisms for thrombus formation on heparin-immobilized segmented polyurethane-ureas. Trans ASAIO, 34:386. 14. Fassina G. 1992. Oriented immobilization of peptide ligands on solid supports. J Chromatogr 591:99. 15. Hubbell JA, Massia SP, Drumheller PD. 1992. Surface-grafted cell-binding peptides in tissue engineering of the vascular graft. Ann NY Acad Sci 665:253. 16. Beer JH, Coller BS. 1989. Immobilized Arg-Gly-Asp (RGD) peptides of varying lengths as structural probes of the platelet glycoprotein IIb/IIIa receptor. Blood 79:117. 17. Axen R, Porath J, Ernback S. Chemical coupling of peptides and proteins to polysaccharides by means of cyanogen halides. Nature 214:1302. 18. Delgado C, Patel JN, Francis GE, et al. 1990. Coupling of poly(ethylene glycol) to albumin under very mild conditions by activation with tresyl chloride: Characterization of the conjugate by partitioning in aqueous two-phase systems. Biotech Appl Biochem 12:119.


19. Massia SP, Hubbell JA. 1991. Human endothelial cell interactions with surface-coupled adhesion peptides on a nonadhesive glass substrate and two polymeric biomaterials. J Biomed Mater Res 25:223. 20. Nakajima K, Hirano Y, Iida T, et al. 1990. Adsorption of plasma proteins on Agr-Gly-Asp-Ser peptide-immobilized poly(vinyl alcohol) and ethylene-acrylic acid copolymer films. Polym J 22:985. 21. Testoff MA, Rudolph AS. 1992. Modification of dry 1,2-dipalmitoylphosatidylcholine phase behavior with synthetic membrane-bound stabilizing carbohydrates. Bioconjugate Chem 3:203. 22. Fontanel M-L, Bazin H, Teoule R. 1993. End attachment of phenololigonucleotide conjugates to diazotized cellulose. Bioconjugate Chem 4:380. 23. Nilsson K, Mosbach K. 1984. Immobilization of ligands with organic sulfonyl chlorides. Methods Enzymol 104:56. 24. Chang Y-A, Gee A, Smith A, et al. Activating hydroxyl groups of polymeric carriers using 4-fluorobenzenesulfonyl chloride. Bioconjugate Chem 3:200. 25. Harris JM, Struck EC, Case MG. 1984. Synthesis and characterization of poly(ethylene glycol) derivatives. J Polym Sci Polym Chem Edn 22:341. 26. Sawhney AS, Hubbell JA. 1992. Poly(ethylene oxide)-graft-poly(L-lysine) copolymers to enhance the biocompatibility of poly(L-lysine)-alginate microcapsule membranes. Biomaterials 13:863. 27. Hearn MTW. 1987. 1,1′-Carbonyldiimidazole-mediated immobilization of enzymes and affinity ligands. Methods Enzymol 135:102. 28. Miron T, Wilchek M. 1993. A simplified method for the preparation of succinimidyl carbonate polyethylene glycol for coupling to proteins. Bioconjugate Chem 4:568. 29. Uy R, Wold F. 1977. 1,4-Butanediol diglycidyl ether coupling of carbohydrates to sepharose: affinity adsorbents for lectins and glycosidases. Anal Biochem 81:98. 30. Sundberg L, Porath J. 1974. Preparation of adsorbents for biospecific affinity chromatography: I. Attachment of group-containing ligands to insoluble polymers by means of bifunctional oxiranes. J Chromatogr 90:87. 31. Kondoh A, Makino K, Matsuda T. 1993. Two-dimensional artificial extracellular matrix: bioadhesive peptide-immobilized surface design. J Appl Polym Sci 47:1983. 32. Annunziato ME, Patel US, Ranade M, et al. 1993. P-Maleimidophenyl isocyanate: A novel heterobifunctional linker for hydroxyl to thiol coupling. Bioconjugate Chem 4:212. 33. Kobayashi H, Ikada Y. 1991. Covalent immobilization of proteins onto the surface of poly(vinyl alcohol) hydrogel. Biomaterials 12:747. 34. Shafer SG, Harris JM. 1986. Preparation of cyanuric-chloride activated poly(ethylene glycol). J Polym Sci Polym Chem Edn 24:375. 35. Kay G, Cook EM. 1967. Coupling of enzymes to cellulose using chloro-s-triazine. Nature (London) 216:514. 36. Bodanszky M. 1988. Peptide Chemistry, Berlin, Springer-Verlag. 37. Staab HA. 1962. Syntheses using heterocyclic amides (azolides). Agnew Chem Internat Edn 7:351. 38. Hirano Y, Okuno M, Hayashi T, et al. 1993. Cell-attachment activities of surface immobilized oligopeptides RGD, RGDS, RGDT, and YIGSR toward five cell lines. J Biomater Sci Polym Edn 4:235. 39. Ozaki CK, Phaneuf MD, Hong SL, et al. 1993. Glycoconjugate mediated endothelial cell adhesion to Dacron polyester film. J Vasc Surg 18:486. 40. Barnwarth W, Schmidt D, Stallard RL, et al. 1988. Bathophenanthroline-ruthenium(II) complexes as non-radioactive labels for oligonucleotides which can be measured by time-resolved fluorescence techniques. Helv Chim Acta 71:2085. 41. Drumheller PD, Elbert DL, Hubbell JA. 1994. Multifunctional poly(ethylene glycol) semi-interpenetrating polymer networks as highly selective adhesive substrates for bioadhesive peptide grafting. Biotech Bioeng 43:772.


42. Liu SQ, Ito Y, Imanishi Y. 1993. Cell growth on immobilized cell growth factor: 9. Covalent immobilization of insulin, transferrin, and collagen to enhance growth of bovine endothelial cells. J Biomed Mater Res 27:909. 43. Wachter E, Machleidt W, Hofner H, Otto J. 1973. Aminopropyl glass and its p-phenylene diisothiocyanate derivative, a new support in solid-phase Edman degradation of peptides and proteins. FEBS Lett 35:97. 44. Maisano F, Gozzini L, de Haen C. 1992. Coupling of DTPA to proteins: A critical analysis of the cyclic dianhydride method in the case of insulin modification. Bioconjugate Chem 3:212. 45. Streeter HB, Rees DA. 1987. Fibroblast adhesion to RGDS shows novel features compared with fibronectin. J Cell Biol 105:507. 46. Singer II, Kawka DW, Scott S, et al. 1987. The fibronectin cell attachment Arg-Gly-Asp-Ser promotes focal contact formation during early fibroblast attachment and spreading. J Cell Biol 104:573. 47. Nathan A, Bolikal D, Vyavahare N, et al. 1992. Hydrogels based on water-soluble poly(ether urethanes) derived from L-lysine and poly(ethylene glycol). Macromolecules 25:4476. 48. Elling L, Kula M-R. 1991. Immunoaffinity partitioning: Synthesis and use of polyethylene glycoloxirane for coupling to bovine serum albumin and monoclonal antibodies. Biotech Appl Biochem 13:354. 49. Pope NM, Kulcinski DL, Hardwick A, et al. 1993. New applications of silane coupling agents for covalently binding antibodies to glass and cellulose solid supports. Bioconjugate Chem 4:166. 50. Werb Z, Tremble PM, Behrendtsen O, et al. 1989. Signal transduction through the fibronectin receptor induces collagenase and stromelysin gene expression. J Cell Biol 109:877. 51. Yamagata M, Suzuki S, Akiyama SK, et al. 1989. Regulation of cell-substrate adhesion by proteoglycans immobilized on extracellular substrates. J Biol Chem 264:8012. 52. Robinson PJ, Dunnill P, Lilly MD. 1971. Porous glass as a solid support for immobilization or affinity chromatography of enzymes. Biochim Biophys Acta 242:659. 53. Harris JM, Dust JM, McGill RA, et al. 1991. New polyethylene glycols for biomedical applications. ACS Symp Ser 467:418. 54. Cholakis CH, Zingg W, Sefton MV. 1989. Effect of heparin-PVA hydrogel on platelets in a chronic arterio-venous shunt. J Biomed Mater Res 23:417. 55. Cholakis CH, Sefton MV. 1984. Chemical characterization of an immobilized heparin: heparinPVA. In SW Shalaby, AS Hoffman, BD Ratner, et al. (eds), Polymers as Biomaterials, New York, Plenum. 56. Gutsche AT, Parsons-Wingerter P, Chand D, et al. 1994. N-Acetylglucosamine and adenosine derivatized surfaces for cell culture: 3T3 fibroblast and chicken hepatocyte response. Biotech Bioeng 43:801. 57. Jagendorf AT, Patchornik A, Sela M. 1963. Use of antibody bound to modified cellulose as an immunospecific adsorbent of antigen. Biochim Biophys Acta 78:516. 58. Collioud A, Clemence J-F, Sanger M, et al. 1993. Oriented and covalent immobilization of target molecules to solid supports: Synthesis and application of a light-activatible and thiol-reactive crosslinking reagent. Bioconjugate Chem 4:528. 59. Bhatia SK, Shriver-Lake LC, Prior KJ, et al. Use of thiol-terminal silanes and heterobifunctional crosslinkers for immobilization of antibodies on silica surfaces. Anal Biochem 178:408. 60. Moeschler HJ, Vaughan M. 1983. Affinity chromatography of brain cyclic nucleotide phosphodiesterase using 3-(2-pyridyldithio)proprionyl-substituted calmodulin linked to thiol-Sepharose. Biochemistry 22:826. 61. Tseng Y-C, Park K. 1992. Synthesis of photoreactive poly(ethylene glycol) and its application to the prevention of surface-induced platelet activation. J Biomed Mater Res 26:373. 62. Yan M, Cai SX, Wybourne MN, et al. 1993. Photochemical functionalization of polymer surfaces and the production of biomolecule-carrying micrometer-scale structures by deep-UV lithography using 4-substituted perfluorophenyl azides. J Am Chem Soc 115:814. 63. Guire PE. 1993. Biocompatible device with covalently bonded biocompatible agent, U.S. Patent 5,263,992.


64. Ito Y, Suzuki K, Imanishi Y. 1994. Surface biolization by grafting polymerizable bioactive chemicals. ACS Symp Ser 540:66. 65. Nakayama Y, Matsuda T, Irie M. 1993. A novel surface photo-graft polymerization method for fabricated devices. ASAIO J 39:M542. 66. Ratner BD, Chilkoti A, Lopez GP. 1990. Plasma deposition and treatment for biomaterial applications. In R d’Agostino (ed), Plasma Deposition, Treatment, and Etching of Polymers, p 463, New York, Academic Press. 67. Ratner BD. 1992. Plasma deposition for biomedical applications: a brief review. J Biomater Sci Polym Edn 4:3. 68. Desai NP, Hubbell JA. 1991. Biological responses to polyethylene oxide modified polyethylene terephthalate surfaces. J Biomed Mater Res 25:829. 69. Lin H-B, Zhao Z-C, Garcia-Echeverria C, et al. 1992. Synthesis of a novel polyurethane co-polymer containing covalently attached RGD peptide. J Biomater Sci Polymer Edn 3:217. 70. Costello CA, McCarthy TJ. 1987. Surface-selective introduction of specific functionalities onto poly(tetrafluoroethylene). Macromolecules 20:2819. 71. Allmer K, Feiring AE. 1991. Photochemical modification of a fluoropolymer surface. Macromolecules 24:5487. 72. Ferguson GS, Chaudhury MK, Biebuyck HA, et al. 1993. Monolayers on disordered substrates: self-assembly of alkyltrichlorosilanes on surface-modified polyethylene and poly(dimethylsiloxane). Macromolecules 26:5870. 73. Plant AL. 1993. Self-assembled phospholipid/alkanethiol biomimetic bilayers on gold. Langmuir 9:2764. 74. Prime KL, Whitesides GM. 1993. Adsorption of proteins onto surfaces containing end-attached oligo(ethylene oxide): A model system using self-assembled monolayers. J Am Chem Soc 115:10714. 75. Biebuyck HA, Whitesides GM. 1993. Interchange between monolayers on gold formed from unsymmetrical disulfides and solutions of thiols: evidence for sulfur-sulfur bond cleavage by gold metal. Langmuir 9:1766. 76. Laibinis PE, Hickman JJ, Wrightson MS, et al. 1989. Orthogonal self-assembled monolayers: Alkanethiols on gold and alkane carboxylic acids on alumina. Science 245:845. 77. Veronese FM, Visco C, Massarotto S, et al. 1987. New acrylic polymers for surface modification of enzymes of therapeutic interest and for enzyme immobilization. Ann NY Acad Sci 501:444. 78. Scouten WH. 1987. A survey of enzyme coupling techniques. Methods Enzymol 135:30. 79. Mech C, Jeschkeit H, Schellenberger A. 1976. Investigation of the covalent bond structure of peptide-matrix systems by Edman degradation of support-fixed peptides. Eur J Biochem 66:133. 80. Iio K, Minoura N, Aiba S, et al. 1994. Cell growth on poly(vinyl alcohol) hydrogel membranes containing biguanido groups. J Biomed Mater Res 28:459. 81. Weinshenker NM, Shen C-M. 1972. Polymeric reagents: I. Synthesis of an insoluble polymeric carbodiimide. Tetrahedron Lett 32:3281. 82. Schnaar RL, Brandley BK, Needham LK, et al. 1989. Adhesion of eukaryotic cells to immobilized carbohydrates. Methods Enzymol 179:542. 83. Sparks MA, Williams KW, Whitesides GM. 1993. Neuraminidase-resistant hemagglutination inhibitors: acrylamide copolymers containing a C-glycoside of N-acetylneuramic acid. J Med Chem 36:778. 84. Ozeki E, Matsuda T. 1990. Development of an artificial extracellular matrix. Solution castable polymers with cell recognizable peptidyl side chains. ASAIO Trans 36:M294. 85. Drumheller PD. 1994. Polymer Networks of Poly(Ethylene Glycol) as Biospecific Cell Adhesive Substrates, PhD dissertation, University of Texas at Austin. 86. Nicol A, Gowda DC, Parker TM, et al. 1993. Elastomeric polytetrapeptide matrices: Hydrophobicity dependence of cell attachment from adhesive (GGIP)n to nonadhesive (GGAP)n even in serum. J Biomed Mater Res 27:801. 87. Kohn J, Gean KF, Nathan A, et al. 1993. New drug conjugates: attachment of small molecules to poly(PEG-Lys). Polym Mater Sci Eng 69:515.


88. Nicol A, Gowda DC, Urry DW. 1992. Cell adhesion and growth on synthetic elastomeric matrices containing ARG-GLY-ASP-SER. J Biomed Mater Res 26:393. 89. Barrera DA, Zylstra E, Lansbury PT, et al. 1993. Synthesis and RGD peptide modification of a new biodegradable copolymer: Poly(lactic acid-colysine). J Am Chem Soc 115:11010. 90. Pulapura S, Kohn J. 1993. Tyrosine-derived polycarbonate: Backbone-modified “pseudo”poly(amino acids) designed for biomedical applications. Biopolymers 32:411. 91. Ngo TT. Coupling capacity of solid-phase carboxyl groups. Determination by a colorimetric procedure. Appl Biochem Biotech 13:207. 92. Ngo TT. 1986. Colorimetric determination of reactive amino groups of a solid support using Traut’s and Ellman’s reagents. Appl Biochem Biotech 13:213. 93. Duan C, Meyerhoff ME. 1994. Separation-free sandwich enzyme immunoassays using microporous gold electrodes and self-assembled monolayer/immobilized capture antibodies. Anal Chem 66:1369. 94. Albeda SM, Buck CA. 1990. Integrins and other cell adhesion molecules. FASEB J 4:2868. 95. Ruoslahti E. 1991. Integrins. J Clin Invest 87:1. 96. Humphries MJ. 1990. The molecular basis and specificity of integrin-ligand interactions. J Cell Sci 97:585. 97. Underwood PA, Bennett FA. 1989. A comparison of the biological activities of the cell-adhesive proteins vitronectin and fibronectin. J Cell Sci 93:641. 98. Yamada KM, Kennedy DW. 1985. Amino acid sequence specificities of an adhesive recognition signal. J Cell Biochem 28:99. 99. Massia SP, Hubbell JA. 1991. An RGD spacing of 44 nm is sufficient for integrin α vβ3-mediated fibroblast spreading and 140 nm for focal contact and stress fiber formation. J Cell Biol 114:1089. 100. Hubbell JA, Massia SP, Desai NP, et al. 1992. Endothelial cell-selective materials for tissue engineering in the vascular graft via a new receptor. Bio/Technology 9:568. 101. Brandley BK, Schnaar RL. 1989. Tumor cell haptotaxis on covalently immobilized linear and exponential gradients of a cell adhesion peptide. Dev Biol 135:74. 102. Drumheller PD, Elbert DL, Hubbell JA. 1994. Multifunctional poly(ethylene glycol) semi-interpenetrating polymer networks as highly selective adhesive substrates for bioadhesive peptide grafting. Biotech Bioeng 43:772. 103. Clapper DL, Daws KM, Guire PE. 1994. Photoimmobilized ECM peptides promote cell attachment and growth on biomaterials. Trans Soc Biomater 17:345. 104. Spinke J, Liley M, Guder H-J, et al. 1993. Molecular recognition at self-assembled monolayers: The construction of multicomponent multilayers. Langmuir 9:1821. 105. Singhvi R, Kumar A, Lopez GP, et al. 1994. Engineering cell shape and function. Science 264:696.


Chinn, J.A., Slack, S. M. “Biomaterials: Protein- Surface Interactions.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

111 Biomaterials: ProteinSurface Interactions Joseph A. Chinn Sulzer Carbomedics

111.1 111.2 111.3

Steven M. Slack The University of Memphis

111.4

Introduction Fundamentals of Protein Adsorption Example Calculations and Applications of Protein Adsorption Summary, Conclusions, and Directions

111.1 Introduction A common assumption in biomaterials research is that cellular interactions with natural and artificial surfaces are mediated through adsorbed proteins. Such diverse processes as thrombosis and hemostasis, hard and soft tissue healing, infection, and inflammation are each affected by protein adsorption to surfaces in vivo. Many in vitro diagnostic analyses, chromatographic separation techniques, and genetic engineering processes also involve protein adsorption at solid-liquid interfaces. The adsorption of fibrinogen, a prevalent blood plasma protein, has been studied extensively because of its role in blood coagulation and thrombosis, as has the adsorption of albumin, because it is thought to inhibit the adhesion of blood platelets [Young et al., 1982]. The amount of protein adsorbed to a substrate is best measured directly using radiolabeled proteins, whereas the thickness of an adsorbed protein film can be calculated from ellipsometric measurements. Further, the importance of the state of an adsorbed protein in mediating cellular interactions is now becoming evident. Molecularly sensitive indirect measurement techniques, e.g., circular dichroism (CD), differential scanning calorimetry (DSC), enzyme-linked immunosorbent assay (ELISA), Fourier transform infrared spectroscopy/attenuated total reflectance (FTIR/ATR), radio-immunoassay (RIA), or total internal reflection fluorescence (TIRF), can be used to characterize the conformation and organization of adsorbed proteins. (The amount of protein adsorbed to a substrate is best measured directly using radiolabeled proteins; the thickness of an adsorbed protein film can be calculated from ellipsometric measurements.) Highly specific monoclonal (MAb) (against specific protein epitopes) [Shiba et al., 1991] and polyclonal (PAb) (against multiple epitopes) [Lindon et al., 1986] antibodies provide direct probes of adsorbed protein conformation and organization. Thus, cellular responses can be compared not only with the amounts of proteins adsorbed but also the organization of the proteins on the surface of the substrate. Whereas previous studies confirmed roles for adsorbed proteins in subsequent cell-surface interactions, much current research aims to better understand cell-protein-surface interactions on a molecular level. Recently, peptide sequences contained within the cell binding domains of adhesive proteins have been identified and characterized, synthesized, and demonstrated to bind cellular receptors known as integrins [Ruoslahti, 1991; Yamada, 1991]. Current and potential applications range from selective or enhanced in vitro cell culture to selective in vivo cellular responses such as endothelialization in the absence of inflammation, infection, or thrombosis [Hubbell et al., 1991].


111.2 Fundamentals of Protein Adsorption Detailed and comprehensive reviews of protein adsorption have been published [Andrade, 1985; Andrade & Hlady, 1986; Horbett, 1982; Norde & Lyklema, 1991]. A thorough understanding of key principles will prove helpful in critically evaluating reports in literature. Particularly important concepts are protein structure and heterogeneity, factors that dramatically affect the thermodynamics and kinetics of adsorption, reversibility of adsorption, and the dynamics of multi-component adsorption. A protein is a complex molecule consisting of amino acid copolymer (polypeptide) chains that interact with each other to give the molecule a three-dimensional structure. Importantly, each amino acid in the polymer contributes to the chemical and physical properties of the protein. A dramatic example of this is the oxygen-carrying protein, hemoglobin, which consists of four polypeptide chains denoted α2 β2 . A single amino acid substitution in the 146 amino acid β-chain results in the conversion of normal hemoglobin (HbA) to sickle-cell hemoglobin (HbS) and underlies the serious consequences of sicklecell disease [Stryer, 1995]. Protein structure and function are relevant to protein adsorption and have been described on four different scales or orders [Andrade & Hlady, 1986]. Primary structure refers to the sequence and number of amino acids in a copolymer chain. The 20 amino acids that are polymerized to make proteins are termed residues. Of these, 8 have non-polar side chains, 7 have neutral polarity side chains, and 5 have charged polar side chains [Stryer, 1995]. Secondary structure results from hydrogen bonding associated with the amide linkages in the polymer chain backbone to form structures such as the α-helix and β-pleated sheet. Tertiary structure results from associations within chains, including hydrogen bonding, ionic and hydrophobic interactions, salt bridges, and disulfide bonds and dictates the three-dimensional structure adopted by protein molecules. Quaternary structure results from associations between chains. Many blood proteins contain polar, non-polar, and charged residues. In polar media such as buffered saline or blood plasma, hydrophilic residues tend to self-associate (often at the outer, water-contacting surface of the protein), as do hydrophobic residues (often “inside” the protein). This results in distinct domains (Fig. 111.1) that dictate higher order protein structure.

FIGURE 111.1 Schematic view of protein interacting with a well characterized surface. The protein has a number of surface domains with hydrophobic, charged, and polar character. The solid surface has a similar domain like character (Andrade & Hlady, 1986.)


When a single, static protein solution contacts a surface, the rate of adsorption depends upon transport of the protein from the bulk to the surface. Andrade and Hlady [1986] identified four primary transport mechanisms, namely diffusion, thermal convection, flow convection, and coupled convection-diffusion. In isothermal, parallel laminar flow or static systems, protein transport to the interface occurs exclusively by diffusion. In turbulent or stirred systems, each of the four transport modes can be significant. When adsorption is reaction limited, the net rate of adsorption can sometimes be described by the classic Langmuir theory of gas adsorption [Smith, 1981]:

(

)

rA = k A ∗ Cb ∗ 1 − Θ − kD ∗ Θ

(111.1)

where rA is the net adsorption rate, kA is the adsorption rate constant, Cb is the bulk concentration of the protein in solution, Θ is fractional surface coverage, and kD is the desorption rate constant. At equilibrium, the net rate of adsorption, rA is zero, and Θ can be calculated from Eq. (111.1) as the Langmuir adsorption isotherm,

Θ=

K ∗ Cb 1 + K ∗ Cb

(111.2)

where K = kA/kD . This model assumes reversible monolayer adsorption, no conformational changes upon adsorption, and no interactions between adsorbed molecules. It is most applicable to dilute solutions and non-hydrophobic substrates. When adsorption is diffusion limited, the initial rate of adsorption is equivalent to the rate of diffusion, described mathematically as [Andrade & Hlady, 1986]

rA = rD = 2 ∗ C D ∗

D πt

(111.3)

where D is the diffusivity of the protein, and t is time. Protein adsorption to hydrophobic substrates differs from that to hydrophilic materials. The primary driving force for adsorption to hydrophilic substrates is often enthalpic, whereas that to hydrophobic substrates is entropic [Norde, 1986]. Water near a hydrophobic surface tends to hydrogen bond to neighboring water molecules, resulting in a highly ordered water structure [Andrade & Hlady, 1986]. Disruption of this structure (dehydration) by adsorption of a protein to the surface increases the entropy of the system and is therefore thermodynamically favored. As a result, adsorption to hydrophilic substrates is generally reversible, whereas that to hydrophobic substrates is not. Denaturation of the adsorbed protein by hydrophobic interactions with the substrate can also contribute to irreversible adsorption [Feng & Andrade, 1994]. The amount of a specific protein adsorbed to a substrate can be measured directly if the protein is radiolabeled. Gamma emitting isotopes are preferred because their signal is directly proportional to the amount of protein present. Radioisotopes of iodine (125I, 129I, and 131I) are commonly used because iodine readily attaches to tyrosine residues [Macfarlane, 1958]. An 125I monochloride radiolabeling technique has been published [Horbett, 1986] but others, such as those using chloramine-T or lactoperoxidase, can also be used. If neither the 125I-protein nor the unlabeled protein preferentially adsorbs to the substrate at the expense of the other, then the amount of that protein adsorbed to a substrate from multi-component media such as plasma can be measured by adding a small amount of the 125I-protein to the adsorption medium. Studies have shown that 125I-fibrinogen generally behaves like its unlabeled analog [Horbett, 1981]. The total amount of the protein adsorbed (both unlabeled and 125I-protein) is calculated by dividing the measured radioactivity by the specific activity of the protein in the medium. (Specific activity is determined by


dividing the gamma activity in a measured aliquot of the adsorption medium by the total amount of that protein, both labeled and unlabeled, in the aliquot.) To verify that neither labeled nor unlabeled fibrinogen preferentially adsorbs to the substrates, the specific activity of the protein in a small aliquot of the plasma dilution from which adsorption was maximum should be increased 10x, and adsorption from that dilution again measured. Changes in calculated adsorption values should be attributable only to the variability in the data and differences in the signal to noise ratio, not the ratio of labeled to unlabeled fibrinogen in the plasma. Similarly, to verify that absorption into the sample materials of free 125I in the buffer is not significant, adsorption from dilute plasma to which only 0.01M unlabeled free iodide is added should be measured. Sample calculations for measuring protein adsorption using this technique are illustrated in the next section. The use of radioisotopes for the measurement of protein adsorption to surfaces offers a significant, though underused, advantage compared with other techniques. Because the radioisotopes, e.g., 125I and 131I, emit unique energy spectra, i.e., their peak radiation emissions occur at distinct energies, one can simultaneously measure the adsorption of two different proteins from a protein mixture. By labeling one protein with 125I and the other with 131I, the adsorption behavior of both can be determined in one experiment. Modern gamma counters come with preset energy windows specific to common isotopes, making such measurements routine. In the absence of such an instrument, one can still perform duallabeling experiments by exploiting the fact that the half-lives of the isotopes differ. For example, the halflife of 125I is sixty days whereas that of 131I is only eight days. After measuring the 131I emission immediately after the experiment, one can allow the radioactivity associated with it to decay to background levels (ten half-lives or 80 days for 131I) and then measure the signal associated solely with 125I [Dewanjee, 1992]. Indirect methods are also used to study proteins adsorbed to a substrate. ELISA and RIA analytical techniques exploit specific antibody-antigen interactions as follows. Antibodies against specific epitopes of an adsorbed protein are either conjugated to an enzyme (ELISA), or radiolabeled (RIA). Substrates are first incubated with the medium, then with a solution containing the antibody or antibody conjugate. In the case of ELISA, the substrates are subsequently incubated with substrate solution. As the substrate is converted to product, the color of the solution changes in proportion to the amount of bound antibody present and is measured spectrophotometrically. However, extensive calibration is required to quantify results. In the case of RIA, the radioactivity originating from the radiolabeled antibody (bound to the protein adsorbed to the substrate) is measured and the amount of antibody bound calculated. With both methods, the relative amount of the adsorbed protein to which the antibody has bound, rather than the total amount of protein adsorbed, is measured. Antibody binding is a function of not only the amount of protein adsorbed, but also the particular antibody used, total protein surface loading, and protein residence time. Thus, although antibody techniques provide direct probes of adsorbed protein conformation and organization, such measurements do not necessarily reflect the absolute amounts of protein adsorbed. Other indirect methods used to study adsorbed proteins include ellipsometry, electron microscopy, high performance liquid chromatography (HPLC), and staining techniques.

111.3 Example Calculations and Applications of Protein Adsorption The following example illustrates how radiolabeled proteins are used to measure the amount of protein adsorbed to a substrate. The fibrinogen concentration in 10 ml plasma is determined to be 5.00 mg/ml by measuring the light absorbance at 280 nm of a redissolved, thrombin-induced clot [Ratnoff & Menzie, 1950]. The concentration and specific activity of a 10 µl aliquot of 125I-fibrinogen are 1.00 mg/ml and 109 cpm/µg, respectively. Fibrinogen adsorption from dilute plasma to a series of polymer samples, each having 1.00 cm2 total surface area (counting both sides of the sample) is to be measured. Based upon reports in literature, maximal adsorption of 250 ng/cm2 is expected. The background signal in the gamma radiation counter is 25 cpm. To achieve a maximum signal/noise ratio of 10, the specific activity of fibrinogen in the plasma should be,


sp. ac . =

(signal noise) ∗ noise = mass adsorbed

10 ∗ 25 cpm 25 ng cm2 ∗1.00 cm

∗ 2

103 ng = 103 cpm µg µg

(111.4)

The volume of 125I-fibrinogen solution to be added to the plasma to obtain 103 cpm/g specific activity (neglecting the mass of 125I-fibrinogen added) is calculated as,

103 cpm µg ∗103 µg mg ∗ 5.00 mg ml ∗10 ml 109 cpm mg ∗1.00 mg ml ∗ ml 103 µl

= 50 µl

(111.5)

Addition of 50 µl 125I-fibrinogen solution should increase the total fibrinogen concentration in the plasma by only a small fraction,

50 µl ∗ ml 103 µl ∗1.00 mg ml = 10−3 10 ml ∗ 5.00 mg ml

(111.6)

To determine the amount of protein adsorbed to the substrate, the radioactivity of samples incubated with plasma is measured and compared with the specific activity of the protein in the plasma. In this example, if the amount of radioactivity retained by a sample is measured to be 137 cpm, then the mass of fibrinogen adsorbed is calculated as,

(137 cpm sample − 25 cpm background) ∗10

3

ng µg

10 cpm µg fibrinogen ∗1.00 cm sample 3

2

= 112 ng cm2

(111.7)

Adsorption of proteins to polymeric substrates is measured because adsorbed proteins influence cellular processes. Adsorbed albumin is proposed to favor biocompatibility, whereas adsorbed fibrinogen is proposed to discourage biocompatibility because of its role in mediating initial adhesion of blood platelets [Young et al., 1982]. This simplified view inadequately describes biocompatibility in vivo for several reasons. First, the relationships between processes involved in thrombosis, hemostasis, inflammation, and healing (e.g., adhesion of platelets, fibroblasts, white blood cells, endothelial cells) and long term biocompatibility remain mostly unknown. For example, Sakariassen and colleagues [1979] proposed that exclusively adsorbed von Willebrand factor (vWF) mediates platelet adhesion to vascular subendothelial structures, yet although some people that lack serum vWF in their blood exhibit bleeding disorders, others remain asymptomatic. Second, biological processes that do not require fibrinogen mediated cell adhesion (e.g., contact activation [Kaplan, 1978], complement activation [Chenowith, 1988]) are also related to material biocompatibility. Third, biological factors and serum proteins other than fibrinogen and albumin (e.g., vWF, fibronectin, vitronectin, laminin) significantly affect cellular processes. Fourth, the reactivity of adsorbed proteins depends upon their organization upon the substrate. Studies of fibrinogen adsorption in vitro illustrate the dynamic nature of protein adsorption from plasma. Vroman and Adams [1969] reported that oxidized silicon and anodized tantalum incubated 2s with plasma bind fibrinogen anti-serum, whereas the same materials incubated 25s with plasma do not. Ellipsometric measurements indicated that the observed decrease in antibody binding was not due to loss of protein. Brash and ten Hove [1984] and Horbett [1984] reported that maximal equilibrium fibrinogen adsorption to different materials occurred from intermediate dilutions of plasma. The adsorption maximum is sometimes called a Vroman peak, which describes the shape of the adsorption versus log (plasma concentration) curve referred to as an adsorption isotherm. Both the location and the magnitude of the peak depend upon the surface chemistry of the substrate. (For this reason, it is wise to fully characterize any substrate prior to measuring protein adsorption. Electron spectroscopy for


chemical analysis, ESCA [Ratner & McElroy, 1986], and secondary ion mass spectroscopy, SIMS [Andrade, 1985], are often appropriate.) Wojciechowski and colleagues [1986] reported that at short contact times, adsorption was greatest from undiluted plasma. As contact time increased, the plasma concentration at which adsorption was greatest decreased. The unusual observed adsorption behavior occurs because fibrinogen adsorption is driven initially by mass action, i.e., a gradient between surface and bulk concentration. However, as coverage of the surface increases, bulk proteins must compete for surface binding sites. The composition of the adsorbed layer continues to change as proteins of higher surface activity displace adsorbed proteins of lower surface activity [Horbett, 1984]. Vroman and colleagues [1980] called this process conversion of the fibrinogen layer, and proposed that fibrinogen adsorbed at early time is at later time displaced by other plasma proteins, possibly high molecular weight kininogen (HMWK). Because Vroman pioneered much of this work, these phenomena are collectively referred to as the Vroman effect [Slack & Horbett, 1995]. This principle is most applicable to hydrophilic substrates and is commonly used in biocompatibility studies to vary the composition of the protein layer adsorbed from plasma in a controlled manner. Slack and Horbett [1989] postulated the existence of two distinct types of adsorbed fibrinogen molecules; displaceable and non-displaceable. Protein adsorption from plasma was modeled as competitive adsorption from a binary solution of fibrinogen and a hypothetical protein H (representing all other plasma components). In this model, protein H adsorbs to unoccupied surface sites in a non-displaceable state, and fibrinogen molecules first adsorb to unoccupied surface sites in a displaceable state, then are displaced by protein H, or spread to become resistant to displacement. The latter process is referred to as fibrinogen transition. Neglecting desorption and mass transfer limitations, rate equations for surface coverage by protein H, displaceable fibrinogen, and non-displaceable fibrinogen were solved simultaneously for a surface initially free of adsorbate. The analytical solution is given by Eq. (111.8) and Eq. (111.9).

()

Θ1 t = β e r1 t − e r2 t   

() (

∗

)

∗

∗ ∗   Θ2 t = β ∗ k2 r1 ∗ r2 ∗ r1 1 − e r2 t  − r2 1 − e r1 t        

(111.8)

(111.9)

where Θ1, Θ2, and Θ3 are fractional coverages by displaceable fibrinogen, non-displaceable fibrinogen, and the hypothetical protein, respectively, k1 is the fibrinogen adsorption rate constant, k2 is the fibrinogen transition rate constant, k3 is the hypothetical protein adsorption rate constant, k4 is the fibrinogen displacement rate constant, CF is the bulk concentration of fibrinogen, CH is the bulk concentration of the hypothetical protein in solution, and β, r1, and r2 are constants related to rate constants and bulk concentrations. This model predicts maximal fibrinogen coverage at intermediate adsorption time (Fig. 111.2), consistent with reported experimental results. Surface exclusion and molecular mobility arguments have also been proposed to explain the Vroman effect [Willems et al., 1991]. Changes in protein conformation upon adsorption have been inferred from a variety of indirect measurements. For example, Castillo and colleagues [1984] inferred substrate, adsorption time, and residence time dependent conformational changes in human serum albumin (HSA) adsorbed to different hydrogels based upon changes in FTIR/ATR and CD spectra. Specific conformational changes, i.e., decreased α-helix and increased random coil and β-pleated sheet content upon adsorption were proposed. Similarly, Norde and colleagues [1986] reported lower α-helix content in HSA first adsorbed to, then desorbed from different substrates compared with native HSA. Castillo and colleagues [1985] also reported that lysozyme became increasingly denatured with increased adsorption time and residence time, and that denatured lysozyme adsorbed irreversibly to contact lens materials. De Baillou and colleagues [1984] proposed conformational changes in fibrinogen upon adsorption to glass based upon DSC measurements. Similarly, Feng and Andrade [1994] proposed that low temperature isotropic (LTI)


FIGURE 111.2 Time course of fibrinogen adsorption to a polymeric substrate as predicted by the model proposed by Slack and Horbett [1989].

carbon significantly denatures adsorbed proteins through hydrophobic interactions. Rainbow and colleagues [1987] proposed conformational changes in albumin upon adsorption to quartz, based upon changes in fluorescence lifetimes calculated from TIRF measurements. Finally, investigators noted that certain enzymes, following their adsorption to solid surfaces, lose a considerable amount of their enzymatic activity compared with enzymes in solution, suggesting that a surface-induced structural rearrangement had occurred [Sandwick & Schray, 1987; Sandwick & Schray, 1988]. Whereas indirect methods provide evidence for changes in protein organization upon adsorption, antibodies against specific protein epitopes provide direct evidence, as antibody-binding measurements directly reflect the availability of different protein epitopes. Proteins at interfaces can undergo both covalent (e.g., conversion of fibrinogen to fibrin) and non-covalent (e.g., denaturation, change in conformation, change in epitope accessibility) organizational changes. Therefore, epitopes inaccessible in solution might become available upon adsorption of the protein, and consequently, the biological activity of the adsorbed protein may differ from that of the same protein in solution. For example, receptorinduced binding site (RIBS) anti-fibrinogens bind to adsorbed but not free fibrinogen molecules [Zamarron et al., 1990]. Because an MAb binds to a single protein epitope, whereas a PAb binds to multiple epitopes on the protein, MAbs rather than PAbs should be more sensitive to such changes. Also, epitopes available at low surface loadings may become unavailable at higher loadings due to steric hindrance by proteins adsorbed to neighboring surface sites. Horbett and Lew [1994] used the Vroman effect principle to maximize the amount of fibrinogen adsorbed from plasma to different polymers, and then measured the binding of different MAbs directed against various epitopes of fibrinogen. Binding was reported to be substrate-dependent and, with some MAbs, changed with protein residence time. Thus, different fibrinogen epitopes become more or less available as the adsorbed molecule reorganizes upon the surface. Soria and colleagues [1985] demonstrated that a MAb directed against fragment D of the fibrinogen molecule, which did not bind to the molecule in solution, did bind to it following its adsorption to a surface. Although this method cannot distinguish between changes in protein conformation (higher order structure) and changes in surface orientation (e.g., rotation, steric effects), several authors reported that with increased adsorption time


or residence time, adsorbed proteins became less readily displaced by plasma or surfactant eluting agents [Balasubramanian et al., In Press; Bohnert & Horbett, 1986; Chinn et al., 1992; Rapoza & Horbett, 1990; Slack & Horbett, 1992]. These results suggest that post-adsorptive transitions in adsorbed proteins are primarily structural. Protein organization is also a function of surface loading. Chinn and colleagues [1992] and Rapoza and Horbett [1990] used the Vroman effect principle to vary both the amounts of fibrinogen and total protein adsorbed from plasma to different polymers. Fibrinogen retention by all substrates was greater when the protein was adsorbed from more dilute than from less dilute plasma. This suggests that at higher total protein loadings, each fibrinogen molecule directly contacts the surface at fewer points (because individual molecules compete for binding sites), and a greater fraction of molecules is displaceable. Conversely, at lower total protein loadings, individual molecules compete less for binding sites and are not hindered from reorganizing on the surface. Because each molecule directly contacts the surface at more points, a greater fraction of adsorbed molecules is non-displaceable. Pettit and colleagues [1994] reported a negative correlation, independent of substrate, between anti-fibronectin binding (normalized to the amount of fibronectin adsorbed), and the amount of fibronectin adsorbed. They proposed that the conformation or orientation of the adsorbed fibronectin molecule favors exposure of the cell binding domain at lower rather than higher fibronectin surface concentrations. Although the implications of protein transitions in long-term in vivo biocompatibility remain largely unknown, changes in the states of adsorbed proteins are related to cellular interactions. More platelets adhered to glass first contacted 5s with plasma than 3 min with plasma [Zucker & Vroman, 1969]. Platelet adhesion in vitro to polymers upon which the Vroman effect principle was used to vary the composition of adsorbed protein layer was reported related not to total fibrinogen binding, but to anti-fibrinogen binding [Lindon et al., 1986; Shiba et al., 1991], as well as the fraction of adsorbed protein that can be eluted by surfactant such as sodium dodecyl sulfate (SDS) [Chinn et al., 1991].

111.4 Summary, Conclusions, and Directions Clearly, adsorbed proteins affect biocompatibility in ways that are not entirely understood. Fibrinogen has been extensively studied because blood platelets involved in thrombosis and hemostasis have a receptor for this protein [Phillips et al., 1988]. Adsorbed fibrinogen is often proposed to discourage biocompatibility, but this view does not consider that adsorbed proteins exist in different states, depending upon adsorption conditions, residence time, and substrate. Evidence suggests that fibrinogen adsorbed from blood to substrates upon which the protein readily denatures, e.g., LTI carbon, may in fact promote biocompatibility. Because it is the organization and not the amount of an adsorbed protein that determines its biological activity, what happens to proteins after they adsorb to the substrate must be determined to properly evaluate the biocompatibility of a material. Further, a material might be made blood compatible if protein organization can be controlled such that the cell binding epitopes become unavailable for cell binding [Horbett et al., 1994]. Much current research aims to understand the relationship between the states of adsorbed proteins and cell-protein-surface interactions. Fundamental to better understanding of material biocompatibility is understanding the importance of protein and surface structure and heterogeneity in determining the organization of proteins at the solid-liquid interface, the dynamics of multi-component adsorption from complex media, e.g., the Vroman effect, and the significance of post-adsorptive events and subsequent cellular interactions. MAbs against specific protein epitopes provide direct evidence of changes in the states of adsorbed proteins, and indirect methods provide corroborative evidence. Molecular imaging techniques such as atomic (AFM), lateral force (LFM), and scanning tunneling (STM) microscopies have been adapted to aqueous systems to better define the states of adsorbed proteins with limited success [Sit et al., 1998], as have surface analysis techniques such as surface-matrix assisted laser desorption ionization (MALDI) [Kingshott et al., 1998].


Identification and characterization of the cell binding domains of adhesive proteins has led to better understanding of cell adhesion at the molecular level. Pierschbacher and colleagues[1981] used monoclonal antibodies and proteolytic fragments of fibronectin to identify the location of the cell attachment site of the molecule. Subsequently, residue sequences within the cell binding domains of other adhesive proteins were identified as summarized by Yamada [1991]. The Arg-Gly-Asp (RGD) amino acid sequence first isolated from fibronectin was later found present within vitronectin, osteopontin, collagens, thrombospondin, fibrinogen, and vWF [Ruoslahti & Pierschbacher, 1987]. Different adhesive peptides exhibit cell line dependent biological activity in vitro, but relatively few in vivo studies of adhesive peptides have been reported. Haverstick and colleagues [1985] reported that addition of RGD-containing peptides to protein-free platelet suspension inhibited thrombin induced platelet aggregation, as well as platelet adhesion in vitro to fibronectin, fibrinogen, and vWF coated polystyrene. These results suggest that binding of the peptide to the platelet renders the platelet’s receptor for the proteins unavailable for further binding. Similarly, Hanson and Harker [1988] used MAbs against the platelet glycoprotein IIb/IIIa (fibrinogen binding) complex to prevent thrombus formation upon Dacron vascular grafts placed within a chronic baboon AV shunt. Controlled release of either MAbs or adhesive peptides at the site of medical device implant might allow localized control of thrombosis. Locally administered adhesive peptides might also be used to selectively control cell behavior, e.g., Ruoslahti [1992] used RGD containing peptides to inhibit tumor invasion in vitro and dissemination in vivo. A recently developed drug delivery system offers great promise with respect to these therapeutic interventions [Markou et al., 1996; Markou et al., 1998]. Alternatively, adhesive peptides can be used to promote cell proliferation. Hubbell and colleagues [1991] reported that immobilization of different adhesive peptides resulted in selective cell response in vitro. Whereas the immobilized RGD and YIGSR peptides both enhanced spreading of human foreskin fibroblasts, human vascular smooth muscle cells, and human vascular endothelial cells, immobilization of REDV enhanced spreading of only endothelial cells. If this concept can be applied in vivo, then endothelialization in the absence of inflammation, infection, or thrombosis might be achieved.

Defining Terms Conformation: Higher order protein structure that describes the spatial relationship between the amino acid chains that a protein comprises. Epitope: Particular regions of a protein to which an antibody or cell can bind. Hemostasis: Mechanism by which damaged blood vessels are repaired without compromising normal blood flow. Integrin: Cellular transmembrane protein that acts as a receptor for adhesive extracellular matrix proteins such as fibronectin. The tripeptide RGD is the sequence recognized by many integrins. Organization: The manner in which a protein resides upon a surface, in particular, the existence, arrangement, and availability of different protein epitopes. Plasma concentration: Not the concentration of total protein in the plasma, but rather the volume fraction of plasma in the adsorption medium when protein adsorption from different dilutions of plasma is measured. Post-adsorptive transitions: Changes in protein conformation and organization that occur when adsorbed proteins reside upon a surface. Residue: The individual amino acids of a peptide or protein. Specific activity: The amount of radioactivity detected per unit mass of a specific protein. State: The reactive state of an adsorbed protein as determined by its conformation and organization. Thrombosis: Formation of plug comprising blood platelets and fibrin that stops blood flow through damaged blood vessels. Embolized thrombus refers to a plug that has detached from the wound site and entered the circulation.


Vroman effect: Collective term describing (1) maximal adsorption of a specific protein from multicomponent medium at early time, (2) maximal equilibrium adsorption from intermediate dilution, and (3) decrease with increased adsorption time in the plasma concentration at which adsorption is maximum. Vroman peak: The adsorption maximum in the adsorption versus log (plasma concentration) curve when protein adsorption from different dilutions of plasma is measured.

References Andrade JD. 1985. Principles of protein adsorption. In JD Andrade (ed), Surface and Interfacial Aspects of Biomedical Polymers: Protein Adsorption, vol 2, pp 1–80, Plenum, New York. Andrade JD, Hlady V. 1986. Protein adsorption and materials biocompatibility: A tutorial review and suggested hypotheses. In Advances in Polymer Science 79. Biopolymers/Non-Exclusion HPLC, pp 1–63, Springer-Verlag, Berlin. Andrade JD. 1985. Polymer surface analysis: Conclusions and expectations. In JD Andrade (ed), Surface and Interfacial Aspects of Biomedical Polymers, vol 1, Surface Chemistry and Physics, pp. 443-460, Plenum Press, Boca Raton. Balasubramanian V, Grusin NK, Bucher RW, et al. In Press. Residence-time dependent changes in fibrinogen adsorbed to polymeric biomaterials. J. Biomed. Mater. Res. Bohnert JL, Horbett TA. 1986. Changes in adsorbed fibrinogen and albumin interactions with polymers indicated by decreases in detergent elutability. J. Coll. Interface Sci. 111:363. Brash JL, ten Hove P. 1984. Effect of plasma dilution on adsorption of fibrinogen to solid surfaces. Thromb. Haemostas. 51:326. Castillo EJ, Koenig JL, Anderson JM et al., 1984. Characterization of protein adsorption on soft contact lenses. I. Conformational changes of adsorbed serum albumin. Biomaterials 5:319. Castillo EJ, Koenig JL, Anderson JM et al., 1985. Characterization of protein adsorption on soft contact lenses. II. Reversible and irreversible interactions between lysozyme and soft contact lens surfaces. Biomaterials 6:338. Chenowith DE. 1988. Complement activation produced by biomaterials. Artif. Organs 12:502. Chinn JA, Posso SE, Horbett TA et al., 1991. Post-adsorptive transitions in fibrinogen adsorbed to Biomer: Changes in baboon platelet adhesion, antibody binding, and sodium dodecyl sulfate elutability. J. Biomed. Mater. Res. 25:535. Chinn JA, Posso SE, Horbett TA et al., 1992. Post-adsorptive transitions in fibrinogen adsorbed to polyurethanes: Changes in antibody binding and sodium dodecyl sulfate elutability. J. Biomed. Mater. Res. 26:757. De Baillou N, Dejardin P, Schmitt A et al., 1984. Fibrinogen dimensions at an interface: Variations with bulk concentration, temperature, and pH. J. Coll. Interf. Sci. 100:167. Dewanjee MK. 1992. Radioiodination: Theory, Practice, and Biomedical Application, Kluwer Academic Press, Norwell. Feng L, Andrade JD. 1994. Protein adsorption on low temperature isotropic carbon. I. Protein conformational change probed by differential scanning calorimetry. J. Biomed. Mater. Res. 28:735. Hanson SR, Harker LA. 1988. Interruption of acute platelet-dependent thrombosis by the synthetic antithrombin D-phenylalanyl-L-prolyl-L-arginyl chloromethylketone. Proc. Natl. Acad. Sci. USA 85:3184. Haverstick DM, Cowan JF, Yamada KM et al., 1985. Inhibition of platelet adhesion to fibronectin, fibrinogen, and von Willebrand factor substrates by a synthetic tetrapeptide derived from the cell-binding domain of fibronectin. Blood 66:946. Horbett TA. 1981. Adsorption of proteins from plasma to a series of hydrophilic-hydrophobic copolymers. II. Compositional analysis with the prelabeled protein technique. J. Biomed. Mater. Res. 15:673.


Horbett TA. 1982. Protein adsorption on biomaterials. In SL Cooper, NA Peppas (eds), Biomaterials: Interfacial Phenomena and Applications, ACS Advances in Chemistry Series, vol 199, pp 233–244, Washington, D.C., American Chemical Society. Horbett TA. 1984. Mass action effects on competitive adsorption of fibrinogen from hemoglobin solutions and from plasma. Thromb. Haemostas. 51:174. Horbett TA. 1986. Techniques for protein adsorption studies. In DF Williams (ed), Techniques of Biocompatibility Testing, pp 183–214, CRC Press, Boca Raton. Horbett TA, Grunkemeier JM, Lew KR. 1994. Fibrinogen orientation of a surface coated with a GP IIb/IIIa peptide detected with monoclonal antibodies. Trans. Soc. Biomater. 17:335. Horbett TA, Lew KR. 1994. Residence time effects on monoclonal antibody binding to adsorbed fibrinogen. J. Biomat. Sci. Polym. Edn. 6:15. Hubbell JA, Massia SP, Desai NP et al., 1991. Endothelial cell-selective materials for tissue engineering in the vascular graft via a new receptor. Biotechnology (NY) 9:568. Kaplan AP. 1978. Initiation of the intrinsic coagulation and fibrinolytic pathway of man: The role of surfaces, Hageman factor, prekallikrein, high molecular weight kininogen, and factor XI. Prog. Hemostas. Thromb. 4:127. Kingshott P, St. John HAW, Vaithianathan T et al., 1998. Study of protein adsorption at monolayer and sub-monolayer levels by surface-MALDI spectroscopy. Trans. Soc. Biomater. 21:253. Lindon JN, McManama G, Kushner L et al., 1986. Does the conformation of adsorbed fibrinogen dictate platelet interactions with artificial surfaces? Blood 68:355. Macfarlane AS. 1958. Efficient trace-labelling of proteins with iodine. Nature 182:53. Markou CP, Chronos NF, Harker LA et al., 1996. Local endovascular drug delivery for inhibition of thrombosis. Circulation 94:1563. Markou CP, Lutostansky EM, Ku DN et al., 1998. A novel method for efficient drug delivery. Ann. Biomed. Engr. 26:502. Norde W, MacRitchie F, Nowicka, G et al. 1986. Protein adsorption at solid-liquid interfaces: reversibility and conformation aspects. J. Colloid Interface Sci. 112:447. Norde W. 1986. Adsorption of proteins from solution at the solid-liquid interface. Adv. Coll. Interf. Sci. 25:267. Norde W, Lyklema J. 1991. Why proteins prefer interfaces. J. Biomater. Sci. Polym. Edn. 2:183. Pettit DK, Hoffman AS, Horbett TA. 1994. Correlation between corneal epithelial cell outgrowth and monoclonal antibody binding to the cell domain of fibronectin. J. Biomed. Mater. Res. 228:685. Phillips DR, Charo IF, Parise LV et al., 1988. The platelet membrane glycoprotein IIb-IIIa complex. Blood 71:831. Pierschbacher MD, Hayman EG, Ruoslahti E. 1981. Location of the cell-attachment site in fibronectin with monoclonal antibodies and proteolytic fragments of the molecule. Cell 26:259. Rainbow MR, Atherton S, Eberhart RE. 1987. Fluorescence lifetime measurements using total internal reflection fluorimetry: Evidence for a conformational change in albumin adsorbed to quartz. J. Biomed. Mater. Res. 21:539. Rapoza RJ, Horbett TA. 1990. Postadsorptive transitions in fibrinogen: Influence of polymer properties. J. Biomed. Mater. Res. 24:1263. Ratner BD, McElroy BJ. 1986. Electron spectroscopy for chemical analysis: Applications in the biomedical sciences. In RM Gendreau RM (ed), Spectroscopy in the Biomedical Sciences, pp 107–140, CRC Press, Boca Raton. Ratnoff OD, Menzie C. 1950. A new method for the determination of fibrinogen in small samples of plasma. J. Lab. Clin. Med. 37:316. Ruoslahti E. 1991. Integrins. J. Clin. Invest. 87:1. Ruoslahti E. 1992. The Walter Herbert Lecture: Control of cell motility and tumour invasion by extracellular matrix interactions. Br. J. Cancer 66:239.


Ruoslahti E, Pierschbacher MD. 1987. New perspectives in cell adhesion: RGD and integrins. Science 238:491. Sakariassen KS, Bolhuis PA, Sixma JJ. 1979. Human blood platelet adhesion to artery subendothelium is mediated by factor VIII-Von Willebrand factor bound to the subendothelium. Nature 279:636. Sandwick RK, Schray KJ. 1987. The inactivation of enzymes upon interaction with a hydrophobic latex surface. J. Coll. Interf. Sci. 115:130. Sandwick RK, Schray KJ. 1988. Conformational states of enzymes bound to surfaces. J. Coll. Interf. Sci. 121:1. Shiba E, Lindon JN, Kushner L et al., 1991. Antibody-detectable changes in fibrinogen adsorption affecting platelet activation on polymer surfaces. Am. J. Physiol. 260:C965. Sit PS, Siedlecki CA, Shainoff JR et al. 1998. Substrate-dependent conformations of human fibrinogen visualized by atomic force microscopy under aqueous conditions. Trans. Soc. Biomater. 21:101. Slack SM, Horbett TA. 1989. Changes in the state of fibrinogen adsorbed to solid surfaces: An explanation of the influence of surface chemistry on the Vroman effect. J. Colloid Interface Sci. 133:148. Slack SM, Horbett TA. 1992. Changes in fibrinogen adsorbed to segmented polyurethanes and hydroxymethacrylate-ethylmethacrylate copolymers. J. Biomed. Mater. Res. 26:1633. Slack SM, Horbett TA. 1995. The Vroman effect: A critical review. In TA Horbett, JL Brash (eds), Proteins At Interfaces II, pp 112–128, Washington, D.C., American Chemical Society. Smith JM. 1981. Chemical Engineering Kinetics, 3rd ed, McGraw Hill, New York. Soria J, Soria C, Mirshahi M et al., 1985. Conformational changes in fibrinogen induced by adsorption to a surface. J. Coll. Interf. Sci. 107:204. Stryer L. 1995. Biochemistry, 4th ed, W.H. Freeman, New York. Vroman L, Adams AL. 1969. Identification of rapid changes at plasma-solid interfaces. J. Biomed. Mater. Res. 3:43. Vroman L, Adams AL, Fischer GL et al., 1980. Interaction of high molecular weight kininogen, factor XII and fibrinogen in plasma at interfaces. Blood 55:156. Willems GM, Hermens WT, Hemker HC. 1991. Surface exclusion and molecular mobility may explain Vroman effects in protein adsorption. J. Biomater. Sci. Polym. Edn. 1:217. Wojciechowski P, ten Hove P, Brash JL. 1986. Phenomenology and mechanism of the transient adsorption of fibrinogen from plasma (Vroman effect). J. Colloid Interface Sci. 111:455. Yamada KM. 1991. Adhesive recognition sequences. J. Biol. Chem. 266:12809. Young BR, Lambrecht LK, Cooper SL. 1982. Plasma proteins: Their role in initiating platelet and fibrin deposition on biomaterials. In SL Cooper, NA Peppas (eds), Biomaterials: Interfacial Phenomena and Applications, ACS Advances in Chemistry Series, vol 199, pp 317–350, Washington, D.C., American Chemical Society. Zamarron C, Ginsberg MH, Plow EF. 1990. Monoclonal antibodies specific for a conformationally altered state of fibrinogen. Thromb. Haemostas. 64:41. Zucker MB, Vroman L. 1969. Platelet adhesion induced by fibrinogen adsorbed on glass. Proc. Soc. Exp. Med. 131:318.

Further Information The American Society for Artificial and Internal Organs (ASAIO) (P.O. Box C, Boca Raton, FL 334290468, Web address http://www.asaio.com) publishes original articles in ASAIO J through Lippincott-Raven Publishers (227 East Washington Square, Philadelphia, PA 19106-3780) and meeting transactions in Trans. Am. Soc. Artif. Intern. Organs. The Society for Biomaterials (SFB) (6518 Walker St., Ste. 150, Minneapolis, MN 55426-4215, Web address http://www.biomaterials.org) publishes original articles in both J. Biomed. Mater. Res. and J. Biomed. Mater. Res. Appl. Biomater. through John Wiley and Sons, Inc. (605 Third Ave., New York, NY 10158) and meeting transactions in Trans. Soc. Biomaterials. The J. Colloid Interface Sci. often contains articles related to protein-surface interactions as well.


The American Association for the Advancement of Science (AAAS) (1200 New York Ave. NW, Washington, D.C. 20002, Web address http://www.aaas.org) publishes original articles in science, and often contains excellent review articles and very current developments in protein research. Nature, published through Macmillan Magazines (Porters South, 4 Crinan St., London N1 9XW), is similar in content but provides a decidedly European perspective. Comprehensive references summarizing applications of protein adsorption and biocompatibility include: 1982. SL Cooper, NA Peppas (eds.), Biomaterials: Interfacial Phenomena and Applications, ACS Advances in Chemistry Series, vol 199, Washington, D.C., American Chemical Society; 1987. TA Horbett and JL Brash (eds.), Proteins at Interfaces: Physicochemical and Biochemical Studies, ACS Symposium Series, vol 343, Washington, D.C., American Chemical Society; and 1993. SM Factor (ed.), Cardiovascular Biomaterials And Biocompatibility. Cardiovasc. Path. 2(3) Suppl.


Mooney, D. J., Langer, R. S. “Engineering Biomaterials for Tissue Engineering: The 10–100 Micron Size Scale.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

112 Engineering Biomaterials for Tissue Engineering: The 10–100 Micron Size Scale David J. Mooney Massachusetts Institute of Technology

Robert S. Langer

112.1 112.2

Fundamentals Applications

Massachusetts Institute of Technology

112.3

Conclusions

Immunoprotective Devices • Devices with Open Structures

A significant challenge in tissue engineering is to take a biomaterial and process it into a useful form for a specific application. All devices for tissue engineering transplant cells and/or induce the ingrowth of desirable cell types from the host organism. The device must provide sufficient mechanical support to maintain a space for tissue to form or serve as a barrier to undesirable interactions. Also, the device can be designed to provide these functions for a defined period before biodegradation occurs or on a permanent basis. Generally speaking, devices can be broken down into two types. Immunoprotective devices contain semipermeable membranes that prevent elements of the host immune system (e.g., IgG antibodies and lymphocytes) from entering the device. In contrast, open devices have large pores (>10 µm) and allow free transport of molecules and cells between the host tissue and transplanted cells. These latter devices are utilized to engineer a tissue that is completely integrated with the host tissue. Both types of devices can range in size from microns to centimeters or beyond, although the larger sizes are usually repetitions on the structure found at the scale of hundreds of microns. A fundamental question in designing a device is whether to use synthetic or natural materials. Synthetic materials (e.g., organic polymers) can be easily processed into various structures and can be produced cheaply and reproducibly; it also is possible to tightly control various properties such as the mechanical strength, hydrophobicity, and degradation rate of synthetic materials. Whereas natural materials (e.g., collagen) sometimes exhibit a limited range of physical properties and can be difficult to isolate and process, they do have specific biologic activity. In addition, these molecules generally do not elicit unfavorable host tissue responses, a condition which is typically taken to indicate that a material is biocompatible. Some synthetic polymers, in contrast, can elicit a long-term inflammatory response from the host tissue [Bostman, 1991]. A significant challenge in fabricating devices is either to develop processing techniques for natural biomaterials that allow reproducible fabrication on a large-scale basis [Cavallaro et al., 1994] or to develop materials that combine the advantages of synthetic materials with the biologic activity of natural biomaterials [Barrera et al., 1993; Massia & Hubbell, 1991]. Computer-aided-design–computer-aided-manufacturing (CAD-CAM) technology may possibly be employed in the future to custom-fit devices with complex structures to patients.


112.1 Fundamentals The interaction of the host tissue with the device and transplanted cells can be controlled by both the geometry of the device and the internal structure. The number of inflammatory cells and cellular enzyme activity around implanted polymeric devices has been found to depend on the geometry of the device [Matlaga et al., 1976], with device geometries that contain sharp angles provoking the greatest response. The surface geometry, or microstructure, of implanted polymer devices also has been found to affect the types and activities of acute inflammatory cells recruited to the device as well as the formation of a fibrous capsule [Taylor & Gibbons, 1983]. The pore structure of a device dictates the interaction of the device and transplanted cells with the host tissue. The pore structure is determined by the size, size distribution, and continuity of the individual pores within the device. Porous materials are typically defined as microporous (pore diameter d < 2 nm), mesoporous (2 nm < d < 50 nm), or macroporous (d > 50 nm) [Schaeffer, 1994]. Only small molecules (e.g., gases) are capable of penetrating microporous materials. Mesoporous materials allow transport of larger molecules, such as small proteins, but transport of large proteins and cells is prevented. Macroporous materials allow free transport of large molecules, and, if the pores are large enough (d > 104 nm), cells are capable of migrating through the pores of the device. The proper design of a device can allow desirable signals (e.g., a rise in serum sugar concentration) to be passed to transplanted cells while excluding molecular or cellular signals which would promote rejections of transplanted cells (e.g., IgG protein). Fibrovascular tissue will invade a device if the pores are larger than approximately 10 µm, and the rate of invasion will increase with the pore size and total porosity of a device [Mikos et al., 1993c; Weslowski et al., 1961; White et al., 1981]. The degree of fibrosis and calcification of early fabric leaflet valves has been correlated to their porosity [Braunwald et al., 1965], as has the nonthrombogenicity of arterial prosthesis [DeBakey et al., 1964] and the rigidity of tooth implants and orthopedic prosthesis [Hamner et al., 1972; Hulbert et al., 1972]. It is important to realize that many materials do not have a unimodal pore size distribution or a continuous pore structure, and the ability of molecules or cells to be transported through such a device will often be limited by bottlenecks in the pore structure. In addition, the pore structure of a device may change over time in a biologic environment. For example, absorption of water into polymers of the lactic/glycolic acid family results in the formation first of micropores, and eventually of macropores as the polymer itself degrades [Cohen et al., 1991]. The porosity and pore-size distribution of a device can be determined utilizing a variety of techniques [Smith et al., 1994]. Specific properties (e.g., mechanical strength, degradability, hydrophobicity, biocompatibility) of a device are also often desirable. These properties can be controlled both by the biomaterial itself and by the processing technique utilized to fabricate the device. An advantage of fabricating devices from synthetic polymers is the variety of processing techniques available for these materials. Fibers, hollow fibers, and porous sponges can be readily formed from synthetic polymers. Natural biomaterials must be isolated from plant, animal, or human tissue and are typically expensive and suffer from large batchto-batch variations. Although the wide range of processing techniques available for synthetic polymers is not available for these materials, cells specifically interact with certain types of natural biomaterials, such as extracellular matrix (ECM) molecules [Hynes, 1987]. The known ability of ECM molecules to mediate cell function in vitro and in vivo may allow precise control over the biologic response to devices fabricated from ECM molecules.

112.2 Applications The applications of tissue engineering are very diverse, encompassing virtually every type of tissue in the human body. However, the devices utilized in these areas can be divided into two broad types. The first type, immunoprotective devices, utilizes a semipermeable membrane to limit communication between


cells in the device and the host. The small pores in these devices (d < 10 nm) allow low-molecular-weight proteins and molecules to be transported between the implant and the host tissue, but they prevent large proteins (e.g., immunoglobulins) and host cells (e.g., lymphocytes) of the immune system from entering the device and mediating rejection of the transplanted cells. In contrast, open structures with large pores are typically utilized (d > 10 µm) if one desires that the new tissue be structurally integrated with the host tissue. Applications that utilize both types of devices are described below.

Immunoprotective Devices Devices that protect transplanted cells from the immune system of the host can be broken down into two types, microencapsulation and macroencapsulation systems [Emerich et al., 1992]. Individual cells or small clusters of cells are surrounded by a semipermeable membrane and delivered as a suspension in microencapsulation systems (Fig. 112.1a). Macroencapsulation systems utilize hollow semipermeable membranes to deliver multiple cells or cell clumps (Fig. 112.1b). The small size, thin wall, and spherical shape of microcapsules all optimize diffusional transport to and from the microencapsulated cells. Macroencapsulation devices typically have greater mechanical integrity than microencapsule devices, and they can be easily retrieved after implantation. However, the structure of these devices is not optimal for diffusional transport. Nonbiodegradable materials are the preferred choice for fabricating both types of devices, as the barrier function is typically required over the lifetime of the implant. A significant effort has been made to cure diabetes by transplanting microencapsulated pancreatic islet cells. Transplantation of nonimmunoprotected islets has led to short-term benefits [Lim & Sun, 1980], but the cells were ultimately rejected. To prevent this, islets have been immobilized in alginate (a naturally occurring polymer derived from seaweed) microbeads coated with a layer of poly(L-lysine) and a layer of polyethyleneimine [Lim & Sun, 1980]. Alginate is ionically crosslinked in the presence of calcium, and the permeability of alginate/poly(L-lysine) microbeads is determined by the formation of ionic or hydrogen bonds between the polyanion alginate and the polycation poly(L-lysine). This processing technique allows cells to be immobilized without exposure to organic solvents or high temperatures. The outer layer of polyethyleneimine was subsequently replaced by a layer of alginate to prevent the formation of fibrous capsules around the implanted microcapsules [O’Shea et al., 1984]. Smaller microbeads (250–400 µm) have been generated with an electrostatic pulse generator [Lum et al., 1992] to improve the in vivo survival and the response time of encapsulated cells [Chicheportiche & Reach, 1988]. These devices have been shown to be effective in a variety of animal models [Lim & Sun, 1980; Lum et al., 1992], and clinical trials of microencapsulated islets in diabetic patients are in progress [Soon-Shiong et al., 1994]. Synthetic analogs to alginate have also been developed [Cohen et al., 1990]. The superior mechanical stability of macroencapsulation devices, along with the possibility of retrieving the entire device, makes these types of devices especially attractive when the transplanted cells have limited lifetimes and/or when one needs to ensure that the transplanted cells are not migrating out of the device. Macroencapsulation devices have been utilized to transplant a variety of cell types, including pancreatic cells [Lacy et al., 1991], NGF-secreting cells [Winn et al., 1994], dopamine-secreting cells [Emerich et al., 1992], and Chromaffin cells [Sagen et al., 1993]. The nominal molecular mass cutoff of the devices was 50 kD, allowing immunoprotection without interfering with transport of therapeutic agents from the encapsulated cells. To prevent cell aggregation and subsequent large-scale cell death due to nutrient limitations, macroencapsulated islets have been immobilized in alginate [Lacy et al., 1991].

Devices with Open Structures Devices with large, interconnected pores (d > 10 µm) are utilized in applications where one wishes the transplanted cells to interact directly with host tissue and form a structurally integrated tissue. The open structure of these devices provides little barrier to diffusional transport and often promotes the ingrowth of blood vessels from the host tissue. Degradable materials are often utilized in these applications, since once tissue structures are formed the device is not needed.


FIGURE 112.1 Examples of microencapsulated cells and a device used for macroencapsulation of cells. (a) Phase contrast photomicrograph of hybridoma cells encapsulated in a calcium crosslinked polyphosphazene gel [Cohen et al., 1993]. Original magnification was 100×. Used with permission of Editions de Sante. (b) A SEM photomicrograph of a poly(acrylonitrile-co-vinyl chloride) hollow fiber formed by phase inversion using a dry-jet wet-spinning technique [Schoichet et al., 1994]. Used with permission of John Wiley and Sons, Inc.

These types of devices typically fall into two categories. The first types is fabrics, either woven or nonwoven, of small-diameter (approximately 10–40 µm) fibers (Fig. 112.2a). High porosity (>95%) and large average pore size can be easily obtained with this type of material, and these materials can be readily shaped into different geometries. Fibers can be formed from synthetic, crystalline polymers such as polyglycolic acid by melt extrusion [Frazza & Schmitt, 1971]. Fibers and fabrics also can be formed from natural materials such as type I collagen, a type of ECM molecule, by extrusion of soluble collagen into a bath where gelling occurs followed by dehydration of the fiber [Cavallaro et al., 1994]. The tensile strength of fibers is dependent on the extent of collagen crosslinking, which can be controlled by the processing technique [Wang et al., 1994]. Processed collagen fibers can be subsequently spooled and


FIGURE 112.2 Examples of a fiber-based fabric and a porous sponge utilized for tissue engineering. (a) A photomicrograph of type I collagen fibers knitted into a fabric. Fiber diameters can be as small as 25 µm, and devices constructed from these fibers can be utilized for a variety of tissue engineering applications [Cavallaro et al., 1994]. Used with permission of John Wiley and Sons, Inc. (b) A SEM photomicrograph of a formaldehyde-crosslinked polyvinyl alcohol sponge. These devices have been utilized for a variety of applications, including hepatocyte transplantation [Uyama et al., 1993].

knitted to form fabrics [Cavallaro et al., 1994]. These devices are often ideal when engineering twodimensional tissues. However, these fabrics typically are incapable of resisting large compressional forces, and three-dimensional devices are often crushed in vivo. Three-dimensional fiber-based structures have been stabilized by physically bonding adjacent fibers [Mikos et al., 1993a; Mooney et al., 1994a; Vacanti et al., 1992]. To engineer three-dimensional tissues, porous sponge devices (Fig. 112.2b) are utilized typically in place of fiber-based devices. These devices are better capable of resisting larger compressional forces


(approximately 104 Pa) [Mikos et al., 1993a] than are unbounded fiber-based devices (approximately 102 Pa) [Mooney et al., 1994a] due to the continuous solid phase and can be designed to have complex, three-dimensional forms [Mikos et al., 1993b; White et al., 1972]. Porous sponges can be fabricated from synthetic polymers utilizing a variety of techniques, including performing the polymerization of a hydrophobic polymer in an aqueous solution [Chirila et al., 1993], exploiting phase separation behavior of dissolved polymers in specific solvents [Lo et al., 1994], and combining solvent casting with particulate leaching [Mikos et al., 1993b; Mikos et al., 1994]. Porous sponges can be formed from type I collagen and other ECM molecules by chemically crosslinking gels or assembling the collagen in nonnatural polymeric structures [Bell et al., 1981; Chvapil, 1979; Stenzel et al., 1974; Yannas et al., 1982]. Perhaps the most significant clinical effort using open devices has been expended to engineer skin tissue to treat burn victims. Both natural [Bell et al., 1981; Yannas et al., 1982] and synthetic degradable materials [Hansbrough et al., 1992] in the form of porous sponges or fiber-based fabrics have been utilized to transplant various cellular elements of skin. One device fabricated from ECM molecules has also been combined with an outer coat of silicone elastomer to prevent dehydration of the wound site [Yannas et al., 1982]. The various approaches have shown efficacy in animal models, and tissue-engineered skin has progressed to clinical trials [Burke et al., 1981; Compton et al., 1989; Heimbach et al., 1988; Stern et al., 1990]. Another area with great clinical potential is the engineering of bone and cartilage tissue. Various ceramics and biodegradable synthetic polymers have been utilized to fabricate devices for these purposes. Porous calcium phosphate devices loaded with mesenchymal stem cells have been shown to promote bone formation when implanted into soft tissue sites of animals [Goshima et al., 1991; Haynesworth et al., 1992]. Ceramics also have been coated onto prosthetic devices (e.g., hip replacements) to promote bone ingrowth and bonding between the prosthetic device and the host tissue [Furlong & Osborn, 1991]. The degradation rate [de Bruijn et al., 1994] and mechanical properties [Yoshinari et al., 1994] of these ceramics can be controlled by the deposition technique, which determines the crystallinity and chemical structure of the deposited ceramic. The brittleness of ceramic materials limits them in certain applications, and to bypass this problem composite ceramic/polymer devices have been developed [Stupp et al., 1993]. Fiber-based fabrics of biodegradable polymers have also been utilized to transplant cells derived from periosteal tissue and form new bone tissue [Vacanti et al., 1993]. To engineer cartilage tissue with specific structures such as an ear or nasal septum, devices have been fabricated from a nonwoven mesh of biodegradable synthetic polymers molded to the size and shape of the desired tissue. These devices, after seeding with chondrocytes and implantation, have been shown to induce the formation of new cartilage tissue with the same structure as the polymer device utilized as the template [Puelacher et al., 1993; Vacanti et al., 1991, 1992]. After tissue development is complete, the device itself degrades to leave a completely natural tissue. Liver tissue [Uyama et al., 1993], ligaments [Cavallaro et al., 1994], and neural tissue [Guenard et al., 1992; Madison et al., 1985] also have been engineered with open devices. Tubular tissues, including blood vessels [Weinberg & Bell, 1986], intestine [Mooney et al., 1994b; Organ et al., 1993], and urothelial structures [Atala et al., 1992] have been engineered utilizing open devices fabricated into a tubular structure.

112.3 Conclusions A variety of issues must be addressed to design and fabricate a device for tissue engineering. Do the transplanted cells need immunoprotection, or should they structurally integrate with the host tissue? If immunoprotection is desired, will a micro- or macroencapsulation device be preferred? If a structurally integrated new tissue is desired, will a fiber-based device or a porous sponge be more suitable? The specific roles that the device will play in a given application and the material itself will dictate the design of the device and the fabrication technique.


Defining Terms Biocompatible: A material which does not elicit an unfavorable response from the host but instead performs with an appropriate host response in a specific application [Williams, 1987]. Biodegradation: The breakdown of a material mediated by a biologic system [Williams et al., 1992]. Biodegradation can occur by simple hydrolysis or via enzyme- or cell-mediated breakdown. Extracellular matrix (ECM) molecules: Various substances present in the extracellular space of tissues that serve to mediate cell adhesion and organization. Immunoprotective: Serving to protect from interacting with the immune system of the host tissue, including cellular elements (e.g., lymphocytes) and proteins (e.g., IgG). Open devices: Devices with large (d > 10 µm) interconnected pores which allow unhindered transport of molecules and cells within the device and between the device and the surrounding tissue.

References Barrera DA, Zylstra E, Lansbury PT, et al. 1993. Synthesis and RGD peptide modification of a new biodegradable copolymer: poly(lactic acid-co lysine). J Am Chem Soc 115:11010. Bell E, Ehrlich HP, Buttle DJ, Nakatsuji T. 1981. Living tissue formed in vitro and accepted as skinequivalent tissue of full thickness. Science 211:1052. Bostman OM. 1991. Absorbable implants for the fixation of fractures. J Bone Joint Surg 73-A(1):148. Braunwald NS, Reis RL, Pierce GE. 1965. Relation of pore size to tissue ingrowth in prosthetic heart valves: an experimental study. Surgery 57:741. Burke JF, Yannas IV, Quinby WC, et al. 1981. Successful use of a physiological acceptable artificial skin in the treatment of extensive burn injury. Ann Surg 194:413. Cavallaro JF, Kemp PD, Kraus KH. 1994. Collagen fabrics as biomaterials. Biotech Bioeng 43:781. Chicheportiche D, Reach G. 1988. In vitro kinetics of insulin release by microencapsulated rat islets: effect of the size of the microcapsules. Diabetologia 31:54. Chirila TV, Constable IJ, Crawford GJ, et al. 1993. Poly(2-hydroxyethyl methacrylate) sponges as implant materials: in vivo and in vitro evaluation of cellular invasion. Biomaterials 14(1):26. Chvapil M. 1979. Industrial uses of collagen. In DAD Parry, LK Creamer (eds), Fibrous proteins: scientific, industrial, and medical aspects, London, L.K. Academic Press. Cohen S, Allcock HR, Langer R. 1993. Cell and enzyme immobilization in ionotropic synthetic hydrogels. In AA Hincal, HS Kas (eds), Recent advances in pharmaceutical and industrial biotechnology, Editions de Sante, Paris. Cohen S, Bano MC, Visscher KB, et al. 1990. Ionically cross-linkable phosphazene: a novel polymer for microencapsulation. J Am Chem Soc 112:7832. Cohen S, Yoshioka T, Lucarelli M, et al. 1991. Controlled delivery systems for proteins based on poly(lactic/glycolic acid) microspheres. Pharm Res 87(6):713. Compton C, Gill JM, Bradford DA. 1989. Skin regenerated from cultured epithelial autografts on fullthickness wounds from 6 days to 5 years after grafting: A light, electron microscopic, and immunohistochemical study. Lab Invest 60:600. DeBakey ME, Jordan GL, Abbot JP, et al. 1964. The fate of dacron vascular grafts. Arch Surg 89:757. De Bruijn JD, Bovell YP, van Blitterswijk CA. 1994. Structural arrangements at the interface between plasma sprayed calcium phosphates and bone. Biomaterials 15(7):543. Emerich DF, Winn SR, Christenson L, et al. 1992. A novel approach to neural transplantation in Parkinson’s disease: Use of polymer-encapsulated cell therapy. Neurosci Biobeh Rev 16:437. Frazza EJ, Schmitt EE. 1971. A new absorbable suture. J Biomed Mater Res Symp 1:43. Furlong RJ, Osborn JE. 1991. Fixation of hip prostheses by hydroxylapatite ceramic coatings. J Bone Jt Surg 73-B(5):741. Goshima J, Goldberg VM, Caplan AI. 1991. The origin of bone formed in composite grafts of porous calcium phosphate ceramic loaded with marrow cells. Clin Orthop Rel Res 191:274.


Guenard V, Kleitman N, Morissey TK, Bunge RP, Aebischer P. 1992. Syngeneic schwann cells derived from adult nerves seeded in semipermeable guidance channels enhance peripheral nerve regeneration. J Neurosci 12:3310–3320. Hamner JE, Reed OM, Greulich RC. 1972. Ceramic root implantation in baboons. J Biomed Mater Res Symp 6:1. Hansbrough JF, Cooper ML, Cohen R, et al. 1992. Evaluation of a biodegradable matrix containing cultured human fibroblasts as a dermal replacement beneath meshed skin grafts on athymic mice. Surgery 111(4):438. Haynesworth SE, Goshima J, Goldberg VM, et al. 1992. Characterization of cells with osteogenic potential from human marrow. Bone 13:81. Heimbach D, Luterman A, Burke J, et al. 1988. Artificial dermis for major burns. Ann Surg 208(3):313. Hulbert SF, Morrison SJ, Klawitter JJ. 1972. Tissue reaction to three ceramics of porous and nonporous structures. J Biomed Mater Res 6:347. Hynes RO. 1987. Integrins: A family of cell surface receptors. Cell 48:549. Lacy PE, Hegre OD, Gerasimidi-Vazeou A, et al. 1991. Maintenance of normoglycemia in diabetic mice by subcutaneous xenografts of encapsulated islets. Science 254:1782. Lim F, Sun AM. 1980. Microencapsulated islets as bioartificial endocrine pancreas. Science 210:908. Lo H, Kadiyala S, Guggino SE, et al. 1994. Biodegradable foams for cell transplantation. In R Murphy, A Mikos (eds), Biomaterials for drug and cell delivery, Materials Research Society Proceedings, vol 331. Lum Z, Krestow M, Tai IT, et al. 1992. Xenografts of rat islets into diabetic mice. Transplantation 53(6):1180. Madison R, Da Silva CR, Dikkes P, et al. 1985. Increased rate of peripheral nerve regeneration using bioresorbable nerve guides and a laminin-containing gel. Exp Neurol 88:767. Massia SP, Hubbell JA. 1991. An RGD spacing of 440 nm is sufficient for integrin αvβ3-mediated fibroblast spreading and 140 nm for focal contact and stress fiber formation. J Cell Biol 115(5):1089. Matlaga BF, Yasenchak LP, Salthouse TN. 1976. Tissue response to implanted polymers: the significance of sample shape. J Biomed Mater Res 10:391. Mikos AG, Bao Y, Cima LG, et al. 1993a. Preparation of poly(glycolic acid) bonded fiber structures for cell attachment and transplantation. J Biomed Mater Res 27:183. Mikos AG, Sarakinos G, Leite SM, et al. 1993b. Laminated three-dimensional biodegradable foams for use in tissue engineering. Biomaterials 14(5):323. Mikos AG, Sarakinos G, Lyman MD, et al. 1993c. Prevascularization of porous biodegradable polymers. Biotech Bioeng 42:716. Mikos AG, Thorsen AJ, Czerwonka LA, et al. 1994. Preparation and characterization of poly(L-lactic) foams. Polymer 35(5):1068. Mooney DJ, Mazzoni CL, Organ GM, et al. 1994a. Stabilizing fiber-based cell delivery devices by physically bonding adjacent fibers. In R Murphy, A Mikos (eds), Biomaterials for Drug and Cell Delivery, Materials Research Society Proceedings, Pittsburgh, Pennsylvania, vol 331, 47–52. Mooney DJ, Organ G, Vacanti JP. 1994b. Design and fabrication of biodegradable polymer devices to engineer tubular tissues. Cell Trans 3(2):203. Organ GM, Mooney DJ, Hansen LK, et al. 1993. Enterocyte transplantation using cell-polymer devices causes intestinal epithelial-lined tube formation. Transplan Proc 25:998. O’Shea GM, Goosen MFA, Sun AM. 1984. Prolonged survival of transplanted islets of Langerhans encapsulated in a biocompatible membrane. Biochim Biophys Acta 804:133. Puelacher WC, Vacanti JP, Kim WS, et al. 1993. Fabrication of nasal implants using human shape specific polymer scaffolds seeded with chondrocytes. Surgical Forum 44:678–680. Sagen J, Wang H, Tresco PA, et al. 1993. Transplants of immunologically isolated xenogeneic chromaffin cells provide a long-term source of pain-reducing neuroactive substances. J Neuroscience 13(6):2415. Schaeffer DW. 1994. Engineered porous materials. MRS Bulletin April 1994:14. Schoichet MS, Winn SR, Athavale S, et al. 1994. Poly(ethylene oxide)-grafted thermoplastic membranes for use as cellular hydrid bio-artificial organs in the central nervous system. Biotech Bioeng 43:563.


Smith DM, Hua D, Earl WL. 1994. Characterization of porous solids. MRS Bulletin April 1994:44. Soon-Shiong P, Sandford PA, Heintz R, et al. 1994. First human clinical trial of immunoprotected islet allografts in alginate capsules. Society for Biomaterials Annual Meeting, Boston, Mass, abstract 356. Stenzel KH, Miyata T, Rubin AL. 1974. Collagen as a biomaterial. Ann Rev Biophys Bioeng 3:231. Stern R, McPherson M, Longaker MT. 1990. Histologic study of artificial skin used in the treatment of full-thickness thermal injury. J Burn Care Rehab 11:7. Stupp SI, Hanson JA, Eurell JA, et al. 1993. Organoapatites: Materials for artificial bone: III. Biological testing. J Biomed Mater Res 27(3):301. Taylor SR, Gibbons DF. 1983. Effect of surface texture on the soft tissue response to polymer implants. J Biomed Mat Res 17:205. Uyama S, Takeda T, Vacanti JP. 1993. Delivery of whole liver equivalent hepatic mass using polymer devices and hepatotrophic stimulation. Transplantation 55(4):932. Vacanti CA, Cima LG, Ratkowski D, et al. 1992. Tissue engineered growth of new cartilage in the shape of a human ear using synthetic polymers seeded with chondrocytes. In LG Cima, ES Ron (eds), Tissue Inducing Biomaterials, pp 367–374, Materials Research Society Proceedings, Pittsburgh, Pennsylvania, vol 252. Vacanti CA, Kim W, Upton J, et al. 1993. Tissue engineered growth of bone and cartilage. Transplan Proc 25(1):1019. Vacanti CA, Langer R, Schloo B, et al. 1991. Synthetic polymers seeded with chondrocytes provide a template for new cartilage formation. Plast Reconstr Surg 88(5):753. Wang MC, Pins GD, Silver FH. 1994. Collagen fibers with improved strength for the repair of soft tissue injuries. Biomaterials 15:507. Weinberg CB, Bell E. 1986. A blood vessel model constructed from collagen and cultured vascular cells. Science 231:397. Weslowski SA, Fries CC, Karlson KE, et al. 1961. Porosity: Primary determinant of ultimate fate of synthetic vascular grafts. Surgery 50(1):91. White RA, Hirose FM, Sproat RW, et al. 1981. Histopathologic observations after short-term implantation of two porous elastomers in dogs. Biomaterials 2:171. White RA, Weber JN, White EW. 1972. Replamineform: A new process for preparing porous ceramic, metal, and polymer prosthetic materials. Science 176:922. Williams DF. 1987. Definitions in Biomedicals. Progress in Biomedical Engineering, vol 4, New York, Elsevier. Williams DF, Black J, Doherty PJ. 1992. Second consensus conference on definitions in biomaterials. In PJ Doherty, RL Williams, DF Williams, et al. (eds), Advances in Biomaterials, vol 10, BiomaterialsTissue Interactions, pp 525–533, New York, Elsevier. Winn SR, Hammang JP, Emerich DF, et al. 1994. Polymer-encapsulated cells genetically modified to secrete human nerve growth factor promote the survival of axotomized septal cholinergic neurons. Proc Natl Acad Sci 91:2324–2328. Yannas IV, Burke JF, Orgill DP, et al. 1982. Wound tissue can utilize a polymeric template to synthesize a functional extension of skin. Science 215:174. Yoshinari M, Ohtsuka Y, Derand T. 1994. Thin hydroxyapatite coating produced by the ion beam dynamic mixing method. Biomaterials 15:529.

Further Information The Society for Biomaterials, American Society for Artificial Internal Organs, Cell Transplantation Society, and Materials Research Society all sponsor regular meetings and/or sponsor journals relevant to this topic. The following Materials Research Society Symposium Proceedings contain a collection of relevant articles: volume 252, Tissue Inducing Biomaterials (1992); volume 331, Biomaterials for Drug and Cell Delivery (1994). Another good source of relevant material is Tissue Engineering, edited by R Skalak and CF Fox, New York, Alan Riss (1988).


Yannas, I.V. “Regeneration Templates.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

113 Regeneration Templates 113.1 113.2 113.3 113.4 113.5 113.6 113.7 113.8

Studies of Nerve Regeneration Using Degradable Tubes

Ioannis V. Yannas Massachusetts Institute of Technology

The Problem of the Missing Organ Search Principles for Identification of Regeneration Templates Structural Specificity of Dermis Regeneration Template (DRT) In Situ Synthesis of Skin with DRT Advantages and Disadvantages of Clinical Treatment of Skin Loss with DRT Modifications of DRT: Use of a Living Dermal Equivalent The Bilayered Skin-Equivalent Graft Structural Specificity of Nerve Regeneration Template (NRT)

113.9

In Situ Synthesis of Meniscus Using a Meniscus Regeneration Template (MRT)

113.1 The Problem of the Missing Organ Drugs typically replace or correct a missing function at the molecular scale; by contrast, regeneration templates replace the missing function at the scale of tissue or organ. An organ may be lost to injury or may fail in disease: The usual response of the organism is repair, which amounts to contraction and synthesis of scar tissue. Tissues and organs in the adult mammal typically do not regenerate. There are exceptions, such as epithelial tissues of the skin, gastrointestinal tract, genitals, and the cornea, all of which regenerate spontaneously; the liver also shows ability to synthesize substantial organ mass, though without recovery of the original organ shape. There are reports that bone and the elastic ligaments regenerate. These exceptions underscore the fact that the loss of an organ by the adult mammal almost invariably is an irreversible process, since the resulting scar tissue largely or totally lacks the structure and function of the missing organ. The most obvious examples involve losses due to injury such as the loss of a large area of skin following a burn accident or the loss of substantial nerve mass following an automobile accident. However, irreversible loss of function can also occur following disease, although over a lengthy time: Examples are the inability of a heart valve to prevent leakage during diastole as a result of valve tissue response to an inflammatory process (rheumatic fever), and the inability of liver tissue to synthesize enzymes due to its progressive replacement by fibrotic tissue (cirrhosis). Five approaches have been used to solve the problem of the missing organ. In autografting, a mass of similar or identical tissue from the patient (autograft) is surgically removed and used to treat the area of loss. The approach can be considered to be spectacularly successful until one considers the long-term cost incurred by the patient. An example is the use of sheet autograft to treat extensive areas of fullthickness skin loss; although the patient incorporates the autograft fully with excellent recovery of function, the “donor” site used to harvest the autograft remains scarred. When the autograft is not


available, as is common in cases of burns extending over more than 30% of body surface area autograft is meshed in order to make it extensible enough to cover the large wound areas. However, the meshed autograft provides cover only where the graft tissue provides direct cover; where there is no cover, scar forms, and the result is one of low cosmetic value. Similar problems of donor site unavailability and scarring must be dealt with in heart bypass surgery, another widespread example of autografting. In transplantation, the donor tissue is typically harvested from a cadaver, and the recipient has to cope with the problems of rejection and the risk of transmission of viruses from this allograft. Another approach has been based on efforts to synthesize tissues in vitro using autologous cells from the patient; this approach has yielded so far a cultured epidermis (a tissue which regenerates spontaneously provided there is a dermal substrate underneath) about 2–3 weeks after the time when the patient was injured. In vitro synthesis of the dermis, a tissue which does not regenerate, has not been accomplished so far. Perhaps the most successful approach from the commercial standpoint has been the one in which engineered biomaterials are used; these materials are typically required by their designers to remain intact themselves without interfering with the patient’s physiologic functions during the entire lifetime; overwhelmingly, this requirement is observed in its breach. A fifth approach is based on the discovery that an analog of the extracellular matrix (ECM) induces partial regeneration of the dermis, rather than of scar, in full-thickness skin wounds in adult mammals (human, guinea pig, pig) where it is well known that no regeneration occurs spontaneously. This fifth approach of solving the problem of organ loss, in situ regeneration, will be described in this chapter. Efforts to induce regeneration have been successful with only a handful of ECM analogs. Evidence of regeneration is sought after the ECM analog has been implanted in situ, i.e., at the lesion marking the site of the missing organ. When morphogenesis is clearly evident, based on tests of recovery both of the original tissue structure and function, the matrix which has induced these physiologic or nearly physiologic tissues is named a regeneration template. In the absence of evidence of such morphogenetic activity of the cell-free matrix, the latter is not referred to as a regeneration template.

113.2 Search Principles for Identification of Regeneration Templates Several parameters have been incorporated in the search for organ regeneration templates. Briefly, these parameters account for the performance of the implant during the early or acute stage following implantation (physicochemical parameters) as well as for the long-term or chronic stage following implantation (biologic parameters). Immediately upon making contact with the wound bed, the implant must achieve physicochemical nanoadhesion (i.e., adhesion at a scale of 1 nm) between itself and the lesion. Without contact of this type it is not possible to establish and maintain transport of molecules and cells between implant and host tissue. The presence of adequate adhesion can be studied by measurements of the force necessary to peel the implant from the wound bed immediately after grafting. Empirical evidence has supported a requirement for an implant which is capable of isomorphous tissue replacement, i.e., the synthesis of new tissue at a rate which is of same order as the rate of degradation of the matrix.

()

tb =O 1 th

(113.1)

In Eq. (113.1), tb denotes a characteristic time constant for biodegradation of the implant at that tissue site, and th denotes a time constant for healing, the latter occurring by synthesis of new tissue inside the implant. A third requirement refers to the critical cell path length lc, beyond which migration of a cell into the implant deprives it of an essential nutrient, assumed to be transported from the host tissue by diffusion alone. The emphasis is on the characteristic diffusion path for the nutrient during the early stages of wound healing, before significant angiogenesis occurs several days later. Calculation of the critical cell


path can be done by use of the cell lifeline number, S, a dimensionless number expressing the relative importance of a chemical reaction, which leads to consumption of an essential nutrient by the cell, and of diffusion of the nutrient which alone makes the latter accessible to the cell. This number is defined as

S=

rl 2 Dco

(113.2)

where r is the rate of consumption of the nutrient by the cell in mole/cm3/s, l is the diffusion length, D is the diffusivity of the nutrient in the medium of the implant, and co is the nutrient concentration at or near the surface of the wound bed, in mole/cm3. When S = O(1) the value of l is the critical path length, lc along which cells can migrate, away from host tissue, without requirement of nutrient in excess of that supplied by diffusion. Eq (113.2) can, therefore, be used to define the maximum implant thickness beyond which cells require the presence of capillaries. The chemical composition of the implant which has induced regeneration was designed on the basis of studies of wound-healing kinetics. In most mammals, full-thickness skin wounds close partly by contraction of the wounds edges and partly by synthesis of scar tissue. Clearly, skin regeneration over the entire area of skin loss cannot occur unless the wound edges are kept apart and, in addition, the healing processes in the wound bed are modified drastically enough to yield a physiologic dermis rather than scar. Although several synthetic polymers, such as porous (poly)dimethyl siloxane, delay contraction to a small but significant extent, they do not degrade and therefore violate isomorphous tissue replacement, Eq. (113.1). Synthetic biodegradable polymers, such as (poly)lactic acid, can be modified, e.g., by copolymerization with glycolic acid, to yield polymers which satisfy Eq. (113.1); however, evidence is lacking that these synthetic polymers delay contraction and prevent synthesis of scar. By contrast, there is considerable evidence that a certain analog of the extracellular matrix (ECM analogs) not only delays contraction significantly but also leads to synthesis of partly physiologic skin. Systematic use of the delay in contraction as an essay has been made to identify the structural features of the dermis regeneration template (DRT), as shown schematically in Fig. 113.1.

FIGURE 113.1 The kinetics of contraction of full-thickness guinea pig skin wounds can be used to separate collagengraft-glycosaminoglycan copolymers into three classes, as shown. The wound half-life, t1/2, is the number of days necessary to reduce the original wound area to 50%. (Courtesy of Massachusetts Institute of Technology.)


Summarized in Fig. 113.1 are three modes of wound-healing behavior, each elicited by an ECM analog of different design. Mode O of healing is described by a very short time for onset of contraction, followed by contraction and definitive closure of the wound with formation of a thin linear scar. Mode I is characterized by a significant delay in onset of contraction, following which contraction proceeds and eventually leads to a linear scar. Mode II is characterized by a significant delay in onset of contraction (somewhat smaller than in mode I), followed by contraction and then by reversal of contraction, with expansion of the original wound perimeter at a rate which exceeds significantly the growth rate of the entire animal. Mode II healing leads to synthesis of a partly physiologic dermis and a physiologic epidermis within the perimeter of the expanded wound bed. Mode O healing occurs when an ECM analog which lacks specificity is used to graft the wound. Mode O is also observed when the wound bed remains ungrafted. Mode I healing occurs when an ECM analog of highly specific structure (the DRT) is grafted on the wound bed. Mode II healing occurs when the DRT, previously identified as the ECM analog which leads to mode I healing, is seeded with autologous epidermal cells before being grafted. Although contraction is a convenient screening method for identification of structural features of ECM analogs which induce skin regeneration, a different procedure has been used to identify the features of an implant that induces regeneration of the peripheral nerve. The structural features of the nerve regeneration template (NRT) were identified using an essay focused on the long-term return of function of the nerve following treatment with candidate ECM analogs. Recent studies have shown that it is possible to achieve regeneration both of a dermis and an epidermis in sequence, rather than simultaneously, provided that the animal model used is one in which wound closure does not take place almost overwhelmingly by wound contraction. The choice of animal model is, therefore, critical in this respect. Mode I behavior, illustrated in Fig. 113.1, is applicable to wound healing in rodents, such as guinea pigs and rats. In these animals, about 90% of wound closure is accounted for fully by wound contraction, the remainder is accounted for by formation of new tissue. Grafting of a full-thickness skin defect in a rodent with the ECM analog which possesses a highly specific structure (the dermis regeneration template or DRT; see below) leads to lengthy delay of the onset of contraction, eventually followed by contraction and formation of a linear scar-like tissue. Although the gross appearance of the tissue formed is that of a modified linear scar (see Fig. 113.1), there is some histological evidence that the small mass of connective tissue layer formed underneath the epidermis is not scar. In contrast, in animals in which contraction contributes about equally to wound closure as does formation of new tissue, such as the swine, approximately one half of the initial wound area is eventually closed by formation of partly regenerated dermis; soon after that, the new dermis is covered by a new epidermis. In wounds which close in large part by formation of new tissue rather than by contraction, as in the swine, grafting of the cell-free DRT leads, therefore, to sequential formation of a dermis and an epidermis as well, which is the same end result that is arrived at simultaneously by grafting with the keratinocyte-seeded DRT. Of several ECM analogs that have been prepared, the most commonly studied is a graft copolymer of type I collagen and chondroitin 6-sulfate. The structure of the latter glycosaminoglycan (GAG) is illustrated below in terms of the repeat unit of the disaccharide, an alternating copolymer of D-glucuronic acid and of an O-sulfate derivative of N-acetyl D-galactosamine:


The principle of isomorphous replacement, Eq. (113.1), cannot be satisfied unless the biodegradation time constant of the network, tb, can be adjusted to an optimal level, about equal to the rate of synthesis of new tissue at that site. Reduction of the biodegradation rate of collagen can be achieved either by grafting GAG chains onto collagen chains or by crosslinking collagen chains to each other. The chemical grafting of GAG chains on polypeptide chains proceeds by previously coprecipitating the two polymers under conditions of acidic pH, followed by covalent crosslinking of the freeze-dried precipitate. A particularly useful procedure for crosslinking collagen chains to GAG, or collagen chains to each other, is a self-crosslinking reaction, requiring no use of crosslinking agent. This condensation reaction principally involves carboxylic groups from glutamyl/aspartyl residues on polypeptide chain P1 and -amino groups of lysyl residues on an adjacent chain P2 to yield covalently bonded collagen chains; as well as condensation of amine groups of collagen with carboxylic groups of glucuronic acid residues on GAG chains to yield graft-copolymers of collagen and GAG:

P1− COOH + P2 − NH 2 → P2 − NHCO − P1 + H 2O

(113.3a)

GAG − COOH + P2 − NH 2 → P2 − NHCO − GAG + H 2O

(113.3b)

In each case above the reaction proceeds to the right, with formation of a three-dimensional crosslinked network when the moisture content of the protein, or protein-GAG coprecipitate, drops below about 1 wt%. As illustrated by Eq. (113.3), removal of water, the volatile product of the condensation, is favored by conditions which drive the reaction towards the right, with formation of a crosslinked network. Thus, the reaction proceeds to the right when both high temperature and vacuum are used. Another crosslinking reaction, used extensively in preparing implants that have been employed in clinical studies as well as in animal studies, involves use of glutaraldehyde. Dehydration crosslinking, which amounts to self-crosslinking as described above, obviously does not lend toxicity to these implants. Glutaraldehyde, on the other hand, is a toxic substance, and devices treated with it require thorough rinsing before use until free glutaraldehyde cannot be detected in the rinse water. Network properties of crosslinked collagen-GAG copolymers can be analyzed structurally by studying the swelling behavior of small specimens. The method is based on the theory of Flory and Rehner, who showed that the volume fraction of a swollen polymer v2 depends on the average molecular weight between crosslinks, Mc, through the following relationship:

(

)

(

)(

)

ln 1 − v2 + v2 + χv2 − ρV1 M c v12 3 − v2 2 = 0

(113.4)

In Eq. (113.4), V1 is the molar volume of the solvent, ρ is the density of the dry polymer, and χ is a constant characteristic of a specific polymer-solvent pair at a particular temperature. Although the chemical identify of collagen-GAG copolymers is a necessary element of their biologic activity, it is not a sufficient one. In addition to chemical composition and the crosslink density, biologic activity also depends strongly on the pore architecture of these ECM analogs. Pores are incorporated first by freezing a very dilute suspension of the collagen-GAG coprecipitate and then by inducing sublimation of the ice crystals by exposing to vacuum at low temperatures. The resulting pore structure is, therefore, a negative replica of the network of ice crystals (dendrites). It follows that control of the conditions of ice crystal nucleation and growth can lead to a large variety of pore structures. In practice, the average pore diameter decreases with decreasing temperature of freezing while the orientation of pore channel axes also depends on the magnitude of the heat flux vector during freezing. The dependence of pore channel orientation on heat transfer parameters is illustrated by considering the dimensionless Mikic number Mi, a ratio of the characteristic freezing time of the aqueous medium of the collagen-GAG suspension, tf, to the characteristic time for entry, te , of a container filled with the suspension which is lowered at constant velocity into a well-stirred cooling bath:


Mi =

tf te

=

ρw h fg rV 10k j ∆T

(113.5)

In Eq. (113.5), ρw is the density of the suspension, hfg is the heat of fusion of the suspension, r is an arbitrary length scale, V is the velocity with which the container is lowered into the bath, kj is the thermal conductivity of the jacket, and ∆T is the difference between freezing temperature and bath temperature. The shape of the isotherms near the freezing front is highly dependent on the value of Mi. The dominant heat flux vector is normal to these isotherms, i.e., ice dendrites grow along this vector. It has been observed that, for Mi < 1 (slow cooling), the isotherms are shallow, flat-shaped parabolae, and the ice dendrites exhibit high axial orientation. For Mi > 1 (rapid cooling), the isotherms are steep parabolae, and ice dendrites are oriented along the radial direction. The structure of the porous matrix is defined by quantities such as the volume fraction, specific surface, mean pore size, and orientation of pores in the matrix. Determination of these properties is based on principles of stereology, the discipline which relates the quantitative statistical properties of three-dimensional structures to those of their two-dimensional sections or projections. In reverse, stereologic procedures allow reconstruction of certain aspects of three-dimensional objects from a quantitative analysis of planar images. A plane which goes through the two-phase structure of pores and collage-GAG fibers may be sampled by random points, by a regular pattern of points, by a near-total sampling using a very dense array of points, or by arranging the sampling points to form a continuous line. The volume fraction of pores, VV , is equal to the fraction of total test points which fall inside pore regions, PP , also equal to the total area fraction of pores, AA , and, finally, equal to the line fraction of pores, LL , for a linear point array in the limit of infinitely close point spacing

VV = PP = AA = LL

(113.6)

Whether cells of a particular type should be part of a regeneration template depends on predictions derived from models of developmental biology as well as empirical findings obtained with well-defined wound-healing models. During morphogenesis of a large variety of organs, an interaction between epithelial and mesenchymal cells, mediated by the basal lamina which is interleaved between the two types of cells, is both necessary and sufficient for development of local physiologic structures involving two types of tissue in juxtaposition. In particular, skin morphogenesis in a full-thickness wound model requires the presence of this interaction between the two cell types and the basal lamina over a critical period. In skin wound–healing experiments with adult mammals, wound healing proceeds with formation of scar, rather than physiologic skin, if the wound bed contains epithelial cells and mesenchymal cells (fibroblasts) but no ECM structure which could act temporarily as a basal lamina. If, by contrast, an analog of the basal lamina is present, wound healing proceeds with nearly physiologic morphogenesis of skin. Furthermore, no epidermis is formed unless epithelial cells become involved in wound healing early during wound healing and continue being involved until they have achieved confluence. It is also known that no dermis forms if fibroblasts are not available early during wound healing. These observations suggest the requirements for a DRT which is an analog of the basal lamina and is designed to encourage the migration and interaction of both epithelial cells and fibroblasts within its volume. Nerve regeneration following injury essentially amounts to elongation of a single nerve cell across a gap resulting from the injury. During nerve development many nerve cells are elongated by processes which eventually become axons. Interaction of elongating processes with basal lamina are credited as being essential in the formation of nerve during development, and it will be assumed here that such interactions are essential in regeneration following adult injury as well. It is also known that Schwann cells, which derive from neural crest cells during development, are essential contributors to regeneration following injury to the adult peripheral nerve. These considerations suggest a nerve regeneration template which is structured as an analog of the basal lamina, interacting with the elongating axons in the presence of Schwann cells.


113.3 Structural Specificity of Dermis Regeneration Template (DRT) The major events accompanying skin wound healing can be summarized as contraction and scar synthesis. Conventional wisdom prescribes the need for a treatment that accelerates wound healing. A large number of devices that claim to speed up various aspects of the healing process have been described in the literature. The need to achieve healing within as short a time as possible is certainly well founded, especially in the clinical setting where the risk to patient’s life as well as the morbidity increase with extension of time to heal. However, the discovery of partial skin regeneration by use of the skin regeneration template has introduced the option of a drastically improved healing result for the patient in exchange for a slightly extended hospital stay. The DRT was optimized in studies with animals in which it was observed that skin regeneration did not occur unless the test ECM analog effectively delayed, rather than accelerated, wound contraction. The length of delay in onset of contraction eventually was used as a quantitative basis for preliminary optimization of the structural features of the DRT. Optimization studies are currently continuing on the basis of a new criterion, namely, the fidelity of regeneration achieved. Systematic use of the criterion of contraction inhibition has been made to select the biodegradation rate and the average pore diameter of DRT. The kinetics of contraction of full-thickness skin wounds in the guinea pig have been studied for each of the three modes of healing schematically presented in Fig. 113.1. The results, presented in Fig. 113.2, show that mode O healing is characterized by early onset of contraction, whereas mode I and mode II healing show a significant delay in onset of contraction. A measure of the delay is the wound half-life t1/2, the time required for the wound area to decrease to 50% of the original value. Use of this index of contraction rate has been made in Figs. 113.3 and 113.4, which present data on the variation of wound half-life with average pore diameter and degradation rate for a

FIGURE 113.2 The kinetics of guinea pig skin wound contraction following grafting with three classes of ECM analogs. Inactive ECM analogs delay the onset of contraction only marginally over the ungrafted wound, whereas active cell-free ECM analogs delay the onset of contraction significantly. When seeded with epithelial cells, not only does an active ECM analog delay the onset of contraction significantly, but it also induces formation of a confluent epidermis and then arrests and reverses the direction of movement of wound edges, leading to expansion of the wound perimeter and to synthesis of partly physiologic skin. (Courtesy of Massachusetts Institute of Technology.)


FIGURE 113.3 The half-life wounds, t1/2, grafted with ECM analogs varies with the average pore diameter of the analog. The range of pore diameters where activity is maximal is shown by broken lines. The half-life of the ungrafted wound is shown for comparison. (Courtesy of Massachusetts Institute of Technology.)

FIGURE 113.4 Variation of wound half-life with degradation rate R of the ECM analog used as graft for a fullthickness guinea pig skin wound. R varies inversely as the biodegradation time constant, tf. The region of maximal activity is indicated by the broken line. The half-life of the ungrafted wound is shown for comparison. (Courtesy of Massachusetts Institute of Technology.)

type I collagen-chondroitin 6-sulfate copolymer. In Fig. 113.2, mode I kinetics are observed using a cellfree copolymer with average pore diameter and biodegradation rate that correspond to the regions of Figs. 113.3 and 113.4 in which the values of half-life are maximal; these regions characterize the structure


of SRT which possesses maximal activity. Maximum delay in wound half-life up to 27 ± 3 days is seen to have occurred when the average pore diameter has ranged from values as low as 20 ± 4 µm to an upper limit of 125 ± 35 µm. In addition, significant delay in wound healing has been observed when the degradation rate has become less than 115 ± 25 enzyme units. The latter index of degradation has been based on the in vitro degradation rate of the copolymer in a standardized solution of bacterial collagenase. The upper limit in degradation rate, defined in Fig. 113.4, is consistent with the requirement of a lower limit in tb , Eq. (113.1), below which the implant is biodegraded too rapidly to function as a scaffold over the period necessary for synthesis of new tissue; the latter is characterized by th . The lower limit in pore diameter, defined in Fig. 113.3, suggests a dimension which is on the order of two cell diameters; adequate space for migration of mesenchymal cells (fibroblasts) from the wound bed into the DRT is thereby guaranteed. At an estimated velocity of 0.2 mm/day for fibroblasts these cells would be expected to migrate across a 0.5-mm thickness of the DRT within very few days, provided that adequate supplies of critical nutrients could be made available to them from host tissue. Use of Eq. (113.2) leads to an estimated critical cell path length lc of order 100 µm. These estimates suggest that fibroblasts can migrate across at least one-fifth the thickness of the porous implant without requiring the presence of capillaries. However, taking into account the observation that wound exudate (comprising primarily serum, with growth factors and nutrients) fills at least one-half the thickness of the implant within no more than a few hours following grafting, we conclude that the boundary of “host” tissue has moved clearly inside the implant. Epithelial cells, as well as fibroblasts and a basal lamina analog, also are required for morphogenesis. They can be supplied in a variety of forms. They are always available as epithelial cell sheets, migrating from the wound edges toward the center of the wound bed. However, when the area of skin loss is several centimeters, as with a severely burned patient, these cell sheets, migrating with speeds of about 0.5 mm/day from opposite edges, would not be expected to cover one-half the characteristic wound dimension in less than the time constant th for synthesis of new tissue. In the absence of epithelial cells, therefore, at the center of the wound bed, Eq. (113.1) wound be violated. To overcome this limitation, which is imposed by the scale of the wound, it has been necessary to resort to a variety of procedures. In the first, uncultured autologous epidermal cells, extracted from a skin biopsy by controlled enzymatic degradation, have been seeded into ECM analogs by centrifugation into the porous matrix prior to grafting of the latter on the wound bed. Tested with animals, the procedure leads to formation of a confluent epidermis by about 2 weeks, provided that at least 5 × 104 epithelial cells per cm2 of DRT area have been seeded. In another procedure, a very thin epidermal layer has been surgically removed from an intact area of the patient and has been grafted on the dermal layer which has been synthesized about 2 weeks after grafting with the DRT. The latter procedure has been tested clinically with reproducible success. A third procedure, studied with animals, has made use of cultured epithelia, prepared by a 2- to 3-week period of culture of autologous cells in vitro and grafted on the newly synthesized dermal bed. Approximately equivalent fidelity of skin regeneration has been obtained by each of these procedures for supplying epithelial cells to the DRT in the treatment of skin wounds of very large area.

113.4 In Situ Synthesis of Skin with DRT The skin regeneration template DRT induces regeneration of skin to a high degree of fidelity. Fidelity of regeneration has been defined in terms of the degree of recovery of structural and functional features which are present in the intact organ. The first test of fidelity of skin regeneration following use of the DRT was a study of treatment of fullthickness skin loss in guineas pigs. In this study the lesion was produced by surgery on healthy animals. The characteristic dimension of the wound was about 3 cm, and the desired period for cover by a confluent epidermis was 2 weeks. Covering a wound of such a scale within the prescribed time would have been out of reach of epithelial cells migrating from the wound edges. Accordingly, autologous epithelial cells were extracted from a skin biopsy and seeded into the DRT under conditions of carefully controlled centrifugation.


A clear and unmistakable difference between healing in the presence and absence of DRT was provided by observing the gross anatomy of the wound (Fig. 113.2). In the absence of the DRT, the wound contracted vigorously and closed up with formation of a linear scar by about day 30. In the presence of a cell-seeded DRT, the wound perimeter started contracting with a delay of about 10 days, and contraction was completely arrested and then reversed between days 30 and 40. The wound then continued to expand at a rate that was clearly higher than that expected from the rate of growth of the animal. The long-term appearance of the wound treated with the cell-seeded DRT was that of an organ that appeared grossly identical in color, texture, and touch to intact skin outside the wound perimeter. However, the newly synthesized skin was totally hairless, and the total area of new skin was smaller in area from the original wound area by about 30% (Fig. 113.2). Morphologic studies of newly synthesized skin in the presence of cell-seeded DRT included comparison with intact skin and scar. Optical microscopy and electron microscopy were supplemented by laser light scattering, the latter used to provide a quantitative measure of collagen fiber orientation in the dermal layer. It was concluded that, in most respects, partly regenerated skin was remarkably similar to intact guinea pig skin. The epidermis in regenerated skin was often hyperplastic; however, the maturation sequence and relative proportion of all cell layers were normal. Keratohyaline granules of the neoepidermis were larger and more irregular in contour than those of the normal granular cell layer. The new skin was characterized by melanocytes and Langerhans cells, as well as a well-formed pattern of rete ridges and interdigitations with dermal papillae, all of which appear in normal skin. Newly synthesized skin was distinctly different morphologically from scar. Scar showed characteristic thinning (atrophy) of the epidermis, with absence of rete ridges and of associated dermal papillae. Elastic fibers in regenerated skin formed a delicate particulate structure, in contrast with scar, where elastic fibers were thin and fragmented. The dermal layer in regenerated skin comprised collagen fibers which were not oriented in the plane of the epidermis, as well as fibroblasts which were not elongated; in scar, collagen fibers were highly oriented in the plane and fibroblasts were elongated. Both normal and regenerated skin comprised unmyelinated nerve fibers within dermal papillae, closely approximated to the epidermis; scar had few, if any, nerves. There were no hair follicles or other skin appendages either in regenerated skin or in scar. Laser light–scattering measurements of the orientation of collagen fibers in tissue sections of the dermal layer were based on use of the Hermans orientation function

f = 2 cos2 α − 1

(113.7)

In Eq. (113.7), α is the angle between an individual fiber and the mean axis of the fibers, and 〈cos2α〉 is the square cosine of α averaged over all the fibers in the sample. For a random arrangement of fibers, 〈cos2α〉 equals 1/2, while for a perfectly aligned arrangement it is equal to 1. Accordingly, S varies from 0 (truly random) to 1 (perfect alignment). Measurements obtained by use of this procedure showed that S took the values 0.20 ± 0.11, 0.48 ± 0.05 and 0.75 ± 0.10 for normal dermis, regenerated dermis, and scar dermis, respectively. These results provided objective evidence that regenerated dermis had a morphology of collagen fibers that was clearly not scar (n = 7; p < 0.001), and was intermediate between that of scar and normal dermis. Functional studies of regenerated skin showed that the moisture permeability of intact skin and regenerated skin had values of 4.5 ± 0.8 and 4.7 ± 1.9 g/cm/h, insignificantly different from each other (n = 4; p < 0.8). Mechanical behavior studies showed a positive curvature for the tensile stress-strain curve of regenerated skin as well as of normal skin. However, the tensile strength of regenerated skin was 14 ± 4 MPa, significantly lower than the strength of intact skin, 31 ± 4 MPa (n = 4; p < 0.01). The second test of fidelity of skin regeneration was a study of treatment of 106 massively burned humans with the cell-free DRT. The characteristic dimension of the wound in this study was of order 15 cm, and the desired period for cover by a confluent epidermis was 2 weeks. The scale of the wound necessitated the introduction of epithelial cells, and this was accomplished by grafting the newly synthesized


dermal layer, 2 weeks after grafting with DRT, with a very thin epidermal layer (epidermal autograft), which was removed from an intact area of the patient. The results of the histologic study of the patient population showed that physiologic dermis, rather than scar, had been synthesized in sites that had been treated with DRT and had later been covered with an epidermal graft. Progress has been made in clarification of the mechanism by which DRT induces regeneration of the dermis. There is considerable evidence, some of which was presented above, supporting the hypothesis that inhibition of wound contraction is required for regeneration. The evidence also suggests strongly that DRT competitively inhibits formation of specific binding interactions between contractile cells and endogenous ECM.

113.5 Advantages and Disadvantages of Clinical Treatment of Skin Loss with DRT When skin is the missing organ, the patient faces threats to life posed by severe dehydration and infection. These threats can be eliminated permanently only if the area of missing skin is covered with a device that controls moisture flux within physiologic limits and presents an effective barrier to airborne bacteria. Both of these functions can be returned over a period of about 2–4 weeks by use of temporary dressings. The latter include the allograft (skin from a cadaver) and a very large variety of membranes based on synthetic or natural polymers. None of these dressings solves the problem of missing skin over the lifetime of the patient: The allograft does not support synthesis of physiologic skin and must be removed to avoid rejection, and the devices based on the vast majority of engineered membranes do not make effective biologic contact with the patient’s tissues, and all lead to synthesis of scar. Three devices have been tested extensively for their ability to provide long-term physiologic cover to patients with massive skin loss: the patient’s own skin (autograft) the dermis regeneration template (DRT), and cultured epithelia (CEA). All three of these treatments have been studied extensively with massively burned patients. Of these, the autograft and DRT are effective even when the loss of skin extends through the full thickness of the dermis (e.g., third-degree burns), whereas CEA is effective when the loss of skin is through part of its thickness only. The basis for the differences among the three treatments lies in the intrinsic response of skin to injury. Skin comprises three layers: the epidermis, a 100-µm thick cellular layer; the dermis, a 1- to 5-mm layer of connective tissue with very few cells; and the subcutis, a 2- to 4-mm layer of primarily adipose tissue. In the adult mammal, an epidermis lost through injury regenerates spontaneously provided that a dermal substrate is present underneath. When the dermis is lost, whether through part thickness or full thickness, none of the injured mass regenerates; instead, a nonphysiologic tissue, scar, forms. Scar is epithelialized and can, therefore, control moisture flux within physiologic limits as well as provide a barrier to bacterial invasion. However, scar does not have the mechanical strength of physiologic skin. Also, scar synthesis frequently proceeds well beyond what is necessary to cover the wound, and the result of such proliferation is hypertrophic scarring, a cosmetically inferior integument which, when extending over a large area or over hands or face, reduces significantly the mobility of the patient’s joints as well as the patient’s cosmetic appearance. Autograft can, if used without meshing, provide an excellent permanent cover; if, however, as commonly practiced, autograft is meshed before grafting in order to extend the wound area that becomes covered, the result is new integument which comprises part physiologic skin and part scar and provides the patient with a solution of much lower quality than unmeshed (sheet) autograft. The clinical use of skin regeneration template has led to a new integument comprising a dermal layer, which has been synthesized by the patient, and an epidermal layer, which has been harvested as a very thin epidermal autograft and has been placed on top of the newly synthesized dermis. The chief advantages DRT over the meshed autograft are the shorter time that it takes to heal the donor site, from which the epidermal graft was harvested, and a superior cosmetic result at the site of the wound. The main disadvantage in the clinical use of DRT is the subjection of the patient to two surgical treatments rather than one (first graft DRT, then graft the epidermal layer after about 2 weeks).


Although it has been shown that a dermis and an epidermis are regenerated sequentially in DRT-treated wounds like that of the swine, in which contraction plays only a modest role in wound closure, the process of sequential skin regeneration is slower than that of simultaneous skin regeneration. Accordingly, the kinetics of wound closure suggest an advantage in the use of keratinocyte-seeded DRT (a two-stage treatment) relative to the unseeded DRT (one-stage treatment). The clinical advantages of CEA are the ability to grow a very large area, about 10,000 times as large as the area of the skin biopsy, thereby providing cover over the entire injured skin area of the patient. The disadvantages of cultured epithelia are the lengthy period required to culture the epidermis from the patient’s biopsy and the inability of the CEA to form a mechanically competent bond with the wound bed.

113.6 Modifications of DRT: Use of a Living Dermal Equivalent The design concept of a skin regeneration template outlined above has been adopted and modified. One modification involves the replacement of the collagen-GAG matrix with a biodegradable mesh consisting of either (poly)glycolic acid (PGA) or polyglactin-910 (PGL) fibers. The latter is used as a matrix for the in vitro culture of human fibroblasts isolated from neonatal skin. Fibroblasts synthesize extracellular matrix inside the synthetic polymeric mesh, and this “living dermal equivalent” has been cryopreserved for a specified period prior to use. The living dermal equivalent has been used to graft full-thickness skin wounds in athymic mice. Following grafting of wounds, these PGA/PGL fibroblast grafts were covered with meshed allograft. The latter is human cadaver skin that was meshed in this study to expand it and achieve coverage of maximum possible wound area. This composite graft became vascularized and that, additionally, epithelial cells from the cadaver graft had migrated to the matrix underneath. After a period of about 100 days following grafting, the reported result was an epithelialized layer covering a densely cellular substratum that resembled dermis. A variant of this design, in which the meshed allograft is not used, has been reported; epidermal cells are cultured with the fibroblast mesh before grafting. Studies of these designs are in progress.

113.7 The Bilayered Skin-Equivalent Graft In one widely reported development, in vitro cell culture procedures have been used to synthesize a bilayered tissue which has been reported to be a useful model of human skin. Briefly, fibroblasts from humans or from rats have been placed inside a collagen gel. Under these conditions, fibroblasts exert contractile forces, trapping the cells inside the contracted collagen lattice. Human epithelial cells, which have been plated onto this contracted dermal equivalent, have been observed to attach to the collagen substrate, multiply, and spread to form a continuous sheet. Differentiation of this sheet has led to formation of specialized epidermal structures, such as a multilayered cell structure with desmosomes, tonofilaments, and keratohyalin granules. Further differentiation events have included the formation of a basement membrane (basal lamina) in vitro when a rat epidermis was formed on top of a dermal equivalent produced from rat fibroblasts. Grafts prepared from the bilayered skin equivalent have been grafted on animals. When grafted on animals, these structures have been reported to become well vascularized with a network of capillaries within 7 to 9 days. It has been reported that the best grafts have blocked wound contraction, but no systematic data have been presented which could be used to compare the in vivo performance of these skin equivalents with grafts based on DRT (see above). Gross observations of the area grafted with skin equivalents have shown pink hairless areas which were not hypertrophically scarred. A systematic comparison between scar and the tissue synthesized in sites that have been grafted with skin equivalents has not yet been made.


113.8 Structural Specificity of Nerve Regeneration Template (NRT) The design principles for regeneration templates presented above have been used to design implants for regeneration of peripheral nerves. The medical problem typically involves the loss of innervation in arms and legs, leading to loss of motor and sensory function (paralysis). The nerves involved are peripheral, and the design problem becomes the regeneration of injured peripheral nerves, with recovery of function. A widely used animal model for peripheral nerve injury is a surgically generated gap in the sciatic nerve of the rat. Interruption of nerve function in this case is localized primarily in the region of plantar muscles of the foot involved. This relatively well defined area of loss of function can then be studied neurologically in relative isolation from other neurologic events. Furthermore, the rate of recovery of function can be studied by electrophysiologic methods, a procedure which provides continuous data over the entire period of healing. In the peripheral nerve the healing period extends to about 10 weeks, clearly longer than healing in skin, which occurs largely within a period of only 3 weeks. When the sciatic nerve is cut and a gap forms between the two nerve ends, the distal part of the nerve is isolated from its cell body in the spinal cord. Communication between the central nervous system and the leg is no longer possible. The lack of muscle innervation leads to inactivity, which in turn leads to muscle atrophy. At the site of injury there is degeneration of the myelin sheath of axons, dissociation of Schwann cells and formation of scar. It has been hypothesized that the formation of scar impedes, more than any other single cause, the elongation of axons across the gap. Axonal elongation through a gap of substantial length becomes, therefore, a parameter of prime importance in the design of a nerve regeneration template (NRT). Intubation of severed nerve ends is a widely used procedure for isolating axons from the tissue environment of the peripheral nerve. The lumen of the tube serves to isolate the process of nerve regeneration from wound-healing events involving connective tissues outside the nerve; the tube walls, for example, prevent proliferation of scar tissue inside the tube and the subsequent obstruction of the regenerating nerve. Silicone tubes are the most widely used method of intubation. These tubes are both nonbiodegradable and nonpermeable to large molecules. In this isolated environment it is possible to study the substrate preferences of elongating axons by incorporating well-defined ECM analogs and studying the kinetics of functional recovery continuously with electrophysiologic procedures for about 40 weeks following implantation. As in studies of dermal regeneration described above, the ECM analogs which were used in the study of substrate preferences of axons were graft copolymers of type I collagen and chondroitin 6-sulfate. Controls used included the autograft, empty silicone tubes, as well as tubes filled with saline. In these studies, it was necessary to work with a gap dimension large enough to preclude spontaneous regeneration, i.e., regeneration in the absence of an ECM analog. It was observed that a gap length of 10 mm was occasionally innervated spontaneously in the rat, whereas no instances of spontaneous regeneration were observed with a 15-mm gap length. Gap lengths of 10 and 15 mm were used in these studies with rats. Three structural parameters of ECM analogs were varied systematically. The degradation rate had a significant effect on the fidelity of regeneration, and an abbreviated optimization procedure led to an ECM analog which degraded much faster than the DRT. In combination with Eq. (113.1), this empirical finding suggests that a healing nerve wound contains a much smaller concentration of the degrading enzyme, collagenase; this suggestion is qualitatively consistent with observations of collagenolytic activity in injured nerves. The average diameter also was found to have a significant effect on fidelity of regeneration, and the optimization procedure led to a value of 5 µm, significantly smaller than the average pore diameter in the DRT. Finally, use of Eq. (113.5) led to procedures of preparing ECM analogs, the pore channels of which were either highly aligned along the tube axis, randomly oriented, or radially oriented. ECM analogs with axially aligned pore channels were found to be superior to analogs with other types of alignment.


TABLE 113.1

Design Parameters for Two Regeneration Templates

Design Parameter of ECM Analog Degradation rate, enzyme units Average pore diameter, µm Pore channel orientation

DRT

NRT

150 5 axial

These results have led to a design for an NRT consisting of a specified degradation rate, average pore diameter, and pore channel alignment as shown in Table 113.1 in which the structural parameters of NRT are contrasted to those of DRT.

Studies of Nerve Regeneration Using Degradable Tubes Porous collagen tubes without matrix content have been extensively studied as guides for peripheral nerve regeneration in rodents and nonhuman primates. The walls of these collagen tubes had an average pore diameter which was considered sufficiently large for transport of molecules as large as bovine serum albumin (MW = 68 kDa). A 4-mm gap in the sciatic nerve of the rat was the standard injury studied. Other injury models that were studied included the 4-mm gap and the 15-mm gap in the adult monkey. The use of an empty tube did not allow for any degree of optimization of tube parameters to be achieved in this study. Nevertheless, the long-term results showed almost complete recovery of motor and sensory responses, at rates that approximated the recovery obtained following use of the nerve autograft, currently the best conventional treatment in cases of massive loss of nerve mass. The success obtained suggests that collagen tubes can be used instead of autografts, the harvesting of which subject the patient to nerve-losing surgery. Even more significant improvement in quality of nerve regeneration is obtained when collagen tubes are filled with NRT. Although lower than normal, the histomorphometric and electrophysiological properties of the regenerate resulting from use of an NRT-filled collagen tube have been statistically indistinguishable from those of the autograft control. Probably, however, the most important effect observed following implantation of the NRT-filled collagen tube has been the observation that, while the total number of myelinated axons in regenerated nerves had reached a plateau by 30 weeks after injury, the number of axons with diameter larger than 6 µm continues to increase at substantial rates through the completion of a recent study (60 weeks). Axons with a diameter larger than 6 µm have been uniquely associated with the A-fiber peak of the action potential, which determines the maximum conduction velocity of the regenerates. Thus, kinetic evidence has been presented which supports the view that the nerve trunk maturation process continues beyond 60 weeks after injury, resulting in a nerve trunk which increasingly approaches the structure of the normal control.

113.9 In Situ of Synthesis of Meniscus Using a Meniscus Regeneration Template (MRT) The meniscus of the knee performs a variety of functions which amount to joint stabilization and lubrication, as well as shock absorption. Its structure is that of fibrocartilage, consisting primarily of type I collagen fibers populated with meniscal fibrochondrocytes. The architecture of collagen fibers is complex. In this tissue, which has a shape reminiscent of one-half a doughnut, the collagen fibers are arranged in a circumferential pattern which is additionally reinforced by radially placed fibers. The meniscus can be torn during use, an event which causes pain and disfunction. Currently, the accepted treatments are partial or complete excision of torn tissue. The result of such treatment is often unsatisfactory, since the treated meniscus has an altered shape which is incompatible with normal joint motion and stability. The longterm consequence of such incompatibility is joint degeneration, eventually leading to osteoarthritis.


An effort to induce regeneration of the surgically removed meniscus was based on the use of a type I collagen-chondroitin 6-sulfate copolymer. The precise structural parameters of this matrix have not been reported, so it is not possible to discuss the results of this study in terms of possible similarities and differences with DRT and NRT. The study focused on the canine knee joint, since the latter is particularly sensitive to biomechanical alterations; in this model, joint instabilities rapidly lead to osteoarthritic changes, which can be detected experimentally. Spontaneous regeneration of the canine meniscus following excision is partial and leads to a biomechanically inadequate tissue which does not protect the joint from osteoarthritic changes. The latter condition provides the essential negative control for the study. In this study, knee joints were subjected to 80% removal (resection) of the medial meniscus, and the lesion was treated either by an autograft or by an ECM analog or was not treated at all. Evaluation of joint function was performed by studying joint stability, gait, and treadmill performance and was extended up to 27 months. No evidence was presented to show that the structure of the ECM analog was optimized to deliver maximal regeneration. The results of this study showed that two-thirds of the joints implanted with the ECM analog, twothirds of the joints which were autografted, and only 25% of the joints which were resected without further treatment showed regeneration of meniscal tissue. These results were interpreted to suggest that the ECM analog, or meniscus regeneration template (MRT), supported significant meniscal regeneration and provided enough biomechanical stability to minimize degenerative osteoarthritis in the canine knee joint. Recently, a clinical trial of the feasibility of MRT treatment was conducted on nine patients with either an irreparable tear of the meniscal cartilage or major loss of meniscal cartilage, and who remained in the study for at least 36 months. Following deletion of irreparably damaged meniscal tissue, MRT was implanted at the site of the missing meniscal tissue mass. The results showed that implantation of MRT induced regeneration of meniscal cartilage while the template was undergoing degradation. Compared to patients who were observed during the same period after meniscectomy (removal of damaged meniscal tissue without MRT treatment), patients who were implanted with MRT showed significant reduction in pain as well as greatly improved resumption of strenuous activity.

Defining Terms Autograft: The patient’s own tissue or organ, harvested from an intact area. Cell lifeline number, S: A dimensionless number that expresses the relative magnitudes of chemical reaction and diffusion. This number, defined as S in Eq. (113.2), can be used to compute the maximum path length lc over which a cell can migrate in a scaffold while depending on diffusion alone for transport of critical nutrients that it consumes. When the critical length is exceeded, the cell requires transport of nutrients by angiogenesis in order to survive. Cultured epithelia: A mature, keratinizing epidermis synthesized in vitro by culturing epithelial cells removed from the patient by biopsy. A relatively small skin biopsy (1 cm2) can be treated to yield an area larger by about 10,000 in 2–3 weeks and can then be grafted on patients. Dermal equivalent: A term which has been loosely used in the literature to describe a device that replaces, usually temporarily, the functions of the dermis following injury. Dermis: A 1–5 mm layer of connective tissue populated with quiescent fibroblasts which lies underneath the epidermis. It is separated from the former by a very thin basement membrane. The dermis of adult mammals does not regenerate spontaneously following injury. Dermis regeneration template (DRT): A graft copolymer of type I collagen and chondroitin 6-sulfate, average pore diameter 20–125 µm, degrading in vivo to an extent of about 50% in 2 weeks, which induces partial regeneration of the dermis in wounds from which the dermis has been fully excised. When seeded with keratinocytes prior to grafting, this analog of extracellular matrix has induced simultaneous synthesis both of a dermis and an epidermis. ECM analog: A model of extracellular matrix, consisting of a highly porous graft copolymer of collagen and a glycosaminoglycan.


Epidermis: The cellular outer layer of skin, about 0.1 mm thick, which protects against moisture loss and against infection. An epidermal graft, e.g., cultured epithelium or a thin graft removed surgically, requires a dermal substrate for adherence onto the wound bed. The epidermis regenerates spontaneously following injury, provided there is a dermal substrate underneath. Isomorphous tissue replacement: A term used to describe the synthesis of new, physiologic tissue within a regeneration template at a rate of the same order as the degradation rate of the template. This relation, described by Eq. (113.1), is the defining equation for a biodegradable scaffold which couples with, or interacts in this unique manner with, the inflammatory response of the wound bed. Meniscus regeneration template (MRT): A graft copolymer of type I collagen and an unspecified glycosaminoglycan, average pore diameter unspecified, which has induced partial regeneration of the knee meniscus in dogs following 80% excision of the meniscal tissue. Mikic number, Mi: Ratio of the characteristic freezing time of the aqueous medium of the collagenGAG suspension, tf , to the characteristic time for entry, te, of a container filled with the suspension which is lowered at constant velocity into a well-stirred cooling bath. Mi, defined by Eq. (113.5), can be used to design implants which have high alignment of pore channels along a particular axis, or no preferred alignment. Morphogenesis: The shaping of an organ during embryologic development or during wound healing, according to transcription of genetic information and local environmental conditions. Nerve regeneration template (NRT): A graft copolymer of type I collagen and chondroitin 6-sulfate, average pore diameter 5 mm, degrading in vivo to an extent of about 50% in 6 weeks, which has induced partial regeneration of the sciatic nerve of the rat across a 15-mm gap. Regeneration: The synthesis of new, physiologic tissue at the site of a tissue (one cell type) or organ (more than one cell type) which either has been lost due to injury or has failed due to a chronic condition. Regeneration template: A biodegradable device which, when attached to a missing organ, induces its regeneration. Scar: The end result of a repair process in skin and other organs. Scar is morphologically different from skin, in addition to being mechanically less extensible and weaker than skin. The skin regeneration template induces synthesis of nearly physiologic skin rather than scar. Scar is also formed at the site of severe nerve injury. Self-crosslinking: A procedure for reducing the biodegradation rate of collagen, in which collagen chains are covalently bonded to each other by a condensation reaction which is driven by drastic dehydration of the protein. This reaction illustrated by Eqs. (113.3a) and (113.3b), is also used to graft glycosaminoglycan chains to collagen chains without use of an extraneous crosslinking agent. Stereology: The discipline which relates the quantitative statistical properties of three-dimensional structures to those of their two-dimensional sections or projections. In reverse, stereologic procedures allow reconstruction of certain aspects of three-dimensional objects from a quantitative analysis of planar images. Its rules are used to determine features of the pore structure of ECM analogs.

References Archibald SJ, Krarup C, Sheffner J, Li S-T, Madison RD. 1991. A collagen-based nerve guide conduit for peripheral nerve repair: An electrophysiological study of nerve regeneration in rodents and nonhuman primates. J Comp Neurol 306:685–696. Archibald SJ, Sheffner J, Krarup C, Madison RD. 1995. Monkey median nerve repaired by nerve graft or collagen nerve guide tube. J Neurosci 15:4109–4123. Butler CE, Orgill DP, Yannas IV, Compton CC. 1998. Effect of keratinocyte seeding of collagen-glycosaminoglycan membranes on the regeneration of skin in a porcine model. Plast Reconstr Surg 101:1572–1579.


Chamberlain LJ, Yannas IV, Hsu H-P, Strichartz G, Spector M. 1998. Peripheral nerve regeneration through a collagen device is comparable with an autograft across 10-mm gaps after 60 weeks. Exp Neurol, in press. Chang AS, Yannas IV. 1992. Peripheral Nerve Regeneration. In B Smith, G Adelman (eds), Neuroscience Year (Supplement 2 to the Encyclopedia of Neuroscience), pp 125–126, Boston, Birkhauser. Compton CC, Butler CE, Yannas IV, Warland G, and Orgill DP. 1998. Organized skin structure is regenerated in vivo from collagen-GAG matrices seeded with autologous keratinocytes. J Invest Dermatol 110:908–916. Hansbrough JF, Boyce ST, Cooper ML, Foreman TJ. 1989. Burn wound closure with autologous keratinocytes and fibroblasts attached to a collagen-glycosaminoglycan substrate. JAMA, vol 262, pp 2125–2141. Hull BE, Sher SS, Rosen S, Church D, Bell E. 1983. Structural integration of skin equivalents grafted to Lewis and Sprague-Dawley rats. J Invest Derm, vol 8, pp 429–436. Orgill DP, Butler CE, Regan JF. 1996. Behavior of collagen-GAG matrices as dermal replacements in rodent and porcine models, Wounds 8:151–157. Stone KR, Rodkey WK, Webber RJ, McKinney L, Steadman JR. 1990. Collagen Based Prostheses for Meniscal Regeneration, Clin Orth, vol 252, pp 129–135. Stone KR, Steadman R, Rodkey WG, Li S-T. 1997. Regeneration of meniscal cartilage with use of a collagen scaffold. J Bone Joint Surg 79A:1770–1777. Yannas IV, Burke JF. 1980. Design of an artificial skin. I. Basic design principles. J Biomed Mater Res, vol 14, pp 65–81. Yannas IV, Burke JF, Orgill DP, Skrabut EM. 1982. Wound tissue can utilize a polymeric template to synthesize a functional extension of skin. Science 215:174–176. Yannas IV, Lee E, Orgill DP, Skrabut EM, Murphy GF. 1989. Synthesis and characterization of a model extracellular matrix that induces partial regeneration of adult mammalian skin. Proc Natl Acad Sci USA, vol 86, pp 933–937. Yannas IV. 1997. Models of organ regeneration processes induced by templates. In A Prokop, D Hunkeler, AD Cherrington (eds), Bioartificial Organs, Ann NY Acad Sci 831:280–293. Yannas IV. 1998. Studies on the biological activity of the dermal regeneration template. Wound Rep Regen, in press.

Further Information Chamberlain LJ, Yannas IV, Arrizabalaga A, Hsu H-P, Norregaard TV, Spector M. 1998. Early peripheral nerve healing in collagen and silicone tube implants. Myofibroblasts and the cellular response. Biomaterials, in press. Chamberlain LJ, Yannas IV, Hsu H-P, Spector M. 1997. Histological response to a fully degradable collagen device implanted in a gap in the rat sciatic nerve. Tissue Eng 3:353–362. Eldad A, Burt A, Clarke JA, Gusterson B. 1987. Cultured epithelium as a skin substitute. Burns 13:173–180. FR Noyes (eds), Biology and Biomechanics of the Traumatized Synovial Joint: The Knee as a Model, pp 221–231, AAOS. Hansbrough JF, Cooper ML, Cohen R, Spielvogel R, Greemnleaf G, Bartel RL, Naughton G. 1992. Evaluation of a biodegradable matrix containing cultured human fibroblasts as a dermal replacement beneath meshed skin grafts on athymic mice. Surg vol 111, pp 438–446. Heimbach D, Luterman A, Burke J, Cram A, Herndon D, Hunt J, Jordan M, McManus W, Solem L, Warden G, Zawacki B. 1988. Artificial dermis for major burns. Ann Surg vol 208, pp 313–320. Madison RD, Archibald SJ, Krarup C. 1992. Peripheral Nerve Injury. In IK Cohen, RF Diegelmann, WJ Lindblad (eds), Wound Healing, pp 450–487, Philadelphia, Saunders. Orgill DP, Yannas IV. 1998. Design of an artificial skin. IV. Use of island graft to isolate organ regeneration from scar synthesis and other processes leading to skin wound closure. J Biomed Mater Res 36:531–535.


Rodkey WG, Stone KR, Steadman JR. 1992. Prosthetic Meniscal Replacement. In GAM Finerman, FR Noyes (eds), Biology and Biomechanics of the Traumatized Synovial Joint: The Knee as a Model, pp 221–231, Amer Acad Orth Surg, Chicago, IL. Yannas IV, Burke JF, Gordon PL, Huang C. 1977. Multilayer membrane useful as synthetic skin. US Patent 4,060,081, Nov. 29. Yannas IV, Burke JF, Orgill DP, Burke JF. 1982. Regeneration of skin following closure of deep wounds with a biodegradable template. Trans Soc Biomat, vol 5, pp 1–38. Yannas IV. 1988. Regeneration of skin and nerves by use of collagen templates. In ME Nimni (ed), Collagen, vol 3, pp 87–115, Boca Raton, CRC Press. Yannas IV. 1989. Skin. Regeneration Templates. Encyclopedia of Polymer Science and Engineering, vol. 15, pp. 317–334. Yannas IV. 1990. Biologically active analogues of the extracellular matrix: Artificial skin and nerve. Angew Chemie Int Ed Engl, vol 29, pp 20–35.


Patrick, C. W., Sampath, R., McIntire, L. V. “Fluid Shear Stress Effects on Cellular Function.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

114 Fluid Shear Stress Effects on Cellular Function 114.1 114.2 114.3 114.4

Charles W. Patrick, Jr. Rice University

Rangarajan Sampath Rice University

114.5 114.6 114.7 114.8

Larry V. McIntire Rice University

114.9

Devices and Methodology Used for in Vitro Experiments Shear Stress-Mediated Cell-Endothelium Interactions Shear Stress Effects on Cell Morphology and Cytoskeletal Rearrangement Shear Stress Effects on Signal Transduction and Mass Transfer Shear Stress Effects on Endothelial Cell Metabolite Secretion Shear Stress Effects on Gene Regulation Mechanisms of Shear Stress-Induced Gene Regulation Gene Therapy and Tissue Engineering in Vascular Biology Conclusions

Cells of the vascular system are constantly exposed to mechanical (hemodynamic) forces due to the flow of blood. The forces generated in the vasculature include the frictional force or a fluid shear stress caused by blood flowing tangentially across the endothelium, a tensile stress caused by circumferential vessel wall deformations, and a net normal stress caused by a hydrodynamic pressure differential across the vessel wall. We will restrict our discussion to examining fluid shear stress modulation of vascular cell function. The endothelium is a biologically active monolayer of cells providing an interface between the flowing blood and tissues of the body. It can synthesize and secrete a myriad of vasoconstrictors, vasodilators, growth factors, fibrinolytic factors, cytokines, adhesion molecules, matrix proteins, and mitogens that modulate many physiologic processes, including wound healing, hemostasis, vascular remodeling, vascular tone, and immune and inflammatory responses. In addition to humoral stimuli, it is now well accepted that endothelial cell synthesis and secretion of bioactive molecules can be regulated by the hemodynamic forces generated by the local blood flow. These forces have been hypothesized to regulate neovascularization and the structure of the blood vessel [Hudlicka, 1984]. Clinical findings further show that arterial walls undergo an endothelium-dependent adaptive response to changes in blood flow, with blood vessels in high flow regions tending to enlarge and vessels in the low flow region having reduced lumen diameter, thereby maintaining a nearly constant shear stress at the vessel wall [Zarins et al., 1987]. In addition to playing an active role in the normal vascular biology, hemodynamic


FIGURE 114.1 Atherosclerotic plaques develop in regions of arteries where the flow rate (and resultant wall shear stress) is relatively low, which is often downstream from a vessel bifurcation, where there can be flow separation from the outer walls and regions of recirculation. These regions of low shear stress are pathologically prone to vessel wall thickening and thrombosis.

forces have also been implicated in the pathogenesis of a variety of vascular diseases. Atherosclerotic lesion-prone regions, characterized by the incorporation of Evans blue dye, enhanced accumulation of albumin, fibrinogen, and LDL, increased recruitment of monocytes, and increased endothelial turnover rates exhibit polygonal endothelial cell morphology typically seen in a low-shear environment, as opposed to nonlesion regions that have elongated endothelial cells characteristic of high-shear regions [Nerem, 1993]. In vivo studies of the distribution of atherosclerosis and the degree of intimal thickening have shown preferential plaque localization in low-shear regions. Atherosclerotic lesions were not seen in random locations but were instead found to be localized to regions of arterial branching and sharp curvature, where complex flow patterns develop, as shown in Fig. 114.1 [Asakura & Karino, 1990; Friedman et al., 1981; Gibson et al., 1993; Glagov et al., 1988; Ku et al., 1985; Levesque et al., 1989; Zarins et al., 1983]. Morphologically, intact endothelium over plaque surfaces showed variation in shape and size and loss of normal orientation, characteristic of low-shear conditions [Davies et al., 1988]. Flow in the vascular system is by and large laminar, but extremely complex time dependent flow patterns can develop in localized regions of complex geometry. Zarins and coworkers [1983] have shown that in the human carotid bifurcation, regions of moderate-to-high shear stress where flow remains unidirectional and axially aligned, were relatively free of intimal thickening, whereas extensive plaque localization was seen in the regions where low shear stress, fluid separation from the wall, and complex flow patterns were predominant. Asakura and Karino [1990] presented a technique for flow visualization by using transparent arteries where they directly correlated flow patterns to regions of plaque formation in human coronary arteries. They observed that the flow divider at the bifurcation point, a region of high shear stress, was relatively devoid of plaque deposition, whereas the outer wall, a region of low shear stress, showed extensive plaque formation. Similar patterns were also seen in curved vessels, where the inner wall or the hip region of the curve, a region of low shear and flow separation, exhibited plaque formation. In addition to the direct shear-stress-mediated effects, the regions of complex flow patterns and recirculation also tend to increase the residence time of circulating blood cells in susceptible regions, whereas the blood cells are rapidly cleared away from regions of high wall shear and unidirectional flow [Glagov et al., 1988]. This increased transit time could influence plaque deposition by favoring margination of


monocytes and platelets, release of vasoactive agents, or altered permeability at the intercellular junctions to extracellular lipid particles and possible concentration of procoagulant materials [Nollert et al., 1991]. The endothelium forming the interface between blood and the surrounding tissues is believed to act as a sensor of the local hemodynamic environment in mediating both the normal response and the pattern of vascular diseases. In vivo studies have shown changes in the actin microfilament localization in a shear-dependent manner. Actin stress fibres have been observed to be aligned with the direction of flow in high-shear regions, whereas they were mostly present in dense peripheral bands in the low-shear regions [Kim et al., 1989]. Langille and Adamson [1981] showed that the cells in large arteries, away from the branches, were aligned parallel to the long axes of the artery in rabbit and mouse. Similar results were shown in coronary arteries of patients undergoing cardiac transplantation [Davies et al., 1988]. Near the branch points of large arteries, however, cells aligned along the flow streamlines. In smaller blood vessels, where secondary flow patterns did not develop, the cell alignment followed the geometry of the blood vessel. In fact, endothelial cell morphology and orientation at the branch may be a natural marker of the detailed features of blood flow [Nerem et al., 1981]. Another cell type that is likely to be affected by hemodynamic forces is the vascular smooth muscle cell (SMC) present in the media. In the early stages of lesion development, SMCs proliferate and migrate into the intima [Ross, 1993; Schwartz et al., 1993]. Since the endothelium is intact in all but the final stages of atherosclerosis, SMCs are unlikely to be directly affected by shear stress. Most of the direct force they experience comes from the cyclical stretch forces experienced by the vessel itself due to pressure pulses [Grande et al., 1989]. Shear forces acting on the endothelium, however, can release compounds such as endothelin, nitric oxide, and platelet-derived growth factors that act as SMC mitogens and can modulate SMC tone. Ono and colleagues [1991] showed that homogenate of shear-stressed endothelial cells contained increased amounts of collagen and stimulated SMC migration in vitro, compared to endothelial cells grown under static conditions. The local shear environment could thus act directly or indirectly on the cells of the vascular wall in mediating a physiologic response. This chapter presents what is currently known regarding how the mechanical agonist of shear stress modulates endothelial cell function in the context of vascular physiology and pathophysiology. In discussing this topic we have adopted an outside-in approach. That is, we first discuss how shear stress affects endothelial cell-blood cell interactions, then progress to how shear stress affects the endothelial cell cytoskeleton, signal transduction, and protein secretion, and finally end with how shear stress modulates endothelial cell gene expression. We close with gene therapy and tissue engineering considerations related to how endothelial cells respond to shear stress. Before proceeding, however, we discuss the devices and methodology used for studying the endothelial cell responses to shear stress in vitro.

114.1 Devices and Methodology Used for in Vitro Experiments In vivo studies aimed at understanding cellular responses to shear forces have the inherent problem that they cannot quantitatively define the exact features of the hemodynamic environment. Moreover, it is very difficult to say if the resultant response is due to shear stress or some other feature associated with the hemodynamic environment. Cell culture studies and techniques for exposing cells to a controlled shear environment in vitro have been increasingly used to elucidate cellular responses to shear stress and flow. Mechanical-force-induced changes in cell function have been measured in vitro using mainly two cell types, cultured monolayers of bovine aortic endothelial cells (BAECs) and human umbilical vein endothelial cells (HUVECs). Shear stress is typically generated in vitro by flowing fluid across endothelial cell monolayers under controlled kinematic conditions, usually in the laminar flow regime. Parallel plate and cone-and-plate geometries have been the most common. Physiologic levels of venous and arterial shear stress range between 1–5 dynes/cm2 and 6–40 dynes/cm2, respectively. The use of the parallel plate flow chamber allows one to have a controlled and well-defined flow environment based on the chamber geometry (fixed) and the flow rate through the chamber (variable).


In addition, individual cells can be visualized in real time using video microscopy. Assuming parallel plate geometry and Newtonian fluid behavior, the wall shear stress on the cell monolayer in the flow chamber is calculated as

τw =

6 Qµ

(bh ) 2

(114.1)

where Q is the volumetric flow rate, µ is the viscosity of the flowing fluid, h is the channel height, b is the channel width, and τw is the wall shear stress. The flow chambers are designed such that the entrance length is very small compared to the effective length of the chamber [Frangos et al., 1985]. Therefore, entry effects can be neglected, and the flow is fully developed and parabolic over nearly the entire length of the flow chamber. Flow chambers usually consist of a machined block, a gasket whose thickness determines in part the channel depth, and a glass coverslip to which is attached a monolayer of endothelial cells (Fig. 114.2a). The individual components are held together either by a vacuum, or by evenly torqued screws, thereby ensuring a uniform channel depth. The flow chamber can have a myriad of entry ports, bubble ports, and exit ports. For short-term experiments, media are drawn through the chamber over the monolayer of cells using a syringe pump. For long-term experiments, the chamber is placed in a flow loop. In a flow loop (Fig. 114.2b), cells grown to confluence on glass slides can be exposed to a welldefined shear stress by recirculation of culture medium driven by gravity [Frangos et al., 1985]. Culture medium from the lower reservoir is pumped to the upper reservoir at a constant flow rate such that there is an overflow of excess medium back into the lower reservoir. This overflow serves two purposes: (1) It maintains a constant hydrostatic head between the two reservoirs, and (2) it prevents entry of air bubbles into the primary flow section upstream of the flow chamber that could be detrimental to the cells. The pH of the medium is maintained at physiologic levels by gassing with a humidified mixture of 95% air and 5% CO2. The rate of flow in the line supplying the chamber is determined solely by gravity and can be altered by changing the vertical separation between the two reservoirs. A sample port in the bottom reservoir allows periodic sampling of the flowing medium for a time-dependent assay. As mentioned above, cone-and-plate geometries can also be utilized. A schematic of a typical coneand-plate viscometer is shown in Fig. 114.2c. Shear stress is produced in the fluid contained between the rotating cone and the stationary plate. Cells grown on coverslips can be placed in the shear field (up to 12 at a time). For relatively small cone angles and low rates of rotation, the shear stress throughout the system is independent of position. The cone angle compensates for radial effects seen in plate-and-plate rheometers. The wall shear stress (τw ) on the cell monolayer in the cone-and-plate viscometer is calculated as

τw =

3T 2 π R3

(114.2)

where T is the applied torque and R is the cone radius. The flow becomes turbulent, however, at the plate’s edge and at high rotational speeds. Modifications from the basic design have allowed use of an optical system with the rheometer, enabling direct microscopic examination and real-time analysis of the cultured cells during exposure to shear stress [Dewey et al., 1981; Schnittler et al., 1993]. For a more complete description of in vitro device design and applications, refer to the text edited by Frangos [Transon-Tay, 1993].

114.2 Shear Stress-Mediated Cell-Endothelium Interactions Cell-cell and cell-substrate interactions in the vascular system are of importance in a number of physiologic and pathologic situations. Lymphocytes, platelets, or tumor cells in circulation may arrest at a


FIGURE 114.2 Devices used for in vitro study of shear stress effects on vascular endothelial cells. (a) Parallel plate flow chamber, (b) flow loop, (c) cone-and-plate viscometer.

particular site as a result of interaction with the endothelium or the subendothelial matrix. While the margination of leukocyte/lymphocyte to the vessel wall be a normal physiologic response to injury or inflammation, adhesion of blood platelets to the subendothelium and subsequent platelet aggression could result in a partial or complete occlusion of the blood vessel leading to thrombosis or stroke. In addition, the adhesion of tumor cells to the endothelium is often the initial step in the development of secondary metastases. The adhesion of leukocytes, platelets, and tumor cells is not only mediated by adhesion molecules on the endothelium but also mediated by the hemodynamic force environment present in the vasculature. In fact, the specific molecular mechanisms employed for adhesion often vary with the local wall shear stress [Alevriadou et al., 1993; Lawrence et al., 1990].


Targeting of circulating leukocytes to particular regions of the body is an aspect of immune system function that is currently of great research interest. This targeting process consists of adhesion of a specific subpopulation of circulating leukocytes to a specific area of vascular endothelium via cell surface adhesion receptors. The large number of receptors involved and the differential regulation of their expression on particular cell subpopulations make this process very versatile but also quite complicated. An extremely important additional complication arises from the fact that these interactions occur within the flowing bloodstream. Study of these various types of adhesive interactions requires accurate recreation of the flow conditions experienced by leukocytes and endothelial cells. Lawrence and coworkers examined neutrophil adhesion to cytokine-stimulated endothelial cells under well-defined postcapillary venular flow conditions in vitro [Lawrence et al., 1987; Lawrence et al., 1990; Lawrence & Springer, 1991]. They also demonstrated that under flow conditions neutrophil adhesion to cytokine-stimulated endothelial cells is mediated almost exclusively by CD18-independent mechanisms but that subsequent neutrophil migration is CD18-dependent [Lawrence & Springer, 1991; Smith et al., 1991]. The initial flow studies were followed by many further studies both in vitro [Abbassi et al., 1991, 1993; Anderson et al., 1991; Hakkert et al., 1991; Jones et al., 1993; Kishimoto et al., 1991] and in vivo [Ley et al., 1991; Perry & Granger, 1991; von Andrian et al., 1991; Watson et al., 1991] which clearly distinguish separate mechanisms for initial adhesion/rolling and firm adhesion/leukocyte migration. Research has further shown that in a variety of systems, selectin/carbohydrate interactions are primarily responsible for initial adhesion and rolling, and firm adhesion and leukocyte migration are mediated primarily by integrin/peptide interactions. Methodology discussed for studying receptor specificity of adhesion for leukocytes under flow can also be utilized for studying red cell–endothelial cell interactions [Barabino et al., 1987; Wick et al., 1987, 1993]. The interaction of tumor cells with endothelial cells is an important step in tumor metastasis. To adhere to the vessel wall, tumor cells that come into contact with the microvasculature must resist the tractive force of shear stress that tends to detach them from the vessel wall [Weiss, 1992]. Hence, studies of the mechanisms involved in the process of tumor metastasis must take into account the physical forces acting on the tumor cells. Bastida and coworkers [1989] have demonstrated that tumor cell adhesion depends not only on tumor cell characteristics and endothelial cell adhesion molecule expression but on shear stress in the interaction of circulating tumor cells with the endothelium. The influence of shear stress on tumor cell adhesion suggests that attachment of tumor cells to vascular structures occurs in areas of high shear stress, such as the microvasculature [Bastida et al., 1989]. This is supported by earlier pathologic observations that indicate preferential attachment of tumor cells on the lung and liver capillary system [Warren, 1973]. Some tumor cell types roll and subsequently adhere on endothelial cells using selectin-mediated mechanisms similar to leukocytes, whereas others adhere without rolling using integrin-mediated receptors on endothelial cells [Bastida et al., 1989; Giavazzi et al., 1993; Kojima et al., 1992; Menter et al., 1992; Patton et al., 1993; Pili et al., 1993]. It has been postulated that some tumor cell types undergo a stabilization process prior to firm adhesion that is mediated by transglutaminase crosslinking [Menter et al., 1992; Patton et al., 1993]. Recently, Pili and colleagues [1993] demonstrated that tumor cell contact with endothelial cells increases Ca2+ release from endothelial cell intracellular stores, which may have a fundamental role in enhancing cell-cell adhesion. The factors and molecular mechanisms underlying tumor cell and endothelial cell interactions remain largely undefined and are certain to be further explored in the future. The endothelium provides a natural nonthrombogenic barrier between circulating platelets and the endothelial basement membrane. However, there are pathologic instances in which the endothelium integrity is compromised, exposing a thrombogenic surface. Arterial thrombosis is the leading cause of death in the United States. Among the most likely possibilities for the initiation of arterial platelet thrombi are: (1) adhesion of blood platelets onto the subendothelium of injured arteries and arterioles or on ruptured atherosclerotic plaque surfaces, containing collagen and other matrix proteins, with subsequent platelet aggregation, or (2) shear-induced aggregation of platelets in areas of the arterial circulation partially constricted by atherosclerosis or vasospasm. Various experimental models have been developed


to investigate the molecular mechanisms of platelet attachment to surfaces (adhesion) and platelet cohesion to each other (aggregation) under shear stress conditions. Whole-blood perfusion studies using annular or parallel-plate perfusion chambers simulate the first proposed mechanism of arterial thrombus formation in vivo, that may occur as a result of adhesion on an injured, exposed subendothelial or atherosclerotic plaque surface [Hubbel & McIntire, 1986; Weiss et al., 1978, 1986]. Platelets arriving subsequently could, under the right conditions, adhere to each other, forming aggregates large enough to partially or completely occlude the blood vessel. In vitro studies have shown an increase in thrombus growth with local shear rate, an event that is believed to be the result of enhanced arrival and cohesion of platelets near the surface at higher wall shear stresses [Turitto et al., 1987]. Video microscopy provides information on the morphologic characteristics of thrombi and enables the reconstruction of threedimensional models of thrombi formed on surfaces coated with biomaterials or endothelial cell basement membrane proteins. Macroscopic analysis of thrombi can provide information on platelet mass transport and reaction kinetics with the surface, and microscopic analysis allows dynamic real-time study of cellsurface and intercellular interactions. Such a technology enables the study of key proteins involved in mural thrombus formation, such as vWF [Alevriadou et al., 1993; Folie & McIntire, 1989]. In addition, tests of antithrombotic agents and investigation of the thrombogenicity of various purified components of the vessel wall or polymeric biomaterials can be performed.

114.3 Shear Stress Effects on Cell Morphology and Cytoskeletal Rearrangement In addition to mediating cell–endothelial cell interactions, shear stress can act directly on the endothelium. It has been demonstrated for almost a decade that hemodynamic forces can modulate endothelial cell morphology and structure [Barbee et al., 1994; Coan et al., 1993; Eskin et al., 1984; Franke et al., 1984; Girard & Nerem, 1991, 1993; Ives et al., 1986; Langille et al., 1991; Levesque et al., 1989; Wechezak et al., 1985]. Conceivably, the cytoskeletal reorganization that occurs in endothelial cells several hours after exposure to flow may, in conjunction with shape change, transduce mechanical signals to cytosolic and nuclear signals (mechanotransduction), thereby playing a role in gene expression and signal transduction [Ingber, 1993; Resnick et al., 1993; Watson, 1991]. In fact, investigators have shown specific gene expression related to cytoskeletal changes [Botteri et al., 1990; Breathnach et al., 1987; Ferrua et al., 1990; Werb et al., 1986]. F-actin has been implicated as the principal transmission element in the cytoskeleton and appears to be required for signal transduction [Watson, 1991]. Actin filaments are anchored in the plasma membrane at several sites, including focal adhesions on the basal membrane, intercellular adhesion proteins, integral membrane proteins at the apical surface, and the nuclear membrane [Davies & Tripathi, 1993]. Substantiating this, Barbee and coworkers [1994] have recently shown, utilizing atomic force microscopy, that F-actin fiber stress bundles are formed in the presence of fluid flow and the fibers are coupled to the apical membrane. Moreover, endothelial cell shape change and realignment with flow can be inhibited by drugs that interfere with microfilament turnover [Davies & Tripathi, 1993]. Although all evidence leads one to believe that the endothelial cell cytoskeleton can respond to flow and that F-actin is involved, the exact mechanism involved remains to be elucidated.

114.4 Shear Stress Effects on Signal Transduction and Mass Transfer Shear stress and resultant convective mass transfer are known to affect many important cytosolic second messengers in endothelial cells. For instance, shear stress is known to cause ATP-mediated increases in calcium ions (Ca2+) [Ando et al., 1988; Dull & Davies, 1991; Mo et al., 1991; Nollert & McIntire, 1992]. Changes in flow influence the endothelial boundary layer concentration of ATP by altering the convective transport of exogenous ATP, thereby altering ATP’s interaction with both the P2y-purinoreceptor and


ecto-ATPase. At low levels of shear stress, degradation of ATP by ecto-ATPase exceeds the rate of diffusion, and the steady state concentration of ATP remains low. In contrast, at high levels of shear stress, convection enhances the delivery of ATP from upstream to the P2y-purinoreceptor, and diffusion exceeds the rate of degradation by the ecto-ATPase [Mo et al., 1991]. Whether physiologic levels of shear stress can directly increase intracellular calcium remains unclear. Preceding the calcium increases are increases in inositol1,4,5 trisphosphate (IP3) [Bhagyalakshmi et al., 1992; Bhagyalakshmi & Frangos, 1989b; Nollert et al., 1990; Prasad et al., 1993], which binds to specific sites on the endoplasmic reticulum and causes release of Ca2+ from intracellular stores, and diacylglycerol (DAG) [Bhagyalakshmi & Frangos, 1989a]. Elevated levels of both DAG and Ca2+ can activate several protein kinases, including protein kinase C (PKC). Changes in Ca2+, IP3, and DAG are evidence that fluid shear stress activated the PKC pathway. In fact, the PKC pathway has been demonstrated to be activated by shear stress [Bhagyalakshmi & Frangos, 1989b; Hsieh et al., 1992, 1993; Kuchan & Frangos, 1993; Milner et al., 1992]. In addition to the PKC pathway, pathways involving cyclic adenosine monophosphate (cAMP) and cyclic guanosine monophosphate (cGMP) may also be modulated by shear stress, as evidenced by increases in cAMP [Bhagyalakshmi & Frangos, 1989b] and cGMP [Kuchan & Frangos, 1993; Ohno et al., 1993] with shear stress. Moreover, it has been shown recently that shear stress causes acidification of cytoslic pH [Ziegelstein et al., 1992]. Intracellular acidification of vascular endothelial cells releases Ca2+ into the cytosol [Ziegelstein et al., 1993]. This Ca2+ mobilization may be linked to endothelial synthesis and release of vasodilatory substances during the pathological condition of acidosis.

114.5 Shear Stress Effects on Endothelial Cell Metabolite Secretion The application of shear stress in vitro is accompanied by alterations in protein synthesis that are detectable within several hours after initiation of the mechanical agonist. Shear stress may regulate the expression of fibrinolytic proteins by endothelial cells; tPA is an antithrombotic glycoprotein and serine protease that is released from endothelial cells. Once released, it rapidly converts plasminogen to plasmin which, in turn, dissolves fibrin clots. Diamond and coworkers [1989] have shown that venous levels of shear stress do not affect tPA secretion, whereas arterial levels increase the secretion rate of tPA. Arterial flows lead to a profibrinolytic state that would be beneficial in maintaining a clot-free artery. In contrast to tPA, neither venous nor arterial levels of shear stress cause significant changes in plasminogen activator inhibitor-1 (PAI-1) secretion rates [Diamond et al., 1989; Kuchan & Frangos, 1993]. Increased proliferation of smooth muscle cells is one of the early events of arteriosclerosis. The secretion rate of endothelin-1 (ET-1), a potent smooth muscle cell mitogen, has been shown to increase in response to venous levels of shear stress and decrease in response to arterial levels of shear stress [Kuchan & Frangos, 1993; Milner et al., 1992; Nollert et al., 1991; Sharefkin et al., 1991; Yoshizumi et al., 1989]. Both the in vitro tPA and ET-1 results are consistent with in vivo observations that atherosclerotic plaque development occurs in low shear stress regions as opposed to high shear stress regions. The low shear stress regions usually occur downstream from vessel branches. In these regions we would expect low secretion of tPA and high secretion of ET-1, leading to locally increased smooth muscle cell proliferation (intimal thickening of the vessel) and periodic problems with clot formation (thrombosis). Both of these observations are observed pathologically and are important processes in atherogenesis [McIntire, 1994]. Important metabolites in the arachidonic acid cascade via the cyclooxygenase pathway; prostacylin (PGI2) and prostaglandin (PGF2a) are also known to increase their secretion rates in response to arterial levels of shear stress [Frangos et al., 1985; Grabowski et al., 1985; Nollert et al., 1989]. In addition, endothelial cells have been shown to increase their production of fibronectin (FN) in response to shear stress [Gupte & Frangos, 1990]. Endothelial cells may respond to shear stress by secreting FN in order to increase their attachment to the extracellular matrix, thereby resisting the applied fluid shear stress. Endothelial cells have been shown to modulate their receptor expression of intercellular adhesion molecule-1 (ICAM-1), vascular cell adhesion molecule-1 (VCAM-1), and monocyte chemotactic peptide-1


(MCP-1) in response to shear stress [Sampath et al., 1995; Shyy et al., 1994]. The adaptive expression of these adhesion molecules in response to hemodynamic shear stress may aid in modulating specific adhesion localities for neutrophils, leukocytes, and monocytes.

114.6 Shear Stress Effects on Gene Regulation In many cases, the alterations in protein synthesis observed with shear stress are preceded by modulation of protein gene expression. The molecular mechanisms by which mechanical agonists alter the gene expression of proteins are under current investigation. Diamond and colleagues [1990] and Nollert and colleagues [1992] have demonstrated that arterial levels of shear stress upregulate the transcription of tPA mRNA, concomitant with tPA protein secretion. The gene expressions of ET-1 and PDGF, both of which are potent smooth muscle cell mitogens and vasoconstrictors, are also known to be modulated by hemodynamic forces. For instance, arterial levels of shear stress cause ET-1 mRNA to be downregulated [Malek et al., 1993; Malek & Izumo, 1992; Sharefkin et al., 1991]. ET-1 mRNA downregulation is sensitive to the magnitude of the shear stress in a dose-dependent fashion, reaching a saturation at 15 dynes/cm2. Conversely, Yoshizumi and coworkers [1989] have shown that venous levels of shear stress cause a transient increase in ET-1 mRNA, peaking at 4 hours and returning to basal levels by 12–24 hours. However, they had previously reported downregulation of ET-1 mRNA expression [Yanagisawa et al., 1988]. PDGF is expressed as a dimer composed of PDGF-A and PDGF-B subunits. There are conflicting reports as to how arterial shear stresses affect PDGF-B mRNA expression. Malek and colleagues [1993] have shown that arterial levels of shear stress applied to BAECs cause a significant decrease in PDGF-B mRNA expression over a 9-hr period. In contrast, Mitsumata and coworkers [1993] and Resnick and coworkers [1993] have shown increases in BAEC mRNA expression of PDGF-B when arterial shear stresses were applied. The discrepancy in the results may be attributed to differences in BAEC cell line origins or differences in the passage number of cells used. In support of the latter two investigators, Hsieh and colleagues [1992] have reported upregulation of PDGF-B mRNA in HUVECs in the presence of arterial shear stress. As with the B chain of PDGF, there are conflicting reports as to the affect of arterial shear stress on PDGF-A. Mitsumata and coworkers [1993] have reported no change in PDGF-A mRNA expression, whereas Hsieh and coworkers [1991, 1992] have reported a shear-dependent increase in the mRNA expression from 0–6 dynes/cm2 which then plateaus from 6–51 dynes/cm2. Endothelial cell expression of ET-1 and PDGF may be important in blood vessel remodeling. In blood vessels exposed to increased flow, the chronic vasculature response is an increase in vessel diameter [Langille & O’Donnell, 1986]. Hence, it is tempting to postulate that hemodynamic-force-modulated alterations in ET-1 and PDGF mRNA expression may account for much of this adaptive change. The gene expression of various adhesion molecules involved in leukocyte recruitment during inflammation and disease has also been investigated. ICAM-1 mRNA has been shown to be transiently upregulated in the presence of venous or arterial shear stresses [Nagel et al., 1994; Sampath et al., 1995]. Its time-dependent response peaked at 1–3 hr following exposure to shear, before declining below basal levels with prolonged exposure of 6–24 hr. In contrast, VCAM-1 mRNA level was downregulated almost immediately upon onset of flow and was found to drop significantly below basal levels within 6 hr of initiation of flow at all magnitudes of shear stresses. E-selectin mRNA expression appeared to be generally less responsive to shear stress, especially at the lower magnitudes. After an initial downward trend 1 hr following exposure to shear stress (2 dynes/cm2), E-selectin mRNA remained at stable levels for up to 6 hr [Sampath et al., 1995]. Recent evidence shows that the expression of MCP-1, a monocyte-specific chemoattractant expressed on endothelial cells, also follows a similar biphasic response with shear stress like ICAM-1 [Shyy et al., 1994]. In addition to adhesion molecules, the gene expression of several growth factors has been investigated. The mRNA expression of heparin-binding epidermal growth factor like growth factor (HB-EGF), a smooth muscle-cell mitogen, transiently increases in the presence of minimal arterial shear stress [Morita et al., 1993]. Transforming growth factor-β1 (TGF-β1) mRNA has been reported to be upregulated in the presence of arterial shear stresses within 2 hr and remain elevated for


12 hr [Ohno et al., 1992]. In addition, Malek and coworkers [1993] have shown that basic fibroblast growth factor (bFGF) mRNA is upregulated in BAECs in the presence of arterial shear stresses. In HUVECs, however, no significant changes in bFGF message were observed [Diamond et al., 1990]. This difference is probably due to differences in cell source (human versus bovine). Proto-oncogenes code for binding proteins that either enhance or repress transcription and, therefore, are ideal candidates to act as gene regulators [Cooper, 1990]. Komuro and colleagues [1990, 1991] were the first to demonstrate that mechanical loading causes upregulation of c-fos expression. Recently, Hsieh and coworkers [1993] investigated the role of arterial shear stress on the mRNA levels of nuclear protooncogenes c-fos, c-jun, and c-myc. Gene expression of c-fos was transiently upregulated, peaking at 0.5 hr and returning to basal levels within an hour. In contrast, both c-jun and c-myc mRNA were upregulated to sustainable levels within an hour. The transcribed protein products of c-fos, c-jun, and c-myc may act as nuclear-signaling molecules for mechanically induced gene modulation [Nollert et al., 1992; Ranjan & Diamond, 1993].

114.7 Mechanisms of Shear Stress-Induced Gene Regulation Although no unified scheme to explain mechanical signal transduction and modulation of gene expression is yet possible, many studies provide insight in an attempt to elucidate which second messengers are involved in gene regulation mediated by hemodynamic forces. There is substantial evidence that the PKC pathway may be involved in the gene regulation. A model of the PKC transduction pathway is shown in Fig. 114.3. Mitsumata and colleagues [1993] have shown that PDGF mRNA modulation by shear stress could be partially attributed to a PKC pathway. Hsieh and coworkers [1992] have shown that shearinduced PDGF gene expression in HUVECs is mainly mediated by PKC activation and requires Ca2+ and the involvement of G proteins. In addition, they demonstrated that cAMP and cGMP dependent protein kinases are not involved in PDGF gene expression. Morita and colleagues [1993] have shown that shearinduced HB-EGF mRNA expression is mediated through a PKC pathway and that Ca2+ may be involved in the pathway. PKC was also found to be an important mediator in flow-induced c-fos expression, with the additional involvement of G proteins, phospholipase C (PLC), and Ca2+ [Hsieh et al., 1993]. Moreover, Levin and coworkers [1988] have shown that tPA gene expression is enhanced by PKC activation and Iba and coworkers [1992] have shown that cAMP is not involved in tPA gene expression. As depicted in Fig. 114.3, the PKC pathway may be involved in activating DNA-binding proteins via phosphorylation. In addition to second messengers, it has been proposed that cytoskeletal reorganization may play a role in regulating gene expression [Ingber, 1993; Resnick et al., 1993]. Morita and colleagues [1993] have shown that shear stress-induced ET-1 gene expression is mediated by the disruption of the actin cytoskeleton and that microtubule integrity is also involved. Gene expression of other bioactive molecules may also be regulated by actin disruption. The actual molecular mechanisms involved in cytoskeletal-mediated gene expression remain unclear. However, the cytoskeleton may activate membrane ion channels and nuclear pores. In addition to second messengers and the cytoskeletal architecture, it has been postulated by Nollert and others [1992] that transcriptional factors that bind to the DNA may play an active role mediating the signal transduction between the cytosol and nucleus (Fig. 114.4). It is known that nuclear translocation of transcriptional factors and DNA-binding activity of the factors can be mediated by phosphorylation [Bohmann, 1990; Hunter & Karin, 1992]. Many of the transcription factors are protein products of proto-oncogenes. It has previously been stated that shear stress increases expression of c-fos and c-jun. The gene products of c-fos and c-jun form protein dimers that bind to transcriptional sites on DNA promoters and act as either transcriptional activators or repressors. Two of the transcriptional sites to which the fos and jun family dimers bind are the TRE (tumor promoting agent response element) and CRE (cAMP response element). These have consensus sequences of TGACTCA and TGACGTCA, respectively. It is known that the promoter regions of tPA, PAI-1, ET, and TGF-β1 possess sequences of homology to the CRE and TRE [Nollert et al., 1992]. In addition to mediating known transcription factor binding


FIGURE 114.3 Model for shear stress induced protein kinase C (PKC) activation leading to modulation of gene expression. Mechanical agonists enhance membrane phosphoinositide turnover via phospholipase C (PLC), producing inositol-1,4,5 trisphosphate (IP3), and diacylglycerol (DAG). DAG can then activate PKC. IP3 may also activate PKC via Ca2+ release. PKC phosphorylates DNA-binding proteins, thereby making them active. The nuclear binding proteins then alter mRNA transcription. R = receptor, G = G protein, PIP2 = phosphatidylinositol 4,5-bisphosphate, (+) = activation.

in perhaps novel ways, mechanical perturbations may initiate molecular signaling through specific stressor strain-sensitive promoters that can activate gene expression. In fact, Resnick and colleagues [1993] have described a cis-acting shear stress response element (SSRE) in the PDGF-B promoter, and this sequence is also found in the promoters of tPA, ICAM-1, TGF-β1, c-fos, and c-jun but not in VCAM-1 or E-Selectin promoters. A core-binding sequence of GAGACC was identified that binds to transcriptional factors found in BAEC nuclear extracts. The identify of transcriptional factors that bind this sequence remains unknown. In addition, Malek and coworkers [1993] have shown that shear-stress mediated ET-1 mRNA expression is not dependent on either the PKC or cAMP pathways, but rather shear stress regulates the transcription of the ET-1 gene via an upstream cis element in its promoter. It remains to be seen if © 2000 by CRC Press LLC

FIGURE 114.4 Gene expression regulated by transcription/translation factors and mechanically sensitive promoters. (a) Phosphorylation of transcription factors allows their translocation to the nucleus and subsequent DNA binding. The transcription factors can activate or inhibit transcription. TRE = tumor promoting agent response element, TREB = tumor promoting agent response element binding protein, CRE = cAMP response element, CREB = cAMP response element binding protein, SSRE = shear sensitive response element. (b) The AU-binding factor binds the AUUUA motif located in the 3′ untranslated regions of some mature mRNA and could play an important role in mediating stability and/or transport of these mRNAs; AUBF = AU binding factor.

