IFMBE Proceedings Series Editor: R. Magjarevic
Volume 29
The International Federation for Medical and Biological Engineering, IFMBE, is a federation of national and transnational organizations representing internationally the interests of medical and biological engineering and sciences. The IFMBE is a non-profit organization fostering the creation, dissemination and application of medical and biological engineering knowledge and the management of technology for improved health and quality of life. Its activities include participation in the formulation of public policy and the dissemination of information through publications and forums. Within the field of medical, clinical, and biological engineering, IFMBE’s aims are to encourage research and the application of knowledge, and to disseminate information and promote collaboration. The objectives of the IFMBE are scientific, technological, literary, and educational. The IFMBE is a WHO accredited NGO covering the full range of biomedical and clinical engineering, healthcare, healthcare technology and management. It is representing through its 58 member societies some 120.000 professionals involved in the various issues of improved health and health care delivery. IFMBE Officers President: Makoto Kikuchi, Vice-President: Herbert Voigt, Former-President: Joachim H. Nagel Treasurer: Shankar M. Krishnan, Secretary-General: Ratko Magjarevic http://www.ifmbe.org
Previous Editions: IFMBE Proceedings MEDICON 2010, “XII Mediterranean Conference on Medical and Biological Engineering and Computing 2010”, Vol. 29, 2010, Chalkidiki, Greece, CD IFMBE Proceedings BIOMAG2010, “17th International Conference on Biomagnetism Advances in Biomagnetism – Biomag2010”, Vol. 28, 2010, Dubrovnik, Croatia, CD IFMBE Proceedings ICDBME 2010, “The Third International Conference on the Development of Biomedical Engineering in Vietnam”, Vol. 27, 2010, Ho Chi Minh City, Vietnam, CD IFMBE Proceedings MEDITECH 2009, “International Conference on Advancements of Medicine and Health Care through Technology”, Vol. 26, 2009, Cluj-Napoca, Romania, CD IFMBE Proceedings WC 2009, “World Congress on Medical Physics and Biomedical Engineering”, Vol. 25, 2009, Munich, Germany, CD IFMBE Proceedings SBEC 2009, “25th Southern Biomedical Engineering Conference 2009”, Vol. 24, 2009, Miami, FL, USA, CD IFMBE Proceedings ICBME 2008, “13th International Conference on Biomedical Engineering” Vol. 23, 2008, Singapore, CD IFMBE Proceedings ECIFMBE 2008 “4th European Conference of the International Federation for Medical and Biological Engineering”, Vol. 22, 2008, Antwerp, Belgium, CD IFMBE Proceedings BIOMED 2008 “4th Kuala Lumpur International Conference on Biomedical Engineering”, Vol. 21, 2008, Kuala Lumpur, Malaysia, CD IFMBE Proceedings NBC 2008 “14th Nordic-Baltic Conference on Biomedical Engineering and Medical Physics”, Vol. 20, 2008, Riga, Latvia, CD IFMBE Proceedings APCMBE 2008 “7th Asian-Pacific Conference on Medical and Biological Engineering”, Vol. 19, 2008, Beijing, China, CD IFMBE Proceedings CLAIB 2007 “IV Latin American Congress on Biomedical Engineering 2007, Bioengineering Solution for Latin America Health”, Vol. 18, 2007, Margarita Island, Venezuela, CD IFMBE Proceedings ICEBI 2007 “13th International Conference on Electrical Bioimpedance and the 8th Conference on Electrical Impedance Tomography”, Vol. 17, 2007, Graz, Austria, CD IFMBE Proceedings MEDICON 2007 “11th Mediterranean Conference on Medical and Biological Engineering and Computing 2007”, Vol. 16, 2007, Ljubljana, Slovenia, CD IFMBE Proceedings BIOMED 2006 “Kuala Lumpur International Conference on Biomedical Engineering”, Vol. 15, 2004, Kuala Lumpur, Malaysia, CD IFMBE Proceedings WC 2006 “World Congress on Medical Physics and Biomedical Engineering”, Vol. 14, 2006, Seoul, Korea, DVD IFMBE Proceedings BSN 2007 “4th International Workshop on Wearable and Implantable Body Sensor Networks”, Vol. 13, 2006, Aachen, Germany
IFMBE Proceedings Vol. 29 Panagiotis D. Bamidis • Nicolas Pallikarakis (Eds.)
XII Mediterranean Conference on Medical and Biological Engineering and Computing 2010 May 27 – 30, 2010 Chalkidiki, Greece
123
Editors Panagiotis D. Bamidis Aristotle University of Thessaloniki Lab of Medical Informatics, Medical Scho 541 24 Thessaloniki Greece E-mail:
[email protected] Nicolas Pallikarakis University of Patras Biomedical Technology Unit 265 04 Rio Patras Greece E-mail:
[email protected] ISSN 1680-0737 ISBN 978-3-642-13038-0
e-ISBN 978-3-642-13039-7
DOI 10.1007/978-3-642-13039-7 Library of Congress Control Number: 2010927933 © International Federation for Medical and Biological Engineering 2010 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permissions for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The IFMBE Proceedings is an Official Publication of the International Federation for Medical and Biological Engineering (IFMBE) Typesetting: Scientific Publishing Services Pvt. Ltd., Chennai, India. Cover Design: deblik, Berlin Printed on acid-free paper 987654321 springer.com
Preface
Over the past three decades, the exploding number of new technologies and applications introduced in medical practice, often powered by advances in biosignal processing and biomedical imaging, created an amazing account of new possibilities for diagnosis and therapy, but also raised major questions of appropriateness and safety. The accelerated development in this field, alongside with the promotion of electronic health care solutions, is often on the basis of an uncontrolled diffusion and use of medical technology. The emergence and use of medical devices is multiplied rapidly and today there exist more than one million different products available on the world market. Despite the fact that the rising cost of health care, partly resulting from the new emerging technological applications, forms the most serious and urgent problem for many governments today, another important concern is that of patient safety and user protection, issues that should never be compromised and expelled from the Biomedical Engineering research practice agenda. Consequently, the associated field of Biomedical Engineering education – and its related branches – is undergoing a rapid evolution characterized by an increasing degree of specialization. This imposes new challenges, while the changing scene at the European level dictates the need for harmonization and standardization of education, with a focus on meeting the emerging needs for appropriately trained young engineers within this new landscape. Therefore, MEDICON 2010 is dedicated to all those young Biomedical Engineers who have to face this very demanding environment and are expected to play a key role in bringing the benefits of this technological evolution right to the patient. The 12th Mediterranean Conference on Medical and Biological Engineering and Computing - MEDICON 2010 has been supported by a limited number of external sponsors to which we would like to express our appreciation. However, this Conference would not have been possible to organize and prove successful without the valuable contribution of the many volunteers involved. Particularly we would like to express our deep thanks to the invited speakers for accepting to come to Chalkidiki on their own expenses and deliver their excellent presentations. It is also a great pleasure for us to acknowledge the dedicated and hard work done by all the members of the International Scientific as well as the Programme Committees, but also the many volunteering reviewers, that performed as a real team, thereby rendering the rigorous reviewing process a quality filter for an indeed noteworthy proceedings publication. All these have created an enthusiastic and challenging atmosphere that has all the necessary ingredients to make MEDICON2010 an unforgettable event. We hope you will appreciate this Proceedings volume as much as we are proud of it!
Nicolas Pallikarakis Panagiotis D. Bamidis MEDICON 2010 Co-chairs
Committees
Conference Chairs Nicolas Pallikarakis Panagiotis D. Bamidis
University of Patras, Greece Aristotle University of Thessaloniki, Greece
Local Organising Committee Chair Costas Pappas
Aristotle University of Thessaloniki, Greece
Local Organising Committee Stavros Panas George Sergiadis Dimitris Koufogiannis Nicholas Dombros
Aristotle University of Thessaloniki, Greece Aristotle University of Thessaloniki, Greece Aristotle University of Thessaloniki, Greece Aristotle University of Thessaloniki, Greece
International Scientific Committee Chairs Ratko Magjarevic Marcello Bracale
University of Zagreb, Croatia University "Federico II" of Naples, Italy
International Scientific Committee James C. H. Goh, Singapore Fernando A. Infantosi, Brazil Akos Jobbagy, Hungary Makoto Kikuchi, Japan Shankar M. Krishnan, USA Luis Kun, USA Mario Medvedec, Croatia Alan Murray, United Kingdom Patrick Pentony, Ireland Niilo Saranummi, Finland Jos A.E. Spaan, The Netherlands Herbert Voigt, USA
Program Committee Chairs Dimitris Koutsouris Nicos Maglaveras
National Technical University of Athens, Greece Aristotle University of Thessaloniki, Greece
VIII
Committees
Program Committee Pantelis Angelidis, Greece Maria Teresa Areddondo, Spain Theo Arvanitis, UK Alexandros Astaras, Greece Joe Barbenel, UK Catherine Chronaki, Greece Asuman Dogac, Tourkey Olaf Dossel, Germany Barry Eaglestone, UK David Elad, Israel Simon G. Fabri, Malta Dimitris Fotiadis, Greece Frederique Frouin, France Demetrius Georgiou, Greece Leontios Hadjileontiadis, Greece Maria Haritou, Greece Jens Haueisen, Germany Jiri Holcik, Czech Republic Andreas Ioannides, Cyprus Eleni Kaldoudi,Greece Panagiotis Ketikidis, Greece Stathis Konstantinidis, Greece Vassilis Koutkias, Greece Periklis Ktonas, Greece Efthyvoulos Kyriakou,Cyprus Igor Lackovic, Croatia Olof Lindahl, Sweden Dimitris Lymperopoulos, Greece Ilias Maglogiannis, Greece Nadia Magnenat-Thalmann, Switcherland Fillia Makedon, USA Andigoni Malousi, Greece Vicky Manthou, Greece Cristina Mazzoleni , Italy Damijan Miklavcic, Slovenia Joe Mizrahi, Israel
Conference Tracks and Chairs 1. Medical Devices & Instrumentation Chairs Alexandros Astaras Andreas Lymberis 2. Education Chairs Eleni Kaldoudi Daniela Giordano Stathis Konstantinidis
Zhivko Bliznakov, Greece Kristina Bliznakova, Greece Ivan Buliev, Boulgaria Babis Bratsas, Greece Ioanna Chouvarda, Greece Per Moeller, Denmark Robert Allen, UK Roger Moore, UK Konstantina Nikita, Greece Marc Nyssen, Belgium Christos Papadelis, Italy Iraklis Paraskakis, Greece Constantinos Pattichis,Cyprus Sotiris Pavlopoulos, Greece Thomas Penzel, Germany Ioannis Pitas, Greece Andriana Prentza, Greece Gunter Rau, Germany Georgios Sakas, Germany Laura Roa, Spain Abdul Roudsari, UK Göran Salerud, Sweden Theodoros Samaras, Greece Mario Sansone, Italy Andres Santos, Spain Christos Schizas, Cyprus Mario Forjaz Secca, Portugal Maria Siebes, The Netherlands Stella Spyrou, Greece Rita Stagni, Italy Gregory-Telemachos Stamkopoulos, Greece Selma Supek, Croatia Panayiotis Tsanakas, Greece Jan Wojcicki, Poland Michalis Zervakis, Greece
Committees
IX
3. Biomedical Imaging Chairs Kristina Bliznakova Ilias Maglogiannis 4. Biosignal Processing Chair Leontios Hadjileontiadis Christos Papadelis 5. Clinical Engineering and Safety Chairs Yadin David Jorge Calil Saide Zhivko Bliznakov 6. E-health Chairs Constantinos Pattichis Efthyvoulos Kyriakou 7. Workshop on BME and MP Education: Current Trends in Europe Chairs Slavik Tabakov Nicolas Pallikarakis
List of Reviewers Mansour Ahmadian Robert Allen Christos-Nikolaos Anagnostopoulos Pantelis Angelidis Antonis Antoniadis Theo Arvanitis Sara Assecondi Alexander Astaras Nizamettin Aydin Branko Babusiak Joe Barbenel Ofer Barnea Katarzyna Blinowska Zhivko Bliznakov Kristina Bliznakova Francesca Bovolo Marcello Bracale Charalampos Bratsas Christoph Braun Maide Bucolo Ivan Buliev Enrico Caiani Giovanni Calcagnini Martin Cerny Aristotelis Chatziioannou Giannis Chatzizisis
Michela Chiappalone Ki-H Chon Ioanna Chouvarda Christodoulos Christodoulou Catherine Chronaki Radu Ciupa Eleni Costaridou Hariton Costin Paul Cristea Eleni Dafli Andriani Daskalaki Kostas Delibasis Gianpaolo Demarchi Aris Dermitzakis Fabrizio De-Vico-Fallani Asuman Dogac Zlatica Dolna Olaf Dossel Charalampos Doukas Barry Eaglestone David Elad George Eleftherakis Miroslawa El-Fray Silvia Erla Simon Fabri Luca Faes
X
Silvia Fantozzi Jocelyne Fayn Mario Forjaz-Secca Dimitrios Fotiadis Christos Frantzidis Monique Frize Michal Gala Demetrius Georgiou Daniela Giordano James Goh Leontios Hadjileontiadis Karel Hana Maria Haritou Jens Haueisen Jan Havlik Martin HoÃbach Jiri Holcik Dimitris Iakovidis Adam Idzkowski Antonio-Fernando Infantosi Andreas Ioannides Robert Istepanian Sriram Iyengar Akos Jobbagy Erik Johannessen Vaggelis Kaimakamis Eleni Kaldoudi Anna Karahaliou Eleni Kargioti Kostas Karpouzis Spyros Kitsiou Manos Klados Athina Kokonozi Antonios Komnidis Evdokimos Konstantinidis Stathis Konstantinidis Dimitrios Kosmopoulos Sophia Kossida Kostas Kostopoulos Vassilis Koutkias Periklis Ktonas Dinesh Kumar Efthyvoulos Kyriacou Igor Lackovic Philip Langley Nikos Laskaris Olof Lindahl Angelika Lingnau Chrysa Lithari Andrej Luneski Andreas Lymberis Ilias Maglogiannis
Committees
Nadia Magnenat-Thalmann Fillia Makedon Andigoni Malousi Vicky Manthou Mattia Marconcini Ioannis Mariolis Michela Masa¨ Georgios Matis Giulia Matrone Veronica Mazza Cristina Mazzoleni Mario Medvedec Damijan Miklavcic Joe Mizrahi Per Moeller Mihaela Morega Antonis Mpillis Brian-Edmond Murphy Alan Murray Tuncay Namli Jila Nazari Marios Neofytou Kleanthis Neokleous Konstantina Nikita Maria Nikolaidou Giandomenico Nollo Leonidas Orfanidis Cristina Oyarzun-Laura Christos Papadelis Kostas Papathanasiou Iraklis Paraskakis Ana Pascoal Constantinos Pattichis Leandro Pecchia Marek Penhaker Krzysztof Penkala Patrick Pentony Thomas Penzel Francesca Pizzorni-Ferrarese Vassilis Plagianakos Vahe Poghosyan Senan Postaci Eugene Postnikov Deborah Prà Michal Prauzek Andriana Prentza Efi Psarouli Chiara Rabotti Dan Rafiroiu Gunter Rau Ulrike Richter Jose-Joaquin Rieta
Committees
Laura Roa Georgios Sakas Goran Salerud Mario Sansone Andres Santos Niilo Saranummi Roberto Sassi Christos Schizas Maurizio Schmid Nicu Sebe George Sergiadis Maria Siebes Pavel Smrcka Delia Soimu Tomasz Soltysinski Jos Spaan Stergiani Spyrou Pascal Staccini Rita Stagni Telemachos Stamkopoulos Stavros Stavrinidis Sebastian Steger
XI
Milan Stork Charalampos Styliadis Selma Supek Slavik Tabakov Luigi Tame Tong-Boon Tang Lucio Tommaso-De-Paolis Lubomir Traikov Riccardo Tranfaglia Ioannis Tsamardinos Styliani Tsigka Aristeidis Vaggelatos Emil Valchinov Teena Vellaramkalayil Vassilios Vescoukis Andreas Voss Wojtek Walendziuk Marta Wasilewska-Radwanska Jozef Wiora Jan Wojcicki Mustafa Yuksel Michalis Zervakis
This page intentionally left blank
Table of Contents
Biosignal Processing Quantitative Analysis of Two-Dimensional Catch-Up Saccades Executed to the Target Jumps in the Time-Continuous Trajectory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vincas Laurutis, Raimondas Zemblys
1
Spike Sorting Based on Dominant-Sets Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D.A. Adamos, N.A. Laskaris, E.K. Kosmidis, G. Theophilidis
5
Differentiation of Human Bone Marrow Stromal Cells onto Gelatin Cryogel Scaffolds . . . . . . . . . L. Fassina, E. Saino, L. Visai, M.A. Avanzini, M.G. Cusella De Angelis, F. Benazzo, S. Van Vlierberghe, P. Dubruel, G. Magenes
9
Simples Coherence vs. Multiple Coherence: A Somatosensory Evoked Response Detection Investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D.B. Melges, A.M.F.L. Miranda de S´ a, A.F.C. Infantosi Measure of Similarity of ECG Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ´ Jobb´ ´ Nagy A. agy, A. Wavelet Phase Synchronization between EHGs at different Uterine Sites: Comparison of Pregnancy and Labor Contractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ´ Alexandersson, J. Terrien, B. Karlsson, C. Marque M. Hassan, A. Dynamic Generation of Physiological Model Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Kretschmer, A. Wahl, K. Moeller
13
17
21
25
Random Forest-Based Classification of Heart Rate Variability Signals by Using Combinations of Linear and Nonlinear Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alan Jovic, Nikola Bogunovic
29
Validation of MRS Metabolic Markers in the Classification of Brain Gliomas and Their Correlation to Energy Metabolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M.G. Kounelakis, M.E. Zervakis, G.J. Postma, L.M.C. Buydens, A. Heerschap, X. Kotsiakis
33
Event-Related Synchronization/Desynchronization for Evaluating Cortical Response Detection Induced by Dynamic Visual Stimuli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P.J.G. Da-Silva, A.F.C. Infantosi, J. Nadal
37
Investigating the EEG Alpha Band during Kinesthetic and Visual Motor Imagery of the Spike Volleyball Movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M.V. Stecklow, M. Cagy, A.F.C. Infantosi
41
Principal Components Clustering through a Variance-Defined Metric . . . . . . . . . . . . . . . . . . . . . . . . . J.C.G.D. Costa, D.B. Melges, R.M.V.R. Almeida, A.F.C. Infantosi
45
XIV
Table of Contents
A Kurtosis-Based Automatic System Using Na¨ıve Bayesian Classifier to Identify ICA Components Contaminated by EOG or ECG Artifacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M.A. Klados, C. Bratsas, C. Frantzidis, C.L. Papadelis, P.D. Bamidis Correlation between Fractal Behavior of HRV and Neurohormonal and Functional Indexes in Chronic Heart Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G. D’ Addio, M. Cesarelli, M. Romano, A. Accardo, G. Corbi, R. Maestri, M.T. La Rovere, Paolo Bifulco, N. Ferrara, F. Rengo On the Selection of Time Interval and Frequency Range of EEG Signal Preprocessing for P300 Brain-Computer Interfacing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N.V. Manyakov, N. Chumerin, A. Combaz, M.M. Van Hulle Development of a Simple and Cheap Device for Movement Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . Csan´ ad G. Erd˝ os, Gerg˝ o Farkas, B´ela Pataki
49
53
57 61
Signal Peptide Prediction in Single Transmembrane Proteins Using the Continuous Wavelet Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I.A. Avramidou, I.K. Kitsas, L.J. Hadjileontiadis
65
Comparison of AM-FM Features with Standard Features for the Classification of Surface Electromyographic Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C.I. Christodoulou, P.A. Kaplanis, V. Murray, M.S. Pattichis, C.S. Pattichis
69
Studying Brain Visuo-Tactile Integration through Cross-Spectral Analysis of Human MEG Recordings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S. Erla, C. Papadelis, L. Faes, C. Braun, G. Nollo
73
Patient-Specific Seizure Prediction Using a Multi-feature and Multi-modal EEG-ECG Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Valderrama, S. Nikolopoulos, C. Adam, Vincent Navarro, M. Le Van Quyen
77
Horizontal Directionality Characteristics of the Bat Head-Related Transfer Function . . . . . . . . . S.Y. Kim, D. Nikoli´c, A.C. Meruelo, R. Allen
81
Assessment of Human Performance during High-Speed Marine Craft Transit . . . . . . . . . . . . . . . . . D. Nikoli´c, R. Collier, R. Allen
85
Effects of Electrochemotherapy on Microcirculatory Vasomotion in Tumors . . . . . . . . . . . . . . . . . . . T. Jarm, B. Cugmas, M. Cemazar
89
Non-linear Modeling of Cerebral Autoregulation Using Cascade Models . . . . . . . . . . . . . . . . . . . . . . N.C. Angarita-Jaimes, O.P. Dewhirst, D.M. Simpson
93
The Epsilon-Skew-Normal Dictionary for the Decomposition of Single- and Multichannel Biomedical Recordings Using Matching Pursuit Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Strohmeier, A. Halbleib, M. Gratkowski, J. Haueisen
97
On the Empirical Mode Decomposition Performance in White Gaussian Noise Biomedical Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Karagiannis, Ph. Constantinou
101
Table of Contents
XV
Simulation of Biomechanical Experiments in OpenSim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I. Symeonidis, G. Kavadarli, E. Schuller, S. Peldschus
107
Comparing Sensorimotor Cortex Activation during Actual and Imaginary Movement . . . . . . . . . A. Athanasiou, E. Chatzitheodorou, K. Kalogianni, C. Lithari, I. Moulos, P.D. Bamidis
111
Graph Analysis on Functional Connectivity Networks during an Emotional Paradigm . . . . . . . . C. Lithari, M.A. Klados, P.D. Bamidis
115
MORFEAS: A Non-Invasive System for Automated Sleep Apnea Detection Utilizing Snore Sound Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Charalampos Doukas, Theodoros Petsatodis, Ilias Maglogiannis
119
Improved Optical Method for Measuring Concentration of Uric Acid Removed during Dialysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Jerotskaja, F. Uhlin, M. Luman, K. Lauri, I. Fridolin
124
Correlations between Longitudinal Corneal Apex Displacement, Head Movements and Pulsatile Blood Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Danielewska, H. Kasprzak, M. Kowalska
128
The Analog Processing and Digital Recording of Electrophysiological Signals . . . . . . . . . . . . . . . . . F. Babarada, J. Arhip, C. Ravariu
132
Parameter Selection in Approximate and Sample Entropy-Complexity of Acute and Chronic Stress Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . T. Loncar Turukalo, O. Sarenac, N. Japundzic-Zigon, D. Bajic
136
The Importance of Uterine Contractions Extraction in Evaluation of the Progress of Labour by Calculating the Values of Sample Entropy from Uterine Electromyogram . . . . . . . . . . . . . . . . . . J. Vrhovec, D. Rudel, A. Macek Lebar
140
Simultaneous Pneumo- and Photoplethysmographic Recording of Oscillometric Envelopes Applying a Local Pad-Type Cuff on the Radial Artery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . R. Raamat, K. Jagom¨ agi, J. Talts, J. Kivastik
144
Estimation of Mean Radial Blood Pressure in Critically Ill Patients . . . . . . . . . . . . . . . . . . . . . . . . . . K. Jagom¨ agi, J. Talts, P. T¨ ahep˜ old, R. Raamat, J. Kivastik Photoplethysmographic Assessment of the Pressure-Compliance Relationship for the Radial Artery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Talts, R. Raamat, K. Jagom¨ agi, J. Kivastik High Frequency Acoustic Properties for Cutaneous Cell Carcinomas In Vitro . . . . . . . . . . . . . . . . . L.I. Petrella, W.C.A. Pereira, P.R. Issa, H.A. Valle, C.J. Martins, J.C. Machado Gender-Related Effects of Carbohydrate Ingestion and Hypoxia on Heart Rate Variability: Linear and Non-linear Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . T. Princi, M. Klemenc, P. Golja, A. Accardo On the Analysis of Dynamic Lung Mechanics Separately in Ins- and Expiration . . . . . . . . . . . . . . K. M¨ oller, Z. Zhao, C.A. Stahl, J. Guttmann
148
152 156
160 164
XVI
Table of Contents
Clinical Validation of an Algorithm for Automatic Detection of Atrial Fibrillation from Single Lead ECG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Triventi, G. Calcagnini, F. Censi, E. Mattei, F. Mele, P. Bartolini
168
Mental and Motor Task Classification by LDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. Gursel Ozmen, L. Gumusel
172
The Human Subthalamic Nucleus – Knowledge for the Understanding of Parkinson’s Disease T. Heida, E. Marani
176
Dissociated Neurons from an Extended Rat Subthalamic Area - Spontaneous Activity and Acetylcholine Addition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . T. Heida, E. Marani
180
Nigro-Subthalamic and Nigro-Trigeminal Projections in the Rat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Marani, N.E. Lazarov, T. Heida, K.G. Usunoff
184
Statistical Estimate on Indices Associated to Atherosclerosis Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . C.M. Ipate, A. Machedon, M. Morega
188
Study of Some EEG Signal Processing Methods for Detection of Epileptic Activity . . . . . . . . . . . R. Matei, D. Matei
192
Continuous Wavelet Transformation of Pattern Electroretinogram (PERG) - A Tool Improving the Test Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . K. Penkala
196
An Interactive Tool for Customizing Clinical Transacranial Magnetic Stimulation (TMS) Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Faro, D. Giordano, I. Kavasidis, C. Pino, C. Spampinato, M.G. Cantone, G. Lanza, M. Pennisi
200
Medical Imaging Measurement Methodology for Tempomandibular Joint Displacement Based on Focus Mutual Information Alignment of CBCT Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . W. Jacquet, E. Nyssen, B. Vande Vannet
204
Computer Aided Diagnosis of Diffuse Lung Disease in Multi-detector CT – Selecting 3D Texture Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I. Mariolis, P. Korfiatis, C. Kalogeropoulou, D. Daoussis, T. Petsas, L. Costaridou
208
Statistical Pre-processing Method for Peripheral Quantitative Computed Tomography Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . T. Cervinka, H. Sievanen, M. Hannula, J. Hyttinen
212
Security and Reliability of Data Transmissions in Biotelemetric System . . . . . . . . . . . . . . . . . . . . . . . M. Stankus, M. Penhaker, V. Srovnal, M. Cerny, V. Kasik A Novel Approach for Implementation of Dual Energy Mapping Technique in CT-Based Attenuation Correction Using Single kVP Imaging: A Feasibility Study . . . . . . . . . . . . . . . . . . . . . . . B. Teimourian, M.R. Ay, H. Ghadiri, M. Shamsaei Zafarghandi, H. Zaidi
216
220
Table of Contents
XVII
Computational Visualization of Tumor Virotherapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X.F. Gao, M. Tangney, S. Tabirca
224
New Approaches for Continuous Non Invasive Blood Pressure Monitoring . . . . . . . . . . . . . . . . . . . . Petr Zurek, Martin Cerny, Michal Prauzek, Ondrej Krejcar, Marek Penhaker
228
Wireless Power and Data Transmission for Robotic Endoscopic Capsules . . . . . . . . . . . . . . . . . . . . . R. Carta, J. Thon´e, R. Puers
232
Ulcer Detection in Wireless Capsule Endoscopy Images Using Bidimensional Nonlinear Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vasileios Charisis, Alexandra Tsiligiri, Leontios J. Hadjileontiadis, Christos N. Liatsos, Christos C. Mavrogiannis, George D. Sergiadis Pre-clinical Physiological Data Acquisition and Testing of the IMAGE Sensing Device for Exercise Guidance and Real-Time Monitoring of Cardiovascular Disease Patients . . . . . . . . . . . . . A. Astaras, A. Kokonozi, E. Michail, D. Filos, I. Chouvarda, O. Grossenbacher, J.-M. Koller, R. Leopoldo, J.-A. Porchet, M. Correvon, J. Luprano, A. Sipil¨ a, N. Maglaveras Thermal Images of Electrically Stimulated Breast: A Simulation Study . . . . . . . . . . . . . . . . . . . . . . . H. Feza Carlak, Nevzat G. Gen¸cer, Cengiz Be¸sik¸ci Magnetic Resonance Current Density Imaging Using One Component of Magnetic Flux Density: An Experimental Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Ers¨ oz, B.M. Ey¨ ubo˘glu
236
240
244
248
Computer-Aided Detection of COPD Using Digital Chest Radiographs . . . . . . . . . . . . . . . . . . . . . . . ´ Horv´ L. Nikh´ azy, G. Horv´ ath, A. ath, V. M¨ uller
252
Localisation, Registration and Visualisation of MRS Volumes of Interest on MR Images . . . . . . Yu Sun, Nigel P. Davies, Kal Natarajan, Theodoros N. Arvanitis, Andrew C. Peet
256
Magnetic Marker Monitoring: A Novel Approach for Magnetic Marker Design . . . . . . . . . . . . . . . S. Biller, D. Baumgarten, J. Haueisen
260
Corneal Nerves Segmentation and Morphometric Parameters Quantification for Early Detection of Diabetic Neuropathy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ana Ferreira, Ant´ onio Miguel Morgado, Jos´e Silvestre Silva Novel Catheters for In Vivo Research and Pharmaceutical Trials Providing Direct Access to Extracellular Space of Target Tissues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Bodenlenz, C. Hoefferer, F. Feichtner, C. Magnes, R. Schaller, J. Priedl, T. Birngruber, F. Sinner, L. Schaupp, S. Korsatko, T.R. Pieber
264
268
Statistical Texture Analysis of MRI Images to Classify Patients Affected by Multiple Sclerosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Faro, D. Giordano, C. Spampinato, M. Pennisi
272
WADEDA: A Wearable Affective Device with On-Chip Signal Processing Capabilities for Measuring ElectroDermal Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E.I. Konstantinidis, C.A. Frantzidis, C. Papadelis, C. Pappas, P.D. Bamidis
276
XVIII
Table of Contents
A Modular Architecture of a Computer-Operated Olfactometer for Universal Use . . . . . . . . . . . . A. Komnidis, E. Konstantinidis, I. Stylianou, M.A. Klados, A. Kalfas, P.D. Bamidis The Role of Geometry of the Human Carotid Bifurcation in the Formation and Development of Atherosclerotic Plaque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P.G. Kalozoumis, A.I. Kalfas, A.D. Giannoukas
280
284
A Wearable Wireless ECG Sensor: A Design with a Minimal Number of Parts . . . . . . . . . . . . . . . E.S. Valchinov, N.E. Pallikarakis
288
Active Contours without Edges Applied to Breast Lesions on Ultrasound . . . . . . . . . . . . . . . . . . . . . W. G´ omez, A.F.C. Infantosi, L. Leija, W.C.A. Pereira
292
Automatic Identification of Trabecular Bone Fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S. Tassani, P.A. Asvestas, G.K. Matsopoulos, F. Baruffaldi
296
Visualization System to Improve Surgical Performance during a Laparoscopic Procedure . . . . . L.T. De Paolis, M. Pulimeno, G. Aloisio
300
The Blood Perfusion Mapping in the Human Skin by Photoplethysmography Imaging . . . . . . . . U. Rubins, R. Erts, V. Nikiforovs
304
Fingerprint Matching with Self Organizing Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.N. Ouzounoglou, T.L. Economopoulos, P.A. Asvestas, G.K. Matsopoulos
307
A Novel Model for Monte Carlo Simulation of Performance Parameters of the Rodent Research PET (RRPET) Camera Based on NEMA NU-4 Standards . . . . . . . . . . . . . . . . . . . . . . . . . . N. Zeraatkar, M.R. Ay, A.R. Kamali-Asl, H. Zaidi
311
Is the Average Gray-Level from Ultrasound B-Mode Images Able to Estimate Temperature Variations in Ex-Vivo Tissue? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C´esar A. Teixeira, A.V. Alvarenga, M.A. von Kr¨ uger, W.C.A. Pereira
315
CT2MCNP: An Integrated Package for Constructing Patient-Specific Voxel-Based Phantoms Dedicated for MCNP(X) Monte Carlo Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Mehranian, M.R. Ay, H. Zaidi
319
Noise Reduction in Fluoroscopic Image Sequences for Joint Kinematics Analysis . . . . . . . . . . . . . T. Cerciello, P. Bifulco, M. Cesarelli, L. Paura, M. Romano, G. Pasquariello, R. Allen
323
The Influence of Patient Miscentering on Patient Dose and Image Noise in Two Commercial ct Scanners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M.A. Habibzadeh, M.R. Ay, A.R. Kamali asl, H. Ghadiri, H. Zaidi
327
A Study on Performance of a Digital Image Acquisition System in Mammography Diagnostic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Dimitric, G. Nisevic, Z. Boskovic, A. Vasic
331
An Efficient Video-Synopsis Technique for Optical Recordings with Application to the Analysis of Rat Barrel-Cortex Responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Tsitlakidis, N.A. Laskaris, G.C. Koudounis, E.K. Kosmidis
335
Table of Contents
XIX
Preoperative Planning Software for Hip Replacement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ˇ c´ M. Michal´ıkov´ a, L. Bednarˇc´ıkov´ a, T. T´ oth, J. Zivˇ ak
339
Entropy: A Way to Quantify Complexity in Calcium Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Fanelli, F. Esposti, J. Ion Titapiccolo, M.G. Signorini
343
A New Fluorescence Image-Processing Method to Visualize Ca2+ - Release and Uptake Endoplasmatic Reticulum Microdomains in Cultured Glia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Ion Titapiccolo, F. Esposti, A. Fanelli, M.G. Signorini
347
Experimental Measurement of Modulation Transfer Function (MTF) in Five Commercial CT Scanners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S.M. Akbari, M.R. Ay, A.R. Kamali asl, H. Ghadiri, H. Zaidi
351
Microcalcifications Segmentation Procedure Based on Morphological Operators and Histogram Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M.A. Duarte, A.V. Alvarenga, C.M. Azevedo, A.F.C. Infantosi, W.C.A. Pereira
355
Segmentation of Anatomical Structures on Chest Radiographs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ´ Horv´ ´ Horv´ S. Juh´ asz, A. ath, L. Nikh´ azy, G. Horv´ ath, A. ath
359
Lung Nodule Detection on Rib Eliminated Radiographs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ´ Horv´ G. Orb´ an, A. ath, G. Horv´ ath
363
An Improved Algorithm for Out-of-Plane Artifacts Removal in Digital Tomosynthesis Reconstructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . K. Bliznakova, Z. Bliznakov, N. Pallikarakis
367
Magnetic Resonance Imaging of Irreversible Electroporation in Tubers . . . . . . . . . . . . . . . . . . . . . . . Mohammad Hjouj and Boris Rubinsky
371
Superposition of Activations of SWI and fMRI Acquisitions of the Motor Cortex . . . . . . . . . . . . . M. Matos, M. Forjaz Secca, M. Noseworthy
376
Medical Devices and Instrumentation A New Optical Method for Measuring Creatinine Concentration during Dialysis . . . . . . . . . . . . . . I. Fridolin, J. Jerotskaja, K. Lauri, F. Uhlin, M. Luman
379
The Role of Viscous Damping on Quality of Haptic Interaction in Upper Limb Rehabilitation Robot: A Simulation Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Oblak, I. Cikajlo, T. Keller, J.C. Perrry, J. Veneman, Z. Matjaˇci´c
383
A New Fibre Optic Pulse Oximeter Probe for Monitoring Splanchnic Organ Arterial Blood Oxygen Saturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Hickey, N. Samuels, N. Randive, R. Langford, P.A. Kyriacou
387
Electrical Properties of Teeth Regarding the Electric Vitality Testing . . . . . . . . . . . . . . . . . . . . . . . . T. Marjanovi´c, Z. Stare, M. Ranilovi´c
391
Stiffness of a Small Tissue Phantom Measured by a Tactile Resonance Sensor . . . . . . . . . . . . . . . . V. Jalkanen, B.M. Andersson, O.A. Lindahl
395
XX
Table of Contents
Vectorial Magnetoencephalographic Measurements for the Estimation of Radial Dipolar Activity in the Human Somatosensory System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Haueisen, K. Fleissig, D. Strohmeier, R. Huonker, M. Liehr, O.W. Witte
399
Registration of Chest X-Rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Csorba, B. Kormanyos, B. Pataki
402
Arterial Pulse Transit Time Dependence on Applied Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . K. Pilt, K. Meigas, M. Viigimaa, J. Kaik, R. Kattai, D. Karai
406
Influence of an Artificial Valve Type on the Flow in the Ventricular Assist Device . . . . . . . . . . . . D. Obidowski, P. Klosinski, P. Reorowicz, K. Jozwik
410
A New Stimulation Technique for Electrophysiological Color Vision Testing . . . . . . . . . . . . . . . . . . M. Zaleski, K. Penkala
414
Novel TiN-Based Dry EEG Electrodes: Influence of Electrode Shape and Number on Contact Impedance and Signal Quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P. Fiedler, S. Brodkorb, C. Fonseca, F. Vaz, F. Zanow, J. Haueisen
418
A Finite Element Method Study of the Current Density Distribution in a Capacitive Intrabody Communication System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ˇ Luˇcev, A. Koriˇcan, M. Cifrek Z.