ET-1 is shown to have the same response element as that proposed by Resnick and colleagues. As mentioned previously, Mitsumata and coworkers [1993] found that PDGF mRNA modulation was not solely dependent on a PKC pathway. The PKC independent mechanism may very well involve the cisacting shear stress response element that Resnick and colleagues described, though other factors are probably important in controlling the complex temporal pattern of gene expression seen in response to mechanical stimuli. The biphasic response of MCP-1 was shown to be regulated at the transcriptional level, possibly involving activation of AP-1, sites and the subsequent binding to the TRE site. The putative SSRE binding site was also identified in the 5′ flanking region of the cloned MCP-1 gene suggesting an interactive mechanism, wherein these cis-acting elements collectively regulate the transcriptional activation of MCP-1 gene under shear stress [Shyy et al., 1994].


Another important determinant in the role played by shear stress on gene regulation could be at the level of mRNA stability. It is not known whether the shear-induced downregulation that has been reported, especially in the case of adhesion molecules, is due to a decrease in the transcriptional rate of the gene or due to a decrease in the stability of the transcript. The β-adrenergic receptors, which are part of a large class of G-protein linked cell surface receptors, are also susceptible to desensitization and downregulation [Hadcock & Malbon, 1988]. A study of the molecular basis for agonist-induced instability revealed the existence of an Mr 35,000 protein that specifically binds mRNA transcripts which contain the sequence AUUUA in their 3′ untranslated region (3′-UTR) [Port et al., 1992]. Agonists that decrease the β-adrenergic RNA transcript levels increased the level of this protein. Further, it has been shown that tandem repeats (three or more) of this sequence are needed for efficient binding of the destabilizer protein [Bohjanen et al., 1992]. Recent reports, however, suggest that AUUUA motif alone does not destabilize the mRNA transcripts and that the presence of additional elements in the AU-rich 3′-UTR may be necessary for the observed mRNA instability [Peppel et al., 1991]. Muller and others [1992] also report that binding of certain AU-binding factors (AUBF) to AUUUA motif accelerates the nuclear export of these mRNAs (Fig. 114.4). Further, their study indicates that this AUBF may protect AU-rich mRNAs against degradation by endoribonuclease V, by binding AUUUA motifs, thereby stabilizing the otherwise unstable mRNA. Analysis of the 3′-UTR of some of the adhesion molecules studied revealed the presence of the AUUUA motif in all three cases. Human VCAM-1 [Cybulsky et al., 1991] has 6 dispersed AUUUA repeats in its 3′-UTR. Human ICAM-1 [Staunton et al., 1988] has 3 AUUUA repeats within 10 bases of each other, and 3-selectin [Hession et al., 1990] has 8 AUUUA repeats, which include 1 tandem repeat and 3 others that are within 10 bases of each other. The constitutively expressed gene human GAPDH [Tokunaga et al., 1987] does not have any AUUUA repeats in the 3′ region analyzed. Of the other shear responsive genes, human PDGF-B (c-sis) gene has 1 AUUUA repeat [Ratner et al., 1985]; human endothelin gene [Bloch et al., 1989] has 4, three within 10 bases of each other; human tissue plasminogen activator (tPA) [Reddy et al., 1987] has 2; and human thrombomodulin [Jackman et al., 1987] has 9 AUUUA boxes, including 1 tandem repeat. The steady state levels of the tPA and PDGF have been shown to be increased, and those of the adhesion molecules and endothelin have been shown to be decreased by arterial shear stress [Diamond et al., 1990; Hsieh et al., 1992; Sampath et al., 1995], suggesting a possible mechanism of shear-induced destabilization of adhesion molecule mRNA. However, the observed changes can be at the transcriptional level, with the AUUUA motifs playing a role only in the transport of these mRNAs to the cytoplasm. As shown by Muller and colleagues [1992], transport of low-abundance mRNAs like cytokine mRNAs, whose cellular concentrations are one-hundredth or less of that of the high abundance mRNAs, are the ones that are prone to modulation by transport-stimulatory proteins such as AUBF. This can explain the presence of AUUUA boxes in the 3′-UTR of adhesion molecules and other modulators of vascular tone and not in that of the constitutively expressed GADPH, which is an essential enzyme in the metabolic pathway and is expressed in high copy numbers. A more detailed study using transcription run-on assays is needed to address these specific issues.

114.8 Gene Therapy and Tissue Engineering in Vascular Biology Endothelial cells, located adjacent to the flowing blood stream, are ideally located for use as vehicles for gene therapy, since natural or recombinant proteins of therapeutic value can be expressed directly into the blood to manage cardiovascular diseases or inhibit vascular pathology. One could drive gene expression only in regions of vasculature where desired by using novel cis-acting stress- or strain-sensitive promoter elements or stress-activated transcription factors. For instance, endothelial cells in regions of low shear stress could be modified to express tPA so as to inhibit atherosclerotic plaque formation. An appropriate vector driven by the ET-1 promoter (active at low stress but downregulated at high stress) attached to the tPA gene would be a first logical construct in this application. Likewise, endothelial cells could be modified to increase proliferation rates in order to endothelialize vascular grafts or other vascular prostheses, thereby inhibiting thrombosis. In addition, endothelial cells could be modified to decrease expression of smooth muscle cell © 2000 by CRC Press LLC

mitogens (ET-1, PDGF) to prevent intimal thickening and restenosis. The endothelial cells could even be modified so as to secrete components that would inactivate toxic substances in the blood stream. Work has already begun to develop techniques to express foreign proteins in vivo [Nabel et al., 1989; Wilson et al., 1989; Zwiebel et al., 1989]. Vical and Gen Vec are currently examining ways to express growth factors in coronary arteries to stimulate angiogenesis following balloon angioplasty [Glaser, 1994]. In addition, Tularik Inc is modifying endothelial cells to over express low density lipoprotein (LDL) receptors to remove cholesterol from the blood stream so as to prevent atherosclerotic plaque formation. The application of gene therapy to cardiovascular diseases is in its infancy and will continue to grow as we learn more about the mechanisms governing endothelial cell gene expression. Since endothelial cells lie in a mechanically active environment, predicting local secretion rates of gene products from transfected endothelial cells will require a knowledge of how these mechanical signals modulate gene expression for each target gene and promoter [McIntire, 1994]. In addition to gene therapy applications, there are tissue engineering applications that can be realized once one gains a fundamental understanding of the function-structure relationship intrinsic to vascular biology. This includes understanding how hemodynamic forces modulate endothelial cell function and morphology. Of primary concern is the development of an artificial blood vessel for use in the bypass and replacement of diseased or occluded arteries [Jones, 1982; Nerem, 1991; Weinberg & Bell, 1986]. This is particularly important in the case of small-diameter vascular grafts (such as coronary bypass grafts), which are highly prone to reocclusion. The synthetic blood vessels must provide the structural support required in its mechanically active environment as well as provide endothelial-like functions, such as generating a nonthrombogenic surface.

114.9 Conclusions This chapter has demonstrated the intricate interweaving of fluid mechanics and convective mass transfer with cell metabolism and the molecular mechanisms of cell adhesion that occur continuously in the vascular system. Our understanding of these mechanisms and how they are modulated by shear stress is really in the initial stages—but this knowledge is vital to our understanding of thrombosis, atherosclerosis, inflammation, and many other aspects of vascular physiology and pathophysiology. Knowledge of the fundamental cellular and molecular mechanisms involved in adhesion and mechanical force modulation of metabolism under conditions that mimic those seen in vivo is essential for real progress to be made in vascular biology and more generally in tissue engineering.

Defining Terms CRE: cAMP response element; its consensus sequence TGACGTCA is found on genes responsive to cAMP agonists such as forskolin. Gene regulation: Transcriptional and posttranscriptional control of expression of genes in eukaryotes where regulatory proteins bind specific DNA sequences to turn a gene either on (positive control) or off (negative control). Gene therapy: The modification or replacement of a defective or malfunctioning gene with one that functions adequately and properly, for instance, addition of gene regulatory elements such as specific stress- or strain-sensitive response elements to specifically drive gene expression only in regions of interest in the vasculature so as to control proliferation, fibrinolytic capacity, etc. Shear: Shear refers to the relative parallel motion between adjacent fluid (blood) planes during flow. The difference in the velocity between adjacent layers of blood at various distances from the vessel wall determines the local shear rate, expressed in cm/s/cm, or s–1. Shear stress: Fluid shear stress, expressed in dynes/cm2, is a measure of the force required to produce a certain rate of flow of a viscous liquid and is proportional to the product of shear rate and blood viscosity. Physiologic levels of venous and arterial shear stresses range between 1–5 dynes/cm2 and 6–40 dynes/cm2, respectively.


SSRE: Shear stress response element; its consensus GAGACC has been recently identified in genes responsive to shear stress. Transcription factor: A protein that binds to a cis-regulatory element in the promoter region of a DNA and thereby directly or indirectly affects the initiation of its transcription to an RNA. TRE: Tumor promoting agent response element: its consensus sequence TGACTCA is commonly located in genes sensitive to phorbol ester stimulation.

References Abbassi O, Kishimoto TK, McIntire LV, et al. 1993. E-selectin supports neutrophil rolling in vitro under conditions of flow. J Clin Invest 92:2719. Abbassi OA, Lane CL, Krater S, et al. 1991. Canine neutrophil margination mediated by lectin adhesion molecule-1 in vitro. J Immunol 147:2107. Alevriadou BR, Moake JL, Turner NA, et al. 1993. Real-time analysis of shear-dependent thrombus formation and its blockade by inhibitors of von Willebrand factor binding to platelets. Blood 81:1263. Anderson DC, Abbassi OA, McIntire LV, et al. 1991. Diminished LECAM-1 on neonatal neutrophils underlies their impaired CD18-independent adhesion to endothelial cells in vitro. J Immunol 146:3372. Ando J, Komatsuda T, Kamiya A. 1988. Cytoplasmic calcium response to fluid shear stress in cultured vascular endothelial cells. In Vitro Cell Devel 24:871. Asakura T, Karino T. 1990. Flow patterns and spatial distribution of atherosclerotic lesions in human coronary arteries. Circ Res 66:1045. Barabino GA, McIntire LV, Eskin SG, et al. 1987. Endothelial cell interactions with sickle cell, sickle trait, mechanically injured, and normal erythrocytes under controlled flow. Blood 70:152. Barbee KA, Davies PF, Lal R. 1994. Shear stress-induced reorganizations of the surface topography of living endothelial cells imaged by atomic force microscopy. Circ Res 74:163. Bastida E, Almirall L, Bertomeu MC, et al. 1989. Influence of shear stress on tumor-cell adhesion to endothelial-cell extracellular matrix and its modulation by fibronectin. Int J Cancer 43:1174. Bhagyalakshmi A, Berthiaume F, Reich KM, et al. 1992. Fluid shear stress stimulates membrane phospholipid metabolism in cultured human endothelial cells. J Vasc Res 29:443. Bhagyalakshmi A, Frangos JA. 1989a. Mechanism of shear-induced prostacyclin production in endothelial cells. Biochem Biophys Res Comm 158:31. Bhagyalakshmi A, Frangos JA. 1989b. Membrane phospholipid metabolism in sheared endothelial cells. Proc 2nd Int Symp Biofluid Mechanics and Biorheology, Munich, Germany 240. Bloch KD, Friedrich SP, Lee ME, et al. 1989. Structural organization and chromosomal assignment of the gene encoding endothelin. J Biol Chem 264:10851. Bohjanen PR, Petryniak B, June CH, et al. 1992. AU RNA-binding factors differ in their binding specificities and affinities. J Biol Chem 267:6302. Bohmann D. 1990. Transcription factor phosphorylation: A link between signal transduction and the regulation of gene expression. Cancer Cells 2:337. Botteri FM, Ballmer-Hofer K, Rajput B, et al. 1990. Disruption of cytoskeletal structures results in the induction of the urokinase-type plasminogen activator gene expression. J Biol Chem 265:13327. Breathnach R, Matrisian LM, Gesnel MC, et al. 1987. Sequences coding part of oncogene-induced transin are highly conserved in a related rat gene. Nucleic Acid Res 15:1139. Coan DE, Wechezak AR, Viggers RF, et al. 1993. Effect of shear stress upon localization of the Golgi apparatus and microtubule organizing center in isolated cultured endothelial cells. J Cell Sci 104:1145. Cooper GM. 1990. Oncogenes, Boston, Jones and Bartlett. Cybulsky MI, Fries JW, Williams AJ, et al. 1991. Gene structure, chromosomal location, and basis for alternative mRNA splicing of the human VCAM-1 gene. Proc Natl Acad Sci USA 88:7859.


Davies MJ, Woolf N, Rowles PM, et al. 1988. Morphology of the endothelium over atherosclerotic plaques in human coronary arteries. Br Heart J 60:459. Davies PF, Tripathi SC. 1993. Mechanical stress mechanisms and the cell: An endothelial paradigm. Circ Res 72:239. Dewey CF Jr, Bussolari SR, Gimbrone MA Jr, et al. 1981. The dynamic response of vascular endothelial cells to fluid shear stress. J Biomech Eng 103:177. Diamond SL, Eskin SG, McIntire LV. 1989. Fluid flow stimulates tissue plasminogen activator secretion by cultured human endothelial cells. Science 243:1483. Diamond SL, Sharefkin JB, Dieffenbach C, et al. 1990. Tissue plasminogen activator messenger RNA levels increase in cultured human endothelial cells exposed to laminar shear stress. J Cell Physiol 143:364. Dull RO, Davies PF. 1991. Flow modulation of agonist (ATP)-response (Ca++) coupling in vascular endothelial cells. Am J Physiol 261:H149. Eskin SG, Ives CL, McIntire LV, et al. 1984. Response of cultured endothelial cells to steady flow. Microvasc Res 28:87. Ferrua B, Manie S, Doglio A, et al. 1990. Stimulation of human interleukin-1 production and specific mRNA expression by microtubule-disrupting drugs. Cell Immunol 131:391. Folie BJ, McIntire LV. 1989. Mathematical analysis of mural thrombogenesis. Biophys J 56:1121. Frangos JA, Eskin SG, McIntire LV, et al. 1985. Flow effects on prostacyclin production by cultured human endothelial cells. Science 227:1477. Franke RP, Grafe M, Schnittler H. 1984. Induction of human vascular endothelial stress fibres by shear stress. Nature 307:648. Giavazzi R, Foppolo M, Dossi R, et al. 1993. Rolling and adhesion of human tumor cells on vascular endothelium under physiological flow conditions. J Clin Invest 92:3038. Girard PR, Nerem RM. 1991. Fluid shear stress alters endothelial cell structure through the regulation of focal contact-associated proteins. Adv Bioeng 20:425. Girard PR, Nerem RM. 1993. Endothelial cell signaling and cytoskeletal changes in response to shear stress. Front Med Biol Eng 5:31. Glagov S, Zarins CK, Giddens DP, et al. 1988. Hemodynamics and atherosclerosis. Arch Pathol Lab Med 112:1018. Glaser V. 1994. Targeted injectable vectors remain the ultimate goal in gene therapy. Genetic Eng News 14:8. Grabowski EF, Jaffe EA, Weksler BB. 1985. Prostacyclin production by cultured endothelial cell monolayers exposed to step increases in shear stress. J Lab Clin Med 105:36. Grande JP, Glagov S, Bates SR, et al. 1989. Effect of normolipemic and hyperlipemic serum on biosynthesis response to cyclic stretching of aortic smooth muscle cells. Arteriosclerosis 9:446. Gupte A, Frangos JA. 1990. Effects of flow on the synthesis and release of fibronectin by endothelial cells. In Vitro Cell Devel Biol 26:57. Hadcock JR, Malbon CC. 1988. Down-regulation of beta-adrenergic receptors: Agonist-induced reduction in receptor mRNA levels. Biochemistry 95:5021. Hakkert BC, Kuijipers TW, Leeuwenberg JFM, et al. 1991. Neutrophil and monocyte adherence to and migration across monolayers of cytokine-activated endothelial cells: the contribution of CD18, ELAM-1, and VLA-4. Blood 78:2721. Hession C, Osborn L, Goff D, et al. 1990. Endothelial leukocyte adhesion molecule-1: Direct expression cloning and functional interactions. Proc Natl Acad Sci USA 87:1673. Hsieh HJ, Li NQ, Frangos JA. 1991. Shear stress increases endothelial platelet-derived growth factor mRNA levels. Am J Physiol H642. Hsieh H, Li NQ, Frangos JA. 1992. Shear-induced platelet-derived growth factor gene expression in human endothelial cells is mediated by protein kinase C. J Cell Physiol 150:552. Hsieh H, Li NQ, Frangos JA. 1993. Pulsatile and steady flow induces c-fos expression in human endothelial cells. J Cell Physiol 154:143.


Hubbel JA, McIntire LV. 1986. Technique for visualization and analysis of mural thrombogenesis. Rev Sci Instrum 57:892. Hudlicka O. 1984. Growth of vessels—historical review. In F Hammerson, O Hudlicka (eds), Progress in Applied Microcirculation Angiogenesis, vol 4, 1–8, Basel, Karger. Hunter T, Karin M. 1992. The regulation of transcription by phosphorylation. Cell 70:375. Iba T, Mills I, Sumpio BE. 1992. Intracellular cyclic AMP levels in endothelial cells subjected to cyclic strain in vitro. J Surgical Res 52:625. Ingber D. 1993. Integrins as mechanochemical transducers. Curr Opin Cell Biol 3:841. Ives CL, Eskin SG, McIntire LV. 1986. Mechanical effects on endothelial cell morphology: In vitro assessment. In Vitro Cell Dev Biol 22:500. Jackman RW, Beeler DL, Fritze L, et al. 1987. Human thrombomodulin gene is intron depleted: Nucleic acid sequences of the cDNA and gene predict protein structure and suggest sites of regulatory control. Proc Natl Acad Sci USA 84:6425. Jones DA, Abbassi OA, McIntire LV, et al. 1993. P-selectin mediates neutrophil rolling on histaminestimulated endothelial cells. Biophys J 65:1560. Jones PA. 1982. Construction of an artificial blood vessel wall from cultured endothelial and smooth muscle cells. J Cell Biol 74:1882. Kim DW, Gotlieb AI, Langille BL. 1989. In vivo modulation of endothelial F-actin microfilaments by experimental alterations in shear stress. Arteriosclerosis 9:439. Kishimoto TK, Warnock RA, Jutila MA, et al. 1991. Antibodies against human neutrophil LECAM-1 and endothelial cell ELAM-1 inhibit a common CD18-independent adhesion pathway in vitro. Blood 78:805. Kojima N, Shiota M, Sadahira Y, Handa K, Hakomori S. 1992. Cell adhesion in a dynamic flow system as compared to static system. J Biol Chem 267:17264–17270. Komuro I, Kaida T, Shibazaki Y, et al. 1990. Stretching cardiac myocytes stimulates protooncogene expression. J Biol Chem 265:3595. Komuro I, Katoh Y, Kaida T, et al. 1991. Mechanical loading stimulates cell hypertrophy and specific gene expression in cultured rat cardiac myocytes. J Biol Chem 266:1265. Kuchan MJ, Frangos JA. 1993. Shear stress regulates endothelin-1 release via protein kinase C and cGMP in cultured endothelial cells. Am J Physiol 264:H150. Langille BL, Adamson SL. 1981. Relationship between blood flow direction and endothelial cell orientation at arterial branch sites in rabbit and mice. Circ Res 48:481. Langille BL, Graham JJ, Kim D, et al. 1991. Dynamics of shear-induced redistribution of F-actin in endothelial cells in vivo. Arterioscler Thromb 11:1814. Langille BL, O’Donnell F. 1986. Reductions in arterial diameter produced by chronic diseases in blood flow are endothelium-dependent. Science 231:405. Lawrence MB, McIntire LV, Eskin SG. 1987. Effect of flow on polymorphonuclear leukocyte/endothelial cell adhesion. Blood 70:1284. Lawrence MB, Smith CW, Eskin SG, et al. 1990. Effect of venous shear stress on CD18-mediated neutrophil adhesion to cultured endothelium. Blood 75:227. Lawrence MB, Springer TA. 1991. Leukocytes roll on a selectin at physiologic flow rates: distinction from and prerequisite for adhesion through integrins. Cell 65:859. Levesque MJ, Sprague EA, Schwartz CJ, et al. 1989. The influence of shear stress on cultured vascular endothelial cells: The stress response of an anchorage-dependent mammalian cell. Biotech Prog 5:1. Levin EG, Santell L. 1988. Stimulation and desensitization of tissue plasminogen activator release from human endothelial cells. J Biol Chem 263:9360. Ley K, Gaehtgens P, Fennie C, et al. 1991. Lectin-like adhesion molecule 1 mediates leukocyte rolling in mesenteric venules in vivo. Blood 77:2553. Malek AM, Gibbons GH, Dzau VJ, et al. 1993a. Fluid shear stress differentially modulates expression of genes encoding basic fibroblast growth factor and platelet-derived growth factor B chain in vascular endothelium. J Clin Invest 92:2013.


Malek AM, Greene AL, Izumo S. 1993b. Regulation of endothelin 1 gene by fluid shear stress is transcriptionally mediated and independent of protein kinase C and cAMP. Proc Natl Acad Sci 90:5999. Malek A, Izumo S. 1992. Physiological fluid shear stress causes downregulation of endothelin-1 mRNA in bovine aortic endothelium. Am J Physiol 263:C389. McIntire LV. 1994. Bioengineering and vascular biology. Ann Biomed Eng 22:2. Menter DG, Patton JT, Updike TV, et al. 1992. Transglutaminase stabilizes melanoma adhesion under laminar flow. Cell Biophys 18:123. Milner P, Bodin P, Loesch A, et al. 1992. Increased shear stress leads to differential release of endothelin and ATP from isolated endothelial cells from 4- and 12-month-old male rabbit aorta. J Vasc Res 29:420. Mitsumata M, Fishel RS, Nerem RM, et al. 1993. Fluid shear stress stimulates platelet-derived growth factor expression in endothelial cells. Am J Physiol 263:H3. Mo M, Eskin SG, Schilling WP. 1991. Flow-induced changes in Ca2+ signaling of vascular endothelial cells: Effect of shear stress and ATP. Am J Physiol 260:H1698. Morita T, Yoshizumi M, Kurihara H, et al. 1993. Shear stress increases heparin-binding epidermal growth factor-like growth factor mRNA levels in human vascular endothelial cells. Biochem Biophys Res Comm 197:256. Muller WEG, Slor H, Pfeifer K, et al. 1992. Association of AUUUA-binding Protein with A + U-rich mRNA during nucleo-cytoplasmic transport. J Mol Biol 226:721. Nabel EG, Plautz G, Boyce FM, et al. 1989. Recombinant gene expression in vivo within endothelial cells of the arterial wall. Science 244:1342. Nagel T, Resnick N, Atkinson W, et al. 1994. Shear stress selectivity upregulates intercellular adhesion molecule-1 expression in cultured vascular endothelial cells. J Clin Invest 94:885. Nerem RM. 1991. Cellular engineering. Ann Biomed Eng 19:529. Nerem RM. 1993. Hemodynamics and the vascular endothelium. J Biomech Eng 115:510. Nerem RM, Levesque MJ, Cornhill JF. 1981. Vascular endothelial morphology as an indicator of the pattern of blood flow. J Biomech Eng 103:172. Nollert MU, Diamond SL, McIntire LV. 1991. Hydrodynamic shear stress and mass transport modulation of endothelial cell metabolism. Biotech Bioeng 38:588. Nollert MU, Eskin SG, McIntire LV. 1990. Shear stress increases inositol trisphosphate levels in human endothelial cells. Biochem Biophys Res Comm 170:281. Nollert MU, Hall ER, Eskin SG, et al. 1989. The effect of shear stress on the uptake and metabolism of arachidonic acid by human endothelial cells. Biochim Biophys Acta 1005:72. Nollert MU, McIntire LV. 1992. Convective mass transfer effects on the intracellular calcium response of endothelial cells. J Biomech Eng 114:321. Nollert MU, Panaro NJ, McIntire LV. 1992. Regulation of genetic expression in shear stress stimulated endothelial cells. Ann NY Acad Sci 665:94. Ohno M, Gibbons GH, Dzau VJ, et al. 1993. Shear stress elevates endothelial cGMP. Role of potassium channel and G protein coupling. Circulation 88:193. Ohno M, Lopez F, Gibbons GH, et al. 1992. Shear stress induced TGFβ1 gene expression and generation of active TGFβ1 is mediated via a K+ channel. Circulation 86:I-87. Ono O, Ando J, Kamiya A, et al. 1991. Flow effects on cultured vascular endothelial and smooth muscle cell functions. Cell Struct Funct 16:365. Patton JT, Menter DG, Benson DM, et al. 1993. Computerized analysis of tumor cells flowing in a parallel plate chamber to determine their adhesion stabilization lag time. Cell Motility and the Cytoskeleton 26:88. Peppel K, Vinci JM, Baglioni C. 1991. The AU-rich sequences in the 3′ untranslated region mediate the increased turnover of interferon mRNA induced by glucocorticoids. J Exp Med 173:349. Perry MA, Granger DN. 1991. Role of CD11/CD18 in shear rate-dependent leukocyte-endothelial cell interactions in cat mesenteric venules. J Clin Invest 87:1798.


Pili R, Corda S, Passaniti A, et al. 1993. Endothelial cell Ca2+ increases upon tumor cell contact and modulates cell-cell adhesion. J Clin Invest 92:3017. Port JD, Huang LY, Malbon CC. 1992. β-adrenergic agonists that down-regulate receptor mRNA upregulate a Mr 35,000 protein(s) that selectively binds to β-adrenergic receptor mRNAs. J Biol Chem 267:24103. Prasad AR, Logan SA, Nerem RM, et al. 1993. Flow-related responses of intracellular inositol phosphate levels in cultured aortic endothelial cells. Circ Res 72:827. Ranjan V, Diamond SL. 1993. Fluid shear stress induces synthesis and nuclear localization of c-fos in cultured human endothelial cells. Biochem Biophys Res Comm 196:79. Ratner L, Josephs SF, Jarrett R, et al. 1985. Nucleotide sequence of transforming human c-sis-cDNA clones with homology to platelet derived growth factor. Nucl Acids Res 13:5007. Reddy VB, Garramone AJ, Sasak H, et al. 1987. Expression of human uterine tissue-type plasminogen activator using BPV vectors. DNA 6:461. Resnick N, Collins T, Atkinson W, et al. 1993. Platelet-derived growth factor B chain promoter contains a cis-acting fluid shear-stress-responsive-element. Proc Natl Acad Sci 90:4591. Ross R. 1993. The pathogenesis of atherosclerosis: a perspective for the 1990s. Nature 362:801. Sampath R, Kukielka GL, Smith CW, et al. 1995. Shear stress mediated changes in the expression of leukocyte adhesion receptors on human umbilical vein endothelial cells in vitro. Ann Biomed Eng. Schnittler HJ, Franke RP, Akbay U, et al. 1993. Improved in vitro rheological system for studying the effect of fluid shear stress on cultured cells. Am J Physiol 265:C289. Schwartz CJ, Valente AJ, Spargue EA. 1993. A modern view of atherogenesis. Am J Cardiol 71:9B. Sharefkin JB, Diamond SL, Eskin SG, et al. 1991. Fluid flow decreases preproendothelin mRNA levels and suppresses endothelin-1 peptide release in cultured human endothelial cells. J Vasc Surg 14:1. Shyy YJ, Hsieh HJ, Usami S, et al. 1994. Fluid shear stress induces a biphasic response of human monocyte chemotactic protein 1 gene expression in vascular endothelium. Proc Natl Acad Sci (USA) 91:4678. Smith CW, Kishimoto TK, Abbassi OA, et al. 1991. Chemotactic factors regulate LECAM-1 dependent neutrophil adhesion to cytokine-stimulated endothelial cells in vitro. J Clin Invest 87:609. Staunton DE, Marlin SD, Stratowa C, et al. 1988. Primary structure of ICAM-1 demonstrates interaction between members of the immunoglobulin and integrin supergene families. Cell 52:925. Tokunaga K, Nakamura Y, Sakata K, et al. 1987. Enhanced expression of a glycearldehyde-3-phosphate dehydrogenase. Cancer Res 47:5616. Tran-son-Tay R. 1993. Techniques for studying the effects of physical forces on mammalian cells and measuring cell mechanical properties. In JA Frangos (ed), Physical Forces and the Mammalian Cell, p 1, New York, Academic Press. Turitto VT, Weiss HJ, Baumgartner HR, et al. 1987. Cells and aggregates at surfaces. In EF Leonord, VT Turitto, L Vroman (eds), Blood in Contact with Natural and Artificial Surfaces, vol 516, pp 453–467, New York, Annals of the New York Academy of Sciences. Von Andrian UH, Chambers JD, McEvoy LM, et al. 1991. Two-step model of leukocyte-endothelial cell interaction in inflammation. Proc Natl Acad Sci 88:7538. Warren BA. 1973. Evidence of the blood-borne tumor embolus adherent to vessel wall. J Med 6:150. Watson PA. 1991. Function follows form: Generation of intracellular signals by cell deformation. FASEB J 5:2013. Watson SR, Fennie C, Lasky LA. 1991. Neutrophil influx into an inflammatory site inhibited by a soluble homing receptor-IgG chimaera. Nature 349:164. Wechezak AR, Viggers RF, Sauvage LR. 1985. Fibronectin and F-actin redistribution in cultured endothelial cells exposed to shear stress. Lab Invest 53:639. Weinberg CB, Bell E. 1986. A blood vessel model constructed from collagen and cultured vascular cells. Science 231:397. Weiss HJ, Turitto VT, Baumgartner HR. 1978. Effect of shear rate on platelet interaction with subendothelium in citrated and native blood. J Lab Clin Med 92:750.


Weiss HJ, Turitto VT, Baumgartner HR. 1986. Platelet adhesion and thrombus formation on subendothelium in platelets deficient in GP IIb-IIIa, Ib, and storage granules. Blood 67:905. Weiss L. 1992. Biomechanical interactions of cancer cells with the microvasculature during hematogenous metastasis. Cancer Metastasis Rev 11:227. Werb Z, Hembry RM, Murphy G, et al. 1986. Commitment to expression of metalloendopeptidases, collagenase, and stromelysin: Relationship of inducing events to changes in cytoskeletal architecture. J Cell Biol 102:697. Wick TM, Moake JL, Udden MM, et al. 1987. ULvWF multimers increase adhesion of sickle erythrocytes to human endothelial cells under controlled flow. J Clin Invest 80:905. Wick TM, Moake JL, Udden MM, et al. 1993. ULvWF multimers preferentially promote young sickle and non-sickle erythrocyte adhesion to endothelial cells. Am J Hematol 42:284. Wilson JM, Birinyi LK, Salomon RN, et al. 1989. Implantation of vascular grafts lined with genetically modified endothelial cells. Science 244:1344. Yanagisawa M, Kurihara H, Kimura S, et al. 1988. A novel potent vasoconstrictor peptide produced by vascular endothelial cells. Nature 332:411. Yoshizumi M, Kurihara H, Sugiyama T, et al. 1989. Hemodynamic shear stress stimulates endothelin production by cultured endothelial cells. Biochem Biophys Res Commun 161:859. Zarins C, Zatina MA, Giddens DP. 1987. Shear stress regulation of artery lumen diameter in experimental atherogenesis. J Vasc Surg 5:413. Zarins CK, Giddens DP, Bharadvaj BK, et al. 1983. Carotid bifurcation atherosclerosis. Circ Res 53:502. Ziegelstein RC, Cheng L, Blank PS, et al. 1993. Modulation of calcium homeostasis in cultured rat aortic endothelial cells by intracellular acidification. Am J Physiol 265:H1424. Ziegelstein RC, Cheng L, Capogrossi MC. 1992. Flow-dependent cytosolic acidification of vascular endothelial cells. Science 258:656. Zwiebel JA, Freeman SM, Kantoff PW, et al. 1989. High-level recombinant gene expression in rabbit endothelial cells transduced by retroviral vectors. Science 243:220.


Lightfoot, E. N., Duca, K. A. “The Roles of Mass Transfer in Tissue Function.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

115 The Roles of Mass Transfer in Tissue Function 115.1 115.2

Self Similarity and Cross-Species Correlations • Time Constants: The Key to Quantitative Modeling • Brownian Motion and Concentration Diffusion • More Complex Situations • Flow, Chemical Reaction, and Boundary Conditions

Edwin N. Lightfoot University of Wisconsin

Karen A. Duca University of Wisconsin

Topology and Transport Characteristics of Living Organisms Fundamentals: The Basis of a Quantitative Description

115.3

Characteristic Behavior: Selected Examples The Energy Cost of Immobility

Mass transfer lies at the heart of physiology and provides major constraints on the metabolic rates and anatomy [Pries et al., 1996; Bunk, 1998] of living organisms, from the organization of organ networks to intracellular structures. Limitations on mass transport rates are major constraints for nutrient supply, waste elimination and information transmission at all of these levels. The primary functional units of metabolically active tissue, e.g., the Krogh tissue cylinders of muscle and brain, liver lobules and kidney nephrons, have evolved to just eliminate significant mass transfer limitations in the physiological state [Lightfoot, 1974]. Turnover rates of highly regulated enzymes are just on the slow side of diffusional limitations [Weisz, 1973]. Signal transport rates are frequently mass transport limited [Lauffenburger and Linderman, 1993], and very ingenious mechanisms have evolved to speed these processes [Berg and von Hippel, 1985; Bray, 1998; Francis and Palsson, 1997; Valee and Sheets, 1996]. In contrast elaborate membrane barriers organize and control intracellular reactions in even the simplest organisms. Understanding tissue mass transport is important both for the engineer and scientist. Examples of engineering interest include the design of extracorporeal devices and biosensors [Fishman et al., 1998]. For the scientist it is important to understand the sometimes complex interactions of transport and reaction to interpret transport based experiments [Bassingthwaighte, Goresky and Linehan, 1998]. Here we shall concentrate on qualitative behavior and orders of magnitude, with particular emphasis on time constants. These are the normal starting points of any serious study. A general background in biology [Campbell, 1996] and physiology [Johnson, 1998] will be helpful.


FIGURE 115.1 Diffusional topology of the mammalian body. Convective transport (bulk flow) dominates in the major blood vessels and airways but becomes comparable to diffusion and reaction in the functional units surrounding capillaries and sinusoids. At the cellular and subcellular levels concentration diffusion complicated by electrical effects and a wide variety of carrier transport processes, interacts in complex ways with enzymatic and genetic reactions. Inspired by Dr. Peter Abbrecht, University of Michigan. Reprinted with permission from Lightfoot (1974). Copyright 1974 by John Wiley and Sons, Inc.

115.1 Topology and Transport Characteristics of Living Organisms [Schmidt-Nielsen 1983, 1984, Calder 1984, Berg 1993 Bassingthwaighte et al. 1994] If one includes viruses, the size scales of living systems range from nanometers to meters: spanning a linear ratio of 109 and a mass ratio of 1027, which is greater than Avogadro’s number! As indicated in Fig. 115.1, the higher animals are organized into spatial hierarchies of discrete structures, which span this whole size range (Table 115.1). At the largest scales animals may be considered as organ networks connected by major blood vessels, with each organ carrying out its own set of specialized tasks. Organs are in turn composed essentially of large numbers of microcirculatory functional units organized in parallel and TABLE 115.1 The Spatial Hierarchy perfused by capillaries or other microscopic ducts which in Mammals, Characteristic Lengths supply oxygen and other nutrients, carry away waste prodEntity Length Scale, m ucts, and interchange via a variety of chemical messengers [Lauffenburger and Linderman, 1993]. The Krogh tissue cylWhole body 10–1–100 inder of Fig. 115.2 is representative and corresponds approxOrgans 10–2–10–1 imately to the functional units of the brain (see, however, Microcirculatory units 10–4 Cells 10–5 (eukaryiots) Federspiel and Popel, 1986; Popel, 1989; Hellums et al., 1996; 10–6 (prokaryiots) Vicini et al., 1998; Bassingthwaighte et al., 1998). Intracellular organelles 10–6 Nutrients and metabolic end products are transported at Molecular complexes 10–8 the size scale of the tissue cylinder axially by convection. © 2000 by CRC Press LLC

FIGURE 115.2 elements.

Krogh tissue cylinder. A circular cylinder is used as an approximation to close-packed hexagonal

Radial transport is primarily by concentration diffusion or Brownian motion [Lightfoot and Lightfoot, 1997] and, to a lesser extent by Starling flow, slow seepage across the proximal region of the capillary driven by the relatively high pressure of entering capillary blood. Flow takes place through clefts between the endothelial cells forming the capillary wall and between the parenchymal cells forming the tissue cylinder. The clefts are narrow and permselective, rejecting large solutes. Starling flow was once thought to reverse direction in the distal (downstream) regions of the tissue cylinder as a result of lower hydrodynamic pressure and increased colloid osmotic pressure caused by the permselectivity of the clefts. However, it is now accepted that such reverse flow does not occur to a significant extent and that the seepage ends up primarily in the lymph ducts and on to the venous blood. The microcirculatory units are in turn comprised of cells which are themselves complex structures. Sketched in Fig. 115.3 is a pancreatic beta cell, which produces, stores, and secretes insulin. Its organization is similar to that of a chemical plant, with raw materials and energy input at the bottom where mitochondria produce the chemical energy for cell metabolism, a synthesis and transport area, plus the

FIGURE 115.3 Structure and organization of a representative eukaryotic cell. Schematic cross-section of a pancreatic beta cell. Reprinted with permission from Lightfoot (1974). Copyright 1974 by John Wiley and Sons. © 2000 by CRC Press LLC

control region of the cell nucleus, at the center, and packaging, storage and export at the top. All of this is accomplished in about 10 micrometers. The cell contains smaller structures known collectively as organelles that segregate and coordinate the many processes required for its operation and maintenance. Prominent in the diagram are mitochondria, which use various carbon sources and oxygen to form high-energy phosphate bonds as in ATP—the energy sources needed for cell metabolism. Also shown are the cell nucleus where DNA is stored, ribosomes in which RNA is used to produce individual proteins, and the endoplasmic reticulum which holds the ribosomes and channels the proteins produced by them to Golgi apparatus (not shown) for packaging. Also not shown are microtubules and other filaments comprising the cytoskeleton [Alberts et al., 1945; Campenot et al., 1996; Maniotis et al., 1997]. The latter provide for chromosome segregation, help to maintain cellular shape, and in some cases provide mobility [Vallee and Sheetz]. They also act as transport networks along with motor proteins, discussed below. Organization and structure of the cells is described in standard works [Alberts et al., 1945]. At the smallest level are enzyme clusters and substructures of cell membranes used for selective transport across the cell boundary [Lauffenberger and Linderman, 1993]. Underlying these structures are biochemical reaction networks that are largely shared by all species, and all are composed of the same basic elements, primarily proteins, carbohydrates, and lipids. As a result there are a great many interspecies similarities [Schmidt-Nielsen, 1983]. This elaborate organization just described is constrained to a very large extent by mass transfer considerations and in particular by the effect of characteristic time and distance scales of the effectiveness of different mass transport mechanisms. At the larger size scales, only flow or convection is fast enough to transport oxygen and major metabolites, and convective transport is a major function of the larger blood vessels. Diffusive transport begins to take precedence at the level of microcirculatory units. At the cellular level and below, diffusion may even be too fast and undirected, and selectively permeable membranes, have evolved to maintain spatial segregation against the randomizing forces of diffusion.

115.2 Fundamentals: The Basis of a Quantitative Description Underneath the bewildering complexity of living organisms are some very simple underlying principles which make a unified description of mass transport feasible. Of greatest utility are observed similarities across the enormous range of system sizes and common magnitudes of key thermodynamic, reaction, and transport parameters. Some simplifying features must still be accepted as justified by repeated observation, and others can be understood from the first principles of molecular kinetic theory. Here we summarize some of the most useful facts and approximations in preparation for the examples of the next section.

Self Similarity and Cross-Species Correlations It is a striking characteristic of life forms that each increase in our mathematical sophistication has shown new regularity in their anatomic and physiologic characteristics. The ability of simple geometric forms to describe morphology has been recognized at least since Leonardo da Vinci, and the definitive work of D’Arcy Thompson summarizes much of this early work. The next step was the concept of allometry, first introduced by J. S. Huxley in 1927 [Schmidt-Nielsen, 1984; Calder, 1984; Lightfoot, 1974]. This is a rudimentary form of self similarity usually expressed as

P = aM b

(115.1)

where P is any property, M is average species mass, and a and b are constants specific to the property. A large number of allometric relations are available, and those of most interest to us deal with metabolic rate. For example, total basal rates of oxygen consumption for whole animals is given to a good approximation by: © 2000 by CRC Press LLC

TABLE 115.2

Characteristic Cerebral Tissue Cylinder

Item

Magnitude 30 µm 3 µm 180 µm 400 µm/s 0.125 atm 8.6 mM 0.12 0.053 atm 5.87 mM 0.05 1.5 E-5 cm2/s (estd.) 0.0372 mmols O2 /liter-s

Outer radius Capillary radius Length Blood velocity Arterial oxygen tension Arterial oxygen concentration (total) Arterial oxygen concentration (dissolved O2) Venous oxygen tension Venous oxygen concentration (total) Venous oxygen concentration (dissolved O2) Tissue oxygen diffusivity Oxygen respiration rate (zero order)

R O ,tot × 3.5M 3 4

(115.2)

Here RO,tot is the oxygen consumption rate in ml O2 (STP)/hr, and M is body mass in grams. Under basal conditions, fat is the primary fuel and heat generation is about 4.7 kcal per milliliter of oxygen (STP) consumed. Small animals have higher specific metabolic rates than large ones, but this is in part because a higher proportion of their body mass is made up of highly active body mass, and for the important case of brain tissue it is invariant at about:

(

)

R O 2 brain ≈ 3.72 ⋅ 10−5 mmolsO2 cm3 , s

(115.3)

for all species. For the liver and kidneys, specific metabolic activity is somewhat lower and falls off slowly with an increase in animal size, but this may be due to an increasing proportion of supporting tissue such as blood vessels and connective tissue. Accurate data valid under physiologic conditions are difficult to find, and to a first approximation specific metabolic activity of parenchymal cells, those actually engaged in the primary activity of the organ, may be close to that of brain for both liver and kidneys. The sizes of both microcirculatory units and cells in vertebrates are also very insensitive to animal size. Capillaries are typically about 3 to 4 µm in radius and about 50 to 60 µm apart. Typical mammalian cells are about 10 to 50 µm in diameter, and organelles such as mitochondria are about the size of prokaryotic cells, about 1 µm in diameter. Approximate characteristics of a cerebral tissue cylinder are given in Table 115.2. The oxygen-carrying capacity of blood, ionic makeup of body fluids, solubility of gases and oxygen diffusivities in body fluids are also largely invariant across species, and some representative data are provided in Tables 115.3, 115.4, and 115.5. TABLE 115.3

Oxygen Solubilities

Solvent

Temperature, °C

Water

25 30 35 40 37 37 37 37 37 37

Plasma Red cell interior (dissolved O2 only) Extracellular tissue (estd.) Oxygen gas Alveolar air Air (0.21 atm of oxygen)


O2 Pressure, atm

Concentration, mM/atm 1.26 1.16 1.09 1.03 1.19 1.18 1.1 39.3

0.136 0.21

TABLE 115.4

Effective Oxygen Diffusivities Temperature, °C

Solvent Water Water Blood plasma Normal blood Red cell interior Air

TABLE 115.5

25 37 37 37 37 25

Pressure

Diffusivity, cm2/s

1 atm

2.1 E-5 3.0 E-5 2.0 E-5 1.4 E-5 0.95 E-5 0.20

Intracellular Diffusion Coefficients

Compound Sorbitol Methylene blue Sucrose Eosin Dextran Inulin Dextran Dextran Actin Bovine serum albumin

MW

Radius, Å

Diffusivity in Water cm2/s × 107

170 320 324 648 3600 5500 10,000 24,000 43,000 68,000

2.5 3.7 4.4 6.0 12.0 13.0 23.3 35.5 23.2 36.0

94 40 52 40 18 15 9.2 6.3 5.3 6.9

Intracell Diffusivity; cm2/s × 107

Diffusivity Ratio

50 15 20 8 3.5 3.0 2.5 1.5 0.03 0.10

1.9 2.6 2.6 5.0 5.0 5.0 3.7 4.2 167 71

Allometric correlations are essentially empirical and cannot be predicted from any fundamental physical principles. Moreover, most measurements, except solubility and diffusivity data, are of doubtful accuracy. Recently a more sophisticated form of self similarity, fractal geometry has been found useful for the description of a wide variety of biological applications of mass transfer interest [Bassingthwaighte et al., 1994], applications of nonlinear dynamics are fast increasing [Griffith, 1996].

Time Constants: The Key to Quantitative Modeling The first step in system modeling is to establish orders of magnitudes of key parameters, and most particularly time constants: estimates of the time required for a given transient process to be “effectively complete.” Time constant or order of magnitude analysis is useful because dynamic response times are insensitive to geometric detail and boundary conditions at the order of magnitude level of approximation, i.e., within a factor of ten. Time constants are, however, essentially heuristic quantities and can only be understood on the basis of experience [Lightfoot and Lightfoot, 1997]. Once characteristic system time scales are established, one can restrict attention to processes with response times of the same order: Those an order of magnitude faster can be treated as instantaneous and those ten times slower as not happening at all. Both fast and slow terms in system description, e.g., the diffusion equation of transport phenomena [Bird et al., 1960], can then be eliminated. Such simplification often provides valuable insights as well as simplifying integration. Quantitative descriptions are particularly valuable at the microcirculatory level, for example in diagnostic procedures, and they have been well studied [Bassingthwaighte and Goresky]. Here we shall stay with relatively simple examples to illustrate selected characteristics of tissue mass transfer, and we begin with diffusion. We then briefly introduce time constants characterizing flow, chemical reaction, and boundary conditions.


Brownian Motion and Concentration Diffusion The basis of most species selective transport is the relatively slow observable motion of molecules or particles resulting from intermolecular collisions [Lightfoot and Lightfoot, 1997]. Such Brownian motion does not take any predictable direction, but the particles under observation do tend to move farther from their starting point with increasing time. The extent of diffusional motion can be described as the probability of finding any reference particle a distance r from its initial position. For an unbounded quiescent fluid this is [Einstein, 1905]:

()

P r = e− r

2

(

 8 πDPFt 

4 D PF t

)

32

  ; 4π 

∫ P(r)r dr = 1 . ∞

2

(115.4a, 4b)

0

This equation defines the Brownian diffusivity, DPF , of a particle or molecule relative to a surrounding fluid. Here P is a spherically symmetrical normalized probability density. Here r is distance (in spherical coordinates) from the initial position, and t is time. The mean net distance traveled from the initial point in unbounded quiescent three-dimensional space is easily determined from the above distribution as:

rm2 = 4 π

∫

0

∞

()

r 2 ⋅ r 2P r dr = 6DPFt, rm2 = 4DPFt, x m2 = 2DPFt

(115.5a, 5b, 5c)

for distances from a point, line, and plane respectively. These results provide useful insight in suggesting characteristic diffusion times t as the time required for “most” of a transient diffusion process to be complete. Some commonly accepted values are shown in Table 115.6. The numbers in the last column are the fractional changes in solute inventory for a sudden change in surface concentration on a particle initially at a different uniform concentration. For the hollow cylinder, the outer surface of radius RT is assumed impermeable to diffusion, and the inner surface of radius RC is permeable. The length L is assumed large compared to either radius. Fractional completion depends in a complex way on the radius ratio, but the diffusion time given is a good approximation. For large numbers of particles or molecules nonuniformly distributed in a moving fluid, following the Brownian motion of individual molecules becomes too cumbersome to use. We must then average behavior over large numbers of molecules to obtain a continuum approximation known as Fick’s law [Bird et al., 1960], which describes the relative motion of solute and solvent and may be written in the form

(v

P

x x  − v F  P F  = −∇x P  DPF 

)

(115.6)

Here vp is observable velocity of species P in a mixture of P and F, xp is mole fraction of species P in the mixture, and DPF is binary mutual mass diffusivity of solute P relative to solvent fluid F. For situations of interest here, the Fick’s law diffusivity may be considered equal to the Brownian diffusivity.

TABLE 115.6 Shape Sphere Cylinder Slab Hollow cylinder


Characteristic Diffusional Response Times Tdif

L

L2/6DPF L2/4DPF L2/2DPF (L2/2DPF) ln (RT /Rc)

Radius Radius Half-thickness Outer radius, RT

Fractional Completion >0.99 >0.99 >0.93

Equation 115.6 is valid for binary or dilute solutions of liquids, gases, and homogeneous, and some typical magnitudes for biologic situations are shown in Tables 115.4 and 115.5. For dilute solutions, hydrodynamic diffusion theory [Bird et al., 1960] provides useful insight:

πR PDPFµ F κT = C

(115.7)

where C is equal to (1/6) if the molecular radius of the solute P is much less than that of the solvent F and (1/4) if they are about equal. Here µ is solvent viscosity, RP and RF are effective spherical solute and solvent radii, κ is the Boltzmann constant, and T is absolute temperature. Hydrodynamic diffusion theory has been extended to solutes in small fluid-filled pores [Deen, 1987] for characterizing transport in microporous membranes. More Complex Situations Most biologic transport processes occur in multicomponent solutions. A generalized diffusion equation [Hirschfelder et al., 1954; Lightfoot and Lightfoot, 1997; Taylor and Krishna, 1993] is available, but lack of data usually forces neglect of multicomponent effects. For our purposes transport of both molecules and particles is adequately described by the simplified equations in Table 115.7, which simply state that relative velocity of a particle or molecule P through a fluid F is proportional to the sum of “driving forces” acting on it. Flow, Chemical Reaction, and Boundary Conditions We now introduce additional time constants in order to characterize the interactions of diffusion with flow and chemical reaction and to show that boundary conditions can have an important bearing on TABLE 115.7

Particle-Molecular Analogs Dilute Binary Diffusion in a Quiescent Continuum

Particles

(v

p

 1  0 − v F = DPF ∇ ln n p + VP 1 − ρ p ρ ∇p κT  

(

)

(

−6 πµρoF

)

12

)( )

∞

(

− Fem + ρop +

1

2

)

ρ f VP

dν P dt

  RP2 I + thermal diffusion   

where: I=

 ∫ v ′ (τ) (t − τ) t

12

p

0

  dτ 

Molecules (interdiffusion of species P and F)

(v

P

)

(

) (

)

− v F = −DPF ∇ ln x P + 1 RT VP 1 − ρoP ρ ∇ p − Fem      

Both: DPF =

κT 6πµ RP

Here: κ = the Boltzmann constant, or molecular gas constant np = number concentration of particles Rp = particle or molecular radius Vp = particle volume or partial molal volume of species P Fem = total electromagnetic force per particle Fem = molar electromagnetic force on species P ρ = density of fluid phase ρop = density of particle or reciprocal of partial specific volume of species P in solution v′ = instantaneous acceleration of particle P The subscript ∞ refers to conditions near the particle but outside its hydrodynamic boundary layer.


system behavior. We begin with flow where mean solute residence time, Tm, forms a convenient time scale. For flow through a cylindrical duct of length L with constant volumetric average velocity , we may write:

Tm ≡ L < v >

(115.8)

More generally, for constant volumetric flow at a rate Q through a system of volume V with a single inlet and outlet and no significant diffusion across either, Tm is equal to V/Q. For chemical reactions we choose as our system, a reactive solid open to diffusion over at least part of its surface where concentration of solute i under consideration is maintained at a uniform value ci0 and where the average volumetric rate of consumption of that solute is . We then define a reaction time constant as:

Trxn ≡ c i 0 < R i >

(115.9)

Note that both ci0 and are measurable, at least in principle [Damköhler, 1937; Weisz, 1973]. Finally we consider as an illustrative situation decay of solute concentration ci in the inlet stream to a flow system according to the expression

(

) (

) (

c i t, inlet = c i 0, inlet exp − t TBC

)

(115.10)

where t is time and TBC is a constant. We now have enough response times for the examples we have selected below, and we turn to illustrating their utility.

115.3 Characteristic Behavior: Selected Examples We now consider some representative examples to illustrate the mass transfer behavior of living tissues. Alveolar Transients and Pulmonary Mass Transport: At the distal ends of the pulmonary network are the alveoli, irregular sacs of air surrounded by blood-filled capillaries, with effective diameters of about 75-300 µm. The blood residence time is of the order of a second, and it is desired to determine whether gas-phase mass transfer resistance is appreciable. To answer, we take the worst possible scenario: flat-plate geometry with a half-thickness of 150 µm or 0.0105 cm. From Table 115.4 we find the oxygen diffusivity to be about 0.2 cm2/s. It follows that alveolar response time is:

(

)

Tdif ≈ 0.015 0.4 = 0.56 ms This is extremely fast, with respect to both the 1s residence time assumed for alveolar blood and the onetwelfth of a minute between breaths: gas-phase mass transfer resistance is indeed negligible. This is typical for absorption of sparingly soluble gases, and a similar situation occurs in cell culture vessels. Mass Transfer Between Blood and Tissue: We next identify blood vessels capable of transferring dissolved solute between blood and surrounding tissue. To be effective we assume that mean blood residence time should at least equal the radial diffusion time:

LDPB L R2 D2 ≥ = ; ≥ 1 16 = 0.0625 < v > 4DPB 16DPB D2 < v > where D is vessel diameter. From the data for a 13-kg dog in Table 115.8, we find the great majority of vessels much too short and that only the three smallest classes—arterioles, capillaries, and venules—are at all capable of transferring appreciable mass. These, especially the capillaries, are quite effective. They


TABLE 115.8

Mass Transfer Effectiveness of Blood Vessels (13-kg Dog, #39)

Vessel

Radius, cm

Lenght, cm

, cm/s

LD2/dPF

0.5 0.15 0.05 0.03 0.001 0.0004 0.0015 0.075 0.12 0.3 0.625

10 20 10 1.0 0.2 0.1 0.2 1.0 10 20 40

50 13.4 8 6.0 0.032 0.07 0.07 1.3 1.48 3.6 33.4

0.00003 0.00003 0.005 0.002 6.25 89.4 12.7 0.0014 0.0047 0.0006 0.00003

Aorta Larger arteries Secondary arteries Terminal arteries Arterioles Capillaries Venules Terminal veins Secondary veins Larger veins Venae cavae

have long been classified as the microcirculation on the basis of being invisible to the unaided eye. This simple order of magnitude analysis provides a functional definition and a guide to the design of hollow fiber cell culture vessels. More refined analyses [Lightfoot, 1974] based on the parameter magnitudes of Table 115.2 shows that lateral diffusional resistance of capillaries is in fact rather small, but recent analyses [Popel, 1989] suggest a more complex picture. Convection and diffusion in parallel [Lightfoot and Lightfoot, 1997] is also of interest, but it is more complicated because of convective dispersion [Lightfoot and Lightfoot, 1997]. However, the lung shows a sharp transition between convective ducts and those which can be assumed well-mixed by axial diffusion [Hobbs and Lightfoot, 1979; Hubal, 1996; Farhi, ]. This has long been known to pulmonary physiologists who often model the adult human lung as a plug-flow channel (“dead space”) of about 500 ml leading to a well-mixed volume of about 6,000 ml. Intercapillary Spacing in the Microcirculation: It has been shown [Damköhler, 1937; Weisz, 1973] that optimized commercial catalysts normally exhibit ratios of diffusion to reaction times, as defined in Table 115.6 in the relatively narrow range:

1 3 < Tdif Trxn < 1 Moreover, Weisz has shown that this ratio holds for many biologic systems as well so that enzyme activities can often be inferred from their location and function. Here we compare this expectation for cerebral tissue cylinders using the data of Tables 115.2 and 115.6. Throughout the tissue cylinder:

Tdif =

9 ⋅ 10−6 cm2 3 ⋅ 10−5 cm2 s

( )

⋅ ln 10 = 0.69s

For venous conditions, which are the most severe,

Trxn =

0.0372 = 0.64s and Tdif Trxn ≈ 1 . 0.0584

These numbers are reasonable considering the uncertainty in the data and the approximation used for diffusion time. The brain, for example, is far from homogeneous. More elaborate calculations suggest somewhat more conservative design. However, these figures are correct in suggesting that the brain is “designed” for oxygen transport and that the safety factor is small [Neubauer and James, 1998]. Sections of the brain become anoxic at about 2/3 of normal blood flow. The body’s control mechanisms will shut down other vital organs such as the gut, liver, and kidneys when blood cardiac output oxygen supply drops to keep the brain as well supplied as possible. All of these comments are for the physiological


(normal) state. Lowering brain temperature can greatly decrease oxygen demand and permit survival under otherwise fatal conditions. This effect is important in cryosurgery and drowning: whereas about 6 minutes of anoxia damages the adult brain at normal temperature, about one hour is required at 62°F. Intracellular Dispersion: We now ask how long it will take a protein initially concentrated in a very small region to disperse through the interior of a cell, and we shall assume the cell to be a sphere with a 1 µm radius. We begin by noting that the cell interior or cytoplasm is a concentrated solution of polymeric species and that the diffusivity of large molecules is considerably slowed relative to water. We therefore assume a diffusivity of 10–8 cm2/s as suggested by the last entry in Table 115.5. If the protein is originally at the center of the sphere, the dispersion time is:

Tdis ≡ Tdif =

(

R 2cell 1 ⋅ 10−4 — 6Dpc

) (6 ⋅ 10 ) ≈ (1 6)s 2

−8

If the protein is initially near the cell periphery, diffusion will take about 4 times as long, or about 2/3 s. Reliable intracellular diffusion coefficients are important: the corresponding numbers for aqueous solution would have been about 2.4 ms and 10 ms, respectively! Cell interiors—the cytoplasm—are crowded. Diffusion Controlled Reaction: We now calculate the rate at which a very dilute protein of 2.5 nm radius solution in a cell of 1 micron radius is adsorbed on a spherical adsorbent particle of radius 2.5 nanometers if the adsorption is diffusion controlled. That is, we assume that the free protein concentration immediately adjacent to the adsorbent surface is always effectively zero because of the speed and strength of the adsorption reaction. In diffusion only the region within a few diameters of a sphere’s surface offers effective resistance to transport [Carslaw and Jaeger, 1959; see also Özisik, 1993]. We can thus assume rapid equilibration over the entire cell volume, except for a thin diffusional boundary layer adjacent to the target surface: the ratio of protein plus adsorbent to cell diameter is only about 0.05. The rate at which protein is adsorbed on the sphere surface per unit area is then [Lightfoot, 1974]:

NP =

DPFc P∞

(R

P

+ R ads

)

 1  ⋅ 1 +   πτ 

where CP∞ is the protein concentration far from the sphere, assumed uniform over the bulk of cell volume since we are about to see that it adjusts rapidly on the time scale of the adsorption process. The radius of the protein is RP , that of the adsorbent Rads, and the dimensionless time τ defined by:

τ≡

(R

tDPF P

+ R ads

)

2

=

t 2.5 ⋅ 10−5 s

Adsorption rate is within 10% of its asymptotic value when πτ = 100, and:

(

)(

)

t1 10 = 2.5 ⋅ 10−5 s 100 π ≈ 0.8 ms This is much smaller than distribution time, and transients can be neglected: the time scales of the transient and distribution are well separated. We now write a mass balance for the rate of change of bulk protein concentration:

− Vcell


[(

dc P∞ dc ≈ − Vcell P = A adsN P ; d ln c P dt = 3 R P + R ads dt dt

)

]

3 — R cell D PF = 1 Trxn

Here Vcell is the volume of the cell, and Aads is the effective area of the adsorbent; T is the characteristic reaction time, that required to reduce cellular concentration by a factor e or about 2.7. For our conditions, reasonably representative of a prokaryotic cell, −1

  3 ⋅ 5 ⋅ 10−7 −8 Trxn =  ⋅ 10  s = 67s 3  10−4   

( )

This is clearly a long time compared to the dispersion time of the last example, and the assumed time scale separations of all three adsorption and dispersion processes are amply justified. Using time scale separations has greatly simplified what was originally a major numerical task [Berg and von Hipple, 1985]. If the protein concentration is now interpreted as the probability of finding a protein in any position, and the target is a DNA site, Trxn is the time required for gene expression, and it can be seen that it is very slow! Comparison of these predictions with experiment showed that both prokaryotic cells respond much faster and this led to a major discovery [Berg and von Hipple, 1985]: DNA exhibits a general binding affinity for promoter molecules which permits the adsorbed protein to undergo one-dimensional diffusion along the DNA chain—thus increasing effective binding site size. Eukaryotes, which are roughly ten times larger in diameter, have elaborate internal barriers [Holsstege et al., 1998] and are much more complex diffusionally. Many other diffusion controlled reactions occur in living systems, for example, protein folding, and here brute force molecular dynamic calculations are difficult even with supercomputers. One can greatly speed calculation by taking judicious time scale separation [Rojnuckarin et al., 1998].

The Energy Cost of Immobility Many cell constituents, from small metabolites to organelles, are free to move in the cytoplasm and yet must be limited in their spatial distribution. Many such entities are transported by mechanochemical enzymes known as protein motors which depend upon consumption of metabolic energy. Here we estimate the energy cost of such processes to maintain spatial segregation, without detailed knowledge of the mechanisms used. The basis of our analysis [Okamoto and Lightfoot, 1992] is the Maxwell-Stefan equation of Table 115.7 and more particularly the fact that migration velocities are related to the motive force of Brownian motion and any mechanically transmitted force through the diffusivity. The heart of our argument is that the mechanical force applied by the protein motors must produce a migration equal and opposite to that resulting from the dispersive force of Brownian motion:

(v − v ) = (v − v ) + (v − v ) (v − v ) = D ∇ ln n (v − v ) P

P

F

dif

F

P

tot PF

F

P

dif

P

P

F

F

mech

= 0,

= FDPF κT

mech

where nP is the number concentration of proteins, and F is the mechanical force required to produce the motion. Now the power PP required for transporting a particle back against diffusional motion is the product of the mechanical force and mechanical migration velocity:

(

PP = F ⋅ v P − v F

)

mech

(

= κT ∇ ln nP

)

2

For a particle of mass mP = (4/3)πRP3 ρP where ρP is particle density, the power requirement per unit mass is:


TABLE 115.9

Energetics of Forced Diffusion Cytoplasm Dilute Solution

Particle Small metabolite Globular protein Typical organelle

Radius, Å

Rel. Diff.

ˆ P/∆G

Rel. Diff.

Estimated ˆ P/∆G

3 30 104

1 1 1

7 × 106 7 × 102 6 × 10–8

1 0.01 10–1/2

7 × 106 7 6 × 10–12

(κT∇ ln n ) ≡ Pˆ =

2

PP mP

P

8πµ eff ρPR 4P

.

Here µeff is the effective viscosity of the cytoplasm, ρp is particle density, and Rp is effective particle radius. This very strong effect of particle radius suggests that protein motors will be most effective for larger particles, and calculations of Okamoto and Lightfoot [1992] bear out this suggestion. If the particle is to be held for the most part within one micrometer of the desired position and the free energy transduction of the cell as a whole corresponds to that of brain tissue, Table 115.9 shows the cost of mechanical motion is negligible for organelles, prohibitive for small metabolites and problematic for proteins. However, for small amounts of critically important metabolites and problematic for proteins. However, for small amounts of critically important metabolites, mechanical transport may still take place. ˆ The quantity, P/∆G, is the ratio of power consumption to mean cellular energy transduction, both per unit mass. It remains to be noted that estimating the cost of organelle transport accurately requires additional information, and some is available in [Okamoto and Lightfoot, 1992]. It is impossible for the cell to transport the large number of small metabolites, and a number of alternate means of segregation have developed. Among them are permselective membranes and compact enzyme clusters, making it difficult for intermediate metabolites to escape into the general cytoplasmic pool. These topics must be discussed elsewhere.

Defining Terms1 Allometry: A special form of similarity in which the property of interest scales across species with some constant power of species mass. Alveolus: A pulmonary air sac at the distal end of an airway. Arterioles: The smallest subdivision of the arterial tree proximal to the capillaries. ATP: Abbreviation of adenosine triphosphate, the source of chemical energy for a wide variety of metabolic reactions. Capillaries: The smallest class of blood vessel, between an arteriole and venule, whose walls consist of a single layer of cells. Cell: The smallest unit of living matter capable of independent functioning, composed of protoplasm and surrounded by a semipermeable plasma membrane. Convection: Mass transport resulting directly from fluid flow. Cytoplasm: The protoplasm or substance of a cell surrounding the nucleus, carrying structures within which most of the life processes of the cell take place. Cytoskeleton: An intracellular network of microtubules. 1

For more complete definitions see the Oxford Dictionary of Biochemistry and Molecular Biology, 1997.


Distal: At the downstream end of a flow system. Endothelium: A single layer of thin flattened cells lining blood vessels and some body cavities. Microcirculation: The three smallest types of blood vessels-arterioles, capillaries, and venules. Mitochondrion: Compartmentalized double-membrane self-reproducing organelle responsible for generating useable energy by formation of ATP In the average cell there are several hundred mitochondria each about 1.5 micrometers in length. Microtubule: Long, generally straight elements of the cytoskeleton, formed of the protein tubulin. Organelle: A specialized cytoplasmic structure of a cell performing a specific function. Parenchyma: The characteristic tissue of an organ or a gland, as distinguished from connective tissue. Proximal: At the upstream end of a flow system. Venules: The smallest vessels of the venous tree, distal to the capillaries.

References Adolph EF. 1949. Quantitative relations in the physiological constitutions of mammals, Science 109: 579. Alberts, B et al., 1945. The Molecular Biology of the Cell, 3rd Ed., Garland, NY. Bassingthwaighte JB, Goresky CA. Modeling in the Analysis of Solute and Water Exchange in the Microvasculature, Chap. 13, Handbook of Physiology-The Cardiovascular System IV. Bassingthwaighte, JB, Liebovitch, LS, and West, BJ, 1994. Fractal Physiology, Oxford. Bassingthwaighte, JB, Goresky, CA, and Linehan, JH. 1998. Whole Organ Approaches to Cellular Metabolism, Springer. Berg, OG. 1993. Random Walks in Biology, expanded edition, Princeton. Berg OG, von Hippel PH. 1985. Diffusion controlled macromolecular interactions. Ann. Rev. Biophys. Chem. 14:13 1. Bird RB, Stewart WE, Lightfoot EN. 1960. Transport Phenomena, Wiley, New York. Bray, D. Signaling Complexes: Biophysical Constraints on Intracellular Communication, Ann. Rev. Biophys. Biomol. Struct. 1998, 27, 59-75. Bunk, S. Do Energy Transport Systems Shape Organisms?, 1998. The Scientist, Dec. 14-15. Calder WA. 1984. Size, Function and Life History, Harvard University Press, Cambridge, Mass. Campenot, RB, Lund, K, and Senger, DL. 1996. Delivery of Newly Synthesized Tubulin to Rapidly Growing Distal Axons of Rat Sympathetic Neurons in Compartmented Cultures, J. Cell Biol., 135, 701-709. Campbell, NA, 1996. Biology, 4th ed, Benjamin/Cummins. Carslaw HS, Jaeger JC. 1959. Conduction of Heat in Solids, 2nd ed, Oxford. Damköhler G. 1937. Einfluss von Diffusion, Strömung und Wärmetransport auf die Ausbeute bei chemische-technische Reaktionen, Der Chemieingenieur, Bd. 3, p 359. Deen WM. 1987. Hindered Transport of Large Molecules in Liquid-Filled Pores, A I Ch E 1, 33 (9): 1409. Einstein A. 1905. Annalen der Physik. 17:549. Farhi LE. Ventilation-perfusion relationships, Chap. 11 In Hdbk. of Physiol.—the Resp. System IV. Federspiel WJ and Popel AS. 1986. A Theoretical Analysis of the Effect of the Particulate Nature of Blood on Oxygen Release in Capillaries, Microvasc. Res., 32:164-189 Fishman, HA, Greenwald DR, Zare RN. 1998. Biosensors in Chemical Separations, Ann. Rev. Biophys. Biol. Mol. Struct., 27:165-198. Francis K and Palsson BO. 1997. Effective intercellular communication distances, etc., Proc. Natl. Acad. Sci. 94:12258-12262. Griffith, TM. 1996. Temporal chaos in the microcirculation, Cardiovasc. Res., 31 (3), 342-358. Gruenberg J and Maxfield FR. 1995. Membrane Transport in the Endocytic Pathway, Current Opinion ion Cell Biology, 7:552-563. Hellums, JD et al., 1996. Simulation of Intraluminal Gas Transport Processes in the Microcirculation, Ann. Biomed. Eng., 24:1-24 Hirschfelder JO, Curtiss CF, Bird RB. 1954. Molecular Theories of Gases and Liquids, Wiley, New York.


Hobbs SH, Lightfoot EN. 1979. A Monte-Carlo simulation of convective dispersion in the large airways. Resp. Physiol. 37:273. Holsstege FCP et al., 1998. Dissecting the Regulatory Circuitry of of a Eukaryotic Genome, Cell, 95:717-728. Hubal, EA et al., 1996, Mass transport models to predict toxicity of inhaled gases in the upper respiratory tract, J. App. Physiol., 80(4):1415-1427. Johnson, LR. 1998. Essential Medical Physiology, 2nd ed., Lippincott-Raven. Lauffenberger DA, Linderman JJ. 1993. Receptors, Oxford. Lightfoot EN. 1974. Transport Phenomena and Living Systems, Wiley-Interscience, New York. Lightfoot EN, Lightfoot EJ. 1997. Mass Transfer in Kirk-Othmer Encyclopedia of Separation. Maniotis, AJ, Chen CS, Ingber DE. 1997. Demonstration of mechanical connections between integrins, cytoskeletal filaments, and nucleoplasm that stabilize nuclear structure, Proc. Natl. Acad. Sci., 94:849-854. Neubauer RA and James P. 1998. Cerebral oxygenation and the recovereable brain, Neurolog. Res. 20, Supplement 1, S33-36. Okamoto GH, Lightfoot EN. 1992. Energy cost of intracellular organization. Ind. Eng. Chem. Res. 31 (3):732. Ozisik MN. 1993. Heat Conduction, 2nd ed, Wiley, New York. Popel AS. 1989. Theory of oxygen transport to tissue. Clin. Rev. Biomed. Eng. 17 (3):257. Pries, AR, TW Secomb and P Gaeghtgens, 1996, Biophysical aspects of blood flow in the microvasculature, Cardiovasc. Res., 32, 654-667. Rojnuckarin A, Kim S, Subramanian S. 1998. Brownian dynamic simulation of protein folding: Access to milliseconds time scale and beyond, Proc. Natl. Acad. Sci. 95:4288-4292. Schmidt-Nielsen Knut. 1983. Animal Physiology: Adaptation and Environment, 3rd ed, Cambridge. Schmidt-Nielsen Knut. 1984. Scaling: Why Is Animal Size So Important?. Cambridge. Suominen, PK et al., 1997. Does water temperature affect outcome of nearly drowned children?, Resuscitation, 35(2), 111-115. Taylor R, Krishna R. 1993. Multicomponent Mass Transfer, New York, Wiley. Thompson DW. 1961. On Growth and Form (an abridged edition, JT Bonner, ed.) Cambridge. Vallee RB, Sheetz MP. 1996.Targeting of Motor Proteins, Science, 271, 1539-1544. Vicini, Paolo et al., 1998. Estimation of Blood Flow Heterogeneity in Skeletal Muscle et al., Ann. Biomed. Eng., 26, 764-774. Weisz PB. 1973. Diffusion and chemical transformation-an interdisciplinary excursion. Science, 179:433. Welling PG. 1997. Pharmacokinetics, American Chemical Society.


Jordan, C. T., Van Zant, G. “The Biology of Stem Cells.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

116 The Biology of Stem Cells Craig T. Jordan Somatix Therapy Corp.

Gary Van Zant University of Kentucky Medical Center

116.1 116.2 116.3 116.4 116.5 116.6

Embryonic Stem Cells Control of Stem Cell Development Adult Stem Cells Aging of Stem Cells Other Types of Stem Cells Summary

Life for most eukaryotes, and certainly all mammals, begins as a single totipotent stem cell, the zygote. This cell contains the same complement of genes—no more and no less—as does every adult cell that will make up the organism once development is complete. Nonetheless, this cell has the unique characteristic of being able to implement every possible program of gene expression and is thus totipotent. How is this possible? It is now known that the selective activation and repression of genes distinguishes cells with different developmental potentials. Unraveling this complex series of genetic changes accompanying the progressive restriction of developmental potential during ontogeny is the realm of modern developmental biology. In contrast to the zygote, which has unlimited developmental potential, an intestinal epithelial cell or a granulocyte, for example, is a highly developed cell type that is said to be differentiated. These cells are fixed with respect to their developmental potential and thus no longer posses the ability to contribute to other tissue types. Indeed, intestinal epithelial cells and granulocytes are incapable of undergoing further division and are said to be terminally differentiated. These mature cells have therefore undergone a process whereby they each have acquired a unique and complex repertoire of functions. These functions are usually associated with the cellular morphologic features and/or enzymatic profiles required to implement a specific developmental or functional program. We will come back to the tissue dynamics of these two cell types in a later section of this chapter. Between the extremes of developmental potency represented by the zygote and terminally differentiated cells, there is obviously a tremendous number of cell divisions (roughly 244) and an accompanying restriction of this potential in zygotic progeny. The human body, for example, is composed of greater than 1013 cells—all ultimately derived from one, the zygote. Where during the developmental sequence, does restriction occur? Is it gradual or quantal? These questions are fundamental to an understanding of developmental biology in general and stem cell biology in particular.

116.1 Embryonic Stem Cells Let us consider first the ultimate human stem cell, the zygote, in more detail. As cellular growth begins, the early embryonic cells start to make a series of irreversible decisions to differentiate along various developmental pathways. This process is referred to as developmental commitment. Importantly, such


decisions do not occur immediately; rather, the zygote divides several times and proceeds to the early blastocyst stage of development while maintaining totipotency in all its daughter cells. This is evident most commonly in the phenomenon of identical twins, where two distinct yet genetically matched embryos arise from the same zygote. The ability of early embryonic cells to maintain totipotency has been utilized by developmental biologists as a means to experimentally manipulate these embryonic stem cells, or ES cells, as they are commonly known. In 1981, two scientists at Cambridge, Evans and Kaufman, were able to isolate ES cells from a blastocyst-stage mouse embryo and demonstrate that such cells could be cloned and grown for many generations in vitro [Kaufman et al., 1983]. Remarkably, under the appropriate culture conditions, such cells remained completely totipotent. That is to say, upon reimplantation into another embryo, the stem cells could grow and contribute to the formation of an adult mouse. Importantly, the ES cells retained the developmental potential to differentiate into all the possible adult phenotypes, thereby proving their totipotency. Thus, in culture, the ES cells were said to self-renew without any subsequent loss of developmental potential. Upon reintroduction into the appropriate environment, the ES cells were able to differentiate into any of the various mature cell types. The basic decision of whether to self-renew or differentiate is a common theme found in stem cells of many developmental systems. Generally self-renewal and differentiation go hand in hand, i.e., the two events are usually inseparable. The important findings of Evans and Kaufman demonstrated that self-renewal and differentiation could be uncoupled and that an extended self-renewal phase of growth was attainable for mammalian embryonic cells. The advent of ES cell technology has had an enormous impact on the field of mammalian molecular genetics. The ability to culture ES cells was quickly combined with molecular techniques which allow for the alteration of cellular DNA. For example, if an investigator were interested in the function of a gene, he or she might elect to mutate the gene in ES cells so that it was no longer functional. The genetically altered ES cells would then be used to generate a line of mice which carry the so-called gene knockout [Koller & Smithies, 1992; Robertson, 1991]. By examining the consequences of such a mutation, clues to the normal function of a gene may be deduced. Techniques such as these have been widely used over the past 10 years and continue to develop as even more powerful means of studying basic cellular function become available.

116.2 Control of Stem Cell Development The concepts of self-renewal and differentiation are central to the description of a stem cell; indeed, the potential to manifest these two developmental options is the only rigorous criterion used in defining what constitutes a true stem cell. Consequently, in studying stem cells, one of the most important questions to consider is how the critical choice whether to self-renew or differentiate is made. As seen in the case of ES cells, the environment of the stem cell or extrinsic signals determine the outcome of the self-renewal decision. In other words, such cells are not intrinsically committed to a particular developmental fate; rather, their environment mediates the differentiation decision. Surprisingly, for ES cells this decision is dictated by a single essential protein, or growth factor, known as Leukemia inhibitory factor (LIF) [Hilton & Gough, 1991; Smith et al., 1992]. In the presence of sufficient concentrations of LIF, ES cells will self-renew indefinitely in culture. Although ES cells eventually lose their totipotency in vitro, this is thought to be due to technical limitations of ex vivo culture rather than developmental decisions by the cells themselves. Interestingly, the default decision for ES cells appears to be differentiation, i.e., unless the cells are prevented from maturing by the presence of LIF, they will quickly lose their totipotent phenotype. Upon beginning the differentiation process, ES cells can be steered along a variety of developmental pathways simply by providing the appropriate extrinsic signal, usually in the form of a growth factor. Unfortunately, the control of other types of stem cells has proven more difficult to elucidate. In particular, the control of blood-forming, or hematopoietic stem cells, has been extensively studied, but as yet the developmental control of these cells is poorly understood.


116.3 Adult Stem Cells As the mammalian embryo develops, various organ systems are formed, and tissue-specific functions are elaborated. For the majority of mature tissues, the cells are terminally differentiated and will continue to function for extended periods of time. However, some tissues operate in a much more dynamic fashion, wherein cells are continuously dying and being replenished. Thee tissues require a population of stem cells in order to maintain a steady flow of fresh cells as older cells turn over. Although there are several examples of such tissue types, perhaps the best characterized are the hematopoietic system and the intestinal epithelia. These two cell types have population parameters that call for their continuous and rapid production: Both cell types occur in very large numbers (approximately 1011 to 1012 for the human hematopoietic system) and have relatively short life spans that can often be measured in days or sometimes even hours [Kroller & Pallson, 1993]. These two tissues in adults are therefore distinct (along with skin epithelium) in that they require tissue-specific stem cells in order to satisfy the inherent population dynamics of the system. Stem cells of this nature represent a population arrested at an intermediate level of developmental potency that permits them to perform the two classic stem cell functions: They are able to replenish and maintain their own numbers through cell divisions that produce daughter cells of equivalent potency, that is, self-renew. And, they have the capacity, depending on need, to differentiate and give rise to some, if not all, of the various mature cell types of that tissue. Stem cells of the small intestine, to the best of our knowledge, give rise to at least four highly specialized lineages found in the epithelium: Paneth, goblet, enteroendocrine, and enterocytes; these stem cells are therefore pluripotent (i.e., they have the potential to give rise to many different, but not all, lineages) [Potten & Loeffler, 1990]. Similarly, pluripotent stem cells of the hematopoietic system give rise to an even wider variety of mature cells, including at least eight types of blood cells: the various lymphocytes, natural killer cells, megakaryocytes, erthroid cells, monocytes, and three types of granulocytes [Metcalf & Moore, 1971]. Proof of the existence of gut and hematopoietic stem cells came from studies of the effects of ionizing radiation on animals. This research, in the 1940s and 1950s, was spurred by concern over military and peaceful uses of atomic energy. It became recognized that the organ/tissue systems most susceptible to radiation damage were those that normally had a high turnover rate and were replenished by stem cell populations, i.e., the gut lining and the hematopoietic system. In particular, the latter was found to be the radiation dose-limiting system in the body. It was subsequently discovered that mice could be rescued from imminent death from radiation “poisoning” by the transfusion of bone marrow cells following exposure [Barnes et al., 1959]. Initially it was not clear whether the survival factor was humoral or cellular, but mounting evidence pointed to a cell-mediated effect, and, as the dose of bone marrow cells was titrated to determine the number required for survival, it was found that low numbers of cells resulted in the development of macroscopic nodules on the spleens of irradiated mice.1 These nodules were composed of cells of some but not all lineages of blood formation—lymphopoiesis was notably missing [Till & McCulloch, 1961]. Low-level radiation was then used to induce unique chromosomal aberrations in bone marrow cells prior to transplantation into lethally irradiated recipients. In this way, unique microscopically identifiable translocations would be passed on to all progeny of an altered cell. It was found that the spleen nodules were, in fact, colonies of hematopoietic cells, all possessing an identical chromosomal marker [Becker al., 1963]. This observation strongly suggested that the nodule was a clonal population, derived from a single stem cell. Since several lineages were represented in the spleen colony, the stem cell responsible was pluripotent. Moreover, single spleen colonies could be isolated and injected into secondary irradiated recipients and give rise to additional spleen colonies. This suggested that the cell giving rise to a spleen colony was capable of some degree of self-renewal as well as multilineage 1Injected bone-marrow-derived stem cells lodge and develop in several types of hematopoietic tissue, including spleen. Apparently, the splenic microenvironment can at least transiently support stem cell growth and development.


differentiation [Till et al., 1964]. The cells which give rise to spleen colonies were termed CFU-S for colony-forming unit-spleen and have been studied extensively in the characterization of pluripotent hematopoietic stem cells. More recent studies employing a similar strategy have used retroviruses to mark stem cells. The site of viral integration in an infected host cell genome is random and is passed with high fidelity to all progeny, thus by molecular means stem cell clones may be identified. Such an approach allows for the analysis not only of spleen colonies but of all anatomically dispersed lymphohematopoietic sites, including bone marrow, spleen, thymus, lymph nodes, and mature blood cells in the circulation [Dick et al,., 1985; Keller et al., 1985; Lemischka et al., 1986]. These analyses unequivocally showed that the same stem cell may give rise to all lineages, including lymphocytes. Repetitive blood cell sampling and analysis gave a temporal picture of the usage and fate of the stem cell population. In the first few weeks and months after transplant of nonlimiting numbers of stem cells, polyclonal hematopoiesis was the rule; however, after approximately 4-6 months, the number of stem cell clones was reduced. In fact, in some cases, a single stem cell clone was responsible for all hematopoiesis for over a year–about half the mouse’s lifetime [Jorday & Lemischka, 1990]. These data were interpreted to mean that many stem cell clones were initially active in the irradiated recipient mice, but over time a subset of stem cells grew to dominate the hematopoietic profile. This implies either that not all the stem cells were equivalent in their developmental potential or that the stem cells were not all seeded in equivalent microenvironments and therefore manifest differing developmental potentials. Both these possibilities have been supported by a variety of subsequent experiments; however, the details of this observation remain cloudy. One piece of evidence suggesting intrinsic differences at the stem cell level comes from studies of allophenic mice [Mintz, 1971]. These artificially generated strains of laboratory mice are created by aggregating the early embryos of two distinguishable mouse strains. As mentioned previously, early embryonic cells are totipotent; thus, upon combining such cells from two strains, a chimeric mouse will arise in which both cell sources contribute to all tissues, including the stem cell population. Patterns of stem cell contribution in allophenic mice show that one strain can cycle more rapidly and thus contribute to mature blood cells early in life, whereas the other slow-growing strain will arise to dominate at later times. Importantly, upon reimplantation of allophenic bone marrow into a secondary irradiated recipient, the two phases of stem cell activity are recapitulated. Thus, the stem cells which had become completely quiescent in the primary animal were reactivated initially, only to be followed by a later phase of activity from the second strain [Van Zant, 1992]. These data suggest that intrinsic differences at the stem cell level rather than the local microenvironment can mediate differences in stem cell activity. Unlike most organ systems, mature cells of the lymphohematopoietic system are dispersed either in the circulation or in scattered anatomic sites such as the thymus, lymph, nodes, spleen, and, in the case of macrophages, in virtually all tissues of the body. The site of production of most of the mature cells, the bone marrow, is a complex tissue consisting of stromal elements, stem and progenitor cells, maturing cells of multiple lineages, and capillaries and sinusoids of the circulatory system into which mature cells are released. Spacial relationships between these different components are not well understood because of their apparently diffuse organization and because of the paucity of some of the critical elements, most notably the stem cells and early progenitors. In contrast, the small intestinal epithelium has a much more straightforward organization that has expedited the understanding of some of the critical issues having to do with stem cell differentiation. Numerous fingerlike projections of the epithelium, called villi, extend into the intestinal lumen to effectively increase the surface area available for absorption. Each villus is covered with approximately 3500 epithelial cells, of which about 1400 are replaced on a daily basis [Potten & Loeffler, 1990]. Surrounding each villus are 6–10 crypts from which new epithelial cells are produced. They subsequently migrate to villi, and as senescent cells are shed from the villus tip, a steady progression of epithelial cells proceed unidirectionally to replace them. Crypts consist of only about 250 cells, including what is now estimated to be 16 stem cells. Since there are about 16 cells in the circumference of the crypt, stem cells occupy one circumferential ring of the crypt interior. This ring has been identified as the fourth from the bottom, directly above the Paneth cells. In addition, the fifth circumferential ring is occupied by


direct progeny of stem cells which retain pluripotency and, in emergencies, may function as stem cells. Given the detailed quantitative information available regarding stem cell numbers and epithelial cell turnover rates, the number of stem cell doublings occurring during a human life span of 70 years has been estimated to be about 5000. Whether this demonstrates that tissue-specific stem cells are immortal is the topic of the following section.

116.4 Aging of Stem Cells Given the zygote’s enormous developmental potential and that ES cells represent cells of apparently equivalent potency that can be propagated as cell lines, it is reasonable to ask whether aging occurs at the cellular level. Put another way, can normal cells, other than ES cells, truly self-replicate without differentiation or senescence? Or do ES cells represent unique examples of cells capable of apparently indefinite self-renewal without differentiation? One of the definitions of hematopoietic stem cells alluded to above, is that they self-replicate. Without self-renewal, it might be argued, a stem cell population may be exhausted in a time-frame less than a lifetime of normal hematopoiesis or in far less time in the event of unusually high hematopoietic demands associated with disease or trauma. If, for example, hematopoietic stem cells can be propagated in vitro without differentiation, it could have tremendous impact on a number of clinically important procedures including bone marrow transplantation and gene therapy. In classic experiments studying fibroblast growth in vitro, Hayflick [1965] observed that there were a finite number of divisions (about 50) that a cell was capable of before reaching senescence. It has been thought that totipotent and pluripotent stem cells may be exempt from this constraint. An analysis above of intestinal epithelial stem cells suggested that in a lifetime they undergo several thousand replications, apparently without any loss in developmental potential. However, several lines of evidence call into question the immortality of hematopoietic stem cells and point to at least a limited self-renewal capacity. For example, studies in which marrow was serially transplanted from primary recipients to secondary hosts, and so on, the number of effective passages is only about four to five [Siminovitch et al., 1964]. After the first transplant, normal numbers of relatively late progenitors are produced, but the number of repopulating stem cells is either diminished or the cells’ developmental potential attenuated, or both, resulting in a requirement for successively larger numbers of transplanted cells to have achieve engraftment. Another interpretation of these results is that the transplantation procedure itself is responsible for the declining repopulating ability of marrow, rather than an intrinsic change in the self-renewal capacity of the stem cells. According to this argument, the repetitive dissociation of stem cells from their normal microenvironmental niches in the marrow, and the required reestablishment of those contacts during seeding and serial engraftment, irreversibly alter their self-renewal capacity [Harrison et al., 1978]. A mechanistic possibility for this scenario is that differentiation is favored when stem cells are not in contact with their stromal microenvironment. In this context, exposure to growth factors has an overwhelming differentiating influence on stem cells in suspension that is normally tempered by stromal associations. Recently, an intriguing series of findings has emerged which may at least partially explain cellular aging. At the end of chromosomes there is a specialized region of DNA known as a telomere. This segment of DNA is comprised of hundreds of short six-nucleotide repeats of the sequence TTAGGG. It has been found that the length of telomeres varies over the life of a cell. Younger cells have longer telomeres, and as replication occurs the telomeres can be seen to shorten. It is thought that via the normal DNA replication process, the last 50-200 nucleotides of the chromosome fail to be synthesized, and thus telomeres are subject to loss with every cell division (reviewed in Blackburn [1992] and Greider [1991]). Consequently, telomeric shortening may act as a type of molecular clock. Once a cell has undergone a certain number of divisions, i.e., aged to a particular extent, it becomes susceptible to chromosome destabilization and subsequent cell death. Importantly, the rate at which telomeric sequence is lost may not be constant. Rather, some cells have the ability to regenerate their telomeres via an enzymatic activity known as telomerase. By controlling the loss of telomeric sequence, certain cell types may be able to extend their ability to replicate. Perhaps primitive tissue such as ES cells, when cultured with LIF, are able to express high levels of telomerase and thereby maintain their chromosomes indefinitely. Similarly, © 2000 by CRC Press LLC

perhaps early hematopoietic stem cells express telomerase, and as differentiation occurs, the telomerase activity is downregulated. Although intriguing, these hypotheses are very preliminary, and much more basic research will be required to elucidate the mechanisms of the stem cell replication and aging. In contrast to normal mechanisms of preserving replicative ability, a type of aberrant self-renewal is observed in the phenomenon of malignant transformation, or cancer. Some hematopoietic cancers are thought to originate with a defect at an early stem or progenitor cell level (e.g., chronic myelogenous leukemia). In this type of disease, normal differentiation is blocked, and a consequent buildup of immature, nonfunctional hematopoietic cells is observed. Malignant or neoplastic growth comes as a consequence of genetic damage or alteration. Such events range from single nucleotide changes to gross chromosomal deletions and translocations. Mechanistically, there appear to be two general types of mutation which cause cancer. One, activation of so-called oncogenes, is a dominant event and only needs to occur in one of a cell’s two copies of the gene. Second, inactivation of tumor-suppressor genes removes the normal cellular control of growth and results in unchecked replication. These two categories of genetic alteration are analogous to stepping on a car’s accelerator versus releasing the brake; both allow movement forward. Importantly, malignancy often comes as the result of a multistep process, whereby a series of genetic alterations occur. This process has been shown to involve different combinations of genes for different diseases [Vogelstein & Kinzler, 1993].

116.5 Other Types of Stem Cells Other tissues may also have stem cell populations contributing to the replacement of effete mature cells. For example, in the liver only a very small faction (2.5–5 × 10–5) of hepatocytes is dividing at any given time, resulting in a complete turnover time of about 1 year [Sell, 1994]. This compares with the complete turnover of intestinal epithelia or granulocytes in a period of a few days. Nonetheless, growing evidence, some of which remains controversial, suggests that hepatic stem cells play a role in this tissue turnover. A moderate loss of liver tissue due to mild or moderate insult is probably replaced by the division of mature hepatocytes. However, a severe loss of hepatic tissue is thought to require the enlistment of the putative stem cell population, morphologically identified as oval cells. Similarly, Noble’s group in London has identified cells in the rat optic nerve that have the requisite functions of stem cells [Wren et al., 1992]. These cells, called oligodendrocyte-type 2 astrocyte (O-2A) progenitors, are capable of long-term self-renewal in vitro and give rise to oligodendrocytes through asymmetric divisions resulting in one new progenitor and one cell committed to differentiation. Conversion in vitro of O-2A progenitors into rapidly dividing and differentiating cells has been shown to be regulated extrinsically by platelet-derived growth factor (PDGF) and the basic form of fibroblast growth factor (bFGF) [Wolswijk & Noble, 1992]. Since these two growth factors are known to be produced in vivo after brain injury, a mechanism is suggested for generation of the large numbers of new oligodendrocytes required subsequent to trauma and demyelination.

116.6 Summary • The zygote is the paradigm of a totipotent stem cell. • ES cells derived from early embryos can be propagated indefinitely and maintain totipotency when cultured in the presence of LIF. Experimental control of differentiation and self-renewal in adult stem cells is being extensively investigated. • Stem cells are defined by two characteristic traits: (1) Stem cells can self-renew, and (2) they can produce large numbers of differentiated progeny. Stem cells possess the intrinsic ability to manifest either trait; however, extrinsic factors mediate their developmental fate. • Tissue-specific stem cells are pluripotent but not totipotent. • Intestinal, epithelial, and hematopoietic tissues are classical self-renewing systems of the adult. In addition, recent studies have indicated the presence of stem cells in several other tissues (e.g., liver, nervous system).


• Although stem cells clearly have an extensive replication potential, it is not clear whether they are truly immortal. Stem cells may possess the ability to circumvent normal cellular processes that determine aging at the cellular level. • Mutations at the DNA level can alter normal cellular control of stem cell replication and differentiation. This type of even can lead to aberrant development and subsequent malignancy.

Defining Terms Commitment: The biologic process whereby a cell decides which of several possible developmental pathways to follow. Differentiation: Expression of cell- or tissue-specific genes which results in the functional repertoire of a distinct cell type. ES cells: Mouse stem cells originating from early embryonic tissue, capable of developing into any of the adult cell types. Gene knockout: Deletion or alteration of a cellular gene using genetic engineering technology (generally performed on ES cells). Hematopoietic: Blood forming. Ontogeny: The process of development, generally referring to development from the zygote to adult stages. Pluripotent: Capable of differentiation into multiple cell types. Self-renew: Term describing cellular replication wherein no developmental commitment or differentiation takes place. Terminally differentiated: The final stage of development in which all cell-specific features have been attained and cell division is no longer possible. Totipotent: Capable of differentiation into all possible cell types.

References Barnes DWH, Ford CE, Gray SM, et al. 1959. Progress in Nuclear Energy, series VI: Spontaneous and Induced Changes in Cell Populations in Heavily Irradiated Mice, London, Pergamon Press. Becker AJ, McCulloch EA, Till JE. 1963. Cytological demonstration of the clonal nature of spleen colonies derived from transplanted mouse marrow cells. Nature 197:452. Blackburn EH. 1992. Telomerases. Annu Rev Biochem 61:113. Dick JE, Magil MC, Huszar D, et al. 1985. Introduction of a selectable gene into primitive stem cells capable of long-term reconstruction of the hemopoietic system of W/Wv mice. Cell 42:71. Evans MJ, Kaufman MH. 1981. Establishment in culture of pluripotent cells from mouse embryos. Nature 292:154. Greider CW. 1991. Telmeres. Curr Opin Cell Biol 3(3):444. Harrison DE, Astle CM, Delaittre JA. 1978. Loss of proliferative capacity in immumohemopoietic stem cells caused by serial transplantation rather than aging. J Exp Med 147:1526. Hayflick L. 1965. The limited in vitro lifetime of human diploid cell strains. Exp Cell Res 37:614. Hilton DJ, Gough NM. 1991 Leukemia inhibitory factor: A biological perspective. J Cell Biochem 46(1):21. Jordan CT, Lemischka IR. 1990. Clonal and systemic analysis of long-term hematopoiesis in the mouse. Genes Dev 4:220. Kaufman MH, Robertson EJ, Handyside AH, et al. 1983. Establishment of pluripotential cell lines from haploid mouse embryos. J Embryol Exp Morphol 73:249. Keller G, Paige C, Gilboa E, et al. 1985. Expression of a foreign gene in myeloid and lymphoid cells derived from multipotent hematapoietic precursors. Nature 318:149. Koller BH, Smithies O. 1992. Altering genes in animals by gene targeting. Annu Rev Immunol 10:705. Koller MR, Palsson BØ. 1993. Tissue engineering: Reconstitution of human hematopoiesis ex vivo. Biotechnol Bioeng 42:909.


Lemischka IR, Raulet DH, Mulligan RC. 1986. Developmental potential and dynamic behavior of hematopoietic stem cells. Cell 45:917. Metcalf D, Moore MAS. 1971. Haemopoietic Cells, Amsterdam, Elsevier/North-Holland. Mintz B. 1971. Methods in Mammalian Embryology, San Francisco, WH Freeman. Potten CS, Loeffler M. 1990. Stem cells: Attributes, cycles, spirals, pitfalls and uncertainties. Lessons for and from the crypt. Develop 110:1001. Robertson EJ. 1991. Using embryonic stem cells to introduce mutations into the mouse germ line. Biol Reprod 44(2):238. Sell S. 1994. Liver stem cells. Mod Pathol 7(1):105. Siminovitch L, Till JE, McCulloch EA. 1964. Decline in colony-forming ability of marrow cells subjected to serial transplantation into irradiated mice. J Cell Comp Physiol 64:23. Smith AG, Nichols J. Robertson M, et al. 1992. Differentiation inhibiting activity (DIA/LIF) and mouse development. Dev Biol 151(2):339. Till JE, McCulloch EA. 1961. A direct measurement of the radiation sensitivity of normal mouse bone marrow cells. Radiat Res 14:213. Till JE, McCulloch EA, Siminovitch L. 1964. A stochastic model of stem cell proliferation based on the growth of spleen colony-forming cells. Proc Natl Acad Sci 51:29. Van Zant G. Scott-Micus K, Thompson BP, et al. 1992. Stem cell quiescence/activation is reversible by serial transplantation and is independent of stromal cell genotype in mouse aggregation chimeras. Exp Heatol 20:470. Vogelstein B, Kinzler KW. 1993. The multistep nature of cancer. Trends Genet 9(4):138. Wolswijk G, Noble M. 1992. Cooperation between PDGF and FGF converts slowly dividing O-2A adult progenitors cells to rapidly dividing cells with characteristics of O-2A perinatal progenitor cells. J Cell Biol 118(4):889. Wren D, Wolswijk G, Noble M. 1992. In vitro analysis of the origin and maintenance of O-2A adult progenitor cells. J Cell Biol 116(10):167.


Dunn, G. A. “Cell Motility and Tissue Architecture.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

117 Cell Motility and Tissue Architecture 117.1

Directed Motile Responses in Vivo Cellular Interactions with the Fluid Phase • Cellular Interactions with the Acellular Solid Phase • Cellular Interactions with Other Cells

117.2

Graham A. Dunn King’s College London

Engineering Directed Motile Responses in Vitro Environments with Specific Properties of the Fluid Phase • Environments with Specific Properties of the Solid Phase • Environments with Specific Arrangements of Neighboring Cells

The characteristic architecture of a tissue results from an interplay of many cellular processes. In addition to the secretion of extracellular matrix, we may distinguish between processes related to the cell cycle—cell growth, division, differentiation, and death—and processes related to cell motility—cell translocation, directed motile responses, associated movements, and remodeling of the extracellular matrix. These processes are controlled and directed by cell-cell interactions, by cell-matrix interactions, and by cellular interactions with the fluid phase of the tissue. It is known that all three types of interactions can control both the speed and direction of cell translocation. This control results in the directed motile responses, which are the main subject of this chapter. Learning how to manipulate these motile responses experimentally will eventually become an essential aspect of tissue engineering. Probably the greatest challenge to tissue engineering lies in understanding the complex dynamic systems that arise as a result of feedback loops in these motile interactions. Not only is cell translocation controlled by the fluid phase, by the matrix, and by other cells of the tissue, but cell motility can profoundly influence the fluid phase, remodel the matrix, and influence the position of other cells by active cell-cell responses or by associated movements. It is often forgotten that, especially in “undifferentiated” types of tissue cells such as fibroblasts, the function of the cell’s motile apparatus is not only to haul the cell through the extracellular matrix of the tissue spaces but also to remodel this matrix and to change the positions of other cells mechanically by exerting tension on cell-matrix and cell-cell adhesions. These complex dynamic systems lie at the heart of pattern formation in the tissue, and indeed in the developing embryo, and understanding them will require not only a knowledge of the motile responses but also an understanding of the mechanism of cell motility itself. In the study of cell motility, a great deal is now known about the relative dispositions of specific molecules that are thought to contribute to the motile process, and the dynamics of their interactions are beginning to be unraveled, yet there appears to have been comparatively little progress toward a satisfactory explanation of how a cell moves. There are some molecular biologists who still believe that it is just a question of time before the current molecular genetic thrust will alone come up with the answers. But there is a rapidly growing climate of opinion, already prevalent among physicists and


engineers, that nonlinear dynamic processes such as cell motility have emergent properties that can never be completely understood solely from a knowledge of their molecular basis. Cell locomotion, like muscle contraction, is essentially a mechanical process, and a satisfactory explanation of how it works will inevitably require a study of its mechanical properties. Unlike muscle, the cellular motile apparatus is a continuously self-organizing system, and we also need to know the overall dynamics of its capacity for reorganization. These outstanding areas of ignorance are essentially problems in engineering. There is a nice analogy for this conceptual gap between the molecular and engineering aspects of biologic pattern formation in Harrison’s new book on the kinetic theory of living pattern [Harrison, 1993]: “…one cannot supplant one end of a bridge by building the other. They are planted in different ground, and neither will ever occupy the place of the other. But ultimately, one has a bridge when they meet in the middle.” This chapter is intended to encourage the building of that bridge.

117.1

Directed Motile Responses in Vivo

Cellular Interactions with the Fluid Phase Cellular responses to the fluid phase of tissue spaces are thought to be mediated largely by specific diffusible molecules in the fluid phase. By far the most important directed response is chemotaxis: the unidirectional migration of cells in a concentration gradient of some chemoattractant or chemorepellent substance. Its study dates back to long before the advent of tissue culture, since it is a widespread response among the free-living unicellular organisms. The most widely studied system in vertebrates is the chemotaxis of neutrophil leukocytes in gradients of small peptides. In the case of tissue cells, it has long been conjectured, usually on the basis of surprisingly little evidence, that the direction of certain cell migrations also might be determined by chemotaxis, and it is recently becoming clear that chemotaxis can in fact occur in several vertebrate tissue cells in culture, particularly in gradients of growth factors and related substances. Yet whether it does occur naturally, and under what circumstances, is still an area of dispute. The concentration gradients themselves are usually speculated to arise by molecular diffusion from localized sources, although a nonuniform distribution of sinks could equally explain them, and they are only likely to arise naturally in conditions of very low convective flow. Besides controlling the migration of isolated cells, there is some evidence that gradients of chemoattractants also may control the direction of extension of organized groups of cells such as blood capillary sprouts, and of cellular processes, such as nerve axons. Although the mechanisms of these responses may be closely allied to those chemotaxis, they should more properly be classed as chemotropism, since the effect is to direct the extension of a process rather than the translocation of an isolated cell. The influence on tissue architecture of chemotaxis and chemotropism is possibly to determine the relative positions of different cell types and to determine the patterns of angiogenesis and innervation—though this is still largely conjectural. Chemotaxis is potentially a powerful pattern generator if the responding cells can modify and/or produce the gradient field. Apart from chemotaxis, there also exists the possibility of mechanically mediated cellular responses to flow conditions in the fluid phase. The principal situation in vivo where this is likely to be important is in the blood vessels, and the mechanical effects of flow on the endothelial cells that line the vessel walls has been investigated.

Cellular Interactions with the Acellular Solid Phase In a tissue, the cellular solid phase is the extracellular matrix, which is usually a fibrillar meshwork, though it may take the laminar form of a basement membrane. One directed response to the fibrillar extracellular matrix is the guidance of cell migrations, during embryonic development, for example, along oriented matrix fibrils. The discovery of this response followed soon after the dawn of tissue culture and is generally attributed to Paul Weiss, who named it contact guidance in the 1930s, though Loeb and Fleisher had already observed in 1917 that cells tend to follow the “path of least resistance” along oriented


matrix fibrils (see [Dunn, 1982]). In culture, contact guidance constrains cell locomotion to be predominantly bidirectional, along the alignment axis of the matrix, whereas many embryonic migrations are predominantly unidirectional. This raises the question of whether contact guidance alone can account for these directed migrations in vivo or whether other responses are involved. The influence on tissue architecture of contact guidance is less conjectural than that of chemotaxis, since matrix alignment is more easily observed than chemical gradients, and the cells themselves are often coaligned with the matrix. Directed motion can be inferred since, in culture, this orientation of the cell shape on aligned surfaces is strongly correlated with an orientation of locomotion. In conjunction with cellular remodeling of the matrix, contact guidance becomes a potentially powerful generator of pattern. Since the cells, by exerting tension, can align the matrix, and the matrix alignment can guide the cells, a mutual interaction can arise whereby cells are guided into regions of higher cell density. Harris and colleagues [1984] have shown that such a feedback loop can result in the spontaneous generation of a regular array of cell clusters from a randomly distributed field of cells and matrix. The pattern of feather formation in birds’ skin, for example, might arise by just such a mechanism. Several other types of directed response may be mediated by the properties of the solid phase of the cell’s environment. One is chemoaffinity, which Sperry [1963] proposed could account for the specific connections of the nervous system by guiding nerve fibres along tracks marked out by specific molecules adsorbed to the surface of substratum. A similar response has been proposed to account for the directional migration of primordial germ cells.

Cellular Interactions with Other Cells Neighboring cells in the tissue environment may be considered as part of the solid phase. Thus it has been reported that when cells are used as a substratum for locomotion by other cells, directed responses such as contact guidance or chemoaffinity may occur, and these responses may persist even if the cells used as a substratum are killed by light fixation. However, the reason that cell-cell interactions are dealt with separately here is that cells show a directed response on colliding with other living cells that they do not show on colliding with cells that have been lightly fixed. Contact inhibition of locomotion, discovered by Abercrombie and Heaysman [1954], is a response whereby a normal tissue cell, on collision with another, is halted and eventually redirected in its locomotion. The effect of the response is to prevent cells from using other cells as a substratum, and a distinguishing feature of the response to living cells is a temporary local paralysis of the cell’s active leading edge at the site of contact, which does not generally occur with a chemoaffinity type of response. The influence on tissue architecture of contact inhibition is probably profound but not easily determined. It is possibly the main response by which cells are kept more or less in place within a tissue, rather than milling around, and a failure of contact inhibition is thought to be responsible for the infiltration of a normal tissue by invasive malignant cells. Contact inhibition can cause cell locomotion to be directed away from centers of population and thus gives rise to the radial pattern of cell orientation that is commonly observed in the outgrowths from explant cultures. A major question is whether this motile response is related, mechanistically, to the so-called contact inhibition of growth, which is also known to fail in malignant cells but which is probably mediated by diffusion of some signal rather than by cell-cell contact.

117.2

Engineering Directed Motile Responses in Vitro

The investigation of the directed motile responses, using tissue culture, can reveal many aspects of the mechanisms that control and direct cell motility in vivo and also give a valuable insight into the mechanism of cell motility. In fact, most of what we know about these responses in vivo is deduced from experiments in culture. On the other hand, many of the responses discovered in culture may never occur naturally, and yet their study is equally important because it may yield further clues to the mechanism of cell motility and result in valuable techniques for tissue engineering. However, responses to properties of the culture environment, such as electric or magnetic fields, that are not yet known to have any counterparts in vivo will not be dealt with here. © 2000 by CRC Press LLC

A general experimental approach to engineering motile responses is first to try to reproduce in culture some response that appears to occur in vivo. The main reason is that the cell behavior can be observed in vitro, whereas this is possible in vivo only in very few instances. Another important reason is that once achieved in vitro, the response may be “dissected,” by progressively simplifying the culture environment, until we isolate one or more well-defined properties that can each elicit a response. To be successful, this approach not only requires the design of culture environments that each isolate some specific environmental property and simultaneously allow the resulting cell behavior to be observed but also requires methods for adequately quantifying this resulting behavior. When designing an artificial environment, it is important to consider its information content. Obviously, a uniform field of some scalar quantity is isotropic and cannot, therefore, elicit a directed motile response. A nondirected motile response, such as a change in speed caused by a change in some scalar property, is generally called a kinesis. Anisotropic uniform fields may be vector fields or may be able to distinguish opposite directions in the field and exhibit a unidirectional response, generally known as a taxis. Or cell movement perpendicular to the vector may predominate, which is sometimes known as a diataxis. In the case of a uniform field of some symmetric second-order tensor, such as strain or surface curvature, there is simply no information to distinguish opposite directions in the field, and yet orthogonal axes may be distinguished. This can give rise to a bidirectional response or, in three-dimensional fields, also to a response where translocation along one axis is suppressed. There is no generally agreed on name to cover all such possible responses, but the term guidance will be used here. Some examples of culture environments with specific physiocochemical properties are given below.

Environments with Specific Properties of the Fluid Phase A Linear Concentration Gradient The most common method of reproducing the chemotaxis response in vitro is to use a Boyden chamber in which a gradient of some specific chemical is formed by diffusion across a membrane filter. The resulting directed cell translocation is inferred from the relative number of cells that migrate through the pores of the filter from its upper to its lower surface. While this system is very useful for screening for potential chemoattractants, its usefulness for investigating the mechanism of the motile response is strictly limited, since it does not fulfill our two main criteria for an in vitro system. In the Boyden chamber, the properties of the environment are not well defined (since the gradient within the narrow, often tortuous pores of the filter is unpredictable), and the cell response cannot be observed directly. The Zigmond chamber was introduced to overcome these difficulties, by allowing the cell behavior to be observed directly, but the gradient is very unstable and cannot be maintained reliably for longer than an hour or two. Zicha and colleagues [1991] have recently developed a direct viewing chemotactic chamber with much greater stability and better optical properties. The chamber is constructed from glass and has an annular bridge separating concentric wells (Fig. 117.1). When covered by a coverslip carrying the cells,

FIGURE 117.1


The Dunn Chemotaxis chamber.

a gap of precisely 20 µm is formed between coverslip and bridge in which the gradient develops. The blind central well confers the greater stability, which allows chemotactic gradients to be maintained for many hours and thus permits the chemotactic responses of slowly moving tissue cells and malignant cells to be studied for the first time. Weber Scientific International, Ltd., manufactures this as the Dunn chamber. In use, both wells of the chamber are initially filled with control medium. The coverslip carrying the cells is then inverted over the wells, firmly seated, and sealed with wax in a slightly offset position (shown by the dashed lines) to allow access to the outer well. The outer well is then emptied using a syringe, refilled with medium containing the substance under test at known concentration, and the narrow opening is sealed with wax. Assuming that diffusion is the only mechanism of mass transport in the 20-µm gap between coverslip and bridge, whereas convection currents keep the bulk contents of the two wells stirred, then the concentration in the 20-µm gap as a function of distance r from the center of the inner well is given by

()

Cr =

( ) ( ) ln (b a )

Ci ln b r + Co ln r a

(117.1)

where Ci and Co are the concentrations in the inner and outer wells, respectively, and a and b are the inner and outer radii of the bridge. Because the bridge is annular, the gradient is slightly convex, but the deviation from linearity is very small. Figure 117.2 shows the formation of the gradient during the first hour for a molecule with diffusion coefficient D = 13.3 × 10–5 mm2/s, such as a small globular protein of molecular weight 17 kD and chamber dimensions a = 2.8 mm, b = 3.9 mm. The equations describing gradient formation are given in Zicha et al. [1991], but it suffices here to show that the gradient is almost linear after 30 minutes. The flux from outer to inner well though the gap of height h ( = 20 µm) is given by

( ) ( )

dQ 2πhD Co − Ci = dt ln b a

FIGURE 117.2


Formation of the gradient in the Dunn chemotaxis chamber (see text).

(117.2)

This flux tends to destroy the gradient, and the half-life of the gradient is equal to (ln 2)/kD, where k is a constant describing the geometric properties of the chamber. For a chamber with volumes vo and vi of the outer and inner wells, respectively, k is given by

k=

(

2πh vi + vo

( )

)

ln b a vivo

(117.3)

Thus, for our small protein, a chamber with vo = 30 µl, vi = 14 µl, and other dimensions as before gives a gradient with a half-life of 33.6 hours. This is ample time to study the chemotaxis of slowly moving tissue cells, with typical speeds of around 1 µm per minute, as well as permitting the study of long-term chemotaxis in the more rapidly moving leukocytes with typical speeds around 10 µm per minute. From the point of view of information content, a linear concentration gradient may be viewed as a nonuniform scalar field of concentration or as a uniform vector field of concentration gradient. Thus, if the cell can distinguish between absolute values of concentration, as well as being able to detect the gradient, then the chemotaxis response may be complicated by a superimposed chemokinesis response. A stable linear concentration gradient is a great advantage when trying to unravel such complex responses. Other sources of complexity are the possibilities that cells can modify the gradient by acting as significant sinks, can relay chemotactic signals by releasing pulses of chemoattractant, or can even generate chemotactic signals in response to nonchemotactic chemical stimulation. These possibilities offer endless opportunities for chaotic behavior and pattern generation. A Gradient of Shear Flow The flow conditions in the fluid phase of the environment can have at least two possible effects on cell motility. First, by affecting mass transport, they can change the distribution of molecules, and second, they can cause a mechanical shear stress to be exerted on the cells. Figure 117.3 shows a culture chamber designed by Dunn and Ireland [1985] that allows the behavior of cells to be observed under conditions of laminar shear flow. If necessary, the cells may be grown on the special membrane in Petriperm culture dishes (Heraeus), which allows gaseous exchange into the medium beneath the glass disk. Laminar shear

FIGURE 117.3


A culture chamber for laminar shear flow.

flow is probably the simplest and best defined flow regime, and the shear flow produced by an enclosed rotating disk has been described in detail [Daily and Nece, 1960]. Very low flow rates, caused by rotating the disk at around 1 rpm with a separation of about 1 mm between disk and cells, are useful for causing a defined distortion to diffusion gradients arising as a result of cell secretion or adsorption of specific molecules. Much information about, for example, chemoattractants produced by cells may be obtained in this way. Higher shear stresses, around 100 times greater, are known to affect the shape and locomotion of cells mechanically, and higher rotational speeds and/or smaller separations will be needed to achieve these.

Environments with Specific Properties of the Solid Phase Aligned Fibrillar Matrices The bidirectional guidance of cells by oriented fibrillar matrices is easily replicated in culture models. Plasma clots and hydrated collagen lattices are the most commonly used substrates, and alignment may be achieved by shear flow, mechanical stress, or strong magnetic fields during the gelling process. The magnitude and direction of orientation may be monitored, after suitable calibration, by measuring the birefringence in a polarizing microscope equipped with a Brace-Köhler compensator. These methods of alignment result in environments that are approximately described as uniform tensor fields, since mechanical strain is a familiar second-order tensor, and they generally give rise to bidirectional motile responses. In three dimensions, the strain tensor that results from applying tension along one axis can be described as a prolate ellipsoid, whereas applying compression along one axis results in an oblate ellipsoid. Thus we might distinguish between prolate guidance, in which bidirectional locomotion predominates along a single axis, and oblate guidance, in which locomotion is relatively suppressed along one of the three axes. But unidirectional information can be imposed on an oriented fibrillar matrix. Boocock [1989] achieved a unidirectional cellular response, called desmotaxis, by allowing the matrix fibrils to form attachments to the underlying solid substratum before aligning them using shear flow. This results in an asymmetrical linkage of the fibrils, and similar matrix configurations may well occur in vivo. Fibrillar environments in culture are probably good models for the type of cell guidance that occurs in vivo, but they are so complex that it is difficult to determine which anisotropic physicochemical properties elicit the cellular response. Among the possibilities are morphologic properties (the anisotropic shape or texture of the matrix), chemical properties (an oriented pattern of adhesiveness or specific chemical affinity) and mechanical properties (an anisotropic viscoelasticity of the matrix). One approach to discovering which properties are dominant in determining the response is to try to modify each in turn while leaving the others unaffected. This is difficult, though some progress has been made. Another approach is to design simpler environments with better defined properties, as described in the sections that follow. Specific Shapes and Textures Ross Harrison, the inventor of tissue culture, placed spiders’ webs in some cultures and reported in 1912 that “The behavior of cells … shows not only that the surface of a solid is a necessary condition but also that when the latter has a specific linear arrangement, as in the spider web, it has an action in influencing the direction of movement.” It is hardly surprising to us today that a cell attached to a very fine fiber, in a fluid medium, is constrained to move bidirectionally. Nevertheless, such stereotropism or guidance by substratum availability is a guidance response and may have relevance to the problem of guidance by aligned fibrillar matrices. Moreover, in the hands of Paul Weiss, the guidance of cells on single cylindrical glass fibers was shown to have a more subtle aspect and it is now known that fibers up to 200 µm in diameter, which have a circumference approximately 10 times greater than the typical length of a fibroblast, will still constrain the locomotion to be parallel with the fiber axis. And so we may distinguish guidance by substratum shape, sometimes known as topographic or morphographic guidance, from guidance by substratum availability. The surface curvature of a cylinder is a uniform tensor field, and since opposite directions within the surface are always equivalent, the cellular response must be bidirectional.


FIGURE 117.4

Diagrammatic cross-section of fibroblasts on substrata of various shapes.

Figure 117.4a is a diagrammatic cross section of a fibroblast attached to a convex cylindrical surface. These surfaces are easily made to any required radius of curvature by pulling glass rod in a flame. Dunn and Heath [1976] speculated that a cell must have some form of straightedge in order to detect slight curvatures of around 100 µm in radius. Obvious candidates were the actin cables that extend obliquely into the cytoplasm from sites of adhesion. These cables or stress fibers are continually formed as the fibroblast makes new adhesions to the substratum during locomotion and are known to contact and thereby to exert a tractive force on the substratum. The bundles of actin filaments are shown in the diagram as sets of parallel straight lines meeting the substratum at a tangent, and it is clear that the cables could not be much longer than this without being bent around the cylinder. Dunn and Heath proposed as an explanation of cell guidance along cylinders that the cables do not form in a bent condition and hence the traction exerted by the cell is reduced in directions of high convex curvature. Further evidence for this hypothesis was found by observing cell behavior on substrata with other shapes. On concave cylindrical surfaces made by drawing glass tubing in a flame, the cells are not guided along the cylindrical axis but tend to become bipolar in shape and oriented perpendicular to the axis [Dunn, 1982]. This is to be expected, since, as shown in Fig. 117.4b, concave surfaces do not restrict formation of unbent cables but allow the cells to spread up the walls, which lifts the body of the cell clear of the substratum and thus prevents spreading along the cylinder axis. On substrata made with a sharp change in inclination like the pitched roof of a house, the hypothesis predicts that locomotion across the ridge is inhibited when the angle of inclination is greater than the angle at which the actin cables normally meet a plane substratum, as in Fig 117.4c. These substrata are more difficult to make than cylinders and require precision optical techniques for grinding and polishing a sharp and accurate ridge angle. Fibroblasts behaved as predicted on these substrata, the limiting angle being about 4 degrees, and high-voltage electron microscopy revealed the actin cables terminate precisely at the ridge on substrata with inclinations greater than this. Substrata with fine parallel grooves have long been known to be very effective for eliciting morphographic guidance and are interesting because they can be a well-defined mimicry of some of the shape properties of an aligned fibrillar matrix while being mechanically rigid. Effective substrata can easily be made by simply scratching a glass surface with a very fine abrasive, but such substrata are not well defined and their lack of uniformity may give rise to variations in macroperties such as wettability, that cannot be ruled out as causing the guidance. Early attempts to make better-defined substrata used ruling engines such as those used to make diffraction gratings, but Dunn and Brown [1986] introduced electron beam


lithography followed by ion milling to make grooves of rectangular cross section down to about 1 µm in width. Clark and colleagues [Clark et al., 1991] have now achieved rectangular grooves with spacings as low as 260 nm using the interference of two wavefronts, obtained by splitting an argon laser beam, to produce a pattern of parallel fringes on a quartz slide coated with photoresist. Groove depths a small as 100 nm can elicit a guidance response from certain cell types, and the main reason for pursuing this line of inquiry is now to discover the molecular mechanism responsible for this exquisite sensitivity. Figure 117.4c shows a diagrammatic cross section of a fibroblast on a substratum consisting of a parallel array of rectangular grooves. One question that has been debated is whether the cells generally sink into the grooves, as shown here, or bridge across them. In the latter case, the wall and floor of the grooves are not an available substratum, and the cellular response might be a form of guidance by substratum availability. Ohara and Buck [1979] have suggested that this might occur, since the focal adhesions of fibroblasts are generally elongated in the direction of cell movement, and if they are forced to become oriented by being confined to the narrow spaces between grooves, this may force the locomotion into the same orientation. On the other hand, if the cells do generally sink into the grooves, then the Dunn and Heath hypothesis also could account for guidance even by very fine grooves, since individual actin filaments in the bundles, shown as dashed lines in the inset to the figure, would become bent and possibly disrupted if the cell made any attempt to pull on them other than in a direction parallel with the grooves. It is still therefore an unresolved issue whether different mechanisms operate in the cases of cylinders and grooves or whether a common mechanism is responsible for all cases of guidance by the shape of the substratum. Other groove profiles, particularly asymmetrical ones such as sawtooth profiles, will be needed for testing these and other rival hypotheses, and an intriguing possibility is that a unidirectional cell response might FIGURE 117.5 A proposed microfabricated subbe achieved on microfabricated substrata with two stratum with two orthogonal arrays of parallel saw orthogonal arrays of parallel sawtooth grooves, as tooth grooves. shown in Fig. 117.5. Specific Patterns of Adhesiveness The equivalent in culture of the chemoaffinity response is guidance by differential adhesiveness, in which cell locomotion is confined to regions of higher adhesiveness patterned on the substrate. As with grooved surfaces, adhesive tracks that guide cells effectively are easily made by a variety of methods, including physically streaking nonadhesive viscous materials on an adhesive substratum or scratching through a nonadhesive film overlying an adhesive substratum. Again, however, these easily made surfaces are not well defined, and in particular, their anisotropic adhesiveness tends to be contaminated by anisotropic surface texture and sometimes by anisotropic mechanical properties. Carter [1965] was probably the first to describe a method of printing a well-defined pattern of adhesiveness onto a substratum; he used the vacuum evaporation of palladium, through a mask, onto a glass substratum made nonadhesive by first coating it with cellulose acetate. Clark and colleagues [1992] have now described a method for fabricating any required pattern of differential adhesiveness by using photolithography to obtain a hydrophobic pattern of methyl groups covalently coupled to a hydrophilic quartz substratum. The most recent developments in their laboratories are to use these patterns of hydrophobicity as templates for patterning specific proteins onto the substratum, and it seems that soon it will be possible to make almost any required pattern in any combination of proteins. The explanation of the guidance of cells along tracks of higher adhesiveness appears to be obvious. In extreme cases, when the cells cannot adhere at all to the substrate outside the track, then the response


is equivalent to guidance by substratum availability, and if the track happens to be sufficiently narrow, cell locomotion is restricted to the two directions along the track. But guidance along tracks of higher adhesiveness may still be very pronounced even when the cells can also adhere to and move on the regions of lower adhesiveness. The explanation in this case is that on encountering boundaries between regions of different adhesiveness, cells will cross them far more frequently in the direction from lower to higher adhesiveness. It is generally assumed that this results from a tug-of-war competition, since traction can be applied more effectively by the parts of the cell overlapping the region of higher adhesiveness. It is not known, however, how the traction fails in those parts of the cell which lose the competition, whether by breakage or slipping of the adhesions or by a relative failure of the contractile apparatus. Another possibility is simply that the cell spreads more easily over the more highly adhesive regions. One reason for studying guidance by differential adhesiveness is to discover whether it can account for guidance by oriented extracellular matrices. An array of very narrow, parallel stripes of alternately high and low adhesiveness mimics the linear arrangement of substratum availability in an aligned matrix. Dunn [1982] found that if the repeat spacing is so small that a cell can span several stripes, there is no detectable cell orientation or directed locomotion even though an isolated adhesive stripe can strongly guide cells. Clark and colleagues [1992] confirmed this observation with one cell type (BHK) but found that cells of another type (MDCK) could become aligned even when spanning several stripes but would become progressively less elongated as the repeat spacing decreased. It is not yet clear, therefore, whether the linear arrangement of substratum availability in an aligned matrix might contribute to the guidance response in some cell types. It is clear from work of Clark and colleagues, however, that the adhesive stripes become less effective in eliciting guidance as their repeat spacing decreases, whereas the opposite is true for grooved surfaces. Thus it seems unlikely that substratum availability is the mechanism of guidance by grooved surfaces and, conversely, unlikely that adhesive stripes guide cells by influencing the orientation of the focal adhesions as suggested for grooved surfaces by Ohara and Buck [1979]. Binary patterns of adhesiveness were not the only ones studied by Carter [1965]. His technique of shadowing metallic palladium by vacuum evaporation onto cellulose acetate also could produce a graded adhesiveness. By placing a rod of 0.5 mm diameter on the substratum before shadowing, he found that the penumbral regions of the rod’s shadow acted as steep gradients of adhesiveness that would cause cultured cells to move unidirectionally in the direction of increasing adhesiveness. This is therefore a taxis as distinct from a guidance response, and he named it haptotaxis. It is still not clear whether haptotaxis plays any role in vivo. Specific Mechanical Properties As yet, there has been no demonstration that anisotropic mechanical properties of the substratum can elicit directed motile responses. However, it is known that isotropic mechanical properties, such as the viscosity of the substratum, can influence cell locomotion[Harris, 1982], and it appears that changing the mechanical properties of aligned matrices can reduce cell guidance [Dunn, 1982; Haston et al., 1983], although it is probable that other properties are altered at the same time. Moreover, the phenomenon of desmotaxis suggest that it is the asymmetrical mechanical linkage of the fibrils that biases the locomotion. Guidance by anisotropic mechanical properties therefore remains a distinct possibility, but further progress is hampered by the difficulty of fabricating well-defined substrata. An ideal substratum would be a flat, featureless, and chemically uniform surface with anisotropic viscoelastic properties, and it is possible that liquid crystal surfaces will provide the answer.

Environments with Specific Arrangements of Neighboring Cells Although contact inhibition of locomotion is of primary importance in determining patterns that develop in populations of cells, it is not easy to control the effects of contact inhibition in culture. If the cells are seeded on the substratum at nonuniform density, the response will generally cause cell locomotion to be biased in the direction of decreasing cell density. This can give a unidirectional bias if superimposed on a guidance response, and it has been conjectured that certain cell migrations in vivo are biased in this


way. Cellular contact responses also can lead to the mutual orientation of confluent neighboring cells, and this can lead to wide regions of cooriented cells arising spontaneously in uniformly seeded cultures. A typical culture arrangement from studying contact inhibition is to seed two dense populations of cells, often primary explants of tissue, about 1 mm apart on the substratum [Abercrombie and Heaysman, 1976]. Homologous contact inhibition causes the cells to migrate radially from these foci, usually as confluent sheets, until the two populations collide. With noninvasive cells, their locomotion is much reduced after the populations have met, and there is little intermixing of the population boundary. If one of the two populations is of an invasive type, however, failure of heterologous contact inhibition will cause it to infiltrate the other population, and their invasiveness can be measured by the depth of interpenetration.

Defining Terms Associated movements: Occur when cells passively change position as a result of external forces, generated by cell motility elsewhere, that are transmitted either through the extracellular matrix or through cell-cell contacts. Cell motility: A blanket term that covers all aspects of movement actively generated by a cell. It includes changes in cell shape, cell contraction, protrusion and retraction of processes, intracellular motility, and cell translocation. Cell translocation, cell locomotion, or cell migration: All describe active changes in position of a cell in relation to its substratum. The translocation of tissue cells always requires a solid or semisolid substratum. In seeking a more rigorous definition, positions must first be defined for both the cell and its substratum. This is not always easy, since both may change continually in shape. Chemoaffinity: The directional translocation of cells or extension of cellular protrusions along narrow tracks of specific molecules adsorbed to the substratum. Chemokinesis: A kinesis (q.v.) in which the stimulating scalar property is the concentration of some chemical. Chemotaxis: The directional translocation of cells in a concentration gradient of some chemoattractant or chemorepellent substance. Chemotropism: The directional extension of a cellular protrusion or multicellular process in a concentration gradient of some chemoattractant or chemorepellent substance. Contact guidance: The directional translocation of cells in response to some anisotropic property of the substratum. Contact inhibition of locomotion: Occurs when a cell collides with another and is halted and/or redirected so that it does not use the other cell as a substratum. Desmotaxis: Describes a unidirectional bias of cell translocation in a fibrillar matrix that is allowed to attach to a solid support and then oriented by shear flow. Diataxis: A taxis (q.v.) in which translocation perpendicular to the field vector predominates. This leads to a bidirectional bias in two dimensions. Directed motile responses: The responses of cells to specific properties of their environment that can control the direction of cell translocation or, in the case of nerve growth, for example, can control the direction of extension of a cellular protrusion. Guidance: Used here to indicate a directed response to some high-order-tensor-like property of the environment in which opposite directions are equivalent. Translocation is biased bidirectionally in two dimensions. Guidance by differential adhesiveness: A form of guidance by substratum availability (q.v.) in which cell locomotion is wholly or largely confined to narrow tracks of higher adhesiveness on the substratum. The response may be absent or much reduced if the cell spans several parallel tracks. Guidance by substratum availability: Occurs when the translocation of a cell is confined to an isolated narrow track either because no alternative substratum exists or because the cell is unable to adhere to it.


Guidance by substratum shape, topographic guidance, or morphographic guidance: All refer to the guidance of cells by the shape or texture of the substratum. It is not known whether all types of morphographic guidance are due to a common mechanism. Haptotaxis: The tendency of cells to translocate unidirectionally up a steep gradient of increasing adhesiveness of the substratum. Heterologous contact inhibition: The contact inhibition of locomotion (q.v.) that may occur when a cell collides with another of different type. Contact inhibition is called nonreciprocal when the responses of the two participating cells are different. A cell type that is invasive with respect to another will generally fail to show contact inhibition in heterologous collisions. Homologous contact inhibition: The contact inhibition of locomotion (q.v.) that may occur when a cell collides with another of the same type. It is not always appreciated that invasive cell types can show a high level of homologous contact inhibition. Kinesis: The dependence of some parameter of locomotion, usually speed or rate of turning, on some scalar property of the environment. In an adaptive kinesis, the response is influenced by the rate of change of the scalar property, and if the environment is stable but spatially nonuniform, this can lead to behavior indistinguishable from a taxis (q.v.). Current nomenclature is inadequate to deal with such situations [Dunn, 1990]. Oblate guidance: A form of guidance in three dimensions in which translocation is suppressed along a single axis. Prolate guidance: A form of guidance in three dimensions in which translocation along a single axis predominates. Stereotropism: A form of guidance by substratum availability (q.v.) in which the only solid support available for locomotion consists of isolated narrow fibers. Taxis: A directed response to some vectorlike property of the environment. Translocation is usually biased unidirectionally, either along the field vector or opposite to it, except in the case of diataxis (q.v.).

References Abercrombie M, Heaysman JEM. 1954. Observations on the social behaviour of cells in tissue culture. Exp Cell Res 6:293. Boocock CA. 1989. Unidirectional displacement of cells in fibrillar matrices. Development 107:881. Carter SB. 1965. Principles of cell motility: The direction of cell movement and cancer invasion. Nature 208:1183. Clark P, Connolly P, Curtis ASG, et al. 1991. Cell guidance by ultrafine topography in vitro. J Cell Sci 99:73. Clark P, Connolly P, Moores GR. 1992. Cell guidance by micropatterned adhesiveness in vitro. J Cell Sci 103:287. Daily JW, Nece RE. 1960. Chamber dimension effects on induced flow and frictional resistance of enclosed rotating disks. Trans AM Soc Mech Engrs D82:217. Dunn GA. 1982. Contact guidance of cultured tissue cells: A survey of potentially relevant properties of the substratum. In R Bellairs, A Curtis, G Dun (eds), Cell Behaviour: A Tribute to Michael Abercrombie, pp 247–280. Cambridge, Cambridge University Press. Dunn GA. 1990. Conceptual problems with kinesis and taxis. In JP Armitage, JM Lackie (eds), Biology of the Chemotactic Response, pp 1–13. Society for General Microbiology Symposium, 46. Cambridge, Cambridge University Press. Dunn GA, Brown AF. 1986. Alignment of fibroblasts on grooved surfaces described by a simple geometric transformation. J Cell Sci 83:313. Dunn GA, Heath JP. 1976. A new hypothesis of contact guidance in tissue cells. EXP Cell Res 101:1. Dunn GA, Ireland GW. 1984. New evidence that growth in 3T3 cell cultures is a diffusion-limited process. Nature 312:63.


Harris AK. 1982. Traction and its relation to contraction in tissue cell locomotion. In R Bellairs, A Curtis, G Dunn (eds), Cell Behaviour: A Tribute to Michael Abercrombie, pp 109–134. Cambridge, Cambridge University Press. Harris AK, Stopak D, Warner P. 1984. Generation of spatially periodic patterns by a mechanical instability: A mechanical alternative to the Turning model. J Emb Exp Morph 80:1. Harrison LG. 1993. Kinetic Theory of Living Pattern, Cambridge, Cambridge University Press. Haston WS, Shields JM, Wilkinson PC. 1983. The orientation of fibroblasts and neutrophils on elastic substrata. Exp Cell Res 146:117. Ohara PT, Buck RC. 1979. Contact guidance in vitro: A light, transmission and scanning electron microscopic study. Exp Cell Res 121:235. Sperry RW. 1963. Chemoaffinity in the orderly growth of nerve fiber patterns and connections. Proc Natl Acad Sci USA 50:703. Zicha D, Dunn GA, Brown AF. 1991. A new direct-viewing chemotaxis chamber. J Cell Sci 99:769.


Long, M.W. “Tissue Microenvironments.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

118 Tissue Microenvironments 118.1

Cellular Elements Immunoglobulin Gene Superfamily • Integrin Gene Superfamily • Selectins

118.2 118.3

Soluble Growth Factors Extracellular Matrix Proteoglycans and Glycosaminoglycans • Thrombospondin (TSP) • Fibronectin • Collagen

Michael W. Long University of Michigan

118.4 118.5

Considerations for ex Vivo Tissue Generation Conclusions and Perspectives

Tissue development is regulated by a complex set of events in which cells of the developing organ interact with each other, with general and specific growth factors, and with the surrounding extracellular matrix (ECM) [Long, 1992]. These interactions are important for a variety of reasons such as localizing cells within the microenvironment, directing cellular migration, and initiating growth-factor-mediated developmental programs. It should be realized, however, that simple interactions such as those between cells and growth factor are not the sole means by which developing cells are regulated. Further complexity occurs via interactions of cells and growth factors with extracellular matrix or via other interactions which generate specific developmental responses. Developing tissue cells interact with a wide variety of regulators during their ontogeny. Each of these interactions is mediated by defined, specific receptor-ligand interactions necessary to stimulate both the cell proliferation and/or motility. For example, both chemical and/or extracellular matrix gradients exist which signal the cell to move along “tracks” of molecules into a defined tissue area. As well, high concentrations of the attractant, or other signals, next serve to “localize” the cell, thus stopping its nonrandom walk. These signals which stop and/or regionalize cells in appropriate microenvironments are seemingly complex. For example, in the hematopoietic system, complexes of cytokines and extracellular matrix molecules serve to localize progenitor cells [Long et al., 1992], and similar mechanisms of cell/matrix/cytokine interactions undoubtedly exist in other developing systems. Thus, the regulation of cell development, which ultimately leads to tissue formation is a complex process in which a number of elements work in cohort to bring about coordinated organogenesis: stromal and parenchymal cells, growth factors and extracellular matrix. Each of these is a key component of a localized and highly organized microenvironmental regulatory system. Cellular interactions can be divided into three classes: cell-cell, cell-extracellular matrix, and cellgrowth factor, each of which is functionally significant for both mature and developing cells. For example, in a number of instances blood cells interact with each other and/or with cells in other tissues. Immunologic cell-cell interactions occur when lymphocytes interact with antigen-presenting cells, whereas neutrophil or lymphocyte egress from the vasculature exemplifies blood cell-endothelial cell recognition. Interactions between cells and the extracellular matrix (the complex prontinaceous substance surrounding


FIGURE 118.1 Hematopoietic cellular interactions. This figure illustrates the varying complexities of putative hematopoietic cell interactions. A conjectural complex is shown in which accessory cell–stromal cell, and stromal cell–PG-growth factor complexes localize developmental signals. ECM = extracellular matrix, gag = glycosaminoglycan side chain bound to proteoglycan (PG) core protein (indicated by cross-hatched curved molecule); IL-1 = interleukin-1; GM-CSF = granulocyte-macrophage colony-stimulating factor. Modified from Long [1992] and reprinted with permission.

cells) also play an important role. During embryogenesis matrix molecules are involved both in cell migration and in stimulating cell function. Matrix components are also important in the growth and development of precursor cells; they also serve either as cytoadhesion molecules for these cells or to compartmentalize growth factors within specific microenvironmental locales. For certain tissues such as bone marrow, a large amount of information exists concerning the various components of the nature of the microenvironment. For others such as bone, much remains to be learned of the functional microenvironment components. Many experimental designs have examined simple interactions (e.g., cell-cell, cell-matrix). However, the situation in vivo is undoubtedly much more complex. For example, growth factors are often bound to matrix molecules which, in turn, are expressed on the surface of underlying stromal cells. Thus, very complex interactions occur (e.g., accessory cell–stromal cell–growth factor–progenitor cell–matrix, see Fig. 118.1), and these can be further complicated by a developmental requirement for multiple growth factors. The multiplicity of tissue-cell interactions requires highly specialized cell surface structures (i.e., receptors) to both mediate cell adhesion and transmit intracellular signals from other cells, growth factors, and/or the ECM. Basically, two types of receptor structures exist. Most cell surface receptors are proteins which consist of an extracellular ligand-binding domain, a hydrophobic membrane-spanning region, and a cytoplasmic region which usually functions in signal transduction. The amino acid sequence of these receptors often defines various families of receptors (e.g., immuno-globulin and integrin gene superfamilies). However, some receptors are not linked to the cell surface by protein, as certain receptors contain phosphotidylinositol-based membrane linkages. This type of receptor is usually associated with signal transduction events mediated by phospholipase C activation [Springer, 1990]. Other cell surface molecules important in receptor-ligand interactions are surface proteins which function as a coreceptors. Coreceptors function with a well-defined receptor, usually to amplify stimulus-response coupling. The goal of this chapter is to examine the common features of each component of the microenvironment (cellular elements, soluble growth factors, and extracellular matrix). As each tissue or organ undergoes its own unique and complex developmental program, this review cannot cover these elements for all organs and tissue types. Rather, two types of microenvironments (blood and bone) will be compared in order to illustrate commonalties and distinctions.

118.1 Cellular Elements Cells develop in a distinct hierarchical fashion. During the ontogeny of any organ, cells migrate to the appropriate region for the nascent tissue to form and there undergo a phase of rapid proliferation and


differentiation. In tissues which retain their proliferative capacity (bone marrow, liver, skin, the gastrointestinal lining, and bone), the complex hierarchy of proliferation cells is retained throughout life. This is best illustrated in the blood-forming (hematopoiesis) system. Blood cells are constantly produced, such that approximately 70-times an adult human’s body weight of blood cells is produced through the human life span. This implies the existence of a very primitive cell type that retains the capacity for self-renewal. This cell is called a stem cell, and it is the cell responsible for the engraphment of hematopoiesis in recipients of bone marrow transplantation. Besides a high proliferative potential, the stem cell also is characterized by its multipotentiality in that it can generate progeny (referred to as progenitor cells) which are committed to each of the eight blood cell lineages. As hematopoietic cells proliferate, they progressively lose their proliferative capacity and become increasingly restricted in lineage potential. As a result, the more primitive progenitor cells in each lineage produce higher colony numbers, and the earliest cells detectable in vitro produce progeny of 2–3 lineages (there is no in vitro assay for transplantable stem cells). Similar stem cell hierarchies exist for skin and other regenerating tissues, but fewer data exist concerning their hierarchical nature. The regulation of bone cell development is induced during bone morphogenesis by an accumulation of extracellular and intracellular signals [Urist et al., 1983a]. Like other systems, extracellular signals are known to be transferred from both cytokines and extracellular matrix molecules [Urist et al., 1983a] to responding cell surface receptors, resulting in eventual bone formation. The formation of bone occurs by two mechanisms. Direct development of bone from mesenchymal cells (referred to as intramembranous ossification, as observed in skull formation) occurs when mesenchymal cells directly differentiate into bone tissue. The second type of bone formation (the endochondrial bone formation of skeletal bone) occurs via an intervening cartilage model. Thus, the familiar cell hierarchy exists in the development and growth of long bones, beginning with the proliferation of mesenchymal stem cells, their differentiation into ostogenic progenitor cells, and then into osteoblasts. The osteoblasts eventually calcify their surrounding cartilage and/or bone matrix to form bone. Interestingly, the number of osteoporgenitor cells in adult bone seems too small to replace all the large mass of bone normally remodeled in the process of aging of the skeleton [Urist et al., 1983a]. Observations from this laboratory confirm this concept by showing that one (unexpected) source of bone osteoprogenitor cells is the bone marrow [Long et al., 1990; Long & Mann, 1993]. This reduced progenitor cell number also implies that there is a disassociation of bone progenitor cell recruitment from subsequent osteogenic activation and bone deposition and further suggests multiple levels of regulation in this process (vide infra). As mentioned, cell-cell interactions mediate both cellular development and stimulus-response coupling. When coupled with other interactions (e.g., cell-ECM), such systems represent a powerful mechanism for directing and/or localizing developmental regulation. Further, combinations of these interactions potentially can yield lineage-specific or organ-specific information. Much of out understanding of cell-cell interactions comes from the immune system and from the study of developing blood cells and their interactions with adjacent stromal cells [Dexter, 1982; Gallatin et al., 1986; Springer, 1990]. For example, the isolation and cloning of immune cell ligands and receptors resulted in the classification of gene families which mediate cell-cell interactions within the immune and hematopoietic systems, and similar systems undoubtedly play a role in the development of many tissues. There are three families of molecules which mediate cell-cell interactions (Table 114.1). The immunoglobulin superfamily is expressed predominantly on cells mediating immune and inflammatory responses (and is discussed only briefly here). The integrin family is a large group of highly versatile proteins which is involved in cell-cell and cell-matrix attachment. Finally, the selectin family is comprised of molecules which are involved in lymphocyte, platelet, and leukocyte interactions with endothelial cells. Interestingly, this class of cell surface molecules utilizes specific glycoconjungates (encoded by glycoslytransferase genes) as their ligands on endothelial cell surfaces.

Immunoglobulin Gene Superfamily These molecules function in both antigen recognition and cell-cell communication. The immunglobulin superfamily (Table 118.2) is defined by a 90–100 base pair immunoglobulinlike domain found within a


TABLE 118.1

Cell Adhesion Molecule Superfamilies

Immunoglobulin superfamily of adhesion receptors LFA 2 (CD2) LFA 3 (CD58) ICAM 1 (CD54) ICAM 2 VCAM-1

T-Cell receptor (CD3) CD4 (TCR coreceptor) CD8 (TCR coreceptor) MHC class I MHC class II

Integrins β1 integrins (VLA proteins) P150,95 (CD11c/CD18) VLA 1–3,6 VLA 4(LPAM 1, CO49d/CO29) Fibronectin receptor (VLA 5, CD-/CD29) LPAM 2 β2 Integrins LFA 1 (CD11a/ CD18) Mac 1 or Mo l (CD11b/CD18) β3 Integrins Vitronectin receptor Platelet gp-IIb/IIa Selectin/LEC-CAMS Mel14 (LE-CAM-1, LHR, LAM-1, Leu 8, Ly 22, gp90 MEL) ELAM-1 (LE-CAM-2) GMP 140 (LE-CAM 3, PADGEM, CD 62) Source: Originally adapted from Springer [1990] and Brandley and coworkers [1990] and reprinted from Long [1992] with permission.

dimer of two antiparallel β strands [Sheetz et al., 1989; Williams & Barclay, 1988]; for a review, see Springer [1990]. Two members of this family, the T-cell receptor and immunoglobulin, function in antigen recognition. The T-cell receptor recognizes antigenic peptides in the context of two other molecules on the surface of antigen-presenting cells: major histocompatibility (MHC) class I and class II molecules [Bierer et al., 1989; Sheetz et al., 1989; Springer, 1990]. Whereas the binding of T-cell receptor to MHC/antigenic peptide complexes seems sufficient for cell-cell adhesion, cellular activation also requires the binding of either of two coreceptors, CD8 or CD4. Neither coreceptor can directly bind the MHC complex, but, rather, each seems to interact with the T-cell receptors to synergistically amplify intercellular signaling [Shaw et al., 1986; Spits et al., 1986; Springer, 1990].

Integrin Gene Superfamily Integrin family members are involved in interactions between cells and extracellular matrix proteins [Giancotti & Ruoslahti, 1990]. Cell attachment to these molecules occurs rapidly (within minutes) and is a result of increased avidity rather than increased expression (see Lawrence and Springer [1991] and references therein). The binding sequence within the ligand for most, but not all, integrins is the tripeptide sequence arganine-glycine-asparagine (RGD) [Ruoslahti & Pierschbacher, 1987]. Structurally, integrins consist of two membrane-spanning alpha and beta chains. The alpha subunits contain three to four tandem repeats of a divalent-ion-binding motif and require magnesium or calcium to function. The alpha chains are (usually) distinct and bind with common or related β subunits to yield functional receptors [Giancotti & Ruoslahti, 1990]. The β subunits of integrins have functional significance, and integrins can be subclassified based on the presence of a given beta chain. Thus, integrins containing β1 and β3 chains are involved predominantly in cell-extracellular matrix interactions, whereas molecules containing the β2 subunits function in leukocyte-leukocyte adhesion (Tables 118.1 and 118.2). The cytoplasmic domain of many integrin receptors interacts with the cytoskeleton. For example, several © 2000 by CRC Press LLC

TABLE 118.2

Cell Surface Molecules Mediating Cell-Cell Interactions

Cell Receptor Ig superfamily MHC I MHC II ICAM-1

ICAM-2 LFA-2 Integrins Mac1 LFA-1 VCAM gp150,95 FN-R IIb/IIIa Selectins LEC-CAM-1 (Mel 14) ELAM-1 (LE-CAM-2) LEC-CAM-3 (GMP-140)

Receptor: Cell Expression

Ligand, Co-, or Counter-Receptor

Ligand Co- or CounterReceptor Cell Expression

References

Macroph, T cell Macroph, T cell Endo, neut. HPC, B cells, T cells, macroph* Endo T cells

CD8,TCR CD4, TCR LFA-1

T cells T cells Mono, T and B cells

Springer 1990 Springer 1990 Springer 1990

LFA-1 LFA-3

Mono, T and B cells T cells, eryth

Springer 1990 Springer 1990

Macroph, neut Macroph, neut Endo

Fibrinogen, C3bi (See above) VLA4

Springer 1990

Macro, neut Eryth lineage Plts, mk

(See above) Fibronectin Fibrinogen, TSP, VN vWF

Endo, plts (See above) Lymphocytes, monocytes B cells (See above) N.A. Endo

Endo

Addressins, neg. charged oligosaccrides sialyl-Lewis X†

Endo Plt gran Weible-Palade bodies, endo

Lewis X (CD15)

Brandley et al., 1990 Miyamake, 1990 see Table 118.3 Springer, 1990

Lymphocytes

Brandley et al., 1990

Endo‡ Neut, tumor cells Endo Neut

Lowe et al., 1990 Brandley et al., 1990

*Source: Modified from Long [1992] and reprinted with permission. macroph = macrophage; Mono = monocyte; Endo = endothelial cell; Eryth = erythroid cells; Plts = platelets; Neut = neutrophil; Mk = megakarocyte. †Constituatively expressed by few cells, upregulated by TNF- and IL-1. ‡Sialylated, fucosylated lactosaminoglycans. Modified from Long [1992] and reprinted with permission.

integrins are known to localize near focal cell contacts were actin bundles terminate [Giancotti & Ruoslahti, 1990; Springer, 1990]. As a result, changes in receptor binding offer an important mechanism for linking cell adhesion to cytoskeletal organization.

Selectins The selectin family of cell-adhesion receptors contains a single N-terminus, calcium-dependent, lectinbinding domain, an EGF receptor (EFGR) domain, and a region of cysteine-rich tandem repeats (from two to seven) which are homologous to complement-binding proteins [Bevilacqua et al., 1989; Springer, 1990; Stoolman, 1989]. Selectins (e.g., MEL14, gp90MEL, ELAM-1, and GMP140/PADGEM, Table 118.2) are expressed on neutrophils and lymphocytes. They recognize specific glycoconjugate ligands on endothelial and other cell surfaces. Early studies demonstrated that fucose or mannose could block lymphocyte attachment to lymph node endothelial cells [Brandley et al., 1990]. Therefore, the observation that the selectin contain a lectin-binding domain [Bevilacqua et al., 1989] led to the identification of the ligands for two members of this family; for review see Brandley and coworkers [1990]. Lowe and coworkers first demonstrated that alpha (1,3/1,4) fucosyltransferase cDNA converted nonmyeloid COS or CHO cells to selectin (sialyl-Lewis X) positive cells which bound to both HL60 cells and neutrophils in an ELAM-1-dependent manner [Lowe et al., 1990]. Conversely, Goelz and coworkers screened an expression library using a monoclonal antibody which inhibited ELAM-mediated attachment which yielded a novel alpha (1,3) fucosyltransferase whose expression conferred ELAM binding activity on target cells [Goelz et al., 1990]. Unlike mature neutrophils and lymphocytes, information on cell-cell interactions among hematopoietic progenitor cells is less well developed (see Table 118.3). Data concerning the cytoadhesive capacities


TABLE 118.3 Hematopoietic Cell–Stromal Cell Interactions (Unknown Receptor-Ligand) Cell Phenotype

Stromal Cell

References

B cells Pre-B cell BFC-E CFC-S CFC-S, B1-CFC CFC-GM CFC-Mk

Fibroblasts Heter stroma Fibroblasts Fibroblasts (NIH 3T3) Heter stroma* Heter stroma Heter stroma

Ryan et al., 1990; Witte et al., 1987 Palacios et al., 1989 Tsai et al., 1986, Tsai et al., 1987 Roberts et al., 1987 Gordon et al., 1985, 1990a, 1990b Tsai et al., 1987; Campbell et al., 1985 Tsai et al., 1987; Campbell et al., 1985

*Methylprednisolone stimulated stromal cells, unstimulated fail to bind—see Gordon and coworkers [1985]. Heter = heterologous; BFC-E = burst forming cell-erythroid; CFC = colony-forming cell; S = spleen; B1 = blast; GM = granulocyte/macrophage; Mk = megakaryocyte. From Long [1992], reprinted with permission.

of hematopoietic progenitor cells deal with the interaction of these cells with underlying, preestablished stromal cell layers [Dexter, 1982]. Gordon and colleagues documented that primitive hematopoietic human blast-colony forming cells (Bl-CFC) adhere to performed stromal cell layers [Gordon et al., 1985, 1990b] and showed that the stromal cell ligand is not one of the known cell adhesion molecules [Gordon et al., 1987a]. Other investigators have shown that hematopoietic (CD34-selected) marrow cell populations attach to stromal cell layers and that the attached cells are enriched for granulocyte-macrophage progenitor cells [Liesveld et al., 1989]. Highly enriched murine spleen colony-forming cells (CFC-S) attach to stromal cell layers, proliferate, and differentiate into hematopoietic cells [Spooncer et al., 1985]. Interestingly, underlying bone marrow stromal cells can be substituted for by NIH 3T3 cells [Roberts et al., 1987], suggesting that these adherent cells supply the necessary attachment ligand for CFU-S attachment [Roberts et al., 1987; Yamazaki et al., 1989].

118.2 Soluble Growth Factors Soluble specific growth factors are an obligate requirement for the proliferation and differentiation of developing cells. These growth factors differ in effects from the endocrine hormones such as anabolic steroids or growth hormone. Whereas the endocrine hormones affect general cell function and are required and/or important to tissue formation, their predominant role is one of homeostasis. Growth factors, however, specifically drive the developmental programs of differentiating cells. Whether these function in a permissive or an inductive capacity has been the subject of considerable past controversy, particularly with respect to blood cell development. However, the large amount of data demonstrating linkages between receptor-ligand interaction and gene activation argues persuasively for an inductive/direct action on gene expression and, hence, cell proliferation and differentiation. Again, a large body of knowledge concerning growth factors comes from the field of hematopoiesis (blood cell development). Hematopoietic cell proliferation and differentiation is regulated by numerous growth factors; for reviews see Metcalf [1989] and Arai and colleagues [1990]. Within the last decade approximately 29 stimulatory cytokines (13 interleukins, M-CSF, erythroprotein, G-CFS, GM-CSF, c-kit ligand, gamma-interferon, and thrombopoietin) have been molecularly cloned and examined for their function in hematopoiesis. Clearly, this literature is beyond the scope of this review. However, the recent genetic cloning of eight receptors for these cytokines has led to the observation that a number of these receptors have amino acid homologies [Arai et al., 1990], showing that they are members of one or more gene families (Table 118.4). Hematopoietic growth factor receptors structurally contain a large extracellular domain, a transmembrane region, and a sequence-specific cytoplasmic domain [Arai et al., 1990]. The extracellular domains of interleukin-1, interleukin-6, and gamma-interferon are homologous with the immunoglobulin gene superfamily, and weak but significant amino-acid homologies exist among the interleukin-2 (beta chain), IL-6, IL-3, IL-4, erythropoietin, and GM-CSF receptors [Arai et al., 1990].


TABLE 118.4

Hematopoietic Growth Factor Receptor Families

Receptors with homology to the immunoglobulin gene family Interleukin-1 receptor Interleukin-6 receptor Gamma-interferon receptor Hematopoietic growth factor receptor family Interleukin-2 receptor (β-chain) Interleukin-3 receptor Interleukin-4 receptor Interleukin-6 receptor Erythropoietin receptor G/M-CSF receptor Source: Modified from Long [1992]. Reprinted with permission.

Like other developing tissues, bone responds to bone-specific and other soluble growth factors. TGF-β is a member of a family of polypeptide growth regulators which affects cell growth and differentiation during developmental processes such as embryogenesis and tissue repair [Sporn & Roberts, 1985]. TGF-β strongly inhibits proliferation of normal and tumor-derived epithelial cells and blocks adipogenesis, myogenesis, and hematopoiesis [Sporn & Roberts, 1985]. However, in bone, TGF-β is a positive regulator. TGF-β is localized in active centers of bone differentiation (cartilage canals and osteocytes) [Massague, 1987], and TGF-β is found in high quantity in bone, suggesting that bone contains the greatest total amount of TGF-β [Gehron Robey et al., 1987; Massague, 1987]. During bone formation, TGF-β promotes chrondrogenesis [Massague, 1987]—an effect presumably related to its ability to stimulate the deposition of extracellular matrix (ECM) components [Ignotz & Massague, 1986]. Besides stimulating cartilage formation, TGF-β is synthesized and secreted in bone cell cultures and stimulates the growth of subconfluent layers of fetal bovine bone cells, thus showing it to be an autocrine regulator of bone cell development [Sporn & Roberts, 1985]. In addition to TGF-β, other growth factors or cytokines are implicated in bone development. Urist and coworkers have been able to isolate various regulatory proteins which function in both in vivo and in vitro models [Urist et al., 1983b]. Bone morphogenic protein (BMP), originally an extract of demineralized human bone matrix, has now been cloned [Wozney et al., 1988] and, when implanted in vivo, results in a sequence of events leading to functional bone formation [Muthukumaran & Reddi, 1985; Wozney et al., 1988]. The implanting of BMP is followed by mesenchymal cell migration to the area of the implant, differentiation into bone progenitor cells, deposition of new bone, and subsequent bone remodeling to allow the establishment of bone marrow [Muthukumaran & Reddi, 1985]. A number of additional growth factors which regulate bone development exist. In particular, bone-derived growth factors (BDGF) stimulate bone cells to proliferate in serum-free media. [Hanamura et al., 1980; Linkhart et al., 1986]. However, these factors seem to function at a different level from BMP [Urist et al., 1983a].

118.3 Extracellular Matrix The extracellular matrix (ECM) varies in its tissue composition throughout the body and consists of various molecules such as laminin, collagens, proteoglycans, and other glycoproteins [Wicha et al., 1982]. Gospodarowicz and coworkers demonstrated that ECM components greatly affect corneal endothelial cell proliferation in vitro [Gospodarowicz et al., 1980; Gospodarowicz & Ill, 1980]. Studies by Reh and coworkers indicate that the ECM protein laminin is involved in inductive interactions which give rise to retinal-pigmented epithelium [Reh & Gretton, 1987]. Likewise, differentiation of mammary epithelial cells is profoundly influenced by ECM components; mammary cell growth in vivo and in vitro requires type IV collagen [Wicha et al., 1982]. A number of investigations elucidated a role for ECM and its components in hematopoietic cell function. These studies have identified the function of both previously known and newly identified ECM components in hematopoietic cell cytoadhesion (Table 118.5).


TABLE 118.5

Protiens and Glycoprotiens Mediating Hematopoietic Cell–Extracellular Matrix Interactions

Matrix Component

Cell Surface Receptor

Cellular Expression

References Patel, 1984, 1986, 1987, Patel et al., 1985; Ryan et al., 1990; Tsai et al., 1987; Van de Water et al., 1988 Giancotti et al., 1987 Silverstein and Nachman, 1987; Leung, 1984 Long and Dixit, 1990 Long and Dixit, 1990 Aruffo et al., 1990, Dorshkind, 1989; Horst et al., 1990; Miyake et al., 1990

Fibronectin

FnR

Erythroid; BFC-E; B cells; Lymphoid cells; HL60 cells

Thrombospondin

IIb/IIIa TSP-R

Hyaluronic acid

CD44

Platelets and megakaryocytes Monocytes and platelets Human CFC, CFC-GEMM T and B cells

Hemonectin

Unk

Neutrophils Tumor cells CFC-GM, BFC-E Immat. neutr. BFU-E

Campbell et al., 1985, 1987, 1990

Proteoglycans: Heparan sulfate Unfract ECM

Unk Unk

B1-CFC B1-CFC, bm Stroma

Gordon, 1988; Gordon et al., 1988 Gordon et al., 1988; Campbell et al., 1985

R = receptor; BFC-E and erythroid progenitor cell = the burst-forming cell-erythrocyte; HL60 = a promyelocytic leukemia cell line; CFC = colony-forming cell; GEMM = granulocyte erythrocyte macrophage megakarocyte; GM = granulocyte/macrophage; unk = unknown; B1 = blast cell; bm = bone marrow. Modified and reprinted from Long [1992] with permission.

As mentioned, soluble factors, stromal cells, and extracellular matrix (the natural substrate surrounding cells in vivo) are critical elements of the hematopoietic microenvironment. Work by Wolf and Trenton on the hematopoietic microenvironment in vivo provided the first evidence that (still unknown) components of the microenvironment are responsible for the granulocytic predominance of bone marrow hematopoiesis and the erythrocytic predominance of spleen [Wolf & Trentin, 1968]. Dexter and coworkers observed that the in vitro development of adherent cell populations is essential for the continued proliferation and differentiation of blood cells in long-term bone marrow cell cultures [Dexter & Lajtha, 1974; Dexter et al., 1976]. These stromal cells elaborate specific ECM components such as laminin, fibronectin, and various collagens and proteoglycans, and the presence of these ECM proteins coincided with the onset of hematopoietic cell proliferation [Zuckerman & Wicha, 1983]. The actual roles for extracellular matrix versus stromal cells in supporting cell development remains somewhat obscure, as it often is difficult to disassociate stromal cell effects from those of the ECM, since stromal cells are universally observed to be enmeshed in the surrounding extracellular matrix. Bone extracellular matrix contains both collagenous and noncollagenous proteins. A number of noncollagenous matrix proteins, isolated from demineralized bone, are involved in bone formation. Osteonectin is a 32 kDa protein which, binding to calcium, hydroxypatite, and collagen, is felt to initiate nucleation during the mineral phase of bone deposition [Termine et al., 1981]. In vivo analysis of osteonectin message reveals its presence in a variety of developing tissues [Holland et al., 1987; Nomura et al., 1988]. However, osteonectin is present in its highest levels in bones of the axial skeleton, skull, and the blood platelet (megakaryocyte) [Nomura et al., 1988]. Bone gla protein (BGP, osteocalcin) is a vitamin K-dependent, 5700 Da calcium-binding bone protein which is specific for bone and may regulate Ca2 + deposition [Price et al., 1976, 1981; Termine et al., 1981]. Other bone proteins seem to function as cytoadhesion molecules [Oldberg et al., 1986; Somerman et al., 1987] or have unresolved functions [Reddi, 1981]. Moreover, bone ECM also contains a number of the more common mesenchymal growth factors such as PDGF, basic, and acidic fibroblast growth factor [Canalis, 1985; Hauschka et al., 1986; Linkhart et al., 1986; Urist et al., 1983a]. These activities are capable of stimulating the proliferation of mesenchymal target cells (BALB/c 3T3 fibroblasts, capillary endothelial cells, and rat fetal osteoblasts). As well, bone-specific proliferating activities such as the BMP exist in bone ECM. Although these general and specific growth factors undoubtedly play a role in bone formation, little is understood concerning the direct inductive/permissive capacity of bone-ECM or bone proteins themselves on human bone cells


or their progenitors. Nor is the role of bone matrix in presenting growth factors understood—such “matricrine” (factor-ECM) interactions may be of fundamental importance in bone cell development. When bone precursor cells are cultured on certain noncollagenous proteins, they show an increase in proliferation and bone protein expression (MWL, unpublished observation). Moreover, we have shown, using the hematopoietic system as a model, that subpopulations of primitive progenitor cells require both a mitogenic cytokine and a specific extracellular matrix (ECM) molecule in order to proliferate [Long et al., 1992]. Indeed, without this obligate matrix-cytokine (“matricrine”) signal, the most primitive of blood precursor cells fail to develop in vitro [Long et al., 1992]. Although poorly understood, a similar requirement exists for human bone precursor cells, and complete evaluation of osteogenic development (or that of other tissues) thus requires additional studies of ECM molecules. For example, we have demonstrated the importance of three bone ECM proteins in human bone cell growth: osteonectin, osteocalcin, and type I collagen [Long et al., 1990; Long et al., 1994]. Additional bone proteins such as bone sialoprotien and osteopontin are no doubt important to bone structure and function, but their role is unknown [Nomura et al., 1988; Oldberg et al., 1986]. The above observations on the general and specific effects of ECM on cell development have identified certain matrix components which seem to appear as a recurrent theme in tissue development. These are proteoglycans, thrombospondin, fibronectin, and the collagens.

Proteoglycans and Glycosaminoglycans Studies on the role of proteoglycans in blood cell development indicate that both hematopoietic cells [Minguell & Tavassoli, 1989] (albeit as demonstrated by cell lines) and marrow stromal cells [Gallagher et al., 1983; Kirby & Bentley, 1987; Spooncer et al., 1983; Wight et al., 1986] produce various proteoglycans. Proteoglycans (PG) are polyanionic macromolecules located both on the stromal cell surface and within the extracellular matrix. They consist of a core protein containing a number of covalently linked glycosaminoglycan (GAG) side chains, as well as one or more O- or N-linked oligosaccharides. The GAGs consist of nonbranching chains of repeating N-acetylglucosamine or N-acetylglactosamine disaccharide units. With the exception of hyaluronic acid, all glycosaminoglycans are sulfated. Interesting, many extracellular matrix molecules (fibronectin, laminin, and collagen) contain glycosaminoglycan-binding sites, suggesting that complex interactions occur within the matrix itself. Proteoglycans play a role in both cell proliferation and differentiation. Murine stromal cells produce hyaluronic acid, heparan sulfate, and chondroitin sulfate [Gallagher et al., 1983], and in vitro studies show PG to be differentially between stromal cell surfaces and the media, with heparan sulfate being the primary cell-surface molecule and chondroitin sulfate the major molecular species in the aqueous phase [Spooncer et al., 1983]. In contrast to murine cultures, the human hematopoietic stromal cells in vitro contain small amounts of heparan sulfate and large amounts of dermatin and chondroitin sulfate, which seem to be equally distributed between the aqueous phase and extracellular matrix [Wight et al., 1986]. The stimulation of proteoglycan/GAG synthesis is associated with an increased hematopoietic cell proliferation, as demonstrated by an increase in the percentage of cells in S-phase [Spooncer et al., 1983]. Given the general diversity of proteoglycans, it is reasonable to expect that they may encode both lineagespecific and organ-specific information. For example, organ-specific PGs stimulate differentiation, as marrow-derived ECM directly stimulates differentiation of human progranulocytic cells (HL60), whereas matrix derived from skin fibroblasts lacks this inductive capacity [Luikart et al., 1987]. Moreover, organspecific effects are seen in studies of human blood precursor cell adhesion to marrow-derived heparan sulfate but not to heparan sulfates isolated from bovine kidney [Gordon et al., 1988]. Interestingly, cell-surface-associated PGs are involved in the compartmentalization or localization of growth factors within the microenvironment. Thus, the proliferation of hematopoietic cells in the presence of hematopoietic stroma is associated with a glycosaminoglycan-bound growth factor (GM-CSF) [Gordon et al., 1987b], and determination of the precise GAG molecules involved in this process (i.e., heparan sulfate) has showed that heparan sulfate side chains bind two blood cell growth factors: GMCSF and interleukin-3 [Roberts et al., 1988]. These data imply that ECM components and growth factors


combine to yield lineage-specific information and indicate that, when PG- or GAG-bound, growth factor is presented to the progenitor cells in a biologically active form.

Thrombospondin (TSP) Thrombospondin is a large, trimeric disulfide-linked glycoprotein (molecular weight 450,000, subunit molecular weight 180,000) having a domainlike structure [Frazier, 1987]. Its protease-resistant domains are involved in mediating various TSP functions such as cell binding and binding of other extracellular matrix proteins [Frazier, 1987]. Thrombospondin is synthesized and secreted into extracellular matrix by most cells; for review see Lawler [1986] and Frazier [1987]. Matrix-bound TSP is necessary for cell adhesion [Varian et al., 1986], cell growth [Majack et al., 1986], and carcinoma invasiveness [Riser et al., 1988] and is differentially expressed during murine embryogenesis [O’Shea & Dixit, 1988]. Work from our laboratories shows that thrombospondin functions within the hematopoietic microenvironment as a cytoadhesion protein for a subpopulation of human hematopoietic progenitor cells [Long et al., 1992; Long & Dixit, 1990]. Interestingly, immunocytochemical metabolic labeling studies show that hematopoietic cells (both normal marrow cells and leukemic cell lines) synthesize TSP, deposit it within the ECM, and are attached to it. The attachment of human progenitor cells to thrombospondin is not mediated by its integrin-binding RGD sequence because this region of the TSP molecule is cryptic, residing within the globular carboxy-terminus of the molecule. Thus, excess concentrations of a tetrapeptide containing the RGD sequence did not inhibit attachment of human progenitor cells [Long & Dixit, 1990], and similar observations exist in other cell systems [Asch et al., 1987; Roberts et al, 1987; Varian et al., 1988]. Other studies from this author’s laboratories show that bone marrow ECM also plays a major role in hematopoiesis in that complex ECM extracts greatly augment LTBMC cell proliferation [Campbell et al., 1985] and that marrow-derived ECM contains specific cytoadhesion molecules [Campbell et al., 1987, 1990; Long et al., 1990, 1992; Long & Dixit, 1990].

Fibronectin Fibronectin is a ubiquitous extracellular matrix molecule that is known to be involved in the attachment of paryenchymal cells to stromal cells [Bentley & Tralka, 1983; Zuckerman & Wicha, 1983]. As with TSP, hematopoietic cells synthesize, deposit, and bind to fibronectin [Zuckerman & Wicha, 1983]. Extensive work by Patel and coworkers shows that erythroid progenitor cells attach to fibronectin in a developmentally regulated manner [Patel et al., 1985; Patel & Lodish, 1984, 1986, 1987]. In addition to cells of the erythroid lineage, fibronectin is capable of binding lymphoid precursor cells and other cell phenotypes [Bernardi et al., 1987; Giancotti et al., 1986]. Structurally, cells adhere to two distinct regions of the fibronectin molecule, one of which contains the RGD sequence; the other is within the carboxy terminal region and contains a high-affinity binding site for heparan [Bernardi et al., 1987].

Collagen The role of various collagens in blood cell development remains uncertain. In vitro marrow cells produce types I, III, IV, and V collagen [Bentley, 1982; Bentley & Foidart, 1981; Castro-Malaspina et al., 1980; Zuckerman and Wicha, 1983], suggesting a role for these extracellular matrix components in the maintenance of hematopoiesis. Consistent with this, inhibition of collagen synthesis with 6-hydroxyprolene blocks or reduces hematopoiesis in vitro [Zuckerman et al., 1985]. Type I collagen is the major protein of bone, comprising approximately 90 percent of its protein.

118.4 Considerations for ex Vivo Tissue Generation The microenvironmental complexities discussed above suggest that the ex vivo generation of human (replacement) tissue (e.g., marrow, liver) will be a difficult process. However, many of the needed tissues (liver, marrow, bone, and kidney) have a degree of regenerative or hyperplastic capacity which allows © 2000 by CRC Press LLC

their in vitro cultivation. Thus, in vivo growth of many of these tissues types is routinely performed, albeit at varying degrees of success. The best example of this is bone marrow. If unfractionated human bone marrow is established in culture, the stromal and hematopoietic cells attempt to recapitulate in vivo hematopoiesis. Both soluble factors and ECM proteins are produced [Dexter & Spooncer, 1987; Long & Dixit, 1990; Zuckerman & Wicha, 1983], and relatively long-term hematopoiesis occurs. However, if long-term bone marrow cultures are examined closely, they turn out to not faithfully reproduce the in vivo microenvironment [Dexter & Spooncer, 1987; Spooncer et al., 1985; Schofield & Dexter, 1985]. Over a period of 8–12 weeks (for human cultures) or 3–6 months (for murine cultures), cell proliferation ceases, and the stromal/hematopoietic cells die. Moreover, the pluripotentiality of these cultures is rapidly lost. In vivo, bone marrow produces a wide variety of cells (erythroid, megakaryocyte/platelet, four types of myeloid cells, and B-lymphocytes). Human long-term marrow cultures produce granulocytes and megakaryocytes for 1–3 weeks, and erythropoiesis is only seen if exogenous growth factors are added. Thereafter, the cultures produce granulocytes and macrophages. These data show that current culture conditions are inadequate and further suggest that other factors such as rate of fluid exchange (perfusion) or that the three-dimensional structure of these cultures is limiting. Recent work by Emerson, Palsson, and colleagues demonstrated the effectiveness of altered medium exchange rates in the expansion of blood cells in vitro [Caldwell et al., 1991; Schwartz et al., 1991a, 1991b]. These studies showed that a daily 50% medium exchange affected stromal cell metabolism and stimulated a transient increase in growth factor production [Caldwell et al., 1991]. As well, these cultures underwent a 10-fold expansion of cell numbers [Schwartz et al., 1991b]. While impressive, thee cultures nonetheless decayed after 10–12 weeks. Recently, this group utilized continuous-perfusion bioreactors to achieve a longer ex vivo expansion and showed a 10–20-fold expansion of specific progenitor cell types [Koller et al., 1993]. These studies demonstrate that bioreactor technology allows significant expansion of cells, presumably via better mimicry of in vivo conditions. Another aspect of tissue formation is the physical structure of the developing organ. Tissues exist as three-dimensional structures. Thus, the usual growth in tissue culture flasks is far removed from the in vivo setting. Essentially, cells grown in vitro proliferate at a liquid/substratum interface. As a result, primary tissue cells grow until they reach confluence and then cease proliferating, a process known as contact inhibition. This, in turn, severely limits the degree of total cellularity of the system. For example, long-term marrow cultures (which do not undergo as precise a contact inhibition as do cells from solid tissues) reach a density of 1–2 × 106 per milliliter. This is three orders of magnitude less than the average bone marrow density in vivo. A number of technologies have been applied to this three-dimensional growth problem (e.g., hollow fibers). However, the growth of cells on or in a nonphysiologic matrix is less than optimal in terms of replacement tissue, since such implants trigger a type of immune reaction (foreign body reaction) or are thrombogenic. Recently, another type of bioreactor has been used to increase the ex vivo expansion of cells. Rotating wall vessels are designed to result in constant, low-shear suspension of tissue cells during their development. These bioreactors thus simulate a microgravity environment. The studies of Goodwin and associates document a remarkable augmentation of mesenchymal cell proliferation in low-shear bioreactors. Their data show that mesenchymal cell types show an average three- to six-fold increase in cell density in these bioreactors, reaching a cellularity of approximately 107 cells/mL [Goodwin et al., 1993a, 1993b]. Importantly, this increase in cell density was associated with a 75% reduction in glucose utilization as well as an approximate 85% reduction in the enzymatic markers of anabolic cellular metabolism (SGOT and LDH) [Goodwin et al., 1993a]. Importantly, further work by Goodwin and colleagues show that the growth of mesenchymal cells (kidney and chondrocyte) under low-shear conditions leads to the formation of tissuelike cell aggregates which is enhanced by growing these cells on collagen-coated microcarriers [Duke et al., 1993; Goodwin et al., 1993a]. The physical requirements for optimal bone precursor cell (i.e., osteoprogenitor cells and preosteoblasts) proliferation both in vivo and ex vivo are poorly understood. In vivo, bone formation most often occurs within an intervening cartilage model (i.e., a three-dimensional framework referred to as endochondral ossification). This well-understood bone histogenesis is one of embryonic and postnatal chondrogenesis,


which accounts for the shape of bone and the subsequent modification and calcification of bone cell ECM by osteoblasts. Recent work in this laboratory has examined the physical requirements for bone cell growth. When grown in suspension cultures (liquid/substratum interface), bone precursor cells develop distinct, clonal foci. These cells, however, express low amounts of bone-related proteins, and they rapidly expand as “sheets” of proliferating cells. As these are nontransformed (i.e., primary) cells, they grow until they reach confluence and then undergo contact inhibition and cease proliferating. We reasoned that growing these cells in a three-dimensional gel might augment their development. It is known from other systems that progenitor cell growth and development in many tissues requires the presence of at least one mitogenic growth factor, and that progenitor cell growth in a three-dimensional matrix results in the clonal formation of cell colonies by restricting the outgrowth of differentiated progeny [Metcalf, 1989]. Thus, bone precursor cells were overlayered with chemically defined, serumfree media containing a biopolymer, thus providing the cells with a three-dimensional scaffold in which to proliferate. In sharp contrast to bone precursor cell growth at a planar liquid/substratum interface, cells grown in a three-dimensional polymer gel show a marked increase in proliferative capacity and an increased per-cell production of the bone-specific proteins (MWL, unpublished observations).