422
Voice Controlled Neuroprosthesis System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D.C. Irimia, M.S. Poboroniuc, M.C. Stefan, Gh. Livint
426
Preoperative Planning Program Tool in Treatment of Articular Fractures: Process of Segmentation Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Tomazevic, D. Kreuh, A. Kristan, V. Puketa, M. Cimerman
430
Neuroimaging of Emotional Activation: Issues on Experimental Methodology, Analysis and Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Styliadis, C. Papadelis, P.D. Bamidis
434
Using Grid Infrastructure for the Promotion of Biomedical Knowledge Mining . . . . . . . . . . . . . . . A. Chatziioannou, I. Kanaris, C. Doukas, Ilias Maglogiannis A Laboratory Scale Facility for the Parametric Characterization of the Intraocular Pressure of the Human Eye . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.V. Michailidou, P. Chatzi, P.G. Kalozoumis, A.I. Kalfas, M. Pappa, I. Tsiafis, E.I. Konstantinidis, P.D. Bamidis
438
442
AM-FM Texture Image Analysis in Multiple Sclerosis Brain White Matter Lesions . . . . . . . . . . . C.P. Loizou, V. Murray, M.S. Pattichis, M. Pantziaris, I. Seimenis, C.S. Pattichis
446
Reliable Hysteroscopy Color Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I. Constantinou, V. Tanos, M. Neofytou, C. Pattichis
450
Comparison of Methods of Measurement of Head Position in Neurological Practice . . . . . . . . . . . P. Kutilek, J. Charfreitag, J. Hozman
455
Table of Contents
XXI
The Nanopous Al2O3 Material Used for the Enzyme Entrapping in a Glucose Biosensor . . . . . C. Ravariu, A. Popescu, C. Podaru, E. Manea, F. Babarada
459
Hand-Held Resonance Sensor Instrument for Soft Tissue Stiffness Measurements – A First Study on Biological Tissue In Vitro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Jalkanen, O.A. Lindahl
463
Head Position Monitoring System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P. Cech, J. Dlouhy, M. Cizek, J. Rozman, I. Vicha
467
Short Range Wireless Link for Data Acquisition in Medical Equipment . . . . . . . . . . . . . . . . . . . . . . . N.M. Roman, S. Gergely, R.V. Ciupa, M.V. Pusca
471
Corneal Quantitative Fluorometry – A Slit-Lamp Based Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . J.P. Domingues, Isa Branco, A.M. Morgado
475
Automatic Detection of Patients’ Spontaneous Activity during Pressure Support Ventilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G. Matrone, F. Mojoli, A. Orlando, A. Braschi, G. Magenes
479
Determination of In Vivo Three-Dimensional Lower Limb Kinematics for Simulation of High-Flexion Squats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P.D. Wong, B. Callewaert, K. Desloovere, L. Labey, B. Innocenti
483
Evaluation of Chronic Diabetic Wounds with the Near Infrared Wound Monitor . . . . . . . . . . . . . . Michael Neidrauer, Leonid Zubkov, Michael S. Weingarten, Kambiz Pourrezaei, Elisabeth S. Papazoglou
487
Non-contact UWB Radar Technology to Assess Tremor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G. Blumrosen, M. Uziel, B. Rubinsky, D. Porrat
490
High Frequency Mechanical Vibrations Stimulate the Bone Matrix Formation in hBMSCs (Human Bone Marrow Stromal Cells) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Pr`e, G. Ceccarelli, M.G. Cusella De Angelis, G. Magenes Mobispiro: A Novel Spirometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Eleni J. Sakka, Pantelis Aggelidis, Markela Psimarnou
494 498
A Computer Program for the Functional Assessment of the Rotational Vestibulo-Ocular Reflex (VOR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. B¨ ohler, M. Mandal´ a, S. Ramat
502
New Application for Automatic Hemifield Damage Identification in Humphrey Field Analyzer (HFA) Visual Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Salonikiou, V. Kilintzis, A. Antoniadis, F. Topouzis
506
The Effect of Mechano– and Magnetochemically Synthesized Magnetosensitive Nanocomplex and Electromagnetic Irradiation on Animal Tumor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V.E. Orel, A.V. Romanov, I.I. Dzyatkovska, M.O. Nikolov, Yu.G. Mel’nik, N.M. Dzyatkovska, I.B. Shchepotin Verification of Measuring System for Automation Intra – Abdominal Pressure Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ˇ c´ T. T´ oth, M. Michal´ıkov´ a, L. Bednarˇc´ıkov´ a, M. Petr´ık, J. Zivˇ ak
510
513
XXII
Table of Contents
Evolution in Bladder Pressure Measuring Implants Developed at K.U. Leuven . . . . . . . . . . . . . . . P. Jourand, J. Coosemans, R. Puers Including the Effect of the Thermal Wave in Theoretical Modeling for Radiofrequency Ablation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J.A. L´ opez Molina, M.J. Rivera, M. Trujillo, V. Romero-Garc´ıa, E.J. Berjano Textile Integrated Monitoring System for Breathing Rhythm of Infants . . . . . . . . . . . . . . . . . . . . . . . H. De Clercq, P. Jourand, R. Puers Comparison between VHDL-AMS and PSPICE Modeling of Ultrasound Measurement System for Biological Medium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. Aouzale, A. Chitnalah, H. Jakjoud, D. Kourtiche, M. Nadi
517
521 525
529
Stimulation Parameter Testing and Verification during Pacing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Martin Augustynek, Marek Penhaker, Pavel Sazel, David Korpas
533
Biosignal Monitoring and Processing for Management of Hypertension . . . . . . . . . . . . . . . . . . . . . . . A. Stan, R. Lupu, M. Ciorap, R. Ciorap
537
Design and Development of an Electrophysiological Signal Acquisition System: A Technological Aid for Research, Teaching and Clinical Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Villavicencio, D. Garc´ıa, L. Navarro, M. Torres, R. Huaman´ı, L.F. Yabar Numerical Models of an Artery with different Stent Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Brand, M. Ryvkin, S. Einav, I. Avrahami, J. Rosen, M. Teodorescu A Macro-quality Field Control of Dual Energy X-Ray Absorptiometry with Anatomical, Chemical and Computed Tomographical Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Scafoglieri, S. Provyn, O. Louis, J.A. Wallace, J. De Mey, J.P. Clarys Digital Filter in Hardware Loop for On Line ECG Signal Baseline Wander Reduction . . . . . . . . A. Petrenas, V. Marozas, S. Daukantas, A. Lukosevicius
541 545
549 554
Biomedical Measurements and Modeling Assessment of a Patient-Specific Silicon Model of the Human Arterial Forearm . . . . . . . . . . . . . . . K. Van Canneyt, F. Giudici, P. Segers, P. Verdonck
558
Numerical Investigations of the Strain-Adaptive Bone Remodeling in the Prosthetic Pelvis . . . A. Bouguecha, I. Elgaly, C. Stukenborg-Colsman, M. Lerch, I. Nolte, P. Wefstaedt, T. Matthias, B.-A. Behrens
562
Development of a System Dynamics Model for Cost Estimation for the Implantation and Revision of Hip Joint Endoprosthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Schr¨ ottner, A. Herzog
566
The Volume Regulation and Accumulation of Synovial Fluid between Articular Plateaus of Knee Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Petrtyl, J. Danesova, J. L´ısal
570
Table of Contents
Anthropometric Measurements and Model Evaluation of Mass-Inertial Parameters of the Human Upper and Lower Extremities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G.S. Nikolova
XXIII
574
Validation of a Person Specific 1-D Model of the Systemic Arterial Tree . . . . . . . . . . . . . . . . . . . . . . P. Reymond, Y. Bohraus, F. Perren, F. Lazeyras, N. Stergiopulos
578
First Trimester Diagnosis of Trisomy-21 Using Artificial Neural Networks . . . . . . . . . . . . . . . . . . . . C.N. Neocleous, K. Nikolaides, K. Neokleous, C.N. Schizas
580
Numerical Analysis of a Novel Method for Temperature Gradient Measurement in the Vicinity of Warm Inflamed Atherosclerotic Plaques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Z. Aronis, E. Massarwa, L. Rosen, O. Rotman, R. Eliasy, R. Haj-Ali, S. Einav A Multilevel and Multiscale Approach for the Prediction of Oral Cancer Reoccurrence . . . . . . . Konstantinos P. Exarchos, G. Rigas, Yorgos Goletsis, Dimitrios I. Fotiadis An Automated Method for Levodopa-Induced Dyskinesia Detection and Severity Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M.G. Tsipouras, A.T. Tzallas, G. Rigas, P. Bougia, D.I. Fotiadis, S. Konitsiotis
584 588
592
Electrospinning Poly(o-methoxyaniline) Nanofibers for Tissue Engineering Applications . . . . . . Wen-Tyng Li, Mu-Feng Shie, Chung-Feng Dai, Jui-MingYeh
596
Diagnosis of Asthma Severity Using Artificial Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Chatzimichail, A. Rigas, E. Paraskakis, A. Chatzimichail
600
Enhanced Stem Cells Characteristic of Fibroblastic Mesenchymal Cells from HHT Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G. Silvani, L. Benedetti, N. Crosetto, C. Olivieri, D. Galli, B. Magnani, G. Magenes, M.G. Cusella De Angelis High Frequency Vibration (HFV) Induces Muscle Hypertrophy in Newborn Mice and Enhances Primary Myoblasts Fusion in Satellite Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G. Ceccarelli, L. Benedetti, D. Pr`e, D. Galli, L. Vercesi, G. Magenes, M.G. Cusella De Angelis Surface Characterization of Collagen Films by Atomic Force Microscopy . . . . . . . . . . . . . . . . . . . . . A. Stylianou, S.B. Kontomaris, M. Kyriazi, D. Yova Changes in Electrocardiogram during Intra-Abdominal Electrochemotherapy: A Preliminary Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Mali, T. Jarm, E. Gadˇzijev, G. Serˇsa, D. Miklavˇciˇc
604
608
612
616
Studying Postural Sway Using Wearable Sensors: Fall Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Turcato, S. Ramat
620
From Biomedical Research to Spin-Off Companies for the Health Care Market . . . . . . . . . . . . . . . O.A. Lindahl, B. Andersson, R. Lundstr¨ om, K. Ramser
624
A Continuous-Time Dynamical Model for the Vestibular Nucleus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Korodi, V. Ceregan, T.L. Dragomir, A. Codrean
627
XXIV
Table of Contents
Fast Optical Signal in the Prefrontal Cortex Correlates with EEG . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.V. Medvedev, J.M. Kainerstorfer, S.V. Borisov, J. VanMeter
631
Using Social Semantic Web Technologies in Public Health: A Prototype Epidemiological Semantic Wiki . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Bratsas, A. Tzalavra, V. Vescoukis, P. Bamidis
635
Patellofemoral Contact during Simulated Weight Bearing Squat Movement: A Cadaveric Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Van Haver, J. Quintelier, M. De Beule, P. Verdonk, F. Almqvist, P. De Baets
639
Rapid Prototype Development for Studying Human Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Fevgas, P. Tsompanopoulou, S. Lalis Rheological and Electrical Properties of RBC Suspensions in Dextran 70. Changes in RBC Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. Antonova, I. Ivanov, Y. Gluhcheva, E. Zvetkova
643
647
Numerical Simulation In Magnetic Drug Targeting. Magnetic Field Source Optimization . . . . . A. Dobre, A.M. Morega
651
Ontology for Modeling Interaction in Ambient Assisted Living Environments . . . . . . . . . . . . . . . . . J.B. Mochol´ı, P. Sala, C. Fern´ andez-Llatas, J.C. Naranjo
655
Protein Surface Atom Neighborhood Functional Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P.D. Cristea, R. Tuduce, O. Arsene
659
Performance Evaluation of a Grid-Based Heart Simulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . R.S. Campos, M.P. Xavier, M. Lobosco, R.W. dos Santos
663
A Web-Based Tool for the Automatic Segmentation of Cardiac MRI . . . . . . . . . . . . . . . . . . . . . . . . . T.H. de Paula, M. Lobosco, R.W. dos Santos
667
Improved Modeling of Lane Intensity Profiles on Gel Electrophoresis Images . . . . . . . . . . . . . . . . . C.F. Maramis, A.N. Delopoulos
671
Affective Learning: Empathetic Embodied Conversational Agents to Modulate Brain Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C.N. Moridis, M.A. Klados, V. Terzis, A.A. Economides, V.E. Karabatakis, A. Karlovasitou, P.D. Bamidis The Role of Electrically Stimulated Endocytosis in Gene Electrotransfer . . . . . . . . . . . . . . . . . . . . . . M. Pavlin, M. Kanduˇser, G. Pucihar, D. Miklavˇciˇc
675
679
A Frequency Synchronization Study on the Temporal and Spatial Evolution of Emotional Visual Processing Using Wavelet Entropy and IAPS Picture Collection . . . . . . . . . . . . . . . . . . . . . . . C.A. Frantzidis, C. Pappas, P.D. Bamidis
683
Frontal EEG Asymmetry and Affective States: A Multidimensional Directed Information Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P.C. Petrantonakis, L.J. Hadjileontiadis
687
Table of Contents
A Game-Like Interface for Training Seniors’ Dynamic Balance and Coordination . . . . . . . . . . . . . A.S. Billis, E.I. Konstantinidis, C. Mouzakidis, M.N. Tsolaki, C. Pappas, P.D. Bamidis
XXV
691
Incorporating Electroporation-Related Conductivity Changes into Models for the Calculation of the Electric Field Distribution in Tissue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I. Lackovi´c, R. Magjarevi´c, D. Miklavˇciˇc
695
Automated Estimation of 3D Camera Extrinsic Parameters for the Monitoring of Physical Activity of Elderly Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . R. Deklerck, B. Jansen, X.L. Yao, J. Cornelis
699
Robotic System for Training of Grasping and Reaching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Podobnik, M. Munih Recognition and Identification of Red Blood Cell Size Using Angular Radial Transform and Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G. Apostolopoulos, S. Tsinopoulos, E. Dermatas Collagen Gel as Cell Extracellular Environment to Study Gene Electrotransfer . . . . . . . . . . . . . . . S. Haberl, D. Miklavˇciˇc, M. Pavlin
703
707 711
The Influence of Seat Pan and Trunk Inclination on Muscles Activity during Sitting on Forward Inclined Seats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Mastalerz, I. Palczewska
715
Radiation Exposure in Routine Practice with PET/CT and Automatic Infusion System – Practical Experience Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P. Tomˇse, A. Biˇcek
719
A Pilot Study for Development of Shoulder Proprioception Training System Using Virtual Reality for Patients with Stroke: The Effect of Manipulated Visual Feedback . . . . . . . . . . . . . . . . . S.W. Cho, J.H. Ku, Y.J. Kang, K.H. Lee, J.Y. Song, H.J. Kim, I.Y. Kim, S.I. Kim
722
Computer Modeling to Study the Dynamic Response of the Temperature Control Loop in RF Cardiac Ablation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Alba, M. Trujillo, R. Blasco, E.J. Berjano
725
Mechanical Properties of Long Bone Shaft in Bending . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S.M. Rajaai, K. PourAkbar Saffar, N. JamilPour
729
Active Behavior of Peripheral Nerves during Magnetic Stimulation . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Cretu, R. Ciupa, L. Darabant
733
Preparations’ Methodology for the Introduction of Information Systems in Hospitals . . . . . . . . . J. Sarivougioukas, A. Vagelatos, Ch. Kalamara
737
Supervised and Unsupervised Finger Vein Segmentation in Infrared Images Using KNN and NNCA Clustering Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Vlachos, E. Dermatas
741
An Echocardiographic Study for Assessment the Indices of Arterial and Ventricular Stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C.M. Stanescu, K. Branidou, A. Dan, I. Daha, C. Baicus, V. Manoliu, C. Adam, A.Gh. Dan
745
XXVI
Table of Contents
Estimation of Linear Parametric Distortions and Motions in the Frequency Domain . . . . . . . . . . D.S. Alexiadis, G.D. Sergiadis
749
Ex Vivo and In Vivo Regulation of Arginase in Response to Wall Shear Stress . . . . . . . . . . . . . . . R.F. da Silva, V.C. Olivon, D. Segers, R. de Crom, R. Krams, N. Stergiopuloss
753
Process Choreography for Interaction Simulation in Ambient Assisted Living Environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Fern´ andez-Llatas, J.B. Mochol´ı, C. S´ anchez, P. Sala, J.C. Naranjo
757
Measuring Device for Determination of Forearm Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P. Hlavoˇ n, J. Krejsa, M. Zezula
761
Subclavian Steal Syndrome – A Computer Model Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Manopoulos, S. Tsangaris
764
Mullins Effect in Human Aorta Described with Limiting Extensibility Evolution . . . . . . . . . . . . . . L. Horny, E. Gultova, H. Chlup, R. Sedlacek, J. Kronek, J. Vesely, R. Zitny
768
A Distribution of Collagen Fiber Orientations in Aortic Histological Section . . . . . . . . . . . . . . . . . . L. Horny, J. Kronek, H. Chlup, R. Zitny, M. Hulan
772
An Innovative Approach for Right Ventricular Volume Calculation during Right Catheterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P. Toumpaniaris, I. Skalkidis, S. Markatis, D. Koutsouris
776
Service Composition to Support Ambient Assisted Living Solutions for the Elderly . . . . . . . . . . . V. Moumtzi, C. Wills, A. Koumpis
780
Oscillations in Subthalamic Nucleus Measured by Multi Electrode Arrays . . . . . . . . . . . . . . . . . . . . J. Stegenga, T. Heida
784
Suitable Polymer Pipe to Modelling the Coronary Veins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Romola Laczko, Tibor Balazs, Eszter Bognar
788
Satisfaction Survey of Greek Inpatients with Brain Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G.K. Matis, O.I. Chrysou, N. Lyratzopoulos, K. Kontogiannidis, T.A. Birbilis
792
End Stage Renal Disease Patients’ Projections Using Markov Chain Monte Carlo Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Rodina, K. Bliznakova, N. Pallikarakis
796
Influence of Bioimplant Surface Electrical Potential on Osteoblast Behavior and Bone Tissue Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yu. Dekhtyar, I. Khlusov, N. Polyaka, R. Sammons, F. Tyulkin
800
Nano-Sized Drug Carrier for Cancer Therapy: Dose-Toxicity Relationship of PEG-PCL-PEG Polymeric Micelle on ICR Mice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J.L. Jiang, N.V. Cuong, S.C. Jwo, M.F. Hsieh
804
“Internet of Things”, an RFID – IPv6 Scenario in a Healthcare Environment . . . . . . . . . . . . . . . . H. Tsirbas, K. Giokas, D. Koutsouris
808
Table of Contents
XXVII
Development of Software Tool for Quantitative Gait Assessment in Parkinsonian Patients with and without Mild Cognitive Impairment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . L. Iuppariello, R. Tranfaglia, M. Amboni, L. Lista, M. Sansone
812
Tensile Stress Analysis of the Ceramic Head Endoprosthesis with different Micro Shape Deviations of the Contact Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Fuis, M. Koukal
815
Modelling of Cancer Dynamics and Comparison of Methods for Survival Time Estimation . . . . Tomas Zdrazil, Jiri Holcik
819
Implications of Data Quality Problems within Hospital Administrative Databases . . . . . . . . . . . . J.A. Freitas, T. Silva-Costa, B. Marques, A. Costa-Pereira
823
Can the EEG Indicate the FiO2 Flow of a Mechanical Ventilator in ICU Patients with Respiratory Failure? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E.G. Peranonti, M.A. Klados, C.L. Papadelis, D.G. Kontotasiou, C. Kourtidou-Papadeli, P.D. Bamidis
827
A European Biomedical Engineering Postgraduate Program – From Evaluation to Continuous Improvement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Griva, N. Pallikarakis
831
Web Based Medical Applications and Telemedicine SOAP/WSDL-Based Web Services for Biomedicine: Demonstrating the Technique with the CancerResource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . T. Meinel, M.S. Mueller, J. Ahmed, R. Yildiriman, M. Dunkel, R. Herwig, R. Preissner
835
A Web-Based Application for the Evaluation of the Healthcare Management . . . . . . . . . . . . . . . . . M. Bava, D. Zotti, R. Zangrando, M. Delendi
839
A System for Acquiring, Transmitting and Distributed EEG Data Processing . . . . . . . . . . . . . . . . . D. Kastaniotis, G. Maragos, N. Fragoulis, A. Ifantis
843
The Umbrella Database on Fever and Neutropenia in Children – Prototype for Internet-Based Medical Data Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Matthias Faix, Daniela Augst, Hans J¨ urgen Laws, Arne Simon, Fritz. Haverkamp, J. Rentzsch
847
A Research Information System (RIS) for Breast Cancer Genetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.L. Leskoˇsek, J. Dimec, K. Gerˇsak, P. Ferk
851
WeCare: Wireless Enhanced Healthcare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hande Ozgur Alemdar, Cem Ersoy
855
The Functionality Control of Horizontal Agitators for Blood Bags . . . . . . . . . . . . . . . . . . . . . . . . . . . . Z. Vasickova, M. Penhaker, M. Darebnikova
859
Experimental Hardware Solutions of Biotelemetric System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dalibor Janckulik, Leona Motalova, Karel Musil, Ondrej Krejcar
863
Modern Tools for Design and Implementation of Mobile Biomedical System for Home Care Agencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dalibor Janckulik, Leona Motalova, Ondrej Krejcar
867
XXVIII
Table of Contents
Application of Embedded System for Sightless with Diabetes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . L. Martinak, M. Penhaker
871
TELEMON – An Embedded Wireless Monitoring and Alert System for Homecare . . . . . . . . . . . C. Rotariu, H. Costin, R. Ciobotariu, F. Adochiei, I. Amariutei, Gladiola Andruseac
875
Graphical Development System Design for Creating the FPGA-Based Applications in Biomedicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Kasik, M. Stankus
879
Low Cost Data Acquisition System for Biomedical Usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Stankus, M. Penhaker, M. Cerny
883
Embedded Programmable Invasive Blood Pressure Simulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Kijonka, M. Penhaker
886
e-Health Generating and Transmitting Ambulatory Electronic Medical Prescriptions . . . . . . . . . . . . . . . . . . . M. Nyssen, K. Thomeer, R. Buyl
890
Estimating Pre-term Birth Using a Hybrid Pattern Classification System . . . . . . . . . . . . . . . . . . . . . M. Frize, N. Yu
893
Design and Implementation of a Radio Frequency IDentification (RFID) System for Healthcare Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.C. Polycarpou, G. Gregoriou, A. Dimitriou, A. Bletsas, I.N. Sahalos, L. Papaloizou, P. Polycarpou Reliability Issues in Regional Health Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S. Spyrou, P. Bamidis, N. Maglaveras Prevention and Management of Risk Conditions of Elderly People through the Home Environment Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . L. Pastor-Sanz, M.M. Fern´ andez-Rodr´ıguez, M.F. Cabrera-Umpi´errez, M.T. Arredondo, E. Bekiaris
897 901
905
Multilevel Access Control in Hospital Information Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Baldas, K. Giokas, D. Koutsouris
909
Managing Urinary Incontinence through Hand-Held Real-Time Decision Support Aid . . . . . . . . Constantinos Koutsojannis, Chrysa Lithari, Eman Alkholy Nabil, Giorgos Bakogiannis, Ioannis Hatzilygeroudis
913
Renal Telemedicine and Telehealth – Where Do We Stand? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Kaldoudi, V. Vargemezis
920
A System for Monitoring Children with Suspected Cardiac Arrhythmias – Technical Optimizations and Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Kyriacou, C. Pattichis, D. Hoplaros, A. Kounoudes, M. Milis, A. Jossif
924
Use of Guidelines and Decision Support Systems within EHR Applications in Family Practice – Croatian Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Kralj, S. Tonkovi´c, M. Konˇcar
928
Table of Contents
A New Concept of the Integrated Care Service for Unstable Diabetic Patients . . . . . . . . . . . . . . . P. L ady˙zy´ nski, P. Folty´ nski, J.M. W´ ojcicki, K. Migalska-Musial, M. Molik, J. Krzymie´ n, G. Rosi´ nski, G. Opolski, K. Czajkowski, M. Tracz, W. Karnafel SHARE: A Meeting Point for the Promotion of Interoperability and Best Practices in eHealth Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Ortega-Portillo, M.M. Fernandez-Rodriguez, M.F. Cabrera-Umpierrez, M.T. Arredondo, G. Carrozza Long Term Evolution (LTE) Technology in e-Health – A Sample Application . . . . . . . . . . . . . . . . . R. Jagusz, J. Borkowski, K. Penkala
XXIX
932
935 939
BME Education EMITEL e-Encyclopaedia of Medical Physics – Project Development and Future . . . . . . . . . . . . . S. Tabakov, P. Smith, F. Milano, S.-E. Strand, C. Lewis, M. Stoeva
943
BME Education Program Following the Expectations from the Industry, Health Care and Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P. Augustyniak, R. Tadeusiewicz, M. Wasilewska-Radwa´ nska
945
Quality Assurance in Biomedical Engineering COOP-Educational Training Program: Planning, Implementation and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Alhamwi, Manal A. Farrag, T. Elsarnagawy
949
Accreditation of Medical Physics and Medical Engineering Programmes in the UK . . . . . . . . . . . S. Tabakov, D. Parker, F. Schlindwein, A. Nisbett Tools Based eLearning Platform to Support the Development and Repurposing of Educational Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . T. Stefanut, M. Marginean, D. Gorgan
953
955
A Feasible Teaching Tool for Physiological Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . R. Stojanovic, D. Karadaglic, B. Asanin, O. Chizhova
959
Repurposing Serious Games in Health Care Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Protopsaltis, D. Panzoli, I. Dunwell, S. de Freitas
963
Geotagged Repurposed Educational Content through mEducator Social Network Enhances Biomedical Engineering Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S.Th. Konstantinidis, N. Dovrolis, Eleni Kaldoudi, P.D. Bamidis MORMED: Towards a Multilingual Social Networking Platform Facilitating Medicine 2.0 . . . . Eleni Kargioti, Dimitrios Kourtesis, Dimitris Bibikas, Iraklis Paraskakis, Ulrich Boes
967 971
Design and Development of a Pilot on Line Electronic OSCE Station for Use in Medical Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E.L. Dafli, P.D. Bamidis, C. Pappas, N. Dombros
975
Review of the Biomedical Engineering Education Programs in Europe within the Framework of TEMPUS IV, CRH-BME Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Z. Bliznakov, N. Pallikarakis
979
XXX
Table of Contents
Virtual Experiments: May VCV Impede Circulation More than PCV in (Virtual) Patients in the Lateral Position? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . T. Golczewski, M. Darowski
983
Clinical Engineering Human Factors Engineering Applied to Risk Management in the Use of Medical Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.P.S. Silva, R.M.A. Almeida, J.A. Ferreira, A. Gibertoni
987
Electromagnetic Interferences (EMI) from Active RFId on Critical Care Equipment . . . . . . . . . . Ernesto Iadanza, Fabrizio Dori, Roberto Miniati, Edvige Corrado
991
The Clinical Data Recorder: What Shall Be Monitored? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . L.N. Nascimento, S.J. Calil
995
Clinical Engineering and Patient Safety: A Forty Year Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Frize, S. Weyand, K. Greenwood
999
Extracorporeal Membrane Oxygenation in the Treatment of Novel Influenza Virus Infection: A Multicentric Hospital-Based Health Technology Assessment in Lombardy Region . . . . . . . . . . 1003 P. Lago, I. Vallone, G. Zarola MRI-Induced Heating on Patients with Implantable Cardioverter-Defibrillators and Pacemaker: Role of Lead Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1007 E. Mattei, G. Calcagnini, M. Triventi, F. Censi, P. Bartolini Adoption and Sophistication of Clinical Information Systems in Greek Public Hospitals: Results from a National Web-Based Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1011 S. Kitsiou, V. Manthou, M. Vlachopoulou, A. Markos Risk Management Process and CE Marking of Software as MD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1017 Fabrizio Dori, Ernesto Iadanza, Roberto Miniati, Samuele Mattei From Laparoscopic Surgery to 3-D Double Console Robot-Assisted Surgery . . . . . . . . . . . . . . . . . . 1021 P. Lago, C. Lombardi, B. Dell’Anna Clinical Engineering and Clinical Dosimetry in Patients with Differentiated Thyroid Cancer Undergoing Thyroid Remnant Ablation with Radioiodine-131 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025 M. Medvedec, D. Dodig Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029 Keyword Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035
Quantitative Analysis of Two-Dimensional Catch-Up Saccades Executed to the Target Jumps in the Time-Continuous Trajectory Vincas Laurutis and Raimondas Zemblys Biomedical Engineering Centre, Siauliai University, Vilniaus st. 141, LT-76353 Šiauliai, Lithuania
Abstract— The purpose of this research was to investigate quantitatively the catch-up saccades occurring during smooth pursuit. In the first experiment, to evoke catch-up saccades we used high velocity predictable two-dimensional timecontinuous target trajectories. In the second experiment, catch-up saccades were evoked using target jump paradigm during sustained two-dimensional pursuit. Target jumps in the different directions were presented at the unexpected moments and positions of the interrupted time-continuous target trajectory. From the experimental results we made a comparison of the main sequences (relationship between peak velocity and amplitude) of the catch-up and refixation saccades and found that they are different. Also we can conclude that the peak velocity of catch-up saccades is strongly correlated with the velocity of the smooth pursuit target component. We found that both position error and retinal slip are taken into account in catch-up saccades programming to predict the future trajectory of the moving target. Ill. 5, bibl. 5 (in English, summaries in English, Russian and Lithuanian). Keywords— Eye movements, Smoth pursuit, Sacadic eye movements, Catch-up saccades.
I. INTRODUCTION Eye movements exist to aid vision by directing gaze towards new objects of interest and, if those objects move, by tracking them. This serves to bring pertinent retinal images into the fovea – the region of the most acute vision. An interesting feature of the brain’s control of eye movements is its modular organization, with different subsystems mediating special functions. The neural subsystem generating the rapid, saccadic eye movements used to capture new object, is quite distinct from that performing the pursuit, tracking movements (second subsystem). The third subsystem – vestibular ocular reflex (VOR) is entirely concerned with generating eye movements that compensate for rotations of head and so tends to stabilize the eyes with respect to the environment. To achieve single vision by two eyes they must be aligned on the target. This has resulted in the evolution of a fourth subsystem to generate vergence eye movements. The nervous system controls all of these eye movements with considerable precision and ability to adapt its performance trough motor learning processes [1].
Recently all four eye movement subsystems are properly studied and many characteristics and parameters of them are well known. More interesting topics of the investigation now are how these subsystems collaborate together or subsequently take in action one after another [2]. This research is dedicated to reveal interaction between saccadic and smooth pursuit eye movements when target tracking is interrupted by catch-up saccades. Saccades are fast, dart-like, conjugate eye movements used to position the fovea of the eyes in a time optimal manner. They could be categorized into refixation saccades and microsaccades, which are seen only during fixation. Normometric saccadic refixations could be either singlestep or multi-step movements. Multi-step (usually doublestep) saccades are executed in two-steps: primary and corrective saccades. Primary saccade could be too small (hypometric) and too large (hypermetric) with respect to the intended target position. Only 30 % of saccades are singlestep and precisely reaches new target. From the rest of the multi-step saccades only 30 % are hypermetric. Normometric saccades demonstrate pretty common trajectories. Saccadic latency, or reaction time, typically refers to the time from onset of the non-predictable step of target movement to onset of the saccadic eye movement initiated to foveate the displaced target. It is approximately 180 to 200 msec, with standard deviation 30 msec. Relationship between peak velocity and amplitude of the saccade, called the main sequence, typically separate these movements from other limb or head movements. Main sequence illustrate that peak velocity is 410 deg/sec for 10 deg amplitude of the saccade, 500 deg/sec for 15 deg amplitude and 650 deg/sec for 20 deg amplitude. Catch-up saccades are seen in the eye tracking trajectory of the smoothly moving target, when target velocity becomes too large for smooth pursuit eye movements. Also catch-up saccades could be executed to the target jumps in the time-continuous trajectory. In this case, at some moments smooth pursuit eye movements are interrupted by quick eye jumps – catch-up saccades, which allow maintain the target on the fovea. Parameters of the catch-up saccades differ from refixation saccades and they are investigated only in one dimensional mode [3].
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 1–4, 2010. www.springerlink.com
2
V. Laurutis and R. Zemblys
II. AIM AND TASKS OF THE WORK
IV. EXPERIMENTAL RESULTS
In this research we investigated quantitative parameters of catch-up saccades such as: main sequence (relationship between peak velocity Ep and amplitude A of catch-up saccades), saccadic latency Td, and the precision of the extrapolation of the target trajectory by smooth pursuit eye movements during time interval Te between target jump and catch-up saccade onset and time interval Tr when after catch-up saccade smooth pursuit behavior is restored. Further investigation was focused on the finding relationship between the quantitative parameters of the catch-up saccades and the velocity of the time-continuous target trajectory. For refixation saccades to stationary targets a sensory signal is the position error between target projection in the periphery of the retina and the fovea. When the target is moving and the eye and target velocities are different, retinal slip took place. To overcome these slip and delay the oculomotor system uses prediction of future target motion to program catch-up saccades to moving target. Previous studies did not clearly distinguish the influence of target velocity, position error, retinal slip and prediction in catchup saccades programming [4].
Four typical examples of the onset positions of catch-up saccades obtained during tracking of the targets which were moving in the clock-wise direction by square-shape and circle-shape trajectories are shown in Fig. 1. There we can see that for 10 times repeated trials catch-up saccades were more concentrated on the corners of the squares.
Fig. 1 Two-dimensional positions of catch-up saccades obtained during tracking of the square-shape (A, B) and circle-shape (C, D) target trajectories. Velocities of the target trajectories were 20 deg/sec for (A, C) and 50deg/sec for (B, D)
III. METHOD Movements of both eyes were recorded with eye tracker EyeGaze System produced by LC Technologies Ltd. Healthy subjects without any known oculomotor abnormalities were recruited after informed consent. Among the five subjects, two authors participated in the experiments. Twodimensional target trajectories were presented on the computer screen. During first experiment, subjects were asked to track target (white spot) moving in the clock-wise direction. Square-shape and circle-shape target trajectories with angular velocities in the range of 10 - 50 deg/sec were used. Two-dimensional positions, amplitudes and velocities of the catch-up saccades were recorded. In the second experiment, subjects were asked to track target moving with non-predictable time-continuous trajectory. At the randomly chosen time, time-continuous trajectory was interrupted by target jumps, amplitudes and directions of which also was random. Trajectories of catch-up saccades executed as the reactions to the target jumps were analyzed. Quantitative parameters of the catch-up saccades, such as reaction time (saccadic latency Td), amplitudes A, peak velocities Vp, were measured. Also the extrapolation ability during time interval between target jump and catch-up saccade onset was evaluated.
At these points the target trajectories change the movement direction from horizontal to vertical. Therefore, the control system of the smooth pursuit eye movements also is forced to change tracking direction. Pier of the horizontal eye globe muscles have to stop horizontal movement and vertical muscles start to act. This elicits tracking errors which are eliminated by catch-up saccades. Two-dimensional positions of the catch-up saccades in the figure 1 (C, D) obtained during tracking of circle-shape target trajectories more clearly demonstrates that even when eyesight moves in the circle catch-up saccades occurs when target changes movement direction from vertical to horizontal and from horizontal to vertical. In the figure 1 C, D, when target velocities were increased to the 50 deg/sec, the shape of eyesight trajectories differs from the target trajectories. They have angular shift in the clock-wise direction what demonstrates anticipation (focused on the future target position) of the predictable target trajectory. Investigation of the relationship between the peak velocities of catch-up saccades and the target velocities during pursuit of the predictable target trajectories reveals close correlation between them (Table 1).
IFMBE Proceedings Vol. 29
Quantitative Analysis of Two-Dimensional Catch-Up Saccades Executed to the Target Jumps in the Time-Continuous Trajectory
3
Table 1 Relationship between target velocity Vt and catch-up saccade peak velocity Vp in deg/sec
Vt Vp
10 43
20 66
30 95.5
40 128
50 164.5
In the figure 2, the trajectories of catch-up saccades in the horizontal x and vertical y directions as reaction to the target jump executed during time-continuous movement are shown. Oculomotor system demonstrates common (reflexive) behavior during ten trials elicited to the same target jump in the non-predictable target trajectory.
Fig. 3 Target trajectory (points 1, 2, 3, 6, 4) and eyesight trajectory (1, 2, 5, 6, 7, 8) obtained during two-dimensional tracking and plotted together
Fig. 2 Ten trajectories of catch-up saccades plotted together in the horizontal (A) and vertical (B) directions Target and eyesight trajectories in the figures 3 and 4 could be divided in the few stages. From point 1 to point 2 we can see precise pursuit of the initial target trajectory. Target trajectory from point 2 to point 3 makes jump. Eyesight in this situation from point 2 to point 5 continues to extrapolate initial target trajectory what is more clearly seen in figure 4. From point 5 to point 6 eyesight jumps to the new position on the further time-continuous target trajectory. At the points 7 and 8 eyesight makes corrective saccades. Most interesting finding in this research is notice that during catch-up saccade eyesight does not respond to the target jump, but catch future position in the time-continuous target trajectory (point 6). It means that vision is active during saccadic latency (points 2, 5).
Fig. 4 The same target and eyesight trajectories as in the figure 3 plotted in the horizontal (A) and vertical (B) directions Fig. 5 illustrates that the main sequences of the refixation and catch-up saccades differ. It means that the refixation and catch-up saccades are controlled by different neural circuits.
IFMBE Proceedings Vol. 29
4
V. Laurutis and R. Zemblys
Fig. 5 Main sequences of the refixation catch-up and saccades In the table 2, the parameters of catch-up saccades are placed. There we can see that saccadic latency T is stable for all five subjects. The extrapolation of the former target trajectory E was evaluated as the angular displacement between the real eyesight and extrapolated target trajectories at the end of the saccadic latency (point 5 in the figure 4)). Table 2 Parameters of catch-up saccades Subject T, sec E, deg
RZ 0.25 2.16
VL 0.28 2.58
GD 0.24 2.27
NR 0.22 2.18
AP 0.26 2.32
V. CONCLUSIONS 1. Catch up saccades aids the oculomotor system to reduce tracking error. Main parameters which induce catch-up saccades are position errors between the target projection on the retina and fovea and the retinal slip.
2. During two-dimensional tracking, when target changes the movement direction and therefore tracking errors increases, appearance of catch-up saccades is seen more often. 3. Saccadic latency for catch-up saccades (Tr = 240 msec) is bigger than for refixation saccades (T = 200 msec) and did not depend on the target velocity. 4. Peak velocity of catch-up saccades is strongly correlated with velocity of the smooth pursuit target component. 5. The relationship between the peak velocity and the amplitude (main sequence) for catch-up saccades differ from main sequence for refixation saccades. 6. Landing points of catch-up saccades do not depend on the amplitudes of the target jumps but is related with the new target position at the offset of catch-up saccade. The prediction of the future trajectory of the moving target requires more calculations and explains why the saccadic latency for catch-up saccades is bigger than for the refixation saccades.
REFERENCES 1. Kiuffedra. K. J., Tannen B. Eye movement Basics for the Clinician. StLouis: Mosby.- 1994.- 266p. 2. Laurutis V., Robinson D. A. Are fixational and pursuit eye movements created by two different neural circuits? // Mechanika. - Kaunas: Technologija, 1996.- No. 2(4). – P. 43-47. 3. Laurutis V., Daunys G. Prediction features of the two-dimensional smooth pursuit eye movements // Medical & Biological Engineering and computing Vol. 34. The 10th Nordic-Baltic conference on biomedical engineering. – 1996. – Finland, Tampere – P. 335 – 336. 4. Bennett S. J., Barnes G. R. Combined smooth and saccadic ocular pursuit during transient occlusion of a moving visual object // Exp. Brain Res. – 2006. – No. 168. – P. 313-321. 5. Skavenski A. A., Steinman R. M. Control of eye position in the dark // Vision Res. – 1970. – No. 10(193). – P. 319-326.
IFMBE Proceedings Vol. 29
Spike Sorting based on Dominant-Sets clustering
D. A. Adamos1, N. A. Laskaris2, E. K. Kosmidis3 and G. Theophilidis1 1
Laboratory of Animal Physiology, School of Biology, Aristotle University of Thessaloniki (AUTh), 54 124, Greece 2 Laboratory of Artificial Intelligence & Information Analysis, Department of Informatics, AUTh, 54 124, Greece 3 Laboratory of Physiology, School of Medicine, AUTh, 54 124, Greece
Abstract—Spike sorting algorithms aim at decomposing complex extracellularly recorded electrical signals to independent events from single neurons in the vicinity of electrode. The decision about the actual number of active neurons in a neural recording is still an open issue, with sparsely firing neurons and background activity the most influencing factors. We introduce a graph-theoretical algorithmic procedure that successfully resolves this issue. Dimensionality reduction coupled with a modern, efficient and progressively-executable clustering routine proved to achieve higher performance standards than popular spike sorting methods. Our method is validated extensively using simulated data for different levels of SNR. Keywords— Spike sorting, dominant sets, graph-theoretic clustering, ISOMAP, manifold learning I. INTRODUCTION
The basis of every spike sorting algorithm is the assumption that all the action potential traces of a particular neuron have nearly the same amplitude and shape. In extracellular recordings, the shapes of recorded spike waveforms mainly depend on neuron’s geometry as well as its distance to the recording electrode. The goal of a spike sorting routine is to process and analyze the usually composite recorded signals in order to identify the number of active neurons and extract detailed time courses of their spiking activity. Related algorithms constitute the core methodological component in various situations ranging from traditional neurophysiological experiments and clinical/neuroscience studies to cortexmachine interfaces. The battery of available spike sorting routines includes mainly automated techniques that analyze the recorded signals by means of their waveforms. At the initial stage, various linear techniques like Principal Component Analysis (PCA) and wavelets are used in order to reduce the dimensionality of the input data and enhance the signal content in the attempted representation. PCA-based projection is often restrained within the subspace spanned by the first two or three principal components, although the employment of more components has recently been reported to carry useful complementary information [1]. Alternatively, the wavelet transform is used for the decomposition of spike waveforms [2], featuring improved discrimination of local-
ized shape divergences. In both the above approaches, the representation scheme serves as a preprocessing for a clustering framework that would take over the detection of distinct signal sources (i.e. active neurons) and the isolation of the corresponding spiking contributions. Regarding clustering, Bayesian [3] and Expectation Maximization [4] methods have been proposed for spike sorting. Assuming a stationary Gaussian profile for the background noise, both methods consider Gaussian properties for the potential clusters residing in the PCA representation subspace. However, in the general form, background noise in neural recordings is non-stationary, non-Gaussian and carries a complex correlated profile with higher power in low frequencies. Synaptic coupling among neurons, superimposed field potentials and bursting neurons are some of the reasons that would make a non-Gaussian cluster profile more plausible. To avoid Gaussian considerations, clustering approaches featuring hierarchical [5] and nearestneighbors [2] algorithms have also appeared. The later employs a stochastic algorithm, known as super-paramagnetic clustering (SPC), which makes no prior assumptions for the statistical properties of the data. There are two important parameters when one evaluates a spike sorting classification process: the number of clusters (i.e. active neurons) being decided by the process and the number of spikes assigned in each cluster. Both are well incorporated by Type I / II errors [4] in the spike sorting domain. Type I (false positive (FP)) / II (false negative (FN)) errors derive from the traditional classification schemes and conceptualize misclassification. For example, the identification of less neurons than expected (underclustering) leads to high false positive errors, while the opposite case (over-clustering) results in a large amount of false negatives. Although a correct estimation of the number of clusters would limit both errors in cluster-delineation, the identification of actual number of active neurons is still an open issue. Even popular methods (like SPC) do not incorporate a sufficient treatment of this issue, leading to inappropriate results [6]. It is worth noting that, in laboratory practice, over-clustering is most often addressed in a less time-consuming way than under-clustering. In a previous work [1] we have stressed the importance of the previous fact and proposed a new cluster error definition that favors over-clustering over under-clustering errors.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 5–8, 2010. www.springerlink.com
6
D.A. Adamos et al.
There are two main reasons why the identification of neurons’ number is still an open issue. The first relates to the adopted clustering techniques, which require the a priori definition of number of groups. The second relates to the low SNR of the signals and the subsequent poor representation (in the original or reduced space) of the waveforms. In this work, we propose a sequential, subtractive clustering algorithm which is based on graph theoretic ideas and in particular the notion of dominant set [7]. The algorithm works in an iterative fashion by operating on a neighborhood graph. It identifies the core of the graph, using replicator dynamics formulation, and then removes it from the graph and feeds back the remaining graph. The procedure terminates when all the data have been assigned to distinct groups or no more compact group can be formed any further. The suggested algorithm is engaged to work within a representation space which is derived via a fully compatible dimensionality reduction technique, namely Isometric feature mapping (ISOMAP) [8]. ISOMAP is known to reveal the intrinsic data variation and is therefore expected to be insensitive to random variations due to noise. Hence, the resulting low-dimensional parameterization of the waveform variation is expected to enhance the clustering performance. Section II describes the proposed methodology. Section III presents the comparative evaluation of our spike sorting technique in relation with a popular alternative [2], using simulated data. Section IV concludes the paper. II. METHOD
A. Low-dimensional Representation The segments, extracted from the time-series of extracellular recordings via a root-mean-square threshold detector, are considered as the first raw representation of spiking activity. The ensemble of these loosely aligned spikewaveforms can be thought as a point-swarm residing in a multidimensional feature space with axes corresponding to signal amplitudes at particular latencies. Following the standard convention, the ith spike waveform is depicted as xi (t), t=1,2,…,T, i=1,2,…,N ( with t denoting discrete time or latency) and represented via the row-vector xi = [xi (1), xi (2),…, xi (t),…, xi (T)] RT. Similarly the whole ensemble is represented in a data-matrix format as X [N x T] = [x1 x2 … xi … xN]. Most often an enhanced representation is sought by employing a dimensionality reduction technique (denoising step). Here, we employ ISOMAP embedding in order to achieve a parsimonious representation, in which the true degrees of freedom can be easily recognized and directly associated with involved neurons.
The algorithmic details of ISOMAP technique can be found elsewhere [9]. It starts by building a neighborhood graph over the data-points in the original feature space. This graph is then used to compute all the geodesic inter-point distances. Multidimensional scaling is finally employed to derive a reduced coordinate space where these distances are preserved and therefore the intrinsic geometry of the data is faithfully represented. In our case, the ISOMAP-routine provides a geometrical picture, within an r-D space, of the spike-waveforms variation. Y [N x r] = [y1 y2… yi … yN ] = ISOMAP (X, r)
(1),
r
where yi = [ yi (1), yi (2),…, yi (r) ] R . The derived point-swarm (for a 2-d example, see Figure 1b) is accompanied with the residual variance, which is a performance index ranging from 0% to 100% and indicating the reliability of the mapping (for an example, see Figure 1c). The optimal dimensionality ro can be sought (as a compromise between accuracy and compression) by computing multiple maps with increasing r (r [1,10]), drawing the diagram of residual variance as a function of r and applying the elbow rule. B. Identifying neurons and classifying the data A recently introduced graph-theoretic algorithm [7] was employed for identifying the most-cohesive groups of vertices given the weighted similarity (adjacency) matrix of a graph. The algorithm is based on the identification of the dominant set of nodes and when repeatedly applied facilitates the effective clustering, in a sequential mode, of pairwise relational data. As stated in [7], the main property of a dominant set is that the overall similarity among internal nodes is higher than that between external and internal nodes, and this fact is the motivation of considering a dominant set as a cluster of nodes. One of its main characteristics is the compact, elegant formulation. In our case, ISOMAP-representation points (Y) are used to build an undirected edge-weighted graph with no self-loops G=(V, E, w), where V = {1,…, N} is the vertex set, E V x V is the edge set and w : E R+* is the positive weight function. Vertices represent data points, edges correspond to neighborhood relationships and edgeweights reflect similarity between pairs of linked vertices. In a preprocessing step, Euclidean distances dij are transformed to similarity weights, which increase with decreasing distances. For this transformation, we use the following form: w(i, j) = exp ( - dij / )
(2),
where is a real positive number estimated as 3 times the average of the mean distance of all dij. Consequently, we
IFMBE Proceedings Vol. 29
Spike Sorting Based on Dominant-Sets Clustering
7
represent the N-node graph G (V, E, w) with the corresponding similarity matrix A, with i representing the i-th data point and the weight of edge (i, j) being set to w(i,j): A =(aij) where ij = w(i, j) if (i, j) E and ij = 0 otherwise. As pointed in [7], the cohesiveness of a cluster is measured by the overall similarity of a dominant set; that is, a good cluster contains elements that have large values connecting one another in the similarity matrix. Hence, the problem of finding a compact cluster is formulated as the problem of finding a vector x that maximizes the following objective function: f (x) = xT A x
(3), n
T
subject to x , where ={x R : x 0 and e x = 1}. Thus, a maximally cohesive cluster denotes the most dominant solution set that is iteratively subtracted from the N-node graph (G). At the end of the iterations, where all the data have been classified, the overall cohesiveness f(x) of each step is exploited for the decision about the actual number of active neurons (for an example, see Figure 1d). III. RESULTS
For the detailed evaluation of our method, we generated
spike waveforms representing neural activity from three separate neurons. Aiming at realistic simulations, we utilized real action potentials from respiratory motoneurons which had been recorded in vitro with a single “hook” electrode from the peripheral nervous system of the beetle Tenebrio molitor [10]. Three such action potential waveforms served as the initial templates. These waveforms were recorded extracellularly with a sampling frequency of 30 kHz and time duration of 4 ms (120 samples). The templates were replicated multiple times and added to segments of background noise extracted from the same recording (randomly extracted from latencies during which the spikedetector was silent). In order to pursue evaluation results under different SNR-levels, each extract of real background noise was modulated by a variable, positive amplitude factor . The SNR of the resulting waveform (template plus noise segment) was then defined as follows:
Three hundreds (300) waveforms per template were generated yielding 900 single spikes corresponding to the three neural classes of our data set. In addition, paired combinations among the three templates were realized by first inducing variable delays and then adding noise segments
Figure 1: a) The overall simulated data set; The SNR is 5 for all waveforms. b) 2-D ISOMAP representation of the data set. c) Residual variance diagram of ISOMAP. The optimal dimensionality (ro=4) is selected using the elbow rule. d) Group-cohesiveness values, iteratively computed for each subtracted cluster. e) Classification results visualized within 2-D ISOMAP space. Colors indicate classes, while unclassified data are left black. f) Classification results presented in the original data domain using the corresponding colors.
IFMBE Proceedings Vol. 29
8
D.A. Adamos et al.
using amplitude factors so as to achieve a given SNR-level. In this way 150 waveforms of double-overlaps were generated; 50 for each template pair. Finally, 50 more waveforms were added corresponding to triple overlaps with their SNRlevel adjusted accordingly. The complete data set of 1100 waveforms is shown in Figure 1a. The 2-D ISOMAP representation of this dataset is shown in Figure 1b. Considering the high density areas of this point-diagram, there are apparently three clusters (pointing at the existence of three active neurons). The residual variance graph, which is a supplementary output of the ISOMAP routine, is included in Fig.1c and clearly designates the first four dimensions as necessary for the faithful low-dimensional representation of the data. Hence, the first four ISOMAP coordinates were used to represent each spike waveforms and dominant-set clustering was applied to the new data matrix Y[1100 x 4]. The overall cohesiveness for every cluster subtracted by the iterative process is shown in Figure 1d. The high ranking of the first three groups denotes the presence of three active neurons in the data. The straight line in the figure depicts the cluster quality for the whole graph taken as single component. For the selected groups, the classification algorithm assigns their waveforms to the corresponding classes, while the rest were left unclassified. This data-sieving procedure step is visualized in ISOMAP space (Figure 1e) and in the original data domain (Figure 1f) using a three-level color-code. Type I/II errors were used for the performance evaluation. The total number of FP and FN is referred to the identified classes. The adopted error-rate accounts for the total number of false positive and false negative spikes:
with i running over the number of single-spike classes (i.e. the number of identified neurons). For comparison purposes, the measurements corresponding to Waveclus [2] are also included. Both Waveclus and our proposed methodology classified the spikes of the data set into three main classes and an additional noise class. The results of this classification are shown in Figure 2, using 10 different realizations of the dataset. Averaged values for the errorrates and the FP-FN errors are shown as a function of SNR level. It can be seen that our proposed algorithm achieves the lowest error rate which is lower than 1% when the SNR is higher than 4.
CONCLUSIONS The present work introduces a graph-theoretical approach to spike sorting; it iteratively pursuits a dominant set in the graph (the most dominant in each iteration) and then
Figure 2: Average error rates for our method and Waveclus.
removes it, until all the data have been clustered. The efficiency of the proposed clustering method has been combined with the robustness of ISOMAP representation. The hybrid scheme has been extensively evaluated. The results indicate high robustness to noise and a measured performance that goes beyond the contemporary standards.
REFERENCES 1.
Adamos DA, Kosmidis EK and Theophilidis G. (2008) Performance evaluation of PCA-based spike sorting algorithms. Computer methods and programs in biomedicine 91(3):232-44 2. Quian Quiroga R, Nadasdy Z, Ben-Shaul Y (2004) Unsupervised Spike Detection and Sorting with Wavelets and Superparamagnetic Clustering. Neural Comp 16:1661-1687. 3. Lewicki M (1998) A review of methods for spike sorting: the detection and classification of neural action potentials. Network: Computation in Neural Systems 9:R53-R78. 4. Harris KD, Henze DA, Csicsvari J, Hirase H, Buzsaki G (2000) Accuracy of Tetrode Spike Separation as Determined by Simultaneous Intracellular and Extracellular Measurements. J Neurophysiol 84:401-414. 5. Fee MS, Mitra PP, Kleinfeld D (1996) Automatic sorting of multiple unit neuronal signals in the presence of anisotropic and non-Gaussian variability. Journal of Neuroscience Methods 69:175-188. 6. Herbst JA, Gammeter S, Ferrero D, Hahnloser RH (2008) Spike sorting with hidden Markov models. J Neurosci Methods 174 (1):12634 7. Pavan M and Pelillo M (2007) Dominant Sets and Pairwise Clustering. IEEE Trans. Pattern Anal. Mach. Intell. 29(1): 167-172 8. Tenenbaum JB, de Silva V, Langford JC (2000) A global geometric framework for nonlinear dimensionality reduction. Science 290(5500):2319-23 9. Laskaris NA and Ioannides AA (2002) Semantic geodesic maps: a unifying geometrical approach for studying the structure and dynamics of single trial evoked responses. Clinical Neurophysiology 113(8):1209-1226 10. Zafeiridou G and Theophilidis G (2004) The action of the insecticide imidacloprid on the respiratory rhythm of an insect: the beetle Tenebrio molitor. Neuroscience Letters 365(3):205-209 Author: Institute: Street: City: Country: Email:
IFMBE Proceedings Vol. 29
Dimitrios A. Adamos School of Biology, Aristotle University Aristotle University Campus, 54124 Thessaloniki Greece
[email protected] Differentiation of human bone marrow stromal cells onto gelatin cryogel scaffolds L. Fassina1,7, E. Saino2,7, L. Visai2,7, M.A. Avanzini3, M.G. Cusella De Angelis4,7, F. Benazzo5,7, S. Van Vlierberghe6, P. Dubruel6, G. Magenes1,7 1
Dipartimento di Informatica e Sistemistica, University of Pavia, Pavia, Italy 2 Dipartimento di Biochimica, University of Pavia, Pavia, Italy 3 Oncoematologia Pediatrica, IRCCS San Matteo, University of Pavia, Pavia, Italy 4 Dipartimento di Medicina Sperimentale, University of Pavia, Pavia, Italy 5 Dipartimento SMEC, IRCCS San Matteo, University of Pavia, Pavia, Italy 6 Polymer Chemistry and Biomaterials Research Group, University of Ghent, Ghent, Belgium 7 Centre for Tissue Engineering (C.I.T., http://cit.unipv.it/cit), University of Pavia, Pavia, Italy Abstract— Biomaterials have been widely used in reconstructive bone surgery to heal critical-size long bone defects due to trauma, tumor resection, and tissue degeneration. In particular, gelatin cryogel scaffolds are promising new biomaterials owing to their biocompatibility; in addition, the in vitro modification of biomaterials with osteogenic signals enhances the tissue regeneration in vivo, suggesting that the biomaterial modification could play an important role in tissue engineering. In this study we have followed a biomimetic strategy where differentiated human bone marrow stromal cells built their extracellular matrix onto gelatin cryogel scaffolds. In comparison with control conditions without differentiating medium, the use of a differentiating medium increased, in vitro, the coating of gelatin cryogel with bone proteins (decorin, osteocalcin, osteopontin, type-I collagen, and typeIII collagen). The differentiating medium aimed at obtaining a better in vitro modification of gelatin cryogel in terms of cell colonization and coating with osteogenic signals, like bone matrix proteins. The modified biomaterial could be used, in clinical applications, as an implant for bone repair. Keywords— Gelatin cryogel, bone marrow stromal cell, cell proliferation, bone extracellular matrix, surface modification, biomimetics.
I. INTRODUCTION One of the key challenges in reconstructive bone surgery is to provide living constructs that possess the ability to integrate in the surrounding tissue. Bone graft substitutes, such as autografts, allografts, xenografts, and biomaterials have been widely used to heal critical-size long bone defects and maxillofacial skeleton defects due to trauma, tumor resection, congenital deformity, and tissue degeneration. The biomaterials used to build 3D scaffolds for bone tissue engineering are, for instance, the hydroxyapatite [1], the partially demineralized bone [2], and biodegradable porous polymer-ceramic matrices [3]. The preceding osteoinductive and osteoconductive biomaterials are ideal in order to follow a typical approach of the tissue engineering, an approach that involves the seeding and the in vitro culturing of cells within a porous scaffold before the implantation. Gorna and Gogolewski [4, 5] have drawn attention to the ideal features of a bone graft substitute: it should be porous with interconnected pores of adequate size allowing for the ingrowth of capillaries and perivascular tissues; it should attract mesenchymal stem cells from the surrounding bone and promote their differentiation into osteoblasts; it should avoid shear forces at the interface between bone and bone graft substitute; and it should be biodegradable.
In this study, following the preceding “golden rules” of Gorna and Gogolewski, we have elected gelatin cryogel [6-8] as bone graft substitute and, applying a differentiating medium to bone marrow stromal cells, we have attempted to populate it with extracellular matrix and differentiated osteoblasts. Gelatin cryogel [6-8] is a promising new biomaterial owing to its biocompatibility. The in vitro modification of gelatin cryogel, with osteogenic signals of the transforming growth factor-E superfamily and with bone morphogenetic proteins, enhances the tissue regeneration in vivo [9] suggesting that the modification of gelatin cryogel could play an important role in tissue engineering. As consequence, aiming, in a future work, at accelerated and enhanced bone regeneration in vivo, in the present study of tissue engineering, we show a particular “biomimetic strategy” that consists in the in vitro modification of gelatin cryogel with differentiated bone marrow stromal cells and their extracellular matrix produced in situ. In other words, using a differentiating medium, our aim was to enhance a bone marrow cell culture onto a gelatin cryogel, that is, to coat the gelatin cryogel with physiological and biocompatible cell-matrix layers. Using this approach, the in vitro cultured material could be theoretically used, in clinical applications, as an osteointegrable implant. II. MATERIALS AND METHODS Gelatin cryogel disks: Bovine gelatin cryogel disks (diameter, 10 mm; height, 2 mm) were kindly provided by Polymer Chemistry and Biomaterials Research Group, University of Ghent (Ghent, Belgium) [6-8] (Fig. 1).
Fig. 1 Unseeded gelatin cryogel [6-8]
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 9–12, 2010. www.springerlink.com
10
L. Fassina et al.
Cells from bone marrow aspirates: Mononuclear cells were isolated from bone marrow aspirates (30 ml) by density gradient centrifugation in Ficoll (density, 1.077 g/ml) (Lymphoprep, Nycomed Pharma) and plated in non-coated 75-175 cm2 polystyrene culture flasks (Corning Costar, Celbio) at a density of 160000 cell/cm2 [10]. The culture condition was based on the basal medium Mesencult (Stem Cell Technologies) supplemented with 2 mM Lglutamine, 50 Pg/ml gentamycin, and 10% fetal calf serum. Cultures were maintained at 37°C in a humidified atmosphere containing 5% CO2. After 48 h, non-adherent cells were discarded and culture medium was replaced twice a week. After reaching 80% confluence as a minimum, the cells were harvested and re-plated for expansion at a density of 4000 cell/cm2 until 5th passage. The colony-forming unit-fibroblast assay (CFU-F) was performed as described previously [11]. CFU-F formation was examined after incubation for 12 days in a humidified atmosphere (37°C, 5% CO2); the clonogenic efficiency was calculated as the number of colonies per 106 bone marrow mononuclear cells seeded. According to the International Society for Cellular Therapy on the nomenclature of mesenchymal progenitors, the cells cultured for this study were defined as multipotent stromal cells [12]. Cell culture: To study the osteogenic differentiation potential, the obtained bone marrow stromal cells were then cultured in DMEM (Invitrogen) supplemented with 10% fetal bovine serum, 50 Pg/ml penicillin-streptomycin, and 1% L-glutamine. After reaching 80% confluence as a minimum, the cells were harvested and replated for expansion at a density of 2.5u104 cell/cm2. The cells were cultured at 37°C with 5% CO2, three fifths of the medium were renewed every 3 days, and then the cells were routinely trypsinized, counted, and seeded onto the gelatin cryogel disks. Cell seeding: To anchor the gelatin cryogel disks to standard well-plates, 3% (w/v) agarose solution was prepared and sterilized in autoclave, and during cooling, at 45°C, 100 Pl of agarose solution were poured inside the wells to hold the placed gelatin disks and to fix them after completed cooling. The well-plates with the biomaterial disks were sterilized by ethylene oxide at 38°C for 8 h at 65% relative humidity. After 24 h of aeration in order to remove the residual ethylene oxide, the disks were ready. A suspension of 5u105 bone marrow stromal cells in 400 Pl was added onto the top of each disk and, after 0.5 h, 600 Pl of culture medium was added to cover the disks. We have utilized two types of culture medium. For the control well-plate, we have used the “proliferative” medium i.e. D-MEM supplemented with 10% fetal bovine serum, 50 Pg/ml penicillinstreptomycin, 1% L-glutamine. Whereas, for the “differentiating” well-plate, we have utilized the proliferative medium only for two weeks and then the “differentiative” one, i.e. the same as above to which ascorbic acid (50 Pg/ml), dexamethasone (10-7 M), and Eglycerophosphate (5 mM from day 21) were added. The duration of the control and differentiating cultures was 6 weeks and the culture media were changed every 3 days. Scanning electron microscopy (SEM) analysis: Gelatin cryogel disks were fixed with 2.5% (v/v) glutaraldehyde solution in 0.1 M Na-cacodylate buffer (pH=7.2) for 1 h at 4°C, washed with Nacacodylate buffer, and then dehydrated at room temperature in a gradient ethanol series up to 100%. The samples were kept in 100% ethanol for 15 min, and then critical point-dried with CO2.
The specimens were sputter coated with gold and observed at 500u magnification with a Leica Cambridge Stereoscan 440 microscope at 8 kV. DNA content: At the end of the culture period, the cells were lysed by a freeze-thaw method in sterile deionized distilled water and the released DNA content was evaluated with a fluorometric method (Molecular Probes). A DNA standard curve [13], obtained from a known amount of cells, was used to express the results as cell number per disk. Set of rabbit polyclonal antisera: L.W. Fisher (National Institutes of Health, Bethesda, MD) presented us, generously, with the following rabbit polyclonal antibody immunoglobulins G: antiosteocalcin, anti-type-I collagen, anti-type-III collagen, antidecorin, and anti-osteopontin (antiserum LF-32, LF-67, LF-71, LF-136, and LF-166, respectively) [14]. Set of purified proteins: Decorin [15], osteocalcin (immunoenzymatic assay kit, BT-480, Biomedical Technologies), osteopontin (immunoenzymatic assay kit, 900-27, Assay Designs), type-I collagen [16], and type-III collagen (Sigma-Aldrich). Confocal microscopy: At the end of the culture period, the disks were fixed with 4% (w/v) paraformaldehyde solution in 0.1 M phosphate buffer (pH=7.4) for 8 h at room temperature and washed with PBS (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na2HPO4, 1.4 mM KH2PO4, pH=7.4) three times for 15 min. The disks were then blocked by incubating with PAT (PBS containing 1% [w/v] bovine serum albumin and 0.02% [v/v] Tween 20) for 2 h at room temperature and washed. L.W. Fisher’s antisera were used as primary antibodies with a dilution equal to 1:1000 in PAT. The incubation with the primary antibodies was performed overnight at 4°C, whereas the negative controls were based upon the incubation, overnight at 4°C, with PAT instead of the primary antibodies. The disks and the negative controls were washed and incubated with Alexa Fluor 488 goat anti-rabbit IgG (H+L) (Molecular Probes) with a dilution of 1:500 in PAT for 1 h at room temperature. At the end of the incubation, the disks were washed in PBS, counterstained with Hoechst solution (2 Pg/ml) to target the cellular nuclei, and then washed. The images were taken by blue excitation with the TCS SPII confocal microscope (Leica Microsystems) equipped with a digital image capture system at 100u magnification. Extraction of the extracellular matrix proteins from the cultured disks and ELISA assay: At the end of the culture period, in order to evaluate the amount of the extracellular matrix constituents over the gelatin surface, the disks were washed extensively with sterile PBS (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na2HPO4, 1.4 mM KH2PO4, pH=7.4) in order to remove the culture medium, and then incubated for 24 h at 37°C with 1 ml of sterile sample buffer (1.5 M Tris-HCl, 60% [w/v] sucrose, 0.8% [w/v] sodium dodecyl sulphate, pH=8.0). At the end of the incubation period, the sample buffer aliquots were removed and the total protein concentration in the two culture systems was evaluated by the BCA Protein Assay Kit (Pierce Biotechnology). The total protein concentration was 198 r 35 Pg/ml in the control culture and 332 r 51 Pg/ml in the differentiating culture (p0.05. Extracellular matrix extraction: In order to evaluate the amount of bone extracellular matrix onto the gelatin cryogel disks, an ELISA of the extracted matrix was performed: at the end of the culture period, in comparison with the control culture, the differentiative stimulation significantly increased the surface coating with decorin, osteocalcin, osteopontin, type-I collagen, and type-III collagen (p 0.05).