118.5 Conclusions and Perspectives As elegantly demonstrated by the composite data above, the molecular basis and function of the various components of tissue microenvironments are becoming well understood. However, much remains to be learned regarding the role of these and other, as yet unidentified molecules in tissue development. One of the intriguing questions to be asked is how each interacting molecule contributes to defining the molecular basis of a given microenvironment—for example, the distinct differences in hematopoiesis as it exists in the marrow versus the spleen (i.e., a predominance granulopoiesis and erythropoiesis, respectively) [Wolf & Trentin, 1968]. Another rapidly advancing area is the dissection of the molecular basis of cell trafficking into tissues. Again, the immune system offers for the assessment of this process. Thus, interaction of the lymphocyte receptors with specific glycoconjugates of vascular addressins [Goldstein et al., 1989; Idzerda et al., 1989; Lasky et al., 1989] suggests that similar recognition systems are involved in other tissues, particularly the bone marrow. Another interesting observation is that hematopoietic progenitor cells synthesize and bind to their own cytoadhesion molecules, independent of the matrix molecules contributed by the stromal cells. For example, developing cells in vitro synthesize and attach to fibronectin, thrombospondin, and hemonectin, suggesting that these molecules function solely in an autochthonous manner to localize or perhaps stimulate development. Such a phenomenon may be a generalized process, as we have noted similar patterns of ECM expression/attachment in osteopoietic cell cultures. Coupled with data from the bioreactor studies, this suggest that under appropriate biology/physical conditions tissue cells may spontaneously reestablish their structure. Finally, the elucidation of the various requirements for optimal progenitor cell growth (cell interactions, specific growth factors and matrix components, and/or accessory cells which supply them) should allow the improvement of ex vivo culture systems to yield an environment in which tissue reconstitution is possible. Such a system would have obvious significance in organ-replacement therapy.

References Arai K, Lee F, Miyajima A, et al. 1990. Cytokines: Coordinators of immune and inflammatory responses. Annu Rev Biochem 59:783. Aruffo A, Staminkovic I, Melnik M, et al. 1990. CD44 is the principal cell surface receptor for hyaluronate. Cell 61:1303. Asch AS, Barnwell J, Silverstein RL, et al. 1987. Isolation of the thrombospondin membrane receptor. J Clin Invest 79:1054. Bentley SA. 1982. Collagen synthesis by bone marrow stromal cells: a quantitative study. Br J Haematol 50:491.


Bentley SA, Froidart JM. 1981. Some properties of marrow derived adherent cells in tissue culture. Blood 56:1006. Bentley SA, Tralka TS. 1983. Fibronectin-mediated attachment of hematopoietic cells to stromal elements in continuous bone marrow cultures. Exp Hematol 11:129. Bernardi P, Patel VP, Lodish HF. 1987. Lymphoid precursor cells adhere to two different sites on fibronectin. J Cell Biol 105:489. Bevilacqua MP, Stengelin S, Gimbrone MA, et al. 1989. Endothelial leukocyte adhesion molecule 1: An inducible receptor for neutrophils related to complement regulatory proteins and lectins. Science 243:1160. Bierer BE, Sleckman BP, Ratnofsky SE, et al. 1989. The biologic roles of CD2, CD4, and CD8 in T-cell activation. Annu Rev Immunol 7:579. Brandley BK, Sweidler SJ, Robbins PW. 1990. Charbohydrate ligands of the LEC cell adhesion molecules. Cell 63:861. Caldwell J, Palsson PB, Locey B, et al. 1991. Culture perfusion schedules influence the metabolic activity and granulocyte-macrophage colony-stimulating factor production rates of human bone marrow stromal cells. J Cell Physiol 147:344. Campbell A, Sullenberger B, Bahou W, et al. 1990. Hemonectin: A novel hematopoietic adhesion molecule. Prog Clin Biol Res 352:97. Campbell A, Wicha MS, Long MW. 1985. Extracellular matrix promotes the growth and differentiation of murine hematopoietic cells in vitro. J Clin Invest 75:2085. Campbell AD, Long MW, Wicha MS. 1987. Haemonectin, a bone marrow adhesion protein specific for cells of granuolocyte lineage. Nature 329:744. Campbell AD, Long MW, Wicha MS. 1990. Developmental regulation of granulocytic cell binding to hemonectin. Blood 76:1758. Canalis E. 1985. Effect of growth factors on bone cell replication and differentiation. Clin Orth Rel Res 193:246. Castro-Malaspina H, Gay RE, Resnick G, et al. 1980. Characterization of human bone marrow fibroblast colony-forming cells and their progeny. Blood 56:289. Coulombel L, Vuillet MH, Tchernia G. 1988. Lineage- and stage-specific adhesion of human hematopoietic progenitor cells to extracellular matrices from marrow fibroblasts. Blood 71:329. Dexter TM. 1982. Stromal cell associated haemopoiesis. J Cell Physiol 1:87. Dexter TM, Allen TD, Lajtha LG. 1976. Conditions controlling the proliferation of haemopoietic stem cells in vitro. J Cell Physiol 91:335. Dexter TM, Lajtha LG. 1974. Proliferation of haemopoietic stem cells in vitro. Br J Haematol 28:525. Dexter TM, Spooncer E. 1987. Growth and differentiation in the hemopoietic system. Annu Rev Cell Biol 3:423. Dorshkind K. 1989. Hemopoietic stem cells and B-lymphocyte differentiation. Immunol Today 10:399. Duke PJ, Danne EL, Montufar-Solis D. 1993. Studies of chondrogenesis in rotating systems. J Cell Biochem 51:274. Frazier WA. 1987. Thrombospondin: A modular adhesive glycoprotein of platelets and nucleated cells. J Cell Biol 105:625. Gallagher JT, Spooncer E, Dexter TM. 1983. Role of the cellular matrix in haemopoiesis: I. Synthesis of glycosaminoglycans by mouse bone marrow cell cultures. J Cell Sci 63:155. Gallatin M, St John TP, Siegleman M, et al. 1986. Lymphocyte homing receptors. Cell 44:673. Gehron Robey P, Young MF, Flanders KC, et al. 1987. Osteoblasts synthesize and respond to transforming growth factor-type beta (TGF-beta) in vitro. J Cell Biol 105:457. Giancotti FG, Comoglio PM, Tarone G. 1986. Fibronectin-plasma membrane interaction in the adhesion of hemopoietic cells. J Cell Biol 103:429. Giancotti FG, Languino LR, Zanetti A, et al. 1987. Platelets express a membrane protein complex immunologically related to the fibroblast fibronectin receptor and distinct from GPIIb/IIIa. Blood 69:1535.


Giancotti FG, Ruoslahti E. 1990. Elevated levels of the alpha 5 beta 1 fibronectin receptor suppress the transformed phenotype of Chinese hamster ovary cells. Cell 60:849. Goelz SE, Hession C. Goff D, et al. 1990. ELFT: A gene that directs the expression of an ELAM-1 ligand. Cell 63:1349. Goldstein LA, Zhou DF, Picker LJ, et al. 1989. A human lymphocyte homing receptor, the hermes antigen, is related to cartilage proteoglycan core and link proteins. Cell 56:1063. Goodwin TJ, Prewett TI, Wolf DA, et al. 1993a. Reduced shear stress: A major component in the ability of mammalian tissues to form three-dimensional assemblies in simulated microgravity. J Cell Biochem 51:301. Goodwin TJ, Schroeder WF, Wolf DA, et al. 1993b. Rotating vessel coculture of small intestine as a prelude to tissue modeling: Aspects of simulated microgravity. Proc Soc Exp Biol Med 202:181. Gordon MY. 1988. The origin of stromal cells in patients treated by bone marrow transplantation. Bone Marrow Transplant 3:247. Gordon MY, Bearpark AD, Clarke D, et al. 1990a. Haemopoietic stem cell subpopulations in mouse and man: discrimination by differential adherence and marrow repopulation ability. Bone Marrow Transplant 5:6. Gordon MY, Clarke D, Atkinson J, et al. 1990b. Hemopoietic progenitor cell binding to the stromal microenvironment in vitro. Exp Hematol 18:837. Gordon MY, Dowding CR, Riley GP, et al. 1987a. Characterization of stroma-dependent blast colonyforming cells in human marrow. J Cell Physiol 130:150. Gordon MY, Hibbin JA, Dowding C, et al. 1985. Separation of human blast progenitors from granulocytic, erythroid, megakaryocytic, and mixed colony-forming cells by “panning” on cultured marrowderived stromal layers. Exp Hematol 13:937. Gordon MY, Riley GP, Clarke D. 1988. Heparan sulfate is necessary for adhesive interactions between human early hemopoietic progenitor cells and the extracellular matrix of the marrow microenvironment. Leukemia 2:804. Gordon MY, Riley GP, Watt SM, et al. 1987b. Compartmentalization of a haematopoietic growth factor (GM-CSF) by glycosaminoglycans in the bone marrow microenvironment. Nature 326:403. Gospodarowicz D, Delgado D, Vlodasvsky I. 1980. Permissive effect of the extracellular matrix on cell proliferation in vitro. Proc Natl Acad Sci USA 77:4094. Gospodarowicz D, Ill C. 1980. Extracellular matrix and control of proliferation of vascular endothelial cells. J Clin Invest 65:1351. Hanamura H, Higuchi Y, Nakagawa M, et al. 1980. Solubilization and purification of bone morphogenetic protein (BMP) from dunn osteosarcoma. Clin Orth Rel Res 153:232. Hauschka PV, Marvrakos AE, Iafarati MD, et al. 1986. Growth factors in bone matrix: Isolation of multiple types by affinity chromatography on heparinsepharose. J Biol Chem 261:12665. Holland PWH, Harmper SJ, McVey JH, et al, 1987. In vivo expression of mRNA for the Ca++-binding protein SPARC (osteonectin) revealed by in situ hybridization. J Cell Biol 105:473. Horst E, Meijer CJML, Radaskiewicz T, Ossekoppele GP, VanKrieken JHJM, and Pals ST. 1990. Adhesion molecules in the prognosis of diffuse large-cell lymphoma: expression of a lymphocyte homing receptor (CD44), LFA-1 (CD11a/18), and ICAM-1 (CD54). Leukemia 4, 595-599. Idzerda RL, Carter WG, Nottenburg C, et al. 1989. Isolation and DNA sequence of a cDNA clone encoding a lymphocyte adhesion receptor for high endothelium. Proc Natl Acad Sci USA 86:4659. Ignotz RA, Massague J. 1986. Transforming growth factor-beta stimulates the expression of fibronectin and collagen and their incorporation into the extracellular matrix. J Biol Chem 261:4337. Kirby SL, Bentley SA. 1987. Proteoglycan synthesis in two murine bone marrow stromal cell lines. Blood 70:1777. Koller MR, Emerson SG, Palsson BO. 1993. Large-scale expansion of human stem and progenitor cells from bone marrow mononuclear cells in continuous perfusion cultures. Blood 82:378. Lasky LA, Singer MS, Yednock TA, et al. 1989. Cloning of a lymphocyte homing receptor reveals a lectin domain. Cell 56:1045.


Lawler J. 1986. The structural and functional properties of thrombospondin. Blood 67, 1197-1209. Lawrence MB, Springer TA. 1991. Leukocytes roll on a selectin at physiologic shear flow rates: Distinction from and prerequisite for adhesion through integrins. Cell 65:859. Leung LLK. 1984. Role of thrombospondin in platelet aggregation. J Clin Invest 74:1764. Liesveld JL, Abbound CN, Duerst RE, et al. 1989. Characterization of human marrow stromal cells: Role in progenitor cell binding and granulopoiesis. Blood 73:1794. Linkhart TA, Jennings JC, Mohan S, et al. 1986. Characterization of mitogenic activities extracted from bovine bone matrix. Bone 7:479. Long, MW. 1992. Blood cell cytoadhesion molecules. Exp Hematol 20:288. Long MW, Ashcraft A, Mann KG. 1994. Regulation of human bone marrow-derived osteoprogenitor cells by osteogenic growth factors. Submitted. Long MW, Briddell R, Walter AW, et al. 1992. Human hematopoietic stem cell adherence to cytokines and matrix molecules. J Clin Invest 90:251. Long MW, Dixit VM. 1990. Thrombospondin functions as a cytoadhesion molecule for human hematopoietic progenitor cells. Blood 75:2311. Long MW, Mann KG. 1993. Bone marrow as a source of osteoprogenitor cells. In MW Long, MS Wicha (eds), The Hematopoietic Microenvironment, pp 110–123, Baltimore, Johns Hopkins University. Long MW, Williams JL, Mann KG. 1990. Expression of bone-related proteins in the human hematopoietic microenvironment. J Clin Invest 86:1387. Lowe JB, Stoolman LM, Nair RP, et al. 1990. ELAM-1-dependent cell adhesion to vascular endothelium determined by a transfected human fucosyl transferase cDNA. Cell 63:475. Luikart SD, Sackrison JL, Maniglia CA. 1987. Bone marrow matrix modulation of HL-60 phenotype. Blood 70:1119. Majack RA, Cook SC, Bornstein P. 1986. Control of smooth muscle cell growth by components of the extracellular matrix: Autocrine role for the thrombospondin. Proc Natl Acad Sci USA 83:9050. Massague J. 1987. The TGF-beta family of growth and differentiation factors. Cell 49:437. Metcalf D. 1989. The molecular control of cell division, differentiation commitment and maturation in haematopoietic cells. Nature 339:27. Minguell JJ, Tavassoli M. 1989. Proteoglycan synthesis by hematopoietic progenitor cells. Blood 73:1821. Miyake K, Medina KL, Hayashi S, et al. 1990. Monoclonal antibodies to Pgp-1/CD44 block lymphhemopoiesis in long-term bone marrow cultures. J Exp Med 171:477. Miyake K, Weissman IL, Greenberger JS, et al. 1991. Evidence for a role of the integrin VLA-4 in lymphohematopoiesis. J Exp Med 173:599. Miyamake K, Medina K, Ishihara K, et al. 1991. A VCAM-like adhesion molecule on murine bone marrow stromal cells mediates binding of lymphocyte precursors in culture. J Cell Biol 114:557. Muthukumaran N, Reddi AH. 1985. Bone matrix-induced local bone induction. Clin Orth Rel Res 200:159. Nomura S, Wills AJ, Edwards DR, et al. 1988. Developmental expression of 2ar (osteopontin) and SPARC (osteonectin) RNA as revealed by in situ hybridization. J Cell Biol 106:441. O’Shea KS, Dixit VM. 1988. Unique distribution of the extracellular matrix component thrombospondin in the developing mouse embryo. J Cell Biol 107:2737. Oldberg A, Franzen A, Heinegard D. 1986. Cloning and sequence analysis of rat bone sialoprotein (osteopontin) cDNA reveals an Arg-Gly-Asp cell-binding sequence. Proc Natl Acad Sci USA 83:8819. Palacios R, Stuber S, Rolink A. 1989. The epigenetic influences of bone marrow and fetal liver stroma cells on the developmental potential of pre-B lymphocyte clones. Eur J Immunol 19:347. Patel VP, Ciechanover A, Platt O, et al. 1985. Mammalian reticulocytes lose adhesion to fibronectin during maturation to erythrocytes. Proc Natl Acad Sci USA 82:440. Patel VP, Lodish HF. 1984. Loss of adhesion of murine erythroleukemia cells to fibronectin during erythroid differentiation. Science 224:996. Patel VP, Lodish HF. 1986. The fibronectin receptor on mammalian erythroid precursor cells: Characterization and developmental regulation. J Cell Biol 102:449.


Patel VP, Lodish HF, 1987. A fibronectin matrix is required for differentiation of murine erythroleukemia cells into reticulocytes. J Cell Biol 105:3105. Ploemacher RE, Brons NHC. 1988. Isolation of hemopoietic stem cell subsets from murine bone marrow: II. Evidence for an early precursor of day-12 CFU-S and cells associated with radioprotective ability. Exp Hematol 16:27. Price PA, Ostuka AS, Poser JW, et al. 1976. Characterization of a gamma-carboxyglutamic acid-containing protein from bone, Proc Natl Acad Sci USA 73:1447. Price PA, Lothringer JW, Baukol SA, et al. 1981. Developmental appearance of the vitamin K-dependent protein of one during calcification. Analysis of mineralizing tissues in human, calf, and rat. J Biol Chem 256:3781. Reddi AH. 1981. Cell biology and biochemistry of endochondral bone development. Coll Res 1:209. Reh TA, Gretton H. 1987. Retinal pigmented epithelial cells induced to transdifferentiate to neurons by laminin. Nature 330:68. Riser BL, Varani J. Carey TE, et al. 1988. Thrombospondin binding and thrombospondin synthesis by human squamous carcinoma and melanoma cells: relationship to biological activity. Exp Cell Res 174:319. Roberts DD, Sherwood JA, Ginsburg V. 1987a. Platelet thrombospondin mediates attachment and spreading of human melanoma cells. J Cell Biol 104:131. Roberts RA, Spooncer E, Parkinson EK, et al. 1987b. Metabolically inactive 3T3 cells can substitute for marrow stromal cells to promote the proliferation and development of multipotent haemopoietic stem cells. J Cell Physiol 132:203. Roberts R, Gallagher J, Spooncer E, et al. 1988. Heparan sulphate bound growth factors: A mechanism for stromal cell mediated haemopoiesis. Nature 332:376. Ruoslahti E, Pierschbacher MD. 1987. New perspectives in cell adhesion: RGD and integrins. Science 238:491. Ryan DH, Nuccie Bl, Abbound CN, et al. 1990. Maturation-dependent adhesion of human B cell precursors to the bone marrow microenvironment. J Immunol 145:477. Schofield R, Dexter TM. 1985. Studies on the self-renewal ability of CFU-S which have been serially transferred in long-term culture or in vivo. Leuk Res 9:305. Schwartz RM, Emerson SG, Clarke MF, et al. 1991a. In vitro myelopoiesis stimulated by rapid medium exchange and supplementation with hematopoietic growth factors. Blood 78:3155. Schwartz RM, Palsson BO, Emerson SG. 1991b. Rapid medium perfusion rate significantly increases the productivity and longevity of human bone marrow cultures. Proc Natl Acad Sci USA 88:6760. Shaw S, Luce GE, Quinones R, et al. 1986. Two antigen-independent adhesion pathways used by human cytotoxic T-cell clones. Nature 323:262. Sheetz MP, Turney S, Qian H, et al. 1989. Nanometre-level analysis demonstrates that lipid flow does not drive membrane glycoprotein movements. Nature 340:248. Silverstein RL, Nachman RL. 1987. Thrombospondin binds to monocytes-macrophages and mediates platelet-monocyte adhesion. J Clin Invest 79:867. Somerman MJ, Prince CW, Sauk JJ, et al. 1987. Mechanism of fibroblast attachment to bone extracellular matrix: Role of a 44kilodalton bone phosphoprotein. J Bone Miner Res 2:259. Sits H, van Schooten W, Keizer H, et al. 1986. Alloantigen recognition is preceded by nonspecific adhesion of cytotoxic T cells and target cells. Science 232:403. Spooncer E, Gallagher JT, Krizsa F, et al. 1983. Regulation of haemopoiesis in long-term bone marrow cultures: IV. Glycosaminoglycan synthesis and the stimulation of haemopoiesis by beta-D-xylosides. J Cell Biol 96:510. Spooncer E, Lord BI, Dexter TM. 1985. Defective ability to self-renew in vitro of highly purified primitive haematopoietic cells. Nature 316:62. Sporn MB, Roberts AB. 1985. Autocrine growth factors and cancer. Nature 313:745. Springer TA. 1990. Adhesion receptors of the immune system. Nature 346:425. Stoolman LM. 1989. Adhesion molecules controlling lymphocyte migration. Cell 56:907.


Termine JD, Kleinman HK, Whison SW, et al. 1981. Osteonectin, a bone-specific protein linking mineral to collagen. Cell 26:99. Tsai S, Sieff CA, Nathan DG. 1986. Stromal cell-associated erythropoiesis. Blood 67:1418. Tsai S, Patel V, Beaumont E, et al. 1987. Differential binding of erythroid and myeloid progenitors to fibroblasts and fibronectin. Blood 69:1587. Urist MR, DeLange RJ, Finerman GAM. 1983a. Bone cell differentiation and growth factors. Science 220:680. Urist MR, Sato K, Brownell AG, et al. 1983b. Human bone morphogenic protein (hBMP). Proc Soc Exp Biol Med 173:194. Van de Water L, Aronson D, Braman V. 1988. Alteration of fibronectin receptors (integrins) in phorbol ester-treated human promonocytic leukemia cells. Cancer Res 48:5730. Varani J, Dixit VM, Fligiel SEG, et al. 1986. Thrombospondin-induced attachment and spreading of human squamous carcinoma cells. Exp Cell Res 156:1. Varani J, Nickloff BJ, Risner BL, et al. 1988. Thrombospondin-induced adhesion of human kertinocytes. J Clin Invest 81:1537. Wicha MS, Lowrie G, Kohn E, et al. 1982. Extracellular matrix promotes mammary epithelial growth and differentiation in vitro. Proc Natl Acad Sci USA 79:3213. Wight TN, Kinsella MG, Keating A, et al. 1986. Proteoglycans in human long-term bone marrow cultures: Biochemical and ultrastructural analyses. Blood 67:1333. Williams AF, Barclay AN. 1988. The immunoglobulin superfamily—domains for cell surface recognition. Annu Rev. Immunol 6:381. Witte PL, Robbinson M, Henley A, et al. 1987. Relationships between B-lineage lymphocytes and stromal cells in long-term bone marrow cultures. Eur J Immunol 17:1473. Wolf NS, Trentin JJ. 1968. Hemopoietic colony studies: V. Effect of hemopoietic organ stroma on differentiation of pluripotent stem cells. J Exp Med 127:205. Wozney JM, Rosen V, Celeste AJ, et al. 1988. Novel regulators of bone formation: Molecular clones and activities. Science 242:1528. Yamazaki K, Roberts RA, Spooncer E, et al. 1989. Cellular interactions between 3T3 cells and interleukin3-dependent multipotent haemopoietic cells.: A model system for stromal-cell-mediated haemopoiesis. J Cell Physiol 139:301. Zuckerman KS, Rhodes RK, Goodrum DD, et al. 1985. Inhibition of collagen deposition in the extracellular matrix prevents the establishment of a stroma supportive of hematopoiesis in long-term murine bone marrow cultures. J Clin Invest 75:970. Zuckerman KS, Wicha MS. 1983. Extracellular matrix production by the adherent cells of long-term murine bone marrow cultures. Blood 61:540.


Naughton, B. A. “The Importance of Stromal Cells.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

119 The Importance of Stromal Cells 119.1

Tissue Composition and Stromal Cells Fibroblasts • Endothelial Cells • Adipocytes, Fat-Storing Cells • Macrophages

119.2

Stromal Cells as “Feeder Layers” for Parenchymal Cell Culture

119.3 119.4

Support of Cultured Cells Using Cell Lines Stereotypic (Three-Dimensional) Culture Versus Monolayer Culture

Bone Marrow • Liver • Tumor Cells

Brian A. Naughton Advanced Tissue Sciences, Inc.

All tissue is composed of parenchymal (from Greek, that poured in beside) and stromal (Greek, framework or foundation) cells. Parenchyma are the functional cells of a tissue (e.g., for liver, hepatic parenchymal cells or hepatocytes; for bone marrow, hematopoietic cells), where stroma comprises primarily connective tissue elements which, together with their products, form the structural framework of tissue. Parenchymal cells can be derivatives of any of the three germ layers, and during development they usually grow into areas populated by stromal cells or their progenitors. Under the strictest definition, stromal cells are derivatives of mesenchyme and include fibroblasts, osteogenic cells, myofibroblasts, and fat cells which appear to arise from a common stem/progenitor cell [Friedenstein et al., 1970; Owen, 1988] (Fig. 119.1). Some investigators apply the term stromal cell to all the nonparenchymal cells that contribute to the microenvironment of a tissue and include endothelial cells and macrophages (histiocytes) in this classification as well [Strobel et al., 1986]. However, the ontogency of both endothelial cells and macrophages is distinct from that of mesenchymal tissue-derived cells [Wilson, 1983]. In this chapter, the more expansive definition of stroma will be used. A partial listing of tissue cells that may influence the function of organ parenchyma is in Table 119.1. For the sake of brevity, migrating cells of bone marrow origin will not be discussed in the text (e.g., mast cells, B lymphocytes, natural killer cells), although these cells can influence parenchyma either directly or via cytokine-mediated modulation of stromal cell function. Stromal and parenchymal cell components are integrated to form a multifunctional tissue in vivo. This chapter will focus on the contribution of stromal cells to the microenvironment and their use in culture to support parenchymal function.

119.1

Tissue Composition and Stromal Cells

Some similarities in the spatial organization of cells of different tissues are apparent. Epithelial cells are a protective and regulatory barrier not just for skin but for all surfaces exposed to blood or to the external environment (e.g., respiratory tract, tubular digestive tract). These cells rest atop a selectively permeable


FIGURE 119.1 Hypothetical relationship between the ontogeny of stromal and parenchymal (hematopoietic) cells of the bone marrow. Note: that stromal cell is defined in this chart as only those cells that are derivatives of mesenchyme. Hematopoietic stem and progenitor cells express CD 34. This epitope also is expressed by stromal cell precursors (CFU-F) [Simmons & Torok-Storb, 1991] and adherent stromal cells develop from cell populations selected for the CD 34 antigen. In addition, fetal marrow elements with CD 34 +, CD 38–, HLA-DR-phenotypes were reported to develop not only the hematopoietic microenvironment but also the hematopoietic cells themselves [Huang & Terstappen, 1992]. * Indicates the possibility that, at least in the fetus, there may be a common stem cell for bone marrow stromal and hematopoietic elements. † There are several schools of thought relating to single or multiple stem cell pools for the lymphoid versus the other lineages. TABLE 119.1

Cells That Contribute to the Tissue Microenvironment

Stromal cells: derivatives of a common precursor cell Mesenchyme Fibroblasts Myofibroblasts Osteogenic/chondrogenic cells Adipocytes Stromal-associated cells: histogenically distinct from stromal cells, permanent residents of a tissue Endothelial cells Macrophages Transient cells: cells that migrate into a tissue for host defense either prior to or following an inflammatory stimulus B lymphocytes/plasma cells Cytotoxic T cells and natural killer (NK) cells Granulocytes Parenchymal cells: cells that occupy most of the tissue volume, express functions that are definitive for the tissue, and interact with all other cell types to facilitate the expression of differentiated function

basement membrane. The composition of the underlying tissue varies, but it contains a connective tissue framework for parenchymal cells. If the tissue is an artery or a vein, the underlying connective tissue is composed of circularly arranged smooth-muscle cells which in turn are surrounded by an adventitia of loose connective tissue. Tissues within organs are generally organized into functional units around


capillaries which facilitate metabolite and blood gas transport by virtue of their lack of a smooth muscle layer (tunica media) and the thinness of their adventitial covering. Connective tissue cells of the tissue underlying capillaries deposit extracellular matrix (ECM) which is a mixture of fibrous proteins (collagens) embedded in a hydrated gel of glycosaminoglycans (GAGs). The GAGs as distinguished by their sugar residues, are divided into five groups: hyaluronic acid, the chondiotin sulfates, dermatan sulfate, heparan and heparin sulfate, and keratan sulfate. All the GAGs except hyaluronic acid link with proteins to form proteoglycans. These proteoglycans bind to long-chained hyaluronic acid cores that interweave the crosslinked collagen fibers that form the framework of the tissue. The basic composition and density of deposition of stromal cell-derived ECM varies from tissue to tissue [Lin and Bissel, 1993]. Figure 119.2 is a scanning electron micrograph (SEM) depicting the intricacies of the ECM deposited by human bone marrow-derived stromal cells growing on a three-dimensional culture template. The ECM forms a sievelike arrangement for diffusion in an aqueous environment and a number of bone marrow-derived migratory cells are present as well. Such an arrangement dramatically enhances the surface area for cell growth. Although these voids are primarily filled by parenchymal cells, normal tissue interstitium also contains migratory immunocompetent cells such as B lymphocyte/plasma cells, T cells, and natural killer cells as well as the ubiquitous macrophage (histiocyte). The numbers of these cells are enhanced during inflammatory episodes when neutrophils or other granulocytes and/or mononuclear leukocytes infiltrate the tissue. Cytokines and other proteins released by these leukocytes recruit stromal cells in the tissue repair process. This phenomenon can be exploited by tumor cells to enhance their invasiveness (see Section 119.2 Tumor Cells). Although tissue function in toto is measured by parenchymal cell output (e.g., for liver, protein synthesis, metabolism; for bone marrow, blood cell production), this is profoundly influenced by and in some instances orchestrated by the stromal cell microenvironment. Stromal cells contribute to parenchymal cell function by synthesizing the unique mix of ECM proteins necessary for cell seeding/attachment of specific types of tissue cells, the modulation of gene expression in parenchymal cells by events triggered by cytoskeleton-mediated transduction [Ben-Ze’ev et al., 1988], the deposition of ECM proteins to sequester growth and regulatory factors for use by developing parenchyma [Roberts et al., 1988], deposition of ECM of the right density to permit diffusion of nutrients, metabolites, and oxygen (and egress of CO2 and waste) to the extent necessary to maintain the functional state of the tissue, by forming the appropriate barriers to minimize cell migration or intrusion, and by synthesizing and/or presenting cytokins that regulate parenchymal cell function, either constituitively or following induction by other humoral agents (see Table 119.2). Creating cultures that contain the multiplicity of cell types found in vivo presents a daunting task for several reasons: (1) culture media that are rich in nutrients select for the most actively mitotic cells at the expense of more mitotically quiescent cells and, (2) cell phenotypic expression and function is related to its location in tissue to some extent. It is difficult, especially using a flat (i.e., two-dimensional) culture template, to create a microenvironment that is permissive or inductive for the formation of tissue-like structures. (3) Localized microenvironmental niches regulate different parenchymal cell functions or, in the case of bone marrow, hematopoietic cell differentiation. These milieu are difficult to reproduce in culture, since all cells are exposed to essentially the same media components. However, a major goal of tissue culture is to permit normal cell-cell associations and the reestablishment of tissue polarity so that parenchymal cell function is optimized. The question of three-dimensionality will be addressed later in this chapter. A brief survey of the various types of stromal elements follows.

Fibroblasts Fibroblasts are responsible for the synthesis of many GAGs and for the deposition and organization of collagens. Although present in most tissues, fibroblasts exhibit specialization with respect to the type of ECM that they secrete. For example, liver tissue contains type I and type IV collagen, whereas type II collagen is found in cartilaginous tissues. Bone marrow contains types I, III, IV, and V collagen [Zuckerman, 1984]. Heterogeneity of collagen deposition as well as GAG composition exists not only between different tissues but in developmentally different stages of the same tissue [Thonar & Kuettner, 1987].


(a)

(b)

FIGURE 119.2 Scanning electron micrographs of bone marrow cultures on nylon screen templates (a) Photograph depicting a portion of a macrophage (M) associated with the numerous, delicate, interweaving strands of ECM (arrows) that are deposited between the openings of the nylon screen of a bone marrow stromal cell culture. (b) Myeloid cells of a hematopoietic colony growing in a coculture of human stromal cells and hematopoietic cells. Note the pattern of attachment of individual cells to matrix and the large open area between the cells for nutrient access. (c) The photograph depicts the intimate association of cells of a myeloid or mixed myeloid/monocytic colony (arrow) with enveloping fibroblastic cells (F) in a bone marrow coculture. A filament of the nylon (n) screen is also present in the field. (d) An erythroid colony (E) in a human bone marrow coculture on nylon screen. Note that the more mature erythroid cells are on the periphery of the colony and that it is in close apposition to a macrophage (M). A nylon filament is also present (n).


(c)

(d)


In addition to matrix deposition, fibroblasts synthesize the cytokines of the fibroblast growth factor (FGF) family, a variety of interleukins, and GM-CSF as well as other regulatory cytokines (Table 119.2). Fibroblast activity can be modulated by circulating or locally diffusable factors including IL-1, TGFβ , TNFα , and a host of other factors [Yang et al., 1988]. For example, IL-1 activated splenic stroma supported the growth of rat natural killer cells in long-term culture [Tjota et al., 1992]. In addition, treatment of bone


TABLE 119.2

Stromal Cell Phenotypes and Secretory Profiles

a

Phenotypes

MHC I (all) MHC II (E, mϕ) CD 10 (Neural endopeptidase)b CD 29 (β2 integrin) (all) CD 34 (CFU-F) CD 36 (IA 7 (mϕ) CD 44 (H-CAM)b CD 49b (α2 chain of VLA-2)b CD 49d (α4 chain of VLA-4)b CD 49e (α5 chain of VLA-5)b CD 51 (vitronectin receptor)b CD 54 (ICAM-1) (all) CD 58 (LFA-3) (all) CD 61 (GP 111a) VCAM-1 (E) α SM actin (F) Vimentin (F) Decorin (Mes) Fibronectin (F, mϕ) Laminin (E) Collagen I (F) Collagen III (F) Collagen IV (F) von Willebrand factor (E) Adipsin (A)

Cytokine/Protein Secretion Il-1β (mϕ, E, F, A) M-CSF (mϕ, E, F) G-CSF (mϕ, F) GM-CSF (mϕ, E) LIF (mϕ, F) GM-colony enhancing factor (mϕ) TGFβ (mϕ, F) TNFα (mϕ, F) PAF (E) BPA (mϕ, E) IFNα (mϕ, F, E) IFNγ (F) Il-6 (mϕ) c-kit ligand (mϕ, F) Acidic and basic FGF (F) Angiotensinogen (A) lipoprotein lipase (A) Adipocyte P2 (A) MIP-1α (mϕ) Complement proteins C1-C5 (mϕ) Factor B, properdin (mϕ) Transcobalamin II (mϕ) Transferrin (mϕ) Arachidonic acid metabolites (mϕ, E) HGF/SF (F) flt3-ligand (S) Prolactin (F) LIF (S)c STR-3 (F) (IL-3, IL-7, IL-9, IL-11, IL-15, neuroleukin)b

a Phenotypic expression of stroma in bone marrow cultures. Parentheses indicate localization according to cell type (mϕ = Macrophage F = Fibroblast, E = endothelial, A = adipocyte, Mes = mesenchyme). Many of these phenotypes were identified in the original work of Cicuttini et al., [1992] and Moreau et al., [1993]. b Indicates the presence of mRNA and /or protein expression in bone marrow cultures but not localized to a particular cell type. c Found in fallopian tube stroma [Keltz et al., 1997]. Abbreviations: BPA—erythroid burst promoting activity, CD—cluster determinant, MHC—major histocompatibility complex, M—monocyte, G—granulocyte, CSF—colony stimulating factor, CFU-F—colony forming unit-fibroblast, MIP—macrophage inflammatory protein, IL—interleukin, IFN—interferon, TGF—transforming growth factor, TNF—tumor necrosis factor, PAF—platelet activating factor, FGF—fibroblast growth factor/scatter factor, LIF-leukemia inhibitory factor, STR-3—stromelysin 3.

marrow stromal cell cultures with the steroid methylprednisolone reduced the concentration of hyaluronic acid and heparan sulfate relative to other proteins [Siczkowski et al., 1993], perhaps making them more conducive to hematopoietic support. Horse serum, presumably because of its high content of hydrocortisone and other steroids, was an essential component of medium used to maintain hematopoiesis in early long term bone marrow cultures [Greenberger et al., 1979]. In our experience, medium supplementation with hydrocortisone or other corticosteroids enhances the ability of liver-derived stromal cells to support parenchymal hepatocyte function for longer-terms in vitro [Naughton et al., 1994], although the precise mechanism(s) has not been defined. Fibroblastic cells in tissue appear to be heterogeneous with respect to support function. Early morphometric studies of bone marrow indicated that stroma supporting different types of hematopoiesis exhibited different staining patterns. The development of highly specific monoclonal antibodies in the © 2000 by CRC Press LLC

intervening years made possible a much more detailed analysis of stromal cells of bone marrow and other tissues and provided a number of avenues to isolate these cells for study. Whereas bone marrow stromal cells can be separated from hematologic cells by virtue of their adherence to plastic, this technique was not optimal for all tissues. Monoclonal antibodies can now be used to “dissect” cells from the stromal microenvironment using a variety of techniques including flow cytometry, immune panning, immunomagnetic microspheres, and affinity column methodologies. Monoclonal antibodies also make it possible to select for stromal cell progenitors like the fibroblast-colony forming unit (CFU-F) which is CD 34 positive.

Endothelial Cells For nonglandular tissues such as bone marrow, the endothelial cells are vascular lining cells and can be distinguished by the expression of or the message for von Willebrand factor and surface major histocompatibility complex-II (MHC-II) antigens. As cells of the vascular tunica intima, they, along with the basement membrane, form a selective barrier to regulate the transport of substances to and from the blood supply. In addition, their expression of integrins following stimulation with IL-1 or other mediators regulates the attachment of immunocompetent cells during acute (neutrophils) or chronic (T lymphocytes, monocytes) inflammation. Endothelial cells also regulate other vascular functions. They synthesize the vasodilatory effector nitric oxide after induction with acetylcholine [Furchgott & Zawadzki, 1980] and secrete regulatory peptides, the endothelins, which counteract this effect [Yanagisawa et al., 1988]. Specialized vascular endothelia are present in bone marrow, where they permit the egress of mature leukocytes from the marrow into the sinusoids, and in the liver, where the surfaces of the cells lining sinusoids are fenestrated to facilitate transport. In a more general context, endothelial cells also produce collagen IV for basement membranes and secrete several cytokines including IL-1, fibronectin, M-CSF, and GM-CSF and release platelet-activating factor (Table 119.2). Vascular endothelia are nonthrombogenic and contribute to angiogenesis by mitotically responding to locally secreted factors such as bFGF and aFGF. These endothelia possess LDL receptors and degrade this lipoprotein at substantially higher rates than other types of cells. These cells also have receptors for Fc, transferrin, mannose, galactose, and Apo-E, as well as for scavenger receptors [Van Eyken & Desmet, 1993]. Endothelia tend to assemble into tubular structures in culture, but their contribution to parenchymal cell growth in vitro is contingent on the generation of the proper tissue polarity. In addition to vascular endothelial cells, some organs possess nonvascular endothelial cells such as the bile duct lining cells of the liver. These cells possess antigenic profiles and secretory potentials that are similar to vascular endothelia.

Adipocytes, Fat-Storing Cells These cells are represented in varying concentrations in different tissues. They are related to fibroblasts (Fig. 119.2), and transformation of fibroblasts to adipocytes in vitro can be induced by supplementation of the medium with hydrocortisone or other steroids [Brockband & van Peer, 1983; Greenberger, 1979] which induce the expression of lipoproteinlipase and glycerolphosphate dehydrogenase as well as an increase in insulin receptors [Gimble et al., 1989]. Like fibroblasts, marrow adipocytes are heterogeneous and display different characteristics related to their distribution in red or yellow marrow [Lichtman, 1984]. These characteristics include insulin independence but glucocorticoid dependence in vitro, positive staining with perfomic acid-Schiff, and higher concentrations of neutral fats and unsaturated fatty acids in triglycerides. In bone marrow in vivo, there appears to be an inverse relationship between adipogenesis and erythropoiesis: phenylhydrazine-induced anemia causes a rapid conversion of yellow marrow (containing adipocytes) to red marrow due to compensatory erythroid heperplasia [Maniatis et al., 1971]. However, the role of the adipocyte in supporting hematopoiesis in culture is controversial. Although Dexter and coworkers [1977] associated declining myelopoiesis in long-term murine bone marrow cultures with the gradual disappearance of adipocytes from the stromal layer, these cells may not be © 2000 by CRC Press LLC

necessary to support hematopoiesis in human bone marrow cultures [Touw & Lowenberg, 1983]. In this regard, IL-11 suppressed adipogenesis but simulated human CD34+ HL-DR + progenitor cells cocultured with human bone marrow stroma and enhanced the numbers of myeloid progenitor cells [Keller et al., 1993]. Hematopoiesis in vivo requires the proper ECM for cell attachment and differentiation as well as regulation by cytokines elaborated by stromal cells [Metcalf, 1993]. Although there is considerable redundancy in the synthesis of regulatory factors by different stromal cell populations, no single cell population has been identified that can provide an entire hematopoietic microenvironment. Further complicating the issue is a recent report indicating that CD 34+ hematopoietic progenitors produce soluble factors that control the production of some cytokines by stromal cells [Gupta et al., 1998]. If “cross-talk” between parenchyma and stroma is established as an important regulatory mechanism, the ratio between the relative numbers of stroma and parenchyma in a tissue culture will assume paramount importance. The use of cell lines to provide hematopoietic support will be discussed later in this chapter. In the liver, purified fat-storing cells are capable of broad-scale synthetic activity that includes collagens I, III, IV, fibronectin, heparan sulfate, chondroitin sulfate, and dermatan sulfate [DeLeeuw et al., 1984]. Adipocytes also synthesize colony-stimulating factors and other regulatory cytokines (Table 119.2). In addition to the above characteristics, adipocytes are desmin positive and therefore are phenotypically related to myogenic cells or myofibroblasts. The function of adipocytes may vary depending on their location. Adipocytes in the bone marrow are in close apposition to the sinusoidal endothelial cells and in the liver are found in the space of Disse under the endothelial cells. Hepatic fat-storing cells are finely integrated into several contiguous parenchymal cells and contain fat droplets that are qualitatively different than those found in hepatic parenchymal cells in that they contain high levels of retinols. Liver adipocytes are responsible for the metabolism and storage of vitamin A [DeLeeuw et al., 1984], provide some of the raw materials for the synthesis of biologic membranes and also contribute to local energy metabolism needs. Adeipocytes of other tissues also act as a type of “progenitor” cell that can be converted to different phenotypes (e.g., osteoblasts or chondroblasts) under the appropriate conditions and are capable of stimulating osteogenesis via their secretion of cytokines [Benayahu et al., 1993].

Macrophages Macrophages are derivatives of peripheral blood monocytes and are, therefore, bone-marrow-derived. They seed and remain on the surfaces of sinusoidal vessels in organs such as the liver and the spleen or migrate into the interstitial spaces of virtually all tissues. Macrophages are quintessential immunocompetent cells and are central components of many defense strategies, including randomized microbial phagocytosis and killing; antibody-dependent cellular cytotoxicity (ADCC), where they are directed against microbial or other cells that are opsonized with antibody; nonrandomized (specific) phagocytosis mediated by the association of immunoglobulins with a multiplicity of Fc receptors on their surfaces; the presentation of processed antigen to lymphocytes; secretion of and reaction to chemotaxins; and an enzymatic profile enabling them to move freely through tissue. The secretory capacity of macrophages is prodigious. In addition to plasma components such as complement proteins C1 through C5, they synthesize the ferric iron- and vitamin B12–building proteins, transferrin and transcobalamin II, as well as a host of locally acting bioreactive metabolites of arachidonic acid. Macrophage secretory activity appears to influence two simultaneous events in vivo, inflammation and tissue repair. One monokine, IL-1, enhances the adhesion of neutrophils to vascular endothelial cells and activates B and T lymphocytes and other macrophages while stimulating the formation of acute phase proteins inducing collagen, ECM, and cytokine synthesis by fibroblasts and other stromal cells. IL-1 as well as a host of other humoral regulatory factors, originate in macrophages, including TNFα , IFN α , GM-CSF, and MIP-1 α (Table 119.2). The ability of macrophages to secrete regulatory cytokines makes them an important contributor to the tissue microenvironment. Macrophages have been intrinsically associated with erythropoiesis in the bone marrow [Bessis & Breton-Gorius, 1959] and [Naughton et al., 1979]. As such, these “nurse cells” destroy defective erythroblasts and the fetal and regenerating liver, and provide recycled iron stores for hemoglobin synthesis by developing red cells. They also synthesize and/or store erythropoietin. © 2000 by CRC Press LLC

119.2

Stromal Cells as “Feeder Layers” for Parenchymal Cell Culture

Irradiated stromal feeder layers were first used by Puck and Marcus [1955] to support the attachment and proliferation of HeLa cells in culture. Direct contact with feeder layers of cells also permits the growth of glioma cells and epithelial cells derived from breast tissue and colon [Freshney et al., 1982]. Enhanced attachment of parenchymal cells and production of factors to regulate growth and differentiation are two important benefits of coculturing parenchymal and stromal cells.

Bone Marrow Bone marrow was the first tissue to be systematically investigated in regard to the influence of ECM and stromal cells on the production of blood cells. By the mid 1970s, the influence of microenvironmental conditions [Trentin, 1970; Wolf & Trentin, 1968] and ECM deposition [McCuskey et al., 1972; Schrock et al., 1973] upon hematopoiesis was well established in the hematology literature. Dexter and coworkers [1977] were the first to apply the feeder-layer-based coculture technology to hematopoietic cells by inoculating mouse bone marrow cells into preestablished, irradiated feeder layers of marrow-derived stromal (adherent) cells. These cultures remained hematopoietically active for several months. By comparison, bone marrow cells cultured in the absence of feeder layers or supplementary cytokines terminally differentiated within the first two weeks in culture; stromal cells were the only survivors [Brandt et al., 1990; Chang & Anderson, 1971]. If the cultures are not supplemented with exogenous growth factors, then myeloid cells, monocytic cells, and nucleated erythroblasts are present for the first 7–10 days of culture. This trilineage pattern is narrowed over successive weeks in culture so that from about 2–4 weeks myeloid and monocytic cells are produced, and, if cultured for longer terms, the products of the culture are almost entirely monocytic. However, the nature of hematopoietic support can be modulated by changing the environmental conditions. In this regard, cocultures established under identical conditions as the above produce B lymphocytes if the steroid supplementation of the medium is removed and the ambient temperature is increased [Whitlock & Witte, 1982]. Static bone marrow coculture systems are generally categorized as declining because of this and the finding that the hematopoietic progenitor cell concentrations decrease as a function of time in culture. However, this trend can be offset by coculturing bone marrow in bioreactors that provide constant media exchange and optimize oxygen delivery [Palsson et al., 1993] or by supplementing the cocultures with cocktails of the various growth factors elaborated by stromal cells [Keller et al., 1993]. Connective tissue feeder layers had originally been hypothesized to act by conditioning medium with soluble factors that stimulated growth and by providing a substratum for the selective attachment of certain types of cells [Puck & Marcus, 1955]. These are functions that have since been proven for bone marrow and other tissue culture systems. It is very difficult to recreate the marrow microenvironment in vitro. One reason is that the marrow stromal cell populations proliferate at different rates; if expanded vigorously using nutrient-rich medium, the culture selects for fibroblastic cells at the expense of the more slowly dividing types of stroma. We established rat bone marrow cocultures using stromal cells that were passaged for 6 months, and these cultures produced lower numbers of progenitor cells (CFU-C) in the adherent zone as compared to cocultures established with stroma that was only passed three to four times (Fig. 119.3). Cells released from cultures using the older stroma were almost entirely monocytic, even at earlier terms of culture. In a related experiment, we suspended nylon screen cocultures of passage 3–4 stroma and bone marrow hematopoietic cells in flasks containing confluent, 6-month-old stromal cell monolayers and found that these monolayer cells inhibited hematopoiesis in the coculture [Naughton et al., 1989]. We use early passage stroma for our cocultures in order to retain a representation of all the stromal cell types. If hematopoietic cells from cocultures are removed after about a month in vitro and reinoculated onto a new template containing passage 3–4 stromal cells and cultured for an additional month, the progenitor cell concentrations of the adherent zones are considerably higher than in cocultures where no transfer took place [Naughton et al., 1994]. This experiment indicated that continued access to “fresh”


FIGURE 119.3 Left. Mean CFU-C progenitor concentration of the adherent zone of a three-dimensional coculture of rat bone hematopoietic cells and rat bone marrow stroma. Stock cultures to generate the stromal cells for the cocultures were seeded onto the three-dimensional template either after passage (P3) or following expansion in monolayer culture for 6 months. Stroma at early passage is substantially more supportive of CFU-C progenitor cells than “old” stroma that is primarily fibroblastic in nature. Vertical lines through the means = ±1 standard error of the mean (sem). Right. Analysis of the cellular content of the nonadherent zone of rat bone marrow hematopoietic cell: stromal cell cocultures on three-dimensional templates by flow cytometry. Cocultures were established either with P3 stroma (closed figures) or stroma that was grown in monolayer culture for 6 months (open figures). The mean percentages of cells recognized by the phenotyping antibodies MOM/3F12/F2 (myeloid), ED-1 (monocytic), and OX-33 (B lymphoid) (Serotec, UK) are depicted. Vertical lines through the means = ±1 sem. Whereas low passage stroma generates myeloid, B lymphoid, and monocytic cells in cocultures and releases them into the nonadherent zone, 6-month-old stroma supports mainly the production of macrophages.

stroma enhances hematopoiesis. The maintenance of mixed populations of stromal cells are desirable for a number of reasons. The same cytokine may be synthesized by different stromal cells, but its mode of action is usually synergistic with other cytokines for the differentiation of a specific blood cell lineage(s). Differentiation may not occur in the absence one of these cytokines, although some degree of redundancy in cytokine expression is intrinsic to the stromal system [Metcalf, 1993]. The fibroblastic cells that “grow out” of later passage human and rodent bone marrow stroma share some phenotypic properties with muscle cells (αSM actin+, CD34-, STRO-1-). These cells (endothelial cells and macrophages are absent or present in negligible quantities) supported hematopoietic progenitors for up to 7 weeks of culture but no longer [Moreau et al., 1993], indicating that other types of cells are necessary for long-term hematopoiesis in vitro. Stromal cells derived from spleen or bone marrow were also shown to support the growth and maturation of rat natural killer cells in long-term three-dimensional culture [Tjota et al., 1992]. Natural killer cells produced by the cultures for more than 2 months in vitro continued to kill YAC-1 (NK sensitive) target cells when activated by IL-2. In more recent experiments, fetal thymus-derived stroma supported the production of CD45+/CD56+ human natural killer cells when cocultured with CD34+linprogenitors [Vaz et al., 1998]. In related work, stromal cell-derived IL-15 was found to enhance the production of CD56+/CD3- natural killer cells from human CD34+ hematopoietic progenitors [Mrózek et al., 1996]. In addition, the process of natural killer cell maturation (but not early progenitor cell proliferation) appears to require direct contact with stromal cells [Tsuji and Pollack, 1995]. Cytoadhesive molecule expression and its modulation is also an important consideration for the engineering of bone marrow and other tissues. In the case of bone marrow, this regulates “homing” or seeding, restricting cells in specific areas within tissues both to expose them to regulatory factors and to the release of cells from the supportive stroma upon maturation. For example, the sequential expression of hemonectin, a cytoadhesion molecule of myeloid cells, in yolk sac blood islands, liver, and bone marrow during embryonic and fetal life parallels the granulopoiesis occurring in these respective organs [Peters et al., 1990]. Hematopoietic stem cells [Simmons et al., 1992] and B-cell progenitors [Ryan et al., 1991]


TABLE 119.3

Cells Contributing to the Hepatic Microenvironment

Cell Type Stroma Kupffer cells

Size, µm

Relative Percent of Total Cells

12–16

8

Vascular endothelia

11-12

9

Biliary endothelia

10–12

5

Fat-storing cells

14–18

3

Fibroblasts

11–14

7

Pit cells Parenchymal cells Mononuclear (type 1)

11-15

1-2(variable)

17-22

35

Binuclear (type II)

20-27

27

Acidophilic (type III)

25-32

5

Characteristics MHC-1+, MHC-II+, Fcr, C3r, mannose and Dacetrylglucosamine receptors, acid phosphatase+, density = 1.076 gm/ml,*1.036 gm/ml† MHC-1+, MHC II+, vWF+, Fcr, TFr , mannose, apo-E, and scavenger receptors, density = 1.06–1.08 gm/ml,*1.036 gm/ml† MHC-I+, Positive for cytokeratins 7, 8, 18, 19, β2 microglobulin+, positive for VLA-2,3,6 integrins, agglutinate with UEA, WGA, SBA, PNA, density = 1.075–1.1 gm/ml,*1.0363gm/ml† MHC-I+, desmin+, retinol+, ECM expression, collagen I, IV expression., density =1.075–1.1 gm/ml* MHC-I+, ECM expression, collagen I, IV expression, density = 1.025 gm/ml,* 1.063 gm/ml† MHC-I,+, asialo-GM+, CD8+, CD-5MHC-I-, MHC-II-, blood group antigen-, density = 1.10–1.14 gm/ml,* 1.067 gm/ml† MHC-I±, MHC-II-, blood group antigen-, density = 1.10-1.14 gm/ml,* 1.071 gm/ml† MHC-1±, MHC-II-, blood group antigen-, density = 1.038 gm/ml†

Abbreviations: ECM = extracellular matrix, MHC – major histocompatibility complex, PNA = peanut agglutinin, SBA = soybean agglutinin, WGA = wheat germ agglutinin, UEA = Ulex europaeus agglutinin.

bind to VCAM-1 on cultures stromal cells by expression of VLA-4. In addition, erythroid progenitor cells bind preferentially to the fibronectin component of the ECM and remain bound throughout their differentiation [Tsai et al., 1987]. A basic requirement of stromal feeder layers is the expression of the proper cytoadhesion molecule profile to permit attachment of parenchyma; in this instance these are hematopoietic stem and progenitor cells.

Liver Liver is an hematopoietic organ during fetal life, and it contains many of the same stromal cell populations found in bone marrow (Table 119.3). Research trends in liver cell and bone marrow culture have followed a somewhat parallel course. Both tissues are difficult to maintain in vitro; the parenchymal cell numbers either declined (hematopoietic progenitors) or lost function and dedifferentiated (hepatocytes) over time in culture. When mixed suspensions of hepatic cells are inoculated into liquid medium, approximately 20% of the total cells attach. The nonadherent population remains viable for about 72 hours. The adherent cells, although they proliferate for only 24–48 hours after inoculation, can survive for substantially longer periods. However, many of these adherent parenchymal cells undergo drastic phenotypic alterations, especially if the medium is conditioned with serum. These changes include a flattening and spreading on plastic surfaces as well as a propensity to undergo nuclear division in the absence of concomitant cytoplasmic division. The appearance of these bizarre, multinucleated, giant cells herald the loss of liverspecific functions such as albumin synthesis and the metabolism of organic chemicals by cytochrome P450 enzymes. The percentage of hepaocytes attaching to the flask, the maintenance of rounded parenchymal cell phenotype, and the expression of specialized hepatic function in vitro can be enhanced by precoating the flasks with ECM components such as type I collagen [Michalopoulos & Pitot, 1975], fibronectin [Deschenes et al., 1980], homogenized liver tissue matrix [Reid et al., 1980], laminin, and type IV collagen [Bissell et al., 1987]. Hepatocyte survival and functional expression also improved when


hepatocytes were cocultured with liver-derived [Fry & Bridges, 1980] and murine 3T3 [Kuri-Harcuch & Mendoza-Figueroa, 1989] fibroblasts as well as a preestablished layer of adherent liver epithelial cells [Guguen-Guillouzo et al., 1983]. As with bone marrow culture on feeder layers, cells derived from the liver itself usually provide the best support in culture. However, we have found that hepatic parenchyma will express a differentiated function in vitro if supported by bone marrow stroma. Conversely, liverderived stromal cells support hematopoiesis in culture but his microenvironment favors erythropoisis (unpublished observations). In contrast to the role of bone marrow stroma on hematopoiesis, the influence of hepatic stromal cells on the function and/or growth of parenchymal hepatocytes has not been exhaustively investigated. However, several studies indicate the Kupffer cells, fat-strong cells, and perhaps other stroma influence parenchymal cell cytochrome P450 enzyme expression [Peterson & Renton, 1984] and act in tandem with parenchymal cells to metabolize lipopolysaccharide [Treon et al., 1979]. In addition, adipocytes and hepatic fibroblasts synthesize collagen type I and the proteoglycans heparan sulfate, dermatan sulfate, and chondroitin sulfate as well as hyaluronic acid, which is unsulfated in the liver and occurs as a simple GAG. Stromal cells as well as parenchyma deposit fibronectin. Liver adipocytes, like their relatives in bone marrow, express phenotypes (vimentin+, actin+, tubulin+) linking them histogenically to fibroblasts and mogenic cells [DeLeeuw et al, 1984]. Fat-storing cells as well as vascular endothelia apparently synthesize the type IV collagen found in the space of Disse, and fibrin originating form parenchymal cell fibrinogen synthesis is a significant part of liver matrix, at least in culture. As with the hematopoietic cells in bone marrow cultures, hepatic parenchymal cell gene expression is modulated by attachment to ECM and influenced by factors released by stromal cells. For example, parenchymal cells attaching to laminin-coated surfaces express the differentiation-associated substance α-fetoprotein, whereas the synthesis of albumin synthesis is favored when parenchymal cells bind to type IV collagen [Michalopoulous & Pitot, 1975]. Hepatocytes bind to fibronectin using the α5β integrin heterodimer and AGp110, a nonintegrin glycoprotein. Cytoadhesion molecule expression by heapatic cells and cells of other tissues changes with and perhaps controls development [Stamatoglou & Hughes, 1994]. Just as differential hemonectin expression occurred during the development of the hematopoietic system [Peters et al., 1990], a differential expression of liver proteins occurs during ontogeny. Although fibronectin and its receptor are strongly expressed in liver throughout life, AGp110 appears later in development and may guide the development of parenchyma into a polarized tissue. It will probably be necessary to incorporate the various stromal support cells found in liver into hepatocyte cultures.

Tumor Cells Stromal cells contribute to the process of neoplastic invasion of tissue by responding in a paracrine fashion to signals released by tumor cells. This includes the tumor-stimulated release of angiogenic factors from fibroblasts such as acid and basic FGF and the release of vascular endothelia growth factor/vascular proliferation factor (VEGF/VPF) by tumor cells to recruit endothelial cells and stimulate their proliferation and development into blood vessels to feed the growing neoplasm (reviewed by Wernert [1997]). In addition to cooperating in the establishment of tumors, stromal cells also contribute to their invasiveness. Activation and/or release of stromal cell-derived Tissue Factor (TF) (whose release by damaged tissue initiates the extrinsic limb of the protease coagulation cascade) has been associated with the progression from early to invasive breast cancer [Vrana et al., 1996]. In this regard, upon direct contact with tumor cells, stromal cells produce stromelysin 3 (STR-3), a matrix metalloproteinase that accelerates the migration of metastatic cells through tissue by degrading tissue matrix [Mari et al., 1998]. STR-3 is overexpressed in a wide variety of tumor stroma in breast, lung, colon, and other cancers. It would be appropriate to include coculture studies (i.e., tumor and stroma) in the investigation of the biology of cancer in vitro as well as its reponsivity to potential treatments. We have previously demonstrated that the presence of stroma alters the hematologic responsiveness to chemotherapy agents such as cyclophosphamide [Naughton et al., 1992]. In related work, primary cultures of murine plasmacytomas requires


TABLE 119.4

Some Representative Stromal Cell Lines Used to Support Hematopoietic Cells or Hepatocytes

Cell line Bone marrow AC-4 ALC GM 1380 GY-30 K-1 MBA-14 MS-1 U2 3T3 10 T1/2 10T1/2 clone D 266 AD Liver 3T3 Detroit 550

Species

Phenotype*

Mouse Mouse Human Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse

A A F A A F/M F F† F F A A

Mouse Human

F F

Support Capability Myeloid, monocytic, B lymphoid Myeloid, monocytic, B lymphoid Short-term myelo- + monopoiesis Myeloid, monocytic, B lymphoid myelopoiesis Enhances CFU-C numbers myelopoiesis Maintenance of CFU-S12, myelopoiesis Stimulates CFU-GEMM, -GM Stimulates CFU-GM only Myeloid, monocytic, B lymphoid Enhanced lipid metabolism, extends period of cytochrome P450 activity† Prolonged cytochrome P450 and NADPH-cytochrome C reductase activity‡

Source: Adapted from Anklesaria et al. [1987] and Gimble [1990]. *Major phenotype of the cell line (F = fibroblastic, M = monocytic, A = capable of supporting adipocyte phenotypes). †A bone marrow-derived cell line that was transformed with the large T oncogene of simian virus 40. ‡Support functions are compared to liquid cultures of rat hepatocytes without stroma. Abbreviations: CFU = colony-forming unit; G = granulocyte; M = macrophage; S = spleen; GEMM = granulocyte, erythrocyte, megakaryoctye, monocyte; NADPH = nicotinamide adenine dinucleotide phosphate.

a feeder layer. Alteration of ECM composition and other support functions of the stroma by antiinflammatory drugs prevented plasmacytoma growth in vivo and in vitro [De Grassi et al., 1993].

119.3

Support of Cultured Cells Using Cell Lines

The precise roles of the various cellular constituents of the tissue microenvironment have not been fully defined. One approach to understanding these mechanisms is to develop stroma cell lines with homogeneous and well-defined characteristics and then ascertain the ability of these cells to support parenchyma in coculture. A brief survey of some of these lines is found in Table 119.4. Stromal cell lines derived from bone marrow are usually either fibroblastic (F) or a mixture of fibroblastic and adipocytic cells which contain a subpopulation of fibroblastic cells that can be induced to undergo lipogenesis with dexamethasone or hydrocortisone (A) (reviewed by Gimble, [1990]). In general, fibroblastic lines support myelopoiesis and monocytopoiesis and stroma with both fibroblastic and adipocytic cells supports myelopoiesis as well as B lymphopoiesis. The MBA-14 cell line, which consists of stroma bearing fibroblastic as well as monocytic phenotypes, stimulated the formation of CFU-C (bipotential myeloid/monocytic) progenitors in coculture. Different cell lines that were transformed using the large T oncogene of simian virus 40 (U2) were used as feeder layers for bone marrow hematopoietic cells. Cocultures established using bone-marrow-derived stroma exhibited considerably better maintenance of the primitive stem cell CFU-S12 in culture than similarly transfected skin, lung, or kidney tissue cells [Rios & Williams, 1990]. Although feeder layers share a number of characteristics [cytokine production, ECM deposition), it is probably best to derive your feeder cells from the tissue that you wish to coculture rather than use xenogeneic feeder cells or cells derived from a completely dissimilar tissue. In this respect, we found that total cell output and progenitor cell concentrations were substantially higher in rat bone marrow cocultures supported with rat bone marrow stroma as compared to those established using immortalized human skin fibroblasts or fetal lung cell lines [Naughton & Naughton, 1989].


119.4

Stereotypic (Three-Dimensional) Culture Versus Monolayer Culture

Tissue is a three-dimensional arrangement of various types of cells that are organized into a functional unit. These cells also are polarized with respect to their position within tissue and the microenvironment, and therefore the metabolic activity and requirements of the tissue are not uniform throughout. Threedimensional culture was first performed successfully by Leighton [1951] using cellulose sponge as a template. Collagen gel frameworks [Douglas, 1980] also have been and are currently being employed to culture tissues such as skin, breast epithelium, and liver. In addition, tumor tissue cocultured with stroma in collagen gels respond to drugs in a manner similar to that observed in vivo [Rheinwald & Beckett, 1981]. We developed three-dimensional coculture templates using nylon filtration screens and felts made of polyester or bioresorbably polyglycolic acid polymers [Naughton et al., 1987, 1994]. Rodent or human bone marrow cocultures retained multilineage hematopoietic expression in these stereotypic cultures, a phenomenon that is possible in plastic flask or suspension cultures only if the medium is supplemented with cocktails of cytokines [Peters et al., 1993]. Cocultures of rat hepatic parenchymal and stromal cells on three-dimensional templates also displayed a number of liver-specific functions for at least 48 days in culture, including the active synthesis of albumin, fibrinogen, and other proteins and the expression of dioxin-inducible cytochrome P450 enzyme activity for up to 2 months in culture [Naughton et al., 1994]. Furthermore, hepatic parenchyma in these stereotypic cocultures proliferated in association with stromal elements and the ECM they deposited until all “open” areas within the template were utilized. Our method is different from others because we use stromal cells derived from the tissue we wish to culture to populate the three-dimensional template. These cells secrete tissue-specific ECM and other matrix components that are indigenous to the normal microenvironment of the tissue. Parenchymal cells associate freely within the template after their inoculation and bind to other cells and/or matrix based upon their natural cytoadhesion molecule profiles. We do not add exogenous proteins. Three-dimensional scaffolds such as nylon screen or polyester felt provide large surface areas for cell attachment and growth. Although mass transfer limitations of diffusion dictate the maximum thickness (density) of a tissue culture, suspended three-dimensional cultures have the advantage of being completely surrounded by medium. This arrangement effectively doubles the maximum tissue thickness that is possible with a plastic flask-based culture. These stereotypic cultures also appear to form tissuelike structure in vitro and, when implanted after coculture on bioresorbable polymer templates, in vivo.

Defining Terms When hematopoietic cells are inoculated into semisolid or liquid medium containing the appropriate growth factors, some of the cells are clonal and will form colonies after approximately 2 weeks in culture. These colonies arise from hematopoietic progenitor cells and are called colony-forming units or CFU. These colonies can consist of granulocytic cells (G), monocytic cells (M), a mixture of these two cell types (GM), or erythroid cells (E). Less mature progenitor cells have a greater potential for lineage expression. For example, a CFU-GEMM contains granulocytic, erythroid, megakaryocytic, and monocytic cells and is therefore the least mature progenitor of this group. The text also mentions CFU-S. Whereas the other assays quantify colony formation in vitro, this is an in vivo assay. Briefly, irradiated mice are infused with meatopoietic cells. Some of these cells will colonize the spleen and will produce blood cells that “rescue” the animal. As with the in vitro assays, colonies that arise later in culture originate from less mature cells. The CFU-S12 therefore is a more primitive hematopoietic cell than the CFU-S9 .


References Anklesaria P, Kase K, Glowacki J, et al. 1987. Engraftment of a clonal bone marrow stromal cell line in vivo stimulates hematopoietic recovery from total body irradiation. Proc Natl Acad Sci USA 84:7681. Beneyahu D, Zipori D, Wientroub S. 1993. Marrow adipocytes regulate growth and differentiation of osteoblasts. Biochem Biophys Res Comm 197:1245. Ben-Ze’ev A, Robinson GS, Bucher NLR, et al. 1988. Cell-cell and cell-matrix interactions differentially regulate the expression of hepatic and cytoskeletal genes in primary cultures of rat hepatocytes. Proc Nat Acad Sci (USA) 85:2161. Bessis M, Breton-Gorius J. 1959. Nouvelles observations sur l’ilot erythroblastique et la rhopheocytose de la ferritin. Rev Hemato 14:165. Bissell DM, Arenson DM, Maher JJ, et al. 1987. Support of cultured hepatocytes by a laminin-rich gel: Evidence for a functionally significant subendothelial matrix in normal rat liver. J Clin Invest 790:801. Brandt J, Srour EF, Van Besien K, et al. 1990. Cytokine-dependent long term culture of highly enriched precursors of hematopoietic progenitor cells from human bone marrow. J Clin Invest 86:932. Brockbank KGM, van Peer CMJ. 1983. Colony stimulating activity production by hemopoietic organ fibroblastoid cells in vitro. Acta Haematol 69:369. Chang VT, Anderson RN. 1971. Cultivation of mouse bone marrow cells: I. Growth of granulocytes. J Reticuloendoth Soc 9:568. Cicuttini FM, Martin M, Ashman L., et al. 1992. Support of human cord blood progenitor cells on human stromal cell lines transformed by SV40 large T antigen under the influence of an inducible (metallothionein) promoter. Blood 80:102. Degrassi A, Hilbert DM, Rudikoff S, et al. 1993. In vitro culture of primary plasmacytomas requires stromal cell feeder layers. Proc Nat Acad Sci (USA) 90:2060. DeLeeuw AM, McCarthy SP, Geerts A, et al. 1984. Purified rat liver fat-storing cells in culture divide and contain collagen. Hepatology 4:392. Dexter TM, Allen TD, Lajtha LG. 1977. Conditions controlling the proliferation of haematopoietic stem cells in vitro. J Cell Physiol 91:335. Douglas WHJ, Moorman GW, Teel RW. 1980. Visualization of cellular aggregates cultured on a threedimensional collagen sponge matrix. In Vitro 16:306. Freshney RI, Hart E, Russell JM. 1982. Isolation and purification of cell cultures from human tumours. In E Reid, GMW Cook, DJ Moore (eds), Cancer Cell Organelles. Methodological Surveys: Biochemistry, pp 97–110, Chichester, England, Horwood Press. Friedenstein AJ, Chailakhyan RK, Gerasimov UV. 1970. Bone marrow osteogenic stem cells: In vitro cultivation and transplantation in diffusion chambers. Cell Tiss Kinet 20:263. Furchgott RF, Zawadzki JV. 1980. The obligatory role of endothelial cells in the relaxation of arterial smooth muscle by actylcholine. Nature 286:373. Gallagher JT, Spooncer E, Dexter TM. 1982. Role of extracellular matrix in haemopoiesis: I. Synthesis of glycosaminoglycans by mouse bone marrow cultures. J Cell Sci 63:155. Gimble JM. 1990. The function of adipocytes in the bone marrow stroma. New Biol 2:304. Gimble JM, Dorheim MA, Cheng Q, et al. 1989. Response of bone marrow stromal cells to adipogenetic antagonists. Mol Cell Biol 9:4587. Greenberger JS. 1979. Corticosteroid-dependent differentiation of human marrow preadipocytes in vitro. In Vitro 15:823. Guguen-Guilluozo C, Clement B, Baffet G, et al. 1983. Maintenance and reversibility of active albumin secretion by adult rat hepatocytes co-cultured with another cell type. Exp Cell Res 143:47. Gupta P, Blazar BR, Gupta K, et al. 1998. Human CD34+ bone marrow cells regulate stromal production of interleukin-6 and granulocyte colony-stimulating factor and increase the colony-stimulating activity of stroma. Blood 91:3724.


Huang S, Terstappen LWMM. 1992. Formation of haematopoietic microenvironment and haematopoietic stem cells from single human bone marrow cells. Nature 360:745. Keller D, Ou XX, Rour EF, et al. 1993. Interleukin-II inhibits adipogenesis and stimulates myelopoiesis in human long-term marrow cultures. Blood 82:1428. Keltz MD, Atton E, Buradagunta S, et al. 1996. Modulation of leukemia inhibitory factor gene expression and protein biosynthesis in human fallopian tube. Am J Gyecol Obstet 175:1611. Kuri-Harcuch W, Mendoza-Figueroa T. 1989. Cultivation of adult rat hepatocytes on 3T3 cells: Expression of various liver differentiated functions. Differentiation 41:148. Leighton J. 1951. A sponge matrix method for tissue culture: Formation of organized aggregates of cells in vitro. J Natl Cancer Inst 12:545. Lictman MA. 1984. The relationship of stromal cells to hemopoietic cells in marrow. In DG Wright, JS Greenberger (eds), Long-Term Bone Marrow Culture, pp 3–30, New York, A.R. Liss. Lin CQ, Bissell MJ. 1993. Multi-faceted regulation of cell differentiation by extracellular matrix. FASEB J 7:737. Maniatis A, Tavassoli M, Crosby WH. 1971. Factors affecting the conversion of yellow to red marrow. Blood 37:581. Mari BP, Anderson IC, Mari SE, et al. 1998. Stromelysin-3 is induced in tumor/stroma cocultures and inactivated via a tumor-specific and basic fibroblast growth factor-dependent mechanism. J Biol Chem 273:618. McCuskey RS, Meineke HA, Townsend SF. 1972. Studies of the hemopoietic microenvironment: I. Changes in the microvascular system and stroma during erythropoietic regeneration and suppression in the spleens of CF1 mice. Blood 5:697. Metcalf D. 1993. Hematopoietic growth factors. Redundancy or subtlety? Blood 82:3515. Michalopoulos G, Pitot HC. 1975. Primary culture of parenchymal liver cells on collagen membranes. Exp Cell Res 94:70. Moreau I, Duvert V, Caux C, et al. 1993. Myofibroblastic stromal cells isolated from human bone marrow induce the proliferation of both early myeloid and B-lymphoid cells. Blood 82:2396. Mrózek E, Anderson P, and Aligiuri MA. 1996. Role of interleukin-15 in the development of human CD56+ natural killer cells from CD34+ human progenitor cells. Blood 87:2632. Naughton BA, Kolks GA, Arce JM, et al. 1979. The regenerating liver: A site of erythropoiesis in the adult long-evans rat. Amer J Anat 156:159. Naughton BA, Naughton GK. 1989. Hematopoiesis on nylon mesh templates. Ann NY Acad Sci 554:125. Naughton BA, San Roman J, Sibanda B, et al. 1994. Stereotypic culture systems for liver and bone marrow: Evidence for the development of functional tissue in vitro and following implantation in vivo. Biotech Bioeng 43:810. Naughton BA, Sibanda B, San Román J, et al. 1992. Differential effects of drugs upon hematopoiesis can be assessed in long-term bone marrow cultures established on nylon screens. Proc Soc Exp Biol Med 199:481. Owen ME. 1988. Marrow stromal stem cells. J Cell Sci 10:63. Palsson BO, Paek S-H, Schwartz RM, et al. 1993. Expansion of human bone marrow progenitor cells in a high cell density continuous perfusion system. Biotechnology 11:368. Peters C, O’Shea KS, Campbell AD, et al. 1990. Fetal expression of hemonectin: An extracellular matrix hematopoietic cytoadhesion molecule. Blood 75:357. Peterson TC, Renton KW. 1984. Depression of cytochrome P-450-dependent drug biotransformation in hepatocytes after the activation of the reticuloendothelial system by dextran sulfate. J Pharmacol Exp Ther 229:229. Puck TT, Marcus PI. 1955. A rapid method for viable cell titration and clone production with HeLa cells in tissue culture: the use of X-irradiated cells to supply conditioning factors. Proc Natl Aca Sci (USA) 41:432. Reid LM, Gaitmaitan Z, Arias I, et al. 1980. Long-term cultures of normal rat hepatocytes on liver biomatrix. Ann NY Acad Sci 349:70.


Rios M, Williams DA. 1990. Systematic analysis of the ability of stromal cell lines derived from different murine adult tissues to support maintenance of hematopoietic stem cells in vitro. J Cell Physiol 145:434. Roberts R, Gallagher J, Spooncer E, et al. 1988. Heparan sulfate bound growth factors: A mechanism for stromal cell-mediated haemopoiesis. Nature 332:376. Ryan DH, Nuccie BL, Abboud CN, et al. 1991. Vascular cell adhesion molecule-1 and the integrin VLA-4 mediate adhesion of human B cell precursors to cultured bone marrow adherent cells. J Clin Invest 88:995. Schrock LM, Judd JT, Meineke HA, et al. 1973. Differences in concentration of acid mucopolysaccharides between spleens of normal and polycythemic CF1 mice. Proc Soc Exp Biol Med 144:593. Siczkowski M, Amos As, Gordon MY. 1993. Hyaluronic acid regulates the function and distribution of sulfated glycosaminoglycans in bone marrow stromal cultures. Exp hematol 21:126. Simmons PJ, Masinovsky B, Longenecker BM, et al. 1992. Vascular cell adhesion molecule-1 expressed by bone marrow stromal cells mediates the binding of hematopoietic progenitor cells. Blood 80:388. Simmons PJ, Torok-Storb B. 1991. CD 34 expression by stromal precursors in normal human adult bone marrow. Blood 78:2848. Strobel E-S, Gay RE, Greenberg PL. 1986. Characterization of the in vitro stromal microenvironment of human bone marrow. Int J Cell Cloning 4:341. Stamatoglou SC, Hughes RC. 1994. Cell adhesion molecules in liver function and pattern formation. FASEB J 8:420. Thonar EJ-MA, Kuttner KE. 1987. Biochemical basis of age-related changes in porteoglycans. In: TN Wight, RP Mecham (eds), Biology of Proteoglycans, pp 211–246, New York, Academic Press. Tjota A, Rossi Tm, Naughton BA. 1992. Stromal cells derived from spleen or bone marrow support the proliferation of rat natural killer cells in long-term culture. Proc Soc Exp Biol Med 200:431. Touw I, Lowenberg B. 1983. No simulative effect of adipocytes on hematopoiesis in long-term human bone marrow cultures. Blood 61:770. Trentin JJ. 1970. Influence of hematopoietic organ stroma (Hematopoietic inductive microenvironment) on stem cell differentiation. In AS Gordon (ed), Regulation of Hematopoiesis, New York, Apppleton-Century-Crofts. Treon SP, Thomas P, Baron J. 1979. Lippopolysaccharide (LPS) processing by Kupffer cells releases a modified LPS with increased hepatocyte binding and decreased tumor necrosis-α stimulatory capacity. Proc Soc Exp Biol Med 202:153. Tsai S, Patel V. Beaumont E, et al. 1987. Differential binding to erythroid and myeloid progenitors to fibroblasts and fibronectin. Blood 69:1587. Tsuji JM and Pollack SB. 1995. Maturation of murine natural killer cells in the absence of exogenous cytokines requires contact with bone marrow stroma. Nat Immun 14:44. Van Eyken P, Desmet VJ. 1993. Bile duct cells. In AV Le Bouton (ed), Molecular and Cell Biology of the Liver, pp 475-524, Boca Raton, Fla, CRC Press. Vaz F, Srour EF, Almeida-Porada G, et al. 1998. Human thymic stroma supports natural killer (NK) cell development from immature progenitors. Cell Immunol 186:133. Vrana JA, Stang MT, Grande JP, et al. 1996. Expression of tissue factor in tumor stroma correlates with progression to invasive human breast cancer: paracrine regulation by carcinoma cell-derived members of the transforming growth factor beta family. Cancer Res 56:5063. Wernert N. 1997. The multiple roles of tumor stroma. Virchows Arch 430:433. Whitlock CA, Witte ON. 1982. Long term culture of B lymphocytes and their precursors from murine bone marrow. Proc Nat Aca Sci (USA) 77:4756. Wight TN, Kinsella MG, Keating A, et al. 1986. Proteoglycans in human long-term bone marrow cultures: Biochemical and ultrastructural analysis. Blood 67:1333. Wilson D. 1983. The origin of the endothelium in the developing marginal vein of the chick wing-bud. Cell Differ 13:63. Wolf NS, Trentin JJ. 1968. Hematopoietic colony studies: V. Effects of hemopoietic organ stroma on differentiation of pluripotent stem cells. J Exp Med 127:205.


Yanagisawa M, Jurihara HJ, Kimura S, et al. 1988. A novel potent vasoconstrictor peptide produced by vascular endothelial cells. Nature 332:411. Yang Y-C, Tsai S, Wong GG, et al. 1988. Interleukin-1 regulation of hematopoietic growth factor production by human stromal fibroblasts. J Cell Physiol 134:292. Zuckerman KS. 1984. Composition and function of the extracellular matrix in the stroma of long-term bone marrow cell cultures. In DG Wright, JS Greenberger (eds), Long-Term Bone Marrow Culture, pp 157–170, New York, A.R. Liss. Zuckerman KS, Rhodes RK, Goodrum DD, et al. 1985. Inhibition of collagen deposition in the extracellular matrix prevents the establishment of stroma supportive of hematopoiesis in long term murine bone marrow cultures. J Clin Invest 75:970.

Further Information There are a number of different methodologies for isolating cells including gradient density centrifugation, sedimentation at unit gravity, lectin agglutination, and reaction with specific antibodies followed by immunoselection via panning or immunomagnetic microspheres. These methods are described and illustrated well in volumes 1–5 of Cell Separation: Methods and Selected Applications, New York, Academic Press, 1987. For additional details concerning the relevance of ECM deposition to the development and functional expression of various types of tissue cells, the reader is referred to the serial reviews of this subject that appeared in The FASEB Journal from volume 7, number 9 (1993) to volume 8, number 4 (1994). Information about the various cell types of the liver, their interaction with matrix components, and their mechanisms of gene expression can be found in Molecular and Cell Biology of the Liver, Boca Raton, Florida, CRC Press (1993). Similarly, for more details concerning bone marrow cells and the mechanisms of hematopoiesis, consult The Human Bone Marrow, volume 1, Boca Raton, Florida, CRC Press (1992).


Koller, M. R., Palsson, B. O. “Tissue Engineering of Bone Marrow.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

120 Tissue Engineering of Bone Marrow 120.1

Biology of Hematopoiesis The Hematopoietic System: Function and Organization • Molecular Control of Hematopoiesis: The Hematopoietic Growth Factors • The Bone Marrow Microenvironment

120.2

Applications of Reconstituted Ex Vivo Hematopoiesis Bone Marrow Transplantation • Alternative Sources of Hematopoietic Cells for Transplantation • Tissue Engineering and Improved Transplantation Procedures • Large-Scale Production of Mature Blood Cells

120.3

Manfred R. Koller Oncosis

The Murine System • The Human System • Tissue Engineering Challenges

120.4

Challenges for Scale-Up Bioreactors and Stroma • Alternatives • Production of Mature Cells

Bernhard Ø. Palsson University of California–San Diego

The History of Hematopoietic Cell Culture Development

120.5

Recapitulation

The human body consumes a staggering 400 billion mature blood cells every day, and this number increases dramatically under conditions of stress such as infection or bleeding. A complex scheme of multilineage proliferation and differentiation, termed hematopoiesis (Greek for blood forming), has evolved to meet this demand. This regulated production of mature blood cells from immature stem cells, which occurs mainly in the bone marrow (BM) of adult mammals, has been the focus of considerable research effort. Ex vivo models of human hematopoiesis now exist that have significant scientific value and promise to have an impact on clinical practice in the near future. This endeavor is spread across many fields, including cell biology, molecular biology, bioengineering, and medicine. This chapter introduces the reader to the fundamental concepts of hematopoiesis, the clinical applications which drive much of the effort to reconstitute hematopoiesis ex vivo, and the progress made to date toward achieving this goal.

120.1 Biology of Hematopoiesis The Hematopoietic System: Function and Organization There are eight major types of mature blood cells which are found in the circulation (see Fig. 120.1). The blood cell population is divided into two major groups; the myeloid and lymphoid. The myeloid


FIGURE 120.1 The hematopoietic system hierarchy. Dividing pluripotent stem cells may undergo self-renewal to form daughter stem cells without loss of potential or may experience a concomitant differentiation to form daughter cells with more restricted potential. Continuous proliferation and differentiation along each lineage results in the production of many mature cells. This process is under the control of many growth factors (GFs) (See Table 120.1). The site of action of some of the better-studied GFs are shown. The mechanisms that determine which lineage a cell will develop into are not understood, although many models have been proposed.

lineage includes erythrocytes (red blood cells), monocytes, the granulocytes (neutrophils, eosinophils, and basophils), and platelets (derived from noncirculating megakaryocytes). Thymus-derived (T) lymphocytes and BM-derived (B) lymphocytes constitute the lymphoid lineage. Most mature blood cells exhibit a limited lifespan in vivo. Although some lymphocytes are thought to survive for many years, it has been shown that erythrocytes and neutrophils have lifespans of 120 days and 8 hours, respectively [Cronkite, 1988]. As a result, hematopoiesis is a highly prolific process which occurs throughout our lives to fulfill this demand. Mature cells are continuously produced from progenitor cells, which in turn are produced from earlier cells which originate from stem cells. There are many levels in this hierarchical system, which is usually diagrammed as shown in Fig. 120.1. At the left are the very primitive stem cells, the majority of which are in a nonproliferative state (Go) [Lajtha, 1979]. These cells are very rare (1 in 100,000 BM cells) but collectively have enough proliferative capacity to last several lifetimes [Boggs et al., 1982; Spangrude et al., 1988]. Through some unknown mechanism, at any given time a small number of these cells are actively © 2000 by CRC Press LLC

proliferating, differentiating, and self-renewing, thereby producing more mature progenitor cells while maintaining the size of the stem cell pool. Whereas stem cells (by definition) are not restricted to any lineage, their progenitor cell progeny do have a restricted potential and are far greater in number. The restricted nature of these progenitors has led to a nomenclature which describes their potential outcome. Those that develop into erythrocytes are called colony-forming unit-erythrocyte (CFU-E, the term colony-forming unit relates to the biological assay which is used to measure progenitor cells). Similarly, progenitors which form granulocytes and macrophages are called CFU-GM. Therefore, as the cells differentiate and travel from left to right in Fig. 120.1, they become more numerous, lose self-renewal ability, lose proliferative potential, become restricted to a single lineage, and finally become a mature cell of a particular type. The biology of stem cells is discussed elsewhere in this Handbook (Chapter 116). The need for identification of the many cell types present in the hematopoietic system has led to many types of assays. Many of these are biologic assays (such as colony-forming assays), which are performed by culturing the cells and examining their progeny, both in number and type [Sutherland et al., 1991a]. Another example is the long-term culture-initiating cell (LTC-IC) assay which measures a very early cell type through 5–16 weeks of in vitro maintenance [Koller et al., 1998a]. In contrast to these biologic assays, which are destructive to the cells being measured, is the real-time technique of flow cytometry. Flow cytometry has been used extensively in the study of the hematopoietic system hierarchy. Antibodies to antigens on many of the cell types shown in Fig. 120.1 have been developed (see [Brott et al., 1995]). Because of the close relation of many of the cell types, often combinations of antigens are required to definitively identify a particular cell. Recently, much effort has been focused on the identification of primitive stem cells, and this has been accomplished by analyzing increasingly smaller subsets of cells using increasingly complex antibody combinations. The first such antigen that was found in CD34, which identifies all cells from the stem through progenitor stage (typically about 2% of BM mononuclear cells (MNC), see Fig. 120.2) [Civin et al., 1984]. The CD34 antigen is stage-specific but not lineage-specific and therefore identifies cells that lead to repopulation of all cell lineages in transplant patients [Berenson et al., 1991]. However, the CD34 antigen is not restricted to hematopoietic cells because it is also found on certain stromal cells in the hematopoietic microenvironment (see Chapters 118 and 119). Although CD34 captures a small population which contains stem cells, this cell population is itself quite heterogeneous and can be fractionated by many other antigens. Over the past several years, many different combinations of antibodies have been used to fractionate the CD34+ population. CD34+ fractions which lack CD33, HLA-DR, CD38, or CD71 appear to be enriched in stem cells [Civin & Gore, 1993]. Conversely, CD34+ populations which coexpress Thy-1 or c-kit appear to contain the primitive cells [Civin & Gore, 1993]. These studies have revealed the extreme rarity of stem cells within the heterogeneous BM population (see Fig. 120.2). Of the mononuclear cell (MNC) subset (~40% of whole BM), only ~2% are CD34+, and of those, only ~5% may be CD38+. Furthermore, this extremely rare population is still heterogeneous with respect to stem cell content. Consequently, stem cells as single cells have not yet been identified.

Molecular Control of Hematopoiesis: The Hematopoietic Growth Factors A large number of hematopoietic growth factors (GFs) regulate both the production and functional activity of hematopoietic cells. The earliest to be discovered were the colony-stimulating factors (CSFs, because of their activity in the colony-forming assay), which include interleukin-3 (IL-3), granulocytemacrophage (GM)-CSF, granulocyte (G)-CSF, and monocyte (M)-CSF. These GFs, along with erythropoeitin, have been relatively well characterized because of their obvious effects on mature cell production and/or activation. The target cells of some of the better-studied GFs are shown in Fig. 120.1. Subsequent intensive research continues to add to the growing list of GFs that affect hematopoietic cell proliferation, differentiation, and function (Table 120.1). However, new GFs have been more difficult to find and characterize because their effects are more subtle, often providing a synergistic effect which potentiates other known GFs. In addition, there appears to be a significant amount of redundancy and pleotropy in this GF network, which makes the discovery of new GFs difficult [Metcalf, 1993]. In fact, more recent © 2000 by CRC Press LLC

FIGURE 120.2 Relative frequency of the different cell subsets within the BM population. A BM aspirate typically contains 99% mature erythroid cells (mostly from blood contamination), and therefore usually only the nucleated cell fraction is studied. Simple density gradient centrifugation techniques, which remove most of the mature erythrocytes and granulocytes, yield what is known as the mononuclear cell (MNC) fraction (about 40% of the nucleated cells). CD34+ cells, about 2% of MNC, can be isolated by a variety of methods to capture the primitive cells, although the population is quite heterogeneous. The most primitive cells are found in subsets of CD34+ cells (e.g., CD38-), which identify about 5% of the CD34+ population. These rare subsets can be obtained by flow cytometry but are still somewhat heterogeneous with respect to stem cell content. Consequently, although methods are available to fractionate BM to a great extent, individual stem cells have not yet been identified. This diagram conveys the heterogeneous nature of different BM populations as well as the incredible rarity of the primitive cell subset, which is known to contain the stem cells.

discoveries have focused on potential receptor molecules on the target cell surface, which then have been used to isolate the appropriate ligand GF. Examples of such recently discovered GFs which exhibit synergistic interactions with other GFs to act on primitive cells include c-kit ligand and flt-3 ligand. Another example is thrombopoietin, a stimulator of platelet production, a factor whose activity was described over 30 years ago but was cloned only recently [De Sauvage et al., 1994]. These and other GFs that act on primitive cells are the subject of intense study because of their potential scientific and commercial value. Already, several of the GFs have been developed into enormously successful pharmaceuticals, used in patients who have blood cell production deficiencies due to many different causes (see below). © 2000 by CRC Press LLC

TABLE 120.1

Hematopoietic Growth Factors

Growth Factor Name Interleukin-1 Interleukin-2 Interleukin-3 Interleukin-4 Interleukin-5 Interleukin-6 Interleukin-7 Interleukin-8 Interleukin-9 Interleukin-10 Interleukin-11 Interleukin-12 Interleukin-13 Interleukin-14 Interleukin-15 Interleukin-16 Interleukein-17 Interleukin-18 Erythropoietin Monocyte-CSF Granulocyte-CSF Granulocyte-macrophage-CF Stem cell factor Interferon-gamma Macrophage inflammatory protein-1 Leukemia inhibitory factor Transforming growth factorbeta Tumor necrosis factor-alpha flk-2 ligand Thrombopoietin

Other Names Hemopoietin-1 Multicolony-stimulating factor B-cell-stimulatory factor-1 B-cell-stimulatory factor-2 Neutrophil activating peptide-1 Cytokine synthesis inhibitory factor NK cell stimulatory factor High molecular weight B cell growth factor Lymphocyte chemoattractant factor IFN-gamma-inducing factor Colony-stimulating factor-1

c-kit ligand, Mast cell growth factor Macrophage activating factor Stem cell inhibitor

Cachetin flk-3 ligand c-mpl ligand; Megakaryocyte growth and development factor

Abbreviations

Reference

IL-1 IL-2 IL-3, Multi-CSF IL-4, BSF-1 IL-5 IL-6, BSF-2 IL-7 IL-8, NAP-1 IL-9 IL-10, CSIF IL-11 IL-12, NKSF IL-13 IL-14, HMW-BCGF IL-15 IL-16; LCF IL-17 IL-18; IGIF Epo M-CSF, CSF-1 G-CSF GM-CSF SCF, KL, MGF IFN-γ ; MAF MIP-1, SCI

Dinarello et al., 1981 Smith, 1988 Ihle et al., 1981 Yokota et al., 1988 Yokota et al., 1988 Kishimoto, 1989 Tushinski et al., 1991 Herbert and Baker, 1993 Donahue et al., 1990 Zlotnik and Moore,1991 Du and Williams, 1994 Wolf et al., 1991 Minty et al., 1993 Ambrus Jr. et al., 1993 Grabstein et al., 1994 Center et al., 1997 Yao et al., 1995 Ushio et al., 1996 Krantz, 1991 Metcalf, 1985 Metcalf, 1985 Metcalf, 1985 Zsebo et al., 1990 Virelizier et al., 1985 Graham et al., 1990

LIF TGF-β

Metcalf, 1991 Sporn and Roberts, 1989

TNF-α Fl Tpo, ML; MGDF

Pennica et al., 1984 Lyman et al., 1994 de Sauvage et al., 1994

The Bone Marrow Microenvironment Hematopoiesis occurs in the BM cavity in the presence of many accessory and support cells (or stromal cells). In addition, like all other cells in vivo, hematopoietic cells have considerable interaction with the extracellular matrix (ECM). These are the chief elements of what is known as the BM microenvironment. Further details on the function of stroma and the microenvironment, and their importance in tissue engineering, are found in Chapters 118 and 119. Bone Marrow Stromal Cells Due to the physiology of marrow, hematopoietic cells have a close structural and functional relationship with stromal cells. Marrow stroma includes fibroblasts, macrophages, endothelial cells, and adipocytes. The ratio of these different cell types varies at different places in the marrow and as the cells are cultured in vitro. The term stromal layer therefore refers to an undefined mixture of different adherent cell types which grow out from a culture of BM cells. In vitro, stem cells placed on a stromal cell layer will attach to and often migrate underneath the stromal layer [Yamakazi et al., 1989]. Under the stromal layer, some of the stem cells will proliferate, and the resulting progeny will be packed together, trapped under the stroma, forming a characteristic morphologic feature known as a cobblestone area (see below). It is widely believed that primitive cells must be in contact with stromal cells to maintain their primitive state. However, much of the effect of stromal cells has been attributed to the secretion of GFs. Consequently, there have been reports of successful hematopoietic cell growth with the addition of numerous soluble © 2000 by CRC Press LLC

GFs in the absence of stroma [Bodine et al., 1989; Brandt et al., 1990,1992; Haylock et al., 1992; Koller et al., 1992a; Verfaillie, 1992]. However, this issue is quite controversial, and stromal cells are still likely to be valuable because they synthesize membrane-bound GFs [Toksoz et al., 1992], ECM components [Long, 1992], and probably some as yet undiscovered cytokines. In addition, stromal cells can modulate the GF environment in a way that would be very difficult to duplicate by simply adding soluble GFs [Koller et al., 1995]. This modulation may be responsible for the observations that stroma can be both stimulatory and inhibitory [Zipori, 1989]. The Extracellular Matrix The ECM of BM consists of collagens, laminin, fibronectin [Zuckerman & Wicha, 1983], victronectin [Coulombel et al., 1988], hemonectin [Campbell et al., 1987], and thrombospondin [Long & Dixit, 1990]. The heterogeneity of this system is further complicated by the presence of various proteoglycans, which are themselves complex molecules with numerous glycosaminoglycan chains linked to a protein core [Minguell & Tavassoli, 1989; Spooncer et al., 1983; Wight et al., 1986]. These glycosaminoglycans include chondroitin, heparan, dermatan, and keratan sulfates and hyaluronic acid. The ECM is secreted by stromal cells of the BM (particularly endothelial cells and fibroblasts) and provides support and cohesion for the marrow structure. There is a growing body of evidence indicating that ECM is important for the regulation of hematopoiesis. Studies have shown that different glycosaminoglycans bind and present different GFs to hematopoietic cells in an active form [Gordon et al., 1987; Roberts et al., 1988]. This demonstrates that ECM can sequester and compartmentalize certain GFs in local areas and present them to hematopoietic cells, creating a number of different hematopoietically inductive microenvironments. Another important ECM function is to provide anchorage for immature hematopoietic cells. Erythroid precursors have receptors which allow them to attach to fibronectin. As cells mature through the BFU-E to the reticulocyte stage, adherence to fibronectin is gradually lost, and the cells are free to enter the circulation [Patel et al., 1985]. It has also been shown that binding to fibronectin renders these erythroid precursors more responsive to the effects of Epo [Weinstein et al., 1989]. Another adhesion protein termed hemonectin has been shown to selectively bind immature cells of the granulocyte lineage in an analogous fashion [Campbell et al., 1987]. This progenitor binding by ECM has led to the general concept for stem cell homing. When BM cells are injected into the circulation of an animal, a sufficient number are able to home to the marrow and reconstitute hematopoiesis [Tavassoli & Hardy, 1990]. It is therefore likely that homing molecules are present on the surface of the primitive cells. Studies suggest that lectins and CD44 on progenitor cells may interact with ECM and stromal elements of the marrow to mediate cellular homing [Aizawa & Tavassoli, 1987; Kansas et al., 1990; Lewinsohn et al., 1990; Tavassoli and Hardy, 1990]. These concepts have been reviewed in detail [Long, 1992].

120.2 Applications of Reconstituted Ex Vivo Hematopoiesis The hematopoietic system, as described above, has many complex and interacting features. The reconstitution of functional hematopoiesis, which has long been desired, must address these features to achieve a truly representative ex vivo system. A functioning ex vivo human hematopoietic system would be a valuable analytic model to study the basic biology of hematopoiesis. The clinical applications of functional ex vivo models of human hematopoiesis are numerous and are just beginning to be realized. Most of these applications revolve around cancer therapies and, more recently, gene therapy. The large-scale production of mature cells for transplantation represents an important goal that may be realized in the more distant future.

Bone Marrow Transplantation In 1980, when bone marrow transplantation (BMT) was still an experimental procedure, fewer than 200 BMTs were performed worldwide. Over the past decade, BMT has become an established therapy for many diseases. In 1996, over 40,000 BMTs were performed, primarily in the United States and Western © 2000 by CRC Press LLC

Europe, for more than a dozen different clinical indications [Horowitz and Rowling, 1997]. The number of BMTs performed annually is increasing at a rate of 20–30% per year and is expected to continue to rise in the foreseeable future. BMT is required as a treatment in a number of clinical settings because the highly prolific cells of the hematopoietic system are sensitive to many of the agents used to treat cancer patients. Chemotherapy and radiation therapy usually target rapidly cycling cells, so hematopoietic cells are ablated along with the cancer cells. Consequently, patients undergoing these therapies experience neutropenia (low neutrophil numbers), thrombocytopenia (low platelet numbers), and anemia (low red blood cell numbers), rendering them susceptible to infections and bleeding. A BMT dramatically shortens the period of neutropenia and thromobocytopenia, but the patient may require repeated transfusions. The period during which the patient is neutropenic represents the greatest risk associated with BMT. In addition, some patients do not achieve engraftment (when cell numbers rise to safe levels). As a result, much effort is focused on reducing the severity and duration of the blood cell nadir period following chemotherapy and radiation therapy. There are several sources of hematopoietic cells for transplantation. BMT may be performed with patient marrow (autologous) that has been removed and cryopreserved prior to administration of chemotherapy or with donor marrow (allogeneic). The numbers of autologous transplants outnumber allogenic transplants by a 2:1 ratio, but there are advantages and disadvantages with both techniques. Autologous Bone Marrow Transplantation Autologous BMTs have been used in the treatment of a variety of diseases including acute lymphoblastic leukemia (ALL), acute myelogenous leukemia (AML), chronic myelogenous leukemia (CML), various lymphomas, breast cancer, neuroblastoma, and multiple myeloma. Currently, autologous BM transplantation is hampered by the long hospital stay that is required until engraftment is achieved and the possibility of reintroducing tumor cells along with the cryopreserved marrow. In fact, retroviral marking studies have proven that tumor cells reinfused in the transplant can contribute to disease relapse in the patient [Rill et al., 1994; Deisseroth et al., 1994]. Long periods of neutropenia, anemia, and thrombocytopenia require parenteral antibiotic administration and repeated blood component transfusions. Autologous transplantation could also be used in gene therapy procedures. The basic concept underlying gene therapy of the hematopoietic system is the insertion of a therapeutic gene into the hematopoietic stem cell, so that a stable transfection is obtained. Engineered retroviruses are the gene carriers currently used. However, mitosis of primitive cells is required for integration of foreign DNA [Bodine et al., 1989]. A culture system which contains dividing stem cells is therefore critical for the enablement of retroviral-based gene therapy of the hematopoietic system. This requirement holds true whether or not the target stem cell population has been purified prior to the transfection step. To date, only very limited success has been achieved with retroviral transfection of human BM cells, whereas murine cells are routinely transfected. A comprehensive and accessible accounting of the status of gene therapy has been presented elsewhere [Mulligan, 1993]. Allogeneic Bone Marrow Transplantation In patients with certain hematologic malignancies, or with genetic defects in the hematopoietic population, allogeneic transplants are currently favored when suitable matched donors are available. With these leukemias, such as CML, it is likely that the patient's marrow is diseased and would not be suitable for autotransplant. A major obstacle in allogeneic transplantation, however, is the high incidence of graftversus-host disease (GVHD), in which the transplanted immune cells attack the host’s tissues as foreign.

Alternative Sources of Hematopoietic Cells for Transplantation Although hematopoiesis occurs mainly in the BM of adult mammals, during embryonic development, pluripotent stem cells first arise in the yolk sac, are later found in the fetal liver, and at the time of delivery are found in high concentrations in umbilical cord blood. In adults, stem cells are found in peripheral blood only at very low concentrations, but the concentration increases dramatically after stem cell mobilization. Mobilization of stem cells into peripheral blood is a phenomenon that occurs in response © 2000 by CRC Press LLC

to chemotherapy or GF administration. Therefore, hematopoietic stem cells can be collected from cord blood [Gluckman et al., 1989; Wagner et al., 1992] or from mobilized peripheral blood [Schneider et al., 1992] as well as from BM. Disadvantages of cord blood are the limited number of cells that can be obtained from the one individual and the question of whether this amount is sufficient to repopulate an adult patient. Mobilized peripheral blood results in more rapid patient engraftment than BM, and as a consequence, its use is becoming more prevalent, particularly in the autologous setting.

Tissue Engineering and Improved Transplantation Procedures BMT would be greatly facilitated by reliable systems and procedures for ex vivo stem cell maintenance, expansion, and manipulation. For example, the harvest procedure, which collects 1–2 liters of marrow, is currently a painful and involved operating room procedure. The complications and discomfort of marrow donation are not trivial and can affect donors for a month or more [Stroncek et al., 1993]. Through cell expansion techniques, a small marrow specimen taken under local anesthesia in an outpatient setting could be expanded into the large number of cells required for transplant, thereby eliminating the large harvest procedure. Engraftment may be accelerated by increasing the numbers of progenitors and immature cells available for infusion. In addition, it may be possible to cryopreserve expanded cells to be infused at multiple time points, thereby allowing multiple cycles of chemotherapy (schedule intensification). Finally, the use of expanded cells may allow increasing doses of chemotherapy (dose intensification), facilitating tumor reduction while ameliorating myeloablative side effects. The expansion of alternative hematopoietic cell sources would also facilitate transplant procedures. For example, multiple rounds of apheresis, each requiring ~4 hours, are required to collect enough mobilized peripheral blood cells for transplant. Expansion of a small amount of mobilized peripheral blood may reduce the number of aphereses required or eliminate them altogether by allowing the collection of enough cells from a volume of blood that has not been apheresed. With cord blood, there is a limit on the number of cells that can be collected from a single donor, and it is currently thought that this number is inadequate for an adult transplant. Consequently, cord blood transplants to date have been performed on children. Expansion of cord blood cells may therefore enable adult transplants from the limited number of cord blood cells available for collection.

Large-Scale Production of Mature Blood Cells Beyond the ability to produce stem and progenitor cells for transplantation purposes lies the promise to produce large quantities of mature blood cells. Large-scale hematopoietic cultures could potentially provide several types of clinically important mature blood cells. These include red blood cells, platelets, and granulocytes. About 12 million units of red blood cells are transfused in the United States every year, the majority of them during elective surgery and the rest in acute situations. About 4 million units of platelets are transfused every year into patients who have difficulty exhibiting normal blood clotting. Mature granulocytes, which constitute a relatively low-usage market of only a few thousand units administered each year, are involved in combating infections. This need arises in situations when a patient's immune system has been compromised and requires assistance in combating opportunistic infections, such as during chemotherapy and the healing of burn wounds. All in all, the market for these blood cells totals about $1–1.5 billion in the United States annually, with a worldwide market that is about three to four times larger. The ability to produce blood cells on demand ex vivo would alleviate several problems with the current blood cell supply. The first of these problems is the availability and stability of the blood cell supply. The availability of donors has traditionally been a problem, and, coupled with the short shelf-life of blood cells, the current supply is unstable and cannot meet major changes in demand. The second problem is the usual blood-type compatibility problem resulting in shortages of certain types at various times. The third problem is the safety of the blood supply. This last issue has received much attention recently due to the contamination of donated blood with the human immunodeficiency virus (HIV). However, the three forms of hepatitis currently pose an even more serious viral contamination threat to the blood supply. © 2000 by CRC Press LLC

Unlike ex vivo expansion of stem and progenitor cells for transplantation, the large-scale production of fully mature blood cells for routine clinical use is less developed and represents a more distant goal. The large market would require systems of immense size, unless major improvements in culture productivity are attained. For example, the recent discovery of thrombopoietin [De Sauvage et al., 1994] may make the large-scale production of platelets feasible. At present, there are ongoing attempts in several laboratories to produce large numbers of neutrophils from CD34-selected cell populations. Thus, largescale production of mature cells may soon become technically feasible, although the economic considerations are still unknown.

120.3 The History of Hematopoietic Cell Culture Development As outlined above, there are many compelling scientific and clinical reasons for undertaking the development of efficient ex vivo hematopoietic systems. Such achievement requires the use of in vivo mimicry, sophisticated cell culture technology, and the development of clinically acceptable cell cultivation devices. The foundations for these developments lie in the BM cell culture methods which have been developed over the past 20 years. The history of BM culture will therefore be described briefly, as it provides the backdrop for tissue engineering of the hematopoietic system. More complete reviews have been published previously [Dexter et al., 1984; Eaves et al., 1991; Greenberger, 1984].

The Murine System In the mid 1970s, Dexter and coworkers were successful in developing a culture system in which murine hematopoiesis could be maintained for several months [Dexter et al., 1977]. The key feature of this system was the establishment of a BM-derived stromal layer during the first three weeks of culture which was then recharged with fresh BM cells. One to 2 weeks after the cultures were recharged, active sites of hematopoiesis appeared. These sites are often described as cobblestone regions, which are the result of primitive cell proliferation (and accumulation) underneath the stromal layer. Traditionally, the cultures are fed by replacement of one-half of the medium either once or twice weekly. In these so-called Dexter cultures, myelopoiesis proceeds to the exclusion of lymphopoiesis. The selection of a proper lot of serum for long-term BM cultures (LTBMCs) was found to be very important. In fact, when using select lots of serum, one-step LTBMCs were successfully performed without the recharging step at week 3 [Dexter et al., 1984]. It is thought that good serum allows rapid development of stroma before the primitive cells are depleted, and once the stroma is developed, the culture is maintained from the remaining original primitive cells without need for recharging. The importance of stroma has often been demonstrated in these Dexter cultures, because the culture outcome was often correlated with stromal development.

The Human System The adaptation of one-step LTBMC for human cells was first reported in 1980 [Gartner & Kaplan, 1980]. A mixture of fetal bovine serum and horse serum was found to be required for human LTBMC, and a number of other medium additives such as sodium pyruvate, amino acids, vitamins, and antioxidants were found to be beneficial [Gartner & Kaplan, 1980; Greenberg et al., 1981; Meagher et al., 1988]. Otherwise, the culture protocol has remained essentially the same as that used for murine culture. Unfortunately, human LTBMCs have never attained the productivity or longevity which is observed in cultures of other species [Dexter et al., 1984; Greenberger et al., 1986]. The exponentially decreasing numbers of total and progenitor cells with time in human LTBMC [Eastment & Ruscetti, 1984; Eaves et al., 1991] renders the cultures unsuitable for cell expansion and indicates that primitive stem cells are lost over time. The discovery of hematopoietic GFs was an important development in human hematopoietic cell culture, because addition of GFs to human LTBMC greatly enhanced cell output. However, GFs did not prolong the longevity of the cultures, indicating that primitive cell maintenance was not © 2000 by CRC Press LLC

improved [Coutinho et al., 1990; Lemoli et al., 1992]. Furthermore, although the total number of progenitors obtained was increased by GF’s, it was still less than the number used to initiate the culture. Therefore, a net expansion in progenitor cell numbers was not obtained. The increased cell densities that were stimulated by GF addition were not well supported by the relatively static culture conditions. The disappointing results from human LTBMCs led to the development of other culture strategies. The development of an increasing number of recombinant GFs was soon joined by the discovery of the CD34 antigen (see above). As protocols for the selection of CD34+ cells became available, it was thought that the low cell numbers generated by enrichment could be expanded in GF-supplemented cultures without the impediment of numerous mature cells in the system. Because the enrichment procedure results in a cell population depleted of stromal cells, CD34+ cell cultures are often called suspension cultures, due to the lack of an adherent stromal layer. A number of groups have reported experiments in which CD34+ cells were incubated with high doses of up to seven recombinant GFs in suspension culture [Brandt et al., 1992; Haylock et al. 1992]. Although 500–1000-fold cell expansion numbers are often obtained, the magnitude of CFU-GM expansion is usually less than 10-fold, suggesting that differentiation, accompanied by depletion of primitive cells, is occurring in these systems. In fact, when LTC-IC have been measured, the numbers obtained after static culture of enriched cells have always been significantly below the input value [Sutherland et al., 1991b, 1993; Verfaillie, 1992]. A further consideration in CD34+ cell culture is the loss of cells during the enrichment procedure. It is not uncommon to experience 70–80% loss of progenitors with most CD34+ cell purification protocols [Traycoff et al., 1994], and this can be very significant when trying to maximize the final cell number obtained (such as in clinical applications, see below). Nevertheless, cultures of purified CD34+ cells, and especially the smaller subsets (e.g., CD33-, CD38-), have yielded valuable information on the biology of hematopoietic stem cells. An alternative approach has also been taken to improve human hematopoietic cell culture. Most of these advances have come from the realization that traditional culture protocols are highly nonphysiologic and that these deficiencies can be corrected by in vivo mimicry. Therefore, these techniques do not involve cell purification or high-dose cytokine stimulation. Because these cultures attain fairly high densities, it was thought that the tradition of changing one-half of the culture medium either once or twice weekly was inadequate. When Dexter-type cultures were performed with more frequent medium exchanges, progenitor cell production was supported for at least 20 weeks [Schwartz et al., 1991b]. This increase in culture longevity indicates that primitive cells were maintained for a longer period, and this was accompanied by an increase in progenitor cell yield. Although the precise mechanisms of increased medium exchange are unknown, the increased feeding rate enhances stromal cell secretion of GFs [Caldwell et al., 1991]. As previously noted, recombinant GFs can significantly improve culture productivity, but GF-stimulated cell proliferation exacerbates the problems of nutrient depletion because cell proliferation and the consumption of metabolites increases many-fold. Therefore, increased feeding protocols also benefit GFstimulated cultures. Cultures supplemented with IL-3/GM-CSF/Epo and fed with 50% daily medium exchanges were found to result in significant cell and progenitor expansion while maintaining culture longevity [Schwartz et al., 1991a]. Optimization of these manually fed cultures has been published for both BM [Koller et al., 1996] and cord blood [Koller et al., 1998b].

Tissue Engineering Challenges Although increased manual feeding significantly enhanced the productivity and longevity of hematopoietic cultures, the labor required to feed each culture is a daunting task. In addition, the cultures are subjected to physical disruption and large discontinuous changes in culture conditions and may be exposed to contamination at each feeding. Thee complications frustrate the optimization of the culture environment for production of hematopoietic cells and limit the clinical usefulness of the cultures. A perfusion system, if properly designed and constructed, would eliminate many of the problems currently associated with these cultures. © 2000 by CRC Press LLC

The success of this manual frequently fed culture approach led to development of continuously perfused bioreactors for human cord blood [Koller et al., 1993a; Koller et al., 1998b], BM [Palsson et al., 1993], and mobilized peripheral blood [Sandstrom et al., 1995] cell culture. Human BM MNC cultures have been performed in spinner flasks in a fed-batch mode as well [Zandstra et al., 1994]. Slow singlepass medium perfusion and internal oxygenation have given the best results to date, yielding cell densities in excess of 107 per ml accompanied by significant progenitor and primitive cell expansion [Koller et al., 1993b]. These systems have also been amenable to scale-up, first by a factor of ten, and then by a further factor of 7.5 [Koller et al., 1998b]. When an appropriate culture substrate is provided [Koller et al., 1998c], perfusion bioreactors support the development and maintenance of accessory cell populations, resulting in significant endogenous growth factor production which likely contributes to culture success [Koller et al., 1995a; Koller et al., 1997]. Importantly, stromal-containing cultures appear to generate greater numbers of primitive cells from a smaller initial cell sample, as measured by in vitro [Koller et al., 1995b; Sandstrom et al., 1995] and in vivo assays [Knobel et al., 1994], as compared with CD34-enriched cell cultures.

120.4 Challenges for Scale-Up To gauge the scale-up challenges, one first needs to state the requirements for clinically useful systems. The need to accommodate stroma that supports active hematopoiesis represents perhaps the most important consideration for the selection of a bioreactor system.

Bioreactors and Stroma If stroma is required, the choices are limited to systems that can support the growth of adherent cells. This requirement may be eliminated in future systems if the precise GF requirements become known and if the microenvironment that the stroma provides is not needed for hematopoietic stem and progenitor cell expansion. Currently, there are at least three culture systems that may be used for adherent cell growth. Fluidized bed bioreactors with macroporous bead carriers provide one option. Undoubtedly, significant effort will be required to develop the suitable bead chemistry and geometry, since the currently available systems are designed for homogeneous cell cultures. Beads for hematopoietic culture probably should allow for the formation of functional colonies comprised of a mixed cell population within each bead. Cell sampling from fluidized beds would be relatively easy, but final cell harvesting may require stressful procedures, such as prolonged treatment with collagenase and/or trypsin. Flatbed bioreactors are a second type of system which can support stromal development. Such units can be readily designed to carry the required cell number, and, further, they can allow for direct microscopic observation of the cell culture. Flatbed bioreactors provide perhaps the most straightforward scaleup and automation of LTBMC. In fact, such automated systems have been developed and used in human clinical trials for the treatment of cancer [Mandalam et al., 1998]. Finally, membrane-based systems, such as hollow fiber units, could be used to carry out hematopoietic cell cultures of moderate size. Special design of the hollow fiber bed geometry with respect to axial length and radial fiberspacing should eliminate all undesirable spatial gradients. Such units have been made already and have proved effective for their use for in vivo NMR analysis of metabolic behavior of homogenous cell cultures [Mancuso et al., 1990]. However, hematopoietic cell observation and harvesting may prove to be troublesome with this approach, as one report has suggested [Sardonini & Wu, 1993]. It is possible that the function of accessory cells may be obtained by the use of spheroids without classical adherent cell growth. This approach to the growth of liver cells in culture has meet with some success (see Chapter 121). In fact, there have been reports of functional heterogeneous cell aggregates within BM aspirates [Blazsek et al., 1990; Funk et al., 1994]. If successfully developed, suspension cultures containing these aggregates could be carried out in a variety of devices, including the rotating wall vessels that have been developed by NASA [Schwarz et al., 1992]. © 2000 by CRC Press LLC

Alternatives The precise arrangement of future large-scale systems will be significantly influenced by continuing advances in the understanding of the molecular and microenvironmental regulation of the hematopoietic process. Currently, the proximity of a supporting stromal layer is believed to be important. It is thought to function through the provision of both soluble and membrane-bound GFs and by providing a suitable microenvironment. The characterization of the microenvironment is uncertain at present, but its chemistry and local geometry are both thought to play a role. If the proximity of stroma is found to be unimportant, one could possibly design culture systems in which the stroma and BM cells are separated [Verfaillie, 1992], resulting in the ability to control each function separately. Finally, if the GF requirements can be defined and artificially supplied, and the stromal microenvironment is found to be unimportant, large-scale hematopoietic suspension cultures would become possible.

Production of Mature Cells Culture systems for generic allogeneic BMT, or for the large-scale production of mature cells, will pose more serious scale-up challenges. The number of cells required for these applications, in particular the latter, may be significantly higher than that for autologous BMT. Of the three alternatives discussed above, the fluidized bed system is the most readily scalable. The flatbed systems can be scaled by a simple stacking approach, whereas hollow fiber units are known for their shortcomings with respect to large-scale use.

120.5 Recapitulation Rapid advances in our understanding of hematopoietic cell biology and the molecular control of hematopoietic cell replication, differentiation, and apoptosis are providing some of the basic information that is needed to reconstitute human hematopoiesis ex vivo. Compelling clinical applications provide a significant impetus for developing systems that produce clinically useful cell populations in clinically meaningful numbers. Use of in vivo mimicry and the bioreactor technologies that were developed in the 1980s are leading to the development of perfusion-based bioreactor systems that will meet some of the clinical needs. Further tissue engineering of human hematopoiesis is likely to continue to grow in scope and sophistication and lead to definition of basic structure-function relationships and the enablement of many needed clinical procedures.

Defining Terms Colony-forming assay: Assay carried out in semisolid medium under GF stimulation. Progenitor cells divide, and progeny are held in place so that a microscopically identifiable colony results after 2 weeks. Differentiation: The irreversible progression of a cell or cell population to a more mature state. Engraftment: The attainment of a safe number of circulating mature blood cells after a BMT. Flow cytometry: Technique for cell analysis using fluorescently conjugated monoclonal antibodies which identify certain cell types. More sophisticated instruments are capable of sorting cells into different populations as they are analyzed. Graft-versus-host disease: The immunologic response of transplanted cells against the tissue of their new host. This response is often a severe consequence of allogeneic BMT and can lead to death (acute GVHD) or long-term disability (chronic GVHD). Hematopoiesis: The regulated production of mature blood cells through a scheme of multilineage proliferation and differentiation. Lineage: Refers to cells at all stages of differentiation leading to a particular mature cell type, i.e., one branch on the lineage diagram shown in Fig. 120.1.


Long-term culture-initiating cell: Cell that is measured by a 7–18 week in vitro assay. LTC-IC are thought to be very primitive, and the population contains stem cells. However, the population is heterogeneous, so not every LTC-IC is a stem cell. Miroenvironment: Refers to the environment surrounding a given cell in vivo. Mononuclear cell: Refers to the cell population obtained after density centrifugation of whole BM. This population excludes cells without a nucleus (erythrocytes) and polymorphonuclear cells (granulocytes). Progenitor cells: Cells that are intermediate in the development pathway, more mature than stem cells but not yet mature cells. This is the cell type measured in the colony-forming assay. Self-renewal: Generation of a daughter cell with identical characteristics as the original cell. Most often used to refer to stem cell division, which results in the formation of new stem cells. Stem cells: Cells with potentially unlimited proliferative and lineage potential. Stromal cells: Heterogeneous mixture of support or accessory cells of the BM. Also refers to the adherent layer which forms in BM cultures.

References Aizawa S, Tavassoli M. 1987. In vitro homing of hemopoietic stem cells is mediated by a recognition system with galactosyl and mannosyl specificities. Proc Nat Acad Sci 84:4485. Ambrus Jr. JL, Pippin J, Joseph A, Xu C, Blumenthal D, Tamayo A, Claypool K, McCourt D, Srikiatchatochorn A, and Ford RJ. 1993. Identification of a cDNA for a human high-molecular weight B-cell growth factor. Proc Natl Acad Sci 90:6330–6334. Armstrong RD, Koller MR, Paul LA, et al. 1993. Clinical scale production of stem and hematopoietic cells ex vivo. Blood 82:296a. Berenson RJ, Bensinger WI, Hill RS, et al. 1991. Engraftment after infusion of CD34+ marrow cells in patients with breast cancer or neuroblastoma. Blood 77:1717. Blazsek I, Misset J-L, Benavides M, et al. 1990. Hematon, a multicellular functional unit in normal human bone marrow: Structural organization, hemopoietic activity, and its relationship to myelodysplasia and myeloid leukemias. Exp Hematol 18:259. Bodine DM, Karlsson S, Nienhuis AW. 1989. Combination of interleukins 3 and 6 preserves stem cell function in culture and enhances retrovirus-mediated gene transfer into hematopoietic stem cells. Proc Natl Acad Sci 86:8897. Boggs DR, Boggs SS, Saxe DF, et al. 1982. Hematopoietic stem cells with high proliferative potential. J Clin Invest 70:242. Brandt JE, Briddell RA, Srour EF, Leemhuis TB, and Hoffman R. 1992. Role of c-kit ligand in the expansion of human hematopoietic progenitor cells. Blood 79:634–641. Brandt JE, Srour EF, Van Besien K, et al. 1990. Cytokine-dependent long-term culture of highly enriched precursors of hematopoietic progenitor cells from human bone marrow. J Clin Invest 86:932. Brott DA, Koller MR, Rummel SA, Palsson BO. 1995. Flow cytometric analysis of cells obtained from human bone marrow cultures. In M Al-Rubeai, AN Emery (eds) Flow cytometry applications in cell culture, pp 121–146, Marcel Dekker, New York. Caldwell J, Palsson BØ, Locey B, et al. 1991. Culture perfusion schedules influence the metabolic activity and granulocyte-macrophage colony-stimulating factor production rates of human bone marrow stromal cells. J Cell Physiol 147:344. Campbell AD, Long MW, Wicha MS. 1987. Haemonectin, a bone marrow adhesion protein specific for cells of granulocyte lineage. Nature 329:744. Center DM, Kornfeld H, Cruikshand WW. 1997. Interleukin-16. Int J Biochem Cell Biol 29:1231–1234. Civin CI, Gore SD. 1993. Antigenic analysis of hematopoiesis: A review. J Hematotherapy 2:137. Civin CI, Strauss LC, Brovall C, et al. 1984. Antigenic analysis of hematopoiesis: III. A hematopoietic progenitor cell surface antigen defined by a monoclonal antibody raised against KG-Ia cells. J Immunol 133:157. © 2000 by CRC Press LLC

Coulombel L, Vuillet MH, Leroy C, et al. 1988. Lineage- and stage-specific adhesion of human hematopoietic progenitor cells to extracellular matrices from marrow fibroblasts. Blood 71:329. Coutinho LH, Will A, Radford J, et al. 1990. Effects of recombinant human granulocyte colony-stimulating factor (CSF), human granulocyte macrophage-CSF, and gibbon interleukin-3 on hematopoiesis in human long-term bone marrow culture. Blood 75:2118. Cronkite EP. 1988. Analytical review of structure and regulation of hemopoiesis. Blood Cells 14:313. De Sauvage FJ, Hass PE, Spencer SD, et al. 1994. Stimulation of megakaryocytopoiesis and thrombopoiesis by the c-Mpl ligand. Nature 369:533. Dexter TM, Allen TD, Lajtha LG. 1977. Conditions controlling the proliferation of haemopoietic stem cells in vitro. J Cell Physiol 91:335. Dexter TM, Spooncer E, Simmons P, et al. 1984. Long-term marrow culture: An overview of techniques and experience 121–146. In DG Wright, JS Greenberger (eds), Long-Term Bone Marrow Culture, pp 57–96, New York, Alan R. Liss. Dinarello CA, Rosenwasser LJ, Wolff SM. 1981. Demonstrating of a circulating suppressor factor of thymocyte proliferation during endotoxin fever in humans. J Immunol 127:2517. Donahue RE, Yang Y-C, Clark SC. 1990. Human P40 T-cell growth factor (interleukin 9) supports erythroid colony formation. Blood 75:2271. Du XX, Williams DA. 1994. Interleukin-11: A multifunctional growth factor derived from the hematopoietic microenvironment. Blood 83:2023. Eastment CE, Ruscetti FW. 1984. Evaluation of hematopoiesis in long-term bone marrow culture: Comparison of species differences. In DG Wright, JS Greenberger (eds), Long-Term Bone Marrow Culture, pp 97–118, New York, Alan R. Liss. Eaves CJ, Cashman JD, Eaves AC. 1991. Methodology of long-term culture of human hemopoietic cells. J Tiss Cult Meth 13:55. Funk PE, Kincade PW, Witte PL. 1994. Native associations of early hematopoietic stem cells and stromal cells isolated in bone marrow cell aggregates. Blood 83:361. Gartner S, Kaplan HS. 1980. Long-term culture of human bone marrow cells. Proc Natl Acad Sci 77:4756. Gluckman E, Broxmeyer HE, Auerback AD, et al. 1989. Hematopoietic reconstitution in a patient with Fanconi's anemia by means of umbilical-cord blood from an HLA-identical sibling. NE J Med 321:1174. Gordon MY, Riley GP, Watt SM, et al. 1987. Compartmentalization of a haematopoietic growth factor (GM-CSF) by glycosaminoglycans in the bone marrow microenvironment. Nature 326:403. Grabstein KH, Eisenman J, Shanebeck K, Rauch C, Srinivasan S, Fung V, Beers C, Richardson J, Schoenborn MA, Ahdieh M. 1994. Cloning of a T cell growth factor that interacts with the beta chain of the interleukin-2 receptor. Science 264:965–968. Graham GJ, Wright EG, Hewick R, et al. 1990. Identification and characterization of an inhibitor of haemopoietic stem cell proliferation. Nature 344:442. Greenberg HM, Newburger PE, Parker LM, et al. 1981. Human granulocytes generated in continuous bone marrow culture are physiologically normal. Blood 58:724. Greenberger JS. 1984. Long-term hematopoietic cultures. In DW Golde (ed), Hematopoiesis, pp 203–242, New York, Churchill Livingstone. Greenberger JS, Fitzgerald TJ, Rothstein L, et al. 1986. Long-term culture of human granulocytes and granulocyte progenitor cells. In Transfusion Medicine: Recent Technological Advances, pp 159–185, New York, Alan R. Liss. Haylock DN, To LB, Dowse TL, et al. 1992. Ex vivo expansion and maturation of peripheral blood CD34+ cells into the myeloid lineage. Blood 80:1405. Herbert CA, Baker JB. 1993. Interleukin-8: A review. Cancer Invest 11:743. Hoffman R, Benz EJ Jr, Shattil SJ, et al. 1991. Hematology: Basic Principles and Practice, New York, Churchill Livingstone. Horowitz MM, Rowlings PA. 1997. An update from the International Bone Marrow Transplant Registry and the Autologous Blood and Marrow Transplant Registry on current activity in hematopoietic stem cell transplantation. Curr Opin Hematol 4:359–400. © 2000 by CRC Press LLC

Ihle JN, Pepersack L, Rebar L. 1981. Regulation of T cell differentiation: In vitro induction of 20 alphahydroxysteroid dehydrogenase in splenic lymphocytes is mediated by a unique lymphokine. J Immunol 126:2184. Kansas GS, Muirhead MJ, Dailey MO. 1990. Expression of the CD11/CD18, leukocyte adhesion molecule 1, and CD44 adhesion molecules during normal myeloid and erythroid differentiation in humans. Blood 76:2483. Kishimoto T. 1989. The biology of interleukin-6. Blood 74:1. Knobel KM, McNally MA, Berson AE, Rood D, Chen K, Kilinski L, Tran K, Okarma TB, Lebkowski JS. 1994. Long-term reconstitution of mice after ex vivo expansion of bone marrow cells: Differential activity of cultured bone marrow and enriched stem cell populations. Exp Hematol 22:1227–1235. Koller MR, Bender JG, Miller WM, et al. 1993a. Expansion of human hematopoietic progenitors in a perfusion bioreactor system with IL-3, IL-6, and stem cell factor. Biotechnology 11:358. Koller MR, Bender JG, Papoutsakis ET, et al. 1992a. Effects of synergistic cytokine combinations, low oxygen, and irradiated stroma on the expansion of human cord blood progenitors. Blood 80:403. Koller MR, Bender JG, Papoutsakis ET, et al. 1992b. Beneficial effects of reduced oxygen tension and perfusion in long-term hematopoietic cultures. Ann NY Acad Sci 665:105. Koller MR, Emerson SG, Palsson BØ. 1993b. Large-scale expansion of human stem and progenitor cells from bone marrow mononuclear cells in continuous perfusion culture. Blood 82:378. Koller MR, Bradley MS, Palsson BØ. 1995. Growth factor consumption and production in perfusion cultures of human bone marrow correlates with specific cell production. Exp Hematol, 23:1275. Koller MR, Manchel I, Palsson BO. 1997. Importance of parenchymal:stromal cell ratio for the ex vivo reconstitution of human hematopoiesis. Stem Cells 15:305–313. Koller MR, Manchel I, Smith AK. 1998a. Quantitative long-term culture-initiating cell assays require accessory cell depletion that can be achieved by CD34-enrichment or 5-fluorouracil exposure. Blood 91:4056. Koller MR, Manchel I, Maher RJ, Goltry KL, Armstrong RD, Smith AK. 1998b. Clinical-scale human umbilical cord blood cell expansion in a novel automated perfusion culture system. Bone Marrow Transplant 21:653. Koller MR, Manchel I, Palsson MA, Maher RJ, Palsson, BØ. 1996. Different measures of human hematopoietic cell culture performance are optimized under vastly different conditions. Biotechnol. Bioeng. 50:505–513. Koller MR, Palsson MA, Manchel I, Palsson, BØ. 1995b. LTC-IC expansion is dependent on frequent medium exchange combined with stromal and other accessory cell effects. Blood 86:1784–1793. Koller MR, Palsson MA, Manchel I, Maher RJ, Palsson BØ. 1998c. Tissue culture surface characteristics influence the expansion of human bone marrow cells. Biomaterials (in press). Krantz SB. 1991. Erythropoietin. Blood 77:419. Lajtha LG. 1979. Stem cell concepts. Differentiation 14:23. Lemoli RM, Tafuri A, Strife A, et al. 1992. Proliferation of human hematopoietic progenitors in longterm bone marrow cultures in gas permeable plastic bags is enhanced by colony-stimulating factors. Exp Hematol 20:569. Lewinsohn DM, Nagler A, Ginzton N, et al. 1990. Hematopoietic progenitor cell expression of the H-CAM (CD44) homing-associated adhesion molecule. Blood 75:589. Long MW. 1992. Blood cell cytoadhesion molecules. Exp Hematol 20:288 Long MW, Dixit VM. 1990. Thrombospondin functions as a cytoadhesion molecule for human hematopoietic progenitor cells. Blood 75:2311. Lyman SD, James L, Johnson L, Brasel K, de Vries P, Escobar SS, Downey H, Splett RR, Beckmann MP, McKenna HJ. 1994. Cloning of human homologue of the murine flt3 ligand: A growth factor for early hematopoietic progenitor cells. Blood 83:2795–2801. Mancuso A, Fernandez EJ, Blanch HW, et al. 1990. A nuclear magnetic resonance technique for determining hybridoma cell concentration in hollow fiber bioreactors. Biotechnology 8:1282.


Mandalam R, Koller MR, Smith AK. 1998. Ex vivo hematopoietic cell expansion for bone marrow transplantation. In R Nordon (ed), Ex vivo cell therapy. Landes Bioscience, Austin, TX. (in press) Meagher RC, Salvado AJ, Wright DG. 1988. An analysis of the multilineage production of human hematopoietic progenitors in long-term bone marrow culture: Evidence that reactive oxygen intermediates derived from mature phagocytic cells have a role in limiting progenitor cell self-renewal. Blood 72:273. Metcalf D. 1985. The granulocyte-macrophage colony-stimulating factors. Science 229:16. Metcalf D. 1991. The leukemia inhibitory factor (LIF). Int J Cell Cloning 9:95. Metcalf D. 1993. Hematopoietic regulators: Redundancy or subtlety? Blood 82:3515. Minguell JJ, Tavassoli M. 1989. Proteoglycan synthesis by hematopoietic progenitor cells. Blood 73:1821. Minty A, Chalon P, Derocq JM, et al. 1993. Interleukin-13 is a new human lymphokine regulating inflammatory and immune responses. Nature 362:248. Mulligan RC. 1993. The basic science of gene therapy. Science 260:926–932. Palsson BØ, Paek S-H, Schwartz RM, et al. 1993. Expansion of human bone marrow progenitor cells in a high cell density continuous perfusion system. Biotechnology 11:368. Patel VP, Ciechanover A, Platt O, et al. 1985. Mammalian reticulocytes lose adhesion to fibronectin during maturation to erythrocytes. Proc Natl Acad Sci 82:440. Pennica D, Nedwin GE, Hayflick JS, et al. 1984. Human tumor necrosis factor: Precursor structure, expression and homology to lymphotoxin. Nature 312:724. Roberts R, Gallagher J, Spooncer E, et al. 1988. Heparan sulphate bound growth factors: A mechanism for stromal cell mediated haemopoiesis. Nature 332:376. Sandstrom CE, Bender JG, Papoutsakis ET, Miller WM. 1995. Effects of CD34+ cell selection and perfusion on ex vivo expansion of peripheral blood mononuclear cells. Blood 86:958–970. Sardonini CA, Wu Y-J. 1993. Expansion and differentiation of human hematopoietic cells from static cultures through small scale bioreactors. Biotechnol Prog. 9:131. Schneider JG, Crown J, Shapiro F, et al. 1992. Ex vivo cytokine expansion of CD34-positive hematopoietic progenitors in bone marrow, placental cord blood, and cyclophosphamide and G-CSF mobilized peripheral blood. Blood 80:268a. Schwartz RM, Emerson SG, Clarke MF, et al. 1991a. In vitro myelopoiesis stimulated by rapid medium exchange and supplementation with hematopoietic growth factors. Blood 78:3155. Schwartz RM, Palsson BØ, Emerson SG. 1991b. Rapid medium perfusion rate significantly increases the productivity and longevity of human bone marrow cultures. Proc Natl Acad Sci 88:6760. Schwarz RP, Goodwin TJ, Wolf DA. 1992. Cell culture for three-dimensional modeling in rotating wall vessels: An application of simulated microgravity. J Tiss Cult Meth 14:51. Smith KA. 1988. Interleukin-2: Inception, impact, and implications. Science 240:1169. Spangrude GJ, Heimfeld S, Weissman IL. 1988. Purification and characterization of mouse hematopoietic stem cells. Science 241:58. Spooncer E, Gallagher JT, Krizsa F, et al. 1983. Regulation of haemopoiesis in long-term bone marrow cultures: IV. Glycosaminoglycan synthesis and the stimulation of haemopoiesis by β-D-xylosides. J Cell Biol 96:510. Sporn MB, Roberts AB. 1989. Transforming growth factor-β: Multiple actions and potential clinical applications. JAMA 262:938. Stroncek DF, Holland PV, Bartch G, et al. 1993. Experiences of the first 493 unrelated marrow donors in the national marrow donor program. Blood 81:1940. Sutherland HJ, Eaves AC, Eaves CJ. 1991a. Quantitative assays for human hemopoietic progenitor cells. In AP Gee (ed), Bone Marrow processing and Purging, pp 155–167, Boca Raton, Fla, CRC Press. Sutherland HJ, Eaves CJ, Lansdorp PM, et al. 1991b. Differential regulation of primitive human hematopoietic stem cells in long-term cultures maintained on genetically engineered murine stromal cells. Blood 78:666. Sutherland HJ, Hogge DE, Cook D, et al. 1993. Alternative mechanisms with and without steel factor support primitive human hematopoiesis. Blood 81:1465. © 2000 by CRC Press LLC

Tavassoli M, Hardy CL. 1990. Molecular basis of homing of intravenously transplanted stem cells to the marrow. Blood 76:1059. Toksoz D, Zsebo KM, Smith KA, et al. 1992. Support of human hematopoiesis in long-term bone marrow cultures by murine stromal cells selectively expressing the membrane-bound and secreted forms of the human homolog of the steel gene product, stem cell factor. Proc Natl Acad Sci 89:7350. Traycoff CM, Abboud CM, Abboud MR, Laver J, et al. 1994. Evaluation of the in vitro behavior of phenotyically defined populations of umbilical cord blood hematopoietic progenitor cells. Exp Hematol 22:215. Tushinski RJ, McAlister IB, Williams DE, et al. 1991. The effects of interleukin 7 (IL-7) on human bone marrow in vitro. Exp Hematol 19:749. Ushio S, Namba M, Okura T, Hattori K, Hukada Y, Akita K, Tanabe F, Konishi K, Micallef M, Fujii M, Torigoe K, Tanimoto T, Fukuda S, Ikeda M, Okamura H, Kurimoto M. 1996. Cloning of the cDNA for human IFN-gamma-inducing factor, espression in Escherichia coli, and studies on the biologic activities of the protein. J Immunol 156:4274–4279. Verfaillie CM. 1992. Direct contact between human primitive hematopoietic progenitors and bone marrow stroma is not required for long-term in vitro hematopoiesis. Blood 79:2821. Virelizier JL, Arenzana-Seisdedos F. 1985. Immunological functions of macrophages and their regulation by interferons. Med Biol 63:149–159. Wagner JE, Broxmeyer HE, Byrd RL, et al. 1992. Transplantation of umbilical cord blood after myeloablative therapy: Analysis of engraftment. Blood 79:1874. Weinstein R, Riordan MA, Wenc K, et al. 1989. Dual role of fibronectin in hematopoietic differentiation. Blood 73:111. Wight TN, Kinsella MG, Keating A, et al. 1986. Proteoglycans in human long-term bone marrow cultures: Biochemical and ultrastructural analyses. Blood 67:1333. Wolf SF, Temple PA, Kobayashi M, et al. 1991. Cloning of cDNA for natural killer cell stimulatory factor, a heterodimeric cytokine with multiple biologic effects on T and natural killer cells. J Immunol 146:3074. Yamakazi K, Roberts RA, Spooncer E, et al. 1989. Cellular interactions between 3T3 cells and interleukin-3 dependent multipotent haemopoietic cells: A model system for stromal-cell-mediated haemopoiesis. J Cell Physiol 139:301. Yao Z, Painter SL, Fanslow WC, Ulrich D, Macduff BM, Spriggs MK, and Armitage RJ. 1995. Human IL17: A novel cytokine derived from T cells. J Immunol 155:5483-5486. Yokota T, Arai N, de Vries JE, et al. 1988. Molecular biology of interleukin-4 and interleukin-5 genes and biology of their products that stimulate B cells, T cells and hemopoietic cells. Immunol Rev 102:137. Zandstra PW, Eaves CJ, Piret JM. 1994. Expansion of hematopoietic progenitor cell populations in stirred suspension bioreactors of normal human bone marrow cells. Biotechnology 12:909–914. Zipori D. 1989. Stromal cells from the bone marrow: Evidence for a restrictive role in regulation of hemopoiesis. Eur J Haematol 42:225. Zlotnik A. Moore KW. 1991. Interleukin 10. Cytokine 3:366. Zsebo KM, Wypych J, McNiece IK, et al. 1990. Identification, purification, and biological characterization of hematopoietic stem cell factor from buffalo rat liver-conditioned medium. Cell 63:195. Zuckerman KS, Wicha MS. 1983. Extracellular matrix production by the adherent cells of long-term murine bone marrow cultures. Blood 61:540.

Further Information The American Society of Hematology (ASH) is the premier organization dealing with both the experimental and clinical aspects of hematopoiesis. The society journal, BLOOD, is published twice per month, and can be obtained through ASH, 1200 19th Street NW, Suite 300, Washington, DC 20036 (phone: 202–857–1118). The International Society of Experimental Hematology (ISEH) publishes Experimental Hematology monthly. © 2000 by CRC Press LLC

Tao Ho Kim, Vacanti, J. P. “Tissue Engineering of the Liver.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

121 Tissue Engineering of the Liver Tao Ho Kim Harvard University and Boston Children’s Hospital

121.1 121.2

Hepatocellular Inject Model • Hepatocyte Transplantation on Polymer

Joseph P. Vacanti Harvard University and Boston Children’s Hospital

Background Hepatocyte Transplantation Systems

121.3

Conclusion

Liver transplantation has been established as a curative treatment for end-stage adult and pediatric liver disease [Starlz et al., 1989], and over recent years, many innovative advances have been made in transplantation surgery. Unfortunately, a fundamental problem of liver transplantation has been severe donor shortage, and no clinical therapeutic bridge exists to abate the progression of liver failure (see Table 121.1). As the demand for liver transplantation surgery increases, still fewer than 3500 donors are available annually for the approximately 25,000 patients who die from chronic liver disease [National Vital Statistic System, 1991, 1992]. Currently, cadaveric and living-related donors are the only available sources. Xenograft [Starzl et al., 1993] and split liver transplantation [Merio & Campbell, 1991] are under experimental and clinical evaluation. The research effort to engineer a functional liver tissue has been vigorous, since tissue engineering of the liver offers, in theory, an efficient use of limited organ availability. Whereas other experimental hepatic support systems such as extracorporeal bioreactors and hemoperfusion devices [Yarmusch et al., 1992] attempt to temporarily support the metabolic functions of liver, transplantation of hepatocyte systems is a possible temporary or permanent alternative therapy to liver transplantation for treatment of liver failure. An experimental model system of hepatocellular transplantation should provide optimal cell survival, proliferation, and maintenance of sufficient functional hepatocyte mass to replace liver function. Direct hepatocellular injection or infusion into various organs or tissues and a complex hepatocyte delivery system utilizing polymer matrices have been two major areas of research interest. Just as the liver is one of the most sophisticated organs in the human body, the science of hepatocyte or liver tissue construction has proved to be equally complex.

121.1

Background

The causes of end-stage liver disease are many, including alcoholic or viral cirrhosis, biliary atresia, inborn errors of metabolism, and sclerosing cholangitis. With chronic and progressive liver injury, hepatic necrosis occurs followed by fatty infiltration and inflammation. Scar tissue and nodular regeneration replace the normal liver architecture and increase the microcirculatory resistance, which results in portal hypertension; the liver is further atrophied as important factors in the portal blood which regulate liver growth and maintenance are shunted away from the liver. Currently, end-stage liver disease must be present to be considered for orthotopic liver transplantation therapy, but a significant difference exists


between alcohol- or viral-induced liver injury and TABLE 121.1 UNOS Liver Transplantation Data congenital liver diseases such as isolated gene Summary from 1989 to 1993 in the United States: Total Number of Liver Transplant Candidates and Deaths defects and biliary atresia. With alcohol- or viralReported per Year While on Transplant Waiting List induced chronic hepatic injury, the degree of liver injury is unknown until metabolic functions are Year No. of Patients No. of Deaths Reported severely impaired and signs of progressive irre1989 3096 39 versible hepatic failure including portal hyperten1990 4008 45 sion, coagulopathy, progressive jaundice, and 1991 4866 67 hepatic encephalopathy have developed. In con1992 5785 104 1993 7040 141 genital liver diseases, normal hepatic metabolic functions exist until dangerous toxins build up Source: United Network for Organ Sharing and the and destroy the liver parenchyma. Hepatocellular Organ Procurement and Transplantation Network. Data transplantation could potentially prevent hepatic as of January 12, 1994. injury and preserve host hepatic function for congenital inborn errors of liver metabolism. As liver transplantation emerged as an important therapeutic modality, research activity intensified to improve or understand many areas of liver transplantation such as immunological tolerance, preservation techniques, and the mechanism of healing after acute and chronic liver injury. Liver growth and regulation, in particular, have been better understood: For example, after partial hepatectomy, several mitogens—epidermal growth factor, alpha fibroblastic growth factor, hepatocyte growth factor, and transforming growth factor-alpha—are produced early after injury to stimulate liver regeneration. Comitogens—including insulin, glucagon, estrogen, norepinephrine, and vasopressin—also aid with liver regeneration [Michalopoulos, 1993]. These stimulation factors help govern the intricate regulatory process of liver growth and regeneration, but much about what controls these factors is still unknown. In vitro and in vivo experiments with mitogens such as hepatocyte growth factor, epidermal growth factor, and insulin have yielded only moderate improvement in hepatic proliferation. The importance of hepatocyte proliferation and hepatic regeneration becomes evident when one considers the difficulty in delivering large numbers of hepatocytes in hepatocyte replacement systems. The potential advantage of hepatocyte cellular transplantation is to take a small number of hepatocytes and proliferate these cells, in vitro or in vivo, to create functional liver equivalents for replacement therapy. Without hepatocyte regeneration, delivery of a very large number of hepatocytes is required. Asonuma and coworkers [1992] have determined that approximately 12% of the liver by heterotopic liver transplantation can significantly correct hyperbilirubinemia in the Gunn rat, which is deficient in uridine diphosphate glucuronyl tranferase. Although long-term efficacy remains unclear, 10–12% of the liver is an approximate critical hepatocellular mass thought necessary to replace the metabolic functions of the liver. The inability to mimic normal liver growth and regeneration in in vitro and in vivo systems has been one significant obstacle for hepatocyte tissue construction thus far.

121.2

Hepatocyte Transplantation Systems

Hepatocyte transplant systems offer the possibilities of creating many functional liver equivalents, storing hepatocytes by cryopreservation for later application [Yarmush et al., 1992], and using autologous cells for gene therapy [Jauregui & Gann, 1991]. The two systems discussed below differ in the amount of hepatocytes delivered, use of implantation devices, and implantation sites and techniques. Yet, in both systems, one significant roadblock in proving the efficacy of hepatocyte transplantation has been in the lack of a definitive, reproducible isolated liver defect model. Previous studies using syngeneic rat models such as the jaundice Gunn rat, the analbuminemic Nagase rat, or acute and chronic liver injury rat models have attributed significant correction of their deficit from hepatocyte transplantation. However, either inconsistencies or variation of animal strains and lack of consistent reproducible results have made accurate scientific interpretations and conclusions difficult. For instance, a study assessing hepatocyte delivery with microcarrier beads reported significant decrease in bilirubin in the hyperbilirubinemic © 2000 by CRC Press LLC

Gunn rat and elevation of albumin in the Nagase analbuminemic rat after intraperitoneal implantation [Demetriou et al., 1986]; other studies have not demonstrated significant hepatocyte survival with intraperitoneal injection of hepatocytes with or without microcarriers; histology showed predominant cell necrosis and granuloma formation after 3 days [Henne-Bruns et al., 1991]. In allogeneic models, the possibility of immunological rejection further complicates the evaluation of the hepatocyte transplant system. Clearly, determining the efficacy of hepatocyte transplantation in liver metabolic-deficient models requires significant chemical results, and histologic correlation without confounding variables.

Hepatocellular Injection Model Hepatocytes require an extracellular matrix for growth and differentiation, and the concept of utilizing existing stromal tissue as a vascular extracellular matrix is inviting. Isolated hepatocytes have been injected directly into the spleen or liver or in the portal or splenic vein. Several studies have reported significant but temporary correction of acute and metabolic liver defects in rat models as a result of intrahepatic, intraportal [Matas et al., 1976], or intrasplenic injections [Vroemen et al., 1985]. However, elucidating the efficacy of hepatocellular injection transplantation has been difficult for three significant reasons: (1) Differentiation of donor transplanted hepatocytes from host liver parenchyma has not been well established in an animal liver defect model, (2) how much hepatocyte mass needed to inject for partial or total liver function replacement has not been determined [Onodera et al., 1992], and (3) establishing a definitive animal liver defect model to prove efficacy of hepatocyte injection has been difficult. Transgenic animal strains have been developed and offer a reproducible model in differentiating host from transplanted hepatocytes. Using transgenic mouse lines, donor hepatocytes injected into the spleen were histologically shown to migrate to the host liver, survive, and maintain function [Ponder et al., 1991]. Recently, a transgenic liver model in a mouse was used to evaluate the replicative potential of adult mouse hepatocytes. Normal adult mouse hepatocytes from two established transgenic lines were injected into the spleen of an Alb-uPA transgenic mouse that had an endogenous defect in hepatic growth potential and function (see Fig. 121.1). The adult mouse hepatocytes were shown to translocate to the liver and undergo up to 12 cell doublings; however, function of the transplanted hepatocytes was not fully reported [Rhim et al., 1994]. Rhim’s study suggests that a hepatocyte has the potential to replicate many-fold so long as the structural and chemical milieu is optimal for survival and regeneration. If a small number of transplanted hepatocytes survive and proliferate many-fold in the native liver, sufficient hepatocyte mass to replace liver function would be accomplished, ameliorating the need to deliver a large quantity of donor hepatocytes. Although many questions about hepatocyte injection therapy remain, the transgenic liver model has been used to test the safety and efficacy of ex vivo gene therapy for metabolic liver diseases. The Watanabe heritable hyperlipidemic rabbit, a strain deficient in low density lipoprotiens (LDL) receptors, has been used to evaluate the possible application of a gene therapy by hepatocyte injection. Autologous hepatocytes were obtained from a liver segment, genetically modified ex vivo, and infused into the inferior mesenteric vein through a catheter placed intraoperatively without postoperative sequelae to the rabbit. Decreased levels of LDL have been reported out to 6 months [Wilson et al., 1992]. Larger animal models have shown engraftment of the genetically altered hepatocytes for as long as 1.5 years. The first clinical application of hepatocellular injection has been performed on a patient diagnosed with homozygous familial hypercholesterolemia. The patient has had decreased levels of cholesterol and has maintained expression of the transfected gene after 18 months [Grossman et al., 1994]. This is an important step forward as we await long-term results.

Hepatocyte Transplantation on Polymer Matrices Since the practical application of implanting few hepatocytes to proliferate and replace function is not yet possible, hepatocyte tissue construction, using polymer as a scaffold, relies on transplanting a large number of hepatocytes to allow survival of enough hepatocyte mass to replace function. The large surface © 2000 by CRC Press LLC

FIGURE 121.1 Control and transgenic liver specimens: a nontransgenic (normal mouse) control with normal liver color (top, left); a transgenic (MT-lacZ) liver with normal function which is stained in blue and served as a positive control (top, center); a notransgenic control transplanted with transgenic MIT-lacZ hepatocytes (top, right); and three livers with a different transgene Alb-uPA transplanted with transgenic (MT-lacZ) hepatocytes. A deficiency in hepatic growth potential and function is induced by the Alb-uPA transgene, resulting in a chronic stimulus for liver growth. The liver with Alb-uPA transgene has the same color as the nontransgenic control liver but is partially replaced by the blue-stained transgencic (MT-lacZ) normal functioning hepatocytes, showing the regenerative response of normal hepatocytes in a mouse liver with a chronic stimulus for live growth. (Rhim, Sandgren, and Brinster, University of Pennsylvania, reprinted with permission from American Association for the Advancement of Science.)

area of the polymer accommodates hepatocyte attachment in large numbers so that many cells may survive initially by diffusion of oxygen and other vital nutrients (see Fig. 121.2). The polymer scaffold is constructed with a high porosity to allow vascular ingrowth, and vascularization of surviving cells then can provide permanent nutritional access [Cima et al., 1991]. The cell-polymer system has been used in several other tissue-engineering applications such as cartilage, bone, intestine, and urologic tissue construction [Langer & Vacanti, 1993]. Hepatocytes adhere to the polymer matrix for growth and differentiation as well as locate into the interstices of the polymer. The polymer-hepatocyte interface can be manipulated with surface proteins such as laminin, fibronectin, and growth factors to improve adherence, viability, function, or growth [Mooney et al., 1992]. Hepatocyte proliferation by attaching mitogenic factors like hepatocyte growth factor, epidermal growth factor, or transforming growth factor-alpha is under current investigation. Synthetic polymer matrices, both degradable and nondegradable, have been evaluated for hepatocytepolymer construction. As degradable polymers were being studied and evaluated for tissue engineering, a nondegradable polymer, polyvinyl (PVA), was used for in vitro and in vivo systems. The polyvinyl alcohol sponge offered one significant advantage: a uniform, noncollapsible structure which allowed quantification of hepatocyte engraftment [Uyama et al., 1993]. In vitro and in vivo studies have demonstrated hepatocyte survival on the polyvinyl alcohol scaffold. However, the polyvinyl alcohol sponge will not degrade and could act as a nidus for infection and chronic inflammation. A degradable polymer conceptually serves as a better implantable scaffold for hepatocyte-polymer transplantation, since the polymer dissolves to leave only tissue. Polyglycolic acid (PGA), polyactic acid (PLA), and copolymer hybrids have been employed in several animal models of liver insufficiency (see Fig. 121.3). Histologic analyses have shown similar survival of hepatocytes when compared to studies with polyvinyl alcohol. The experimental design for the hepatocyte-polymer model has been standardized as follows: (1) an endto-side portacaval shunt is performed to provide hepatotrophic stimulation to the graft, (2) hepatocytes


FIGURE 121.2

Hepatocytes are seen adhering to polyglycolic acid polymer in culture (Mooney, unpublished data).

are isolated and seeded onto degradable polyglycolic acid polymer, (3) the hepatocyte-polymer construct is implanted into the abdominal cavity on vascular beds of small intestinal mesentery and omentum, (4) pertinent chemical studies are obtained and analyzed at periodic intervals, and (5) histologic analysis in the specimens is performed at progressive time points. Early studies have shown that a large percentage of hepatocytes perish from hypoxia within 6–24 hours after implantation. To improve hepatocyte survival, implantation of hepatocytes onto large vascular surface areas for engraftment and exposing the hepatocytes to hepatotrophic factors were two important maneuvers. The small intestinal mesentery and omentum have offered large vascular surface areas. Hepatocyte survival has been reported at other sites such as the peritoneum [Demetriou et al., 1991], renal capsule [Ricordi et al., 1989], lung [Sandbichler et al., 1992], and pancreas [Vroemen et al., 1988], but these sites do not provide enough vascular surface area to allow survival of a large number of hepatocytes. The concept of hepatotrophic factors originated when atrophy and liver insufficiency was observed with heterotopic liver transplantation. Later studies have confirmed that important factors regulating liver growth and maintenance existed in the portal blood [Jaffe et al., 1991]. Thus, when a portacaval shunt to redirect hepatotrophic factors from the host liver to the hepatocyte-polymer construct was performed, survival of heterotopically transplanted hepatocytes significantly improved [Uyama et al., 1993]; consequently, portacaval shunts were instituted in all experimental models. Studies thus far have shown survival of functional hepatocytes over 6 months in rat and dog models (unpublished data); replacement of liver function has been of shorter duration. A study using the Dalmatian dog model of hyperuricosuria typifies the current status of the hepatocytepolymer system. The hepatocyte membrane of the Dalmatian dog has a defect in the uptake of uric acid, which results in hyperuricemia and hyperuricosuria [Giesecke & Tiemeyer, 1984]. The study has shown a significant but temporary correction of the liver uric acid metabolic defect after implantation of normal beagle hepatocytes, which have normal uric acid metabolism, on degradable polyglycolic acid polymer [Takeda et al., 1994]. Cyclosporine was administered for immunosuppression. The results of the Dalmatian dog study suggest that (1) successful engraftment of the hepatocyte-polymer construct occurred,


FIGURE 121.3 Scanning electron microscopic photographs of polyglyclic acid (top), polyvinyl alcohol (middle), and polylactic acid (bottom) demonstrate the high porosity of these polymers (Mooney, reprinted with permission from W.B. Saunders Company).

(2) maintenance of functional hepatocytes under current conditions was temporary, and (3) loss of critical hepatocyte mass to replace the liver uric acid metabolism defect occurred at 5–6 weeks after hepatocytepolymer transplantation. The temporary effect could be attributed to suboptimal immunosuppression, suboptimal regulation of growth factors and hormones involved with hepatic growth, regeneration, and maintenance, or both. Coculture of hepatocytes with pancreatic islet cells also has been under investigation to aid in hepatocyte survival, growth, and maintenance. Trophic factors from islet cells have been shown to improve hepatocyte survival [Ricordi et al., 1988], and cotransplantation of hepatocytes with islet cells on a polymer matrix has been shown to improve hepatocyte survival as well [Kaufmann et al., 1994]. Coculture


with other cell types such as the biliary epithelial cell also may improve hepatocyte survival. Other studies with biliary epithelial cells have shown ductular formation in in vitro and in vivo models [Sirica et al., 1990], and vestiges of ductular formation in hepatocyte-polymer tissue have been histologically observed [Hansen & Vacanti, 1992]. A distinct advantage of the cell-polymer engineered construct is that one can manipulate the polymer to direct function. Thus far, diseases involving the biliary system, primarily biliary atresia, are not amenable to hepatocyte transplantation. An attempt to develop a biliary drainage system with biliary epithelial cells, hepatocytes, and polymer has been initiated. In the future, the potential construction of a branching polymer network could serve as the structural cues for the development of an interconnecting ductular system. Ex vivo gene therapy with the hepatocyte-polymer system is also an exciting potential application as demonstrated by the recent clinical trial with hepatocyte injection therapy. Genetically altered hepatocytes transplanted on polymer constructs have been studied with encouraging results [Fontaine et al., 1993].

121.3

Conclusion

Studies in hepatocyte transplantation through tissue engineering methods have made important advances in recent years. The research in the hepatocellular injection and the research in the hepatocyte-polymer construct models have complemented each other in understanding the difficulties as well as the possibilities of liver replacement therapy. In order to make further advances with hepatocyte replacement systems, the process of liver development, growth, and maintenance need to be better understood. Currently, the amount of hepatocyte engraftment, proliferation, and the duration of hepatocyte survival remain undetermined in both systems. The amount of functional hepatocyte engraftment necessary may vary for different hepatic diseases. For instance, isolated gene defects of the liver may require a small number of functional transplanted hepatocytes to replace function, whereas end-stage liver disease may require a large amount of hepatocyte engraftment. Current hepatic replacement models have both advantages and disadvantages. For hepatocellular injection, the application of ex vivo gene therapy for an isolated gene defect of liver metabolism is promising. However, the small amount of hepatocyte delivery and significant potential complications for patients with portal hypertension may preclude application of the hepatocellular injection method for end-stage liver disease. A significant amount of intrapulmonary shunting of hepatocytes was observed in rats with portal hypertension after intrasplenic injection of hepatocytes, which resulted in increased portal pressures, pulmonary hypertension, pulmonary infarction, and reduced pulmonary compliance [Gupta et al., 1993]. With the hepatocyte-polymer system, delivery of a large number of hepatocytes is possible. In patients with portal hypertension, portal blood-containing hepatotrophic factors are shunted away from the liver, obviating the need for a portacaval shunt. Thus, patients with end-stage liver disease and portal hypertension may need only transplantation of the hepatocyte-polymer construct. However, an end-to-side portacaval shunt operation is needed in congenital liver diseases with normal portal pressures to deliver hepatotrophic factors to the heterotopically place hepatocytes. In the future, each hepatocyte transplant system could have specific and different clinical applications. More important, both offer the hope of increasing therapeutic options for patients requiring liver replacement therapy. Approximately 260,000 patients out of 634,000 patients hospitalized for liver diseases have liver diseases which could have been considered for hepatic transplantation. The total acute care nonfederal hospital cost for liver diseases in 1992, which does not include equally substantial outpatient costs, exceeded $9.2 billion [HCIA Inc., 1992].

Defining Terms Allogeneic: Pertaining to different genetic compositions within the same species. Cadaveric: Related to a dead body. In transplantation, cadaveric is related to a person who has been declared brain dead; organs should be removed prior to cardiac arrest to prevent injury to the organs.


Heterotopic: Related to a region or place where an organ or tissue is not present in normal conditions. Mitogens: Substances that stimulate mitosis or growth. Orthotopic: Related to a region or place where an organ or tissue is present in normal conditions. Portacaval shunt: A surgical procedure to partially or completely anastomose the portal vein to the inferior vena cava to divert portal blood flow from the liver to the systemic circulation. Stroma: The structure or framework of an organ or gland usually composed of connective tissue. Transgenic: Referred to introduction of a foreign gene into a recipient which can be used to identify genetic elements and examine gene expression. Xengraft: A graft transferred from one animal species to another species.

References Asonuma K, Gilber JC, Stein JE, et al. 1992. Quantitation of transplanted hepatic mass necessary to cure the Gunn rat model of hyperbilirubinemia. J Ped Surg 27(30):298. Asonuma K, Vacanti JP. 1992. Cell transplantation as replacement therapy for the future. Pediatr Transplantation 4(2):249. Cima L, Vacanti JP, Vacant C, et al. 1991. Tissue engineering by cell transplantation using degradable polymer substrates. J Biomech Eng 113:143. Demetriou AA, Felcher A, Moscioni AD. 1991. Hepatocyte transplantation. A potential treatment for liver disease. Dig Dis Sci 12(9):1320. Demetriou AA, Whiting JF, Feldman D, et al. 1986. Replacement of liver function in rats by transplantation of microcarrier-attached hepatocytes. Science 233:1190. Fontaine MJ, Hansen, LK, Thompson S, et al. 1993. Transplantation of genetically altered hepatocytes using cell-polymer constructs leads to sustained human growth hormone secretion in vivo. Transplant Proc 25(1):1002–4. Giesecke D, Tiemeyer W. 1984. Defect of uric acid in Dalmatian dog liver. Experientia 40:1415. Grossman M, Roper SE, Kozarsky K, et al. 1994. Successful ex vivo gene therapy directed to liver in a patient with familial hypercholesterolemia. Nature Genetics 6:335. Gupta S, Yereni PR, Vemuru RP, et al. 1993. Studies on the safety of intrasplenic hepatocyte transplantation: relevance to ex vivo gene therapy and liver repopulation in acute hepatic failure. Hum Gene Ther 4(3):249. Hansen LK, Vacanti JP. 1992. Hepatocyte transplantation using artificial biodegradable polymers. 1992. In MA Hoffman (ed), Current Controversies in Biliary Atresia. The Medical Intelligence Unit Series pp 96–106, (CRC Press), Austin, Tex, R.G. Landes. HCIA Inc. 1992. Survey of costs in non-Federal, acute care hospitals in the United States prepared for the American Liver Foundation, Baltimore. Henne-Bruns D, Kruger U, Sumpelman D, et al. 1991. Intraperitoneal hepatocyte transplantation: morphological results. Virchows Arch A, Path Anat Histopathol 419(1):45. Jaffe V, Darby H, Bishop A, et al. 1991. The growth of liver cells in the pancreas after intrasplenic implantation: the effects of portal perfusion. Int J Exp Pathol 72(3):289. Jauregui HO, Gann KL. 1991. Mammalian hepatocytes as a foundation for treatment in human liver failure [Review]. J Cell Biochem 45(4):359. Kaufmann P-M, Sano K, Uyama S, et al. 1994. Heterotopic hepatocyte transplantation using three dimensional polymers. Evaluation of the stimulatory effects by portacaval shunt or islet cell co-transplantation. Second International Congress of the Cell Transplant Society, Minneapolis, Minnesota. Langer R, Vacanti J. 1993. Tissue Engineering Science 260:920. Matas AJ, Sutherland DER, Steffes MW, et al. 1976. Hepatocellular transplantation for metabolic deficiencies: decrease of plasma bilirubin in Gunn rats. Science 192:892. Merion RM, Campbell DA Jr. 1991. Split liver transplantation: One plus one doesn’t always equal two. Hepatology 14(3):572. Michalopoulos G. 1993. HGF and liver regeneration. Gasterologica Japonica 28(suppl 4):36.


Mooney DJ, Hansen LK, Vacanti JP, et al. 1992. Switching from differentiation to growth in hepatocytes: Control by extracellular matrix. J Cell Physiol 151:497. National Vital Statistic System. 1991, 1992. Data derived from National Center for Health Statistics. Onodera K, Ebata H, Sawa M, et al. 1992. Comparative effects of hepatocellular transplantation in the spleen, portal vein, or peritoneal cavity in congenitally ascorbic acid biosynthetic enzyme-deficient rats. Transplant Proc 24(6):3006. Ponder KP, Gupta S, Leland F, et al. 1991. Mouse hepatocytes migrate to liver parenchyma and function indefinitely after intrasplenic transplantation. Proc Natl Acad Sci USA 88(4):1217. Rhim JA, Sandgren EP, Degen JL, et al. 1994. Replacement of diseased mouse liver by hepatic cell transplantation. Science 263:1149. Ricordi C, Lacy PE, Callery MP, et al. 1989. Trophic factors from pancreatic islets in combined hepatocyteislet allografts enhance hepatocellular survival. Surgery 105:218. Sandbichler P, Then P, Vogel W, et al. 1992. Hepatocellular transplantation into the lung for temporary support of acute liver failure in the rat. Gastroenterology 102(2):605. Sirica AE, Mathis GA, Sano N, et al. 1990. Isolation, culture, and transplantation of intrahepatic biliary epithelial cells and oval cells. Pathobiology 58:44. Starzl TE, Demetris AJ, Van Thiel D. 1989. Chronic liver failure: Orthotopic liver transplantation. N Eng J Med 321:1014. Takeda T, Kim TH, Lee SK, et al. 1994. Hepatocyte transplantation in biodegradable polymer scaffolds using the Dalmatian dog model of hyperuricosuria. Fifteenth Congress of the Transplantation Society, Kyoto Japan. Submitted. Uyama S, Takeda T, Vacanti JP. 1993. Delivery of whole liver equivalent hepatic mass using polymer devices and hetertrophic stimulation. Transplantation 55(4):932. Uyama S, Takeda T, Vacanti JP. In press. Hepatocyte transplantation equivalent to whole liver mass using cell-polymer devices. Polymer Preprints. Vacanti JP, Morse MA, Saltzman WM, et al. 1988. Selective cell transplantation using bioabsorbable artificial polymers as matrices. J Ped Surg 23(1):3. Vroemen JPAM, Blanckaert N, Buurman WA, et al. 1985. Treatment of enzyme deficiency by hepatocyte transplantation in rats. J Surg Res 39:267. Vroemen JPAM, Buurman WA, van der Linden CJ, et al. 1988. Transplantation of isolated hepatocytes into the pancreas. Eur Surg Res 20:1. Wilson JM, Grossman M, Raper SE, et al. 1992. Ex vivo gene therapy of familial hypercholesterolemia. Human Gene Ther 3(2):179. Yarmush ML, Toner M, Dunn JCY, et al. 1992. Hepatic tissue engineering: Development of critical technologies. Ann NY Acad Sci 665:238.


Bellamkonda, R., Aebischer, P. “Tissue Engineering in the Nervous System.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

122 Tissue Engineering in the Nervous System 122.1

Delivery of Neuroactive Molecules to the Nervous System Pumps • Slow-Release Polymer-Release Polymer Systems • Cell Transplantation

122.2

The Active Use of Channel Properties • Intraluminal Matrices for Optimal Organization of Regeneration Microenvironment • Cell-Seeded Lumens for Trophic Support • Polyethyleneglycol-Induced Axon Fusion • CNS Nerve Regeneration

Ravi Bellamkonda Lausanne University Medical School

Patrick Aebischer Lausanne University Medical School

Tissue Reconstruction: Nerve Regeneration

122.3 122.4

In Vitro Neural Circuits and Biosensors Conclusion

Tissue engineering in the nervous system facilitates the controlled application and/or organization of neural cells to perform appropriate diagnostic, palliative, or therapeutic tasks. As the word tissue implies, tissue engineering in general involves cellular components and their organization. Any cell, given its broad genetic program, expresses a particular phenotype in a manner that is dependent on its environment. The extracellular environment consists of cells, humoral factors, and the extracellular matrix. Research in genetic engineering and the intense focus on growth factors and extracellular matrix biology have made it possible to manipulate both the cell’s genetic program and its phenotypic expression. In the nervous system, degeneration or injury to neurons or glia and/or an aberrant extracellular environment can cause a wide variety of ailments. Diseases such as Parkinson’s may required the replacement of diminished levels of a particular neurochemical, e.g., dopamine. Other pathologies such as injured nerves or reconnection of served neural pathways may require regeneration of nervous tissue. Tissue engineering efforts in the nervous system have currently been addressing the following goals: 1. Functional replacement of a missing neuroactive component 2. Rescue or regeneration of damaged neural tissue 3. Human-machine interfaces: neural coupling elements

122.1 Delivery of Neuroactive Molecules to the Nervous System Deficiency of specific neuroactive molecules has been implicated in several neurologic disorders. These factors may be neurotransmitters, neurotrophic agents, or enzymes. For example, part of the basal ganglia circuitry that plays a role in motor control consists of striatal neurons receiving dopaminergic input from


Table 122.1

Engineering Solutions for Parkinson’s Disease

Mode of Delivery Infusion

Rodents

Monkey

Human

Rat striatum Cerebroventricular Systemic

Polymer slow release

Cell transplantation Fetal Substantia nigra Ventral mesencephalon Human DA neurons Autologous primary Genetically altered skin fibroblasts Genetically altered skin fibroblasts with myoblasts Encapsulated xenogeneic tissue Bovine adrenal chromaffin cells PC12 cells

Rat striatum (EVA rods) Rat subcutaneous (EVA rods) Rat striatum (Si pellet) Rat striatum (liposomes)

Rat striatum Rat striatum Putamen

Hargraves et al., 1987 De Yebenes et al., 1988 Hardie et al., 1984 Winn et al., 1989 Sabel et al., 1990 Becker et al., 1990 During et al., 1992

Björklund et al., 1980 Freund et al., 1985 Bolam et al., 1987 Lindvall et al., 1990

Rat striatum

Chen et al., 1991

Rat striatum

Jiao et al., 1993

Striatum (microcapsules) Rat striatum (microcapsules) Rat striatum (macrocapsule)

Aebischer et al., 1991 Winn et al., 1991 Aebischer et al., 1988 Aebischer et al., 1994

Striatum (macrocapsule) Mouse mesencephalon

Reference

Rat parietal cortex (macrocapsule)

Aebischer et al., 1988

the mesencephalic substantia nigra neurons. It has been shown that a lesioned nigrostriatal dopaminergic pathway is responsible for Parkinson’s disease [Ehringer & Hornykiewicz, 1960]. In chronic cancer patients, the delivery of antinociceptive neurotransmitters such as enkephalins, endorphins, catecholamines, neuropeptide Y, neurotensin, and somatostatin to the cerebrospinal fluid may improve the treatment of severe pain [Akil et al., 1984; Joseph et al., in press]. Neurotrophic factors may play a role in the treatment of several neurodegenerative disorders. For example, local delivery of nerve growth factor (NGF) [Hefti et al., 1984; Williams et al., 1986] and/or brain-derived growth factor (BDNF) may be useful in the treatment of Alzheimer’s disease [Anderson et al., 1990, Knüsel et al., 1991]. BDNF [Hyman et al., 1991; Knüsel et al., 1991] and glial cell line-derived nerve growth factor (GDNF) may be beneficial in Parkinson’s’ disease [Lin et al., 1993]. Other neurotrophins such as ciliary neurotrophic factor (CNTF) [Oppenheim et al., 1991]; Sendtner et al., 1990], BDNF [Yan et al., 1992], neurotrophin-3 (NT-3), neurotrophin-4 (NT-4/5) [Hughes et al., 19993] and GDNF [Zurn et al., 1994] also could have an impact on amyotrophic lateral sclerosis (ALS) or Lou Gehrigs’s disease. Therefore, augmentation or replacement of any of the above-mentioned factors in the nervous system would be a viable therapeutic strategy for the treatment of the pathologies listed above. There are several issues that ought to be considered in engineering a system to deliver these factors. The stability of the factors, the dosage required, the solubility, the target tissue, and possible side effects all are factors that influence the choice of the delivery mode. Pumps, slow-release polymer systems, and cells from various sources that secrete the compound of interest are the main modes by which these factors can be delivered. Table 122.1 lists all the means employed to deliver dopamine to alleviate the symptoms of Parkinson’s disease.

Pumps Pumps are used to deliver opiates epidurally to relieve severe pain [Ahlgren et al., 1987]. Pumps also have been used to deliver neurotrophic factors such as NGF intraventricularly as a prelude or supplement


to transplantation of chromaffin cells for Parkinson’s disease [Olson et al., 1991] or as a potential therapy for Alzheimer’s disease [Olson et al., 1992]. Pumps also have been used to experimentally deliver dopamine or dopamine receptor agonists in Parkinson’s disease models [DeYebenes et al., 1988; Hargraves et al., 1987]. While pumps have been employed successfully in these instances, they may need to be refilled every 4 weeks, and this may be a limitation. Other potential drawbacks include susceptibility to “dumping” of neuroactive substance due to presence of a large reservoir of neuroactive element, infections, and diffusion limitations. Also, some factors such as dopamine and ciliary neurotrophic factor (CNTF) are very unstable chemically and have short half-life periods, rendering their delivery by such devices difficult. Some of these problems may be eliminated by well designed slow-release polymer systems.