Indentation measurements were performed with the resonance sensor system on the gelatin tissue phantom. Five locations on the tissue phantom were subjected for a total of six measurements each (Figure 2). During each measurement the measured parameters F, ǻf, and d were registered (Figure 3) and from the linear relation between F and ǻf, the stiffness parameter (equation (1)) was calculated as the slope of the least squares line fit. IFMBE Proceedings Vol. 29
Stiffness of a Small Tissue Phantom Measured by a Tactile Resonance Sensor
at the periphery, the resistance to force of the sample was less due to the limited sample geometry. However, similar dependence was observed between ǻf and the sample geometry. In this case, the limited sample geometry was assumed to affect the load sensitive ǻf. The results of F and ǻf were shown at d = 1mm as an example (Figure 4), but similar results were observed at other d-values. At higher d the effect was more pronounced. The calculated stiffness parameter (equation (1)) measured the homogenous stiffness of the tissue phantom, unaffected by the limited sample geometry (Figure 4). This was assumed to be due to it being a combination of F and ǻf, where the geometry dependence was minimized. The variation in the measured parameters (Figure 4) includes effects due to measurement time, i.e. drying of the gelatin sample. These results suggest that the stiffness parameter is suitable for measuring the stiffness of prostate tissue samples which are of limited size. Therefore, in earlier studies [9, 10], it could be assumed that the effect of the limited prostate sample size on the estimated stiffness would be less. In this study a gelatin tissue phantom was used as a soft tissue mimicking phantom material because of it being relatively easy to prepare, and relatively easy to handle and cut into desired shape. In addition, the gelatin tissue phantom was homogeneous in stiffness. These properties of the tissue phantom were important in conducting this study.
F
d = 1 mm
(mN)
6 5 4 3 2
1
2 3 4 5 measurement location
Δf
d = 1 mm
(Hz)
180 160 140 120 100
1 2 3 4 5 measurement location
dF/dΔf (mN/Hz)
0.04
0.03
0.02
0.01
397
V. CONCLUSIONS
1 2 3 4 5 measurement location
Fig. 4 The mean and standard deviation of the force (F), frequency change (ǻf), and stiffness parameter (F/ǻf) shown for the five measurement locations (n = 6 on each location). Both F and ǻf are at an impression depth of d = 1 mm, while F/ǻf was calculated from the impression depth interval d = 0 to 1 mm.
IV.
The findings of this study suggest that the stiffness of a homogenous small tissue phantom can be measured with a tactile resonance sensor. The observed geometry independency of the measured stiffness parameter was promising for measurements on small biological tissue samples in vitro. Further studies must be done to determine the full value of the method and the reported findings. ACKNOWLEDGMENT
DISCUSSION
In this study the tactile resonance sensor technique was used to measure the stiffness of a tissue phantom sample of limited size. It was shown that the measured parameters of force and frequency change were dependent on the location of the indentation measurement. However, it was observed that the calculated stiffness sensitive parameter was independent of the indentation location. The force measurement alone will falsely estimate the elastic stiffness of the tissue phantom sample at the peripheral indentation locations (Figure 4). During an indentation
The study was supported by grants from Objective 2 Norra NorrlandEU structural Fund.
REFERENCES 1.
2.
Krouskop TA, Wheeler TM, Kallel F, Garra BS, Hall T (1998) Elastic moduli of breast and prostate tissues under compression. Ultrason Imaging 20:260-74 Greenleaf JE, Fatemi M, Insana M (2003) Selected methods for imaging elastic properties of biological tissues. Annu Rev Biomed Eng 5:57-78
IFMBE Proceedings Vol. 29
398 3.
4.
5.
6.
7.
8.
V. Jalkanen, B.M. Andersson, and O.A. Lindahl Murayama Y, Haruta M, Hatakeyama Y, Shiina T, Sakuma H, Takenoshita S, Omata S Constantinou CE (2008) Development of a new instrument for examination of stiffness in the breast using haptic sensor technology. Sensors Actuators 143:430-8 Omata S and Terunuma Y (1992) New tactile sensor like the human hand and its applications. Sensors Actuators 35:9-15 DOI 10.1016/0924-4247(92)87002-X Lindahl OA, Constantinou CE, Eklund A, Murayama Y, Hallberg P, Omata S (2009) Tactile resonance sensors in medicine. J Med Eng Technol 33:263-73 Jalkanen V, Andersson BM, Bergh A, Ljungberg B, Lindahl OA (2008) Explanatory models for a tactile resonance sensor system – elastic and density-related variations of prostate tissue in vitro. Physiol Meas 29:729-745 DOI 10.1088/0967-3334/29/7/003 Kusaka K, Harihara Y, Torzilli G, Kubota K, Takayama T, Makuuchi M, Mori M, Omata S (2000) Objective evaluation of liver consistency to estimate hepatic fibrosis and functional reserve for hepatectomy. J Am Coll Surg 191:47-53 Miyaji K, Furuse A, Nakajima J, Kohno T, Ohtsuka T, Yagyu K, Oka T, Omata S (1997) The stiffness of lymph nodes containing lung carcinoma metastases – a new diagnostic parameter measured by a tactile sensor. Cancer 80:1920-5
9.
Jalkanen V, Andersson BM, Bergh A, Ljungberg B, Lindahl OA (2006) Resonance sensor measurements of stiffness variations in prostate tissue in vitro – a weighted tissue proportion model. Physiol Meas 27:1373-1386 DOI 10.1088/0967-3334/27/12/009 10. Jalkanen V, Andersson BM, Lindahl OA (2009) Instrument towards faster diagnosis and treatment of prostate cancer – Resonance sensor stiffness measurements on human prostate tissue in vitro, IFMBE Proc. vol. 25(7), World Congress on Med. Phys. & Biomed. Eng., Münich, Germany, 2009, pp 145-48 The address of the corresponding author: Author: Institute: City: Country: Email:
IFMBE Proceedings Vol. 29
Ville Jalkanen Dept. of Applied Physics and Electronics, Umeå University 901 87, Umeå Sweden
[email protected] Vectorial magnetoencephalographic measurements for the estimation of radial dipolar activity in the human somatosensory system J. Haueisen1,2, K. Fleissig2 , D. Strohmeier1, R. Huonker2, M. Liehr2, and O.W. Witte2 1
Institute of Biomedical Engineering and Informatics, Technical University Ilmenau, Germany 2 Biomagnetic Center, Department of Neurology, University Hospital Jena, Germany
Abstract— Radial dipolar activity is difficult to estimate with standard single-component magnetoencephalographic measurements. However, recent technological developments allow for vectorial magnetoencephalographic measurements. We tested the hypothesis that radial dipolar activity can be estimated on the basis of vectorial magnetoencephalographic measurements. Eleven healthy participants received right median nerve stimulation, while the biomagnetic field was measured over the contralateral hemisphere with a novel vector-biomagnetometer. In this measurement system, SQUID based magnetometer sensors are arranged in perpendicular triplets at each measurement location. Subsequently, source analysis of radial and tangential cortical dipoles was performed. We found that both radial and tangential dipolar activity could be estimated in ten out of eleven participants. Dipole locations were found in the vicinity of the central sulcus and dipole orientations were predominantly tangential for the first cortical activity N20m and predominantly radial for the second cortical activity P25m. The mean location difference between the tangential and the radial dipole was 11.9 mm and the mean orientation difference was 97.5 degree. We conclude that radial dipolar activity can be estimated from vectorial magnetoencephalographic measurements. Keywords— magnetoencephalography, somatosensory cortex, median nerve, magnetic field measurements.
[4] and in simulation studies [5] that radially oriented dipoles produce 4 to 10 times weaker magnetic fields outside the head than tangentially oriented dipoles at the same position. Thus, MEG is considered to be not appropriate for the localization of radial sources. However, all previous studies used only one vectorial component (either Bz or Br values) of the vectorial biomagnetic field. Recently, vector-biomagnetometers were developed [6-8], which allow for the recording of all three vectorial components of the biomagnetic field. It has been shown in measurements [6] and simulations [9] that the information content is higher for measurements performed with a vector-biomagnetometer as compared to standard biomagnetometers. The aim of this study is to investigate experimentally the feasibility of reconstructing radially oriented dipoles based on vectorial measurements of biomagnetic fields. We have chosen the somatosensory system for its well known generator structure of a radial (Brodmann area 1) and tangential (Brodmann area 3b) dipolar source overlapping in time.
MEG, primary three-component
II.
METHODS
A. Participants and measurements I. INTRODUCTION
Electroencephalography (EEG) and magnetoencephalography (MEG) are non-invasive modalities for recording brain activity with high temporal resolution. Source reconstruction based on EEG and MEG is a widely used technique in the neurosciences allowing for spatio-temporal disentanglement of overlapping brain activity. Commonly, dipolar source models are employed in order to describe the electrical activity in a certain brain area. EEG and MEG have different sensitivities with respect to superficial and deep sources and also radial and tangential sources (e.g. [1]). Consequently, the combination of MEG and EEG is used to distinguish radial and tangential as well as superficial and deep sources [2, 3]. One specific property of MEG is its lower sensitivity to radial sources as compared to tangential sources. It has been shown both experimentally
Eleven healthy volunteers (6 males, 5 females), 10 righthanded and one left-handed, underwent examination with the ARGOS 200 vector-biomagnetometer (AtB SrL, Pescara, Italy) positioned over the somatosensory cortex. Contralateral to the magnetic recording site the median nerve was electrically stimulated (stimulation strength: motor plus sensor threshold; constant current square wave impulses with a length of 200 s). The stimulus onset asynchrony was randomized between 0.7 and 1.4 s. A total of 512 epochs was averaged. Data were sampled with 1 kHz. For artifact detection ECG, horizontal and vertical EOG was recorded. Isotropic T1-weighted magnetic resonance (MR) images with 1 mm resolution were taken of each participant’s head to provide realistic head modeling for the source localization procedure. Co-registration between MRI and MEG coordinate systems was obtained by digitizing and
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 399–401, 2010. www.springerlink.com
400
J. Haueisen et al.
rigidly transforming anatomical landmarks (nasion, left and right pre-auricular point). The study was approved by the ethics committee of the medical faculty of the Friedrich-Schiller-University Jena. All participants gave their written informed consent. B. Vector-biomagnetometer The used vector-biomagnetometer includes 195 Superconducting QUantum Interference Devices (SQUIDs). The sensors are fully integrated planar SQUID magnetometers produced using Nb technology with integrated pick-up loops. The sensing area is a square of 8 mm side length. The intrinsic noise level of the SQUIDs is below 5 fT Hz–1/2 at 10 Hz. Triaxial vector magnetometers are formed by grouping three basic sensor elements to a triplet. In each triplet the magnetic field vector, Bx, By, and Bz (with respect to the local coordinate system of the triplet) is measured, since the three square sensor elements are arranged perpendicular to each other on the three adjacent planes of the corner of a cube (Fig. 1 inset). In order to have a similar distance of all three SQUIDs to the bottom of the measurements system, the corner of the cube is placed closest to the bottom of the cryostat (the cube is standing on its corner). An additional advantage of this arrangement is that one can obtain the commonly measured Bz component (i.e. perpendicular to the cryostat bottom) of the magnetic field by simply adding the magnetic flux measured by all three SQUIDs.
The triplets are distributed over four levels. The lower level, the main measurement plane, is a planar sensor array consisting of 56 sensor triplets, laid out on a hexagonal grid, covering a circular planar surface with a diameter of about 25 cm (Fig. 1). The second level contains seven triplets, and the third and fourth level one triplet each. The second level is on a plane which is positioned parallel to the measurement plane at a distance of 98 mm. The third level is 196 mm above the first plane, and the fourth level is 254 mm above the first plane. The centre (cube corner point) of the triplets in the third and fourth level is located at the x, y position (0,0). The triplets located in the second, third, and fourth level are used for noise cancellation. The dynamic range of the SQUID electronics is 22 bit with a lowest resolution of 2.05 fT and a range of ±4.31 nT. The system is installed in a magnetically shielded room with three layers of highly permeable material and one layer of aluminum. For visualization purposes, the local Bx, By, and Bz at each triplet (note the different orientation of the triplets in Fig. 1) were transformed to global Bx, By, and Bz values at each center of gravity of the triplet. C. Source localization For each participant a realistic three compartment boundary element model to account for skin, skull, and brain was derived by segmenting the MR images. The triangle side length was set to 7 mm for each of the surfaces. We assumed conductivities of 0.33, 0.0042 and 0.33 S/m (skin, skull, brain). MEG data preprocessing consisted of artifact rejection, filtering (3rd order Butterworth 0.1 – 170 Hz), and baseline correction (-100 to 0 ms). A two step spatio-temporal dipole localization procedure was performed considering the time interval of the first cortical components N20m and P25m. First in the upstroke of the N20m a single dipole was fitted. This source activity was then projected out for the entire time interval and a second dipole was fitted representing mainly the activity of the P25m component. Source localization was done with Curry version 4.6 (Compumedics NeuroScan, Charlotte, NC, USA). III. RESULTS
Fig. 1: ARGOS 200 sensor configuration.
Fig. 2 shows an example of the measured magnetic field distributions. The Bz field pattern is similar to the field pattern measured with standard biomagnetometers and shows a typical dipolar arrangement for the tangential component and a monopolar arrangement for the radial component. For the tangential component, the Bx field pattern shows a slightly quadrupolar arrangement (two IFMBE Proceedings Vol. 29
Vectorial Magnetoencephalographic Measurements for the Estimation of Radial Dipolar Activity in the Human Somatosensory System
negative maxima in the center of the field pattern) and the By field pattern a tri-polar arrangement. For the radial component, both Bx and By show dipolar arrangements. The center points of the field patterns of the tangential and radial components are slightly different.
401
might provide more insight into brain activity as compared to standard magnetoencephalography.
ACKNOWLEDGMENT This work was in part supported by the Deutsche Forschungsgemeinschaft (DFG grant Ha 2899/7/8-1) and the BMBF grant 03IP605.
REFERENCES
tang
rad
Bx
By
Bz
Fig. 2: Measured magnetic field distributions of one volunteer for tangential N20m (top row) and radial P25m (bottom row). Small squares indicate triplet positions and small crosses indicate omitted channels due to artifacts (if one channel was omitted always the entire triplet was switched off). Line increment is 50 fT in the top row and 20 fT in the bottom row. We found the expected source locations for both dipoles (N20m and P25m) in the vicinity of the central sulcus for ten out of eleven participants. In one out of eleven participants the radial P25m component could not be localized reliably. The mean distance between the radial and the tangential dipole was 11.9 ±5.4 mm, which is within the expected range for the distance between the underlying Brodmann areas 3b an 1. The angle between the two dipoles was found to be 97.5 ±28.5 degrees, which is also in accordance with the expected range of angles.
[1] Goldenholz DM, Ahlfors SP, Hämäläinen MS et al. (2009) Mapping the signal-to-noise-ratios of cortical sources in magnetoencephalography and electroencephalography. Hum Brain Mapp 30(4):1077-86 [2] Wood CC, Cohen D, Cuffin BN, et al. (1985) Electrical Sources in Human Somatosensory Cortex - Identification by Combined Magnetic and Potential Recordings. Science 227(4690):1051-1053 [3] Jaros U, Hilgenfeld B, Lau S, et al. (2008) Nonlinear interactions of high-frequency oscillations in the human somatosensory system. Clinical Neurophysiology, 119(11):2647-57 [4] Melcher JR, Cohen D. (1988) Dependence of the MEG on dipole orientation in the rabbit head. Electroencephalogr Clin Neurophysiol 70(5):460-72 [5] Haueisen J, Ramon C, Czapski P et al. (1995) On the Influence of Volume Currents and Extended Sources on Neuromagnetic Fields: A Simulation Study. Ann Biomed Eng 23(11):728 - 739 [6] Kobayashi K and Uchikawa Y (2001) Estimation of multiple sources using spatio-temporal data on a three dimensional measurement of MEG. IEEE Trans Magn 37:2915–2917 [7] Schnabel A, Burghoff M, Hartwig, F et al. (2004) A sensor configuration for a 304 SQUID vector magnetometer. Neurol Clin Neurophysiol 70 [8] Liehr M and Haueisen J (2008) Influence of anisotropic compartments on magnetic field and electric potential distributions generated by artificial current dipoles inside a torso phantom. Phys Med Biol 53:245–254 [9] Arturi CM, Di Rienzo L and Haueisen J (2004) Information Content in Single Component versus Three Component Cardiomagnetic Fields. IEEE Trans Magn, 40(2):631-634 Corresponding author:
IV. CONCLUSIONS
To our knowledge this is the first study which reconstructs radially oriented dipoles based on vectorial biomagnetic measurements. We conclude that radial dipolar activity can be estimated from such measurements and that three-component magnetoencephalographic measurements
Author: Jens Haueisen Institute: Institute of Biomedical Engineering and Informatics, Technical University Ilmenau Street: Gustav-Kirchhoff Str. 2 City: Ilmenau Country: Germany Email:
[email protected] IFMBE Proceedings Vol. 29
Registration of Chest X-Rays J. Csorba, B. Kormanyos, and B. Pataki Budapest University of Technology and Economics, Department of Measurement and Information Systems, Budapest, Hungary
[email protected],
[email protected],
[email protected] Abstract— Doctors often want to compare X-ray images of one patient made at different times. In case of chest X-ray analysis it is hard to find the typically small differences of the two images. A software supporting this method could help them a lot. We analyzed the possible methods and developed a prototype of a feature-based registration software. Characteristic pairs of points were used for the registration transformation. These points were found automatically on the contour and in the inner part of the lung. Several transformations (rigid and non-rigid ones) were examined to find the best one for registration. Keywords— chest X-ray, medical image registration, feature-based registration, rigid and non-rigid transformations, disorder detection.
I. INTRODUCTION Lung cancer is a leading cause of death worldwide. It is responsible for 1.000.000 deaths yearly according to WHO statistics. Cancer death rate could be decreased if cases were detected and treated early. X-ray is one of the cheapest common tools for diagnosis. During examinations the doctor always compares the patient’s past X-ray images, if available, with the present one. It is a difficult task, and even expert physicians can easily make a mistake, especially when high number of patients must be examined. A little fault can lead to misdiagnosis so it is important to develop supporting software for this job. The core of the support software is the registration of two X-ray images. We analyzed and rated the possible methods of the registration and created a prototype of the software which can support doctors' work. An automatic method without any human interaction was chosen because a large number of images are to be processed and doctors have no time to help the program. There are intensity-based and feature based methods of registration. [1] Because the different images have different exposition characteristics the feature based method based on representative points of the lung was chosen.
II. ALGORITHM STRUCTURE A. Main Structure First we preprocessed the two X-ray images made at different dates from the same person. Then pairs of characteristic points were looked for: one representative point in the current image and its counterpart in the older one. First we searched these points on the contour of the lungs rather than in the inner parts of the lungs. We cleared from the set the pairs which are possibly mismatches and then transformed the older image to the same position as the current one using the cleared set of the point-pairs. B. Preprocessing In this first step the brightness of the images were normalized based on the area of the lungs. After that we used a recently developed tool [2] which can remove the clavicles and the ribs from X-ray images. The removing of the bones is important because they could also generate some false pointpairs. These pairs can not be used for registration because the bones are often moving independently from the lung tissue. C. Contour Points First we searched points on the contour of the lungs. On the different areas there were different difficulties such as a low signal-noise ratio near the mediastinum or the air bubbles in the stomach which could be a problem in finding the inner or the bottom edge of the lung. Because of these problems several methods had to be used to get the full contour of the lung. Topmost point: The first point that we determined was the topmost point of the lung. To find this we searched candidate points from the image which are in the expected region (in the top of the lung) and there is a big intensity difference between their upper and bottom neighbours. When we got the candidates we had to choose the real top point. There have to be a lot of similar points in its surrounding to make sure it is not just an isolated candidate because of some random noise and this candidate also have to be in as top position as possible because it is the top of the lung.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 402–405, 2010. www.springerlink.com
Registration of Chest X-Rays
403
Bottom line: After the top point we determined the bottom contour of the lung. For that we need a reliable point of that contour part. We searched candidates here too. These points were in the expected region of the bottom of the lung and their vertical neighbours had a big intesity difference. There also had to be a lot of other candidates in its surroundings to avoid false hits. The next step was a Sobel transformation on the image to heighten the vertical contours. In the new image the original contour is a high intensity ridge. In this ridge we can go along on the high intensity path from the reliable starting point to both the left and the right side. Because of the big noise we used the following algorithm. From a point already assigned to the ridge we search the angle in which the average intensity on a vector with a given length is the biggest and we step to the next contour point along that vector. We continue this algorithm till the average intensity is big enough. After we had found points on the bottom contour on the two images we had to pair them. The algorithm almost always finds the most outer points of the bottom line but sometimes it makes some mistakes at the most inner points (closest to the mediastinum) because of the low signal to noise ratio near the mediastinum. That is why we started pairing the points on the two images from the most outer points going along the contour. Sides: Now we know the top point and the bottom outermost one of the contour. The outer side of the contour is a line connecting them. It is also true for the inner side. To find these lines we searched points with proper intensity differences in there surrondings. These candidates could be points of the contour line. Knowing the ends of the line and the possible points of it (the candidates) we can find the whole line. We had to find point pairs in the contour lines of the two images. On the outer side we did not have characteristic points because the outer part of the lung is very homogeneus. Our only information was the position of a point on the line. We also did not have much information about the inner points because of the low signal to noise ratio near the mediastinum. Considering these problems we simply split every section of the contour to some smaller parts with equal length. The selected points were the ends of these small parts and we could pair them in the two images one after the other. D. Inner Points Looking for characteristic points: Finding characteristic inner points is a difficult task also for the human brain. But there are some clearly identifiable parts of the lung: branches of the vessels, enlargements of (blood) vessels,
and different pathological disorders of the lung. The common in these helpful parts is that they all appear on the Xray as a nodule. The detection of these nodules could be a starting point in order to find coherent pairs of points. To detect nodules we apply a small fixed sized window – called scanning window – to every pixel [3]. We count the average I avg and the minimal I min grey levels inside the scanning window. If the grey level of the actual (central) pixel is less than a suitable threshold, then we mark the pixel as a part of a suspected nodule. The threshold is calculated as a weighted average of I min and I avg . Thus the pixel (i , j ) is possibly part of a nodule, if its intensity:
I (i, j ) ≤ w1 I avg + w2 I min
(1)
The size of the scanning window determines the size of nodules we want to find while the threshold determines how large a change in the contrast we would like to detect. Our preliminary selection is the following: the length and height of the window are 1/50 of the original size of the images, and the weights are w1 = 0.25, w2 = 0.75 .
Parameters of the points: We may assign to each nodule its center of gravity calculated from the points belonging to this nodule. We can use these center points: this step does not cause significant inaccuracy, because the diameter of the nodules is small. Different features were defined in order to pair the nodule center points in, which were found on the two images independently: • •
•
Morphological parameters of the nodules (area, diameter and a commonly used shape factor – the circularity). “Characteristic” parameter of the nodule: This is calculated from the difference between the intensity of the points in the nodule and the above threshold. The more „characteristic“ a nodule is the better it is marked out from the image. We may assign a one variable function to every single point in the following way: Scan the picture with radial lines from the center of a nodule with a given length. Count the average of the intensity of the pixels, which can be found along the line. If we do this by different angles then we get a function. This function could be used to pair the center of the nodules on different images, because its values are similar on corresponding points and significantly different on non-corresponding ones.
Pairing the points: If we want to find the best fitting pair of a point marked in the old image without having any preliminary information, we theoretically need to check all characteristic points found in the new image. This situation
IFMBE Proceedings Vol. 29
404
J. Csorba, B. Kormanyos, and B. Pataki
could be significantly improved, when the contour point pairs are used to calculate an average translation between the old and new lung images. To determine the pair of a point we just need to look in the neighborhood of the translated position. In this neighborhood that point is chosen, which has the smallest distance in the space of the above defined features. The reliability of the pair is inversely proportional to the distance.
The pairs of points found can be used to create the registration transformation. Different transformations were examined. In the following part we will present the mappings, which were analyzed. We start with different well known rigid transformations and continue with some non-rigid mappings.
E. Data Cleaning
λ , rotates by α degrees and translates with a vector
F. Transformations
Linear transformation: This map enlarges with a factor
A consistency test is made with the existing pairs of points for the sake of reducing the mismatched pairs. For this purpose we define to every pair of points the “shift vector” as the difference of the coordinates of the points of a pair (new – old). We expect that for two different point pairs, if the two points on the old image are close to each other, then the difference of the two “shift vectors” cannot be large. Based on this we do the following: Start with an arbitrary pair and calculate the “shift vector”. After that, take the pairs found one by one. In each step choose a new pair. It is accepted if its shift vector does not differ too much from the shift vectors of any previously chosen pairs. “Not too much” means here that the length of the difference of the shift vectors is not larger than 10% of the distance of the points of the pairs on the old image. If a point is not accepted than there is inconsistency (between our pairs). At this situation we throw away the less reliable pair from the two pairs in conflict. After this procedure we get a set of consistent pairs.
( d x , d y ) , so it has 4 free parameters.
Affine transformation: The image of the point ( x0 , y 0 ) is given by:
⎛ A B⎞ ⎟ ⎜ (u 0 , v0 ) =( x0 , y 0 ,1) × ⎜ C D ⎟ ⎜E F ⎟ ⎠ ⎝
(2)
So it has 6 free parameters. Projective transformation: The image of the point ( x h , y h , w) , which is given in a homogeneous coordinate can be obtained in the frame of reference:
⎛A D G⎞ ⎟ ⎜ (u , v, h) = ( x h , y h , w) × ⎜ B E H ⎟ ⎜C F I ⎟ ⎠ ⎝
(3)
This transformation is given by 8 free parameters. Polynomial transformation: These can be considered as a generalization of the affine transformation. In affine case the coordinate functions are linear combinations of first degree polynomials, while here the coordinate functions are linear combinations of higher degree polynomials. Local weighted mean transformation: If a point ( x, y ) in the original image is close to a control point (characteristic point found and paired) in this image then we expect the point corresponding to ( x, y ) in the transformed image to be close to the control point (in the other image) [4]. In the original image take the N closest control points to ( x, y ) and denote them by: P , P … P . Now we can calculate 1
2
N
the (u , v ) point, which is corresponding to ( x, y ) using fittingly weighted average of P , P … P . 1
Fig. 1 The pairs of points that have been found
2
N
Interpolation based transformations: A transformation could be imagined as a two-variable function with two values. This function is known at the already found pairs of
IFMBE Proceedings Vol. 29
Registration of Chest X-Rays
405
points. A two variable function with two values can be considered as two two-variable functions with one value. They are surfaces, which are defined in some dots. The surfaces, which are important to the transformation, can be determined with a chosen interpolation procedure.
III. TEST RESULTS 8 pairs of images (an old and a new X-ray image of the same patient) were used for testing. In each pair a human expert marked 8-8 inner and 8-8 contour points, which were coherent. We could measure the difference (in pixel) between the marked and transformed control old points and the corresponding marked points in the new image. The mean square error was calculated from the differences. This procedure describes the accuracy of the whole registration and compares the transformations indirectly through this. The resolution of the original images was 2000*2000 pixels. We got the following errors of the different transformations (errors are averaged over the 8 images).
IV. CONCLUSIONS Based on our test results, the second degree polynomial transformation is the closest to the real deformation of the lung among the examined mappings. The reached 20 pixels imprecision is totally acceptable for pictures with similar resolution. The mistake is just barely visible to the human eye. Thus the second degree polynomial is proposed to model the shift of the lungs. The sufficiently accurate result permits the algorithm to be used in different applications. An opportunity is alternate display of the transformed old image and the new one. Another possibility is to display the “difference picture” in order to detect the change of the hardly noticeable nodules. Both tools are also useful to make doctors’ diagnostic work easier, and more efficient. The algorithms of these two functions were implemented. Results of this application are reassuring to create highprecision automated detection software in the future.
Table 1 Results of registrations by transformations Transformation Linear
Average error (in pixels) 25.38
Affine
25.13
Projective
26.75
Polynomial(2 order)
23.88
Polynomial(3 order)
31.75
Local weighted mean
31.13
Interpolation(linear)
27.50
Interpolation(cubic)
29.50
Fig. 2 New image, old image and difference image
ACKNOWLEDGMENT This work was partly supported by the National Development Agency under contract KMOP-1.1.1-07/1-2008-0035.
The rigid transformations performed better than the nonrigid mappings. Among these transformations the second degree polynomial turned to be the most efficient one. This can be explained by the degree of freedom (DOF) of the linear, affine and projective transformation being too small to model the deformations of the lung. However, the third degree polynomial (mapping) with its 20 DOF could already make too special changes. One possibly wrongly found pair of points in case of non-rigid transformations could cause a mistake in the neighborhood of the point. Applying it is too risky. In contrast to this, rigid transformations are fault-tolerant, which is an important aspect of decision support software for doctors.
REFERENCES [1] Medical Image Computing and Computer-Assisted Intervention — MICCAI 2002, 5th International Conference Tokyo, Springer, 517524, 2002 [2] G. Horváth, G. Orbán, Á. Horváth, G. Simkó, B. Pataki, P. Máday, S. Juhász and Á. Horváth, “A CAD System for Screening X-ray Chest Radiography” World Congress 2009 Medical Physics and Biomedical Engineering. Vol. 25, 210-213, 2009 [3] Kim Le, "Automated Detection of Early Lung Cancer and Tuberculosis Based on X-Ray Image Analysis", the 6th World Scientific & Engineering Academy and Society (WSEAS) International Conference, 2006 [4] Ardeshir Goshtasby, "Image Registration by Local Approximation Methods", Image and Vision Computing. vol. 6, no. 4, 255-261, 1988
IFMBE Proceedings Vol. 29
Arterial Pulse Transit Time Dependence on Applied Pressure K. Pilt, K. Meigas, M. Viigimaa, J. Kaik, R. Kattai, and D. Karai Tallinn University of Technology/Technomedicum/Department of Biomedical Engineering, Tallinn, Estonia Abstract— Arterial pulse transit time dependence on applied pressure is analyzed experimentally. The pressure was applied on brachial artery. Pulse transit times between left and right hand were compared by calculating the correlation on different applied pressures. In addition the pulse transit time characteristics were analyzed on different pressures. It revealed that pulse transit time is not influenced by applied pressure, when it is lowered to certain level. The level can be located from piezoelectric signal amplitude. Keywords— Pulse transit time, PPG, piezoelectric transducer, applied pressure.
I. INTRODUCTION Pulse transit time (PTT) is the time of a pulse wave to travel between two arterial sites. The time delay between two registered pressure waves, which are measured from the same artery, is directly proportional to blood pressure [1]. PTT depends on arterial stiffness. When arterial wall becomes stiffer it causes the PTT to be shortened and opposite. In addition to blood pressure rise, the arterial stiffness is also increased with age, arteriosclerosis and diabetes mellitus, which resulting in a shortening of the PTT [2-4]. Different methods have been used to register the pressure wave, such as piezoelectric pressure sensor, PPG, Doppler ultrasound, etc. PPG is a non-invasive optical technique for measuring changes in blood circulation. This method is widely used in oxygen saturation measurement, but it can be used also to register the pressure wave. The optical radiation from the light emitting diode (LED) is directed to the skin, which is often red or infrared. The light is absorbed, reflected and scattered in the tissue and blood. Only a small fraction of light intensity changes are received by the photodiode. Pulsating AC component of the registered PPG signal corresponds to pressure wave. There are two main PPG sensor design modes: the reflection and the transmission mode. In the reflection mode, a photodiode is placed adjacent to the LED and directed toward skin. The photodiode measures the reflected and scattered light intensity from the skin surface. In the transmission mode, the photodiode and the light source are placed on opposite sides of the measured volume. The photodiode measures the transmitted light intensity. The transmission sensor
measurement sites are limited, because of its geometrics, whereas reflection mode sensor can be placed to any point on the skin surface. Still the PPG signal has been easier to obtain from fingers, toes and earlobes by using transmission mode sensor (unpublished observations). Mechanical pulsation of the arteries can be registered with piezoelectric transducer. It is relatively simple method for the detection of the pressure wave. Piezoelectric transducer generates measurable voltage when a deforming mechanical force is applied. Piezoelectric transducer can be used to measure pressure wave from the measurement sites, where it is difficult to obtain the PPG signal. For example on elbow surface the brachial artery is hidden behind other type of tissues and the obtained PPG signal is with low signal-to-noise ratio (SNR). This article is about preliminary study for wider research in the area of diagnosis of cardiovascular diseases by using PTT as one of the parameter. The pressure wave measurement sites are wrist and elbow. The pressure wave signal from the brachial artery is planned to obtain with piezoelectric transducer. It is placed above the artery and should be fixed with ribbon, which is connected around elbow. In this article has been analyzed how additional stress from piezoelectric transducer influences the brachial artery and the pressure wave transit time.
II. METHODS To obtain piezoelectric signal with high SNR the transducer should be applied on the skin surface above the brachial artery with additional pressure on it. The brachial artery is closed for blood flow, when there is applied certain amount of pressure. When the pressure is lowered the blood flow recovers and it is possible to register the pulsating pressure wave. On certain pressure the piezoelectric signal is with maximum amplitude. Signal amplitude starts to lower again, when the pressure is lowered. At the optimal pressure the piezoelectric signal should be registered with as high signal amplitude as possible. At the same time the transducer pressure, which is applied on brachial artery, should not influence the pressure wave velocity, as the second sensor is placed on the same hand radial artery.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 406–409, 2010. www.springerlink.com
Arterial Pulse Transit Time Dependence on Applied Pressure
407
Fig. 1 Experimental device to analyze the dependency between PTT and applied arterial pressure The time interval, which is measured between the peak of R-wave on the electrocardiographic (ECG) and raising front of the pressure wave signal, can be taken as PTT [5]. In this article we denote this time interval as R-wave gated pulse transit time (RWPTT). In simplified conditions the arteries, which are leading from the heart to left and right hand radial arteries in wrist, are with identical parameters. The RWPTTs for both hands are comparable and with high correlation. Left hand brachial artery mechanical properties are influenced by applying pressure on it. Due to it the correlation, between left and right hand RWPTTs, starts to lower. At the moment, when the brachial artery is closed, there is no correlation between two time intervals. The experimental device was built to analyze the PTT dependency on piezoelectric transducer pressure. The device schematic is given on Figure 1. It consists of solid plastic tube, which is with enough large, that human hand can fit through it. Piezoelectric transducer is fixed inside the tube and to the opposite side of it is placed cuff, what can be filled with air. The pressure, what is applied on cuff is measured with manometer. Stem is fixed inside the tube and top of it is connected piezoelectric transducer. During the experiment subject's hand is placed through the tube and placed on cuff. The hand is placed in the way that transducer is aimed on the brachial artery above elbow. The position of the brachial artery is located through palpation previously. Cuff starts to expand under hand by pumping the air into it and the piezoelectric transducer is getting contact with skin surface. By continuing with air pumping the pressure under elbow starts to rise and the direct force from transducer is applied on artery. Three different pressure wave signals are registered synchronously during the experiment. Piezoelectric signal is registered from transducer, which is fixed inside plastic tube and placed on left hand. Two PPG signals are measured with reflectance sensors from the left and right hand wrist.
Fig. 2 Three different R-wave gated pulse transit times are measured between ECG signal and pressure waves In addition ECG signal, which describes the electrical activity of the heart, was registered synchronously. For all three registered pressure wave signals the corresponding RWPTTs are measured: RWPTTPPG - time interval between ECG and left hand PPG signal; RWPTTrefPPG – time interval between ECG and right hand PPG signal; RWPTTpiezo - time interval between ECG and piezoelectric signal (Figure 2). According to previous research, the RWPTT is suggested to be measured between 50% of the PPG signal raising front and ECG signal R-peak [6]. Pressure wave front from piezoelectric signal was detected from its every period maximum point. RWPTTrefPPG corresponds to pressure influenced time interval and RWPTTPPG corresponds to reference time interval. To analyze the pressure influence on artery the correlation are calculated between RWPTTPPG and RWPTTrefPPG.
III. RESULTS Experiments were conducted on 4 healthy volunteers. On every volunteer was carried out 3 experiments. The ages of the subjects were 22, 31, 38 and 61. During the experiments the room temperature was around 24 degrees and subjects were in resting position.
IFMBE Proceedings Vol. 29
408
K. Pilt et al.
Fig. 3 Four conducted experiments results. Correlations: rPPG, rpiezo, rstart, rstop; and the amplitude of piezoelectric signal on different applied pressures. a) 38 years old subject b) 61 years old subject c) 31 years old subject d) 23 years old subject
Raw PPG and piezoelectric signals were obtained with lab built circuit. For the PPG signals the Nellcor Max-Fast reflectance sensors were used. PPG sensors were placed on left and right hand wrist. Sensors were fixed on skin surface above radial artery, which was located through palpation. For the piezoelectric signal the ADInstruments MP100 transducer was used. PPG and piezoelectric signals were digitalized with National Instruments PCI MIO-16-E1 data acquisition card and registered with LabVIEW environment. ECG signal was obtained and digitalized with ADInstruments PowerLab 4/20T device and registered with ADInstruments Chart software. All the signals were digitized with sampling frequency 1kHz. Before experiments it was ensured, that signals are registered synchronously. The signals were recorded continuously during the whole experiment. The subject's left hand was fixed inside previously described experimental device. The cuff under hand was filled with air. At certain pressure piezoelectric sensor closed the blood flow in brachial artery. The closure of artery was detected from left hand PPG signal. The pressure in cuff, which was measured with manometer, was lowered step-by-step. Each pressure level was marked up synchronously with recorded signals. Before and after every experiment the signals were recorded without pressure on artery to determine the correlation between RWPTTPPG and RWPTTrefPPG afterwards. Post processing of the signals was carried out in MATLAB. The signals were filtered with high- and lowpass FIR filters. The cut-off frequencies were 0.1Hz and 30Hz respectively. R-peaks of the ECG signal were detected by using Hamilton-Tompkins algorithm [8]. Pulse wave rising fronts were detected from PPG and piezoelectric signals as it was described previously. Three different intervals: RWPTTPPG, RWPTTpiezo and RWPPTrefPPG; were measured between R-peak of the ECG signal and pressure waves. For each pressure the two correlations were calculated between RWPTTs: rpiezo is correlation between RWPTTpiezo and RWPTTrefPPG; rPPG is correlation between RWPTTPPG and RWPTTrefPPG. In addition reference correlations rstart and rstopp were calculated between RWPTTPPG and RWPTTrefPPG. Time intervals for rstart and rstop calculation were measured from part of the signals, which were recorded before and after the pressure was applied respectively. All correlations were calculated by using constant number of time intervals. On Figure 3 has been shown for each subject the average amplitude of the piezoelectric signal and two correlations rpiezo and rPPG dependency on different applied pressures on cuff. In addition on every graph has been given two level marker lines, which corresponds to rstart and rstopp.
IFMBE Proceedings Vol. 29
Arterial Pulse Transit Time Dependence on Applied Pressure
Fig. 4 From
23 year old subject measured RWPPTs on different applied
pressures
The correlations rPPG and rpiezo are given starting from the pressure, when there was detectable pulsating PPG signal. It is visible, that in every experiment when the rPPG is below 0.5 the piezoelectric signal amplitude is near its maximal value. Low correlation means that pressure on artery influences the pulse transit time. By lowering the pressure on the artery the rPPG increases near 1. The correlation varies from pressure to pressure around rstart and rstop. It can be assumed that pulse transit time is not influenced by the pressure on artery. On Figure 3 b) the rPPG follows the rpiezo. Near cuff pressure of 15mmHg the rPPG is decreased. It means that the properties of artery in right hand differ from left hand. The difference of the arteries is not caused by the transducer applied pressure, because also the rpiezo is decreased. rpiezo describes the correlation between RWPTTs in left and right hand without any applied pressure. On Figure 3 b) and d) the correlation is firstly decreases to its minimum and then starts to increase while the pressure is decreased. On Figure 4 is given RWPTTs which correlations are shown on Figure 3 d). The RWPTTPPG is higher than RWPTTrefPTT in the beginning. After decreasing the pressure the time interval transition is taking place. It means that high correlation around pressures 21-25mmHg, which is shown on Figure 3 b), does not describe the applied pressure influence. It is visible that the high pressure on artery causes constant increase in RWPTT. Similarly the high correlation around pressure of 23mmHg is explained on Figure 3 b).
IV. CONCLUSIONS The PTT dependence on applied arterial pressure was analyzed experimentally. The pressure was applied on left
409
hand brachial artery with piezoelectric transducer. The RWPTTs were measured between ECG signal R-peak and pressure waves, which were registered from left and right hand radial arteries. The correlations rpiezo and rPPG were calculated on different applied pressures. In addition the RWPTT characteristics were analyzed on different applied pressures. It can be conclude that PTT is not influenced by applied pressure, when it is lowered to certain level. The pressure level can be set according to piezoelectric signal amplitude. The piezoelectric signal amplitude reaches its maximum when the pressure is lowered. On lower pressures, than the maximum amplitude is located, the pulse transit time is not dependent on applied pressure. As future work the pulse shape should be analyzed on different pressures. In addition the problem should be analyzed analytically.
ACKNOWLEDGMENT This study was supported by the Estonian Science Foundation (grant No. 7506), by the Estonian targeted financing project SF0140027s07, and by the European Union through the European Regional Development Fund.
REFERENCES 1. Allen J (2007) Photoplethysmography and its applications in clinical physiological measurement. Physiol Meas 28:R1-R39 2. Smith R. P, Argod J, Pepin J. L, Levy P. A (1999) Pulse transit time: An appraisal of potential clinical applications. Thorax 54:452-457 3. Hlimonenko I, Meigas K, Viigimaa M, Temitski K (2008) Aortic and Arterial Pulse Wave Velocity in Patients with Coronary Heart Disease of Different Severity. Estonian J Eng, 14(2):167-176 4. O’Rourke M. F, Hayward C. S (2003) Arterial stiffness, gender and heart rate. J Hypertens 21:487-490 5. Naschitz J. E et al (2005) Pulse transit time by R-wave-gated infrared photoplethysmography: Review of the literature and personal experience. J Clin Monit Comput 18:333-342 6. Lass J, Meigas K, Kattai R, Karai D, Kaik J, Rosmann M (2004) Optical and electrical methods for pulse wave transit time measurement and its correlation with arterial blood pressure. Proc Estonian Acad Sci Eng 10:123-136 7. Hamilton P. S, Tompkins W. J (1986) Quantitative investigation of QRS detection rules using the MIT/BIH arrhythmia database. IEEE Trans Eng Biomed Eng 12:1157-1165
Author: Institute: Street: City: Country: Email:
IFMBE Proceedings Vol. 29
Kristjan Pilt Department of Biomedical Engineering Ehitajate tee 5 Tallinn, 19086 Estonia
[email protected] Influence of an Artificial Valve Type on the Flow in the Ventricular Assist Device D. Obidowski, P. Klosinski, P. Reorowicz, and K. Jozwik Institute of Turbomachinery, Technical University of Lodz, Lodz, Poland Abstract— Numerical fluid mechanics methods play a significant role in the design process of many devices, thus increasing comfort of life. One of the fields where their application is of the widest interest is medicine. The authors used the latest Computer Aided Design and Computational Fluid Dynamics software to analyze the flow in the pneumatic Ventricular Assist Device (VAD) with two different types of valves applied. In the study, a MEDTRONIC-HALL tilting disc mechanical artificial heart valve was compared with a three-leaflet polyurethane artificial heart valve designed at the Foundation of Cardiac Surgery Development, Zabrze. A comparison was made on the basis of the flow visualization inside the VAD chamber and the size of stagnation regions where the flowing blood may coagulate. The presented results were obtained for the steady state flow conditions and on the assumption that walls of the assist device, adapters and valves were rigid. The simulated fluid was blood. Dynamic viscosity of blood was defined according to the Non-Newtonian Power Law. Simulations were preformed for systole and diastole. The Ansys CFX v12 code was used to perform preprocessing, solving and postprocessing stages. Deformations of the threeleaflet polyurethane valve were obtained in SolidWorks 2009 and imported to Ansys ICEM v.12. On the basis of the preformed analysis, it has been proven that the disc mechanical heart valve generates better flow conditions inside the heart chamber, especially when a risk of coagulation is concerned. Moreover, the flow observed inside the chamber when the disc valve was used is more homogenous and a single swirl occurring in the central part enables good washing of the connection of the diaphragm and chamber regions. The analysis presented here is an integral part of the investigations conducted within the Polish Artificial Heart Programme. Keywords— ventricular assist device, artificial heart, artificial heart valves, numerical fluid mechanics.
I. INTRODUCTION Heart diseases are one of the most frequent cause of death and they are serious threats to human life. A proper diagnosis procedure may help to treat heart diseases. In many cases the only available procedure to rescue the patient’s life is a heart transplantation. An insufficient number of donors is a significant problem not only in the heart illness treatment, but since heart malfunction is always
dangerous to life, many efforts are made to create an artificial heart that may be implanted in place of the human heart. This will make the treatment independent of a limited number of donors. One of methods that may help to heal heart or to extend the time the recipient is waiting for a transplantation is to use an external artificial heart that supports the patient’s heart by forcing an extra blood flow rate. The presented research is devoted to examination of an effect of the artificial heart valve type on the flow inside the chamber and in the adaptor. The experiment was conducted on the original Ventricular Assist Device developed at the Foundation of Cardiac Surgery Development (FCSD). Two different types of artificial valves were introduced into VAD. A Medtronic-hall tilting disc mechanical valve was the first type and a three-leaflet polyurethane valve developed at the Foundation (FCSD) was the second one. An evaluation of different solutions was made on the basis of the flow visualization inside the VAD chamber and the spread of blood flow stagnation regions. The analysis was based on the numerical experiment performed with the Computational Fluid Dynamics software Ansys v12. The non-Newtonian model of blood, based on the Power Law, was applied. Two stages of the device operation was simulated. Diastole – the diaphragm is in its most external position and the volume of blood inside the chamber is the highest as well. In this particular time of the heart cycle, the inlet valve is totally open and the outlet valve is almost closed (5% open). The opposite deflection of the diaphragm – systole – occurs when most of blood is discharged from the chamber. The outlet valve is totally open whereas the inlet valve is almost closed (5% open). All the computations were performed at the Institute of Turbomachinery, Technical University of Lodz, Poland.
II. NUMERICAL EXPERIMENT A. Computational Domain A model of the pneumatically driven Ventricular Assist Device – POLVADEXT – has been used in the presented study. The whole model consists of a blood chamber, an air chamber separated by a diaphragm, two artificial heart valves, and two adapters (Fig. 1). The pneumatic part of VAD is irrelevant in this study, hence it is omitted.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 410–413, 2010. www.springerlink.com
Influence of an Artificial Valve Type on the Flow in the Ventricular Assist Device
The CFD software solves the Navier-Stokes equations, employing the Finite Volume Method, thus the computational domain has to be divided into small volumes. The geometry has been imported into Ansys ICEM v12, in which it has been discretized. Due to the complexity of the geometry, an unstructured mesh has been used. In the boundary layer, in the vicinity of walls, prismatic elements have been employed to solve accurately the flow in the regions of highest velocity gradients. In all the presented cases, a number of elements in the whole domain have exceeded 14 millions.
411
the Power Law in the range of high strain values underestimates values of dynamic viscosity. Thus, for strains larger than 327 s-1, the Newtonian model is proposed, for which: μ = 0.00345 Pas. Table 1 Boundary conditions for all cases under investigations Inlet maximum velocity [m/s] Diastole Systole
3.815 0.315
Outlet static pressure [kPa] 13.65 -0.2821
The dynamic viscosity used in the study is described by equation (2). ∂v ⎧ −9 ⎪ μ = 0,554712 for ∂y < 1e ⎪ n −1 ⎛ ∂v ⎞ ∂v ⎪ −9 ⎜ ⎟ = μ μ ⎨ 0⎜ ⎟ for 1e ≤ ∂y < 327 ∂ y ⎝ ⎠ ⎪ ∂v ⎪ ≥ 327 μ = 0,00345 for ⎪ ∂y ⎩
Fig. 1
(2)
where: μ – dynamic viscosity, Computational domain with described model elements
The Shear Stress Transport turbulence model has been used. The quality of the mesh has been checked in a mesh independence study and with the Yplus parameter that was smaller than 8 in all cases. Low residual levels have been achieved in the solution. Boundary conditions were estimated on the basis of the literature survey. At the inlet, a velocity profile has been introduced with the maximum value listed in Table (1). The profile has based on the cork shape, typical of turbulent flows. It is described by the following formula (1): r⎞ ⎛ V = Vmax ⎜1 − ⎟ ⎝ R⎠
1
7
∂v – strain. ∂y The blood density was assumed to be equal to 1045 kg/m3. B. Numerical Study Results The most dangerous complication connected to an application of the VAD is blood coagulation. It may occur when blood platelets are activated by deformation or contact with air or material of the implanted device.
(1)
where: V – velocity at the particular node of the mesh at a distance r measured from the axis of the adapter, Vmax – velocity along the axis of the adapter, R – radius of the adapter, r – distance from the axis of the adapter. At the outlet cross-section, pressure has been assigned. Velocity and pressure values used in the presented study are listed in Table (1). The non-Newtonian blood model based on the Power Law [1, 2] model has been used in the study. The blood dynamic viscosity has been described as a function of strain. The Basic Power Law model has been limited in the range of very low strain values below 1e-9 [s-1]. It is known that
Fig. 2 Streamlines with velocity vectors – a disc valve during diastole
IFMBE Proceedings Vol. 29
412
D. Obidowski et al.
A risk of coagulation increases when the exposure time is elongated. This leads to a conclusion that regions of very low velocity are potentially most dangerous, especially in the vicinity of VAD walls. Two types of visualization methods are used, namely: streamlines with velocity vectors (Figs 2, 4, 6) and surfaces enclosing low velocity regions (Figs 3, 5, 7, 8).
It has been shown in the earlier studies [3] that the angular positioning of the disc valve plays a significant role in the proper stream optimization. In this paper, the best angular position is presented. For the same angular position, velocity streamlines during systole are illustrated in Fig. 4. High velocity (greenish lines) in the connection of the diaphragm and the chamber is worth noticing. This ensures a low risk of coagulation due to short contact time of blood platelets with the material of the chamber. Stagnation regions are depicted in Fig. 5. Although those regions are large, it is necessary to remember that stagnation occurs in limited time of the heart cycle. The most dangerous is a situation when stagnation is observed for both diastole and systole in the same regions, especially in the vicinity of the diaphragm. This conclusion has been drawn from clinical observations.
Fig. 3 Stagnation regions – a disc valve during diastole Swirls that can be seen in Fig. 2 have an axis around which blood circulates. Blood flows in the region where the diaphragm is connected to the artificial heart chamber. In Fig. 3 no significant regions of stagnation are visualized. The volume of regions where velocity is lower than 0.01 m/s is marked in colours depicted in the legend. In Figure 3 only very small spots, in the vicinity of the inlet valve, are visible, which proves that the velocity in the whole domain is larger than the value mentioned.
Fig. 4 Streamlines with velocity vectors – a disc valve during systole
Fig. 5 Stagnation regions – a disc valve during systole
Fig. 6 Streamlines with velocity vectors – a three-leaflet valve during diastole
IFMBE Proceedings Vol. 29
Influence of an Artificial Valve Type on the Flow in the Ventricular Assist Device
The main aim of this study was to compare a disc heart valve and a polyurethane three-leaflet valve operating under the same conditions in the same chamber. Streamlines of diastole only are presented in this paper as the flow conditions are better, which has been shown in the case of the disc valve.
413
The most important is that there are regions of stagnation visible in systole and diastole that occur between leaflets of the valve and the adaptor wall. This occurs only for the inlet valve. In those regions, a risk of blood coagulation is considerable. Another problematic region observed in the case of the three-leaflet valve during systole is a connection of the diaphragm and the chamber. This is an additional disadvantage of an application of the polyurethane threeleaflet valve for the pneumatic Ventricular Assist Devices described in the paper.
III. CONCLUSIONS
Fig. 7
Stagnation regions – a three-leaflet valve during diastole
When compared to the tilting disc valve, the polyurethane valve shows highly worse flow conditions. The streamlines vector plots shown in Fig. 6 present several regions of swirls within the chamber that do not lie along one axis. Moreover, separation is observed near the wall just after the blood enters the chamber. Backflows that can be seen on the external sides of leaflets cause stagnation regions visualized in Fig. 7 and Fig. 8.
In the presented numerical analysis, it is shown that the three-leaflet valve exhibits a high risk of blood coagulation in the place where the leaflet is attached to the adapter wall. Some significant regions of stagnation were observed for diastole and systole. The tilting disc mechanical valve when inserted in the optimal angular position ensures good flow circulation during diastole and systole as well. In the case of diastole, no significant regions of stagnation have been observed. Angular positioning of the three-leaflet valve would not provide any improvement in flow conditions as the regions of stagnation are not related to the geometry of the chamber but to the design of the valve itself. A further study for time-dependant boundary conditions and, possibly, involving fluid structure interaction simulations [4] should be carried out to support the conclusions drawn here.
ACKNOWLEDGMENT This work has been supported by the “Polish Artificial Heart” governmental project.
REFERENCES
Fig. 8 Stagnation regions – a three-leaflet valve during systole
1. K. Jozwik, D. Obidowski (2010), Numerical simulations of the blood flow through vertebral arteries, Journal of Biomechanics, Vol. 43, Issue 2: 177-185. 2. Johnston, B., Johnson, P., Corney, S., Kilpatrick, D. (2004), NonNewtonian Blood Flow in Human Right Coronary Arteries: Steady State Simulation, Journal of Biomechanics 37: 709-720. 3. K. Jozwik, D. Obidowski, P. Klosinski et al (2009), Modifications of an Artificial Ventricle Assisting Heart Operation on the Basis of Numerical Methods, Turbomachinery, Vol. 135: 61-68. 4. De Hart J., Baaijens F.P.T., Peters G.W.M., Schreurs P.J.G., (2003), A computational fluid-structure interaction analysis of a fiber-reinforced stentless aortic valve, Journal of Biomechanics, Vol. 36: 699–712.
IFMBE Proceedings Vol. 29
A New Stimulation Technique for Electrophysiological Color Vision Testing M. Zaleski and K. Penkala West Pomeranian University of Technology/Faculty of Electrical Engineering/Department of Systems, Signals and Electronics Engineering, Szczecin, Poland
Abstract— In the paper a novel kind of stimuli for ERG (Electroretinography) and VEP (Visual Evoked Potentials) color vision tests are presented, along with a special-purpose generator. The base of these stimuli is color change with constant luminance (isoluminance). Sample results obtained in laboratory experiments as well as possible applications of this stimulation technique are also discussed. Keywords— Electroretinography (ERG), Visual Evoked Potentials (VEP), color vision testing, color stimulation, objective anomaloscopy.
Table 1
Main features and parameters of the generator
Feature No. of channels LED current Flicker frequency Light pulse duration Optical isolation
Value (range) 3 0 – 30 1 – 50 1 – 1000 >1000
Unit mA Hz ms V
The generator’s software works in two main modes: the luminance equalization mode and the test mode (Figure 1).
I. INTRODUCTION Objective color vision testing is a very important issue in ophthalmology. Electrophysiological methods and techniques like Electroretinography (ERG) and Visual Evoked Potentials (VEP) are commonly used in clinical practice. However, color stimulation tools implemented in commercially available equipment for these tests cannot be used in objective investigations of color vision mechanisms, particularly in diagnosing several kinds of abnormalities of this visual function (“objective anomaloscopy”), because electrical responses recorded from the visual system are dependent on the luminance component of light stimuli. The method presented in the paper is free of this disadvantage – this technique is based on color alterations of strictly defined spectral characteristics, with equal, constant luminance (isoluminant color stimulation).
II. MATERIALS AND METHODS A. The Generator A microcontroller-based generator is used to drive a tricolor LED. The hardware is based on Atmel ATMega8 microcontroller and a three-channel adjustable current source. The LED drive current can be independently adjusted in a wide range. A typical BNC connector allows external triggering from an electrophysiology system, e.g. UTAS E-2000, LKC, USA which was used in the preliminary tests. The input is optically coupled to match safety requirements.
Fig. 1 Simplified algorithm for green-red color alterations
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 414–417, 2010. www.springerlink.com
A New Stimulation Technique for Electrophysiological Color Vision Testing
415
1. Luminance equalization mode
C. The Colors
In this mode the patient is exposed to a two-color flicker, with the 50% duty cycle. The luminance of one of the colors (LED) is set by the examiner. The luminance of the other color and the flicker frequency is controlled by the patient. As described in [1], the maximum color flicker recognition frequency is slightly smaller than the critical fusion frequency for luminance flicker. This allows the patient to equalize the luminance of the two given colors. In [2], where the idea of the isoluminant color test was presented, it was shown that this subjective method was reliable and offered repeatable results.
Three basic colors are created by a light emitting diode (LED). It is important to note, that the LED red, green and blue wavelengths partially correspond with sensitivity curves of cones (Table 2 and Figure 3). Table 2
Light emission maxima of the LED
Color
Maximum intensity wavelength
Blue Green Red
470 nm 530 nm 630 nm
2. The test mode In this mode, one of the previously chosen colors is the basic color, and during the test it changes to the other color when a trigger pulse comes from UTAS E-2000 (or any other ERG/VEP unit). The light pulse duration is controlled in a wide range from 1 ms to 1000 ms in 1 ms steps, and duty factor can be set according to requirements of the test. If one of the colors is switched off, the generator can be simply used as a source of color flash stimuli of equal luminance. B. The Optical Module The optical part consists of a high luminance tri-color SMD LED and Maxwellian View optics [3]. Fig. 3 Relative spectral sensitivity curves of three types of cones population in human retina (S – short wavelength, M – medium wavelength and L – long wavelength) D. Testing Procedure The generator is capable of creating a wide range of color stimuli:
Fig. 2 Maxwellian View optics – a simplified diagram Maxwellian View optics provides high luminance by focusing the light coming from source directly on the patient’s pupil (Figure 2). It consists of a black cylinder made of nonreflecting material, a lens and diaphragm system. Thanks to such arrangement - together with the electronic generator direct light stimulation of 100 of the central retina is obtained, with a wide range of flexible brightness adjustment for three different colors.
-
blue flash, green flash, red flash, isoluminant color alterations: all possible combinations (red-to-green, green-to-red, green-to-blue, blue-to-green, red-to-blue, blue-to-red).
In the preliminary experiments, simultaneous monocular ERG and VEP responses were recorded using the UTAS E2000 system (LKC Inc., USA), in a patient with normal color vision. Electrodes for ERG recordings were placed at the cornea (active, DTL), ipsilateral outer canthus (reference, gold cup) and forehead Fz (ground, gold cup), according to the International Society for Clinical Electrophysiology of Vision (ISCEV) ERG standard [4].
IFMBE Proceedings Vol. 29
416
M. Zaleski and K. Penkala
Electrodes for VEP recordings (Figure 4) were placed in standard locations for ISCEV Flash VEP [5] – occipital scalp Oz (active), forehead Fz (reference) and earlobe A1 or A2 (ground) - all gold cup electrodes.
Fig. 5 Sample VEP responses to color flash stimuli of equal luminance
Fig. 4 Electrodes placement (adapted from [5]) The recording time was 1 s with 1 ms time resolution. Amplifiers were set to 10 μV/div sensitivity and 0.3-500 Hz bandwidth. In each test 80 responses were averaged with proper artifact rejection, what required ca. one hundred test cycles. In single color flash responses maximum available luminance was used. For color alteration responses, luminance of respective colors was equalized as described above. This allowed recording of responses to color (spectral content) change with no influence of luminance. Fig. 6 Sample VEP recording in the isoluminant color alteration mode (green-to-red, two cycles are shown)
III. RESULTS Sample results for three color flash VEP and one color alteration (red – green) responses are shown on Figures 5 and 6, respectively. It can be seen that cortical responses to color flashes (Figure 5) are different in morphology (polarity, amplitude and time parameters). The first wave in the red response is positive, and in green as well as blue signal is negative. The blue response voltage is much higher than in red and green recordings, and blue as well as green first waves appear faster than the red one. We obtained also clear responses to isoluminant color changes (Figure 6). They seem to have similar properties to single color flash responses: green stimulus evokes negative signal polarity, while red makes it positive. The ERG recordings (Figure 7) are smaller in amplitude and, as a result of noise, more difficult in interpretation. However, at the retinal stage of signal and information processing, the morphological differences between responses to color flashes of equal luminance seem to be much less pronounced.
Fig. 7 Sample ERG recordings obtained in color flash stimulation with equal luminance
IFMBE Proceedings Vol. 29
A New Stimulation Technique for Electrophysiological Color Vision Testing
IV. DISCUSSION Although the first results are new and rather incomparable to anything recorded before, they seem to be very promising in further research on color vision processes [6,7]. What is important, the electrical responses of the visual system may be recorded dependent only of the spectral characteristics of the light stimulus, because the luminance contribution in this technique is eliminated. Recently an attempt has been made to extract isolated responses from L, M and S cones. The extraction, performed in the MatLab environment, is based on comparing cones sensitivity curves with the LED emission curves.
417
3. Beer RD, Macleod IAD, Miller PT (2005) The extended Maxwellian View (BIGMAX): A high-intensity, high-saturation color display for clinical diagnosis and vision research. Behavior Research Methods 37 (3):513-521 4. Marmor MF, Fulton AB et al (2009) Standard for clinical Electroretinography. Doc Ophthalmol 118:69-77 5. Odom JV, Bach M et al (2004) Visual Evoked Potentials standard. Doc Ophthalmol 108:115-123 6. Zaleski M, Brykalski A, Lubinski W, Penkala K (2009) A new approach to ERG/VEP stimulation in colour vision testing. Proc. ISTET’09 XV Int. Symp. on Theor. Electr. Eng., Lubeck, Germany, 2009, at http://www.istet09.de 7. Penkala K, Zaleski M, Lubinski W (2009) Isoluminant colour stimuli for electrophysiological tests. 47th ISCEV Symp., Padova - Abo Terme, Italy, p 61, at http://www.iscev2009.org/iscev2009 Corresponding author:
V. CONCLUSIONS The new kind of biosignals obtained in isoluminant color-alteration mode may be useful in investigations of color vision mechanisms (particularly in solving the “color coding” phenomenon), in modeling those processes as well as in clinical diagnosis of color vision disorders “electrophysiological, objective anomaloscopy”.
Author: Krzysztof Penkala Institute: Department of Systems, Signals and Electronics Engineering, Faculty of Electrical Engineering, West Pomeranian University of Technology Street: Sikorskiego 37 City: 70-313 Szczecin Country: Poland Email:
[email protected] REFERENCES 1. Le Grand Y (1960) Les yeux et la vision. Dunod, Paris 2. Penkala K (1989) A model of the photostimulator with light emitting diodes for electrophysiological evaluation of colour vision (in Polish) Probl Techn Med XX (3):141-149
IFMBE Proceedings Vol. 29
Novel TiN-based dry EEG electrodes: Influence of electrode shape and number on contact impedance and signal quality P. Fiedler1, S. Brodkorb1, C. Fonseca2, 3, F. Vaz4, F. Zanow5 and J. Haueisen1, 6 1
Institute for Biomedical Engineering and Informatics, Ilmenau University of Technology, Ilmenau, Germany INEB – Instituto de Engenharia Biomédica, Divisão de Biomateriais, Universidade do Porto, Porto, Portugal 2 Faculdade de Engenharia, Departamento de Engenharia Metalúrgica e de Materiais, Universidade do Porto, Porto, Portugal 4 Departamento de Física, Universidade do Minho, Guimarães, Portugal 5 eemagine Medical Imaging Solutions GmbH, Berlin, Germany 6 Biomagnetic Center / Department of Neurology, University Hospital Jena, Jena, Germany 2
Abstract—Usability of conventional wet electrodes for electroencephalography (EEG) is depending on a set of requirements, including time consuming and complex preparation of the skin of a subject, thus limiting possible application. A new class of “dry” electrodes without the need for electrolyte gels or pastes is being investigated. The dry application scenario of these novel electrodes requires a stable and reliable contact with the subject’s skin. In order to develop an electrode shape with large contact surface for low electrode-skin impedance while also ensuring a sufficient hair layer penetration, several studies were performed. In this paper a distinct titanium electrode substrate shape for titanium nitride (TiN) coated electrodes was analyzed regarding influences of the number of interconnected electrodes and contact surface on electrodeskin impedance and biosignal quality. As a result 10 interconnected TiN pins had the lowest impedance values of 14 to 55 kΩ (depending on signal frequency) in comparison to 2 to 44 kΩ using conventional Ag/AgCl electrodes. Also the mean average deviation (MAD) of 5 seconds long EEG episodes were computed. The lowest MADs of 2.00 to 2.25 µV were determined using three interconnected TiN pins. In comparison to MADs of 2.13 to 2.54 µV, using a second set of Ag/AgCl electrodes, this leads to the conclusion that most of the error was related to spatial distance. This first step in optimization of electrode shape for dry TiN based electrodes showed very promising results and enable their use for EEG acquisition. Keywords— titanium nitride, biomedical electrodes, bioelectric signals, electrochemical characterization, electroencephalography I. INTRODUCTION
In clinical routine and medical research electroencephalography (EEG) is a common technique for investigating the human central nervous system. In the majority of cases 32 to 256 electrodes are placed on the head of the subject in order to measure the electrical activity of the brain by detecting and recording electrical potential fluctuations at different areas of the human scalp [1]. Direct signal acquisition at the skin is still a difficult, time-consuming process and therefore susceptible to many
error sources. Conventional silver/silver-chloride electrodes (Ag/AgCl) are the most commonly used type of electrodes and can be considered to be the “gold standard”. They have to be used in combination with different types of pastes and / or gels containing necessary electrolytes. Preliminary to the application of electrodes the intended areas of the scalp must be cleaned and prepared in order to lower the skin-electrode impedance. Additional technological limitations, e.g. limited long-time stability of the pastes and gels as well as a risk of conductive bridges between adjacent electrodes can lead to changing electrochemical characteristics and therefore can directly lead to faulty measurements. Due to such drawbacks of the current technology, alternative types of electrodes and materials are being investigated [1, 2, 3]. This new class of electrodes is called “dry electrode” since there is no need for application of electrolyte gels or pastes. Additionally the preparation time and complexity is strongly decreased by means of elimination of the need for cleaning and other types of skin preparation. In the current study the authors focused on the conductive titanium nitride (TiN), deposited as thin film on titanium substrate. TiN is well-known for its chemical and mechanical stability as well as outstanding biocompatibility [4, 5]. The need for low contact impedance and therefore large contact surface conflicts with hair layer penetration. As a possible solution several single electrode pins, separately penetrating the hair layer, can be interconnected, thus combining the single contact surfaces. Electrode shape, pin number and hair layer penetration as well a stable and reliable electrode-skin contact still needs further investigation. The aim of the present study was to analyze the influence of electrode pin number on electrode-skin impedance and signal quality. Therefore impedance measurements were performed and EEG biosignal episodes were recorded using electrodes with different numbers of single pins on a biological subject. The results were compared to simultaneously acquired data of conventional Ag/AgCl electrodes.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 418–421, 2010. www.springerlink.com
Novel TiN-Based Dry EEG Electrodes: Influence of Electrode Shape and Number on Contact Impedance and Signal Quality
419
II. MATERIALS AND METHODS
A. TiN-coated biosignal electrodes Identical TiN films were deposited on titanium pin substrates cut and turned from a rod of 99.96 % pure titanium (GoodFellow Metals, London, UK) and coated by reactive DC magnetron sputtering in a custom made laboratory deposition system. The pin shape is shown in figure 1b. Preparation of the blank pins included abrasion, polishing, cleaning and drying. For creation of a homogenous coating a custom, turning substrate holder was used and positioned in front of a Ti target. During the sputtering a gas atmosphere composed of argon and nitrogen was applied. The distinct coating parameters were selected according to the results of a full series of previous tests for different kinds of TiN films [5]. Shielded copper signal cables were glued to the backside of the TiN electrode pins using conductive silver glue (Elecolit 325) in order to connect the electrodes.