Slow-Release Polymer Systems Slow-release polymer systems essentially trap the molecule of interest in a polymer matrix and release it slowly by diffusion over a period of time. Proper design of the shape and composition of the polymer matrix may achieve stabilization of the bioactive molecule and facilitate a steady, sustained release over a period of time. For instance, it has been shown that the dopamine precursor L- dopa may be effective in alleviating some motor symptoms of Parkinson’s disease [Birkmayer, 1969], fluctuations in the plasma levels of L-dopa due to traditional, periodic oral administration and difficulties in converting L-dopa to dopamine may cause the clinical response to fluctuate as well [Albani et al., 1986; Hardie et al., 1984; Muenter et al., 1977; Shoulson et al., 1975]. It has been demonstrated that a slow-release ethylene vinyl acetate polymer system loaded with L-dopa can sustain elevated plasma levels of L-dopa for at least 225 days when implanted subcutaneously in rats [Sabel et al., 1990]. Dopamine can only be released directly in the CNS, since it does not pass the blood-brain barrier. It has been demonstrated that experimental parkinsonism in rats can be alleviated by intrastriatal implantation of dopamine-releasing ethylene vinyl acetate rods [Winn et al., 1989]. Silicone elastomer as well as resorbable polyester pellets loaded with dopamine also have been implanted intrastriatally and shown to induce a behavioral recovery in parkinsonian rats [Becker et al., 1990; McRae et al., 1991]. Slow-release systems also may be employed to deliver trophic factors to the brain either to avoid side affects that may come about due to systemic administration or to overcome the blood-brain barrier, getting the factor delivered to the brain directly [During et al., 1992; Hoffman et al, 1990]. Therefore, polymeric systems with the molecule of interest trapped inside may be able to achieve many of the goals of an ideal delivery system, including targeted local delivery and zero-order continuos release [Langer, 1981; Langer et al., 1976]. However, some of the disadvantages of this system are the finite amounts of loaded neuroactive molecules and difficulties in shutting off release or adjusting rate of release once the polymer is implanted. Also, for the long-term release in humans, the device size may become a limiting factor. Some of these limitations may be overcome by the transplantation of cells that release neuroactive factors to the target site.

Cell Transplantation Advances in molecular biology and gene transfer techniques have given rise to a rich array of cellular sources, which have been engineered to secrete a wide range of neurologic compounds. These include cells that release neurotransmitters, neurotrophic factors, and enzymes. Transplantation of these cells leads to functional replacement or augmentation of the original source of these compounds in the host. They can deliver neuroactive molecules as long as they survive, provided they maintain their phenotype and/or transgene expression, in the case of gene therapy. Some of the disadvantages of the slow-release systems, such as the presence of a large reservoir for long-term release or reloading of an exhausted reservoir, can thus be overcome. Some of the tissues transplanted so far may be classified in the following manner. Transplantation of Autologous Primary Cells This technique involves the procurement of primary cells from the host, expanding them if necessary to generate requisite amounts of tissue, engineering them if so required by using gene transfer techniques,


and then transplanting them back into the “donor” at the appropriate site. For instance, autologous Schwann cell shave been isolated and transplanted experimentally into the brain and have been shown to enhance retinal nerve regeneration, presumably by the release of factors that influence regeneration [Morrissey et al., 1991]. Schwann cells also express various neurologically relevant molecules [Bunge & Bunge, 1983; Muir et al., 1989]. Autologous Schwann cells also can work as nerve bridges to help reconstruction of rat sciatic nerve after axotomy [Guénard et al., 1992]. Primary skin fibroblasts also have been genetically engineered to secrete L-dopa and transplanted successfully into the autologous host’s striatum in an experimental model of Parkinson’s disease in rats [Chen et al., 1991]. The same group also has reported that nerve growth factor, tyrosine hydroxylase, glutamic acid decarboxylase, and choline acetyltransferase genes may be introduced successfully and expressed in primary fibroblasts [Gage et al., 1991]. More recently, muscle cells have been engineered to express tyrosine hydroxylase and transplanted successfully in a rat model of Parkinson’s disease [Jiao et al., 1993]. Thus nonneural cells such as a fibroblast or muscle cells may be selected, in part, for the ease with which they can be engineered genetically and made neurologically relevant. It is therefore possible to engineer cells to suit particular pathologies as a step toward being able to engineer biomimetic tissues and then place them in appropriate locations and contexts in vivo. However, it may not always be technically possible to procure sufficient amounts of autologous tissue. Other sources of tissues have therefore been explored, and these include fetal tissue, used usually in conjunction with immunosuppression. Fetal Tissue Transplantation One of the important advantages of fetal tissue is its ability to survive and integrate into the host adult brain. Transplantation of fetal neural tissue allografts might be useful in treating Parkinson’s disease [Lindvall et al., 1990]. While promising results have been reported using neural fetal tissue disease [Björklund, 1991; Björklund et al., 1980], availability of donor tissue and potential ethical issues involved with the technique may be potential shortcomings. One promising strategy to obtain allogeneic fetal tissue in large quantities is to isolate neural stem cells and have them proliferate in vitro. It has been demonstrated recently that CNS progenitor cells may be selected using epidermal growth factor (EGF). The cells thus selected have been shown to proliferate in vitro under appropriate culture conditions. With time and appropriate culture conditions, they differentiate into mature CNS neurons and glia [Reynolds et al., 1992; Vescovi et al., 1993]. When optimized, this technique could be useful to select and expand fetal neurons of interest in vitro and then transplant them. Transplantation of xenogeneic fetal tissue with immunosuppression using cyclosporin A is another alternative to using fetal tissue [Wictorin et al., 1992], and this approach might yield more abundant amounts of tissue for transplantation and overcome some of the possible ethical dilemmas in using human fetal tissue. However, immunosuppression may not be sufficient to prevent long-term rejection and may have other undesirable side-effects. Transplantation of Encapsulated Xenogeneic Tissue Polymeric encapsulation of xenogeneic cells might be a viable strategy to transplant cells across species in the absence of systemic immunosuppression [Aebischer et al., 1988; Tresco et al., 1992]. Typically, the capsules have pores large enough for nutrients to reach the transplanted tissue and let the neuroactive factors out, but the pores are too small to let the molecules and the cells of the immune system reach the transplanted tissue (Fig. 122.1). At the same time, this strategy retains all the advantages of using cells as controlled, local manufacturers of the neuroactive molecules. The use of encapsulation also eliminates the restriction of having to use postmitotic tissue for transplantation to avoid tumor formation. The physical restriction of the polymeric capsule prevents escape of the encapsulated tissue. Should the capsule break, the transplanted cells are rejected and eliminated by the host immune system. However, no integration of the transplanted tissue into the host is possible with this technique. Some of the tissue engineering issues involved in optimizing the encapsulation technique are (1) the type and configuration of the encapsulating membrane, (2) the various cells to be used for encapsulation, and (3) the matrix in which the cells are immobilized.


FIGURE 122.1 (a)Schematic illustration of the concept of immunoisolation involved in the transplantation of xenogenic tissue encapsulated in semipermeable polymer membranes. (b) Light micrograph of a longitudinal section shown baby hamster kidney cells encapsulated in a polyacryolonitrile-vinylchloride membrane and transplanted in an axotomized rat pup after 2 weeks in vivo. W denotes the capsule’s polymeric wall, and Y shows the cells.

Type and Configuration of the Encapsulating Membrane. One important factor determining the size of the device is oxygen diffusion and availability to the encapsulated cells. This consideration influences the device design and encourages situations where the distances between the oxygen source, usually a capillary, and the inner core of transplanted tissue are kept as minimal as possible. The size and configuration of the device also influence the kinetics of release of the neuroactive molecules, the “response time” being slower in larger capsules with thicker membrane walls.


The capsule membrane may be a water-soluble system stabilized by ionic or hydrogen bonds formed between two weak polyelectrolytes—typically an acidic polysaccharide, such as alginic acid or modified cellulose, and a cationic polyaminoacid, such as polylysine or polyornithine [Goosen et al., 1985; Lim & Sun, 1980; Winn et al., 1991]. Gelation of the charged polyelectrolytes is caused by ionic cross-linking in the presence of di- or multivalient counterions. However, the stability and mechanical strength of these systems are questionable in the physiologic ionic environments. The major advantage of using these systems is that they obviate the need of organic solvent use in the making of the capsules and might be less cytotoxic in the manufacturing process. The capsule membrane also may be a thermoplastic, yielding a more mechanically and chemically stable membrane. This technique involves loading of cells of interest in a preformed hollow fiber and then sealing the ends either by heat or with an appropriate glue. The hollow fibers are typically fabricated by a dry jet wet-spinning technique involving a phase-inversion process [Aebischer et al., 1991]. The use of thermoplastic membranes allows the manipulation of membrane structure, porosity, thickness, and permeability by appropriate variation of polymer solution flow rates, viscosity of the polymer solution, the nonsolvent used etc. Long-term cross-species transplants of dopaminergic xenogeneic tissues and functional efficacy in the brain have been reported [Aebischer et al., 1991, 1994] using the preceding system. PC12 cells, a catecholaminergic cell line derived from a rat pheochromocytoma, ameliorated experimental Parkinson’s disease when encapsulated within thermoplastic PAN/PVC capsules and implanted in the striatum of adult guinea pigs [Aebischer et al., 1991]. The Choice of Cells and Tissues for Encapsulation. Three main types of cells can be used for encapsulation. Cells may be postmitotic, cell lines that differentiate under specific conditions, or slow-dividing cells. The latter two types of cells lend themselves to genetic manipulation. Postmitotic Cells. Primary xenogeneic tissue may be encapsulated and transplanted across species. For instance, chromaffin cells release various antinociceptive substances such as enkephalins, catecholamines, and somatostatin. Allografts of adrenal chromaffin cells have been shown to alleviate pain when transplanted in the subarachnoid space in rodent models and terminal cancer patients [Sagen et al., 1993]. Transplantation of encapsulated xenogeneic chromaffin cells may provide a long-term source of painreducing neuroactive substances [Sagen et al., 1993]. In our laboratory, clinical trials are currently underway using the transplantation of encapsulated bovine chromaffin tissue into human cerebrospinal fluid as a strategy to alleviate chronic pain in terminal cancer patients [Aebischer et al., 1994]. This may circumvent the problem of the limited availability of human adrenal tissue in grafting procedures of chromaffin tissue. Cell Lines that Differentiate under Specific Conditions. Postmitotic cells are attractive for transplantation applications because the possibility of tumor formation, loss of phenotypic expression, is potentially lower. Also, when encapsulated, there is less debris accumulation inside the capsule due to turnover of dividing cells. However, the disadvantage is that the amount of postmitotic tissue is usually too limited in quantity for general clinical applications. Therefore, the use of cells that are mitotic and then are rendered postmitotic under specific conditions is attractive for transplantation applications. Under appropriate culture conditions, primary myoblasts undergo cell division for at least 50 passages. Fusion into resting myoblasts can be obtained by controlling the culture conditions. Alternatively, a transformed myoblast cell line C2C12 derived from a C2H mouse thigh differentiates and forms myotubes when cultured under low serum conditions [Yaffe & Saxel, 1977]. Therefore, these cells could be genetically altered, expanded in vitro and then made postmitotic by varying the culture conditions. They can then be transplanted with the attendant advantages of using postmitotic tissue. Since myoblasts can be altered genetically, they have the potential to be a rich source for augmentation of tissue function via cell transplantation. Slow-Dividing Cells. Slow-dividing cells that reach a steady state in a capsule either due to contact inhibition or due to the encapsulation matrix are an attractive source of cells for transplantation. Potentially, their division can result in a self-renewing supply of cells inside the capsule. Dividing cells


are also easier to transfect reliably with retroviral methods and therefore lend themselves to genetic manipulation. It is therefore possible to envisage the transplantation of various genetically engineered cells for the treatment of several neurologic disorders. This technique allows access to an ever-expanding source of xenogeneic tissues that have been engineered to produce the required factor of interest. For instance, Horellou and Mallet [1989] have retrovirally transferred the human TH cDNA into mouse anterior pituitary AtT-20 cell line, potentially resulting in a plentiful supply of dopamine-secreting cells that can then be encapsulated and transplanted into the striatum. Another promising area for the use of such techniques is in the treatment of some neurodegenerative disorders where a lack of neurotrophic factors is believed to be part of the pathophysiology. Neurotrophic factors are soluble proteins that are required for the survival of neurons. These factors often exert a trophic effect; i.e., they have the capability of attracting growing axons. The “target” hypothesis describes the dependence of connected neurons on a trophic factor that is retrogradely transported along the axons after release from the target neurons. In the absence of the trophic factors, the neurons shrink and die, presumably to avoid potential misconnections. Experimentally, fibroblast lines have been used in CNS transplantation studies because of their convenience for gene transfer techniques [Gage et al., 1991]. The transplantation of encapsulated genetically engineered fibroblasts to produce NGF has been shown to prevent lesion-induced loss of septal ChaT expression following a fimbria-fornix lesion [Hoffman et al., 1993]. The fimbria-fornix lesion is characterized by deficits in learning and memory resembling those of Alzheimer’s disease. Matrices for Encapsulation. The physical, chemical, and biologic properties of the matrix in which the cells have been immobilized may play an important role in determining the transplanted cell’s state and function. Broadly, matrices can be classified into the following types: cross-linked polyelectrolytes, collagen in solution or as porous beads, naturally occurring extracellular matrix derivatives such as Matrigel, fibrin clots, and biosynthetic hydrogels with appropriate biologic cues bound to them to elicit a specific response from the cells of interest. The matrix has several functions: It can prevent the formation of large cell aggregates that lead to the development of central necrosis as a consequence of insufficient oxygen and nutrient access; it may allow anchorage-dependent cells to attach and spread on the matrix substrate; and it may induce differentiation of a cell line and therefore slow down or stop its division rate. Negatively charged polyeletrolytes such as alginate have been used successfully for the immobilization of adrenal chromaffin cells [Aebischer et al., 1991]. Positively charged substrates, such as those provided by the amine groups of chitosan, allow attachment and spreading of fibroblasts [Zielinski et al., in press]. Biologically derived Matrigel induces differentiation of various cell lines such as Chinese hamster ovary (CHO) cells, astrocyte lines, or fibroblast lines (unpublished observations). Spongy collagen matrices, as well as fibrin matrices, seem to possess similar qualities. Our laboratory is currently evaluating the use of biosynthetic hydrogel matrices with biologically relevant peptides covalently bound to the polymer backbone. It is hypothesized that these matrices may elicit a specific designed response from the encapsulated cells.

122.2 Tissue Reconstruction: Nerve Regeneration Most of the techniques described above were attempts at identifying the missing molecules of various neuropathologies, finding an appropriate source for these molecules and, if necessary, designing a cellular source via genetic engineering, and ultimately, choosing an optimal mode of delivery of the molecules, be it chemical or cellular. This approach, however, falls short of replacing the physical neuroanatomic synaptic circuits in the brain, which, in turn, may play an important role in the physiologic feedback regulating mechanisms of the system. Attempts to duplicate in vivo predisease neuronal structure have been made. For instance, the bridging of the nigrostriatal pathway, which when disrupted may cause Parkinson’s disease, and the septohippocampal pathway, which may serve as a model for Alzheimer’s disease, has attracted some attention. Wictorin and colleagues [1992] have reported long-distance directed axonal growth from human dopaminergic mesencephalic neuroblasts implanted along the nigrostriatal pathway in 6-hydroxydopamine-lesioned rats. In this section we shall examine the use of synthetic


FIGURE 122.2 Schematic illustration of a nerve guidance channel and some of the possible strategies for influencing nerve regeneration.

guidance channels and extracellular matrix cues to guide axons to their appropriate targets. Thus a combination of all these techniques may render the complete physical and synaptic reconstruction of a degenerated pathway feasible. The promotion of nerve regeneration is an important candidate task for tissue reconstruction in the nervous system. Synthetic nerve guidance channels (NGCs) have been used to study the underlying mechanisms of mammalian peripheral nerve regeneration after nerve injury and enhance the regeneration process. Guidance channels may simplify end-to-end repair and may be useful in repairing long nerve gaps. The guidance channel reduces tension at the suture line, protects the regenerating nerve from infiltrating scar tissue, and directs the sprouting axons toward their distal targets. The properties of the guidance channel can be modified to optimize the regeneration process. Nerve guidance channels also may be used to create a controlled environment in the regenerating site. In the peripheral nervous system, NGCs can influence the extent of nerve gap that can be bridged and the quality of regeneration. The channel properties, the matrix filling the NGC, the cells seeded within the channel lumen, and polymerinduced welding of axons all can be strategies used to optimize and enhance nerve regeneration and effect nervous tissue reconstruction (see Fig. 122.2 for a schematic). Table 122.2 lists some of the kinds of nerve guidance channels used so far.

The Active Use of Channel Properties In the past, biocompatability of a biomaterial was evaluated by the degree of its passivity or lack of “reaction” when implanted into the body. However, the recognition that the response of the host tissue is related to the mechanical, chemical, and structural properties of the implanted biomaterial has led to the design of materials that promote a beneficial response from the host. In the context of a synthetic nerve guidance channel, this may translate to manipulation of its microstructural properties, permeability, electrical properties, and the loading of its channel wall with neuroactive components that might then be released locally into the regenerating environment. The strategy here is to engineer a tailored response from the host and take advantage of the natural repair processes. Surface Microgeometry The morphology of regenerating peripheral nerves is modulated by the surface microgeometry of polymeric guidance channels [Aebischer et al., 1990]. Channels with smooth inner walls give rise to organized, longitudinal fibrin matrices, resulting in discrete free-floating nerve cables with numerous myelinated axons. The rough inner surface channels, however, give rise to an unorganized fibrin matrix with nerve


TABLE 122.2

Nerve Guidance Channels

I. The Channel Wall 1. Passive polymeric channels Silicone elastomer Polyvinyl chloride Polyethylene 2. Permeable polymer channels Acrylonitrile vinychloride copolymer Collagen Expanded polytetrafluroethylene 3. Resorbable polymer channels Polyglycolic acid Poly-L-lactic acid Collagen 4. Electrically shaped polymer channels Silicone channels with electrode cuffs Polyvinylidenefluoride (piezoelectric) Polytetrofluoroethylene (electret) 5. Polymer channels releasing trophic factors Ethylene vinylacetate copolymer II. Intrachannel, luminal matrices 1. Fibrin matrix 2. Collagen-glycosaminoglycan template 3. Matrigel III. Cell seeded lumens for trophic support 1. Schwann cell-seeded lumens (PNS) 2. Schwann cell-seeded lumens (CNS)

Lundborg et al., 1982 Scaravalli, 1984 Madison et al., 1988 Uzman and Villegas, 1983 Archibald et al., 1991 Young et al., 1984 Molander et al., 1989 Nyilas et al., 1983 Archibald et al., 1991 Kerns and Freeman, 1986; Kerns et al., 1991 Aebischer et al., 1987 Valentini et al., 1989 Aebischer et al., 1989 Williams et al., 1987 Yannas et al., 1985 Valentini et al., 1987 Guénard et al., 1992 Guénard et al., 1993; Kromer and Cornbrooks, 1985, 1987; Smith and Stevenson, 1988

fascicles scattered in a loose connective tissue filling the entire channel’s lumen. Thus the physical textural properties and porosity of the channel can influence nervous tissue behavior and may be used to elicit a desirable reaction from the host tissue. Molecular Weight Cutoff The molecular weight cutoff of the NGCs influences peripheral nerve regeneration in rodent models [Aebischer et al., 1989]. The molecular weight cutoff may influence nerve regeneration possibly by controlling the exchange of molecules between the channel lumen and the external wound-healing environment. This may be important because the external environment consists of humoral factors that can play a role in augmenting regenerative processes in the absence of a distal stump. Electrical Properties In vivo regeneration following transection injury in the peripheral nervous system has been reported to be enhanced by galvantropic currents produced in silicone channels fitted with electrode cuffs [Kerns & Freeman, 1986; Kerns et al., 1991]. Polytetrafluoroethylene (PTFE) “electret” tubes show more myelinated axons compared with uncharged tubes in peripheral nerves [Valentini et al., 1989]. Dynamically active piezoelectric polymer channels also have been shown to enhance nerve regeneration in the sciatic nerves of adult mice and rats [Aebischer et al., 1987; Fine et al., 1991]. Release of Bioactive Factors from the Channel Wall Polymer guidance channels can be loaded with various factors to study and enhance nerve regeneration. Basic fibroblast growth factor released from an ethylene-vinyl acetate copolymer guidance channel facilitates peripheral nerve regeneration across long nerve gaps after a rat sciatic nerve lesion [Aebischer et al., 1989]. The possible influence of interleukin-1 (IL-1) on nerve regeneration also was studied by the release of IL-1 receptor antagonist (IL-1ra) from the wall of an EVA copolymer channel [Guénard et al., 1991]. It is conceivable that the release of appropriate neurotrophic factors from the channel wall may enhance


specifically subsets of axons, e.g., ciliary neurotrophic factor on motor neurons and nerve growth factor on sensory neurons.

Resorbable Channel Wall Bioresorbable nerve guidance channels are attractive because once regeneration is completed, the channel disappears without further surgical intervention. Mice sciatic nerves have been bridged with poly-L-lactic acid channels [Nyilas et al., 1983] and polyester guidance channels [Henry et al., 1985]. Rabbit tibial nerves also have been bridged with guidance channels fabricated from polyglycolic acid [Molander et al., 1989]. Resorbable guidance channels need to retain their mechanical integrity over 4 to 12 weeks. At the same time, their degradation products should not interfere with the regenerative processes of the nerve. These issues remain the challenging aspects in the development of bioresorbable nerve guidance channels for extensive use in animals and humans.

Intraluminal Matrices for Optimal Organization of Regeneration Microenvironment The physical support structure of the regenerating environment may play an important role in determining the extent of regeneration. An oriented fibrin matrix placed in the lumen of silicone guidance channels accelerates the early phases of peripheral nerve regeneration [Williams et al., 1987]. Silicone channels filled with a collagen-glycosaminoglycan template bridged a 15-mm nerve gap in rats, whereas no regeneration was observed in unfilled tubes [Yannas et al., 1985]. However, even matrices known to promote neuritic sprouting in vitro may impede peripheral nerve regeneration in semipermeable guidance channels if the optimal physical conditions are not ensured [Madison et al., 1988; Valentini et al., 1987]. Therefore, the structural, chemical, and biologic aspects of the matrix design may all play a role in determining the fate of the regenerating nerve. The importance of the effect of the physical environment on regeneration, mediated by its influence on fibroblast and Schwann cell behavior, has been demonstrated in several studies and has been reviewed by Schwartz [1987] and Fawcett & Keynes [1990]. Thus the choice of a hydrogel with physical, chemical, and biologic cues conducive to nerve regeneration may enhance nerve regeneration. This strategy is currently being explored [Belamkonda et al., in press]. Neurite-promoting oligopeptides from the basement membrane protein laminin (LN) were covalently coupled to agarose hydrogels. Agarose gels derivatized with LN oligopeptides specifically enhance neurite extension from cells that have receptors to the LN peptides in vitro [Bellamkonda et al., in press]. Preliminary results show that the presence of an agarose gel carrying the LN peptide CDPGYIGSR inside the lumen of a synthetic guidance channel enhances the regeneration of transected peripheral nerves (Fig. 122.3) in rats. Thus it is feasible to tailor the intraluminal matrices with more potent neurite-promoting molecules such as the cell adhesion molecules (CAMs) L1, N-CAM, or tenascin and “engineer” a desired response from the regenerating neural elements.

Cell-Seeded Lumens for Trophic Support Cells secreting various growth factors may play an important role in organizing the regeneration environment, e.g., Schwann cells in the peripheral and central nervous system. It has been reported that regenerating axons do not elongate through acellular nerve grafts if Schwann cell migration was impeded [Hall et al., 1986]. Syngeneic Schwann cells derived from adult nerves and seeded in semipermeable guidance channels enhance peripheral nerve regeneration [Guénard et al., 1992]. Schwann cells in the preceding study orient themselves along the axis of the guidance channel, besides secreting various neurotrophic factors. Schwann cells could play a role in organizing the fibirin cable formed during the initial phases of nerve regeneration. Schwann cells may be effective in inducing regeneration in the CNS, too [Kromer & Cornbrook, 1985, 1987; Smith & Stevenson, 1988]. The use of tailored intraluminal matrices and presenting exogeneic Schwann cells to the regeneration environment in a controlled matter


FIGURE 122.3 Light micrograph of a cross-sectional cut of a sural nerve regenerating through a polymer guidance channel 4 weeks after transection. The nerve guidance channel had been filled with a CDPGYIGSR derivatized agarose gel. E is the epineurium; V shows neovascularization; and MA is myelinated axon.

are strategies aimed at engineering the desired tissue response by creating the optimal substrate, trophic, and cellular environments around the regenerating nerves. CNS glial cells have a secretory capacity that can modulate neuronal function. Astrocytes release proteins that enhance neuronal survival and induce neuronal growth and differentiation. When a silicone channel was seeded with astrocytes of different ages, ranging from P9 to P69 (postnatal), it was observed that while P9 astrocytes did not interfere with peripheral nerve regeneration, adult astrocytes downregulate axonal growth [Kalderon, 1988]. However, the presence of Schwann cells reverses the inhibition of PNS regeneration due to adult astrocytes [Guénard et al., 1994]. Thus the cellular environment in the site of injury may play an important role in determining the extent of regeneration. Knowledge of these factors also may be employed in designing optimal environments for nerve regeneration. Polyethyleneglycol-Induced Axon Fusion Rapid morphologic fusion of severed myelinated axons may be achieved by the application of polyethylene glycol (PEG) to the closely apposed ends of invertebrate-myelinated axons [Krause & Bittner, 1990]. Selection of appropriate PEG concentration and molecular mass, tight apposition, and careful alignment of the cut ends of the nerve may facilitate the direct fusion of axons. However, this technique is only applicable when the two ends of the severed nerve are closely apposed to each other, before the onset of wallerian degeneration.

CNS Nerve Regeneration Most of the preceding studies have been conducted in the peripheral nervous system (PNS). In the CNS, however, endogenous components express poor support for axonal elongation. Significant regeneration may, however, occur with supporting substrates. Entubulation with a semipermeable acrylic copolymer tube allows bridging of a transected rabbit optic nerve with a cable containing myelinated axons [Aebischer et al., 1988]. Cholinergic nerve regeneration into basal lamina tubes containing Schwann cells has been reported in a transected septohippocampal model in rats [Kromer & Cornbrooks, 1985].


Thus the appropriate combination of physical guidance, matrices, and growth factors can create the right environmental cues and may be effective in inducing regeneration in the CNS. Therefore, both in the PNS and CNS, manipulation of the natural regenerative capacities of the host either by guidance factor or stimulation by electrical or trophic factors or structural components of the regenerating microenvironment can significantly enhance regeneration and help the reconstruction of severed or damaged neural tissue.

122.3 In Vitro Neural Circuits and Biosensors The electrochemical and chemoelectrical transduction properties of neuronal cells can form the basis of a cell-based biosensing unit. The unique information-processing capabilities of neuronal cells through synaptic modulation may form the basis of designing simple neuronal circuits in vitro. Both the preceding applications necessitate controlled neuronal cell attachment, tightly coupled to the substrate and a sensitive substrate to monitor changes in the cell’s electrical activity. The use of bioactive material systems tailored to control neuronal cell attachment on the surface and still amenable to the incorporation of electrical sensing elements like a field effect transistor (FET) could be one feasible design. Therefore, composite material systems, which might incorporate covalently patterned bioactive peptides on their surface to control cell attachment and neurite extension, may be a step toward the fulfillment of the preceding goal. Oligopeptides derived from larger extracellular proteins like laminin have been shown to mediate specific cell attachment via cell surface receptors [Graf et al., 1987; Iwamoto et al., 1987; Kleinman et al., 1988]. Cell culture on polymeric membranes modified with the preceding bioactive peptidic components may give rise to a system where neuronal cell attachment and neuritic process outgrowth may be controlled. This control may help in designing microelectronic leads to complete the cell-electronic junction. Preliminary recordings from a FET-based neuron-silicon junction using leech Retzius cells [Fromherz et al., 1991] have been reported. Though there are many problems, such as attaining optimal coupling, this could form the basis of a “neural chip.” A neural chip could potentially link neurons to external electronics for applications in neuronal cell-based biosensors, neural circuits, and limb prosthesis. Polymer surface modification and intelligent use of extracellular matrix components through selective binding could help attain this goal. Studies in our laboratory have been trying to understand the underlying mechanisms involving protein adsorption onto polymeric substrates and their role in influencing and controlling nerve cell attachment [Ranieri et al., 1993]. Controlled neuronal cell attachment within a tolerance range of 20 µm may be achieved either nonspecifically monoamine surfaces or specifically via oligopeptides derived from ECM proteins like laminin and fibronectin [Fig. 122.4], mediated by integrin cell surface receptors [Ranieri et al., in press]. Molecular control of neuronal cell attachment and interfacing neuronal cells with electrodes may find applications in the design and fabrication of high-sensitivity neuron-based biosensors with applications in detection of low level neurotransmitters. Studies are also currently in progress involving polymeric hydrogels and controlling neuronal cell behavior in a three-dimensional (3D) tissue culture environs as a step toward building 3D neuronal tissues [Bellamkonda et al., in press]. The choice of an appropriate hydrogel chemistry and structure, combined with the possibility of the gel serving as a carrier for ECM proteins or their peptidic analogues, can enable one to enhance regeneration when seeded in a nerve guidance channel. Also, the use of appropriate hydrogel chemistries in combination with the chemical modification of the polymer backbone by laser-directed photochemistry may be feasible in controlling the direction and differentiation of neuronal cells in three dimensions. Covalent binding of bioactive components like the laminin oligopeptides to the hydrogel backbone gives a specific character to the gel so that it elicits specific responses from anchorage-dependent neuronal cells [Bellamkonda et al., in press] (Fig. 122.5). Such a system could be useful in reorganizing nerves in 3D either for bridging different regions of the brain with nerve cables or for the 3D organization of nerves for optimal coupling with external electronics in the design of artificial limb prosthesis. In either case, development of such systems presents an interesting challenge for tissue engineering. © 2000 by CRC Press LLC

FIGURE 122.4 (a) Schematic of fluorinated ethylene propylene membrane modified in a “striped” fashion with bioactive oligopeptides using carbonyldiimidazole homobifunctional linking agent. (b) Light micrograph of Ng10815 cells “striping” on FEP membrane surfaces selectively modified with CDPGYIGSR oligopeptide.

122.4 Conclusion Advances in gene transfer techniques and molecular and cell biology offer potent tools in the functional replacement of various tissues of the nervous system. Each of these cells’ functions can be optimized with the design and selection of its optimal extracellular environment. Substrates that support neuronal differentiation in two and three dimensions may play an important role in taking advantage of the advances in molecular and cell biology. Thus research aimed at tailoring extracellular matrices with the


FIGURE 122.5 (a) Schematic of hydrogel derivatized with bioactive peptides with an anchorage-dependent neuron suspended in 3D. (b) Light micrograph of a E14 chick superior cervical ganglion suspended in 3D and extending neurites in an agarose gel derivatized with the laminin oligopeptide CDPGYIGSR.

appropriate physical, chemical, and biologic cues may be important in optimizing the function of transplanted cells, inducing nerve regeneration, or in the construction of neuronal tissues in two and three dimensions in a controlled fashion. Controlled design and fabrication of polymer hydrogels and polymer scaffolds on a scale that is relevant for single cells also may be important. This would presumably control the degree and the molecular location of permissive, attractive, and repulsive regions of the substrate and in turn control cellular and tissue response in vitro and in vivo. Biologic molecules like laminin, collagen, fibronectin, and tenascin may provide attractive and permissive pathways for axons to grow. On the other hand, some sulfated proteoglycans have been shown to inhibit or repulse neurites [Snow & Letourneau, 1992]. The use of these molecules coupled with a © 2000 by CRC Press LLC

clearer understanding of protein-mediated material-cell interaction may pave the way for neural tissue engineering, molecule by molecule, in three dimensions. Thus it is possible to tailor the genetic material of a cell to make it neurologically relevant and to control its expression by optimizing its extracellular environment. All the preceding factors make tissue engineering in the nervous system an exciting and challenging endeavor.

Acknowledgments We wish to than Mr. Nicolas Boche for the illustrations.

References Aebischer P, Goddard M, Signore A, Timpson R. 1994. Functional recovery in MPTP lesioned primates transplanted with polymer encapsulated PC12 cells. Exp. Neurol 26:1. Aebischer P, Buchser E, Joseph JN. 1994. Transplantation in humans of encapsulated xenogeneic cells without immunosuppression. Transplantation 58:1275. Aebischer P, Guénard V, Brace S. 1989a. Peripheral nerve regeneration through blind-ended semipermeable guidance channels: effect of the molecular weight cutoff. J Neurosci 9:3590. Aebischer P, Guénard V, Valentini RF, 1990. The morphology of regenerating peripheral nerves is modulated by the surface microgeometry of polymeric guidance channels. Brain Res 531:211. Aebischer P, Salessiotis AN, Winn SR. 1989b. Basic fibroblast growth factor released from synthetic guidance channels facilitates peripheral nerve regeneration across long nerve gaps. J Neurosci Res 28:282. Aebischer P, Tresco PA, Winn SR, et al. 1991a. Long-term cross-species brain transplantation of a polymerencapsulated dopamine-secreting cell line. Exp Neurol 111:269. Aebischer P, Tresco PA, Sagen J, Winn SR. 1991b. Transplantation of microencapsulated bovine chromaffin cells reduces lesion-induced rotational asymmetry in rats. Brain Res 560:43. Aebsicher P, Wahlberg L, Tresco PA, Winn SR. 1991c. Macroencapsulation of dopamine secreting cells by coextrusion with an organic polymer solution. Biomaterials 12:50. Aebischer P, Winn SR, Galletti PM. 1998a. Transplantation of neural tissue in polymer capsules. Brain Res 448:364. Aebischer P, Valentini RF, Dario P, et al. 1987. Piezoelectric guidance channels enhance regeneration in the mouse sciatic nerve after axotomy. Brain Res 436:165. Aebischer P, Valentini RF, Winn SR, Galletti PM. 1988b. The use of a semipermeable tube as a guidance channel for a transected rabbit optic nerve. Brain Res 78:599. Ahlgren FI, Ahlgren MB. 1987. Epidural administration of opiates by a new device. Pain 31:353. Akil H, Watson SJ, Young E, et al. 1984. Endogenous opioids: biology and function. Annu Rev Neurosci 7:223. Albani C, Asper R, Hacisalihzade SS, Baumgartner F. 1986. Individual levodopa therapy in Parkinson’s disease. In Advances in Neurology: Parkinson’s Disease, pp 497–501. New York, Raven Press. Anderson RF, Alterman AL, Barde YA, Lindsay RM. 1990. Brain-derived neurotrophic factor increases survival and differentiation of septal cholinergic neurons in culture. Neuron 5:297. Archibald SJ, Krarup C, Shefner J, et al. 1991. A collagen-based nerve guide conduit for peripheral nerve repair: an electrophysiological study of nerve regeneration in rodents and nonhuman primates. J Comp Neurol 306–685. Becker JB, Robinson TE, Barton P, et al. 1990. Sustained behavioral recovery from unilateral nigrostriatal damage produced by the controlled release of dopamine from a silicone polymer pellet placed into the denervated striatum. Brain Res 508:60. Bellamkonda R, Ranieri JP, Aebischer P. Laminin oligopeptide derivatized agarose gels allow threedimensional neurite outgrowth in vitro. J Neurosci Res. In press. Bellamkonda R, Ranieri JP, Bouche N, Aebischer P. In press. A hydrogel-based three-dimensional matrix for neural cells. J Biomed Nat Res. In press. © 2000 by CRC Press LLC

Birkmayer W. 1969. Experimentalle ergebnisse uber die kombinationsbehandlung des Parkinsonsyndroms mit 1-dopa und einem decarboxylasehemmer. Wiener Klin Wochenschr 81:677. Björklund A. 1991. Neural transplantation—An experimental tool with clinical possibilities. TINS 14:319. Björklund A, Dunnett SB, Stenevi U, et al. 1980. Reinnervation of the denervated striatum by substantia nigra transplants: Functional consequences as revealed by pharmacological and sensorimotor testing. Brain Res 199:307. Bolam JP, Freund TF, Björklund A, et al. 1987. Synaptic input and local output of dopaminergic neurons in grafts that functionally reinnervate the host neostriatum. Exp Brain Res 68:131. Bunge RP, Bunge MB. 1983. Interrelationship between Schwann cell function and extracellular matrix production. TINS 6:499. Chen LS, Ray J, Fisher LJ, et al. 1991. Cellular replacement therapy for neurologic disorders: potential of genetically engineered cells. J Cell Biochem 45:252. De Yebenes JG, Fahn S, Jackson-Lewis V, et al. 1988. Continuous intracerebroventricular infusion of dopamine and dopamine agonists through a totally implanted drug delivery system in animal models of Parkinson’s disease. J Neural Transplant 27:141. During MJ, Freese A, Deutch AY, et al. 1992. Biochemical and behavioral recovery in a rodent model of Parkinson’s disease following stereotactic implantation of dopamine-containing liposomes. Exp Neurol 115:193. Ehringer H, Hornykiewicz O. 1960. Vetreilung von Noradrenalin und Dopamin (3-Hydroxtyramin) im Gehirn des menschen und ihr Verhalten bei Erkrankungen des extrapyramidalen systems. Klin Ther Wochenschr 38:1236. Fawcett JW, Keynes RJ. 1990. Peripheral nerve regeneration. Annu Rev Neurosci 13:43. Fine EG, Valentini RF, Bellamkonda R, Aebischer P. 1991. Improved nerve regeneration through piezoelectric vinylidenefluoride-trifluoroethylene copolymer guidance channels. Biomaterials 12:775. Fromherz P, Offenhausser A, Vetter T, Weis J. 1991. A neuron-silicon junction: A Retzius cell of the leech on an insulated-gate field-effect transistor. Science 252:1290. Gage FH, Kawaja MD, Fisher LJ. 1991. Genetically modified cells: Applications for intracerebral grafting. TINS 14:328. Goosen MFA, Shea GM, Gharapetian HM, et al. 1985. Optimization of microencapsualtion parameters: semipermeable microcapsules as a bioartificial pancreas. Biotech Bioeng 27:146. Graf J, Ogle RC, Robey FA, et al. 1987. A pentrapeptide from the laminin B1 chain mediates cell adhesion and binds the 67,000 laminin receptor. Biochemistry 26:6896. Guénard V, Aebischer P, Bunge R. 1994. The astrocyte inhibition of peripheral nerve regeneration is reversed by Schwann cells. Exp Neurol 126:44. Guénard V, Dinarello CA, Weston PJ, Aebischer P. 1991. Peripheral nerve regeneration is impeded by interleukin 1 receptor antagonist released from a polymeric guidance channel. J Neurosci Res 29:396. Guénard V, Kleitman N, Morrissey TK, et al. 1992. Syngeneic Schwann cells derived from adult nerves seeded in semipermeable guidance channels enhance peripheral nerve regeneration. J Neurosci 2:3310. Guénard V, Xu XM, Bunge MB. 1993. The use of Schwann cell transplantation to foster central nervous system repair. Semin Neurosci 5:401. Hall SM. 1986. The effect of inhibiting Schwann cells mitosis on the re-innervation of acellular autografts in the peripheral nervous system of the mouse. Neuropathol Appl Neurobiol 12:27. Hardie RJ, Lees AJ, Stern GM. 1984. On-off fluctuations in Parkinson’s disease: A clinical and neuropharmacological study. Brain 107:487. Hargraves R, Freed WJ. 1987. Chronic intrastriatal dopamine infusions in rats with unilateral lesions of the substantia nigra. Life Sci 40:959. Hefti F, Dravid A, Hartikka J. 1984. Chronic intraventricular injections of nerve growth factor elevate hippocampal choline acetyltransferase activity in adult rats with partial septo-hippocampal lesions. Brain Res 293:305.


Henry EW, Chiu TH, Nyilas E, et al. 1985. Nerve regeneration through biodegradable polyester tubes. Exp Neurol 90:652. Hoffman D, Breakefield XO, Short MP, Aebischer P. 1993. Transplantation of a polymer-encapsulated cell line genetically engineered to release NGF. Exp Neurol 122:100. Hoffman D, Wahlberg L, Aebischer P. 1990. NGF released from a polymer matrix prevents loss of ChaT expression in basal forebrain neurons following a fimbria-fornix lesion. Exp Neurol 110:39. Horellou P, Guilbert B, Leviel V, Mallet J. 1989. Retroviral transfer of a human tyrosine hydroxylase cDNA in various cell lines: Regulated release of dopamine in mouse anterior pituitary AtT-20 cells. Proc Natl Acad Sci USA 86:7233. Hughes RA, Sendtner M, Thoenen H. 1993. Members of several gene families influence survival of rat motoneurons in vitro and in vivo. J Neurosci Res 36:663. Hyman C, Hofer M, Barde YA, et al. 1991. BDNF is a neurotrophic factor for dopaminergic neurons of the substantia nigra. Nature 350:230. Iwamoto Y, Robey FA, Graf J, et al. 1987. YIGSR, a synthetic laminin pentapeptide, inhibits experimental metastasis formation. Science 238:1132. Jiao S, Gurevich V, Wolff JA. 1993. Long-term correction of rat model of Parkinson’s disease by gene therapy. Nature 362:450. Joseph JM, Goddard MB, Mills J, et al. 1994. Transplantation of encapsulated bovine chromaffin cells in the sheep subarachnoid space: a preclinical study for the treatment of cancer pain. Cell Transplant 3:355. Kalderon N. 1988. Differentiating astroglia in nervous tissue histogenesis regeneration: studies in a model system of regenerating peripheral nerve. J Neurosci Res 21:501. Kerns JM, Fakhouri AJ, Weinrib HP, Freeman JA. 1991. Electrical stimulation of nerve regeneration in the rat: The early effects evaluated by a vibrating probe and electron microscopy. J Neurosci 40:93. Kleinman H, Ogle RC, Cannon FB, et al. 1988. Laminin receptors for neurite formation. Proc Natl Acad Sci USA 85:1282. Knüsel B, Winslow JW, Rosenthal A, et al. 1991. Promotion of central cholinergic and dopaminergic neuron differentiation by brain derived neurotrophic factor but not neurotrophin-3. Proc Natl Acad Sci USA 88:961. Krause TL, Bittner GD. 1990. Rapid morphological fusion of severed myelinated axons by polyethylene glycol. Proc Natl Acad Sci USA 87:1471. Kromer LF, Cornbrooks CJ. 1985. Transplants of Schwann cell culture cultures promote axonal regeneration in adult mammalian brain. Proc Natl Acad Sci USA 82:6330. Kromer LF, Cornbrooks CJ. 1987. Identification of trophic factors and transplanted cellular environments that promote CNS axonal regeneration. Ann NY Acad Sci 495:207. Langer R. 1981. Polymers for sustained release of macromolecules: Their use in a single-step method for immunization. IN JJ Langone, J Van Vunakis (eds), Methods of Enzymology pp 57–75. San Diego, Academic Press. Langer R, Folkman J. 1976. Polymers for sustained release of proteins and other macromolecules. Nature 263:797. Lim F, Sun AM. 1980. Microencapsulated islets as bioartificial endocrine pancreas. Science 210:908. Lin HL-F, Doherty DH, Lile JD, et al. 1993. GDNF: A glial derived neurotrophic factor for midbrain dopaminergic neurons. Science 260:1130. Lindvall O. 1991. Prospects of transplantation in human neurodegenerative diseases. TINS 14:376. Lindvall O, Brundin P, Widner H, et al. 1990. Grafts of fetal dopamine neurons survive and improve motor function in Parkinson’s disease. Science 247:574. Lundborg G, Dahlin LB, Danielsen N, et al. 1982. Nerve regeneration in silicone chambers: Influence of gap length and of distal stump components. Exp Neurol 76:361. Madison RD, Da Silva CF, Dikkes P. 1988. Entubulation repair with protein additives increases the maximum nerve gap distance successfully bridged with tubular prosthesis. Brain Res 447:325.


McRae A, Hjorth S, Mason DW, et al. 1991. Microencapsulated dopamine (DA)-induced restitution of function in 6-OHDA denervated rat striatum in vivo: Comparison between two microsphere excipients. J Neural Transplant Plast 2:165. Molander H, Olsson Y, Engkvist O, et al. 1989. Regeneration of peripheral nerve through a polygalactin tube. Muscle Nerve 5:54. Morrissey TK, Kleitman N, Bunge RP. 1991. Isolation and functional characterization of Schwann cells derived from adult nerve. J Neurosci 11:2433. Muenter MD, Sharpless NS, Tyce SM, Darley FL. 1977. Patterns of dystonia (I-D-I) and (D-I-D) in response to 1-dopa therapy for Parkinson’s disease. Mayo Clin Proc 52:163. Muir D, Gennrich C, Varon S, Manthorpe M. 1989. Rat sciatic nerve Schwann cell microcultures: Responses to mitogens and production of trophic and neurite-promoting factors. Neurochem Res 14:1003. Nyilas E, Chiu TH, Sidman RL, et al. 1983. Peripheral nerve repair with bioresorbable prosthesis. Trans Am Soc Artif Intern Organs 29:307. Olson L, Backlund EO, Ebendal T, et al. 1991. Intraputaminal infusion of nerve growth factor to support adrenal medullary autografts in Parkinson’s disease: One year follow-up of first clinical trial. Arch Neurol 48:373. Olson L, Nordberg A, Von-Holst H, et al. 1992. Nerve growth factor affects 11C-nicotine binding, blood flow, EEG, and verbal episodic memory in an Alzheimer patient (case report). J Neural Transm Park Dis Dement Sect 4:79. Oppenheim RW, Prevette D, Yin QW, et al. 1991. Control of embryonic motoneuron survival in vivo by ciliary neurotrophic factor. Science 251:1616. Ranieri JP, Bellamkonda R, Bekos E, et al. 1994. Spatial control of neural cell attachment via patterned laminin oligopeptide chemistries. Int J Dev Neurosci 12:725. Ranieri JP, Bellamkonda R, Jacob J, et al. 1993. Selective neuronal cell attachment to a covalently patterned monoamine of fluorinated ethylene propylene films. J Biomed Mater Res 27:917. Reynolds BA, Tetzlaff W, Weiss S. 1992. A multipotent EGF_responsive striatal embryonic progenitor cell produces neurons and astrocytes. J Neurosci 12:4565. Sabel BA, Dominiak P, Hauser W, et al. 1990. Levodopa delivery from controlled release polymer matrix: Delivery of more than 600 days in vitro and 225 days elevated plasma levels after subcutaneous implantation in rats. J Pharmacol Exp Ther 255:914. Sagen J, Pappas GD, Winnie AP. 1993a. Alleviation of pain in cancer patients by adrenal medullary transplants in the spinal subarachnoid space. Cell Transplant 2:259. Sagen J, Wang H, Tresco PA, Aebischer P. 1993b. Transplants of immunologically isolated xenogeneic chromaffin cells provide a long-term source of pain-reducing neuroactive substances. J Neurosci 13:2415. Scaravalli F. 1984. Regeneration of peineurium across a surgically induced gap in a nerve encased in a plastic tube. J Anat 139:411. Schwartz M. 1987. Molecular and cellular aspects of nerve regeneration. CRC Crit Rev Biochem 22:89. Sendtner M, Kreutzberg GW, Thoenen H. 1990. Ciliary neurotrophic factor prevents the degeneration of motor neurons after axotomy. Nature 345:440. Shoulson I, Claubiger GA, Chase TN. 1975. On-off response. Neurology 25:144. Smith GV, Stevenson JA. 1988. Peripheral nerve grafts lacking viable Schwann cells fail to support central nervous system axonal regeneration. Exp Brain Res 69:299. Snow DM, Letourneau PC. 1992. Neurite outgrowth on a step gradient of chondroitin sulfate proteoglycan (CS-PG). J Neurobiol 23:322. Tresco PA, Winn SR, Aebischer P. 1992. Polymer encapsulated neurotransmitter secreting cells; Potential treatment for Parkinson’s disease. ASAIO J 38:17. Uzman BG, Villegas GM. 1983. Mouse sciatic nerve regeneration through semipermeable tubes: A quantitative model. J Neurosci Res 9:325.


Valentini RF, Aebischer P, Winn SR, Galletti PM. 1987. Collagen- and laminin-containing gels impede peripheral nerve regeneration through semipermeable nerve guidance channels. Exp Neurol 98:350. Valentini RF, Sabatini AM, Dario P, Aebischer P. 1989. Polymer electret guidance channels enhance peripheral nerve regeneration in mice. Brain Res 48:300. Vescovi AL, Reynolds BA, Fraser DD, Weiss S. 1993. BFGF regulates the proliferative fate of unipotent (neuronal) and bipotent (neuronal/astroglial) EGF-generated CNS progenitor cells. Neuron 11:951. Wictorin K, Brundin P, Sauer H, et al. 1992. Long distance directed axonal growth from human dopaminergic mesencephalic neuroblasts implanted along the nigrostriatal pathway in 6-hydroxydopamine lesioned rats. J Comp Neurol 323:475. Williams LR, Danielsen N, Muller H, Varon S. 1987. Exogenous matrix precursors promote functional nerve regeneration across a 15-mm gap within a silicone chamber in the rat. J Comp Neurol 264:284. Williams LR, Varon S, Peterson GM, et al. 1986. Continuos infusion of nerve growth factor prevents basal forebrain neuronal death after fimbria-fornix transection. Proc Natl Acad Sci USA 83:9231. Winn SR, Tresco PA, Zielinski B, et al. 1991. Behavioral recovery following intrastriatal implantation of microencapsulated PC12 cells. Exp Neurol 113:322. Winn SR, Wahlberg L, Tresco PA, Aebischer P. 1989. AN encapsulated dopamine-releasing polymer alleviates experimental parkinsonism in rats. Exp Neurol 105:244. Yaffe D, Saxel O. 1977. Serial passaging and differentiation of myogenic cells isolated from dystrophic mouse muscle. Nature 270:725. Yan Q, Elliott J, Snider WD. 1992. Brain-derived neurotrophic factor rescues spinal motor neurons from axotomy-induced cell death. Nature 360:753. Yannas EV, Orgill DP, Silver J, et al. 1985. Polymeric template facilitates regeneration of sciatic nerve across 15 mm gap. Trans Soc Biomater 11:146. Young BL, Begovac P, Stuart D, Glasgow GE. 1984. An effective sleeving technique for nerve repair. J Neurosci Methods 10:51. Zielinski B, Aebischer P. 1994. Encapsulation of mammalian cells in chitosan-based microcapsules: effect of cell anchorage dependence. Biomaterials. Zurn AD, Baetge EE, Hammang JP, et al. 1994. Glial cell line-derived neurotrophic factor (GDNF): A new neurotrophic factor for motoneurones. Neuroreport 6:113.


Brooks, S.V.,Faulkner, J.A. “Tissue Engineering of Skeletal Muscle.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

123 Tissue Engineering of Skeletal Muscle 123.1 123.2 123.3 123.4

Introduction Skeletal Muscle Structure Skeletal Muscle Function Injury and Repair of Skeletal Muscle Injury of Skeletal Muscle • Repair of Injured Skeletal Muscles • Satellite Cell Activation • Myogenic Regulatory Factors

123.5

Susan V. Brooks University of Michigan

John A. Faulkner University of Michigan

123.6

Reconstructive Surgery of Whole Skeletal Muscles Myoblast Transfer and Gene Therapy Myoblast Transfer Therapy • Gene Therapy • Transgenic Mice • Direct Intramuscular Injection • RetrovirusMediated Gene Transfer • Adenovirus-Mediated Gene Transfer

123.1 Introduction Contractions of skeletal muscles generate the stability and power for all movement. Consequently, any impairment in skeletal muscle function results in at least some degree of instability or immobility. Muscle function can be impaired as a result of injury, disease, or old age. The goal of tissue engineering is to restore the structural and functional properties of muscles to permit the greatest recovery of normal movement. Impaired movement at all ages, but particularly in the elderly, increases the risk of severe injury, reduces participation in the activities of daily living, and impacts on the quality of life. Contraction is defined as the activation of muscle fibers with a tendency of the fibers to shorten. Contraction occurs when an increase in the cytosolic calcium concentration triggers a series of molecular events that includes the binding of calcium to the muscle regulatory proteins, the formation of strong interactions between the myosin cross-bridges and the actin filaments, and the generation of the crossbridge driving stroke. In vivo, muscles perform three types of contractions depending on the interaction between the magnitude of the force developed by the muscle and the external load placed on the muscle. When the force developed by the muscle is greater than the load on the muscle, the fibers shorten during the contraction. When the force developed by the muscle is equal to the load, or if the load is immovable, the overall length of the muscle remains the same. If the force developed by the muscle is less than the load placed on the muscle, the muscle is stretched during the contraction. The types of contractions are termed miometric, isometric, and pliometric, respectively. Most normal body movements require varying proportions of each type of contraction.


FIGURE 123.1 Levels of anatomical organization within a skeletal muscle. Source: Bloom, W. and Fawcett, D.W. 1968. A Textbook of Histology, 9th ed., Saunders, Philadelphia. With permission.

123.2 Skeletal Muscle Structure Each of the 660 skeletal muscles in the human body is composed of hundreds to hundreds of thousands of single muscle fibers (Fig. 123.1). The plasma membrane of a muscle fiber is termed the sarcolemma. Contractile, structural, metabolic, regulatory, and cytosolic proteins, as well as many myonuclei and other cytosolic organelles are contained within the sarcolemma of each fiber (Fig. 123.2). The contractile proteins myosin and actin are incorporated into thick and thin myofilaments, respectively, which are arrayed in longitudinally repeated banding patterns termed sarcomeres (Fig. 123.1). Sarcomeres in series form myofibrils, and many parallel myofibrils exist within each fiber. The number of myofibrils arranged in parallel determines the cross-sectional area (CSA) of single fibers, and consequently, the force generating capability of the fiber. During a contraction, the change in the length of a sarcomere occurs as thick and thin filaments slide past each other, but the overall length of each actin and myosin filament does not change. An additional membrane, referred to as the basement membrane or the basal lamina, surrounds the sarcolemma of each fiber (Fig. 123.2). In mammals, the number of fibers in a given muscle is determined at birth and changes little throughout the life span except in cases of injury or disease. In contrast, the number of myofibrils can change dramatically, increasing with normal growth or hypertrophy induced by strength training and decreasing


FIGURE 123.2 Drawing of a muscle fiber-satellite cell complex. Note that the satellite cell is located between the muscle fiber sarcolemma and the basal lamina. Source: Carlson and Faulkner, 1983. With permission.

with atrophy associated with immobilization, inactivity, injury, disease, or old age. A single muscle fiber is innervated by a single branch of a motor nerve. A motor unit is composed of a single motor nerve, its branches, and the muscle fibers innervated by the branches. The motor unit is the smallest group of fibers within a muscle that can be activated volitionally. Activation of a motor unit occurs when action potentials emanating from the motor cortex depolarize the cell bodies of motor nerves. The depolarization generates an action potential in the motor nerve that is transmitted to each muscle fiber in the motor unit, and each of the fibers contracts more or less simultaneously. Motor units range from small slow units to large fast units dependent on the CSA of the motor nerve.

123.3 Skeletal Muscle Function Skeletal muscles may contract singly or in groups, working synergistically. On either side of limbs, muscles contract against one another or antagonistically. The force or power developed during a contraction depends on the frequency of stimulation of the motor units, the total number of motor units, and the size of the motor units recruited. The frequency of stimulation, particularly for the generation of power, is normally on the order of the frequency-power relationship. Consequently, the total number of motor units recruited is the major determinant of the force or power developed.


Motor units are classified into three general categories based on their functional properties [Burke et al., 1973]. Slow (S) units have the smallest single muscle fiber CSAs, the fewest muscle fibers per motor unit, and the lowest velocity of shortening. The cell bodies of the S units are the most easily depolarized to threshold [Henneman, 1965]. Consequently, S units are the most frequently recruited during tasks that require low force or power but highly precise movements. Fast-fatigable (FF) units are composed of the largest fibers, have the most fibers per unit, and have the highest velocities of shortening. The FF units are the last to be recruited and are recruited for high force and power movements. The fast fatigueresistant (FR) units are intermediate in terms of the CSAs of their fibers, the number of fibers per motor unit, the velocity of shortening, and the frequency of recruitment. The force normalized per unit CSA is ~280 kN/m2 for each type of fiber, but the maximum normalized power (W/kg) developed by FF units is as much as four-fold greater than that of the S units due to a four-fold higher velocity of shortening for FF units. Motor units may also be identified by histochemical techniques as Type I (S), IIA (FR), and IIB (FF). Classifications based on histochemical and functional characteristics are usually in good agreement with one another, but differences do exist, particularly following experimental interventions. Consequently, in a given experiment, the validity of this interpretation should be verified.

123.4 Injury and Repair of Skeletal Muscle Injury to skeletal muscles may occur as a result of disease, such as dystrophy; exposure to myotoxic agents, such as bupivacaine or lidocaine; sharp or blunt trauma, such as punctures or contusions; ischemia, such as that which occurs with transplantation; exposure to excessively hot or cold temperatures; and contractions of the muscles. Pliometric contractions are much more likely to injure muscle fibers than are isometric or miometric contractions [McCully and Faulkner, 1985]. Regardless of the factors responsible, the manner in which the injuries are manifested appears to be the same, varying only in severity. In addition, the processes of fiber repair and regeneration appear to follow a common pathway regardless of the nature of the injurious event [Carlson and Faulkner, 1983].

Injury of Skeletal Muscle The injury may involve either some or all of the fibers within a muscle [McCully and Faulkner, 1985]. In an individual fiber, focal injuries, localized to a few sarcomeres in series or in parallel (Fig. 123.3), as well as more widespread injuries, spreading across the entire cross section of the fiber, are observed using electron microscopic techniques [Macpherson et al., 1997]. Although the data are highly variable, many injuries also give rise to increases in serum levels of muscle enzymes, particularly creatine kinase, leading to the conclusion that sarcolemmal integrity is impaired [McNeil and Khakee, 1992; Newham et al., 1983]. This conclusion is further supported by an influx of circulating proteins, such as serum albumin [McNeil and Khakee, 1992], and of calcium [Jones et al., 1984]. An increase in intracellular calcium concentration may activate a variety of proteolytic enzymes leading to further degradation of sarcoplasmic proteins. In cases when the damage involves a large proportion of the sarcomeres within a fiber, the fiber becomes necrotic. If blood flow is impaired, fibers remain as a necrotic mass of noncontractile tissue [Carlson and Faulkner, 1983]. In contrast, in the presence of an adequate blood supply, the injured fibers are infiltrated by monocytes and macrophages [McCully and Faulkner, 1985]. The phagocytic cells remove the disrupted myofilaments, other cytosolic structures, and the damaged sarcolemma (Fig. 123.4). The most severe injuries result in the complete degeneration of the muscle fiber, leaving only the empty basal lamina. The basal lamina appears to be highly resistant to any type of injury and generally remains intact [Carlson and Faulkner, 1983]. An additional indirect measure of injury is the subjective report by human beings of delayed onset muscle soreness, common following intense or novel exercise [Newham et al., 1983]. Because of the focal nature of the morphological damage, the variability of serum enzyme levels and the subjectivity of reports


FIGURE 123.3 Focal areas of damage to single sarcomeres after a single 40% stretch of a single maximally activated rat skeletal muscle fiber (average sarcomere length 2.6 µm). Two types of damage are observed in this electron micrograph. The arrow indicates Z-line streaming and asterisks show disruption of the thick and thin filaments in the region of overlap, the A-band. A third type of damage, the displacement of thick filaments to one Z-line is not shown. Note that with the return of the relaxed fiber to an average sarcomere length of 2.6 µm, the sarcomeres indicated by asterisks are at 5.1 µm and 3.8 µm while sarcomeres in series are shortened to ~1.8 µm. Scale bar is 1.0 µm. Source: Modified from Macpherson et al., 1997. With permission.

of soreness, the most quantitative and reproducible measure of the totality of a muscle injury is the decrease in the ability of the muscle to develop force [McCully and Faulkner, 1985; Newham et al., 1983].

Repair of Injured Skeletal Muscles Under circumstances when the injury involves only minor disruptions of the thick or thin filaments of single sarcomeres, the damaged molecules are likely replaced by newly synthesized molecules available in the cytoplasmic pool [Russell et al., 1992]. In addition, contraction-induced disruptions of the sarcolemma are often transient and repaired spontaneously, allowing survival of the fiber [McNeil and Khakee, 1992]. Following more severe injuries, complete regeneration of the entire muscle fiber will occur. Satellite Cell Activation A key element in the initiation of muscle fiber regeneration following a wide variety of injuries is the activation of satellite cells [Carlson and Faulkner, 1983]. Satellite cells are quiescent myogenic stem cells located between the basal lamina and the sarcolemma (Fig. 123.2). Upon activation, satellite cells divide


FIGURE 123.4 Schematic representation of the cellular responses during the processes of degeneration and regeneration following transplantation of extensor digitorum longus muscles in rats and cats. The diagram is divided into segments that represent the histological appearance of the muscle cross-section at various times after transplantation. The times given in days refer to rat muscles and those given in weeks refer to the larger cat muscles. The letters indicate groups of muscle fibers with similar histological appearances. A: surviving fibers, B: fibers in a state of ischemic necrosis, C: fibers invaded by phagocytic cells, D: myoblasts and early myotubes, E: early myofibers, F: immature regenerating fibers, G: mature regenerated fibers, H: normal control muscle fibers [Carlson and Faulkner, 1983]. Source: Mauro, A. 1979. Muscle Regeneration, p. 493-507, Raven Press, New York, NY. With permission.

mitotically to give rise to myoblasts. The myoblasts can then fuse with existing muscle fibers, acting as a source of new myonuclei. This is the process by which a muscle fiber increases the total number of myonuclei in fibers that are increasing in size [Moss and Leblond, 1971] and may be necessary to repair local injuries. Alternatively, the myoblasts can fuse with each other to form multinucleated myotubes inside the remaining basal lamina of the degenerated fibers [Carlson and Faulkner, 1983]. The myotubes then begin to produce muscle specific proteins and ultimately differentiate completely into adult fast or slow muscle fibers [Carlson and Faulkner, 1983]. Recent evidence also suggests that in addition to the activation of resident satellite cells, regenerating muscle may recruit undifferentiated myogenic precursor cells from other sources [Ferrari et al., 1998] Following a closed contusion injury, mitotic activation of satellite cells has been observed within the first day after the injury and is correlated in time with the appearance of phagocytes and newly formed capillaries [Hurme and Kalimo, 1992]. Similarly, DNA synthesis by the satellite cells is observed within the first day following crush injuries [Bischoff, 1986] and exercise-induced injuries [Darr and Schultz, 1987] in rats. These observations are consistent with the hypothesis that the factors that activate satellite cells may be endogenous to the injured tissue itself or synthesized and secreted by platelets at the wound site, infiltrating neutrophils, and macrophages [reviewed in Husmann et al., 1996]. The primary candidates for the factors that activate and regulate satellite cell function include the fibroblast growth factors (FGFs), platelet-derived growth factor (PDGF), transforming growth factor beta (TGF-β), and insulinlike growth factors I and II (IGF-I, II). The effects of these factors on muscle satellite cell proliferation and differentiation have been studied extensively in cell culture [reviewed in Florini and Magri, 1989]. FGF is a powerful mitogen for myogenic cells from adult rat skeletal muscle, but a potent inhibitor of terminal differentiation, i.e., myoblast fusion and expression of the skeletal muscle phenotype. PDGF also shows a strong stimulating effect on proliferation and inhibitory effect on differentiation of satellite cells [Jin et al., 1990]. Similarly, the presence


of TGF-β prevents myotube formation as well as muscle-specific protein synthesis by rat embryo myoblasts and by adult rat satellite cells. In contrast to the previously mentioned growth factors, all of which inhibit myoblast differention, IGFs stimulate both proliferation and differentiation of myogenic cells [Ewton and Florini, 1980]. Despite our extensive knowledge of the actions of many individual growth factors in vitro and in vivo, the interactions between different growth factors have been less thoroughly investigated [Rosenthal et al., 1991]. A better understanding of the interactions between growth factors and the mechanisms that guide satellite cells through the regeneration process is necessary. Myogenic Regulatory Factors The conversion of pluripotent embryonic stem cells to differentiated muscle cells involves the commitment of these cells to the myogenic lineage and the subsequent proliferation, differentiation and fusion to form multinucleated myotubes and ultimately mature muscle cells. New muscle cell formation from muscle satellite cells resembles embryonic development in the sense that in regenerating muscle cells embryonic isoforms of the muscle proteins are expressed [Whalen et al., 1985]. The conversion of stem cells to mature fibers, in both developing muscle and regenerating muscle is directed by a group of related regulatory factors. These so-called muscle regulatory factors (MRFs) are part of a superfamily of basic helix-loop-helix DNA binding proteins that interact to regulate the transcription of skeletal muscle genes [reviewed in Weintraub et al., 1991]. MyoD was the first MRF identified followed by three other related genes including myogenin, MRF4 (also called herculin or Myf-6), and Myf-5 [Weintraub et al., 1991]. The observations that each of the MRFs has the ability to independently convert cultured fibroblasts into myogenic cells led to the original conclusion that functionally the MRFs were largely redundant. Subsequent gene targeting experiments, in which null mutations were introduced in each of the MRF genes, support separate and distinct roles in myogenesis for each MRFs [reviewed in Rudnicki and Jaenisch, 1995]. For example, mice that lack either Myf-5 or MyoD apparently have normal skeletal muscle but deletion of both genes results in the complete absence of skeletal myoblasts [Rudnicki et al., 1992; 1993; Braun et al., 1992]. While these observations suggest that Myf-5 and MyoD do have overlapping functions, characterization of the temporal-spatial patterns of myogenesis in Myf-5- and MyoDdeficient mouse embryos support the hypothesis that during normal development Myf-5 and MyoD primarily regulate epaxial (paraspinal and intercostal) and hypaxial (limb and abdominal wall) muscle development, respectively [Kablar et al., 1997]. In mice lacking myogenin, the number of myoblasts is not different from that of control mice, but skeletal muscles in myogenin-deficient mice display a marked reduction in the number of mature muscle fibers [Hasty et al., 1993; Nabeshima et al., 1993]. In summary, the MRF family can be divided into two functional groups. MyoD and Myf-5 are referred to as primary factors and are required for the determination of skeletal myoblasts whereas myogenin and MRF4 are secondary factors that act later and are necessary for differentiation of myoblasts into myotubes. How the MRFs control the series of events required for myoblast determination and differentiation are important questions for the future. In addition, the roles played by this family of proteins in regeneration, adaptation, and changes in skeletal muscle with aging are areas of active investigation [Jacobs-El et al., 1995; Marsh et al., 1997; Megeney et al., 1996].

123.5 Reconstructive Surgery of Whole Skeletal Muscles When an injury or impairment is so severe that the total replacement of the muscle is required, a whole donor muscle must be transposed or transplanted into the recipient site [Faulkner et al., 1994a]. One of the most versatile muscles for transpositions is the latissimus dorsi (LTD) muscle. LTD transfers have been used in breast reconstruction [Moelleken et al., 1989], to restore elbow flexion, and to function as a heart assist pump [Carpentier and Chachques, 1985]. Thompson [1974] popularized the use of small free standard grafts to treat patients with partial facial paralysis. The transplantation of large skeletal muscles in dogs with immediate restoration of blood flow through the anastomosis of the artery and vein provided an operative technique with numerous applications [Tamai et al., 1970]. Coupled with cross-face nerve grafts, large skeletal muscles are transplanted with microneurovascular repair to correct


deficits in the face [Harii et al., 1976] and adapted for reconstructive operations to treat impairments in function of the limbs, anal and urinary sphincters, and even the heart [Freilinger and Deutinger, 1992]. Transposition and transplantation of muscles invariably results in structural and functional deficits [Guelinckx et al., 1992]. The deficits are of the greatest magnitude during the first month and then a gradual recovery results in the stabilization of structural and functional variables between 90 and 120 days [Guelinckx et al., 1992]. In stabilized vascularized grafts ranging from 1 to 3 grams in rats to 90 grams in dogs, the major deficits are a ~30% decrease in muscle mass and in most grafts a ~40% decrease in maximum force [Faulkner et al., 1994a]. The decrease in power is more complex since it depends on both the average shortening force and the velocity of shortening. As a consequence, the deficit in maximum power may be either greater, or less than the deficit in maximum force [Kadhiresan et al., 1993]. Tenotomy and repair are major factors responsible for the deficits [Guelinckx et al., 1988]. When a muscle is transplanted to act synergistically with other muscles, the action of the synergistic muscles may contribute to the deficits observed [Miller et al., 1994]. Although the data are limited, skeletal muscle grafts appear to respond to training stimuli in a manner not different from that of control muscles [Faulkner et al., 1994a]. The training stimuli include traditional methods of endurance and strength training [Faulkner et al., 1994b], as well as chronic electrical stimulation [Pette and Vrbova, 1992]. In spite of the deficits, transposed and transplanted muscles develop sufficient force and power to function effectively to maintain posture and patent sphincters and to move limbs or drive assist devices in parallel or in series with the heart [Faulkner et al., 1994a].

123.6 Myoblast Transfer and Gene Therapy Myoblast transfer and gene therapy are aimed at delivering exogenous genetic constructs to skeletal muscle cells. The implications of myoblast and gene therapy hold great promise for skeletal muscle research and for those afflicted with inherited myopathies such as Duchenne and Becker muscular dystrophy (DMD and BMD). Myoblast transfer is a cell-mediated technique designed to treat inherited myopathies by intramuscular injection of myoblasts containing a normal functional genome. The goal is to correct for a defective or missing gene in the myopathic tissue through the fusion of normal myoblasts with growing or regenerating diseased cells. Gene therapy presents a more complex and flexible approach, whereby genetically engineered DNA constructs are delivered to a host cell to specifically direct production of a desired protein. By re-engineering the coding sequence of the gene and its regulatory regions, the function, the quantity of expression, and the protein itself can be altered. For many years, cells have been genetically altered to induce the production of a variety of useful proteins, such as human growth hormone and interferon. As a focus in skeletal muscle tissue engineering, DMD is an X-linked recessive disorder characterized by progressive muscle degeneration resulting in debilitating muscle weakness and death in the second or third decade of life as a result of respiratory failure [Emery, 1988]. DMD and the milder BMD are due to genetic defects that lead to the absence or marked deficiency in the expression or functional stability of the protein dystrophin [Bonilla et al., 1988; Hoffman et al., 1988]. Similarly, a mutation in the dystrophin gene leads to the complete absence of the protein in muscle and brain tissues of the mdx mouse [Bulfield et al., 1984; Sicinski et al., 1989]. The homology of the genetic defects in DMD patients and mdx mice support the use of mdx mice as a model of dystrophin deficiency to explore the processes of the dystrophic disease and test proposed therapies or cures.

Myoblast Transfer Therapy The concept of myoblast transfer is based on the role satellite cells play in muscle fiber growth and repair [Carlson and Faulkner, 1983]. As a therapy for DMD, the idea is to obtain satellite cells containing a functional dystrophin gene from a healthy compatible donor, have the cells multiply in culture, and then inject the “normal” myoblasts into the muscles of the patient. The objective is for the injected myoblasts


to fuse with growing or regenerating muscle fibers to form a mosaic fiber in which the cytoplasm will contain normal myoblast nuclei capable of producing a functional form of dystrophin. Experiments involving mdx mice have had varying success. Several investigators have reported that implantation of healthy myoblasts into muscles of mdx mice led to the production of considerable quantities of dystrophin [Morgan et al., 1990; Partridge et al., 1989]. Others found that myoblasts injected into limb muscles of mdx and control host mice showed a “rapid and massive” die off shortly after injection [Fan et al. 1996] with large and permanent decreases in muscle mass and maximum force [Wernig et al., 1995]. The successful transfer and fusion of donor myoblasts may be enhanced by X-ray irradiation of mdx muscles prior to myoblast injection to prevent the proliferation of myoblasts endogenous to the host and encourage the growth of donor myoblasts [Morgan et al., 1990]. In contrast to the studies with mice, delivery of myoblasts to DMD patients has shown very low levels of fusion efficiency, transient expression of dystrophin, and immune rejection [Gussoni et al., 1992; Karpati et al., 1993; Mendell et al., 1995; Morgan, 1994]. The use of X-ray irradiation in an attempt to enhance transfer efficiency is not applicable to DMD boys due to substantial health risks. Furthermore, immunosuppression may be necessary to circumvent immune rejection, which carries risks of its own. A better understanding of the factors that govern the survival, fusion, and expression of donor myoblasts is required before the viability of myoblast transfer as a treatment of DMD can be evaluated.

Gene Therapy The aim of gene therapy for DMD is to transfer a functional dystrophin gene directly into the skeletal muscle. The challenge behind gene therapy is not only obtaining a functional genetic construct of the dystrophin gene and regulatory region but the effective delivery of the gene to the cell’s genetic machinery. Methods to transfer genetic material into a muscle cell include direct injection, and the use of retroviral and adenoviral vectors. Each of these strategies presents highly technical difficulties that to date remain unresolved. Transgenic Mice To explore the feasibility of gene therapy for DMD, Cox and colleagues [1993] examined the introduction of an exogenous dystrophin gene into the germ line of mdx mice to produce transgenic animals. The transgenic mdx mice expressed nearly 50 times the level of endogenous dystrophin found in muscles of control C57BL/10 mice and displayed the complete absence of any morphological, immunohistological, or functional symptoms of the murine muscular dystrophy with no deleterious side effects. Although transgenic technology does not provide an appropriate means for treating humans, these results demonstrated the efficacy of gene therapy to correct pathological genetic defects such as DMD. Direct Intramuscular Injection The straightforward gene delivery method of direct injection of plasmid DNA into skeletal and heart muscles [Lin et al., 1990; Wolff et al., 1990] has been proposed as a treatment for DMD and BMD. The idea is that dystrophic cells will incorporate the genetic constructs, whereby the genes will use the cell’s internal machinery to produce the protein dystrophin. The advantages of direct injection of DNA as a gene delivery system are its simplicity, and it presents no chance of viral infection or the potential of cancer development that can occur with viral vectors [Morgan, 1994]. Although this approach is appealing in principal, direct intramuscular injection of human dystrophin plasmid DNA into the quadriceps muscles of mdx mice led to the expression of human dystrophin in only 1% to 3% of the muscle fibers [Acsadi et al., 1991]. In order for this method to be clinically effective, a much larger number of transfected myofibers must be achieved. Retrovirus-Mediated Gene Transfer Retroviruses reverse the normal process by which DNA is transcribed into RNA. A single-stranded viral RNA genome enters a host cell and a double helix comprised of two DNA copies of the viral RNA is


created by the enzyme reverse transcriptase. Catalyzed by a viral enzyme, the DNA copy then integrates into a host cell chromosome where transcription, via the host cell RNA polymerase, produces large quantities of viral RNA molecules identical to the infecting genome [Alberts et al., 1989]. Eventually, new viruses emerge and bud from the plasma membrane ready to infect other cells. Consequently, retroviral vectors used for gene therapy are, by design, rendered replication defective. Once they infect the cell and integrate into the genome, they cannot make functional retroviruses to infect other cells. After the infective process is completed, cells are permanently altered with the presence of the viral DNA that causes the synthesis of proteins not originally endogenous to the host cell. A primary obstacle to the efficiency of a retroviral gene delivery system is its dependence on host cell division. The requirement that a cell must be mitotically active for the virus to be incorporated into the host cell genetic machinery [Morgan, 1994] presents a problem for skeletal muscle tissue engineering since skeletal muscle is in a post-mitotic state. Nonetheless, as described above, recovery from injury in skeletal muscle involves the proliferation of myogenic precursor cells and either the incorporation of these cells into existing muscle fibers or the fusion of these cells with each other to form new fibers. Furthermore, since degeneration and regeneration of muscle cells are ongoing in DMD patients, viral transfection of myoblasts may be an effective route for gene delivery to skeletal muscle for treatment of DMD. Another limitation of the retroviral gene delivery system is the small carrying capacity of the retroviruses of approximately 7 kilobases. This size limitation precludes the delivery of a full-length dystrophin construct of 12-14 kilobases [Dunckley et al., 1993; Morgan, 1994]. A 6.3 kilobase dystrophin construct, containing a large in-frame deletion resulting in the absence of ~40% of the central domain of the protein, has received a great deal of attention since its discovery in a BMD patient expressing a very mild phenotype. A single injection of retrovirus containing the Becker dystrophin minigene into the quadriceps or tibialis anterior muscle of mdx mice led to the sarcolemmal expression of dystrophin in an average of 6% of the myofibers [Dunckley et al., 1993]. Restoration of the 43-kDalton dystrophin-associated glycoprotein was also observed and expression of the recombinant dystrophin was maintained for up to nine months. Transduction of the minigene was significantly enhanced when muscles were pretreated with an intramuscular injection of the myotoxic agent bupivacaine to experimentally induce muscle regeneration. Adenovirus-Mediated Gene Transfer Adenoviruses have many characteristics that make the adenovirus-mediated gene transfer the most promising technology for gene therapy of skeletal muscle. The primary advantages are the stability of the viruses allowing them to be prepared in large amounts and the ability of adenoviral vectors to infect nondividing or slowly proliferating cells. In addition, adenoviral vectors have the potential to be used for systemic delivery of exogenous DNA. Through the use of tissue-specific promoters, specific tissues such as skeletal muscle may be targeted for transfection via intravenous injection. Initial studies using adenoviral vectors containing the Becker dystrophin minigene driven by the Rous Sarcoma Virus promoter demonstrated that after a single intramuscular injection in newborn mdx mice, 50% of muscle fibers contained dystrophin [Ragot et al., 1993]. The truncated dystrophin was correctly localized to the sarcolemmal membrane and appeared to protect myofibers from the degeneration process characteristic of mdx muscles [Vincent et al., 1993]. Six months after a single injection, expression of the minigene was still observed in the treated muscle. More recently, these same investigators demonstrated that the injection of the adenoviral vector containing the dystrophin minigene into limb muscles of newborn mdx mice provided protection from the fiber damage and force deficit associated with a protocol of pliometric contractions that was administered at four months of age [Deconinck et al., 1996]. Despite the promise of adenovirus-mediated gene therapy, a number of limitations to its usefulness remain to be resolved. One major drawback of the system is the relatively brief duration of transgene expression observed following injection of adenoviral vectors into adult immunocompetent animals [Kass-Eisler et al., 1994]. The lack of long-term exogenous gene expression is likely the result of low level expression of endogenous viral proteins triggering an inflammatory response that attacks infected cells [Yang et al., 1994]. In addition to the difficulties resulting from the potent immune response triggered


by adenovirus, direct cytotoxic effects of adenovirus injection on skeletal muscle have been reported [Petrof et al., 1995]. Current generation adenoviruses are also limited by their relatively small cloning capacity of ~8 kilobases. The development of adenoviral vectors with increased cloning capacity, the ability to evade host immune rejection, and no toxic effects are areas of active investigation [KumarSingh and Chamberlain, 1996; Petrof et al., 1996; Hauser et al., 1997].

Acknowledgments The research in our laboratory and the preparation of this chapter was supported by grants from the United States Public Health Service, National Institute on Aging, AG-15434 (SVB) and AG-06157 (JAF) and the Nathan Shock Center for Basic Biology of Aging at the University of Michigan.

References Acsadi, G., Dickson, G., Love, D.R., Jani, A., Walsh, F.S., Gurusinghe, A., Wolff, J.A., and Davies, K.E. 1991. Human dystrophin expression in mdx mice after intramuscular injection of DNA constructs. Nature 352:815-818. Alberts, B., Bray, D., Lewis, J., Raff, M., Roberts, K., and Watson, J.D. 1989. Molecular Biology of the Cell, 2nd ed. Garland Publishing, Inc., New York, NY, p. 254. Bischoff, R. 1986. A satellite cell mitogen from crushed adult muscle. Dev. Biol. 115:140-147. Bonilla, E., Samitt, C.E., Miranda, A.F., Hays, A.P., Salviati, G., DiMauro, S., Kunkel, L.M., Hoffman, E.P., and Rowland, L.P. 1988. Duchenne muscular dystrophy: deficiency of dystrophin at the muscle cell surface. Cell 54:447-452. Braun, T., Rudnicki, M.A., Arnold, H.H., and Jaenisch, R. 1992. Targeted inactivation of the muscle regulatory gene Myf-5 results in abnormal rib development and perinatal death. Cell 71:369-382. Bulfield, G., Siller, W.G., Wight, P.A.L., and Moore, K.J. 1984. X chromosome-linked muscular dystrophy (mdx) in the mouse. Proc. Natl. Acad. Sci. USA 81:1189-1192. Burke, R.E., Levin, D.N., Tsairis, P. and Zajac, F.E., III. 1973. Physiological types and histochemical profiles in motor units of the cat gastrocnemius muscle. J. Physiol. (Lond.) 234:723-748. Carlson, B.M. and Faulkner, J.A. 1983. The regeneration of skeletal muscle fibers following injury: a review. Med. Sci. Sports Exer. 15:187-198. Carpentier, A. and J.C. Chachques. 1985. Myocardial substitution with a stimulated skeletal muscle: first successful clinical case. Lancet 1(8440):1267. Cox, G.A., Cole, N.M., Matsumura, K., Phelps, S.F., Hauschka, S.D., Campbell, K.P., Faulkner, J.A., and Chamberlain, J.S. 1993. Overexpression of dystrophin in transgenic mdx mice eliminates dystrophic symptoms without toxicity. Nature 364:725-729. Darr, K.C. and Schultz, E. 1987. Exercise-induced satellite cell activation in growing and mature skeletal muscle. J. Appl. Physiol. 63:1816-1821. Deconinck, N., Ragot, T., Maréchal, G., Perricaudet, M., and Gillis, J.M. 1996. Functional protection of dystrophic mouse (mdx) muscles after adenovirus-mediated transfer of a dystrophin minigene. Proc. Natl. Acad. Sci. USA 93:3570-3574. Dunckley, M.G., Wells, D.J., Walsh, F.S., and Dickson, G. 1993. Direct retroviral-mediated transfer of a dystrophin minigene into mdx mouse muscle in vivo. Hum. Mol. Genet. 2:717-723. Emery, A.E.H. 1988. Duchenne Muscular Dystrophy, 2nd ed. Oxford University Press, New York, NY. Ewton, D.A., and Florini, J.R. 1980. Relative effects of the somatomedins, multiplication-stimulating activity, and growth hormone on myoblast and myotubes in culture. Endocrinol. 106:577-583. Fan, Y., Maley, M., Beilharz, M., and Grounds, M. 1996. Rapid death of injected myoblasts in myoblast transfer therapy. Muscle Nerve 19:853-860. Faulkner, J.A., Carlson, B.M., and Kadhiresan, V.A. 1994a. Whole muscle transplantation: Mechanisms responsible for functional deficits. Biotech. and Bioeng. 43:757-763.


Faulkner, J.A., Green, H.J., and White, T.P. 1994b. Response and adaptation of skeletal muscle to changes in physical activity. In Physical Activity Fitness and Health, ed. C. Bouchard, R.J. Shephard, and T. Stephens, p. 343-357. Human Kinetics Publishers, Champaign, IL. Ferrari, G., Cusella-De Angelis, G., Coletta, M., Paolucci, E., Stornaiuolo, A., Cossu, G., and Mavilio, F. 1998. Muscle regeneration by bone marrow-derived myogenic progenitors. Science 279:1528-1530. Florini, J.R. and Magri, K.A. 1989. Effects of growth factors on myogenic differentiation. Am. J. Physiol. 256 (Cell Physiol. 25):C701-C711. Freilinger, G. and Deutinger, M. 1992. Third Vienna Muscle Symposium, Blackwell-MZV, Vienna, Austria. Guelinckx, P.J, Faulkner, J.A., and Essig, D.A. 1988. Neurovascular-anastomosed muscle grafts in rabbits: Functional deficits result from tendon repair. Muscle Nerve 11:745-751. Guelinckx, P.J., Carlson, B.M., and Faulkner, J.A. 1992. Morphologic characteristics of muscles grafted in rabbits with neurovascular repair. J. Recon. Microsurg. 8:481-489. Gussoni, E., Pavlath, G.K., Lancot, A.M., Sharma, K.R., Miller, R.G., Steinman, L., and Blau, H.M. 1992. Normal dystrophin transcrips detected in Duchenne muscular dystrophy patients after myoblast transplantation. Nature 356:435-438. Harii, K., Ohmori, K., Torii, S. 1976. Free gracilis muscle transplantation with microneurovascular anastomoses for the treatment of facial paralysis. Plast. Recon. Surg. 57:133-143. Hasty, P., Bradley, A., Morris, J.H., Edmondson, J.M., Venuti, J.M., Olson, E.N., and Klein, W.H. 1993. Muscle deficiency and neonatal death in mice with a targeted mutation in the myogenin gene. Nature 364:501-506. Hauser, M.A., Amalfitano, A., Kumar-Singh, R., Hauschka, S.D., and Chamberlain, J.S.1997. Improved adenoviral vectors for gene therapy of Duchenne muscular dystrophy. Neuromusc. Dis. 7:277-283. Henneman, E., Somjen, G., and Carpenter, D. 1965. Functional significance of cell size in spinal motor neurons. J. Neurophysiol. 28:560-580. Hoffman, E.P., Fischbeck, R.H., Brown, R.H., Johnson, M., Medori, R., Loike, J.D., Harris, J.B., Waterson, R., Brooke, M., Specht, L., Kupsky, W., Chamberlain, J., Caskey, C.T., Shapiro, F., and Kunkel, L.M. 1988. Characterization of dystrophin in muscle-biopsy specimens from patients with Duchenne’s or Becker’s muscular dystrophy. N. Eng. J. Med. 318:1363-1368. Hurme, T. and Kalimo, H. 1992. Activation of myogenic precursor cells after muscle injury. Med. Sci. Sports Exer. 24:197-205. Husmann, I., Soulet, L., Gautron, J., Martelly, I., and Barritault, D. 1996. Growth factors in skeletal muscle regeneration. Cytokine Growth Factor Rev. 7:249-258. Jacobs-El, J., Zhou, M.Y., and Russell, B. 1995. MRF4, Myf-5, and myogenin mRNAs in the adaptive responses of mature rat muscle. Am. J. Physiol. 268:C1045-C1052. Jin, P., Rahm, M., Claesson-Wesh, L., Heldin, C.-H., Sejerson, T. 1990. Expression of PDGF A chain and β-receptor genes during rat myoblast differentiation. J. Cell Biol. 110:1665-1672. Jones, D.A., Jackson, M.J., McPhail, G., and Edwards, R.H.T. 1984. Experimental mouse muscle damage: the importance of external calcium. Clin. Sci. 66:317-322. Kablar, B., Krastel, K., Ying, C., Asakura, A., Tapscott, S.J., and Rudnicki, M.A. 1997. MyoD and Myf-5 differentially regulate the development of limb versus trunk skeletal muscle. Development 124:4729-4738. Kadhiresan, V.A., Guelinckx, P.J., and Faulkner, J.A. 1993. Tenotomy and repair of latissimus dorsi muscles in rats: implications for transposed muscle grafts. J. Appl. Physiol. 75:1294-1299. Karpati, G., Ajdukovic, D., Arnold, D., Gledhill, R.B., Guttmann, R., Holland, P., Koch, P.A., Shoubridge, E., Spence, D., Vanasse, M., Watters, G.V., Abrahamowicz, M., Duff, C., and Worton, R.G. 1993. Myoblast transfer in Duchenne muscular dystrophy Ann. Neurol. 34:8-17. Kass-Eisler, A., Falck-Pedersen, E., Elfenbein, D.H., Alvira, M., Buttrick, P.M., and Leinwand, L.A. 1994. The impact of developmental stage, route of administration and the immune system on adenovirusmediated gene transfer. Gene Therapy 1:395-402. Kumar-Singh, R., and Chamberlain, J.S. 1996. Encapsidated adenovirus minichromosomes allow delivery and expression of a 14 kb dystrophin cDNA to muscle cells. Hum. Mol. Genet. 5:913-921.


Lin, H., Parmacek, M.S., Morle, G., Bolling, S., and Leiden, J.M. 1990. Expression of recombinant genes in myocardium in vivo after direct injection of DNA. Circulation 82:2217-2221. Macpherson, P.C.D., Dennis, R.G. and Faulkner, J.A. 1997. Sarcomere dynamics and contraction-induced injury to maximally activated single muscle fibres from soleus muscles of rats. J. Physiol. (Lond.), 500:523-533. Marsh, D.R., Criswell, D.S., Carson, J.A., and Booth, F.W. 1997. Myogenic regulatory factors during regeneration of skeletal muscle in young, adult, and old rats. J. Appl. Physiol. 83:1270-1275. McCully, K.K.and Faulkner, J.A. 1985. Injury to skeletal muscle fibers of mice following lengthening contractions. J. Appl. Physiol. 59:119-126. McNeil, P.L. and Khakee, R. 1992. Disruptions of muscle fiber plasma membranes. Role in exerciseinduced damage. Am. J. Path. 140:1097-1109. Megeney, L.A., Kablar, B., Garrett, K., Anderson, J.E., and Rudnicki, M.A. 1996. MyoD is required for myogenic stem cell function in adult skeletal muscle. Genes & Develop. 10:1173-1183. Mendell, J.R., Kissel, J.T., Amato, A.A., King, W., Signore, L, Prior, T.W., Sahenk, Z., Benson, S., McAndrew, P.E., Rice, R., Nagaraja, H., Stephens, R., Lantry L., Morris, G.E., and Burghes A.H.M. 1995. Myoblast transfer in the treatment of Duchenne’s muscular dystrophy. New Engl. J. Med. 333:832-838. Miller, S.W., Hassett, C.A., White T.P., and Faulkner, J.A. 1994. Recovery of medial gastrocnemius muscle grafts in rats: implications for the plantarflexor group. J. Appl. Physiol., 77:2773-2777. Moelleken, B.R.W., Mathes, S.A., and Chang, N. 1989. Latissimus dorsi muscle-musculocutaneous flap in chest-wall reconstruction. Surgical Clinics of North America 69(5):977-990. Morgan, J.E. 1994. Cell and gene therapy in Duchenne muscular dystrophy. Hum. Gene Ther. 5:165-173. Morgan, J.E., Hoffman, E.P., and Partridge, T.A. 1990. Normal myogenic cells from newborn mice restore normal histology to degenerating muscles of the mdx mouse. J. Cell. Biol. 111:2437-2449. Moss, F.P. and Leblond, C.P. 1971. Satellite cells as the source of nuclei in muscles of growing rats. Anat. Rec. 170:421-436. Nabeshima, Y., Hanaoka, K., Hayasaka, M., Esumi E., Li, S., Nonaka, I., and Nabeshima, Y. 1993. Myogenin gene disruption results in perinatal lethality because of a severe muscle defect. Nature 364:532-535. Newham, D.J., McPhail, G., Jones, D.A., and Edwards, R.H.T. 1983. Large delayed plasma creatine kinase changes after stepping exercise. Muscle Nerve 6:380-385. Partridge, T.A., Morgan. J.E., Coulton, G.R., Hoffman, E.P., and Kunkel, L.M. 1989. Conversion of mdx myofibers from dystrophin-negative to -positive by injection of normal myoblasts. Nature 337:176-179. Petrof, B.J., Acsadi, G., Jani, A., Bourdon, J., Matusiewicz, N., Yang, L., Lochmüller, H., and Karpati, G. 1995. Efficiency and functional consequences of adenovirus-mediated in vivo gene transfer to normal and dystrophic (mdx) mouse diaphragm. Am. J. Respir. Cell Mol. Biol. 13:508-517. Petrof, B.J., Lochmüller, H., Massie, B., Yang, L., Macmillan, C., Zhao, J.-E., Nalbantoglu, J., and Karpati, G. 1996. Impairment of force generation after adenovirus-mediated gene transfer to muscle is alleviated by adenoviral gene inactivation and host CD8+ T cell deficiency. Human Gene Ther. 7:1813-1826. Pette, D., and Vrbova, G. 1992. Adaptation of mammalian skeletal muscle fibers to chronic electrical stimulation. Rev. Physiol. Biochem. Pharmcol., 120:115-202. Ragot, T., Vincent, N., Chafey, P., Vigne, E., Gilgenkrantz, H., Couton, D., Cartaud, J., Briand, P., Kaplan, J., Perricaudet, M., and Kahn, A. 1993. Efficient adenovirus-mediated transfer of a human minidystrophin gene to skeletal muscle of mdx mice. Nature 361:647-650. Rosenthal, S.M., Brown, E.J., Brunetti, A., and Goldfine, J.D.1991. Fibroblast growth factor inhibits insulin-like growth factor-II (IGF-II) gene expression and increases IGF-I receptor abundance in BC3H-1 muscle cells. Mol. Endocrinol. 5:678-684. Rudnicki, M.A., Braun, T., Hinuma, S., and Jaenisch, R. 1992. Inactivation of MyoD in mice leads to upregulation of the myogenic HLH gene Myf-5 and results in apparently normal muscle development. Cell 71:383-390. Rudnicki, M.A., and Jaenisch, R. 1995. The MyoD family of transcription factors and skeletal myogenesis. BioEssays 17:203-209.


Rudnicki, M.A., Schnegelsberg, P.N.J., Stead, R.H., Braun, T., Arnold, H.H., and Jaenisch, R. 1993. MyoD or Myf-5 is required for the formation of skeletal muscle. Cell 75:1351-1359. Russell, B., Dix, D.J., Haller, D.L., and Jacobs-El, J. 1992. Repair of injured skeletal muscle: a molecular approach. Med. Sci. Sports Exer. 24:189-196. Sicinski, P., Geng, Y., Ryder-Cook, A.S., Barnard, E.A., Darlison, M.G., and Barnard, P.J. 1989. The molecular basis of muscular dystrophy in the mdx mouse: A point mutation. Science 244:1578-1580. Tamai, S., Komatsu, S., Sakamoto, H., Sano, S., and Sasauchi, N. 1970. Free-muscle transplants in dogs with microsurgical neuro-vascular anastomoses. Plast. Recon. Surg. 46:219-225. Thompson, N. 1974. A review of autogenous skeletal muscle grafts and their clinical applications. Clin. Plast. Surg. 1:349-403. Vincent, N., Ragot, T, Gilgenkrantz, H., Couton, D., Chafey, P., Gregoire, A., Briand, P., Kaplan, J., Kahn, A., and Perricaudet, M. 1993. Long-term correction of mouse dystrophic degeneration by adenovirus-mediated transfer of a minidystrophin gene. Nature Genetics 5:130-134. Weintraub, H., Davis, R., Tapscott, S., Thayer, M., Krause, M., Benezra, R., Blackwell, T.K., Turner, D., Rupp, R., Hollenberg, S., Zhuang, Y., and Lassar, A. 1991. The myoD gene family: nodal point during specification of the muscle cell lineage. Science 251:761-766. Wernig, A., Irintchev, A., and Lange, G. 1995. Functional effects of myoblast implantation into histoincompatible mice with or without immunosuppression. J. Physiol. (Lond.) 484:493-504. Whalen, R.G., Butler-Browne, G.S., Bugaisky, L.B., Harris, J.B., and Herliocoviez, D. 1985. Myosin isozyme transitions in developing and regenerating rat muscle. Adv. Exp. Med. Biol. 182:249-257. Wolff, J.A., Malone, R.W., Williams, P., Chong, W., Acsadi, G., Jani, A., and Felgner, P.L. 1990. Direct gene transfer into mouse muscle in vivo. Science 247:1465-1468. Yang, Y., Nunes, F.A., Berencsi, K., Gönczöl, E., Englelhardt, J.F., and Wilson, J.M. 1994. Inactivation of E2a in recombinant adenoviruses improves the prospect for gene therapy in cystic fibrosis. Nature Genet. 7:362-369.


Freed, L.E., Vunjak-Novakovic, G. “Tissue Engineering of Cartilage.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

124 Tissue Engineering of Cartilage 124.1 124.2

Scope Cell-Based Approaches to Cartilage Tissue Engineering

124.3

Cell-Polymer-Bioreactor System

L.E. Freed Harvard University

In vivo Cartilage Repair • In vitro Chondrogenesis

G. Vunjak-Novakovic Massachusetts Institute of Technology

Experimental Methods • Developmental Studies • Modulation of Cartilaginous Structure and Function

124.4

Summary and Future Directions

124.1 Scope This chapter reviews the current state-of-the-art of articular cartilage tissue engineering, and focuses on a cell-polymer-bioreactor model system which can be used for controlled studies of chondrogenesis and the modulation of engineered cartilage by environmenal factors.

124.2 Cell-Based Approaches to Cartilage Tissue Engineering Articular cartilage is an avascular tissue that contains only one cell type (the chondrocyte), and has a very limited capacity for self-repair. The chondrocytes are responsible for the synthesis and maintenance of their extracellular matrix (ECM), which is composed of a hydrated collagen network (~60% of the tissue dry weight, dw), a highly charged proteoglycan gel (PG, ~25% dw), and other proteins and glycoproteins (~15% dw).1 Its high water content (70 to 80% of the tissue wet weight, ww) enables cartilage to withstand the compressive, tensile and shear forces associated with joint loading. None of the methods conventionally used for cartilage repair (e.g., tissue auto- or allografts) can predictably restore a durable articular surface to an osteoarthrotic joint.2 Cell-based therapies, i.e., implantation of cells or engineered cartilage, represent an alternative approach to articular cartilage repair. Figure 124.1 shows the cell-polymer-bioreactor system for cartilage tissue engineering. Constructs based on chondrogenic cells and polymer scaffolds are cultured in bioreactors to form 3-dimensional (3D) cartilaginous tissues. Engineered cartilage can be used either in vivo, to study articular cartilage repair or in vitro, for controlled studies of cell and tissue-level responses to molecular, mechanical, and genetic manipulations. Table 124.1 lists selected, representative examples of recent articular cartilage tissue engineering studies in which the stated research goals were either in vivo cartilage repair or in vitro studies of chondrogenesis. Several key parameters varied from study to study: (i) cell source and expansion in culture, (ii) scaffold material and structure, (iii) in vitro cultivation (conditions and duration), (iv) additional components (e.g., perichondrium), (v) experimental animal (species and age), (vi) surgical model (defect location


FIGURE 124.1 Cell-based approach to articular cartilage repair. Isolated cells, e.g., chondrocytes or bone marrow stromal cells (BMSC) are seeded onto 3-dimensional scaffolds, e.g., polymers formed as fibrous meshes or porous sponges, to form cell-polymer constructs. Constructs are first cultured in vitro, e.g., in bioreactors, and then used either as implants or for in vitro research.

and dimensions), (vii) graft type (autograft or allograft), (viii) duration of in vivo implantation, and (viii) methods used to assess the resulting tissue (e.g., histological, biochemical, and/or mechanical).

In Vivo Cartilage Repair Brittberg et al.3 reported a new cell-based procedure for the repair of human knee injuries that could potentially eliminate the need for more than half a million arthroplastic procedures and joint replacements performed annually in the United States. Autologous articular cartilage was harvested from a minor load-bearing area on the distal femur, and component cells were isolated, expanded in vitro, and reimplanted back into the patient. In particular, after excising the damaged cartilage down to the subchondral bone, defects (1.65 to 6.5 cm2) were sutured closed with a periosteal flap creating a pocket into which the cultured cells were injected. Disabling symptoms (e.g., knee pain, swelling, locking) were markedly reduced over a follow-up period of up to 5 years. Femoral transplants were clinically graded as good or excellent in 88% of the cases, with 73% of the biopsy specimens resembling hyaline cartilage. In contrast, patellar transplants were graded as good or excellent in only 29% of the cases, with only 14% of the biopsies resembling hyaline cartilage. A therapy based on this technique was approved for clinical use by the Food and Drug Administration (FDA) in August, 1997 and is being marketed under the tradename Carticel®.4 However, the original clinical study report has been criticized for not being a prospective controlled, randomized study and for its lack of a standard outcome analysis (i.e., for subjective evaluation with little quantitative biochemical or mechanical data).5,6 The Carticel® technique was also assessed in rabbits and dogs. In one study, cultured autologous chondrocytes were transplanted into patellar defects in conjunction with periosteal flaps and in some cases carbon fiber pads.7 The repair tissue was characterized as hyaline-like and reported to develop in vivo, as assessed by an increasing degree of cellular columnarization. Histological scores were highest in defects treated with transplanted chondrocytes and periosteum, as compared to the combination of chondrocytes, carbon fibers and periosteum or periosteum only. The finding that the use of carbon fibers did not improve longterm repair was attributed to scaffold-induced diffusional limitations.


species, age

Mixed vial

2 hours

Rabbit 9-12 months old

Chick 3 years old

Hyaluronic acid gel

Poly(D,D,L,L) lactic acid

Goat adult

Rabbit 3 mo.–adult

Hyaluronic acid gel

None or collagen gel

Dog 2yr–adult

Rabbit 4 months old

Several hours

duration

None or carbon fibers

Static dish

conditions

Human 14–48 years old

3.7 mm dia × 10 mm thick

dimensions

In vitro culture

None


Precursor cells from bone marrow (goat, expanded) Precursor cells from bone marrow (embryonic and adult chick); chondrocytes (chick, expanded) Perichondral chondrocyte (rabbit, expanded 3 weeks)

Articular chondrocyte (human, expanded 2-3 wks) Articular chondrocyte (rabbit, expanded 2 wks) and periosteum Articular chondrocyte (dog, expanded 3 wks) and periosteum Precursor cells from bone marrow and periosteum; chondrocytes (rabbit, expanded); glue

material

Scaffold

Representative Studies of Articular Cartilage Tissue Engineering

1 In vivo cartilage repair

Cells (source, expansion) and other components

Table 124.1

Distal femur 3.7 mm dia, 5 mm deep

Proximal tibia 3 mm dia × 2 mm deep

Distal femur 4 mm dia, full thickness

Distal femur up to 3 × 6 × 3 mm ± protease treatment

Distal femur, 4 mm dia, full thickness

Distal femur, patella 1.6-6.5 cm2, full thickness Patella 3 mm dia, full thickness

implant site, defect size

Animal model

H; amounts of GAG and collagen; IH (collagen types I, II); Confined compression modulus

H (semiquantitative) IH (collagen types I, II)

Auto- and allografts 3 months

Allograft 6 weeks– 12 months

H (quantitative) IH (collagen type II, KS, PG)

H (semiquantitative); TEM mechanics (microindentation)

Autograft precursor cells Allograft chondrocytes 1 week–1 year

Auto- and allografts 3 months

H (semiquantitative) IH (collagen types I, II)

H (semiquantitative)

Arthroscopy clinical signs

Assessment

Autograft 1.5-18 months

Autograft 2–12 months

Autograft 16–66 months

graft type, duration

Cell alignment determined by scaffold geometry Hyaline cartilage; collagen mainly type II GAG subnormal, mechanical properties normal.

Autograft better than allograft; Embryonic cells better than adult.

Hyaline cartilage (surface), new bone (deep); Precursor cells better than chondrocytes; Bone marrow better than periosteum; Integration improved by protease treatment. Autograft better than allograft.

Fibroblastic and chondrocytic cells, collagen II early repair better than late repair.

Cells plus periosteum better than cells on carbon meshes.

Symptomatic relief Femur repair better than patella.

Results

Chu et al.15,16

Nevo et al.14

ButnariuEphrat et al.13

*Caplan et al.10 Wakitani et al.11

Shortkroff et al.8 Breinan et al.9

Brittberg et al.7

Brittberg et al.3

References (* = review)

Alginate beads

Fibrous PGA; porous PLLA fibrous PGA

Fibrous PGA and nylon; PLGA mesh porous collagen

Articular chondrocyte (human, 5-42 yr old)

Articular chondrocyte (bovine)

Articular chondrocyte (bovine; expanded up to P2)


Agarose gel


2 In vitro chondrogenesis

Static and mixed dishes Mixed dish

10 mm × (5-10) mm × (2-3) mm 10 mm dia × (1-5) mm thick 2 mm thick Perfused teflon bag Static dish

Static dish

Static dish; compression chamber

Not reported

16 mm dia × 1 mm thick

Static dish

Static dish


Fibrous PGA

Collagen gel

Static dish


Fibrous PGA

Static dish

conditions

Articular chondrocytes (2-8 month old rabbits) Muscle mesenchymal stem cell (expanded 3 wks) Articular chondrocyte (1 month old rabbits)

dimensions

5 weeks

6-8 weeks

6-8 weeks

30 days

10 weeks 7 weeks

2 weeks

2.5 weeks

3-4 weeks

2 weeks

duration

In vitro culture

Collagen gel

material

Scaffold

Rabbit 6 months old

Rabbit 8 months old

Rabbit 8 months old

Rat 8-9 weeks old

species, age


Articular chondrocyte (3 week old rats)


Table 124.1 (continued)

distal femur 4 mm dia × 4 mm deep

distal femur 3 mm dia, full thickness

distal femur 3 mm dia, full thickness

distal femur 1.5 mm dia × 1.5 mm deep


Animal model


allograft 1 day– 6 months

H; GAG content incorporation of 35S and 3H

H; TEM; IH (keratan sulfate) Amounts of DNA, aggrecan, GAG, Incorporation of 35S and 3H H; IH (type II collagen) amounts of GAG, DNA, collagen and undegraded PGA

H; SEM; GAG and DNA content Incorporation of 35S and 3H Mechanical properties

H (qualitative)



Assessment

allograft 6 weeks– 3 months

allograft 1–6 months

auto- and allografts 2 weeks– 12 months


Development of mechanically functional matrix (25% of normal) after 7 wks; Dynamic compression enhanced synthesis of GAG and collagen Formation of cartilaginous matrix composed of two compartments with different rates of proteoglycan turnover PGA better than PLLA 1.2-3.5 mm thick cartilaginous constructs; Mixing and high initial cellularity improved construct structures Scaffold materials affected GAG and collagen synthesis; Perfused bag better than static dish

Hyaline cartilage without endochondral ossification

Fairly good repair at 3 months.

Slight inflammation that resolved by 8 weeks 6-12 month autografts and allografts comparable. Fairly good repair at 6 months.

Results

Grande et al.26

Freed et al.24,25,32,34

Haeuselmann et al.23

Buschmann et al.20,21

Kawamura et al.12

Grande et al.19

Freed et al.18

Noguchi et al.17


Fibrous PGA

None

Fibrous PGA

Fibrous PGA

Fibrous PGA

Articular chondrocyte (equine; young and adult)

Articular chondrocyte (human fetal)





Articular chondrocyte (rabbit)

PLGA and polydioxanon meshes coated with adhesion factors and agarose Fibrous PGA

Articular chondrocyte (human, 30-65 yr old)

Static and mixed dishes Static and mixed flasks

Rotating vessel

Rotating vessel


5 mm × 5 mm × 2 mm 5 mm dia × 2 mm thick


Static dish

Perfused cartridge (1.2 mL) Static dish Compression chamber (3.4 or 6.9 MPa; 5s on, 30s off)

10 mm dia × 2 mm thick 10 mm × 10 mm × 1 mm

Perfused chamber

Not reported

6 weeks

1 wk– 7 months

8 weeks

4 months

5 weeks

4 weeks

2 weeks

SEM; amounts of DNA, GAG and collagen (total and types II, IX and X); collagen pyridinium crosslinks

H; TEM; IH (collagen types II, IX) amounts of DNA, GAG, and collagen Incorporation of 35S and 3H Mechanical properties

H, TEM; IH (collagen types I, II, IX; CS, KS) amounts of GAG and total collagen H; amounts of GAG, DNA, and total collagen

H; amounts of GAG and total collagen compressive stiffness

H, IH (type II collagen, CS) amounts of GAG and collagen

H; TEM; IH (collagen types I and II)

2.7-4.8 mm thick constructs; Seeding and cultivation in mixed flasks yielded largest constructs and highest fractions of ECM 3-8 mm thick constructs. After 6 wks ECM was continuously cartilaginous. After 7 months, compressive stiffness became comparable to normal cartilage Structure and composition of collagen network in 6 wk constructs comparable to native cartilage; collagen content and crosslinking subnormal.

Cyclical loading promoted the production of GAG and collagen and increased compressive stiffness in constructs based on young cells Fetal chondrocytes cultured without serum formed 1.5-2 mm thick layer of hyaline cartilage.

2 mm thick constructs More tissue at edge than center

Evidence of collagen fibrils and proteoglycan

Riesle et al.38

Freed et al.36,37

VunjakNovakovic et al.33,35

Adkisson et al.31

Carver and Heath29,30

Dunkelman et al.28

Sittinger et al.27


BMSC CS ECM GAG H IH KS

Static and mixed flasks Rotating vessel

Mixed dish



PG PGA PLGA PLLA SEM TEM

conditions

species, age


Animal model

proteoglycan polyglycolic acid polylactic-co-glycolic acid poly (L) lactic acid scanning electron microscopy transmission electron microscopy

4 weeks

6 weeks

duration

In vitro culture

dimensions

bone marrow stromal cells chondroitin sulfate extracellular matrix glycosaminoglycan histology immunohistochemistry keratan sulfate

Fibrous PGA Porous PLGA

BMSC (chick embryo; bovine)

Note:

Fibrous PGA

material

Scaffold




Table 124.1 (continued)


H; IH (collagen types II and X) amounts of DNA, GAG and collagen

H; amounts of DNA, GAG and collagen (total and type II); mechanical properties

Assessment

Construct structure and function could be modulated by the conditions of flow and mixing. Mechanical parameters correlated with wet weight fractions of GAG, collagen and water. Selective cell expansion and the cultivation on 3-dimensional scaffolds resulted in the formation of cartilaginous tissues

Results

Martin et al.42-44

VunjakNovakovic et al.39


In other studies, cultured autologous chondrocytes in conjunction with periosteal flaps were transplanted into femoral defects in dogs and compared to periosteal flaps alone and to untreated defects.8,9 Three phases of healing were demonstrated: formation of repair tissue (at 1.5 months), remodeling (at 3 and 6 months), and degradation (at 12 and 18 months). Neither periosteum nor transplanted chondrocytes enhanced healing after 1 year, at which time hyaline-like repair tissue appeared to be displaced by fibrocartilage. The differences between results obtained in rabbit and dog studies were attributed to several variables including species, subject age, defect location, surgical technique, and postoperative animal activity. Low retention of implanted cells in the lesion and failure of immature repair tissue subjected to high mechanical forces were listed as possible causes of degradation. The group of Caplan et al. pioneered the use of autologous mesenchymal stem cells in the repair of osteochondral defects.10 Osteoprogenitor cells were selected from whole bone marrow based on their ability to adhere to Petri dishes and expanded in monolayer culture prior to implantation. Autologous precursor cells were delivered in collagen gels into defects that were in some cases pretreated with proteolytic enzymes. Other implants were based on either autologous precursor cells derived from periosteum or allogeneic chondrocytes derived from articular cartilage. The dimensions of defects were up to 3 mm wide, 6 mm long, and 3 mm deep, among the largest reported for repair studies in rabbits. Repair was assessed histologically and in some cases mechanically, e.g., relative compliance using a microindentation probe.11 Autologous precursor cells in collagen gels formed a cartilaginous surface zone while tissue at the base of the defect hypertrophied, calcified, and was replaced by host-derived vasculature, marrow, and bone.10 It was postulated that the different biological milieus of the synovial fluid at the joint surface and the osseus recepticle in the underlying bone played key roles in architecturally appropriate precursor cell differentiation. Improved integration of the graft with the surrounding host tissue was reported following partial digestion at the defect site (e.g., with trypsin) prior to implantation; this finding was attributed to enhanced interdigitation of newly synthesized ECM molecules. However, progressive thinning of the cartilaginous surface zone was observed over 6 months, which was more pronounced with periosteallyderived than with bone marrow-derived precursor cells. In contrast, allografted chondrocytes in collagen gels rapidly formed plugs of hyaline cartilage that filled the entire defect but failed to develop a region of subchondral bone at the base or to integrate with the surrounding host tissue even after 6 months. Similar results were obtained when chondrocytes were cultured in vitro in collagen gels prior to in vivo implantation12 (as described below). Nevo et al.13,14 have explored the use of both chondrocytes and osteoprogenitor cells derived from bone marrow delivered in a hyaluronic acid gel. Autografts or allografts were used to repair tibial defects in goats and femoral defects in chicks. The autografts were superior to the allografts, which evoked a typical immune response and progressive arthrosis. In both animal models, autografted defects were repaired with a well-integrated tissue that resembed hyaline cartilage at the surface and bone at the base. Amiel et al. qualitatively (i.e., histologically) and quantitatively (i.e., biochemically and mechanically) assessed osteochondral defects repaired with perichondral cells and porous polylactic acid in a rabbit allograft model.15,16 After 6 weeks, the alignment of cells in the repair tissue followed the architecture of the scaffold. After 1 year, the repair tissue had variable histological appearance (only 50% of the subchondral bone reformed) and biochemical composition (dry weight fractions of GAG were 55% those of the host cartilage), but mechanical properties (i.e., modulus and permeability in radially confined compression) were comparable to those measured for the host cartilage. In this case, mechanical testing appeared to be less sensitive than histological and biochemical assessments; this finding was attributed to difficulties in mechanical measurements and the derivation of intrinsic parameters, due to the geometry and the non-homogeneous nature of the repair tissue, respectively. Noguchi et al.17 compared autologous and allogenic repair using chondrocytes cultured in collagen gels prior to implantation in inbred rats. After 2 to 4 weeks, the repair tissue consisted of hyaline articular cartilage with slight inflammatory cell infiltration in both groups; the immune response was somewhat more conspicuous in the allografts. However, inflammation resolved by 8 weeks and auto- and allografted repair tissues were almost identical after 6 months.


Freed et al.18 cultured chondrocytes expanded by serial passage (P4) on polyglycolic acid (PGA) scaffolds for 3 to 4 weeks in vitro prior to implantation as allografts in rabbits. Compared to native rabbit articular cartilage, constructs contained 25% as much total collagen and 86% as much GAG per gram dry weight at the time of implantation. After 1 and 6 months, the histological scores of defects repaired with cell-PGA constructs did not differ significantly from those following implantation of PGA alone, except for qualitatively better surface smoothness, cell columnarization, and spatial uniformity of GAG in the defects grafted with cell-PGA constructs. Grande et al.19 cultured osteoprogenitor cells derived from skeletal muscle on PGA scaffolds for 2 to 3 weeks in vitro prior to implantation as allografts in rabbits. At the time of implantation, cells were attached to the scaffold but had not undergone chondrogenesis. After 3 months, defects repaired with cell-PGA constructs consisted of a cartilaginous surface region similar in thickness to the host cartilage and normal appearing subchondral bone, while implantation of PGA alone resulted in a patchy mixture of fibrous and hyaline cartilage. Kawamura et al.12 cultured chondrocytes in collagen gels for 2 weeks in vitro prior to implantation as allografts in rabbits. At the time of implantation, constructs appeared stiffer than uncultured cells in collagen gels and histologically resembled hyaline cartilage. After 6 months, the repair tissue still consisted of a thick plug of hyaline cartilage, i.e., the region at the base was neither vascularized nor replaced by bone. It was suggested that a mechanical mismatch between the thin layer of host cartilage and the thick plug of cartilaginous repair tissue, and the lack of regeneration of a proper subchondral bony base would contribute to long term implant failure.

In Vitro Chondrogenesis Buschmann et al.20,21 showed that high cell density cultures of calf articular chondrocytes in agarose gels formed a mechanically functional cartilaginous matrix. After 7 weeks in vitro, construct compositions and mechanical properties (i.e., fractions of DNA and GAG, compressive modulus, streaming potential) reached values of about 25% those of native calf cartilage. Constructs responded to mechanical forces in a manner similar to that of native cartilage: static compression supressed ECM synthesis by an amount that increased with increasing compression amplitude and culture time, while dynamic compression stimulated ECM synthesis by an amount that increased with ECM accumulation and culture time. The authors postulated that several mechanisms could be involved in mechanotransduction, including cellECM interactions and/or changes in interstitial fluid flow and streaming potential. Hauselmann et al.22 demonstrated the phenotypic stability of calf articular cartilage after prolonged (8 month) cultivation in alginate beads. The ECM formed by adult human chondrocytes cultured in gels for 30 days was similar to that of human articular cartilage and could be used to study PG turnover.23 The ECM consisted of two compartments: a small amount of cell-associated matrix (corresponding to the pericellular and territorial ECM), and a larger amount of matrix further removed from the cells (corresponding to interterritorial ECM). Aggregated PGs in the cell-associated ECM turned over relatively quickly (t1/2 of 29 days) as compared to those in the further-removed ECM (t1/2 >100 days). These findings were attributed to the effects of proteolytic enzymes located at the cell membrane. Freed et al.24 characterized cartilaginous constructs based on calf articular or human rib chondrocytes and synthetic polymer scaffolds. Fibrous polyglycolic acid (PGA) meshes yielded constructs with higher cellularities and GAG production rates as compared to porous polylactic acid (PLLA) sponges; these findings were attributed to differences in polymer geometry (mesh vs. sponge, pore size distribution) and degradation (faster for PGA than PLLA). PGA scaffolds were subsequently characterized in detail and produced on a commercial scale.25 Grande et al.26 showed that the choice of scaffold affected ECM synthesis rates, e.g., PGA and collagen scaffolds respectively enhanced GAG and protein synthesis rates of calf chondrocytes cultured in a perfused system. Sittinger et al.27 observed PG and collagen deposition by human articular chondrocytes cultured on synthetic polyester meshes coated with poly-L-lysine or type II collagen and embedded in


agarose gels. Dunkelman et al.28 and Grande et al.26 both reported that cell-polymer constructs were more likely to form cartilaginous ECM if cultured in perfused vessels rather than statically. Carver and Heath29,30 demonstrated that physiologicial levels of compression enhanced GAG deposition and improved the compressive moduli of engineered cartilage. In particular, constructs based on equine chondrocytes and PGA scaffolds were subjected to intermittent hydrostatic pressures of 500 and 1000 psi (3.4 and 6.9 MPa) in a semi-continuous perfusion system for 5 weeks. Structural and functional improvements were observed in constructs based on cells obtained from young but not adult horses. In a recent study, fetal human chondrocytes cultured at high density in serum-free medium formed 1.5 to 2.0 mm thick hyaline cartilage over 120 days whereas otherwise identical cultures containing serum could not be maintained for more than 30 days.31 Freed and Vunjak-Novakovic studied in vitro chondrogenesis using a variety of methods to seed and cultivate calf chondrocytes on polymer scaffolds as follows. Static seeding of scaffolds two or more mm thick resulted in bilaminar constructs with a fibrous upper region and a cartilaginous lower region.32 In contrast, dynamic cell seeding in spinner flasks allowed relatively uniform cell seeding of scaffolds 2-5 mm thick at an essentially 100% yield.33 An increase in the initial density of seeded cells resulted in comparable construct cellularities and collagen contents, but markedly higher GAG contents.34 Mixing during 3D culture markedly improved construct morphology and composition.32,35 For example, 10 mm diameter × 5 mm thick scaffolds seeded and cultured in mixed flasks weighed twice as much as those seeded and cultured in mixed dishes, and contained about 2.5 times more of each GAG and total collagen.35 However, constructs grown in mixed flasks formed outer capsules that were up to 300 µm thick and contained high concentrations of elongated cells and collagen and little GAG; this finding was attributed to the effects of turbulent flow conditions on cells at the construct surface.35 In contrast, constructs cultured in rotating vessels had relatively uniform distributions of cells and ECM and were up to 5 mm thick after 6 weeks in culture.36 With increasing culture time, construct ECM biosynthesis rates and the fractional loss of newly synthesized macromolecules into the culture medium both decreased36 but construct GAG fractions and mechanical properties improved.37 The collagen networks of constructs and native cartilage were similar with respect to fibril density and diameter and fractions of collagen types II, IX and X.38 Vunjak-Novakovic et al.39 studied the relationships between construct compositions (fractions of water and ECM components) and mechanical properties in static and dynamic compression (equilibrium modulus, dynamic stiffness, hydraulic permeability, streaming potential) using three different culture environments: static flasks, mixed flasks, and rotating vessels. Constructs cultured in static and mixed flasks had lower concentrations of ECM and worse mechanical properties as compared to constructs cultured in rotating vessels. The structure-function relationships detected for chondrocyte-PGA constructs appeared consistent with those previously reported for chondrocytes cultured in agarose gels,20 native calf cartilage,40 and adult human cartilage.41 Martin et al.42-44 demonstrated that bone marrow stromal cells (BMSC) expanded in monolayers, seeded onto scaffolds and cultured in mixed Petri dishes formed large, 3D cartilaginous tissues. The presence of a 3D scaffold was required, as demonstrated by the small size and noncartilaginous nature of control cell pellet cultures. The presence of fibroblast growth factor 2 (FGF) during 2D expansion promoted differentiation of BMSC during 3D cultivation. As compared to avian (embryonic chick) BMSC, mammalian (calf) BMSC required a more structurally stable 3D scaffold and the presence of additional biochemical signals.

124.3 Cell-Polymer-Bioreactor System The literature reviewed in Section 124.2 above demonstrates the feasibility of using cell-based tissue engineering approaches to in vivo articular cartilage repair and in vitro studies of chondrogenesis. In the next section we will first describe the in vitro cell-polymer-bioreactor system and then use selected examples from our own work to illustrate its use in studies of the development and modulation of


FIGURE 124.2 Cell seeding of polyglycolic acid (PGA) mesh. (a) PGA ultrastructure (scanning electron micrograph). (b) chondrocytes attach to PGA fibers and maintain their spherical morphology (H&E stained histological section). (c) dynamic cell seeding: magnetic stirring generates convective motion of cells into porous scaffolds that are fixed to needles embedded in the stopper of the flask. (d) chondrocytes attached to PGA scaffolds with an effective yield of 100% over 24 h.

construct structure and function. The unifying hypothesis of our work is that isolated chondrocytes or precursor cells can form functional cartilage tissue in vitro, if cultured on a 3D structural template in an environment that provides the necessary biochemical and physical signals. Ideally, engineered cartilage should display the key structural features of native cartilage, be able to withstand physiological loads, and be able to integrate with adjacent host tissues following in vivo implantation.

Experimental Methods Cells: Cell types studied included articular chondrocytes and their precursors. Chondrocytes were obtained from the articular cartilage of 2 to 3 week old bovine calves20,24 and used immediately after isolation. Bone marrow stromal cells (BMSC) were obtained from the marrow of 2 to 3 week old bovine calves43 or 16 day embryonic chicks.42 Scaffolds: Our best characterized scaffolds are made of polyglycolic acid (PGA) in form of 97% porous nonwoven meshes of 13 µm diameter fibers (Fig. 124.2a).25 Other scaffolds that have been studied include polylactic acid (PLLA)24 and polylactic-co-glycolic acid (PLGA) sponges.45 Cell seeding: Polymer scaffolds (5 to 10 mm in diameter, 2 to 5 mm thick) were fixed in place and seeded with cells using well-mixed spinner flasks 33 (Fig. 124.2c). Bioreactor cultivation: Tissue culture systems under investigation are schematically presented in Fig. 124.3. Static and mixed dishes contained one construct per well in 6 to 7 mL medium; mixed dishes were placed on an orbital shaker at 75 rpm.32 In spinner flasks, scaffolds were fixed to needles (8 to 12 per flask, in 100 to 120 cm3 medium), seeded with isolated cells under mixed conditions and cultured either statically or exposed to unidirectional turbulent flow using a non-suspended magnetic bar at 50-80 rpm.35 Rotating vessels included the slow turning lateral vessel (STLV) and the high aspect ratio vessel (HARV), each of which was rotated around its central axis such that the constructs (8 to 12 per vessel, in 100 to


FIGURE 124.3 Tissue culture bioreactors. (a) static and orbitally mixed Petri dishes. (b) static and magnetically stirred flasks. (c) rotating vessels: the slow turning lateral vessel (above) and the high aspect ratio vessel (below). (d) perfused chamber in which cultured constructs are monitored using magnetic resonance imaging (MRI, above); perfused rotating vessel (below).

120 cm3 medium) settled freely in a laminar flow field.46,47 Perfused vessels included columns used to culture constructs and non-invasively monitor the progression of chondrogenesis using magnetic resonance imaging (MRI),48 and a rotating perfused vessel developed by NASA for flight studies.37 In all vessels, medium was exchanged batchwise (at a rate of ~3 cm3 per construct and day), and gas was exchanged continuously, by surface aeration or by diffusion through a silicone membrane. Analytical techniques: Construct size and distributions of cells and tissue components were assessed by image analysis.33,36 Constructs for histological assessment were fixed in neutral buffered formalin, embedded in paraffin, sectioned (5-8 µm thick), and stained with hematoxylin and eosin (H&E for cells), safranin-O (for GAG), and monoclonal antibodies (for collagen types I, II, IX, and X).36,38,42 Ultrastructural analyses included scanning and transmission electron microscopy (SEM, TEM).38 Biochemical compositions were measured following papain or protease-K digestion of lyophilized constructs.24,25,36 Cell number was assessed by measuring the amount of DNA using Hoechst 33258 dye.49 Sulfated GAG content was determined by dimethylmethylene blue dye binding.50 Total collagen content was determined from hydroxyproline content after acid hydrolysis and reaction with p-dimethylaminobenzaldehyde and chloramine-T.51 Type II collagen content was determined by inhibition ELISA.36,38,52 The presence of other collagen types (e.g., I, IX, X) was demonstrated and semiquantitatively measured using SDS-PAGE and Western blots.38 GAG distribution was determined by MRI.48,53 Synthesis rates of GAG, total protein and collagen were measured by incorporation and release of radiolabeled tracers.36 Cell metabolism was assessed based on the ratio of lactate production and glucose consumption, and ammonia production rates.54 Mechanical construct properties (e.g., compressive modulus, dynamic stiffness, hydraulic permeability, streaming potential) were measured in static and dynamic radially confined compression.39,55

Developmental Studies As described above (Section 124.2, Table 124.1), articular cartilage repair has been enhanced following the transplantation of chondrocytes and osteochondral progenitor cells in conjunction with polymer scaffolds. However, identification of specific factors that influence in vivo regeneration and/or repair can be difficult due to host systemic responses (e.g., neuro-endocrinological, immunological). Controlled


FIGURE 124.4 Temporo-spatial pattern of ECM deposition. Calf chondrocytes cultured on PGA scaffolds in rotating vessels. Chondrogenesis was initiated peripherally in 12 day constructs (left panels) and proceeded appositionally resulting in a continous cartilaginous ECM in 40 day constructs (right panels). Histological cross-sections were stained with safranin-O for GAG (top panels) or with a monoclonal antibody to collagen type II (bottom panels).

in vitro studies carried out in the cell-polymer-bioreactor system can thus provide useful, complementary information regarding the process of cartilage matrix regeneration starting from isolated cells (herein referred to as developmental studies). Chondrogenesis in vivo and in vitro is thought to be initiated by precursor cell aggregation.56 Cartilaginous ECM is deposited, with concurrent increases in the amounts of GAG and collagen types II and IX, both in mesenchymal cells obtained from embryonic chick limbs57 and in human limbs at various developmental stages.58 ECM deposition starts at the center of the developing limb bud and spreads peripherally,57 in conjunction with cellular conditioning of their microenvironment.59 In calf chondrocyte-PGA constructs, amounts of GAG and collagen type II also increased concomittantly36 (Fig. 124.4). By culture day 12, cartilaginous tissue had formed at the construct periphery; by day 40, constructs were continuously cartilaginous matrix over their entire cross sections (6.7 mm diameter × 5 mm thick). Chondrogenesis was initiated peripherally and progressed both inward towards the construct center and outward from its surface; these findings could be correlated with construct cell distributions, as described below. The temporo-spatial patterns of GAG distribution could also be monitored in living constructs using MRI.48 Constructs cultured 6 weeks contained an interconnected network of collagen fibers that resembled that of calf articular collagen with respect to overall organization and fiber diameter (Fig. 124.5A). Type II collagen represented more than 90% of the total collagen, as quantitated by an ELISA36,38 and demonstrated qualitatively in Fig. 124.5B. Construct collagen type II was not susceptible to extraction with a chaotropic agent (guanidine hydrochloride, GuHCl), suggesting some degree of cross-linkage. However, construct collagen had only 30% as many pyridinium crosslinks as native calf cartilage.38 Construct wet weight fractions of cartilaginous ECM increased, the amount of polymer decreased, and construct cellularity plateaued over 6 weeks of culture in rotating vessels36 (Fig. 124.6). Over the same time interval, construct ECM synthesis rates decreased by approximately 40% and the fraction of newly synthesized macromolecules released into the culture medium decreased from about 25% of total at 4 days to less than 4%; the latter can be attributed in part to collagen network development.36 The mass of the PGA scaffold decreased to about 40% of initial over 6 weeks.25 The time constant for scaffold degradation


FIGURE 124.5 Collagen structure and type. Calf chondrocytes cultured on PGA scaffolds for 6 weeks in rotating vessels. (a) Scanning electron micrographs showed similar collagen network ultrastructures for constructs and native calf articular cartilage. (b) The presence of type II collagen in constructs was demonstrated using pepsin-digested constructs in conjunction with SDS-PAGE, and confirmed by Western blot (with a monoclonal antibody to αII). Lanes on SDS-PAGE: (1) molecular weight standards, (2) collagen type I, (3) GuHCl extract of construct, (4) pepsin extract of construct, (5) collagen type II; Lanes on Western blot: (1) collagen type II, (2) GuHCl extract of construct, (3) pepsin extract of construct.

was of the same order of magnitude as the time constant of ECM deposition, a situation associated with enhanced tissue regeneration according to the hypothesis of isomorphous tissue replacement.60

Modulation of Cartilaginous Structure and Function The structure and function of articular cartilage are determined, at least in part, by environmental factors.61 It is likely that the same factors that affect in vivo cartilage development, maintenance and remodeling also affect in vitro chondrogenesis. In the selected examples described below, we will describe the effects of: (1) cell, scaffold, and biochemical factors and (2) cultivation conditions and time on the structure and function of engineered cartilage. In principle, in vitro cultivation of a cell-polymer construct prior to in vivo implantation can help localize cell delivery and promote device fixation and survival, while maintaining an ability for integration at the graft-host interface. In contrast, cells transplanted in the absence of a carrier vehicle tended to leak away from the defect site,9 while cells implanted immediately after loading in/on biomaterials were more vulnerable to mechanical forces and metabolic changes experienced during fixation or following implantation.12 Cells, the Scaffold and Biochemical Signals Chondrogenesis depends on the cells themselves (e.g., type, in vitro expansion, density, spatial arrangement), the scaffold (e.g., material, structure, degradation rate), and the presence of biochemical factors


FIGURE 124.6 Tissue regeneration kinetics. Calf chondrocytes cultured on PGA scaffolds in rotating vessels. ECM components (GAG and collagen type II) accumulated, cellularity remained constant, and polymer mass decreased with increasing culture time. Data represent the average ± SD of 3 independent measurements.

(e.g., growth factors, hormones). The choice of cell type, which includes differentiated chondrocytes isolated from cartilage (articular or rib) and osteochondral progenitor cells isolated from bone marrow (herein referred to as bone marrow stromal cells, BMSC), can affect in vitro culture requirements (e.g., medium supplements) and in vivo construct function (e.g., integration potential). Polymer scaffolds vary with respect to surface properties (e.g., chemistry and wettability, which affect cell spreading and proliferation), geometry (e.g., dimensions, porosity, and pore size, which affect the spatial cell arrangement and the transmission of biochemical and mechanical signals), and physical properties (e.g., mechanical integrity and degradation rate, which determine whether the polymer can provide a structurally stable template for tissue regeneration). Watt 61 suggested that a cell cultured in vitro will tend to retain its differentiated phenotype under conditions that resemble its natural in vivo environment. In the case of chondrocytes, phenotypic stability is enhanced by cultivation in alginate or agarose.22,23,63 Moreover, chondrocytes dedifferentiated by serial passage in monolayers redifferentiated (i.e., reacquired a spherical shape, ceased dividing, and resumed the synthesis of GAG and collagen type II) when transferred into 3D cultures.64,65 When cultured on 3D fibrous PGA scaffolds, chondrocytes retained their spherical shape and formed cartilaginous tissue.25,66 Chondrocytes cultured at high cell densities tended to express their differentiated phenotype.63,67 It was hypothesized that the effective cross-talk between cells depends on the presence of homotypically differentiated cells in the immediate cell environment.67 The term “community effect” was later coined by Gurdon69 who suggested that the ability of a cell to respond to phenotypic induction is enhanced by, or even dependent on, other neighboring cells differentiating in the same way at the same time. A postulated underlying mechanism involves changes in gene transcription and translation caused by cellcell/ECM interactions.62,70,71 In cell-polymer constructs, the density of cells initially seeded at the construct periphery was sufficient to initiate chondrogenesis in that region. In particular, corresponding tempero-spatial patterns of cartilaginous ECM deposition and cell distribution were observed (compare Figs. 124.4 and 124.7). The finding that cell densities were initially higher peripherally than centrally, which implies that cell seeding density and proliferation rate were relatively higher peripherally, can be attributed to enhanced rates of nutrient and gas transfer at the construct surface. Over the 6 week culture period, self-regulated cell


FIGURE 124.7 Temporo-spatial pattern of cell density. Calf chondrocytes cultured on PGA scaffolds in rotating vessels. Cell density was initially higher at the construct periphery (white bars) than at its center (shaded bars), while cell distribution in 40 day constructs was spatially uniform. Number of cells/mm2 was assessed by image analysis of H&E stained histological sections (average ± SD of 24 independent measurements).

proliferation and ECM deposition resulted in constructs with physiological cellularities and spatially uniform distributions of components36 (Figs. 124.4 and 124.7). Increases in the size, cellularity, and absolute amount of cartilaginous ECM were observed when chondrocyte-PGA constructs were cultured in medium supplemented with insulin-like growth factor I (IGF-I, 10-300 ng/mL) and serum (10%).72 In the case of osteochondral progenitor cells, Caplan et al.10 suggested that principles of skeletal tissue engineering should be governed by the same motifs as embryonic development. BMSC differentiation is thought to be regulated by cell-to-cell contacts in an environment capable of activating the differentiation program.73 In vitro (in BMSC aggregates), the induction of chondrogenesis depended on the presence of transforming growth factor-β1 (TGF β) and dexamethasone.74 In vivo (in rabbits), osteochondral repair recapitulated embryonic events and depended on the spatial arrangement and density of precursor cells and the presence of specific bioactive factors.10 In cell-polymer constructs, chondrocytic differentiation of BMSC depended both on scaffold-related factors and on exogenous biochemical signals.42-44 Avian BMSC cultured in the absence of polymer scaffolds formed small bilaminar tissues in which the lower region contained GAG and upper region appeared undifferentiated (Fig. 124.8A). The same BMSC formed constructs consisting of a single tissue phase when cultured on PGA scaffolds in mixed Petri dishes (Fig. 124.8B), while BMSC expansion in the presence of FGF prior to culture on PGA scaffolds resulted in the most cartilaginous ECM (Fig. 124.8C). When FGF-expanded mammalian BMSC were cultured on nonwoven PGA mesh (Fig. 124.2A), constructs first contracted and then collapsed (Fig. 124.8D), while the same cells cultured on a scaffold consisting of a continuous polymer phase (polylactic-co-glycolic acid and polyethylene glycol (PLGA/PEG) sponge)45 maintained their original dimensions (Fig. 124.8E). When FGF-expanded mammalian BMSC were cultured on PLGA/PEG scaffolds in mixed dishes, chondrogenesis was observed in media supplemented with transforming growth factor beta 1 (TGF β), insulin, and dexamethasone (Fig. 124.8F) while osteogenesis was observed in media supplemented with betaglycerophosphate and dexamethasone (Fig. 124.8G). In the absence of these supplements, constructs consisted mainly of type I collagen and resembled loose connective tissue.


FIGURE 124.8 Chondrogenesis starting from bone marrow stromal cells (BMSC). (a-c) avian (chick embryo) BMSC cultured in media containing serum (10%) and ascorbic acid (50 µg/mL): (a) without polymer scaffolds (as cell pellets); (b) on PGA scaffolds, or (c) after expansion in the presence of FGF (1 ng/mL). (d-e) FGF-expanded mammalian (bovine calf) BMSC cultured on (d) PGA scaffolds (nonwoven meshes) or (e) PLGA/PEG scaffolds (sponges consisting of a continuous polymer phase). (f-g) FGF-expanded mammalian BMSC were cultured on PLGA/PEG sponges in media containing serum (10%), ascorbic acid (50 µg/mL) and either: (f) TGFβ (10 ng/mL), insulin (5 µg/mL), dexamethasone (100 mM), or (g) bGP (7 mM) and dexamethasone (10 mM).

Cultivation Conditions and Time Tissue culture bioreactors permit the in vitro cultivation of larger, better organized engineered cartilage than can be grown in static Petri dishes.47 Flow and mixing within bioreactors are expected to affect tissue formation in at least two ways: by enhancing mass transfer (e.g., of gases and nutrients) and by direct physical stimulation of the cells (e.g., by hydrodynamic forces). The transport of chemical species lies at the heart of physiology and to a large extent determines tissue structure.75,76 Cells communicate with each other by a combination of diffusion and convective flow, which are in turn driven by hydrodynamic, concentration and osmotic gradients. In vivo, mass transfer to chondrocytes involves diffusion and convective transport by the fluid flow that accompanies tissue loading77; the presence of blood vessels in immature cartilage can further enhance mass transfer. In vitro, mixing-induced convection can enhance mass transport at construct surfaces. In contrast, mass transfer within constructs, which occurs by diffusion only, can become the limiting factor in the cultivation of a large construct with a dense ECM.78 As compared to constructs grown statically, constructs grown in orbitally mixed Petri dishes and in mixed spinner flasks were larger and contained higher amounts of tissue components.32,35 Cell metabolism in constructs cultured in mixed and static flasks were respectively found to be aerobic and anaerobic, as assessed by lactate to glucose ratios and ammonia production rates.54 The form of a skeletal tissue represents a diagram of underlying forces transmitted across the ECM to the individual cells.61 Mechanotransduction is thought to involve four steps: mechanocoupling, biochemical coupling, signal transmission, and effector cell response.79 In vivo, load-bearing and immobilized articular cartilage respectively contained high and low GAG fractions.80,81 In vitro, physiological levels of dynamic compression increased the rates of GAG and protein synthesis in cartilage explants,82,83 while static loading supressed GAG synthesis.82,84 Physiological levels of dynamic compression also increased the GAG content and improved the mechanical properties of engineered cartilage.29,30 The motion of


FIGURE 124.9 Culture conditions affect construct morphology. Representative cross sections of 6 week constructs (calf chondrocytes/PGA) cultured in static flasks, mixed flasks, and rotating vessels are compared to fresh calf cartilage (low and high power H&E stained histological sections).

medium in roller bottles stimulated chondrocytes to form cartilaginous nodules,85,86 and fluid shear enhanced PG size and synthesis rate in chondrocyte monolayers.87 Turbulent mixing in spinner flasks induced the formation of a fibrous outer capsule at the construct surface, the thickness of which increased with both the mixing intensity and the duration of tissue cultivation.35 This finding was attributed to direct effects of mechanical forces, i.e., cells exposed to external forces tend to flatten and activate stressprotection mechanisms in order to remain firmly attached to their substrate88 and increase their stiffnesses by cytoskeletal rearrangements.89 The effects of culture conditions on the morphology, composition, and mechanical properties of chondrocyte-PGA constructs were studied over 6 weeks using three different bioreactors: static flasks, mixed flasks, and rotating vessels.39 In static cultures, GAG accumulated mostly at the periphery, presumably due to diffusional constraints of mass transfer (Fig. 124.9a). In mixed flasks, turbulent shear caused the formation of a thick outer capsule with little or no GAG (Fig. 124.9b). Only in rotating vessels, were GAG fractions high and spatially uniform (Fig. 124.9c). Construct fractions of GAG (Fig. 124.10a) and total collagen (Fig. 124.10b) increased in the following order: static flasks, mixed flasks, rotating vessels, native cartilage.39 Construct equilibrium moduli and hydraulic permeabilities (Figs. 124.10c,d) varied in a manner consistent with sample composition.39 As compared to native calf cartilage, 6-week constructs cultured in rotating bioreactors had similar cellularities, 75% as much GAG and 40% as much total collagen per unit wet weight, but only 20% the compressive modulus and 5-fold higher hydraulic


FIGURE 124.10 Culture conditions affect construct compositions and mechanical properties. Constructs (calf chondrocytes/PGA) cultured for 6 weeks bioreactors were compared with respect to: (a,b) wet weight fractions of GAG and total collagen, (c) equilibrium modulus, and (d) hydraulic permeability. As compared to static and mixed flasks, rotating vessels yielded constructs with the best properties, but these remained inferior to native calf cartilage. Data represent the average ± SD of 3-6 independent measurements.

permeability. The apparent lack of functional organization of ECM in constructs may be explained either by the use of immature cartilage (2 to 4 week old calves), or by the absence of specific factors in the in vitro culture environment which are normally present in vivo. It is possible that the mechanisms by which the dynamic fluctuations in shear and pressure in rotating vessels enhanced chondrogenesis resembled those associated with dynamic loading in vivo. However, the acting hydrodynamic forces were different in nature and several orders of magnitude lower than those resulting from joint loading.46,90,91 Studies of engineered cartilage subjected to physiological levels of dynamic compression29,30 and shear should thus allow a more direct comparison of in vitro and in vivo tissue responses. Chondrocytes can be phenotypically stable for prolongued periods of time in 3D cultures (e.g., for 8 months in alginate beads).22 The effect of prolongued cultivation (7 months in rotating bioreactors) on the structure and function of chondrocyte-PGA constructs is shown in Fig. 124.11. As compared to native calf articular cartilage, 7 month constructs had comparable GAG fractions (Fig. 124.11a), 30% as much total collagen (Fig. 124.11b), comparable equilibrium moduli (Fig. 124.11c), and comparable hydraulic permeability (Fig. 124.11d). At 7 months, constructs were phenotypically stable (75% of the total construct collagen was type II) and consisted of metabolically active cells (component cells attached and spread in Petri dishes and were enzymatically active).37 A successful approach to cartilage tissue engineering must also provide the potential for constructs to integrate with the adjacent cartilage and subchondral bone. Most of the in vivo studies described in Section 2 addressed this issue. In general, the implantation of chondrocytes and BMSC without a cartilaginous ECM resulted in repair tissue that deteriorated with time in vivo,9,11 while implantation of cartilaginous constructs integrated relatively poorly with adjacent host tissues.10,12 In vitro systems can be used to study the effects of specific factors on construct integration in the absence of uncontrollable


FIGURE 124.11 Culture time affects construct structure and function. Constructs (calf chondrocytes/PGA) cultured for 3 days (3 d), 6 weeks (6 w), 3 or 7 months (3 mo, 7 mo) in rotating vessels were compared with respect to: (a,b) wet weight fractions of GAG and total collagen, (c) equilibrium modulus, and (d) hydraulic permeability. All properties improved with culture time and approached values measured for native calf cartilage (normal ranges denoted by dotted lines). Data represent the average ± SD of 3 independent measurements.

variables intrinsic to in vivo studies. In one such study, constructs were cultured for various times, sutured into ring-shaped explants of native cartilage, and cultured for an additional period of time as composites.92 The integration process involved cell migration into the construct-explant interface and the formation of a new tissue which was initially fibrous but became progressively cartilaginous with increasing culture time (Fig. 124.12a). Construct equilibrium modulus, which was negligible at the beginning of cultivation, increased to approximately 15% of native calf cartilage after 6 weeks (Fig. 124.11c). The adhesive strength at the construct-explant interface was approximately 65% higher for composites made with 6 day constructs, which consisted mainly of cells, as compared to composites made with 5 week constructs, which had a well-formed ECM (Fig. 124.12b).

124.4 Summary and Future Directions Tissue engineering offers a cell-based approach to articular cartilage repair. In this chapter, we reviewed the state of the art of cartilage tissue engineering and focused on a cell-polymer-bioreactor system which can be used for controlled studies of chondrogenesis and the modulation of engineered cartilage by environmental factors. A procedure in which autologous chondrocytes are obtained from an articular cartilage biopsy, expanded in culture, and transplanted in conjunction with a periosteal flap3 is currently the only FDAapproved cell-based treatment for articular cartilage repair. However, in vivo studies in dogs had variable results and showed long-term degradation of the repair tissue.8,9 The clinical study has been viewed with some skepticism5,6 and long-term, prospective randomized clinical studies are needed to better evaluate the potential of this technique. Alternatively, autologous BMSC have been isolated from bone marrow aspirates, expanded, and implantated in conjunction with various gels to repair osteochondral defects in experimental animals.10,14 Following implanatation, the cells underwent a site-specific differentiation and formed a cartilaginous zone © 2000 by CRC Press LLC

FIGURE 124.12 Culture time affects construct integration potential. (a) Composites made from 6 day constructs (calf chondrocytes/PGA) and cultured for an additional 2 or 4 weeks (upper or lower panel, respectively; safranin-O stained histological sections). (b) Adhesive strengths of composites cultured 4 weeks, including: construct-cartilage composites made from 6 day or 5 week constructs and cartilage-cartilage composites (adhesion was estimated by fixing the outer ring in place and uniaxially loading the newly integrated central portion to failure as shown in the inset).

at the surface of the defect and a vascularized bony tissue at its base. In contrast to grafts based on mature chondrocytes, which failed to develop subchondral bone and to fully integrate with the host tissue, grafts based on precursor cells suffered from the progressive thinning of the cartilaginous surface zone. A 3D scaffold permits the in vitro cultivation of cell-polymer constructs that can be readily manipulated, shaped, and fixed to the defect site.18 As compared to such a pre-formed cartilaginous construct, cells injected under a periosteal flap or immobilized in a hydrated gel are more vulnerable to environmental factors and mechanical forces.12 The selection of an appropriate scaffold depends in part on the starting cell type. For example, nonwoven fibrous PGA meshes supported chondrogenesis starting from chondrocytes25 and avian BMSC,42 while a scaffold with more structural stability appeared to be required for chondrogenesis starting from mammalian BMSC (Fig. 124.8). Tissue constructs resembling native cartilage were engineered in vitro using isolated cells, 3D polymer scaffolds, and bioreactors (Figs. 124.4 through 124.6). Construct structure (histological, biochemical) and function (mechanical, metabolic) depended on cultivation conditions and duration.39 The cellpolymer-bioreactor system can thus provide a basis for controlled in vitro studies of the effects of time and specific biochemical and physical factors on chondrogenesis (Figs. 124.7 through 124.11). Moreover, in vitro studies can be used to assess the potential of a cell-polymer construct to integrate with adjacent tissues (Fig. 124.12). Some of the current research needs include: 1. Development of design criteria. Specific construct design criteria (e.g., required cell metabolic activity, ECM composition, and mechanical properties) need to be established based on the results of further in vitro and in vivo experimentation and phenomenological modeling of cell and tissue responses to environmental signals. 2. Selection of a cell source. Autologous BMSC and mature chondrocytes are the most likely immediate candidates; allogeneic fetal chondrocytes represent an option for the future. Both BMSC and chondrocytes maintain their chondrogenic potential when expanded in the presence of FGF and then cultured on 3D scaffolds.42,43 BMSC-based grafts recapitulate embryonic events of endochondral bone formation in response to local environmental factors10 resulting in repair tissue that


is relatively well integrated but can undergo ossification, leading to progressive thinning at the articular surface.11 On the other hand, chondrocyte-based grafts do not integrate as well12 but consist of phenotypically stable cartilage with a mechanically functional ECM37 that may promote survival in the presence of mechanical loading. Human fetal articular chondrocytes were recently demonstrated to regenerate hyaline cartilage when cultured at high densities in serum-free medium.31 However, this approach has unresolved ethical issues. 3. Selection of an appropriate scaffold for human chondrocytes and BMSC. Ideally, a tissue engineering scaffold should meet all of the following criteria: (1) reproducible processing into complex, 3D shapes, (2) highly porous structure that permits a spatially uniform cell distribution during cell seeding and minimizes diffusional constraints during in vitro cultivation, (3) controlled degradation at a rate matching that of cellular deposition of ECM (to provide a stable template during in vitro chondrogenesis) followed by complete elimination of the foreign material (to maximize long-term in vivo biocompatibility). 4. Development of methods for in vitro seeding and cultivation of human cells. Ideally, a tissue culture bioreactor should provide: (1) a means to achieve efficient, spatially uniform cell distribution throughout a scaffold (e.g., dynamic seeding), (2) uniform concentrations of biochemical species in the bulk phase and their efficient transport to the construct surface (e.g., well-mixed culture conditions), (3) steady state conditions (e.g., automated control based on biosensors triggering appropriate changes in medium and gas supply rates), to more closely mimic the in vivo cellular environment, and (4) applied physical forces (e.g., hydrostatic pressure and shear, dynamic compression), to more closely mimic the in vivo tissue environment. 5. Development of methods to promote graft survival, integration, and maturation. Articular cartilage repair refers to healing that restores a damaged articular surface without actually replicating the complete structure and function of the tissue while articular cartilage regeneration refers to the formation of a new tissue that is indistinguishable from normal articular cartilage, including the zonal organization, composition, and mechanical properties.2 Ideally, engineered cartilage should meet the criteria for regeneration, since any other repair tissue represents a mechanical discontinuity likely to cause long-term device failure.10 Local pre-treatment of the host tissues at the site of the defect (e.g., with proteolytic enzymes) may enhance graft integration.10 Further in vivo studies are needed to evaluate whether implanted constructs develop characteristic architectural features of articular cartilage in conjunction with physiological loading. This chapter describes technologies that can potentially lead to articular cartilage regeneration in vitro and in vivo. Representative studies are summarized in which cells (autologous or allogeneic chondrocytes or BMSC) were isolated, cultivated, and in some cases used to repair large full-thickness cartilage defects. Extension of these results to a human cell source and scale is expected to have a major clinical impact. At this time, tissue engineering studies are mainly observational. The increasing use of models to describe specific aspects of tissue formation (e.g., patterns of ECM deposition, structure-function correlations) is expected to help in the design of hypothesis-driven experiments and interpretation of their results.

Acknowledgments This work was supported by the National Aeronautics and Space Association (grant NAG9-836). The authors would like to thank R. Langer for general advice and I. Martin for reviewing the manuscript.

Defining Terms Bioreactors: Tissue culture vessels mixed by magnetic stirring or rotation. Bone marrow stromal cell (BMSC): A bone marrow-derived precursor cell with the potential to differentiate into various tissues including cartilage. Chondrocyte: A cartilage cell. Chondrogenesis: The process of cartilage formation.


Extracellular matrix (ECM): The biochemical components present in the extracellular space of a tissue, e.g., collagen type II and glycosoaminoglycan (GAG) in articular cartilage. Polymer scaffold: A synthetic material designed for cell cultivation, characterized by its specific chemical composition and 3D structure. Tissue construct: The tissue engineered in vitro using isolated cells, polymer scaffolds, and bioreactors.

References 1. Buckwalter, J.A., Mankin, H.J.: Articular cartilage. Part I: tissue design and chondrocyte-matrix interactions. J. Bone Joint Surg. 79-A, 600, 1997a. 2. Buckwalter, J.A., Mankin, H.J. Articular cartilage repair and transplantation. Arthrit. Rheum. 41, 1331, 1998. 3. Brittberg, M., Lindahl, A., Nilsson, A., Ohlsson., C., Isaksson, O. Peterson, L. Treatment of deep cartilage defects in the knee with autologous chondrocyte transplantation. NEJM 331, 889, 1994. 4. Arnst, C., Carey, J. Biotech Bodies. Business Week, July 27, 56, 1998. 5. Messner, K., Gillquist, J. Cartilage repair: a critical review. Acta Orthop.Scand. 67, 523, 1996. 6. Newman, A.P. Articular cartilage repair. Am. J. Sports Med. 26, 309, 1998. 7. Brittberg, M., Nilsson, A., Lindahl, A., Ohlsson, C., Peterson, L. Rabbit articular cartilage defects treated with autologous cultured chondrocytes. Clin. Orthop. Rel. Res. 326, 270, 1996. 8. Shortkroff, S., Barone, L., Hsu, H.P., Wrenn, C., Gagne, T., Chi, T., Breinan, H., Minas, T., Sledge, C.B., Tubo, R., Spector, M. Healing of chondral and osteochondral defects in a canine model: the role of cultured chondrocytes in regeneration of articular cartilage. Biomat. 17, 147, 1996. 9. Breinan, H.A., Minas, T., Barone, L., Tubo, R., Hsu, H.P., Shortkroff, S., Nehrer, S., Sledge, C.B., Spector, M. Histological evaluation of the course of healing of canine articular cartilage defects treated with cultured autologous chondrocytes. Tissue Eng. 4, 101, 1998. 10. Caplan, A. I., Elyaderani, M., Mochizuki, Y., Wakitani, S., Goldberg, V. M. Principles of cartilage repair and regeneration. Clin. Orthop. Rel. Res. 342, 254, 1997. 11. Wakitani, S., Goto, T., Pineda, S.J., Young, R.G., Mansour, J.M., Caplan, A. I., Goldberg, V.M. Mesenchymal cell-based repair of large, full-thickness defects of articular cartilage. J. Bone Joint Surg.. 76A, 579, 1994. 12. Kawamura, S, Wakitani, S, Kimura, T, Maeda, A., Caplan, A.I., Shino, K, Ochi, T. Articular cartilage repair-rabbit experiments with a collagen gel-biomatrix and chondrocytes cultured in it. Acta Orthop. Scand. 69, 56, 1998. 13. Butnariu-Ephrat, M., Robinson, D., Mendes, D.G., Halperin, N., Nevo, Z. Resurfacing of goat articular cartilage from chondrocytes derived from bone marrow. Clin. Orthop. Rel. Res. 330, 234, 1996. 14. Nevo, Z., Robinson, D., Horowitz, S., Hashroni, A., Yayon, A. The manipulated mesenchymal stem cells in regenerated skeletal tissues. Cell Transpl. 7, 63, 1998. 15. Chu, C., Coutts, R.D., Yoshioka, M., Harwood, F.L., Monosov, A.Z., Amiel, D. Articular cartilage repair using allogeneic perichondrocyte-seeded biodegradable porous polylactic acid (PLA): A tissue-engineering study. J. Biomed. Mat. Res. 29, 1147, 1995. 16. Chu, C., Dounchis, J.S., Yoshioka, M., Sah, R.L., Coutts, R.D., Amiel, D. Osteochondral repair using perichondrial cells: a 1 year study in rabbits. Clin. Orthop. Rel. Res. 340, 220, 1997. 17. Noguchi, T., Oka,M., Fujino, M., Neo, M., Yamamuro, T. Repair of osteochondral defects with grafts of cultured chondrocytes: comparison of allografts and isografts. Clin. Orthop. Rel. Res. 302, 251, 1994. 18. Freed, L.E., Grande, D.A., Emmanual, J., Marquis, J.C., Lingbin, Z., Langer, R. Joint resurfacing using allograft chondrocytes and synthetic biodegradable polymer scaffolds, J. Biomed. Mat. Res. 28, 891, 1994. 19. Grande, D.A., Southerland, S.S., Manji, R., Pate, D.W., Schwartz, R.E., Lucas, P.A. Repair of articular cartilage defects using mesenchymal stem cells. Tissue Eng. 1, 345, 1995.


20. Buschmann, M.D., Gluzband, Y.A., Grodzinsky, A.J., Kimura, J.H., Hunziker, E.B. Chondrocytes in agarose culture synthesize a mechanically functional extracellular matrix. J. Orthop. Res. 10, 745, 1992. 21. Buschmann, M.D., Gluzband, Y.A., Grodzinsky, A.J., Hunziker, E.B. Mechanical compression modulates matrix biosynthesis in chondrocyte/agarose culture. J. Cell Sci. 108, 1497, 1995. 22. Hauselmann, H.J., Fernandes, R.J., Mok, S.S., Schmid, T.M., Block, J.A., Aydelotte, M.B., Kuettner, K.E., Thonar, E.J.-M. Phenotypic stability of bovine articular chondrocytes after long-term culture in alginate beads. J. Cell Sci. 107, 17, 1994. 23. Haeuselmann, H.J., Masuda, K., Hunziker, E.B., Neidha, M., Mok, S.S., Michel, B.A., Thonar E.J.-M. Adult human chondrocytes cultured in alginate form a matrix similar to native human articular cartilage. Am. J. Physiol. 271, C742, 1996. 24. Freed, L.E., Marquis, J.C., Nohria, A., Mikos, A.G., Emmanual, J., Langer, R. Neocartilage formation in vitro and in vivo using cells cultured on synthetic biodegradable polymers. J. Biomed. Mat. Res. 27, 11, 1993. 25. Freed, L.E., Vunjak-Novakovic, G., Biron, R., Eagles, D., Lesnoy, D., Barlow, S., Langer, R. Biodegradable polymer scaffolds for tissue engineering. Bio/Technology 12, 689, 1994. 26. Grande, D.A., Halberstadt, C., Naughton, G., Schwartz, R., Manji, R. Evaluation of matrix scaffolds for tissue engineering of articular cartilage grafts. J. Biomed. Mat. Res. 34, 211, 1997. 27. Sittinger, M., Bujia, J., Minuth, W.W., Hammer, C., Burmester, G.R. Engineering of cartilage tissue using bioresorbable polymer carriers in perfusion culture. Biomaterials, 15, 451, 1994. 28. Dunkelman, N.S., Zimber, M.P., LeBaron, R.G., Pavelec,R., Kwan, M., Purchio, A.F. Cartilage production by rabbit articular chondrocytes on polyglycolic acid scaffolds in a closed bioreactor system. Biotech. Bioeng. 46, 299, 1995. 29. Carver, S.E., Heath, C.A. A semi-continuous perfusion system for delivering intermittent physiological pressure to regenerating cartilage. Tissue Eng. (in press) 1998. 30. Carver, S.E., Heath, C.A. Increasing extracellular matrix production in regenerating cartilage with intermittent physiological pressure. Biotech. Bioeng. (in press) 1998. 31. Adkisson, H.D., Maloney, W.J., Zhang, J., Hruska, K.A. Scaffold-independent neocartilage formation: a novel approach to cartilage engineering. Trans. Orth. Res. Soc. 23, 803, 1998. 32. Freed, L.E., Marquis, J.C., Vunjak-Novakovic, G., Emmanual, J., Langer, R. Composition of cellpolymer cartilage implants, Biotech. Bioeng. 43, 605. 1994. 33. Vunjak-Novakovic, G., Obradovic, B., Bursac, P., Martin, I., Langer, R. Freed, L.E. Dynamic seeding of polymer scaffolds for cartilage tissue engineering. Biotechnol. Progress 14, 193, 1998. 34. Freed, L.E., Vunjak-Novakovic, G., Marquis, J.C., Langer, R. Kinetics of chondrocyte growth in cell-polymer implants, Biotech. Bioeng. 43, 597, 1994. 35. Vunjak-Novakovic, G., Freed, L.E., Biron, R.J., Langer, R. Effects of Mixing on the Composition and Morphology of Tissue Engineered Cartilage, J. A.I.Ch.E. 42, 850, 1996. 36. Freed, L.E., Hollander, A.P., Martin, I., Barry, J., Langer, R., Vunjak-Novakovic, G. Chondrogenesis in a cell-polymer-bioreactor system. Exp. Cell Res. 240, 58, 1998. 37. Freed, L.E., Langer, R., Martin, I., Pellis, N., Vunjak-Novakovic, G., Tissue engineering of cartilage in space, PNAS 94, 13885, 1997. 38. Riesle, J., Hollander, A.P., Langer, R., Freed, L.E., Vunjak-Novakovic, G. Collagen in tissue engineered cartilage: types, structure and crosslinks. J. Cell. Biochem. 71, 313, 1998. 39. Vunjak-Novakovic, G., Martin, I., Obradovic, B., Treppo, S, Grodzinsky, A.J., Langer, R., Freed, L. Bioreactor cultivation conditions modulate the composition and mechanical properties of tissue engineered cartilage. J. Orthop. Res. (in press) 1998. 40. Sah, R.L., Trippel, S.B., Grodzinsky, A.J. Differential effects of serum, insulin-like growth factor-I, and fibroblast growth factor-2 on the maintenance of cartilage physical properties during longterm culture. J. Orthop. Res. 14, 44, 1996. 41. Armstrong, C.G., Mow, V.C. Variations in the intrinsic mechanical properties of human articular cartilage with age, degeneration, and water content. J. Bone Joint Surg. 44A, 88, 1982.


42. Martin, I., Padera, R.F., Vunjak-Novakovic, G., Freed, L.E. In vitro differentiation of chick embryo bone marrow stromal cells into cartilaginous and bone-like tissues. J. Orthop. Res. 16, 181, 1998. 43. Martin, I., Shastri, V.P., Langer, R., Vunjak-Novakovic, G., Freed, L.E. Engineering autologous cartilaginous implants. BMES Annual Fall Meeting, Ann. Biomed. Eng. 26, S-139, 1998. 44. Martin, I., Shastri, V.P., Padera, R.F., Langer, R., Vunjak-Novakovic G., Freed, L.E. Bone marrow stromal cell differentiation on porous polymer scaffolds. Trans. Orth. Res. Soc. 24, 57, 1999. 45. Shastri, V.P., Martin, I., Langer, R. A versatile approach to produce 3-D polymeric cellular solids. 4th US-Japan Symposium on Drug Delivery Systems, 1, 36, 1997. 46. Freed, L.E., Vunjak-Novakovic, G. Cultivation of cell-polymer constructs in simulated microgravity. Biotechnol. Bioeng. 46, 306, 1995. 47. Freed, L.E., Vunjak-Novakovic, G. Tissue Culture Bioreactors: Chondrogenesis as a Model System, In Principles of Tissue Engineering, R.P. Lanza, R. Langer and W.L. Chick, (eds.), Landes & Springer, 1997, chap. 11. 48. Williams, S.N.O., Burstein, D., Gray, M.L., Langer, R., Freed, L.E., Vunjak-Novakovic, G. MRI measurements of fixed charge density as a measure of glycosaminoglycan content and distribution in tissue engineered cartilage. Trans. Orth. Res. Soc. 23, 203, 1998. 49. Kim, YJ, Sah, RL, Doong, JYH et al. Fluorometric assay of DNA in cartilage explants using Hoechst 33258. Anal. Biochem. 174, 168, 1988. 50. Farndale, RW, Buttler, DJ, Barrett, AJ. Improved quantitation and discrimination of sulphated glycosaminoglycans by the use of dimethylmethylene blue. Biochim. Biophys. Acta. 883, 173, 1986. 51. Woessner, J.F. The determination of hydroxyproline in tissue and protein samples containing small proportions of this amino acid. Arch. Biochem. Biophys. 93, 440, 1961. 52. Hollander, AP, Heathfield, TF, Webber, C, Iwata, Y, Bourne, R, Rorabeck, C, Poole, RA: Increased damage to type II collagen in osteoarthritic articular cartilage detected by a new immunoassay. J. Clin. Invest. 93, 1722, 1994. 53. Bashir, A., Gray, M.L., Burstein, D. Gd-DTPA2 as a measure of cartilage degradation. Magn. Res. Med. 36, 665, 1996. 54. Obradovic, B., Freed, L.E., Langer, R., Vunjak-Novakovic, G. Bioreactor studies of natural and engineered cartilage metabolism. Fall meeting of the AIChE, Los Angeles, November, 1997. 55. Frank, E.H., Grodzinsky, A.J. Cartilage electromechanics: II. A continuum model of cartilage electrokinetics and correlation with experiments. J. Biomech. 20, 629, 1987. 56. Tachetti, C., Tavella, S., Dozin, B., Quarto, R., Robino, G. and Cancedda, R. Cell condensation in chondrogenic differentiation. Exp. Cell Res. 200, 26, 1992. 57. Kulyk, W.M., Coelho, C.N.D., Kosher, R.A. Type IX collagen gene expression during limb cartilage differentiation. Matrix 11, 282, 1991. 58. Treilleux, I., Mallein-Gerin, F., le Guellec, D., Herbage, D. Localization of the expression of type I, II, III collagen and aggrecan core protein genes in developing human articular cartilage. Matrix 12, 221, 1992. 59. Gerstenfeld, L.C., Landis, W.J. Gene expression and extracellular matrix ultrastructure of a mineralizing chondrocyte cell cul;ture system. J. Cell Biol. 112, 501, 1991. 60. Yannas, I.V. In vivo synthesis of tissues and organs, In Principles of Tissue Engineering, R.P. Lanza, R. Langer, and W. Chick (eds.), Academic Press & Landes, Austin, 1997, chap. 12. 61. Thompson, D.W. On Growth and Form. Cambridge University Press, New York, 1977. 62. Watt, F. The extracellular matrix and cell shape. TIBS 11, 482, 1986. 63. Bruckner, P., Hoerler, I., Mendler, M., Houze, Y., Winterhalter, K.H., Eiach-Bender, S.G., Spycher, M.A. Induction and prevention of chondrocyte hypertrophy in culture J. Cell Biol. 109, 2537, 1989. 64. Benya, P.D., Shaffer, J.D. Dedifferentiated chondrocytes reexpress the differentiated collagen phenotype when cultured in agarose gels. Cell 30, 215, 1982. 65. Bonaventure, J., Kadhom, N., Cohen-Solal, L., Ng, K.H., Bourguignon, J., Lasselin, C., Freisinger, P. Reexpression of cartilage-specific genes by dedifferentiated human articular chondrocytes cultured in alginate beads. Exp. Cell Res. 212, 97, 1994.


66. Vacanti, C., Langer, R., Schloo, B., Vacanti, J.P. Synthetic biodegradable polymers seeded with chondrocytes provide a template for new cartilage formation in vivo. Plast. Reconstr. Surg. 88, 753, 1991. 67. Watt, F. Effect of seeding density on stability of the differentiated phenotype of pig articular chondrocytes. J. Cell Sci. 89, 373, 1988. 68. Abbot, J., Holtzer, H. The loss of phenotypic traits by differentiated cells J. Cell Biol. 28, 473, 1966. 69. Gurdon, J.B. A community effect in animal development. Nature 336, 772, 1988. 70. Brockes, J.P. Amphibian limb regeneration: rebuilding a complex structure. Science 276, 81, 1997. 71. Zanetti, N.C., Solursh, M. Effect of cell shape on cartilage differentiation, In Cell Shape: Determinants, Regulation and Regulatory Role. Academic Press, New York, 1989, chap. 10. 72. Blunk, T., Sieminski, A.L., Nahir, M., Freed, L.E., Vunjak-Novakovic, G., Langer, R. Insulin-like growth factor-I (IGF-I) improves tissue engineering of cartilage in vitro. Proc. Keystone Symp. on Bone and Collagen: Growth and Differentiation, 1997. Paper #19 73. Osdoby, P., Caplan, A.I. Scanning electron microscopic investigation of in vitro osteogenesis. Calcif. Tissue Int. 30, 45, 1980. 74. Johnstone, B., Hering, T.M., Caplan, A.I., Goldberg, V.M., Yoo, J.U. In vitro chondrogenesis of bone marrow-derived mesenchymal progenitor cells. Exp. Cell Res. 238, 265, 1998. 75. Grodzinsky, A.J., Kamm, R.D., Lauffenburger, D.A. Quantitative aspects of tissue engineering: basic issues in kinetics, transport and mechanics, In Principles of Tissue Engineering, R.P.Lanza, R. Langer, and W. Chick (eds.), Academic Press & Landes, Austin, 1997, chap. 14. 76. Lightfoot, E.N. The roles of mass transfer in tissue function, In The Biomedical Engineering Handbook., J.D. Bronzino (ed.), CRC Press, Boca Raton, 1995, chap. 111. 77. O’Hara, B.P., Urban, J.P.G., Maroudas, A. Influence of cyclic loading on the nutrition of articular cartilage. Ann. Rheum. Dis. 49, 536, 1990. 78. Bursac, P.M., Freed, L.E., Biron, R.J., Vunjak-Novakovic, G. Mass transfer studies of tissue engineered cartilage. Tissue Eng. 2, 141, 1996. 79. Dunkan, R.L., Turner, C.H. Mechanotransduction and the functional response of bone to mechanical strain. Calc. Tissue Int. 57, 344, 1995. 80. Slowman, S.D. and Brandt, K.D. Composition and glycosaminoglycan metabolism of articular cartilage from habitually loaded and habitually unloaded sites. Arthrit. Rheum. 29, 88, 1986. 81. Kiviranta, I., Jurvelin, J., Tammi, M., Saamanen, A.-M., Helminen, H.J. Weight-bearing controls glycosaminoglycan concentration and articular cartilage thickness in the knee joints of young beagle dogs. Arthritis Rheum. 3, 801, 1987. 82. Sah, R.L.-Y., Kim, Y.J., Doong, J.Y.H., Grodzinsky, A.J., Plaas, A.H., Sandy, J.D. Biosynythetic response of cartilage explants to dynamic compression. J. Orthoped. Res. 7, 619, 1989. 83. Parkinnen, J.J., Ikonen, J., Lammi, M.J., Laakkonen, J., Tammi, M., Helminen, H.J. Effect of cyclic hydrostatic pressure on proteoglycan synthesis in cultured chondrocytes and articular cartilage explants. Arch. Biochem. Biophys. 300, 458, 1993. 84. Schneidermann, R., Keret, D., Maroudas, A. Effects of mechanical and osmotic pressure on the rate of glycosaminoglycan synthesis in the human adult femoral head cartilage: an in vitro study. J. Orthop. Res. 4, 393, 1986. 85. Kuettner, K.E., Pauli, B.U., Gall, G., Memoli, V.A., Schenk, R. Synthesis of cartilage matrix by mammalian chondrocytes in vitro. Isolation, culture characteristics, and morphology. J. Cell Biol. 93, 743, 1982. 86. Kuettner, K.E., Memoli, V.A., Pauli, B.U., Wrobel, N.C., Thonar, E.J.-M.A., Daniel, J.C. Synthesis of cartilage matrix by mammalian chondrocytes in vitro. maintenance of collagen and proteoglycan. J. Cell Biol. 93, 751, 1982. 87. Smith, R.L., Donlon, B.S., Gupta, M.K., Mohtai, M., Das, P., Carter, D.R., Cooke, J., Gibbons, G., Hutchinson, N., Schurman, D.J. Effect of fluid-induced shear on articular chondrocyte morphology and metabolism in vitro. J. Orthop. Res. 13, 824, 1995. 88. Franke, R.P., Grafe, M., Schnittler, H., Seiffge, D., Mittermayer, C. Induction of human vascular endothelial stress fibers by fluid shear stress. Nature 307, 648, 1984.


89. Wang, N., Ingber, D.E. Control of cytoskeletal mechanics by extracellular matrix, cell shape and mechanical tension. Biophys. J. 66, 2181, 1994. 90. Berthiaume, F., Frangos, J. Effects of flow on anchorage-dependent mammalian cells-secreted products, In Physical Forces and the Mammalian Cell, J. Frangos (ed.), Academic Press, San Diego, 1993, chap. 5. 91. Buckwalter, J.A., Mankin, H.J. Articular cartilage, part II: degeneration and osteoarthrosis, repair, regeneration, and transplantation. J. Bone Joint Surg. 79A, 612, 1997. 92. Obradovic, B., Martin, I., Padera, R.F., Treppo, S., Freed, L. E., Vunjak-Novakovic, G. Integration of engineered cartilage into natural cartilage, Annual Meeting of the AIChE, Miami, Nov. 1998.

Further Information 1. Buckwalter, J.A., Ehrlich, M.G., Sandell, L.J., Trippel, S.B. (eds.): Skeletal Growth and Development: Clinical Issues and Basic Science Advances, American Academy of Orthopaedic Surgeons, 1998. 2. Buckwalter, J.A., Mankin, H.J. Articular cartilage repair and transplantation. Arthrit. Rheum. 41, 1331, 1998. 3. Caplan, A. I., Elyaderani, M., Mochizuki, Y., Wakitani, S., Goldberg, V. M. Principles of cartilage repair and regeneration. Clin. Orthop. Rel. Res. 342, 254, 1997. 4. Comper, W.D. (ed): Extracellular Matrix: Molecular Components and Interactions, Hardwood Academic Publishers, the Netherlands, 1996. 5. Newman, A.P. Articular cartilage repair. Am. J. Sports Med. 26, 309, 1998.


Humes, H. D. “Tissue Engineering of the Kidney.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

125 Tissue Engineering of the Kidney 125.1

Fundamentals of Kidney Function Glomerular Ultrafiltration • Tubule Reabsorption • Endocrine

125.2

Tissue-Engineering Formulation Based Upon Fundamentals Bioartificial Glomerulus: The Filter • Bioartificial Tubule: The Reabsorber • Bioartificial Kidney • Bioartificial Endocrine Gland

H. David Humes University of Michigan

125.3 125.4

Clinical and Economic Implications Summary

Tissue engineering is one of he most intriguing and exciting areas in biotechnology due to its requirements for state-of-the-art techniques from both biologic and engineering disciplines. This field is on the threshold of the development of an array of products and devices comprised of cell components, biologic compounds, and synthetic materials to replace physiologic function of diseased tissues and organs. Successful tissue and organ constructs depend on a thorough understanding and creative application of molecular, cellular, and organ biology and the principles of chemical, mechanical, and material engineering to produce appropriate structure-function relationships to restore, maintain, or improve tissue or organ function. This approach depends on the most advanced scientific methodologies, including stem cell culture, gene transfer, growth factors, and biomaterial technologies. The kidney was the first solid organ whose function was approximated by a machine and a synthetic device. In fact, renal substitution therapy with hemodialysis or chronic ambulatory peritoneal dialysis (CAPD) has been the only successful long-term ex vivo organ substitution therapy to date [1]. The kidney was also the first organ to be successfully transplanted from a donor individual to an autologous recipient patient. However, the lack of widespread availability of suitable transplantable organs has kept kidney transplantation from becoming a practical solution to most cases of chronic renal failure. Although long-term chronic renal replacement therapy with either hemodialysis or CAPD has dramatically changed the prognosis of renal failure, it is not complete replacement therapy, since it only provides filtration function (usually on an intermittent basis) and does not replace the homeostatic, regulatory, metabolic, and endocrine functions of the kidney. Because of the nonphysiologic manner in which dialysis performs or does not perform the most critical renal functions, patients with ESRD on dialysis continue to have major medical, social, and economic problems [2]. Accordingly, dialysis should be considered as renal substitution rather than renal replacement therapy. Tissue engineering of a biohybrid kidney comprised of both biologic and synthetic components will most likely have substantial benefits for the patient by increasing life expectancy, increasing mobility and flexibility, increasing quality of life with large savings in time, less risk of infection, and reduced costs. This approach could also be considered a cure rather than a treatment for patients.