(a) Electrode pin array (grey) and wiring schematic (top view)
Fig. 1:
(b) TiN electrode pin schematic (side view)
Fig. 2: Electrode placement on the head of the subject: star-shaped TiN array (black) and Ag/AgCl (grey) electrodes The Ag/AgCl electrodes were constantly used in combination with EEG paste (D.O. Weaver and Co. Ten20 conductive) applied to preliminary cleaned positions on the head. No additional liquid, gel or paste was added at the position of the TiN array thus creating a realistic dry application scenario for these electrodes. C. Impedance measurement A Hewlett Packard 4192A LF impedance analyzer was used for impedance measurement. The impedance between one frontal Ag/AgCl electrode and the TiN array at parietal position was measured. In order to evaluate the impedance drop due to increasing electrode number and therefore contact surface, the number of interconnected electrode pins at the TiN array was varied between one and ten electrode pins. As a reference an additional measurement between the frontal and a parietal Ag/AgCl electrode was carried out. All impedance measurements were executed sequentially using parameters according to table 1. Table 1: Impedance measurement parameters
TiN electrode array schematic and pin schematic
B. Electrode setup The ten TiN pins were arranged in a fixed star-like shape on an acrylic base plate according to arrangement and wiring scheme in figure 1a. For direct comparison of the results using different electrode setups and conventional Ag/AgCl electrodes, regarding differences in impedance characteristics and EEG signal quality, both types of electrodes were placed nearby on the head of a subject. In order to minimize artifacts due to relative movement between electrode and skin, a silicone headband was used for fixation. Figure 2 shows the electrode positions of the conventional Ag/AgCl reference electrodes as well as the TiN array at positions Fp2 and POz according to the international 10-20 system [6].
Frequency range 1 Frequency range 2
Start frequency
Stop frequency
Frequency steps
5 Hz
200 Hz
10 Hz
200 Hz
5 kHz
100 Hz
The total measurement time per sample was two seconds. Measurement control and recording as well as subsequent calculations were done using custom implemented MATLAB tools. D. Biosignal acquisition Two inputs of a commercial 12 channel DC amplifier (TheraPRAX from neuroConn GmbH, Germany) were used for amplification and recording of the biosignals. The frontal electrode was used as patient ground in a bipolar mea-
IFMBE Proceedings Vol. 29
420
P. Fiedler et al.
surement setup using patient ground reference. Therefore temporally parallel EEG acquisition with two different types of electrodes was possible while also minimizing the spatial distance between adjacent electrodes. The actual distance was approx. 4 cm (center to center). Due to the large and plane TiN array fixation on an acrylic plate a sufficient contact between all TiN electrode pins and the scalp was only possible in distinct regions of the subject’s head including parietal position. A reference measurement for the evaluation of signal differences caused by spatial electrode distance was performed by replacing the TiN array by a third Ag/AgCl electrode. During the measurements different signal episodes were recorded containing eye movement and eye blinking as well as alpha activity provoked by closed eyes. All signals were recorded using the neuroConn Software. Further investigation was performed using MATLAB.
B. Biosignal acquisition After filtering all EEG signals using a band pass with cut-off frequencies at 2 Hz and 30 Hz the results of TiN electrode array and Ag/AgCl electrode can be compared as shown in figure 4.
III. RESULTS
A. Impedance measurement The averaged results of the impedance measurements are plotted in figure 3. It is clearly visible that the impedance is decreasing with increased number of electrode pins. Using a single TiN pin leads to an impedance maximum of 220 kΩ for 5 Hz and a minimum of 60 kΩ for 5 kHz. The conventional Ag/AgCl electrodes show the lowest impedances, compared to all electrode setups, with a maximum of 44 kΩ for 5 Hz and a minimum of 2 kΩ for 5 kHz. The 10 pin TiN setup has slightly higher impedances with 55 kΩ and 14 kΩ respectively, being the setup having lowest impedances of all TiN arrays.
Fig. 4: Overlay plot of potentials produced by eye blinking and recorded using an array of three dry TiN electrode pins in comparison to Ag/AgCl electrodes in combination with EEG paste. Corresponding episodes X (TiN array / second Ag/AgCl) and Y (reference Ag/AgCl) of 5 seconds (N samples) were selected from both signals and the mean absolute deviation (MAD) according to equation 1 as well as the standard deviation of the MAD (SD-MAD) were calculated.
MAD =
1 ⋅∑ x − y N N
i=1
i
i
(1)
Due to too high impedance it was impossible to record an EEG with one TiN pin only. The results of the remaining electrode arrangements are shown in table 2. Table 2: Mean absolute deviation (MAD) and standard deviation of MAD (SD-MAD) between recorded signals of reference and TiN array setup Electrode setup Ag/AgCl TiN - 2 Pins TiN - 3 Pins TiN - 5 Pins TiN - 10 Pins
Fig. 3: Averaged impedance measurement results for different numbers of electrode pins
Eye blinking MAD SD-MAD [µV] [µV] 2.54 2.31 4.01 3.34 2.25 2.03 3.53 3.51 2.64 2.27
Alpha activity MAD SD-MAD [µV] [µV] 2.13 1.62 2.55 2.04 2.00 2.03 2.12 1.44 2.40 1.81
With MADs of 4.01 µV (eye blinking) and 2.55 µV (alpha activity) the combination of two TiN pins shows the highest MADs of all electrode setups while the combination of three TiN pins shows the lowest MADs of 2.25 µV (eye
IFMBE Proceedings Vol. 29
Novel TiN-Based Dry EEG Electrodes: Influence of Electrode Shape and Number on Contact Impedance and Signal Quality
blinking) and 2.00 µV (alpha activity). The MAD of the 10pin array is lower than the MAD of the five-pin arrangement for eye blinking, while in the case of alpha activity the MAD relation is reversed. The MADs of the EEG signals acquired using a second set of Ag/AgCl electrodes are in fact higher than those of the arrangement of three TiN pins. For all electrode setups the MAD of eye blinking recordings is higher than with alpha activity. This is caused by higher signal amplitudes. The fact that higher signal amplitudes lead to higher differences between both signals is also visible in the overlay plot in figure 4.
tion and optimization in order to develop an optimal dry electrode system based on a low cost, time effective and environmental friendly manufacturing process. An electrode design incorporating several electrode pins on a single base plate and increased pin density will be developed, thus further decreasing electrode-skin impedance and increasing signal quality. Future developments will also include the assembly of a custom multichannel EEG cap system. This will enable the use of the novel dry electrodes in clinical routine as well as in research scenarios including source localization and functional coupling [7, 8].
ACKNOWLEDGMENT
IV. DISCUSSION
In relation to the used measurement setup and method, the resulting impedance values are cumulated impedances of several components which can be simplified summarized according to equation 2. Herein ZFp2 is the impedance of the electrode at frontal position Fp2 while ZPOz is the impedance of the electrode at parietal position POz. Z = Z Fp 2 + Z EEG paste + Z Scalp + Z POz
421
This work was supported by the company ANT B.V., Enschede, The Netherlands. Additional financial support was granted by the Landesentwicklungsgesellschaft Thüringen mbH and the European Regional Development Fund (TNA XII-1/2009), the German Academic Exchange Service (D/07/13619) as well as the German Federal Ministry of Education and Research (03IP605).
(2)
Possible further impedance reduction could be achieved using more than 10 pins in an optimized arrangement for higher electrode pin density. Due to the dependency between signal quality and spacial distance, which was proven by comparison of signals from two adjacent Ag/AgCl electrode setups, signal quality can be improved by increased density of the electrode pin arrangement.
REFERENCES 1. 2. 3. 4. 5.
V. CONCLUSION
In the present study different TiN electrode setups varying in the number of electrode pins were analyzed regarding their influence on electrode skin impedance and EEG signal quality in order to evaluate the applicability for EEG biosignal acquisition as well as finding an optimal number of pins per electrode. Therefore the novel TiN electrodes were investigated under a realistic dry TiN films were selected because of their known excellent chemical, mechanical and biocompatibility properties. The used electrode shape is capable of sufficiently penetrating the hair layer of a biological subject. The TiN coatings revealed excellent electrochemical characteristics and enable EEG biosignal acquisition with adequate signal quality. Due to the low difference between the acquired signals, compared to Ag/AgCl electrodes, it is possible to summarize that the novel TiN coated electrodes are appropriate for EEG acquisition. This fact is suggesting further investiga-
6. 7.
8.
Taheri B A, Knight R T, Smith R L (1994) A dry electrode for EEG recording. Electroen Clin Neuro 90:376–383. Searle A, Kirkup L (2000) A direct comparison of wet, dry and insulating bioelectric recording electrodes. Physiol Meas 21:271-283 Ng W C, Seet H L, Lee K S et al. (2009) Micro-spike EEG electrode and vacuum-casting technology for mass production. J Mater Process Tech 209:4434-4438 Vaz F, Ferreira J, Ribeiro E et al. (2005) Influence of nitrogen content on the structural, mechanical and electrical properties of TiN thin films. Surf Coat Tech 191:317-323 Cunha L T, Pedrosa P, Tavares C J et al. (2009) The role of composition, morphology and crystalline structure in the electrochemical behavior of TiNx thin films for dry electrode sensor materials. Electrochim Acta 55:59-67 Jasper H H (1958) The ten-twenty electrode system of the International Federation. Electroen Clin Neuro 10:371–375 Haueisen J, Leistritz L, Süsse T et al. (2007) Identifying mutual information transfer in the brain with diffenential-algebraic modeling: Evidence for fast oscillatory coupling between cortical somatosensory areas 3b and 1. NeuroImage 37:130-136 Graichen U, Witte H, Haueisen J (2009) Analysis of induced components in electroencephalograms using a multiple correlation method. BioMed Eng OnLine 8:21 Corresponding author: Author: Patrique Fiedler Institute: Institute for Biomedical Engineering and Informatics, Ilmenau University of Technology Street: Gustav-Kirchhoff Str. 2 City: Ilmenau Country: Germany Email:
[email protected] IFMBE Proceedings Vol. 29
A finite element method study of the current density distribution in a capacitive intrabody communication system Ž. Luþev, A. Koriþan and M. Cifrek Faculty of Electrical Engineering and Computing, University of Zagreb, Zagreb, Croatia Abstract— In this paper we present the finite element method (FEM) study of a capacitive intrabody communication (IBC) system. We analyze current density distribution at the frequencies of 100 kHz, 1 MHz, and 10 MHz. We investigate the ratio between the capacitive and resistive current density component inside the human body and the influence of skin humidity, as well as the electrode size on the total current density distribution. We showed that the highest total current density is achieved inside the muscle tissue, and that the total current density increases with frequency, skin humidity and the size of the excitation electrodes. The surface potential shows the same trend and is in order of microvolts. At the frequency of 100 kHz the safety limits on the total current density are exceeded for wet skin and for larger electrodes. At the higher frequencies (1 MHz and 10 MHz) maximum allowed current density is not exceeded. Keywords—finite element method (FEM), intrabody communication (IBC), human arm model, current density I. INTRODUCTION
In the intrabody communication (IBC) the human body is used as a signal transmission medium. The signal is sent through the body using the transmitter excitation electrodes, and is measured by the receiver. The received signal strength is affected by the orientation of the transmitter with respect to the receiver and the number of ground electrodes connected to the body [1–6]. Moreover, the signal transmission path highly depends on the surrounding environment [1–3]. In vivo measurements [1, 2] showed that the highest received signal strength is achieved when both transmitter electrodes and a receiver signal electrode are connected to the human body, and the receiver ground electrode remains disconnected. The dielectric properties of the human body, electrical conductivity and relative permittivity, determine the flow of electric current and the magnitude of polarization effects, respectively. The most relevant source of dielectric properties of the human tissues for a frequency range from 10 Hz to 10 GHz and different tissues is given by Gabriel et al. [7]. It is shown that the dielectric properties of tissues depend on the type of tissue, frequency, temperature, and the amount of water in a particular tissue. At lower frequencies the permittivities of biological tissues are mostly high, so in
order to develop a realistic human arm model, tissue capacitive properties should be taken into account, together with the tissue resistive properties. In order to understand the signal propagation through the human body and to improve the IBC hardware, it is essential to investigate and simulate current pathway through the human tissue and to assess the influence of the human anatomy on signal propagation. In [5, 6] the authors developed a human arm model to investigate the electrode structure and the effects of ground electrode on capacitive intrabody signal transmission. The proposed model is formed as a parallelepiped with a 5 cm x 5 cm base and the results are calculated at 5 MHz frequency using FDTD-based EM simulator, under the assumption that the human body is a lossy dielectric material. In [3] the author modeled a galvanic IBC system and investigated the influence of the distance between the coupler and detector, the influence of joints, the sensitivity to resistivity changes of the tissue layers, and different coupling by wet, dry and a combined electrode interface. It was shown that the majority of the current flows between the coupler electrodes in the fat and muscles without penetrating into the bone structure. In this paper we extend the study of the capacitive IBC system to include conditions of wet and dry skin and lower frequency range (. 2007 4. Jacobs P. M. Vitreous loss during cataract surgery: prevention and optimal management. Eye. Online: . Feb. 2008 5. Vitreous–retina–macula consultants of New York. Vitrectomy. Online: < http://www.vrmny.com/pe/ vitrectomy.html >. New York, 2006
Author: Petr Cech Institute: Department of Biomedical Engineering, Brno University of Technology Street: Kolejni 4 City: Brno Country: Czech Republic Email:
[email protected] IFMBE Proceedings Vol. 29
Short range wireless link for data acquisition in medical equipment N.M.Roman1, S.Gergely2, R.V. Ciupa1, M.V.Pusca1 1
2
Technical University of Cluj-Napoca/ Biomedical Engineering Department, Cluj-Napoca, Romania National Institute for Research and Development of Isotopic and Molecular Technologies, Cluj-Napoca, Romania
Abstract— The majority of patient monitoring equipments do have numerous cable connections to the main data processing unit. Therefore, especially the emergency rooms are usually packed with a lot of uncomfortable equipments which could create a difficult access to the patient. One solution to this issue is the using of low-micro power radio remote data transmission. Keywords— wireless, link, MSP430F5419, MSP430F5500, CC1000
I. INTRODUCTION The signal types which are suitable for a radio link are the respiratory, the ECG, EMG and temperature signals. This paper describes an application for a 12 channel analog or digital signals transmission using a short range radio link. The wireless link has two modules respectively the acquisition module and the host module. There is one transceiver on each module having the same program structure. The structure of the designed wireless link is shown in figure 1. ADC converter
Signal encoding
• Controlled Signal multiplexing • MSP430F5419
Signal decoding
• Radio channel link • Transceiver • CC1000
• USB port controlled • Transceiver • MSP430F5500
Fig. 1 Link structure II. DESCRIPTION OF THE SYSTEM The PC controls the host interface through the full speed USB port. A good solution for the host microcontroller is the Texas Instruments MSP430F5500. The Texas Instruments MSP430 family of ultralow-power microcontrollers consists of several devices featuring different sets of peripherals well targeted for this application. The architecture, combined with five low-power modes is optimized to achieve extended battery life in portable measurement applications. The de-
vice features a powerful 16-bit RISC CPU, 16-bit registers, having high code efficiency. The USB module is a fully integrated USB interface that is compliant with the USB 2.0 specification. The module supports full-speed operation of control, interrupt, and bulk transfers. The module includes an integrated LDO, PHY, and PLL. The PLL is highly-flexible and can support a wide range of input clock frequencies. USB RAM, when not used for USB communication, can be used by the system. The used USB protocol does not require the installing of a device specific driver. It uses the Human Interface Device (HID) protocol which programs the USB registers through the embedded firmware USB protocol. The below listed registers must be programmed [1] to achieve the desired data packets dimension; correspondingly by programming the endpoint registers: USB Configuration Registers (Base Address: 0900h) USB key/ID USB module configuration
USBKEYID USBCNF
00H 02H
USB Control Registers (Base Address: 0920h)
Offset
Input endpoint#0 configuration Input endpoint #0 byte count Output endpoint#0 configuration Output endpoint #0 byte count Input endpoint interrupt enables Output endpoint interrupt enables Input endpoint interrupt flags Output endpoint interrupt flags USB interrupt vector USB frame number
00H 01H 02H 03H 0EH 0FH 10H 11H 12H 1AH
IEPCNF_0 IEPCNT_0 OEPCNF_0 OEPCNT_0 IEPIE OEPIE IEPIFG OEPIFG USBIV USBFN
At start up the host is programmed for the desired number of channels, transmission rate and data resolution as a number of bit/sample. Also the microcontroller’s ports are programmed for the data exchange with the host-transceiver, using a very simple routine. During transmission, data is embedded in a structure that allows the correct per channel
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 471–474, 2010. www.springerlink.com
472
N.M. Roman et al.
distribution in the PC-running application. In our case the structure is presented in figure 2.
Fig. 2 Transmitted data structure After the host establishes the connection with the PC, the next step is to program the host-transceiver. The first step at the opening of the communication channel is to program the desired number of channels at the acquisition module. Data structure is the same like in picture 2 with the observation that after the DC balanced preamble, the packet start-ID cod which contains a command-code is repeated until the interface is completely programmed. The end-data section contains the acknowledged code transmitted by the interface to the host module by reversing the transmission direction.
In receive mode CC1000 is configured as a traditional superheterodyne receiver. The RF input signal is amplified by the low noise amplifier (LNA) and converted down to the intermediate frequency (IF) by the mixer (MIXER). In the intermediate frequency stage (IF STAGE) this downconverted signal is amplified and filtered before being fed to the demodulator (DEMOD). As an option a RSSI signal, or the IF signal after the mixer is available at the RSSI/IF pin. After demodulation CC1000 outputs the digital demodulated data on the pin DIO. Synchronization is done on-chip providing data clock at DCLK. In transmit mode the voltage controlled oscillator (VCO) output signal is fed directly to the power amplifier (PA). The RF output is frequency shift keyed (FSK) by the digital bit stream fed to the pin DIO. The internal T/R switch circuitry makes the antenna interface and matching very easy. The frequency synthesizer generates the local oscillator signal which is fed to the MIXER in receive mode and to the PA in transmit mode. The frequency synthesizer is programmed to achieve the desired transmission respectively reception frequency. The device is programmed through the 3 wire digital serial interface (CONTROL) at a very fast rate up to 10 MHz. Throughout the data transmission the registers of both transceivers are programmed accordingly to the desired transmission way. Because of the bandwidth limitation we have preferred the synchronous NRZ transmission mode. Data transmission mode is shown in figure 4.
Transceiver module Our application uses the Texas Instruments CC1000 transceiver. CC1000 is a true single-chip UHF transceiver designed for very low power and very low voltage wireless applications. [2] The circuit is mainly intended for the ISM (Industrial, Scientific and Medical) and SRD (Short Range Device) frequency bands at 315, 433, 868 and 915 MHz. We selected the value for transmission frequency to 433MHz. In figure 3 is shown the simplified block diagram of the CC1000:
Fig. 4 Synchronous NRZ mode
Fig. 3 Simplified block diagram of the CC1000
There are 28 8-bit configuration registers, each addressed by a 7-bit address. A Read/Write bit initiates a read or writes operation. A full configuration of CC1000 requires sending 22 data frames of 16 bits each (7 address bits, R/W bit and 8 data bits). At reception data is detected through a digital demodulator. The IF signal is sampled and its instantaneous frequency is detected. The result is decimated and filtered. In the data
IFMBE Proceedings Vol. 29
Short Range Wireless Link for Data Acquisition in Medical Equipment
slicer the data filter output is compared to the average filter output to generate the data output. The averaging filter is used to find the average value of the incoming data. While the averaging filter is running and acquiring samples, it is important that the number of high and low bits received is equal. Therefore all modes, also synchronous NRZ mode, need a DC balanced preamble for the internal data slicer to acquire correct comparison level from the averaging filter. The used preamble is a ‘010101…’ bit pattern. This is necessary for the bit synchronizer to synchronize correctly. The averaging filter must be locked before any NRZ data can be received. If the averaging filter is locked (MODEM 1.LOCK_AVG_MODE=’1’), the acquired value will be kept also after Power Down or Transmit mode. After a modem reset (MODEM1.MODEM_RESET_N), or a main reset (using any of the standard reset sources), the averaging filter is reset. In a polled receiver system the automatic locking can be used. The programming of CC1000 requires an initializing routine which consist of all registers programming, starting with address 00H to 46H, followed by PLL and VCO calibration routine. Finally the TX/RX-routine allows data transmission. Due to bandwidth requirement the programmed baud rate is set to the maximum value which is 76.8 KBaud in NRZ synchronous mode. The data acquisition module The acquisition board consists of the MSP430F5419 microcontroller. The main reason why we used this microcontroller is because of the 12 bit on board ADC. Also the multi I/O architecture allows the direct to chip analog signal inputting of all 12 cannels without external multiplexing. The 12 external and 4 internal analog signals are selected as the channel for conversion by the analog input multiplexer. The input multiplexer is a break-before-make type to reduce input-to-input noise injection resulting from channel switching. The assigned transceiver is programmed using the serial data output from the CC1000 transceiver. In figure 5 it shows the simplified block diagram of the interface module:
473
sion end. The cyclic redundancy check (CRC) module provides a signature for a given data sequence. The CRC module produces a signature for a given sequence of data values. The signature is generated through a feedback path from data bits 0, 4, 11, and 15. The CRC signature is based on the polynomial given in the CRC-CCITT-BR polynomial shown in eq1: The CRC generator is first initialized by writing a 16-bit word (seed) to the CRC Initialization and Result (CRCINIRES) register. Any data that should be included into the CRC calculation must be written to the CRC Data Input (CRCDI or CRCDIRB) register in the same order that the original CRC signature was calculated. The actual signature can be read from the CRCINIRES register to compare the computed checksum with the expected checksum. To allow parallel processing of the CRC, the linear feedback shift register (LFSR) functionality is implemented with an XOR tree. The power management module (PMM) includes an integrated voltage regulator that supplies the core voltage to the device and contains programmable output levels to provide for power optimization. The PMM also includes supply voltage supervisor (SVS) and supply voltage monitoring (SVM) circuitry, as well as brownout protection. The brownout circuit is implemented to provide the proper internal reset signal to the device during power-on and poweroff. The SVS/SVM circuitry detects if the supply voltage drops below a user-selectable level and supports both supply voltage supervision (the device is automatically reset) and supply voltage monitoring (SVM, the device is not automatically reset). SVS and SVM circuitry is available on the primary supply and core supply. In standby mode (LPM3 RTC Mode) the overall measured current consumption of the microcontroller was 2.80 μA. This is a very good value for a portable interface unit. After the preconditioning of the signals, the implemented software filtering routine is capable of a real time noise reduction for a maximum of 3 allocated channels. Noise filtering is done by n=25 point moving average routine. A tremendous advantage of the moving average filter is that it can be implemented with an algorithm that is very fast. The moving average filter operates by averaging a number of points from the input signal to produce each point in the output signal. In equation form, this is written in eq2:
Fig. 5 Acquisition interface module block diagram The End-Data section from the data transmission structure contains a computed CRC value at the end of each transmis-
where ci are the known smoothing coefficients for a symmetric averaging. The moving average filter is in fact a convolution using a very simple filter kernel. The only limitation
IFMBE Proceedings Vol. 29
474
N.M. Roman et al.
of such a filter is that it is not suitable for the frequency domain because the moving average filter cannot separate one band of frequencies from another. An improvement in computational efficiency can be achieved if we perform the calculation of the mean in a recursive fashion. A recursive solution [3] is one which depends on a previously calculated value. Suppose that at any instant k, the average of the latest n samples of a data sequence xi is given by eq3:
Similarly at the previous time instant, k-1, the average of the latest n samples is:
even at the lowest sampling rate. These sampling rate values are correlated with the number of used channels. Regarding the averaging filtering method, if the spectrum of the noise and signal components do not overlap in the frequency domain, one can simply design a filter that keeps or enhances the desired signal term, x(t), and discards the unwanted noise term, n(t).While this is a simple and useful way of cleaning up a signal, this approach does not work in many instances because the biological signal and possible noise spectrums overlap. This is especially true in the case of an acquisition tool designed for in deep signal analysis or in other words for medical research. As a future development of this wireless link we intend to add to the host PC a wavelet set of routines for the rejection of signal base line wobbling and also a selective power supply noise removal. Also a multi address wireless network may monitor more patients using a single host computer.
ACKNOWLEDGMENT We would like to thank Texas Instruments for their electronic samples and software which we used to be able to produce the experimental model.
on rearrangement gives:
In fact this is a very simple filtering method, but in practice the results are great and the low computation time gave us the chance to use the routine for 3 acquisition channels. These above mentioned channels do also have the possibility of an automatic signal level scaling. All inputs are calibrated for 1V. In case that the noise reduction capable 3 channels are programmed for the differential mode, then obviously the maximum channel number is reduced to 9 channels. Further signal processing is done by the host PC having much powerful DSP routines.
REFERENCES 1. 2. 3. 4. 5. 6.
Texas Instruments, Mixed Signal Microcontrollers Chicon Products, Texas Instruments, Single Chip Very Low Power RF Transceiver Sanjit K. Mitra, Digital Signal Prodessing. A computer based Approach, Mc Graw-Hill Jerry Luecke, Analog and Digital Circuits for Electronic Control System Applications using the MSP 430 microcontroller, Elsevier, ISBN 0-7506-7810-0 John Enderle,Susan Blanchard, Joseph Branzino, Introduction to Biomedical Engineering, Elsevier Academic Press, ISBN 0-12238662-0 David Prutchi, Michael Norris, Design and Development of Medical Electronic Instrumentation, Willey-Interscience ISBN 0-471-67623-3
III. CONCLUSIONS This wireless link is fairly simple and is suitable for patient monitoring than for diagnosing. Because the resulted overall bandwidth is in the mid range, the obtained acquisition rate may be programmed from 400 up to 4000 samples/s. As a result of the sampling theorem which states that:
Author: Nicolae Marius Roman Institute: Technical University of Cluj Napoca/ Biomedical Engineering Department Street: 26-28 Gh. Baritiu City: Cluj Napoca Country: Romania Email:
[email protected] [email protected] the frequency content of an analog ECG signal which is in the 0.5–100 Hz domain, is acquired without significant loses
IFMBE Proceedings Vol. 29
Corneal Quantitative Fluorometry – A Slit-Lamp Based Platform J.P. Domingues1,2, Isa Branco2, and A.M. Morgado1,2 1
Biomedical Institute for Research on Light and Imaing – University of Coimbra, Coimbra, Portugal 2 Department of Physics University of Coimbra
Abstract— Ocular Fluorometry has long been used (since early eighties) to measure non-invasively the presence and concentration of tracers in ocular tissues and fluids. The most common tracer has been sodium fluorescein, after systemic administration, but tissue native fluorescence has also been clinically valuable. Our goal is the development of a cooled CCD-camera based instrument configured as an accessory to a slit-lamp - a common instrument in ophthalmic observation of anterior eye - capable of measuring fluorescence in the eye from cornea to anterior vitreous with enough sensitivity and spatial resolution. Sensitivity of 0.1 ng/ml fluorescein equivalent concentration and 100 µm axial spatial resolution have been achieved with in vitro tests. This represent a crucial step forward in slitlamp based quantitative measurements as several new clinical issues can be addressed: Cornea auto-fluorescence and its relation with Diabetic Retinopathy and corneal function evaluation are two of them. With those figures of sensitivity and spatial resolution one can use narrower excitation bands in different wavelengths to address different fluorophores and also reduce slit widths and optimize angular positioning in order to reach inner locations in the eye with enough axial resolution. Accurate corneal in vivo fluorescence quantification evaluating its relation with age and pathologies like diabetes is our first step. Some results have already been achieved and are presented. These developments will also make it possible to improve quantification of Blood-Aqueous Barrier (BAB) leakage into anterior chamber and to assess anterior vitreous fluorescence resulting from Blood-Retinal Barrier (BRB) breakdown. Both are closely related with Diabetes progression. Keywords— Ocular Fluorometry, Slit-Lamp, CCD multielement sensors.
I. INTRODUCTION Diabetes has gained a strong social and economic impact in health care over the last decades and years. Its prevalence worldwide – mainly in developed countries – has dramatically increased: The 2007 National Diabetes Fact Sheet [1], most recent data available in US, states that in the age group of 20 years or older 23.5 million (or 10.7% of total) have diabetes and 57 million have pre-diabetes. The undiagnosed diabetes is 5.7 million and 1.6 million new cases are diagnosed each year (mostly in the age group of 40 – 59 years old). Loose of vision is one of major complications and Diabetes is the leading cause of new cases of
blindness among adults aged 20 – 74 years (Diabetic Retinopathy causes 12,000 to 24,000 new cases of blindness each year in the US). In European Union, according to a report issued by the International Diabetes Federation [2], the number of people suffering from diabetes increased by almost 20% (to 31 million) during the 2003-2006 period. The prevalence rate forecast for 2025 made in 2003 have been reached and surpassed in some countries in 2006. Only 13 of the EU's 27 member states have national plans to address diabetes. In some of them the direct costs reach 18% of total heath care spending. Diabetic Retinopathy (DR) and Diabetic Neuropathy (DN) are among the major chronic complications of diabetes mellitus. DR is the leading cause of blindness among adult population and peripheral DN is responsible for 5075% of non traumatic amputations. In both diseases early diagnosis is a key factor to define treatments and new therapies. Ocular fluorometry has long been used as an early diagnostic tool to DR [3][4][5] by quantification the Blood-Retinal Barrier leakage. In the eighties the first commertial Ocular fluorometer became available (Fluorotron Master, Ocumetrics, USA) and since then several research data has been published relating the amount of sodium fluorescein leakage into vitreous (after systemic administration of the sodium fluorescein tracer) with the grade of diabetic retinopathy and even before DR disease in diabetic patients[3][4][5][7]. More recently this line of research has been reinforced using more sophisticated instrumentation including ocular fundus angiographs [6]. Also Blood-Aqueous Barrier (BAB) proved to be a good indication of alterations in blood vessels permeability and, consequently, of the possibility of measuring DR progression[3][7][13]. Finally, consistent studies indicate that the corneal auto-fluorescence is also a good indicator of metabolic control in diabetic patients and, so, related with DR grading [8][9][11][14]. On measuring natural occurring fluorescence like corneal autofluorescence there is no need of tracer injection which is an enormous advantage as some adverse reactions occur to fluorescein. Also there is no need to take blood samples as n the case of Blood-Ocular barriers permeability evaluation. Therefore, accurate quantification of corneal auto-fluorescence which is usually as low as 10
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 475–478, 2010. www.springerlink.com
476
J.P. Domingues, I. Branco, and A.M. Morgado
ng/ml fluorescein equivalent concentration is well needed and opens new possibilities in clinic diagnosis.
Optical amplification is given by the combination of objective lens and focusing lens on a classic two-lens system (M =
II. MATERIALS AND METHODS A. Ocular Fluorometer Hardware A slit-lamp based Ocular Fluorometer has been developed by our group (US patent 06,013,034, EP 0 656 759 B1) with a multi-element sensor to quantify fluorescence along line segments of ocular globe by electronic scanning. A clinical study to measure fluorescein leakage into anterior chamber after systemic administration has already been conducted [13]. A new data acquisition system has been developed to improve sensitivity, measurement resolution, portability and to increase programmability. This is achieved by using a dsPIC Microcontroller (dsPIC30F6012A, Microchip, USA) together with a 16-bit, 1.25 MSPS ADC and allowing the possibility of using cooled CCD cameras. The communications with the PC (for PIC programming and data reading) are done either by USB or RS-232 and a robust power supply for overall instrument has been coupled. Next picture depicts a simplified block diagram. In vitro performance tests have been performed with this new architecture. This new hardware setup and mainly the possibility of using a cooled high sensitive CCD camera represent a crucial advancement in performance (sensitivity and spatial resolution) which will allow to address other locations in the eye (cornea and vitreous) and, mainly, to test the possibility of quantitatively evaluate different clinical situations (corneal auto-fluorescence and its relation with progression of Diabetes, corneal epithelial and endothelial function, effects of contact lenses in cornea, inflammation follow-up, vitreous fluorophotometry). Of course, it keeps the compatibility with a set of lower grade NMOS multi-element image sensors. The system is configured as an add-on to an ordinary slit-lamp - the most common ophthalmic equipment for anterior segment observation which gives the system the capability of widespread clinical use (Fig. 1). B. Optical Setup To reach the best results making use of the slit-lamp basic optics some additional components can be used. Excitation filters must be selected according with application. The standard slit-lamp filter set does not usually fit our demands. For preliminary tests we used a band pass filter (460-490 nm) with peak transmission (90%) at 480 nm. At emission side we introduced a standard high pass filter, HP 500 nm. A Zeiss 30SL/M slit-lamp was used.
f2 ) which can be used in conjunction with the slitf1
lamp built-in Galilean system. Another possibility is the use of a cylindrical lens to increase the image illuminance over the one-dimensional array detector for better detectivity. With high sensitivity camera sensors narrower excitation bands can be used improving fluorophore selectivity and also narrower slit widths can be selected improving spatial resolution. Also lower pixel pitch can, of course, contribute to a better resolution.
Fig. 1 Overall system diagram
III. RESULTS A. In Vitro A graphical user interface using MatLab and some software tools have been developed for data collection and analysis and preliminary tests have been performed using an Hamamatsu C5809 Multichannel Detector Head. This uses a thermoelectrically cooled FFT-CCD sensor with 24 µm pixel size in line binning operation, using charge integration method. Figures 2 and 3 show results of in vitro measurements to determine linearity at low fluorescein concentrations (of the order of corneal auto-fluorescence equivalent values, about 10 ng/ml fluorescein equivalent). Linearity was determined to be 2% (large percentage deviation of experimental points from best line fit).
IFMBE Proceedings Vol. 29
Corneal Quantitative Fluorometry – A Slit-Lamp Based Platform
477
geometry measurements can be done which is the case of corneal measurements.
Fig. 2 Linearity of measurements for low fluorescein concentrations
Fig. 4 Measurements of lateral spatial resolution with USAF 1951 test target B. Preliminary Corneal Measurements
Fig. 3 Output response for in vitro mesurements Another important parameter is the Lowest Level of Detection (LLOD) defined as background value (0 ng/ml) added to twice the standard deviation of its measurement distribution. It was found to be 0.1 ng/ml with the cooled CCD camera and 1.5 ng/ml fluorescein with conventional NMOS sensors. We also measured the lateral spatial resolution using the USAF 1951 target. Figure 4 depicts one of the results obtained with Group 1, Element 6 (3.56 LP/mm). Measurement resolution depends on Galilean system optical amplification and on the focusing lenses used. Of course pixel-to-pixel pitch and overall optics quality are important. We were able to reach 100 µm resolution. This relates directly with ocular axial resolution as long as 90º slit-lamp
The fluorometer proved to be accurate enough to measure lens auto-fluorescence and anterior chamber fluorescence (after either IV or oral fluorescein administration). Cornea auto-fluorescence has significantly lower intensity and its axial extent is only about 500 µm. We used the already mentioned cooled CCD camera with very low dark charge and tested different measurement geometry, optical amplification, filter set and slit width to obtain the best results. Also signal amplification must be accurately defined. We were able to perform preliminary in vivo corneal measurements in some volunteers with different measurements setup but measurement calibration and definition of standard protocols must proceed and be established. However promising results can already be foreseen. Figure 5 shows measurements with 90º slit-lamp geometry and 1 second integration time. The optical setup was the one already briefly described on which we used a 65 mm focal length focusing lens and 30 × Galilean amplification. Left-end peak is the beginning of lens autofluorescence peak – not shown. Two consecutive scans are shown with Y-scale representing sensor output in ADC relative units. Inter-scan reproducibility was found to be 7%. Much remains to be done towards calibration standards and geometry/optical settings optimization and, finally, in vivo validated studies to ensure the efficiency of the instrument to report for clinically relevant parameters and diagnosis.
IFMBE Proceedings Vol. 29
478
J.P. Domingues, I. Branco, and A.M. Morgado
cornea
Fig.
5 Preliminary results of quantification of corneal auto-fluorescence (two scans)
REFERENCES 1. 2008 National Diabetes Fact Sheet, general information and national estimates on diabetes in the United States, Atlanta, GA: US Department of Health and Human Services 2008. 2. Hall, Michael, Together we are stronger, report presented to the IDF Europe General Assembly, Sep. 2008. 3. Yoshida Akitoshi et al., Permeability of blood-ocular barriers in adolescent and adult diabetic patients. British Journal of Ophthalmology 1993; 77: 158-161 4. Cunha-Vaz JG et al. Blood-Retinal Barrier permeability and its relation to progression of retinopathy in patients with type 2 diabetes. A four-year follow-up study. Graefe’s Arch Clin Exp Ophthalmol (1993) 231:141-145. 5. Cunha-Vaz, J. G. The Blood-ocular barriers: past, present and future, Documenta Ophthalmologica 93: 149-157, 1997.
6. C. Lobo R. Bernardes, J. Figueira et al. Three-year follow-up study of blood-retinal Barrier and retinal thickness alterations in patients with type-2 diabetes mellitus and mild non-proliferative diabetic retinopathy. Arch. Ophthalmol., 122:211-217, 2004 7. Schalnus, Rainer, Christian Ohrloff, Eckart Jungmann, Kerstin Maaβ, Stephan Rinke Anette Wagner - Permeability of the Blood-retinal Barrier and the Blood Aqueous Barrier in Type I diabetes without diabetic Retinopathy: Simultaneous Evaluation with Fluorophotometry, German J. Ophthalmol 2: 202-206, 1993. 8. Stolwijk TR, van Best JA. Corneal auto fluorescence by fluorophotometry as indicator of diabetic retinopathy. Invest Ophthalmol Vis Sci 32 (Suppl.):1067 (1991). 9. van Schaik HJ, Coppens J, van den Berg TJ, van Best JA. Autofluorescence distribution along the corneal axis in diabetic and healthy humans. Exp Eye Res 69(5):505-10 (1999). 10. Cunha-Vaz, J. Domingues, JPP, Correia, CMBA Ocular Fluorometer, EP 0 656 759 B1, European patent (1998). 11. Van Best, J et al. Simple, low-cost, portable corneal fluorometer for detection of the level of diabetic retinopathy, Applied Optics, Vol. 37 No. 19, 4303-4311. 12. Cunha-Vaz, J. Domingues, JPP, Correia, CMBA Ocular Fluorometer, US patent 06,013,034 (2000). 13. Domingues J. P. P.,Figueira J., Correia C. M. , Cunha-Vaz J. G. Blood-Aqueous Barrier Permeability Assessment By Ocular Fluorescence Measurements After Oral And Iv Fluorescein Administration, , IFMBE Proceedings, Vol. 11. Prague: IFMBE, 2005. ISSN 17271983. Editors: Jiri Hozman, Peter Kneppo (Proceedings of the 3rd European Medical & Biological Engineering Conference EMBEC’05. Prague, Czech Republic, 20-25.11.2005). 4752 p. 14. Satoshi Ishito et al. Corneal and Lens autofluorescence in young insulin-dependent diabetic patients, Ophthalmologica, 212: 301-5 (1998) Author: Institute: Street: City: Country: Email:
IFMBE Proceedings Vol. 29
José Paulo Domingues Biomedical Institute For Research on Light and Image Az. Sta Comba - Celas Coimbra Portugal
[email protected] Automatic Detection of Patients’ Spontaneous Activity During Pressure Support Ventilation G. Matrone1, F. Mojoli2, A. Orlando2, A. Braschi2 and G. Magenes1 1
2
Dept. of Computer Engineering and Systems Science, University of Pavia, Pavia, Italy Dept. of Surgical Sciences - Anaesthesia and Intensive Care, University of Pavia, Pavia, Italy
Abstract— The occurrence of significant patient-ventilator asynchronies in assisted ventilation modes is an impellent problem in clinical practice. Addressing this question, an original software has been developed and is here proposed. This tool implements a new automatic technique to identify the beginning and the end of the patient’s respiratory effort, events that are sometimes missed or detected with significant delay by the ventilator. Its improved skills have been evaluated on a set of signals coming from 6 ICU patients and including 6445 respiratory acts, and proved to outperform the machine in increasing the amount of respiratory acts assisted without significant delay from 22 to 70%. The presented tool is the first step in the development a hardware-software device to be directly interfaced with the ventilator in order to represent a monitoring aid for the clinician and possibly to directly drive the device activity. Keywords— Pressure Support Ventilation, asynchrony, inspiratory trigger, expiratory trigger.
I. INTRODUCTION
Positive Pressure mechanical Ventilation (PPV) is the fundamental assistive resource in modern Intensive Care Units (ICU) for patients suffering from respiratory failure. Basically, it aims to provide the patient with adequate oxygenation and removal of carbon dioxide. Mechanical ventilation devices are said to work in controlled mode when the breathing cycle is completely machine-controlled, or instead in assisted mode when its intervention is synchronized to the patient’s spontaneous breathing activity. Among these, Pressure Support Ventilation (PSV) is the most widespread assisted ventilation mode in clinical practice. In assisted mode, the ventilator synchronizing systems (i.e. inspiratory and expiratory triggers), should detect respiratory muscles activations and relaxations in order to correctly drive the device activity [1], which consists in the opening/closing of the inspiratory and expiratory valves. In modern ventilators, both inspiratory and expiratory triggering systems are flow-based. The first one opens the inspiratory valve, providing pressure support to the patient whenever his/her spontaneous activity generates an inspiratory flow overcoming a set threshold. Expiratory
trigger instead causes the inspiratory valve closure while opening the expiratory one, only when the inspiratory flow is below an adjustable threshold (usually defined as a fraction of peak inspiratory flow). Particularly, when treating patients with obstructive pulmonary disease patientmachine asynchronies are frequently prone to occur. In these cases the machine is often unable to recognize the patient’s spontaneous breathing activity or recognizes it but with a significant delay. As a matter of fact, surveys in the literature assess that asynchronies can be noticed in 10 to 97% of respiratory acts during assisted ventilation [2]. A non-optimal interaction between patient and ventilator can either directly damage the respiratory muscles or, in any case, cause complications leading to prolonged mechanical ventilation, longer ICU stay and worse outcome [3] [4]. Alternative ventilation modes have been developed in order to address such a problem [4] [5]. Actually though, they are not always implemented by common ICU ventilators and their primacy hasn’t been assessed in everyday clinical practice yet. In this paper a new software for the automatic detection of the patient’s spontaneous respiratory activity – i.e. contraction and subsequent relaxation of respiratory muscles – is presented. The automatic detection technique here introduced is based on the identification of sudden changes in the flow trajectory due to the patient’s respiratory activity. Such an improved skill will be employed to real-time monitor patient-ventilator synchronization during traditional (flow-based) trigger operation, and it could also be used to directly drive the ventilator activity.
II.
MATERIALS AND METHODS
A. Patients recruitment Data presented in this paper refer to a set of signals recorder from 6 ICU patients, including even 6445 respiratory acts. All the enrolled subjects were hospitalized at the Intensive Care Unit, Policlinico S. Matteo, Pavia, Italy, and assisted by means of the Galileo ICU ventilator (Hamilton Medical AG, Rähzüns, Switzerland), running in PSV mode. Selected patients were characterized by an high-
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 479–482, 2010. www.springerlink.com
480
G. Matrone et al.
frequency asynchronies occurrence, as it was detectable on the ventilator monitor screen. These patients were all difficult to wean from mechanical ventilation; as a matter of fact, they underwent PSV for about 15 days. Patients respiratory data recordings, used for the subsequent analysis, lasted about one hour on the average. The ventilator respiratory cycling operations, i.e. inspiratory/expiratory triggers, were flow-controlled. Their respective threshold values were set to 3 l/min and to the 30% of inspiratory flow peak value. The ventilator was connected via RS-232 serial port to a stand-alone personal computer (Intel Core2 Duo, 2.4 GHz) running the data acquisition software. Pressure, flow and volume data recorded from the machine sensors were acquired with a sampling rate of 68 Hz. While running, the software stored all data records in a text file, which was available to the clinician at the end of the recording for further analyses. B. Visual analysis Modern ventilators are able to display the patient’s airway pressure and airflow variations in real-time, allowing the clinicians to visually evaluate patient-ventilator interaction [3]. In particular, patient-ventilator asynchronies are usually identified by the clinician by accurately observing the airflow curve (Fig. 1) [1]. Such events can be broadly classified as inspiratory or expiratory asynchronies. Concerning these first ones, the ventilator may be significantly late in assisting the patient inspiratory act (inspiratory delay) or may not deliver any support at all (ineffective effort). Inspiratory delay is defined as the time between the beginning of the inspiratory effort and the inspiratory valve opening by the ventilator. The real beginning of the inspiratory effort can be recognized as a sudden and sustained upward deviation of the flow signal (Fig. 1). On the other hand, expiratory asynchronies take place when the expiratory valve is opened too early (early
Fig. 1 Typical patient-ventilator asynchrony patterns visible on the airflow waveform: inspiratory delay, expiratory delay and ineffective effort.
Fig. 2 The designed software graphical interface. Airways pressure (top), flow (middle) and the synthetic signal (bottom) waveforms are displayed. cycling-off) or late (delayed cycling-off) with respect to the patient’s inspiratory musculature relaxation. Also in this case, the patient’s relaxation (i.e. the end of inspiratory effort) can be detected as a sudden change in the flow trajectory. Similarly to inspiratory delay, the expiratory one represents the time between the typical flow trajectory change and the expiratory valve opening (Fig. 1). In this work, visual inspection has been performed by a single operator (A. O.) and used as the standard reference for software validation. C. Data visualization and analysis software A new software has been developed in order to graphically represent and process respiratory data coming from mechanically ventilated ICU patients (Fig. 2). Up to now, it has been used as an off-line processing tool, working on data recordings already stored in text files by the ventilator data acquisition software (sampling rate = 68 Hz). At the moment, a real-time implementation is being developed, in order to conjugate data acquisition with online visualization and elaboration. The visualization and analysis software has been ad-hoc designed using LabVIEW (National Instruments Corp., Austin, TX, USA). For what concerns mere signals display, after choosing the data file to be processed, our tool is able to represent airway pressure (cmH2O), volume (ml) and airflow (ml/s) curves, dynamically evolving in time. The clinician can select which of these waveforms are to be displayed and set the width of the temporal window per screen to be shown (5-15-30 sec). The evolution of the selected waveforms on the screen can be paused in order to switch to a more accurate visual analysis mode. Moveable cursors are made visible on each one of the three paused graphs, and the user can always see which value (both time and amplitude) the cursor is pointing at while being moved over the curve. Cursors can also be used to mark any point of interest on
IFMBE Proceedings Vol. 29
Automatic Detection of Patients’ Spontaneous Activity during Pressure Support Ventilation
plots to be saved into a text file for subsequent analysis (e.g. contractions and relaxations of respiratory muscles). Not only the system is able to represent physiological curves: the user can also choose to visualize an additional waveform, synthetically generated by the software itself (Fig. 2, third graph). This new signal will represent the basis for the development of an innovative and more accurate triggering system; from here forth, it will be called flow trajectory trigger. The conceived algorithm mainly involves the airflow signal derivative computation. Since flow data can be affected by high frequency noise, digital low-pass prefiltering is necessary in order to avoid undesired spikes generation and alterations of the new signal shape. Different filter implementations were tested in order to find out the most appropriate one and to determine a trade-off between noise filtering and signal delaying, after a previous Fourier analysis. Particularly, both moving average and a 3rd order digital Butterworth IIR low-pass filter, with a cut-off frequency of 2 Hz (significant frequency band of the flow signal almost under 5 Hz) allowed to obtain the desired result. However, considering that IIR filters have non-linear phase response and can also alter the signal shape, in the end the airflow signal was smoothed using a moving average filter (span = 25 samples). In order to generate the synthetic signal (ml/s2), the digitally filtered airflow signal must be numerically differentiated over time. We have defined two temporal windows in which the synthetic signal is calculated by the system (otherwise it is set to zero). The first time window is placed between the inspiratory peak of filtered flow and the cycling-off time (expiratory valve opening and inspiratory valve closure). The second one starts when the filtered flow reaches the 90% of peak expiratory (filtered) flow and lasts until the opening of the inspiratory valve. If Δd is the variation of the filtered flow derivative, computed in each of these temporal windows, the signal value is equal to Δd
Fig. 3 Behavior of the new monitoring system during inspiratory/expiratory delays and ineffective efforts. Typical asynchornies patterns are shown by the flow curve (top) and by the corresponding synthetic signal (bottom). Green and red dots are produced by the system to identify patient’s activity.
481
when the derivative is positive; the signal equals –Δd when the derivative is negative. This way, the signal is positive whenever the flow trajectory deviation represents an inspiratory muscles contraction; the signal is negative whenever the flow trajectory deviation represents an inspiratory muscles relaxation. A simple threshold comparison can be then used to identify triggering events. An inspiratory trigger occurrence is “announced” by our automatic system when the variation of the synthetic signal significantly exceeds a user-defined positive threshold (e.g. 150 ml/s2). This event is pointed out by the software displaying a green dot (Fig. 2, 3) just below the flow curve. Similarly, an expiratory trigger is signaled by a red dot, whenever the signal variation goes significantly below a negative threshold (e.g. -100 ml/s2) (Fig. 2, 3). This new expiratory trigger works together with the traditional flowbased one in a competitive manner. As a matter of fact, a red dot appears also when the flow decreases under a userdefined flow threshold. All the mentioned threshold values can be set and modified by the clinician using the graphical interface. Yellow horizontal line segments, instead, correspond to the mechanical inspiratory time, that is the time between inspiratory valve opening and closure (Fig. 3).
III.
RESULTS
Patient recordings analyzed in this paper last about 6 hours altogether. Visual analysis succeeded in identifying 6445 respiratory acts with their corresponding muscular contraction and relaxation times. In some cases, the visual analysis wasn’t able to recognize the patient’s activity with sufficient reliability. Thus, these data (less than 2%) were neglected. The operator was able to visually identify 1758 ineffective efforts (27% of all patients’ efforts). Among 4687 assisted acts, the average inspiratory delay was 330±241 ms while the expiratory delay was 65±15 ms. A significant inspiratory delay (>200 ms) occurred in 3164 acts (49%); expiratory delays greater than 200 ms were instead observed in 650 acts (10%). 51% of respiratory acts were assisted by the ventilator with significant delay; therefore, only 22% of patients’ efforts were correctly supported. On the other hand, our software recognized 6425 patients’ respiratory acts (99.7%). In 5798 cases, both inspiratory muscles contraction and relaxation were detected. In 647 acts (10% of 6445) only the relaxation was identified. The beginning of patients’ inspiratory effort was recognized 126±96 ms after the operator did (visual inspection); in 1535 acts (17.6% of 6445) the delay was more than 200 ms. Considering the mechanically supported respiratory acts, the new system anticipated the ventilator
IFMBE Proceedings Vol. 29
482
G. Matrone et al.
Fig. 4 Patient–ventilator interactions during traditional flow triggering versus flow trajectory triggering.
by 224±16 ms. Muscular relaxation was identified 15±17 ms after the operator did. Only in 158 cases (2.5%) this delay was more than 200 ms. The percentages of non/well/delayed assisted acts are resumed in Figure 4, when operating in two different conditions: traditional flow-based triggering (real condition) and our flow trajectory-based trigger (hypothetical condition). In this last case the new triggering system was assumed to drive the ventilator.
IV.
DISCUSSION AND CONCLUSIONS
In our case study, the ventilator device correctly gave its support to the patient only in a little more than one case out of five. A high number of ineffective efforts was observed (in almost one out of four respiratory acts) and inspiratory/expiratory delays were detected in about 50% of assisted breaths. The newly developed software has been used by clinicians as a visual and elaboration aid, in order to identify patient-ventilator asynchronies and to evaluate the quality of the mechanical ventilation device behavior. Our visualization tool is highly sensitive in immediately highlighting ineffective efforts (which actually are the most typical asynchronies in PSV) almost in 99% of occurrences; it behaves similarly or sometimes even better than previously developed methods [6] [7]. For what concerns triggering, our software identifies more than the 63% of efforts not assisted by the ventilator (Fig. 3), thanks to the automatic identification algorithm previously introduced and based on the synthetic signal computation. Thus, if this tool were running in synergy with the machine itself, probably almost nine out of ten spontaneous breathing acts would be recognized and assisted. Even when the ventilator supports the patient’s breathing activity, the software allows to detect the muscular activity in advance (~220 ms) with respect to the machine; this way, inspiratory delay could be significantly reduced. For example, we can see in Figure 3 that the first respiratory act is recognized by the ventilator but with meaningful inspiratory and expiratory delays
(yellow horizontal segment). The second respiratory act is completely missed by the ventilator. Our automatic system instead promptly identifies both the beginning and the end of the two patient’s inspiratory efforts (green and red dots). Altogether, if our triggering system were to drive the ventilator, the correctly (without delays) assisted acts would be 70% compared to 22% of the real case. Summarizing all the obtained results, we can assert that the implemented algorithm outperforms the machine flow-based triggering system. So far, our software has proved to be a reliable offline monitoring system which is likely to improve patientventilator interaction. Further developments are foreseen in the immediate future. First of all, the described software functionalities are being improved in order to supply it with an automatic system for both asynchronies detection and classification. Next, a real-time implementation will be developed, including data acquisition operations. Connecting the PC running our program to the ventilator will provide the clinician with an additional but more reliable monitoring and data analysis system. The long-term objective of this work is to provide the ventilator machine with a new flow trajectory based triggering system.
REFERENCES 1.
2.
3.
4.
5.
6.
7.
Mojoli F, Venti A, Pozzi M, Via G, Braschi A (2009) Patientventilator interaction during Pressure Support Ventilation: how to monitor and to improve it. Proc. 63rd SIAARTI Conf., Florence, Italy, 2009, 75(7-8):533-536 Thille AW, Rodriguez P, Cabello B, Lellouche F, Brochard L (2006) Patient-ventilator asynchrony during assisted mechanical ventilation. Intensive Care Med 32(10):1515-1522 Georgopoulos G, Prinianakis G, Kondili E (2006) Bedside waveforms interpretation as a tool to identify patient-ventilator asynchronies. Intensive Care Med 32:34 Xirouchaki N, Kondili E, Vaporidi K et al. (2008) Proportional assist ventilation with load-adjustable gain factors in critically ill patients: comparison with pressure support. Intensive Care Med 34(11):20262034 Brander L, Leong-Poi H, Beck J et al. (2009) Titration and implementation of neurally adjusted ventilatory assist in critically ill patients. Chest 135(3):695-703 Mulqueeny Q, Ceriana P, Carlucci A et al. (2007) Automated detection of ineffective triggering and double triggering during mechanical ventilation. Intensive Care Med 33:2014-2018 Younes M, Brochard L, Grasso S et al. (2007) A method for monitoring and improving patient: ventilator interaction. Intensive Care Med 33:1337-1346
Author: Giulia Matrone Institute: Dept. of Computer Engineering and Systems Science, University of Pavia Street: via Ferrata, 1 City: 27100 Pavia Country: Italy Email:
[email protected] IFMBE Proceedings Vol. 29
Determination of In Vivo Three-Dimensional Lower Limb Kinematics for Simulation of High-Flexion Squats P.D. Wong1, B. Callewaert2, K. Desloovere2, L. Labey1, and B. Innocenti1 1
2
European Centre for Knee Research, Smith & Nephew, Leuven, Belgium University Hospital Pellenberg, Katholieke Universiteit Leuven, Leuven, Belgium
Abstract—In vitro and numerical simulations of the knee require reasonable kinematic and load inputs and boundary conditions, in order to help ensure their clinical relevance. However, previous simulations of high-flexion squats often have applied loads and motions that possibly oversimplify the true knee kinematics. This study aimed to improve future simulations of squatting by obtaining three-dimensional squat kinematics from a cohort of healthy adults. Seventeen subjects (age range 24-75) underwent motion capture sessions using a standard, systematic clinical procedure. Joint positions were normalized versus femur and tibia segment lengths, and ground reaction forces were normalized versus body weight. Range of motion and velocity decreased with age. The ankle was more anterior to the hip with decreasing hip height. Dynamic squat kinematics were reported.
of joint loads, which then can better aid the evaluation of knee pathology and treatments. However, the literature lacks attempts to define “average” or standard squat kinematics. Without this data, the design of better test systems is based more on assumption rather than a population. Considering this problem, this study attempted to define three-dimensional lower-body kinematics of typical adult subjects while they performed high-flexion body-weight squats. It would normalize the data and report it in a general form that can act as inputs to knee kinematics simulations, particularly for electromechanical machines that are more complex than the first Oxford Rig.
Keywords— squat, high flexion, motion analysis, knee simulator, healthy subjects
I. INTRODUCTION
Researchers often perform in vitro or computational studies to simulate in vivo knee biomechanics. This gives an alternative when in vivo studies are impractical or invasive to patients. The clinical relevance of these simulations then relies on the definition of plausible load inputs and boundary conditions. For example, previous studies often use electromechanical systems and computer modeling to simulate the knee joint during a squat, which requires various assumptions about motion curves, loads, and muscle connections [1,2]. Although such studies have produced much useful information so far, their clinical relevance still may be limited. The squat kinematic simulators in the literature today are modeled off the “Oxford Rig” design reported in 1997 [3]. This machine advanced research capabilities at the time, as it was a controllable six-degree-of-freedom joint simulator that could produce vertical motion. However, it lacked the ability to control anteroposterior or mediolateral motion, and therefore could only simulate a simplified squat, more like squat up against a wall (Fig 1). Better knee simulations should incorporate the full threedimensional motions of the lower limb to be more realistic. This could hypothetically allow more accurate simulations
Fig. 1 Schematic lateral view of a simulated deep squat where the hip lies directly over the ankle, versus more realistic squat kinematics
II.
MATERIALS AND METHODS
Seventeen adult subjects (age range 24-75, 6 female, 11 male) with no reported musculoskeletal pathologies volunteered for this study after giving informed consent. They each underwent one motion analysis session, using a 14camera optical motion tracking system (Vicon, Oxford, UK), two forceplates (AMTI, Watertown, MA, USA), and a standard clinical kinematic model (Plug-in-Gait [4] with Knee Alignment Device [5], Vicon, Oxford, UK). In each session, a subject was asked to stand with their feet over two separate forceplates, which were spaced 115mm apart, and then perform a high-flexion squat. This consisted of descending as far down as comfortably possible, and then
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 483–486, 2010. www.springerlink.com
484
P.D. Wong et al.
rising back up to standing position, without using upper limb support (e.g.. no holding the thighs with the hands). Beyond these instructions, all subjects used self-selected speeds and postures. Three repeated squat trials were taken, and one trial for each subject with no loss of balance was identified for further analysis. Subject age, height, and mass were recorded. The motion tracking system measured ground reaction forces, calculated joint centers based on skin marker trajectories, and calculated joint rotations with Euler angles. Femur and tibia segment lengths were measured from the motion tracking data with automated algorithms (Matlab, Mathworks, Natick, MA, USA). Femur length was taken as the average distance between the hip joint center and knee joint center throughout the squat, and tibia length was taken similarly between the knee and ankle. The femur-to-tibia length ratio (Fem/Tib) and total femur-plus-tibia leg length (Fem+Tib) were recorded for each subject. Data were then generalized, so that they could be used as inputs into typical load- and motion-controlled knee simulators, which often use linear actuators. Linear translations were measured as follows (Fig 2).
III.
RESULTS
Average subject characteristics (n=17) are summarized in Table 1. Subjects had a healthy average body-mass index but still exhibited a wide variety of characteristics. Table 1 Subject characteristics Mean
SD 15.3
Min 23.9
Max
Age (y)
49.8
75.4
Mass (kg)
71.8
13.5
45.3
97.6
Height (cm)
172.9
9.5
158.0
190.0
Body-mass index
23.9
3.8
17.7
30.8
Femur length (mm)
404.1
30.0
364.0
455.5
Tibia length (mm)
407.1
26.8
364.7
454.1
Fem+Tib (mm)
811.1
54.5
733.4
909.6
Fem/Tib ratio
0.993
0.040
0.931
1.067
Squat cycle time (s)
4.207
1.411
2.320
7.050
For each year older, squat times slowed by 0.0634 s, minimum HH increased by 0.50% of Fem+Tib length, and maximum knee flexion decreased by 0.70° (p 0.75, indicating adequate blood supply to the wounded limb. All wounds were debrided of necrotic tissue before entering into the study. Patients diagnosed with osteomyelitis had excision of the infected bone and treatment with antibiotics before enrollment. Each of the sixteen patients underwent a standard wound care routine for their foot ulcers, which consisted of weekly or biweekly debridement, offloading when possible, and treatment with moist wound healing protocols. When indicated, active wound healing modalities such as hyperbaric oxygen, negative pressure wound healing, and active biosynthetic skin substitutes were used. Optical measurements were done prior to weekly or biweekly debridement. Serial measurements were obtained at every patient visit to the clinic from the time of enrollment until complete wound closure, amputation of the limb, or a
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 487–489, 2010. www.springerlink.com
488
M. Neidrauer et al.
maximum of 20 visits without closure or amputation. A wound was classified as “non-healing” for the purposes of this study if it did not heal by the 20th visit or if the limb was amputated. The optical measurements of wounds were conducted at varying sites, depending on the geometry and location of each wound. Locations measured included the area (1) directly on the wound (2) on intact skin at the edge of the wound (3) on non-wound tissue on the wounded limb at a distance of at least 2 cm from the wound, and (4) nonwound tissue on the contralateral limb as symmetric to wound location as possible. The control site chosen varied between a location on the wounded or contralateral limb, depending on access to the site due to other wounds or previous amputations. Tegaderm transparent sterile dressing (3M Health Care) was used to cover the fiber optic probe
during all measurements. During every visit a digital photograph was taken of the wound after the near infrared data was collected. Crosspolarizing filters were used to reduce surface reflections. A paper ruler was used in each photograph to correct for variations in the distance between the camera and wound. The wound boundary was traced and the surface area was calculated using an image analysis program created with MATLAB computing software (MathWorks, Inc.). III. RESULTS AND DISCUSSION
Of the 16 wounds studied, 7 wounds completely healed
OxyhemoglobinConcentration OxyhemoglobinConcentration
0.25
0.35
0.20
Conc.(mM)
Conc.(mM)
0.30 0.25 0.20 0.15
0.15 0.10
0.10
0.05
0.05
0.00
0.00
0 0
10
20
30
40
50
12
Area(cm2)
Area(cm2)
10 8 6 4 2 0 30
Fig
1. Example of data obtained from a healing wound that closed after 41 weeks. (upper) Oxyhemoglobin concentration [HbO2] as measured by the NIR device at the wound center (Ɣ), wound edge (ǻ), and control site on the same limb as the wound (+). The solid line is the linear trendline associated with data obtained from the wound center (slope = -3.9 ȝM/wk); the dashed line is the linear trendline associated with data obtained from the wound edge (slope = -2.9 ȝM/wk). The slopes of both trend lines are negative, as is characteristic of the healing wounds in this study. (lower) Wound sizes measured on each day.
10
20
30
time(weeks)
40
time(weeks)
Fig
45 40 35 30 25 20 15 10 5 0 0
20
30
WoundArea
WoundArea
10
20
time(weeks)
time(weeks)
0
10
2. Example of data obtained from a non-healing wound that resulted in amputation of the limb after 28 weeks of participation in the study. (a) Oxyhemoglobin concentration [HbO2] as measured by the NIR device at the wound center (Ɣ), wound edge (ǻ), control site on the same limb as the wound (+), and control site on the contralateral limb (x). The solid line is the linear trendline associated with data obtained from the wound center (slope = 0.4 ȝM/wk); the dashed line is the linear trendline associated with data obtained from the wound edge (slope = 0.0 ȝM/wk). The slopes of both trend lines are nearly zero or slightly positive, as is characteristic of the non-healing wounds in this study. (b) Wound sizes measured on each day.
IFMBE Proceedings Vol. 29
Evaluation of Chronic Diabetic Wounds with the Near Infrared Wound Monitor
and 9 wounds remained unhealed or resulted in amputation. In both healing and non-healing wounds, oxyhemoglobin concentration [HbO2] and total hemoglobin concentration [Tot Hb] during the initial measurement session was greater at the wound centers and wound edges than at the control sites. In the seven wounds that healed, wound measurements of [HbO2] and [Tot Hb] decreased gradually over time and converged with control site values of [HbO2] and [Tot Hb]. An example of data obtained from a healing wound is given in Figure 1. In the nine non-healing wounds, [HbO2] and [Tot Hb] at the wound sites remained elevated throughout the duration of the study and did not converge with control site values. An example of data obtained from a non-healing wound is given in Figure 2. The rates of change in hemoglobin concentration over time were quantified by fitting a linear trend line to the measured values. The [HbO2] and [Tot Hb] slopes for all healing wounds were negative while the slopes of nonhealing wounds were nearly zero or slightly positive. The mean and standard error of slopes obtained from healing and non-healing wounds are compared in Figure 3. The healing and non-healing groups were compared using two– tailed heteroscedastic t-tests, and a significant difference was found between the [HbO2] and [Tot Hb] slopes of healed and non-healing wounds (p < 0.05).
rate of chnage (ȝM/wk)
0.0
*
moglobin concentration over time were used to differentiate healing from non-healing wounds in a study of human diabetic foot ulcers, indicating that this method may be able to help wound care clinicians in the assessment of overall wound health when treating diabetic foot ulcers.
IV. ACKNOWLEDGMENTS The authors would like to thank Varshana Gurusamy, Sarah Kralovic, Usha Kumar, and Xiang Mao for their help with wound measurements. This research was made possible by the generous support of the Wallace H. Coulter Foundation and the U.S. Army Medical Research Acquisition Activity. This research was funded in part by The U.S. Army Medical Research Acquisition Activity, 820 Chandler Street, Fort Detrick, MD 21702-5014 is the awarding and administering acquisition office. This investigation was funded under a U.S. Army Medical Research Acquisition Activity; Cooperative Agreement W81XWH 04-1-0419. The content of the information herein does not necessarily reflect the position or the policy of the U.S. Government or the U.S. Army and no official endorsement should be inferred.
V. REFERENCES
MeanSlopesofHemoglobin concentration 2.0
1.
*
2.
Ͳ2.0 Ͳ4.0
3.
Ͳ6.0 Ͳ8.0
J. Mobley and T. Vo-Dinh, Optical properties of tissue, in Biomedical Photonics Handbook. 2003, CRC Press, Boca Raton, Fla. E. S. Papazoglou, M. S. Weingarten, L. Zubkov, M. Neidrauer, L. Zhu, S. Tyagi, and K. Pourrezaei, Changes in optical properties of tissue during acute wound healing in an animal model. Journal of Biomedical Optics, 2008. 13: p. 044005. E. S. Papazoglou, M. S. Weingarten, L. Zubkov, L. Zhu, S. Tyagi, and K. Pourrezaei, Optical Properties of Wounds: Diabetic Versus Healthy Tissue. IEEE Transactions on Biomedical Engineering, 2006. 53(6): p. 1047-1055.
Ͳ10.0 [TotHb]
[OxyHb]
489
[DeoxyHb]
HealingWounds(N=7) NonͲHealingWounds(N=9)
Fig. 3: Mean ± standard error of the temporal slopes of total, oxy-, and deoxy-hemoglobin concentration for healing and non-healing wounds. A significant difference was found between the [HbO2] and [Tot Hb] slopes of healed and non-healing wounds (*p < 0.05, two–tailed heteroscedastic t-tests).
These results indicate that temporal changes in the concentration of hemoglobin derived from diffuse near infrared measurements of the optical absorption coefficient in diabetic foot ulcers can be used to monitor healing progress. Changes in the calculating the linear rate of change of heIFMBE Proceedings Vol. 29
Non-contact UWB Radar Technology to Assess Tremor G. Blumrosen1, M. Uziel2, B. Rubinsky1, and D. Porrat1 1
Hebrew University of Jerusalem, School of Engineering and Computer Science, Jerusalem, Israel 2 Hebrew University of Jerusalem, Applied Physics Department, Jerusalem, Israel
Abstract— This work quantifies and analyzes tremor using Ultra Wide Band (UWB) radio technology. The UWB technology provides a new technology for non contact tremor assessment with extremely low radiation and penetration through walls. Tremor is the target symptom in the treatment of many neurological disorders such as Parkinson’s disease (PD), midbrain tremor, essential tremor (ET) and epilepsy. The common instrumental approaches for the assessment of tremor are motion capture devices and video tracking systems. The new tremor acquisition system is based on transmission of a wideband electromagnetic signal with extremely low radiation, and analysis of the received signal composed of many propagation paths reflected from the patient and its surroundings. An efficient UWB radar detection technique adapted to tremor detection is developed. Periodicity in the time of arrival of the received signal is detected to obtain tremor characteristics. For a feasibility test we built an UWB acquisition system and examined the performance with an arm model that fluctuated in the range of clinical tremor frequencies (3-12 Hz). A devlpoment of this work can lead to a monitoring system installed at any home, hospital or school to continuously asses and report tremor conditions during daily life activities. Keywords— Tremor, UWB, human radar signature, detection techniques.