125.1 Fundamentals of Kidney Function The kidneys are solid organs located behind the peritoneum in the posterior abdomen and are critical to body homeostasis because of their excretory, regulatory, metabolic, and endocrinologic functions. The excretory function is initiated by filtration of blood at the glomerulus, which is an enlargement of the proximal end of the tubule incorporating a vascular tuft. The structure of the glomerulus is designed to provide efficient ultrafiltration of blood to remove toxic waste from the circulation yet retain important circulating components, such as albumin. The regulatory function of the kidney, especially with regard to fluid and electrolyte homeostasis, is provided by the tubular segments attached to the glomerulus. The ultrafiltrate emanating from the glomerulus courses along the kidney tubule, which resorbs fluid and solutes to finely regulate their excretion in various amounts in the final urine. The kidney tubules are segmented, with each segment possessing differing transport characteristics for processing the glomerular ultrafiltrate efficiently and effectively to regulate urine formation. The segments of the tubule begin with the proximal convoluted and straight tubules, where most salt and water are resorbed. This segment leads into the thin and thick segments of Henle’s loop, which are critical to the countercurrent system for urinary concentration and dilution of water. The distal tubule is next and is important for potassium excretion. The final segment is the collecting duct, which provides the final regulation of sodium, hydrogen, and water excretion. The functional unit of the kidney is therefore composed of the filtering unit (the glomerulus) and the regulatory unit (the tubule). Together they form the basic component of the kidney, the nephron. In addition to these excretory and regulatory functions, the kidney is an important endocrine organ. Erythropoietin, active forms of vitamin D, renin, angiotensin, prostaglandins, leukotrienes, and kallikrein-kinins are some of the endocrinologic compounds produced by the kidney. In order to achieve its homeostatic function for salt and water balance in the individual, the kidney has a complex architectural pattern (Fig. 125.1). This complex organization is most evident in the exquisitely regulated structure-function interrelationships between the renal tubules and vascular structures to coordinate the countercurrent multiplication and exchange processes to control water balance with urinary concentration and dilution [3]. This complex structure is the result of a long evolutionary process in which animals adapted to many changing environmental conditions. Although well suited for maintaining volume homeostasis, the mammalian kidney is an inefficient organ for solute and water excretion. In humans, approximately 180 liters of fluid are filtered into the tubules daily, but approximately 178 liters must be reabsorbed into the systemic circulation. Because of the importance of maintaining volume homeostasis in an individual to prevent volume depletion, shock, and death, multiple redundant physiologic systems exist within the body to maintain volume homeostasis. No such redundancy, however, exists to replace renal excretory function of soluble metabolic toxic by products of metabolic activity. Accordingly, chronic renal disease becomes a clinical disorder due to loss of renal excretory function and buildup within the body of metabolic toxins which require elimination by the kidneys. Because of the efficiency inherent in the kidneys as an excretory organ, life can be sustained with only 5–10% of normal renal excretory function. With the recognition that the complexity of renal architectural organization is driven by homeostatic rather than excretory function and that renal excretory function is the key physiologic process which must be maintained or replaced to treat clinical renal failure, the approach to a tissue engineering construct becomes easier to entertain, especially since only a fraction of normal renal excretory function is required to maintain life. For elimination of solutes and water from the body, the kidney utilizes simple fundamental physical principles which govern fluid movement (Fig. 125.2). The kidney’s goal in excretory function is to transfer solutes and water from the systemic circulation into tubule conduits in order to eliminate toxic byproducts from the body in a volume of only several liters. Solute and fluid removal from the systemic circulation is the major task of the renal filtering apparatus, the glomerulus. The force responsible for this filtration process is the hydraulic pressure generated within the system circulation due to myocardial contraction and blood vessel contractile tone. Most of the filtered fluid and solutes must be selectively reabsorbed by the renal tubule as the initial filtrate courses along the renal tubules. The reabsorptive process depends


upon osmotic forces generated by active solute transport by the renal epithelial cell and the colloid oncotic pressure within the peritubular capillary (Fig. 125.2). The approach of a tissue-engineering construct for renal replacement is to mimic these natural physical forces to duplicate filtration and reabsorption processes to attain adequate excretory function lost in renal disorders.

Glomerular Ultrafiltration The process of urine formation begins within the capillary bed of the glomerulus [4]. The glomerular capillary wall has evolved into a structure with the property to separate as much as one-third of the plasma entering the glomerulus into a solution of a nearly ideal ultrafitrate. The high rate of ultrafiltration across the glomerular capillary is a result of hydraulic pressure generated by the pumping action of the heart and the vascular tone of the preglomerular and postglomerular vessels as well as the high hydraulic permeability of the glomerular capillary walls. This hydraulic pressure as well as the hydraulic permeability of the glomerular capillary bed is at least two times and two orders of magnitude higher, respectively, than most other capillary networks within the body [5]. Despite this high rate of water and solute flux across the glom- FIGURE 125.1 Representation of the complex morerular capillary wall, this same structure retards the phologic architecture of a section of a mammalian kidney in which these components are schematized separately: filtration of important circulating macromolecules, (a) arterial and capillary blood vessels; (b) venous drainespecially albumin, so that all but the lower-molec- age; (c) two nephrons with their glomeruli and tubule ular-weight plasma proteins are restricted in their segments (Source: From [3] with permission.) passage across the filtration barrier. A variety of experimental studies and model systems have been employed to characterize the sieving properties of the glomerulus. Hydrodynamic models of solute transport through pores have been successfully used to describe the size selective barrier function of this capillary network to macromolecules [6]. This pore model, in its simplest form, assumes the capillary wall to contain cylindrical pores of identical size and that macromolecules are spherical particles. Based upon the steric hindrances that macromolecules encounter in the passage through small pores, whether by diffusion or by convection (bulk flow of fluid), definition of the glomerular capillary barrier can be a fluid-filled cylindrical pore (Fig. 125.3). This modeling characterizes the glomerular capillary barrier as a membrane with uniform pores of 50Å radius [7]. This pore size predicts that molecules with radii smaller than 14 Å appear in the filtrate in the same concentration as in plasma water. Since there is no restriction to filtration, fractional clearance of this size molecule is equal to one. The filtration of molecules of increasing size decreases progressively, so that the fractional clearance of macromolecules the size of serum albumin (36 Å) is low. The glomerular barrier, however, does not restrict molecular transfer across the capillary wall only on the basis of size (Fig. 125.4). This realization is based upon the observation that the filtration of the circulating protein, albumin, is restricted to a much greater extent than would be predicted from size alone. The realization that albumin is a polyanion at physiologic pH suggests that molecular charge, in addition to molecular size, is another important determination of filtration of macromolecules [8]. The


FIGURE 125.2

Physical forces which govern fluid transfer within the kidney.

greater restriction to the filtration of circulating polyanions, including albumin, is due to the electrostatic hindrance by fixed negatively charged components of the glomerular capillary barrier. These fixed negative charges, as might be expected, simultaneously enhance the filtration of circulating polycations. Thus, the formation of glomerular ultrafiltrate, the initial step in urine formation, depends upon the pressure and flows within the glomerular capillary bed and the intrinsic permselectivity of the glomerular capillary wall. The permselective barrier excludes circulating macromolecules from filtration based upon size as well as net molecular charge, so that for any given size, negatively charged macromolecules are restricted from filtration to a greater extent than neutral molecules.


FIGURE 125.3 Scanning electron micrograph of the glomerular capillary wall demonstrating the fenestrae (pores) of the endothelium within the glomerulus (mag × 50,400). Reprinted with permission from Schrier RW, Gottschalk CW. 1988. Diseases of the Kidney, p 12, Boston, Little, Brown.

FIGURE 125.4 Fractional clearances of negatively charged (sulfate) dextrans, neutral dextrans, and positively charged (DEAE) dextrans of varying molecular size. These data demonstrate that the glomerular capillary wall behaves as both a size-selective and charge-selective barrier [8].

Tubule Reabsorption Normal human kidneys form approximately 100 ml of filtrate every minute. Since daily urinary volume is roughly 2 L more than 98% of the glomerular ultrafiltrate must be reabsorbed by the renal tubule. The bulk of the reabsorption, 50–65%, occurs along the proximal tubule. Similar to glomerular filtration,


fluid movement across the renal proximal tubule cell is governed by physical forces. Unlike the fluid transfer across the glomerular capillary wall, however, tubular fluid flux is principally driven by osmotic and oncotic pressures rather than hydraulic pressure (Fig. 125.2). Renal proximal tubule fluid reabsorption is based upon active Na+ transport, requiring the energy-dependent Na+K+ ATPase located along the basolateral membrane of the renal tubule cell to promote a small degree of luminal hypotonicity [9]. This small degree of osmotic difference (2–3 mOsm/kgH2O) across the renal tubule is sufficient to drive isotonic fluid reabsorption due to the very high diffusive water permeability of the renal tubule cell membrane. Once across the renal proximal tubule cell, the transported fluid is taken up by the peritubular capillary bed due to the favorable oncotic pressure gradient. This high oncotic pressure within the peritubular capillary is the result of the high rate of protein-free filtrate formed in the proximate glomerular capillary bed [10]. As can be appreciated, the kidney has evolved two separate capillary networks to control bulk fluid flow from various fluid compartments of the body. The glomerular capillary network has evolved an efficient structure to function as a highly efficient filter to allow water and small solutes such as urea and sodium, to cross the glomerular capillary wall while retaining necessary macromolecules, such as albumin. This fluid transfer is driven by high hydraulic pressures within the glomerular capillary network generated by the high blood flow rates to the kidney and a finely regulated vascular system. The permselectivity of the filter is governed by an effective pore size to discriminate macromolecular sieving based upon both size and net molecular charge. The high postglomerular vascular resistance and the protein-free glomerular filtrate results, respectively, in low hydraulic pressure and high oncotic pressure within the peritubular capillary system which follows directly in series from the glomerular capillary network. The balance of physical forces within the peritubular capillaries, therefore, greatly favors the uptake of fluid back into the systemic circulation. The addition of a renal epithelial monolayer with high rates of active Na+ transport and high hydraulic permeability assists further in the high rate of salt and water reabsorption along the proximal tubule. Thus, an elegant system has evolved in the nephron to filter and reabsorb large amounts of fluid in bulk to attain high rates of metabolic product excretion while maintaining regulatory salt and water balance.

Endocrine As an endocrine organ, the kidney has been well recognized as critical in the production of erythropoietin, a growth factor for red blood cell production, and vitamin D, a compound important in calcium metabolism, along with, but not limited to, prostaglandins, kinins, and renin. For the purposes of this chapter, this discussion will be limited to erythropoietin production as an example of a potential formulation of a tissue-engineering construct to replace this lost endocrine function in chronic end-stage renal disease. More than 40 years ago erythropoietin was shown to be the hormone that regulates erythropoiesis, or red blood cell production, in the bone marrow [11]. In adults, erythropoietin is produced primarily (greater than 90%) by specialize interstitial cells in the kidney [12]. Although the liver also synthesizes erythropoietin, the quantity is not adequate (less than 10%) to maintain adequate red cell production in the body [13]. The production of erythropoietin by the kidney is regulated by a classic endocrinologic feedback loop system. As blood flows through the kidney, the erythropoietin-producing cells are in an ideal location to sense oxygen delivery to tissues by red cells in the bloodstream, since they are located adjacent to peritubular capillaries in the renal interstitium. As demonstrated in Fig. 125.5, erythropoeitin production is inversely related to oxygen delivery to the renal interstitial cells. With hypoxemia or decline in red blood cell mass, a decline in oxygen delivery occurs to these specialized cells, and increased erythropoietin production develops. Upon return to normal oxygen delivery with normoxia and red blood cell mass, the enhanced production of erythropoeitin is suppressed, closing the classic feedback loop. Of importance, the regulation of erythropoietin in the kidney cells depends upon transcriptional control, not upon secretory control, as seen with insulin [14]. The precise molecular mechanism of the oxygen sensor for tissue oxygen availability has not been delineated but appears to depend on a heme protein.


FIGURE 125.5

Endocrinologic back loop which regulates erythropoietin production by the kidney.

Once erythropoietin is released, it circulates in the bloodstream to the bone marrow, where it signals the marrow to produce red blood cells. In this regard, all blood cells, both white and red cells, originate from a subset of bone marrow cells, called multipotent stem cells. These stem cells develop during embryonic development and are maintained through adulthood via self-regulation. Under appropriate stimulation, stem cells proliferate and produce committed, more highly differentiated progenitor cells which are then destined to a specific differentiated line of blood cells, including neutrophils, lymphocytes, platelets, and red cells. Specifically, erythropoietin binds to receptors on the outer membrane of committed erythroid progenitor cells, stimulating the terminal steps in erythroid differentiation. Nearly 200 years ago, it was first recognized that anemia is a complication of chronic renal failure. Ordinarily an exponential increase in serum erythropoietin levels are observed when hemoglobin level in patients declines below 10 gm/dl. In the clinical state of renal disease, however, the normal increase in erythropoietin in response to anemia is impaired [14]. Although the oxygen delivery (or hemoglobin) to erythropoietin feedback loop is still intact, the response is dramatically diminished. In patients with endstage renal disease (ESRD), the hematocrit levels are directly correlated to circulating erythropoietin levels. In fact, ERSD patients with bilateral nephrectomies have the lowest hematocrits and lowest rates of erythropoiesis, demonstrating that even the small amount of erythropoietin produced by the endstage kidney is important. Thus, the loss of renal function due to chronic disease results in an endocrine deficiency of a hormone normally produced by the kidney and results in a clinical problem that complicates the loss of renal excretory function.

125.2 Tissue-Engineering Formulation Based upon Fundamentals In designing an implantable bioartificial kidney for renal replacement function, essential functions of kidney tissue must be utilized to direct the design of the tissue-engineering project. The critical elements of renal function must be replaced, including the excretory, regulatory (reabsorptive), and endocrinologic functions. The functioning excretory unit of the kidney, as detailed previously, is composed of the filtering unit, the glomerulus, and the regulatory or reabsorptive unit, the tubule. Therefore, a bioartificial kidney requires two main units, the glomerulus and the tubule, to replace renal excretory function.


Bioartificial Golmerulus: The Filter The potential for a bioartificial glomerulus has been achieved with the use of polysulphone fibers ex vivo with maintenance of ultrafiltration in humans for several weeks with a single device [15, 16]. The availability of hollow fibres with high hydraulic permeability has been an important advancement in biomaterials for replacement function of glomerular ultrafiltration. Conventional hemodialysis for ESRD has used membranes in which solute removal is driven by a concentration gradient of the solute across the membranes and is, therefore, predominantly a diffusive process. Another type of solute transfer also occurs across the dialysis membrane via a process of ultrafiltration of water and solutes across the membrane. This convective transport is independent of the concentration gradient and depends predominantly on the hydraulic pressure gradient across the membrane. Both diffusive and convective processes occur during traditional hemodialysis, but diffusion is the main route of solute movement. The development of synthetic membranes with high hydraulic permeability and solute retention properties in convenient hollow fiber form has promoted ERSD therapy based upon convective homofiltration rather than diffusive hemodialysis [17, 18]. Removal of uremic toxins, predominantly by the convective process, has several distinct advantages, because it imitates the glomerular process of toxin removal with increased clearance of higher-molecular-weight solutes and removal of all solutes (up to a molecular weight cutoff) at the same rate. The comparison of the differences between diffusive and convective transport across a semipermeable membrane is detailed in Fig. 125.6. This figure demonstrates the relationship between molecular size and clearance by diffusion and convection. As seen, the clearance of a molecule by diffusion is negatively correlated with the size of the molecule. In contrast, clearance of a substance by convection is independent of size up to a certain molecular weight. The bulk movement of water carries passable solutes along with it in approximately the same concentration as in the fluid.

FIGURE 125.6 Relationship between solute molecular size and clearance by diffusion (open circles) and convection (closed circles). Left curve shows solute clearance for a 1.0 m2 dialyzer with no ultrafiltration where solute clearance is diffusion. Right curve shows clearance for a 1.6 m2 ultrafilter where solute clearance is by convection. Smaller molecules are better cleared by diffusion; larger molecules by convection. Normal kidneys clear solutes in a pattern similar to convective transport. (Figure adapted from [18].)


FIGURE 125.7

Conceptual schematization of bioartificial glomerulus.

Development of an implantable device which mimics glomerular filtration will thus depend upon convective transport. This physiologic function has been achieved clinically with the use of polymeric hollow fibers ex vivo. Major limitations to the currently available technology for long-term replacement of filtration function include bleeding associated with required anticoagulation, diminution of filtration rate due to protein deposition in the membrane over time or thrombotic occlusion, and large amounts of fluid replacement required to replace the ultrafiltrate formed from the filtering unit. The use of endothelial-cell-seeded conduits along filtration surfaces may provide improved long-term hemocompatibility and hemofiltaration in vivo [19, 20. 21], as schematized in Fig. 125.7. In this regard, endothelial cell seeding of small-caliber vascular prosthesis has been shown experimentally to reduce long-term platelet deposition, thrombus formation, and loss of graft patency [21]. Recent results in humans have demonstrated success in autologous endothelial cell seeding in small caliber grafts after growth of these cells along the graft lumen ex vivo to achieve a confluent monolayer prior to implantation. Long-term persistent endothelialization and patency of the implanted graft has been reported [19]. A potential rate-limiting step in endothelial-cell-lined hollow fibers of small caliber is thrombotic occlusion, which limits the functional patency of this filtration unit. In this regard, gene transfer into seeded endothelial cells for constitutive expression of anticoagulant factors can be envisioned to minimize clot formation in these small-caliber hollow fibers. Since gene transfer for in vivo protein production has been clearly achieved with endothelial cells [22, 23], gene transfer into endothelial cells for the production of an anticoagulant protein is clearly conceivable. For differentiated endothelial cell morphology and function, an important role for various components of the extracellular matrix (ECM) has been demonstrated [24, 25]. The ECM has been clearly shown to dictate phenotype and gene expression of endothelial cells, thereby modulating morphogenesis and growth. Various components of ECM, including collagen by type I, collagen type IV, laminin, and fibronectin, have been shown to affect endothelial cell adherence, growth, and differentiation. Of importance, ECM produced by MDCK cells, a permanent renal epithelial cell line, has the ability to induce capillary endothelial cells to produce fenestrations [25, 26]. Endothelial cell fenestrations are large openings which act as channels or pores for convective transport though the endothelial monolayer and are important in the high hydraulic permeability and sieving characteristics of glomerular capillaries. Thus, the ECM component on which the endothelial cells attach and grow may be critical in the functional characteristics of the lining monolayer.

Bioartificial Tubule: The Reabsorber As detailed above, the efficiency of reabsorption, even though dependent upon natural physical forces governing fluid movement across biologic as well as synthetic membranes, requires specialized epithelial cells to perform vectorial solute transport. Critical to the advancement of this tubule construct, as well as for the tissue-engineering field in general, is the need for the isolation and growth in vitro of specific cells, referred to as stem progenitor cells, from adult tissues.


Stem or Progenitor Cells These cells are those that possess stem-cell-like characteristics with a high capacity for self-renewal under defined conditions into specialized cells to develop correct structure and functional components of a physiologic organ system [27, 28, 29]. Stem cells have been extensively studied in three adult mammalian tissues: the hematopoietic system, the epidermis, and the intestinal epithelium. Recent work has also suggested that stem cells may also reside in the adult nervous system [30]. Little insight into possible renal tubule stem cells had been developed until recent data demonstrating methodology to isolate and grow renal proximal tubule stem or progenitor cells from adult mammalian kidneys [31, 32]. This series of studies was promoted by the clinical and experimental observations suggesting that renal proximal tubule progenitor cells must exist, because they have the ability to regenerate after severe neophrotoxic or ischemic injury to form a fully functional and differentiated epithelium [33, 34]. Whether proximal tubule progenitor cells are pluripotent, possessing the ability to differentiate into cells of other segments (such as loop of Henle, distal convoluted tubule) as in embryonic kidney development, is presently unclear; the clinical state of acute tubular necrosis certainly supports the idea that proximal tubule progenitor cells have the ability to replicate and differentiate into proximal tubule cells with functionally and morphologically differentiated phenotypes. In this regard, recent data have demonstrated, using renal proximal tubule cells in primary culture, that the growth factors transforming growth factor-β1 (TGF-β1) and the epidermal growth factor (EGF), along with the retinoid, retinoic acid, promoted tubulogenesis in renal proximal tubule progenitor cells in tissue culture [31]. These observations defined a coordinated interplay between growth factors and retinoids to induce pattern formation and morphogenesis. This finding is one of the first definitions of inductive factors which may be important in the organogenesis of a mammalian organ. In addition, using immunofluorescence microscopy, retinoic acid induced laminin A- and B1-chain production in these cells and purified soluble laminin completely substituted for retinoic acid in kidney tubulogenesis. These results clearly demonstrate the manner in which retinoic acid, as a morphogen, can promote pattern formation and differentiation by regulating the production of an extracellular matrix molecule. Further work has demonstrated, in fact, that a population of cells resides in the adult mammalian kidney which have retained the capacity to proliferate and morphogenically differentiate into tubule structures in vitro [32]. These experiments have identified non-serum-containing growth conditions, which select for proximal tubule cells with a high capacity for self-renewal and an ability to differentiate phenotypically, collectively and individually, into proximal tubule structures in collagen gels. Regarding the high capacity for self-renewal, genetic marking of the cells with a recombinant retrovirus containing the lacZ gene and dilution analysis demonstrated that in vitro tubulogenesis often arose from clonal expansion of a single genetically tagged progenitor cell. These results suggest that a population of proximal tubule cells exist within the adult kidney in a relatively dormant, slowly replicative state, but with a rapid potential to proliferate, differentiate, and pattern-form to regenerate the lining proximal tubule epithelium of the kidney following severe ischemic or toxic injury. Bioartificial Tubule Formulation The bioartificial renal tubule is now clearly feasible when conceived as a combination of living cells supported on polymeric substrata [35]. A bioartificial tubule uses epithelial progenitor cells cultured on water and solute-permeable membranes seeded with various biomatrix materials so that expression of differentiated vectorial transport and metabolic and endocrine function is attained (Figs. 125.8 and 125.9). With appropriate membranes and biomatrices, immunoprotection of cultured progenitor cells can be achieved concurrent with long-term functional performance as long as conditions support tubule cell viability [35, 36]. The technical feasibility of an implantable epithelial cell system derived from cells grown as confluent monolayers along the luminal surface of polymeric hollow fibers has been achieved [35]. These previously constructed devices, however, have used permanent renal cell lines which do not have differentiated transport function. The ability to purify and grow renal proximal tubule progenitor cells with the ability to differentiate morphogenically may provide a capability for replacement renal tubule function.


FIGURE 125.8 Ligh micrograph of an H&E section (100×) of hollow fiber lined with collagen type IV and confluent monolayer of human renal tubule epithelial cells along the inner component of the fiber. In this fixation process, the hollow fiber is clear with the outer contour of the hollow fiber identified by the irregular line (disregard artifact in lower left quadrant).

A bioartificial proximal tubule satisfies a major requirement of reabsorbing a large volume of filtrate to maintain salt and water balance within the body. The need for additional tubule equivalents to replace another nephronal segment function, such as the loop of Henle, to perform more refined homeostatic elements of the kidney, including urine concentration or dilution, may not be necessary. Patients with moderate renal insufficiency lose the ability to finely regulate salt and water homeostasis—because they are unable to concentrate or dilute, yet are able to maintain reasonable fluid and electrolyte homeostasis due to redundant physiologic compensation via other mechanisms. Thus, a bioartificial proximal tubule, which reabsorbs iso-osmotically the majority of the filtrate, may be sufficient to replace required tubular function to sustain fluid electrolyte balance in a patient with end-stage renal disease.

Bioartificial Kidney The development of a bioartificial filtration device and a bioartificial tubule processing unit would lead to the possibility of an implantable bioartificial kidney, consisting of the filtration device followed in series by the tubule unit (Fig. 125.10). The filtrate formed by this device will flow directly into the tubule unit. The tubule unit should maintain viability, because metabolic substrates and low-molecular-weight growth factors are delivered to the tubule cells from the ultrafiltration unit. Furthermore, immunoprotection of the cells grown within the hollow fiber is achievable due to the impenetrance of immunologically competent cells through the hollow fiber. Rejection of transplanted cells will, therefore, not occur. This arrangement thereby allows the filtrate to enter the internal compartments of the hollow fiber network, which are lined with confluent monolayers of renal tubule cells for regulated transport function. This device could be used either extracoporeally or implanted within a patient. In this regard, the specific implant site for a bioartificial kidney will depend upon the final configuration of both the bioartificial filtration and tubule device. As presently conceived, the endothelial-line bioartificial filtration hollow fibers can be placed into an arteriovenous circuit using the common iliac artery and vein, similar to the surgical connection for a renal transplant. The filtrate is connected in series to a bioartificial proximal tubule, which is embedded into the peritoneal membrane, so that reabsorbate will be transported into the peritoneal cavity and reabsorbed into the systemic circulation. The processed filtrate


FIGURE 125.9 Electorn micrograph of a single hollow fiber lined with extracellular matrix and a confluent monolayer (see Fig. 125.8) of renal tubule cells along the inner component of the fiber. As displayed, the differentiated phenotype of renal tubule cells on the hollow fiber prelined with matrix is apparent. The well developed microvilli and apical tight junctions can be appreciated (14,000×).

exiting the tubule unit is then connected via tubing to the proximate ureter for drainage and urine excretion via the recipient’s own urinary collecting system. Although an implantable form of this therapy is in the conceptual phase of development, a functioning extracorporeal bioartificial renal tubule has been developed. The technical feasibility of an epithelial cell system derived from cells grown as confluent monolayers along the luminal surface of polymeric hollow fibers has been recently achieved [38]. Because of its anatomic and physiological similarities with humans and the relative simplicity with which it can be bred in large numbers, the pig is currently considered the best source of organs for both human xenotransplantation and immuno-isolated cell therapy devices [39–41]. Yorkshire breed pig renal tubule progenitor cells have been expanded and cultured on semipermeable hollow fiber membranes seeded with various biomatrix materials, so that expression of differentiated vectorial transport, metabolic, and endocrine function has been attained [38, 42, 43]. This


FIGURE 125.10 Conceptual schematization of an implantable tissue engineered bioartificial kidney with an endothelial-cell-lined hemofilter in series with a proximal tubule-cell-lined reabsorber.

bioartificial renal tubule has been evaluated in uremic and non-uremic large animals [44]. A customized extracorporeal circuit using standard arterial venous blood tubing with a dialysis machine has been developed which delivers the post-filtered blood through the extracapillary space and the ultrafiltrate through the liminal space of the bioartificial tubule. Experiments have confirmed the functional metabolic performance and active tubule transport properties of the bioartificial renal tubule using this extracorporeal circuit [44]. Proximal tubule cells derive embryonically from mesodermal progenitors closely related to bone marrow precursor cells and have retained many elements of immunologically competent cells, including the ability for antigen presentation and production of a variety of immunologically active cytokines [45]. Since the kidney participates in this complex and dynamic network of pro- and antiinflammatory cytokines, cell replacement therapy with an extracorporeal bioartificial renal tubule assist device may play a critical role in the future treatment of renal failure.

Bioartificial Endocrine Gland The Erythropoietin Generator Because specialized cells are programmed to carry out specific biologic tasks, cell therapy may deliver several key proteins in a coordinated cascade to promote a biologic or physiologic process. Targeted delivery of a specific deficient protein, hormone, or neurotransmitter may be achieved with site-specific implantation of cells which can produce this deficient compound after being encapsulated in special polymeric membranes. The membranes allow cell nutrients into the encapsulated space to maintain cell viability and allow cellular metabolic wastes to exit along with the desired protein, hormone, or neurotransmitter while shielding the cells from the host’s destructive immune response. This strategy is being employed, for example, to deliver dopamine produced by bovine adrenal cells to the substantia nigra where a deficiency of this neurotransmitter at this site leads to Parkinson’s Disease [46]. Regulated and homeostatic drug dosing may also be achieved with cell therapy. For hormonal therapy, such as insulin for diabetes mellitus, appropriate insulin levels within the body are only crudely attained with a once-a-day or twice-a-day dosing. Any hormone-producing cell has a highly evolved biologic sensing system to monitor the ambient environment and respond with graded production and release of the hormone to regulate the sensed level of the moiety which is regulated. The circulating level of a protein or a hormone may be regulated at several levels: at the gene level by transcriptional mechanisms, at the protein level by translational processes, or at the secretory level by cellular processes. The complexity of regulation increases several-fold, as control progresses from transcriptional to translational to excretory processes. Accordingly, a more refined differentiated cell phenotype is required to maintain a regulated


secretory process compared to a transcriptional process. The lack of success of encapsulation of insulinproducing cells is due to the fact that the cells are unable to maintain a viable, highly differentiated state to sense ambient glucose levels and release performed insulin in a regulated, differentiated secretory pathway. In contrast, since erythropoietin production is regulated by transcriptional mechanisms, the ability to identify and perhaps grow cells from adult mammalian kidneys with the ability to regulate erythropoietin production in response to oxygen delivery may allow the design of an implantable cell therapy device to sense circulating oxygen levels and regulate erythropoeitin production based upon a biologic sensing mechanism. Recently, genetically engineered polymer encapsulated myoblasts have been shown to continuously deliver human and mouse erythropoietin in mice [47].

125.3 Clinical and Economic Implications Although long-term chronic renal replacement therapy with either hemodialysis or CAPD has dramatically changed the prognosis of renal failure, it is not complete replacement therapy, since it only provides filtration function (usually on an intermittent basis) and does not replace the excretory, regulatory, and endocrine functions of the kidney. Because of the nonphysiologic manner in which dialysis performs or does not perform the most critical renal functions, patients with ESRD on dialysis continue to have major medical, social, and economic problems. Renal transplant addresses some of these issues, but immunologic barriers and organ shortages keep this approach from being ideal for a large number of ESRD patients. Although dialysis or transplantation therapies can prolong the life of a patient with ESRD disease, it is still a serious medical condition, with ESRD patients having only one-fifth the life expectancy of a normal age-matched control group. ESRD patients also experience significantly greater morbidity. Patients with ESRD have five times the hospitalization rate, nearly twice the disability rate, and five times the unemployment rate of age-matched non-ESRD individuals [2]. Accordingly, this new technology based upon the proposed bioengineering prototypes would most likely have substantial benefits to the patient by increasing life expectancy, increasing mobility and flexibility, increasing quality of life with large savings in time, less risk of infection, and reduced costs. Besides the personal costs to the patient and family, care of chronic kidney failure is monetarily expensive on a per-capita basis in comparison to most forms of medical care [1, 2]. The 1989 estimated Medicare payment (federal only) per ESRD patient during the entire year averaged $30,900. The patient and private insurance obligations were an addition $6900 per patient. The total cost of a patient per year with end-stage renal disease, therefore, is approximately $40,000. In 1988 the expected life span after beginning dialysis for an ESRD patient was approximately four years; therefore, the total cost per patient of ESRD during his or her lifetime was approximately $160,000 in 1998. The total cost in direct medical payments for ESRD, by both public and private payers, increased from $6 billion in 1989 to over $14 billion in 1996 [48]. These estimates do not include a number of indirect cost items, since they do not include patient travel costs and lost labor production. The number of patients receiving chronic dialytic therapy in the U.S. is presently over 250,000 with a current growth rate at 8 to 10% per year. It is conceivable that in the not too distant future, tissue engineering technology could supplant current treatments for ESRD. Although it is difficult to estimate the value of a technology, on a purely economic basis there may be an opportunity for major cost savings with this technology.

125.4 Summary Three technologies will most likely dominate medical therapeutics in the next century. One is “cell therapy”—the implantation of living cells to produce a natural substance in short supply from the patient’s own cells due to injury and destruction from various clinical disorders. Erythropoietin cell therapy is an example of this approach to replace a critical hormone deficiency in end-stage renal disease. A second therapy is tissue engineering, wherein cells are cultured to replace masses of cells that normally function


in a coordinated manner. Growing a functional glomerular filter and tubule reabsorber from a combination of cells, biomaterials, and synthetic polymers to replace renal excretory and regulatory functions is an example of this formulation. Over the last few years, an extracorporeal bioartificial renal tubule has become a reality; demonstrating physiologic and biochemical regulatory function in large animal studies. Finally, a third technology that will dominate future therapeutics is gene therapy, in which genes are transferred into living cells either to deliver a gene product to a cell in which it is missing or to produce a foreign gene product by a cell to promote a new function. The use of genes which encode for anticoagulant proteins as a means to deliver in a targeted and local fashion an anticoagulant to maintain hemocompatibility of a tissue engineered hemofilter is an example of the application of this third technology. The kidney was the first organ whose function was substituted by an artificial device. The kidney was also the first organ to be successfully transplanted. The ability to replace renal function with these revolutionary technologies in the past was due to the fact that renal excretory function is based upon natural physical forces which govern solute and fluid movement from the body compartment to the external environment. The need for coordinated mechanical or electrical activities for renal substitution was not required. Accordingly, the kidney may well be the first organ to be available as a tissue-engineered implantable device as a fully functional replacement part for the human body.

References 1. Iglehart JK. 1993. The American Health Care System: The End Stage Renal Disease Program. N Engl J Med 328:366. 2. Excerpts from United States Renal Data System 1991 Annual Data Report. Prevalence and cost of ESRD therapy. Am J Kidney Diseases 18(5)(supp)2:21. 3. Kriz W, Lever AF. 1969. Renal countercurrent mechanisms: Structure and function. Am Heart J 78:101. 4. Brenner BM, Humes HD. 1977. Mechanisms of glomerular ultra-filtration. N Engl J Med 297:148. 5. Landis EM, Pappenheimer JR. Exchange of substances through the capillary walls. In WF Hamilton, P Dow (eds), Handbook of Physiology: Circulation, sec 2, vol 2, p 961, Washington DC, American Physiological Society. 6. Anderson JL, Quinn JA. 1974. Restricted transport in small pores. A model for steric exclusion and hindered particle motion. Biophys J 14:130. 7. Chang RLS, Robertson CR, Deen WM, et al. 1975. Permselectivity of the glomerular capillary wall to macromolecules: I. Theoretical considerations. Biophys J 15:861. 8. Brenner BM, Hostetter TH, Humes HD. 1978. Molecular basis of proteinuria of glomerular origin. N Engl J Med 298:826. 9. Andreoli TE, Schafer JA. 1978. Volume absorption in the pars recta: III. Luminal hypotonic-sodium reabsorption. Circ Res 52:491. 10. Knox FG, Mertz JI, Burnett JC, et al. 1983. Role of hydrostatic and oncotic pressures in renal sodium reabsorption. Circ Res 52:491. 11. Jacobson LO, Goldwasser E, Fried W, et al. 1957. Role of kidney in erythropoiesis. Nature. 179:633. 12. Maxwell PH, Osmond MK, Pugh CW, et al. 1993. Identification of the renal erythropoietinproducing cells using transgenic mice. Kidney Int 441:1149. 13. Fried W. 1972. The liver as a source of extrarenal erythropoeitin production. Blood 49:671. 14. Jelkmann W. 1992. Erythropoietin: Structure, control of production, and function. Physiolog Rev 72(2):449. 15. Golper TA. 1986. Continuous arteriorvenous hemofiltration in acute renal failure. Am J Kidney Dis 6:373. 16. Kramer P, Wigger W, Rieger J, et al. 1977. Arterior-venous hemofiltration: A new and simple method for treatment of overhydrated patients resistant to diuretics. Klin Wochenschr 55:1121. 17. Colton CK, Henderson LW, Ford CA, et al. 1975. Kinetics of hemodiafiltration. In vitro transport characteristics of a hollow-fiber blood ultrafilter. J Lab Clin Med 85:355.


18. Henderson LW, Colton CK, Ford CA. 1975. Kinetics of hemodiafiltration: II. Clinical characterization of a new blood cleansing modality. J Lab Clin Med 85:372. 19. Kadletz M, Magometschnigg H, Minar E., et al. 1992. Implantation of in vitro endothelialized polytetrafluoroethylene grafts in human beings. J Thorac Cardiovasc Surg 104:736. 20. Schnider PA, Hanson SR, Price TM, et al. 1988. Durability of confluent endothelial cell monolayers of small-caliber vascular prostheses in vitro. Surgery 103:456. 21. Shepard AD, Eldrup-Jorgensen J, Keough EM, et al. 1986. Endothelial cell seeding of small caliber synthetic grafts in the baboon. Surgery 99:318. 22. Zweibel JA, Freeman SM, Kantoff PW, et al. 1989. High-level recombinant gene expression in rabbit endothelial cells transduced by retroviral vectors. Science 243:220. 23. Wilson JM, Birinyi LK, Salomon RN, et al. 1989. Implantation of vascular grafts lined with genetically modified endothelial cells. Science 244:1344. 24. Carey DJ. 1991. Control of growth and differentiation of vascular cells by extacellular matrix proteins. Annu Rev Physiol 53:161. 25. Carley WW, Milici AJ, Madri JA. 1988. Extracellular matrix specificity for the differentiation of capillary endothelial cells. Exp Cell Res 178:426. 26. Milici AJ, Furie MB, Carley WW. 1985. The formation of fenestrations and channels by capillary endothelium in vitro. Proc Natl Acad Sci 82:6181. 27. Garlick JA, Katz AB, Fenjves ES, et al. 1991. Retrovirus-mediated transduction of cultured epidermal keratinocytes. J Invest Dermatol 97:824. 28. Hall PA, Watt FM. 1989. Stem cells: The generation and maintenance of cellular diversity. Development 106:619. 29. Potten CS, Lieffler M. 1990. Stem cells; lessons for and from the crypt. Development 110:1001. 30. Reynolds BA, Weiss S. 1992. Generation of neurons and astrocytes from isolated cells of the adult mammalian central nervous system. Science 255:1707. 31. Humes HD, Cieslinski DA. 1992. Interaction between growth factors and retinoic acid in the induction of kidney tubulogenesis. Exp Cell Res 201:8. 32. Humes HD, Krauss JC, Cieslinski DA, et al. Tubulogenesis from isolated single cells of adult mammalian kidney: Clonal analysis with a recombinant retrovirus (submitted). 33. Coimbra T, Cieslinski DA, Humes HD. 1990. Exogenous epidermal growth factor enhances renal repair in mercuric chloride-induced acute renal failure. AM J Physiol 259:F 483. 34. Humes HD, Cieslinski DA, Coimbra T, et al. 1989. Epidermal growth factor enhances renal tubule cell regeneration and repair and accelerates the recovery of renal function in postischemic acute renal failure. J Clin Invest 84:1757. 35. Ip TK, Aebischer P. 1989. Renal epithelial-cell-controlled solute transport across permeable membranes as the foundation for a bioartificial kidney. Artif Organs 13:58. 36. Aebischer P, Wahlberg L, Tresco PA, et al. 1991. Macroencapsulation of dopamine-secreting cells by coextrusion with an organic polymer solution. Biomaterials 12:50. 37. Tai IT, Sun AM. 1993. Microencapsulation of recombinant cells: A new delivery system for gene therapy. FASEB J 7:1061. 38. McKay SM, Funke AJ, Buffington DA, Humes HD. Tissue engineering of a bioartificial renal tubule. ASAIO Journal. 44:179–183, 1998. 39. Cozzi E and White D. The generation of transgenic pigs as potential organ donors for humans. Nature Medicine 1:965–966, 1995. 40. Cooper DKC, Ye Y, Rolf JLL and Zuhdi N. The pig as potential organ donor for man. In: Cooper DKC, Kemp E, Reemtsma K, White DJG (eds): Xeno-Transplantation. Springer, Berlin. 1991. pp481–500. 41. Calne RY. Organ transplantation between widely disparate species. Transplant Proc 2:550–553, 1970. 42. Humes HD and Cieslinski DA. Interaction between growth factors and retinoic acid in the induction of kidney tubulogenesis. Exp Cell Res 201:8–15, 1992.


43. Humes HD, Krauss JC, Cieslinski DA and Funke AJ. Tubulogenesis from isolated single cells of adult mammalian kidney: clonal analysis with a recombinant retrovirus. Am J Phsyiol: Renal 271(40):F42–F49, 1996. 44. Weitzel WF, Browning A, Buffington DA, Funke AJ, Gupte A, MacKay S, Humes HD. Analysis of a Renal Proximal tubule Assist Device (RAD) during CVVH on uremic dogs, J Am Soc of Nephrol Abstract, 31st Annual Meeting, 1998 (In press). 45. Ong ACM and Fine LG. Tubular-derived growth factors and cytokines in the pathogenesis of tubulointerstitial fibrosis: implications for human renal disease progression. Am J Kidney Dis 23:205–209, 1994. 46. Aebischer P, Tresco PA, Winn SR, et al. Long-term cross-species brain transplantation of a polymerencapsulated dopamine-secreting cell line. Exp Neurol 111:269. 47. Relulier E, Schneider BL, Deglon N, Beuzard Y, Aebischer P. Continuous delivery of human and mouse erythropoietin in mice by genetically engineered polymer encapsulated myoblasts. Gene Therapy 5:1014–1022, 1998. 48. Excerpts from the United States Renal Data System 1998 Annual Data Report. The economic cost of ESRD and Medicare spending for alternative modalities of treatment. J Kidney Diseases 32(2)(supp)1:S118.


Galletti, P. M., Nerem , R. M. “Hard Tissue Replacement.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

XIII Prostheses and Artificial Organs Pierre M. Galletti (deceased) and Robert M. Nerem 126 Artificial Heart and Circulatory Assist Devices

Gerson Rosenberg

Engineering Design • Engineering Design of Artificial Heart and Circulatory Assist Devices • Conclusions

127 Cardiac Valve Prostheses

Ajit P. Yoganathan

A Brief History of Heart Valve Prostheses • Current Types of Prostheses • Tissue Versus Mechanical • Engineering Concerns and Hemodynamic Assessment of Prosthetic Heart Valves • Implications for Thrombus Deposition • Durability • Current Trends in Valve Design • Conclusion

128 Vascular Grafts

David N. Ku, Robert C. Allen

History of Vascular Grafts • Synthetic Grafts • Regional Patency • Thrombosis • Neointimal Hyperplasia • Graft Infections

129 Artificial Lungs and Blood-Gas Exchange Devices Clark K. Colton

Pierre M. Galleti,

Gas Exchange Systems • Cardiopulmonary Bypass • Artifical Lung Versus Natural Lung • Oxygen Transport • CO2 Transport • Coupling of O2 and CO2 Exchange • Shear-Induced Transport Augmentation and Devices for Improved Gas Transport

130 Artificial Kidney

Pierre M. Galletti, Clark K. Colton, Michael J. Lysaght

Structure and Function of the Kidney • Kidney Disease • Renal Failure • Treatment of Renal Failure • Renal Transplantation • Mass Transfer in Dialysis • Clearance • Filtration • Permeability • Overall Transport • Membranes • Hemofiltration • Pharmacokinetics • Adequacy of Dialysis • Outlook

131 Peritoneal Dialysis Equipment

Michael J. Lysaght, John Moran

Therapy Format • Fluid and Solute Removal • The Peritoneal Membrane: Physiology and Transport Properties • Transport Modeling • Emerging Developments

132 Therapeutic Apheresis and Blood Fractionation

Andrew L. Zydney

Plasmapheresis • Plasma Perfusion • Cytapheresis • Summary

133 Liver Support Systems

Pierre M. Galletti, Hugo O. Jauregui

Morphology of the Liver • Liver Functions • Hepatic Failure • Liver Support Systems • Global Replacement of Liver Function • Hybrid Replacement Procedures • Outlook


134 Artificial Pancreas Gerard Reach

Pierre M. Galletti, Clark K. Colton, Michel Jaffrin,

Structure and Function of the Pancreas • Endocrine Pancreas and Insulin Secretion • Diabetes • Insulin • Insulin Therapy • Therapeutic Options in Diabetes • Insulin Administration Systems • Insulin Production Systems • Outlook

135 Nerve Guidance Channels

Robert F. Valentini

Peripheral Nervous System • PNS Response to Injury • PNS Regeneration • Surgical Repair of Transected Peripheral Nerves • Repair with Nerve Guidance Channels • Recent Studies with Nerve Guidance Channels • Enhancing Regeneration by Optimizing Nerve Guidance Channel Properties • Summary

136 Tracheal, Laryngeal, and Esophageal Replacement Devices Yasuhiko Shimizu

Tatsuo Nakamura,

Tracheal Replacement Devices • Laryngeal Replacement Devices • Artificial Esophagi (Esophageal Prostheses)

137 Artificial Blood

Marcos Intaglietta, Robert M. Winslow

Oxygen Carrying Plasma Expanders and the Distribution of Transport Properties in the Circulation • The Distribution of Oxygen in the Circulation • Oxygen Gradients in the Arteriolar Wall • Mathematical Modeling of Blood/Molecular Oxygen Carrying Plasma Expander • Blood Substitutes and Hemodilution • Hematocrit and Blood Viscosity • Regulation of Capillary Perfusion During Extreme Hemodilution in Hamster Skin Fold Microcirculation by High Viscosity Plasma Expander • Crystalloid and Colloidal Solutions as Volume Expanders • Artificial Oxygen Carriers • Fluorocarbons • Hemoglobin for Oxygen Carrying Plasma Expanders • Hemoglobin-Based Artificial Blood • Results in the Microcirculation with Blood Substitution with αα-Hemoglobin • Hemoglobin and Nitric Oxide (NO) Binding • Rational Design of an Oxygen Carrying Molecular Plasma Expander • Conclusions

138 Artificial Skin and Dermal Equivalents

Ioannis V. Yannas

The Vital Functions of Skin • Current Treatment of Massive Skin Loss • Two Conceptual Stages in the Treatment of Massive Skin Loss by Use of the Artificial Skin • Design Principles for a Permanent Skin Replacement • Clinical Studies of a Permanent Skin Replacement (Artificial Skin) • Alternative Approaches: Cultured Epithelia Autographs (CEA) and Skin Equivalents (SE)

XIII.1

Substitutive Medicine

O

ver the past 50 years, humanity has progressively discovered that an engineered device or the transplantation of organs, tissues, or cells can substitute for most, and perhaps all, organs and body functions. The devices are human-made, whereas the living replacement parts can be obtained from the patient, a relative, a human cadaver, or a live animal or can be prospectively developed through genetic engineering. The concept that a disease state may be addressed not only by returning the malfunctioning organ to health using chemical agents or physical means but also by replacing the missing function with a natural or an artificial counterpart has brought about a revolution in therapeutics. Currently in the United States alone, 2 to 3 million patients a year are treated with a human-designed spare part (assist device, prosthesis, or implant), with the result that over 20 million people enjoy a longer or better quality of life thanks to artificial organs. In comparison, a shortage of donor organs limits the number of transplantation procedures to about 20,000 a year, and the total population of transplant survivors is on the order of 200,000. The fundamental tenet of substitutive medicine is that beyond a certain stage of failure, it is more effective to remove and replace a malfunctioning organ than to seek in vain to cure it. This ambitious proposition is not easy to accept. It goes against the grain of holistic views of the integrity of the person. It seems at odds with the main stream of twentieth-century scientific medicine, which strives to elucidate pathophysiologic mechanisms at the cellular and molecular level and then to correct them through a specific biochemical key. The technology of organ replacement rivals that of space travel in complexity and fanfare and strikes the


popular imagination by its daring, its triumphs, and its excesses. Although the artificial organ approach does not reach the fundamental objective of medicine, which is to understand and correct the disease process, it is considerably more effective than drug therapy or corrective surgery in the treatment of many conditions, e.g., cardiac valve disease, heart block, malignant arrhythmia, arterial obstruction, cataract. A priori, functional disabilities due to the destruction or wear of body parts can be addressed in two ways: the implantation of prosthetic devices or the transplantation of natural organs. For a natural organ transplant, we typically borrow a spare part from a living being or from an equally generous donor who before death volunteered to help those suffering from terminal organ failure. Transplanted organs benefit from refinements acquired over thousands of years of evolution. They are overdesigned, which means they will provide sufficient functional support even though the donated part may not be in perfect condition at the time of transfer to another person. They have the same shape and the same attachment needs as the body part they replace which means that surgical techniques are straightforward. The critical problem is the shortage of donors, and therefore only a small minority of patients currently benefit from this approach. Artificial organs have different limitations. Seen on the scale of human evolution, they are still primitive devices, tested for 40 years at most. Yet they have transformed the prognosis of many heretofore fatal diseases, which are now allowed to evolve past what used to be their natural termination point. In order to design artificial organs, inventive engineers, physiologists, and surgeons think in terms of functional results, not anatomical structures. As a result, artificial organs have but a distant similarity to natural ones. They are mostly made of synthetic materials (often called biomaterials) which do not exist in nature. They use different mechanical, electrical, or chemical processes to achieve the same functional objectives as natural organs. They adapt but imperfectly to the changing demands of human activity. They cannot easily accommodate body growth and therefore are more beneficial to adults than to children. Most critically, artificial organs, as is the case for all machines, have a limited service expectancy because of friction, wear, or decay of construction materials in the warm, humid, and corrosive environment of the human body. Such considerations limit their use to patients whose life expectancy matches the expected service life of the replacement part or to clinical situations where repeated implantations are technically feasible. In spite of these obstacles, the astonishing reality is that millions of people are currently alive thanks to cardiac pacemakers, cardiac valves, artificial kidneys, or hydrocephalus drainage systems, all of which address life-threatening conditions. An even larger number of people enjoy the benefits of hip and knee prostheses, vascular grafts, intraocular lenses, and dental implants, which correct dysfunction, pain, inconvenience, or merely appearance. In short, the clinical demonstration of the central dogma of substitutive medicine over the span of two generations can be viewed demographically as the first step in a evolutionary jump which humans cannot yet fully appreciate. Hybrid artificial organs, or bioartificial organs, are more recent systems which include living elements (organelles, cells, or tissues) as part of a device made of synthetic materials. They integrate the technology of natural organ transplantation and the refinements which living structures have gained through millions of years of evolution with the purposeful design approach of engineering science and the promises of newly developed synthetic materials. Table XIII.1 provides a current snapshot in the continuing evolution of substitutive medicine. Depending upon medical needs and anticipated duration of use, artificial organs can be located outside of the body yet attached to it (paracorporeal prostheses or assist devices) or implanted inside the body in a appropriate location (internal artificial organs or implants). The application of artificial organs may be temporary, i.e., a bridge procedure to sustain life or a specific biologic activity while waiting for either recovery of natural function (e.g., the heart-lung machine), or permanent organ replacement (e.g., left ventricular assist devices). It can be intermittent and repeated at intervals over extended periods of time when there is no biologic necessity for continuous replacement of the missing body functions (e.g., artificial kidney). It can pretend to be permanent, at least within the limits of a finite life span. Up to 1950, organ replacement technology was relatively crude and unimaginative. Wooden legs, corrective glasses, and dental prostheses formed the bulk of artificial organs. Blood transfusion was the only accepted form of transplantation of living tissue. Suddenly, within a decade, the artificial kidney,


TABLE XIII.1

Evolution of Organ Replacement Technology: A 1995 perspective

Current Status Broadly accepted clinically

Accepted with reservations

Limited clinical application

Experimental stage

Conceptual stage

Artificial Organs Heart-lung machine Large-joint prostheses Bone fixation systems Cardiac pacemakers Implantable defibrillators Large vascular grafts Prosthetic cardiac valves Intra-aortic balloon pump Intraocular lenses Middle ear ossicle chain Hydrocephalus shunts Dental implants Skin and tissue expanders Maintenance hemodialysis Chronic ambulatory peritoneal dialysis Breast implants Sexual prostheses Small joint prostheses ECMO in children ECMO in adults Ventricular assist devices Cochlear prostheses Artificial tendons Artificial skin Artificial limbs Artificial pancreas Artificial blood Intravenous oxygenation Artificial esophagus Total artificial heart Nerve guidance channels Artificial eye Neurostimulator Blood pressure regulator Implantable lung Artificial trachea Artificial gut Artificial fallopian tube

Transplantation Blood transfusion Corneal transplants Banked bone Bone marrow Kidney—living related donor Kidney—cadaveric donor Heart Liver

Whole pancreas Single and double lung Combined heart-lung Cardiomyoplasty Pancreatic islets Liver lobe or segment Small Intestine

Bioartificial pancreas Bioartificial liver CNS implants of secreting tissue Gene therapy products

Striated muscle implants Smooth muscle implants Cardiac muscle implants Functional brain implants Bioartificial kidney

the heart-lung machine, the cardiac pacemaker, the arterial graft, the prosthetic cardiac valve, and the artificial hip joint provided the first sophisticated examples of engineering in medicine. More recently, the membrane lung, the implantable lens, finger and tendon prostheses, total knee replacements, and soft-tissue implants for maxillo-facial, ear, or mammary reconstruction have reached the stage of broad clinical application. Ventricular assist devices and the total artificial heart have been extensively tested in animals and validated for clinical evaluation. Artificial skin is increasingly used in the treatment of ulcers and burns. Soft- and hard-tissue substitutes function effectively for several years. Sexual and sensory prostheses offer promises for the replacement of complex human functions. Interfacing of devices with the peripheral and central nervous systems appears as promising today as cardiovascular devices were 30 years ago. Perhaps the brightest future belongs to “information prostheses” which bring to the human body, signals which the organism can no longer generate by itself (e.g., pacemaker functions), signals which need to be modulated differently to correct a disease state (e.g., electronic blood pressure regulators) or signals which cannot be perceived by the nervous system through its usual channels of information gathering (e.g., artificial eye or artificial ear).


XIII.2

Biomaterials

The materials of the first generation of artificial organs—those which are widely available at the moment—are for the most part standard commodity plastics and metals developed for industrial purposes. Engineers have long recognized the limitations of construction materials in the design and performance of machines. However, a new awareness arose when they started interacting with surgeons and biologic scientists in the emerging field of medical devices. In many cases the intrinsic and well established physical properties of synthetic materials such as mechanical strength, hardness, flexibility, or permeability to fluids and gases were not as immediately limiting as the detrimental effects deriving from the material’s contact with living tissues. As a result, fewer than 20 chemical compounds among the 1.5 million candidates have been successfully incorporated into clinical devices. Yet some functional implants require material properties which exceed the limits of current polymer, ceramic, or metal alloy technology. This is an indirect tribute to the power of evolution, as well as a challenge to scientists to emulate natural materials with synthetic compounds, blends, or composites. The progressive recognition of the dominant role of phenomena starting at the tissue-material interface has led to two generalizations: 1. All materials in contact with body fluids or living tissue undergo almost instantaneous and then continuing surface deposition of body components which alter their original properties. 2. All body fluids and tissues in contact with foreign material undergo a dynamic sequence of biologic reactions which evolve over weeks or months, and these reactions may remain active for as long as the contact persists and perhaps even beyond. The recognition of biologic interactions between synthetic materials and body tissues has been translated into the twin operational concepts of biomaterials and compatibility. Biomaterials is a term used to qualify materials which can be placed in intimate contact with living structures without harmful effects. Compatibility characterizes a set of material specifications and constraints which address the various aspects of material-tissue interactions. More specifically, hemocompatibility defines the ability of a biomaterial to stay in contact with blood for a clinically relevant period of time without causing alterations of the formed elements and plasma constituents of the blood or substantially altering the composition of the material itself. The term biocompatibility is often used to highlight the absence of untoward interactions with tissues other than blood (e.g., hard or soft tissues). It is worth observing that hemocompatibility and biocompatibility are virtues demonstrated not by the presence of definable favorable properties but rather by the absence of adverse effects on blood or other tissues. Although these terms imply positive characteristics of the material, the presumption of compatibility is actually based on the accumulation of negative evidence over longer and longer periods of time, using an increasingly complex battery of tests, which must eventually be confirmed under the conditions of clinical use. The clinical success of materials incorporated into actual devices is altogether remarkable, considering how limited our understanding is of the physical and biologic mechanisms underlying tissue-material interactions. Indeed the most substantial conclusion one can draw from a review of records of literally millions of implants is how few major accidents have been reported and how remarkably uncommon and benign have been the side effects of implanting substantial amounts of synthetic substances into the human body. Artificial organs are by no means perfect. Their performance must be appreciated within the same limits that the inexorable processes of disease and aging impose on natural organs.

XIII.3

Outlook for Organ Replacement

Now emerging is a second generation of implantable materials through the confluence of biomaterial science and cell biology (Fig. XIII.1). Cell culture technology, taking advantage of biotechnology products and progressing to tridimensional tissue engineering on performed matrices, now provides building


FIGURE XIII.1 Schematic description of the advances in engineering, biologic, and medical technology which led to the first generation of artificial organs (read from the top) and the newer developments in body replacement parts (read from the bottom).

blocks which incorporate the peptide or glycoprotein sequences responsible for cell-to-cell interactions. This combination leads to a new class of biohybrid devices which includes 1. Cellular transplants for continuing secretion of bioactive substances (e.g., transplants of insulinproducing xenograft tissue protected against immune rejection by permselective envelopes) 2. Composites of synthetic materials with living cells (often called organoids) to accelerate implant integration within the body (e.g. endothelial cell-lined polymer conduits designed for vascular grafts) 3. Replacement parts in which natural tissue regeneration is activated by the presence of supportive cells (e.g., Schwann cell-seeded nerve guidance channels) 4. Vehicles for gene therapy in which continued gene expression is enhanced by a synthetic polymer substrate with appropriate mechanical, chemical, or drug release properties (e.g., epicardial transplants of genetically modified skeletal or cardiac muscle grown on a distensible polymer matrix) In many respects, the new wave of organ replacement exemplifies the synergy of the two original currents in substitutive medicine: prostheses and transplants. It expands the feasibility of cell and tissue transplantation beyond the boundaries of autografts and related donor allografts, opening the way to xenogeneic


and engineered replacement parts. It also confronts the “foreign body” limitations of human-made synthetic implants by adding the molecular and cellular elements that favor permanent biointegration.

XIII.4

Design Considerations

Natural organ transplants, if ideally preserved, should be able to fulfill all functions of the original body part except for those mediated by the nervous system, since a transplanted organ is by definition a denervated structure. In actuality, transplants always present some degree of ischemic damage caused by interruption of the blood supply during transfer from donor to recipient. This may be reflected by a temporarily impaired function in the postoperative period or by permanent necrosis of the most delicate components of the transplant, resulting in some degree of functional limitation. In the long run, transplanted organs may also exhibit functional alterations because of cell or tissue damage associated with an underlying systemic disease. They may be damaged by the immunosuppression protocol, which at the current stage is needed for all organ replacements except for autografts, identical-twin homografts, and some types of fetal tissue transplants. The second-order limitations of transplanted organs are usually ignored, and the assumption is made that all original functions are restored in the recipient. Artificial organs, however, can only replace those bodily functions which have been incorporated into their design because these functions were scientifically described and known to be important. Therefore, in the design of an artificial organ, the first task is to establish the specifications for the device, i.e., to describe in quantitative terms the function or functions which must be fulfilled by a human-made construct and the physical constraints that apply because the device must interface with the human body. Each human organ fulfills multiple functions of unequal importance in terms of survival. Consequently, it is critical to distinguish the essential functions which must be incorporated into an effective spare part from those which can be neglected. Defining specifications and constraints is the first step in the conceptualization of an artificial organ. Only when this is done can one think realistically about design alternatives, the limitations of available materials, and the clinical constraints which will apply, of which the key ones are connections to the body and duration of expected service. Once all these considerations have been integrated (modeling is often useful at that stage), the next step is typically the construction of a prototype. Ideally the device should achieve everything it was expected to do, but usually it exhibits some level of performance and durability which falls short of design specifications, either because of some misjudgment in terms of required function or because of some unanticipated problem arising at the interface between the device and the body. The following step of development may be called optimization, if the specifications were well defined from the outset, or reevaluation, if they were not. More commonly it is the reconciliation of competition and at times contradictory design criteria which leads to a second prototype. At this point, new experiments are needed to establish the reliability and effectiveness of the device in animal models of the target disease (if such exist) or at least in animals in which the natural organ can be removed or bypassed. This is the stage of validation of the device, which is first conducted in acute experiments and must later be extended to periods of observation approximating the duration of intended use in humans. These criteria, however, cannot always be met for long-term implants, since the life expectancy of most animals is shorter than that of humans. By this point, the diverse vantage points of the theoretician, the manufacturer, the performance evaluator, and the clinical user have been articulated for some specific devices and generalized in terms of quality control for classes of devices. The final stage of design, for many artificial organs, is individualization, i.e., the ability to fit the needs of diverse individuals. Humans come in a wide range of body sizes. In some cases, the prostheses must fit very strict dimensional criteria, which implies that they must be fabricated over an extended range of sizes (e.g., cardiac valves). In other cases, there is enough reserve function in the device that one pediatric model and one adult size model may suffice (e.g., blood oxygenator for cardiac surgery).


XIII.5

Evaluation Process

The evaluation of an artificial organ typically is done in six phases: 1. 2. 3. 4. 5. 6.

In vitro bench testing Ex vivo appraisal In vivo studies with healthy experimental animals In vivo studies with animal models of disease Controlled clinical trials General clinical use

In Vitro Bench Testing In vitro bench testing of a completed prototype has three major purposes: 1. To observe the mode of operation of the device and assess its performance under tightly controlled circumstances 2. To define performance in quantitative terms over a wide range of environmental or input conditions 3. To assess the device’s reliability and durability in a manner which can be extrapolated to the intended clinical use For all its value, there are limitations to in vitro testing of devices. Devices are made to work while in contact with body fluids or body tissues. This complex environment modifies materials in ways which are not always predictable. To duplicate this effect as closely as possible a laboratory bench system can be made to match the body’s environment in terms of temperature and humidity. Operating pressures and external forces can also be imitated but not perfectly reproduced (e.g., the complex pulsatile nature of cardiovascular events). Other fluid dynamic conditions such as viscosity, wall shear stress, and compliance of device-surrounding structures call for sophisticated laboratory systems and can only be approximated. The chemical environment is the most difficult to reproduce in view of the complexity of body fluids and tissue structures. Some in vitro testing systems make use of body fluids such as plasma or blood. This in turn brings in additional intricacies because these fluids are not stable outside of the body without preservatives and must be kept sterile if the experiment is to last more than a few hours. Accelerated testing is a standard component in the evaluation of machines. It is critical for permanent implants with moving parts which are subject to the repeated action of external forces. Fatigue testing provides important information on progressive wear or catastrophic failure of device components. For example, the human heart beats about 40 million times per year. Manufacturers and regulatory agencies conduct testing of prosthetic cardiac valves over at least 400 million cycles. With a testing apparatus functioning at 1200 cycles per minute, this evaluation can be compressed by a factor of about 15, i.e., to about a year.

Ex Vivo Appraisal Because of the difficulty of keeping blood in its physiologic state in a container, the evaluation of some blood processing or blood contacting devices is performed by connecting them through the skin to an artery or vein or both if the blood must be returned to the cardiovascular system to avoid excessive hemorrhage. Such experiments retain the advantage of keeping the device under direct observation while allowing longer experiments than are feasible in vitro, particularly if the animal does not require general anesthesia. It is also possible in some cases to evaluate several devices in parallel or sequentially under quite realistic conditions and therefore to conduct comparative experiments under reasonably standardized conditions.


In Vivo Evaluation with Healthy Experimental Animals There comes a stage in the development of most devices where they must be assessed in their target location in a living body. The matching of device size and shape with available experimental sites in the appropriate animal species is a necessary condition. Such experiments typically last weeks, months, or years and provide information about body-device and tissue-material interactions either through noninvasive measurement techniques or through device retrieval at the end of the observation period. Rodents, felines, and dogs raised for research purposes are usually too small for the evaluation of humansized devices. Farm animals such as sheep, goats, pigs, and calves are commonly used. Here again the limited life expectancy of experimental animals prevents studies for periods of service as long as can be expected with permanent implants in man.

In Vivo Evaluation with Animal Models of Disease A first approximation of the effectiveness of a device in replacing a physiologic function can be obtained after removing the target organ in a normal animal. However, when the organ failure is only the cardinal sign of a complex systemic disease, the interactions between the device and the persisting manifestations of the disease occur spontaneously in some species and in other cases can be obtained by chemical, physical, or surgical intervention. Where such models of disease exist in animals which can be fitted with a device, useful information is obtained which helps to refine the final prototypes.

Controlled Clinical Trials Although some devices can be evaluated with little risk in normal volunteers who derive no health benefit from the experiment, our culture frowns on this approach and legal considerations discourage it. Once reliability and effectiveness have been established through animal experiments and the device appears to meet a recognized clinical need, a study protocol is typically submitted to an appropriate ethics committee or institutional review board and, upon their approval, a series of clinical trials is undertaken. The first step often concentrates on the demonstration of safety of the device with a careful watch for side effects or complications. If the device passes this first hurdle, a controlled clinical trial will be carried out with patients to evaluate effectiveness as well as safety on a scale which allows statistical comparison with a control form of treatment. This protocol may extend from a few months to several years depending upon the expected benefits of the device and the natural history of the disease.

General Clinical Use Once a device is deemed successful by a panel of experts, it may be approved by regulatory agencies for commercial distribution. Increasingly a third stage of clinical evaluation appears necessary, namely postmarket surveillance, i.e., a system of clinical outcomes analysis under conditions of general availability of the device to a wide range of doctors and patients. Postmarket surveillance is a new concept which is not yet uniformly codified. It may take the form of a data collection and analysis network, a patient registry to allow continuing follow-up and statistical analysis, a device-tracking system aimed at early identification of unforseen types of failure, or ancillary controls such as inspection of facilities and review of patient histories in institutions where devices are used. Protocols of surveillance on a large scale are difficult and costly to implement and their costeffectiveness is therefore open to question. They are also impaired by the shortage of broadly available and minimally invasive diagnostic methods for assessing the integrity or function of a device prior to catastrophic failure. Worthwhile postmarket surveillance requires a constructive collaboration between patients, doctors, device manufacturers, government regulatory agencies, and study groups assessing health care policy issues in the public and private sectors.


Acknowledgments The chapters that follow are derived to a substantial extent from lecture notes used in graduate courses at Brown University, the Massachusetts Institute of Technology, and the Georgia Institute of Technology. Colleagues at other institutions have also contributed chapters in their own areas of specialization. The authors are also indebted to successive generations of students of biomedical engineering who through a fresh, unencumbered look at the challenges of organ replacement have demonstrated their curiosity, their creativity, and their analytical skills.

In Memoriam This introduction was written by Pierre Galletti, Ph.D., who passed away in 1997. It is retained here as a memorial to his significant contributions to the field of Biomedical Engineering.

Defining Terms Artificial organs: Human-made devices designed to replace, duplicate, or augment, functionally or cosmetically, a missing, diseased, or otherwise incompetent part of the body, either temporarily or permanently, and which require a nonbiologic material interface with living tissue. Assist device: An apparatus used to support or partially replace the function of a failing organ. Bioartificial organ: Device combining living elements (organelles, cells, or tissues) with synthetic materials in a therapeutic system. Bicompatibility: The ability of a material to perform with an appropriate host tissue response when incorporated for a specific application in a device, prosthesis, or implant. Biomaterial: Any material or substance (other than a drug) or combination of materials, synthetic or natural in origin, which can be used as a whole or as a part of a system which treats, augments, or replaces any tissue, organ, or function of the body. Compatibility: A material property which encompasses a set of specifications and constraints relative to material-tissue interactions. Device: Defined by Congress as “…an instrument, apparatus, implement, machine, contrivance, implant, in vitro reagent, or other similar or related article, including any component, part, or accessory, which is…intended for use in the diagnosis of disease or other conditions, or in the cure, mitigation, treatment, or prevention of disease, in man or other animals,…and which does not achieve any of its principal intended purposes through chemical action within or on the body of man or other animals and which is not dependent upon being metabolized for the achievement of any of its principal intended purposes.” Hemocompatibility: The ability of a biomaterial to stay in contact with blood for a clinically relevant period of time without causing alterations of the blood constituents. Hybrid artificial organs: Synonym of bioartificial organs, stressing the combination of cell transplantation and artificial organ technology. Implant: Any biomaterial or device which is actually embedded within the tissues of a living organism. Organs: Differentiated structures or parts of a body adapted for the performance of a specific operation or function. Organoid: An organlike aggregate of living cells and synthetic polymer scaffolds or envelopes, designed to provide replacement or support function. Organ transplant: An isolated body part obtained from the patient, a living relative, a compatible cadaveric donor, or an animal and inserted in a recipient to replace a missing function. Prosthesis: An artificial device to replace a missing part of the body. Substitutive medicine: A form of medicine which relies on the replacement of failing organs or body parts by natural or human-made counterparts. Tissue-material interface: The locus of contact and interactions between a biomaterial, implant, or device and the tissue or tissues immediately adjacent.


References Cauwels JM. 1986. The Body Shop: Bionic Revolutions in Medicine, St. Louis, Mosby. Galleti PM. 1991. Organ replacement: A dream come true. In R Johnson-Hegyeli, AM Marmong du Haut Champ (eds), Discovering New Worlds in Medicine, pp 262–277, Milan, Farmitalia Carlo Erba. Galletti PM. 1992. Bioartificial organs. Artif Organs 16(1):55. Galletti PM. 1993. Organ replacement by man-made devices. J Cardiothor Vasc Anesth 7:624. Harker LA, Ratner BD, Didisheim P. 1993. Cardiovascular biomaterials and biocompatibility. A guide to the study of blood-tissue material interactions. Cardiovasc Pathol 2(3)(suppl):IS–2245. Helmus MN. 1992. Designing critical medical devices without failure. Spectrum, Diagnostics, Medical Equipment Supplies Opthalmics, pp 37–1–37–17, Decision Resources, Inc. Richardson PD. 1976. Oxygenator testing and evaluation: Meeting ground of theory, manufacture and clinical concerns. In WM Zapol, J Qvist (eds), Artificial Lungs for Acute Respiratory Failure. Theory and Practice. pp 87–102, New York, Academic Press.

Further Information The articles and books listed above provide different overviews in the filed of organ replacement. Historically most contributions to the filed of artificial organs were described or chronicled in the 40 annual volumes of the Transactions of the American Society of Artificial Internal Organs (1955 to present). The principal journals in the field of artificial organs are Artificial Organs (currently published by Blackwell Scientific Publications, Inc.), the ASAIO Journal (published by J. B. Lippincott Company), and the International Journal of Artificial Organs (published by Wichtig Editore s.r.l. Milano, Italy). Publications of the Japanese Society for Artificial Organs (typically in Japanese with English abstracts, Artificial Organs Today, for example) contain substantial information.


Photograph of Carpentier-Edwards pericardial valve.


Rosenberg, G. “Artificial Heart and Circulatory Assist Devices.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

8594/S13/ch126/frame Page 1 Friday, March 03, 2000 11:18AM

126 Artificial Heart and Circulatory Assist Devices 126.1 126.2

Define the Problem—Clarification of the Task • Conceptual Design—Plan Treatment • Detailed Design—Execute the Plan • Learn and Generalize

Gerson Rosenberg Pennsylvania State University

Engineering Design Engineering Design of Artificial Heart and Circulatory Assist Devices

126.3

Conclusions

In 1812, LeGallois [1813] postulated the use of mechanical circulatory support. In 1934, DeBakey proposed a continuous flow blood transfusion instrument using a simple roller pump [DeBakey, 1934]. In 1961, Dennis et al., performed left heart bypass by inserting cannulae into the left atrium and returning blood through the femoral artery [Dennis, 1979]. In 1961, Kolff and Moulopoulos developed the first intra-aortic balloon pump [Moulopoulos et al., 1962]. In 1963, Liota performed the first clinical implantation of a pulsatile left ventricular assist device [Liotz et al., 1963]. In 1969, Dr. Denton Cooley performed the first total artificial heart implantation in a human [Cooley et al., 1969]. Since 1969, air driven artificial hearts and ventricular assist devices have been utilized in over 200 and 600 patients, respectively. The primary use of these devices has been as a bridge to transplant. Recently, electrically powered ventricular assist devices have been utilized in humans as a bridge to transplant. These electrically powered systems utilize implanted blood pumps and energy converters with percutaneous drive cables and vent tubes. Completely implanted heart assist systems requiring no percutaneous leads have been implanted in calves with survivals over eight months. The electric motor-driven total artificial hearts being developed by several groups in the United States are completely implanted systems employing transcutaneous energy transmission and telemetry [Rosenberg, 1991. Appendix C, The Artificial Heart, Prototypes, Policies, and Patients]. One total artificial heart design has functioned in a calf for over five months [Snyder, 1993]. There has been steady progress in the development of artificial heart and circulatory assist devices. Animal survivals have been increasing, and patient morbidity and mortality with the devices has been decreasing. It does not appear that new technologies or materials will be required for the first successful clinical application of long-term devices. There is no doubt that further advances in energy systems, materials, and electronics will provide for smaller and more reliable devices, but for the present, sound engineering design using available materials and methods appear to be adequate to provide devices satisfactory for initial clinical trials.



126.1

Engineering Design

A definition for design is given by Pahl and Beitz [1977]: “Designing is the intellectual attempt to meet certain demands in the best possible way. It is an engineering activity that impinges on nearly every sphere of human life, relies on the discoveries and laws of science, and creates the conditions for applying these laws to the manufacture of useful products.” Designing is a creative activity that calls for sound grounding in mathematics, physics, chemistry, mechanics, thermodynamics, hydrodynamics, electrical engineering, production engineering, materials technology, and design theory together with practical knowledge and experience in specialist fields. Initiative, resolution, economic insight, tenacity, optimism, sociability, and teamwork are all qualities that will assist the designer. Engineering design has been broken into many steps or phases by various authors. In general, though, each of these definitions includes some common tasks. No matter what device is being designed, be it a complicated device such as a totally implantable artificial heart or a simpler device, such as a new fastener, sound engineering design principles utilizing a methodical approach will help ensure satisfactory outcome of the design process. The engineering design process can be broken down into at least four separate stages. 1. Define the problem—clarification of the task. At first it may appear that defining the problem or clarifying the task to be accomplished is an easy matter. In general, this is not the case. Great care should be taken in defining the problem, being very specific about the requirements and specifications of the system to be designed. Very often, a complex system can be reduced to the solution of one core problem. An excellent way to begin defining the problem or clarifying the task to be accomplished is to begin by writing a careful design specification or design requirement. This is a document that lists all the requirements for the device. Further, this design requirement may elucidate one or two specific problems which, when solved, will yield a satisfactory design. 2. Conceptual design—plan treatment. After the problem has been defined, the designer must plan the treatment of the problem and begin conceptual design. In this phase possible solutions to the problem at hand are examined. Various methods of determining possible solutions include brainstorming, focus groups, Delphi method. This is the phase of the design process where a thorough review of the literature and examination of similar problems is valuable. In this phase of design each of the proposed solutions to the problem should be examined in terms of a hazard analysis or failure modes and effects analysis to determine which solution appears the most feasible. Economic considerations should also be examined in this phase. 3. Detailed design—Execute the plan. In this phase of engineering design, a detailed design is formulated. Perhaps two designs may be evaluated in the initial detailed design phase. As the detailed design or designs near completion, they must be examined with reference to the design specifications. Here, each of the proposed designs may be evaluated with regard to its ability to perform. Such aspects as system performance, reliability, manufacturability, cost, and user acceptance are all issues which must be considered before a final design is chosen. 4. Learn and generalize. Finally, after the design is complete, the designer should be able to learn and generalize from the design. This educational process will include manufacturing of prototypes and testing. General concepts and principles may be gleaned from the design process that can be applied to further designs. The remainder of this chapter will deal with the application of this engineering design method to the design of artificial heart and circulatory assist devices.

126.2

Engineering Design of Artificial Heart and Circulatory Assist Devices

Define the Problem—Clarification of the Task In the broadest sense, this step can best be accomplished by writing the detailed design requirement or specification for the device. In defining the problem, it is often easiest to begin with the most obvious



and imperative requirements and proceed to the subtler and less demanding. A general statement of the problem for a total artificial heart or assist device is “to develop a device (perhaps totally implantable) that when implanted in the human will provide a longer and better quality of life than conventional pharmacologic or transplant therapy.” The devices considered will be assumed to be permanent implantable devices, not necessarily totally implantable; they may utilize percutaneous wires or tubes. In general, it will be assumed that they will be intended for long-term use (one year or longer) but may also be utilized for short-term applications. Fit of the System One must first decide who the device is intended for. Will it be used in men and women, and of what size? No matter how good the device is, it must first “fit” the patient. For our example, let us assume the device will be used in men and women in the size range of 70–100 kg. The device must then fit in these patients and cause minimal or no pathologic conditions. When considering the fit of the implanted device, one must consider the volume and mass of the device, as well as any critical dimension such as the length, width, or height and the location of any tubes, conduits, or connectors. Careful consideration must be given to the physical attributes of the system such as whether the system should be hard, soft, rough, or smooth and the actual shape of the system in terms of sharp corners or edges that may damage tissue or organs. The design specification must give the maximum length, height, and width of the device, these dimensions being dictated by the patient’s anatomy. The designer must not be limited by existing anatomy; the opportunity to surgically alter the anatomy should be considered. Nontraditional locations for devices should be considered. The device should not project heat in such a way that surfaces in contact with tissue or blood are subjected to a temperature rise 5°C above core temperature on a chronic basis. The use of heat spreaders or fins, along with insulation, may be required. Heat transfer analysis may be helpful. The effect of device movement and vibration should be considered in the design specification. The acceptable sound levels at various frequencies must be specified. A device should meet existing standards for electromagnetic interference and susceptibility. The use of any percutaneous tubes will require the choice of an exit site. This site must not be in a location of constant or excessive movement or tissue breakdown will be experienced at the interface. Pump Performance Pump performance must be specified in terms of cardiac output range. In a 70-kg person, a normal resting cardiac output is approximately 70ml/min/kg or 5 1/min. Choosing a maximum cardiac output of approximately 8 1/min will allow the patient the ability to perform light exercise. The cardiac output performance must be obtained at physiologic pressures, that is, the device, be it a heart assist or total artificial heart device, must be able to pump a cardiac output ranging up to 8 1/min with physiologic inlet and outlet pressures (central venous pressure ∼ 5 mmHg mean, left atrial pressure ∼ 7 mmHg mean, pulmonary artery pressure ∼ 15 mmHg mean, aortic pressure ∼ 100 mmHg mean). Control of the device is critical and must be included in the design specification. For an assist pump, it may be as simple as stating that the pump will pump all the blood delivered to it by the native heart while operating under physiologic inlet and outlet pressures. Or, the design specification may include specific requirements for synchronization of the device with the natural heart. For the total artificial heart, the device must always maintain balance between the left and right pumps. It must not let left atrial pressure rise above a value that will cause pulmonary congestion (approximately 20 mmHg). The device must respond to the patient’s cardiac output requirements. The device must either passively or though an active controller vary its cardiac output upon patient demand. Bicompatibility Bicompatibility has already been alluded to in the design requirements by saying that the device must not cause excessive damage to the biologic system. Specifically, the device must be minimally thrombogenic and minimally hemolytic. It should have a minimal effect on the immune system. It should not promote infection, calcification, or tissue necrosis. Meeting these design requirements will require careful design of the pumping chamber and controller and careful selection of materials.



Reliability The design specification must assign a target reliability for the device. For total artificial hearts and circulatory assist devices, the NIH has proposed a reliability of 80% with an 80% confidence for a 2-year device life. This is the value that the NIH feels is a reliability goal to be achieved before devices can begin initial clinical trials, but the final design reliability may be much more stringent, perhaps 90% reliable with 95% confidence for a 5-year life. The design specification must state which components of the system could be changed if necessary. The design specification must deal with any service that the device may require. For instance, the overall design life of the device may be 5 years, but battery replacement at 2-year intervals may be allowed. The reliability issue is very complex and involves moral, ethical, legal, and scientific issues. A clear goal must be stated and is necessary before the detailed design can begin. Quality of Life The design specification must address the quality of life for the patient. The designers must specify what is a satisfactory quality of life. Again, this is not an easy task. Quantitative measures of the quality of life are difficult to achieve. One person’s interpretation of a satisfactory quality of life may not be the same as another’s. It must always be kept in mind that the quality of life for these patients without treatment would generally be considered unsatisfactory by the general public. The prognosis for patients before receiving artificial hearts and circulatory assist devices is very poor. The quality of life must thus be considered in relation to the patient’s quality of life without the device without ignoring the quality of life of individuals unaffected by cardiac disease [Rosenberg, 1991]. The design specification must state the weight of any external power supplies. The designer must consider how much weight an older patient will be able to carry. How long should this patient be able to be free-roaming and untethered? How often will the energy source require a “recharge?” What sound level will be acceptable? These are all issues that must be addressed by the designer. All the foregoing must be considered and included in the definition of the problem and clarification of the task. Each of these issues should be clearly described in the design specification or requirement. In many instances there are no right and wrong answers to these questions.

Conceptual Design—Plan Treatment In the conceptual design phase, the designer must plan the treatment of the problem and consider various designs that meet the design specification. In the design specification, it must be stated whether the blood pump is to be pulsatile or nonpulsatile. If there is no requirement in the design specification, then in the conceptual design phase, the designer may consider various nonpulsatile and pulsatile flow devices. Nonpulsatile devices include centrifugal pumps, axial flow pumps, shear flow pumps, and peristaltic pumps. Pulsatile pumps have traditionally been sac- or diaphragm-type devices. At the present time there is no definitive work describing the absolute requirement for pulsatility in the cardiovascular system; thus, both types of devices can be considered for assist and total artificial hearts. In the conceptual design phase, the designer should consider other nontraditional solutions to the problem such as devices that employ micro-machines or magneto-hydrodynamics. Careful consideration must be given to source of energy. Sources that have been considered include electrical energy stored in batteries or derived from piezoelectric crystals, fuel cells, and thermal energy created either thermonuclearly or through thermal storage. The performance of each of these energy sources must be considered in the conceptual design phase. Public considerations and cost for a thermonuclear source, at the present time, have essentially eliminated this as an implantable energy source. Thermal storage has been shown to be a feasible energy source, and adequate insulation has been developed and demonstrated to be reliable for short periods of time [Uringer, 1989]. Steady flow devices that employ seals within the blood stream deserve very careful consideration of the seal area. These seals have been prone to failure, causing embolization. Active magnetic levitation has been proposed for component suspension and may be considered as a possible design. A device that utilities a rotating member such as a rotating electric motor that would drive an impeller or mechanical mechanism



will create forces on the tissue when the patient moves due to gyroscopic or coriolis accelerations. These forces must be considered. Pump performance, in terms of purely hydraulic considerations, can be achieved with any of the pumping systems described. Control of these devices may be more difficult. Designs that include implanted sensors such as pressure or flow sensors must deal with drift of these signals. Very often, signalto-noise ratios are poor in devices that must operate continuously for several years. Designs that employ few or no sensors are preferable. Systems such as sac-type blood pumps have been described as having intrinsic automatic control. That is, these devices can run in a fill limited mode, and, as more blood is returned to the pump, it will increase its stroke volume. Intrinsic control is desirable, but, unfortunately, this generally provides for only a limited control range. Nearly all devices currently being developed employ some form of automatic electronic control system. These automatic control systems are utilized on both total artificial hearts and assist devices. In some designs system parameters such as blood pump fill time, electric current, voltage, or speed are utilized to infer the state of the circulatory system. These types of systems appear to demonstrate good long-term stability and eliminate the potential for device malfunction associated with transducer drift. Consideration of pump performance and the interaction with the biologic system is important in the conceptual design phase. Two pump designs may be capable of pumping the same in a hydrodynamic sense in terms of pressures and flows, but one pump may have much higher shear stresses than the other and thus be more hemolytic. One device may have much more mechanical vibration or movement compromising surrounding tissue. The subject of biocompatibility for circulatory assist and artificial hearts is a very complex one. No matter what type device is designed, its interaction with the environment is paramount. Blood-contacting materials must not cause thrombosis and should have a minimal effect on formed elements in the blood. In terms of tissue biocompatibility, the device should not have sharp corners or areas where pressure necrosis can occur. Both smooth and rough exterior surfaces of devices have been investigated. It appears that devices in contact with the pleura tend to form a thinner encapsulation when they are rough surfaced. Compliance chambers form much thinner capsules when they have a rough surface. In other areas, it is not entirely clear if a rough surface is advantageous. Tissue ingrowth into a rough surface makes removal of the device difficult. The selection of materials in the design of these devices is limited. The designer’s job would be made much easier if there were several completely biocompatible materials available. If the designer only had a perfectly nonthrombogenic material that had outstanding fatigue properties, the design of these devices would be greatly simplified! Unfortunately, at the present time the designer is limited to existing materials. Traditionally, metals that have been employed in blood pumps include various stainless steels, cobalt, cobalt chromium alloys, and titanium. Each of these materials has adequate performance when in contact with tissue and blood under certain circumstances. Ceramic materials such as pyrolytic carbon, alumina, and zirconia have been used in contact with both tissue and blood with varying degrees of success. The range of polymeric materials that have been utilized for these devices is much greater. These materials include various polyurenthanes, silicone rubber, Kel-F, Teflon, Delrin, butyl rubber, Hexsyn rubber, polysulfone, polycarbonate, and others [Williams, 1981]. The designer must carefully consider all the properties of these materials when employing them. The designer must look at the strength, durability, hardness, wear resistance, modulus of elasticity, surface energy, surface finish, and biocompatibility before choosing any of these materials. The interaction between the biologic system and these materials is complex and may be strongly influenced by fluid mechanics if in contact with blood. The designer should carefully consider surface modification of any of these materials. Surface modification can be performed, including ion implantation or grafting of various substances to the surface to promote improved compatibility. Manufacturing and fabrication processes have a profound effect on the properties of these materials and must be carefully analyzed and controlled.



The designer must give a great deal of consideration to the fluid mechanics involved. It is well known that excessive shear stresses can promote hemolysis and activation of the clotting system, as well as damage to the other formed elements in the blood. Not only the actual design of the blood pump, but the operation of the blood pump can affect phenomena such as cavitation, which can be hemolytic and destructive to system components. Thrombosis is greatly affected by blood flow. Regions of stagnation, recirculation, and low wall shear stress should be avoided. The magnitude of “low” shear stress is reported in the literature with wide range [Folie & McIntire, 1989; Hashimoto et al., 1985; Hubbel & McIntire, 1986]. Many of the analytical tools available today in terms of computational fluid dynamics and finite element analysis are just beginning to be useful in the design of these devices. Most of these systems have complex flow (unsteady, turbulent, non-Newtonian with moving boundaries) and geometries and are not easily modeled. Once a reliability goal has been established, the designer must ensure that this goal is met. The natural heart beats approximately 40 million times a year. This means that an artificial heart or assist device with a 5-year life may undergo as may as 200,000,000 cycles. The environment in which this device is to operate is extremely hostile. Blood and extracellular fluids are quite corrosive and can promote galvanic or crevice corrosion in most metallic materials. Devices that employ polymeric materials must deal with diffusion of mass across these materials. Water, oxygen, nitrogen, carbon dioxide, and so on may all diffuse across these polymeric materials. If temperature fluctuations occur or differences exist, liquid water may form due to condensation. With carbon dioxide present, a weak acid can be formed. Many polymeric materials are degraded in the biologic environment. A careful review of the literature is imperative. The use of any “new” materials must be based upon careful testing. The designer must be aware of the difficulty in providing a sealed system. The designer may need to utilize hermetic connectors and cables which can tolerate the moist environment. All these affect the reliability of the device. External components of the system which may include battery packs or monitoring and perhaps control functions have reliability requirements that differ from the implanted components. System components which are redundant may not require the same level of reliability as do nonredundant implanted components. Externally placed components have the advantage of being amenable to preventative maintenance or replacement. Systems that utilize transcutaneous energy transmission through inductive coupling may have advantages over systems that utilize a percutaneous wire. Although the percutaneous wire can function for long periods of time, perhaps up to 1 year, localized infection is almost always present. This infection can affect the patient’s quality of life and be a source of systemic infection. In the conceptual design phase, one must carefully weigh quality of life issues when evaluating solutions to the problem. Careful consideration needs to be given to the traditional quality of life issues such as the patient’s emotional, social, and physical well-being, as well as to the details of day-to-day use of the device. What are the frequency and sound level requirements for patient prompts or alarms, and are certain kinds of alarms useful at all? For visible information for the patient, from what angles can this information be viewed, how bright does the display need to be, should it be in different languages, or should universal symbols be used? A great deal of thought must be given to all aspects of device use, from charging of the batteries to dealing with unexpected events. All these issues must be resolved in the conceptual design phase so that the detailed phase will be successful.

Detailed Design—Execute Plan This is the phase of engineering design where the designer and other members of the team must begin to do what is generally considered the designer’s more traditional role, i.e., calculations and drawings. This phase of design may require some initial prototyping and testing before the detailed design can be complete. Akutsu and Koyangi [1993] provide several results of various groups’ detailed designs for artificial hearts and circulatory assist devices.



Learn and Generalize Substantial research and development of artificial hearts and circulatory assist devices has been ongoing for almost 30 years. During this period there have been literally thousands of publications related to this research, not only descriptions of device designs, but experimental results, detailed descriptions fluid dynamic phenomena, hemolysis, thrombosis, investigation of materials selection and processing, consideration of control issues, and so on. The artificial organs and biomaterials literature has numerous references to materials utilized in these devices. A thorough review of the literature is required to glean the general principles that can be applied to the design of these devices. Many of these principles or generalities apply only to specific designs. In some design circulatory assist devices, a smooth surface will function satisfactorily, whereas a rough surface may cause thrombosis or an uncontrolled growth of neointima. Yet, in other design devices, a textured or rough surface will have performance superior to that of an extremely smooth surface. Wide ranges of sheer stresses are quoted in the literature that can be hemolytic. Although considerable research has been performed examining the fluid mechanics of the artificial hearts and circulatory assist devices, there is really no current measure of what is considered a “good” blood flow within these devices. We know that the design must avoid regions of stasis, but how close to stasis one can come, or for how long, without thrombosis is unknown. It is up to the designer to review current literature and determine what fundamental principles are applicable to his or her design. Then, when the design is complete, the designer can learn and generalize specific principles related to her or his device. Hopefully, general principles that can apply to other devices will be elucidated.

126.3

Conclusions

The design of artificial hearts and circulatory assist devices is a very complex process involving many engineering disciplines along with medicine and other life science areas. Social issues must enter into the design process. The design of such devices requires sound engineering design principles and an interdisciplinary design team dedicated to the development and ultimate clinical application of these devices.

References Akutsu T, Koyanagi H. 1993. Heart Replacement. Artificial Heart 4, Tokyo, Springer-Verlag. Cooley DA, Liota D, Hallman GL, et al. 1969. Orthotopic cardiac prosthesis for 2-stage cardiac replacement. AM J Card 24:723. DeBakey ME. 1934. A simple continuous flow blood transfusion instrument. New Orleans Med Surg J 87:386. Folie BJ, McIntire LV. 1989. Mathematical analysis of mural thrombogenesis, concentration profiles of platelet-activating agents and effects of viscous shear flow. Biophys J 56:1121. Hashimoto S, Maeda H, Sasada T. 1985. Effect of shear rate on clot growth at foreign surfaces. Artif Organs 9:345. Hubbell JA, McIntire LV. 1986. Visualization and analysis of mural thrombogenesis on collagen, polyurethane and nylon. Biomaterials 7:354. LeGallois CJJ. 1813. Experience on the Principles of Life, Philadelphia, Thomas. (Translation of CJJ Le Gallois 1812. Experiences sur les Principles de Vie. Paris, France.) Liota D, Hall CW, Walter SH, et al. 1963. Prolonged assisted circulation during and after cardiac and aortic surgery. Prolonged partial left ventricular bypass by means of an intra-corporeal circulation. AM J Card 12:399. Moulopoulos D, Topaz SR, Kolff WJ. 1962. Extracorporeal assistance to the circulation at intra-aortic balloon pumping. Trans AM Soc Artif Intern Organs 8:86. Pahl G, Beitz W. 1977. Engineering Design, Berlin, Heidelberg, Springer-Verlag (First English edition published 1984 by The Design Council, London.)



Rosenberg G. 1991. Technological opportunities and barriers in the development of mechanical circulatory support systems (Appendix C). In Institute of Medicine, The Artificial Heart, Prototypes, Policies, and Patients, National Academy Press. Snyder AJ, Rosenberg G, Weiss WJ, et al. 1993. In vivo testing of a completely implanted total artificial heart system. ASAIO J 39(3):M415. Unger Felix 1989. Assisted Circulation 3, Berlin, Heidelberg, Springer-Verlag. Williams DF. 1981. Biocompatibility of Clinical Implant Materials, Boca Raton, Fla, CRC Press.


Yoganathan, A. P. “Cardiac Valve Prostheses.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

127 Cardiac Valve Prostheses 127.1

A Brief History of Heart Valve Prostheses

127.2 127.3 127.4

Current Types of Prostheses Tissue Versus Mechanical . Engineering Concerns and Hemodynamic Assessment of Prosthetic Heart Valves

Mechanical Valves • Tissue Valves

Pressure Grandient • Effective Orifice Area EOA) • Regurgitation • Flow Patterns and Turbulent Shear Stresses

Ajit P. Yoganathan Georgia Institute of Technology

127.5 127.6

Implications for Thrombus Deposition Durability

127.7 127.8

Current Trends in Valve Design Conclusion

Wear • Fatigue • Mineralization

The first clinical use of a cardiac valvular prosthesis took place in 1952, when Dr. Charles Hufnagel implanted the first artificial caged ball valve for aortic insufficiency. The Plexiglas cage contained a ball occluder and was inserted into the descending aorta without the need for cardiopulmonary bypass. It did not cure the underlying disease, but it did relieve regurgitation from the lower two-thirds of the body. The first implant of a replacement valve in the anatomic position took place in 1960, with the advent of cardiopulmonary bypass. Since then, the achievements in valve design and the success of artificial heart valves as replacements have been remarkable [Roberts, 1976]. More than 50 different cardiac valves have been introduced over the past 35 years. Unfortunately, after many years of experience and success, problems associated with heart valve prostheses have not been eliminated. The most serious problems and complications [Bodnar, Frater, 1991; Butchart, Bodnar, 1992; Giddens et al. 1993; Roberts, 1976] are: • • • • • •

Thrombosis and thromboembolism Anticoagulant-related hemorrhage Tissue overgrowth Infection Paravalvular leaks due to healing defects, and Valve failure due to material fatigue or chemical change.

New valve designs continue to be developed. Yet to understand the future of valve replacements, it is important to understand their history.


127.1 A Brief History of Heart Valve Prostheses This section on replacement valves highlights a relatively small number of the many various forms which have been made. However, those that have been included are either the most commonly used today or those which have made notable contributions to the advancement of replacement heart valves [Brewer, 1969; Yoganathan et al., 1992].

Mechanical Valves The use of the caged-ball valve in the descending aorta became obsolete with the development in 1960 of what today is referred to as the Starr-Edwards ball-and-cage valve. Similar in concept to the original Hufnagel valve, it was designed to be inserted in place of the excised diseased natural valve. This form of intracardiac valve replacement was used in the mitral position and for aortic and multiple replacements. Since 1962 the Starr-Edwards valve has undergone many modifications to improve its performance in terms of reduced hemolysis and thromboembolic complications. However, the changes have involved materials and techniques of construction and have not altered the overall concept of the valve design in any way (Fig. 127.1a). Other manufacturers have produced variations of the ball and cage valve, notably the Smeloff-Cutter valve and the Magovern Prosthesis. In the case of the former, the ball is slightly smaller than the orifice. A subcage on the proximal side of the valve retains the ball in the closed position with its equator in the plane of the sewing ring. A small clearance around the ball ensures easy passage of the ball into the orifice. This clearance also gave rise to a mild regurgitation which was felt, but not proven, to be beneficial in preventing thrombus formation. The Magovern valve is a standard ball-and-cage format which incorporates two rows of interlocking mechanical teeth around the orifice ring. These teeth are used for inserting the valve and are activated by removing a special valve holder once the valve has been correctly located in the prepared tissue annulus. The potential hazard of dislocation from a calcified annulus due to imperfect placement was soon observed. This valve is no longer in use. Due to the high-profile design characteristics of the ball valves, especially in the mitral position, lowprofile caged disc valves were developed in the mid-1960s. Examples of the caged disc designs are the Kay-Shiley and Beall prostheses, which were introduced in 1965 and 1967, respectively (Fig. 127.1b). These valves were used exclusively in the atrioventricular position. However, due to their inferior hemodynamic characteristics, caged disc valves are rarely used today.

FIGURE 127.1 (a) Photograph of Starr-Edwards ball and cage valve; (b) photograph of Kay-Shiley disc valve; (c) photograph of Bjork-Shiley tilting disc valve; (d) photograph of Medtronic-Hall tilting disc valve; (e) photograph of St. Jude bileaflet valve; (f ) photograph of CarboMedics bileaflet valve; (g) photograph of Parallel bileaflet valve.




Even after 35 years of valve development, the ball-and-cage format remains the valve of choice for some surgeons. However, it is no longer the most popular mechanical valve, having been superseded, to a large extent, by tilting-disc and bileaflet valve designs. These valve designs overcome two major drawbacks of the ball valve, namely, high profile heights and excessive occluder-induced turbulence in the flow through and distal to the valve. The most significant developments in mechanical valve design occurred in 1969 and 1970 with the introduction of the Bjork-Shiley and Lillehei-Kaster tilting-disc valves (Fig. 127.1c). Both prostheses involve the concept of a free-floating disc which in the open position tilts to an angle depending on the design of the disc-retaining struts. In the original Bjork-Shiley valve, the angle of the tilt was 60˚ for the aortic and 50˚ for the mitral model. The Lillehei-Kaster valve has a greater angle of tilt of 80˚ but in the closed position is preinclined to the valve orifice plane by an angle of 18˚. In both cases the closed valve configuration permits the occluder to fit into the circumference of the inflow ring with virtually no overlap, thus reducing mechanical damage to erythrocytes. A small amount of regurgitation backflow induces a “washing out” effect of “debris” and platelets and theoretically reduces the incidence of thromboemboli. The obvious advantage of the tilting-disc valve is that in the open position it acts like an aerofoil in the blood flowing through the valve, and induced flow disturbance is substantially less than that obtained with a ball occluder. Although the original Bjork-Shiley valve employed a Delrin occluder, all presentday tilting-disc valves use pyrolitic carbon for these components. It should also be noted that the freefloating disc can rotate during normal function, thus preventing excessive contact wear from the retaining components on one particular part of the disc. Various improvements to this form of mechanical valve design have been developed but have tended to concentrate on alterations either to the disc geometry as in the Bjork-Shiley convexo-concave design or to the disc-retaining system as with the Medtronic-Hall and Omniscience valve designs (Fig. 127.1d). The Medtronic-Hall prosthesis was introduced in 1977. It is characterized by a central, disc-control strut, with a mitral opening angle of 70˚ and an aortic opening of 75˚. An aperture in the flat, pyrolitic carbon-coated disc affixes it to the central guide strut. This strut not only retains the disc but controls its opening angle and allows it to move downstream 1.5–2.0 mm; this movement is termed disc translation and improves flow velocity between the orifice ring and the rim of the disc. The ring and strut combination is machined from a single piece of titanium for durability. All projections into the orifice (pivot points, guide struts, and disc stop) are open-ended, streamlined, and in the region of highest velocity to prevent the retention of thrombi by valve components. The sewing ring is of knitted Teflon. The housing is rotatable within the sewing ring for optimal orientation of the valve within the tissue annulus. Perhaps the most interesting development has been that of the bileaflet all-pyrolitic carbon valve designed by St. Jude Medical, Inc. and introduced in 1978 (Fig. 127.1e). This design incorporates two semicircular hinged pyrolitic carbon occluders (leaflets) which in the open position are intended to provide minimal disturbance to flow. The leaflets pivot within grooves made in the valve orifice housing. In the fully open position the flat leaflets are designed to open to an angle of 85˚. The Duromedics valve is similar in concept to the St. Jude except that it incorporates curved leaflets. The CarboMedics bileaflet prosthesis gained FDA approval for U.S. distribution in 1993 (Fig. 127.1f ). The CarboMedics valve is also made of Pyrolite, which is intended for durability and thromboresistance. The valve has a recessed pivot design and is rotatable within the sewing ring. The two leaflets are semicircular, radiopaque, and open to an angle of 78˚. A titanium stiffening ring is used to lessen the risk of leaflet dislodgment or impingement. The most recent bileaflet design is the Parallel valve from Medtronic, Inc. (Fig. 127.1g). The significant design aspect of the Parallel valve is the ability of its leaflets to open to a position parallel to flow. This is intended to reduce the amount of turbulence that is created in the blood and therefore should improve hemodynamics and reduce thromboembolic complications. European clinical implants began in the Spring of 1994. The most popular valve design in use today is the bileaflet. Approximately 75% of the valves implanted today are bileaflet prostheses.


Tissue Valves Two major disadvantages with the use of mechanical valves is the need for life-long anticoagulation therapy and the accompanying problems of bleeding [Butchart & Bodnar, 1992]. Furthermore, the hemodynamic function of even the best designed valves differs significantly from that of the natural healthy heart valve. An obvious step in the development of heart valve substitutes was the use of naturally occurring heart valves. This was the basis of the approach to the use of antibiotic or cryotreated human aortic valves (homografts: from another member of the same species) removed from cadavers for implantation in place of a patient’s own diseased valve. The first of these homograft procedures was performed by Ross in 1962, and the overall results so far have been satisfactory. This is, perhaps, not surprising since the homograft replacement valve is optimum both from the point of view of structure and function. In the open position these valves provide unobstructed central orifice flow and have the ability to respond to deformations induced by the surrounding anatomical structure. As a result, such substitutes are less damaging to the blood when compared with the rigid mechanical valve. The main problem with these cadaveric allografts, as far as may be ascertained, is that they are no longer living tissue and therefore lack that unique quality of cellular regeneration typical of normal living systems. This makes them more vulnerable to long-term damage. Furthermore, they are only available in very limited quantities. An alternative approach is to transplant the patient’s own pulmonary valve into the aortic position. This operation was also the first carried out by Ross in 1967, and his study of 176 patients followed up over 13 years showed that such transplants continued to be viable in their new position with no apparent degeneration [Wain et al., 1980]. This transplantation technique is, however, limited in that is can only be applied to one site. The next stage in development of tissue valve substitutes was the use of autologous fascia lata (a wide layer of membrane that encases the thigh muscles) either as free or frame-mounted leaflets. The former approach for aortic valve replacement was reported by Senning in 1966, and details of a frame-mounted technique were published by Ionescu and Ross in 1966 [Ionescu, 1969] The approach combined the more natural leaflet format with a readily available living autologous tissue. Although early results seemed encouraging, Senning expressed his own doubt on the value of this approach in 1971, and by 1978 fascia lata was no longer used in either of the above, or any other, form of valve replacement. The failure of this technique was due to the inadequate strength of this tissue when subjected to long-term cyclic stressing in the heart. In parallel with the work on fascia lata valves, alternative forms of tissue leaflet valves were being developed. In these designs, however, more emphasis was placed on optimum performance characteristics than on the use of living tissue. In all cases the configuration involved a three-leaflet format which was maintained by the use of a suitably designed mounting frame. It was realized that naturally occurring animal tissues, if used in an untreated form, would be rejected by the host. Consequently, a method of chemical treatment had to be found which prevented this antigenic response but did not degrade the mechanical strength of the tissue. Formaldehyde has been used by histologists for many years to arrest autolysis and “fix” tissue in the state in which it is removed. It had been used to preserve biologic tissues in cardiac surgery but, unfortunately, was found to produce shrinkage and increase the stiffness of the resulting material. For these reasons, formaldehyde was not considered ideal as a method of tissue treatment. Glutaraldehyde is another histologic fixative which has been used especially for preserving fine detail for electron microscopy. It is also used as a tanning agent by the leather industry. In addition to arresting autolysis, glutaraldehyde produces a more flexible material due to increased collagen crosslinking. Glutaraldehyde has the additional ability of reducing the antigenicity of xenograft tissue to a level at which it can be implanted into the heart without significant immunologic reaction. In 1969, Kaiser and coworkers described a valve substitute using an explanted glutaral dehyde-treated porcine aortic valve which was mounted on to a rigid support frame. Following modification in which the rigid frame was replaced by a frame having a rigid base ring with flexible posts, this valve became


FIGURE 127.2

(a) Photograph of Hancock porcine valve; (b) photograph of Carpentier-Edwards pericardial valve.

commercially available as the Hancock Porcine Xenograft in 1970 (Fig. 127.2a). It remains one of the two most popular valve substitutes of this type, the other being the Carpentier-Edwards Bioprosthesis introduced commercially by Edwards Laboratories in 1976.This latter valve uses a totally flexible support frame. In 1977 production began of the Hancock Modified Orifice (M.O.) valve, a refinement of the Hancock Standard valve. The Hancock M.O. is of a composite nature—the right coronary leaflet containing the muscle shelf is replaced by a noncoronary leaflet of the correct size from another porcine valve. This highpressure fixed valve is mounted into a Dacron-covered polypropylene stent. The Hancock II and CarpentierEdwards supra-annular porcine bioprostheses are second-generation bioprosthetic valve designs which were introduced in the early 1980s. The porcine tissue is initially fixed at 1.5 mmHg and then at high pressure. This fixation method is designed to ensure good tissue geometry. Both valves are treated with antimineralization treatments. Neither valve has been FDA approved for clinical use in the United States. In porcine prostheses, the use of the intact biologically formed valve makes it unnecessary to manufacture individual valve cusps. Although this has the obvious advantage of reduced complexity of construction, it does require a facility for harvesting an adequate quantity of valves so that an appropriate range of valve sizes of suitable quality can be made available. This latter problem did not occur in the production of the three-leaflet calf pericardium valve developed by Ionescu and colleagues; the construction of this valve involved the molding of fresh tissue to a tricuspid configuration around a support frame. As the tissue is held in this position, it is treated with a glutaraldehyde solution. The valve, marketed in 1976 as the Ionescu-Shiley Pericardial Xenograft, was discontinued in the mid-1980s due to structural failure problems. Early clinical results obtained with tissue valves indicated their superiority to mechanical valves with respect to a lower incidence of thromboembolic complications [Bodnar & Yacoub 1991]. For this reason the use of tissue valves increased significantly during the late 1970s. The Carpentier-Edwards pericardial valve consists of three pieces of pericardium mounted completely within the Elgiloy wire stent to reduce potential abrasion between the Dacron-covered frame and the leaflets. The pericardium is retained inside the stent by a Mylar button rather than by holding sutures. Its clinical implantation began in July 1980, and it is currently approved for clinical use in the United States (Fig. 127.2b). Clinical experiences with different tissue valve designs have increasingly indicated time-dependent (5- to 7-year) structural changes such as calcification and leaflet wear, leading to valve failure and subsequent replacement [Ferrans et al., 1980; Oyer et al., 1979; Bodnar, Yacoub, 1986]. The problem of valve leaflet calcification is more prevalent in children and young adults. Therefore, tissue valves are rarely used in children and young adults at the present time. Such problems have not been eliminated by the glutaraldehyde tanning methods so far employed, and it is not easy to see how these drawbacks are to


be overcome unless either living autologous tissue is used or the original structure of the collagen and elastin are chemically enhanced. On the latter point there is, as yet, much room for further work. For instance, the fixing of calf pericardium under tension during the molding of the valve cusps will inevitably produce “locked-in” stresses during fixation, thus changing the mechanical properties of the tissue.

127.2 Current Types of Prostheses At present, over 175,000 prosthetic valves are implanted each year throughout the world. Currently, the four most commonly used basic types of prostheses are: 1. 2. 3. 4.

Caged ball Tilting disc Bileaflet Bioprostheses (tissue valves)

Valve manufacturers continue to develop new designs of mechanical and tissue valves. The ideal heart valve prosthesis does not yet exist and may never be realized. However, the characteristics of the “perfect” prostheses should be noted. The ideal heart valve should: • Be fully sterile at the time of implantation and be nontoxic • Be surgically convenient to insert at or near the normal location of the heart • Conform to the heart structure rather than the heart structure conforming to the valve (i.e., the size and shape of the prosthesis should not interfere with cardiac function) • Show a minimum resistance to flow so as to prevent a significant pressure drop across the valve • Have a minimal reverse flow necessary for valve closure, so as to keep the incompetence of the valve at a low level • Show long resistance to mechanical and structural wear • Be long-lasting (25 years) and maintain its normal functional performance (i.e., must not deteriorate over time) • Cause minimal trauma to blood elements and the endothelial tissue of the cardiovascular structure surrounding the valve • Show a low probability for thromboembolic complications without the use of anticoagulants • Be sufficiently quiet so as not to disturb the patient • Be radiographically visible • Have an acceptable cost

127.3 Tissue Versus Mechanical Tissue prostheses gained widespread use during the mid-1970s. The major advantage of tissue valves compared to mechanical valves is that tissue valves have a lower incidence of thromboembolic complications [Butchart & Bodnar, 1992]. Therefore, most patients receiving tissue valves do not have to take anticoagulants long-term. The major disadvantages to tissue valves are large pressure drops compared to some mechanical valves (particularly in the smaller valve sizes), jetlike flow through the valve leaflets, material fatigue and/or wear of valve leaflets, and calcification of value leaflets, especially in children and young adults. Valve deterioration, however, usually takes place slowly with tissue valves, and patients can be monitored by echocardiography and other noninvasive techniques. The clear advantage of mechanical valves is their long-term durability. Current mechanical valves are manufactured from a variety of materials, such as pyrolitic carbon and titanium. Structural failure of mechanical valves is rare, but, when it occurs, is usually catastrophic [Giddens et al., 1993]. One major disadvantage of the use of mechanical valves is the need for continuous, life-long anticoagulation therapy to minimize the risk of thrombosis and thromboembolic complications. Unfortunately, the anticoagulation © 2000 by CRC Press LLC

therapy may lead to bleeding problems; therefore, careful control of anticoagulation medication is essential for the patient’s well-being and quality of life. Another concern is the hemodynamic performance of the prosthesis. The hemodynamic function of even the best designs of mechanical valves differs significantly from that of normal heart valves.

127.4 Engineering Concerns and Hemodynamic Assessment of Prosthetic Heart Valves In terms of considerations related to heart valve design, the basic engineering concerns are • Hydrodynamics/hemodynamics • Durability (structural mechanics and materials) • Biologic response to the prosthetic implant The ideal heart valve design from the hemodynamic point of view should [Giddens et al., 1993] • • • • •

Produce minimal pressure gradient Yield relatively small regurgitation Minimize production of turbulence Not induce regions of high shear stress Contain no stagnation or separation regions in its flow field, especially adjacent to the valve superstructure

No valve as yet, other than normal native valves, satisfies all these criteria.

Pressure Gradient The heart works to maintain adequate blood flow through a prosthetic valve; a well-designed valve will not significantly impede that blood flow and will therefore have as small a pressure gradient as possible across the valve. Because of the larger separation region inherent in flow over bluff bodies, configurations such as the caged disc and caged ball have notably large pressure gradients. Porcine bioprostheses have relatively acceptable pressure gradients for larger diameter valves because they more closely mimic natural valve geometry and motion, but the small sizes ( 0.8, which can be achieved using a membrane with an effective pore size greater than about 160 Å (albumin has a molecular weight of 69,000 and a StokesEinstein radius of 36 Å). This membrane would be able to retain about 80% of the immunglobulin M (which has a molecular weight of about 900,000 and a Stokes-Einstein radius of 98 Å), but it would retain less than 40% of the immunglobulin G (with MW = 155,000 and a radius of 55 Å). The protein retention obtained during an actual plasma filtration is substantially more complex than indicated by the above discussion. The polymeric membranes used in these devices actually have a broad distribution of irregularly shaped (noncylindrical) pores. Likewise, the proteins can have very different (nonspherical) conformations, and their transport characteristics also can be affected by electrostatic, hydrophobic, and van der Waals interactions between the proteins and the polymeric membrane, in addition to the steric interactions that are accounted for in the development leading to Eq. (132.14). Protein-protein interactions can also significantly alter the observed protein retention. Finally, the partially retained proteins will tend to accumulate at the upstream surface of the membrane during filtration (analogous to the concentration polarization effects described previously in the context of blood cell filtration). This type of secondary plasma filtration, which is generally referred to in the literature as cascade filtration, is primarily effective at removing large immune complexes (molecular weight of approximately 700,000) and immunglobulin M (MW of 900,000) from smaller proteins such as albumin. Several studies have, however, found a higher degree of albumin-immunoglobulin G separation than would be expected based on purely steric considerations [Eq. (132.14)]. This enhanced selectivity is probably due to some type of long-range (e.g., electrostatic) interaction between the proteins and the membrane. A number of different techniques have been developed to enhance the selectivity of these plasma filtration devices. For example, Malchesky and coworkers at the Cleveland Clinic [Malchesky et al., 1980] developed the process of cryofiltration in which the temperature of the plasma is lowered to about 10°C prior to filtration. A number of diseases are known to be associated with the presence of large amounts of cryo- (cold-) precipitable substances in the plasma, including a number of autoimmune diseases such as systemic lupus erythematosus and rheumatoid arthritis. Lowering the plasma temperature causes the aggregation and/or gelation of these cryoproteins, making it much easier for these components to be


8594/S13/ch132/frame Page 12 Friday, March 03, 2000 03:08 PM

removed by the membrane filtration. About 10 g of cryogel can be removed in a single cryofiltration, along with significant amounts of the larger-molecular-weight immune complexes and IgM. The actual extent of protein removal during cyrofiltration depends on the specific composition of the plasma and thus on the nature as well as the severity of the particular disease state [Sawada et al., 1990]. There is thus considerable uncertainty over the actual components that are removed during cryofiltration under different clinical and/or experimental conditions. The cryogel layer that accumulates on the surface of the membrane also affects the retention of other plasma proteins, which potentially could lead to unacceptable losses even of small proteins such as albumin. It is also possible to alter the selectivity of the secondary membrane filtration by heating the plasma up to or even above physiologic temperatures. This type of thermofiltration has been shown to increase the retention of low- (LDL) and very low (VLDL) density lipoproteins, and this technique has been used for the online removal of these plasma proteins in the treatment of hypercholesterolemia. LDL removal can also be enhanced by addition of a heparin/acetate buffer to the plasma, which causes precipitation of LDL and fibrinogen with the heparin [Sawada et al., 1990]. These protein precipitates can then be removed relatively easily from the plasma by membrane filtration. The excess heparin is subsequently removed from the solution by adsorption, with the acetate and excess fluid removed using bicarbonate dialysis. An attractive alternative to secondary membrane filtration for the selective removal of plasma components is the use of sorbent columns such as: (1) activated charcoal or anion exchange resins for the removal of exogenous toxins, bile acids, and bilirubin; (2) dextran sulfate cellulose for the selective removal of cholesterol, LDL, and VLDL; (3) immobilized protein A for the removal of immunoglobulins (particularly IgG) and immune complexes; and (4) specific immobilized ligands like DNA (for the removal of anti-DNA Ab), tryptophan (for the removal of antiacetylcholine receptor antibodies), and insulin (for the removal of anti-insulin antibodies). These sorbents provide a much more selective separation than is possible with any of the membrane processes; thus they have the potential to significantly reduce the side effects associated with the depletion of needed plasma components. The sorbent columns generally are used in combination with membrane plasmaphersis, since the platelets that are present in the plasma collected from available centrifugal devices can clog the columns and interfere with the subsequent protein separation. The development of effective sorbent technology for online plasma treatment has been hindered by the uncertainties regarding the actual nature of the plasma components that must be removed for the clinical efficacy of therapeutic apheresis in the treatment of different disease states. In addition, the use of biologic materials in these sorbent systems (e.g., protein A or immobilized DNA) presents particular challenges, since these materials may be strongly immunogenic if they desorb from the column and enter the circulation.

132.3

Cytapheresis

Cytapheresis is used to selectively remove one (or more) of the cellular components of blood, with the other components (including the plasma) returned to the patient. For example, leukocyte (white cell) removal has been used in the treatment of leukemia, autoimmune diseases with a suspected cellular immune mechanism (e.g., rheumatoid arthritis and myasthenia gravis), and renal allograft rejection. Erythrocyte (red cell) removal has been used to treat sickle cell anemia, severe autoimmune hemolytic anemia, and severe parasitemia. Plateletapheresis has been used to treat patients with thrombocythemia. Most cytapheresis is performed using either continuous or intermittent flow centrifuges, with appropriate software and/or hardware modifications used to enhance the collection of the specific cell fraction. It is also possible to remove leukocytes from whole blood by depth filtration, which takes advantage of the strong adherence of leukocytes to a variety of polymeric materials (e.g., acrylic, cellulose acetate, polyester, or nylon fibers). Leukocyte adhesion to these fibers is strongly related to the configuration and the diameter of the fibers, with the most effective cell removal obtained with ultrafine fibers less than 3 µm in diameter. Available leukocyte filters (Sepacel, Cellsora, and Cytofrac from Asahi Medical Co.)



have packing densities of about 0.1–0.15 g fiber/cm3 and operate at blood flow rates of 20–50 ml/min, making it possible to process about 2 L of blood in 1.5 hr. Leukocyte filtration is used most extensively in blood-banking applications to remove leukocytes from the blood prior to transfusion, thereby reducing the likelihood of antigenic reactions induced by donor leukocytes and minimizing the possible transmission of white-cell associated viral diseases such as cytomegalovirus. The absorbed leukocytes can also be eluted from these filters by appropriate choice of buffer solution pH, making it possible to use this technique for the collection of leukocytes from donated blood for use in the subsequent treatment of leukopenic recipients. Depth filtration has also been considered for online leukocyte removal from the extracorporeal circuit of patients undergoing cardiopulmonary bypass as a means to reduce the likelihood of postoperative myocardial or pulmonary reperfusion injury which can be caused by activated leukocytes. A new therapeutic technique that involves online cytapheresis is the use of extracorporeal photochemotherapy, which is also known in the literature as photopheresis. Photopheresis can be used to treat a variety of disorders caused by aberrant T-lymphocytes [Edelson, 1989], and it has become an established therapy for the treatment of advanced cutaneous T-cell lymphoma in the U.S. and several European countries. In this case, the therapy involves the use of photoactivated 8-methoxypsoralen, which blocks DNA replication causing the eventual destruction of the immunoactive T-cells. The psoralen compound is taken orally prior to the phototherapy. Blood is drawn from a vein and separated by centrifugation. The white cells and plasma are collected, diluted with a saline solution, and then pumped through a thin plastic chamber in which the cells are irradiated with a high-intensity UV light that activates the psoralen. The treated white cells are then recombined with the red cells and returned to the patient. Since the photoactivated psoralen has a half-life of only several microseconds, all its activity is lost prior to reinfusion of the cells, thereby minimizing possible side effects on other organs. The removal of the red cells (which have a very high adsorptivity to UV light) makes it possible to use a much lower energy UV light, thereby minimizing the possible damage to normal white cells and platelets. Photopheresis has also been used in the treatment of scleroderma, systemic lupus erythematosus, and pemphigus vulgaris. The exact mechanism for the suppression effect induced by the photo-therapy in these diseases is uncertain, although the T-cell destruction seems to be highly specific for the immunoactive T-cells [Edelson, 1989]. The response is much more involved than simple direct photoinactivation of the white cells; instead, the photo-treated cells appear to undergo a delayed form of cell death which elicits an immunologic response possibly involving the production of anti-idiotypic antibodies or the generation of clone-specific suppressor T-cells. This allows for an effective “vaccination” against a particular T-cell activity without the need for isolating or even identifying the particular cells that are responsible for that activity [Edelson, 1989]. Phototherapy has also been used for virus inactivation, particularly in blood-banking applications prior to transfusion. This can be done using high-intensity UV light alone or in combination with specific photoactive chemicals to enhance the virus inactivation. For example, hematoporphyrin derivatives have been shown to selectively destroy hepatitis and herpes viruses in contaminated blood. This technique shows a high degree of specificity toward this type of enveloped virus, which is apparently due to the affinity of the photoactive molecules for the lipids and glycolipids that form an integral part of the viral envelope. Another interesting therapeutic application involving cytapheresis is the ex vivo activation of immunologically active white cells (lymphokine-activated killer cells, tumor-infiltrating lymphocytes, or activated killer macrophages) for the treatment of cancer. The detailed protocols for this therapy are still being developed, and there is considerable disagreement regarding its actual clinical efficacy. A pool of activated cells is generated in vivo by several days of treatment with interleukin-2. These cells are then collected from the blood by centrifugal cytapheresis and further purified using density gradient centrifugation. The activated cells are cultured for several days in a growth media containing additional interleukin-2. These ex-vivo activated cells are then returned to the patient, where they have been shown to lyse existing tumor cells and cause regression of several different metastatic cancers.



132.4 Summary Apheresis is unique in terms of the range of diseases and metabolic disorders which have been successfully treated by this therapeutic modality. This broad range of application is possible because apheresis directly alters the body’s immunologic system though the removal or alteration of specific immunologically active cells and/or proteins. Although there are a number of adverse reactions that can develop during apheresis (e.g., fluid imbalance, pyrogenic reactions, depletion of important coagulation factors, and thrombocytopenia), the therapy is generally well tolerated even by patients with severely compromised immune systems. This has, in at least some instances, led to the somewhat indiscriminate use of therapeutic apheresis for the treatment of diseases in which there was little physiologic rationale for the application of this therapy. This was particularly true in the 1980s, where dramatic advances in the available technology for both membrane and centrifugal blood fractionation allowed for the relatively easy use of apheresis in the clinical milieu. In some ways, apheresis in the 1980s was a medical treatment that was still looking for a disease. Although apheresis is still evolving as a therapeutic modality it is now a fairly well-established procedure for the treatment of a significant number of diseases (most of which are relatively rare) in which the removal of specific plasma proteins or cellular components can have a beneficial effect on the progression of that particular disease. Furthermore, continued advances in the equipment and procedures used for blood fractionation and component removal have, as discussed in this chapter, provided a safe and effective technology for the delivery of this therapy. The recent advances in sorbent-based systems for the removal of specific immunologically active proteins and in the development of treatment for the activation or inactivation of specific cellular components of the immune system has provided exciting new opportunities for the alteration and even control of the body’s immunologic response. This includes: (1) the direct removal of specific antibodies or immune complexes (using membrane plasmapheresis with appropriate immunosorbent columns), (2) the inactivation or removal of specific lymphocytes (using centrifugal cytapheresis in combination with appropriate extracorporeal phototherapy or chemotherapy), and/or (3) the activation of a diseasespecific immunologic response (using cytapheresis and ex vivo cell culture with appropriate lymphokines and cell stimuli). New advances in our understanding of the immune system and in our ability to selectively manipulate and control the immunologic response should thus have a major impact on therapeutic apheresis and the future development of this important medical technology.

Defining Terms Autoimmune diseases: A group of diseases in which pathological antibodies are produced that attack the body’s own tissue. Examples include glomerulonephritis (characterized by inflammation of the capillary loops in the glomeruli of the kidney) and myasthenia gravis (characterized by an inflammation of the nerve/muscle junctions). Cascade filtration: The combination of plasmapheresis with a second online membrane filtration of the collected plasma to selectively remove specific toxic or immunogenic components from blood based primarily on their size. Cytapheresis: A type of therapeutic apheresis involving the specific removal of red blood cells, white cells (also referred to as leukapheresis), or platelets (also referred to as plateletapheresis). Donor apheresis: The collection of a specific component of blood (either plasma or one of the cellular fractions), with the return of the remaining blood components to the donor. Donor apheresis is used to significantly increase the amount of plasma (or a particular cell type) that can be donated for subsequent use in blood banking and/or plasma fractionation. Immune complexes: Antigen-antibody complexes that can be deposited in tissue. In rheumatoid arthritis this deposition occurs primarily in the joints, leading to severe inflammation and tissue damage. Photopheresis: The extracorporeal treatment of diseases characterized by aberrant T-cell populations using visible or ultraviolet light therapy, possibly in combination with specific photoactive chemicals. Plasma exchange: The therapeutic process in which a large volume of plasma (typically 3 L) is removed and replaced by an equivalent volume of a replacement fluid (typically fresh frozen plasma, a plasma substitute, or an albumin-containing saline solution).



Plasma perfusion: The therapeutic process in which a patient’s plasma is first isolated from the cellular elements in the blood and then subsequently treated to remove specific plasma components. This secondary treatment usually involves a sorbent column designed to selectively remove a specific plasma component or a membrane filtration designed to remove a broad class of plasma proteins. Plasmapheresis: The process in which plasma is separated from the cellular components of blood using either centrifugal or membrane-based devices. Plasmapheresis can be employed in donor applications for the collection of source plasma for subsequent processing into serum fractions or in therapeutic applications for the treatment of a variety of disorders involving the presence of abnormal circulating components in the plasma. Therapeutic apheresis: A process involving the separation and removal of a specific component of the blood (either plasma, a plasma component, or one of the cellular fractions) for the treatment of a metabolic disorder or disease state.

References Edelson RL. 1989. Photopheresis: A new therapeutic concept. Yale J Biol Med 62:565. Kambic HE, Nosé Y. 1993. Plasmapheresis: Historical perspective, therapeutic applications, and new frontiers. Artif Organs 17(10):850. Malchesky PS, Asanuma Y, Zawicki I, et al. 1980. On-line separation of macromolecules by membrane filtration with cryogelation. Artif Organs 400:205. Nosé Y, Kambic HE, Matsubara S. 1983. Introduction to therapeutic apheresis. In Y Nosé, PS Malchesky, JW Smith, et al. (eds), Plasmapheresis: Therapeutic Applications and New Techniques, pp 1–22, New York, Raven Press. Rock G. 1983. Centrifugal apheresis techniques. In Y Nosé, PS Malchesky, JW Smith, et al. (eds), Plasmapheresis: Therapeutic Applications and New Techniques, pp 75–80, New York, Raven Press. Sawada K, Malchesky P, Nosé Y. 1990. Available removal systems: State of the art. IN UE Nydegger (ed), Therapeutic Hemapheresis in the 1990s, pp 51–113, New York, Karger. Solomon BA, Castino F, Lysaght MJ, et al. 1978. Continuous flow membrane filtration of plasma from whole blood. Trans AM Soc Artif Intern Organs 24:21. Zeman LJ, Zydney AL. 1986. Microfiltration and Ultrafiltration: Principles and Applications, pp. 471–489, New York, Marcel Dekker. Zydney AL, Colton CK. 1982. Continuous flow membrane plasmaphersis: Theoretical models for flux and hemolysis prediction. Trans Am Soc Artif Intern Organs 28:408. Zydney AL, Colton CK. 1986. A concentration polarization model for filtrate flux in cross-flow microfiltration of particulate suspensions. Chem Eng Commun 47:1. Zydney AL, Colton CK. 1987. Fundamental studies and design analyses of cross-flow membrane plasmapheresis. In JD Andrade, JJ Brophy, DE Detmer (eds), Artificial Organs, pp 343-358, VCH Publishers.

Further Information Several of the books listed above provide very effective overviews of both the technical and clinical aspects of therapeutic apheresis. In addition, the Office of Technology Assessment has published Health Technology Case Study 23: The Safety, Efficacy, and Cost Effectiveness of Therapeutic Apheresis, which has an excellent discussion of the early clinical development of apheresis. Several journals also provide more detailed discussions of current work in apheresis, including Artificial Organs and the Journal of Clinical Apheresis. The abstracts and proceedings from the meetings of the International Congress of the World Apheresis Association and the Japanese Society for Apheresis also provide useful sources for current research on both the technology and clinical applications of therapeutic apheresis.


Galletti, P. M., Jauregui, H. O. “Liver Support Systems.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

133 Liver Support Systems

Pierre M. Galletti (deceased)

133.1 133.2 133.3 133.4 133.5 133.6

Source of Functional Cells • Supporting Structures

Hugo O. Jauregui Rhode Island Hospital

133.1

Morphology of the Liver Liver Functions Hepatic Failure Liver Support Systems Global Replacement of Liver Function Hybrid Replacement Procedures

133.7

Outlook

Morphology of the Liver

The liver is a complex organ that operates both in series and in parallel with the gastrointestinal tract. After entering the portal system, the products of digestion come in contact with the liver parenchymal cells, or hepatocytes, which remove most of the carbohydrates, amino acids, and fats from the feeder circulation, therefore preventing excessive increases throughout the body after a meal. In the liver, these products are then stored, modified, and slowly released to the better advantage of the whole organism. The liver can be considered a complex large-scale biochemical reactor, since it occupies a central position in the metabolism, i.e., the sum of the physical and chemical processes by which living matter is produced, maintained, and destroyed, and whereby energy is made available for the functioning of liver cells as well as tissues from all other organs. The adult human liver (weighing 1500 g) receives its extensive blood supply (on the order of 1 L/min or 20% of cardiac output) from two sources: the portal vein (over two-thirds) and the hepatic artery (about one-third). Blood from the liver drains through the hepatic veins into the inferior vena cava. Macroscopically, the liver is divided into 4 or 5 lobes with individual blood supply and bile drainage channels. Some of these lobes can be surgically separated, although not without difficulty. Microscopically, human hepatocytes (250-500 × 109 in each liver) are arranged in plates (Fig. 133.1) that are radially distributed around the central (drainage) vein [Jones & Spring-Mills, 1977] and form somewhat hexagonal structures, or liver lobules, which are much more clearly demarcated in porcine livers. Present in the periphery of these lobules are the so-called portal triads, in the ratio of three triads for each central vein. In each portal triad, there are tributaries of the portal vein, branches of the hepatic artery, and collector ducts for the bile (Fig. 133.1). Blood enters the liver lobule at the periphery from terminal branches of the portal vein and the hepatic arteries and is distributed into capillaries which separate the hepatocyte plates. These capillaries, called sinusoids, characteristically have walls lined by layers of endothelial cells that are not continuous but are perforated by small holes (fenestrae). Other cells are present in the sinusoid wall, e.g., phagocytic Kuppfer cells, fat-storing Ito cells, and probably a few yet undefined mesenchymal cells. It is important to emphasize that blood-borne products (with the exception of blood cells) have free access to the perisinusoidal space, called the space of Disse, which can be visualized by electron microscopy as a gap separating the sinusoidal wall from the hepatocyte plasma


FIGURE 133.1

FIGURE 133.2

The liver lobule.

Hepatocyte relationships with the space of Disse and the sinusoid wall.

membrane (Fig. 133.2). In this space, modern immunomicroscopic studies have identified three types of collagens: Type IV (the most abundant), Type I, and Type III. Fibronectin and glycosaminoglycans are also found there, but laminin is only present in the early stages of liver development not in adult mammalian livers [Martinez-Hernandez, 1984]. © 2000 by CRC Press LLC

The hepatocytes themselves are large (each side about 25 microns), multifaceted, polarized cells with an apical surface which constitutes the wall of the bile canaliculus (the channel for bile excretion) and basolateral surfaces which lie in close proximity to the blood supply. Hepatocytes constitute 80–90% of the liver cell mass. Kuppfer cells (about 2%) belong to the reticulo-endothelial system, a widespread class of cells which specialize in the removal of particulate bodies, old blood cells, and infectious agents from the blood stream. The cytoplasm of hepatocytes contains an abundance of smooth and rough endoplasmic reticulum, ribosomes, lysosomes, and mitochondria. These organelles are involved in complex biochemical processes: fat and lipid metabolism, synthesis of lipoproteins and cholesterol, protein metabolism, and synthesis of complex proteins, e.g., serum albumin, transferrin, and clotting factors from amino acid building blocks. The major aspects of detoxification take place in the cisternae of the smooth endoplasmic reticulum, which are the site of complex oxidoreductase enzymes known collectively as the cytochrome P-450 system. In terms of excretion, hepatocytes produce bile, which contains bile salts and conjugated products. Hepatocytes also store large pools of essential nutrients such as folic acid, retinol, and cobalamin.

133.2 Liver Functions The liver fulfills multiple and finely tuned functions that are critical for the homeostasis of the human body. Although individual pathways for synthesis and breakdown of carbohydrates, lipids, amino acids, proteins, and nucleic acids can be identified in other mammalian cells, only the liver performs all these biochemical transformations simultaneously and is able to combine them to accomplish its vital biologic task. The liver is also the principal site of biotransformation, activation or inactivation of drugs and synthetic chemicals. Therefore, this organ displays a unique biologic complexity. When it fails, functional replacement presents one of he most difficult challenges in substitutive medicine. Under normal physiologic requirements, the liver modifies the composition and concentration of the incoming nutrients for its own usage and for the benefit of other tissues. Among the major liver functions, the detoxification of foreign toxic substances (xenobiotics), the regulation of essential nutrients, and the secretion of transport proteins and critical plasma components of the blood coagulation system are probably the main elements to evaluate in a successful organ replacement [Jauregui, 1991]. The liver also synthesizes several other critical proteins, excretes bile, and stores excess products for later usage, functions that can temporarily be dispensed with but must eventually be provided. The principal functions of the liver are listed in Table 133.1. The challenge of liver support in case of organ failure is apparent from the complexity of functions served by liver cells and from our still imperfect ability to rank these functions in terms of urgency of replacement. TABLE 133.1

Liver Functions

Carbohydrate metabolism: Glyconeogenesis and glycogenolysis Fat and lipid metabolism: Synthesis of lipoproteins and cholesterol Synthesis of plasma proteins, for example: Albumin Globulins Fibrinogen Coagulation factors Transferrin α-fetoprotein Conjugation of bile acids; conversion of heme to bilirubin and biliverdin Detoxification: Transformation of metabolites, toxins, and hormones into water-soluble compounds (e.g., cytochrome P-450 P-450 oxidation, glucuronyl transferase conjugation) Biotransformation and detoxification of drugs Metabolism and storage of vitamins Storage of essential nutrients Regeneration


133.3

Hepatic Failure

More than any other organ, the liver has the property of regeneration after tissue damage. Removal or destruction of a large mass of hepatic parenchyma stimulates controlled growth to replace the missing tissue. This can be induced experimentally, e.g., two thirds of a rat liver can be excised with no ill effects and will be replaced within 6 to 8 days. The same phenomenon can be observed in humans and is a factor in the attempted healing process characteristic of the condition called liver cirrhosis. Recent attempts at liver transplantation using a liver lobe from a living donor rely on the same expectation of recovery of lost liver mass. Liver regeneration is illustrated by the myth of Prometheus, a giant who survived in spite of continuous partial hepatectomy through the good auspices of a vulture (a surgical procedure inflicted on him as punishment for having stolen the secret fire from the gods and passing it on to humanity). Hepatic failure may be acute or chronic according to the time span it takes for the condition to develop. Mechanisms and toxic by-products perpetuating these two conditions are not necessarily the same. Acute fulminant hepatic failure (FHF) is the result of massive necrosis of hepatocytes induced over a period of days or weeks by toxic substances or viral infection. It is characterized by jaundice and mental confusion which progresses rapidly to stupor or coma. The latter condition, hepatic encephalopathy (HE), is currently thought to be associated with diminished hepatic catabolism. Metabolites have been identified which impair synaptic contacts and inhibit neuromuscular and mental functions (Table 133.2). Although brain impairment is the rule in this condition, there is no anatomic damage to any of the brain structures, and therefore, the whole process is potentially reversible. The mortality rate of FHF is high (70–90%), and death is quite rapid (a week or two). Liver transplantation is currently the only effective form of treatment for FHF. Transplantation procedures carried out in life-threatening circumstances are much more risky than interventions in relatively better compensated patients. The earlier the transplantation procedure takes place, the greater is the chance for patient survival. However, 10–30% of FHF patients will regenerate their liver under proper medical management without any surgical intervention. Hence, liver transplantation presents the dilemma of choosing between an early intervention, which might be unnecessary in some cases, or proceeding to a late procedure with a statistically higher mortality [Jauregui et al., 1994]. Chronic hepatic failure, the more common and progressive form of the disease, is often associated with morphologic liver changes known as cirrhosis in which fibrotic tissue gradually replaces liver tissue as the result of long-standing toxic exposure (e.g., alcoholism) or secondary to viral hepatitis. More than 30,000 people died of liver failure in the United States in 1990. In chronic hepatic failure, damaged hepatocytes are unable to detoxify toxic nitrogenous products that are absorbed by intestinal capillaries and carried to the liver by the portal system. Ammonia probably plays the major role in the deterioration of the patient’s mental status, leading eventually to “hepatic coma.” An imbalance of conventional amino acids (some abnormally high, some low) may also be involved in the pathogenesis of the central nervous system manifestation of hepatic failure, the most dramatic of which is cerebral edema. Impaired blood coagulation (due to decreased serum albumin and clotting factors), hemorrhage in the gastro-intestinal system (increased resistance to blood flow through the liver leads to portal hypertension, ascites formation, and bleeding from esophageal varices), and

TABLE 133.2

Metabolic Products with Potential Effects in Acute Liver Failure

Substance

Mode of action

Ammonia

Neurotoxic interaction with other neurotransmitters Contributes to brain edema Neural inhibition Neural inhibition Inhibition of Na-K ATPase Acts as a false neurotransmitter

Benziodiazepinelike substances GABA Mercaptans Octopamine


hepatic encephalopathy with glial cell damage in the brain are the standard landmarks of chronic hepatic failure. In fact, HE in chronic liver failure is often precipitated by episodes of bleeding and infection, and progression to deep coma is an ominous sign of impending death. Intensive management of chronic liver failure includes fluid and hemodynamic support, correction of electrolyte and acid-base abnormalities, respiratory assistance, and treatment of cerebral edema if present. Aggressive therapy can diminish the depth of the coma and improve the clinical signs, but the outcome remains grim. Eventually, 60–90% of the patients require transplantation. About 2500 liver transplants are performed every year in the United States, with a survival rate ranging from 68–92%. The most serious limitation to liver transplantation (besides associated interrelated diseases) remains donor scarcity. Even if segmented transplants and transplants from living related donors become acceptable practices, it is unlikely that the supply of organs will ever meet the demand. Further, the problem of keeping a patient alive with terminal hepatic failure, either chronic or acute, while waiting for an adequately matched transplant is much more difficult than the parallel problem in end-stage renal disease, where dialysis is a standardized and effective support modality. An appreciation of the modalities of presentation of the two types of hepatic coma encountered in liver failure is needed for a definition of the requirements for the proper use of liver assist devices. In the case of FHF, the hepatologist wants an extracorporeal device that will circulate a large volume of blood through a detoxifying system [Jauregui & Muller, 1992] allowing either the regeneration of the patient’s damaged liver (and the avoidance of a costly and risky liver transplantation procedure) or the metabolic support needed for keeping the patient alive while identifying a cadaveric donor organ. In the first option, the extracorporeal liver assist device functions as an organ substitute for the time it takes the liver to regenerate and recover its function; in the second, it serves as temporary bridge to transplantation. In the case of chronic liver failure today, spontaneous recovery appears impossible. The damaged liver needs to be replaced by a donor organ, although not with the urgency of FHF. The extracorporeal liver assist device (LAD) is used as a bridge while waiting for the availability of a transplant. It follows that the two different types of liver failure may require different bioengineering designs.

133.4

Liver Support Systems

The concept of artificial liver support is predicated on the therapeutic benefit of removing toxic substances accumulating in the circulation of liver failure patients. These metabolites reflect the lack of detoxification by damaged hepatocytes, the lack of clearance of bacterial products from the gut by impaired Kupffer cells, and possibly the release of necrotic products from damaged cells which inhibit liver regeneration. Systemic endotoxemia as well as massive liver injury give rise to an inflammatory reaction with activation of monocytes and macrophages and release of cytokines which may be causally involved in the pathogenesis of multiorgan failure commonly encountered in liver failure. Technologies for temporary liver support focus on the detoxifying function, since this appears to be the most urgent problem in liver failure. The procedures and devices which have been considered for this purpose include the following. Hemodialysis Hemodialysis with conventional cellulosic membranes (cut-off point around 2000 daltons) or more permeable polysulfone or polyacrylonitrile [de Groot et al., 1984] (cut-off 1500–5000 daltons) helps to restore electrolyte and acid-base balance and may decrease the blood ammonia levels but cannot remove large molecules and plasma protein-bound toxins. Improvement of the patient’s clinical condition (e.g., amelioration of consciousness and cerebral edema) is temporary. The treatment appears to have no lasting value and no demonstrated effect on patient survival. In addition, hemodialysis may produce a respiratory distress syndrome caused by a complement-mediated poly-morphonuclear cell aggregation in the pulmonary circulatory bed. Because some of the clinical benefit seems related to the removal of toxic molecules, more aggressive approaches focused on detoxification have been attempted.


Hemofiltration Hemofiltration with high cut-off point membranes (around 50,000 daltons with some polyacrylonitrilepolyvinyl chloride copolymers, modified celluloses, or polysulfones) clears natural or abnormal compounds within limits imposed by convective transport across the exchange membrane. These procedures again have a temporary favorable effect on hepatic encephalopathy (perhaps because of the correction of toxic levels of certain amino acids) with reversal of coma, but they do not clearly improve survival rates. Hemoperfusion Hemoperfusion, i.e., extracorporeal circulation of blood over nonspecific sorbents (e.g., activated charcoal) [Chang, 1975] or more complex biochemical reactors which allow the chemical processing of specific biologic products, such as ammonia, have not yet met clinical success in spite of encouraging experimental results, except in the case of hepatic necrosis induced by poisonous mushrooms such as Amanita phalloides. Anion exchange resins and affinity columns similar to those used in separative chromatography may help in removing protein-bound substances (e.g., bilirubin) which would not pass through dialysis or hemofiltration membranes, but nonspecific sorbents may also deplete the plasma of biologically important substances. Further, these techniques are complicated by problems of hemocompatibility, related in part to the entrainment of dust (“fines”) associated with the sorbent material itself and in part to platelet activation in patients with an already compromised coagulation status. To minimize this problem direct blood or plasma contact with the sorbent material can be avoided by polymer coating of the sorbent particles using either albumin, cellulose nitrate, or similar thin films, but hemocompatibility remains a concern. Here again, there is anecdotal evidence of clinical improvement of hepatic failure with hemoperfusion, with some reports claiming a higher survival rate in hepatic encephalopathy, but these reports have not been supported by well-controlled studies. As is the case for hemodialysis and hemofiltration, the possible beneficial effect of hemoperfusion should be evaluated in the context of the clinical variability in the course of FHF. Lipophilic Membrane Systems Because lipophilic toxins dominate in fulminant hepatic failure, it is conceibable to eliminate such compounds with a hydrophobic (e.g., polysulfone) membrane featuring large voids filled with a nontoxic oil [Brunner & Tegtimeier, 1984]. After diffusion, the toxins can be made water-soluble through reaction with a NaOH-based acceptor solution, thereby preventing their return to the blood stream. A standard, high-flux dialyzer in series with the lipophilic membrane device allows the removal of hydrophilic solutes. Such a system has proved effective in removing toxins such as phenol and p-cresol as well as fatty acids without inducing detrimental side effects of its own. Immobilized Enzyme Reactors To address the problem of specificity in detoxification, enzymes such as urease, tyrosinase, L-asparaginase, glutaminase, and UDP-glucuronyl transferase have been attached to hollow fibers or circulated in the closed dialysate compartment of an artificial kidney or still incorporated into microcapsules or “artificial cells” exposed to blood. There is considerable in vitro evidence for the effectiveness of this approach, and some indication of therapeutic value from in vivo animal experiments [Brunner et al., 1979]. However, no clinical report has documented the superiority of enzyme reactors over the various modalitites of dialysis. Again there are clinical observations of clearing of the mental state of patients in hepatic coma, but no statistically demonstrated effect on survival. It is unclear whether the lack of success is due to the inability of specific enzymes to remove all offending toxins or is evidence of the need for more than detoxification for effective treatment. Parabiotic Dialysis Also referred to as cross-dialysis, parabolic dialysis is a variant of hemodialysis in which the dialysate compartment of a solute exchange device is perfused continuously with blood from a living donor. Because of membrane separation of the two blood streams, the procedure can be carried out even if the


two subjects belong to different blood groups or different animal species. However, the risk of the procedure to a human donor (control of blood volume, transfer of toxic substances, mixing of blood streams in case of dialyzer leak) and the difficulty of introducing a live animal donor into the hospital environment have relegated this approach to the class of therapeutic curiosities. Exchange Transfusion Exchange transfusion, i.e., the quasi-total replacement of the blood volume of a patient in liver failure by alternating transfusion and bleeding, is occasionally used in severe hyperbilirubinemia of the newborn, which used to carry an ominous prognosis because of its association with cerebral edema. The rationale is that exchange transfusion will reduce the level of toxins and replenish the deficient factors in the blood stream while the underlying condition is corrected by natural processes or drugs [Trey et al., 1966]. With the advent of blood component therapy, specific plasma components can also be administered to treat identified deficiencies. Mortality rates of patients treated with exchange blood transfusions have been reported as greater than those observed with conventional therapies. Plasmapheresis Plasmapheresis, i.e., the combination of withdrawal of blood, centrifugation, or membrane processing to separate and discard the patient’s plasma, and return of autologous cells diluted with donor plasma, was practiced initially as a batch process. Techniques now exist for a continuous exchange process, in which plasma and cells are separated by physical means outside of the body (membrane separation or centrifugation), and the patient’s plasma replaced by banked plasma (up to 5000 ml per day) [Lepore et al., 1972]. There is evidence from controlled clinical trials for the effectiveness of this form of therapy, but the mortality rate remains high in patients with hepatic failure, whether from insufficient treatment or the risks of the procedure. It appears, however, that plasma exchange can be beneficial in the preoperative period prior to liver transplantation so as to correct severe coagulopathy. Plasmapheresis is used in conjunction with the placement of a hepatocyte-seeded extracorporeal hollow-fiber device to treat acute and chronic liver failure [Rozga et al., 1993]. Combined Therapy Endotoxins and cytokines can be removed by hemoperfusion over activated charcoal and absorbent resins, but it may be more effective to process plasma than whole blood. This has led to the concept of combining plasmapheresis with continuous plasma treatment for removal of substances such as tumor necrosis factor (TNF), interleukin-6 (IL-6), and bile acids by a resin column, and then ultrafiltration or dialysis for fluid removal, since patients with liver failure often develop secondary renal failure.

133.5

Global Replacement of Liver Function

Because of the complexity and interplay of the various functions of the liver, more success can be expected from global approaches, which allow many or all hepatic functions to be resumed. These include the following. Cross-Criculation Cross-circulation of the patient in hepatic coma with a compatible, healthy donor is one approach. This procedure is more than a prolonged exchange transfusion since it allows the donor’s liver to substitute for the patient’s failing organ and to process chemicals from the patient’s blood stream as long as the procedure lasts [Burnell, 1973]. It had been attempted in isolated cases, but reports of effectiveness are entirely anecdotal and the procedure has not been accepted clinically because of ambiguous results and the perceived risk for the donor. Hemoperfusion over Liver Tissue Slices The incorporation of active hepatocytes in a hemoperfusion circuit was suggested by the laboratory practice of biochemists who, since Warburg, have investigated metabolic pathways in tissue slices. For


liver replacement, this technology has been pioneered primarily in Japan as a substitute for organ transplantation, which is culturally frowned upon in that country in spite of a major incidence of severe liver disease. The procedure may improve biochemical markets of liver failure but has no demonstrated clinical value [Koshino et al., 1975]. Ex Vivo Perfusion Ex vivo perfusion uses an isolated animal liver (pig or baboon) connected to the patient’s cardiovascular system [Saunders et al., 1968]. This is a cleaner and more acceptable form of treatment than crosscirculation of hemoperfusion over tissue pieces. Nevertheless, it is limited by the need for thorough washing of the animal’s blood from the excised liver, the requirement for a virus-free donor organ source, the limited survival capacity of the excised, perfused organ, which must be replaced at intervals of approximately 24 hours, and the cost of the procedure. Success has recently been reported in isolated clinical trials [Chari, 1994]. Heterotopic Hepatocyte Transplantation This procedure may someday offer an alternative to whole organ transplantation, especially in cases of chronic liver failure, if a sufficient number of cells can be grafted. Freshly isolated hepatocytes have damaged cell surfaces and must be cultured to regain their integrity and display the surface receptors needed for attachment or binding of xenobiotics or endogenous toxic products [McMillan et al., 1988]. At the clinical level, this procedure could rely in part on banking frozen hepatocytes from livers which are not usable for whole organ transplants but constitute a reliable source of cells [Bumgardener et al., 1988]. The procedure does not require removal of the recipients’s liver and provides the followng advantages: (1) minimal surgery, (2) repeatability as needed, and (3) interim strategy to whole organ transplant. There is no agreement as to the best anatomic site for hepatocyte transplantation, the type of matrix needed for cell attachment and differentation, and the number of cells needed. It has been reported that hepatocyte culture supernatants were as effective as transplanted hepatocytes in treating rats with chemically induced liver failure [LaPlante-O’Neill et al., 1982]. The tridemensional reconstruction of high-density, functional liver tissue will be the greatest challenge in view of the cell mass required. Structural organization and differentiated functions may be achieved by using an asialoglycoprotein model polymer as the synthetic substrate for a primary culture of hepatocytes to develop functional modules for implantation in humans [Akaike et al., 1989]. Whole Organ in Situ Transplantation This is currently the procedure of choice, most particularly in children. Although progress in clinical and surgical skills certainly accounts in large part for the growing success rate of this procedure over the past 20 years, it is worth noting that the introduction of extracorporeal ciruclation techniques in the surgical protocol to support the donor organ while waiting for completion of the anastomosis has paralleled the steepest increase in success rate of liver transplantation in the past few years. Whereas most liver transplants rely on the availability of cadaver organs, the recent advent of segmented transplants allows consideration of living related donors and possibly the sharing of a donor organ among two or more recipients.

133.6

Hybrid Replacement Procedures

The complexity of hepatic functions, coupled with shortage of human donor organs, has encouraged the development of procedures and devices which rely on xenogeneic living elements attached to synthetic structures and separated from the most host by a permselective membrane to replace temporarily, and perhaps someday permanently, the failing organ. The incorporation of liver microsomes in microcapsules, hydrogels, or polymer sheets, with or without additonal enzymes or pharmacologic agents, goes one step beyond the immobilized enzyme reactor inasmuch as it calls on cell components endowed with a variety of enzymatic properties to process blood


or other body fluids. The feasibility of this technique has been demonstrated in vitro [Denti et al., 1976] but has not been investigated extensively in animal experiments. Cellular hybrid devices—the incorporation of functional cells in a device immersed in body fluids or connected to the vascular system (extracorporeally or somewhere inside the body of the recipient)—are a promising application of the concept of bioartificial organs. However, the problems faced by the “hepatocyte bioreactor” are formidable: 1. The mass of functional cells required is much larger than in the case of secretory or endocrine organs, since a normal liver weighs about 1.5 kg and since as much as 10–30% of that mass (i.e., several billion cells) may be required for life-sustaining replacement of function. Taking into account the need for supporting structures, the sheer size of the device will be an obstacle to implantation. 2. The liver features a double feeder circulation (portal and hepatic) and a complex secretoryexcretory system which utilizes both blood and bile to dispose of its waste products. How to duplicate such a complex manifolding in an artificial organ and whether it is worth the resulting complexity are questions not yet resolved. 3. With membrane separation of recipient and donor cells, the size of some natural macromolecules to be exchanged (e.g., low-density lipoproteins) precludes the use of standard diffusion membranes. Allowing relatively free solute exchange between the compartments in a device without endangering the immune sequestration of the donor tissue remains a challenge for membrane technology. However, the low immunogenicity of hepatocytes (which lack type I HLA antigens) allows the consideration of relatively open membranes for the construction of extracorporeal reactors. At the moment, most of the extracorporeal liver assist devices (ELAD) utilize xenogeneic mammalian hepatocytes seeded on solid, isotropic hollow fibers. Membrane selectivity limits the rate of diffusion of liposoluble toxins which are bound to plasma proteins for transport. Hence manipulation of membrane transport properties and concentrations of acceptor protiens can affect the clearance of polar materials. Also, most devices focus on replacing the detoxification function of the natural liver and avoid the more complex “cascade” dialysis circuitry which could either (1) allow the macro-molecules synthesized by hepatocytes to return to the blood stream (or another body fluid compartment) though a high cutoff point membrane or (2) provide an excretory path to clear toxic products manufactured by the hepatoctyes, on the model of the bile excretory system. Nonetheless, such circuitry might be valuable to prolong the life of the seeded cells, since the combination of bile salts and bile acids has a damaging detergent effect on the lipid components of the hepatocyte membrane. The development of bioreactors including cells capable of performing liver functions, and therefore capable of providing temporary liver support, finds applications in the treatment of acute, reversible liver failure or as a bridge for liver transplantation. Designs for a bioartificial liver can be classified according to (1) the type of cells selected to replace the hepatic functions or (2) the geometry and chemical nature of the polymer structure used organize the hepatocytes.

Source of Functional Cells Two main methods of hepatocyte isolation (mechanical and enzymatic) can be used, separately or in combination. Mechanical methods (tissue dissociation) have largely failed in terms of long-term cell viability, although this is not always recognized by investigators developing liver assist systems. Collagenase perfusion [Seglen, 1976] is today the method of choice, yet there is evidence that hepatocytes lose some of their oligosaccharide-lectin binding capacity in the process and do not recover their glycocalyx until after a day or two in culture. Collagenase is thought to loosen cell junctions and secondarily to digest the connective tissue around the hepatocytes. Chemical methods using citrate, EDTA, and similar substances weaken cell junctions by depleting calicum ions without altering the cell membranes and presumably result in better preservation of natural enzymatic functions. Although the yield of chemical methods is lower than that of enzymatic dissociation methods, it may be of interest once a reliable technology for separating viable from nonviable cells on a large scale is perfected.


FIGURE 133.3

Cellular choices based on proliferation and differentiation.

A priori, an effective bioartificial liver would require either all the multifunctional characteristics of normal hepatocytes in vivo or only the specific functional activity that happens to be missing in the patient. Unfortunately, in most cases, clinical signs do not allow us to distinguish between these two extremes, justifying a preference for highly diffrentiated cells. Differentiation and proliferation are usually at opposite ends of a biologic continuum in most cell types (Fig 133.3) [Jauregui, 1991]. Hence, there is a difficulty of obtaining large numbers of multifunctional hepatocytes. Several options are available: 1. The simplest approach is to use adult mammalian liver cells isolated in sufficient number from large animals. Porcine hepatocytes are preferred for clinical applications because of the availability of virus-free donors. Large-scale isolation of porcine hepatocytes is becoming a routine procedure, and Demetriou has been able to treat several patients with such a system. The expectation is that hepatocytes in suspension or attached to synthetic microcarriers will remain metabolically active even though separated from neighboring cells and normal supportive structures. This approach, which appeared almost beyond practicality a few years ago, seems now less formidable because of expanded knowledge of the molecular factors which favor both hepatocyte attachment to polymeric substrates and functional differentiation. The potential contribution of Kupffer cells to a bioartificial liver has not yet been extensively investigated. 2. The use of liver tumor cells—preferably nonmalignant hepatoma—has been pioneered by Wolfe and Munkelt [1975] because such cells can proliferate indefinitely and therefore require a minimal seeding dose. They are also less anchor-dependent than normal cells. The drawbacks are the loss of specialized hepatic functions often encountered in tumor cell lines, and the theroretical risk of escape of tumor cells in the recipient. 3. A modified, functionally differentiated, human hepatoblastoma cell line capale of growing to high densities, yet strongly contact-inhibited by containment with membranes, has been patented and used in animals and man by Sussman and colleagues [1992]. Evidence for metabolic effectiveness has been reported, and clinical trials are now in progress. The uncertainty associated with the use of a human tumor cell line remains a source of concern. 4. Potentially replicating hepatocytes, such as those obtained from embryonic liver, neonatal animals, or recently hepatectomized adults, have been proposed as a means to obtain a stem cell population with enhanced proliferation capacity. Neonatal hepatocyte-based bioartificial devices have been built on that principle [Hager et al., 1983] and have shown to produce albumin, to metabolize urea, to deaminate cytidine, and to detoxify drugs such as a diazepam. The reliance on “juvenile” cells has become less important with the identification of molecular mechanisms for growth control mitogens such as EGF, TGFa, hepatopoieten B, HSS (hepatic stimulatory substance) and HGF


TABLE 133.3

Present Choice of the Cellular Component for Extracorporeal Liver Assist Devices

Sussman & Kelly (1993)

Neuzil et al.

Jauregui & Muller (1992)

Hepatoblastomaderived cell line Cells divide indefinitely Cells are cultured in the device

Porcine hepatocytes separated via portal vein perfusion Hepatocyte division has not been tested (5 to 8) Hepatocytes are seeded on microcarriers and introduced in the device immediately before clinical use

Porcine hepatocytes separated via hepatic vein perfusion Limited number of hepatocyte doublings Hepatocytes may be seeded or cultured in the device through a proprietary technology

(hepatoctye growth factor), co-mitogenic factors such as norepinephrine and growth inhibitors such as TGFβ, and interleukin-1β. 5. Transfected or transgeic hepatocytes may provide the ultimate solution to the cell supply problem. However, they are usually selected for the monoclonal expression of a single function and therefore may not be suitable except when a single cause of hepatic failure has been identified. Alternatively, a combination of different transfected hepatocytes may be considered. Table 133.3 illustrates the choices made by three different groups of investigators for clinical liver assist devices.

Supporting Structures Microencapsulation in a nutrient liquid or a polymer gel surrounded by a conformal membrane provides a suspension of metabolically active units which can either be placed in a container for extracorporeal hemoperfusion, introduced into the peritoneal cavity for implantation, or even infused into the portal vein for settling in the patient’s liver. The blood or tissue reaction to multiple implants and the longterm in vivo stability of hydrogels based on polyelectrolytes remain unresolved problems. A flat sheet of membrane or a spongy matrix coated with attachment factors can be used to anchor a suspension of functional cells. The limitation of this approach is the need for vascularization of the implant, since the cells are metabolically active and therefore are quite avid of oxygen and nutrients. Microcarrier-attached hepatocytes are attractive because the technology of suspension cultures is amenable to the proliferation of a large number of cells. The bulk of the carrier beads probably limits this approach to the extracorporeal circuits similar to those used for hemoperfusion [Rozga et al., 1993]. Hollow-fiber ELADs are constructed by filling the interstices within a bundle of parallel hollow fibers with liver cells, using the lumen of the tubes to provide metabolic support and an excretion channel, typically by circulating blood from an extracorporeal circuit. Alternatively, a hybrid organ can be built by filling the lumen of the hollow fibers with functional cells and implanting the bundle in the body cavity to allow exchanges with the surrounding fluid. One could combine such a device with an oxygenation system, e.g., by filling the peritoneal cavity with a high-oxygen-capacity fluid such as a fluorocarbon and a gas exchange system for that fluid analogous to the intravenous oxygenator (IVOX).

133.7

Outlook

Medical trials with extracorporeal hollow-fiber systems are already in progress, although there are some unanswered questions in the proper design of a hepatocyte-seeded ELAD. A review of Table 133.4 will indicate that there is no consensus for implementing cellular choices in the construction of such systems. Some researchers believe that the culture of hepatocytes on a synthetic matrix prior to chemical application is a complicated proposition that impairs the practical application of this technology. An alternative could be the isolation, freezing, and shipping of porcine hepatocytes to the medical centers treating patients with acute and chronic liver failure (usually transplantation centers). This approach may offer a direct and economically sound solution. Unfortunately, many FHF patients require emergency liver support and are managed in secondary care medical institutions which have neither a transplantation


TABLE 133.4

Consideration in the Choice of the Cellular Component of Extracorporeal Liver Assist Devices

Advantages Hepatoblastoma cell lines are easy to culture and are free of other cell contaminants Tumor cell lines are not anchorage dependent Porcine or primary hepatocytes respond to physiologic stimulation Porcine hepatocytes express P450 (detoxification activity)

Disadvantages Hepatoblastoma cell lines are tumorigenic Tumor cell lines may not respond to physiologic regulation Porcine hepatocytes have limited proliferation ability Porcine hepatocytes show limited life span

program nor a tissue culture facility for the seeding of hollow-fiber devices. Part of the argument for the porcine hepatocyte isolation, shipment, seeding, and immediate use in a bioartificial liver system relies on the earlier concept that primary mammalian hepatocytes do not grow in vitro or are very difficult to maintain in hollow-fiber devices. In fact, rodent hepatocytes have shown excellent detoxification activities when cultured on perfused hollow fiber cultures. Rabbit hepatocytes also survive in hollow-fiber bioreactors which have proved successful in treating one of the most representative animal modes of human FHF. The technology for manufacturing ELADs may also need further attention. For instance, the efficiency of hollow fibers in maintaining hybridoma cultures and producing specific proteins is known to be inferior to that of cellular bioreactors based on microcarrier technology. The surface of hollow fibers has been optimized for blood compatibility but not for hepatocyte attachment, and therefore new materials or structures may have to be developed. At the conceptual level, primary hepatocytes may need to be cultured either in combination with other cells that will provide parabiotic support or on substrates that imitate the composition of the extracellular matrix found in situ in the space of Disse. Our own experience [Naik et al., 1990] and that of others [Singhvi et al., 1994] raise some questions about the role of substrates in maintaining long-term hepatocyte viability in vitro. In fact, most of them operate in a rather indiscriminate fashion by providing an anchor for hepatocytes. Collagen, types I, III, IV, and fibonectin contribute to cytoplasmic spreading which in the long term does not favor maintenance of the pheontypic expression of hepatocytes. Polymer compositions expressing surface sugar residues responsible for hepatocyte attachment through the asailoglycoprotein receptor (a plasma membrane complex present mainly in the bile canalicular area which internalizes plasma asiaglycoprotein) appear able to maintain hepatocytes in culture with extended functional activities [Akaike et al., 1989]. Other investigators have shown the value of extracellular glycoproteins and glycosaminoglycans rich in laminin (e.g., Matrigel) [Caron & Bissell, 1989]. Hepatocytes immobilized on these substraes do not spread but maintain their tridimensional morphology, as well as their functional activity. Such observations suggest that long-term expression of hepaotcytespecific functions depend on maintaining the in situ cell shape and their spacial interrelations [Koide et al., 1990]. Future ELAD designs should provide not only ideal polymer substrates for cell attachment but also a special configuration that will maintain the tridimensional structure of hepatic tissue. Experience with artificial orans shows that the development of these devices tends to underestimate the effort and the technical advances needed to pass from a “proof of principle” prototype for animal evaluation to fabrictaion of a clinically acceptable system for human use. Full-scale design, cell procurement, cell survival, and device storage are all major bottlenecks on the way to a clinical product. One of the gray areas remains in the molecular weight cut-off of the hollow fiber wall. Which substances are responsible for the development of FHF and must therefore be cleared? Are they protein-bound, middlemolecular-, or low-molecular-size compounds? In the absence of clear answers to these questions, the designer of separation membranes is in an ambigous position; some investigators use low molecular cutoff membranes to guarantee immunoseparation of the xenograft, whereas others prefer high-molecular cut-off to enhance functionality. The use of microporous membranes in human subjects has not led, as of yet, to any hypersensitivity reactions in spite of potential passage of immunglobulins. A reliable way to restore consciousness in animal models of FHF is to cross-circulate the blood with a normal animal. To the extent that ELADs may function in patients as in these animal models, they may prove


successful only in relieving the symptoms of the disorder without increasing the survival rate. For instance, when 147 patients with advanced stages of HE were treated with charcoal hemoperfusion over a 10-year period, all showed symptomatic improvement, but the survival rate of 32% was the same as in five control groups. Therefore, enthusiasm generated by preliminary human trials with hollow-fiber devices should be tempered with a cautious approach. Without reliable control studies, we will remain in the position defined by Benhamou and coworkers [1972], “The best future one can wish for a sufferer from severe acute hepatic failure is to undergo a new treatment and have his case published—Be published or perish!”

Defining Terms Bioartificial liver: A liver assist or liver replacement device incorporating living cells in physical or chemical processes normally performed by liver tissue. Catabolism: The aspect of metabolism in which substances in living tissues are transformed into waste products or solutes of simpler chemical composition. (The opposite process is called anabolism.) Cirrhosis: A degenerative process in the liver marked by excess formation of connective tissue destruction of functional cells, and, often, contraction of the organ. Conjugated: The joining of two compounds to produce another compound, such as the combination of a toxic product with some substance in the body to form a detoxified product, which is then eliminated. ELAD: Extracorporeal liver assist device. Homeostasis: A tendency of stability in the normal body states (internal environment) of the organism. This is achieved by a system of control mechanisms activated by negative feedback, e.g., a high level of carbon dioxide in extracellular fluid triggers increased pulmonary ventilation, which in turn causes a decrease in carbon dioxide concentration. Mesenchymal cells: The meshwork of embryonic connective tissue in the mesoderm from which are formed the connective tissues of the body and the blood vessels and lymphatic vessels. Metabolism: The sum of all the physical and chemical processes by which living organized substance is produced and maintained (anabolism) and the transformation by which energy is made available for the uses of the organism (catabolism). Parenchymal: The essential elements of an organ. Phagoctyic: Pertaining to or produced by any cells that ingest microrganisms of other cells and foreign particles. Portal triad: These are microscopic areas of collagen type I-III fibroblasts as well as other connective tissue elements. These triads have a branch of the portal vein, a branch of the hepatic artery, and intermediate caliber bile ducts. Xenobiotic: A chemical foreign to the biologic system.

References Akaike T, Kobayashi A, Kobayashi K, et al. 1989. Separation of parenchymal liver cells using a lactosesubstituted styrene polymer substratum. J Bioact Compat Polym 4:51. Benhamou JP, Rueff B, Sicot C. 1972. Severe hepatic failure: A critical study of current therapy. In F Orlandi, AM Jezequel (eds), Liver and Drugs, New York, Academic Press. Brunner G, Holloway CJ, Lösgen H. 1979. The application of immobilized enzymes in an artificial liver support system. Artif Organs 3:27. Brunner G, Tegtimeier F. 1984. Enzymatic detoxification using lipophilic hollow fiber membranes. Artif Organs 8:161. Bumgardner GL, Fasola C, Sutherland DER. 1988. Prospects for hepatocyte transplantation. Hepatology 8:1158. Burnell JM, Runge C, Saunders FC, et al. 1973. Acute hepatic failure treated by cross circulation. Arch Intern Med 132:493.


Caron JM, Bissel DM. 1989. Extracellular matrix induces albumin gene expression in cultured rat hepatocytes. Hepatology 10:636. Chang TMS. 1975. Experience with the treatment of acute liver failure patients by hemoperfusion over biocompatible microencapsulated (coated) charcoal. In R Williams, IM Murray-Lyon (eds), Artificial Liver Support, Tunbridge Wells, England, Pitman Medical. Chari RS, Collins BH, Magee JC, et al. 1994. Brief Report: Treatment of hepatic failure with ex vivo pigliver perfusion followed by liver transplantation. N Engl J Med 331:134. De Groot GH, Schalm SW, Schicht I, et al. 1984. Large-pore hemodialytic procedures in pigs with ischemic hepatic necrosis; a randomized study. Hepatogastroenterology 31:254. Denti E, Freston JW, Marchisi M, Et al. 1976. Toward a bioartificial drug metabolizing system: Gel immobilized liver cell microsomes. Trans Am Soc Artif Intern Orans 22:693. Hager JC, Carman R, Porter LE, et al. 1983. Neonatal hepatocyte culture on artificial capillaries: A model for drug metabolism and the artificial liver. ASAIO J 6:26–35. Jauregui HO. 1991. Treatment of hepatic insufficiency based on cellular therapies. Int J Artif Organs 14:407. Jauregui HO, Muller TE. 1992. Long-term cultures of adult mammalian hepatocytes in hollow fibers as the cellular component of extracorporeal (hybrid) liver assist devices. Artif Organs 16(2):209. Jauregui HO, Naik S, Santangini H, et al. 1994. Primary cultures of rat hepatocytes in hollow fiber chambers. In Vitro Cell Dev BIol 30A:23. Jones Al, Spring-Mills E. 1977. The liver and gallbladder. In Weiss, RO Greep, (eds), Histology, 4th ed, New York, McGraw-Hill. Koide NH, Sakaguchi K, Koide Y, et al. 1990. Formation of multicellular spheroids composed of adult rat hepatocytes in dishes with positively charged surfaces and under other nonadherent environments. Exper Cell Res 186:227. Koshino I, Castino F, Yoshida K, et al. 1975. A biological extracorporeal metabolic device for hepatic support. Trans Amer Soc Artif Intern Organs 21:492. LaPlante-O’Neill P, Baumgarner D, Lewis WI, et al. 1982. Cell-free supernatant from hepatocyte cultures improves survival of rats with chemically induced acute liver failure. J Surg REs 32:347. Lepore MJ, Stutman LJ, Bonnano CA, et al. 1972. Plasmapheresis with plasma exchange in hepatic coma. II. Fulminant viral hepatitis as a systemic disease. Arch Intern Med 129:900. Martinez-Hernandez A. 1984. The hepatic extracellular matrix. I. Electron immunohistochemical studies in normal rat liver. Lab Invest 51:57. McMillan PN, Hevey KA, Hixson DC, et al. 1988. Hepatocyte cell surface polarity as demonstrated by lectin binding to isolate and cultured hepatocytes. J Histochem Cytochem 36:1561. Naik S, Santangini H, Jauregui HO. 1990. Culture of adults rabbit hepatoctyes in perfused hollow membranes. In Vitro Cell Dev Biol 26:107. Neuzil DF, Rozga J, Moscioni AD, Ro MS, hakim R, Arnaout WS, Demetriou AA. 1993. Use of a novel bioartificial liver in a patient with acute liver insufficiency. Surgery 113:340. Rozga J, Williams F, Ro M-S, et al. 1993. Development of a bioartificial liver: Properties and function of a hollow-fiber module inoculated with liver cells. Hepatology 17:258. Saunders SJ, Bosman SCW, Terblanche J, et al. 1968. Acute hepatic coma treated by cross ciruclation with a baboon and by repeated exchange transfusions. Lancet 2:585. Seglen PO. 1976. Preparation of isolated rat liver cells. Meth Cell Biol 13:29. Singhvi R, Kumar A, Lopez GP, et al. 1994. Engineering cell shape and function. Science 264:696. Sussman NL, Chong MG, Koussayer T, et al. 1992. Reversal of fulminant hepatic failure using an extracorporeal liver assist device. Hepatology 16:60. Sussman NL, Kelly JH. 193. Extracorporeal liver assist in the treatment of fulminant hepatic failure. Blood Purif 11:170. Trey C, Burns DG, Saunders SJ. 1966. Treatment of hepatic coma by exchange blood transfusion. N Eng J Med 274:473. Wolf CFW, Munkelt BE. 1975. Bilirubin conjugation by an artificial liver composed of cultured cells and synthetic capillaries. Trans Amer Soc Artif Internal Organs 21:16.


Further Information Two books of particular interest in hepatic encephalopathy are Hepatic Encephalopathy: Pathophysiology and Treatment edited by Roger F. Butterworth and Gilles Pomier Layrargues, published by The Humana Press (1989), and Hepatology: A Textbook of Liver Disease edited by David Zakim and Thomas D. Boyer, published by W. B. Saunders Company (1990). Articles pertaining to extracorporeal liver assist devices appear periodically in the following journals: ASAIO Journal, the official journal for the American Society for Artificial Internal Organs (for subscription information, write to J.B. Lippincott, P.O. Box 1600, Hagerstown, MD 21741-9932); Artificial Organs, official journal of the International Society for Artificial Organs (for subscription information, write to Blackwell Scientific Publications, Inc., 328 Main Street, Cambridge, MA 02142); and Cell Transplantation, official journal of the Cell Transplantation Society (for subscription information, write to Elsevier Science, P.O. Box 64245, Baltimore, MD, 21264-4245).


Galletti, P. M., Colton, C. K., Jaffrin, M., Reach, G. “Artificial Pancreas.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

134 Artificial Pancreas

Pierre M. Galletti (deceased) Clark K. Colton Massachusetts Institute of Technology

134.1 134.2 134.3 134.4 134.5 134.6 134.7

Syringes and Pens • Reservoirs, Depots, and Sprays • Insulin Pumps • Servo-Controlled Insulin Delivery Systems

Michel Jaffrin Université de Technolgie de Compiègne

134.8

134.1

Insulin Production Systems Pancreas Transplantation • Human Islet Transplantation • Xenogeneic Islet Transplantation

Gerard Reach Hôpital Hotel-Dieu Paris

Structure and Function of the Pancreas Endocrine Pancreas and Insulin Secretion Diabetes Insulin Insulin Therapy Therapeutic Options in Diabetes Insulin Administration Systems

134.9

Outlook

Structure and Function of the Pancreas

The pancreas is a slender, soft, lobulated gland (ca. 75 g in the adult human), located transversally in the upper abdomen in the space framed by the three portions of the duodenum and the spleen. Most of the pancreas is an exocrine gland which secretes proteolytic and lipolytic enzymes, conveying more than 1 L per day of digestive juices to the gastrointestinal tract. This liquid originates in cell clusters called acini and is collected by a system of microscopic ducts, which coalesce in a channel (the canal of Wirsung) which courses horizontally through the length of the organ and opens into the duodenum next to, or together with, the main hepatic duct (choledocous). Interspersed throughout the exocrine tissue are about 1 million separate, highly vascularized and individually innervated cell clusters known as islets of Langerhans, which together constitute the endocrine pancreas (1–2% of the total pancreatic mass). Blood is supplied to the islets by the pancreatic artery and drained into the portal vein. Therefore the entire output of pancreatic hormones is first delivered to the liver. Human islets average 150 µm in diameter, each cluster including several endocrine cell types. Alpha cells (about 15% of the islet mass) are typically located at the periphery of the cluster. They are the source of glucagon, a fuel-mobilizing (catabolic) hormone which induces the liver to release into the circulation energy-rich solutes such as glucose and aceto-acetic and β-hydroxybutyric acids. Beta cells, which occupy the central portion of the islet, comprise about 80% of its mass. They secrete insulin, a fuel-storing (anabolic) hormone that promotes sequestration of carbohydrate, protein, and fat in storage depots in the liver, muscle, and adipose tissue. Delta cells, interposed between alpha and beta cells, produce somatostatin, one role of which seems to be to slow down the secretion of insulin and digestive juices, thereby prolonging the absorption of food. The PP cells secrete a “pancreatic polypeptide” of yet unknown significance. Other cells have the potential to produce gastrin, a hormone that stimulates the production of gastric juice.


134.2

Endocrine Pancreas and Insulin Secretion

The term artificial pancreas is used exclusively for systems aimed at replacing the endocrine function of that organ. Although the total loss of exocrine function (following removal of the gland for polycystic disease, tumor, or trauma) can be quite debilitating, no device has yet been designed to replace the digestive component of the pancreas. Since insulin deficiency is the life-threatening consequence of the loss of endocrine function, the artificial pancreas focuses almost exclusively on insulin supply systems, even though some of the approaches used, such as islet transplantation, of necessity include an undefined element of delivery of other hormones. Insulin is the most critical hormone produced by the pancreas because, in contradistinction to glucagon, it acts alone in producing its effects, and survival is not possible in its absence. Endogenous insulin secretion is greatest immediately after eating and is lowest in the interval between meals. Coordination of insulin secretion with fluctuating demands for energy production results from beta-cell stimulation by metabolites, other hormones, and neural signals. The beta cells monitor circulating solutes (in humans, primarily blood glucose) and release insulin in proportion to needs, with the result that in normal individuals, blood glucose levels fluctuate within quite narrow limits. The response time of pancreatic islets to an increase in blood glucose is remarkably fast: 10 to 20 sec for the initial burst of insulin release (primarily from intracellular stores), which is then followed, in a diphasic manner, by a gradual increase in secretion of newly synthesized hormone up to the level appropriate to the intensity and duration of the stimulus. Under fasting condition the pancreas secretes about 40µg (1 unit) of insulin per hour to achieve a concentration of insulin of 2–4 µg/ml (50 to 100µU/ml) in portal blood and of 0.5 µg/ml (12µU/ml) in the peripheral circulation.

134.3

Diabetes

Insulin deficiency leads to a disease called diabetes mellitus, which is the most common endocrine disease in advanced societies, affecting as many as 3–5% of the population in the United States, with an even higher incidence in specific ethnic groups. Diabetes is a chronic systemic disease resulting from a disruption of fuel metabolism either because the body does not produce enough insulin or because the available insulin is not effective. In either case, the result is an accumulation of glucose in the blood or hyperglycemia, and, once the renal tubular threshold for glucose is exceeded, a spillover of glucose into the urine or “glycosuria.” Hyperglycemia is thought to be the main determinant of microvascular alterations which affect several organs (renal glomerulus, retina, myocardium). It is considered as an important factor in large-blood-vessel pathology (aorta and peripheral arteries, such as carotid and lower limb vessels) often observed in diabetes. Neuropathies of the autonomic, peripheral, and central nervous systems are also common in diabetics, although the pathogenic mechanism is not known. When blood glucose levels are abnormally high , an increasing fraction of hemoglobin in the red blood cells tends to conjugate with glucose forming a compound identifiable by chromatography and called HbAIC. This reaction serves as a tool in diabetes management and control. The fraction of hemoglobin that is normally glycosylated (relative to total Hb) is between 6% and 9%. If elevated, HbAIC reflects the timeaveraged level of hyperglycemia over the lifetime of the red blood cells (3–4 months). It can be interpreted as an index of the severity of diabetes or the quality of control by dietary measures and insulin therapy. There are two main forms of diabetes: Type I, juvenile onset diabetes (typically diagnosed before age 30), or insulin-dependent diabetes; and type II, maturity onset diabetes (typically observed after age 40), or non-insulin-dependent diabetes. Type I appears abruptly (in a matter of days or weeks) in children and young people, although there is evidence that the islet destruction pattern may start much earlier but remain undiagnosed until close to 90% of the islets have been rendered ineffective. Type I represents only 10% of all cases of diabetes, yet it still affects close to 1 million patients in the United States alone. Endogenous insulin is almost totally absent, and an exogenous supply is therefore required immediately to avoid life-threatening metabolic © 2000 by CRC Press LLC

accidents (ketoacidosis) as well as the more insidious TABLE 134.1 Pathogenesis of Diabetes degenerative processes which affect the cardiovascular Type I Diabetes Type II Diabetes system, the kidneys, the peripheral nervous system Autoimmune/genetic/ Unknown and the retina. For some unknown reason, 15% of environmental pathologic/genetic juvenile diabetics never develop complications. factors factors Because the cellular and molecular mechanisms of | | hyperglycemic damage appear experi-mentally to be Destruction of beta cells Insulin resistance reversible by tight control of glucose levels, and the | progress of vascular complications is said to be proAbsolute insulin deficiency portional to the square of the blood glucose concentration, early and rigorous control of insulin | administration is believed to be essential to minimize Inability to maintain Inability to maintain normoglycemia normoglycemia the long term consequences of the disease. In contrast, type II diabetes, which affects close to | | 90% of all patients, develops slowly and can often be Secondary lesions Secondary lesions controlled by diet alone or by a combination of weight loss and/or oral hypoglycemic drugs. Endogenous insulin is often present in normal and sometimes exaggerated amounts, reflecting an inability of cells to make use of available insulin rather than a true hormonal deficiency. Only 20–30% of type II diabetics benefit from insulin therapy. Keto-acidosis is rare, and the major problems are those associated with vascular wall lesions and arterial obstruction. Adult onset diabetics develop the same vascular, renal, ocular, and neural complications as type I diabetics, suggesting that even if the origin of the disease is not identical, its later evolution is fundamentally the same as in juvenile diabetics (Table 134.1). Diabetes is the leading cause of blindness in the United States. It is also responsible for the largest fraction of the population of patients with end-stage renal disease who are being keep alive by maintenance dialysis. It is one of the most frequent causes of myocardial infraction and stroke, and the most common factor in arterial occlusion and gangrene of the lower extremities leading to limb amputation. With the diabetic population increasing by about 5% per year, diabetes has become one of the major contributors to health expenditures in advanced societies. It is currently thought that destruction of the pancreatic islets in juvenile diabetes is the result of an autoimmune process, occurring in genetically predisposed individuals, perhaps in relation to an intercurrent infectious disease. This has led to clinical attempts to salvage the remaining intact islets at the earliest recognizable stage of the disease using standard immunosuppressive drugs such as cyclosporin A. The effectiveness of this approach has not been demonstrated. Diabetologists have also recognized that, at the onset of juvenile diabetes, hyperglycemia exerts a deleterious effect on the function of the still surviving islets, and that in the short term, intensive exogenous insulin therapy may actually lead to an apparent cure of the disease. However, this phenomenon, occasionally referred to as the honeymoon period of insulin treatment is transitory. After a period of weeks to months, diabetes reappears, and an exogenous source of insulin administration becomes necessary for survival. The need for insulin to protect borderline functioning islets against hyperglycemic damage has also been recognized in the early stage of islet transplantation.

134.4

Insulin

Insulin is the mainstay for the treatment of diabetes type I and, to a lesser extent, type II. Since it is the key substance produced by beta cells, it is important to define its chemical nature and its mode of action. Insulin is a 51 amino acid peptide with a molecular weight of about 5800 daltons, made up of two chains (A and B) connected by disulfide bridges. Insulin is formed in the beta cell as a cleavage product of a larger peptide, proinsulin, and is stored there as crystals in tiny intracellular vesicles until released into the blood. The other cleavage product, the “connecting” or C-peptide, though biologically inactive,


passes into the blood stream together with insulin, and therefore its concentration reflects, in a patient receiving exogenous insulin, the amount of insulin secreted by the pancreas itself, if any. Active extracts of islet tissue, obtained from animals following ligation of the pancreation duct to avoid autolysis, were first prepared by Banting and Best in 1922. Insulin was crystallized by Abel in 1926. In 1960, Sanger established the sequence of amino acids that make up the molecule shortly thereafter the hormone was synthesized. The extract composition of insulin varies slightly among animal species. The presence of disulfide bonds between the A and the B chain is critical. Loss of the C-terminal alanine of the B chain by carboxypeptidase hydrolysis results in no loss of biologic activity. The octapeptide residue that remains after splitting off the last eight amino acids of the B chain has a biologic activity amounting to about 15% of the original insulin molecule. About one-half of the insulin disappears in its passage through the liver. Only a small fraction normally goes to peripheral tissues. Plasma half-life is on the order of 40 min, most of the degradation occurring in the liver and kidneys. No biologically active insulin is eliminated in the urine. Since insulin is a polypeptide, it cannot be taken orally because it would be digested and inactivated in the gastrointestinal tract. In an emergency, insulin can be administered intravenously, but standard treatment relies on subcutaneous administration. This port of entry differs from physiologic secretion of insulin in two major ways: The adsorption is slow and does not mimic the rapid rise and decline of secretion in response to food ingestion, and the insulin diffuses into the systemic veins rather than into the portal circulation.

134.5

Insulin Therapy

Preparations of insulin are classified according to their species of origin and their duration of action. Human insulin (synthesized by recombinant DNA techniques) and porcine insulin (which is obtained by chemical extraction from slaughterhouse-retrieved organs and differs from human insulin by only one amino-acid at the carboxyl-terminus of the B chain) are in principle preferable to bovine insulin (which differs from the human form by three amino acid residues). In practice, all three are equipotent, and all three can, in a minority of patients, stimulate an immune response and cause hypersensitivity reactions. Because of insulin’s relatively short duration of action, formulations delaying the absorption of subcutaneously injected hormone and hence prolonging its effectiveness have greatly facilitated the treatment of diabetes. The mode of administration of insulin influences its plasma concentration and bioavailability. The usual treatment schemes rely on the use of one or more types of insulin (Table 134.2). Pharmacologists classify the available insulin formulations according to their latency and duration of action as fast acting, intermediate-acting, and slow acting using a terminology of regular, semilente, lente, and ultralente (Table 134.3). Crystalline insulin is prepared by precipitating the polypeptide in the presence of zinc in a suitable buffer solution. Insulin complexed with a strongly basic protein, protamine, and stabilized with zinc is relatively insoluble at physiological pH and is therefore released only slowly from the site of injection. However, one must keep in mind that onset and duration of action may be quite variable from one patient to another. Whereas insulin was hailed as a life-saving drug in the years following its discovery (which it was, as evidenced by the much reduced incidence of hyperglycemic coma and the much longer life expectancy of juvenile diabetics, as compared to the period before 1930), it has not turned out to be quite the universal panacea it was expected to be. Some of the problems that have surfaced relate to the limitations associated with the mode of administration of the drug. Other problems are thought to derive from our inability to mimic the finely tuned feedback system which normally maintains the blood glucose levels within very narrow limits (Table 134.4).


TABLE 134.2

Types of Insulin

Fast acting Intermediate Slow acting

Multiple injections (basal and preprandial) Two injections (morning and evening) One injection (to prevent hyperglycemia at night)

TABLE 134.3

Properties of Insulin Preparations

Type

Appearance

Rapid Crystalline (regular) Insulin-zinc suspension (semilente) Intermediate NPH (Isophane insulin suspension) Lente (insulin zinc suspension) Slow Ultalente (extended insulin zinc suspension) PZI (Protamine zinc insulin suspension)

Added Protein

Time of Action (h)

Zinc Content mg/100 U

Onset

Peak

Duration

Clear Cloudy

None None

0.01–0.04 0.2–0.25

0.3–0.7 0.5–1.0

2–4 2–8

5–8 12–16

Cloudy Cloudy

Protamine None

0.016–0.04 0.2–0.25

1–2 1–2

6–12 6–12

18–24 18–24

Cloudy Cloudy

None Protamine

0.2–0.25 0.2–0.25

4–6 4–6

16–18 14–20

20–36 24–36

Source: Modified from Table 71-1 in The Pharmacological Basis of Therapeutics, 3rd ed., Goodman and Gillman

TABLE 134.4

Problems with Insulin Treatment

Problem

Answers

Poor compliance with multiple injections

Inadequate pharmacokinetics of insulin preparations Hyperinsulinemia Lack of feedback control

134.6

Better syringes and needles Subcutaneous ports Intravenous ports Insulin pens Enhance immediate effect Prolong duration of effect Release insulin in portal circulation Servo-controlled administration Natural control by transplant Bioartificial pancreas

Therapeutic Options in Diabetes

Experimental studies of blood glucose regulation and empirical observations in diabetic patients have revealed three major characteristics of blood glucose regulation by the pancreas: 1. The natural system operates as a closed-loop regulatory mechanism within narrow limits. 2. Portal administration is more effective than systemic administration because insulin reaches the liver first. 3. Pulsatile administration of insulin is more effective than continuous administration. There are several stages in the evolution of diabetes where a medical intervention might be helpful (Table 134.5). Many interventions have been attempted with various degrees of success. Since diabetes TABLE 134.5

Potential Approaches to Diabetes Treatment

Prevent destruction of beta cells Prevent beta cell exhaustion Increase insulin output Amplify glucose signal Overcome insulin resistance Replace beta cells by: Insulin administration Electromechanical delivery system Whole organ transplantation Islet transplantation Biohybrid device


TABLE 134.6

Biologic and Engineered Insulin Delivery Systems

Insulin Administration Systems Standard Insulin

Insulin release systems

Insulin delivery pumps Open loop

Closed loop

Routes of administration (subcutaneous, intravenous, intraperitoneal, nasal spray) Injection systems (syringes, pens) Passive release from depot forms Bioresponsive insulin depot Implanted, permeable reservoirs Programmed release systems Osmotic pumps Piston and syringe pumps Peristaltic (roller) pumps Bellow frame pumps Pressurized reservoir pumps Glucose sensors Electromechanical, servo-controlled delivery systems

Insulin Synthesis Systems Pancreas transplantation

Islet transplantation

Encapsulated islets or beta cell transplants Genetically engineered cell transplants

Simultaneous with kidney transplant After kidney transplant Before kidney transplant Autologous tissue Allogeneic tissue Xenogeneic tissue Microencapsulated cells Macroencapsulated cells Unprotected gene therapy Protected gene therapy

expresses a disturbance of biologic feedback mechanism between blood glucose levels and insulin secretion, there might be possibilities to influence the biologic regulation process through its natural sensing or amplifying mechanisms. However, no practical solution has yet emerged from this approach. As a result the accent has been placed primarily on pharmacologic forms of treatment (diverse insulin formulations and routes of administration), engineered delivery systems (extracorporeal or implanted pumps), and substitution by natural insulin production sources (organ, islet, or individual beta cell transplantation, with or without genetic manipulation and with or without membrane immunoprotection). An overview of the recently investigated approaches is given in Table 134.6.

134.7

Insulin Administration Systems

Syringes and Pens Insulin traditionally has been administered subcutaneously by means of a syringe and needle, from a vial containing insulin at a concentration of 40 units per ml. When insulin was first proposed to treat diabetic patients in 1922, glass syringes were used. The burden linked to repeated sterilization disappeared when disposable syringes became available in the early 1970s. Yet patients still had to perform the boring and, for some, difficult task of refilling the syringe from a vial. From a pragmatic viewpoint the development of pens in which insulin is stored in a cartridge represented a major advance in diabetes management. In France, for instance, a recent survey indicates that more than half of the patients use a pen for daily therapy. A needle is screwed on the cartridge and should ideally be replaced before each injection. Cartridges containing either 1.5 or 3 ml of 100U/ml regular or intermediate insulin are available. Unfortunately, long-acting insulin cannot be used in pens


because it crystallizes at high concentrations. Attempts are being made toward developing soluble, slowactin analogs of insulin to overcome this problem. A common insulin regimen consists of injecting regular insulin before each meal with a pen and longacting insulin at bedtime with a syringe. Typically half of the daily dosage is provided in the form of regular insulin. Since patients often need 50 units of insulin per day, a pen cartridge of 150 units of regular insulin must be replaced every 6 days or so. The setting of the appropriate dose is easy. However, most pens cannot deliver more than 36 units of insulin at a time, which in some patients may place a limit on their use. In those countries where insulin comes in vials of 40 U/ml for use with syringes, patients must be aware that insulin from cartridges is 2.5 times more concentrated than standard vial insulin. The major advantage of insulin pens is that they do not require refilling and therefore can be used discretely. Pens are particularly well accepted by teenagers.

Reservoirs, Depots, and Sprays Attempts have been made to develop distensible, implantable reservoirs or bags made of silicone elastomers, fitted with a small delivery catheter, and refillable transcutaneously. In the preferred embodiment, insulin drains at a constant rate into the peritoneal cavity. Assuming uptake by the capillary beds in the serosal membranes which line the gastrointestinal tract, insulin reaches first the portal valve and hence the liver which enhances its effectiveness. This is an attractive concept but not without serious handicaps. The implanted reservoir may elicit an untoward tissue reaction or become a site of infection. The delivery catheter may be plugged by insulin crystals if a high insulin concentration is used, or the catheter may be obstructed by biologic deposits at the tip. The system only provides a baseline insulin delivery and needs to be supplemented by preprandial injections. Its overall capacity constitutes a potential risk should the reservoir accidentally rupture and flood the organism with an overdose of insulin. Insulin depots, in the form of bioerodible polymer structures in which insulin—amorphous or in crystalline form—is entrapped and slowly released as hydrolytic decay of the carrier of the polymer progressively liberates it, present some of the same problems as reservoirs. In addition, the initial burst which often precedes zero-order release can prove unpredictable and dangerous. No reliable long-term delivery system has yet been developed on that principle. New routes of introduction of insulin are being investigated, in particular, transmucosal adsorption. Nasal sprays of specially formulated insulin may become a practical modality of insulin therapy if means are found to control reliably the dose administered.

Insulin Pumps Externally carried portable pumps, based on the motorized syringe or miniature roller pump designs, were evaluated in the 1970s with the anticipation that they could be both preprogrammed for baseline delivery and overridden for bolus injections at the time of meals without the need for repeated needle punctures. In actual practice, this system did not find wide acceptance on the part of patients, because it was just too cumbersome and socially unacceptable. The first implantable insulin pump was evaluated clinically in 1980. Unlike heparin delivery or cancer chemotherapy, insulin administration requires programmable pumps with adjustable flow rates. Implantable pumps provide better comfort for the patient than portable pumps, since the former are relatively unobtrusive and involve no danger of infection at the skin catheter junction. They also permit intraperitoneal insulin delivery, which is more efficient than subcutaneous administration. However, these pumps operate without feedback control, since implantable glucose sensors are not yet available, and the patient must program the pump according to his/her needs. The vapor pressure driven Infusaid pump relies on a remarkably astute mechanism for an implantable device. A rigid box is separated in two compartments by a metal bellow with a flat diaphragm and an accordion-pleated seal (bellow frame). One compartment is accessible from the outside through a subcutaneously buried filling port and contains a concentrated insulin solution. The other compartment


(the bellows itself) is filled with a liquid freon which vaporizes at 37ºC with a constant pressure of 0.6 bar. With no other source of energy but body heat (which is continuously replenished by metabolism and blood circulation), the freon slowly evaporates, and the pressure developed displaces the diaphragm and forces insulin at a constant flow rate through a narrow delivery catheter. The freon energy is restored each time the insulin reservoir is refilled, since the pressure needed to move the bellow liquefies the freon vapor within a smaller compartment. The housing is a disc-shaped titanium box. The insulin reservoir contains 15 to 25 ml of insulin stabilized with polygenol at a concentration of either 100 or 400 U/ml, providing an autonomy of 1 to 3 months. The self-sealing septum on the filling port can be punctured through the skin up to 500 times. The pump is implanted under general or local anesthesia between skin and muscle in the lateral abdominal area. The tip of the catheter can be located subcutaneously or intraperitoneally. The original device provided a constant flow of insulin which still needed to be supplemented by the patient at meal time. The most recent model (Infusaid 1000) weighs 272 g, contains 22 ml of insulin, and has a diameter of 9 cm and a height of 2.2 cm. It is designed for 100 U/ml insulin and is equipped with a side port which allows flushing of the catheter when needed. To dissipate the pressure generated by freon evaporation before it reaches the catheter, the insulin must cross a 0.22µm bacteriologic filter and pass through a steel capillary 3 cm long and 50µm in diameter. Basal flow rate can be adjusted from 0.001–0.5 ml/h, and a bolus can be superposed by releasing a precise amount of insulin stored in a pressurized accumulator through the opening of a valve leading to the catheter. These control features require an additional source of energy in the form of lithium thionile batteries with a service life of about 3 years. Control is achieved with a handheld electronic module or programs connected to the pump by telemetry. The current Minimed pump (INIP 2001) has a dry weight of 162 g and a diameter of 8.1 cm and contains 14 ml of insulin at a concentration of 400 U/ml, allowing up to 3 months’ autonomy between refills. This pump relies on a reservoir from which insulin can be delivered in a pulsatile form by a piston mechanism under the control of solenoid-driven valves. The basal rate can be adjusted 0.13–30 U/h (0.0003 to 0.07 ml/h), and boluses from 1 to 32 U over intervals 1–60 min can be programmed. The reservoir is permanently under negative pressure to facilitate filling. The Siemens Infusor (1D3 model) replaced an earlier type based on a peristaltic miniature roller pump (discontinued in 1987). It is somewhat similar in design and dimensions to the minimed pumps, but it has now been withdrawn from the market. The Medtronic Synchromed pump, which was originally designed for drug therapy and relied on a roller pump controlled by an external programmer, was evaluated for insulin therapy but is seldom used for that purpose because it lacks the required flexibility. Between 1989 and 1992, 292 insulin pumps were implanted in 259 patients in France, where most of the clinical experience has been collected (205 Minimed, 47 Infusaid, and 7 Siemens). The treatment had to be permanently discontinued in 14 patients (10 because of poor tissue tolerance, 3 due to catheter obstruction, and 1 due to pump failure). The pump was replaced in 33 patients, with 28 cases of component failure; battery, microelectronic control, flow rate decline due to insulin precipitation in Infusaid pumps, insulin reflux in Minimed pumps, and perceived risk of leak from septum. The overall frequency of technical problems was 0.10 per patient-year. There were 46 surgical interventions without pump replacement because of catheter obstructions, and 24 related to poor tissue tolerance at the site of implantation. Glucose regulation was satisfactory in all cases, and the mean blood glucose concentration dropped appreciably after 6 and 18 months of treatment. A major gain with the intraperitoneal delivery was the reduction in the incidence of severe hypoglycemia associated with improved metabolic control. For the sake of comparison, other methods used for intensive insulin therapy showed a threefold increase in the frequency of serious hypoglycemic episodes.

Servo-Controlled Insulin Delivery Systems Sensor-actuator couples which allow insulin delivery in the amount needed to maintain a nearly constant blood glucose level are sometimes designated as the artificial beta cell because of the analogy between


TABLE 134.7

Component Mechanisms of Insulin Delivery Systems

Sensor of glucose level Energy source Logic Insulin reserve Delivery system Set point Fail-safe

Natural β cell

Artificial β cell

Intracellular glycolysis Mitochondria Nucleus Insulin granules Cell membrane Normally 80–120 mg% Glucagon and hunger

Polarographic or optoelectronic sensor Implantable battery Minicomputer chip Insulin solution reservoir Insulin pump Adaptable Glucose infusion

the components of the electromechanical system and the biologic mechanisms involved in sensing, controlling, and delivering insulin from the pancreas (Table 134.7). From a technology viewpoint, the key aspects are the glucose sensor, the control systems applicable to a biologic situation where both hyperglycemia and hypoglycemia must be prevented, and the practicality of the overall system. Glucose Sensors A glucose sensor provides a continuous reading of glucose concentration. The device consists of a detection part, which determines for the specificity of the glucose measurement, and a transducing element, which transforms the chemical or physical signals associated with glucose recognition into an electric signal. The most advanced technology is based on enzymatic and amperometric detection of glucose. Glucose is recognized by a specific enzyme, usually glucose oxidase, layered on the anode of an electrode, generating hydrogen peroxide which is then oxidized and detected by the current generated in the presence of a fixed potential between the anode and the cathode. Alternatively, a chemical mediator may serve as a shuttle between the enzyme and the electrode to avoid the need for a potential difference at which other substances, such as ascorbate or acetaminophen, may be oxidized and generate an interfering current. Other approaches such as the direct electrocatalytic oxidation of glucose on the surface of the electrode or the detection of glucose by a combination with a lectin have not reached the stage of clinical evaluation. Ideally a blood glucose sensor would be permanently installed in the bloodstream. Because of the difficulty of building a hemocompatible device and the inconvenience of a permanent blood access, several investigators have turned to the subcutaneous site for glucose sensing. Indeed, glucose concentration in the interstitial fluid is very close to that of blood and reflects quite directly changes associated with meals or physical activity as observed in diabetic patients. Two types of subcutaneous glucose sensors have been developed. Those with electrodes at the tip of a needle may be shielded from their environment by the inflammatory reaction in tissues contacting the electrode membrane and occasionally cause pain or discomfort. Nonetheless, reliable measurements have been obtained for up to 10 days, whereupon the sensor needed to be changed. Sensing can also be based on microdialysis through an implanted, permeable hollow fiber, providing fluid which is then circulated to a glucose electrode. This system has not yet been miniaturized and industrialized. In either case, the concept is to develop an external monitor such as a wrist watch, which could receive its analyte from the implanted catheter and/or sensor and trigger a warning when glucose concentration is abnormally high or low. Such a monitor could eventually be incorporated in a closed-loop insulin delivery system, but this objective is not yet realizable. Noninvasive technology has been suggested to monitor blood glucose, e.g., by near-infrared spectroscopy or optical rotation applied to transparent fluid media of the eye. No reliable technology has yet emerged. The Artificial Beta Cell The standard equipment for feedback-controlled insulin administration is the artificial beta cell, also referred to in the literature as the extracorporeal artificial pancreas. It consists of a system for continuous


blood sampling, a blood glucose sensor, and minicomputer which, through preestablished algorithms, drives an insulin infusion pump according to the needs of the patient. Accessories provide a minute-byminute recording of blood glucose levels, insulin delivery, and glucose delivery. In the first clinically oriented artificial beta cell (Miles Laboratories’ Bioslator), the glucose sensor was located in a flow-through chamber where blood from a patient’s vein was sucked continuously, at the rate of 2 ml/h, through a double-lumen catheter. The double lumen allowed the infusion of a minute dose of heparin at the tip of the catheter to prevent thrombosis in the glucose analyzer circuit. Glucose diffused through a semipermeable membrane covering the electrode and was oxidized by the enzyme. The hydrogen peroxide generated by this reaction crossed a second membrane, the role of which was to screen off substances such as urate and ascorbate. The resulting electric current was fed to a computer which controlled the flow rate of the insulin pump when the readings were in the hyperglycemic range and administered a glucose solution in case of hypoglycemia. Three forms of oversight of blood glucose levels and insulin needs can be achieved with the artificial beta cell: Blood glucose monitoring is the technique employed to investigate the time course of blood glucose levels in a particular physiologic situation without any feedback control from the artificial beta cell, that is, without insulin administration to correct the fluctuations of blood glucose levels. Blood glucose control is the standard feedback mode of application of the artificial beta cell, which entails an arbitrary choice of the level of blood glucose desired and provides a recording of the rate of administration and the cumulative dose of intravenous insulin needed to achieve a constant blood glucose level. Blood glucose clamp involves intravenous insulin administration at a constant rate to obtain a stable, high blood insulin level. The desired value of blood glucose level is arbitrarily selected (typically in the normal range), and the feedback-control capability of the artificial beta cell is used to measure the amount of exogenous glucose needed to “clamp” blood glucose at the desired level in the presence of a slight excess of insulin. With this technique, a decrease in the rate of glucose administration signals a decrease in insulin biologic activity. The artificial beta cell is primarily a clinical research tool, applicable also for therapy in high-risk conditions such as pregnancy in poor controlled diabetic mothers or cardiac surgery in brittle diabetics [Galletti et al., 1985], where hypothermia and a massive outpouring of adrenergic agents alter the response to insulin [Kuntschen et al., 1986]. The primary drawbacks of current servo-controlled insulin delivery systems are the instability of the glucose sensor (in terms of both risk of thrombosis and the need for intermittent recalibration), the complexity of the system (up to four pumps can be needed for continuous blood sampling, heparinization, insulin administration, and glucose administration), and the risks involved. The dose of intravenous insulin required for rapid correction of hyperglycemia may cause an overshoot, which in turn calls for rapid administration of glucose. Such fluctuations may also influence potassium uptake by cells and in extreme conditions lead to changes in potassium levels which must be recognized and corrected [Kuntschen et al., 1983]. Finally, the general cumbersomeness of the extracorporeal system, to which the patient must be tethered by sampling and infusion lines, limits its applicability. Modeling of blood glucose control by insulin has allowed the development of algorithms that are remarkably efficient in providing metabolic feedback in most clinical situations. The physiologic limits of closed-loop systems have been lucidly analyzed by Sorensen and coworkers [1982] and Kraegen [1983].

134.8

Insulin Production Systems

The exacting requirement placed on insulin dosage and timing of administration in diabetic patients, as well as the many years of safe and reliable treatments expected from the insulin delivery technology, have pointed to the advantages of implantable systems in which insulin would be synthesized as needed and made available to the organism on demand.


As already outlined in Table 134.6 four avenues have been considered and undergone chemical evaluation: whole organ transplantation, human islet and xenogeneic islet transplantation, immunisolation of normal or tumoral insulin-secreting tissue, and transplantation of cells genetically engineered to replace the functions of the beta cells.

Pancreas Transplantation Human allograft transplantation, first attempted over 20 years ago, has been slow in reaching clinical acceptance because of the difficulty of identifying healthy cadaver organs (the pancreas is also a digestive gland which undergoes autolysis soon after death) and the need to deal with the gland exocrine secretion, which serves no useful purpose in diabetes but nonetheless must be disposed of. After many ingenious attempts to plug the secretory channels with room-temperature vulcanizing biopolymers, a preferred surgical technique has been developed whereby the pancreas is implanted in the iliac fossa, its arterial supply and venous drainage vessels anastomosed to the iliac artery and vein, and the Wirsung canal implanted in the urinary bladder. Surprisingly, the bladder mucosa is not substantially damaged by pancreatic enzymes, and the exocrine secretion dries up after a while. The success rate of pancreatic tranplantation is now approaching that of heart and liver transplantation. Since it is mostly offered to immunosuppressed uremic diabetic patients who previously received or concurrently receive a kidney transplant, preventative therapy against organ rejection does not constitute an additional risk. The main limitation remains the supply of donor organs, which in the United States will probably not exceed 2000 per year, that is to say, very far from meeting the incidence of new cases of juvenile diabetes [Robertson & Sutherland, 1992]. Therefore, human organ transplantation is not a solution to the public health problems of diabetes and ensuing complications.

Human Islet Transplantation Interest in the transplantation of islets of Langerhans, isolated from exocrine, vascular, and connective tissue by enzymatic digestion and cell separation techniques, has received a boost following the clinical demonstration that it can lead to insulin independence in type I diabetic patients [Scharp et al., 1990]. However, this “proof of principle” has not been followed by widespread application. Not only is the overall supply of cadaveric organs grossly insufficient in relating to the need (by as much as two orders of magnitude), but the islet separation techniques are complex and incompletely standardized. Therefore, it is difficult to justify at this point cutting into the limited supply of human pancreata for whole organ transplantation, the more so that whole organ replacement has become increasingly successful. Autologous islet transplantation can be successfully performed in relatively uncommon cases of pancreatic and polyscystic disease, where the risk of spilling pancreatic juice in the peritoneal cavity necessitates the removal of the entire organ. Allogeneic islet transplantation under the cover of pharmacologic immunosuppression is useful inasmuch as it provides a benchmark against which to evaluate islet transplantation from animal sources and the benefits of membrane immunoisolation.

Xenogeneic Islet Transplantation The term bioartificial pancreas refers to any insulin-production-glycemia-regulation system combining living pancreatic beta cells or equivalents with a synthetic polymer membrane of gel to protect the transplant against immune rejection. A wide variety of device designs, cell sources and processing techniques, implant location, and biomaterial formulation and characterization have been investigated (Table 134.8). Common to all is the belief that if transplantation of insulin-producing tissue is to serve the largest number of insulin-dependent diabetic patients possible, xenogeneic tissue sources will have to be identified. In that context, protection by a semipermeable membrane may be the most effective way to dispense with drug immunosuppression therapy [Lysaght et al., 1994]. However, a number of issues must be addressed and resolved before the bioartificial pancreas becomes a clinically acceptable treatment modality [Colton, 1994].


TABLE 134.8

Membrane-Encapsulated Cell Transplants

Technology

Location

Microencapsulation Macroencapsulation Coextrusion Vascular release Portal release

Systemic release Cell source

Cell processing

Membrane characterization

 Human islets or beta cells  fetal or adult Animal islets or beta cells  Functionally active tumor tissue Immortalized cell lines Genetically engineered cell lines Isolation and purification Culture and banking Cryopreservation Implant manufacturing Envelope mechanical stability Envelope chemical stability Diffusion and filtration kinetics Immunoportection Bioacceptance

Intravascular Perivascular Intraperitoneal Intrahepatic Intrasplenic Subcutaneous Intramuscular

Tissue Procurement In the evolution of type I diabetes, the oral glucose tolerance test does not become abnormal until about 70% of the islets are destroyed. Therefore, roughly 250,000 human islets, or about 3500 islets per kilogram of body weight, should suffice to normalize blood glucose levels. In reality, clinical observations of transplanted patients indicate a need for a considerably larger number (3500–6000 islets/kg, and perhaps more). This suggests that many processed islets are either nonviable or not functional because they have been damaged during the isolation procedure or by the hyperglycemic environment of the transplanted patient [Eizirik et al., 1992]. To procure such large numbers of islets an animal source must be identified. Porcine islets are favored because of the size of the animal, the availability of virus-free herds, and the low antigenicity of porcine insulin. Pig islet separation procedures are not yet fully standardized, and there are still considerable variations in terms of yield, viability, and insulin secretory function. Quality control and sterility control present serious challenges to industrialization. Alternative tissue sources are therefore being investigated. Insulinomas provided the first long-term demonstration of the concept of encapsulated endocrine tissue transplant [Altman et al., 1984]. Genetically engineered cell lines which can sense glucose concentration and regulate it with-in the physiologic range have been reported, but so far none has matched the secretory ability of normal pancreatic tissue or isolated islets. This approach remains nonetheless attractive for large-scale device production, even though no timetable can be formulated for successful development. Device Design Immunoisolation of allogeneic or xenogeneic islets can be achieved by two main classes of technology: microencapsulation and macroencapsulation. Microencapsulation refers to the formation of a spherical gel around each islet, cell cluster, or tissue fragment. Calcium alginate, usually surrounded with a polyanion such as poly-L-lysine to prevent biodegradation, and at times overcoated with an alginate layer to improve the biocompatibility, has been the most common approach, although other polymers may


be substituted. Suspension of microcapsules are typically introduced in the peritoneal cavity to deliver insulin to the portal circulation. Technical advances have permitted the fabrication of increasingly smaller microcapsules (order of 150–200 µm) which do not clump, degrade, or elicit too violent a tissue reaction. Control of diabetes has been achieved in mice, rats, dogs, and, on pilot basis, in humans [Soon-Siong et al., 1994]. An obstacle to human application has been the difficulty, if not the impossibility, of retrieving the very large number of miniature implants, many of which adhere to tissue. There is also a concern for the antigenic burden that may arise upon polymer degradation and liberation of islet tissue, whether living or necrotic. Macroencapsulation refers to the reliance on larger, prefabricated envelopes in which a slurry of islets or cell clusters is slowly introduced and sealed prior to implantation. The device configuration can be tubular or planar. The implantation site can be perivascular, intravascular, intraperitoneal, subcutaneous, or intravascular. Intravascular Devices In the most successful intravascular device (which might be more accurately designated as perivascular), a semipermeable membrane is formed into a wide-bore tube connected between artery and vein or between the severed ends of an artery as if it were a vascular graft. The islets are contained in a gel matrix within a compartment surrounding the tube through which blood circulates. In other embodiments, the islets are contained in semipermeable parallel hollow fibers or coiled capillaries attached to the external wall of a macroporous vascular prosthesis. In all cases, a rising glucose concentration in the blood stream leads to glucose diffusion into the islet compartment, which stimulates insulin production, raises its concentration in the gel matrix, and promotes diffusion into the blood stream. In an earlier type of perivascular device, blood is forced to circulate through the lumina of a tight bundle of hollow fibers, and the islet suspension is placed in the extracapillary space between fibers and circumscribed by a rigid shell, as was the case in the design which established the effectiveness of immunoprotected islets [Chick et al., 1987]. A truly intravascular device design has been proposed in which the islets are contained within the lumina of a bundle of hollow fibers which are plugged at both ends and placed surgically in the blood stream of a large vein or artery, in a mode reminiscent of the intravenous oxygenator (IVOX). Hemocompatibility has been the major challenge with that approach. There have also been attempts to develop extracorporeal devices with a semipermeable tubular membrane connected transcutaneously, through flexible catheters, to an artery or vein. The islets are seeded in the narrow “extravascular” space separating the membrane from the device external wall and can therefore be inspected and, in case of need, replaced. This concept has evolved progressively to widerbore (3–6 mm ID), spirally coiled tubes in a disc-shaped plastic housing. Such devices have remained patent for several months in dogs in the absence of anticoagulation and have shown the ability to correct hyperglycemia in spontaneous or experimentally induced diabetes. In order to accelerate glucose and insulin transport across synthetic membranes and shorten the reactive time lag of immuno-protective islets, Reach and Jaffrin [1984] have proposed to take advantage of a Starling-type ultrafiltration cycle made possible, in blood-perfused devices, by the arteriovenous pressure difference between device inlet and outlet. An outward-directed ultrafiltration flux in the first half of the blood conduit is balanced by a reverse readsorption flux in the second half, since the islet compartment is fluid-filled, rigid, and incompressible. An acceleration of the response time to fluctuations in blood glucose levels has been demonstrated both by modeling and experimentally. This system may also enhance the transport of oxygen and nutrients and improve the metabolic support of transplanted tissue. The primary obstacle to clinical acceptance of the intravascular bioartificial pancreas is the risk of thrombosis (including obstruction and embolism) of a device expected to function for several years. The smaller the diameter of the blood channel, and the greater the surface of polymeric material exposed to blood, the more likely are thrombotic complications. These devices also share with vascular grafts the risk of a small but definite incidence of infection of the implant and the potential sacrifice of a major blood vessel. Therefore, their clinical application is likely to remain quite limited.


Intratissular Devices Tubular membranes with diameters on the order of 0.5–2 mm have also been evaluated as islet containers for subcutaneous, intramuscular, or intracavitary implantation: The islets are contained inside the membrane envelope at a low-volume fraction in a gel matrix. A problem is that small-diameter tubes require too much length to be practical. Large-diameter tubes are mechanically fragile and often display a core of necrotic tissue because diffusion distances are too long for adequate oxygen and nutrient transport to the islets. Both systems are subject to often unpredictable foreign body reactions and the development of scar tissue which further impairs metabolic support of the transplanted tissue. However materials that elicit a minimal tissue reaction have been identified [Galletti, 1992], and further device design and evaluation is proceeding at a brisk pace [Scharp et al., 1994]. Some microporous materials display the property of inducing neovascularization at the tissue-material interface and in some cases within the voids of a macroporous polymer. This phenomenon is thought to enhance mass transport for nutrient and secretory products by bringing capillaries in closer proximity to immuno-seperated cells. Pore size, geometry, and interconnections are critical factors in vascularization [Colton, 1994]. Some encapsulated cell types also stimulate vacularization beyond the level observed with empty devices. Chambers made by laminating a 5-µm porosity expanded polytetrafluorethylene (Gore PTFE) on a 9.45-µm Biopore (Millipore) membrane have shown the most favorable results [Brauker et al., 1992]. The laminated, vascualrized membrane structure has been fabricated in a sandwichlike structure that can be accessed through a port to inject an islet suspension once the empty device has been fully integrated in the soft tissues of the host. This organoid can nonetheless be separated from adjacent tissues, exteriorized and retrieved. Polymers for Immunoisolation The polymeric materials used in bioartificial endocrine devices serve two major purposes: (1) As a scaffold and an extracellular matrix, they favor the attachment and differentiation of functional cells or cell clusters, and keep them separate from one another, which in some cases has proven critical; (2) as permselective envelopes they provide immunoisolation of the transplant from the host, while inducing a surrounding tissue reaction which will maximize the diffusional exchange of solutes between the transplant and its environment. A number of materials can be used for these purposes (Table 134.9). The matrix materials typically are gels made of natural or synthetic polyelectrolytes, with quite specific requirements in terms of viscosity, porosity, and electric charge. In some cases specific attachment or growth factors may be added. For immunoisolation, the most commonly used envelopes are prepared from polyacrylonitrile-polyvinyl chloride copolymers. These membranes display the anistropic structure typical of most ultrafiltration membranes, in which a thin retentive skin is supported on a spongy matrix [Colton, 1994]. The TABLE 134.9

Polymer Technology for Cell Transplantation

Component Scaffold

Function Synthetic or semisynthetic extracellular matrix

Physical separation of cells or cell aggregates Envelope Microencapsulation

Stabilization of cell suspension and immunoseperation

Macroencapsulation

Physical immunoseperation

Transport-promoting tissue-material interface


Polymer Collagen Gelatin Alginate Agarose Chitosan Hyaluronan Alginate Poly-L-lysine Polyacrylonitrile Polyvinyl copolymers Polysulfones Modified celluloses PTFE-Biopore laminate

shape and dimensions of the interconnecting voids and the microarchitecture of the inner and outer surfaces of the membrane are often critical characteristics for specific cell types and implant locations. Protection from Immune Rejection The central concept of immunoisolation is the placement of a semipermeable barrier between the host and transplanted tissue. It has been tacitly assumed that membranes with a nominal 50,000–100,000daltons cutoff would provide adequate protection, because they would prevent the passage of cells from the host immune system and impair the diffusion of proteins involved in the humoral component of immune rejection. This belief has been largely supported by empirical observations of membrane-encapsulated graft survival, with occasional failures rationalized as membrane defects. However, the cut-off point of synthetic semipermeable membranes is not as sharp as one may believe, and there is often a small number of pores which theoretically at least allow the transport of much larger molecules than suggested by the nominal cut-off definition. Therefore, complex issues of rate of transport, threshold concentration of critical proteins, adsorption and denaturation in contact with polymeric materials, and interactions between proteins involved in immune reactions in an environment where their relative concentrations may be far from normal may all impact in immunoprotection. The possibility of antigen release from living or necrotic cells in the sequestered environment must also be considered. The cellular and luminal mechanisms of immunoseperation by semipermeable membranes and the duration of the protection they afford against both immune rejection of the graft and sensitizations of the host call for considerably more study.

134.9

Outlook

Replacement of the endocrine functions of the pancreas presents a special challenge in substitutive medicine. The major disease under consideration, diabetes, is quite common, but a reasonably effective therapy already exists with standard insulin administration. The disease is not immediately life-threatening, and therefore optimization of treatment is predicated on the potential for reducing the long-term complications of diabetes. Therefore, complete clinical validation will require decreases of observations, not merely short-term demonstration of effectiveness in controlling blood glucose levels. Standard insulin treatment is also relatively inexpensive. Competitive therapeutic technologies will therefore be subject to a demanding cost-benefit analysis before they are widely recognized. Finally, the patient self-image will be a major factor in the acceptance of new diagnostic or treatment modalities. Already some demonstrably useful devices, such as subcutaneous glucose sensors, portable insulin pumps, and the extracorporeal artificial beta cell, have failed in the marketplace for reasons of excessive complexity, incompatibility with all activities of daily life, physicians’ skepticism, or cost. Newer technologies involving the implantation of animal tissue or genetically engineered cells will bring about a new set of concerns, whether justified or imaginary. There is perhaps no application of the artificial organ concept where human factors are so closely intertwined with the potential of science and technology as is already the case with the artificial pancreas.

Defining Terms Anabolism: The aspect of metabolism in which relatively simple building blocks are transformed into more complex substances for the purpose of storage or enhanced physiologic action. Artificial beta cell: A system for the control of blood glucose levels based on servo-controlled administration of exogenous insulin based on continuous glucose level monitoring. Artificial pancreas: A device or system designed to replace the natural organ. By convention, this term designates substitutes for the endocrine function of the pancreas and specifically glucose homeostasis though the secretion of insulin. Autolysis: Destruction of the components of a tissue, following cell death, mediated by enzymes normally present in that tissue. © 2000 by CRC Press LLC

Bioartificial pancreas: A device or implant containing insulin-producing, glycemia-regulating cells in combination with polymeric structures for mechanical support and/or immune protection. Catabolism: The aspect of metabolism in which substances in living tissues are transformed into waste products or solutes of simpler chemical composition. Endogenous: Originating in body tissues. Exogenous: Introduced in the body from external sources. Extracorporeal artificial pancreas: An apparatus including a glucose sensor, a minicomputer with appropriate algorithms, an insulin infusion pump, and a glucose infusion pump, the output of which are controlled so as to maintain a constant blood glucose level. (Synonymous with artificial beta cell.) Hyperglycemia: Abnormally high blood glucose level (in humans, above 140 mg/100 ml). Hypoglycemia: Abnormally low blood glucose level (in humans, below 50 mg/100 ml). Immunoisolation: Separation of transplanted tissue from its host by a membrane or film which prevents immune rejection of the transplant by forming a barrier against the passage of immunologically active solutes and cells. Insulinoma: A generally benign tumor of the pancreas, originating in the beta cells and functionally characterized by a secretion of insulin and the occurrence of hypoglycemic coma. Insulinoma cells are thought to have lost the feedback function and blood-glucose-regulating capacity of normal cells, or to regulate blood glucose around an abnormally low set point. Ketoacidosis: A form of metabolic acidosis encountered in diabetes mellitus, in which fat is used as a fuel instead of glucose (because of lack of insulin), leading to high concentration of metabolites such as acetoacetic acid, β hydroxybutyzic acid, and occasionally acetone in the blood and intestinal fluids. Organoid: A device—typically an implant—in which cell attachment and growth in the scaffold provided by a synthetic polymer sponge or mesh leads to a structure resembling that of a natural organ including, in many cases, revascularization.

References Altman JJ, Houlbert D, Callard P, et al. 1986. Long-term plasma glucose normalization in experimental diabetic rats using microencapsulated implants of benign human insulinomas. Diabetes 35:625. Altman JJ, Houlbert D, Chollier A, et al. 1984. Encapsulated human islet transplantation in diabetic rats. Trans Am Soc Artif Intern Organs 30:382. Brauker JH, Martinson LA, Young S, et al. 1992. Neovascularization at a membrane-tissue interface is dependent on microarchitecture. Abstracts, Fourth World Biomaterials Congress, Berlin FRG: 685. Chick WL, Like AA, Lauris V, et al. 1975. Artificial pancreas using living beta cells: effects on glucose homeostasis in diabetic rats. Science 197:780. Chick WL, Perla JJ, Lauris V, et al. 1977. Artificial pancreas using living beta cells: Effects on glucose homeostasis in diabetic rats. Science 197:780. Colton CK.1992. The engineering of xenogeneic islet transplantation by immunoisolation. Diab Nutr Metab 5:145. Colton CK. 1994. Engineering issues in islet immunoisolation. In R Lanza, W Chick (eds), Pancreatic Islet Transplantation, vol III: Immunoisolation of Pancreatic Islets, RG Landes. Colton CK, Avgoustiniatos ES. 1991. Bioengineering in development of the hybrid artificial pancreas. J Biomech Eng 113:152. Dionne KE, Colton CK, Yarmush ML. 1993. Effect of hypoxia on insulin secretion by isolated rat and canine islets of Langerhans. Diabetes 42:12. Eizirik DL, Korbutt GS, Hellerstrom C. 1992. Prolonged exposure of human pancreatic islets to high glucose concentrations in vitro impairs the β-cell function. J Clin Invest 90:1263. Galletti PM. 1992. Bioartificial organs. Artif Organs 16(1):55. Galletti PM, Altman JJ. 1984. Extracorporeal treatment of diabetes in man. Trans AM Soc Artif Intern Organs 30:675.


Galletti PM, Kuntschen FR, Hahn C. 1985. Experimental and clinical studies with servo-controlled glucose and insulin administration during cardiopulmonary bypass. Mt Sinai J Med 52:500. Jaffrin MY, Reach G, Notelet D. 1988. Analysis of ultrafiltration and mass transfer in a bioartificial pancreas. ASME J Biomech Eng 110:1. Kraegen EW. 1983. Closed loop systems: Physiological and practical considerations. In P Brunetti, KGMM Albeti, AM Albisser, et al. (eds), Artificial Systems for Insulin Delivery, New York, Raven Press. Kuntschen FR, Galletti PM, Hahn C. 1986. Glucose-insulin interactions during cardiopulmonary bypass. Hypothermia versus normothermia. J Thorac Cardiovas Surg 91:45. Kuntschen FR, Taillens C, Hahn C, et al. 1983. Technical aspects of Biostator operation during coronary artery bypass surgery under moderate hypothermia. In Artificial Systems for Insulin Delivery, pp 555–559, New York, Raven Press. Lanza RP, Butler DH, Borland KM, et al. 1991. Xenotransplantation of canine, bovine, and porcine islets in diabetic rats without immunosuppression. PNAS 88:11100. Lanza RP, Sullivan SJ, Chick WL. 1992. Islet transplantation with immunoisolation. Diabetes 41:1503. Lysaght MJ, Frydel B, Winn S, et al. 1994. Recent progress in immunoisolated cell therapy. J Cell Biochem 56:1. Mikos AG, Papadaki MG, Kouvroukogiou S, et al. 1994. Mini-review: Islet transplantation to create a bioartificial pancreas. Biotechnol Bioeng 43:673. Pfeiffer EF, Thum CI, Clemens AH. 1974. The artificial beta cell. A continuous control of blood sugar by external regulation of insulin infusion (glucose controlled insulin infusion system). Horm Metab Res 6:339. Reach G, Jaffrin MY, Desjeux J-F. 1984. A U-shaped bioartificial pancreas with rapid glucose-insulin kinetics. In vitro evaluation and kinetic modeling. Diabetes 33:752. Ricordi C. (ed). 1992. Pancreatic Islet Transplantation, Austin, Tex, RG Landes. Robertson RP, Sutherland DE. 1992. Pancreas transplantation as therapy for diabetes mellitus. Annu Rev Med 43:395. Scharp DW, Lacy PE, Santiago JV, et al. 1990. Insulin independence after islet transplantation into a Type I diabetes patient. Diabetes 39:515. Scharp DW, Swanson CJ, Olack BJ, et al. 1994. Protection of encapsulated human islets implanted without immunosuppression in patients with Type I and II diabetes and in nondiabetic controls. Diabetes (accepted for publication in September 1994 issue). Soon-Siong P, Heintz RE, Meredith N, et al. 1994. Insulin independence in a type 1 diabetic patient after encapsulated islet transplantation. Lancet 343:950. Sorensen JT, Colton CK, Hillman RS, et al. 1982. Use of physiologic pharmacokinetic model of glucose homeostasis for assessment of performance requirements for improved requirements for improved insulin therapies. Diabetes Care 5:148. Sullivan SJ, Maki T, Boreland KM, et al. 1991. Biohybrid artificial pancreas: Long-term implantation studies in diabetic, pancreatectomized dogs. Science 252:718.

Further Information A useful earlier review of insulin delivery systems is to be found in: Artificial Systems for Insulin Delivery, edited by P. Brunetti, K.G.M.M. Alberti, A.M. Albisser, K.D. Hepp, and M. Massi Benedetti, Serono Symposia Publications from Raven Press, New York, 1983. An upcoming review, focused primarily on islet transplantation will be found in Pancreatic Islet Transplantation, vol. III: Immunoisolation of Pancreatic Islets, edited by R.P. Lanza and W.L. Chick, R.G. Landes Company, Austin, Texas, 1994. A volume entitled Implantation Biology: The Host Response and Biomedical Devices, edited by R.S. Greco, from CRC Press, 1994, provides a background in the multiple aspects of tissue response to biomaterials and implants. Individual contributions to the science and technology of pancreas replacement are likely to be found in the following journals: Artificial Organs, the Journal of Cell Transplantation, and Transplantation. Reports on clinically promising devices are often published in Diabetes and Lancet.


Valentini, R. F. “ Nerve Guidance Channels.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

135 Nerve Guidance Channels 135.1 135.2 135.3 135.4 135.5 135.6 135.7

Transmural Permeability • Surface Texture or Microgeometry • Electric Charge Characteristics • Release of Soluble Factors • Inclusion of Insoluble Factors • Seeding Neuronal Support Cells

Robert F. Valentini Brown University

Peripheral Nervous System PNS Response to Injury PNS Regeneration Surgical Repair of Transected Peripheral Nerves Repair with Nerve Guidance Channels Recent Studies with Nerve Guidance Channels Enhancing Regeneration by Optimizing Nerve Guidance Channel Properties

135.8

Summary

In adult mammals, including humans, the peripheral nervous system (PNS) is capable of regeneration following injury. The PNS consists of neural structures located outside the central nervous system (CNS), which is comprised of the brain and spinal cord. Unfortunately, CNS injuries rarely show a return of function, although recent studies suggest a limited capacity for recovery under optimal conditions. Neural regeneration is complicated by the fact that neurons, unlike other cell types, are not capable of proliferating. In successful regeneration, sprouting axons from the proximal nerve stump traverse the injury site, enter the distal nerve stump, and make new connections with target organs. Current surgical techniques allow surgeons to realign nerve ends precisely when the lesion does not require excision of a large nerve segment. Nerve realignment increases the probability that extending axons will encounter an appropriate distal neural pathway, yet the incidence of recovery in the PNS is highly variable, and the return of function is never complete. Surgical advances in the area of nerve repair seem to have reached an impasse, and biologic rather than technical factors now limit the quality of regeneration and functional recovery. The use of synthetic nerve guidance channels facilitates the study of nerve regeneration in experimental studies and shows promise in improving the repair of injured human nerves. Advances in the synthesis of biocompatible polymers have provided scientists with a variety of new biomaterials which may serve as nerve guidance channels, although the material of choice for clinical application has not yet been identified. The purpose of this chapter is to review the biologic aspects of PNS regeneration (CNS regeneration will be discussed in a subsequent chapter) and the influence of nerve guidance channels on the regeneration process. The biologic mechanisms and the guidance channel characteristics regulating regeneration will be emphasized, since the rational design of guidance systems hinges on the integration of engineering (polymer chemistry, materials science, and so on) and biologic (cellular and molecular events) principles.


FIGURE 135.1 Spinal cord in cross-section. Relationship between motor and sensory cell bodies and their axons. Motor neurons (black) are located in the anterior horn of the gray matter within the spinal cord. Sensory neurons (white) are located in dorsal root ganglia just outside the spinal cord. Axons exit via dorsal (sensory) and ventral (motor) spinal roots and converge to form peripheral nerves, which connect to target structures (sensory endings and muscles). Note that the distance between axons and their corresponding cell bodies can be quite long.

135.1

Peripheral Nervous System

Peripheral nerve trunks are responsible for the innervation of the skin and skeletal muscles and contain electrically conductive fibers termed axons, whose cell bodies reside in or near the spinal cord (Fig. 135.1). Nerve cells are unique in that their cellular processes may extend for up to one meter or longer (e.g., for axons innervating the skin of the foot). Three types of neuronal fibers can be found in peripheral nerves; (1) motor fibers, whose axons originate in the anterior horn of the spinal cord and terminate in the neuromuscular ending of skeletal muscle; (2) sensory fibers, which are the peripheral projections of dorsal root ganglion neurons and terminate at the periphery either freely or in a variety of specialized receptors; and (3) sympathetic fibers, which are the postganglionic processes of neurons innervating blood vessels or glandular structures of the skin. These three fiber types usually travel within the same nerve trunk, and a single nerve can contain thousands of axons. All axons are wrapped by support cells named Schwann cells. Larger axons are surrounded by a multilamellar sheath of myelin, a phospholipidcontaining substance which serves as an insulator and enhances nerve conduction. An individual Schwann cell may ensheath several unmyelinated axons but only one myelinated axon, within its cytoplasm. Schwann cells are delineated by a fine basal lamina and are in turn surrounded by a complex structure made of collagen fibrils interspersed with fibroblasts and small capillaries, forming a tissue termed the endoneurium. A layer of flattened cells with associated basement membrane and collagen constitute the perineurium, which envelops all endoneurial constituents. The presence of tight junctions between the perineurial cells creates a diffusion barrier for the endoeurium and thus functions as blood-nerve barrier. The perineurium and its endoneurial contents constitute a fascicle, which is the basic structural unit of a peripheral nerve. Most peripheral nerves contain several fascicles, each containing numerous myelinated and unmyelinated axons. Since the fascicular tracts branch frequently and follow a tortuous pathway, the cross-sectional fascicular pattern changes significantly along a nerve trunk. Outside the perineurium is the epineurium, a protective, structural connective tissue sheath made of several layers of flattened fibroblastic cells interspersed with collagen, blood vessels, and adipocytes (fat cells). Peripheral nerves


are well vascularized, and their blood supply is derived either from capillaries located within the epineurium and endoneurium (i.e., vasa nervosum) or from peripheral vessels which penetrate into the nerve from surrounding arteries and veins.

135.2

PNS Response to Injury

Nerves subjected to mechanical, thermal, chemical, or ischemic insults exhibit a typical combination of degenerative and regenerative responses. The most severe from of injury results from complete transaction of the nerve, which interrupts communication between the nerve cell body and its target, disrupts the interrelations between neurons and their support cells, destroys the local blood-nerve barrier, and triggers a variety of cellular and humoral events. Injuries close to the nerve cell body are more detrimental than injuries occurring more peripherally. The cell bodies and axons of motor and sensory nerves react in a characteristic fashion to transection injury. The central cell body (which is located within or just outside the spinal cord) and its nucleus swell. Neurotransmitter (e.g., the chemicals which control neuronal signaling) production is diminished drastically, and carbohydrate metabolism is shifted to the pentose-phosphate cycle. These changes indicate a metabolic shift toward production of the substrates necessary for reconstituting the cell membrane and other structural components. For example, the synthesis of the protein tubulin, which is the monomeric elements of the microtubule, the structure responsible for axoplasmic transport, is increased dramatically. Following injury, tubulin and other substrates produced in the cell soma are transported, via slow and fast axoplasmic transport, to the distal nerve fiber. Immediately after injury, the tip of the proximal stump (i.e., the nerve segment closest to the spinal cord) swells to two or three times its original diameter, and the severed axons retract. After several days, the proximal axons begin to sprout vigorously and growth cones emerge. Growth cones are axoplasmic extrusions of the cut axons which flatten and spread when they encounter a solid, adhesive surface. They elicit numerous extensions (fillipodia) which extend outward in all directions until the first sprout reaches an appropriate target (i.e., Schwann cell basal lamina). The lead sprout is usually the only one to survive, and the others quickly die back. In the distal nerve stump, the process of Wallerian degeneration begins within 1 to 2 hours after transection. The isolated axons and their myelin sheaths are degraded and phagocytized by Schwann cells and macrophages, leading to a complete dissociation of neurotubules and neurofilaments. This degradation process is accompanied by a proliferation of Schwann cells which retain the structure of the endoneurial tube by organizing themselves into ordered columns termed bands of Bunger. While the Schwann cells multiply, the other components of the distal nerve stump atrophy or are degraded, resulting in a reduction in the diameter of the overall structure. Severe retrograde degeneration can lead to cell atrophy and, eventually, cell death. Peripheral nerve transection also leads to marked atrophy of the corresponding target muscle fibers.

135.3

PNS Regeneration

In successful regeneration, axons sprouting from the proximal stump bridge the injury gap and encounter an appropriate Schwann cell column (band of Bunger). In humans, axons elongate at an average rate of 1 mm per day. The bridging axons are immediately engulfed in Schwann cell cytoplasm, and some of them become myelinated. Induction of myelogenesis by the Schwann cell is thought to depend on axonal contact. Some of the axons reaching the distal stump traverse the length of the nerve to form functional synapses, but changes in the pattern of muscular and sensory reinnervation invariably occur. The misrouting of growing fibers occurs primarily at the site of injury, since once they reach the distal stump their paths do not deviate further. The original Schwann cell basal lamina in the distal stump persists and may aid in the migration of axons and proliferating Schwann cells. Axons which fail to reach an appropriate end organ or fail to make a functional synapse eventually undergo Wallerian degeneration.


Abnormal regeneration may lead to the formation of a neuroma, a dense, irregular tangle of axons which can cause painful sensations. Some nerve injuries result from clean transections of the nerve which leave the fascicular pattern intact. More often, a segment of nerve is destroyed (e.g., during high-energy trauma or avulsion injuries in which tissue is torn), thus creating a nerve deficit. The resulting nerve gap is the separation between the proximal and distal stumps of a damaged nerve that is due to elastic retraction of the nerve stumps and tissue loss. The longer the nerve gap, the less likely regeneration will occur.

135.4

Surgical Repair of Transected Peripheral Nerves

In the absence of surgical reconnection, recovery of function following a transection or gap injury to a nerve is negligible, since (1) the cell bodies die due to severe retrograde effects, (2) the separation between nerve ends precludes sprouting axons from finding the distal stump, (3) connective tissue ingrowth at the injury site acts as physical barrier to neurite elongation, and (4) the proximal and distal fascicular patterns differ so much that extending axons cannot find an appropriate distal pathway. Efforts to repair damaged peripheral nerves by surgical techniques date back many years although histologic evidence of regeneration was not reported by Ramon y Cajal until the first part of this century. Early attempts to reconnect severed peripheral nerves utilized a crude assortment of materials, and anatomic repair rarely led to an appreciable return of function. In the 1950s, the concept of end-to-end repair was refined by surgeons who directly reattached individual nerve fascicles or groups of fascicles. Further refinements occurred during the 1960s with the introduction of the surgical microscope and the availability of finer suture materials and instrumentation. Microsurgical nerve repair led to a significant improvement in the return of motor and sensory function in patients to the point that success rates of up to 70% can currently be achieved. In microsurgical nerve repair the ends of severed nerves are apposed and realigned by placing several strands of very fine suture material through the epineurial connective tissue without entering the underlying nerve fascicles. Nerve-grafting procedures are performed when nerve retraction or tissue loss prevents direct end-to-end repair. Nerve grafts are also employed when nerve stump reapproximation would create tension along the suture line, a situation which is known to hinder the regeneration process. In grafting surgery, an autologous nerve graft from the patient, such as the sural nerve (whose removal results in little functional deficit), is interposed between the ends of the damaged nerve.

135.5

Repair with Nerve Guidance Channels

In repair procedures using nerve guidance channels, the mobilized ends of a severed nerve are introduced in the lumen of a tube and anchored in place with sutures (Fig 135.2). Tubulation repair provides: (1) a direct, unbroken path between nerve stumps; (2) prevention of scar tissue invasion into the regenerating environment; (3) directional guidance for elongation neurites and migrating cells; (4) proximal-distal stump communication without suture-line tension in cases of extensive nerve deficit; (5) minimal number of epineurial stay sutures, which are known to stimulate connective tissue proliferation; and (6) preservation, within the guidance channel lumen, of endogenous trophic or growth factors released by the traumatized nerve ends. Guidance channels are also useful from an experimental perspective: (1) the gap distance between the nerve stumps can be precisely controlled; (2) the fluid and tissue entering the channel can be evaluated; (3) the properties of the channel can be varied; and (4) the channel can be filled with various drugs, gels, and the like. Nerves with various dimensions from several mammalian species, including mice, rats, rabbits, hamsters, and nonhuman primates have been tested. The regenerated tissue in the guidance channel is evaluated morphologically to quantify the outcome of regeneration. Parameters analyzed include the cross-sectional area of the regenerated nerve cable, the number of myelinated and unmyelinated axons, and the relative percentages of cellular constituents (i.e., epineurium, endoneurium, blood vessels, and so on). Electrophysiologic and functional evaluation can also be performed in studies conducted for long periods (e.g., several weeks or more).


FIGURE 135.2

Tube to repair nerve. Surgical placement of nerve guidance channel.

The frequent occurrence of nerve injuries during the world wars of this century stimulated surgeons to seek simpler, more effective means of repairing damaged nerves. A variety of biologic and synthetic materials shaped into cylinders were investigated (Table 135.1). Bone, collagen membranes, arteries, and veins (either fresh or freeze-dried to reduce antigenicity) were used from the late 1800s through the 1950s to repair nerves. These materials did not enhance the rate of nerve regeneration when compared to regular suturing techniques, so clinical applications were infrequent. Magnesium, rubber, and gelatin tubes were evaluated during World War I, and cylinders of parchment paper and tantalum were used during World War II. The poor results achieved with these materials can be attributed to poor biocompatibility, since the channels elicited an intense tissue response which limited the ability of growing axons to reach the distal nerve stump. Following World War II, polymeric materials with more stable mechanical and chemical properties became available. Millipore (cellulose acetate) and Silastic (silicone elastomer) tubing received the greatest attention. Millipore, a filter material with a maximum pore size of 0.45µm, showed early favorable results. In human trials, however, Millipore induced calcification and eventually fragmented several months after implantation so that its use was discontinued. Silastic tubing, a biologically inert polymer with rubberlike properties, was first tested in the 1960s. Thin-walled Silastic channels were reported to support regeneration over large gaps in several mammalian species. Thick-walled tubing was


TABLE 135.1

Materials Used for Nerve Guidance Channels

Synthetic materials Nonresorbable Nonporous Ethylene-Vinyl Acetate Copolymer (EVA) Polytetrafluoroethylene (PTFE) Polyethylene (PE) Silicone elastomers (SE) Polyvinyl chloride (PVC) Polyvinylidene fluoride (PVDF) Microporous Gortex, expanded polytetrafluoroethylene (ePTFE) Millipore (cellulose filter) Semipermeable Polyacrylonitrile (PAN) Polyacrylonitrile/Polyvinyl chloride (PAN/PVC) Polysulfone (PS) Bioresorbable Polyglycolide (PGA) Polylactide (PLLA) PGA/PLLA blends Biologic materials Artery Collagen Hyaluronic acid derivatives Mesothelial tubes Vein Metals Stainless steel Tantalum

associated with nerve necrosis and neuroma production. The material showed no long-term degradation nor did it elicit a sustained inflammatory reaction. As a result, thin-walled Silastic tubing has been used, on a very limited basis, in the clinical repair of severed nerves.

135.6

Recent Studies with Nerve Guidance Channels

The availability of a variety of new biomaterials has led to a resurgence of tubulation studies designed to elucidate the mechanisms of nerve regeneration. The spatial-temporal progress of nerve regeneration across a 10-mm rat sciatic nerve gap repaired with a silicone elastomer tube has been analyzed in detail (Fig. 135.3). During the first hours following repair, the tube fills with a clear, protein-containing fluid exuded by the cut blood vessels in the nerve ends. The fluid contains the clot-forming protein, fibrin, as well as factors known to support nerve survival and outgrowth. By the end of the first week, the lumen is filled with a longitudinally oriented fibrin matrix which coalesces and undergoes syneresis to form a continuous bridge between the nerve ends. The fibrin matrix is soon invaded by cellular elements migrating from the proximal and distal nerve stumps, including fibroblasts (which first organize along the periphery of the fibrin matrix), Schwann cells, macrophages, and endothelial cells (which form capillaries). At 2 weeks, axons advancing from the proximal stump are engulfed in the cytoplasm of Schwann cells. After 4 weeks some axons have reached the distal nerve stump, and many have become myelinated. The number of axons reaching the distal stump is related to the distance the regenerating nerve has to traverse and the length of original nerve resected. Silicone guidance channels do not support regeneration if the nerve gap is greater than 10 mm and if the distal nerve stump is left out of the guidance channel. The morphology and structure of the regenerated nerve is far from normal. The size and number of axons and the thickness of myelin sheaths are less than normal. Electroymyographic evaluation of


FIGURE 135.3

Regeneration process. Nerve regeneration through a guidance channel.

nerves regenerated through silicone tubes reveals that axons can make functional synapses with distal targets, although nerve conduction velocities and signal amplitudes are slower than normal, even after many months. Attempts to improve the success rate and quality of nerve regeneration have led to the use of other tubular biomaterials as guidance channels (Table 135.1). biodurable materials such as acrylic copolymers, polyethylene, and porous stainless steel and bioresorbable materials including polyglycolides, polylactides, and other polyesters have investigated. There is some concern that biodurable materials may cause long-term complications via compression injury to nerves or soft tissues. Biodegradable materials offer the advantage of disappearing once regeneration is complete, but success thus far has been limited by swelling, increased scar tissue induction, and difficulty in controlling degradation rates. In all cases, these materials have displayed variable degrees of success in bridging transected nerves, and the newly formed nerves are morphologically quite different from normal peripheral nerves. The general spatialtemporal progress of nerve regeneration, however, resembles that described for the silicone channel model.


135.7

Enhancing Regeneration by Optimizing Nerve Guidance Channel Properties

Manipulating the physical, chemical, and electrical properties of guidance channels allows control over the regenerating environment and optimization of the regeneration process. The following features of synthetic nerve guidance channels have been studied: (1) transmual permeability, (2) surface texture or microgeometry, (3) electric charges, (4) release of soluble factors, (5) inclusion of insoluble factors, and (6) seeding with neuronal support cells.

Transmural Permeability The synthetic nerve guidance channel controls the regeneration process by influencing the cellular and metabolic aspects of the regenerating environment. Since transected nerves lose the integrity of their blood-nerve barrier (which controls oxygen and carbon dioxide tensions, pH, and the concentrations of nutrients and essential proteins), the guidance channel’s transmural mass transfer characteristics modulate solute exchange between the regenerating tissue and the surrounding fluids Nerves regenerated through permselective tubes display more normal morphologic characteristics than nerves regenerated in impermeable silicone elastomer (SE) and polyethylene (PE) or freely permeable expanded polytetrafluoroethylene (ePTFE) tubes. Nerves found in semipermeable tubes feature more myelinated axons and less connective tissue. Nerve cables regenerated in semipermeable or impermeable tubes are both round-shaped and free from attachment to the inner wall of the guidance channel. Nerves regenerated in highly porous, open structures do not form a distinct cable but contain connective tissue and dispersed neural elements. The range of permeselectivity is very important, and optimal regeneration is observed with a molecular weight (MW) cut-off of 50,000–100,000 daltons (D). Permeselective PAN/PVC channels with an MW cut-off of 50,000 D support regeneration even in the absence of a distal nerve stump. These observations suggest that controlled solute exchange between the internal regenerative and external wound-healing environments is essential in controlling regeneration. The availability of oxygen and other nutrients may minimize connective tissue formation in permeable PAN/PVC and PS tubes. Decreased oxygen levels and waste buildup may increase connective tissue formation in SE and PE tubes. Regeneration may also be modulated by excitatory and inhibitory factors released by the wound-healing environment. Semipermeable channels may sequester locally generated growth factors while preventing the inward flux of molecules inhibitory to regeneration.

Surface Texture or Microgeometry The microgeometry of the luminal surface of the guidance channel plays an important role in regulating tissue outgrowth. Expanded microfibrilar polytetrafluorethylene (ePTFE) tubes exhibiting different internodal distances (1, 5, and 10 µm) were compared to smooth-walled impermeable PTFE tubes. Larger internodal distances result in greater surface irregularity and increased transmural porosity. Rough-walled tubes contained isolated fascicles of nerves disperses within a loose connective tissue stroma. The greater the surface roughness, the greater the spread of fascicles. In contrast, smooth-walled, impermeable PTFE tubes contained a discrete nerve cable delineated by an epineurium and located within the center of the guidance channel. Similar results were observed with semipermeable PAN/PVC tubes with the same chemistry and MW cut-off but with either smooth or rough surfaces. Nerves regenerated in tubes containing alternating sections of smooth and rough inner walls showed similar morphologies with an immediate change from single-cable to numerous fascicle morphology at the interface of the smooth and rough segments. These studies suggest that the microgeometry of the guidance channel lumen also modulates the nerve regeneration process. Wall structure changes may alter the protein and cellular constituents of the regenerating tissue bridge. For example, the orientation of the fibrin matrix is altered in the presence of a rough inner wall. Instead of forming a single, longitudinally oriented bridge connecting the nerve ends,


the fibrin molecules remain dispersed throughout the lumen. As a result, cells migrating in from the nerve stumps loosely fill the entire lumen rather than form a dense central structure.

Electric Charge Characteristics Applied electric fields and direct dc stimulation are known to influence nerve regeneration in vitro and in vivo. Certain dielectric polymers may be used to study the effect of electric activity on nerve regeneration in vivo and in vitro. These materials are advantageous in that they provide electric charges without the need for an external power supply, can be localized anatomically, are biocompatible, and can be formed into a variety of shapes including tubes and sheets. Electrets are a broad class of dielectric polymers which can be fabricated to display surface charges because of their unique molecular structure. True electrets, such as polytetrafluorethylene (PTFE), can be modified to exhibit a static surface charge due to the presence of stable, monopolar charges located in the bulk of the polymer. The sign of the charge depends on the poling conditions. Positive, negative, or combined charge patterns can be achieved. The magnitude of surface charge density is related to the number and stability of the monopolar charges. Charge stability is related to the temperature at which poling occurs. Crystalline piezoelectric materials such as polyvinylidene fluoride (PVDF) display transient surface charges related to dynamic spatial reorientation of molecular dipoles located in the polymer bulk. The amplitude of charge generation depends on the degree of physical deformation (i.e., dipole displacement) of the polymer structure. The sign of the charge is dependent on the direction of deformation, and the materials show no net charge at rest. Negatively and positively poled PVDF and PTFE tubes have been implanted as nerve guidance channels. Poled PVDF and PTFE channels contain significantly more myelinated axons than unpoled, but otherwise identical, channels. In general, positively poled channels contained larger neural cables with more myelinated axons than negatively poled tubes. It is not clear how static or transient charge generation affects the regeneration process. The enhancement of regeneration may be due to electrical influences on protein synthesis, membrane receptor mobility, growth core mobility, cell migration, and other factors.

Release of Soluble Factors The release of soluble agents, including growth factors and other bioactive substances from synthetic guidance channels, may improve the degree and specificity of neural outgrowth. Using single or multiple injections of growth factors has disadvantages including early burst release, poor control over local drug levels, and degradation in biologic environments. Guidance channels can be prefilled with drugs or growth factors, but the aforementioned limitations persist. Advantages of using a local, controlled delivery system are that the rate and amount of factor release can be controlled and that the delivery can be maintained for long periods (several weeks). Channels composed of an ethylene-vinyl acetate (EVA) copolymer have been fabricated and designed to release incorporated macromolecules in a predictable manner. The amount of drug loaded, its molecular weight, and geometry of the drug-releasing structure affect the drug release kinetics. It is also possible to restrict drug release to the luminal side of the guidance channel by coating its outer wall with a film of pure polymer. Growth or neurontrophic factors that ensure the survival and general growth of neurons are produced by support cells (e.g., Schwann cells) or by target organs (e.g., muscle fibers). Some factors support neuronal survival, other support nerve outgrowth, and some do both. Numerous growth factors have been identified, purified, and synthesized through recombinant technologies (Table 135.2). In vivo, growth factors are found in solution in the serum or extracellular fluid or bound to extracellular matrix (ECM) molecules. Nerve guidance channels fabricated from EVA and designed to slowly release basic fibroblast growth factor (bFGF) or nerve growth factor (NGF) support regeneration over a 15-mm gap in a rat model. Control EVA tubes supported regeneration over a maximum gap of only 10 mm with no regeneration in 15-mm gaps. The concurrent release of growth factors which preferentially control the survival and outgrowth of motor and sensory neurons may further enhance regeneration, since the


TABLE 135.2

Growth Factors Involved in Peripheral Nerve Regeneration

Growth Factor

Possible function

NGF—nerve growth factor BDNF—brain-derived neurotophic factor CNTF—ciliary neuronotrophic factor NT-3—neuronotrophin 3 NT-4—neuronotrophin 4 IGF-1—insulinlike growth factor-1 IGF-2—Insulinlike growth factor-2 PDGF—platelet-derived growth factor aFGF—acidic fibroblast growth factor bFGF—basic fibroblast growth factor

TABLE 135.3

Neuronal survival, axon-Schwann cell interaction Neuronal survival Neuronal survival Neuronal survival Neuronal survival Axonal growth, Schwann cell migration Motoneurite sprouting, muscle reinnervation Cell proliferation, neuronal survival Neurite regeneration, cell proliferation Neurite regeneration, neovascularization

Neuronal Attachment and Neurite-Promoting Factors

Factor Collagen Fibronectin Laminim Neural cell adhesion molecule (N-CAM) N-caherin

Minimal Peptide Sequence RGD RGD RGD, SIKVAV, YIGSR Unknown Unknown

majority of peripheral nerves contain both populations. For example, NGF and b-FGF control sensory neuronal survival and outgrowth, whereas brain-derived growth factor (BDGF) and ciliary neurontrophic factor (CNTF) control motor neuronal survival and outgrowth. Growth factors released by guidance channels may also allow regeneration over large nerve deficits, and important consideration in nerve injuries with severe tissue loss. The local release of other pharmacologic agents (e.g., anti-inflammatory drugs) may also be useful in enhancing nerve growth.

Inclusion of Insoluble Factors Several neural molecules found on cell membranes and in the extracellular matrix are potent modulators of neural attachment and outgrowth (Table 135.3). Proteins responsible for eliciting and stimulating axon elongation are termed neurite promoting factors. The glycoprotein laminin, an ECM component present in the balsa lamina of Schwann cells, has been reported to promote nerve elongation in vitro and in vivo. Other ECM products, including the glycoprotein fibronectin and the proteoglycan heparan sulphate, also have been reported to promote nerve extension in vitro. Some subtypes of the ubiquitous protein collagen also support neural attachment. Filling guidance channels with gels of laminin and collagen has been shown to improve and accelerate nerve repair. The addition of longitudinally oriented fibrin gels to SE tubes has also been shown to accelerate regeneration. Collagen, laminin, and glycosaminoglycans (another ECM component) introduced in the guidance channel lumen support some degree of regeneration over a 15–20 mm gap in adult rats. The concentration of ECM gel is important, as thicker gels impede regeneration in semipermeable tubes. The activity of these large, insoluble ECM molecules (up to 106 D MW) can be mimicked by short sequences only 3–10 amino acids long (e.g., RGD, YIGSR) (Table 135.3). The availability of small, soluble, bioactive agents allows more precise control over the chemistry, conformation, and binding of neuronspecific substances. Additionally, their stability and linear structure facilitate their use instead of labile (and more expensive ) proteins which require three-dimensional structure for activity. Two- and threedimensional substrates containing peptide mimics have been shown to promote neural attachment and regeneration in vitro and in vivo.


Seeding Neuronal Support Cells Adding neural support cells to the lumen of guidance channels is another strategy being used to improve regeneration or to make regeneration possible over otherwise irreparable gaps. For example, Schwann cells cultured in the lumen of semipermeable guidance channels have been shown to improve nerve repair in adult rodents. Cells harvested from inbred rats were first cultured in PAN/PVC tubes using various ECM gels as stabilizers. The Schwann cells and gel formed a cable at the center of the tube after several days in culture. Once implanted, the cells were in direct contact with the nerve stumps. A dose-dependent relationship between the density of seeded cells and the extent of regeneration was noted. Another approach toward nerve repair involves the use of Schwann cells, fibroblasts, and the like which are genetically engineered to secrete growth factors. The use of support cells which release neuronotrophic and neurite-promoting molecules may enable regeneration over large gaps. There is increasing evidence that PNS elements, especially Schwann cells, are capable of supporting CNS regeneration as well. For example, Schwann-cell-seeded semipermeable channels support regeneration at the level of the dorsal and ventral spinal cord roots and the optic nerve, which are CNS structures.

135.8

Summary

The permeability, textural, and electrical properties of nerve guidance channels can be optimized to impact favorably on regeneration. The release of growth factors, addition of growth substrates, and inclusion of neural support cells and genetically engineered cells also enhance regeneration through guidance channels. Current limitations in PNS repair, especially the problem of repairing long gaps, and in CNS repair, where brain and spinal cord trauma rarely result in appreciable functional return, may benefit from advances in engineering and biology. The ideal guidance system has not been identified but will most likely be a composite device that contains novel synthetic or bioderived materials and that incorporates genetically engineered cells and new products from biotechnology.

References Aebischer P, Salessiotis AN, Winn SR. 1989. Basic fibroblast growth factor released from synthetic guidance channels facilitates peripheral nerve regeneration across nerve gaps. J Neurosci Res 23:282. Dyck PJ, Thomas PK. 1993. Peripheral Neuropathy, 3rd ed, vol 1, Philadelphia, London, WB Saunders. Guenard V, Kleitman N, Morrissey TK, et al. 1992. Syngeneic Schwann cells derived from adult nerves seeded in semi permeable guidance channels enhance peripheral nerve regeneration. J Neurosci 12:3310–3320. LeBeau JM, Ellisman MH, Powell HC. 1988. Ultrastructural and morphometric analysis of long-term peripheral nerve regeneration through silicone tubes. J Neurocytol 17:161. Longo FM, Hayman EG, Davis GE, et al. 1984. Neurite-promoting factors and extracellular matrix components accumulating in vivo within nerve regeneration chambers. Exp Neurol 81:756. Lundborg G. 1987. Nerve regeneration and repair: A review. Acta Orthop Scand 58:145. Lundborg G, Dahlin LB, Danielsen N, et al. 1982. Nerve regeneration in silicone chambers: Influence of gap length and of distal stump components. Exp Neurol 76:361. Lundborg G, Longo FM, Varon S. 1982. Nerve regeneration model and trophic factors in vivo. Brain Res 232:157. Raivich G, Kreutzberg GW. 1993. Peripheral nerve regeneration: role of growth factors and their receptors. Int J Dev Neurosci 11(3):311. Sunderland S. 1991. Nerve Injuries and Their Repair, Edinburgh, London, Churchill Livingstone. Valentini RF, Aebischer P. 1991. The role of materials in designing nerve guidance channels and chronic neural interfaces. In P Dario, G Sandini, P Aebischer (eds), Robots and Biological Systems: Towards a New Bionics?, pp 625–636, Berlin, Germany, Springer-Verlag.


Valentini RF, Aebischer P, Winn SR, et al. 1987. Collagen- and laminin-containing gels impede peripheral nerve regeneration through semi-permeable nerve guidance channels. Exp Neurol 98:350. Valentini RF, Sabatini AM, Dario P, et al. 1989. Polymer electret guidance channels enhance peripheral nerve regeneration in mice. Brain Res 480:300. Weiss P. 1944. The technology of nerve regeneration: A review. Sutureless tubulation and related methods of nerve repair. J Neurosurg 1:400. Williams LR, Longo FM, Powell HC, et al. 1983. Spatial-temporal progress of peripheral nerve regeneration within a silicone chamber: Parameters for a bioassay. J Comp Neurol 218:460.


Nakamura, T., Shimizu, Y. “Tracheal, Laryngeal, and Esophageal Replacement Devices.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

136 Tracheal, Laryngeal, and Esophageal Replacement Devices 136.1

Tracheal Replacement Devices Designs of Artificial Tracheae • Nonporous Tube-Type Artificial Tracheae • Mesh-Type Artificial Tracheae

Tatsuo Nakamura

136.2

Pneumatic Larynxes • Electric Artificial Larynxes • Voice Prostheses

Kyoto University

Yasuhiko Shimizu

Laryngeal Replacement Devices

136.3

Kyoto University

Artificial Esophagi (Esophageal Prostheses) Development of New Artificial Esophagi

As the ability to reconstruct parts of the body has increased, so has the potential for complications associated with the replacement devices used to do so. Some of the most significant complications associated with replacement devices, such as vascular prostheses, cardiac valves, and artificial joints, are caused by infections at the implantation site. It is well known that the presence of a foreign material impairs host defenses and that prolonged infections cannot be cured unless the foreign material is removed from the site of infection. As the trachea, larynx, and esophagus are located at sites facing the “external environment,” these prostheses are exposed to a high risk of infections and severe complications. The development of an artificial trachea, esophagus, and larynx is way behind that of artificial vascular grafts even though they are all tubular organs. In this chapter, we review conventional prostheses and their problems and limitations and discuss the current state of developments in this field.

136.1 Tracheal Replacement Devices End-to-end anastomosis has been one of the standard operations for tracheal reconstruction. However, in patients in whom a large length of trachea has to be resected, this procedure is often difficult (the resectable limit is now considered to be approximately 6 cm). In such patients, alternative methods are required to reconstruct the airway. Such reconstructive methods can be classified into the following three categories: (1) reconstruction with autologous tissue, (2) reconstruction with nonautologous trachea (that is, transplantation), and (3) reconstruction with artificial material. The only one of these operative techniques which achieves good clinical results in the long term is the first (reconstruction with autologous tissue), particularly cervical tracheal reconstruction using an autologous skip flap conduit.

Designs of Artificial Tracheae The artificial tracheae developed previously were designed according to one of two concepts. One is that the implanted prosthesis alone replaces the resected area of trachea, and the inner surface of the reconstructed


FIGURE 136.1

Neville artificial trachea constructed with a nonporous silicone tube with Dacron suture rings.

trachea is not endothelialized. The other is that the implanted prosthesis is incorporated by the host tissue and its inner surface is endothelialized. These two types of prosthesis are called nonporous and mesh types, respectively, which reflect the materials from which they are made.

Nonporous Tube—Type Artificial Tracheae The study of nonporous tube-type trachea has a long history, and many materials have been tried repeatedly for artificial tracheae. However, the results have been unsatisfactory, and now only one prosthesis of this type remains on the market, the Neville artificial trachea, which comprises a silicone tube with two suture rings attached to each of its ends (Fig 136.1). Neville used this trachea in 62 patients, from 1970 to 1988, some of whom survived for a long time [1]. However, late complications, such as migration of the artificial trachea and granular tissue formation at the anastomosis, were inevitable in many cases and occurred within several months and even as late as 2 years after implantation. Therefore, the Neville prosthesis is not used for patients with benign tracheal disease, because clinically, even if a large portion of the trachea up to the bifurcation has to be resected, the airway can be reconstructed easily by tracheotomy using a skin flap with sternal resection. For the alleviation of tracheal stenosis, silicone T-tubes (Fig. 136.2) are widely indicated. The major advantage of such nonporous tubular prostheses is that airway patency can be ensured. Therefore, for patients for whom end-to-end anastomosis is impossible, nonporous prostheses may be used at last resort to avoid threatened suffocation (asphyxiation). Tracheal replacement using the Neville artificial trachea requires the following operative procedure: 1. Right-sided posterolateral skin incision and 4th or 5th intercoastal thoracotomy. Up to the stage of tracheal resection, the operative procedure is similar to that for standard tracheal resection. However, because the tension at anastomosis is not as high as that with an end-to-end anastomosis, neither hilar nor laryngeal release is often necessary. 2. Reconstruction with an artificial trachea. After resection of the tracheal lesion, the oral intubation tube is drawn back, and the trachea is reintubated via the operative field (Fig. 136.3). In patients who require resection that reaches to near the bifurcation, the second intubation tube should be placed in the left main bronchus. An artificial trachea with a diameter similar to that of the tracheal end is used. Any small differences in their diameters can be compensated for by suturing. Anastomosis is carried out using 4–0 absorbable sutures, 2 mm apart, with interrupted suturing. The tracheal sutures are attached to the tracheal cartilage. On completion of the anastomosis on one


FIGURE 136.2

FIGURE 136.3

Silicone T-Tube.

Operation proceeding of tracheal reconstruction using an artificial trachea.

side, the oral intubation tube is readvanced, and the other side is anastomosed under oral ventilation. After anastomosis, an air leakage test at pressure of 30 cmH2O is carried out. In order to avoid rupture of the great vessels near the implanted artificial trachea, they should not be allowed to touch each other. In cases at risk of this occurring, wrapping of the artificial trachea with the grater omentum is also recommended.


3. Postoperative care. Frequent postoperative boncofiberscopic checks should be performed to ensure sputum does not come into contact with the anastomosis sites. The recurrent laryngeal nerves on both sides are often injured during the operation, so the movement of the vocal cord should be checked at extubation. The major reason for the poor results with the Neville tracheal prosthesis during long-term followup was considered to be the structure of the prosthesis, that is, the nonelasticity of the tube and suture rings. A variety of improvements to this prosthesis have been tried, especially at the areas of interface with the host tissue by suture reinforcement with mesh skirts, increasing the flexibility of the tube, and application of hydroxyapatite to the suture rings. However, the problems at the anastomosis have not been conquered yet. In attempt to overcome the problems described above, a new mesh-type artificial trachea has been designed that is intended to be incorporated into the host tissue so that eventually there is no foreignbody-external-environment interface.

Mesh-Type Artificial Tracheae Porous artificial tracheae are called mesh-type because the prosthetic trunk is made of mesh. In the 1950s, several trials of tracheal reconstruction using metallic meshes made of tantalum and stainless steel were conducted. In the 1960s, heavy Marlex mesh was used clinically for tracheal reconstruction, and good short-term results were reported. However, long-term observations showed that this mesh caused rupture of the adjacent graft vessels, which was fatal, so it fell gradually out of use for tracheal reconstruction. When used clinically, because the heavy mesh was rough and not air-tight, other tissues, such as autografted pericardium, fascia, or dura mater, were applied to seal it until the surrounding tissue grew into it and made it air-tight. The pore size of materials conventionally used for artificial vessels, such as expanded PTFE (polytetrafluoroethylene, pore size of 15 ~ 30µm), is so small that the host tissue cannot penetrate the mesh, which is rejected eventually. The optimal pore size for tracheal replacement mesh is 200 ~ 300 µm. Fine Marlex mesh is made of polypropylene with a pore size 200 ~ 300 µm (Fig. 136.4). It is now widely used clinically for abdominal wall reconstruction and reinforcement after inguinal

FIGURE 136.4


Marlex Mesh, fine (left) and heavy (right) (division of the scale; mm).

FIGURE 136.5

New artificial trachea made from collagen-conjugated fine Marlex mesh.

herniation. Collagen-grafted fine Marlex mesh is air-tight, and clinically, good tissue regeneration is achieved when it is used to patch-graft of the trachea. The grafted collagen has excellent biocompatibility and promotes connective tissue infiltration into the mesh. However, the fine mesh alone is too soft to keep the tube open, so a tracheal prosthesis was made of collagen-grafted fine Marlex mesh reinforced with a continuous polypropylene spiral (Fig. 136.5). In dogs, complete surgical resection of a 4-cm length of trachea, which was replaced with a 5-cm long segment of this type of artificial trachea, was performed, and the prostheses were incorporated completely by the host trachea and confluent formation of respiratory epithelium on each prosthetic lumen was observed (Fig. 136.6) [2]. These results indicate that this artificial trachea is highly biocompatible and promising for clinical application.

FIGURE 136.6 Macroscopic inner view of the reconstructed trachea 6 months after operation. Inner surface is covered with smooth and lustrous soft regenerated tissue.


136.2

Laryngeal Replacement Devices

Total laryngectomy is one of the standard operations for laryngeal carcinomas. As radiation therapy and surgery have progressed, the prognosis associated with laryngeal carcinoma has improved. The curability of total laryngeal carcinoma is now almost 70%, and therefore many patients survive for a long time after surgery. Individuals who have undergone laryngectomy are called laryngectomees or laryngetomized patients, and for them, laryngeal reconstruction is of the utmost importance. However, because the larynx is situated just beneath the oral cavity, where the danger of infection is high, successful reconstruction with foreign materials is very difficult. As yet, no total replacement device for the larynx has been developed, and laryngeal transplantation, although apparently feasible, is still at the animal experimental stage. The larynx has three major functions: (1) phonation, (2) respiration, and (3) protection of the lower airway during swallowing. Of these, phonation is considered to be the most important. The conventional so-called artificial larynx can only substitute phonation. A variety of methods have been developed to recover phonation after total larygectomy, which is called vocal rehabilitation. Methods for vocal rehabilitation are classified as (1) esophageal speech, (2) artificial larynx, and (3) surgical laryngoplasty. Twothirds of laryngectomees learn esophageal speech as their means of communication. For the rest, an artificial larynx and surgical laryngoplasty are indicated. The typical devices are the pneumatic and electrical larynx which are driven by the expiratory force and electric energy, respectively. Tracheoesophageal (T-E) fistula with voice prosthesis is the most popular method in surgical laryngoplasty. Fig. 136.7 illustrates the mechanical structures of typical artificial larynxes.

Pneumatic Larynxes The first pneumatic mechanical device was developed by Tapia in 1883. Several variations of pneumatic larynxes are now used. In Fig. 136.8, the pneumatic device uses expired air from the tracheo stoma to vibrate a rubber band or reed to produce a low frequency sound, which is transmitted to the mouth via a tube. Pneumatic transoral larynxes produce excellent natural speech, which is better than that with other artificial larynxes, but their disadvantages are that they are conspicuous and that regular cleaning and mopping of saliva leakage is necessary.

Electric Artificial Larynxes The transcervical electrolarynx is an electric, handheld vibrator that is placed on the neck to produce sound (Fig. 136.9). The frequency used is 100 ~ 200 Hz. The vibrations of the electrolarynx are conducted to the neck tissue and create a low-frequency sound in the hypopharynx. This is the most popular artificial larynx. The transoral artificial laryngeal device is a handheld electric device that produces a low-pitched sound which is transmitted to the back of the mouth by a connecting tube placed in the patient’s mouth. As microelectronic science progresses, great hopes of applying microelectric techniques to the artificial larynx are now entertained, and some devices have been designed, although implantable laryngeal prostheses have not achieved widespread use.

Voice Prostheses As well as the artificial larynxes described above, tracheo-esophageal (T-E) fistula prostheses are now widely used for vocal rehabilitation, and excellent speech and voice results have been achieved. In 1980, Singer and Blom developed and introduced the first simple method, which is called Blom and Singer’s voice prosthesis. The principle of the tracheo-esophageal fistula technique is to shunt expired pulmonary air through a voice prosthesis device into the esophagus to excite the mucosal tissue to vibrate. A fistula is made by puncturing the posterior wall of the trachea 5 mm below the upper margin of the tracheal stoma, and when a patient speaks, he/she manually occludes the stoma to control the expiratory flow


FIGURE 136.7 (a) Sagital views of the laryngectomee (left) and esophageal speech (right). Air flow from the esophagus makes the sound. (b) Pneumatic larynx (reed type) (left); electric artificial larynx (transcervical type), (center); voice prosthesis (T-E shut) of Blom & Singer method, right.

through the fistula to the oral cavity. The voice prosthesis has a one-way value to prevent saliva leakage (Fig. 136.10).

136.3

Artificial Esophagi (Esophageal Prostheses)

In patients with esophageal cancer, the esophagus is resected and reconstructed using a piece of pediculated alimentary tract, such as the gastric conduit, colon, or ileum. However, in some cases, it is impossible to use autologous alimentary tract, for example, in patients who have undergone gastrectomy. In such cases, an artificial esophagus is indicated. However, with the exception of extracorporeal-type esophagi and intraesophageal stent tubes, the esophageal replacement devices developed have remained far from useful in clinical reconstructive practice. The main reason is anastomosis dehiscence, which often leads to fatal infections, as well as prosthetic dislodgment, migration, and narrowing at a late stage. The conventional artificial esophagi now used in the clinic can be broadly classified into two types: extracorporeal and intraesophageal stents. Extracorporeal artificial esophagi are used as bypasses from


FIGURE 136.8 Pneumatic larynxes: Myna (left) (Nagashima Medical Instrument Tokyo, Japan) and Okumura Artificial larynx (Okumura, Osaka, Japan) (right). Arrow marks indicate the portions of reed and rubber band, respectively.

FIGURE 136.9


Electric artificial larynx (transcervical type): Servox (Dr Kuhn & Co. GmBH, Köln, Germany).

FIGURE 136.10 Voice prosthesis (Bivona, Indiana, USA). This valved tube is inserted into a surgically placed tracheo-esophageal fistula for voice restoration following laryngectomy.

FIGURE 136.11

Extracorporeal artificial esophagus (Tokyo University type).

cervical esophageal to gastric fistulae (Fig. 136.11). They are used during the first stage of two-stage esophageal reconstruction. They are made of latex rubber or silicone, but since the development of IVH method, their use is now extremely rare. Intraesophageal stent types of artificial esophagus are made from a rubber or plastic tube, which is inserted in the stenotic part of the esophagus (Fig. 136.12). They are used only for unresectable esophageal carcinoma, i.e., they are not indicated for resectable cases. Therefore, conventional artificial esophagi are only palliative devices for use as stop-gap measures.


FIGURE 136.12

Intraesophageal type of artificial esophagus. (Sumitomo Bakelite Co., Ltd., Tokyo, Japan).

Development of New Artificial Esophagi In contrast to the palliative artificial esophagi, the ideal artificial esophagus would replace the resected part of the esophagus by itself. The artificial esophagi of this type are classified according to the materials from which they are made, namely natural substances, artificial materials, and their composites (hybrids). Artificial Esophagi Made of Natural Substances The first report of an artificial esophagus made of a natural substance was the skin conduit developed by Bircher in 1907. Subsequently, esophageal reconstruction using a variety of natural substances—muscle fasciae, isolated jejunum, autologous aorta, aortic homograft, autologous esophagus, connective tissue conduit, trachea, and freeze-dried dura mater homograft—has been reported. However, non of these has overcome the problems of stenosis, which necessitates continuous bougieing, and other complications at the anastomosis sites. Artificial Esophagi Made of Artificial Materials The first trial of an esophagus made of an artificial material was carried out by Neuhof, who, in 1922, used a rubber tube to reconstruct the cervical esophagus. Subsequently, several artificial materials have


been tried repeatedly, such as polyethylene, Dacron, stainless steel mesh, tantalum mesh, dimethylopolysiloxane, polyvinylformal sponge, Teflon, nylon, silicone rubber, collagen-silicone, Dacron-silastic, acrylresin, and expanded PTFE tubes, but all the trials ended in failure. These materials were foreign bodies, after all, and the host tissues continuously rejected them, which caused anastomotic dehiscence followed by infections. Even rare cases, whose artificial esophagi escaped early rejection, did not avoid the late complications, such as prosthetic migration and esophageal stenosis. Hybrid Artificial Esophagi Cultured cells are now widely used for hybrid artificial organs, especially for metabolic organs, such as the liver and pancreas. A hybrid artificial esophagus comprising a latissimus dorsi muscle tube, the inner surface of which is epithelialized by human cultured esophageal epithelial cells, is being studied by the Keio University group [3]. Human esophageal epithelial cells cultured on collagen gel for 10 days were transplanted onto the surfaces of the lattisimus dorsi muscles of athymic mice, and new epithelial cells were observed, which indicates it may be possible to develop an artificial esophagus incorporating cultured epithelial cells. New Concepts in Artificial Esophageal Design The previous studies on artificial esophagi are a history of how to prevent the host tissue from recognizing the artificial esophagus as a foreign body, and all the trials, without exception, resulted in failure. Currently, an artificial esophagus made according to a completely new design concept is undergoing trials. The outer collagen sponge layer of the prosthesis is intended to be replaced with host tissue over a period of time, and its inner tube acts as a palliative (temporary) stent until the outer layer has been replaced by host tissue. This type of artificial esophagus comprises an inner silicone tube and an outer tube of collagen sponge (Fig. 136.13), which is nonantigenic and has excellent biocompatibility, as well as promoting tissue regeneration in the form of an extracellular matrix. In dogs implanted with a 5-cm length of this artificial esophagus, esophageal regeneration was accomplished in 3 weeks, and all the early

FIGURE 136.13

An artificial esophagus made of collagen sponge which is intended to be replaced with host tissue.


FIGURE 136.14 Pathologic finding of a regenerated esophageal tissue in a dog using the artificial esophagus (Fig. 136.13). Continuous epithelial layer and the muscle and esophageal gland were regenerated at the interposed area.

complications at the anastomosis were overcome. However, in dogs from which the inner stents were removed as soon as mucosal regeneration was accomplished, stenosis developed rapidly, and the regenerated esophagus began to constrict. Such stenosis and constriction could not be overcome, even when autologous buccal mucosal cells were seeded onto the collagen sponge. Accordingly, on the basis of the hypothesis that stenosis and constriction depend on the maturity of the regenerated submucosal tissue, stent removal was postponed for at least 1 week after mucosal epithelialization of the artificial esophagus was complete. Late stenotic complications did not occur, and regeneration of muscle tissue and esophageal glands was observed (Fig. 136.14) [4]. The regenerated esophagus showed adequate physiologic function and pathologically satisfactory results. The limit to the length of the esophagus that can be resected successfully was reported to be 9 cm, and the longest successful artificial esophagus developed hitherto was 5 cm. In order to achieve more widespread use, artificial esophagi which can be used for longer reconstructions are needed. Although still at the stage of animal experiments, the long-term observations indicate that typical difficulties, such as anastomotic leakage, ablation, and dislocation of the prosthesis, have been overcome, which suggests this type of prosthesis will have a promising future in clinical practice.

Defining Terms Collagen: A main supportive protein of connective tissue, bone, skin, and cartilage. One-third of protein of vertebrate animal consists of collagen. End-to-end anastomosis: An operative union after resecting the lesion, each end to be joined in a plane vertical to the ultimate flow through the structures. Expanded PTFE (polytetrafluorethylene): PTFE is polymer made from tetrafluoroethylene (CF2 = CF2), with the structure of –CF2 –CF2–. It provides excellent stability chemically and thermally. Teflon is the trade mark of PTFE of Du Pont Co. Expanded PTFE has a microporous structure which has elasticity and antithrombogenisity in the body and is medically applied for vascular grafts or surgical seats. Extracellular matrix (ECM): The substances which are produced by connective tissue cells. The two major components of the extracellular matrix are collagen and proteoglycans. There are a number of other macromolecules which provide important functions such as tissue growth, regeneration or aging. © 2000 by CRC Press LLC

Heavy Marlex mesh and fine Marlex mesh: Surgical mesh is used for reconstruction of the defect that results from massive resection. Most popular surgical meshes are Marlex meshes. Heavy mesh is made of high-density polyethylene and has been used for reconstruction of chest wall. Fine mesh is made of polypropylene and is now widely used for abdominal wall reconstruction or reinforcement of inguinal herniation operation. Hilar release and laryngeal release: The pulmonis hilus is the depression of the mediastinal surface of the lung where the blood vessels and the bronchus enter. The hilus is fixed by the pulmonary ligament to downward. In order to reduce the tension at the tracheal anastomoses, the pulmonary ligament is released surgically. This method is called hilar release. At the resection and reconstruct of upper trachea, the larynx can be also released from its upper muscular attachment. This method is called laryngeal release and up to 5 cm of tracheal mobilization may be achieved. Hydroxyapatite: An inorganic compound, Ca10(PO4)6(OH)2, found in the matrix of bone and the teeth which gives rigidity to these structures. The biocompatibility of hydroxyapatite has attracted special interest. IVH (intravenous hyperalimentation): An alimentation method for patients who cannot eat. A catheter is placed in the great vessel, through which high concentration alimentation is given continuously. Silicone T-tube: A self-retaining tube in the shape of a T which is made of silicone. Tracheal T-tube is used popularly for tracheal stenotic disease and serves both as a tracheal stent and tracheotomy tube. One side branchi of T-tube projects from the tracheotomy orifice.

References 1. Neville WE, Bolanowski PJP, Kotia GG. 1990. Clinical experience with the silicone tracheal prosthesis. J Thorac Cardiovasc Surg 99:604. 2. Okumura N, Nakamura T, Takimoto Y, et al. 1993. A new tracheal prosthesis made form collagen grafted mesh. ASAIO J 39:M475. 3. Sato M, Ando N, Ozawa S, et al. 1993. A hybrid artificial esophagus using cultured human esophageal cells. ASAIO J 39:M554. 4. Takimoto Y, Okumura N, Nakamura T, et al. 1993. Long-term follow up of the experimental replacement of the esophagus with a collagen-silicone composite tube. ASAIO J 39:M736.

Further Information Proceedings of the American Society of Artificial Internal Organs Conference are published annually by the American Society of Artificial Internal Organs (ASAIO). These proceedings include the latest developments in the field of reconstructive devices each year. The monthly journal Journal of Thoracic and Cardiovascular Surgery reports advances in tracheal and esophageal prosthetic instruments. An additional reference is H.F. Mahieu, “Voice and speech rehabilitation following larygectomy, Groningen,” the Netherlands, Rijksuniversiteit Groningen, 1988.


Intaglietta, M., Winslow, R. M. “Artificial Blood.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

137 Artificial Blood 137.1 137.2 137.3 137.4 137.5 137.6 137.7

137.8 137.9 137.10 137.11 137.12 137.13

University of California

137.14 137.15

Robert M. Winslow

137.16

Marcos Intaglietta

Oxygen Carrying Plasma Expanders and the Distribution of Transport Properties in the Circulation The Distribution of Oxygen in the Circulation Oxygen Gradients in the Arteriolar Wall Mathematical Modeling of Blood/Molecular Oxygen Carrying Plasma Expander Blood Substitutes and Hemodilution Hematocrit and Blood Viscosity Regulation of Capillary Perfusion During Extreme Hemodilution in Hamster Skin Fold Microcirculation by High Viscosity Plasma Expander Crystalloid and Colloidal Solutions as Volume Expanders Artificial Oxygen Carriers Fluorocarbons Hemoglobin for Oxygen Carrying Plasma Expanders Hemoglobin-Based Artificial Blood Results in the Microcirculation with Blood Substitution with αα-Hemoglobin Hemoglobin and Nitric Oxide (NO) Binding Rational Design of an Oxygen Carrying Molecular Plasma Expander Conclusions

SANGART Inc.

Blood is a key component of surgery and treatment of injury, and its availability is critical for survival in the presence of severe blood losses. Its present use under conditions of optimal medical care delivery is one unit (0.5 L) per 20 person-year. On a world-wide basis, however, its availability and use is limited to one unit per 100 person-year according to statistics from the World Health Organization. The gap between current optimal need and actual availability is further aggravated by the fact that the blood supply in many locations is not safe. Even in conditions of optimal testing and safeguards, blood per se has a number of inherent risks [Dodd, 1992], ranging from about 3% (3 adverse outcomes per 100 units transfused) for minor reactions, to a probability of 0.001% of undergoing a fatal hemolytic reaction. Superimposed on these risks is the possibility of transmission of infectious diseases such as hepatitis B (0.002%) and hepatitis non-A non-B (0.03%). The current risk of becoming infected with the human immunodeficiency virus (HIV) is about 1 chance in 400,000 under optimal screening conditions. These are risks related to the transfusion of one unit of blood, and become magnified in surgical interventions requiring multiple transfusions.


The HIV epidemic caused many fatalities due to blood transfusion before stringent testing was introduced in 1984. The danger of contamination of the blood supply led the U.S. military to develop blood substitutes starting in 1985 with the objective of finding a product for battlefield conditions that was free of contamination, could be used immediately, and did not need special storage. The end of the Cold War, changes in the nature of military engagements, the slow progress in the development of a blood substitute, and the fact that blood testing caused the blood supply to be again safe and abundant in both the U.S. and Europe lowered the interest of the U.S. military in artificial blood, and the development passed almost exclusively into the hands of private industry. While about a dozen products have been tested in humans, at present only three are in clinical trials for blood replacement. These products have side effects whose relevance is difficult to assess in the absence of massive clinical trials. A side effect common to hemoglobin-based products is hypertension, which was proposed to be of additional therapeutic value in the treatment of hypotensive conditions associated with severe blood losses and hemorrhagic shock. However, extensive clinical trials led by Baxter Healthcare, with their product HemAssist™ showed that their product caused twice the mortality found with the use of conventional therapies for volume resuscitation. This unfortunate outcome was fully predicted by results in academic research showing that hypertension per se causes a severe impairment of microvascular function. The key transport parameters that determine the exchange of oxygen in the microcirculation were, until recently, incompletely understood. Most experimental studies underlying the development of artificial blood emphasized systemic, whole organ methods in order to assess efficacy and biocompatibility, without quantitative analysis of transport phenomena inherent to blood function, or data from the microcirculation. Advances in instrumentation and in vivo methods (Fig. 137.1) during the past few years have provided new understanding of how oxygen is distributed to the tissues by the microscopic blood vessels, and this information has proven to be crucial in identifying the transport characteristics needed for artificial blood to work effectively, and to counteract some of its inherent problems.

FIGURE 137.1 Experimental hamster skin fold model for the analysis of the microvasculature in the awake animal, without anesthesia and after the effects of surgery have subsided. This technology allows to quantify blood flow, blood and tissue pO2, leukocyte activation, blood vessel permeability to macromolecules, and blood vessel tone throughout the arterioles, capillaries, and venules. The tissue under study consists of skeletal muscle and subcutaneous connective tissue. Arterial and venous catheters are used to monitor systemic data.


137.1 Oxygen Carrying Plasma Expanders and the Distribution of Transport Properties in the Circulation To the present it has not been possible to obtain an artificial fluid with the same properties as blood, primarily due to its cellular nature. Consequently, artificial blood or blood substitutes are plasma expanding fluids that carry more oxygen than that dissolved in water or plasma. Their physical properties are significantly different from those of natural blood, and therefore it is necessary to determine the positive or negative effects on circulatory function and tissue oxygenation that may arise from these differences. Under normal conditions, the transport properties of blood, such as hydraulic pressure, partial pressure of oxygen, oxygen content and viscosity, are distributed (Fig. 137.2). The distribution of transport properties of blood is matched to anatomical features of the macro- and microcirculation and set within narrow limits throughout the life of the organism. Hydraulic blood pressure, which changes continuously in the circulation due to viscous losses, is an example of this situation. Each branching order of the vasculature is adapted to recognize a narrow range of hydraulic blood pressures and react if pressure changes and the normal set point is not met. The distribution of blood pressure influences the cellular composition of blood vessels, and is regulated by the so-called myogenic response [Johnson, 1986], which causes vessels to constrict, thus increasing viscous losses, if pressure increases and vice versa. The same type of matching is present for the distribution of oxygen tension and blood oxygen content in the vasculature. The shape of the oxygen dissociation curve for hemoglobin and the distribution of

FIGURE 137.2 Distributed nature of transport properties in the hamster skinfold microcirculation. Arterial blood vessels are classified according to their hierarchical position starting with the A0 small arteries (100 µm and greater diameter), and ending with A4 terminal arterioles. Blood pressure, blood viscosity, and blood pO2 fall continuously from the systemic value. Furthermore, the characteristics of the oxygen saturation of hemoglobin curve correspond to specific microvascular locations. This distribution is maintained through the life of an organism and is autoregulated by central and peripheral controls, including adrenergic nerve endings which have the highest density in the A3 arterioles. Introduction of a molecular plasma expander fundamentally changes the distribution of viscosity, blood pressure, shear stress, and the availability of oxygen to the vessel wall. Autoregulatory processes sense these changes and cause the microcirculation to react. Data on viscosity is derived from [Lipowsky, 1987], data on blood pressure is from the hamster cheek pouch from [Joyner et al., 1972], other data is from our studies.


enervation of arterioles is such that the knee of the oxygen dissociation curve for hemoglobin is located in arterioles that present the highest density of adrenergic enervation [Saltzman et al., 1992]. Blood viscosity and blood flow velocity in the circulation are also distributed in such a fashion that together with the anatomical features of the local blood vessels, shear stress is caused to be virtually uniform throughout the circulation. This is also the consequence of the continuous variation of hematocrit from the systemic value in the larger blood vessels, to about half this value in the capillaries, due to the Faraheus-Lindquist effect (a property due to the cellular composition of blood) and the very strong dependance of blood viscosity on hematocrit. In larger arterioles, blood viscosity is about 3.5 to 4.0 centipoise (cP), while in the smaller arterioles viscosity falls to about that of plasma (1.1 to 1.2 cP). This has large effects on shear stress and shear stress dependant release of endothelial dependant relaxing factors (EDRF, NO, and prostaglandins), since the circulation is designed to maintain shear stress constant. In other words, shear stress is also present in a prescribed way in the microcirculation when the system is perfused by normal blood. A molecular oxygen carrying plasma expander is not subjected to the Fahraeus-Lindquist effect (which determines the progressive decrease of hematocrit in the microcirculation), causing the whole microcirculation to be exposed to a uniform viscosity (see the mathematical model) and significantly changing shear stress distribution. The significance of the mechanical traction at the interface between flowing blood and tissue is that shear stress modulates the production of vasoactive materials such as nitric oxide (NO) and prostacyclin [Frangos et al., 1985; Kuchan et al., 1994] and therefore is an active control of blood flow and the oxygen supply. Furthermore, NO is also a major regulator of mitochondrial metabolism [Shen et al., 1995], acting as a brake for tissue oxygen consumption. These considerations show that the transport properties of blood are intertwined with the physical and biological regulation of oxygen delivery and the level of tissue oxygen consumption.

137.2 The Distribution of Oxygen in the Circulation Oxygenation of tissue as a whole has been extrapolated from the so-called Krogh model, which is focused on how gases are exchanged between blood flowing in a cylindrical conduit, the single capillary, and a surrounding tissue cylinder. It is generally assumed that most of the oxygen is exchanged at this level, implying the existence of large blood/tissue oxygen gradients in a substantial portion of tissue capillaries. However, capillary/tissue O2 gradients are maximal in the lung (50 mmHg/µm) and minimal in the tissues (0.5 mmHg/µm). Most tissue capillaries appear to be in near oxygen concentration equilibrium with the tissue; thus, large oxygen gradients are not present, suggesting that capillaries may not be the primary mechanism for tissue oxygenation. The technology of phosphorescence quenching for measuring oxygen partial pressure optically in the microvessels and the surrounding tissue [Torres & Intaglietta, 1994; Wilson, 1993), when used in conjunction with the awake hamster skinfold model, allows for in vivo analysis of the assumptions in the Krogh model. Current findings indicate that in the hamster skin fold connective tissue and skeletal muscle at rest, most of the oxygen is delivered by the arterioles and that little oxygen is contributed by the capillaries. In this tissue average capillary pO2 is about 25 mmHg and venular pO2 is higher. This indicates that at least half of the oxygen in blood exits the blood vessels prior to arrival in the tissue. In summary, these results and previous findings from other laboratories [Intaglietta et al., 1996] show that: 1. Capillary blood pO2 is only slightly higher (about 5 mmHg) than tissue pO2 . 2. Arterio/venous capillary pO2 differences are very small because tissue pO2 is essentially uniform, and capillaries are close to pO2 equilibrium with the tissue. 3. The only tissue domain where pO2 exhibits large gradients is the immediate vicinity of the microvessels (vessels with diameter 80 µm and smaller), a tissue compartment whose main constituent is the microvascular wall. 4. A major portion of blood oxygen exits the circulation via the arterioles.


137.3 Oxygen Gradients in the Arteriolar Wall The phosphorescence optical technology allows us to make an accurate mass balance between the decrease of oxygen content in the arterioles and the diffusion flux of oxygen out of the microvessels determined by the oxygen gradients in the surrounding tissue. These measurements made simultaneously inside and outside of the vessels show that oxygen exiting from arterioles is driven by steep oxygen gradients at the arteriolar wall, which is consistent with the hypothesis that the arteriolar wall is a high metabolism tissue and therefore a large oxygen sink [Tsai et al., 1998]. These steep gradients are present in arterioles, but not in capillaries and venules. The large oxygen consumption is due to biological activity in the endothelium and smooth muscle [Kjellstrom et al., 1987]. These findings lead to the following conceptualization for the design of artificial blood: 1. Endothelium and smooth muscle serve as a metabolic barrier to the passage of oxygen from blood to tissue, which in part protects the tissue from the high oxygen content (pO2) of blood. 2. One of the goals of basal tissue perfusion is to supply oxygen to the endothelium and smooth muscle. 3. Oxygenation of working tissue (exercising skeletal muscle) results from three events: • Lowering of the vessel wall metabolic barrier. • Increased perfusion with oxygenated blood. • Deployment of a biochemical process that protects the tissue from high pO2 levels. 4. Under basal conditions, tissue capillaries only partially serve to supply oxygen to the tissue. They may be a structure to expose the endothelium to blood in order to fulfill the large oxygen demand of these cells and provide for the extraction of CO2. 5. The physical and biological properties of blood affect the oxygen consumption of the vessel walls.

137.4 Mathematical Modeling of Blood/Molecular Oxygen Carrying Plasma Expander The previous discussion shows that tissue oxygenation is the result of the interplay of quantifiable physical events, and therefore it may be subjected to analytical modeling to identify key parameters and set the stage for rational design. The critical issue is to understand the changes due to the presence of molecular oxygen carriers. These changes are: (a) change in availability of oxygen to the vessel wall due to the molecular nature of the carrier; (b) distortion of the pattern of intraluminal oxygen distribution in the microcirculation; (c) lowered viscosity and decreased generation of EDRF due to lowered shear stress; (d) increased colloid osmotic pressure; and (e) increased vessel wall pO2 gradient. The mathematical model is based on oxygen mass balance in subsequent vessel segments, starting with a major arterial vessel and terminating in the capillaries, with the objective of calculating the oxygen tension of blood in the capillaries, which should correspond to tissue pO2. The result of this analysis is the following equation, which expresses how the total loss of oxygen from the arterial and arteriolar network Kn is related to anatomical and transport features of the blood vessels and blood [Intaglietta, 1997]. Total oxygen exit is the summation of losses from n individual vascular segments i. This equation gives the functional relationship between transport parameters that determine capillary blood pO2 for given changes in the physical properties of blood:

Kn =

∑

ki =

(

128 µ

)

F Htc , C mt

g oαD 2

n

∑ n d ∆P 1

L2i

3 i i

(137.1)

i

where µ is blood viscosity, F(Htc,C) is the concentration of hemoglobin (red blood cells + molecular), mt is the maximum amount of oxygen that can be dissolved in blood under normal conditions, D is the


diffusion constant for oxygen in tissue, α is the solubility of oxygen in tissue, go is the oxygen consumption by the vessel wall, Li is the length of each vessel segment, di is the diameter of each segment, and ni is the slope of the oxygen dissociation curve for hemoglobin. The summation shows two distinct groups of terms. One group is common to all vessel segments and includes blood viscosity, hematocrit or blood oxygen carrying capacity, and vessel wall metabolism. The second group is a summation where each term is specific to each vascular segment. This expression shows that, in principle, better capillary oxygenation results from lowering viscosity. However, increased capillary oxygenation is a signal that triggers auto regulatory mechanisms which strive to maintain capillary oxygen constant through vasoconstriction. Thus, lowering blood viscosity should be expected to lead to vasoconstriction. The factor go represents vessel wall metabolism which is directly affected by the composition of blood and flow velocity. Furthermore go has a basal value representative of the baseline activity necessary for the living processes of the tissue in the microvessel wall and alteration of this activity, such as by an inflammatory process or increased tone, leads to an increase in tissue metabolism and therefore lowering of tissue oxygenation (since the increase in Kn lowers capillary pO2).

137.5 Blood Substitutes and Hemodilution In general, it is possible to survive very low hematocrits, corresponding to losses of the red blood cell mass of the order of 70%; however, our ability to compensate for comparatively smaller losses of blood volume is limited. A 30% deficit in blood volume can lead to irreversible shock if not rapidly corrected. Maintenance of normovolemia is the objective of most forms of blood substitution or replacement, leading to the dilution of the original blood constituents. This hemodilution produces systemic and microvascular phenomena that underlie all forms of blood replacement and provides a physiological reference for comparison for blood substitutes. The fluids available to accomplish volume restitution can be broadly classified as crystalloid solutions, colloidal solutions, and oxygen carrying solutions. All of these materials significantly change the transport properties of blood, and therefore it is important to determine how these changes affect tissue oxygenation, a phenomenon that takes place in the microvasculature. Therefore, hemodilution must be analyzed in terms of systemic effects and how these, coupled with the altered composition of blood, influence the transport properties of the microcirculation.

137.6 Hematocrit and Blood Viscosity Blood viscosity is primarily determined by the hematocrit in the larger vessels while it is a weaker function of the systemic hematocrit in the microcirculation. At a given shear rate, blood viscosity is approximately proportional to the hematocrit squared according to the relationship:

µ = a s + bs H 2 while microvascular blood viscosity can be empirically described by a relation of the form:

µ = am + bm H where µ is the blood viscosity in centipoise and ai and the bi are parameters that are shear rate and vessel size dependent [Dintenfass, 1979; Quemada, 1978]. In the microcirculation, blood viscosity is relatively insensitive to shear rate. These relationships show that when hematocrit is reduced, systemic viscous pressure losses decrease much more rapidly than those in the microvasculature, while in the microcirculation the A-V pressure drop is not very much affected. The net result is that if arterial pressure remains constant, hemodilution produces a significant pressure re-distribution in the circulation [Mirhashemi et al., 1987].


Hemodilution increases central venous pressure, which improves cardiac performance and increases cardiac output [Richardson and Guyton, 1959]. This effect causes increased blood flow velocity and therefore the maintenance of oxygen delivery capacity, since a lesser number of red blood cells arrive at a greater frequency. As a consequence of hemodilution, both systemic and capillary oxygen carrying capacities improve up to hematocrit 33% and are at the normal value at arterial hematocrits of the order of 27%. The maximum improvement in oxygen carrying capacity is about 10%. This behavior of the heart and the circulation as determined by the viscosity of the circulating blood has been subjected to exhaustive experimental and clinical verification. The consequence of the adaptability of the circulation to changes in blood viscosity is that, in general, oxygen delivery capacity is not compromised up to red blood cell losses of about 50%. This fact determines the transfusion trigger, currently set at about 7 g Hb/dl. It should be noted that up to the transfusion trigger blood losses can be corrected with the use of a molecular plasma expander. Furthermore, if artificial blood is based on a molecular oxygen carrying material, its introduction in the circulation once the transfusion trigger is passed should be analyzed also as a phenomenon of extreme hemodilution. Conversely, introduction of an oxygen carrying plasma expander prior to reaching the transfusion trigger should show no improvement on oxygenation, since this point is the limit for compensatory adjustments in the circulation.

137.7 Regulation of Capillary Perfusion During Extreme Hemodilution in Hamster Skin Fold Microcirculation by High Viscosity Plasma Expander Although capillaries do not appear to be the determinant structure for the supply of oxygen, our studies [Kerger et al., 1996] have shown that maintenance of functional capillary density (FCD, defined as the number of capillaries that possess red blood cell transit in a mass of tissue) in shock is a critical parameter in determining the outcome in terms of survival vs. non-survival, independent of tissue pO2 , suggesting that extraction of products of metabolism may be a more critical function of capillaries than oxygenation (Fig. 137.3). When hemodilution is carried to extreme conditions, defined as the replacement of more than 80% of the red blood cell mass, blood viscosity falls to near plasma levels. A concomitant effect is the reduction of FCD to near pathological levels. The decrease in FCD observed in blood substitutions with plasma expanders is in part due to the lowered viscosity of the resulting blood in circulation. This has been demonstrated with conventional plasma expanders such as dextran 70 kDa, where reductions of systemic hematocrit were to 75% of control and final plasma viscosities of 1.38 cP lowered FCD to 53% of control, while hemodilution with a combination of dextrans of different molecular weights yielding a final plasma viscosity of 2.19 cP maintained FCD to near normal levels. In these experiments, vessel wall shear stress was not changed after low viscosity hemodilution while it increased by a factor of 1.3 in arterioles and 2.0 in venules relative to baseline with high viscosity [Tsai et al., 1998] exchanges. Therefore increased shear stress dependant release of endothelium derived relaxing factors are a possible mechanism that reverses the constrictor effects and decreased FCD due to low blood viscosity.

137.8 Crystalloid and Colloidal Solutions as Volume Expanders Crystalloids are among the most widely used fluids for volume replacement. Ringer’s lactate, for example, is administered in volumes that are as much as three times the blood loss, since the dilution of plasma proteins lowers the plasma oncotic pressure causing an imbalance of the fluid exchange favoring microvascular fluid extravasation and edema. The advantage of crystalloid solutions is that large volumes can be given over a short period of time with low danger of increasing pulmonary wedge pressure. Excess volumes are rapidly cleared from the circulation by diuresis, which in many instances is a beneficial side effect in the treatment of trauma. Blood volume replacement with Ringer’s lactate lowers blood viscosity.


FIGURE 137.3 Changes of functional capillary density (FCD) in the hamster skin fold model during shock and resuscitation. FCD is the number of capillaries with red blood cell transit in a microscopic field of view. Hamsters were subjected to 4 h hemorrhagic shock to 40 mmHg and resuscitated with their own blood. FCD was the principal microvascular parameter that was the predictor of outcome, while microvascular and tissue pO2 was the same for survivors and non-survivors. This result indicates that resuscitation solutions should ensure the restoration of FCD as well as re-establish tissue oxygenation [Kerger et al., 1996]. Bsln: Baseline. Timepoint: Time after induction of hemorrhage. B30: 30 min. after resuscitation. B24: 24 h after resuscitation. ∗: Significantly different from control, p > 0.05. +: Significantly different between survivors and non-survivors, p > 0.05.

The rapid clearance of crystalloid solutions from the intravascular compartment is at the expense of the expansion of the tissue compartment and edema. Peripheral edema has been speculated to impair oxygen delivery, wound healing, and the resistance to infection. Microvascular flow may be impaired by capillary compression from edema. However, given the wide clinical experience with the use of crystalloids, it appears that these are not major effects. A more important effect is the development of pulmonary edema, or adult respiratory distress syndrome (ARDS); however, the relationship between the use of large volumes of salt solutions and ARDS is not firm. Larger colloidal molecular weight materials such as albumin have the advantage of longer retention time, they do not extravasate and therefore do not cause edema. Albumin is used as a plasma expander for emergency volume restitution, but due to cost considerations the synthetic colloids dextran and hydroxyethyl starch are used more frequently. These materials are free of viral contaminations, but may cause anaphylactic reactions and have a tendency to alter platelet function thus interfering with hemostasis. It is not well established whether these materials increase bleeding per se or the effect is due to improved perfusion. The beneficial effect of this form of volume replacement resides in the high oncotic pressure which they generate, maintaining intravascular volume in the amount of about 20 ml of fluid per gram of circulating colloid. In this context, dextran 70 kDA and hydroxyethyl starch have the highest retention capacity, 22 and 17 ml/g, respectively, while albumin is 15 ml/g. Furthermore, this form of volume replacement lowers blood viscosity with all the associated positive transport effects described for hemodilution [Messmer et al., 1972].


137.9 Artificial Oxygen Carriers The amount of oxygen carried in simple aqueous solutions is inconsequential in the overall oxygen requirements of the tissue, and therefore when intrinsic oxygen carrying capacity has to be restored, it is necessary to include an oxygen carrier. This can be presently accomplished through the use of modified hemoglobins which bind oxygen chemically and reversibly, and fluorocarbons which have a high capacity for dissolving oxygen. These compounds present important differences which will ultimately determine in what form of blood substitution and replacement they will be used. Hemoglobin separated from the red cell membrane, i.e., stroma-free hemoglobin, carries oxygen with a high affinity. These materials have intrinsically high oxygen carrying capacity; however, they are based on a biologically active molecule, which in most instances requires specialized storage procedures. By contrast, fluorocarbons carry a limited amount of oxygen under normal atmospheric conditions, but are biologically inert, can be stored at room temperature, and are excreted as gas through the lungs.

137.10

Fluorocarbons

Fluorocarbons (perfluorochemicals, PFCs) are compounds that have a very high solubility of oxygen and their oxygen transport capacity is directly proportional to their concentration and pO2. They are formulated in stable water soluble emulsions that are metabolically inert. These PFC emulsions consist in droplets of fluorochemicals in the range of 0.1 to 0.2 µm diameter coated by a thin film of egg yolk phospholipids. These products are heat sterilized and can be stored under ambient conditions ready for use. The technology for large scale manufacture is well established and the necessary materials for production are well characterized and commercially available. Their half life in the circulation is in the range of 6 to 9 h. Pure PFCs have oxygen solubilities of the order of 50 ml O2 /100 ml at 37°C and 760 mmHg oxygen pressure. The solubility varies linearly as a function of gas pressure. These materials carry about 30% more oxygen than blood when diluted to 60% (the recommenced clinical form of administration) and with 100% inspired oxygen 100%. They have been given clinically at a maximal dosage of 2 g PFC/kg (vs. 15 g hemoglobin/kg for normal blood in humans) [Lamy et al., 1998]. At this dosage, in an isovolemic exchange they increase intrinsic oxygen carrying capacity of blood by about 20%, provided that 100% oxygen is inspired. Since normal tissues and the microcirculation operate in the range of 20 to 50 mmHg, the effect of this fundamentally altered blood pO2 and oxygen content on the normal distribution of oxygen tension in the circulation is not known. However, since tissues regulate their blood oxygen supply and pO2, it is unlikely that microvascular blood pO2 would be much greater than 50 mmHg without eliciting autoregulatory adjustments; therefore, in the absence of hypertensive effects it may be assumed that microvascular blood pO2 in the presence of PFCs does not exceed 50 mmHg and therefore in terms of tissue oxygen delivery the contribution is small relative to hemoglobin since at pO2 50 mmHg they carry about 2 ml O2 /dl, while hemoglobin carries 15 ml O2 /dl. The viscosity of PFCs is only slightly higher than water; therefore, when used in the amounts described they behave as low viscosity plasma expanders, and therefore should exhibit some of the beneficial effects to be derived from moderate hemodilution.

137.11

Hemoglobin for Oxygen Carrying Plasma Expanders

The hemoglobin molecule in solution presents unique characteristics as an oxygen carrying plasma expander, since it has a molecular weight similar to that of albumin and therefore comparable intravascular retention. Its ability to chemically bind oxygen determines that large amounts of oxygen can be associated chemically at relatively low pO2 , and its molecular weight is such that it does not extravasate. Consequently, it should provide an extended period of colloid osmotic effects that prevent the passage of water into the interstitium and therefore edema. However, when hemoglobin is diluted from the


concentration present in red blood cells, the tetrameric molecule tends to spontaneously dissociate into smaller molecular weight dimers and monomers which rapidly extravasates from the vascular compartment. Until recently it was assumed that when hemoglobin is free in the circulation it has an oxygen affinity that is too high for tissue oxygenation. As a consequence, strategies for the modification of hemoglobin by chemical means at present are aimed at prolonging intravascular retention and reducing oxygen affinity [Winslow, 1997]. Hemoglobin can be prevented from dissociating by chemically crosslinking the tetramers such as the widely used agent glutaraldehyde. Direct crosslinking with this agent produces a spectrum of different high molecular weight hemoglobin derivatives since glutaraldehyde is non-specific and reacts with any amino group. A more homogenous product can be obtained by reacting human hemoglobin first with the 2,3-DPG analog PLP, which reduces oxygen affinity and then polymerizing the compound with glutaraldehyde. This product is called PLP-polyHB and has near normal oxygen carrying capacity and a plasma half-time retention (in rats) of about 20 h. Diaspirins are another group of crosslinking agents. The reaction of native hemoglobin with bis(3,5dibromosalicy)fumarate (DBBF) leads to the linking of the two α chains of the hemoglobin molecule, producing a compound called ααHb which has an oxygen affinity (P50 ) of 30 mm Hg and an intravascular retention time of the order of 12 h in rats. Phosphorylated compounds added to purified hemoglobin also change the P50 value of the resulting compound and have been studied extensively. There is an increasing variety of hemoglobin compounds that is being developed; however, at this time ααHb is the most widely studied compound since it was produced in quantity by the U.S. Army Letterman Institute, San Francisco, and a similar compound was developed and used in extensive clinical trials by Baxter Healthcare Inc. Hemoglobin can be produced as a recombinant protein leading to the construction of either the naturally occurring molecule or variants that have different crosslinking and oxygen affinity. It is presently possible to produce both alpha and beta chains of human hemoglobin in bacteria, yeast, and transgenic animals. While it is in principle possible to produce large amounts of a specific product by these means, there remain problems of purification, elimination of other bacterial products, and endotoxins that may be expressed in parallel. Human hemoglobin genes have been induced into transgenic mammals, leading to the potential production of large amounts of human hemoglobin. However, the production is not 100% pure and some of the original animal material is also present, causing an important purification problem. This problem may be compounded by potential immunogenicity of animal hemoglobin and the potential transmission of animal diseases to humans. A unique form of hemoglobin modification is the process of pegylation, consisting of attaching polyethylene glycol polymeric strands to the surface of the hemoglobin molecule. The resulting product termed PEG-hemoglobin interacts with the surrounding water, producing an aqueous barrier that surrounds the molecule. This phenomenon causes several beneficial effects, namely (1) renders the molecule invisible to the immune system; (2) draws water into the circulation increasing blood volume; (3) increases molecular size of the molecule, thus further limiting its capacity to extravasate; (4) decreases the diffusive oxygen delivery from HbO2 by increasing path length; and (5) increases plasma viscosity. This type of molecule is presently manufactured by Sangart, Inc., La Jolla, and Apex Bioscience Inc., North Carolina, who uses human hemoglobin and Enzon Inc., New Jersey, who uses bovine hemoglobin.

137.12

Hemoglobin-Based Artificial Blood

Many hemoglobin-based artificial blood substitutes are being developed which may be classified relative to the technology used and whether the hemoglobin is either of human, animal, or bacterial origin. It is now established that hemoglobin containing red blood cell substitutes deliver oxygen to the issue and maintain tissue function. However, problems of toxicity and efficacy are not yet resolved, and are shared by many of the hemoglobin solutions presently available. Renal toxicity is one of the primary problems


because hemoglobin toxicity to the kidney is a classic model for kidney failure. The mechanism of toxicity includes tubular obstruction, renal arterial vasospasm, and direct toxicity to tubular cells. It is not clear whether hemoglobin must be filtered by the kidney to be toxic, and the tolerance of the kidney to different types of hemoglobins and concentrations is not established [Paller, 1988]. There are many reports of anaphylaxis/anaphylactoid reactions to hemoglobin infusion. This may result from heme release from hemoglobin, the presence of endotoxins bound to hemoglobin or preexisting in the circulation, or from oxygen-derived free radicals released after exposure to hemoglobin. Vasoconstriction of coronary, cerebral, and renal vessels has also been attributed to the presence of free hemoglobin in the circulation, a phenomenon that may be due to the intrinsic capacity of hemoglobin to scavenge nitric oxide or metabolic autoregulation. Toxicity of hemoglobin may be due to special characteristics of the manufacturing procedure, the given molecule, or problems due to purification. It may also be inherent to the presence of large quantities of a molecular species for which the circulation has a very limited tolerance when present in solution. This problem may be circumvented by separating hemoglobin from the circulation by encapsulating this material in a synthetic membrane, in the same way that the red cell membrane encapsulates hemoglobin. Encapsulation of hemoglobin has been accomplished by introducing the molecule into a liposome. To the present this technology presents several problems, namely (1) poor efficiency of hemoglobin incorporation into the liposome; (2) physical instability of liposomes in storage; and (3) chemical instability of the lipid-hemoglobin interface leading to lipid peroxidation and hemoglobin denaturation. These factors translate in decreased oxygen carrying capacity, increased viscosity of the circulating blood/liposome mixture, and reticuloendothelial blockade. This situation notwithstanding, lipid encapsulation may ultimately provide the artificial blood of choice once a quality hemoglobin becomes consistently available and liposome production can be scaled up to quantities commensurate with demand.

137.13 Results in the Microcirculation with Blood Substitution with αα-Hemoglobin The effectiveness of αα-Hb (cell-free o-raffinose cross-linked) as a blood carrying plasma expander was tested in the microcirculation by implementing isovolemic hemodilution up to an exchange of 75% of the original cell mass. Exchanges with dextran 70 kDa molecular weight were used as control. Experiments were carried out in the microcirculation of the hamster window preparation, which was evaluated in terms of hemodynamic parameters and oxygen distribution at three successive levels of exchange. The oxygen delivery capacity of the microcirculation was determined in terms of the rate of arrival of hemoglobin (in red cell cells + molecular) to the capillaries. It was found that using molecular solutions of about 10 g/dl concentration, the oxygen delivery capacity of the microcirculation was identical to that of dextran 70 for substitutions up to 60% of the red blood cell mass. This was a consequence of the lack of increased cardiac output with the Hb solutions, which should be expected when viscosity is lowered, as is the case with colloidal solutions (starch and dextran). This hemoglobin caused hypertension, and at the highest level of exchange exhibited abnormally low tissue pO2, which fell from a normal value of 22.4 mmHg to about 5 mmHg. This lowered tissue pO2 was accompanied by a significant decrease in functional capillary density, and increased vessel wall gradients. Consequently, this hemoglobin does not appear to function as a tissue oxygenator in the microcirculation, although the total amount of hemoglobin and its capacity to transport oxygen is similar to that of natural blood. The principal reasons for this outcome are hypertension elicited by lowered blood viscosity with consequent decrease of production of NO, increased scavenging of NO, and increased availability of oxygen to the vessel wall, which promotes autoregulatory responses aimed at limiting over oxygenation of the vessel wall. The linkage between blood supply and oxygen demand is due to the existence of chemical vasodilator signals transmitted from the tissue cells to the resistance vessels, which relax (increase their diameter) causing blood flow to increase when tissue oxygen tension falls below critical levels. Conversely, increasing blood oxygen tension over normal levels elicits vasoconstriction.


The most critical side effect in terms of microvascular function is the finding that molecular hemoglobin solutions significantly lower functional capillary density.

137.14

Hemoglobin and Nitric Oxide (NO) Binding

The principal adverse effect of hemoglobin in solution is to produce vasoconstriction. This is based on two fundamental observations: (1) Hb has a very high affinity to bind NO at heme sites, and (2) Hb has been shown to produce constriction of isolated aortic rings. However, there is no direct proof that these mechanisms exists because NO has a very short half-life in the circulation (1.8 ms) [Liu et al., 1998], making its detection difficult. Further complicating the story is the observation that NO can also bind to hemoglobin at its sulfhydryl residues [Stamler et al., 1997], and that this mechanism might be oxygenlinked. It should also be noted that NO binding as a cause for vasoconstriction has been primarily deduced from studies that have shown constriction of isolated aortic rings when exposed to hemoglobin solutions. However, in vivo the situation is more complicated, particularly because most of the important diameter reductions occur in specific microvessels. Alternative explanations are that this effect can be produced directly on the blood side of the endothelium, where the presence of molecular hemoglobin in the red cell free plasma layer could significantly distort the diffusion field from the endothelial cell, diverting NO from smooth muscle into blood. Another effect could arise if the protein leaves the vascular space and enters the interstitium, there to interfere with the diffusion of NO from endothelium to its target, vascular smooth muscle. Thus, Rohlfs et al. [1997], found that hemoglobins with differing effects on mean arterial blood pressure all had similar reaction kinetics with NO. These findings are consistent with the autoregulation hypothesis, and suggest that there is more to the story than NO scavenging. A different explanation is that altered blood properties in extreme hemodilution affect the production of NO and prostacyclin (and other cytokines) in endothelial cells, which is dependent on the maintenance of some level of shear stress. Such shear stress, in a flowing system such as blood, is dependent on a number of factors, including the viscosity of the solution. Thus, it would be expected that hemoglobin solutions with low viscosity, such as αα-Hb, would decrease NO synthesis and induce vasospasm. In fact, vasoconstriction is probably a natural reaction of the organism to reduced blood viscosity. The difficulties associated with the measurement of NO directly in biological systems has led to the formulation of theoretical models. This type of analysis shows that NO distribution between blood and smooth muscle is the consequence of its high diffusivity, the fact that its half-life is limited by its reaction with oxygen and hemoglobin, its diffusion into red cells, and the geometry of vessels in the microcirculation. A tentative conclusion, in part supported by direct studies in the microcirculation, indicates that effects should be maximal in arterioles of the order of 30 to 100 µm in diameter. Thus far, the NO physiology has not been completely translated into the design of blood substitutes [Winslow, 1998]. The group at Somatogen have designed a series of hemoglobin mutants with a range of NO affinity and shown that there is a correlation between NO binding and the blood pressure response in rats [Doherty et al., 1998]. However, Rohlfs et al. [1998], have shown that a different series of chemically modified hemoglobins with different vasoactivities have essentially the same NO reactivity.

137.15 Rational Design of an Oxygen Carrying Molecular Plasma Expander The products presently in clinical trials reflect the physiological know-how of the of 1970s, when their effects could not be analyzed at the level of microscopic blood vessels, blood flow regulation by NO was unknown, the theory of microvascular autoregulation had just been formulated, and the endothelium was viewed as a passive cellular lining which prevented the extravasation of plasma proteins. At that time, systemic experimental findings and the clinical experience with low viscosity plasma expanders used in hemodilution established that lowering blood viscosity was a safe and efficacious method for restoring


blood volume, particularly if oncotic pressure was maintained within the physiological range through the use of colloids. These well-established tenets were incorporated into the design of artificial blood, or blood substitute, which was conceived as a fluid of low viscosity (i.e., lower than blood and closer to plasma in order to obtain the beneficial effects of hemodilution), moderate oncotic pressure (i.e., of the order of 25 mmHg), a right-shifted oxygen dissociation curve, and a concentration of hemoglobin, or equivalent oxygen carrying capacity, of 10 g Hb/dl. This scenario was firmly established because there were no practical methods for assessing tissue pO2 clinically, other than direct measurement of venous blood pO2, while in experimental conditions tissue pO2 could only be determined by extremely laborious and difficult to implement microelectrode methods. Thus, the distribution of oxygen in the tissue, even at the experimental level, was not known and related fundamental mechanisms directly involved with survival at the level of the microcirculation were equally unknown or misunderstood. High resolution methods for measuring tissue pO2 became available at the beginning of the 1990s, and when these were applied to the analysis of the effects of hemoglobin-based blood substitutes, they reversed several of the established principles. It was found that oxygen is delivered by the arterioles and not the capillaries, and that this rate of delivery was modulated by the consumption of oxygen in the arteriolar wall. Vasoconstrictor effects were found to increase the oxygen consumption of the arterioles to the detriment of tissue oxygenation. This was particularly true for hemoglobin solutions formulated with small molecules, such as the αα-crosslinked hemoglobin, which also presented another of the presumed required properties, namely low viscosity. To the present, the vasoconstrictor effect has been attributed to NO scavenging by hemoglobin; however, other mechanisms may be involved and superimposed on the scavenging effect. The molecular nature of the presently developed hemoglobin solutions determines that the oxygen source in blood is very close to the endothelium, since the barrier determined by the plasma layer is no longer present. This configuration increases the oxygen availability to the microvascular wall, potentially eliciting autoregulatory vasoconstriction aimed at regulating oxygen delivery. It is now apparent that right-shifted materials further increase oxygen availability, thus enhancing the vasoconstrictor stimulus. Concerning NO scavenging, it may well be that this shall remain an intrinsic property of the hemoglobin molecule dictated by the physical similarity of NO and O2 . In this context, two solutions to the problem may be implemented directly at the level of the oxygen carrier: One consists of producing an elevated (relative to normal) plasma viscosity, in such a fashion that shear stress dependant NO production remains constant or is elevated. Second, physically separating the hemoglobin molecule from the endothelial barrier as may be obtained with a pegylated material may retard hemoglobin–NO interaction. There is increasing evidence that low viscosity is of no benefit once the transfusion trigger is passed, particularly if combined with vasoconstrictor effects that impede the increase of cardiac output. In fact, conditions of extreme hemodilution, as those that obtain purely on the basis of viscosity considerations when the red blood cell mass is reduced, can only survive when plasma viscosity is increased, so that blood viscosity is near normal, which has the effect of maintaining functional capillary density at basal levels. This beneficial effect is further augmented by added production of shear stress dependant NO, which induces vasodilation and adequate blood flow. The physiological consequences of introducing molecular solutions of hemoglobin in the circulation can be seen in full perspective when the products are ranked according to size (Fig. 137.4), where it becomes apparent that side effects such as hypertension and retention time are a direct result of molecular dimensions, which in turn determine viscosity and diffusivity. In these conditions, tissue oxygenation is enhanced by the use of a left-shifted hemoglobin which effectively “hides” oxygen from the arterioles and facilitates its unloading in the regions of the microcirculation where oxygen consumption by the vessel wall is smaller. Finally, a critical issue is how much hemoglobin is necessary to obtain the desired effects. Experimental evidence presently indicates that when the “counter-intuitive” formulation of product is implemented, the actual amount of hemoglobin needed may be as low as 3 g Hb/dl (Winslow et al., 1998). This


FIGURE 137.4 The physiological reaction to the presence of even small amounts of circulating hemoglobin can be predicted by grouping the products according to molecular dimension. Systemic blood pressure rise is maximal for the smallest molecules and virtually absent for the largest molecules. The effect on blood viscosity correlates with both the rise of blood pressure and the maintenance of functional capillary density, where small molecules, which lower blood viscosity, also significantly lower FCD, while the larger molecules, causing blood viscosity to remain similar to that of whole blood, have no effect on FCD. Intravascular retention time is also directly proportional to molecular size, varying from about 12 h for the smaller molecules to more than 48 h for PEG-hemoglobin. Products are identified according to their manufacturer, namely: Baxter Healthcare, Round Lake, IL; Somatogen, Inc., Boulder, CO; Northfield Laboratories Inc., Chicago, IL; Hemosol Ltd., Toronto, Canada; Biopure Pharmaceuticals Corp., Boston, MA; Apex Bioscience Inc., Research Triangle Park, NC.; Enzon Inc., Piscataway, NJ; SANGART Inc., La Jolla, CA. (Data from Vandegriff et al., 1997.)

consideration impacts on the cost of the product and the potential for production, since most hemoglobin modifications have at most a 50% yield. This is critical for products based on human blood, where the principal source is outdated material, which would impose an inherent limitation on the total production capacity if the product is formulated with 7 to 10 g Hb/dl concentration.

137.16

Conclusions

Lethality consequent to blood losses results from hypovolemia and not anemia, and a number of plasma expanders exist that restitute volume and ensure survival even though they carry oxygen only to the extent of its plasma solubility. This is a consequence of the “derived” increase of oxygen carrying capacity resulting from lowered viscosity and increased cardiac output. This approach serves as hemoglobin concentration reaches the transfusion trigger, beyond which restoration of oxygen carrying capacity is needed. The use of oxygen carrying plasma expanders made with molecular solutions of hemoglobin, when the red blood cell mass is reduced below the transfusion trigger, causes extreme hemodilution associated with a significant reduction of blood viscosity and NO production, and reflex vasoconstriction, leading to decreased functional capillary density. This combination of events is difficult to survive because decreased NO availability increases the intrinsic oxygen consumption of the tissue. An oxygen carrying volume replacement with a capacity commensurate to blood is presently the perceived sine qua necessity for extreme blood losses. However, new findings indicate that this may not necessarily be the case. Recent developments in the understanding of the physiology of extreme hemodilution, and the physical events associated with the substitution of red blood cells with molecular hemoglobin solutions have


determined a shift in paradigm, indicating that a viable “artificial blood” will be obtained from a counterintuitive formulation of the product. In this formulation, viscosity is near normal for blood, the dissociation curve is left-shifted, oncotic pressure is high, and the concentration of hemoglobin is in the range of 3 to 5 g Hb/dl [Winslow and Intaglietta, 1998].

Acknowledgment This work was supported in part by USPHS Grant HLBI 48018.

Defining Terms Hemodilution: The replacement of natural blood with a compatible fluid that reduces the concentration of red blood cells Stroma-free hemoglobin: Hemoglobin derived from red blood cells where all materials related to cell membrane and other components within the red blood cells have been removed. Liposome: Microscopic phospholipid vesicle used to encapsulate materials for slow release. The use of liposomes to encapsulate hemoglobin is a departure from the conventional use of liposomes, exploiting encapsulation, and requires their modification to ensure sustained entrapment. Oncotic: Refers to colligative properties due to the presence of macromolecules, for instance oncotic pressure, which is differentiated from osmotic pressure, which is that due to the presence of all molecular species in solution. Oxygen carrying capacity: The total amount of oxygen that may be transported by blood or the fluid in the circulation. Differentiated from oxygen delivery capacity which involves considerations of flow rate. Plasma viscosity: Viscosity of blood devoid of cellular elements, a parameter that is critically impacted by the introduction of molecular hemoglobin solutions. Shear stress: Force per unit area parallel to the vessel wall or traction, experienced by the endothelium due to blood flow and blood viscosity. Transfusion trigger: Concentration of blood hemoglobin beyond which blood is required to restore circulatory function.

References Dintenfass, L. Blood Microrheology: Viscosity Factors in Blood Flow, Ischemia and Thrombosis. AppletonCentury-Crofts, New York, 1971. Doherty, D. H., Doyle, M. P., Curry, S. R., Vali, R. J., Fattor, T. J., Olson, J. S., and Lemon, D. D. Rate of reaction with nitric oxide determines the hypertensive effect of cell-free hemoglobin. Nature Biotechnology 16:672, 1998. Frangos, J. A., Eskin, S. G., McIntire, L. V., and Ives, C. L. Flow effects on prostacyclin production in cultured human endothelial cells. Science 227:1477, 1985. Intaglietta, M., Johnson, P. C., and Winslow, R. M. Microvascular and tissue oxygen distribution. Cardiovasc. Res. 32:632, 1996. Intaglietta, M. Whitaker Lecture 1996: Microcirculation, biomedical engineering and artificial blood. Ann. Biomed. Eng. 25:593, 1997. Johnson, P. C. Brief Review: Autoregulation of blood flow. Circ. Res. 59:483, 1996. Joyner, W. L., Davis, M. J., and Gilmore, J. P. Intravascular pressure distribution and dimensional analysis of microvessels in the hamsters with renovascular hypertension. Microvasc. Res. 22:1, 1974. Kerger, H., Saltzman, D. J., Menger, M. D., Messmer, K., and Intaglietta, M. Systemic and subcutaneous microvascular pO2 dissociation during 4-h hemorrhagic shock in conscious hamsters. Am. J. Physiol. 270:H827, 1996. Kjellstrom, B. T., Ortenwall, P., and Risberg, R. Comparison of oxidative metabolism in vitro in endothelial cells from different species and vessels. J. Cell. Physiol. 132:578, 1987. © 2000 by CRC Press LLC

Kuchan, M. J., Jo, H., and Frangos, J. A. Role of G proteins in shear stress-mediated nitric oxide production by endothelial cells. Am. J. Physiol. 267:C753, 1994. Lamy, M., Mathy-Hartert, M., and Deby-Dupont, G. Perfluorocarbons as oxygen carriers. In Update in Intensive Care Medicine 33, J.-L. Vincent, Ed., Springer-Verlag, Berlin, 1998, 332. Lipowsky, H. H. Mechanics of blood flow in the microcirculation. In Handbook of Bioengineering, R. Skalak and S. Chien, Eds., McGraw-Hill, New York, 1987, ch. 8. Liu, X., Miller, M. J., Joshi, M. S., Sadowska-Krowicka, H., Clark, D. A., and Lancaster, J. R. J. Diffusionlimited reaction of free nitric oxide with erythrocytes. J. Biol. Chem. 273:18709, 1998. Messmer, K., Sunder-Plasman, L., Klövekorn, W. P., and Holper, K. Circulatory significance of hemodilution: Rheological changes and limitations. Adv. Microcirculation, 4:1, 1972. Mirhashemi, S., Messmer, K., and Intaglietta, M. Tissue perfusion during normovolemic hemodilution investigated by a hydraulic model of the cardiovascular system. Int. J. Microcirc.: Clin. Exp. 6:123, 1987. Paller, M. S. Hemoglobin and myoglobin-induced acute renal failure: Role of iron nephrotoxicity. Am. J. Physiol. 255:F539, 1988. Quemada, D. Rheology of concentrated dispersed systems: III. General features of the proposed nonNewtonian model: Comparison with experimental data. Rheol. Acta 17:643, 1978. Richardson, T. Q. and Guyton, A.C. Effects of polycythemia and anemia on cardiac output and other circulatory factors. Am. J. Physiol. 197:1167, 1959. Rohlfs, R., Vandegriff, K., and Winslow, R. The reaction of nitric oxide with cell-free hemoglobin based oxygen carriers: Physiological implications. In: Industrial Opportunities and Medical Challenges, R. M. Winslow, K. D. Vandegriff, M. Intaglietta, Eds., Birkhäuser, Boston, 1997, 298. Rohlfs, R. J., Bruner, E., Chiu, A., Gonzales, A., Gonzales, M. L., Magde, D., Magde, M. D. J., Vandegriff, K. D., and Winslow, R. M. Arterial blood pressure responses to cell-free hemoglobin solutions and the reaction with nitric oxide. J. Biol. Chem. 273:12128, 1998. Saltzman, D., DeLano, F. A., and Schmid-Schönbein, G. W. The microvasculature in skeletal muscle. VI. Adrenergic innervation of arterioles in normotensive and spontaneously hypertensive rats. Microvasc. Res. 44:263, 1992. Shen, W., Hintze, T. H., and Wolin, M. S. Nitric Oxide. An important signaling mechanism between vascular endothelium and parenchymal cells in the regulation of oxygen consumption. Circulation 92:3505, 1995. Stamler, J. S., Jia, L., Eu, J. P., McMahon, T. J., Demchenko, I. T., Bonaventura, J., Gernert, K., and Piantadosi, C. A. Blood flow regulation by S-nitrosohemoglobin in the physiological oxygen gradient. Science 276:2034, 1997. Torres Filho, I. P. and Intaglietta, M. Micro vessel pO2 measurements by phosphorescence decay method. Am. J. Physiol. 265(34):H1434, 1994. Tsai, A. G., Friesenecker, B., Mazzoni, M. C., Kerger, H., Buerk, D. G., Johnson, P. C., and Intaglietta, M. Microvascular and tissue oxygen gradients in the rat mesentery. Proc. Natl. Acad. Sci. U.S.A. 95:6590, 1998. Tsai, A. G., Friesenecker, B., McCarthy, M., Sakai, H., Intaglietta, M. Plasma viscosity regulates capillary perfusion during extreme hemodilution in hamster skin fold mode. Am. J. Physiol., 275:H2170, 1998. Vandegriff, K. D., Rohlfs, R. J., and Winslow, R. M. Colloid osmotic effects of hemoglobin-based oxygen carriers. In Advances in Blood Substitutes. Industrial Opportunities and Medical Challenges, R. M. Winslow, K. D. Vandegriff, and M. Intaglietta, Eds., Birkhäuser, Boston, MA, 1997. Wilson, D. F. Measuring oxygen using oxygen dependent quenching of phosphorescence: a status report. Adv. Exp. Med. Biol. 333:225, 1993. Winslow, R. M. Blood substitutes. Sci. Med. 4:54, 1997. Winslow, R. M. and Intaglietta, M. U.S. Patent 5,814,601. Methods and compositions for optimization of oxygen transport by cell free systems, 1998. Winslow, R. M. Artificial blood: Ancient dream, modern enigma. Nature Biotechnology. 16:621, 1998.


Further Information Blood Substitutes. Physiological Basis of Efficacy, R. M. Winslow, K. D. Vandegriff, and M. Intaglietta, Eds., Birkhäuser, Boston, 1995. Blood Substitutes. New Challenges, R. M. Winslow, K. D. Vandegriff, and M. Intaglietta, Eds., Birkhäuser, Boston, 1996. Advances in Blood Substitutes. Industrial Opportunities and Medical Challenges, R. M. Winslow, K. D. Vandegriff, and M. Intaglietta, Eds., Birkhäuser, Boston, 1997.


Yannas, I. V. “Artificial Skin and Dermal Equivalents.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

138 Artificial Skin and Dermal Equivalents 138.1 138.2 138.3 138.4

The Vital Functions of Skin Current Treatment of Massive Skin Loss Two Conceptual Stages in the Treatment of Massive Skin Loss by Use of the Artificial Skin Design Principles for a Permanent Skin Replacement Stage 1 Design Parameters • Stage 2 Design Parameters

138.5

Clinical Studies • Clinical Parameters Used in the Evaluation • Short-Term Clinical Evaluation of Artificial Skin • Long-Term Clinical Evaluation of Artificial Skin • Clinical Results • Summary

Ioannis V. Yannas Massachusetts Institute of Technology

Clinical Studies of a Permanent Skin Replacement (Artificial Skin)

138.6

Alternative Approaches: Cultured Epithelial Autografts (CEA) and Skin Equivalents (SE)

138.1 The Vital Functions of Skin Skin is a vital organ, in the sense that loss of a substantial fraction of its mass immediately threatens the life of the individual. Such loss can result suddenly, either from fire or from a mechanical accident. Loss of skin can also occur in a chronic manner, as in skin ulcers. Irrespective of the time scale over which skin loss is incurred, the resulting deficit is considered life threatening primarily for two reasons: Skin is a barrier to loss of water and electrolytes from the body, and it is a barrier to infection from airborne organisms. A substantial deficit in the integrity of skin leaves the individual unprotected either from shock, the result of excessive loss of water and electrolytes, or from sepsis, the result of a massive systemic infection. It has been reported that burns alone account for 2,150,000 procedures every year in the United States. Of these, 150,000 refer to individuals who are hospitalized, and as many as 10,000 die. Four types of tissue can be distinguished clearly in normal skin. The epidermis, outside, is a 0.1-mmthick sheet, comprising about 10 layers of keratinocytes at levels of maturation which increase from the inside out. The dermis, inside, is a 2–5-mm-thick layer of vascularized and innervated connective tissue with very few cells, mostly quiescent fibroblasts. The dermis is a massive tissue, accounting for 15–20% of total body weight. Interleaved between the epidermis and the dermis is the basement membrane, an approximately 20–nm-thick multilayered membrane (Fig. 138.1). A fourth layer, the subcutis, underneath the dermis and 0.4–4-mm in thickness, comprises primarily fat tissue. In addition to these basic structural elements, skin contains several appendages (adnexa), including hair follicles, sweat glands, and sebaceous glands. The latter are mostly embedded in the dermis, although they are ensheathed in layers of epidermal tissue.


FIGURE 138.1 Schematic view of skin which highlights the epidermis, the basement membrane interleaved between the epidermis and the dermis, and the dermis underneath. Only a small fraction of the thickness of the dermis is shown. (Redrawn with permission from J. Darnell, J. Lodish, and D. Baltimore, Molecular Cell Biology, Scientific American Books, New York, Chapter 5, Fig. 552, 1986.)

The functions of skin are quite diverse, although it can be argued that the single function of skin is to provide a viable interface with the individual’s environment. In addition to the specific vital functions mentioned above (protection from water and electrolyte loss, and from infection), skin also protects from heat and cold, mechanical friction, chemicals, and from UV radiation. Skin is responsible for a substantial part of the thermoregulatory and communication needs of the body, including the transduction of signals from the environment such as touch, pressure, and temperature. Further, skin transmits important emotional signals to the environment, such as paleness or blushing of the face and the emission of scents (pheromones). Far from being a passive membrane that keeps the internal organs in shape, skin is a complex organ.

138.2 Current Treatment of Massive Skin Loss The treatment of skin loss has traditionally focused on the design of a temporary wound closure. Attempts to cover wounds and servere burns have been reported from historical sources at least as far back as 1500 B.C., and a very large number of temporary wound dressings have been designed. These include membranes or sheets fabricated from natural and synthetic polymers, skin grafts from human cadavers (homografts, or allografts), and skin grafts from animals (heterografts, or xenografts). Although a satisfactory temporary dressing helps to stem the tide, it does not provide a permanent cover. Polymeric membranes which lack specific biologic activity, such as synthetic polymeric hydrogels, have to be removed after several days due to incidence of infection and lack of formation of physiologic structures. Patients with cadaver allografts and xenografts are frequently immunosuppressed to avoid rejection; however, this is a stop-gap operation which is eventually terminated by removal of the graft after several


FIGURE 138.2 Comparision between treatment with the meshed autograft (R) and treatment with the artificial skin (L). Autograft is usually meshed before grafting; scar forms in areas coinciding with the open slits of the autograft. The artificial skin treatment consists of grafting the excised wound bed with a skin regeneration template, followed by grafting on about day 14 with a very thin epidermal autograft. (Photo courtesy of J.F. Burke.)

days. In all cases where temporary dressings have been used, the routine result has been an open wound. Temporary dressings are useful in delaying the time at which a permanent graft, such as an autograft, is necessary and are therefore invaluable aids in the management of the massively injured patient. The use of an autograft has clearly shown the advantages of a permanent wound cover. This treatment addresses not only the urgent needs but also the long-term needs of the patient with massive skin loss. The result of treatment of a third-degree burn with a split-thickness autograft is an almost fully functional skin which has become incorporated into the patient’s body and will remain functional over a lifetime. Autografts usually lack hair follicles and certain adnexa as well. However, the major price paid is the removal of the split thickness graft from an intact area of the patient’s body: The remaining dermis eventually becomes epithelialized but not without synthesis of scar over the entire area of trauma (donor site). To alleviate the problem associated with the limited availability of autograft, surgeons have resorted to meshing, a procedure in which the sheet autograft is passed through an apparatus which cuts slits into the sheet autograft, allowing the expansion of the graft by several times and thereby extending greatly the area of use. An inevitable long-term result of use of these meshed autografts is scar synthesis in areas coinciding with the open slits of the meshed graft and a resulting pattern of scar which greatly reduces the value of the resulting new organ (Fig. 138.2). An important aspect of the use of the autograft is the requirement for early excision of dead tissue and the provision, thereby, of a viable wound bed to “take” the autograft. The term “artificial skin” has been used to describe a cell-free membrane comprising a highly porous graft copolymer of type I collagen and chondroitin 6-sulfate which degrades at a specific rate in the wound and regenerates the dermis in dermis-free wounds in animal models and patients (see below: dermis regeneration template, DRT). “Skin equivalent” (SE) refers to a collagen lattice which has been prepared by contraction of a collagen gel by heterologous fibroblasts (“dermal equivalent” or DE) and has subsequently been overlayed with a keratinocyte culture to induce formation of a mature, cornified epidermis in vitro prior to grafting of skin wounds. Cultured epithelial autografts (CEA) consist of a mature, cornified epidermis which has been produced by culturing keratinocytes in vitro, prior to grafting on skin wounds. The major goal of these treatments has been to replace definitively the use of the autograft in the treatment of patients with massive skin loss.


138.3 Two Conceptual Stages in the Treatment of Massive Skin Loss by Use of the Artificial Skin Loss of the epidermis alone can result from a relatively mild burn, such as an early exposure to the sun. Controlled loss of epidermis in a laboratory experiment with an animal can result from the repeated use of adhesive tape to peel off the keratinocyte layers. In either case, the long-term outcome is an apparently faithful regeneration of the epidermis by migration of epithelial cells from the wound edge, and from roots of hair follicles, over the underlying basement membrane and dermis. It has been shown that the epidermis can regenerate spontaneously provided there is a dermal substrate over which epithelial migration and eventual anchoring to the underlying connective tissue can occur. Loss of the dermis has quite a different outcome. Once lost, the dermis does not regenerate spontaneously. Instead, the wound closes by contraction of the wound edges toward the center of the skin deficit, and by synthesis of scar. Scar is a distinctly different type of connective tissue than is dermis. The depth of skin loss is, therefore, a critical parameter in the design of a treatment for a patient who has a skin deficit. In the treatment of burns, physicians distinguish among a first-degree burn (loss of epidermis alone), a second-degree burn (loss of epidermis and a fraction of the thickness of the dermis), and a thirddegree burn (loss of the epidermis and the entire dermis down to muscle tissue). A similar classification, based on depth of loss, is frequently applied to mechanical wounds, such as abrasion. The area of skin which has been destroyed needs also to be specified in order to assess the clinical status of the patient. A massively injured patient, such as a patient with about 30% body surface area or more destroyed from fire through the full thickness of skin, presents an urgent problem to the clinician, since the open wound is an ongoing threat to survival. A large number of temporary wound coverings have been used to help the patient survive through this period while waiting for availability of autografts which provide permanent cover. If autograft is unavailable over a prolonged period while the patient has survived the severe trauma, contraction and scar synthesis occur over extensive areas. In the long run, the patient therefore has to cope with deep, disfiguring scars or with crippling contractures. Thus, even though the patient has been able to survive the massive trauma and has walked out of the clinic, the permanent loss of skin which has been sustained prevents, in many cases, resumption of an active, normal life.

138.4 Design Principles for a Permanent Skin Replacement The analysis of the plight of the patient who has suffered extensive skin loss, presented above, leads logically to a wound cover which treats the problem in two stages. Stage 1 is the early phase of the clinical experience, one in which protection against severe fluid loss and against massive infection are defined as the major design objectives. Stage 2 is the ensuing phase, one in which the patient needs protection principally against disfiguring scars and crippling contractures. Even though the conceptual part of the design is separated in two stages for purposes of clarity, the actual treatment is to be delivered continuously, as will be become clear below. The sequential utilization of features inherent in stages 1 and 2 in a single device can be ensured by designing the graft as a bilayer membrane (Fig. 138.3). In this approach, the top layer incorporates the features of a stage 1 device, while the bottom layer delivers the performance expected from a stage 2 device. The top layer is subject to disposal after a period of about 10–15 days, during which time the bottom layer has already induced substantial synthesis of new dermis. Following removal of the top layer, the epidermal cover is provided either by covering with a thin epidermal graft or by modifying the device (cell seeding) so that an epidermis forms spontaneously by about 2 weeks after grafting.

Stage 1 Design Parameters The overriding design requirement at this stage is based on the observation that air pockets (“dead space”) at the graft-wound bed interface readily become sites of bacterial proliferation. Such sites can be prevented


FIGURE 138.3 Schematic of the bilayer membrane which has become known as the artificial skin. The top layer is a silicone film which controls moisture flux through the wound bed to nearly physiologic levels, controls infection of the wound bed by airborne bacteria, and is strong enough to be sutured on the wound bed. The bottom layer is the skin regeneration template, which consists of a graft copolymer of type I collagen and chondroitin 6-sulfate, with critically controlled porosity and degradation rate. About 14 days after grafting, the silicone layer is removed and replaced with a thin epidermal autograft. The bottom layer induces synthesis of a nearly physiologic dermis and eventually is removed completely by biodegradation. (From Yannas IV, Burke JF, Orgill DP, et al. 1982. Science 215:74.)

from forming if the graft surface wets, in the physicochemical sense, the surface of the wound bed on contact and thereby displaces the air from the graft-tissue interface (Fig. 138.4). It follows that the physicochemical properties of the graft must be designed to ensure that this leading requirement is met, not only when the graft is placed on the wound bed but for several days thereafter, until the function of the graft has moved clearly into its stage 2, in which case the graft-wound bed interface has been synthesized de novo and the threat of dead space has been thereby eliminated indefinitely. First, the flexural rigidity of the graft, i.e., the product of Young’s modulus and moment of inertia of a model elastic beam, must be sufficiently low to provide for a flexible graft which drapes intimately over a geometrically nonuniform wound bed surface and thus ensures that the two surfaces will be closely apposed. In practice, these requirements can be met simply by adjusting both the stiffness in tension and the thickness of the graft to appropriately low values. Second, the graft will wet the wound bed if the surface energy of the graft-wound bed interface is lower than that of the air-wound bed surface, so that the graft can adequately displace air pockets from the air-wound bed surface. Although the measurement of a credible value of the surface energy is not a simple matter when the graft is based on certain natural polymers in the form of a hydrated gel, the requirement of adequate adhesion can be met empirically by chemical modification of the surface or by proper use of structural features such as porosity. Third, the moisture flux through the graft must be maintained within bounds which are set by the following considerations. The upper bound to the moisture flux must be kept below the level where excessive dehydration of the graft occurs, thereby leading to alteration of the surface energy of the graftwound bed interface and loss of the adhesive bond between graft and wound bed. Further, when the moisture flux exceeds the desired level, the graft is desiccated, and shrinkage stesses develop which pull


FIGURE138.4 Certain physicochemical and mechanical requirements in the design of an effective closure for a wound bed with full-thickness skin loss. (a) The graft (cross-hatched) does not displace air pockets (arrows) efficiently from the graft-wound bed interface. (b) Flexural rigidity of the graft is excessive. The graft does not deform sufficiently, under its own weight and the action of surface forces, to make good contact with depressions on the surface of the wound bed; as a result, air pockets form (arrows). (c) Shear stresses τ (arrows) cause buckling of the graft, rupture of the graft-wound bed bond and formation of an air pocket. (d) Peeling force P lifts the graft away from the wound bed. (e) Excessively high moisture flux rate J through the graft causes dehydration and development of shrinkage stresses at the edges (arrows), which cause lift-off away from the wound bed. (f ) Very low moisture flux J causes fluid accumulation (edema) at the graft-wound bed interface and peeling off (arrows). (From Yannas IV, Burke JF. 1980. J Biomed Mater Res 14:65.)

the graft away from the wound bed. An estimate of the maximum normal stress σm can be obtained by modeling the desiccating graft in one dimension as a shrinking elastic beam bonded to a rigid surface

(

)

σ m = 0.45α V2 − V1 E

(138.1)

In Eq. (138.1), α is the coefficient of expansion of a graft which swells in water, V1 and V2 are initial and final values of the volume fraction of moisture in the graft, and E is Young’s modulus of the graft averaged over the range V1 to V2, the latter range being presumed to be narrow. If, by contrast, the moisture flux through the graft is lower than the desired low bound, water accumulates between the graft and the wound bed, and edema results with accompanying loss of the adhesive bond between the two surfaces.

Stage 2 Design Parameters The leading design objectives in this stage are two: synthesis of new, physiologic skin and the eventual disposal of the graft. The lifetime of the graft, expressed as the time constant of biodegradation tb , was modeled in relation to the time constant for normal healing of a skin incision th . The latter is about 25 days. In preliminary studies with animals, it was observed that when matrices were synthesized to degrade at a very rapid rate, amounting to tb th, the initially insoluble matrix was reduced early to a liquidlike state, which was incompatible with an effective wound closure. At the other extreme, matrices were synthesized which degraded with exceptional difficulty within 3–4 weeks, compatible with tb th . In these preliminary studies it was observed that a highly intractable matrix, corresponding to the latter condition, led to formation of a dense fibrotic tissue underneath the graft which eventually led to loss of the bond between graft and wound bed. Accordingly, it was hypothesized that a rule of isomorphous matrix replacement, equivalent to assuming a graft degradation rate of order of magnitude similar to the synthesis rate for new tissue, and represented by the relation


tb =1 th

(138.2)

would be optimal. Control of tb is possible by adjustment of the crosslink density of the matrix. Equation (138.2) is the defining equation for a biodegradable scaffold which is coupled with, and therefore interacts with, the inflammatory process in a wound. Migration of cells into the matrix is necessary for synthesis of new tissue. Such migration can proceed very slowly, defeating Eq. (138.2), when fibroblasts and other cells recruited below the wound surface are required to wait until degradation of a potentially solid-like matrix has progressed sufficiently. An easier pathway to migrating cells can be provided by modifying a solid-like matrix into one which has an abundance of pore channels, where the average pore is at least as large as one cell diameter (about 10 µm) for ready access. Although this rationale is supported by experiment, results with animal studies have shown that not only is there a lower limit to the average pore diameter, but there is also an upper limit (see below). Migration of cells into the porous graft can proceed only if nutrients are available to these cells. Two general mechanisms are available for transport of nutrients to the migrating cells, namely, diffusion from the wound bed and transport along capillaries which may have sprouted within the matrix (angiogenesis). Since capillaries would not be expected to form for at least a few days, it is necessary to consider whether a purely diffusional mode of transport of nutrients from the wound bed surface into the graft could immediately supply the metabolic needs of the invading cells adequately. The cell has been modled as a reactor which consumes a critical nutrient with a rate r, in units of mole/cm3/s; the nutrient is transported from the wound bed to the cell by diffusion over a distance l, the nutrient concentration at or near the surface of the wound bed is c0, in units of mole/cm3, and the diffusivity of the nutrient is D, in cm2/s. The appropriate conditions were expressed in terms of a dimensionless number S, the cell lifeline number, which expresses the relative importance of reaction rate for consumption of the nutrient by the cell to rate of transport of the nutrient by diffusion alone:

S=

rl 2 Dc0

(138.3)

Eq. (138.3) suggests that when S = 1, the critical value of the path length, lc , corresponds to the maximum distance along which cells can migrate inside the graft without requiring angiogenesis (vascularization) for nutrient transport. The value of lc defines the maximum thickness of graft that can be populated with cells within a few hours after grafting, before angiogenesis has had time to occur. These conceptual objectives have been partially met by designing the graft as an analog of extracellular matrix (ECM) which possesses morphogenetic activity since it leads to partial regeneration of dermis. The discovery of the specific ECM analog that possesses this activity has been based on the empirical observation that, whereas the vast majority of ECM analogs apparently do not inhibit wound contraction almost at all, one of the analogs does. The activity of this analog, for which the term regeneration template has been coined, is conveniently detected as a significant delay in the onset of wound contraction. When seeded with (uncultured) autologous keratinocytes, an active regeneration template is capable of inducing simultaneous synthesis both of a dermis and an epidermis in the guinea pig and in the swine (Yorkshire pig). The regeneration is almost complete; however, hair follicles and other skin adnexa are not formed. The resulting integument performs the two vital functions of skin, i.e., control of infection and moisture loss, while also providing physiologic mechanical protection to the internal organs and, additionally, providing a cosmetic effect almost identical to that of intact skin. The morphogenetic specificity of the dermis regeneration template depends sensitively on retention of certain structural characteristics. The overall structure is that of an insoluble, three-dimensional covalently crosslinked network. The primary structure can be described as that of a graft-copolymer of type I collagen and a glycosaminoglycan (GAG) in the approximate ratio 98/2. The GAG can be either © 2000 by CRC Press LLC

chondroitin 6-sulfate or dermatan sulfate; other GAGs appear capable of contributing approximately equal increments to morphogenetic specificity. The collagen fibers lack banding almost completely although the integrity of the triple helical structure is retained through the network. The resistance of the network to collagenase degradation is such that approximately two-thirds of the mass of the network becomes solubilized in vivo within about 2 weeks. The structure of the network is highly porous. The pore volume fraction exceeds 95% while the average pore diameter is maintained in the range 20–125 µm. The regeneration template loses its activity rapidly when these structural features are flawed deliberately in control studies. The dermis regeneration template, a porous matrix unseeded with cells, induces synthesis of a new dermis and solves this old surgical problem. Simultaneous synthesis of a new, confluent epidermis occurs by migration of epithelial cell sheets from the wound edges, over the newly synthesized dermal bed. With wounds of relatively small characteristic dimension, e.g., 1 cm, epithelial cells migrating at speeds of about 0.5 mm/day from each wound edge can provide a confluent epidermis within 10 days. In such cases, the unseeded template fulfills all the design specifications set above. However, the wounds incurred by a massively burned patient are typically of characteristic dimension of several centimeters, often more than 20–30 cm. These wounds are large enough to preclude formation of a new epidermis by cell migration alone within a clinically acceptable timeframe, say 2 weeks. Wounds of that magnitude can be treated by seeding the porous collagen-GAG template, before grafting, with at lest 5 × 104 keratinocytes per cm2 wound area. These uncultured, autologous cells are extracted by applying a cell separation procedure, based on controlled trypsinization, to a small epidermal biopsy. Details of the synthesis of the dermis regeneration template, as well as of other templates which regenerate peripheral nerves and the knee meniscus, are presented elsewhere in this Handbook (see Chapter 13). The dermis regeneration template described in this section was first reported as a synthetic skin and as an artificial skin.

138.5 Clinical Studies of a Permanent Skin Replacement (Artificial Skin) Clinical Studies The skin regeneration template has been tested clinically on two occasions. In the first, conducted in the period 1979–1980, one clinical center was involved, and 10 severely burned patients were studied. In the second, conducted during 1986–1987, 11 clinical centers were involved, and 106 severely burned patients were treated in a prospective, randomized manner. In each case the results have been published in some detail. The second study led to a surgical report, a histologic report, and an immunologic report. There is now adequate information available to discuss the advantages and disadvantages of this prototype artificial skin in the treatment of the severely burned patient. The artificial skin used in clinical studies so far consists of the bilayer device illustrated in Fig. 138.3. The outer layer is a silicone film, about 100 µm in thickness, which fulfills the requirements of stage 1 of the design (see above), and the inner layer is the skin regeneration template. In these clinical studies this device has not been seeded with keratinocytes. Closure of the relatively large wounds by formation of an epidermis has been achieved instead by use of a 100-µm-thin layer of the patient’s epidermis (autoepidermal graft). The latter has been excised from an intact area of the patient’s skin; the donor site can, however, be harvested repeatedly, since the excised epidermis regenerates spontaneously in a few days over the relatively intact dermal bed. Briefly, the entire procedure consists of preparation of the wound bed prior to grafting by excision of thermally injured tissue (eschar), followed by immediate grafting of the unseeded template on the freshly excised wound and ending, 3 weeks later, by replacing the outer, silicone layer of the device with a thin epidermal graft. The results of studies with a guinea pig model and a swine model have shown that seeding of the dermis regeneration template with fresh, uncultured autologous keratinocytes prior to grafting leads to simultaneous synthesis of an epidermis as


well as a dermis in about 2 weeks. However, definitive clinical studies of the keratinocyte-seeded template have yet to be performed. The discussion below focuses on the advantages and disadvantages of the (unseeded) artificial skin, as these emerge from clinical observations during the treatment as well as from a limited number of follow-up observations extending over several years after the treatment. The controls used in the clinical studies included meshed autograft, allograft, and xenografts. Comparative analysis of the clinical data will focus on each of the two stages of treatment for the massively burned patient, i.e., the early (acute) stage and the long term stage, the conceptual basis for which has been discussed above.

Clinical Parameters Used in the Evaluation The clinical parameters during the early stage of treatment (about 3 weeks) include the take of the graft, expressed as a percentage of graft area which formed an adhesive bond of sufficient strength with the wound bed and became vascularized. In the case of the artificial skin treatment, two different measures of take are reported, namely, that of the bilayer membrane on the freshly excised wound bed and the take of the epidermal graft applied on the neodermal bed about 3 weeks later. Another parameter is the thickness of dermis that has been excised from the donor site in order to obtain the autograft that is used to close the wound definitively. An additional parameter which characterizes the cost of the donor site to the patient is the time to heal the donor site. The surgeon’s overall qualitative evaluation of the treatment (relative to controls) during the early stage is also reported. The long-term evaluation has extended at least 1 year in approximately one-quarter of the patients. The first long-term parameter is based on the patients’ reports of the relative incidence of nonphysiologic sensations, including itching, dryness, scaliness, lack of elasticity (lack of deformability), sweating, sensation, and erythema. The second parameter is based on the physicians’ report of the relative presence of hypertrophic scarring in the grafted area. A third parameter is the patient’s evaluation of the physiologic feel and appearance of the donor sites. Finally, there is an overall evaluation and preference of the patients for a given grafted site as well as the physicians’ evaluation of the same grafted site.

Short-Term Clinical Evaluation of Artificial Skin The median percentage take of the artificial skin was 80%, compared with the median take of 95% for all controls. Use of the Wilcoxin Rank Sum Test for the bimodally distributed data led to the conclusion that the take of the artificial skin was lower than that of all controls with a p value of 0.10). The take of the epidermal autograft was 86%. Mean donor site thickness was 0.325 ± 0.045mm for control sites and only 0.15 ± 0.0625mm for epidermal grafts which were harvested for placement over the newly synthesized dermis; the difference was found to be significant by t test with a p value of < 0.001. The thinner donor sites used in the artificial skin procedures healed, as expected, significantly faster, requiring 10.6 ± 5.8 d compared to 14.3 ± 6.9 d for control sites, with a p value of < 0.001 by t test. It is worth noting that donor sites used in the artificial skin procedure were frequently reharvested sites from previous autografting; reharvested donor sites healed more slowly than primary sites. The subjective evaluation of the operating surgeons at the conclusion of the acute stage of treatment was a response to the question, “Was artificial dermis (artificial skin) advantageous in the management of this particular patient?” Sixty-three percent of the comments were affirmative, whereas in 36% of the responses, the acute (early) results were believed to be no better than by use of routine methods. The physicians who responded positively to the use of artificial skin commented on the ability to use thin donor sites that healed quickly, relative to the thicker donor sites which were harvested in preparation for an autograft. Positive comments also cited the handling characteristics of the artificial skin relative to the allograft as well as the ability to close the wound without fear of rejection while awaiting healing


of the donor site. Negative comments included a less-than-adequate drainage of serum and blood through the unperforated silicone sheet, the seemingly poor resistance of the artificial skin to infection, and the need for a second operation.

Long-Term Clinical Evaluation of Artificial Skin One year after treatment, the allografted and xenografted sites had been long ago covered definitively with autograft; therefore, the experimental sites included, in the long term, either autografts (referred to occasionally as controls below) or the test sites, comprising the new integument induced as a result of treatment with the artificial skin (partially regenerated dermis closed with an epidermal graft). The patients reported that itching was significantly less (Wilcoxin Rank Sum test p < 0.02) in the artificial skin site than in control sites. Dryness, scaliness, elasticity (deformability), sweating, sensation, and erythema were similar at both control and artificial skin sites. Hypertrophic scarring was reported to be less in artificial skin in 42% of sites and was reported to be equivalent on test and control sites 57% of the time. No patient reported that the artificial skin sites had more hypertrophic scar than the autografted sites. Even though donor sites that were used during treatment with the artificial skin were harvested repeatedly (recropping), 72% of patients reported that these artificial skin donor sites felt “more normal,” 17% felt that there was no difference, and 11% felt that the control donor site was “more normal.” The results of the histologic study on this patient population showed that, in sites where the artificial skin was used, an intact dermis was synthesized as well as definitive closure of a complete epidermal layer with a minimum of scarring had occurred. The results of the immunologic study led to the conclusion that, in patients who had been treated with the artificial skin, there was increased antibody activity to bovine skin collagen, bovine skin collagen with chondroitin sulfate, and human collagen; however, it was concluded that these increased levels of antibodies were not immunologically significant. The overall evaluation by the patients showed that 26% preferred the new integument generated by use of the artificial skin whereas 64% found that the sites were equivalent and 10% showed preference for the autografted site. Physician’s overall evaluation showed that 39% preferred the artificial skin site, 45% found the sites to be equivalent, and 16% preferred the autografted site.

Clinical Results The take of the artificial skin was comparable to all other grafts and was inferior only to the meshed autograft. The latter showed superior take in part because meshing reduces drastically the flexural rigidity of the graft (see above) leading thereby to greater conformity with the wound bed (see Fig. 138.4). The interstices in the meshed autograft also provided an outlet for drainage of serum and blood from the wound, thereby allowing removal of these fluids. By contrast, the continuity of the silicone sheet in the artificial skin accounted for the increased flexural rigidity of the graft and prevented drainage of wound fluids with a resulting increased incidence of fluid accumulation underneath the graft. Fluid accumulation was probably the cause of the reduced take of the artificial skin, since immediate formation of a physiochemical bond between the graft and the wound bed was thereby prevented (see Fig. 138.4). The development of infection underneath the artificial skin, noted by physicians in certain cases, can also be explained as originating in the layer of wound fluid which presumptively collected underneath the artificial skin. This analysis suggests that meshing of the silicone layer of the artificial skin, without affecting the continuity of the collagen-GAG layer, could lead to improved take and probably to reduced incidence of infection. The healing time for donor sites associated with use of the artificial skin was shorter by about 4 days than for donor sites that were used to harvest autograft. An even shorter healing time for donor sites for artificial skin can be realized by reducing the thickness of the epidermal graft which is required to close the dermal bed. The average epidermal graft thickness reported in this study, 0.15mm, was significantly higher than thicknesses in the range 0.05–0.07mm, corresponding to an epidermal graft with adequate continuity but negligible amount of attached dermis. Increasing familiarity of surgeons with the procedure for harvesting these thin epidermal grafts is expected to lead to harvesting of thinner grafts in future studies. The importance of harvesting a thin graft cannot be overestimated, since the healing time of the © 2000 by CRC Press LLC

donor site decreases rapidly with decreasing thickness of harvested graft. It has been reported that the mean healing time for donor sites for the artificial skin reported in this study, about 11 days, is reduced to 4 days provided that a pure epidermal graft, free of dermis, can be harvested. Not only does the time to heal increase, but the incidence of hypertrophic scarring at a donor site also increases with the thickness of the harvested graft. This observation explains the higher incidence of hypertrophic scarring in donor sites associated with harvesting of autografts, since the latter were thicker by about 0.175mm than the epidermal grafts used with the artificial skin. An additional advantage associated with use of a thin epidermal graft is the opportunity to reharvest (recropping) within a few days; this reflects the ability of epithelial tissues to regenerate spontaneously provided there is an underlying dermal bed. When frequent recropping of donor graft is possible, the surface area of a patient that can be grafted within a clinically acceptable period increases rapidly. In the clinical study described here, a patient with deep burns over as much as 85% body surface area was covered with artificial skin grafts for 75 days while the few donor sites remaining were being harvested several times each. In the long term rarely did a patient or a physician in this clinical study prefer the new skin provided by the autograft to that provided by the artificial skin treatment. This result is clearly related to the use of meshed autografts, a standard procedure in the treatment of massively burned patients. Meshing increases the wound area which can by autografted by between 1.5 and 6 times, thereby alleviating a serious resource problem. However, meshing destroys the dermis as well as the epidermis; although the epidermis regenerates spontaneously and fills in the defects, the dermis does not. The long-term result is a skin site with the meshed pattern permanently embossed on it. The artificial skin is a device that, in principle, is available in unlimited quantity; accordingly, it does not suffer from this problem (Fig. 138.2). The result is a smooth skin surface which was clearly preferred on average by patients and physicians alike. It has been established that the artificial skin regenerates the dermis and, therefore, its use leads to complete inhibition of scar formation in full-thickness skin wounds. The regeneration is partial because skin adenexa (hair follicles, sweat glands) are not recovered. The results of studies of the mechanism by which the artificial skin regenerates the dermis in full-thickness skin wounds in animal models have been described elsewhere (see Chapter 113, Regeneration Templates).

Summary The artificial skin leads to a new skin which appears closer to the patient’s intact skin than does the meshed autograft. Take of the artificial skin is as good as all comparative materials except for the unmeshed autograft, which is superior in this respect. Donor sites associated with the artificial skin treatment heal faster, can be recropped much more frequently, and eventually heal to produce sites that look closer to the patient’s intact skin than do donor sites harvested for the purpose of autografting. In comparison to the allograft, the artificial skin is easier to use, has the same take, does not get rejected, and is free of the risk of viral infection associated with use of allograft.

138.6 Alternative Approaches: Cultured Epithelial Autografts (CEA) and Skin Equivalents (SE) The use of cultured epithelial autografts has been studied clinically. In this approach autologous epidermal cells are removed by biopsy and are then cultured in vitro for about 3 weeks until a mature keratinizing epidermis has formed; the epidermis is then grafted onto the patient. The epithelial cells spread and cover the dermal substrate, eventually covering the entire wound. Early reports on two pediatric patients were very encouraging, describing the life-saving coverage of one-half of body surface with cultured epithelial autografts and the technique was eventually used in a very large number of clinical centers in several countries. Later studies showed that the “take” of CEA was very good on partial thickness wounds but was questionable in full thickness wounds. In particular, blisters formed within 2 weeks in areas grafted by CEA, a problem which has recurred persistently in clinical studies. The mechanical fragility of the resulting integument resulting from use of CEA has been traced to lack of three structural features


which are required for formation of a physiological dermal-epidermal junction at the grafted site, namely, the 7-S domain of type IV collagen, anchoring fibrils, and rete ridges. Early studies of the connective tissue underlying the CEA grafts have shown lack of a convincing dermal architecture as well as lack of elastin fibers. In another development, a skin equivalent has been prepared by populating a collagen lattice with heterlologous fibroblasts, observing the contraction of the lattice by the cells and finally seeding the surface of the lattice with a suspension of epidermal cells from an autologous source or from cell banks. The latter attach, proliferate, and differentiate to form a multilayered epidermis in 7 to 10 days of exposure to the atmosphere and the resulting skin equivalent is then grafted on wounds. Clinical studies of the SE have been limited. In an early study (1988), the SE was used to cover partially full-thickness burn wounds covering over 15% of body surface area on eight patients. In every patient grafted with SE, an extensive lysis of the SE grafts was observed at the first dressing (48 h). In one patient only, a significant percentage of “take” (40%) was observed 14 days after grafting. It was concluded that the SE was not completely appropriate to serve routinely as a substitute for the autograft. In a later study (1995), the wounds treated were acute, mostly the result of excision of skin cancers. Twelve patients had clinical “takes” at the time of grafting and there was no evidence of rejection or toxicity following grafting with the SE. The wounds grafted with SE contracted by 10 to 15%, an extent larger than that observed following grafting with full-thickness skin. Biopsies of the grafted sites showed formation of scar tissue. The authors hypothesized that the SE was eventually replaced by host tissue. Recent studies of the SE have focused on patients with venous ulcers; these studies are in progress. An attempt has been made to correct the erratic cover provided by split-thickness autografts; the latter are normally applied in a meshed form (meshed autograft) and, consequently, fail to cover the entire wound bed with a dermal layer. The attempted improvement consisted in grafting underneath the meshed autograft a living dermal tissue replacement, consisting of a synthetic polymeric mesh (polyglactin-910) which had been cultured in vitro over a period of 2 to 3 weeks with fibroblasts isolated from neonatal foreskin. Seventeen patients with full-thickness burn wounds were included in a preliminary clinical trial. Epithelialization of the interstices of the meshed autograft led to complete wound closure in 14 days in sites where the dermal living replacement had been grafted underneath the meshed autograft (experimental sites) and in those where it was omitted (control sites). Take of the meshed autograft was slightly reduced when the living dermal tissue replacement was underneath. Basement membrane structures developed both in control and experimental sites. Elastic fibers (elastin) were not observed in neodermal tissue either in control or experimental sites at periods up to one year after grafting. A subsequent clinical study explored the use of this device for the temporary closure of excised burn wounds. This 66-patient multicenter trial showed that the biosynthetic skin replacement was equivalent or superior to cadaver skin graft (frozen human cadaver allograft) with respect to its ability to prepare wounds for eventual closing with autograft. Allograft is a human cadaver skin, which is frequently stored in frozen state in a skin bank. It is a temporary cover. If left on the wound longer than about 2 weeks, allograft is rejected by the severely burned patient, even though the latter is in an immunocompromised condition. When rejection is allowed to occur, the wound bed is temporarily ungraftable and is subject to infection. In a modification of this basic use of the allograft, the latter has been used as a dermal equivalent prior to grafting with cultured epithelia. Since the allograft is rejected if allowed to remain on the wound long enough for the epithelia to spread across the wound bed, the allograft has been treated in a variety of media in an effort to eliminate its immunogenicity.

Defining Terms Adnexa: Accessory parts or appendages of an organ. Adnexa of skin include hair follicles and sweat glands. Allograft: Human cadaver skin, usually maintained frozen in a skin bank and used to provide a temporary cover for deep wounds. About 2 weeks after grafting, the allograft is removed and replaced with autograft which has become available by that time. Previously referred to as homograft.


Artificial skin: A bilayer membrane consisting of an upper layer of silicone and a lower layer of dermis regeneration template. The template is a cell-free, highly porous analog of extracellular matrix. Autograft: The patient’s own skin, harvested from an intact area in the form of a membrane and used to graft an area of severe skin loss. Basement membrane: An approximately 20-nm-thick multilayered membrane interleaved between the epidermis and the dermis. Cell lifeline number: A dimensionless number which compares the relative magnitudes of chemical reaction and diffusion. This number, defined as S in Eq. (134.3) above, can be used to compute the maximum path length, lc , over which a cell can migrate in a scaffold while depending on diffusion alone for transport of critical nutrients that it consumes. When the critical length is exceeded, the cell requires transport of nutrients by angiogenesis in order to survive. Cultured epithelial autografts: A mature, keratinizing epidermis synthesized in vitro by culturing epithelial cells removed from the patient by biopsy. A relatively small skin biopsy (1 cm2) can be treated to yield an area larger by about 10,000 in 2–3 weeks and is then grafted on patients with burns. Dermal equivalent: A term which has been loosely used to describe a device that replaces, usually temporarily, the functions of the dermis following injury. Dermis: A 2–5mm-thick layer of connective tissue populated with quiescent fibroblasts which lies underneath the epidermis. It is separated from the former by a very thin basement membrane. The dermis of adult mammals does not regenerate spontaneously following injury. Dermis regeneration template: A graft copolymer of type I collagen and chondroitin 6-sulfate, average pore diameter 20–125 µm, degrading in vivo to an extent of about 50% in 2 weeks, which induces partial regeneration of the dermis in wounds from which the dermis has been fully excised. When seeded with keratinocytes prior to grafting, this analog of extracellular matrix has induced simultaneous synthesis both of a dermis and an epidermis. Donor site: The skin site from which an autograft has been removed with a dermatome. Epidermis: The cellular outer layer of skin, about 0.1-mm thick, which protects against moisture loss and against infection. An epidermal graft, e.g., cultured epithelium or a thin graft removed surgically, requires a dermal substrate for adherence onto the wound bed. The epidermis regenerates spontaneously following injury, provided there is a dermal substrate underneath. Eschar: Dead tissue, typically the result of a thermal injury, which covers the underlying, potentially viable tissue. Extracellular matrix: A largely insoluble, nondiffusible macromolecular network, consisting mostly of glycoproteins and proteoglycans. Isomorphous matrix replacement: A term used to describe the synthesis of new, physiologic tissue within a skin regeneration template at a rate which is of the same order as the degradation of the template. This relation, Eq. (138.2) above, is the defining equation for a biodegradable scaffold which biologically interacts with the inflammatory process of the wound bed. Living dermal replacement: A synthetic biodegradable polymeric mesh, previously cultured with fibroblasts, which is placed underneath a conventional meshed autograft. Meshed autograft: A sheet autograft which has been meshed and then expanded by a factor of 1.5–6 to produce grafts with a characteristic pattern. Regeneration template: A biodegradable scaffold which, when attached to a missing organ, induces its regeneration. Scar: The result of a repair process in skin and other organs. Scar is morphologically different from skin, in addition to being mechanically less extensible and weaker than skin. The skin regeneration template induces synthesis of nearly physiologic skin rather than scar. Sheet autograft: A layer of the patient’s skin, comprising the epidermis and about one-third of the dermal thickness, which has not been meshed prior to grafting areas of severe skin loss. Skin: A vital organ which indispensably protects the organism from infection and dehydration while also providing other functions essential to physiologic life, such as assisting in thermoregulation and providing a tactile sensor for the organism.


Skin equivalent: A collagen gel which has been contracted by fibroblasts cultured therein. Following culturing with keratinocyte, until a cornified epidermis is formed over the contracted collagen lattice, it is grafted on skin wounds. Split-thickness autograft: An autograft which is about one-half or one-third as thick as the full thickness of skin. Subcutis: A layer of fat tissue underneath the dermis. Synthetic skin: A term used to describe the artificial skin in the early literature. Take: The adhesion of a graft on the woundbed. Without adequate take, there is no physicochemical or biologic interaction between graft and wound bed. Xenograft: Skin graft obtained from a different species: e.g., pig skin grafted on human. Synthetic polymeric membranes are often referred to as xenografts. Previously referred to as heterograft.

References Boykin JV, Jr., Molnar JA. 1992. Burn scar and skin equivalents. In IK Cohen, RF Diegelmann and WJ Lindblad (eds). Wound Healing, pp523–540, Philadelphia, Saunders. Burke JF, Yannas IV, Quinby WC, Jr., Bondoc CC, Jung WK. 1981. Successful use of a physiologically acceptable artificial skin in the treatment of extensive burn injury. Ann. Surg. 194:413–428. Compton CC. 1992. Current concepts in pediatric burn care: the biology of cultured epithelial autografts: An eight-year study in pediatric burn patients. Eur. J. Pediatr. Surg. 2:216–222. Compton CC, Gill JM, Bradford DA, Regauer S, Gallico GG, O’Connor NE. 1989. Skin regenerated from cultured epithelial autografts on full-thickness burn wounds from 6 days to 5 years after grafting. Lab. Invest. 60:600–612. Eaglstein WH, Iriondo M, Laszlo K. 1995. A composite skin substitute (Graftskin) for surgical wounds. Dermatol. Surg. 21:839–843. Gallico, GG, O’Connor NE, Compton CC, Kehinde O, Green H. 1984. Permanent coverage of large burn wounds with autologous cultured human epithelium, N. Engl. J. Med. 311:448–451. Hansbrough, J.F., C. Dore and W.B. Hansbrough (1992b). Clinical trials of a living dermal tissue replacement placed beneath meshed, split-thickness skin grafts on excised burn wounds, J. Burn Care Rehab. 13:519–529. Heimbach D, Luterman A, Burke J, Cram A, Herndon D, Hunt J, Jordan M, McManus W, Solem L, Warden G, Zawacki B. 1988. Artificial dermis for major burns. Ann. Surg. 208:313–320. Michaeli D, McPherson M. 1990. Immunologic study of artificial skin used in the treatment of thermal injuries, J. Burn Care Rehab. 11:21–26. Purdue GF, Hunt JL, Still JM Jr, Law EJ, Herndon DN, Goldfarb IW, Schiller WR, Hansbrough JF, Hickerson WL, Himel HN, Kealey P, Twomey J, Missavage AE, Solem LD, Davis M, Totoritis M, Gentzkow GD. 1997. A multicenter clinical trial of a biosynthetic skin replacement, DermagraftTC, compared with cryopreserved human cadaver skin for temporary coverage of excised burn wounds. J Burn Care Rehab. 18:52–57. Sabolinski ML, Alvarez O, Auletta M, Mulder G, Parenteau NL. 1996. Cultured skin as a ‘smart material’ for healing wounds: experience in venous ulcers. Biomaterials 17:311–320. Stern R, McPherson M, Longaker MT. 1990. Histologic study of artificial skin used in the treatment of full-thickness thermal injury, J. Burn Care Rehab. 11:7–13. Tompkins RG, Burke JF. 1992. Artificial Skin. Surg. Rounds October: 881–890. Waserman D, Sclotterer M, Toulon A, Cazalet C, Marien M, Cherruau B, Jaffray P. 1988. Preliminary clinical studies of a biological skin equivalent in burned patients. Burns 14:326–330. Yannas IV, Burke JF. 1980. Design of an artificial skin I. Basic design principles, J. Biomed, Mater. Res. 14:65–81. Yannas IV, Burke JF, Warpehoski M, Stasikelis P, Skrabut EM, Orgill D, Giard DJ. 1981. Prompt, longterm functional replacement of skin. Trans. Am. Soc. Artif. Intern. Organs 27:19–22.


Further Information Bell E, Erlich HP, Buttle DJ, T. Nakatsuji. 1981. Living skin formed in vitro and accepted as skin-equivalent tissue of full thickness. Science 211:1052–1054. Bell E, Ivarsson B, Merrill C. 1979. Production of a tissue-like structure by contraction of collagen lattices by human fibroblasts of different proliferative potential in vitro. Proc. Natl. Acad. Sci. USA 76:1274–1278. Boyce ST, Hansbrough JF. 1988. Biologic attachment, growth, and differentiation of cultured human keratinocytes on a graftable collagen and chondroitin 6-sulfate substrate. Surg. 103:421–431. Eldad A, Burt A, Clarke JA. 1987. Cultured epithelium as a skin substitute, Burns. 13:173–180. Green H, Kehinde O, Thomas and J. 1979. Growth of cultured human epidermal cells into multiple epithelia suitable for grafting. Proc. Natl. Acad. Sci. USA 76:5665–5668. Langer R, Vacanti JP. 1993. Tissue Engineering. Science. 260:920–926. O’Connor, NE, Mulliken JB, Banks-Schlegel S, Kehinde O, Green H. 1981. Grafting of burns with cultured epithelium prepared from autologous epidermal cells. Lancet 1: 75–78. Woodley, DT, Peterson HD, Herzog SR, Stricklin GP, Bergeson RE, Briggaman RA, Cronce DJ, O’Keefe EJ. 1988. Burn wounds resurfaced by cultured epidermal autografts show abnormal reconstitution of anchoring fibtils. JAMA 259:2566–2571. Woodley, DT, Briggaman RA, Herzog SR, Meyers AA, Peterson HD, O’Keefe EJ. 1990. Characterization of “neo-dermis” formation beneath cultured human epidermal autografts transplanted on muscle fascia. J. Invest. Dermatol. 95:20–26. Yannas, IV, Lee E, Orgill DP, Skrabut EM, Murphy GF. 1989. Synthesis and characterization of a model extracellular matrix that induces partial regeneration of adult mammalian skin. Proc. Natl. Acad. Sci. USA 86:933–937. Yannas IV, Burke JF, Gordon PL, Huang C. 1977. Multilayer membrane useful as synthetic skin, US Patent 4,060,081, Nov. 29. Yannas IV. 1982. Wound tissue can utilize a polymeric template to synthesize a functional extension of the skin. Science 215:174–176.


Robinson, C. J. “Rehabilitation Engineering.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

XIV Rehabilitation Engineering Charles J. Robinson Louisiana Tech University and the Overton Brooks VA Medical Center 139 Rehabilitation Engineering, Science, and Technology

Charles J. Robinson

Introduction • Rehabilitation Concepts • Engineering Concepts in Sensory Rehabilitation • Engineering Concepts in Motor Rehabilitation • Engineering Concepts in Communication Disorders • Appropriate Technology • The Future of Engineering in Rehabilitation

140 Orthopedic Prosthetics and Orthotics in Rehabilitation Marilyn Lord, Alan Turner-Smith Fundamentals • Applications • Summary

141 Wheeled Mobility: Wheelchairs and Personal Transportation Rory A. Cooper Introduction • Categories of Wheelchairs • Wheelchair Structure and Component Design • Ergonomics of Wheelchair Propulsion • Power Wheelchair Electrical Systems • Personal Transportation

142 Externally Powered and Controlled Orthotics and Prosthetics Dejan B. Popović FES Systems • Active Prostheses

143 Sensory Augmentation and Substitution Kurt A. Kaczmarek Visual System • Auditory System • Tactual System

144 Augmentative and Alternative Communication Gregg Vanderheiden, Katya Hill

Barry Romich,

Introduction • Language Representation Methods and Acceleration Techniques • User Interface • Outputs • Outcomes, Intervention, and Training • Future

145 Measurement Tools and Processes in Rehabilitation Engineering George V. Kondraske Fundamental Principles • Measurement Objectives and Approaches • Decision-Making Processes • Current Limitations

146 Rehabilitation Engineering Technologies: Principles of Application Douglas Hobson, Elaine Trefler The Conceptual Frameworks • The Provision Process • Education and Quality Assurance • Specific Impairments and Related Technologies • Future Developments


E

ngineering advances have resulted in enormous strides in the field of rehabilitation. Individuals with reduced or no vision can be given “sight”; those with severe or complete hearing loss can “hear” by being provided with a sense of their surroundings; those unable to talk can be aided to “speak” again; and those without full control of a limb (or with the limb missing) can, by artificial means, “walk” or regain other movement functions. But the present level of available functional restoration for seeing, hearing, speaking, and moving still pales in comparison to the capabilities of individuals without disability. As is readily apparent from the content of many of the chapters in this Handbook, the human sensory and motor (movement) systems are marvelously engineered, both within a given system and integrated across systems. The rehabilitation engineer thus faces a daunting task in trying to design augmentative or replacement systems when one or more of these systems is impaired. Rehabilitation engineering had its origins in the need to provide assistance to individuals who were injured in World War II. Rehabilitation engineering can be defined in a number of ways. Perhaps the most encompassing (and the one adopted here) is that proposed by Reswick [1982]—Rehabilitation engineering is the application of science and technology to ameliorate the handicaps of individuals with disabilities. With this definition, any device, technique, or concept used in rehabilitation that has a technological basis falls under the purview of rehabilitation engineering. This contrasts with the much narrower view that is held by some that rehabilitation engineering is only the design and production phase of a broader field called Assistive Technology. Lest one consider this distinction trivial, consider that the U.S. Congress has mandated that rehabilitation engineering and technology services be provided by all states; and an argument has ensued among various groups of practitioners about who can legally provide such services because of the various interpretations of what rehabilitation engineering is. There is a core body of knowledge that defines each of the traditional engineering disciplines. Biomedical engineering is less precisely defined; but, in general, a biomedical engineer must be proficient in a traditional engineering discipline and have a working knowledge of things biological or medical. The rehabilitation engineer is a biomedical engineer who must not only be technically proficient as an engineer and know biology and medicine, but must also integrate artistic, social, financial, psychological, and physiological considerations to develop or analyze a device, technique, or concept that meets the needs of the population the engineer is serving. In general, rehabilitation engineers deal with musculoskeletal or sensory disabilities. They often have a strong background in biomechanics. Most work in a multidisciplinary team setting. Rehabilitation engineering deals with many aspects of rehabilitation including applied, scientific, clinical, technical, and theoretical. Various topics include, but are not limited to, assistive devices and other aids for those with disability, sensory augmentation and substitution systems, functional electrical stimulation (for motor control and sensory-neural prostheses), orthotics and prosthetics, myoelectric devices and techniques, transducers (including electrodes), signal processing, hardware, software, robotics, systems approaches, technology assessment, postural stability, wheelchair seating systems, gait analysis, biomechanics, biomaterials, control systems (both biological and external), ergonomics, human performance, and functional assessment [Robinson, 1993]. In this section of the Handbook, we focus only on applications of rehabilitation engineering. The concepts of rehabilitation engineering, rehabilitation science, and rehabilitation technology are outlined in Chapter 139. Chapter 141 discusses the importance of personal mobility and various wheeled modes of transportation (wheelchairs, scooters, cars, vans, and public conveyances). Chapter 142 looks at other non-wheeled ways to enhance mobility and physical performance. Chapter 143 covers techniques available to augment sensory impairments or to provide a substitute way to input sensory information. Conversely, Chapter 144 looks at the output side. For the purposes of this Handbook, many topics that partially fall under the rubric of rehabilitation engineering are covered elsewhere. These include chapters on Electrical Stimulation (Durand), Hard Tissue Replacement—Long Bone Repair and Joints (Goel), Biomechanics (Schneck), Musculoskeletal Soft Tissue Mechanics (Lieber), Analysis of Gait (Davis), Sports Biomechanics/Kinesiology (Johnson),


Biodynamics (Diggs), Cochlear Mechanics (Steele), Measurement of Neuromuscular Performance Capabilities (Smith), Human Factors Applications in Rehabilitation Engineering (Strauss and Gunderson), Electrical Stimulators (Peckham), Prostheses and Artificial Organs (Galletti), Nerve Guidance Channels (Valentini), and Tracheal, Laryngeal, and Esophageal Replacement Devices (Shimizu). Rehabilitation engineering can be described as an engineering systems discipline. Imagine being the design engineer on a project that has an unknown, highly non-linear plant, with coefficients whose variations in time appear to follow no known or solvable model, where time (yours and your client’s) and funding are severely limited, where no known solution has been developed (or if it has, will need modification for nearly every client so no economy of scale exists). Further, there will be severe impedance mismatches between available appliances and your client’s needs. Or, the low residual channel capacity of one of your client’s senses will require enormous signal compression to get a signal with any appreciable information content through it. Welcome to the world of the rehabilitation engineer!!

References Reswick, J. 1982. What is a Rehabiliation Engineer? in Annual Review of Rehabiltation, Vol 2, E.L. Pan, T.E. Backer, and C.L. Vash, Eds., Springer-Verlag, New York. Robinson, C.J., 1993. Rehabilitation engineering—An editorial, IEEE Trans. Rehab. Eng., 1(1):1-2.


Robinson, C. J. “Rehabilitation Engineering, Science, and Technology.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

139 Rehabilitation Engineering, Science, and Technology

Charles J. Robinson Louisiana Tech University and the Overton Brooks VA Medical Center

139.1 139.2 139.3 139.4 139.5 139.6 139.7

Introduction Rehabilitation Concepts Engineering Concepts in Sensory Rehabilitation Engineering Concepts in Motor Rehabilitation Engineering Concepts in Communication Disorders Appropriate Technology The Future of Engineering in Rehabilitation

139.1 Introduction Rehabilitation engineering requires a multidisciplinary effort. To put rehabilitation engineering into its proper context, we need to review some of the other disciplines with which rehabilitation engineers must be familiar. Robinson [93] has reviewed or put forth the following working definitions and discussions: Rehabilitation is the (re)integration of an individual with a disability into society. This can be done either by enhancing existing capabilities or by providing alternative means to perform various functions or to substitute for specific sensations. Rehabilitation engineering is the “application of science and technology to ameliorate the handicaps of individuals with disabilities” [Reswick, 82]. In actual practice, many individuals who say that they practice “rehabilitation engineering” are not engineers by training. While this leads to controversies from practitioners with traditional engineering degrees, it also has the de facto benefit of greatly widening the scope of what is encompassed by the term “rehabilitation engineering.” Rehabilitation medicine is a clinical practice that focuses on the physical aspects of functional recovery, but that also considers medical, neurological and psychological factors. Physical therapy, occupational therapy, and rehabilitation counseling are professions in their own right. On the sensory-motor side, other medical and therapeutical specialties practice rehabilitation in vision, audition, and speech. Rehabilitation technology (or assistive technology) narrowly defined is the selection, design, or manufacture of augmentative or assistive devices that are appropriate for the individual with a disability. Such devices are selected based on the specific disability, the function to be augmented or restored, the user’s wishes, the clinician’s preferences, cost, and the environment in which the device will be used. Rehabilitation science is the development of a body of knowledge, gleaned from rigorous basic and clinical research, that describes how a disability alters specific physiological functions or anatomical structures, and that details the underlying principles by which residual function or capacity can be measured and used to restore function of individuals with disabilities.


139.2 Rehabilitation Concepts Effective rehabilitation engineers must be well versed in all of the areas described above since they generally work in a team setting, in collaboration with physical and occupational therapists, orthopedic surgeons, physical medicine specialists and/or neurologists. Some rehabilitation engineers are interested in certain activities that we do in the course of a normal day that could be summarized as activities of daily living (ADL). These include eating, toileting, combing hair, brushing teeth, reading, etc. Other engineers focus on mobility and the limitations to mobility. Mobility can be personal (e.g., within a home or office) or public (automobile, public transportation, accessibility questions in buildings). Mobility also includes the ability to move functionally through the environment. Thus, the question of mobility is not limited to that of getting from place to place, but also includes such questions as whether one can reach an object in a particular setting or whether a paralyzed urinary bladder can be made functional again. Barriers that limit mobility are also studied. For instance, an ill-fitted wheelchair cushion or support system will most assuredly limit mobility by reducing the time that an individual can spend in a wheelchair before he or she must vacate it to avoid serious and difficult-to-heal pressure sores. Other groups of rehabilitation engineers deal with sensory disabilities, such as sight or hearing, or with communications disorders, both in the production side (e.g., the non-vocal) or in the comprehension side. For any given client, a rehabilitation engineer might have all of these concerns to consider (i.e., ADLs, mobility, sensory and communication dysfunctions). A key concept in physical or sensory rehabilitation is that of residual function or residual capacity. Such a concept implies that the function or sense can be quantified, that the performance range of that function or sense is known in a non-impaired population, and that the use of residual capacity by a disabled individual should be encouraged. These measures of human performance can be made subjectively by clinicians or objectively by some rather clever computerized test devices. A rehabilitation engineer asks three key questions: Can a diminished function or sense be successfully augmented? Is there a substitute way to return the function or to restore a sense? And is the solution appropriate and cost-effective? These questions give rise to two important rehabilitation concepts: orthotics and prosthetics. An orthosis is an appliance that aids an existing function. A prosthesis provides a substitute. An artificial limb is a prosthesis, as is a wheelchair. An ankle brace is an orthosis. So are eyeglasses. In fact, eyeglasses might well be the consumate rehabilitation device. They are inexpensive, have little social stigma, and are almost completely unobtrusive to the user. They have let many millions of individuals with correctable vision problems lead productive lives. But in essence, a pair of eyeglasses is an optical device, governed by traditional equations of physical optics. Eyeglasses can be made out of simple glass (from a raw material as abundant as the sands of the earth!) or complex plastics such as those that are ultraviolet sensitive. They can be ground by hand or by sophisticated computer-controlled optical grinders. Thus, crude technology can restore functional vision. Increasing the technical content of the eyeglasses (either by material or manufacturing method) in most cases will not increase the amount of function restored, but it might make the glasses cheaper, lighter and more prone to be used.

139.3 Engineering Concepts in Sensory Rehabilitation Of the five traditional senses, vision and hearing most define the interactions that permit us to be human. These two senses are the main input channel through which data with high information content can flow. We read; we listen to speech or music; we view art. A loss of one or the other of these senses (or both) can have a devastating impact on the individual affected. Rehabilitation engineers attempt to restore the functions of these senses either through augmentation or via sensory substitution systems. Eyeglasses and hearing aids are examples of augmentative devices that can be used if some residual capacity remains. A major area of rehabilitation engineering research deals with sensory substitution systems [Kaczmarek, this Handbook]. The visual system has the capability to detect a single photon of light, yet also has a dynamic range that can respond to intensities many orders of magnitude greater. It can work with high contrast items © 1997 by CRC Press LLC

and with those of almost no contrast, and across the visible spectrum of colors. Millions of parallel data channels form the optic nerve that comes from an eye; each channel transmits an asynchronous and quasi-random (in time) stream of binary pulses. While the temporal coding on any one of these channels is not fast (on the order of 200 bits per sec or less), the capacity of the human brain to parallel process the entire image is faster than any supercomputer yet built. If sight is lost, how can it be replaced? A simple pair of eyeglasses will not work, since either the sensor (the retina), the communication channel (the optic nerve and all of its relays to the brain), or one or more essential central processors (the occipital part of the cerebral cortex for initial processing; the parietal and other cortical areas for information extraction) has been damaged. For replacement within the system, one must determine where the visual system has failed and whether a stage of the system can be artificially bypassed. If one uses another sensory modality (e.g., touch or hearing) as an alternate input channel, one must determine whether there is sufficient bandwidth in that channel and whether the higher-order processing hierarchy is plastic enough to process information coming via a different route. While the above discussion might seem just philosophical, it is more than that. We normally read printed text with our eyes. We recognize words from their (visual) letter combinations. We comprehend what we read via a mysterious processing in the parietal and temporal parts of the cerebral cortex. Could we perhaps read and comprehend this text or other forms of writing through our fingertips with an appropriate interface? The answer surprisingly is yes! And, the adaptation actually goes back to one of the earliest applications of coding theory—that of the development of Braille. Braille condenses all text characters to a raised matrix of 2 by 3 dots (26 combinations), with certain combinations reserved as indicators for the next character (such as a number indicator) or for special contractions. Trained readers of Braille can read over 250 words per minute of grade 2 Braille (as fast as most sighted readers can read printed text!). Thus, the Braille code is in essence a rehabilitation engineering concept where an alternate sensory channel is used as a substitute and where a recoding scheme has been employed. Rehabilitation engineers and their colleagues have designed other ways to read text. To replace the retina as a sensor element, a modern high resolution, high sensitivity, fast imaging sensor (CCD, etc.) is employed to capture a visual image of the text. One method, used by various page scanning devices, converts the scanned image to text by using optical character recognition schemes, and then outputs the text as speech via text-to-speech algorithms. This machine essentially recites the text, much as an sighted helper might do when reading aloud to the blind individual. The user of the device is thus freed of the absolute need for a helper. Such independence is often the goal of rehabilitation. Perhaps the most interesting method presents an image of the scanned data directly to the visual cortex or retina via an array of implantable electrodes that are used to electrically activate nearby cortical or retinal structures. The visual cortex and retina are laid out in a topographic fashion such that there is an orderly mapping of the signal from different parts of the visual field to the retina, and from the retina to corresponding parts of the occipital cortex. The goal of stimulation is to mimic the neural activity that would have been evoked had the signal come through normal channels. And, such stimulation does produce the sensation of light. Since the “image” stays within the visual system, the rehabilitation solution is said to be modality-specific. However, substantial problems dealing with biocompatibility and image processing and reduction remain in the design of the electrode arrays and processors that serve to interface the electronics and neurological tissue. Deafness is another manifestation of a loss of a communication channel, this time for the sense of hearing. Totally deaf individuals use vision as a substitute input channel when communicating via sign language (also a substitute code), and can sign at information rates that match or exceed that of verbal communication. Hearing aids are now commercially available that can adaptively filter out background noise (a predictable signal) while amplifying speech (unpredictable) using autoregressive, moving average (ARMA) signal processing. With the recent advent of powerful digital signal processing chips, true digital hearing aids are now available. Previous analog aids, or digitally programable analog aids, provided a set of tunable filters and amplifiers to cover the low, mid and high frequency ranges of the hearing spectrum. But the digital aids can be specifically and easily tailored (i.e., programmed) to compensate for the specific losses of each individual client across the frequency continuum of hearing, and still provide automatic


gain control and one or more user-selectable settings that have been adjusted to perform optimally in differing noise environments. An exciting development is occurring outside the field of rehabilitation that will have a profound impact on the ability of the deaf to comprehend speech. Electronics companies are now beginning to market universal translation aids for travellers, where a phrase spoken in one language is captured, parsed, translated, and restated (either spoken or displayed) in another language. The deaf would simply require that the visual display be in the language that they use for writing. Deafness is often brought on (or occurs congenitally) by damage to the cochlea. The cochlea normally transduces variations in sound pressure intensity at a given frequency into patterns of neural discharge. This neural code is then carried by the auditory (eighth cranial) nerve to the brainstem where it is preprocessed and relayed to auditory cortex for initial processing and on to the parietal and other cortical areas for information extraction. Similar to the case for the visual system, the cochlea, auditory nerve, auditory cortex and all relays in between maintain a topological map, this time based on tone frequency (tonotopic). If deafness is solely due to cochlear damage (as is often the case) and if the auditory nerve is still intact, a cochlear implant can often be substituted for the regular transducer array (the cochlea) while still sending the signal through the normal auditory channel (to maintain modality-specificity). At first glance, the design of a cochlear prosthesis to restore hearing appears daunting. The hearing range of a healthy young individual is 20 to 16,000 Hz. The transducing structure, the cochlea, has 3500 inner and 12000 outer hair cells, each best activated by a specific frequency that causes a localized mechanical resonance in the basilar membrane of the cochlea. Deflection of a hair cell causes the cell to fire an all-or-none (i.e., pulsatile) neuronal discharge, whose rate of repetition depends to a first approximation on the amplitude of the stimulus. The outputs of these hair cells have an orderly convergence on the 30,000 to 40,000 fibers that make up the auditory portion of the eighth cranial nerve. These afferent fibers in turn go to brainstem neurons that process and relay the signals on to higher brain centers [Klinke, 1983]. For many causes of deafness, the hair cells are destroyed, but the eighth nerve remains intact. Thus, if one could elicit activity in a specific output fiber by means other than the hair cell motion, perhaps some sense of hearing could be restored. The geometry of the cochlea helps in this regard as different portions of the nerve are closer to different parts of the cochlea. Electrical stimulation is now used in the cochlear implant to bypass hair cell transduction mechanisms [Loeb, 1985; Clark et al., 1990]. These sophisticated devices have required that complex signal processing, electronic and packaging problems be solved. One current cochlear implant has 22 stimulus sites along the scala tympani of the cochlea. Those sites provide excitation to the peripheral processes of the cells of the eighth cranial nerve, which are splayed out along the length of the scala. The electrode assembly itself has 22 ring electrodes spaced along its length and some additional guard rings between the active electrodes and the receiver to aid in securing the very flexible electrode assembly after it is snaked into the cochlea’s very small (a few mm) round window (a surgeon related to me that positioning the electrode was akin to pushing a piece of cooked spaghetti through a small hole at the end of a long tunnel). The electrode is attached to a receiver that is inlaid into a slot milled out of the temporal bone. The receiver contains circuitry that can select any electrode ring to be a source and any other electrode to be a sink for the stimulating current, and that can rapidly sequence between various pairs of electrodes. The receiver is powered and controlled by a radiofrequency link with an external transmitter, whose alignment is maintained by means of a permanent magnet imbedded in the receiver. A digital signal processor stores information about a specific user and his or her optimal electrode locations for specific frequency bands. The object is to determine what pair of electrodes best produces the subjective perception of a certain pitch in the implanted individual, and then to associate a particular filter with that pair via the controller. An enormous amount of compression occurs in taking the frequency range necessary for speech comprehension and reducing it to a few discrete channels. At present, the optimum compression algorithm is unknown, and much fundamental research is being carried out in speech processing, compression and recognition. But, what is amazing is that a number of totally deaf individuals can relearn to comprehend speech exceptionally well without speech-reading through the use of these implants. Other individuals find that the implant aids in speech-reading. For some only an


awareness of environmental sounds is apparent; and for another group, the implant appears to have had little effect. But, if you could (as I have been able to) finally converse in unaided speech with an individual who had been rendered totally blind and deaf by a traumatic brain injury, you begin to appreciate the power of rehabilitation engineering.

139.4 Engineering Concepts in Motor Rehabilitation Limitations in mobility can severely restrict the quality of life of an individual so affected. A wheelchair is a prime example of a prosthesis that can restore personal mobility to those who cannot walk. Given the proper environment (fairly level floors, roads, etc.), modern wheelchairs can be highly efficient. In fact, the fastest times in one of man’s greatest tests of endurance, the Boston Marathon, are achieved by the wheelchair racers. Although they do gain the advantage of being able to roll, they still must climb the same hills, and do so with only one-fifth of the muscle power available to an able-bodied marathoner. While a wheelchair user could certainly go down a set of steps (not recommended), climbing steps in a normal manual or electric wheelchair is a virtual impossibility. Ramps or lifts are engineered to provide accessibility in these cases, or special climbing wheelchairs can be purchased. Wheelchairs also do not work well on surfaces with high rolling resistance or viscous coefficients (e.g., mud, rough terrain, etc.), so alternate mobility aids must be found if access to these areas is to be provided to the physically disabled. Hand-controlled cars, vans, tractors and even airplanes are now driven by wheelchair users. The design of appropriate control modifications falls to the rehabilitation engineer. Loss of a limb can greatly impair functional activity. The engineering aspects of artificial limb design increase in complexity as the amount of residual limb decreases, especially if one or more joints are lost. As an example, a person with a mid-calf amputation could use a simple wooden stump to extend the leg, and could ambulate reasonably well. But such a leg is not cosmetically appealing and completely ignores any substitution for ankle function. Immediately following World War II, the United States government began the first concerted effort to foster better engineering design for artificial limbs. Dynamically lockable knee joints were designed for artificial limbs for above-knee amputees. In the ensuing years, energy-storing artificial ankles have been designed, some with prosthetic feet so realistic that beach thongs could be worn with them! Artificial hands, wrists and elbows were designed for upper limb amputees. Careful design of the actuating cable system also provided for a sense of hand grip force, so that the user had some feedback and did not need to rely on vision alone for guidance. Perhaps the most transparent (to the user) artificial arms are the ones that use electrical activity generated by the muscles remaining in the stump to control the actions of the elbow, wrist and hand [Stein et al., 1988]. This electrical activity is known as myoelectricity, and is produced as the muscle contraction spreads through the muscle. Note that these muscles, if intact, would have controlled at least one of these joints (e.g., the biceps and triceps for the elbow). Thus, a high level of modality-specificity is maintained since the functional element is substituted only at the last stage. All of the batteries, sensor electrodes, amplifiers, motor actuators and controllers (generally analog) reside entirely within these myoelectric arms. An individual trained in the use of a myoelectric arm can perform some impressive tasks with this arm. Current engineering research efforts involve the control of simultaneous multi-joint movements (rather than the single joint movement now available) and the provision for sensory feedback from the end effector of the artificial arm to the skin of the stump via electrical means.

139.5 Engineering Concepts in Communications Disorders Speech is a uniquely human means of interpersonal communication. Problems that affect speech can occur at the initial transducer (the larynx) or at other areas of the vocal tract. They can be of neurological (due to cortical, brainstem or peripheral nerve damage), structural, and/or cognitive origin. A person might only be able to make a halting attempt at talking, or might not have sufficient control of other motor skills to type or write. © 1997 by CRC Press LLC

If only the larynx is involved, an externally applied artificial larynx can be used to generate a resonant column of air that can be modulated by other elements in the vocal tract. If other motor skills are intact, typing can be used to generate text, which in turn can be spoken via text-to-speech devices described above. And the rate of typing (either whole words or via coding) might be fast enough so that reasonable speech rates could be achieved. The rehabilitation engineer often becomes involved in the design or specification of augmentative communication aids for individuals who do not have good muscle control, either for speech or for limb movement. A whole industry has developed around the design of symbol or letter boards, where the user can point out (often painstakingly) letters, words or concepts. Some of these boards now have speech output. Linguistics and information theory have been combined in the invention of acceleration techniques intended to speed up the communication process. These include alternative language representation systems based on semantic (iconic), alphanumeric, or other codes; and prediction systems, which provide choices based on previously selected letters or words. A general review of these aids can be found in Chapter 144, while Goodenough-Trepagnier [1994] edited a good publication dealing with human factors and cognative requirements. Some individuals can produce speech, but it is dysarthric and very hard to understand. Yet the utterance does contain information. Can this limited information be used to figure out what the individual wanted to say, and then voice it by artificial means? Research labs are now employing neural network theory to determine which pauses in an utterance are due to content (i.e., between a word or sentence) and those due to unwanted halts in speech production.

139.6 Appropriate Technology Rehabilitation engineering lies at the interface of a wide variety of technical, biological and other concerns. A user might (and often does) put aside a technically sophisticated rehabilitation device in favor of a simpler device that is cheaper, and easier to use and maintain. The cosmetic appearance of the device (or cosmesis) sometimes becomes the overriding factor in acceptance or rejection of a device. A key design factor often lies in the use of the appropriate technology to accomplish the task adequately given the extent of the resources available to solve the problem and the residual capacity of the client. Adequacy can be verified by determining that increasing the technical content of the solution results in disproportionately diminishing gains or escalating costs. Thus, a rehabilitation engineer must be able to distinguish applications where high technology is required from those where such technology results in an incremental gain in cost, durability, acceptance and other factors. Further, appropriateness will greatly depend on location. What is appropriate to a client near a major medical center in a highly developed country might not be appropriate to one in a rural setting or in a developing country. This is not to say that rehabilitation engineers should shun advances in technology. In fact, a fair proportion of rehabilitation engineers work in a research setting where state-of-the-art technology is being applied to the needs of the disabled. However, it is often difficult to transfer complex technology from a laboratory to disabled consumers not directly associated with that laboratory. Such devices are often designed for use only in a structured environment, are difficult to repair properly in the field, and often require a high level of user interaction or sophistication. Technology transfer in the rehabilitation arena is difficult, due to the limited and fragmented market. Advances in rehabilitation engineering are often piggybacked onto advances in commercial electronics. For instance, the exciting developments in text-to-speech and speech-to-text devices mentioned above are being driven by the commercial marketplace, and not by the rehabilitation arena. But such developments will be welcomed by rehabilitation engineers no less.

139.7 The Future of Engineering in Rehabilitation The traditional engineering disciplines permeate many aspects of rehabilitation. Signal processing, control and information theory, materials design, computers are all in widespread use from an electrical engineering © 1997 by CRC Press LLC

perspective. Neural networks, microfabrication, fuzzy logic, virtual reality, image processing and other emerging electrical and computer engineering tools are increasingly being applied. Mechanical engineering principles are used in biomechanical studies, gait and motion analysis, prosthetic fitting, seat cushion and back support design, and the design of artificial joints. Materials and metalurgical engineers provide input on newer biocompatable materials. Chemical engineers are developing implantable sensors. Industrial engineers are increasingly studying rehabilitative ergonomics. The challenge to rehabilitation engineers is to find advances in any field, engineering or otherwise, that will aid their clients who have a disability.

Defining Terms [n.b., The first five terms below have been proposed by the National Center for Medical Rehabilitation and Research (NCMRR) of the US National Institutes of Health (NIH)]. Disability: Inability or limitation in performing tasks, activities, and roles to levels expected within physical and social contexts Functional Limitation: Restriction or lack of ability to perform an action in the manner or within the range consistent with the purpose of an organ or organ system Impairment: Loss or abnormality of cognitive, emotional, physiological or anatomical structure or function, including all losses or abnormalities, not just those attributed to the initial pathophysiology Pathophysiology: Interruption or interference with normal physiological and developmental processes or structures Societal Limitation: Restriction, attributable to social policy or barriers (structural or attitudinal), which limits fulfillment of roles, or denies access to services or opportunities that are associated with full participation in society Activities of daily living (ADL): Personal activities that are done by almost everyone in the course of a normal day including eating, toileting, combing hair, brushing teeth, reading, etc. ADLs are distinguished from hobbies and from work-related activities (e.g., typing). Appropriate technology: the technology that will accomplish a task adequately given the resources available. Adequacy can be verified by determining that increasing the technological content of the solution results in diminishing gains or increasing costs. Modality-specific: A task that is specific to a single sense or movement pattern. Orthosis: A modality-specific appliance that aids the performance of a function or movement by augmenting or assisting the residual capabilities of that function or movement. An orthopaedic brace is an orthosis. Prosthesis: an appliance that substitutes for the loss of a particular function, generally by involving a different modality as an input and/or output channel. An artificial limb, a sensory substitution system, or an augmentative communication aid are prosthetic devices. Residual function or residual capacity: Residual function is a measure of the ability to to carry out one or more general tasks using the methods normally used. Residual capacity is a measure of the ability to to carry out these tasks using any means of performance. These residual measures are generally more subjective than other more quantifiable measures such as residual strength.

References Clark, G.M., Y.C. Tong, and J.F. Patrick, 1990. Cochlear Prostheses, Edinburgh, Churchill Livingstone. Goodenough-Trepagnier, C., 1994. Guest Editor of a special issue of Assistive Technology 6(1) dealing with mental loads in augmentative communication. Kaczmarek, K.A., J.G. Webster, P. Bach-y-Rita and W.J. Tompkins, 1991. “Electrotactile and Vibrotactile Displays for Sensory Substitution”, IEEE Trans. Biomed. Engr., 38:1-16. Klinke, R., 1983. “Physiology of the Sense of Equilibrium, Hearing and Speech.” Chapter 12 in: Human Physiology (eds: R.F. Schmidt and G. Thews), Berlin, Springer-Verlag.


Loeb, G.E., 1985. “The Functional Replacement of the Ear,” Scientific American, 252:104-111. Reswick, J. 1982. “What is a Rehabiliation Engineer?” in Annual Review of Rehabiltation, Vol 2 (eds. E.L. Pan, T.E. Backer, C.L. Vash), New York, Springer-Verlag. Robinson, C.J. 1993. “Rehabilitation Engineering—an Editorial,” IEEE Transactions on Rehabilitation Engineering 1(1):1-2. Stein, R.B., D. Charles, and K.B. James, 1988. Providing Motor Control for the Handicapped: A Fusion of Modern Neuroscience, Bioengineering, and Rehabilitation,” Advances in Neurology, Vol. 47: Functional Recovery in Neurological Disease, (ed. S.G. Waxman), Raven Press, New York.

Further Information Readers interested in rehabilitation engineering can contact RESNA—an interdisciplinary association for the advancement of rehabilitation and assistive technologies, 1101 Connecticut Ave., N.W., Suite 700, Washington, D.C. 20036. RESNA publishes a quarterly journal called Assistive Technology. The United States Department of Veterans Affairs puts out a quarterly Journal of Rehabilitation R&D. The January issue of each year contains an overview of most of the rehabilitation engineering efforts occurring in the U.S. and Canada, with over 500 listings. The IEEE Engineering in Medicine and Biology Society publishes the IEEE Transactions on Rehabilitation Engineering, a quarterly journal. The reader should contact the IEEE at PO Box 1331, 445 Hoes Lane, Piscataway, NJ 08855-1331 U.S.A. for further details.


Lord, M. L., Turner-Smith, A . “ Orthopedic Prosthetics and Orthotics in Rehabilitation.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

140 Orthopedic Prosthetics and Orthotics in Rehabilitation Marilyn Lord King’s College London

140.1 140.2

Computer-Aided Engineering in Customized Component Design • Examples of Innovative Component Design

Alan Turner-Smith King’s College London

Fundamentals Applications

140.3

Summary

An orthopedic prosthesis is an internal or external device that replaces lost parts or functions of the neuroskeletomotor system. In contrast, a orthopedic orthosis is a device that augments a function of the skeletomotor system by controlling motion or altering the shape of body tissue. For example, an artificial leg or hand is a prosthesis, whereas a calliper (or brace) is an orthosis. This chapter addresses only orthoses and external orthopedic prostheses; internal orthopedic prostheses, such as artificial joints, are a subject on their own. When a human limb is lost through disease or trauma, the integrity of the body is compromised in so many ways that an engineer may well feel daunted by the design requirements for a prosthetic replacement. Consider the losses from a lower limb amputation. Gone is the structural support for the upper body in standing, along with the complex joint articulations and muscular motor system involved in walking. Lost also is the multimode sensory feedback, from inter alia pressure sensors on the sole of the foot, length and force sensors in the muscles, and position sensors in the joints, which closed the control loop around the skeletomotor system. The body also has lost a significant percentage of its weight and is now asymmetrical and unbalanced. We must first ask if it is desirable to attempt to replace all these losses with like-for-like components. If so, we need to strive to make a bionic limb of similar weight embodying anthropomorphic articulations with equally powerful motors and distributed sensors connected back into the wearer’s residual neuromuscular system. Or, is it better to accept the losses and redefine the optimal functioning of the new unit of person-plus-technology? In many cases, it may be concluded that a wheelchair is the optimal solution for lower limb loss. Even if engineering could provide the bionic solution, which it certainly cannot at present despite huge inroads made into aspects of these demands, there remain additional problems inherent to prosthetic replacements to consider. Of these, the unnatural mechanical interface between the external environment and the human body is one of the most difficult. Notable, in place of weight bearing through the structures of the foot that are well adapted for this purpose, load must now be transferred to the skeletal structures via intimate contact between the surface of residual limb and prosthesis; the exact distribution of load becomes critical. To circumvent these problems, an alternative direct transcutaneous fixation to the bone has been attempted in limited experimental trials, but this


brings its own problems of materials biocompatability and prevention of infection ingress around the opening through the skin. Orthotic devices are classified by acronyms that describe the joint which they cross. Thus an AFO is an ankle-foot orthosis, a CO is a cervical orthosis (neck brace or collar), and a TLSO is a thoracolumbosacral orthosis (spinal brace or jacket). The main categories are braces for the cervix (neck), upper limb, trunk, lower limb, and foot. Orthoses are generally simpler devices than prostheses, but because orthoses are constrained by the existing body shape and function, they can present an equally demanding design challenge. Certainly the interaction with body function is more critical, and successful application demands an in-depth appreciation of both residual function and the probable reaction to external interference. External orthotics are often classified as structural or functional, the former implying a static nature to hold an unstable joint and the latter a flexible or articulated system to promote the correct alignment of the joints during dynamic functioning. An alternative orthotic approach utilizes functional electrical stimulation (FES) of the patient’s own muscles to generate appropriate forces for joint motion; this is dealt with in Chapter 142.

140.1 Fundamentals Designers of orthotic and prosthetic devices are aware of the three cardinal considerations—function, structure, and cosmesis. For requirements of function, we must be very clear about the objectives of treatment. This requires first an understanding of the clinical condition. Functional prescription is now a preferred route for the medical practitioner to specify the requirements, leaving the implementation of this instruction to the prosthetist, orthotist, or rehabilitation technologist. The benefits of this distinction between client specification and final hardware will be obvious to design engineers. Indeed, the influence of design procedures on the supply process is a contribution from engineering that is being appreciated more and more. The second requirement for function is the knowledge of the biomechanics that underlies both the dysfunction in the patient and the function of proposed device to be coupled to the patient. Kinematics, dynamics, energy considerations, and control all enter into this understanding of function. Structure is the means of carrying the function, and finally both need to be embodied into a design that is cosmetically acceptable. Some of the fundamental issues in these concepts are discussed here. To function well, the device needs an effective coupling to the human body. To this end, there is often some part that is molded to the contours of the wearer. Achieving a satisfactory mechanical interface of a molded component depends primarily on the shape. The internal dimensions of such components are not made an exact match to the external dimensions of the limb segment, but by a process of rectification, the shape is adjusted to relieve areas of skin with low load tolerance. The Shapes are also evolved to achieve appropriate load distribution for stability of coupling between prosthetic socket and limb or, in orthotic design, a system of usually three forces that generates a moment to stabilize a collapsing joint (Fig. 140.1). Alignment is a second factor influencing the interface loading. For lower limb prostheses particularly, the alignment of the molded socket to the remainder of the structural components also will be critical in determining the moments and forces transmitted to the interface when the foot is flat on the ground. The same is true for lower limb orthoses, where the net action of the ground reaction forces and consequent moments around the natural joints are highly dependent on the alignment taken up by the combination of orthosis and shoe. Adjustability may be important, particularly for children or progressive medical conditions. Functional components that enable desirable motions are largely straightforward engineering mechanisms such as hinges or dampers, although the specific design requirements for their dynamic performance may be quite complex because of the biomechanics of the body. An example of the design of knee joints is expanded below. These motions may be driven from external power sources but more often are passive or body-powered mechanisms. In orthoses where relatively small angular motions are needed, these may be provided by material flexibility rather than mechanisms. The structural requirements for lower-limb prosthetics have been laid down at a consensus meeting (1978) bases on biomechanical measurement of forces in a gait laboratory, referred to as the Philadelphia standards and soon to be incorporated into an ISO standard (ISO 13404,5; ISO 10328).Not only are the © 1997 by CRC Press LLC

FIGURE 140.1 Three-force system required in an orthosis to control a valgus hindfoot due to weakness in the hindfoot supinators.

load level and life critical, but so is the mode of failure. Sudden failure of an ankle bolt resulting in disengagement of an artificial foot is not only potentially life-threatening to an elderly amputee who falls and breaks a hip but also can be quite traumatic to unsuspecting witnesses of the apparent event of autoamputation. Design and choice of materials should ensure a controlled slow yielding, not brittle fracture. A further consideration is the ability of the complete structure to absorb shock loading, either the repeated small shocks of walking at the heel strike or rather more major shocks during sports activities or falls. This minimizes the shock transmitted through the skin to the skeleton, known to cause both skin lesions and joint degeneration. Finally, the consideration of hygiene must not be overlooked; the user must be able to clean the orthosis or prosthesis adequately without compromising its structure or function. Added to the two elements of structure and function, the third element of cosmesis completes the trilogy. Appearance can be of great psychological importance to the user, and technology has its contribution here, too. As examples, special effects familiar in science fiction films also can be harnessed to provide realistic cosmetic covers for hand or foot prostheses. Borrowing from advanced manufacturing technology, optical shape scanning linked to three-dimensional (3D) computer-aided design, and CNC machining can be pressed into service to generate customized shapes to match a contralateral remaining limb. Up-to-date materials and component design each contribute to minimize the “orthopedic appliance” image of the devices (Fig. 140.2). In providing cosmesis, the views of the user must remain paramount. The wearer will often choose an attractive functional design in preference to a life like design that is not felt to be part of his or her body. Upper limb prostheses are often seen as a more interesting engineering challenge than lower limb, offering the possibilities for active motor/control systems and complex articulations. However, the market is an order of magnitude smaller and cost/benefit less easy to prove—after all, it is possible to function


FIGURE 140.2 The ARGO reciprocating-gait orthosis, normally worn under the clothing, with structural components produced from 3D CAD. (Courtesy of Hugh Steeper, Ltd., U.K.)

fairly well with one arm, but try walking with one leg. At the simplest end, an arm for a below-elbow amputee might comprise a socket with a terminal device offering a pincer grip (hand or hook) that can be operated through a Bowden cable by shrugging the shoulders. Such body-powered prostheses may appear crude, but they are often favored by the wearer because of a sense of position and force feedback from the cable, and they do not need a power supply. Another, more elegant method of harnessing body power is to take a muscle made redundant by an amputation and tether its tendon through an artificially fashioned loop of skin: the cable can then be hooked through the loop [Childress, 1989]. Externally powered devices have been attempted using various power sources with degrees of success. Pneumatic power in the form of a gas cylinder is cheap and light, but recharging is a problem that exercised the ingenuity of early suppliers: where supplies were not readily available, even schemes to involve the local fire services with recharging were costed. Also, contemplate the prospect of bringing a loaded table fork toward your face carried on the end of a position-controlled arm powered with spongy, low-pressure pneumatic actuators, and you will appreciate another aspect of difficulties with this source. Nevertheless, gas-powered grip on a hand can be a good solution. Early skirmishes with stiffer hydraulic servos were largely unsuccessful because of power supply and actuator weight and oil leakage. Electric actuation, heavy and slow at first, has gradually improved to establish its premier position. Input control to these powered devices can be from surface electromyography or by mechanical movement of, for example, the shoulder or an ectromelic limb. Feedback can be presented as skin pressure, movement of a sensor over the skin, or electric stimulation. Control strategies range from position control around a



single joint or group of related joints through combined position and force control for hand grip to computer-assisted coordination of entire activities such as feeding. The physical designs in prosthetic and orthotic devices has changed substantially over the past decade. One could propose that this is solely the introduction of new materials. The sockets of artificial limbs have always been fashioned to suit the individual patient, historically by carving wood, shaping leather, or beating sheet metal. Following the introduction of thermosetting fiber-reinforced plastics hand-shaped over a plaster cast of the limb residuum, substitution of thermoforming plastics that could be automatically vacuum-formed made a leap forward to give light, rapidly made, and cosmetically improved solutions. Polypropylene is the favored material in this application. The same materials permitted the new concept of custom-molded orthoses. Carbon fiber composites substituted for metal have certainly improved the performance of structural components such as limb shanks. But some of the progress owes much to innovative thinking. The flex foot is a fine example, where a traditional anthropomorphic design with imitation ankle joint and metatarsal break is completely abandoned and a functional design adopted to optimize energy storage and return. This is based on two leaf springs made from Kevlar, joined together at the ankle with one splaying down toward the toes to form the forefoot spring and the other rearward to form the heel spring (Fig. 140.3). Apart from the gains for the disabled athletes for whom the foot was designed–and these are so remarkable that there is little point in competing now without this foot–clients across all age groups have benefited from the adaptability to rough ground and shockabsorption capability.


140.2 Applications Computer-Aided Engineering in Customized Component Design Computer-aided engineering has found a fertile ground for exploitation in the process of design of customized components to match to body shape. A good example is in sockets for artificial limbs. What prosthetists particularly seek is the ability to produce a well-fitting socket during the course of a single patient consultation. Traditional craft methods of casting the residual limb in plaster of paris, pouring a positive mold, manual rectification, and then socket fabrication over the rectified cast takes too long. By using advanced technology, residual limb shapes can be captured in a computer, rectified by computer algorithms, and CNC machined to produce the rectified cast in under an hour so that with the addition of vacuum-formed machinery to pull a socket rapidly over the cast, the socket can be ready for trial fitting in one session. There are added advantages too, in that the shape is now stored in digital form in the computer and can be reproduced or adjusted whenever and wherever desired. Although such systems are still in an early stage of introduction, many practicing prosthetists in the United States have now had hands-on experience of this technology, and a major evaluation by the Veterans Administration has been undertaken [Houston et al., 1992]. Initially, much of the engineering development work went into the hardware components, a difficult brief in view of the low cost target for a custom product. Requirements are considerably different from those of standard engineering, e.g., relaxation in the accuracies required (millimeters, not microns);a need to measure limb or trunk parts that are encumbered by the attached body, which may resist being orientated conveniently in a machine and which will certainly distort with the lightest pressure; and a need to reproduce fairly bulky items with strength to be used as a sacrificial mold. Instrumentation for body shape scanning has been developed using methods of silhouettes, Moiré fringes, contact probes measuring contours of plaster casts, and light triangulation. Almost universally the molds are turned by “milling on a spit” [Duncan & Mair, 1983], using an adapted lathe with a milling head to spiral down a large cylindrical plug of material such as plaster of paris mix. Rehabilitation engineers watch with great interest, some with envy, the developments in rapid prototyping manufacture, which is so successful in reducing the cycle time for one-off developments elsewhere in industry, but alas the costs of techniques such as stereolithography are as yet beyond economic feasibility for our area. Much emphasis also has been placed on the graphics and algorithms needed to achieve rectification. Opinions vary as to what extent the computer should simply provide a more elegant tool for the prosthetist to exercise his or her traditional skills using 3D modeling and on-screen sculpting as a direct replacement for manual plaster rectification or to what extent the computer system should take over the bulk of the process by an expert systems approach. Systems currently available tend to do a little of each. A series of rectification maps can be held as templates, each storing the appropriate relief or buildup to be applied over a particular anatomic area of the limb. Thus the map might provide for a ridge to be added down the front of the shin of a lower limb model so that the eventual socket will not press against the vulnerable bony prominence of the tibia (Fig. 140.4). Positioning of the discrete regions to match individual anatomy might typically be anchored to one FIGURE 140.3 The Flex Foot. or more anatomic features indicated by the prosthetist.


FIGURE 140.4 A rectification require defined over the tibia of lower limb stump using the Shapemaker application for computer-aided socket design.

FIGURE 140.5 Adjusting a socket contour with reference to 3D graphics and cross-sectional profiles in the UCL CASD system. (Reproduced from Reynolds and Lord [1992] with permission.)

The prosthetist is also able to free-form sculpt a particular region by pulling the surface interactively with reference to graphic representation (Fig. 140.5); this is particularly useful where the patient has some unusual feature not provided for in the templates. As part of this general development, finite-element analysis has been employed to model the soft tissue distortion occurring during limb loading and to look at the influence of severity of rectification in the resultant distribution of interface stress [e.g., Reynolds & Lord, 1992] (Fig. 140.6). In engineering terms, this modeling is somewhat unusual and decidedly nonlinear. For a start, the tissues are highly deformable but nearly incompressible, which raises problems of a suitable Poisson ratio to apply in the modeling. Values of n = 0.3 to n = 0.49 have been proposed, based on experimental matching of stress-strain curves from indentation of limb tissue in vivo. In reality, though, compression (defined as a loss of volume) may be noted in a limb segment under localized external pressure due to loss of mass as first the blood


FIGURE 140.6 Finite-element analysis employed to determine the sensitivity of interface pressure to socket shape rectification: (a) limb and socket; (b) elements in layers representing idealized geometry of bone, soft tissue, and socket liner; (c) rectification map of radial differences between the external free shape of the limb and the internal dimensions of socket; and (d) FE predictions of direct pressure. (Courtesy of Zhang Ming, King’s College London.)

is rapidly evacuated and then interstitial fluids are more slowly squeezed out. Also, it is difficult to define the boundaries of the limb segment at the proximal end, still attached to the body, where soft tissues can easily bulge up and out. This makes accurate experimental determination of the stress-strain curves for the tissue matrix difficult. A nonlinear model with interface elements allowing slip to occur between skin and socket at the limit of static friction may need to be considered, since the frictional conditions at the interface will determine the balance between shear and direct stresses in supporting body weight against the sloping sidewalls. Although excessive shear at the skin surface is considered particularly damaging, higher pressures would be required in its complete absence. In a similar vein, computer-aided design (CAD) techniques are also finding application in the design of bespoke orthopedic footwear, using CAD techniques from the volume fashion trade modified to suit the one-off nature of bespoke work. This again requires the generation of a customized mold, or shoe last, for each foot, in addition to the design of patterns for the shoe uppers [Lord et al., 1991]. The philosophy of design of shoe lasts is quite different from that of sockets, because last shapes have considerable and fundamental differences from foot shapes. In this instance, a library of reference last shapes is held, and a suitable one is selected both to match the client’s foot shape and to fulfill the shoemaking needs for the particular style and type of shoe. The schematic of the process followed in development of the Shoemaster system is shown in Fig. 140.7. Design of shoe inserts is another related application, with systems to capture, manipulate, and reproduce underfoot contours now in commercial use. An example is the Ampfit system, where the foot is placed on a platform to which preshaped arch supports or other wedges or domes may first be attached. A matrix of round-ended cylinders is then forced up by gas pressure through both platform and supports, supporting the foot over most of the area with an even load distribution. The shape is captured from the cylinder locations and fed into a computer, where rectification can be made similar to that described for prosthetic sockets. A benchtop CNC machine then routs the shoe inserts from specially provided blanks while the client waits.




FIGURE 140.7 Schematic of operation of the Shoemaster shoe design system based on selection of a basis last from a database of model lasts. A database of styles is also employed to generate the upper patterns.

Examples of Innovative Component Design An Intelligent Prosthetic Knee The control of an artificial lower limb turns out to be most problematic during the swing phase, during which the foot is lifted off the ground to be guided into contact ahead of the walker. A prosthetic lower limb needs to be significantly lighter than its normal counterpart because the muscular power is not present to control it. Two technological advances have helped. First, carbon fiber construction has reduced the mass of the lower limb, and second, pneumatic or hydraulically controlled damping mechanisms for the knee joint have enabled adjustment of the swing phase to suit an individual’s pattern of walking. Swing-phase control of the knee should operate in three areas: 1. Resistance to flexion at late stance during toe-off controls any tendency to excessive heel rise at early swing. 2. Assistance to extension after midswing ensures that the limb is fully extended and ready for heel strike. 3. Resistance before a terminal impact at the end of the extension swing dampens out the inertial forces to allow a smooth transition from flexed to extended knee position. In conventional limbs, parameters of these controls are determined by fixed components (springs, bleed valves) that are set to optimum for an individual’s normal gait at one particular speed, e.g., the pneumatic controller in Fig. 140.8. If the amputee subsequently walks more slowly, the limb will tend to lead, while if the amputee walks more quickly, the limb will tend to fall behind; the usual compensatory actions are, respectively, an unnatural tilting of the pelvis to delay heel contact or abnormal kicking through of the leg. In a recent advance, intelligence is built into the swing-phase controller to adjust automatically for cadence variations (Fig. 140.9). A 4-bit microprocessor is used to adjust a needle valve, via a linear stepper


FIGURE 140.8 Pneumatic cylinder action in a swing phase controller. (Reprinted with permission from S. Zahedi, The Results of the Field Trial of the Endolite Intelligent Prosthesis, internal publication, Chas. A. Blatchford & Sons, U.K.)

motor, according to duration of the preceding swing phase [Zahedi, in press]. The unit is programmed by the prosthetist to provide optimal damping for the particular amputee’s swing phase at slow, normal, and fast walking paces. Thereafter, the appropriate damping is automatically selected for any intermediate speed. A Hierarchically Controlled Prosthetic Hand Control of the intact hand is hierarchical. It starts with the owner’s intention, and an action plan is formulated based on knowledge of the environment and the object to be manipulated. For gross movements, the numerous articulations rely on “preprogrammed” coordination from the central nervous system. Fine control leans heavily on local feedback from force and position sensors in the joints and tactile information about loading and slip at the skin. In contrast, conventional prostheses depend on the conscious command of all levels of control and so can be slow and tiring to use. Current technology is able to provide both the computing power and transducers required to recreate some of a normal hand’s sophisticated proprioceptive control. A concept of extended physiologic proprioception (EPP) was introduced for control of gross arm movement [Simpson & Kenworthy, 1973] © 1997 by CRC Press LLC

FIGURE 140.9

The Endolite intelligent prosthesis in use, minus its cosmetic covers.

whereby the central nervous system is retrained through residual proprioception to coordinate gross actions applying to the geometry of the new extended limb. This idea can be applied to initiate gross hand movements while delegating fine control to an intelligent controller. Developments by Chappell and Kyberd [1991] and others provide a fine example of the possibilities. A suitable mechanical configuration is shown in Fig. 140.10. Four 12-V dc electric motors with gearboxes control, respectively, thumb adduction, thumb flexion, forefinger flexion, and flexion of digits 3,4, and 5. Digits 3, 4, and 5 are linked together by a double-swingletree mechanism that allows all three to be driven together. When one digit touches an object the other two can continue to close until they also touch or reach their limit of travel. The movement of the digits allows one of several basic postures: • Three-point chuck: Precision grip with digits 1, 2, and 3 (thumb set to oppose the midline between digits 2 and 3); digits 4 and 5 give additional support. • Two-point grip: Precision grip with digits 1 and 2 (thumb set to oppose forefinger); digits 3, 4, and 5 fully flexed and not used or fully extended. • Fist: As two-point grip but with thumb fully extended to allow large objects to be grasped. • Small Fist: As fist but with thumb flexed and abducted to oppose side of digit 2. • Side, or key: Digits 2 to 5 half fully flexed with thumb opposing side of second digit. • Flat hand: Digits 2 to 5 fully extended with thumb abducted and flexed, parked beside digit 2. The controller coordinates the transition between these positions and ensures that trajectories do not tangle. Feedback to the controller is provided by several devices. Potentiometers detect the angles of flexion of the digits; touch sensors detect pressure on the palmer surfaces of the digits; and a combined contact force (Hall effect) and slip sensor (from acoustic frequency output of force sensor) is mounted at the fingertips. The latter detects movement of an object and so controls grip strength appropriate to the task–whether holding a hammer or an egg [Kyberd & Chappell, 1993].


The whole hand may be operated by electromyographic signals from two antagonistic muscles in the supporting forearm stump, picked up at the skin surface. In response to tension in one muscle, the hand opens progressively and then closes to grip with an automatic reflex. The second muscle controls the mode of operation as the hand moves between the states of touch, hold, squeeze, and release. A Self-Aligning Orthotic Knee Joint Knee orthosis are often supplied to resist knee flexion during standing and gait at an otherwise collapsing joint. The rigid locking mechanisms on these devices are manually released to allow knee flexion during sitting. Fitting is complicated by the difficulty of attaching the orthosis with its joint accurately aligned to that of the knee. The simple diagram in Fig. 140.11 shows how misplacement of a simple hinged orthosis with a notional fixed knee axis would cause the cuffs on the thigh and calf to press into the soft tissues of the limb (known as pistoning). The human knee does not have a fixed axis, though, but is better represented as a polycentric joint. In a sagittal (side) view, it is easy to conceptualize the origin of these kinematics from the anatomy of the cruciate ligaments running crisscross across the joint, which together with the base of the femur and the head of the tibia form a classic four-bar linkage. The

FIGURE 140.10 The Southampton hand prosthesis with four degrees of freedom in a power grip. An optical/acoustic sensor is mounted on the thumb. (Reprinted from Kyberd and Chappell [1993], Fig. 1.)

FIGURE 140.11 The problem caused by misplacement of a single-axis orthotic joint (a) is overcome by an orthosis (b) with a self-aligning axis. (The Laser system, courtesy of Hugh Steeper, Ltd., U.K.)


polycentric nature of the motion can therefore be mimicked by a similar geometry of linkage on the orthosis. The problem of alignment still remains, however, and precision location of attachment points is not possible when gripping through soft tissues. In one attempt to overcome this specific problem, the knee mechanism has been designed with not one but two axes (Fig. 140.11). The center of rotation is then free to self-align. This complexity of the joint while still maintaining the ability to fixate the knee and meeting low weight requirements is only achieved by meticulous design in composite materials.

140.3 Summary The field of prosthetics and orthotics is one where at present traditional craft methods sit alongside the application of high technology. Gradually, advanced technology is creeping into most areas, bringing vast improvements in hardware performance specifications and aesthetics. This is, however, an area where the clinical skills of the prosthetist and orthotist will always be required in specification and fitting and where many of the products have customized components. The successful applications of technology are those which assist the professional to exercise his or her judgment, providing him or her with good tools and means to realize a functional specification. Since the demand for these devices is thankfully low, their design and manufacture are small scale in terms of volume. This taxes the skills of most engineers, both to design the product at reasonable upfront costs and to manufacture it economically in low volume. For bespoke components, we are moving from a base of craft manufacture through an era when modularization was exploited to allow smallbatch production toward the use of CAD. In the latter, the engineering design effort is then embodied in the CAD system, leaving the prosthetist or orthotist to incorporate clinical design for each individual component. Specific examples of current applications have been described. These can only represent a small part of the design effort that is put into prosthetics and orthotics on a continuing basis, making advances in materials and electronics in particular available. We are also aware that in the space available, it has not been possible to include a discussion of the very innovative work that is being done in intermediate technology for the third world, for which the International Society of Prosthetics and Orthotics (address below) currently has a special working group.

Defining Terms Biocompatability: Compatibility with living tissue, e.g., in consideration of toxicity, degradability, and mechanical interfacing. CNC machining: Use of a computer numerically controlled machine. Cosmesis: Aesthetics of appearance. Ectromelia: Congenital gross shortening of the long bones of a limb. Functional prescription: A doctor’s prescription for supply of a device written in terms of its function as opposed to embodiment. Neuroskeletomotor system: The skeletal frame of the body with the muscles, peripheral nerves, and central nervous system of the spine and brain, which together participate in movement and stabilization of the body. Rectified, rectification: Adjustment of a model of body shape to achieve a desirable load distribution in a custom-molded prosthesis or orthosis. Soft tissues: Skin, fat, connective tissues, and muscles which, along with the hard tissues of bone, teeth, etc. and the fluids, make up the human body. Transcutaneous: Passing through the skin.


References Chappell PH, Kyberd PJ. 1991. Prehensile control of a hand prosthesis by a microcontroller. J Biomed Eng 13:363. Childress DS. 1989. Control philosophies for limb prostheses. In J Paul, et al. (eds), Progress in Bioengineering, pp 210–215. New York, Adam Hilger. Duncan JP, Mair SG. 1983. Sculptured Surfaces in Engineering and Medicine. Cambridge, England, Cambridge Univ. Press. Houston VL, Burgess EM, Childress DS, et al. 1992. Automated fabrication of mobility aids (AFMA): Below-knee CASD/CAM testing and evaluation. J Rehabil Res Dev 29:78. Kyberd PJ, Chappell PH. 1993. A force sensor for automatic manipulation based on the Hall effect. Meas Sci Technol 4:281. Lord M, Foulston J, Smith PJ. 1991. Technical evaluation of a CAD system for orthopaedic shoe-upper design. Eng Med Proc Instrum Mech Eng 205:109. Reynolds DP, Lord M. 1992. Interface load analysis for computer-aided design of below-knee prosthetic sockets. Med Biol Eng Comput 30:419. Simpson DC, Kenworthy G. 1973. The design of a complete arm prosthesis. Biomed Eng 8:56. Zahedi S. In press, 1994. Evaluation and biomechanics of the intelligent prosthesis: A two-year study. Orthop Tech.

Further Information Bowker P, Condie DN, Bader DL, Pratt DJ (eds). Biomechanical Basis of Orthotic Management. Oxford, Butterworth-Heinemann, 1993. Murdoch G, Donovan RG (eds). 1988. Amputation Surgery and Lower Limb Prosthetics. Boston, Blackwell Scientific Publications. Nordin M, Frankel V. 1980. Basic Mechanics of the Musculoskeletal System, 2d ed. Philadelphia, Lea & Febiger, 1980. Smidt GL (ed). 1990. Gait in Rehabilitation. New York, Churchill-Livingstone.

Organizations International Society of Prosthetics and Orthotics (ISPO), Borgervaenget 5,2100 Copenhagen Ø, Denmark [tel (31) 20 72 60]. Department of Veterans Affairs, VA Rehabilitation Research and Development Service, 103 Gay Street, Baltimore, MD 21202-4051. Rehabilitation Engineering Society of North America (RESNA), Suite 1540, 1700 North Moore Street, Arlington, VA 22209-1903.


Cooper, R. A. “Wheeled Mobility: Wheelchairs and Personal Transportation.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

141 Wheeled Mobility: Wheelchairs and Personal Transportation 141.1 141.2 141.3

Introduction Categories of Wheelchairs Wheelchair Structure and Component Design

141.4

Ergonomics of Wheelchair Propulsion

Materials • Frame Design • Wheels and Casters Kinematics • Kinetics • Net Joint Forces and Moments

141.5

Power Wheelchair Electrical Systems User Interface • Integrated Controls • Power System • Electromagnetic Compatibility

Rory A. Cooper University of Pittsburgh VA Pittsburgh Health Care System

141.6

Personal Transportation Vehicle Selection • Lift Mechanisms • Wheelchair Restraint Mechanisms • Hand Controls

141.1 Introduction Centuries ago, people with disabilities who survived for an extended period of time were transported on hammocks slung between poles which were carried by others. This was the preferred means of transportation of the upper class and thus carried no stigma. Later the wheelbarrow was developed and soon became a common mode of transportation for people with disabilities. Because wheelbarrows were used to transport materials, during this period in history, people with disabilities were looked upon as outcasts from society. During the renaissance, the French court popularized the first wheelchairs. Wheelchairs were overstuffed arm chairs with wheels placed upon them. This enabled movement, with assistance, indoors. Later the wooden wheelchair with wicker matting was developed. This type of chair remained the standard until the 1930s. Franklin D. Roosevelt was not satisfied with the wooden wheelchair and had many common metal kitchen chairs modified with wheels. In the 1930s a young mining engineer, named Everest experienced an accident that left him mobility impaired. He worked with a fellow engineer Jennings to develop steel wheelchairs. Within a few years, they formed a company Everest & Jennings to manufacture wheelchairs. Following World War II, medical advances saved the lives of many veterans with spinal cord injuries or lower limb amputations, who would have otherwise died. Veterans medical centers issued these veterans steel framed wheelchairs with 18 inch seat widths. These wheelchairs were designed to provide the veteran some mobility within the hospital and home, and not to optimize ergonomic variables. Just as among the ambulatory population, mobility among people with disabilities varies. Mobility is more of a functional limitation than a disability related condition. Powered mobility can have tremendous positive psycho-social effects on an individual. Power wheelchairs provide greater independence to thousands of people with severe mobility impairments.


Power wheelchairs began in the 1940’s as standard cross-brace folding manual wheelchairs adapted with automobile starter motors and an automobile battery. The cross-braced wheelchair remained the standard for a number of years. When the rigid power wheelchair frame was developed, space became available under the seat for electronic controls, respirators, communication systems, and reclining devices. By the mid-1970s, wheelchairs had evolved to the point where people had acquired a significant level of mobility. A personal automobile has a profound affect on a persons’ mobility and ability to participate in society. A wheelchair is suitable for short distances, and for many situations where an unimpaired person would walk. Modifications to vehicles may be as simple as a lever attached to the brake and accelerator pedals or as complex as a complete joystick controlled fly-by-wire system. Modifications to other components of the vehicle may be required to provide wheelchair access. An automobile may not be appropriate for some people who travel distances too long to be convenient with a wheelchair, but not long enough to warrant an adapted automobile. Micro-cars, enlarged wheelchairs which travel at bicycle speeds, are convenient for many people who wish only to travel to the local grocery store or post office. Micro-cars are also useful for people who like to travel along bicycle paths or drive short distances off-road.

141.2 Categories of Wheelchairs There are two basic classes of wheelchairs: manually powered, and externally powered. For practical purposes, externally powered wheelchairs are electrically powered wheelchairs. There are approximately 200,000 wheelchairs sold annually within the US of which about 20,000 are powered wheelchairs. Most wheelchairs are purchased by third-party-payers (e.g., insurance companies, government agencies). This requires the market to be responsive to wheelchairs user’s needs, prescriber expertise and experience, third-party-payer purchase criteria, and competition from other manufacturers. Despite the complicated interaction between these components, and the regulation of products by several government agencies, a variety of wheelchairs and options are available. Depot wheelchairs are intended for institutional use where several people may use the same wheelchair. Generally, these wheelchairs are inappropriate for active people who use wheelchairs for personal mobility. Depot wheelchairs are designed to be inexpensive, accommodate large variations in body size, to be low maintenance, and often to be attendant propelled. They are heavy and their performance is limited. A typical depot wheelchair will have swing away footrests, removable armrests, a single cross-brace frame and solid tires. People who have impairment of an arm and one or both lower extremities may benefit from a onearm drive wheelchair that uses a linkage connecting the rear wheels. This allows the user to push upon the pushrim of one wheel and to propel both wheels. To effectively turn the wheelchair, the user must have the ability to disengage the drive mechanism. Some people have weakness of the upper and lower extremities and can gain maximal benefit from wheelchair propulsion by combining the use of their arms and legs or by using only their legs. The design and selection of a foot-drive wheelchair depends greatly upon how the user can take greatest advantage of their motor abilities. Indoor spaces are more limited and one is often required to get close to furnishings and fixtures to use them properly. Indoor wheelchairs often use rear castors because of the manueverability of these designs. However, rear castor designs make the wheelchair less stable in lateral directions. Indoor wheelchairs typically have short wheelbases. All wheelchairs are not propelled by the person sitting in the wheelchair. In many hospitals and longterm care facilities, wheelchairs are propelled by attendants. Attendant propelled wheelchair designs must consider the rider and the attendant as users. The rider must be transported safely and comfortably. The attendant must be able to operate and easily maneuver safely and with minimum physical strain. Active users often prefer highly maneuverable and responsive wheelchairs which fit their physical and psychosocial character. The ultralight wheelchair evolved from the desire of wheelchair users to develop functional ergonomic designs for wheelchairs. Ultralight wheelchairs are made of such materials as aluminum, alloy steel, titanium, or composites. The design of ultralight wheelchairs allows a number of © 1997 by CRC Press LLC

features to be customized by the user or be specified for manufacture. The most common features of all are the light weight, the high quality of materials used in their construction, and their functional design. Many people can benefit from ultralight wheelchair designs. The desire to achieve better performance has led wheelchair users, inventors, and manufacturers to constantly develop specialized wheelchairs for sports. There is no real typical sports wheelchair as the design depends heavily on the sport. Basketball and tennis wheelchairs are often thought to typify sports wheelchair design. However, racing, field events or shooting wheelchairs have little in common with the former. Some wheelchairs are made to change configuration from reclining to sitting, and from sitting to standing. Most stand-up wheelchairs cannot be driven in the stand-up setting in order to insure safe and stable operation. Standing gives the user the ability to reach cabinet and counter spaces otherwise inaccessible. Standing has the additional advantage of providing therapeutic benefits, i.e., hemodynamic improvements, and amelioration of osteoporosis. Stairs and other obstacles persist despite the progress made in universal design. Stair-climbing wheelchairs are electrically powered wheelchairs designed to ascend and descend stairs safely under the occupant’s control. Stair-climbing wheelchairs are quite complicated, and often reconfigure themselves while climbing stairs. The additional power required to climb stairs often reduces the range of the wheelchair when compared to standard power wheelchairs.

141.3 Wheelchair Structure and Component Design Several factors must be considered when designing a wheelchair frame: what are the intended uses, what are the abilities of the user, what are the resources available, and what are the existing products available. These factors determine if and how the frame will be designed and built. Successful designs of wheelchairs can only be accomplished with continuous input from and interaction with wheelchair users. The durability, aesthetics, function, ride comfort, and cost of the frame are dependent on the materials for construction, the frame geometry, and fabrication methods. One of the issues that makes wheelchair design more complicated is the fact that many users are dependent upon wheeled mobility everyday, nearly all day.

Materials Most wheelchairs are made of either aluminum or steel. Some chairs are made of titanium or advanced composite materials, primarily carbon fiber, and in the future composite frames will probably begin to become more available. All of these materials have their strengths and weaknesses. Common aluminum wheelchairs are Tungsten Inert Gas (TIG) welded (i.e., electrically welded together in a cloud of inert gas). They are sometimes bolted together using lugs. Most aluminum wheelchair frames are constructed of round drawn 6061 aluminum tubing. This is one of the least expensive and most versatile of the heat treatable aluminum alloys. It has most of the desirable qualities of aluminum, has good mechanical properties and high corrosion resistance. It can be fabricated using most standard techniques. Most steel wheelchairs are made of mild steel (1040 or 1060) or chromium-molybdenum alloy (4130 or 4140) seamless tubing commonly called chro-moly. Mild steel is very inexpensive and easy to work with. It is wildly available, and performs well for many applications. However, it has a low strength to weight ratio compared to other materials. Chro-moly is widely used because of its weldability, ease of fabrication, mild hardenability, and high fatigueability. Commonly wheelchairs are made of tubing 0.028-0.035 inches in wall thickness but diameters vary depending on the expected loads from between 0.25 to 1.25 in. More and more of the high-end wheelchairs are made of titanium. Titanium is a lightweight, strong, nonferrous metal. Titanium wheelchair frames are TIG welded. Titanium is the most exotic of the metals used in production wheelchairs and the most expensive. Titanium requires special tooling and skill to be machined and welded. It has very good mechanical properties and high corrosion resistance. It is resilient to wear and abrasion. Titanium is used because of its availability, appearance, corrosion resistance, very good strength and light weight. A drawback of titanium, besides cost, is that titanium once worn or if flawed may break rapidly (i.e., it has a tendency towards brittle fractures).


Advanced composites have been in use in aerospace and industrial applications for a number of years. These materials include Kevlar, carbon fiber, and polyester limestone composite. These materials are now making the transition to wheelchair design. Kevlar is an organic fiber which is yellow in color and soft to the touch. It is extremely strong and tough. It is one of the lightest structural fabrics on today’s market. Kevlar is highly resistant to impact, but its compression strength is poor. Carbon fibers are made by changing the molecular structure of Rayon fibers by extreme stretching and heating. Carbon fiber is very stiff (high modulus of elasticity), very strong (high tensile strength) and has very low density (weight for a given volume). Composites come as cloth or yarn. Composite cloth is woven into bidirectional or unidirectional cloth. Unidirectional weaves can add strength along a particular direction. Composites must be bound together by resin or epoxy. Generally, polyester resins or various specialty epoxies (e.g., Safe-T-Poxy) are used. To achieve greatest strength a minimum amount of epoxy must be used while wetting all of the fibers. This is often achieved through a process called bagging. To increase the strength and stiffness of structural components a foam (e.g., styrofoam, urethane or PVC) core is used. The strengthening occurs because of the separation of the cloth layers (it now becomes more like a tube than a flat sheet). Polyesther limestone composites have been used widely in industrial high voltage electrical component enclosures. A blend of polyester and limestone are used to form a mixture which can be molded under pressure and heated to form a stiff and durable finished product. Polyester limestone composites have high impact strength, and hold tolerances well, but have substantially lower strength to weight ratios than other composites. Their primary advantage is cost, polyester limestone composites are very inexpensive and readily available. Composites can be molded into elaborate shapes which opens a multitude of possibilities for wheelchair design.

Frame Design Presently all common wheelchair frames center around tubular construction. The tubing can either be welded together, or bolted together using lugs. There are two basic common frame styles: box frame and the cantilever frame. The box frame is named such because of its rectangular shape, and that tubes outline the edges of the “box”. Box frames can be very strong and very durable. A cantilever frame is named so because the front and rear wheels, when viewing the chair from the side, appear to be connected by only one tube; this is similar to having the front wheels attached to a cantilever beam fixed at the rear wheels. Both frame types require cross bracing to provide adequate strength and stiffness. The box frame provides great strength and rigidity. If designed and constructed properly the frame only deflects minimally during normal loading, and most of the suspension is provided by the seat cushion, the wheels and the wheel mounting hardware. Many manufacturers do not triangulate their box frame designs to allow some flexibility. The cantilever frame is based upon a few basic principles: (1) the frame can act as a suspension, (2) there are fewer tubes and they are closer to the body which may make the chair less conspicuous, and (3) there are fewer parts and fewer welds which makes the frame easier to construct.

Wheels and Casters Casters can be as small as 2 in. in diameter or as large as 12 in. in diameter for wheelchairs designed for daily use. Casters are either pneumatic, semi-pneumatic, or solid (polyurethane). Pneumatic casters offer a smoother ride at the cost of increased maintenance, whereas polyurethane casters are very durable. Semi-pneumatic tires offer a compromise. Most active users prefer 5 in. polyurethane casters or 8 inch pneumatic casters for daily use. An 8-in. caster offers a better ride comfort at the expense of foot clearance. Caster foot clearance is maximized with a 2-in. “Roller Blade” caster often used for court sports (e.g., basketball, tennis, and racquetball). Rear wheels come in three common sizes 22, 24, and 26 in. They come in two styles: spoked and MAG. MAG wheels are typically made of plastic and are die cast. MAG wheels require minimal maintenance, and wear well. However, spoked wheels are substantially lighter, more responsive, and are generally preferred by active manual wheelchair users. Rear tires can be two


types; pneumatic or puncture proof. Pneumatic tires can either use a separate tube and tire or a combined tube and tire (sew-up). Commonly, a belted rubber tire with a Butyl tube (65 psi) is used. However, those desiring higher performance prefer sew-up tires or Kevlar belted tires with high pressure tubes (180 psi). Puncture proof tires are heavier, provide less suspension, and are less lively than pneumatic tires. The chair must be designed to optimize the interaction of the wheels with the ground. Four critical performance factors need to be considered: (1) Castor Flutter, (2) Castor Float, (3) Tracking, and (4) Alignment. Castor flutter is the shimmy (rapid vibration of the front wheels) that may occur on some surfaces above certain speeds. When one of the castors does not touch the floor when on level ground, the wheelchair has castor float. Castor float decreases the stability and performance of the wheelchair. A manual wheelchair uses rear wheel steering via differential propulsion torque. Tracking is the tendency of the wheelchair/rider to maintain its course once control has been relinquished. Tracking is important, as the rider propels the handrims periodically (about every second) and if the chair does not track well it will drift from its course between pushes and force the rider to correct heading. This will waste valuable energy and reduce control over the chair. Alignment generally refers to the orientation of the rear wheels with respect to one another. Typically, it is desirable to have the rear wheels parallel to one another without any difference between the distance across the two rear wheels at the front and back. Misalignment on the order of 1/8 inch can cause a noticeable increase in the effort required to propel the wheelchair.

141.4 Ergonomics of Wheelchair Propulsion The most important area of wheelchair design and prescription is determining the proper interaction between the wheelchair and the user. This can lead to reducing the risk of developing repetitive strain injury while maximizing mobility. Cardiovascular fitness can be improved through exercise which requires a properly fitted wheelchair.

Kinematics Kinematic data by itself does not provide sufficient information for the clinician to implement appropriate rehabilitation intervention strategies or for the engineer to incorporate this information into wheelchair design changes. Kinematic data are commonly collected at 60 Hz, which is the maximum frequency of many videotape-based systems. Kinematic data analysis shows that experienced wheelchair users contact the pushrim behind top-dead-center and push to nearly 90° in front of top-dead-center. This is significantly longer than non-wheelchair users. Lengthening the stroke permits lowering the propulsion moment and may place less stress on the users’ joints. An important aspect of the evaluation, and possible retraining of wheelchair users is to determine the optimal stroke kinetics and kinematics [2]. However, there is typically some degree of variation from one stroke to another. Wheelchair propulsion kinematic data are typically cyclic (i.e., a person repeats or nearly repeats his/her arm motions over several strokes). Each marker of the kinematic model (e.g., shoulder, elbow, wrist, knuckle) of each subject generates an x and y set of data which is periodic. The frequencies of the x and y data are dependent upon the anthropometry of the individual, the construction of the wheelchair, and the speed of propulsion. The periodic nature of the kinematic data for wheelchair propulsion can be exploited to develop a characteristic stroke from a set of kinematic data (with the rear hub of the wheelchair chosen as the origin) including several strokes.

Kinetics The SMARTwheel was developed to measure the pushrim forces required for evaluating net joint forces and moments to allow the clinician and researcher to study the level of stress experienced by the joint structures during wheelchair propulsion. The SMARTwheel uses a standard wheelchair wheel fitted with three beams 120° apart and each has two full strain gage bridges. The strain gage bridges are each interfaced through an instrumentation amplifier to a micro-controller which transmits the data through


a mercury slip-ring to the serial port of a computer. Kinetics of wheelchair propulsion are affected by speed of propulsion, injury level, user experience, and wheelchair type and fit. Van Der Woude et al. have reported on an ergometer which detected torque by way of a force transducer located in the wheel center and attached to what is referred to as the “wheel/hand rim construction” [7]. The ergometer was adjusted for each subject’s anthropometric measurements. Data were sampled at 100 Hz for 7.5 second periods with a digital filter cut-off frequency of 10 Hz. Mean and peak torque increased with mean velocity, a maximum mean peak torque of 31 N-m occurred at 1.27 m/s. Torque curves of inexperienced subjects showed an initial negative deflection and a dip in the rising portion of the curve. Torque curves were reported to be in agreement with results of previous investigations [8,9,10]. Brauer and Hertig [12] measured the static torque produced on push-rims which were rigidly restrained by springs and mounted independent of the tires and rims of the wheelchair. The spring system was adjustable for the subject’s strength. The wheels were locked in a fixed position. Torque was measured using slide-wire resistors coupled to the differential movements between the push-rim and wheels and recorded using a strip chart recorder. Subjects were asked to grasp the push-rim at six different test positions (-10, 0, 10, 20, 30, and 40 degrees relative to vertical) and to use maximal effort to turn both wheels forward. Male subjects (combined ambulatory and wheelchair users) produced torques of 27.9 to 46.6 N-m and female subjects produced torques of 17.1 to 32.1 N-m [12]. Grip location, handedness, grip strength, and how well the test wheelchair fit the anthropometric measurements of the individual affected the torque generated. Problems encountered were slipperiness of the push-rims due to a polished finish and limited contact due to the small diameter of the push-rim tubing (12.7 mm or 1/2 in). The use of one wheelchair for all subjects presented the problem of variations due to inappropriate fit for some individuals. Brubaker, Ross, and McLaurin [13] examined the effect of horizontal and vertical seat position (relative to the wheel position) on the generation of static push-rim force. Force was measured using a test platform with a movable seat and strain gauged beams to which the push-rims were mounted. Pushing and pulling forces were recorded using a strip chart recorder. Static force was measured for four grip positions (–30, 0, 30, and 60 degrees) with various seat positions. Push-rim force ranged from approximately 500 to 750 N and varied considerably with seat position and rim position [13].

Net Joint Forces and Moments Net joint forces and moments acting at the wrist, elbow and shoulder during wheelchair propulsion provide scientists and clinicians with information related to the level of stress borne by the joint structure. Joint moments and forces are calculated using limb segment and joint models, anthropometric data, kinetic data, and kinematic data. Joint moments data shows that forces at each joint vary among subjects in terms of peak forces, where they occur during the propulsion phase, and how quickly they develop. Peak net joint moments occur at different joint angles for different subjects, and conditions (e.g., speed, resistance). Convention for joint angles is that 180 degrees at the elbow represents full extension; while at the wrist, this is the hand in the neutral position (flexion less than 180 degrees and extension greater than 180 degrees). Joint angles at the shoulder are determined between the arm and the trunk, with zero measured at the point where the trunk and arm are aligned. Wheelchair users show maximum net shoulder moment between 20° and 40° of extension. Some wheelchair users also show a rapid rise in the elbow extensor moment at the beginning of the stroke with the elbow at about 120°. This moment value begins to decrease between 150° and 170°. At the wrist, the peak moments occur between 190° and 220°. Net joint moments and force models need to account for hand center of pressure, inaccuracies in anthropometric data, and joint models related to clinical variables.

141.5 Power Wheelchair Electrical Systems Some people are impaired to an extent that they would have no mobility without a power wheelchair. However, some people may have limited upper body mobility and may have the ability to propel a manual wheelchair for short distances. These people may liken using a power wheelchair to admitting defeat.


However, a power wheelchair may provide greater mobility. In such cases it may be best to suggest a power wheelchair for longer excursions, and a manual wheelchair for in home and recreational use.

User Interface Power wheelchairs are often used in conjunction with a number of other adaptive devices. For people with severe mobility impairments, power wheelchairs may be used with communication devices, computer access devices, respirators, and reclining seating systems. The integration of the users’ multiple needs must also be considered when designing or prescribing a power wheelchair. The joystick is the most common control interface between the user and the wheelchair. Joysticks produce voltage signals proportional to displacement, force, or switch closures. Displacement joysticks are most popular. Displacement joysticks may use either potentiometers, variable inductors (coils) or optical sensors to convert displacement to voltage. Inductive joysticks are most common as they wear well and they can be made quite sensitive. Joysticks can be modified to be used for chin, foot, elbow, tongue or shoulder control. Typically, short throw joysticks are used for these applications. Force sensing joysticks use three basic transducers: simple springs and dampeners on a displacement joystick, cantilever beams with strain gages, and fluid with pressure sensors. Force sensing joysticks which rely on passive dampeners or fluid pressure generally require the user to have a range of motion within normal values for displacement joysticks users. Beam-based force sensing joysticks require negligible motion, and hence may be used for people with limited motion abilities. People who exhibit intention or spastic tremor or with multiple sclerosis may require special control considerations. Signal processing techniques are often required to grant the user greater control over the wheelchair. Typically, signal averaging or a low pass filter with a cut-off frequency of below 5 Hertz is used. The signal processing is typically incorporated into the controller. Some people lack the fine motor control to effectively use a joystick. An alternative for these people is to use a switch control or a head position control. A switch control simply uses either a set of switches or a single switch and a coded input, i.e., Morse code or some other simple switch code. The input of the user is latched by the controller and the wheelchair performs the task commanded by the user. The user may latch the chair to repeatedly perform a task a specified number of times, e.g., continue straight until commanded to do otherwise. Switch control is quite functional, but it is generally slower than joystick control. Switch inputs can be generated in many ways. Typically, low pressure switches are used. The input can come from a sip-and-puff mechanism which works off of a pressure transducer. A switch contact is detected when the pressure exceeds or drops below a threshold. The pressure sensor may be configured to react to pressure generated by the user blowing into or sipping from an input or by the user simply interrupting the flow in or out of a tube. Sip-and-puff may also be used as a combination of proportional and switch control. For example, the user can put the control in the “read speed” mode and then the proportional voltage output from the pressure transducer will be latched as the user-desired speed. Simple switches of various sizes can be used to control the chair using many parts of the body. Switches may be mounted on the armrests or a lap tray for hand or arm activation, on the footrest(s) for foot activation, or on a head rest for head activation. The motion of the head can also be used for proportional control by using ultrasonic sensors. Ultrasonic sensors can be mounted in an array about the headrest. The signal produced by the ultrasonic sensors is related to the position of the head. Hence, motion of the head can be used to create a proportional control signal. Ultrasonic head control and switch control can be combined to give some users greater mastery over their power wheelchair. Switches can be used to select the controller mode, whereas the ultrasonic sensors give a proportional input signal. A critical consideration when selecting or designing a user interface is that the ability of the user to accurately control the interface is heavily dependent upon the stability of the user within the wheelchair. Often custom seating and postural support systems are required for a user interface to be truly effective. The placement of the user interface is also critical to its efficacy as a functional control device.


Integrated Controls People with severe physical impairments may only be able to effectively manipulate a single input device. Integrated controls are used to facilitate using a single input device (e.g., joystick, head switches, voice recognition system) to control multiple actuators (e.g., power wheelchair, environmental control unit, manipulator). This provides the user with greater control over the environment. The M3S-Multiple Master Multiple Slave bus is designed to provide simple, reliable access to a variety of assistive devices. Assistive devices include input devices, actuators, and end-effectors. M3S is based on the Computer Area Network (CAN) standard. A wide range of organizations provide assistive devices which offer the opportunity of functioning in a more independent manner. However, many of these devices and systems are developed without coordination resulting in incompatible products. Clinicians and users often desire to combine products from various sources to achieve maximal independence. The result is to have several devices with their own input devices and overlapping functions. Integrated controls provide access to various end-effectors with a single input device. The M3S provides an electronic communication protocol so that the system operates properly. M3S is an interface specification with a basic hardware architecture, a bus communication protocol, and a configuration method. The M3S standard incorporates CAN plus two additional lines for greater security (i.e., 7 wire bus, 2 power lines, 2 CAN lines, safety lines, 1 shield, and 1 harness line). The system can be configured to each individual’s needs. An M3S system consists of a microcontroller in each device and a control and configuration module (CCM). The CCM insures proper signal processing, system configuration, and safety monitoring. The CCM is linked to a display (e.g., visual, auditory, tactile) which allows the user to select and operate each end-effector. Any M3S compatible device can communicate with another M3S compatible device. M3S is an International Organization of Standards (ISO) open system communication implementation.

Power System To implement a motor controller, a servo-amplifier is required to convert signal level power (volts at milliamps) to motor power (volts at amps). Typically, a design requirement for series, shunt, and brushless motor drives is to control torque and speed, and hence, power. Voltage control can often be used to control speed for both shunt and series motors. Series motors require feedback to achieve accurate control. Either a linear servo-amplifier or a chopper can be used. Linear servo amplifiers are not generally used with power wheelchairs primarily because of their lower efficiency than chopper circuits. A motor can be thought of as a filter to a chopper circuit, in this case, the switching unit can be used as part of a speed and current control loop. The torque ripple and noise associated with phase control drives can be avoided by the use of high switching frequencies. The response of the speed control loop is likewise improved with increasing switching frequency. Motor torque is proportional to the armature current in shunt motors and to the square of the current in series motors. The conduction loss of the motor and servo-amplifier are both proportional to the current squared. Optimal efficiency is achieved by minimizing the form factor, (Irms /Imean ). This can be done by increasing the switching frequency to reduce the amplitude of the ripple. Benefits of increased efficiency are increased brush life, gear life, and lower probability of field permanent magnet demagnetization. Switching or chopper drives are classified as either unidirectional or bi-directional. They are further divided by whether they use dynamic braking. Typically, power wheelchairs use bi-directional drives without dynamic braking. However, scooters may use unidirectional drives. The average voltage delivered to the motor from a switching drive is controlled by varying the duty cycle of the input waveform. There are two common methods of achieving this goal: (1) fixed pulse width, variable repetition rate, and (2) pulse-width modulation (PWM). Power wheelchair servo-amplifiers typically employ PWM. Pulse width modulation at a fixed frequency has no minimum on-time restriction. Therefore, current peaking and torque ripple can be minimized. For analysis, a d.c. motor can be modeled as a RL circuit, resistor and inductor, or in series with a voltage source. If the motor current is assumed continuous, then the minimum and maximum motor current can be represented by Eq. (141.1).


Imin = Imax =

e

( )

(1 − e (

) − R L t −t 1 − e ( ) ( on off )

− R L t off

− R L t on

) V −V s

R

gen

R

( ) Vs Vgen − − ( R L ) ( t on − t off ) R R 1− e 1− e

− R L t on

Two basic design principles are used when designing switching servo-amplifiers: (1) Imax should be limited to five times the rated current of the motor to ensure that demagnetization does not occur, (2) the ripple, (Imax–Imin)/Iavg , should be minimized to improve the form factor, and reduce the conduction loss in the switching devices. To achieve low ripple, either the inductance has to be large or the switching frequency has to be high. Permalloy powder cores can be used to reduce core loss at frequencies above a few kilohertz. However, this comes at the cost of the electrical time constant of the motor, degrading the motor response time. Hence, raising the switching frequency is most desirable. A power MOSFET has the ability to switch rapidly without the use of load-shaping components. There are several motor types that may be suitable for use with power wheelchairs. Most current designs use permanent magnet direct current motors. These motors provide high torque, high starting torque and are simplest to control. Permanent magnet direct current motors can either be controlled in what are commonly called current mode or voltage mode. These modes developed out of designs based upon controlling torque and speed, respectively. Alternating current motors can be designed to be highly efficient and can be controlled with modern power circuitry. Because of the development and wide spread dissemination of switching direct current converters, it is quite feasible to use alternating current motors with a battery supply. To date, alternating current motors have only been used in research on power wheelchairs. The output of the motor is controlled by varying the phase or the frequency. The battery energy storage system is recognized as one of the most significant limiting factors in powered wheelchair performance. Battery life and capacity are important. If battery life can be improved, the powered wheelchair user will have longer, reliable performance from his/her battery. An increase in battery capacity will allow powered wheelchair users to travel greater distances with batteries that weigh and measure the same as existing wheelchair batteries. Most importantly, increases in battery capacity will enable the use of smaller and lighter batteries. Because batteries account for such a large proportion of both the weight and volume of current powered wheelchair systems, wheelchair manufacturers must base much of their design around the battery package. Power wheelchairs typically incorporate 24 volt d.c. energy systems. The energy for the wheelchair is provided by two, deep cycle lead-acid batteries connected in series. Either wet cell or gel cell batteries are used. Wet cell batteries also cost about one-half as much as gel cell batteries. Gel cells may be required for transport by commercial air carriers. Battery technology for wheelchair users remains unchanged despite the call for improvements by power wheelchair users. This may be in part due to the relatively low number of units purchased, about 500,000 per annum, when compared to automotive applications with about 6.6 million per annum by a single manufacturer. Wheelchair batteries are typically rated at 12 Volts and 30-90 ampere-hours capacity at room temperature. A power wheelchair draws about 10 amps during use. The range of the power wheelchair is directly proportional to the ampere-hour rating for the operating temperature. Batteries are grouped by size. Group size is indicated by a standard number. The group size defines the dimensions of the battery as shown in Table 141.1. The ampere-hour rating defines the battery’s capacity. It is important that the appropriate charger be used with each battery set. Many battery chargers automatically reduce the amount of current delivered to the battery as the battery reaches full charge. This helps to prevent damage to the battery from boiling. The rate at which wet and gel cell batteries charge is significantly different. Some chargers are capable of operating with both types of batteries.


Table 141.1 Standard Power Wheelchair Battery Group Sizes Group number

Length

Width

Height

U1 22NF 24 27

7-3/4 9-7/16 10-1/4 12-1/16

5-3/16 5-1/2 6-13/16 6-13/16

7-5/16 8-15/16 8-7/8 8-7/8

Note: Units are in inches.

Many require setting the charger for the appropriate battery type. Most wheelchair batteries connected in series are charged simultaneously with a 24 Volt battery charger.

Electromagnetic Compatibility Powered wheelchairs have been reported to exhibit unintended movement. Wheelchair manufacturers and the US Food and Drug Administration Center for Devices and Radiological Health (FDA-CDRH) have examined the susceptibility of powered wheelchairs and scooters to interference from radio and microwave transmissions. These devices are tested at frequencies ranging from 26 MHz to 1 GHz, which is common for transmissions (e.g., radio, TV, microwave, telephones, mobile radios). Power wheelchairs incorporate complex electronics and microcontrollers which are sensitive to electromagnetic (EM) radiation, electrostatic discharge (ESD), and other energy sources. Electric powered wheelchairs may be susceptible to electromagnetic interference (EMI) present in the ambient environment. Some level of EMI immunity is necessary to ensure the safety of power wheelchair users. Electromagnetic compatibility (EMC) is the term used to describe how devices and systems behave in an electromagnetic environment. Because of the complexity of power wheelchairs and scooters, and the interaction with an electromagnetic environment, susceptibility to interference cannot be calculated or estimated reliably. A significant number of people attach accessories (e.g., car stereos, computers, communication systems) to their power wheelchairs which also share the batteries. This may increase the susceptibility of other system components to EMI. A number of companies make electric powered devices designed to operate on power wheelchairs to provide postural support, pressure relief, environmental control, and motor vehicle operation. These devices may alter the EMI compatibility of power wheelchairs as provided by the original equipment manufacturer (OEM). Wheelchairs and accessories can be made to function properly within EM environments through testing. Field strengths have been measured at 20 V/m from a 15 Watt hand held cellular telephone, and 8 V/m from a 1 Watt hand held cellular telephone [2]. The FDA-CDRH has tested power wheelchairs and scooters in a Gigahertz Transverse Electromagnetic (GTEM) cell, and in an anechoic chamber with exposure strengths from 3 V/m to 40 V/m. The US FDA requires that a warning sticker be placed on each power wheelchair or scooter indicating the risk due to EMI. Two tests are commonly performed on chairs: brake release and variation in wheel speed. The device(s) used for measuring wheel speed and brake release must not significantly alter the field. The brakes shall not release or the wheels are not to move with a wheel torque equivalent to a 1:6 slope with a 100 kilogram rider when the wheelchair is exposed to EM radiation. Non-electrical contact methods (e.g., audio sensing, optical sensing) of measuring brake release or wheelchair movement are preferable. Nominal wheel speed may drift over the length of the test. This drift is primarily due to drop in battery charge over the test interval. Wheel speed must be recorded without EM interference between test intervals. The percentage change in wheel speed during exposure to EM interference shall be referenced to the nominal wheel speed for that test interval. The variation in absolute forward speed, (vemR + vemL )/2, is to be within 30% of the nominal forward speed, (vnomR + vnomL )/2. The differential speed between the two wheels should be within 30% of each other, 2•(vemR – vemL )/(vnomR + vnomL ).The test frequency must be held long enough to accommodate the slowest time constant (time required to reach 63% of maximum or minimum) of parameters related to wheelchair driving behavior. Currently two seconds is used by the FDA-CDRH test laboratory.


141.6 Personal Transportation Special adaptive equipment requirements increase with the degree of impairment and desired degree of independence in areas such as personal care, mobility, leisure, personal transportation, and employment. People are concerned that they receive the proper equipment for them to safely operate their vehicle. Access and egress equipment have the greatest maintenance requirements. Other devices such as hand controls, steering equipment, securement mechanisms, and interior controls require less maintenance. Most users of adaptive driving equipment are satisfied with the performance of such equipment. Most frequent equipment problems are minor and are repaired by consumers themselves. Physical functional abilities such as range of motion, manual muscle strength, sensation, grip strength, pinch strength, fine motor dexterity, and hand-eye coordination all may be related to driving potential. Driving characteristics must also be evaluated when determining an individual’s potential to safely operate a motor vehicle. Throttle force, brake force, steering force, brake reaction time, and steering reaction time are all factors which influence an individual’s driving potential.

Vehicle Selection While it is often a difficult task to find an automobile which meets the specific needs of a particular wheelchair user, currently, no automobile meets the needs of all wheelchair users. Automotive consumers with disabilities are also concerned about ease of entry, stowage space for the wheelchair, and seat positioning. Reduced size, increasingly sloping windshields, lower roofs, and higher sills of new cars make selecting a new vehicle difficult for wheelchair users. The ability to load the wheelchair into a vehicle is essential. Some individuals with sufficient strength and a suitable vehicle are able to stow their wheelchairs inside the vehicle without the use of assistive devices. Many people must rely on an external loading device. The selection of the appropriate vehicle should be based upon the client’s physical abilities and social needs. An approach some people have used to overcome the problems associated with a smaller car is to use a car-top wheelchair carrying device. These devices lift the wheelchair to the top of the car, fold, and stow it. They have been designed to work with four door sedans, light trucks, and compact automobiles. There are several critical dimensions to an automobile when determining wheelchair accessibility. The wheelbase of the automobile is often used by auto manufacturers to determine vehicle size (e.g., fullsize, mid-size, compact). Typical ranges for passenger vehicles are presented in Table 141.2.

Lift Mechanisms Many wheelchair users who cannot transfer into a passenger vehicle seat or prefer a larger vehicle, drive vans equipped with wheelchair lifts. Platform lifts may use a lifting track, a parallelogram lifting linkage, or a rotary lift. Lift devices are either electromechanically or electrohydraulically powered. The platform often folds into the side doorway of the van. Crane lifts, also called swing-out lifts, have a platform which elevates and folds or rotates into the van. Lifts may either be semi-automatic or automatic. In many cases TABLE 141.2 Typical Ranges of Accessibility Dimensions for Sedans Wheelbase Door height Door width Headroom Max. Space behind seat Min. Space behind seat Seat-to-ground distance Width of door opening Note: Units in inches.


93–108 33–47 41–47 36–39 9–19 4–9 18–22 38–51

semi-automatic lifts require the user to initiate various stages (e.g., unlocking door, door opening, lowering lift) of the lifting process. Automatic lifts are designed to perform all lift functions. They usually have an outside key operated control box, or an interior radio controlled control box. Some lifts use electrohydraulic actuators to lift and fold, with valves and gravity used to lower the lift. Crane lifts may swing out from a post at the front or rear of the side door. Interlocking mechanisms are available with some lifts to prevent the lift from being operated while the door is closed. The Society of Automotive Engineers (SAE) has developed guidelines for the testing of wheelchair lift devices for entry and exit from a personal vehicle. The standards are intended to set an acceptable level of reliability and performance for van lifts.

Wheelchair Restraint Mechanisms Securement systems are used to temporarily attach wheelchairs to vehicles during transport. Many wheelchair users can operate a motor vehicle from their wheelchair, but are unable to transfer into a vehicle seat. Auto safety standards have reduced the number of US automobile accident fatalities despite an increase in the number of vehicles. The crash pulse determines the severity of the collision of the test sled, and hence, simulates real-world conditions. Securement systems are tested with a surrogate wheelchair at 30 miles per hour (48 +2/–0 kilometers per hour) with a 20g deceleration. Wheelchairs must be safely restrained when experiencing an impact of this magnitude and no part of the wheelchair shall protrude into the occupant space where it might cause injury. Proper use of lap and shoulder belts is critical to protecting passengers in automobiles seats. A similar level of crash protection is required for individuals who remain in their wheelchairs during transportation. Wheelchairs are flexible, higher than a standard automobile seat, and not fixed to the vehicle. The passenger is restrained using a harness of at least one belt to provide pelvic restraint and two shoulder or torso belts that restrain both shoulders. A head support may also be used to prevent rearward motion of the head during impact or rebound. A three point restraint is the combination of a lap belt and a shoulder belt (e.g., pelvic torso restraint, lap-sash restraint, lap-shoulder restraint). The relationship between injury criteria and the mechanics of restraint systems are important to insure the safety of wheelchair users in motor vehicles. Hip and head deflection are often used criteria for determining potential injury. The automotive industry has invested considerable effort for research and development to protect vehicle passengers. Research is not nearly extensive for the passenger who remains seated in a wheelchair while traveling. Many wheelchair and occupant restraint systems copy the designs used for standard automobile seats. However, this type of design may not be appropriate. Crash tests have shown that for 10 or 20 g impacts of 100 millisecond duration, people may sustain injuries despite being restrained. When shoulder belts mounted 60 inches above the floor were used to restrain a 50th percentile male dummy, it was found that the torso was well controlled, and head and chest excursions were limited. When shoulder belts were anchored 36 inches above the floor they were ineffective in controlling torso movement. Kinematic results and head injury criteria (HIC) can be used to estimate the extent of injury sustained by a human passenger. A HIC of 1000 or greater indicates a serious or fatal head injury. Generally, a HIC value approaching or exceeding 1000 is indicative of head impact with some portion of the vehicle interior. The open space typically surrounding a wheelchair user in a public bus precludes impact with the buses’ interior. High HIC values may occur when the torso is effectively restrained and there is a high degree of neck flexion. If the chin strikes the chest then there may be an impact great enough to cause head injury.

Hand Controls Hand-controls are available for automatic and manual transmission vehicles. However, hand-controls for manual transmission automobiles must be custom made. There are also portable hand-controls, and long-term hand controls. Portable hand-controls are designed to easily attach to most common automobiles with a minimal number of tools. Hand-controls are available for either left or right hand control.


Many hand controls are attached to the steering column. Hand-controls either clamp to the steering column or are attached to a bracket which is bolted to the steering column or dash, typically where the steering column bolts to the dash. Installation of the hand-controls should not interfere with driver safety features (e.g., air bags, collapsible steering columns). The push rods of the hand-control either clamp directly to the pedals or the levers connected to them. Most systems activate the brakes by having the driver push forward on a lever with a hand grip. This allows the driver to push against the back of the seat, creating substantial force, and braces the driver in the event of a collision. The throttle, or gas pedal, is operated in a number of ways. Some systems use a twist knob or motorcycle type throttle. Other systems actuate the throttle by pulling on the brake throttle lever. Another method is to rotate the throttle-brake lever downwards (i.e., pull the lever towards the thigh at a right angle to the brake application to operate the throttle). It is common to have the same vehicle driven by multiple people which may require the vehicle to be safely operated with hand-controls, and the OEM foot controls. Care must be taken to insure that the lever and brackets of the hand-controls do not restrict the driving motions of foot control drivers. Many people have the motor control necessary to operate a motor vehicle, but they do not have the strength required to operate manual hand-controls. Automatic or Fly-By-Wire hand-controls use external actuators (e.g., air motors, servo mechanisms, hydraulic motors) to reduce the force required to operate various vehicle primary controls. Power steering, power brakes, six-way power seats, and power adjustable steering columns can be purchased as factory options on many vehicles. Six-way power seats are used to provide greater postural support and positioning than standard automotive seats. They can be controlled by a few switches to move fore-aft, incline-recline, and superiorinferior. This allows the user to position the seat for easy entry and exit, and for optimal driving comfort. Power adjustable steering columns also make vehicles more accessible. By using a few buttons, the steering column can be tilted upwards or downwards allowing positioning for entry/exist into the vehicle, and for optimal driving control. Custom devices are available for people who require more than the OEM options for power assistance. Microprocessor and electronic technology have dramatically changed how motor vehicles are designed. Many functions of an automobile are controlled electronically or with electromechanical-electrohydraulic controls. This change in vehicle design has made a wide variety of options available for people who require advanced vehicle controls. Many automobiles use electronic fuel injection. Electronic fuel injection systems convert the position of the accelerator pedal to a serial digital signal which is used by a microcontroller to inject the optimal fuel-air mixture into the automobile at the proper time during the piston stroke. The electronic signal for the accelerator position can be provided by another control device (e.g., joystick, slide-bar).

References 1. Adams T.C., Reger S.I. 1993. Factors affecting wheelchair occupant injury in crash simulation. Proc. 16th Annual RESNA Conference, Las Vegas, NV, pp. 80-82. 2. Adams T.C., Sauer B., Reger S.I. 1992. Kinematics of the wheelchair seated body in crash simulation. Proc. RESNA Intl. ’92, Toronto, Ontario, Canada, pp. 360-362.; 3. Asato K.T., Cooper R.A., Robertson R.N., and Ster J.F. 1993. SMARTWheels: Development and Testing of a System for Measuring Manual Wheelchair Propulsion Dynamics, IEEE Trans. Biomed. Eng. 40(12):1320-1324. 4. Aylor J.H., Thieme A., Johnson B.W. 1992. A battery state-of-charge indicator. IEEE Trans. Ind. Electr., 39(5):398-409. 5. Boninger M.L., Cooper R.A., Robertson R.N., Shimada S.D. 1997, Three-dimensional pushrim forces during two speeds of wheelchair propulsion, Am. J. Phys. Med. Rehab., 76:420-426. 6. Brauer R.L. and Hertig B.A. 1981. Torque generation on wheelchair handrims, Proc. 1981 Biomechanics Symp., ASME/ASCE Mechanics Conference, pp. 113-116.


7. Brienza D.M., Cooper R.A., Brubaker C.E. 1996, Wheelchairs and seating, Curr. Opinion in Orthopedics, 7(6):82-86. 8. Brubaker C.E.,Ross S., and McLaurin C.A. 1982. Effect of seat position on handrim force, Proc. 5th Annual Conference on Rehabilitation Engineering, p. 111. 9. Cooper R.A. 1998, Wheelchair Selection and Configuration, Demos Medical Publishers, New York, NY. 10. Gray D.B., Quatrano L.A., Lieberman M.L. 1998, Designing and Using Assistive Technology, Brookes Publishing Company, Baltimore, MD. 11. Cooper R.A., Trefler E., Hobson D.A. 1996, Wheelchairs and seating: issues and practice, Technology and Disability, 5:3-16. 12. Cooper R.A. 1995, Intelligent control of power wheelchairs, IEEE Engineering in Medicine and Biology Magazine, 15(4):423-431. 13. Cooper R.A. 1995, Rehabilitation Engineering Applied to Mobility and Manipulation, Institute of Physics Publishing, Bristol, United Kingdom. 14. Cooper R.A., Gonzalez J.P., Lawrence B.M., Rentschler A., Boninger M.L., VanSickle D.P. 1997, Performance of selected lightweight wheelchairs on ANSI/RESNA tests, Arch. Phys. Med. Rehab., 78:1138-1144. 15. Cooper R.A. 1996, A perspective on the ultralight wheelchair revolution, Technology and Disability, 5:383-392. 16. Cooper R.A., Robertson R.N., Lawrence B., Heil T., Albright S.J., VanSickle D.P., Gonzalez J.P. 1996, Life-cycle analysis of depot versus rehabilitation manual wheelchairs, J. Rehab. Res. Dev., 33(1):4555. 17. Cooper R.A. 1993. Stability of a Wheelchair Controlled by a Human Pilot, IEEE Trans. on Rehabilitation Engineering. 1(4):193-206. 18. Cooper R.A., Baldini F.D., Langbein W.E., Robertson R.N., Bennett P., and Monical S. 1993. Prediction of Pulmonary Function in Wheelchair Users, Paraplegia. 31:560-570. 19. Cooper R.A., Horvath S.M., Bedi J.F., Drechsler-Parks D.M., and Williams R.E. 1992. Maximal Exercise Responses of Paraplegic Wheelchair Road Racers, Paraplegia. 30:573-581. 20. Cooper R.A. 1991. System Identification of Human Performance Models, IEEE Trans. on Systems, Man, and Cybernetics. 21(1):244-252. 21. Cooper R.A. 1991. High Tech Wheelchairs Gain the Competetive Edge, IEEE Engineering in Medicine and Biology Magazine. 10(4):49-55. 22. Kauzlarich J.J., Ulrich V., Bresler M., Bruning T. 1983. Wheelchair Batteries: Driving cycles and testing. J. Rehab. Res. Dev. 20(1):31-43. 23. MacLeish M.S., Cooper R.A., Harralson J., and Ster J.F. 1993. Design of a Composite Monocoque Frame Racing Wheelchair, J. Rehab. Res. Dev. 30(2):233-249. 24. Powell F.,Inigo R.M. 1992. Microprocessor based D.C. brushless motor controller for wheelchair propulsion. Proc. RESNA Intl. ‘92, Toronto, Canada, pp. 313-315. 25. Riley P.O., Rosen M.J. 1987. Evaluating manual control devices for those with tremor disability. J. Rehab. Res. Dev. 24(2):99-110. 26. Sprigle S.H., Morris B.O., Karg P.E. 1992. Assessment of transportation technology: Survey of driver evaluators. Proc. RESNA Intl. ’92, Toronto, Ontario, Canada, pp. 351-353. 27. Sprigle S.H., Morris B.O., Karg P.E. 1992. Assessment of transportation technology: Survey of Equipment vendors. Proc. RESNA Intl. ’92, Toronto, Ontario, Canada, pp. 354-356. 28. Schauer J., Kelso D.P., Vanderheiden G.C. 1990. Development of a serial auxiliary control interface for powered wheelchairs. Proc. RESNA 13th Annual Conf., Washington, D.C., pp.191-192.


Popović, D. B. “Externally Powered and Controlled Orthotics and Prosthetics.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

142 Externally Powered and Controlled Orthotics and Prosthetics 142.1

FES Systems Restoration of Hand Functions • Restoration of Standing and Walking • Hybrid Assistive Systems (HAS)

Dejan B. Popovic´ University of Belgrade, Belgrade

142.2

Active Prostheses Active Above-Knee Prostheses • Myoelectric Hand and Arm Prostheses

Rehabilitation of humans with disabilities requires effective usage of assistive systems for restoration of motor functions. Important features of an effective assistive system are: (1) reliability; (2) minimum increase of energy rate and cost with respect to able-bodied subjects performing the same task; (3) minimum disruption of normal activities when employing the assistive system; (4) cosmetics; and (5) practicality. The system should be easy to don and doff and available for daily use at home. These requirements and available technology have led to the development of externally powered orthoses and prostheses that interface directly or indirectly with the human neuromuscular system. These devices include battery-powered actuators or muscle stimulators, microprocessor-based controller and reliable biologic-like sensors. Hand, arm, and lower extremities assistive systems for amputees and humans with paralysis are elaborated in this chapter. Two approaches for the restoration of movements in humans with paralysis are described: functional activation of paralyzed muscles called Functional Electrical Stimulation (FES) or Functional Neuromuscular Stimulation (FNS), and the combined usage of an FES and a mechanical orthosis. The latter is called Hybrid Assistive System (HAS). The section dealing with prosthesis relates to externally controlled and powered artificial limbs, with specific emphasis on so-called myoelectric controlled devices. Available assistive systems meet many of the requirements listed above, but still have some drawbacks. The small number of externally powered system users on a daily basis is a significant indicator that these systems are not perfected. Recent neurophysiological findings on muscle properties and strengthening techniques, in addition to improved percutaneous or implantable stimulators, may increase the applicability of many assistive systems, specifically in cases of paralyzed limbs. Major limitations in daily application relate to insufficiently adaptive and robust control methods. The complexity of central nervous system (CNS) control and the interface between voluntary control and external artificial control are still challenging, unanswered questions. Hierarchical control methods, combining symbolic models at the highest level and analytic models at lower levels, give hope for further progress (see Chapter 165—Control of Movements).


Sensory feedback is essential for effective control of FES systems. In addition to artificial sensors, hopes are directed towards the use of natural sensors and the development of an “intelligent” movement controller.

142.1 FES Systems Functional Electrical Stimulation (FES) can help in regaining functional movements in numerous paralyzed humans. FES activates innervated but paralyzed muscles, by using an electronic stimulator to deliver trains of pulses to neuromuscular structures. The basic phenomenon of the stimulation is a contraction of muscle due to the controlled delivery of electric charge to neuromuscular structures. FES systems can restore (1) goal-oriented (hand and arm) movements, and (2) cyclic (walking and standing) movements.

Restoration of Hand Functions The objective of an upper extremity assistive system is directed toward establishing independence to the user. Efforts in developing upper extremity FES systems were targeted toward individuals with diminished, but preserved, shoulder and elbow functions, with lack of wrist control and grasping ability [1]. There have been several designs of FES systems. These systems can be divided among the source of control signals to trigger or regulate the stimulation pattern: shoulder control [2], voice control [3,4], respiratory control [5], joystick control [6], and position transducers [7, 8] or trigger [9, 6]. The division can be made based upon the method to which patterned electrical stimulation is delivered: one to three channel surface electrode systems [6-10], multichannel surface stimulation system [4], multichannel percutaneous systems with intramuscular electrodes [2,3,5], and fully implanted systems with epimysial electrodes [1]. Only a small number of FES grasping systems has been used outside the laboratory. The first grasping system used to provide prehension and release [11] used a splint with a spring for closure and electrical stimulation of the thumb extensor for release. This attempt was unsuccessful mostly because of the state-of-the-art technology used, but also because of muscle fatigue and erratic contractile response. Rudel et al. [12], following the work of Vodovnik [8], suggested the use of a simple two-channel stimulation system and a position transducer (sliding potentiometer). The shift of the potentiometer forward from its neutral position causes opening by stimulating the dorsal side of the forearm; a shift backward causes closing of the hand by stimulating the volar side of the forearm. The follow-up of initial FES system use was systematically continued in Japan. Japanese groups succeeded in developing the FES clinic in Sendai, Japan, where many subjects are implanted with up to 30 percutaneous intramuscular electrodes that are used for therapy, but not to assist grasping. The Japanese research approach in functional grasping relates to subjects lacking not only hand functions but also the elbow control (e.g., C4 complete CNS injuries), and the system uses either a voice or suck/puff control and preprogrammed EMG-based stimulation patterns. This preprogrammed EMG-based stimulation is developed by detailed studies of muscle activities with intramuscular electrodes in able-bodied subjects while reaching and grasping [3]. The approach taken at Ben Gurion University, Israel [4] used a voice controlled multichannel surface electrode system. As many as 12 bipolar stimulation channels and a splint are used to control elbow, wrist, and hand functions. There is very little practical experience with the system, and the system has to be tuned for the needs of every single user. Surface stimulation most probably does not allow control of small hand and forearm muscles necessary to provide dexterity while grasping. Daily mounting and fitting of the system is problematic. The work of Nathan resulted with a commercial device called Handmaster NMS1 [9,13]. This device is approved as a therapeutic device claiming that it is improving grasping functionality in humans after stroke. A detailed analysis of effects of FES systems after stroke is presented by Hines and colleagues [14]. The group at the Institut for Biokibernetik, Karlsruhe, Germany suggested the use of EMG recordings from the muscle, which is stimulated [15]. The aim of this device is to enhance the grasping by using


weak muscles. Hence, in principle it could be possible to use retained recordings from the volar side of the forearm to trigger on and off the stimulation of the same muscles. In this case, it is essential to eliminate the stimulation artifact and the evoked potential caused by the stimulus in order to eliminate positive feedback effects which will generate a tetanic contraction that cannot be turned off using the method presented. This approach is further developed in Denmark [16]. The Case Western Reserve University (CWRU) fully implantable system has a switch to turn the system on and off and select the grasp and a joystick to proportionally control aperture of the hand for palmar and lateral grasp [17-19]. These two grasps are synthesized using a preprogrammed synergy. The joystick mounted on the contralateral shoulder voluntarily controls the preprogrammed sequence of stimulation. The palmar grasp starts from the extended fingers and thumb (one end position of the joystick), followed by the movement of the thumb to opposition and flexing of fingers (other end position of the joystick). The lateral grasp starts from the full extension of fingers and the thumb, followed by the flexion of fingers and adduction of the thumb. The system is applicable if the following muscles can be stimulated: extensor pollicis longus, flexor pollicis longus, adductor pollicis, opponens, flexor digitorum profundus and superficialis, and extensor digitorum communis. It is possible to surgically change the grasp permanently [1] (pining some joints, tendon transfer, etc.). An important feature of the grasping system is related to daily fitting of the joystick (zeroing the neutral position), and going to hold mode from the movement mode. The hold mode is the regime where the muscle nerves are stimulated at the level that the same force is maintained, and the user selects the level. At this time, the CWRU system uses the joystick with two degrees of freedom or velocity sensor software to switch from active control to hold mode. When the system is in the hold mode, joystick movements do not effect the grasp. The initial CWRU system [20] suggested the use of myoelectric signals obtained from a site with some regaining voluntary activity. The CWRU implantable system is the only system used for assistance in daily living function [1]. The functional evaluation of the system [21] showed that there is substantial improvement in simple grasping tasks, which is important to a person with tetraplegia, and an increasing number are in use around North America and Australia. Prochazka [7] suggested the use of wrist position to control the stimulation of muscles to enhance the tenodesis grasping, and designed a device called the Bionic Glove (Fig. 142.1). A sensor is used to detect wrist movement, and trigger opening and closing of the hand. A microcomputer is built into the battery operated stimulation unit, which detects movements and controls three channels to stimulate thumb extension and flexion, and finger flexors. Clinical evaluation of the Bionic Glove [22] showed that the stimulation is beneficial for persons with tetraplegia both therapeutically and as an orthosis, but that the overall acceptance rate is still low. Based on four-channel surface electrical stimulation for walking triggered from the lesion EMG recordings [23,24], a grasping system with up to three channels of surface stimulation provides similar function to the Prochazka’s system [25]. The use of myoelectric signals from above lesion sites (forearm wrist extensors) to trigger the stimulation of thumb and finger flexors was presented. A threshold discrimination of the EMG is done, which activates the proper stimulus pattern to be applied to the surface electrodes. A new laboratory version of the myoelectric control of a grasping system has been developed and tried at the ETH Zurich [26]. An interesting approach of using myoelectric control to drive the CWRU system was evaluated by Scott and colleagues [27]. Using muscles that are not affected and easily controllable, a bilateral control of the CWRU system might be possible. The idea of controlling the whole arm and assisting manipulation in humans lacking shoulder and elbow control has been getting more attention recently from the research team at the Case Western Reserve University in Cleveland, Ohio [28-31]. The system is designed to combine a fully implantable grasping system with some additional channels to control elbow extension, flexion, and shoulder movements. A different approach is taken by using surface stimulation to control elbow movements [32,33]. The clinical study at the Rehabilitation Institute in Belgrade with the so-called Belgrade grasping/reaching system showed that most persons with tetraplegia responding to stimulation of the biceps and triceps brachii muscles improve their function within six weeks to the stage at which they become reluctant to use FES assistance for manipulation [32,33].


FIGURE 142.1 Bionic glove for restoration of grasping in persons with tetraplegia used for multicentre trials [7,22]. The upper panel shows electrodes, stimulator, and the glove, while the bottom panel shows the glove mounted and ready for use.

The development of implantable cuff electrodes to be used for sensing contact, slippage, and pressure [34-38] opens a new prospective in controlling grasping devices, and the Center for Sensory Motor Interactions in Aalborg, Denmark is pursuing a series of experiments combining their sensing technique with the fully implantable CWRU system.

Restoration of Standing and Walking The application of FES to the restoration of gait was first investigated systematically in Ljubljana, Slovenia [39-44]. Currently, FES for gait rehabilitation is used in a clinical setting in several rehabilitation centers [45-57] and there is a growing trend for the design of devices for home use. Current surface FES systems use various numbers of stimulation channels. The simplest one, from a technical point, is a single-channel stimulation system. This system is only suitable for stroke patients and a limited group of incomplete spinal cord injury patients. These individuals can perform limited ambulation with assistance of the upper extremities without an FES system, although this ambulation may be both modified and/or impaired. The FES in these humans is used to activate a single muscle group. The first demonstrated application of this technique was in stroke patients [40], even though the


original patent came from the Liberson patent in 1961. The stimulation is applied to ankle dorsiflexors so the “foot-drop” can be eliminated. A commercial system has been designed by Stein and colleagues [58], which integrates a single-channel stimulator and a tilt sensor; thus, eliminating a foot switch which was proved for easily generating false triggering and malfunctioning. Single and dual channel correcting foot-drop is now a regular clinical treatment in some rehabilitation institutions [59]. A multichannel system with a minimum of four channels of FES is required for ambulation of a patient with a complete motor lesion of lower extremities and preserved balance and upper body motor control [49]. Appropriate bilateral stimulation of the quadriceps muscles locks the knees during standing. Stimulating the common peroneal nerve on the ipsilateral side, while switching off the quadriceps stimulation on that side, produces a flexion of the leg. This flexion, combined with adequate movement of the upper body and use of the upper extremities for support, allows ground clearance and is considered the swing phase of the gait cycle. Hand or foot switches can provide the flexion–extension alternation needed for a slow forward or backward progression. Sufficient arm strength must be available to provide balance in parallel bars (clinical application), and with a rolling walker or crutches (daily use of FES). These systems evolved into a commercial product called Parastep-1R (Sigmedics, Chicago, IL) which was approved for home usage in 1994 by the Food and Drug Administration. Multichannel percutaneous systems for gait restoration were suggested [46,50,61,62]. The main advantage of these systems is the plausibility to activate many different muscle groups. A very similar preprogrammed stimulation pattern to the one in a human with no motor disorders is delivered to ankle, knee, and hip joints, as well as to paraspinal muscles. The experience of the Cleveland research team suggested that 48 channels are required for a complete SCI walking system to achieve a reasonable walking pattern, but they recently changed their stimulation strategy and included some external bracing. Fine-wire intramuscular electrodes are cathodes positioned close to the motor point within selected muscles. Knee extensors (rectus femoris, vastus medialis, vastus lateralis, vastus intermedius), hip flexors (sartorius, tensor fasciae latae, gracilis, iliopsoas), hip extensors (semimembranosus, gluteus maximus), hip abductors (gluteus medius), ankle dorsiflexors (tibialis anterior, peroneus longus), ankle plantar flexors (gastrocnemius lateralis and medialis, plantaris and soleus), and paraspinal muscles are selected for activation. A surface electrode is used as a common anode. Interleaved pulses are delivered with a multichannel, battery-operated, portable stimulator. The hand controller allows the selection of gait activity. These systems were limited to the clinical environment. The application was investigated in complete spinal cord lesions and in stroke patients [61,62]. The same strategy and selection criteria for implantation was used for both stroke and SCI patients. Recent developments use the CWRU system with eight channels per leg to be activated and improved control [61-63]. A multichannel totally implanted FES system [55] was proposed and tested in a few subjects. This system uses a 16-channel implantable stimulator attached to the epineurium electrodes. Femoral and gluteal nerves were stimulated for hip and knee extension. The so-called “round-about” stimulation was applied in which four electrodes were located around the nerve and stimulated intermittently. This stimulation method reduces muscle fatigue. The development of the stimulation technology is providing new hope for patients and researchers alike. Two new techniques are especially important: (1) application of remotely controlled wireless microstimulators [64-66], and (2) so-called “stimulator for all seasons” [67]. There are several attempts to design effective wireless stimulators that are believed to be capable of selectively stimulating fascicles [68,69]. The technology used for cochlear implants is finding its way into applications for standing and walking restoration [70]. However, there are some unanswered questions which limit the effectiveness of FES systems that deal with muscle fatigue, reduced joint torques generated through FES in comparison to CNS activated torques in healthy subjects, modified reflex activities, spasticity, joint contractures, osteoporosis, and stress fractures. From the engineering point of view, the further development of FES systems has to address the following issues: the interface between an FES system and neuromuscular structures in the organism, biocompatibility of the FES system, and overall practicality. The least resolved questions in FES systems


deal with control. Open-loop control, often called reference trajectory control, is not efficient because of the disturbances and model errors. Closed-loop control requires that we know in advance the desired trajectory and use multiple sensors within the system in which the actuators have a delay in the same time range as duration of some locomotion states. In principle, closed-loop systems offer substantial increases in input–output linearity and repeatability, along with a substantial decrease in the sensitivity of the system to parameter variations (internal disturbances) and load changes (external disturbances). Digital closed-loop methods using proportional (P) and proportional plus integral (PI) controllers were studied for both recruitment and fatigue of muscles [71]. A non-conventional method based on symbolic representation of motion in animals and humans, called artificial reflex control, was proposed [72-76]. Different solutions are discussed in great detail in Chapter 165 of this book. Many recent studies for improving control [77-82] aim to prolong standing with FES systems without arm supports. This is a very difficult task, considering that the human body acts as an inverted chain, and that FES systems generate non-physiological, fast fatiguing muscle contractions. The research results indicate that the usual classification of CNS lesions is applicable as a guideline, but is not convenient as a selection criterion for FES candidates. FES users and an appropriate FES system must be selected through a functional diagnosis. The term “functional diagnosis” means that the functional status after the injury determines the type of treatment. General statements about FES systems indicate that it is suitable for subjects with preserved peripheral neuromuscular structures, moderate spasticity, without joint contractures and limited osteoporosis. Subjects should be able to control their balance and upper body posture using the upper extremities for assistance (parallel bars, walker, crutches, etc.). Subjects with pathologies that affect the heart, lungs, or circulation should be treated with extreme care, and often they should not be included in an FES walking program. A satisfactory mental and emotional condition is extremely important because an FES treatment requires a certain degree of intelligence and understanding of the technical side of the system. High motivation and good cooperation with medical staff is a significant aspect for the efficacy of FES. Surface stimulation may be applied only in subjects with limited sensation because it activates pain receptors. Subjects suitable for FES can include those with head and spinal cord injuries, cerebral paralysis, multiple sclerosis, and different types of myelitis and others. The final current drawback to FES ambulation is the excessive energy cost while walking. FES walking should be described as a sequence of static states; long stance phase (several seconds) followed by brisk flexion movement and short step, followed by the same movement at the contralateral side. Dynamic walking is a necessary development which will introduce better use of inertial and gravity forces, and reduce energy rate and cost, increase the walking distance, and increase the speed of progression in parallel with decreased fatigue of muscles because the muscles will not be stimulated for long periods. Table 142.1 summarizes differences in walking performances between able-bodied persons and persons with paralysis (energy cost, energy rate, amount of the use of upper extremities, cardiovascular stress measured through pulse rate, and blood pressure).

Hybrid Assistive Systems (HAS) A specific approach to integrating two assistive systems (FES and an external mechanical orthosis) has been proposed (Fig. 142.2). These systems are called hybrid assistive systems (HAS) or hybrid orthotic systems (HOS) [83,84]. A few possibilities for HAS design have been suggested, which combine relatively simple rigid mechanical structures for passive stabilization of lower limbs during stance phase and FES systems. These systems combine the use of a reciprocating gait orthosis with multichannel stimulation [53,84], the use of an ankle–foot orthosis or an extended ankle–foot orthosis with a knee cage [85], or the use of a self-fitting modular orthosis [86]. A few more sophisticated laboratory systems were demonstrated [85,87-89]. Each trend in the design of HAS implies different applications as well as specific hardware and control problems. On the basis of accumulated experience, the following features can serve as criteria for a closer description of various HAS designs [88]: (1) partial mechanical support, (2) parallel operation of the biological and mechanical system, and (3) sequential operation of the biological and


TABLE 142.1 Best Performance with the Usage of an Assistive System Compared to Performance of Able-Bodied Subjects Oxygen Rate Oxygen Cost ml/kg/min ml/kg/m Paralyzed 13.6 0.6 subjects Able-bodied 12.1 0.15 subjects Control method Voluntary, Voluntary, hand-switch sensory driven Assistive system FES+SFMO FES+SFMO used FES+RGO

Speed vMAX m/min 60

Heart Rate b/min

118 (84 rest) 85 95 (81 rest) Automatic, Voluntary, handpreprogrammed switch or sensory driven FES implant FES+RGO FES+SFMO

Blood Pressure mmHg

Use of Upper Extremities [%]

119/85

28

120/80

0

Voluntary, hand- Voluntary, handswitch or sensory switch or sensory driven driven FES+RGO FES+RGO

The data is collected from literature [53, 83, 86, 88; 115-119]. Abbreviations used in the table are: RGO—reciprocating gait orthosis; SFMO—self-fitting modular orthosis; FES— functional electrical stimulation. Data for able-bodied subjects are from Waters et al. [1989].

FIGURE 142.2 An incomplete tetraplegic subject walking with a hybrid assistive system. The particular system incorporates a six-channel surface stimulation system and a powered self-fitting modular orthosis [88].

the mechanical system. The partial mechanical support refers to the use of braces to assist FES only at specific events within a walking cycle [85]. The advanced version of powered orthoses to be used with FES is being developed by Durfee and colleagues [89]. Control of joints in mechanical orthosis is again becoming a target for research and development mainly because of new technological tools [90,91].


142.2 Active Prostheses The role of active prosthesis is to extend the function provided by a “non-externally” powered and controlled artificial organ; hence, to improve the overall performance of motor function, ultimately providing a better quality of life.

Active Above-Knee Prostheses Effective restoration of walking and standing of handicapped humans is an important element to improve the quality of life. Artificial legs of different kinds have been in use for a long time, but in many cases they are inadequate for the needs of amputees, specifically for high above-knee amputations (e.g., hip disarticulation), bilateral amputees, and patients who have demanding biomechanical requirements (e.g., subjects involved in sports). Modern technology has led to improved designs of below-knee prostheses (BKP) in recent years. Below-knee amputees perform many normal locomotor activities and participate in many sports requiring running, jumping, and other jerky movements [92]. The biggest progress was made using readily available and easy-to-work-with plastic and graphite alloys for building the artificial skeletal portion of the shank and foot [93]. Below-knee prostheses are light, easy to assemble, and reliable. They provide good support and excellent energy absorption, which prevents impact, jerks, and allows the push-off phase in the gait cycle. Existing below-knee prostheses, even those without ankle joints, duplicate the behavior of the normal foot–ankle complex during swing and stance phases of the step cycle. The same technology has been introduced into the design of above-knee prostheses (AKP). However, commercially available AKPs suffer from several drawbacks. The requirements for an AKP were stated by Wagner and Catranis [94]. The prosthesis must support the body weight of the amputee like the normal limb during the stance phase of level walking, on slopes and on soft or rough terrain. This implies that the prosthesis provides “stability” during weight bearing activity, i.e., it prevents sudden or uncontrolled flexion of the knee. The second requirement is that the body is supported such that undesirable socket/stump interface pressures and gait abnormalities due to painful socket/stump contact are prevented. The analysis of biomechanical factors that influence the shaping, fitting, and alignment of the socket is a problem in itself. If the fitting has been accomplished, allowing the amputee to manipulate and control the prosthesis in an active and comfortable manner, the socket and stump can be treated as one single body. The third requirement, which is somewhat controversial, is that the prosthesis should duplicate as nearly as possible the kinematics and dynamics of normal gait. The amputee should walk with a normallooking gait over a useful range of speeds associated with typical activities for normal persons of similar age. The latter requirement has received attention in recent years and fully integrated systems, so called “self-contained” active AKPs are being incorporated into modern rehabilitation. The self-contained principle implies that the artificial leg contains the energy source, actuator, controller, and sensors (Fig. 142.3). Devices such as the polycentric knee mechanism [95], the polycentric knee mechanism with a hydraulic valve, or the AKP with friction type brake [96] satisfy some of the above performance requirements. A logically controlled AKP with a hydraulic valve represents a further bridge between purely passive and fully controllable assistive devices [97,98]. Studies of the knee joint performance in the stance phase have been done [99-101]. From those studies it became clear that for several gait modes (ramp, stairs), active knee control in the swing phase is desirable. To meet these requirements, self-contained, multi-mode AKPs have been introduced [102-105]. The final prototype version of an active self-contained AKP allowing controlled flexion and extension throughout the gait cycle was developed by James [97] and is a product of Otto Bock Company, Germany. This leg uses an efficient knee controller with a custombuilt hydraulic valve and a single chip microcontroller with a rule-based control scheme (see Chapter 165: Control of Movements). The advantages of the externally powered and controlled leg (see, e.g., [104]) are an increased speed of locomotion, better gait symmetry, and lower energy cost and rate. These accomplishments are due to


the fact that the powered leg allows controlled knee bounce after heel contact, limited push-off at the end of the stance phase, and effective flexion and controlled knee extension during the swing phase of gait. The external control allows the amputee to walk almost without any circumduction, which helps gait symmetry considerably. The knee joint of a standard endoskeletal prosthesis is fitted with an actuator having two independent braking units (friction drum type brake with an azbestum lining to allow control of the knee joint stiffness), and the extension-flexion driver with a ball-screw mechanism. The rechargeable battery power supply is designed for up to three hours of continuous level walking. The Motorola 68HC11-based microcontroller has been fitted into the interior of the prosthesis fulfilling self-containing principles. A hierarchical controller allows for intention recognition (volitional actions of the subject), adaptation to environmental changes, cyclical triggering throughout locomotion, and minimal jerk actuator control.

Myoelectric Hand and Arm Prostheses The popularity of this control strategy comes from FIGURE 142.3 A self-contained, battery powered, the fact that it is “very elegant” because the control microcomputer controlled above-knee prosthesis using signals going to the prosthesis may come from the artificial reflex control [104]. myoelectric activity of the muscle which formerly controlled the lost limb [23,24,106,107]. Therefore, in some sense, the brain’s own signals are used to control the motion of the prosthesis [109]. Although the validity of myoelectric control-based techniques seems reasonable, in practice the performances are somewhat disappointing. There are a number of reasons for this. In the cases of high level amputation, the electrical activity of the remaining muscles have only minor correlation with the normal arm movements and this makes coordinated control of several functions very difficult. A more important problem is that any EMG controlled prosthesis is inherently open-loop (preprogrammed) [110]. The absence of position proprioception and other related problems of reliability in executing movements [111] has contributed to the poor success rate of myoelectrically controlled prostheses. Presently two systems based on myoelectric control are used with success (Fig. 142.4). The Otto Bock myoelectric hand [120] uses the technique of detection of the fixed threshold of the integrated rectified surface electromyography. Extension/flexion of the targeted muscle group serves as the on/off control, which causes opening and closing of the hand, respectively. Two sets of electrodes are used to detect this differential signal. The Utah arm [107,108] utilizes the myoelectric signals from two antagonists, as the proportional elbow control (Fig. 142.5). These signals are full wave rectified and differentiated to give the command signal. The Utah arm has a wrist rotator and it can be combined with different grasping systems. Most of the artificial hands used in everyday activities do not include the five finger dexterity, except in the devices that are mechanically preprogrammed. A new two-degree of freedom hand prosthesis with hiearachical grip was designed by Kyberd [112]. A new appealing version of a controller was designed by Kurtz [113].


FIGURE 142.4

The Waseda University prototype of a myoelectric artificial hand.

FIGURE 142.5

The Utah arm [107].

An interesting approach to overcoming the problems encountered with EMG control has been proposed by Simpson [110]. The control strategy is termed “extended physiological proprioception (EPP)”, and uses the positions of intact joints as the controlling input signals to a prosthesis. This strategy is based on recognition of the fact that intact joints possess inherent position feedback, and that most goal oriented tasks are highly synergistic [32,33]. By establishing a one-to-one relationship between the position of the intact joint and the position in the space of the terminal device (prosthesis), the natural


feedback of the intact joint can be “extended” to the prosthesis, and thereby provide it with proprioceptive information. Based on this idea, an externally controlled elbow–wrist–hand prosthesis was evaluated [110]. The shoulder joint position is measured with a potentiometer, mounted between the upper arm and trunk. An important feature of the system is that disabled humans can learn how to control the system within minutes using trial-and-error procedure. Bertos [114] presented a very elegant solution for position control based on EPP.

Defining Terms Functional electrical stimulation or Functional neuromuscular stimulation: Patterned electrical stimulation of neuromuscular structures dedicated to restore motor functions. Externally controlled assistive system: Assistive system for restoration of motor functions with automatic control. Externally powered assistive system: Assistive system for restoration of motor functions which uses external power to either stimulate muscles or drive actuators. Hybrid assistive systems: Combination of a functional electrical stimulation and a mechanical orthosis. Myoelectric (EMG) control: Use of voluntary generated myoelectric activity as control signals for an externally controlled and powered assistive system. Artificial reflex control: A sensory-driven control algorithm based on knowledge representation (production rule based system). Reciprocating gait orthosis: A walking and standing assistive system with a reciprocating mechanism for hip joints which extends the contralateral hip when the ipsilateral hip is flexed. Self-fitting modular orthosis: A modular, self-fitting, mechanical orthosis with a soft interface between the human body and the orthosis.

References 1. Peckham, P. H., and Keith, M. W., Motor prostheses for restoration of upper extremity function. In Neural Prostheses: Replacing Motor Function After Disease or Disability, In Stein, R. B., Peckham, P. H., and Popovic, ´ D., Oxford University Press, New York, 1992, 162-190. 2. Buckett, J. R., Peckham, H. P., et al., A flexible, portable system for neuro-muscular stimulation in the paralyzed upper extremities, IEEE Trans. Biomed. Eng., BME-35:897-904, 1988. 3. Handa, Y., Handa, T., et al., Functional electrical stimulation (FES) systems for restoration of motor function of paralyzed muscles-versatile systems and a portable system, Front. Med. Biol. Eng., 4:241255, 1992. 4. Nathan, R.H. and Ohry, A., Upper limb functions regained in quadriplegia: A hybrid computerized neuromuscular stimulation system, Arch. Phys. Med. Rehabil., 71:415-421, 1990. 5. Hoshimiya, N., Naito, N., Yajima, M., and Handa, Y., A multichannel FES system for the restoration of motor functions in high spinal cord injury patients: A respiration-controlled system for multijoint upper extremity, IEEE Trans. Biomed. Eng., BME-36:754-760 1989. 6. Popovic, ´ D., Popovic, ´ M., et al., Clinical evaluation of the Belgrade grasping system, Proc. V Vienna Intern. Workshop on Functional Electrical Stimulation, Vienna, 1998. 7. Prochazka, A., Gauthier, M., Wieler, M., and Kenwell, Z., The bionic glove: an electrical stimulator garment that provides controlled grasp and hand opening in quadriplegia, Arch Phys. Med. Rehabil., 78:608-614, 1997. 8. Reber˘sek, S. and Vodovnik, L., Proportionally controlled functional electrical stimulation of hand, Arch. Phys. Med. Rehabil., 54:378-382, 1973. 9. Nathan, R., Handmaster NMS—present technology and the next generation, In Popovic, ´ D., (Ed.) Proc. II Intern. Symp. on FES, Burnaby, pp. 139-140, 1997. 10. Vodovnik, L., Bajd, T., et al., Functional electrical stimulation for control of locomotor systems, CRC Crit. Rev. Bioeng., 6:63-131, 1981.


11. Long, C. II and Masciarelli, C. V., An electrophysiologic splint for the hand, Arch. Phys. Med. Rehabil., 44:499-503, 1963. 12. Rudel, D., Bajd, T., Reber˘sek, S., and Vodovnik, L., FES assisted manipulation in quadriplegic patients, In Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities VIII, pp. 273-282, ETAN, Belgrade, 1984. 13. Nathan, R.H., An FNS-based system for generating upper limb function in the C4 quadriplegic, Med. Biol. Eng. Comp., 27:549-556, 1989. 14. Hines, A. E., Crago, P. E., and Billian, C., Hand opening by electrical stimulation in patients with spastic hemiplegia, IEEE Trans. Rehab. Eng., TRE-3:193-205, 1995. 15. Holländer, H. J., Huber, M., and Vossius, G., An EMG controlled multichannel stimulator, In Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities IX, pp. 291-295, ETAN, Belgrade, 1987. 16. Sennels, S., Biering-Soerensen, F., Anderson, O. T., and Hansen, S. D., Functional neuromuscular stimulation control by surface electromyographic signals produced by volitional activation of the same muscle: adaptive removal of the muscle response from the recorded EMG-signal, IEEE Trans. Rehab. Eng., TRE-5:195:206, 1997. 17. Lan, N., Feng, H. Q., and Crago, P. E., Neural network generation of muscle stimulation patterns for control of arm movements, IEEE Trans. Rehab. Eng., TRE-4:213-224, 1994. 18. Hart, R. L., Kilgore, K. L., and Peckham, P. H., A comparison between control methods for implanted FES hand-grasp systems, IEEE Trans. Rehab. Eng., TRE-6:208-218, 1998. 19. Smith, B. T., Tang, Z., Johnson, M. W., et al., An externally powered multichannel, implantable stimulator-telemeter for control of paralyzed muscles, IEEE Trans. Biomed. Eng., BME-45:463-475, 1998. 20. Peckham, P. H., Mortimer, J. T., and Marsolais, E. B., Controlled prehension and release in the C5 quadriplegic elicited by functional electrical stimulation of the paralyzed forearm muscles, Ann. Biomed. Eng., 8:369-388, 1980. 21. Wijman, A. C., Stroh, K. C., Van Doren, C. L., Thrope, C. L., Peckham C. H., and Keith, M. W., Functional Evaluation of Quadriplegic Patients Using a Hand Neuroprosthesis, Arch. Phys. Med. Rehabil., 71:1053-1057, 1990. 22. Popovic, ´ D., Stojanovic,´ A., et al., Clinical evaluation of the bionic glove, Arch. Phys. Med. Rehabil., 80, 1999. 23. Graupe, D. and Kline, W.K., Functional separation of EMG signals via ARMA identification methods for prosthesis control purposes, IEEE Trans. System, Man Cybern., SMC-5:252-259, 1975. 24. Graupe, D., EMG pattern analysis for patient-responsive control of FES in paraplegics for walkersupported walking, IEEE Trans. Biomed. Eng., BME-36:711-719, 1989. 25. Saxena, S., Nikolic, S., and Popovic, ´ D., An EMG controlled FES system for grasping in tetraplegics, J. Rehabil. Res. Develop., 32:17-23, 1995. 26. Keller, T., Curt, A., Dietz, V., and Morari, M., EMG controlled grasping neuroprosthesis for high lesioned tetraplegic patients, In Popovic, ´ D., (Ed.) Proc. II Intern. Symp. on FES, Burnaby, pp. 129-130, 1997. 27. Scott, T. R. D., Peckham, P. H., and Kilgore, K. L., Tri-state myoelectric control of bilateral upper extremity neuroprosthesis for tetraplegic individuals, IEEE Trans. Rehab. Eng., TRE-4:251-263, 1996. 28. Grill, J. H. and Peckham, P. H., Functional neuromuscular stimulation for combined control of elbow extension and hand grasp in C5 and C6 quadriplegics, IEEE Trans. Rehab. Eng., TRE6:190-199, 1998. 29. Crago, P. E., Memberg, W. D., Usey, M. K. et al., An elbow extension neuroprosthesis for individuals with tetraplegia, IEEE Trans. Rehab. Eng., TRE-6:1-6, 1998. 30. Lemay, M. A. and Crago, P. E., Closed loop stabilization in C4 and C4 tetraplegia, IEEE Trans. Rehab. Eng., TRE-5:244-252, 1997. 31. Smith, B. T., Mulcahey, M. J., and Betz, R. R., Development of an upper extremity FES system for individuals with C4 tetraplegia, IEEE Trans. Rehab. Eng., TRE-4:264-270, 1996.


32. Popovic, ´ M. and Popovic, ´ D., A new approach to reaching control for tetraplegic subjects, J. Electromyog. Kinesiol., 4:242-253, 1994. 33. Popovic, ´ D. and Popovic, ´ M., Tuning of a nonanalytic hierarchical control system for reaching with FES, IEEE Trans. Biomed. Eng., BME-45:203-212, 1998. 34. Haugland, M. and Sinkjaer, T., Cutaneous whole nerve recordings used for correction of foot-drop in hemiplegic man, IEEE Trans. Rehab. Eng., TRE-3:307-317, 1995. 35. Haugland, M. K. and Hoffer, J. A., Slip information provided by nerve cuff signals: application in closed-loop control of functional electrical stimulation, IEEE Trans. Rehab. Eng., TRE-2:29-37, 1994. 36. Haugland, M. K., Hoffer, J. A., and Sinkjaer, T., Skin contact force information in sensory nerve signals recorded by implanted cuff electrodes, IEEE Trans. Rehab. Eng., TRE-2:18-27, 1994. 37. Sahin, M. and Durand, D. M., Improved nerve cuff electrode recordings with subthreshold anode currents, IEEE Trans. Biomed. Eng., BME-45:1044-1050, 1998. 38. Upshaw, B. and Sinkjaer, T., Digital signal processing algorithms for the detection of afferent nerve recorded from cuff electrodes, IEEE Trans. Rehab. Eng., TRE-6:172-181, 1998. 39. Bajd, T., Kralj, A., and Turk, R., Standing-up of a healthy subject and a paraplegic patients, J. Biomech., 15:1-10, 1982. 40. Gra˘canin, F., Prevec, T., and Trontelj, J., Evaluation of use of functional electronic peroneal brace in hemiparetic patients, in Advances in External Control of Human Extremities III, ETAN, Belgrade, 1967, 198-210. 41. Kralj, A., Bajd, T., and Turk, R., Electrical stimulation providing functional use of paraplegic patient muscles, Med. Prog. Technol., 7:3-9, 1980. 42. Kralj, A., Bajd, T., Turk, R., and Benko, H., Results of FES application to 71 SCI patients, in Proc. RESNA 10th Ann. Conf. Rehabil. Tech., San Jose, 1987, 645- 647. 43. Vodovnik, L. and McLeod, W. D., Electronic detours of broken nerve paths, Med. Electr., 20:110-116, 1965. 44. Vodovnik, L., Crochetiere, W. J., and Reswick, J. B., Control of a skeletal joint by electrical stimulation of antagonists, Med. Biol. Eng., 5:97-109, 1967. 45. Andrews, B. J., Baxendale, R. H., et al., Hybrid FES orthosis incorporating closed loop control and sensory feedback, J. Biomed. Eng., 10:189-195, 1988. 46. Brindley, G. S., Polkey, C. E., and Rushton, D. N., Electrical splinting of the knee in paraplegia, Paraplegia, 16:428-435, 1978. 47. Hermens, H. J., Mulder, A. K., et al., Research on electrical stimulation with surface electrodes, in Proc. 2nd Vienna Intern. Workshop on Functional Electrostimulation, Vienna, 1986, 321-324. 48. Jaeger, R., Yarkony, G. Y., and Smith, R., Standing the spinal cord injured patient by electrical stimulation: Refinement of a protocol for clinical use, IEEE Trans. Biomed. Eng., BME-36:720-728, 1989. 49. Kralj, A. and Bajd, T., Functional Electrical Stimulation, Standing and Walking After Spinal Cord Injury CRC Press, Boca Raton, FL, 1989. 50. Marsolais, E. B. and Kobetic, R., Functional walking in paralyzed patients by means of electrical stimulation, Clin. Orthop., 175:30-36, 1983. 51. Mizrahi, J., Braun, Z., Najenson, T., and Graupe, D., Quantitative weight bearing and gait evaluation of paraplegics using functional electrical stimulation, Med. Biol. Eng. Comput., 23:101-107, 1985. 52. Petrofsky, J. S. and Phillips, C. A., Computer controlled walking in the paralyzed individual, J. Neurol. Orthop. Surg., 4:153-164, 1983. 53. Solomonow, M., Biomechanics and physiology of a practical powered walking orthosis for paraplegics, in Neural Prostheses: Replacing Motor Function After Disease or Disability, (Eds.) Stein, R. B., Peckham, H. P. and Popovic, ´ D., (Eds.), Oxford University Press, New York, 1992, 202-230. 54. Stein, R. B., Prochazka, A., et al., Technology transfer and development for walking using functional electrical stimulation, in Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities X, Nauka, Belgrade, 1990, 161-176.


55. Thoma, H., Frey, M., et al., Functional neurostimulation to substitute locomotion in paraplegia patients, in Andrade, D. et al. (Eds.), Artificial Organs, VCH Publishers, 1987, 515-529. 56. Vossius, G., Mueschen, U., and Hollander, H. J., Multichannel stimulation of the lower extremities ´ D. (Ed.), with surface electrodes, in Advances in External Control of Human Extremities IX, Popovic, ETAN, Belgrade, 1987, 193-203. 57. Waters, R. L., McNeal, D. R., Fallon W., and Clifford, B., Functional electrical stimulation of the peroneal nerve for hemiplegia, J. Bone Joint Surg, 67:792-793, 1985. 58. Wieler, M., Stein, R.B., et al., A Canadian multicentre trial using functional electrical stimulation for assisting gait, in Popovic, ´ D., (Ed.), Proc. II Intern. Symp. on FES, Burnaby, 1997, 160-161. 59. Acimovi ´ c,´ R., Stani˘c, U., et al., Routine clinical use of functional electrical stimulation in hemiplegia, in Popovic, ´ D. (Ed.), Proc. II Intern. Symp. on FES, Burnaby, 1997, 182-183. 60. Graupe, D. and Parastep I, Sigmedics, IL—the therapy for paraplegics or an orthosis for walking, personal communication, 1997. 61. Marsolais, E. B. and Kobetic, R., Implantation techniques and experience with percutaneous intramuscular electrode in the lower extremities, J. Rehabil. Res, 23:1-8, 1987. 62. Kobetic, R. and Marsolais, E. B., Synthesis of paraplegic gait with multichannel functional electrical stimulation, IEEE Trans. Rehab. Eng., TRE-2:66-79, 1994. 63. Abbas, J. J. and Triolo, R. J., Experimental evaluation of an adaptive feedforward controller for use in functional neuromuscular stimulation systems, IEEE Trans. Rehab. Eng., TRE-5:12-22, 1997. 64. Cameron, T., Loeb, G. E., Peck, R., et al., Micromodular implants to provide electrical stimulation of paralyzed muscles and limbs, IEEE Trans. Biomed. Eng., BME-44:781-790, 1997. 65. Cameron, T., Liinama, T., Loeb, G. E., and Richmond, F. J. R., Long term biocompatibility of a miniature stimulator implanted in feline hind limb muscles, IEEE Trans. Biomed. Eng., BME45:1024-1035, 1998. 66. Cameron, T., Richmond, F. J. R., and Loeb, G. E., Effects of regional stimulation using a miniature stimulator implanted in feline posterior bicpes femoris, IEEE Trans. Biomed. Eng., BME-45:10361045, 1998. 67. Strojnik, P., Whitmoyer, D., and Schulman, J., An implantable stimulator for all season, in Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities X, Nauka, Belgrade, 1990, 335-344. 68. Ziaie, B., Nardin, M. D., Coghlan, A. R., and Najafi. K., A single channel implantable microstimulator for functional neuromuscular stimulation, IEEE Trans. Biomed. Eng., BME-44:909-920, 1997. 69. Haugland, M. K., A miniature implantable nerve stimulator, in Popovic, ´ D. (Ed.), Proc. II Intern. Symp. on FES, Burnaby, 1997, 221-222. 70. Houdayer, T., Davis, R., et al., Prolonged closed-loop standing in paraplegia with implanted cochlear FES-22 stimulator and Andrews ankle-foot orthosis, in Popovic, ´ D. (Ed.), Proc. II Intern. Symp. on FES, Burnaby, 1997, 168-169. 71. Chizeck, H. J., Adaptive and nonlinear control methods for neuroprostheses, in Stein, R. B., Peckham, H. P., and Popovic, ´ D., (Eds.) Neural Prostheses: Replacing Motor Function After Disease or Disability, Oxford University Press, New York, 1992, 298-328. 72. Andrews, B. J., Baxendale, R. M., et al., A hybrid orthosis for paraplegics incorporating feedback control, in Advances in External Control of Human Extremities IX, ETAN, Belgrade, 1987, 297-310. 73. Tomovic,´ R., Control of assistive systems by external reflex arcs, in Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities IX, ETAN, Belgrade, 1984, 7-21. 74. Tomovic,´ R., Popovic, ´ D., and Tepavac, D., Adaptive reflex control of assistive systems, in Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities IX, ETAN, Belgrade, 1987, 207-214. 75. Veltink, P. H., Koopman, A. F. M., and Mulder, A. J., Control of cyclical lower leg movements ´ D. (Ed.), generated by FES, in Advances in External Control of Human Extremities X, Popovic, Nauka, Belgrade, 1990, 81-90. 76. Tomovic,´ R., Popovic, ´ D., and Stein, R. B., Nonanalytical Methods for Motor Control, World Scientific Publishing, Singapore, 1995.


77. Riener, R. and Fuhr, T., Patient-driven control of FES supported standing up: A simulation study, IEEE Trans. Rehab. Eng., TRE-6:113-124, 1998. 78. Veltink, P. H. and Donaldson, N., A perspective on the controlled FES-supported standing, IEEE Trans. Rehab. Eng., TRE-6:109-112, 1998. 79. Matja˘cic,´ Z. and Bajd, T., Arm-free paraplegic standing — part I: Control model synthesis and simulation, part II: Experimental results, IEEE Trans. Rehab. Eng., TRE-6: 125-150, 1998. 80. Davoodi, R. and Andrews, B. J., Computer simulation of FES standing up in paraplegia: A selfadaptive fuzzy controller with reinforcement learning, IEEE Trans. Rehab. Eng., TRE-6:151-161, 1998. 81. Donaldson, N. and Yu, C., A strategy used by paraplegics to stand up using FES. IEEE Trans. Rehab. Eng., TRE-6:162-166, 1998. 82. Hunt, K. J., Munih, M., and Donaldson, N., 1997. Feedback control of unsupported standing in paraplegia: Part I: Optimal control approach; Part II: Experimental results, IEEE Trans. Rehab. Eng., TRE-5:331-352, 1997. 83. Popovic, ´ D., Tomovic, ´ R., and Schwirtlich, L., Hybrid assistive system—Neuroprosthesis for motion, IEEE Trans. Biomed. Eng., BME-37:729-738, 1989. 84. Solomonow, M., Baratta, R., et al., Evaluation of 70 paraplegics fitted with the LSU RGO/FES, in Popovic, ´ D. (Ed.), Proc. II Intern. Symp. on FES, Burnaby, 1997, 159. 85. Andrews, B. J., Barnett, R. W., et al., Rule-based control of a hybrid FES orthosis for assisting paraplegic locomotion, Automedica, 11:175-199, 1989. 86. Schwirtlich, L. and Popovic, ´ D., Hybrid orthoses for deficient locomotion, in Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities VIII, ETAN, Belgrade, 1984, 23-32. 87. Phillips, C. A., An interactive system of electronic stimulators and gait orthosis for walking in the spinal cord injured, Automedica, 11:247-261, 1989. 88. Popovic, ´ D., Schwirtlich, L., and Radosavljevic,´ S., Powered hybrid assistive system, in Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities X, Nauka, Belgrade, 1990, 191-200. 89. Goldfarb, M. and Durfee, W. K., Design of a controlled-brake orthosis for FES-aided gait, IEEE Trans. Rehab. Eng., TRE-4:13-24, 1996. 90. Irby, S. E., Kaufman, K. R., and Sutherland, D.H, A digital logic controlled electromechanical long leg brace, Proc. 15 Southern Biomed. Eng. Conf, Dayton, OH, 1996, 28. 91. Kaufman, K. R., Irby, S. E., Mathewson, J. W., et al., Energy efficient knee-ankle-foot orthosis, J. Prosthet. Orthot., 8:79-85, 1996. 92. Inman, V. T., Ralston, J. J., and Todd, F., Human Walking. Williams & Wilkins, Baltimore, MD, 1981. 93. Doane, N. E. and Holt, L. E., A comparison of the SACH foot and single axis foot in the gait of the unilateral below-knee amputee, Prosth. Orthot. Intern, 7:33-36, 1983. 94. Wagner, E. M. and Catranis, J. G., New developments in lower-extremity prostheses, in Human Limbs and Their Substitutes, Klopsteg, P. E., Wilson, P. D., et al. (Eds.), McGraw-Hill, New York, 1954, (reprinted 1968). 95. Cappozzo, A., Leo, T., and Cortesi, S., A polycentric knee-ankle mechanism for above-knee prostheses, J. Biomech., 13:231-239, 1980. 96. Aoyama, F., Lapoc system leg, Proc. Rehab. Eng. Seminar REIS ‘80, Tokyo, Japan, 1980, 59-67. 97. James, K., Stein, R.B., Rolf, R., and Tepavac, D., Active suspension above-knee prosthesis, in Goh, D. and Nathan, A. (Eds.), Proc VI Intern. Conf. Biomed. Eng., 1991, 317-320. 98. Bar, A., Ishai, P., Meretsky, P., and Koren, Y., Adaptive microcomputer control of an artificial knee in level walking, J. Biomed. Eng, 5:145-150, 1983. 99. Flowers, W. C. and Mann, R.W., An Electrohydraulic knee-torque controller for a prosthesis simulator, J. Biomechan. Eng., 99:3-8, 1977. 100. Flowers, W. C., Rowell, D., Tanquary, A., and Cone, A., A microcomputer controlled knee mechanism for A/K prostheses, Proc. 3rdCISM-IFToMM Int. Symp: Theory and Practice of Robots and Manipulators, Udine, Italy, 1978, 28-42.


101. Stein, J. L. and Flowers, W. C., Above-knee prosthesis: A case study of the interdependency of effector and controller design, ASME Winter Annual Meeting, Chicago, IL, 1980, 275-277. 102. Koganezawa, E., Fujimoto, H., and Kato, I., Multifunctional above-knee prosthesis for stairs walking, Pros. Orth. Intern., 11:139-145, 1987. 103. Kuzhekin, A. P., Jacobson, J. S., and Konovalov, V. V., Subsequent development of motorized aboveknee prosthesis, in Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities VIII, ETAN, Belgrade, 1984, 525-530. 104. Popovic, ´ D. and Schwirtlich, L., Belgrade active A/K prosthesis, in deVries, J. (Ed.), Electrophysiological Kinesiology, 804:337-343, Excerpta Medica, Amsterdam, Intern. Cong. Ser, 1988. 105. Tomovic,´ R., Popovic, ´ D., Turajlic S., and McGhee, R.B., Bioengineering actuator with non-numerical control, Proc. IFAC Conf. Orthotics & Prosthetics, Columbus, OH, Pergamon Press, 1982, 145-151. 106. Lyman, J. H., Freedy, A., Prior, R., and Solomonow, M., Studies toward a practical computer-aided arm prosthesis system, Bull. Prosth. Res., 213-224, 1974. 107. Jacobsen, S. C., Knutti, F. F., Johnson, R. T., and Sears, H. H., Development of the Utah artificial arm, IEEE Trans. Biomed. Eng., BME-29:249-269, 1982. 108. Park, E. and Meek, S. G., 1995. Adaptive filtering of the electromyographic signal for prosthetic and force estimation, IEEE Trans. Biomed. Eng., BME-42:1044-1052. 109. Sheridan, T. B. and Mann, R. W., Design of control devices for people with severe motor impairment, Human Factors, 20:321-3389, 1978. 110. Simpson, D. C., The choice of control system for the multimovement prosthesis: extended physiological proprioception (e.p.p.), in The Control of Upper-Extremity Prostheses and Orthoses, Herberts, P. (Ed.), C. Thomas. 1974, chap. 15. 111. Gibbons, D. T., O’Riain, M. D., and Philippe-Auguste, J. S., An above-elbow prosthesis employing programmed linkages, IEEE Trans. Biomed. Eng., BME-34:251-258, 1987. 112. Kyberd, P. J., Holland, O. E., Chappel, P. H. et al., MARCUS: A two degree of freedom hand prosthesis with hierarchical grip control, IEEE Trans. Rehab. Eng., TRE-3:70-76, 1995. 113. Kurtz, I., Programmable prosthetic controller, Proc. MEC ‘97, Frederciton, NB, 33, 1997. 114. Bertos, Y. A., Hechathorne, C. H., Weir, R. F., and Childress, D. S., Microprocessor-based EPP position controller for electric powered upper limb prostheses, Proc. IEEE Intern. Conf. EMBS, Chicago, IL, 1997. 115. Bowker, P., Messenger, N., Oglivie, C., and Rowley, D. I., Energetics of paraplegic walking, J. Biomed. Eng., 14:344-350, 1992. 116. Nene, A. V. and Jennings, S. J., Physiological cost index of paraplegic locomotion using the ORLAU ParaWalker, Paraplegia, 30:246-252, 1992. 117. Petrofsky, J. S. and Smith, J. B., Physiologic cost of computer-controlled walking in persons with paraplegia using a reciprocating-gait orthosis, Arch. Phys. Med. Rehabil., 72:890-896, 1991. 118. Waters, R. L. and Lunsford, B. R., Energy Cost of Paraplegic Locomotion, J. Bone Joint Surg., 67A:1245-1250, 1985. 119. Waters, R. L., Yakura, J. S., Adkins, R., and Barnes, G., Determinants of gait performance following spinal cord injury, Arch. Phys. Med. Rehabil., 70:811-818, 1980. 120. Pike, personal communication.

Further Information Stein, R. B., Peckham, H. P., and Popovic, ´ D., Neural Prostheses: Replacing Motor Function After Disease or Disability, Oxford University Press, New York, 1992. Kralj, A. and Bajd, T., Functional Electrical Simulation: Standing and Walking After Spinal Cord Injury, CRC Press, Boca Raton, FL, 1989. Agnew, W. V. and McCreery, D. B., Neural Prostheses: Fundamental Studies, Prentice-Hall, Englewood Cliffs, NJ, 1990.


Proceedings: Advances in External Control of Human Extremities I-IX, Yugoslav Committee for ETAN, Belgrade, Yugoslavia, 1962-1987 (9 books). Popovic, ´ D. (Ed.), Advances in External Control of Human Extremities X, Nauka, Belgrade, 1990. Tomovic,´ R., Popovic, ´ D., and Stein, R. B., Nonanalytical Methods for Motor Control, World Scientific Publishers, Singapore, 1990.


Kaczmarek, K. A. “Sensory Augmentation and Substitution.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

143 Sensory Augmentation and Substitution 143.1

Visual System Visual Augmentation • Tactual Vision Substitution • Auditory Vision Substitution

143.2

Auditory System Auditory Augmentation • Visual Auditory Substitution • Tactual Auditory Substitution

Kurt A. Kaczmarek

143.3

University of Wisconsin at Madison

Tactual System Tactual Augmentation • Tactual Substitution

This chapter will consider methods and devices used to present visual, auditory, and tactual (touch) information to persons with sensory deficits. Sensory augmentation systems such as eyeglasses and hearing aids enhance the existing capabilities of a functional human sensory system. Sensory substitution is the use of one human sense to receive information normally received by another sense. Braille and speech synthesizers are examples of systems that substitute touch and hearing, respectively, for information that is normally visual (printed or displayed text). The following three sections will provide theory and examples for aiding the visual, auditory, and tactual systems. Because capitalizing on an existing sensory capability is usually superior to substitution, each section will consider first augmentation and then substitution, as shown below: Human Sensory Systems Visual Visual augmentation Tactual vision substitution Auditory vision substitution

Auditory

Tactual

Auditory augmentation Visual auditory substitution Tactual auditory substitution

Tactual augmentation Tactual substitution

143.1 Visual System With a large number of receptive channels, the human visual system processes information in a parallel fashion. A single glimpse acquires a wealth of information; the field of view for two eyes is 180 degrees horizontally and 120 degrees vertically [Mehr & Shindell, 1990]. The spatial resolution in the central (foveal) part of the visual field is approximately 0.5 to 1.0 minute of arc [Shlaer, 1937], although Vernier acuity, the specialized task of detecting a misalignment of two lines placed end to end, is much finer, approximately 2 seconds of arc [Stigmar, 1970]. Low-contrast presentations substantially reduce visual acuity.


The former resolution figure is the basis for the standard method of testing visual acuity, the Snellen chart. Letters are considered to be “readable” if they subtend approximately 5 minutes of arc and have details one-fifth this size. Snellen’s 1862 method of reporting visual performance is still used today. The ratio 20/40, for instance, indicates that a test was conducted at 20 ft and that the letters that were recognizable at that distance would subtend 5 minutes of arc of 40 ft (the distance at which a normally sighted, or “20/20,” subject could read them). Although the standard testing distance is 20 ft, 10 ft and even 5 ft may be used, under certain conditions, for more severe visual impairments [Fonda, 1981]. Of the approximately 6 to 11.4 million people in the United States who have visual impairments, 90% have some useful vision [NIDRR, 1993]. In the United States, severe visual impairment is defined to be 20/70 vision in the better eye with best refractive correction (see below). Legal blindness means that the best corrected acuity is 20/200 or that the field of view is very narrow ( 100 dB) are often still perceived as loud, slightly softer sounds are lost. Looked at from this perspective, it is easy to understand why the amplification and automatic gain control of conventional hearing aids do not succeed in presenting the 30-dB or so dynamic range of speech to persons with 70+ dB of sensorineural impairment. Most hearing aids perform three basic functions. (1) Amplification compensates for the reduced sensitivity of the damaged ear. (2) Frequency-domain filtering compensates for hearing loss that is not spectrally uniform. For example, most sensorineural loss disproportionately affects frequencies over 1 kHz or so, so high-frequency preemphasis may be indicated. (3) Automatic gain control (ACG) compresses the amplitude range of desired sounds to the dynamic range of the damaged ear. Typical AGC systems respond to loud transients in 2 to 5 ms (attack time) and reduce their effect in 100 to 300 ms (recovery time). Sophisticated multiband AGC systems have attempted to normalize the ear’s amplitude/frequency response, with the goal of preserving intact the usual intensity relationships among speech elements. However, recent research has shown that only certain speech features are important for intelligibility [Moore, 1990]. The fundamental frequency (due to vocal cord vibration), the first and second formants (the spectral peaks of speech that characterize different vowels), and place of articulation are crucial to speech recognition. In contrast, the overall speech envelope (the contour connecting the individual peaks in the pressure wave) is not very important; articulation information is carried in second-formant and high-frequency spectral information and is not well-represented in the envelope [Van Tasell, 1993]. Therefore, the primary design goal for hearing aids should be to preserve and make audible the individual spectral components of speech (formants and high-frequency consonant information). The cochlear implant could properly be termed an auditory augmentation device because it utilizes the higher neural centers normally used for audition. Simply stated, the implant replaces the function


of the (damaged) inner ear by electrically stimulating the auditory nerve in response to sound collected by an external microphone. Although the auditory percepts produced are extremely distorted and noiselike due to the inadequate coding strategy, many users gain sufficient information to improve their lipreading and speech-production skills. The introductory chapter in this section provides a detailed discussion of this technology.

Visual Auditory Substitution Lipreading is the most natural form of auditory substitution, requiring no instrumentation and no training on the part of the speaker. However, only about one-third to one-half of the 36 or so phonemes (primary sounds of human speech) can be reliably discriminated by this method. The result is that 30% to 50% of the words used in conversational English look just like, or very similar to, other words (homophenes) [Becker, 1972]. Therefore, word pairs such as buried/married must be discriminated by grammar, syntax, and context. Lipreading does not provide information on voice fundamental frequency or formants. With an appropriate hearing aid, any residual hearing (less than 90-dB loss) often can supply some of this missing information, improving lipreading accuracy. For the profoundly deaf, technological devices are available to supply some or all of the information. For example, the Upton eyeglasses, an example of a cued-speech device, provide discrete visual signals for certain speech sounds such as fricatives (letters like f of s, containing primarily high-frequency information) that cannot be readily identified by sight. Fingerspelling, a transliteration of English alphabet into hand symbols, can convey everyday words at up to 2 syllables per second, limited by the rate of manual symbol production [Reed et al., 1990]. American Sign Language uses a variety of upper body movements to convey words and concepts rather than just individual letters, at the same effective rate as ordinary speech, 4 to 5 syllables per second. Closed captioning encodes the full text of spoken words on television shows and transmits the data in a nonvisible part of the video signal (the vertical blanking interval). Since July of 1993, all new television sets sold in the United States with screens larger than 33 cm diagonal have been required to have builtin decoders that can optionally display the encoded text on the screen. Over 1000 hours per week of programming is closed captioned [National Captioning Institute, 1994]. Automatic speech-recognition technology may soon be capable of translating ordinary spoken discourse accurately into visually displayed text, at least in quiet environments; this may eventually be a major boon for the profoundly hearing impaired. Presently, such systems must be carefully trained on individual speakers and/or must have a limited vocabulary [Ramesh et al., 1992]. Because there is much commercial interest in speech command of computers and vehicle subsystems, this field is advancing rapidly.

Tactual Auditory Substitution Tadoma is a method of communication used by a few people in the deaf-blind community and is of theoretical importance for the development of tactual auditory substitution devices. While sign language requires training by both sender and receiver, in Tadoma, the sender speaks normally. The trained receiver places his or her hands on the face and neck of the sender to monitor lip and jaw movements, airflow at the lips, and vibration of the neck [Reed et al., 1992]. Experienced users achieve 80% keyword recognition of everyday speech at a rate of 3 syllables per second. Using no instrumentation, this the highest speech communication rate recorded for any tactual-only communication system. Alternatively, tactile vocoders perform a frequency analysis of incoming sounds, similarly to the ear’s cochlea [Békésy, 1955], and adjust the stimulation intensity of typically 8 to 32 tactile stimulators (vibrotactile or electrotactile) to present a linear spectral display to the user’s abdominal or forehead skin. Several investigators [Blamey & Clark, 1985; Boothroyd & Hnath-Chisolm, 1988; Brooks & Frost, 1986; Saunders et al., 1981] have developed laboratory and commercial vocoders. Although vocoder users cannot recognize speech as well as Tadoma users, research has shown that vocoders can provide enough


“auditory” feedback to improve the speech clarity of deaf children and to improve auditory discrimination and comprehension in some older patients [Szeto & Riso, 1990] and aid in discrimination of phonemes by lipreading [Hughes, 1989; Rakowski et al., 1989]. An excellent review of earlier vocoders appears in Reed et al. [1982]. The most useful information provided by vocoders appears to be the second-formant frequency (important for distinguishing vowels) and position of the high-frequency plosive and fricative sounds that often delineate syllables [Bernstein et al., 1991].

143.3 Tactual System Humans receive and combine two types of perceptual information when touching and manipulating objects. Kinesthetic information describes the relative positions and movements of body parts as well as muscular effort. Muscle and skin receptors are primarily responsible for kinesthesis; joint receptors serve primarily as protective limit switches [Rabischong, 1981]. Tactile information describes spatial pressure patterns on the skin given a fixed body position. Everyday touch perception combines tactile and kinesthetic information; this combination is called tactual or haptic perception. Loomis and Lederman [1986] provide an excellent review of these perceptual mechanisms. Geldard [1960] and Sherrick [1973] lamented that as a communication channel, the tactile sense is often considered inferior to sight and hearing. However, the tactile system possess some of the same spatial and temporal attributes as both of the “primary” senses [Bach-y-Rita, 1972]. With over 10,000 parallel channels (receptors) [Collins & Saunders, 1970], the tactile system is capable of processing a great deal of information if it is properly presented. The human kinesthetic and tactile senses are very robust and, in the case of tactile, very redundant. This is fortunate, considering their necessity for the simplest of tasks. Control of movement depends on kinesthetic information; tremors and involuntary movements can result from disruption of this feedback control system. Surgically repaired fingers may not have tactile sensation for a long period or at all, depending on the severity of nerve injuries; it is known that insensate digits are rarely used by patients [Tubiana, 1988]. Insensate fingers and toes (due to advanced Hansen’s disease or diabetes) are often injured inadvertently, sometimes requiring amputation. Anyone who has had a finger numbed by cold realizes that it can be next to useless, even if the range of motion is normal. The normal sensitivity to touch varies markedly over the body surface. The threshold forces in dynes for men (women) are lips, 9 (5); fingertips, 62 (25); belly 62 (7); and sole of foot, 343 (79) [Weinstein, 1968]. The fingertip threshold corresponds to 10-µm indentation. Sensitivity to vibration is much higher and is frequency-and area-dependent [Verillo, 1985]. A 5-cm2 patch of skin on the palm vibrating at 250 Hz can be felt at 0.16-µm amplitude; smaller areas and lower frequencies require more displacement. The minimal separation for two nonvibrating points to be distinguished is 2 to 3 mm on the fingertips, 17 mm on the forehead, and 30 to 50 mm on many other locations. However, size and localization judgments are considerably better than these standard figures might suggest [Vierck & Jones, 1969].

Tactual Augmentation Although we do not often think about it, kinesthetic information is reflected to the user in many types of human-controlled tools and machines, and lack of this feedback can make control difficult. For example, an automobile with power steering always includes some degree of “road feel” to allow the driver to respond reflexively to minor bumps and irregularities without relying on vision. Remote-control robots (telerobots) used underwater or in chemical- or radiation-contaminated environments are slow and cumbersome to operate, partly because most do not provide force feedback to the operator; such feedback enhances task performance [Hannaford & Wood, 1992]. Tactile display of spatial patterns on the skin uses three main types of transducers [Kaczmarek & Bachy-Rita, 1995; Kaczmarek et al., 1991]. Static tactile displays use solenoids, shape-memory alloy actuators, and scanned air or water jets to indent the skin. Vibrotactile displays encode stimulation intensity as the


amplitude of a vibrating skin displacement (10 to 500 Hz); both solenoids and piezoelectric transducers have been used. Electrotactile stimulation uses 1- to 100-mm2-area surface electrodes and careful waveform control to electrically stimulate the afferent nerves responsible for touch, producing a vibrating or tingling sensation. Tactile rehabilitation has received minimal attention in the literature or medical community. One research device sensed pressure information normally received by the fingertips and displayed it on the forehead using electrotactile stimulation [Collins & Madey, 1974]. Subjects were able to estimate surface roughness and hardness and detect edges and corners with only one sensor per fingertip. Phillips [1988] reviews prototype tactile feedback systems that use the intact tactile sense to convey hand and foot pressure and elbow angle to users of powered prosthetic limbs, often with the result of more precise control of these devices. Slightly more attention has been given to tactile augmentation in special environments. Astronauts, for example, wear pressurized gloves that greatly diminish tactile sensation, complicating extravehicular repair and maintenance tasks. Efforts to improve the situation range from mobile tactile pins in the fingertips to electrotactile stimulation on the abdomen of the information gathered from fingertip sensors [Bach-y-Rita et al., 1987].

Tactual Substitution Because of a paucity of adequate tactual display technology, spatial pressure information from a robot or remote manipulator is usually displayed to the operator visually. A three-dimensional bar graph, for example, could show the two-dimensional pressure pattern on the gripper. While easy to implement, this method suffers from two disadvantages: (1) the visual channel is required to process more information (it is often already heavily burdened), and (2) reaction time is lengthened, because the normal human tactual reflex systems are inhibited. An advantage of visual display is that accurate measurements of force and pressure may be displayed numerically or graphically. Auditory display of tactual information is largely limited to warning systems, such as excessive force on a machine. Sometimes such feedback is even inadvertent. The engine of a bulldozer will audibly slow down when a heavy load is lifted; by the auditory and vibratory feedback, the operator can literally “feel” the strain. The ubiquity of such tactual feedback systems suggests that the human-machine interface on many devices could benefit from intentionally placed tactual feedback systems. Of much current interest is the virtual environment, a means by which someone can interact with a mathematic model of a place that may or may not physically exist. The user normally controls the environment by hand, head, and body movements; these are sensed by the system, which correspondingly adjusts the information presented on a wide-angle visual display and sometimes also on a spatially localized sound display. The user often describes the experience as “being there,” a phenomenon known as telepresence [Loomis, 1992]. One can only imagine how much the experience could be enhanced by adding kinesthetic and tactile feedback [Shimoga, 1993], quite literally putting the user in touch with the virtual world.

Defining Terms Distal attribution: The phenomenon whereby events are normally perceived as occurring external to our sense organs—but also see Loomis’ [1992] engaging article on this topic. The environment or transduction mechanism need not be artificial; for example, we visually perceive objects as distant from our eyes. Electrotactile: Stimulation that evokes tactile (touch) sensations within the skin at the location of the electrode by passing a pulsatile, localized electric current through the skin. Information is delivered by varying the amplitude, frequency, etc. of the stimulation waveform. Also called electrocutaneous stimulation.


Illuminance: The density of light falling on a surface, measured in lux. One lux is equivalent to 0.0929 foot-candles, an earlier measure. Illuminance is inversely proportional to the square of the distance form a point light source. A 100-W incandescent lamp provides approximately 1280 lux at a distance of 1 ft (30.5 cm). Brightness is a different measure, depending also on the reflectance of the surrounding area. Kinesthetic perception: Information about the relative positions of and forces on body parts, possibly including efference copy (internal knowledge of muscular effort). Sensory augmentation: The use of devices that assist a functional human sense; eyeglasses are one example. Sensory substitution: The use of one human sense to receive information normally received by another sense. For example, Braille substitutes touch for vision. Sound pressure level (SPL): The root-mean-square pressure difference from atmospheric pressure (≈100 kPa) that characterizes the intensity of sound. The conversion SPL = 20 log (P/P0) expresses SPL in decibels, where P0 is the threshold pressure of approximately 20 Pa at 1 kHz. Static tactile: Stimulation that is a slow local mechanical deformation of the skin. It varies the deformation amplitude directly rather than the amplitude of vibration. This is “normal touch” for grasping objects, ect. Tactile perception: Information about spatial pressure patterns on the skin with a fixed kinesthetic position. Tactual (haptic) perception: The seamless, usually unconscious combination of tactile and kinesthetic information; this is “normal touch.” Vibrotactile: Stimulation that evokes tactile sensations using mechanical vibration of the skin, typically at frequencies of 10 to 500 Hz. Information is delivered by varying the amplitude, frequency, etc. of the vibration. Virtual environment: A real-time interactive computer model that attempts to display visual, auditory, and tactual information to a human user as if he or she were present at the simulated location. The user controls the environment with head, hand, and body motions. A airplane cockpit simulator is one example.

References Bach-y-Rita P. 1972. Brain Mechanisms in Sensory Substitution. New York, Academic Press. Bach-y-Rita P., Kaczmarek, KA, Tyler, M. and Garcia-Lara, M. 1998. Form perception with a 49-point electrotactile stimulus array on the tongue. J. Rehab. Res. Dev. 35:427–430. Bach-y-Rita P, Webster JG, Tompkins WJ, Crabb T. 1987. Sensory substitution for space gloves and for space robots. In Proceedings of the Workshop on Space Telerobotics, Jet Propulsion Laboratory, Publication 87–13, pp 51–57. Barfield, W., Hendrix, C., Bjorneseth, O., Kaczmarek, KA and Lotens, W. 1996. Comparison of human sensory capabilities with technical specifications of virtual environment equipment. Presence 4:329–356. Becker KW. 1972. Speechreading: Principles and Methods. Baltimore, National Educational Press. Békésy GV. 1955. Human skin perception of traveling waves similar to those of the cochlea. J Acoust Soc AM 27:830. Bernstein LE, Demorest ME, Coulter DC, O’Connell MP. 1991. Lipreading sentences with vibrotactile vocoders: Performance of normal-hearing and hearing-impaired subjects. J Acoust Soc Am 90:2971. Blamey PJ, Clark GM. 1985. A wearable multiple-electrode electrotactile speech processor for the profoundly deaf. J Acoust Soc Am 77:1619. Boothroyd A, Hnath-Chisolm T. 1988. Spatial, tactile presentation of voice fundamental frequency as a supplement to lipreading: Results of extended training with a single subject. J Rehabil Res Dev 25(3):51. Boyd LH, Boyd WL, Vanderheiden GC. 1990. The Graphical User Interface Crisis: Danger and Opportunity. September, Trace R&D Center, University of Wisconsin-Madison.


Brooks PL, Frost BJ. 1986. The development and evaluation of a tactile vocoder for the profoundly deaf. Can J Public Health 77:108. Collins CC. 1985. On mobility aids for the blind. In DH Warren, ER Strelow (eds), Electronic Spatial Sensing for the Blind, pp 35–64. Dordrecht, The Netherlands, Matinus Nijhoff. Collins CC, Madey JMJ. 1974. Tactile sensory replacement. In Proceedings of the San Diego Biomedical Symposium, pp 15–26. Collins CC, Saunders FA. 1970. Pictorial display by direct electrical stimulation of the skin. J Biomed Syst 1:3–16. Cook AM. 1982. Sensory and communication aids. In AM Cook, JG Webster (eds), Therapeutic Medical Devices: Application and Design, pp 152–201. Englewood Cliffs, NJ, Prentice-Hall. Fonda GE. 1981. Management of Low Vision. New York, Thieme-Stratton. Geldard FA. 1960. Some neglected possibilities of communication. Science 131:1583. Hannaford B, Wood L. 1992. Evaluation of performance of a telerobot. NASA Tech Briefs 16(2): item 62. Hughes BG. 1989. A New Electrotactile System for the Hearing Impaired. National Science Foundation final project report, ISI-8860727, Sevrain-Tech, Inc. Kaczmarek KA, Bach-y-Rita P. 1995. Tactile displays. In W Barfield, T Furness (eds), Virtual Environments and Advanced Interface Design. New York, Oxford University Press. Kaczmarek, KA, Tyler, ME and Bach-y-Rita, P. 1997. Pattern identification on a fingertip-scanned electrotactile display. Proc. 19th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. pp. 1694–1697. Kaczmarek KA, Webster JG, Bach-y-Rita, Tompkins WJ. 1991. Electrotactile and vibrotactile displays for sensory substitution systems. IEEE Trans Biomed Eng 38:1. Loomis JM. 1992. Distal attribution and presence. Presence: Teleoperators Virtual Environ 1(1):113. Loomis JM, Lederman SJ. 1986. Tactual perception. In KR Boff et al (eds), Handbook of Perception and Human Performance, vol II: Cognitive Processes and Performance, pp 31.1–31.41. New York, Wiley. Mehr E, Shindell S. 1990. Advances in low vision and blind rehabilitation. In MG Eisenberg, RC Grzesiak (eds), Advances in Clinical Rehabilitation, vol 3, pp 121–147. New York, Springer. Moore BCJ. 1990. How much do we gain by gain control in hearing aids? Acta Otolaryngol (Stockh) Suppl. 469:250. Mountcastle VB (ed). 1980. Medical Physiology. St. Louis, Mosby. National Captioning Institute, Falls Church, Va, 1994. Personal communication. NIDRR. 1993. Protocols for choosing low vision devices. U.S. Department of Education. Consensus Statement 1(1–28). Phillips CA. 1988. Sensory feedback control of upper- and lower-extremity motor prostheses. CRC Crit Rev Biomed Eng 16:105. Rabischong P. 1981. Physiology of sensation. In R Tubiana (ed), The Hand, pp 441–467. Philadelphia, Saunders. Rakowski K, Brenner C, Weisenberger JM. 1989. Evaluation of a 32-channel electrotactile vocoder (abstract). J Acoust Soc Am 86(suppl 1):S83. Ramesh P, Wilpon JG, McGee MA, et al. 1992. Speaker independent recognition of spontaneously spoken connected digits. Speech Commun 11:229. Reed CM, Delhorne LA, Durlach NI, Fischer SD. 1990. A study of the tactual and visual reception of fingerspelling. J Speech Hear Res 33:786. Reed CM, Durlach NI, Bradia LD. 1982. Research on tactile communication of speech: A review. AHSA Monogr 20:1. Reed CM, Rabinowitz WM, Durlach NI, et al. 1992. Analytic study of the Tadoma method: Improving performance through the use of supplementary tactile displays. J Speech Hear Res 35:450. Sataloff J, Sataloff RT, Vassallo LA. 1980. Hearing Loss, 2d ed. Philadelphia, Lippincott. Saunders FA, Hill WA, Franklin B. 1981. A wearable tactile sensory aid for profoundly deaf children. J Med Syst 5:265. Sherrick CE. 1973. Current prospects for cutaneous communication. In Proceedings of the Conference on Cutaneous Communication System Development, pp 106–109.


Shimoga KB. 1993. A survey of perceptual feedback issues in dextrous telemanipulation: II. Finger touch feedback. In IEEE Virtual Reality Annual International Symposium, pp 271–279. Shlaer S. 1937. The relation between visual acuity and illumination. J Gen Physiol 21:165. Smeltzer CD. 1993. Primary care screening and evaluation of hearing loss. Nurse Pract 18:50. Stigmar G. 1970. Observation on vernier and stereo acuity with special reference to their relationship. Acta Ophthalmol 48:979. Szeto AYJ, Riso RR. 1990. Sensory feedback using electrical stimulation of the tactile sense. In RV Smith, JH Leslie Jr (eds), Rehabilitation Engineering, pp 29–78. Boca Raton, Fla, CRC Press. Tubiana R. 1988. Fingertip injuries. In R Tubiana (ed), The Hand, pp 1034–1054. Philadelphia, Saunders. Van Tasell DJ. 1993. Hearing loss, speech, and hearing aids, J Speech Hear Res 36:228. Verrillo RT. 1985. Psychophysics of vibrotactile stimulation. J Acoust Soc Am 77:225. Vierck CJ, Jones MB. 1969. Size discrimination on the skin. Science 163:488. Webster JG, Cook AM, Tompkins WJ, Vanderheiden GC (eds). 1985. Electronic Devices for Rehabilitation. New York, Wiley. Weinstein S. 1968. Intensive and extensive aspects of tactile sensitivity as a function of body part, sex and laterality. In DR Kenshalo (ed), The Skin Senses, pp 195–218. Springfield, Ill, Charles C Thomas.

Further Information Presence: Teleoperators and Virtual Environments is a bimonthly journal focusing on advanced humanmachine interface issues. In an effort to develop tactile displays without moving parts, our laboratory has demonstrated simple pattern recognition on the fingertip [Kaczmarek et al., 1997] and tongue [Bachy-Rita et al., 1998] using electrotactile stimulation. The Trace Research and Development Center, Madison, Wisc., publishes a comprehensive resource book on commercially available assistive devices, organizations, etc. for communication, control, and computer access for individuals with physical and sensory impairments. Electronic Devices for Rehabilitation, edited by J. G. Webster (Wiley, 1985), summarizes the technologic principles of electronic assistive devices for people with physical and sensory impairments.


Romich, B., Vanderheiden, G., Hill, K. “Hard Tissue Replacement.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

144 Augmentative and Alternative Communication Barry Romich Prentke Romich Company

Gregg Vanderheiden

144.1 144.2

University of Wisconsin

Katya Hill Edinboro University of Pennsylvania and University of Pittsburgh

144.3 144.4 144.5 144.6

Introduction Language Representation Methods and Acceleration Techniques User Interface Outputs Outcomes, Intervention and Training Future

144.1 Introduction The inability to express oneself through either speech or writing is perhaps the most limiting of physical disabilities. Meaningful participation in life requires the communication of basic information, desires, needs, feelings, and aspirations. The lack of full interpersonal communication substantially reduces an individual’s potential for education, employment, and independence. Today, through multidisciplinary contributions, individuals who cannot speak or write effectively have access to a wide variety of techniques, therapies and systems designed to ameliorate challenges to verbal communication. The field of augmentative and alternative communication (AAC) consists of many different professions, including speech-language pathology, regular and special education, occupational and physical therapy, engineering, linguistics, technology, and others. Experienced professionals now can become certified as Assistive Technology practitioners and/or suppliers (Minkel, 1996). Individuals who rely on AAC, as well as their families and friends, also contribute to the field (Slesaransky-Poe, 1998). Engineering plays a significant role in the development of the field of AAC and related assistive technology. Engineering contributions range from relatively independent work on product definition and design to the collaborative development and evaluation of tools to support the contributions of other professions, such as classical and computational linguistics and speech-language pathology. Augmentative communication can be classified in a variety of ways ranging from unaided communication techniques, such as gestures, signs, and eye pointing, to highly sophisticated electronic devices employing the latest technology. This chapter focuses on the review of technology-aided techniques and related issues. Technology-based AAC systems have taken two forms: hardware designed specifically for this application and software that runs on mass market computer hardware. Three basic components comprise AAC systems. These are the language representation method (including acceleration techniques), the user interface, and the outputs. Generally, a multidisciplinary team evaluates the current and projected


skills and needs of the individual, determines the most effective language representation method(s) and physical access technique(s), and then selects a system with characteristics that are a good match.

144.2 Language Representation Methods and Acceleration Techniques The ultimate goal of AAC intervention is functional, interactive communication. The four purposes that communication fulfills are (1) communication of needs/wants, (2) information transfer, (3) social closeness, and (4) social etiquette (Light, 1988). From the perspective of the person who uses AAC, communicative competence involves the ability to transmit messages efficiently and effectively in all four of the interaction categories, based on individual interests, circumstances, and abilities (Buekelman & Miranda, 1998). To achieve communication competence, the person using an AAC system must have access to language representation methods capable of handling the various vocabulary and message construction demands of the environment. Professionals rely on the theoretical models of language development and linguistics to evaluate the effectiveness of an AAC language representation method. Language is defined as an abstract system with rules governing the sequencing of basic units (sounds, morphemes, words, and sentences), and rules governing meaning and use (McCormick & Schiefelbusch, 1990). An individual knows a language when he or she understands and follows its basic units and rules. Knowledge of language requires both linguistic competence (understanding the rules), and linguistic performance (using these rules). Most language models identify the basic rules as phonology, semantics, morphology, syntax and pragmatics. The basic rules of language apply to AAC. For example, AAC research on vocabulary use has documented the phenomenon of core and fringe vocabulary (Yorkston et al., 1988) and the reliance on a relatively limited core vocabulary to express a majority of communication utterances (Vanderheiden & Kelso, 1987). Research on conversations has documented topic and small talk patterns (Stuart et al., 1993; King et al., 1995). An awareness of how a given AAC system handles these basic rules enables one to be critical of AAC language representation methods. In addition to the language representation method, the acceleration technique(s) available in an AAC system contribute(s) to communication competence. Communication by users of AAC systems is far slower than that of the general population. Yet the speed of communication is a significant factor influencing the perceptions of the user’s communication partner, and potential for personal achievement for the person relying on AAC. The development and application of techniques to accelerate communication rates is critical. Further, these techniques are most effective when developers pay attention to human factors design principles (Goodenough-Trepagnier, 1994). Alphabet-based language representation methods involve the use of traditional orthography and acceleration techniques that require spelling and reading skills. AAC systems using orthography require the user to spell each word using a standard keyboard or alphabet overlay on a static or dynamic display. A standard or customized alphabet overlay provides use for all the rules and elements of a natural language; however, spelling letter-by-letter is a slow and inefficient AAC strategy without acceleration techniques. Abbreviation systems represent language elements using a number of keystrokes typically smaller than that required by spelling. For example, words or sentences are abbreviated using principled approaches based on vowel elimination or the first letters of salient words. Abbreviation systems can be fast, but require not only spelling and reading skills, but memory of abbreviation codes. Typically, people with spelling and reading skills have large vocabulary needs and increased demands on production of text. Demasco (1994) offers additional background and proposes some interesting work in this area. Word prediction is another acceleration technique available from many sources. Based on previously selected letters and words, the system presents the user with best guess choices for completing the spelling of a word. The user then chooses one of the predictions or continues spelling, resulting in yet another set of predictions. Prediction systems have demonstrated a reduction in the number of keystrokes, but recent research (Koester and Levine, 1994) reports that the actual communication rate does not represent a statistical improvement over spelling. The reason for word predication’s failure to improve rate is that


increased time is needed to read and select the word table choices. The cost of discontinuity and increased cognitive load in the task seems to match the benefits of reduced keystrokes. Picture Symbol-based language representation methods involve the use of graphic or line drawn symbols to represent single word vocabulary or messages (phrases, sentences, and paragraphs). A variety of AAC symbol sets are available on devices or software depict the linguistics elements available through the system. Picture Communication Symbols (PCS), DynaSyms, and Blissymbols are popular symbol sets used in either dedicated or computer-aided systems. One taxonomy differentiates symbols according to several subordinate levels including static/dynamic, iconic/opaque, and set/system (Fuller, et.al., 1998) Universal considerations regarding the selection of a symbol set include research on symbol characteristics such as size, transparency, complexity and iconicity (Fuller et al., 1997, Romski & Sevcik, 1988). Understanding symbol characteristics is necessary to make clinical decisions about vocabulary organization and system selection. The choice of picture communication language representation methods is facilitated when teams use a goal-driven graphic symbol selection process (Schlosser et al., 1996). For example, identification of vocabulary and language outcomes assists the selection of an AAC graphic symbol system. Vocabulary size and representation of linguistic rules (grammar) are concerns for users relying on graphic symbol sets. Users of static display systems have a limited number of symbols available on any one overlay; however, they have ready access to that vocabulary. Dynamic display users have an almost unlimited number of symbols available as vocabulary for message generation; however, they must navigate through pages or displays to locate a word. Frequently, morphology and syntax are not graphically represented. Since research has strongly supported the need for users to construct spontaneous, novel utterances (Beukelman et al., 1984), neither method is efficient for interactive communication. Semantic Compaction or Minspeak is perhaps the most commonly used AAC language representation method (Baker, 1994). With this method, language is represented by a relatively small set of multimeaning icons. The specific meaning of each icon is a function of the context in which it is used. Semantic compaction makes use of a meaningful relationship between the icon and the information it represents; it does not require spelling and reading skills, and yet is powerful even for people with these skills. The performance of Minspeak stems from its ability to handle both vocabulary and linguistic structures as found in the Minspeak Application Programs (MAPS) of Words Strategy and Unity. Both MAPS support the concept of a core and fringe vocabulary. They provide the architecture for handling rules of grammar and morphology. The number of required keystrokes is reduced relative to spelling. Predictive selection is an acceleration technique used with Minspeak. When this feature is enabled, only those choices that complete a meaningful sequence can be selected. With scanning, for example, the selection process can be significantly faster because the number of possible choices is automatically reduced.

144.3 User Interface Most AAC systems employ a user interface based on the selection of items that will produce the desired output (Vanderheiden and Lloyd, 1986). Items being selected may be individual letters, as used in spelling, or whole words or phrases expressing thoughts, or symbols that represent vocabulary. The numerous techniques for making selections are based on either direct selection or scanning. Direct selection refers to techniques by which a single action from a set of choices indicates the desired item. A common example of this method is the use of a computer keyboard. Each key is directly selected by finger. Expanded keyboards accommodate more gross motor actions, such as using the fist or foot. In some cases, pointing can be enhanced through the use of technology. Sticks are held in the mouth or attached to the head using a headband or helmet. Light pointers are used to direct a light beam at a target. Indirect pointing systems might include the common computer mouse, trackball, or joystick. Alternatives to these for people with disabilities are based on the movement of the head or other body part. Figure 144.1 depicts direct selection in that the desired location is pointed to directly.


Scanning refers to techniques in which the individual is presented with a time sequence of choices and indicates when the desired choice appears. A simple linear scanning system might be a clock face type display with a rotating pointer to indicate a letter, word, or picture. Additional dimensions of scanning can be added to reduce the selection time when the number of possible choices is larger. A common tech- FIGURE 144.1 Direct selection. nique involves the arrangement of choices in a matrix of rows and columns. The selection process has two steps. First, rows are scanned to select the row containing the desired element. The selected row is then scanned to select the desired element. This method is called row-column scanning. See Fig. 144.2. Either by convention or by the grouping of the elements, the order might be reversed. For example, in the U.K., column-row scanning is preferred over row-column scanning. Other scanning techniques also exist and additional dimensions can be employed. Both direct selection and scanning are used to select elements that might not of themselves define an output. In these cases the output may be defined by a code of selected elements. A common example is Morse code by which dots and dashes are directly selected but must be combined to define letters and numbers. Another example, more common to AAC, has an output defined by a sequence of two or more pictures or icons. Scanning, and to some degree direct selection, can be faster when the choices are arranged such that those most frequently used are easiest to access. For example, in a row-column scanning spelling system that scans top to bottom and left to right, the most frequently used letters are usually grouped toward the upper left corner. Generally the selection technique of choice should be that which results in the fastest communication possible. Consideration of factors such as cognitive load (Cress and French, 1994), environmental changes, fatigue, aesthetics, and stability of physical skill often influence the choice of the best selection technique.

FIGURE 144.2


Row-column scanning.

144.4 Outputs AAC system outputs in common use are speech, displays, printers, beeps, data, infrared, and other control formats. Outputs are used to facilitate the interactive nature of “real time” communication. Users may rely on auditory and visual feedback to enhance vocabulary selection and the construction of messages. The speech and display outputs may be directed toward the communication partner to support the exchange of information. Auditory feedback such as beeps and key clicks, while useful to the individual using the system, also provide the communication partner with the pragmatic information that a user is in the process of generating a message. Finally, outputs may be used to control other items or devices such as printers. AAC speech output normally consists of two types: synthetic and digitized. Synthetic speech usually is generated from text input following a set of rules. Synthetic speech is usually associated with AAC systems that are able to generate text. These systems have unlimited vocabulary and are capable of speaking any word or expression that can be spelled. Commonly used synthetic speech systems like DECtalk™ offer a variety of male, female, and child voices. Intelligibility has improved over recent years to the point that it is no longer a significant issue. With some systems it is actually possible to sing songs, again a feature that enhances social interaction. Most synthetic speech systems are limited to a single language, although bilingual systems are available. Further limitations relate to the expression of accent and emotion. Research and development in artificial speech technology is attacking these limitations. Digitized speech is essentially speech that has been recorded into digital memory. Relatively simple AAC systems typically use digitized speech. The vocabulary is entered by speaking into the system through a microphone. People who use these systems can say only what someone else said in programming them. They are independent of language and can replicate the song, accent, and emotion of the original speaker. RS-232c serial data output is used to achieve a variety of functional outcomes. Serial output permits the AAC system to replace the keyboard and mouse for computer access, a procedure known as emulation. The advantage of emulation is that the language representation method and physical access method used for speaking is used for writing and other computer tasks (Buning and Hill, 1998). Another application for the serial output is environmental control. Users are able to operate electrical and electronic items in their daily-living surroundings. Particular sequences of characters, symbols or icons are used to represent commands such as answering the telephone, turning off the stereo, and even setting a thermostat. In addition, the serial output also may be used to monitor the language activity for purposes of clinical intervention, progress reporting, and research. Infrared communication now is available in many AAC devices. Infrared output supports the same functions as the RS-232c serial data output, but without requiring direct wiring or linking between operating devices. Infrared interfaces providing computer access improve the independence of AAC users because no physical connection needs to be manipulated to activate the system. Infrared interfaces can perform as environmental control units by learning the codes of entertainment and other electronic systems.

144.5 Outcomes, Intervention and Training The choice of an AAC system should rely primarily on the best interests of the person who will be using the system. Since outcomes and personal achievement will be related directly to the ability to communicate, the choice of a system will have lifelong implications. The process of choosing the most appropriate system is not trivial and can be accomplished best with a multidisciplinary team focusing on outcomes. Team members should realize that the interests of the individual served professionally are not necessarily aligned with those of the providers and/or payers of AAC services. The temptation to select a system that is easy to apply or inexpensive to purchase frequently exists, especially with untrained teams. Teams that identify outcomes and have the goal of achieving interactive communication make informed decisions.


FIGURE 144.3 Wooster, OH.

Individual using an electronic communication system. Photo courtesy of Prentke Romich Company,

Following the selection (and funding) of a system, the next step is the actual intervention program. For successful outcomes to be achieved, intervention must go beyond the technical use of the AAC system and include objectives for language development and communication pragmatics. Early in the history of AAC, Rodgers postulates (1984) that AAC systems are more like tools than appliances. To be effective, there is a need for much more than simply “plugging them in and turning them on.” The individual who relies on AAC must develop the skills to become a communication craftsperson. In general, as well as specific to AAC, the fastest, most efficient use of a system occurs when the individual operates from knowledge that is in the head, rather than knowledge that must be gathered from the world (Norman, 1980). The implication for AAC is that an intervention program must include a significant training component to assure that the needed knowledge is put into the head (Romich, 1994). Further, a drill component of training develops automaticity (Treviranus, 1994). Perhaps the single factor limiting the widespread use of assistive technology in general and AAC in particular is the lack of awareness of its availability and the impact it can have on the lives of the people who could benefit (Romich, 1993). This situation exists within not only the general public but also in many of the professions providing services to this population. Training opportunities continue to be limited for a majority of professionals working with persons who could benefit from AAC technology. University programs have lagged behind in their integration of this information into professional curricula. Intervention and training should extend beyond the initial application of the system. People who rely on AAC frequently live unstable lives. Consequently they have a need for on-going services. For people with congenital (from birth) conditions, developmental delays may have occurred that result in educational and training needs past the normal age of public school services. For people with progressive neurological disorders, such as Lou Gehrig’s disease, the physical skill level will change and additional accommodations will need to be evaluated and implemented.

144.6 Future Technological development in and of itself will not solve the problems of people with disabilities. Technology advancements, however, will continue to provide more powerful tools allowing the exploration of new and innovative approaches to support users and professionals. As access to lower cost, more powerful computer systems increases, the availability of alternative sources of information, technical support and training is possible either through software or the internet. Users and professionals will have increased opportunities to learn and exchange ideas and outcomes. © 2000 by CRC Press LLC

A related development is the collection and analysis of data describing the actual long term use of AAC devices. Currently, clinicians and researchers have not utilized this information in the clinical intervention process, progress reporting, or research. Tools for monitoring, editing, and analyzing language activity are just now becoming available (Hill and Romich, 1998).

References Baker BR. 1994. Semantic Compaction: An Approach to a Formal Definition. Proceedings of the Sixth Annual European Minspeak Conference, Swinstead, Lincs, UK: Prentke Romich Europe. Beukelman DR, Yorkson K, Poblete M, Naranjo C. 1984. Frequency of word occurrence in communication samples produced by adult communication aid users, J. Speech Hearing Dis., 49, 360-367. Beukelman DR, Mirenda P. 1998. Augmentative and Alternative Communication: Management of Severe Communication Disorders in Children and Adults, Baltimore, Paul H. Brookes Publishing Co. Buning ME, Hill K. 1999. An AAC device as a computer keyboard: More bang for the buck. AOTA Annual Conference, Indianapolis. Cress CJ, French GJ. 1994. The Relationship between Cognitive Load Measurements and Estimates of Computer Input Control Skills, Ass. Tech. 6.1, 54-66. Demasco P. 1994. Human Factors Considerations in the Design of Language Interfaces in AAC, Ass. Tech. 6.1, 10-25. Fuller D, Lloyd L, Schlosser R. 1992. Further development of an augmentative and alterntive communication symbol taxonomy, Aug. Alt. Comm. 8:67-74. Fuller D, Lloyd L, Stratton M. 1997. Aided AAC symbols. In L Lloyd, D Fuller, & H Arvidson (eds.), Augmentative and alternative communication: Principles and practice, pp 48-79, Needham Heights, MA, Allyn & Bacon. Goodenough-Trepagnier D. 1994. Design Goals for Augmentative Communication, Ass. Tech. 6.1, 3-9. Hill K, Romich B. 1998. Language Research Needs and Tools in AAC, Ann. Biomed. Eng. 26, 131. Horstmann Koester H, Levine SP. 1994. Learning and Performance of Able-Bodied Individuals Using Scanning Systems with and without Word Prediction, Ass. Tech. 6.1, 42-53. King J, Spoeneman T, Stuart S, Beukelman D. 1995. Small talk in adult conversations, Aug. Alt. Comm. 11, 244-248. Light J. 1988. Interaction involving individuals using augmentative and alternative communication systems: State of the art and future directions, Aug. Alt. Comm. 4, 66-82. McCormick L, Schiefelbusch RL. 1990. Early Language Intervention: An Introduction, 2nd ed., Columbus, Merrill Publishing Company. Minkel J. 1996. Credentialing in Assistive Technology: Myths and Realities, RESNA News, 8(5), 1. Norman DA. 1980. The Psychology of Everyday Things, New York, Basic Books, Inc.. Rodgers B. 1984. Presentation at Discovery ‘84 Conference, Chicago, IL. Romich B. 1993. Assistive Technology and AAC: An Industry Perspective, Ass. Tech. 5.2, 74-77. Romich B. 1994. Knowledge in the World vs. Knowledge in the Head: The Psychology of AAC Systems, Comm. out. 14(4). Romski M, Sevcik R. 1988. Augmentative and alternative communication systems: Considerations for individuals with severe intellectual disabilities, Aug. Alt. Comm. 4, 83-93. Schlosser RW, Lloyd LL, McNaughton S. 1996. Graphic symbol selection in research and practice: Making the case for a goal-driven process, Communication… Naturally: Theoretical and Methodological Issues in Aug. Alt. Comm. E. Bjorck-Akesson and P. Lindsay (eds.), Proc. of the Forth ISAAC Research Symposium, pp. 126-139. Slesaransky-Poe G. 1998. ACOLUG: The Communication On-Line User’s Group, ISAAC Proc., 51-52. Stuart S, Vanderhoof D, Beukelman D. 1993. Topic and vocabulary use patterns of elderly women, Aug. Alt. Comm. 9, 95-110. Treviranus J. 1994. Mastering Alternative Computer Access: The Role of Understanding, Trust, and Automaticity, Ass. Tech. 6.1, 26-41.


Vanderheiden GC, Lloyd LL. 1986. In SW Blackstone (ed.), Augmentative Communication: An Introduction, Rockville, MD: American Speech-Language-Hearing Association. Vanderheiden GC, Kelso D. 1987. Comparative analysis of fixed-vocabulary comunication acceleration techniques. Aug. Alt. Comm. 3:196-206. Yorkston KM, Dowden PA, Honsinger MJ, Marriner N, Smith K. 1988. A Comparison of standard and user vocabulary list, Aug. Alt. Comm. 4:189-210.

Further Information There are a number of organizations and publications that relate to AAC. AAC (Augmentative and Alternative Communication) is the quarterly refereed journal of ISAAC. It is published by Decker Periodicals Inc., PO Box 620, L.C.D. 1, Hamilton, Ontario, L8N 3K7 CANADA, Tel. 905-522-7017, Fax. 905-522-7839. http://www.isaac-online.org ACOLUG (Augmentative Communication On-Line User Group) is a listserve with primary participants being people who rely on AAC. Topics include a wide range of issues of importance to this population, their relatives, and friends, and those who provide AAC services. http://nimbus.ocis.temple.edu/~kcohen/listserv/homeacolug.html American Speech-Language-Hearing Association (ASHA) is the professional organization of speechlanguage pathologists. ASHA has a Special Interest Division on augmentative communication. ASHA, 10801 Rockville Pike, Rockville, MD, 20852, Tel. 301-897-5700, Fax. 301-571-0457. http://www.asha.org. Augmentative Communication News is published bi-monthly by Augmentative Communication, Inc., 1 Surf Way, Suite #215, Monterey, CA, 93940, Tel. 408-649-3050, Fax. 408-646-5428. CAMA (Communication Aid Manufacturers Association) is an organization of manufacturers of AAC systems marketed in North America. CAMA, 518-526 Davis St., Suite 211-212, Evanston, IL, 60201, Tel. 800-441-2262, Fax. 708-869-5689. http://www.aacproducts.org. Communicating Together is published quarterly by Sharing to Learn as a means of sharing the life experiences and communication systems of augmentative communicators with other augmentative communicators, their families, their communities, and those who work with them. Sharing to Learn, PO Box 986, Thornhill, Ontario, L3T 4A5, CANADA, Tel. 905-771-1491, Fax. 905-771-7153. http://www.isaaconline.org. Communication Outlook is an international quarterly addressed to the community of individuals interested in the application of technology to the needs of persons who experience communication handicaps. It is published by the Artificial Language Laboratory, Michigan State University, 405 Computer Center, East Lansing, MI, 48824-1042, Tel. 517-353-0870, Fax. 517-353-4766. http://www.msu.edu/~artlang/CommOut.html. Trace Research & Development Center, University of Wisconsin—Madison. http://www.trace.wisc.edu. ISAAC is the International Society for Augmentative and Alternative Communication). USSAAC is the United States chapter. Both can be contacted at PO Box 1762 Station R, Toronto, Ontario, M4G 4A3, CANADA, Tel. 905-737-9308, Fax 905-737-0624. http://www.isaac-online.org. RESNA is an interdisciplinary association for the advancement of rehabilitation and assistive technologies. RESNA has many Special Interest Groups including those on Augmentative and Alternative Communication and Computer Applications. RESNA, 1700 North Moore Street, Suite 1540, Arlington, VA, 22209-1903, Tel. 703-524-6686.


Kondraske, G. V. “Measurement Tools and Processes in Rehabilitation Engineering.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

145 Measurement Tools and Processes in Rehabilitation Engineering 145.1

Fundamental Principles Structure, Function, Performance, and Behavior • Subjective and Objective Measurement Methods • Measurements and Assessments

145.2

Measurement Objectives and Approaches Characterizing the Human System and Its Subsystems • Characterizing Tasks • Characterizing Assistive Devices • Characterizing Overall Systems in High-Level-Task Situations

George V. Kondraske University of Texas at Arlington

145.3 145.4

Decision-Making Processes Current Limitations Quality of Measurements • Standards • Rehabilitation Service Delivery and Rehabilitation Engineering

In every engineering discipline, measurement facilitates the use of structured procedures and decisionmaking processes. In rehabilitation engineering, the presence of "a human," the only or major component of the system of interest, has presented a number of unique challenges with regard to measurement. This is especially true with regard to the routine processes of rehabilitation that either do or could incorporate and rely on measurements. This, in part, is due to the complexity of the human system's architecture, the variety of ways in which it can be adversely affected by disease or injury, and the versatility in the way it can be used to accomplish various tasks of interest to an individual. Measurement supports a wide variety of assistive device design and prescription activities undertaken within rehabilitation engineering (e.g., Leslie and Smith [1990], Webster et al. [1985], and other chapters within this section). In addition, rehabilitation engineers contribute to the specification and design of measurement instruments that are used primarily by other service providers (such as physical and occupational therapists). As measurements of human structure, performance, and behavior become more rigorous and instruments used have taken advantage of advanced technology, there is also a growing role for rehabilitation engineers to assist these other medical professionals with the proper application of measurement instruments (e.g., for determining areas that are most deficient in an individual's performance profile, objectively documenting progress during rehabilitation, etc.). This is in keeping with the team approach to rehabilitation that has become popular in clinical settings. In short, the role of measurement in rehabilitation engineering is dynamic and growing.


In this chapter, a top-down overview of measurement tools and processes in rehabilitation engineering is presented. Many of the measurement concepts, processes, and devices of relevance are common to applications outside the rehabilitation engineering context. However, the nature of the human population with which rehabilitation engineers must deal is arguably different in that each individual must be assumed to be unique with respect to at least a subset of his or her performance capacities and/or structural parameters; i.e., population reference data cannot be assumed to be generally applicable. While there are some exceptions, population labels frequently used such as "head-injured" or "spinal cordinjured" represent only a gross classification that should not be taken to imply homogeneity with regard to parameters such as range of motion, strength, movement speed, information processing speed, and other performance capacities. This is merely a direct realization that many different ways exist in which the human system can be adversely affected by disease or injury and recognition of the continuum that exists with regard to the degree of any given effect. The result is that in rehabilitation engineering, compared with standard human factors design tasks aimed at the average healthy population, many measurement values must be acquired directly for the specific client. Measurement in the present context encompasses actions that focus on (1) the human (e.g., structural aspects and performance capacities of subsystems at different hierarchical levels ranging from specific neuromuscular subsystems to the total person and his or her activities in daily living, including work), (2) assistive devices (e.g., structural aspects and demands placed on the human), (3) tasks (e.g., distances between critical points, masses of objects involved, etc.), and (4) overall systems (e.g., performance achieved by a human-assistive device-task combination, patterns of electrical signals representing the timing of muscle activity while performing a complex maneuver, behavior of an individual before and after being fitted with a new prosthetic devices, etc). Clearly, an exhaustive treatment is beyond the scope of this chapter. Measurements are embedded in every specialized subarea of rehabilitation engineering. However, there are also special roles served by measurement in a broader and more generic sense, as well as principles that are common across the many special applications. Emphasis here is placed on these. There is no lack of other literature regarding the types of measurement outlined to be of interest here and their use. However, it is diffusely distributed, and gaps exist with regard to how such tools can be integrated to accomplish goals beyond simply the acquisition of numeric data for a given parameter. With rapidly changing developments over the last decade, there is currently no comprehensive source that describes the majority of instruments available, their latest implementations, procedures for their use, evaluation of effectiveness, etc. While topics other than measurement are discussed, Leslie and Smith [1990] produced what is perhaps the single most directly applicable source with respect to rehabilitation engineering specifically, although it too is not comprehensive with regard to measurement, nor does it attempt to be.

145.1 Fundamental Principles Naturally, the fundamental principles of human physiology manifest themselves in the respective sensory, neuromuscular, information-processing, and life-sustaining systems and impact approaches to measurement. In addition, psychological considerations are vital. Familiarization with this material is essential to measurement in rehabilitation; however, treatment here is far beyond the scope of this chapter. The numerous reference works available may be most readily found by consulting relevant chapters in this Handbook and the works that they reference. In this section, key principles that are more specific to measurement and of general applicability are presented.

Structure, Function, Performance, and Behavior It is necessary to distinguish between structure, function, performance, and behavior and measurements thereof for both human and artificial systems. In addition, hierarchical systems concepts are necessary


both to help organize the complexity of the systems involved and to help understand the various needs that exist. Structural measures include dimensions, masses (of objects, limb segments), moments of inertia, circumferences, contours, compliances, and any other aspects of the physical system. These may be considered hierarchically as being pertinent to the total human (e.g., height, weight, etc.), specific body segments (e.g., forearm, thigh, etc.), or components of basic systems such as tendons, ligaments and muscles. Function is the purpose of the system of interest (e.g., to move a limb segment, to communicate, to feed and care for oneself). Within the human, there are many single-purpose systems (e.g., those that function to move specific limb segments, process specific types of information, etc.). As one proceeds to higher levels, such as the total human, systems that are increasingly more multifunctional emerge. These can be recognized as higher-level configurations of more basic systems that operate to feed oneself, to conduct personal hygiene, to carry out task of a job, etc. This multilevel view of just functions begins to help place into perspective the scope over which measurement can be applied. In rehabilitation in general, a good deal of what constitutes measurement involves the application of structured subjective observation techniques (see also the next subsection) in the form of a wide range of rating scales [e.g., Fuhrer, 1987; Granger & Greshorn, 1984; Potvin et al., 1985]. These are often termed functional assessment scales and are typically aimed at obtaining a global index of an individual's ability to function independently in the world. The global index is typically based on a number of items within a given scale, each of which addresses selected, relatively high-level functions (e.g., personal hygiene, mobility, etc.). The focus of measurement for a given item is often in estimate of the level of independence or dependence that the subject exhibits or needs to carry out the respective function. In addition, inventories of functions that an individual is able or not able to carry out (with and without assistance) are often included. The large number of such scales that have been proposed and debated is a consequence of the many possible functions and combinations thereof that exist on which to base a given scale. Functional assessment scales are relatively quick and inexpensive to administer and have a demonstrated role in rehabilitation. However, the nature and levels of measurements obtained are not sufficient for many rehabilitation engineering purposes. This latter class of applications generally begins with a function at the level and of the type used as a constituent component of functional assessment scales, considers the level of performance at which that function is executed more quantitatively, and incorporates one or more lower levels in the hierarchy (i.e., the human subsystems involved in achieving the specific functions of daily life that are of interest and their capacities for performance). Where functions can be described and inventoried, performance measures directly characterize how well a physical system of interest executes its intended function. performance is multidimensional (e.g., strength, range, speed, accuracy, steadiness, endurance, etc.). Of special interest are the concepts of performance capacity and performance capacity measurement. Performance capacity represents the limits of a given system's ability to operate in its corresponding multidimensional performance space. In this chapter, a resource-based model for both human and artificial system performance and measurement of their performance capacities is adopted [e.g., Kondraske, 1990, 1995]. Thus the maximum knee flexor strength available (i.e., the resource availability) under a stated set of conditions represents one unique performance capacity of the knee flexor system. In rehabilitation, the terms impairment, disability, and handicap [World Health Organization, 1980] have been prominently applied and are relevant to the concept of performance. While these terms place an emphasis on what is missing or what a person cannot do and imply not only a measurement but also the incorporation of an assessment or judgment based on one or more observations the resource-based performance perspective focuses on "what is present" or "what is right" (i.e., performance resource availability). From this perspective, an impairment can be determined to exist if a given performance capacity is found to be less than a specified level (e.g., less than 5th percentile value of a health reference population). A disability exists when performance resource insufficiency exists in a specified task.


While performance relates more to what a system can do (i.e., a challenge or maximal stress is implied), behavior measurements are used to characterize what a system does naturally. Thus a given variable such as movement speed can relate to both performance and behavior depending on whether the system (e.g., human subsystem) was maximally challenged to respond "as fast as possible" (performance) or simply observed in the course of operation (behavior). It is also possible to observe a system that it is behaving at one or more of its performance capacities (e.g., at the maximum speed possible, etc.) (see Table 145.1).

Subjective and Objective Measurement Methods Subjective measurements are made by humans without the aid of instruments and objective measurements result from the use of instruments. However, it should be noted that the mere presence of an instrument does not guarantee complete objectivity. For example, the use of a ruler requires a human judgment in reading the scale and thus contains a subjective element. A length-measurement system with an integral data-acquisition system would be more objective. However, it is likely that that even this system would involve human intervention in its use, e.g., the alignment of the device and making the decision as to exactly what is to be measured with it by selection of reference points. Measures with more objectivity (less subjectivity) are preferred to minimize questions of bias. However, measurements that are intrinsically more objective are frequently more costly and time-consuming to obtain. Well-reasoned tradeoffs must be made to take advantage of the ability of a human (typically a skilled professional) to quickly "measure" many different items subjectively (and often without recording the results but using them internally to arrive at some decision). It is important to observe that identification of the variable of interest is not influenced by whether it is measured subjectively or objectively. This concept extends to the choice of instrument used for objective measurements. This is an especially important concept in dealing with human performance and behavior, since variables of interest can be much more abstract than simple lengths and widths (e.g., coordination, postural stability, etc.) In fact, many measurement variables in rehabilitation historically have tended to be treated as if they were inextricably coupled with the measurement method, confounding debate regarding what should be measured with what should be used to measure it in a given context.

Measurements and Assessments The basic representation of a measurement itself in terms of the actual units of measure is often referred to as the raw form. For measures of performance, the term raw score is frequently applied. Generally, some form of assessment (i.e., judgment or interpretation) is typically required. Assessments may be applied to (or, viewed from a different perspective, may require) either a single measure of groups of them. Subjective assessments are frequently made that are based on the practitioner's familiarity with values for a given parameter in a particular context. However, due to the large number of parameters and the amount of experience that would be required to gain a sufficient level of familiarity, a more formal and objective realization of the process that takes place in subjective assessments is often employed. This process combines the measured value with objectively determined reference values to obtain new metrics, or scores, that facilitate one or more steps in the assessment process. For aspects of performance, percent normal scores are computed by expressing subject Y's availability of performance resource k[RAk(Y)] as a fraction of the mean availability of that resource in a specified reference population [RAk(pop)]. Ideally, the reference population is selected to match the characteristics of the individual as closely as possible (e.g., age range, gender, handedness, etc.).

Percent normal =

()

RA k Y

( )

RAk pop

× 100

(145.1)

Aside from the benefit of placing all measurements on a common scale, a percent normal representation of a performance capacity score can be loosely interpreted as a probability. Consider grip strength as the © 2000 by CRC Press LLC


• Height • Weight • Postures • Subjective and instrumented methods

• Dimensions • Shape • Etc. • Instrumented methods

• Dimensions • Shape • Masses • Moments of inertia • Instrumented methods

• Mechanical properties • Instrumented methods/imaging

Complex Body Systems • Cognitive • Speech • Lifting, gait • Upper extremity • Cardiovascular/respiratory • Etc.

Basic Systems • Visual information processors • Flexors, extensors • Visual sensors • Auditory sensors • Lungs • Etc.

Components of basic systems • Muscle • Tendon • Nerve • Etc.

Structure

Global/Composite • Total human • Human with artificial systems

Hierarchical Level

Function

• Generally single-function • Component-specific functions

• Single function • System-specific functions

• Multifunction, reconfigurable systems • System-specific functions

• Multifunction, reconfigurable system • High-level functions: tasks of daily life (working, grooming, recreation, etc.) • Functional assessment scales Single-number global index Level of indep. estimates

The Scope of Measurement in Rehabilitation is Broad

• Difficult to assess for individual subjects • Infer from measures at “basic system level” • Direct measurement methods with lab samples, research applications

• Subjective estimates by clinician for diagnostic and routine monitoring purposes • Instrumented measures of performance capacities (e.g., strength, extremes/range of motion, speed, accuracy, endurance, etc.)

• Function specific subjective rating scales Often based on impairment/ disability concepts Relative metrics • Some instrumented performance capacity measures Known also as “functional capacity” (misnomer)

• No single-number direct measurement is possible •Possible models to integrate lower-level measures • Direct measurement (subjective and instrumented) of selected performance attribute for selected functions

Performance

• Difficult to assess for individual subject • Direct measurement methods with lab samples, research applications

• Instrumented systems Measure and log electrophysiologic biomechanical, and other variables vs. time Post-hoc parameterization

• Subjective and automated (objective) videotape evaluation • Instrumented measures of physical quantities vs. time (e.g., forces, angles, motions) • Electromyography (e.g., muscle timing patterns, coordination)

• Subjective self- and family reports • Instrumented ambulatory activity monitors (selected attributes) • See notes under “function”

Behavior

Note: Structure, function, performance, and behavior are encompassed at multiple hierarchical levels. Both subjective and objective, instrumented methods of measurement are employed.

TABLE 145.1

performance resource. Assume that there is a uniform distribution of demands of demands placed on grip strength across a representative sample of tasks of daily living, with requirements ranging from zero to the value representing mean grip strength availability in the reference population. Further assuming that grip strength was the only performance resource that was in question for subject Y (i.e., all other were available in nonlimiting amounts), the percent normal score would represent the probability that a task involving grip strength, randomly selected from those which average individuals in the reference population could execute (i.e., those for which available grip strength would be adequate), could be successfully executed by subject Y. While the assumptions stated here are unlikely to be perfectly true, this type of interpretation helps place measurements that are most commonly made in the laboratory into daily-life contexts. In contrast to percent normal metrics, z-scores take into account variability within the selected reference population. Subject Y's performance is expressed in terms of the difference between it and the reference population mean, normalized by a value corresponding to one standard deviation unit (σ) of the reference population distribution:

z=

()

( )

RAk Y − RAk pop σ

(145.2)

It is important to note that valid z-scores assume that the parameter in question exhibit a normal distribution in the reference population. Moreover, z-scores are useful in assessing measures of structure, performance, and behavior. With regard to performance (and assuming that measures are based on a resource construct, i.e., a larger numeric value represents better performance), a z-score of zero is produced when the subject's performance equals that of the mean performance in the reference population. Positive z-scores reflect performance that is better than the population mean. In a normal distribution, 68.3% of the samples fall between z-scores of –1.0 and +1.0, wile 95.4% of these samples fall between z-scores of –2.0 and +2.0. Due to variability of a given performance capacity within a healthy population (e.g., some individuals are stronger, faster, more mobile that others), a subject with a raw performance capacity score that produces a percent normal score of 70% could easily produce a z-score of –1.0. Whereas this percent normal score might raise concern regarding the variable of interest, the z-score of –1.0 indicates that a good fraction of healthy individuals exhibit lower level of performance capacity. Both percent normal and z-scores require reference population data to compute. The best reference (i.e., most sensitive) is data for that specific individual (e.g., preinjury or predisease onset). In most cases, these data do not exist. However, practices such as preemployment screenings and regular checkups are beginning to provide individualized reference data in some rehabilitation contexts. In yet another alternative, it is frequently desirable to use values representing demands imposed by tasks [RDk(task A)] as the reference for assessment of performance capacity measures. Demands on performance resources can be envisioned to vary over the time course of a task. In practice, an estimate of the worst-case value (i.e., highest demand) would be used in assessments that incorporate task demands as reference values. In one form, such assessments can produce binary results. For example, availability can be equal to or exceed demand (resource sufficiency),or it can be less than demand (resource insufficiency). These rule-based assessments are useful in identifying limiting factors, i.e., those performance resources that inhibit a specified type of task from being performed successfully or that prevent achievement of a higher level of performance in a given type of task.

( ) ( ) R (subject Y ) is sufficient, else R (subject Y ) is insufficient

If RAk subject Y ≥ RDk task A , then Ak

Ak


(145.3)

These rule-based assessments represent the basic process often applied (sometimes subliminally) by experienced clinicians in making routine decisions, as evidenced by statements such as "not enough strength," "not enough stability," etc. It is natural to extend and build on these strategies for use with objective measures. Extreme care must be employed. It is often possible, for example, for an individual to substitute another performance resource that is not insufficient for one that is. Clinicians take into account many such factors, and objective components should be combined with subjective assessments that provide the required breadth that enhances validity of objective components of a given assessment. Using the same numeric values employed in rule-based binary assessments, a preference capacity stress metric can be computed:

( )

Performance capacity stress % =

(

RDk task A

(

)

RAk subject Y

)

× 100

(145.4)

Binary assessments also can be made using this metric and a threshold of 100%. However, the stress value provides additional information regarding how far (or close) a given performance capacity value is from the sufficiency threshold.

145.2 Measurement Objectives and Approaches Characterizing the Human System and Its Subsystems Figure 145.1 illustrates various points at which measurements are made over the course of a disease or injury, as well as some of the purposes for which they are made. The majority of measurements made in rehabilitation are aimed at characterizing the human system. Measurements of human structure [e.g., Pheasant, 1986] play a critical role in the design and prescription of components such as seating, wheelchairs, workstations, artificial limbs, etc. Just like clothing, these items must "fit" the specific individual. Basic tools such as measuring tapes and rulers are becoming supplemented with three-dimensional digitizers and devices found in computer-aided manufacturing.

FIGURE 145.1 Measurements of structure, performance, and behavior serve many different purposes at different points over the course of a disease or injury that results in the need for rehabilitation services.


Measurements of structure (e.g., limb segment lengths, moments of inertia, etc.) are also used with computer models [Vasta & Knodraske, 1995] in the process of analyzing tasks to determine demands in terms of performance capacity variables associated with basic systems such s flexors and extensors. After nearly 50 years, during which a plethora of mostly disease- and injury-specific functional assessment scales were developed, the functional independence measure (FIM) [Hamilton et al., 1987; Keith et al., 1987] is of particular note. It is the partial result of a task-force effort to produce a systematic methodology (Uniform Data System for Medical Rehabilitation) with the specific intent of achieving standardization throughout the clinical service-delivery system. This broad methodology uses subjective judgements exclusively, based on rigorous written guidelines, to categorize demographic, diagnostic, functional, and cost information for patients within rehabilitation settings. Its simplicity to use once learned and its relatively low cost of implementation have helped in gaining a rather widespread utilization for tracking progress of individuals from admission to discharge in rehabilitation programs and evaluating effectiveness of specific therapies within and across institutions. In contrast, many objective measurement tools of varying degrees of technological sophistication exist [Jones, 1995; Kondraske, 1995, Smith & Leslie, 1990; Potvin et al, 1985; Smith, 1995] (also see Further Information below. A good fraction of these have been designed to accomplish the same purposes as corresponding subjective methods, but with increased resolution, sensitivity, and repeatability. The intent is not always to replace subjective methods completely but to make available alternatives with the advantages noted for situations that demand superior performance in the aspects noted. There are certain measurement needs, however, that cannot be accomplished via subjective means (e.g., measurement of a human's visual information-processing speed, which involves the measurement of times of less than 1 second with millisecond resolution). These needs draw on the latest technology in a wide variety of ways, as demonstrated in the cited material. With regard to instrumented measurements that pertain to a specific individual, performance capacity measures at both complex body system and basic system levels (Fig. 145.1) constitute a major area of activity. A prime example is methodology associated with the implementation of industrial lifting standards [NIOSH, 1981]. Performance-capacity measures reflect the limits of availability of one or more selected resources and require test strategies in which the subject is commanded to perform at or near a maximum level under controlled conditions. Performance tests typically last only a short time (seconds or minutes). To improve estimates of capacities, multiple trials are usually included in a given "test" from which a final measure is computed according to some established reduction criterion (e.g., average across five trials, best of three trials). This strategy also tends to improve test-retest repeatability. Performance capacities associated with basic and intermediate-level systems are important because they are "targets of therapy" [Tourtellotte, 1993], i.e., the entities that patients and service providers want to increase to enhance the chance that enough will be available to accomplish the tasks of daily life. Thus measurements of baseline levels and changes during the course of a rehabilitation program provide important documentation (for medical, legal, insurance, and other purposes) as well as feedback to both the rehabilitation team and the patient. Parameters of human behavior are also frequently acquired, often to help understand an individual's response to a therapy or new circumstance (e.g., obtaining a new wheelchair or prosthetic device). Behavioral parameters reflect what the subject does normally and are typically recorded over longer time periods (e.g., hours or days) compared with that required for a performance capacity measurement under conditions that are more representative of the subject's natural habitat (i.e., less laboratory-like). The general approach involves identifying the behavior (i.e., an event such as "head flexion," "keystrokes," "steps," "repositionings," etc.) and at least one parametric attribute of it. Frequency, with units of "events per unit time," and time spent in a given behavioral or activity state [e.g., Gonapthy & Kondraske, 1990] are the most commonly employed behavioral metrics. States may be detected with electromyographic means or electronic sensors that respond to force, motion, position, or orientation. Behavioral measures can be used as feedback to a subject as a means to encourage desired behaviors or discourage undesirable behaviors.


Characterizing Tasks Task characterization or task analysis, like the organization of human system parameters, is facilitated with a hierarchical perspective. A highly objective, algorithmic approach could be delineated for task analysis in any given situation [Imrhan, 1995; Maxwell, 1995]. The basic objective is to obtain both descriptive and quantitative information for making decisions about the interface of a system (typically a human) to a given task. Specifically, function, procedures, and goals are of special interest. Function represents the purpose of a task (e.g., to flex the elbow, to lift an object, to communicate). In contrast, task goals relate to performance, or how well the function is to be excuted, and are quantifiable (e.g., the mass of an object to be lifted, the distance over which the lift must occur, the speed at which the lift must be performed, etc.). In situations with human and artificial systems, the term overall task goals is used to distinguish between goals of the combined human-artificial system and goals associated with the task of operating the artificial system. Procedures represent the process by which goals are achieved. Characterization of procedures can include descriptive and quantitative components (e.g., location of a person's hands at beginning and end points of a task, path in three-dimensional space between beginning and end points). Partial or completely unspecified procedures allow for variations in style. Goals and procedures are used to obtain numeric estimates of task demands in terms of the performance resources associated with the systems anticipated to be used to execute the task. Task demands are time dependent. Worst-case demands, which may occur only at specific instants in time, are of primary interest in task analysis. Estimates of task demand can be obtained (1) direct measurement (i.e., of goals and procedures), (2) the use of physics-based models to map direct measurements into parameters that relate more readily to measurable performance capacities of human subsystems, or (3) inference. Examples of direct measurement include key dimensions and mass of objects, three-dimensional spatial locations between "beginning" and "end points" of objects in tasks involving the movement of objects, etc. Instrumentation supporting task analysis is available (e.g., load cells, video and other systems for measuring human position and orientation in real-time during dynamic activities), but it is not often integrated into systems for task analysis per se. Direct measurements of forces based on masses of objects and gravity often must be translated (to torques about a given body joint): this requires the use of static and dynamic models and analysis [e.g., Vista & Kondraske, 1995; Winter, 1990]. An example of an inferential task-analysis approach that is relatively new is nonlinear causal resource analysis (NCRA)[Kondraske, 1999; Kondraske et al., 1997; Kondraske, 1988]. This method was motivated by human performance analysis situations where direct analysis is not possible (e.g., determination of the amount of visual information-processing speed required to drive safely on a highway). Quantitative task demands, in terms of performance variables that characterize the involved subsystems, are inferred from a population data set that includes measures of subsystem performance, resource availabilities (e.g., speed, accuracy, etc.), and overall performance on the task in question. This method is based on the simple observation that the individual with the least amount of the given resource (i.e., the lowest performance capacity) who is still able to accomplish a given goal (i.e., achieve a given level of performance in the specified high-level task) provides the key clue. That amount of availability is used to infer the amount of demand imposed by the task. The ultimate goal to which task characterization contributes is to identify limiting factors or unsafe conditions when a specific subject undertakes the task in question; this goal must not be lost while carrying out the basic objectives of task analysis. While rigorous algorithmic approaches are useful to make evident the true detail of the process, they are generally not performed in this manner in practice at present. Rather, the skill and experience of individuals performing the analysis are used to simplify the process, resulting in a judicious mixture of subjectives estimates and objective measurements. For example, some limiting factors (e.g., grip strength) may be immediately identified without measurement of the human or the task requirements because the margin between availability and demand is so great that quick subjective "measurements" followed by an equally quick "assessment" can be used to arrive at the proper conclusion (e.g., "grip strength is a limiting factor in this task").


Characterizing Assistive Devices Assistive devices can be viewed as artificial systems that either completely or partially bridge a gap between a given human (with his or her unique profile of performance capacities, i.e., available performance resources) and a particular task or class of tasks (e.g., communication, mobility, etc.). It is thus possible to consider the aspects of the device that constitute the user-device interface and those aspects which constitute, more generally, the device-task interface. In general, measurements supporting assessment of the user-device interface can be viewed to consist of (1) those which characterize the human and (2) those which characterize tasks (i.e., "operating" the assistive device). Each of these was described earlier. Measurements that characterize the device-task interface are often carried out in the context of the complete system, i.e., the human-assistive device-task combination (see next subsection).

Characterizing Overall Systems in High-Level-Task Situations This situation generally applies to a human-artificial system-task combination. Examples include an individual using a communication aid to communicate, an individual using a wheelchair to achieve mobility, etc. Here, concern is aimed at documenting how well the task (e.g., communication, mobility, etc.) is achieved by the composite or overall system. Specific aspects or dimensions of performance associated with the relevant function should first be identified. Examples include speech, accuracy, stability, efficiency, etc. The total system is then maximally challenged (tempered by safety considerations) to operate along one or more of these dimensions of performance (usually not more than two dimensions are maximally challenged at the same time). For example, a subject with a communication device may be challenged to generate a single selected symbol "as fast as possible" (stressing speed without concern for accuracy). Speed is measured (e.g., with units of symbols per second) over the course of short trial (so as not to be influenced by fatigue). Then the "total system" may be challenged to generate a subset of specific symbols (chosen at random from the set of those available with a given device) one at a time, "as accurately as possible" (stressing accuracy while minimizing stress on speed capacities). Accuracy is then measured after a representative number of such trials are administered (in terms of "percent correct," for example). To further delineate the speed-accuracy performance envelope, "the system" may be challenge to select symbols at a fixed rate while accuracy is measured. Additional dimensions can be evaluated similarly. For example, endurance (measure in units of time) can be determined by selecting an operating point (e.g., by reference to the speed-accuracy performance envelope) and challenging the total system "to communicate" for "as long as possible" under the selected speed-accuracy condition. In general, it is more useful if these types of characterizations consider all relevant dimensions with some level of measurements (i.e., subjective or objective) than it would be to apply a high resolution, objective measurement in a process that considers only one aspect of performance.

145.3 Decision-Making Processes Measurements that characterize the human, task, assistive device, or combination thereof are themselves only means to an end; the end is typically a decision. As noted previously, decisions are often the result of assessment processes involving one or more measurements. Although not exhaustive, many of the different types of assessments encountered are related to the following questions: (1) Is a particular aspect of performance normal (or impaired)? (2) Is a particular aspect of performance improving, stable, or getting worse? How should therapy be modified? (3) Can a given subject utilize (and benefit from) a particular assistive device? (4) Does a subject possess the required capacity to accomplish a given higher level task (e.g., driving, a particular job after a work-related injury, etc.)? In Fig. 145.2, several of the basic concepts associated with measurement are used to illustrate how they enter into and facilitate systematic decision-making processes. The upper section shows raw score values as well as statistics for a healthy normal reference population in tabular form (left). It is difficult to reach any decision by simple inspection of just the raw performance capacity values. Tabular data are used to


obtain percent normal (middle) and z-score (right) assessments. Both provide a more directly interpretable result regarding subject A's impairments. By examining the "right shoulder flexion extreme of motion" item in the figure, it can be seen that a raw score value corresponding to 51.2% normal yields a very large-magnitude, negative z-score (–10.4). This z-score indicates that virtually no one in the reference population would have a score this low. In contrast, consider similar scores for the "grip strength" item (56.2% normal, z-score = –1.99). On the basis of percent normal scores, it would appear that both of these resources are similarly affected, whereas the z-score basis provides a considerably different perspective due to the fact that grip strength is much more variable in healthy populations than the extreme angle obtained by a given limb segment about a joint, relatively speaking. As noted, z-scores account for this variability. The lower section of Fig. 145.2 considers a situation in which the issue is a specific individual (subject A) considered in a specific task. Tabular data now include raw score values (which are the same as in upper section of the figure) and quantitative demands (typically worst case) imposed on the respective performance resources by task X. The lower-middle plot illustrates the process of individually assessing sufficiency of each performance resource in this task context using a rule-based assessment that incorporates the idea of a threshold (i.e., availability must exceed demand for sufficiency). The lower-right plot illustrates an analogous assessment process that is executed after computation of a stress metric for each of the performance capacities. Here, any demand that corresponds to more than a 100% stress level is obviously problematic. In addition to binary conclusions regarding whether a given capacity is or is not a limiting factor, it is possible to observe that of the two limiting resources (e.g., grip strength and right shoulder flexion extreme of motion), the former is more substantial. This might suggest, for example, that the task be modified so as to decrease the grip-strength demand (i.e., gains in performance capacity required would be substantial to achieve sufficiency) and the use of focused exercise therapy to increase shoulder flexion mobility (i.e., gains in mobility required are relatively small).

145.4 Current Limitations Quality of Measurements Key issues are measurement validity, reliability (or repeatability), accuracy, and discriminating power. At issue in terms of current limitations is not necessarily the quality of measurements but limitations with regard to methods employed to determine the quality of measurements and their interpretability. A complete treatment of these complex topics is beyond the present scope. However, it can be said that standards are such [Potvin et al., 1985] that most published works regarding measurement instruments do address quality of measurements to some extent. Validity (i.e., how well does the measurement reflect the intended quantity) and reliability are most often addressed. However, one could easily be left with the impression that these are binary conditions (i.e., measurement is or is not reliable or valid), when in fact a continuum is required to represent these constructs. Of all attributes that relate to measurement quality, reliability is most commonly expressed in quantitative terms. This is perhaps because statistical methods have been defined and promulgated for the computation of so-called reliability coefficients [Winer, 1971]. Reliability coefficients range from 0.0 to 1.0, and the implication is that 1.0 indicates a perfectly reliable or repeatable measurement process. Current methods are adequate, at best, for making inferences regarding the relative quality of two or more methods of quantifying "the same thing." Even these comparisons require great care. For example, measurement instruments that have greater intrinsic resolving power have a great opportunity to yield smaller-reliability coefficients simply because they are capable of measuring the true variability (on repeated measurement) of the parameter in question within the system under test. While there has been widespread determination or reliability coefficients, there has been little or no effort directed toward determination of what value of a reliability coefficient is "good enough" for a particular application. In fact, reliability coefficients are relatively abstract to most practitioners.



FIGURE 145.2 Examples of different types of assessments that can be performed by combining performance capacity measures and reference values of different types. The upper section shows raw score values as well as statistics for a healthy normal reference population in tabular form (left). It is difficult to reach any decision by simple inspection of just the raw performance capacity values. Tabular data are used to obtain a percent normal assessment (middle) and a z-score assessment (right). Both of these provide a more directly interpretable result regarding subject A’s impairments. The lower section shows raw score values (same as in upper section) and quantitative demands (typically worst case) imposed on the respective performance resources by task X. The lower-middle plot illustrates the process of individually assessing sufficiency of each performance resource in this task context using a threshold rule (i.e., availability must exceed demand for sufficiency). The lower-right plot illustrates a similar assessment process after computation of a stress metric for each of the performance capacities. Here, any demand that corresponds to more than a 100% stress level is obviously problematic.

Methods for determining the quality of a measurement process (including the instrument, procedures, examiner, and actual noise present in the variable of interest) that allow a practitioner to easily reach decisions regarding the use of a particular measurement instrument in a specific application and limitations thereof are currently lacking. At the use of different measurements increases and the number of options available for obtaining a given measurement grows, this topic will undoubtedly receive additional attention. Caution in interpreting literature, common sense, and the use of simple concepts such as "I need to measure range of motion to within 2 degrees in my application" are recommended in the meantime [Mayer et al., 1997]. Standards Measurements, and concepts with which they are associated, can contribute to a shift from experiencebased knowledge acquisition to rule-based, engineering-like methods. This requires (1) a widely accepted conceptual framework (i.e. known to assistive device manufacturers, rehabilitation engineers, and other professionals within the rehabilitation community), (2) a more complete set of measurement tools that are at least standardized with regard to the definition of the quantity measured, (3) special analysis and assessment software (that removes the resistance to the application of more rigorous methods by enhancing the quality of decisions as well as the speed with which they can be reached), and (4) properly trained practitioners. Each is a necessary but not sufficient component. Thus balanced progress is required in each of these areas.

Rehabilitation Service Delivery and Rehabilitation Engineering In a broad sense, it has been argued that all engineers can be considered rehabilitation engineers who merely work at different levels along a comprehensive spectrum of human performance, which itself can represent a common denominator among all humans. Thus an automobile is a mobility aid, a telephone is a communication aid, and so on. Just as in other engineering disciplines, measurement must be recognized not only as an important end in itself (in appropriate instances) but also as an integral component or means within the overall scope of rehabilitation and rehabilitation engineering processes. The service-delivery infrastructure must provide for such means. At present, one should anticipate and be prepared to overcome potential limitations associated with factors such as third-party reimbursement for measurement procedures, recognition of equipment and maintenance costs associated with obtaining engineering-quality measurements, and education of administrative staff and practitioners with regard to the value and proper use of measurements.

Defining Terms Behavior: A general term that relates to what a human or artificial system does while carrying out its function(s) under given conditions. Often, behavior is characterized by measurement of selected parameters or identification of unique system states over time. Function: The purpose of a system. Some systems map to a single primary function (e.g., process visual information). Others (e.g., the human arm) map to multiple functions, although at any given time multifunction systems are likely to be executing a single function (e.g., polishing a car). Functions can be described and inventoried, whereas level of performance of a given function can be measured. Functional assessment: The process of determining, from a relatively global perspective, an individual's ability to carry out tasks in daily life. Also, the result of such a process. Functional assessments typically cover a range of selected activity areas and include (at a minimum) a relatively gross indication (e.g., can or can't do; with or without assistance) of status in each area. Goal: A desired endpoint (i.e., result) typically characterized by multiple parameters, at least one of which is specified. Examples include specified task goals (e.g., move an object of specified mass from point A to point B in 3 seconds) or estimated task performance (maximum mass, range, speed of movement obtainable given a specified elemental performance resource availability pro-


file), depending on whether a reverse or forward analysis problem is undertaken. Whereas function describes the general process of task, the goal directly relates to performance and is quantitative. Limiting resource: A performance resource at any hierarchical level (e.g., vertical lift strength, knee flexor speed) that is available in an amount that is less than the worst-case demand imposed by a task. Thus a given resource can be "limiting" only when considered in the context of a specific task. Overall task goals: Goals associated with a task to be executed by a human-artificial system combination (to be distinguished from goals associated with the task of operating the artificial system). Performance: Unique qualities of a human or artificial system (e.g., strength, speed, accuracy, endurance) that pertain to how well that system executes its function. Performance capacity: A quantity in finite availability that is possessed by a system or subsystem, drawn on during tasks, and limits some aspect (e.g., speed, force, production, etc.) of a system's ability to execute tasks, or, the limit of that aspect itself. Performance capacity measurement: A general class of measurements, performed at different hierarchical levels, intended to quantify one or more performance capacities. Procedure: A set of constraints placed on a system in which flexibility exists regarding how a goal (or set of goals) associated with a given function can be achieved. Procedure specification requires specification of initial intermediate, and/or final states or conditions dictating how the goal is to be accomplished. Such specification can be thought of in terms of removing some degrees of freedom. Structure: Physical manifestation and attributes of a human or artificial system and the object of one type of measurements at multiple hierarchical levels. Style: Allowance for variation within a procedure, resulting in the intentional incomplete specification of a procedure or resulting from either international or unintentional incomplete specification of procedure. Task: That which results from (1) the combination of specified functions, goals, and procedures or (2) the specification of function and goals and the observation of procedures utilized to achieve the goals.

References Fuhrer MJ. 1987. Rehabilitation Outcomes: Analysis and Measurement. Baltimore, Brookes. Ganapathy G, Kondraske GV. 1990. Microprocessor-based instrumentation for ambulatory behavior monitoring. J Clin Eng 15(6):459. Granger CV, Greshorn GE. 1984. Functional Assessment in Rehabilitation Medicine. Baltimore, Williams & Wilkins. Hamilton BB, Granger CV, Sherwin FS, et al. 1987. A uniform national data system for medical rehabilitation. In MJ Fuhrer (ed), Rehabilitation Outcomes: Analysis and Measurement, pp 137–147. Baltimore, Brookes. Imrhan S. 2000. Task analysis and decomposition: Physical components. In JD Bronzino (ed), Handbook of Biomedical Engineering, 2nd ed., Boca Raton, Fla, CRC Press. Jones RD. 1995. Measurement of neuromotor control performance capacities. In JD Bronzino (ed), Handbook of Biomedical Engineering. Boca Raton, Fla, CRC Press. Keith RA, Granger CV, Hamilton BB, Sherwin FS. 1987. The functional independence measure: A new tool for rehabilitation. In MG Eisenberg, RC Grzesiak (eds), Advances in Clinical Rehabilitation, vol 1, pp 6–18. New York, Springer-Verlag. Kondraske GV. 1988. Experimental evaluation of an elemental resource model for human performance. In Proceedings of the Tenth Annual IEEE Engineering in Medicine and Biology Society Conference, New Orleans, pp 1612–13. Kondraske GV. 1990. Quantitative measurement and assessment of performance. In RV Smith, JH Leslie (eds), Rehabilitation Engineering, pp 101–125. Boca Raton, Fla, CRC Press. Kondraske GV. 2000. A working model for human system-task interfaces. In JD Bronzino (ed), Handbook of Biomedical Engineering, 2nd ed., Boca Raton, Fla, CRC Press.


Kondraske GV, Johnston C, Pearson, A & Tarbox L. 1997. Performance Prediction and limiting resource identification with nonlinear causal resource analysis. Proceedings, 19th Annual Engineering in Medicine and Biology Society Conference, pp 1813–1816. Kondraske GV, Vasta PJ. 2000. Measurement of information processing performance capacities. In JD Bronzino (ed), Handbook of Biomedical Engineering, 2nd ed., Boca Raton, Fla, CRC Press. Maxwell KJ. 2000. High-level task analysis: Mental components. In JD Bronzino (ed), Handbook of Biomedical Engineering, 2nd ed., Boca Raton, Fla, CRC Press. Mayer T, Kondraske GV, Brady Beals, S., & Gatchel RJ. 1997. Spinal range of motion: accuracy and sources of error with inclinometric measurement. Spine, 22(17), 1976–1984. National Institute of Occupational Safety and Health (NIOSH). 1981. Work Practices Guide for Manual Lifting (DHHS Publication No. 81122). Washington, US Government Printing Office. Pheasant ST. 1986. Bodyspace: Anthropometry, Ergonomics and Design. Philadelphia, Taylor & Francis. Potvin AR, Tourtellotte WW, Potvin JH, et al. 1985. The Quantitative Examination of Neurologic Function. Boca Raton, Fla, CRC Press. Smith RV, Leslie JH. 1990. Rehabilitation Engineering. Boca Raton, Fla, CRC Press. Smith SS. 2000. Measurement of neuromuscular performance capacities. In JD Bronzino (ed), Handbook of Biomedical Engineering, 2nd ed., Boca Raton, Fla, CRC Press. Tourtellotte WW. 1993. Personal communication. Vasta PJ, Kondraske GV. 1994. Performance prediction of an upper extremity reciprocal task using nonlinear causal resource analysis. In Proceedings of the Sixteenth Annual IEEE Engineering in Medicine and Biology Society Conference, Baltimore. Vasta PJ, Kondraske GV. 2000. Human performance engineering: Computer based design and analysis tools. In JD Bronzino (ed), Handbook of Biomedical Engineering, 2nd ed., Boca Raton, Fla, CRC Press. Webster JG, Cook AM, Tompkin WJ, Vanderheiden GC. 1985. Electronic Devices for Rehabilitation. New York, Wiley. Winer BJ. 1971. Statistical Principles in Experimental design, 2d ed. New York, McGraw-Hill. Winter DA. 1990. Biomechanics and Motor Control of Human Movement, 2d ed. New York, Wiley. World Health Organization. 1980. International Classification of Impairments, Disabilities, and Handicaps. Geneva, World Health Organization.

Further Information The section of this Handbook entitled "Human Performance Engineering" contains chapters that address human performance modeling and measurement in considerably more detail. Manufacturers of instruments used to characterize different aspects of human performance often provide technical literature and bibliographies with conceptual backgrounds, technical specifications and application examples. A partial list of such sources is included below. (No endorsement of products is implied.) Baltimore Therapeutic Equipment Co. 7455-L New Ridge Road Hanover, MD 21076-3105 http://www.bteco.com/ Chattanooga Group 4717 Adams Road Hixson, TN 37343 http://www.chattanoogagroup.com/ Henley Healthcare 120 Industrial Blvd. Sugarland, TX 77478 http://www.henleyhealth.com/


Human Performance Measurement, Inc. P.O. Box 1996 Arlington, TX 76004–1996 http://www.flash.net/~hpm/ Lafayette Instrument 3700 Sagamore Parkway North Lafayette, IN 47904–5729 http://www.lafayetteinstrument.com The National Institute on Disability and Rehabilitation Research (NIDRR), part of the Department of Education, funds a set of Rehabilitation Engineering Research Centers (RERCs) and Research and Training Centers (RTCs). Each has a particular technical focus; most include measurements and measurements issues. Contact NIDRR for a current listing of these centers. Measurement devices, issues, and application examples specific to rehabilitation are included in the following journals: IEEE Transactions on Rehabilitation Engineering IEEE Service Center 445 Hoes Lane P.O. Box 1331 Piscataway, N.J. 08855–1331 http://www.ieee.org/index.html Journal of Rehabilitation Research and Development Scientific and Technical Publications Section Rehabilitation Research and Development Service 103 South Gay St., 5th floor Baltimore, MD 21202–4051 http://www.vard.org/jour/jourindx.htm Archives of Physical Medicine and Rehabilitation Suite 1310 78 East Adams Street Chicago, IL 60603–6103 American Journal of Occupational Therapy The American Occupational Therapy Association, Inc. 4720 Montgomery Ln., Bethesda, MD 20814–3425 http://www.aota.org/ Physical Therapy American Physical Therapy Association 1111 North Fairfax St. Alexandria, VA 22314 http://www.apta.org/ Journal of Occupational Rehabilitation Subscription Department Plenum Publishing Corporation 233 Spring St. New York, NY 10013 http://www.plenum.com/


Hobson, D., Trefler, E. “Rehabilitation Engineering Technologies: Principles of Application.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

146 Rehabilitation Engineering Technologies: Principles of Application 146.1 146.2

The Conceptual Frameworks The Provision Process The Shifting Paradigm • The Evaluation • Service Delivery Models

146.3

Education and Quality Assurance RESNA

146.4 University of Pittsburgh

Elaine Trefler University of Pittsburgh

Specific Impairments and Related Technologies Mobility • Sitting • Sensation • Access (Person-Machine Interface) • Communication • Transportation • Activities of Daily Living (ADL) • School and Work • Recreation • Community and Workplace Access

Douglas Hobson

146.5

Future Developments

Rehabilitation engineering is the branch of biomedical engineering that is concerned with the application of science and technology to improve the quality of life of individuals with disabilities. Areas addressed within rehabilitation engineering include wheelchairs and seating systems, access to computers, sensory aids, prosthetics and orthotics, alternative and augmentative communication, home and work-site modifications, and universal design. Because many products of rehabilitation engineering require careful selection to match individual needs and often require custom fitting, rehabilitation engineers have necessarily become involved in service delivery and application as well as research, design, and development. Therefore, as we expand on later, it is not only engineers that practice within the field of rehabilitation engineering. As suggested above, and as in many other disciplines, there are really two career tracks in the field of rehabilitation engineering. There are those who acquire qualifications and experience to advance the state of knowledge through conducting research, education, and product development, and there are others who are engaged in the application of technology as members of service delivery teams. At one time it was possible for a person to work in both arenas. However, with the explosion of technology and the growth of the field over the past decade, one must now specialize not only within research or service delivery but often within a specific area of technology. One can further differentiate between rehabilitation and assistive technology. Rehabilitation technology is a term most often used to refer to technologies associated with the acute-care rehabilitation process. Therapy evaluation and treatment tools, clinical dysfunction measurement and recording instrumentation,


and prosthetic and orthotic appliances are such examples. Assistive technologies are those devices and services that are used in the daily lives of people in the community to enhance their ability to function independently, examples being specialized seating, wheelchairs, environmental control devices, workstation access technologies and services are now communication aids. Recognition and support of assistive technology devices and services are now embedded in all the major disability legislation that has been enacted over the last decade. The primary focus of this chapter is on the role of the rehabilitation engineering practitioner as he or she carries out the responsibilities demanded by the application of assistive technology. Before launching into the primary focus of this chapter, let us first set a conceptual framework for the raison d'etre for assistive technology and the role of the assistive technology professional.

146.1 The Conceptual Frameworks The application of assistive technology can be conceptualized as minimizing the functional gap between the person and his or her environment. This reality is what technology does for all of us to varying degrees. For example, if you live in a suburban area that has been designed for access only by car and your car breaks down, you are handicapped. If your house has been designed to be cooled by air conditioning in the "dog days" of summer and you lose a compressor, your comfort is immediately compromised by your incompatibility with your environment. Similarly, if you live in a home that has only access by steps and you have an impairment requiring the use of a wheelchair, you are handicapped because you no longer have abilities that are compatible with your built environment. Because our environments, homes, workplaces, schools, and communities have been designed to be compatible with the abilities of the norm, young children, persons with disabilities, and many elderly people experience the consequences of their mismatch as a matter of course. The long-term utopian solution would be to design environments and their contents so that they can be used by all people of all ages, which is the essence of the universal design concept. However, given that today we do not have very many products and environments that have been universally designed, rehabilitation engineers attempt to minimize the effects of the mismatch by designing, developing, and providing technologies that will allow persons with disabilities to pursue their life goals in a manner similar to any other person. Of course, the rehabilitation engineer cannot accomplish this working in isolation but rather must function as a part of a consumerresponsive team that can best deal with the multiplicity of factors that usually impact on the successful application of assistive technology. Let us now move to another conceptual framework, one that conceptualizes how people actually interact with technology. The following conceptualization has been adapted from the model proposed by Roger Smith [Smith, 1992]. In Fig. 146.1, Smith suggests that there are three cyclic elements that come into play when humans interact with technology: the human and his or her innate sensory, cognitive, and functional abilities; the human factor's characteristics of the interface between the human and the technology; and the technical characteristics of the technology itself in terms of its output as a result of a specific input by the user. People with disabilities may have varying degrees of dysfunction in their sensory, cognitive, and functional abilities. The interface will have to be selected or adapted to these varying abilities in order to allow the person to effectively interact with the technology. The technology itself will need to possess specific electronic or mechanical capabilities in order to yield the desired outcome. The essence of assistive technology applications is to integrate all three of these element into a functional outcome that meets the specific needs of a user. This is usually done by selecting commercially available devices and technologies at a cost that can be met by either the individual or his or her third-party payment source. When technologies are not available, then they must be modified from existing devices or designed and fabricated as unique custom solutions. It is particularly in these latter activities that a rehabilitation engineer can make his or her unique contribution to team process. In 1995, Cook and Hussey [Cook and Hussey, 1995], published an excellent text, Assistive Technologies—Principles and Practice. As well as comprehensively addressing many of the assistive technologies © 2000 by CRC Press LLC

FIGURE 146.1

Conceptual framework of technology and disability. (Modified from Smith [1992].)

briefly covered in this chapter, they also present a conceptual framework which builds on the one developed by Smith above. They introduce the additional concepts of activity and context. That is, understanding of a person’s activity desires and the context (social, setting, physical) in which they are to be carried out are essential components to successful assistive technology intervention. It should be realized that there are several levels of assistive technology. The first level might be termed fundamental technology in contrast to advanced technology. Fundamental technologies, such as walkers, crutches, many wheelchairs, activities of daily living (ADL) equipment, etc., usually do not require the involvement of the rehabilitation engineer in their application. Others on the team can better assess the need, confirm the interface compatibility, and verify that the outcome is appropriate. The rehabilitation engineer is most often involved in the application of advanced technologies, such as powered wheelchairs, computerized workstation designs, etc., that require an understanding of the underlying technological principles in order to achieve the best match with the abilities and needs of the user, especially if custom modifications or integration of devices are required to the original equipment. The rehabilitation engineer is usually the key person if a unique solution is necessary. Let's now discuss a few fundamental concepts related to the process by which assistive technology is typically provided in various service delivery programs.

146.2 The Provision Process The Shifting Paradigm In the traditional rehabilitation model of service delivery, a multidisciplinary team of professionals is already in place. Physicians, therapists, counselors, and social worker meet with the client and, based on © 2000 by CRC Press LLC

the findings of a comprehensive evaluation, plan a course of action. In the field of assistive technology, the rules of the team are being charted anew. First, the decision making often takes place in a nonmedical environment and often without a physician as part of the team. Second, the final decision is rapidly moving into the hands of the consumer, not the professionals. The third major change is the addition of a rehabilitation engineer to the team. Traditional team members have experience working in groups and delegating coordination and decision making to colleagues depending on the particular situation. They are trained to be team players and are comfortable working in groups. Most engineers who enter the field of rehabilitation engineering come with a traditional engineering background. Although well versed in design and engineering principles, they often do not receive formal training in group dynamics and need to learn these skills if they are to function effectively. As well, engineers are trained to solve problems with technical solutions. The psychosocial aspects of assisting people with disabilities to make informed choices must be learned most often outside the traditional education stream. Therefore, for the engineer to be a contributing member of the team, not only must he or she bring engineering expertise, but it must be integrated in such a manner that it supports the overall objectives of the technology delivery process, which is to respond to the needs and desires of the consumer. People with disabilities want to have control over the process and be informed enough to make good decisions. This is quite different from the traditional medical or rehabilitation model, in which wellmeaning professionals often tell the individual what is best for him or her. Within this new paradigm, the role is to inform, advise, and educate, not to decide. The professional provides information as to the technical options, prices, etc. and then assists the person who will use the technology to acquire it and learn how to use it.

The Evaluation An evaluation is meant to guide decision-making for the person with a disability toward appropriate and cost-effective technology. Often, more than one functional need exists for which assistive technology could be prescribed. Costly, frustrating, and time-consuming mistakes often can be avoided if a thorough evaluation based on a person's total functional needs is performed before any technology is recommended. Following the evaluation, a long-range plan for acquisition and training in the chosen technology can be started. For example, suppose a person needs a seating system, both a powered and manual wheelchair, an augmentative communication device, a computer workstation, and an environmental control unit (ECU). Where does one begin? Once a person's goals and priorities have been established, the process can begin. First, a decision would likely be made about the seating system that will provide appropriate support in the selected manual chair. However, the specifications of seating system should be such that the components also can be interfaced into the powered chair. The controls for the computer and augmentative communication device must be located so that they do not interfere with the display of the communication device and must in some way be compatible wit the controls for the ECU. Only if all functional needs are addressed can the technology be acquired in a logical sequence and in such a manner that all components will be compatible. The more severely disabled the individual, the more technology he or she will need, and the more essential is the process of setting priorities and ensuring compatibility of technical components. In summary, as suggested by the conceptual model, the process begins by gaining an understanding of the person's sensory, cognitive, and functional abilities, combined with clarification of his or her desires and needs. These are then filtered through the technology options, both in terms of how the interface will integrate with the abilities of the user and how the technology itself will be integrated to meet the defined needs. This information and the associated pros and cons are conveyed to the user, or in some cases their caregiver, who then has the means to participate in the ultimate selection decisions.

Service Delivery Models People with disabilities can access technology through a variety of different service delivery models. A team of professionals might be available in a university setting where faculty not only teach but also deliver technical services to the community. More traditionally, the team of rehabilitation professionals, © 2000 by CRC Press LLC

including a rehabilitation engineer, might be available at a hospital or rehabilitation facility. More recently, technology professionals might be in private practice either individually, as a team, or part of the university, hospital, or rehabilitation facility structure. Another option is the growing number of rehabilitation technology suppliers (RTSs) who offer commercial technology services within the community. They work in conjunction with an evaluation specialist and advise consumers as to the technical options available to meet their needs. They then sell and service the technology and train the consumer in its use. Local chapters of national disability organizations such as United Cerebral Palsy and Easter Seals also may have assistive technology services. In recent years, a growing number of centers for independent living (CILs) have been developed in each state with federal support. Some of these centers have opted to provide assistive technology services, in addition to their information and referral services, which are common to all CILs. And finally, there are volunteers, either in engineering schools or community colleges (student supervised projects) or in industry (high-technology industries often have staff interested in doing community service), such as the Telephone Pioneers. Each model has its pros and cons for the consumer, and only after thoroughly researching the options will the person needing the service make the best choice as to where to go with his or her need in the community. A word of caution. Only if there is timely provision and follow-up available is a service delivery system considered appropriate, even if the cost of the service is less. A more extensive description of service delivery options may be reviewed in a report that resulted from a RESNA-organized conference on service delivery [ANSI/RESNA, 1987].

146.3 Education and Quality Assurance Professionals on the assistive technology team have a primary degree and credential in their individual professions. For example, the occupational or physical therapist will have a degree and most often state licensure in occupational or physical therapy. The engineer will have recognized degrees in mechanical, electrical, biomedical, or some other school of engineering. However, in order to practice effectively in the field of assistive technology, almost all will need advanced training. A number of occupational therapy curriculums provide training in assistive technology, but not all. The same is true of several of the others. Consumers and payers of assistive technology need to know that professionals practicing in the field of assistive technology have a certain level of competency. For this reason, all professionals, including rehabilitation engineers, are pursuing the ATP (assistive technology practitioner) credential through RESNA.

RESNA RESNA, an interdisciplinary association of persons dedicated to the advancement of assistive technology for people with disabilities, has a credentialing program that credentials individuals on the assistive technology team. As part of the process, the minimum skills and knowledge base for practitioners is tested. Ties with professional organizations are being sought so that preservice programs will include at least some of the knowledge and skills base necessary. Continuing education efforts by RESNA and others also will assist in building the level of expertise of practitioners and consumers. At this time RESNA has a voluntary credentialling process to determine if a person meets a predetermined minimal standard of practice in the field of Assistive Technology. Persons who meet the prerequisite requirements, pass a written exam, and agree to abide by the RESNA Standards of Practice can declare themselves as RESNA certified. They can add the ATP if they are practitioners or ATS if they are suppliers of assistive technology. Payment for technology and the services required for its application is complex and changing rapidly as health care reform evolves. It is beyond the scope of this discussion to detail the very convoluted and individual process required to ensure that people with disabilities receive what they need. However, there are some basic concepts to be kept in mind. Professionals need to be competent. The documentation of need and the justification of selection must be comprehensive. Time for a person to do this must be allocated if there is to be success. Persistence, creativity, education of the payers, and documentation of need and outcomes are the key issues.


146.4 Specific Impairments and Related Technologies Current information related to specific technologies is best found in brochures, trade magazines (Report Rehab), exhibit halls of technology-related conferences, and databases such as ABLEDATA. Many suppliers and manufacturers are now maintaining Websites, which provides a quick means to locate information on current products. What follows is only a brief introduction to specific disabilities areas to which assistive technology applications are commonly used.

Mobility Mobility technologies include wheelchairs, walkers, canes, orthotic devices, FES (functional electrical stimulation), laser canes, and any other assistive device that would assist a person with a mobility impairment, be it motor or sensory, to move about in his or her environment. There are very few people who have a working knowledge of all the possible commercial options. Therefore, people usually acquire expertise in certain areas, such as wheelchairs. There are hundreds of varieties of wheelchairs, each offering a different array of characteristics that need to be understood as part of the selection process. Fortunately, there are now several published ways that the practitioner and the consumer can obtain useful information. A classification system has been developed that sets a conceptual framework for understanding the different types of wheelchairs that are produced commercially [Hobson, 1990]. Paraplegic News and Sports and Spokes annually publish the specifications on most of the manual and powered wheelchairs commonly found in the North American marketplace. These reviews are based on standardized testing that is carried out by manufacturers following the ANSI/RESNA wheelchair standards [ANSI/RESNA, 1990]. Since the testing and measurements of wheelchairs are now done and reported in a standard way, it is possible to make accurate comparisons between products, a tremendous recent advancement for the wheelchair specialist and the users they serve [Axelson et al., 1994]. Possibly the most significant advancement in wheelchairs is the development and application of industry, on an international scale, for testing the safety and durability of their products. These standards also mandate what and how the test information should be made available in the manufacturer’s presale literature. The Rehabilitation Engineering Research Center and University of Pittsburgh [RERC, 1999] maintains a large Website, where among its many resources is a listing of wheelchair research publications and a general reference site, termed Wheelchairnet [Wheelchairnet, 1999]. The RERC site also tracks the current activities occurring in many of the wheelchair standards working groups. Finally, Cooper [1995, 1998] has published two excellent reference texts on rehabilitation engineering with emphasis on wheeled mobility.

Sitting Many people cannot use the wheelchairs as they come from the manufacturer. Specialized seating is required to help persons to remain in a comfortable and functional seated posture for activities that enable them to access work and attend educational and recreational activities. Orthotic supports, seating systems in wheelchairs, chairs that promote dynamic posture in the workplace, and chairs for the elderly that fit properly, are safe, and encourage movement all fit into the broad category of sitting technology.

Sensation People with no sensation are prone to skin injury. Special seating technology can assist in the preventions of tissue breakdown. Specially designed cushions and backs for wheelchairs and mattresses that have pressure-distributing characteristics fall into this category. Technology also has been developed to measure the interface pressure. These tools are now used routinely to measure and record an individual’s pressure profile, making cushion selection and problem solving more of a science than an art. Again, a classification system of specialized seating has been developed that provides a conceptual framework for understanding the features of the various technologies and their potential applications. The same reference also discusses the selection process, evaluation tools, biomechanics of supported sitting, and materials properties of weight-relieving materials [Hobson, 1990]. © 2000 by CRC Press LLC

Access (Person-Machine Interface) In order to use assistive technology, people with disabilities need to be able to operate the technology. With limitations in motor and/or sensory systems, often a specially designed or configured interface system must be assembled. It could be as simple as several switches or a miniaturized keyboard or as complex as an integrated control system that allows a person to drive a wheelchair and operate a computer and a communication device using only one switch.

Communication Because of motor or sensory limitations, some individuals cannot communicate with spoken or written word. There are communication systems that enable people to communicate using synthesized voice or printed output. Systems for people who are deaf allow them to communicate over the phone or through computer interfaces. Laptop computers with appropriate software can enable persons to communicate faster and with less effort than previously possible. Some basic guidelines for selecting an augmentative communication system, including strategies for securing funding, have been proposed in an overview chapter by James Jones and Winifred Jones [Jones & Jones, 1990].

Transportation Modified vans and cars enable persons with disabilities to independently drive a vehicle. Wheelchair tiedowns and occupant restraints in personal vehicles and in public transportation vehicles are allowing people to be safely transported to their chosen destination. Fortunately, voluntary performance standards for restraint and tie-down technologies have been developed by a task group within the Society for Automotive Engineers (SAE). Standards for car hand controls, van body modifications, and wheelchair lifts are also available from SAE. These standards provide the rehabilitation engineer with a set of tools that can be used to confirm safety compliance of modified transportation equipment. Currently in process and still requiring several more years of work are transport wheelchair and vehicle power control standards.

Activities of Daily Living (ADL) ADL technology enables a person to live independently as much as possible. Such devices as environmental control units, bathroom aids, dressing assists, automatic door openers, and alarms are all considered aids to daily living. Many are inexpensive and can be purchased through careful selection in stores or through catalogues. Others are quite expensive and must be ordered through vendors who specialize in technology for independent living. Ron Mace, now deceased and creator of the Center for Universal Design at the North Carolina State University, is widely acknowledged as the father of the Universal Design concept. The concept of universal design simply means that if our everyday built environments and their contained products could be designed to meet the needs of a wider range of people, both young and old, then the needs of more persons with disabilities would be met without the need for special adaptions [Center for Universal Design, 1999]. Others like Paul Grayson have also published extensively regarding the need to re-think how we design our living environments [Grayson, 1991]. Vanderheiden and Denno have prepared human factors guidelines that provide design information to allow improved access by the elderly and persons with disabilities [Denno et al., 1992; Vanderheiden & Vanderheiden, 1991; Trace Center, 1999].

School and Work Technology that supports people in the workplace or in an educational environment can include such applications as computer workstations, modified restrooms, and transportation to and from work or school. Students need the ability to take notes and do assignments, and people working have a myriad of special tasks that may need to be analyzed and modified to enable the employee with the disability to


be independent and productive. Weisman has presented an extensive overview of rehabilitation engineering in the workplace, which includes a review of different types of workplaces, the process of accommodation, and many case examples [Weisman, 1990].

Recreation A component of living that is often overlooked by the professional community is the desire and, in fact, need of people with disabilities to participate in recreational activities. Many of the adaptive recreational technologies have been developed by persons with disabilities themselves in their effort to participate and be competitive in sports. Competitive wheelchair racing, archery, skiing, bicycles, and technology that enables people to bowl, play pool, and fly their own airplanes are just a few areas in which equipment has been adapted for specific recreational purposes.

Community and Workplace Access There is probably no other single legislation that is having a more profound impact on the lives of people with disabilities then the Americans with Disabilities Act (ADA), signed into law by President Bush in August of 1990. This civil rights legislation mandates that all people with disabilities have access to public facilities and that reasonable accommodations must be made by employers to allow persons with disabilities to access employment opportunities. The impact of this legislation is now sweeping America and leading to monumental changes in the way people view the rights of persons with disabilities.

146.5 Future Developments The field of rehabilitation engineering, both in research and in service delivery, is at an important crossroad in its young history. Shifting paradigms of services, reduction in research funding, consumerism, credentialing, health care reform and limited formal educational options all make speculating on what the future may bring rather hazy. Given all this, it is reasonable to say that one group of rehabilitation engineers will continue to advance the state of the art through research and development, while another group will be on the front lines as members of clinical teams working to ensure that individuals with disabilities receive devices and services that are most appropriate for their particular needs. The demarcation between researchers and service providers will become clearer, since the latter will become credentialed. RESNA and its professional specialty group (PSG) on rehabilitation engineering are working out the final credentialing steps for the Rehabilitation Engineer RE and the Rehabilitation Engineering Technologist RET. Both must also be an ATP. They will be recognized as valued members of the clinical team by all members of the rehabilitation community, including third-party payers, who will reimburse them for the rehabilitation engineering services that they provide. They will spend as much or more time working in the community as they will in clinical settings. They will work closely with consumer-managed organizations who will be the gatekeepers of increasing amounts of governmentmandated service dollars. If these predictions come to pass, the need for rehabilitation engineering will continue to grow. As medicine and medical technology continue to improve, more people will survive traumatic injury, disease, and premature birth, and many will acquire functional impairments that impede their involvement in personal, community, educational, vocational, and recreational activities. People continue to live longer lives, thereby increasing the likelihood of acquiring one or more disabling conditions during their lifetime. This presents an immense challenge for the field of rehabilitation engineering. As opportunities grow, more engineers will be attracted to the field. More and more rehabilitation engineering education programs will develop that will support the training of qualified engineers, engineers who are looking for exciting challenges and opportunities to help people live more satisfying and productive lives.


References ANSI/RESNA. 1990. Wheelchair Standards. RESNA Press, RESNA, 1700 Moore St., Arlington, VA 22209–1903. Axelson P, Minkel J, Chesney D. 1994. A Guide to Wheelchair Selection: How to Use the ANSI/RESNA Wheelchair Standards to Buy a Wheelchair. Paralyzed Veterans of America (PVA). Bain BK, Leger D. 1997. Assistive Technology. An Interdisciplinary Approach. Churchill Livingstone, New York. Center for Universal Design, 1999. http://www.design.ncsu.edu/cud/ Cook AM, Hussey SM. 1995. Assistive Technologies: Principles and Practice. Mosby, St. Louis, MO. Cooper RA. 1995. Rehabilitation Engineering Applied to Mobility and Manipulation. Institute of Physics Publishing, Bristol, U.K. Cooper RA. 1998. Wheelchair Selection and Configuration. Demos Medical Publishing, New York. Deno JH, et al. 1992. Human Factors Design Guidelines for the Elderly and People with Disabilities. Honeywell, Inc., Minneapolis, MN 55418 (Brian Isle, MN65–2300). Galvin JC, Scherer MJ. 1996. Evaluating, Selecting, and Using Appropriate Assistive Technology, Aspen Publishers, Gaithersburg, MD. Hobson DA. 1990. Seating and mobility for the severely disabled. In R Smith, J Leslie (eds), Rehabilitation Engineering, pp 193–252. CRC Press, Boca Raton, FL. Jones D, Jones W. 1990. Criteria for selection of an augmentative communication system. In R Smith, J Leslie (eds), Rehabilitation Engineering, pp 181–189. CRC Press, Boca Raton, FL. Medhat M, Hobson D. 1992. Standardization of Terminology and Descriptive Methods for Specialized Seating. RESNA Press, RESNA, 1700 Moore St., Arlington, VA 22209–1903. Rehabilitation Technology Service Delivery—A Practical Guide. 1987. RESNA Press, RESNA, 1700 Moore St., Arlington, VA 22209–1903. Smith RO. 1992. Technology and disability. AJOT 1(3):22. Society for Automotive Engineers. 1994. Wheelchair Tie-Down and Occupant Restraint Standard (committee draft). SAE. Warrendale, PA. Trace Center, 1999. http://trace.wisc.edu/ Vanderheiden G, Vanderheiden K. 19991. Accessibility Design Guidelines for the Design of Consumer Products to Increase their Accessibility to People with Disabilities or Who Are Aging. Trace R&D Center, University of Wisconsin, Madison, WI. Weisman G. 1990. Rehabilitation engineering in the workplace. In R Smith, J Leslie (eds), Rehabilitation Engineering, pp 253–297. CRC Press, Boca Raton, FL. WheelchairNet, 1999. http://www.wheelchairnet.org

Further Information ABLEDATA, 8455 Colesville Rd., Suite 935, Silver Spring, Md. 20910–3319.


Geddes, L. G. “Historical Perspectives 4 - Electromyography .” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

Historical Perspectives 4

Electromyography Leslie A. Geddes Purdue University

Early Investigations Clinical Electromyography

Early Investigations The study of bioelectricity started with the Galvani-Volta controversy over the presence of electricity in frog muscle [see Geddes and Hoff, 1971]. Galvani likened the sciatic nerve-gastrocnemius muscle to a charged Leyden jar (capacitor) in which the nerve was the inner conductor and the surface of the muscle was the outer conductor. Therefore, Galvani thought that by joining the two with an arc of dissimilar metals, the biologic capacitor was discharged and the muscle twitched. Volta proved conclusively that it was the dissimilar metals in contact with tissue fluid that was the stimulus. Interestingly, it was found that when the sciatic nerve of a nerve-muscle preparation was laid on the cut end of another frog muscle and the nerve was touched to the intact surface, the muscle of the nervemuscle preparation twitched. Here was evidence of stimulation without metal conductors; this experiment was performed by Matteucci [1842]. With the first galvanometers, it was shown that current would be indicated when one electrode was placed on the cut end of a frog muscle and the other on the intact surface. This phenomenon became known as the injury current or frog current, the cut surface being negative to the intact surface. Whereas the foregoing experiments showed that skeletal muscle possessed electricity, little was known about its relation to contraction. Matteucci [1842] conducted an ingenious experiment in which he placed the nerve of a second nerve-muscle preparation on the intact muscle of a first such preparation and stimulated the nerve of the first using an inductorium. He discovered that both muscles contracted. Here is the first evidence of the electric activity of contracting skeletal muscle. Matteucci and DuBois-Reymond both found that the injury current disappeared when a muscle was contracted tetanically. This observation led directly to the concept of a resting membrane potential and its disappearance with activity [see Hoff and Geddes, 1957]. That human muscle, as well as frog muscle, produced electric activity was demonstrated by Du BoisReymond [1858] in the manner shown in Fig. HP4.1. With electrodes in saline cups connected to a galvanometer, Du Bois-Reymond stated that as soon as the fingers were placed in the cups, the galvanometer needle deflected, and it required some time for a position of equilibrium to be attained. Du Bois-Reymond [1858] stated: As soon as this state [equilibrium] is attained, the whole of the muscles of one of the arms must be so braced that an equilibrium may be established between the flexors and the extensors of all the articulations of the limb, pretty much as in a gymnastic school is usually done when one wants to let a person feel the development of one's muscles. As soon as this is done, the [galvanometer] needle is thrown into movement, its deflection being uniformly in such a sense as to indicate in the braced arm "an inverse current," according to Nobili's nomenclature; that is to say, a current passing from the hand to the shoulder. The braced arm then


FIGURE HP4.1 The first evidence contracting skeletal muscle in man produces an electrical signal. (From Du BoisReymond [1858].)

acts the part of the copper in the compound arc of zinc and copper mentioned above. [Du BoisReymond was referring to the polarity of a voltaic cell in which zinc and copper are the positive and negative electrodes, respectively.] The rheotome and slow-speed galvanometer were used to reconstruct the form of the muscle action potential. However, it was desired to know the time course of the electric change associated with a single muscle contraction (twitch), as well as its relationship to the electrical event (action potential). A second item of interest was the so-called latent period, that time between the stimulus and the onset of muscle contraction, which Helmholtz [1853] reported to be 10 ms for frog muscle. Waller [1887] set himself the task of measuring the latent period and the relationship between the action potential and the force developed by frog gastrocnemius muscle in response to a single stimulus. He found that the onset of the twitch was later than the onset of the action potential, as shown in Fig. HP4.2. However, the true form of the muscle action potential and its relationship to the onset of the twitch had to await the development of the micropipet electrode, the vacuum-tube amplifier, and the cathode-ray oscilloscope. In 1957, Hodgkin and Horowicz [1957] recorded the twitch and action potential of a single muscle fiber of the frog. Figure HP4.3 is a copy of their record. Note that the onset of the action potential precedes the onset of muscle contraction by about 4 ms. We know that it is the action potential that triggers the release of mechanical energy.


FIGURE HP4.2 Relationship between the twitch (recorded with a myograph, m) and the action potential (recorded with a capillary electrometer, e) of a frog gastrocnemius muscle. The time marks (t) are 1/20 s. (From Waller [1887].)

FIGURE HP4.3 The relationship between the muscle action potential and the force of concentration in a single skeletal muscle fiber in the frog. (From Hodgkin and Horowicz [1957].)

Clinical Electromyography It was well known that when a nerve that innervates a skeletal muscle is cut, the muscle is paralyzed immediately; however, days to weeks later (depending on the species), on careful visual examination, the individual muscle fibers are seen to be contracting and relaxing randomly, i.e., fibrillating. The first to bring the facts together regarding normal muscle action potentials and denervation-fibrillation potentials were Denny-Brown and Pennybacker in the United Kingdom [1939]; the date and locale are highly significant. They distinguished between involuntary twitching of innervated muscle and fibrillation of denervated muscle by recording both the electric and mechanical activity of muscles. Two instrumental advances made their study possible: (1) the use of a hypodermic needle electrode inserted into the muscle and (2) the use of a rapidly responding, mirror-type photographic recorder, the Matthews [1928] oscillograph. The cathode-ray tube was not generally available in the United Kingdom when Matthews [1928] constructed his moving-tongue mirror oscillograph. The device consisted of a piece of soft iron mounted on a steel leaf spring, as shown in Fig. HP4.4. A strong electromagnet attracted the soft iron, which bent


FIGURE HP4.4 The Matthews moving-tongue oscillograph and amplifier used with it. The electromagnetic coil (A) provided an attractive force on the tongue (soft iron and steel spring); the signal current applied to the small coils aided or opposed this force and caused the tongue to bend more or less and hence move the mirror which reflected a beam of light on to a recording surface. (From BHC Matthews, 1928, J. Physiol (Lond) 65:225, with permission.)

the steel (leaf-spring) support. Two coils mounted on the pole faces were connected to the output tubes of a five-stage, single-sided, resistance-capacitance coupled amplifier. The amplified potentials altered the current in the electromagnet coils, causing more or less bending of the leaf spring, thereby tilting the mirror mounted on the leaf spring and permitting photographic recording of action potentials. With the Matthews oscillograph, Denny-Brown and Pennybacker [1939] laid the foundation for clinical electromyography when they reported as follows: Denervated muscle fibers contract periodically, and the confused medley of small twitches constitutes true fibrillation. The movement is so slight that it can seldom be seen in the clinic. The twitchings appear to be due to heightened excitability of the sarcolemma or rapidly conducting portion of the muscle fibre to traces of free acetylcholine in the tissues. Reinnervated muscle is free from such involuntary excitation, except for the curious "contracture" of facial muscle, which consists of periodic, intense repetitive discharges which suggest a central mechanism.


Earlier it was stated that 1939 was significant; this was the year when World War II broke out in Europe. Soon motor nerve injuries due to shrapnel wounds began to appear in large numbers, and the need for electromyography to identify denervation fibrillation potentials and their gradual disappearance with reinnervation was urgent. The first electromyograph in North America was developed by Herbert Jasper at McGill University (Montreal Neurological Institute). Starting in 1942, design concepts were developed, and by 1944 prototypes had been built and used clinically. In his report to the Committee on Army Medical Research, Capt. Jasper [1945] stated: The present equipment has been developed over a period of about 18 months experimentation with different designs of electromyograph for use on hospital wards. Numerous modifications of design have been incorporated in the present model in order to provide simplicity of operation, portability, freedom from electrical interference, and perfection of both the audible and visible analysis of the electrical activity of normal and diseased muscles. A portable clinical electromyograph has been developed which has proven to be practical in operation on hospital wards to aid in the diagnosis and prognosis of muscles paralyzed by traumatic injuries of their nerve supply. Four complete units have been constructed for use in special centers for the treatment of nerve injuries. The Royal Canadian Army Medical Corps (RCAMC) electromyograph had many unique design features that were incorporated in all later commercially available electromyographs. It consisted of three units, a small battery-operated three-stage differential amplifier (Fig. HP4.5) and an oscilloscope (Fig. HP4.6), both placed on a loudspeaker cabinet on rubber-wheel casters (Fig. HP4.7). Thus simultaneous visual display and aural monitoring of normal motor units and fibrillation potentials was possible. The preamplifier (Fig. HP4.5) was very carefully constructed, the input tubes being supported by rubber-mounted antimicrophonic sockets. The grid resistors (R9) and plate resistors (R8) were wire wound and carefully matched. A high common-mode rejection ratio was obtained by matching the input tubes and adjustment of the potentiometer (R7) in the screen-grid supply. A common-mode rejection ratio in excess of 10,000 was easily achieved. The overall frequency response extended from 3 to 10,000 Hz. The cathode-ray oscilloscope (Fig. HP4.6) was of unusual design for that time because the sweep velocity was independent of the number of sweeps per second, a feature to appear much later in oscilloscopes. A linear sweep (time base) was obtained by charging a capacitor (0.1 µF) through a pentode (6U7G) which acted as a constant-current device. The sweep was initiated at a rate of about 7 times per second by the multivibrator (6N7), which also provided an output to enable stimulating a nerve, the stimulus occurring at the beginning of each sweep, thereby permitting nerve conduction-time measurements. The oscilloscope also contained the audio amplifier (6L6). The cathode-ray tube had a shortpersistence, blue-white phosphor that produced brilliant blue-white images of remarkable clarity. A camera was used to obtain photographic records of the waveforms, which were optimized by listening to them via the loudspeaker (as advocated by Adrian) as the needle electrode was being inserted and adjusted. Fig. HP4.8 illustrates typical action potentials. At the end of the war (1945), oscilloscopes became available, and Fig. HP4.7 shows the RCAMC electromyograph with a Cossor oscilloscope (right) and the high-gain differential amplifier (left), both on the loudspeaker cabinet, which was on casters. The recessed opening at the top of the loudspeaker cabinet face provided access to the on-off and volume controls. In addition to the creation of a high-performance EMG unit, Jasper introduced the monopolar needle electrode system used in all subsequent EMGs. The needle electrode was insulated with varnish down to its tip and was paired with a skin-surface electrode of silver. The patient was grounded by another electrode taped to the same member that was being examined. Figure HP4.8 illustrates application of the electrodes and typical motor-unit and fibrillation potentials. The report by Jasper and Ballem [1949] summarized the experience with the RCAMC electromyograph and laid the foundation for diagnostic EMG. World War II ended in 1945, after which electromyographs became available commercially. Their features were essentially the same as those embodied in the RCAMC electromyograph. From the beginning, these units were completely power-line-operated, the author's master's thesis describing the first of these units.


FIGURE HP4.5 The three-stage resistance-capacity coupled differential amplifier used is the RCAMC electromyography. (From Jasper et al. [1945].)


FIGURE HP4.6 The oscilloscope, loudspeaker amplifier, and stimulator unit of the RCAMC electromyograph. (From Jasper et al. [1945].)


FIGURE HP4.7 A later version of the RCAMC electromyograph showing the high-gain preamplifier (left) and oscilloscope (right) on top of the loud-speaker cabinet.

FIGURE HP4.8 Electrode arrangement and action potentials with the RCAMC electromyograph. (Redrawn from Jasper and Ballem [1949].)


References Denny-Brown D, Pennybacker JB. 1939. Fibrillation and fasciculation in voluntary muscle. Brain 61:311. Du Bois-Reymond E. 1858. Untersuchungen uber thierische Elekticitat. Moleschott's Untersuch. Z Natur Mensch 4:1. Geddes LA, Hoff HE. 1971. The discovery of bioelectricity and current electricity (the Galvani-Volta controversy). IEEE Spect 8(12):38. Helmholtz H. 1853. On the methods of measuring very small portions of time and their application to physiological purposes. Philos Mag J Sci 6:313. Hodgkin AL, Horowicz P. 1957. The differential action of hypertonic solutions on the twitch and action potential of a muscle fiber. J Physiol (Lond) 136:17P. Hoff HE, Geddes LA. 1957. The rheotome and its prehistory: A study in the historical interrelation of electrophysiology and electromechanics. Bull Hist Med 31(3):212. Jasper HH, Ballem G. 1949. Unipolar electromyograms of normal and denervated human muscle. J Neurophysiol 12:231. Jasper HH. 1945. The RCAMC electromyograph, Mark II. With the technical assistance of Lt. RH Johnston and LA Geddes. Report submitted to the Associate Committee on Army Medical Research, National Research Council of Canada, 27 April 1945. Matteucci C. 1842. Duxieme memoire sur le courant electrique propre de la grenouille et sur celui des animaux a sang chaud. Ann Chim Phys 3S(6):301. Matthews BHC. 1928. A new electrical recording system for physiological work. J Physiol (Lond) 65:225. Waller AD. 1887. A demonstration on man of electromotive changes accompanying the heart's beat. J Physiol (Lond) 8:229.


Kondraske, G. V. “Human Performance Engineering.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

XV Human Performance Engineering George V. Kondraske University of Texas at Arlington, Human Performance Institute 147 A Working Model for Human System-Task Interfaces George V. Kondraske Background • Basic Principles • Application Issues • Conclusion

148 Measurement of Neuromuscular Performance Capacities Susan S. Smith Neuromuscular Functional Units • Range of Motion and Extremes of Motion • Strength • Speed of Movement • Endurance • Reliability, Validity, and Limitations in Testing • Performance Capacity Space Representations • Conclusions

149 Measurement of Sensory-Motor Control Performance Capacities: Tracking Tasks Richard D. Jones Basic Principles • Measurement of Sensory-Motor Control Performance • Analysis of Sensory-Motor Control Performance

150 Measurement of Information-Processing Performance Capacities George V. Kondraske, Paul J. Vasta Basic Principles • General Measurement Paradigms • Measurement Instruments and Procedures • Present Limitations

151 High-Level Task Analysis: Mental Components

Kenneth J. Maxwell

Fundamentals • Mental Task Analysis: Process and Methods • Models of Human Mental Processing and Performance • Models of Machine Processing Capabilities • HumanMachine Task Analytic Framework • Brief Example: Analysis of Supervisory Control Task

152 Task Analysis and Decomposition: Physical Components

Sheik N. Imrhan

Fundamental Principles • Early Task-Analysis Methods • Methods of Physical Task Analysis • Factors Influencing the Conduct of Task Analysis • Measurement of Task Variables • Uses and Applications of Task Analysis • Future Developments

153 Human-Computer Interface Design Issues

Kenneth J. Maxwell

Fundamentals • A Quantitative Performance-Based Model of Usability • Selected HCI Design Issues, Goals, and Resource Requirements


154 Applications of Human Performance Measurements to Clinical Trials to Determine Therapy Effectiveness and Safety Pamela J. Hoyes Beehler, Karl Syndulko Basic Principles: Types of Studies • Methods • Representative Application Examples • Future Needs and Anticipated Developments

155 Applications of Quantitative Assessment of Human Performance in Occupational Medicine Mohamad Parnianpour Principles • Low Back Pain and Trunk Performance • Clinical Applications • Conclusions

156 Human Performance Engineering: Computer-Based Design and Analysis Tools Paul J. Vasta, George V. Kondraske Selected Fundamentals • Scope, Functionality, and Performance • Functional Overview of Representative Packages • Anticipated Development

157 Human Performance Engineering: Challenges and Prospects for the Future George V. Kondraske Models • Measurements • Databases and Data Modeling • Summary

T

HE ULTIMATE GOAL OF HUMAN performance engineering is enhancement of the performance and safety of humans in the execution of tasks. The field (in a more formalized sense) was fueled initially by military applications but has become an important component in industrial settings as well. In a biomedical engineering context, the scope of definition applied to the term human not only encompasses individuals with capabilities that differ from those of a typical healthy individual in many possible different ways (e.g., individuals who are disabled, injured, unusually endowed, etc.) but also includes those who are “healthy” (e.g., health care professionals). Consequently , one finds a wide range of problems in which human performance engineering and associated methods are employed. Some examples include • Evaluation of an individual's performance capacities for determining the efficacy of new therapeutic interventions or so-called level of disability for worker's compensation and other medicallegal purposes. • Design of assistive devices and/or work sites in such a way that a person with some deficiency in his or her “performance resource profile” will be able to accomplish specified goals. • Design of operator interfaces for medical instruments that promote efficient, safe, and error-free use. In basic terms, each of these situations involves one or more of the following: (1) a human, (2) a task or tasks, and (3) the interface of a human and task(s). Human performance engineering emphasizes concepts, methods, and tools that strive toward treatment of each of these areas with the engineering rigor that is routinely applied to artificial systems (e.g., mechanical, electronic, etc.). Importance is thus placed on models (a combination of cause-and-effect and statistical), measurements (of varying degrees of sophistication that are selected to fit needs of a particular circumstance), and various types of analyses. Whereas many specialty areas within biomedical engineering begin with an emphasis on a specific subsystem and then proceed to deal with it at lower levels of detail (sometimes even at the molecular level) to determine how it functions and often why it malfunctions, human performance engineering emphasizes subsystems and their performance capacities (i.e., how well a system functions), the integration of these into a whole and their interactions, and their operation in the execution of tasks that are of ultimate concern to humans. These include tasks of daily living, work, and recreation. In recent years, there has been an increased concern within medical communities on issues such as quality of life, treatment outcome measures, and treatment cost-effectiveness. By linking human subsystems into the “whole” and discovering objective quantitative relationships between the human and tasks, human performance engineering can play an important role in addressing these and other related concerns. Human performance engineering combines knowledge, concepts, and methods from across many disciplines (e.g., biomechanics, neuroscience, psychology, physiology, and many others) which, in their


overlapping aspect, all deal with similar problems. Among current difficulties is that these wide-ranging efforts are not linked by a conceptual framework that is commonly employed across contributing disciplines. In fact, few candidate frameworks exist even within the relevant disciplines. One attempt to provide some unification and commonality is presented in Chapter 147 as basis for readers to integrate material in subsequent chapters and from other sources. In a further attempt to enhance continuity across this section, chapter authors have been requested to consider this perspective and to incorporate basic concepts and terms where applicable. Chapters 148 through 150 look “toward the human” and focus on measurement of human performance capacities and related issues. Owing to a combination of the complexity of the human system (even when viewed as a collection of rather high-level subsystems) and limited space available, treatment is not comprehensive. For example, measurement of sensory performance capacities (e.g., tactile, visual, auditory) is not included in this edition. Both systems and tasks can be viewed at various hierarchical levels. Chapters 148 and 149 focus on a rather “low” systems level and discuss basic functional units such as actuator, processor, and memory systems. Chapter 150 moves to a more intermediate level, where speech, postural control gait, and hand-eye coordination systems could be considered. Measurement of structural parameters, which play important roles in many analyses, also is not allocated the separate chapter it deserves (as a minimum) due to space limitations. Chapter 151 and 152 then shift focus to consider the analysis of different types of tasks in a similar, representative fashion. Chapters 153 through 155 are included to provide insight into a representative selection of application types. Space constraints, the complexity of human performance, and the great variety of tasks that can be considered limit the level of detail with which such material can reasonably be presented. Work in all application areas will begin to benefit from emerging computer-based tools, which is the theme of chapter 156. The section concludes with a look to the future (Chapter 157) that summarizes selected current limitations, identifies some specific research and development needs, and speculates regarding the nature of some anticipated developments. Many have contributed their talents to this exciting field in terms of both research and applications, yet much remains to be done. I am indebted to the authors not only for their contributions and cooperation during the preparation of this section but also for their willingness to accept the burdens of communicating complex subject matter reasonably, selectively, and as accurately as possible within the imposed constraints.


Kondraske, G. V. “A Working Model for Human System-Task Interfaces.” The Biomedical Engineering Handbook: Second Edition. Ed. Joseph D. Bronzino Boca Raton: CRC Press LLC, 2000

147 A Working Model for Human System-Task Interfaces 147.1 147.2

Background Basic Principles General Systems Performance Theory • Monadology • The Elemental Resource Model

147.3

Conceptual, Low-Tech Practical Application • Conceptual, Theoretical Application • Application “in Part” • Rigorous, High-Tech Application

George V. Kondraske University of Texas at Arlington

Application Issues

147.4

Conclusion

Humans are complex systems. Our natural interest in ourselves and things that we do has given rise to the study of this complex system at every conceivable level ranging from genetic, through cellular and organ systems, to interactions of the total human with the environment in the conduct of purposeful activities. At each level, there are corresponding practitioners who attempt to discover and rectify or prevent problems at the respective level. Some practitioners are concerned with specific individuals, while others (e.g., biomedical scientists and product designers) address populations as a whole. Problems dealt with span medical and nonmedical contexts, often with interaction between the two. Models play a key role not only in providing conceptual understanding of key issues at each level but also in describing relationships between various levels and providing frameworks that allow practitioners to obtain reasonably predictable results in a systematic and efficient fashion. In this chapter, a working model for human system-task interfaces is presented. Any such model must, of course, not only consider the interface per se but also representations of the human system and tasks. The model presented here, the elemental resource model (ERM), represents the most recent effort in a relatively small family of models that attempt to address similar needs.

147.1 Background The interface between a human and a task of daily living (e.g., work, recreation, or other) represents a level that is quite high in the hierarchy noted above. One way in which to summarize previous efforts directed at this level, across various application contexts, is to recognize two different lines along which study has evolved: (1) bottom-up and (2) top-down. Taken together, these relative terms imply a focus of interest at a particular level of convergence. Here, this special level is termed the human-task interface level. It is emphasized that bottom-up and top-down terms are used here to characterize the general


course of development and not specific approaches applied at a particular instant in time. A broad view is necessary to grapple with the many previous efforts that either are, or could be, construed to be pertinent. The biomedical community has approached the human-task interface largely along the bottom-up path. This is not surprising given the historical evolution of interest first in anatomy (human structure) and then physiology (function). The introduction of chemistry (giving rise to biochemistry) and the refinement of the microscope provided motivations to include even lower hierarchical levels of inquiry and of a substantially different character. Models in this broad bottom-up category begin with anatomic components and include muscles, nerves, tendons (or subcomponents thereof), or subsets of organs (e.g., heart, lungs, vasculature, etc.). They often focus on relationships between components and exhibit a scope that stays within the confines of the human system. Many cause-and-effect models have been developed at these lower levels for specific purposes (e.g., to understand lines of action of muscle forces and their changes during motion about a given joint). As a natural consequence of linkages that occur between hierarchical levels and our tendency to utilize that which exists, consideration of an issue at any selected level (in this case, the human-task interface level) brings into consideration all lower levels and all models that have been put forth with the stated purpose of understanding problems or behaviors at the original level of focus. The amount of detail that is appropriate or required at these original, lower levels results in great complexity when applied to the total human at the human-task interface level. In addition, many lower-level modeling efforts (even those which are quantitative) are aimed primarily at obtaining a basic scientific understanding of human physiology or specific pathologies (i.e., pertaining to populations of humans). In such circumstances, highly specialized, invasive, and cumbersome laboratory procedures for obtaining the necessary data to populate models are justified. However, it is difficult and sometimes impossible to obtain data describing a specific individual to utilize in analyses when such models are extended to the human-task interface level. Another result of drawing lower-level models (and their approaches) into the human-task interface context is that the results have a specific and singular character (e.g., biomechanical versus neuromuscular control versus psychologic, etc.) [e.g., Card et al., 1986; Delp et al., 1990; Gottlieb et al., 1989; Hemami, 1988; Schoner & Kelso, 1988]. Models that incorporate most or all of the multiple aspects of the human system or frameworks for integrating multiple lower-level modeling approaches have been lacking. Lowerlevel models that serve meaningful purposes at the original level of focus have provided and will continue to provide insights into specific issues related to human performance at multiple levels of considerations. However, their direct extension to serve general needs at the human-task interface level has inherent problems; a different approach is suggested. A top-down progression can be observed over the major history in human factors/ergonomic [Gilbreth & Gilbreth, 1917; Taylor, 1911] and vocational assessment [e.g., Botterbusch, 1987] fields (although the former has more recently emphasized a “human-centered” concept with regard to design applications). In contrast to the bottom-up path, in which anatomic components form the initial basis of modeling efforts, the focus along the top-down path begins with consideration of the task or job that is to be performed by the total human. The great variety in the full breadth of activities in which humans can be engaged gives rise to one aspect of complexity at this level that pertains to taxonomies for job and task classification [e.g., Fleishman & Quaintance, 1984; Meister, 1989; U.S. Department of Labor, 1992]. Another enigmatic aspect that quickly adds complexity with respect to modeling concerns the appropriate level to be used to dissect the items at the highest level (e.g., jobs) into lower-level components (e.g., tasks and subtasks). In fact, the choice of level is complicated by the fact that no clear definition has evolved for a set of levels from which to choose. After progressing down through various levels at which all model elements represent tasks and are completely outside the confines of the human body, a level is eventually reached where one encounters the human. Attempts to go further have been motivated, for example, by desires to predict performance of a human in a given task (e.g., lifting an object, assembling a product, etc.) from a set of measures that characterizes the human. At the human-task interface level, difficulty is encountered with regard to the strategy for approaching a system as complex, multifaceted, and multipurpose as a human [Fleishman


& Quaintance, 1984; Wickens, 1984]. In essence, the full scope of options that have emerged from the bottom-up development path is now encountered from the opposite direction. Options range from relatively gross analyses (e.g., estimates of the “fraction” of a task that is physical or mental) to those which are much more detailed and quantitative. The daunting prospect of considering a “comprehensive quantitative model” has led to approaches and models, argued to be “more practical,” in which sets of parameters are often selected in a somewhat mysterious fashion based on experience (including previous research) and intuition. The selected parameters are then used to develop predictive models, most of which have been based primarily on statistical methods (i.e., regression models) [Fleishman, 1967; Fleishman & Quaintance, 1984]. Although the basic modeling tools depend only on correlation, it is usually possible to envision a causal link between the independent variables selected (e.g., visual acuity) and the dependent variable to be predicted (e.g., piloting an aircraft). Models (one per task) are then tested in a given population and graded with regard to their prediction ability, the best of which have performed marginally [Kondraske and Beehler, 1994]. Another characteristic associated with many of the statistically based modeling efforts from the noted communities is the almost exclusive use of healthy, “normal” subjects for model development (i.e., humans with impairments were excluded). Homogeneity is a requirement of such statistical models, leading to the need for one model per task per population (at best). Also, working with a mindset that considers only normal subjects can be observed to skew estimates regarding which of the many parameters that one might choose for incorporation in a model are “most important.” The relatively few exceptions that employ cause-and-effect models (e.g., based on physical laws) at some level of fidelity [e.g., Chaffin & Andersonn, 1984] often adopt methods that have emerged from the bottom-up path and are, as noted above, limited in character at the “total human” level (e.g., “biomechanical” in the example cited). The issue is not that no useful models have emerged from previous efforts but rather that no clear comprehensive strategy has emerged for modeling at the human-task interface level. A National Research Council panel on human performance modeling [Baron et al., 1990] considered the fundamental issues discussed here and also underscored needs for models at the human-task interface level. While it was concluded that an all-inclusive model might be desirable (i.e., high fidelity, in the sense that biomechanical, information processing, sensory and perceptual aspects, etc. are represented), such a model was characterized as being highly unlikely to be achieved and perhaps ultimately not useful because it would be overly complex for many applications. The basic recommendation made by this panel was to pursue development of more limited scope submodels. The implication is that two or more submodels could be integrated to achieve a broader range of fidelity, with the combination selected to meet the needs of particular situations. The desire to “divide efforts” due to inherent complexity of the problem also surfaces within the histories of the bottom-up and top-down development paths discussed above. While a reasonable concept in theory, one component in the division of effort that has consistently been underrepresented is the part that ties together the so-called submodels. Without a conceptual framework for integration of relatively independent modeling efforts and a set of common modeling constructs, prospects for long-term progress are difficult to envision. This, along with the recognition that enough work had been undertaken in the submodel areas so that key issues and common denominators could be identified, motivated development of the ERM. The broad objectives of the ERM [Kondraske, 1994] are most like those of Fleishman and colleagues [Fleishman, 1966, 1972,1982; Fleishman & Quaintance, 1984], whose efforts in human performance are generally well known in many disciplines. These are the only two efforts known that (1) focus on the total human in a task situation (i.e., directly address the human-task interface level); (2) consider tasks in general, and not only a specific task such as gait, lifting, reading, etc.; (3) incorporate all aspects of the total human system (e.g., sensory, biomechanical, information processing, etc.); and (4) aim at quantitative models. There are also some similarities with regard to the incorporation of the ideas of “abilities” (of humans) and “requirements” (of tasks). The work of Fleishman and colleagues has thus been influential in shaping the ERM either directly or indirectly through its influence of others. However, there are several substantive conceptual differences that have resulted in considerably different endpoints. Fleishman’s work emerged from “the task” perspective and is rooted in psychology, whereas the ERM


emerges from the perspective of “human system architecture” and is rooted in engineering methodology with regard to quantitative aspects of system performance but also incorporates psychology and physiology. Both approaches address humans and tasks, and both efforts contain aspects identifiable with psychology and engineering, as they ultimately must. These different perspectives, however, may explain in part some of the major differences. Aspects unique to the ERM include (1) the use of a resource construct for modeling and measurement of all aspects of a system’s performance, (2) the use of causeand-effect resource economic principles (i.e., the idea of threshold “costs” for achieving a given level of performance in any given high-level task), (3) the concept of monadology (i.e., the use of a finite set of “elements” to explain a complex phenomenon), and (4) a consistent strategy for identifying performance elements at different hierarchical levels. The ERM attempts to provide a quantitative and relatively straightforward framework for characterizing the human system, tasks, and the interface of the human to tasks. It depends in large part on, and evolves directly from, a separate body of material referred to collectively as general systems performance theory (GSPT). GSPT was developed first and independently, i.e., removed from the human system context. It incorporates resource constructs exclusively for modeling of the abstract idea of system performance, including specific rules for measuring performance resource availability and resource economic principles to provide a cause-and-effect analysis of the interface of any system (e.g., humans) to tasks. The concept of a performance model is emphasized and distinguished from other model types.

147.2 Basic Principles The history of the ERM and the context in which it was developed are described elsewhere [Kondraske, 1987a, 1990b, 2000]. It is important to note that the ERM is derived from the combination of GSPT with the philosophy of monadology and their application to the human system. As such, these two constituents are briefly reviewed before presenting and discussing the actual ERM.

General Systems Performance Theory The concept of performance now pervades all aspects of life, especially decision-making processes that involve both human and artificial systems. Yet it has not been well understood theoretically, and systematic techniques for modeling and its measurement have been lacking. While a considerable body of material applicable to general systems theory exists, the concept of performance has not been incorporated in it, nor has performance been addressed in a general fashion elsewhere. Most of the knowledge that exists regarding performance and its quantitative treatment has evolved within individual application contexts, where generalizations can easily be elusive. Performance is multifaceted, pertaining to how well a given system executes an intended function and the various factors that contribute to this. It differs from behavior of a system in that “the best of something” is implied. The broad objectives of GSPT are 1. To provide a common conceptual basis for defining and measuring all aspects of the performance of any system. 2. To provide a common conceptual basis for the analysis of any task in a manner that facilitates system-task interface assessments and decision making. 3. To identify cause-and-effect principles, or laws, that explain what occurs when any given system is used to accomplish any given task. While GSPT was motivated by needs in situations where the human is “the system” of interest and it was first presented in this context [Kondraske, 1987a], application of it has been extended to the context of artificial systems. These experiences range from computer vision and sensor fusion [Yen & Kondraske, 1992] to robotics [Kondraske & Khoury, 1992; Kondraske & Standridge, 1988]. A succint statement of GSPT designed to emphasize key constructs is presented below in a steplike format. The order of steps is intended to suggest how one might approach any system or system-task


interface situation to apply GSPT. While somewhat terse and “to-the-point,” it is nonetheless an essential prerequisite for a reasonably complete understanding the ERM. 1. Within a domain of interest, select any level of abstraction and identify the system(s) of interest (i.e., the physical structure) and its function (i.e., purpose). 2. Consider “the system” and “the task” separately. 3. Use a resource construct to model the system’s performance. First, consider the unique intangible qualities that characterize how well a system executes its function. Each of these is considered to represent a unique performance resource associated with a specific dimension of performance (e.g., speed, accuracy, stability, smoothness, “friendliness,” etc.) of that system. Each performance resource is recognized as a desirable item (e.g., endurance versus fatigue, accuracy versus error, etc.) “possessed” by the system in a certain quantitative amount. Thus one can consider quantifying the amount of given quality available. As illustrated, an important consequence of using the resource construct at this stage is that confusion associated with duality of terms is eliminated. 4. Looking toward the system, identify all I dimensions of performance associated with it. In situations where the system does not yet exist (i.e., design contexts), it is helpful to note that dimensions of performance of the system are the same as those of the task. 5a. Keeping the resource construct in mind, define a parameterized metric for each dimension of performance (e.g., speed, accuracy, etc.). If the resource construct is followed, values will be produced with these metrics that are always nonnegative. Furthermore, a larger numeric value will consistently represent more of a given resource and therefore more performance capacity. 5b. Measure system performance with the system removed from the specific intended task. This is a reinforcement of Step 2. The general strategy is to maximally stress the system (within limits of comfort and/or safety, when appropriate) to define its performance envelope, or more specifically, the envelope that defines performance resource availability, RAS(t). Note that RAS(t) is a continuous surface in the system’s nonnegative, multidimensional performance space. Also note that unless all dimensions of performance and parameterized metrics associated with each are defined using the resource construct, a performance envelope cannot be guaranteed. Addressing the issue of measurement more specifically , consider resource availability values RAi |Qi,k(t) for i = 1 to I, associated with each of the I dimensions of performance. Here, each Qi,k represents a unique condition, in terms of a set of values Ri along other identified dimensions of performance, under which a specific resource availability (RAi) is measured; i.e., Qi,k = {R1,k , R2,k ,…, Rp,k} for all p ≠ i (1 ≥ p ≥ I). The subscript k is used to distinguish several possible conditions under which a given resource availability (RA1, for example) can be measured. These values are measured using a set of “test tasks,” each of which is designed to maximally stress the system (within limits of comfort and/or safety, when appropriate): (a) along each dimension of performance individually (where Qi,k = Qi,0 = {0,0, …,0}) or (b) along selected subsets of dimensions of performance simultaneously (i.e., Qi,k = Qi,n , where each possible Qi,n has one or more nonzero elements). The points obtained [RAi |Qi,k(t)] provide the basis to estimate the performance envelope RAS(t). Note that if only onaxis points are obtained (e.g., maximally stress one specific performance resource availability with minimal or no stress on other performance resources, or the Qi,0 condition), a rectangular or idealized performance envelope is obtained. A more accurate representation, which would be contained within the idealized envelope, can be obtained at the expense of making additional measurements or the use of known mathematic functions based on previous studies that define the envelope shape in two or more dimensions. 5c. Define estimates of single-number system figures-of-merit, or composite performance capacities, as the mathematical product of all or any selected subset of RAi |Qi,k(t). If more accuracy is desired and a sufficient number of data points is available from the measurement process described in Step 5b, composite performance capacities can be determined by integration over RAS(t) to determine the volume enclosed by the envelope. The composite performance capacity is a measure of performance at a higher level of abstraction than any individual dimension of performance at the “system”


level, representing the capacity of the system to perform tasks that place demands on those performance resources availabilities included in the calculation. Different composite performance capacities can be computed for a given system by selecting different combinations of dimensions of performance. Note that the definition of a composite performance capacity used here preserves dimensionality; e.g., if speed and accuracy dimensions are included in the calculation, the result has units of speed × accuracy. (This step is used only when needed, e.g., when two general-purpose systems of the same type are to be compared. However, if decision making that involves the interface of a specific system to a specific task is the issue at hand, a composite performance capacity is generally of any use only to rule out candidates). 6. Assess the “need for detail.” This amounts to a determination of the number of hierarchical levels included in the analysis. If the need is to determine if the currently identified “system” can work in the given task or how well it can execute its function, go to Step 7 now. If the need is to determine the contribution of one or more constituent subsystem or why a desired level of performance is not achieved at the system level, repeat Steps 1 to 5 for all J functional units (subsystems), or a selected subset thereof based on need, that form the system that was originally identified in Step 1; i.e., go to the next-lowest hierarchical level. 7. At the “system” level, look toward the task(s) of interest. Measure, estimate, or calculate demands on system performance resources (e.g., the speed, accuracy, etc. required), RDi |Q′i,k(t), where the notation here is analogous to that employed in Step 5b. This represents the quantitative definition and communication of goals, or the set of values PHLT representing level of performance P desired in a specific high-level task (HLT). Use a worst-case or other less-conservative strategy (with due consideration of the impact of this choice) to summarize variations over time. This will result in a set of M points (RDm , for m = 1 to M) that lie in the multidimensional space defined by the set of I dimensions of performance. Typically, M ≥ I. 8. Use resource economic principles (i.e., require RA ≥ RD for “success”) at the system level and at all system-task interfaces at the subsystem level (if included) to evaluate success/failure at each interface. More specifically, for a given system-task interface, all task demand points (i.e., the set of RDm associated with a given task or subtask) must lie within the performance resource envelope RAS(t) of the corresponding system. This is the key law that governs system-task interfaces. If a two-level model is used (i.e., “system” and “subsystem” levels are incorporated), map system-level demands to demands on constituent subsystems. That is, functional relationships between PHLT and demands imposed on constituent subsystems [i.e., RDi,j(t, PHLT)] must be determined. The nature of these mappings depends on the type of systems in question (e.g., mechanical, information processing, etc.). The basic process includes application of Step 7 to the subtasks associated with each subsystem. If resource utilization flexibility (i.e., redundancy in subsystems of similar types) exists, select the “best” or optimal subsystem °configuration (handled in GSPT with the concept of performance resource substitution) and procedure (i.e., use of performance resources over time) as that which allows accomplishment of goals with minimization of stress on available performance resources across all subsystems and over the time of task execution. Thus redundancy is addressed in terms of a constrained performance resource optimization problem. Stress on individual performance resources is defined as 0 < RDi,j(t, PHLT)/ RAi,j(t) < 1. It is also useful to define and take note of reserve capacity, i.e., the margin between available and utilized performance resources. The preceding statement is intended to reflect the true complexity that exists in systems, tasks, and their interfaces when viewed primarily from the perspective of performance. This provides a basis for the judicious decision making required to realize “the best” practical implementation in a given situation where many engineering tradeoffs must be considered. While a two-level approach is described above, it should be apparent that it can be applied with any number of hierarchical levels by repeating the steps outlined in an iterative fashion starting at a different level each time. A striking feature of GSPT is the threshold effect associated with the resource economic principle. This nonlinearity has important implications in quantitative human performance modeling, as well as interesting ramifications in practical


applications such as rehabilitation, sports, and education. Note also that no distinction is made as to whether a given performance resource is derived from a human or artificial system; both types of systems, or subcomponents thereof, can be incorporated into models and analyses.

Monadology Monadology dates back to 384 B.C. [Neel, 1977] but was formalized and its importance emphasized by Gottfried Wilhelm Leibniz, inventor of calculus, in his text Monadologia in 1714. This text presents what is commonly called Leibniz’s Monadology and has been translated and incorporated into contemporary philosophy texts (e.g., Leibniz and Montgomery, 1992). It is essentially the idea of “basic elements” vis a vis chemistry, alphabets, genetic building blocks, etc. The concept is thus already well accepted as being vital to the systematic description of human systems from certain perspectives (i.e., chemical, genetic). Success associated with previous applications of monadology, whether intentional or unwitting (i.e., discovered to be at play after a given taxonomy has emerged), compels its serious a priori consideration for other problems. Insight into how monadology is applied to human performance modeling is perhaps more readily obtained with reference to a widely known example in which monadology is evident, such as chemistry. Prior to modern chemistry (i.e., prior to the introduction of the periodic table), alchemy existed. The world was viewed as being composed of an infinite variety of unique substances. The periodic table captured the notion that this infinite variety of substances could all be defined in terms of a finite set of basic elements. Substances have since been analyzed using the “language” of chemistry and organized into categories of various complexity, i.e., elements, simple compounds, complex compounds, etc. Despite the fact that this transition occurred approximately 200 years ago, compounds remain that have yet to be analyzed. Furthermore, the initial periodic table was incorrect and has undergone revision up to relatively recent times. Analogously, in the alchemy of human performance, the world is viewed as being composed of an infinite variety of unique tasks. A “chemistry” can be envisioned that first starts with the identification of the “basic elements” or, more specifically, basic elements of performance. Simple and complex tasks are thus analogous to simple and complex compounds, respectively. The analogy carries over to quantitative aspects of GSPT as well. Consider typical equations of chemical reactions with resources on the left and products (i.e., tasks) on the right. Simple compounds (tasks) are realized by drawing on basic elements in the proper combination and amounts. The amount of “product” (level of performance in a high-level task) obtained depends on availability of the limiting resource. Another informative aspect of this analogy is the issue of how to deal with the treatment of hierarchical level. Clearly, the chemical elements are made up of smaller particles (e.g., protons, neutrons, and electrons). Physicists have identified even smaller, more elusive entities, such as bosons, quarks, etc. Do we need to consider items at this lowest level of abstraction each time a simple compound such as hydrochloric acid is made? Likewise, the term basic in basic elements of performance is clearly relative and requires the choice of a particular hierarchical level of abstractions for the identification of systems or basic functional units, a level that is considered to be both natural and useful for the purpose at hand. Just as it is possible but not always necessary or practical to map chemical elements down to the atomic particle level, it is possible to consider mapping a basic element of performance (see below) such as elbow flexor torque production capacity down to the level of muscle fibers, biochemical reactions at neuromuscular junctions, etc.

The Elemental Resource Model The resource and resource economic constructs used in GSPT specifically to address performance have employed and have become well established in some segments of the human performance field, specifically with regard to attention and information processing [Navon & Gopher, 1979; Wickens, 1984]. However, in these cases, the term resource is used mostly conceptually (in contrast to quantitatively), somewhat softly defined, and applied to refer in various instances to systems (e.g., different processing


FIGURE 147.1 The elemental resource model contains multiple hierarchical levels. Performance resources (i.e., the basic elements) at the “basic element level” are finite in number, as dictated by the finite set of human subsystems and the finite set of their respective dimensions of performance. At higher levels, new “systems” can be readily created by configuration of systems at the basic element level. Consequently, there are an infinite number of performance resources (i.e., higher-level elements) at these levels. However, rules of general systems performance theory (refer to text) are applied at any level in the same way, resulting in the identification of the system, its function, dimensions of performance, performance resource availabilities (system attributes), and performance resource demands (task attributes).

centers), broad functions (e.g., memory versus processing), and sometimes to infer a particular aspect of performance(e.g., attentional resources). In the ERM, through the application of GSPT, these constructs are incorporated universally (i.e., applied to all human subsystems) and specifically to model “performance” at both conceptual and quantitative levels. In addition to the concept of monadology, the insights of others [Shoner & Kelso, 1988, Turvey et al., 1978] were valuable in reinforcing the basic “systems architecture” employed in the ERM and in refining description of more subtle, but important aspects. As illustrated in Fig. 147.1, the ERM contains multiple hierarchical levels. Specifically, three levels are defined: (1) the basic element level, (2) the generic intermediate level, and (3) the high level. GSPT is to define performance measures at any hierarchical level. This implies that to measure performance, one must isolate the desired system and then stress it maximally along one dimension of performance (or more, if interaction effects are desired) to determine performance resource availability. For example, consider the human “posture stabilizing” system (at the generic intermediate level), which is stressed maximally along a stability dimension. As further illustrated below, the basic element level represents


measurable stepping stones in the human system hierarchy between lower-level systems (i.e., ligaments, tendons, nerves, etc.) and higher-level tasks. A summary representation emphasizing the basic element level of the ERM is depicted in Fig. 147.2. While this figure is intended to be more or less self-explanatory, a brief walk-through is warranted. Looking Toward the Human. The entire human (lower portion of Fig. 147.2) is modeled as a pool of elemental performance resources that are grouped into one of four different domains: (1) life-sustaining, (2) environmental interface (containing purely sensory and sensorimotor components), (3) central processing, and (4) information. Within each of the first three domains, physical subsystems referred to as functional units are identified (see labels along horizontal aspect of grids) through application of fairly rigorous criteria [Kondraske, 1990b]. GSPT is applied to each functional unit, yielding first a set of dimensions of performance (defined using a resource construct) for each unit. A single basic element of performance (BEP) is defined by specifying two items: (1) the basic functional unit and (2) one of its dimensions of performance. Within a domain, not every dimension of performance indicated in Fig. 147.2 is applicable to every functional unit in that domain. However, there is an increasing degree of “likeness” among functional units (i.e., fewer fundamentally different types) in this regard as one moves from life-sustaining to environmental-interface to central-processing domains. The fourth domain, the information domain, is substantially different than the other three. Whereas the first three represent physical systems and their intangible performance resources, the information domain simply represents information. Thus, while memory functional units are located within the central-processing domain, the contents of memory (e.g., motor programs and associated reference information) are partitioned into the information domain. As illustrated, information is grouped, but within each group there are many specific skills. The set of available performance resources [RAi,j(t)|Q] consist of both BEPs (i = dimension of performance, j = functional unit) and information sets (e.g., type i within group j). Although intrinsically different, both fit the resource construct. This approach permits even the most abstract items such as motivation and friendliness to be considered with the same basic framework as strength and speed. Note that resource availability in GSPT and thus in the ERM is potentially a function of time, allowing quantitative modeling of dynamic processes such as child development, aging, disuse at

Biomedical Engineering Handbook

Clinical Engineering Handbook (Biomedical Engineering)

Clinical Engineering Handbook (Biomedical Engineering)

Biomedical Information Technology (Biomedical Engineering)

Handbook of Neuroprosthetic Methods (Biomedical Engineering)

Standard Handbook of Biomedical Engineering and Design

Sensors, Nanoscience, Biomedical Engineering, and Instruments: Sensors Nanoscience Biomedical Engineering (The Electrical Engineering Handbook Third Edition)

Biomedical Engineering Fundamentals

APPLIED BIOMEDICAL ENGINEERING MECHANICS

Fundamentals of Biomedical Engineering

Introduction to Biomedical Engineering

Introduction to Biomedical Engineering

Introduction to biomedical engineering

Fundamentals of Biomedical Engineering

Biomedical Engineering Desk Reference

Advances in Biomedical Engineering

Biomedical Engineering for Global Health (Cambridge Texts in Biomedical Engineering)

Biomedical Engineering and Design Handbook, Volume 1: Second Edition, Biomedical Engineering Fundamentals

Biomedical Engineering and Design Handbook, Volume 2: 2nd Edition, Biomedical Engineering Applications

Handbook of Biomedical Fluorescence

Biomedical Photonics Handbook

Biomedical Photonics Handbook

Numerical Methods in Biomedical Engineering

Engineering Materials for Biomedical Applications

New Developments in Biomedical Engineering

Computational Intelligence in Biomedical Engineering

Engineering Materials for Biomedical Applications

Handbook of Medical Imaging: Processing and Analysis (Biomedical Engineering)

Biomedical Engineering & Design Handbook, Volumes I and II

Biomedical Applications of Hydrogels Handbook

The Biomedical Quality Auditor Handbook

Biomedical Engineering Handbook

Clinical Engineering Handbook (Biomedical Engineering)

Clinical Engineering Handbook (Biomedical Engineering)

Biomedical Information Technology (Biomedical Engineering)

Handbook of Neuroprosthetic Methods (Biomedical Engineering)

Standard Handbook of Biomedical Engineering and Design

Sensors, Nanoscience, Biomedical Engineering, and Instruments: Sensors Nanoscience Biomedical Engineering (The Electrical Engineering Handbook Third Edition)

Biomedical Engineering Fundamentals

APPLIED BIOMEDICAL ENGINEERING MECHANICS

Fundamentals of Biomedical Engineering

Introduction to Biomedical Engineering

Introduction to Biomedical Engineering

Introduction to biomedical engineering

Fundamentals of Biomedical Engineering

Biomedical Engineering Desk Reference

Advances in Biomedical Engineering

Biomedical Engineering for Global Health (Cambridge Texts in Biomedical Engineering)

Biomedical Engineering and Design Handbook, Volume 1: Second Edition, Biomedical Engineering Fundamentals

Biomedical Engineering and Design Handbook, Volume 2: 2nd Edition, Biomedical Engineering Applications

Handbook of Biomedical Fluorescence

Biomedical Photonics Handbook

Biomedical Photonics Handbook

Numerical Methods in Biomedical Engineering

Engineering Materials for Biomedical Applications

New Developments in Biomedical Engineering

Computational Intelligence in Biomedical Engineering

Engineering Materials for Biomedical Applications

Handbook of Medical Imaging: Processing and Analysis (Biomedical Engineering)

Biomedical Engineering & Design Handbook, Volumes I and II

Biomedical Applications of Hydrogels Handbook

The Biomedical Quality Auditor Handbook

Recommend Documents