I. INTRODUCTION Tremor is the target symptom in the treatment of numerous neurological disorders such as Parkinson’s disease (PD), midbrain tremor, and essential tremor (ET) [1]. Quantification and analysis of tremor is significant for diagnosis and establishment of treatments. For clinical research purposes, a number of scales have been developed for semiquantitative assessment of frequency and magnitude [2] of tremor. Motion capture devices such as accelerometers [2] or gyroscopes [3], are the most popular for tremor assessment. But must be attached to patient’s body and have limited capabilities on giving precise tremor amplitude due to amplitude drift [3]. Video recording is another popular technology for tremor assessment in gait analysis laboratories [4], but requires the patient to be inside the range of the video camera lens and consequently cannot be used for continuous assessment of tremor during daily life activities. A narrow band radar has been used [5] for the detection and classification of people’s movements and location based
on the Doppler signatures. When humans walk, the motion of various components of the body including the torso, arms, and legs produce a characteristic Doppler signature. Fourier transform techniques were used to analyze these signatures and identified key features representative of the human walking motion. [6] uses a classifier on the human body radar signature to characterize gait, in particular step rate and mean velocity. Radar techniques based on Doppler cannot detect tremor as the signal bandwidth they use, and correspondingly the temporal and spatial resolution, is usually too low to detect typical tremor. Ultra-wideband (UWB) is a radio technology that can be used with very low energy levels for short-range highbandwidth communications, by using a large portion of the radio spectrum. The potential strength of the UWB radio technique lies in its use of extremely wide transmission bandwidths, which result in accurate position location and ranging, and material penetration. Most recent applications target sensor data collection and locating and tracking applications such as [7]. [8] suggests the use of biomedical applications of UWB radar for cardiac biomechanic assessment and chest movements assessment, OSA (obstructive sleep apnoea), and SID (sudden infant death syndrome) monitoring. We suggest to quantify and analyze tremor with UWB radar technology. The UWB radar technology is based on transmission of a wideband electromagnetic signal and analysis of the received signal reflected from the patient to assess tremor characteristics. We provide data analysis tools for the UWB tremor acquisition system and give preliminary results for an UWB tremor acquisition system prototype we built. This paper is organized as follows. Section 2 describes the UWB tremor acquisition system and efficient algorithms to assess tremor characteristics. Section 3 describes the experimental set-up which consists of an arm model with tremor and an UWB tremor acquisition system. We performed a series of experiments with different distances in the range of 1-2 meters between the acquisition systems and the arm with different sources of disturbance. Section 4 analyzes the performance of the UWB acquisition system. Section 5 concludes the work and gives suggestions for future research.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 490–493, 2010. www.springerlink.com
Non-contact UWB Radar Technology to Assess Tremor
491
II. SYSTEM AND METHODS
Linear Minimum Mean Square Error (MMSE) criterion with the tremor amplitude and frequency constraints is:
A. System Model Our system is composed of a transmitter and a receiver. A high bandwidth pulse is transmitted into the medium where the patient is located. The signal that has propagated through a wireless channel consists of multiple replicas (echoes, mainly caused by reflections from objects in the medium) of the originally transmitted signal, named Multipath Components (MPCs). Each MPC is characterized by attenuation and a delay. The received signal at time instance t is: r t
∑
,
β t
mT p t
mT
τ
,
n t
(1)
Where p(t) is a pulse with typical duration around 10 ns, m is the pulse index and T is the pulse repetition time, τm,k is the k’th MPC delay in the m’th pulse, βm,k is its related attenuation factor which is assumed constant for a short observation time, n(t) is an additive noise component. The noise includes thermal and amplifier noise which can be modeled by white Gaussian processes, distortion from non linearity of amplifiers and interference from other radio signals from narrow band systems. The received signal can be further separated to desired MPCs reflected from Tremorring Body Parts (TBPs), non desired MPCs from other reflectors in the medium, and the noise. We sample the received signal in (1) every period of T . Per each pulse, we reduce the observation period to N temporal samples that include only reflections from around the patient center body (torso) which can be obtained by any UWB tracking mechanism. The received signal for M consecutive pulses, are stored in an observation matrix r of size Nh . The column dimension of r represents the time dimension of pulse repetition. The row dimension represents delay, which is equivalent to the spatial dimension since for a given MPC, multiplying the MPC’s delay by the speed of light c gives us twice the distance the transmitted pulse propagated in space from the reflecting object to the acquisition system. B. Data Analysis Algorithm We divide the data analysis to two stages. First we extract from the received signal MPCs’ delays which relate to TBP’s displacements. Then we analyze the MPCs’ delays and obtain tremor characteristics. If we choose an observation period small enough so that the patient is stationary and the TBP’s displacements are around center location, the MPCs related to the TBP differ in time mainly by a weight time shift. This weight time shift is in a range that is determined by the tremor amplitude and the frequency of the change of the weight time shift in the observation period is determined by tremor frequency. A
s. t.
, A 2c
argmin , E ∑ A τ , 2Hz 2c
s |
(2)
τ |
12 z
Where r is an N length vector of the sampled received signal for pulse index m, 1≤m≤M, sl is a scalar representing the signal energy reflected from the l’th TBP surface for pulse index m, τ is the m’th weight time shift, w is an N length weight vector w shifted by τ , w n w n τ , τ is an M long vector that includes the weight time shifts, τ1, τ2 . . τM , A is the maximal clinical tremor displacement (in range of 1-4cm) and E · is the expectation operator in the observation period. The first constraint in (2) operates on observation matrix rows and limits the solution to the clinical tremor amplitude range, this is a spatial constraint. The second constraint operates on observation matrix columns and limits the solution during observation period to changes in tremor in the range of clinical frequencies, this is a temporal constraint. An MMSE optimal solution to (2) is based on match filtering of the received signal with the transmitted pulse shape and combining the result with an optimal MMSE weights. It can be shown that the constraints can be translated to ones that satisfy Karush–Kuhn–Tucker (KKT) conditions. A solution derived by methods of nonlinear programming (NLP) [9] is optimal. An optimal solution is cumbersome, requires nonlinear programming and unavailable statistics and is sensitive to distortion in pulse shape. A suboptimal solution to the problem, with no significant sacrifice in performance, is to apply to the matched filter outputs, the constraints in (2) one after another. A further efficient approximation uses instead of the MMSE weights, the Maximal Ratio Combining (MRC) weights, which combine the MPCs according to their Signal-to-Noise Ratios (SNRs), and is optimal if MPCs are well separated [10]. MRC has the advantage that it does not require the usually unavailable a-priory statistical information. Full tremor characteristics can be derived by the approximated weight time shifts vector τ τ , τ . . τ . A set of tremor frequencies and amplitudes is obtained by Fourier transform of τ. For a single dominant tremor frequency the estimated tremor frequency and amplitude are: |
| ,
|
|
(3)
Where c is the speed of light, and are the approximated tremor amplitude and frequency. More advanced pattern matching algorithms based of the spectrum of
IFMBE Proceedings Vol. 29
492
G. Blumrosen et al.
known tremor patterns pathologies over time can be applied in the future. In the common case of multiple TBPs we need to map the MPCs to the different TBPs. One way to map is according to the proximity of the MPCs where paths with similar delays are more likely to be related to the same TBP. This mapping is not accurate in a medium rich with scatterers that has no direct paths. Another way to map is according to MPCs pattern change in time. With a metal marker attached to the TBP of interest, the related MPCs amplitudes are enhanced and become more distinct then MPCs related to other TBPs.
processing unit was a common notebook computer (Lenovo T61) and the SW we used for processing was Matlab.
III. EXPIREMENTAL SETUP The experimental setup consisted of a UWB prototype system and an arm model to model a TBP of a patient. For modeling arm tremor we used a conduction coil, an AC generator source and a solid arm model with a small magnet attached. The AC generator induced periodic electrical current. The generator was wired to a transformer, which created a varying magnetic field in its core that induced a varying electromotive force. The force acted on a magnet attached to the solid arm model and generated periodic movement of the arm in the AC generator frequency. We attached a metal strip to the arm to magnify the UWB reflection. The UWB sensor node prototype consisted of a transmitter, a receiver, a processing unit and a storage unit. The transmitter was based on pulse generator (Picosecond Pulse Labs 4015D). The Pulse width Tp was 100ps, the pulse amplitude was 1.35V, the pulse repetition frequency was 85 Hz, and the bandwidth was 8GHz, similar to commercial UWB dongles. The pulse generator was connected to an omni-directional antenna (EM-6865 Elector Metrix) via an 30 dB amplifier (Herotek AF2 1828A). The transmission power was extremely low with peak power of 52mWatt/cm2 and average power of 105µWatt/cm2 measured at distance of 1 meter from the antenna. The UWB signal was received by an omni-directional antenna (EM-6865 Elector Metrix) with an amplifier (Herotek AF2 1828A) and then fed to the receiver. The receiver was based on an oscilloscope (Agilent DSO81304A) with sampling rate Ts of 20GS/s. The receiver was synchronized to the transmitter by a trigger from the pulse Generator. The raw data was sent to a storage unit. To enhance antenna gain and improve directionality, we added a metal cover structure over both transmit (Tx) and receive (Rx) antennas. We isolated the received a nd transmit antenna with a Carton board wrapped by aluminum foil to avoid a direct path. We used the segmented memory oscilloscope memory for storage. The
Fig. 1 UWB tremor acquisition system prototype and arm model. The arm model in the back was moving back and forth from and to the UWB acquisition system in a way that maximized the reflection surface The arm surface was moving back and forth from and toward the UWB acquisition system in a single axis. The UWB antennas were placed in an optimal orientation to capture the maximal reflection from the arm model. The arm model was placed in 1 to 2 meters from the acquisition system. Figure 1 shows the UWB tremor acquisition system prototype, the arm model can be seen in the back. We performed two sets of experiments. The first set was performed with different distances between the arm and the acquisition system. For each distance, the arm model trembled with a single frequency which varied from 3 to 12Hz. The second set of experiments was performed with different disturbers. We recorded for 20 seconds with relatively stationary channel conditions (arm fixed to one place, no change in environment).
IV. RESULTS For each experiment we followed the steps that is described in Section II-B. First we approximated the matched filtering to pulse shape by a simple peak detector. Then the constraints in (2) were applied one after another to the approximated matched filtered outputs. The approximated weight time shifts related to the arm (having similar frequency content) were combined with MRC weights. From the weight time shifts τ we estimated the tremor frequency and amplitude according to (3). Figure 2 shows the
IFMBE Proceedings Vol. 29
Non-contact UWB Radar Technology to Assess Tremor
estimated amplitude as a function of tremor frequency for distances between the arm and the acquisition system of 1, 1.5 and 2 meters. The approximated amplitude decreases with frequency. Near the frequency of 5Hz there is a peak which indicates the arm model resonance frequency. The amplitude estimations for the different distances are correlated with correlation factor of 0.97 and the average deviation from a video reference estimation was 0.1cm. The amplitude estimation at distance of 1.5meters was lower by factor of 2 than the other amplitude estimations. This difference is explained by the variation of tremor amplitude along the arm as the tremor’s amplitude near the body is lower. With a smaller marker surface, the amplitude variation of the TBP can be minimized and the amplitude estimation variance can be improved. The average tremor frequency estimation error was 0.01Hz. The accuracy achieved by the tremor frequency estimation is explained by the single tremor frequency present in our arm model (unlike the tremor amplitude that varied along the arm).
0.8
Tremor Amplitude(cm)
of less than 0.03Hz. Amplitude estimation error from a video reference was less than 1 mm.
V. CONCLUSIONS We suggested UWB technology to quantify tremor for diagnosis of different patient pathologies. We built a UWB tremor acquisition system prototype and provided data analysis tools for the acquisition system. A feasibility test that was performed showed accurate tremor frequency and amplitude estimations. This new technology can offer non contact tremor assessment, utilizing extremely low radiation that can penetrate walls, work in any light condition, and can collect accurate data continuously. This new technology can be utilized in the future to work at any home and transmit the collected data to a remote hospital for continuous tremor monitoring and analysis with minimum cost. Directional antennas, higher sampling rates, higher bandwidth, higher radiation power, smaller marker size and more advanced equalization techniques can all improve the system performance.
1 meter meas 1.5 meter meas 2 meter meas
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 2
493
4
6
8
10
12
Tremor freqency(Hz)
Fig. 2 Tremor
amplitude approximation for distances of 1, 1.5 and 2 meters between the acquisition system and the arm model. The tremor amplitude declines with tremor frequency but near 5 Hz, the resonance frequency, there is an increase in amplitude. The amplitude estimation at a distance of 1.5meters is lower by factor of 2 due to the variation of tremor amplitude along the arm surface
We verified system performance with different noise sources. We used the following noise sources: static metal reflectors, a wooden partition that separated the arm model from the acquisition device and a person with his hand covered with metal in the background. We tested the system at a distance of 1.5 meter with tremor frequency of 5 Hz. The system has shown tolerance to all noise sources. In all cases the frequency estimation was excellent with absolute error
REFERENCES 1. M. Ivan, “Electromyographic differentiation of tremors,” Clinical neurophysiology : official journal of the International Federation of Clinical Neurophysiology, vol. 112, no. 9, 2001, pp. 1626-1632. 2. J. Jankovic and J.D. Frost, Jr., “Quantitative assessment of parkinsonian and essential tremor: Clinical application of triaxial accelerometry,” Neurology, vol. 31, no. 10, 1981, pp. 1235-. 3. A. Salarian, et al., “Quantification of Tremor and Bradykinesia in Parkinson's Disease Using a Novel Ambulatory Monitoring System,” Biomedical Engineering, IEEE Transactions on, vol. 54, no. 2, 2007, pp. 313-322. 4. A. Jobbagy and G. Hamar, “PAM: passive marker-based analyzer to test patients with neural diseases,” Proc. Engineering in Medicine and Biology Society, 2004. IEMBS '04. 26th Annual International Conference of the IEEE, 2004, pp. 4751-4754. 5. S.S. Ram, et al., “Doppler-based detection and tracking of humans in indoor environments,” Journal of the Franklin Institute, vol. 345, no. 6, 2008, pp. 679-699. 6. C. Hornsteiner and J. Detlefsen, “Characterisation of human gait using a continuous-wave radar at 24 GHz,” Adv. Radio Sci., vol. 6, 2008, pp. 67-70. 7. C. Chang and A. Sahai, “Object tracking in a 2D UWB sensor network,” Proc. Signals, Systems and Computers, 2004. Conference Record of the Thirty-Eighth Asilomar Conference on, 2004, pp. 12521256 Vol.1251. 8. E.M. Staderini, “UWB radars in medicine,” Aerospace and Electronic Systems Magazine, IEEE, vol. 17, no. 1, 2002, pp. 13-18. 9. M. Avriel, Nonlinear Programming: Analysis and Methods., 2003. 10. J.G. Proakis, “Digital communications,” McGRAW-HILL INTERNATIONAL EDITIONS, New York. 3rd ed, 1995., pp. 780-782.
IFMBE Proceedings Vol. 29
High frequency mechanical vibrations stimulate the bone matrix formation in hBMSCs (human Bone Marrow Stromal Cells) D.Prè1,3, G.Ceccarelli2,3, M.G.Cusella De Angelis2,3 and G.Magenes1,3 1
Department of Computer and System Science, University of Pavia, Pavia, Italy 2 Department of Experimental Medicine, University of Pavia, Pavia, Italy 3 C.I.T., Tissue Engineering Centre, Pavia, Italy
Abstract— The aim of this work is to test the effects of a specific mechanical stimulation (Low Amplitude, high Frequency Vibrations) on the bone matrix formation of hBMSCs. Previous studies demonstrated that chemical culture conditions could influence the differentiation of hBMSCs toward bone: by plating the cells in appropriate osteogenic culture medium hBMSCs differentiate into osteoblasts [1-3]. In our experiment the cells were treated for 21 and 40 days by vibrating the wells for 45 minutes a day at a working frequency of 30 Hz. In order to separate the effects of the induction toward bone caused by the osteogenic culture medium and the effects of the high frequency vibrations, we divided the cells in four samples: in normal medium with or without mechanical treatment and in osteogenic medium with or without mechanical treatment. Afterwards, in order to measure the level of calcium deposition and consequently, the formation of bone matrix, the Alizarin Red assay was performed. The results express a strong increase in the deposition calcium for the extracellular matrix for the vibrated samples with respect to the non-mechanically treated ones for the 40 days treatment.
chemical factors, by adding the osteogenic improver in the culture medium and we tested the effects of low amplitude, high frequency vibration on differentiation of cells. In particular, the deposition of calcium has been investigated because it is the first step of extracellular bone matrix formation. Our previous studies demonstrated that high frequency vibrations increase the expression of many osteogenic proteins in SAOS-2 human osteoblasts [9]. Thus, our aim is to improve the differentiation of the hBMSCs and to reduce the time required to deposit mineralized bone matrix with the association of chemical and mechanical factors. In fact, the possibility to create a bone matrix is on one hand a strong indicator of differentiation of BMSCs toward bone, and on the other hand it makes possible to create prosthesis by removing the cellular part from the composite and leaving the only extracellular one: so the structure will be patient-independent, and it could be used as a general prosthesis.
Keywords— Bioreactor, differentiation, hBMSCs, Bone matrix.
I. INTRODUCTION
Bone tissue engineering offers innovative therapeutic opportunities for the repair of bone tissues damaged by diseases or injuries. Despite many advances in cell-based tissue engineering, significant challenges remain with respect to cell sourcing, expansion, and differentiation toward bone tissue for the application on patients. The identification of various adult stem cells retaining the capability to differentiate into multiple cell types has been a critical step in providing potential cell sources for bone tissue engineering. One of the first adult stem cell types investigated to reproduce osteoblasts were hBMSCs (human Bone Marrow Stromal Cells). Good results have been obtained for the differentiation of this cell line toward bone tissue [1, 4-7], but hBMSCs are difficult to harvest, very delicate and slow in their proliferation [8]. Subsequently, it’s useful to reduce the time to obtain osteoblasts from BMSCs. In order to accelerate their differentiation, we used
II. MATERIALS AND METHODS
A. Cell cultures We used BMSCs (bone marrow stromal cells) taken from a male young patient at the 3rd passage. Iliac crest bone marrow aspirates were collected. The bone marrow sample was centrifuged at 500 g for 10 min and about 25 ml of plasma was collected. Plasma obtained by this low speed centrifugation contained the platelet fraction. Buffy coat (pellet) was resuspended in an equal volume of DMEM and the resulting suspension was centrifuged on a Ficoll separating solution for 20 min at 1200 × g. The mononucleated cells were recovered from the interphase and counted on a hemocytometer. They were then plated in flasks of 75 cm2 surface containing the proliferative medium and incubated at 37°C in a humidified atmosphere (95% air, 5% CO2), till reaching the confluence.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 494–497, 2010. www.springerlink.com
High Frequency Mechanical Vibrations Stimulate the Bone Matrix Formation in hBMSCs (Human Bone Marrow Stromal Cells)
B. The bioreactor
495
E. Cell counting
A device to produce the vibrating stimuli to in vitro cells cultures was used. Its detailed description has already been reported in [10]. The system is composed by an eccentric motor to produce the displacement (Maxon Motor™ Brushless E-Series) with a Voltage of 24V and a diameter of 22 mm. The imposed displacement is thus a half of the diameter (11 mm). The angular velocity (i.e. vibration frequency) of the motor is voltage-controlled and can be modified through an electronic controller. By changing the Voltage supply from 2.5 up to 24 V the vibration frequency varies between 1 Hz and 120 Hz. A 3D accelerometer was added to detect the acceleration forced on the platform by the motor. C. Bioreactor Cultures The cells were divided in four groups. The first one with proliferative medium (PM) and subjected to the mechanical treatment of high frequency vibration (TP), the second one with the same medium but without any mechanical treatment (CP), the third one with osteogenic medium (OM) and subjected to high frequency vibration (TO) and the last one with OM but without any mechanical treatment (CO). The treated samples were subjected to high frequency vibration treatment at 30 Hz for 45 minutes every day. The four groups of cells were stopped and subjected to the tests after 21 and 40 days. The media were replaced every 6 days. All cells were plated in 9 cm diameter dishes (so, with an Area of 28.27 cm2 each one) at a density of 5000 cells/cm2. After the end of the experiment the Alizarin Red test were performed and the results were normalized on the cell number. D. Alizarin Red Test Alizarin red test was used as biochemical assay to quantitatively determine by colorimetry the presence of calcific deposition of an osteogenic lineage. It is an early stage marker of matrix mineralization, a crucial step towards the formation of calcified extracellular matrix associated with true bone. The cells were stained with pH-adjusted (4.1–4.3) 2% Alizarin Red solution (Electron Microscopy Sciences, Fort Washington, PA), washed, and then photographed using transmitted light. The stain was eluted by adding 1 ml of 10% cetylpyridinium chloride per well for 10 min at room temperature in gentle agitation. Afterwards the intensity of the color, proportional to the calcium deposition, was measured with the Nanodrop™ (Nanodrop Technologies, Wilmington, USA) at a wavelength of 562 nm. The level of Alizarin was normalized to the number of cells.
The number of cells to normalize the alizarin red values was evaluated by counting cells by Burker’s chamber. To follow the evolution of the number of cells during the experiment this analysis was performed after 0, 21 and 40 days of treatment. Cells were detached by using tripsin 1X and resuspended in an appropriate volume of PBS before counting. F. Statistical Analysis The experiments were repeated three times in order to obtain a better statistical confidence. To evaluate the effects on the vibrated cells with respect to the controls, a two-way analysis of variance (ANOVA) was performed for the XTT test and Molecular Biology tests. We used the statistical tool of Matlab 7.1. Statistical significance was declared at p≤0.05, in which p is the null hypothesis.
III.
RESULTS
The results of Alizarin Red test on 21 days cultures are shown in Figure 1 for the proliferative medium (1A) and osteogenic medium (1B). In the proliferative medium the value of the concentration of Calcium deposition in the control sample is equal to the treated one (20 pg/(ml*cell)), and in differentiative medium, the concentration is slightly higher for the control sample (48.75 pg/(ml*cell) with respect to 46.25 pg/(ml*cell) of the treated), even if it is not statistically significant (p>0.05). The graphics that summarize the results obtained with the Alizarin Red Test after 40 days of treatment are expressed in Figure 2.
Figure 1: Graphic of Alizarin Red level at 21 days for treated and controls in both the culture media (in proliferative medium-A- and in differentiative medium -B). The vertical blue lines represent the standard deviations on data.
IFMBE Proceedings Vol. 29
496
D. Prè et al.
The red colouring represents a high concentration of calcium deposition. By analyzing the samples at optical microscopy, it’s possible to observe the difference in red-coloured calcium deposition between treated and control samples, as reported in Figure 4. IV. DISCUSSION
Figure 2: Alizarin Red Test at 40 days for proliferative (A) and osteogenic (B) medium on hBMSCs. The p-values are reported on the graphics and the vertical blue lines represent the standard deviation on data.
The results show a strong increase in the deposition of extracellular matrix for the mechanically treated samples with respect to the non-treated ones. In fact, the normalized level of Alizarin red in the treated samples in proliferative medium is 27% higher with respect to the non-mechanically treated sample in the same medium (Figure 2A). The difference of the Alizarin Red Level for the samples kept in osteogenic medium is even more significant (Figure 2B). In fact, the normalized level of alizarin red (and, consequently, of the calcium deposition for the extracellular matrix) of the mechanically treated samples after 40 days is 3120 pg/(ml*cell), compared with a value of 1610 pg/(ml*cell) for the control samples. Thus, the mechanical treatment increases the deposition of Calcium (i.e. the production of extracellular matrix) of 93%. All the p-values are less that 0.001 As it is possible to observe from Figure 3 the difference between the mechanically treated and the control samples in osteogenic medium is clearly visible.
Previous experiments [9] have demonstrated an inductive effect of high frequency vibrations on the differentiation toward bone on SAOS-2 cell line. However, SAOS-2 is a tumoral cell line and it cannot be considered for future clinical application. Consequently, we decided to test the effect of the same stimulation on normal human stem cell lines: the BMSC line has been the first choice, because of their demonstrated capability to differentiate into bone. We changed the duration of the treatment as a consequence of the different proliferation rate between SAOS-2 and BMSCs, following the mathematical modeling proposed in [8]. Since the inductive properties of the osteogenic medium have already been demonstrated, we studied separately the effects of the mechanical treatment by creating four groups of samples: in this way, we can highlight the effect of the high frequency vibration by itself on differentiation of cells. One of the first parameters to understand the differentiation level is the amount of calcium deposition, by the alizarin red test. Our results demonstrate the strong differentiative effect of the treatment, in particular if it is associated with the inductive effect of the osteogenic medium. By observing the results on the deposition of calcium at 21 days, it is possible to conclude that the high frequency vibration treatment does not have any effect on this parameter for both culture media. On the contrary, the effect of osteogenic medium on the calcium deposition at 21 days is evident. However, 21 days is a period of early differentiation for this cell line, so probably the cells are not yet in the phase of mineralization. In fact, to understand the effects of the mechanical treatment on the deposition of calcium, it is important to evaluate the difference between treated and control samples in the mineralization phase, after 40 days. The results show the positive effects of the combination between the osteogenic medium and the mechanical treatment on BMSCs, and it is the first strong signal of the differentiative effect of the treatment.
Figure 3: Picture of the plates of BMSCs after 40 days of culture in differentiative medium for a control sample (C, on the left) and a mechanically treated one (T45, on the right).
IFMBE Proceedings Vol. 29
High Frequency Mechanical Vibrations Stimulate the Bone Matrix Formation in hBMSCs (Human Bone Marrow Stromal Cells)
497
VI. REFERENCES
1.
Figure 4: Pictures with light microscopy of the plates of BMSCs in differentiative medium after 40 days. (A) Control sample at a magnification of 20x. (B) Treated sample at a magnification of 20x. (C) Control sample at a magnification of 10x. (D) Treated sample at a magnification of 10x.
Further analyses on the osteogenic genes and on the following translation in proteins should confirm these preliminary results. Afterwards, it will be possible to affirm the efficacy of the high frequency vibration treatment on in vitro cell culture.
Cheng, S.L., et al., Differentiation of human bone marrow osteogenic stromal cells in vitro: induction of the osteoblast phenotype by dexamethasone. Endocrinology, 1994. 134(1): p. 277-86. 2. Anselme, K., et al., In vitro control of human bone marrow stromal cells for bone tissue engineering. Tissue Eng, 2002. 8(6): p. 941-53. 3. Agata, H., et al., Feasibility and efficacy of bone tissue engineering using human bone marrow stromal cells cultivated in serumfree conditions. Biochem Biophys Res Commun, 2009. 382(2): p. 353-8. 4. Ashman, R.B., et al., A continuous wave technique for the measurement of the elastic properties of cortical bone. J Biomech, 1984. 17(5): p. 349-61. 5. Gomes, M.E., et al., Influence of the porosity of starch-based fiber mesh scaffolds on the proliferation and osteogenic differentiation of bone marrow stromal cells cultured in a flow perfusion bioreactor. Tissue Eng, 2006. 12(4): p. 801-809. 6. Jiang, Y., et al., Pluripotency of mesenchymal stem cells derived from adult marrow. Nature, 2002. 418(6893): p. 41-9. 7. Marolt, D., et al., Bone and cartilage tissue constructs grown using human bone marrow stromal cells, silk scaffolds and rotating bioreactors. Biomaterials, 2006. 27(36): p. 6138-49. 8. Prè, D., Ceccarelli, G., Benedetti, L., Cusella De Angelis, M.G., Magenes, G. . A comparison between the proliferation rate of SAOS-2 human osteoblasts and BMSCs (Bone Marrow Stromal Cells) using mathematical models. in World Congress 2009Medical Physics and Biomedical Engineering. 2009. Munich, Germany. 9. Pre, D., et al., Effects of Low Amplitude, High Frequency Vibrations on Proliferation and Differentiation of SAOS-2 Human Osteogenic cell line. Tissue Eng Part C.Methods, 2009. 10. Pre, D., et al. A high frequency vibration system to stimulate cells in bone tissue engineering. 2008. Shanghai, China.
V. CONCLUSIONS
The results obtained suggest a differentiative effect of the mechanical treatment on hBMSCs: in particular, the association with the osteogenic medium increases the efficacy of the treatment. The following steps require the association of the cell culture with a scaffold: after that, an in vivo study on immunodeficient mice and finally on humans could be performed. At the end, it will be possible Co to treat the hBMSCs of the patient with the HFV treatment to induce their differentiation toward bone in order to build autologus prosthesis.
Corresponding Author: Author: PRÉ DEBORAH Institute: UNIVERSITY OF PAVIA Street: VIA FERRATA 1, 27100 City: PAVIA (ITALY) e-mail address:
[email protected] IFMBE Proceedings Vol. 29
Mobispiro: A Novel Spirometer Eleni J. Sakka, Pantelis Aggelidis, and Markela Psimarnou Ρ.Α. Mobihealth LTD, Nicosia, Cyprus
Abstract— Respiratory function monitoring and pulmonary diseases evaluation usually involve patient examination with spirometers, for the recording and monitoring of two principal parameters: the total air volume that the lungs can inhale and exhale, and the peek of the exhaled air flow. The purpose of implementing a portable spirometer device is to be used by patients who suffer from chronic pulmonary diseases, and need frequently to record and monitor their respiratory function. Mobispiro is a novel spirometer that enables recording and transmission of vital signs over GSM/GPRS networks and supports patient-centric models of healthcare provision. The core of the device includes all the essential hardware components, as well as software providing the necessary functionality and it is expected to fulfill all expert needs for monitoring patient’s disease. Mobispiro measurements are transferred to the doctor through GSM network. In addition, the use of the GPRS service for data transmission supports the ability for the device to communicate with receiving points, e.g. PC, PDA or a cell phone. It is anticipated that the volume of data to be transferred is relatively low and thus, the above connectivity solutions are satisfactory for this kind of medical use. The device is programmed to send data to predefined access points. The solution is integrated with a web server application which gives the ability to access data simply by using an internet connection. The web server application is responsible for the service administration, the retrieval of measurements from the database and their presentation. Mobispiro is also provided as an OEM solution and complies with the international standard for medical data transmission and storage and the ATS recommendations for diagnostic spiromenters. Keywords— Spirometry, GSM/GPRS.
I. INTRODUCTION Pulmonary diseases such as chronic obstructive pulmonary disease (COPD) and asthma can be evaluated with the help of spirometry. In this examination the monitoring of the respiratory function involves the recording of two principal parameters: the total air volume that the lungs can inhale and exhale, and the peek of the exhaled air flow. Chronic obstructive pulmonary disease (COPD) is a lung ailment that is characterized by a persistent blockage of airflow from the lungs. It is an under-diagnosed, life-threatening lung disease that interferes with normal breathing and is not fully reversible. Chronic obstructive pulmonary disease is more than a “smoker’s cough”. An
estimated 210 million people have COPD worldwide, whereas more than 3 million people died of COPD in 2005, which is equal to 5% of all deaths globally that year. According to WHO, almost 90% of COPD deaths occur in low- and middle-income countries. The primary cause of COPD is tobacco smoke, either through tobacco use or second-hand smoke. The disease now affects men and women almost equally, due in part to increased tobacco use among women in high-income countries. COPD is not curable, but treatment can slow the progress of the disease. Total deaths from COPD are projected to increase by more than 30% in the next 10 years without interventions to cut risks, particularly exposure to tobacco smoke. Asthma is a chronic disease characterized by recurrent attacks of breathlessness and wheezing, which vary in severity and frequency from person to person. Symptoms may occur several times in a day or week in affected individuals, and for some people become worse during physical activity or at night. During an asthma attack, the lining of the bronchial tubes swell, causing the airways to narrow and reducing the flow of air into and out of the lungs. Recurrent asthma symptoms frequently cause sleeplessness, daytime fatigue, reduced activity levels and school and work absenteeism. Asthma has a relatively low fatality rate compared to other chronic diseases. WHO estimates that 300 million people currently suffer from asthma. Asthma is the most common chronic disease among children. It is a public health problem not just for high-income countries; it occurs in all countries regardless of the level of development. Most asthma-related deaths occur in low- and lower-middle income countries. Often asthma is under-diagnosed and under-treated. It creates substantial burden to individuals and families and often restricts individuals’ activities for a lifetime. For the monitoring of both COPD and asthma, spirometry is an imperative solution. Most of the spirometers that are available in market are designed to record the air flow data and show the measurements either on a display or send it to the doctor’s PC via cable. Contrarily to them, Mobispiro has embedded communication capabilities that enable it to communicate measurements to remote destinations. At least two use case scenarios show the potential added value of Mobispiro. In the first, the device is operated by a
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 498–501, 2010. www.springerlink.com
Mobispiro: A Novel Spirometer
499
GP, and the mobile unit is programmed to send a) data packets to an FTP server which collects the measurements, and b) notification SMS to a predefined medical expert’s cell phone, who is getting informed by this way for the transmission. In the second use case scenario Mobispiro is operated by the patient and the communication module is programmed to send a) data packets to the medical expert’s email address and b) SMS with specific measurement parameters to his cell phone. The measurements’ receiving node can be either a PC or an advanced cell phone or PDA, where the control, advanced processing and print of data are possible, through a handy application.
B. Implementation The implementation of the spirometer was based on the standards of the American Thoracic Society (ATS) for diagnostic spirometers [2]. According to these standards these medical devices should comply with the values presented on table 1. Table 1 Minimal recommendations for diagnostic spirometry Test VC
II. MOBISPIRO A. Mobispiro Spirometer Overview
FVC
Mobispiro is a high quality spirometer that facilitates patient-centric, continuous monitoring services of healthcare provision models. Mobispiro can be used to perform a comprehensive set of spirometry measurements (e.g. FEV1, FVC, FEF 25-75%, etc). Measurements are locally stored on the device's memory and also presented on the device's screen. The device has diverse communication capabilities, both wired (e.g. Ethernet) and wireless (GSM/GPRS). Mobispiro is a patented spirometer that is designed and developed to comply with international standards. It is a portable medical device targeting chronic patients with pulmonary diseases and easy to use. Its user friendliness makes it suitable even for users with limited possibilities or special needs.
FEV1
Time zero
PEF
FEF2575%
V
MVV
Fig. 1
Mobispiro prototype
Range/Accuracy (STPS) 0.5 to 8 L ±3% of reading or ±0.050 L whichever is greater 0.5 to 8 L ±3% of reading or ±0.050 L whichever is greater
Flow Time Range (s) (L/s) zero to 30 14 zero to 15 14
0.5 to 8 L ±3% of reading or ±0.050 L whichever is greater
zero to 1 14
The time point from which all FEVt measurements are taken Accuracy: ±10% of reading or ±0.400 Us, whichever is greater Precision: ±5% of reading or ±0.200 Us, whichever is greater 7.0 Us ±5% of reading or ± 0.200 Us, whichever is greater ±14 Us ±5% of reading or ±0.200 Us, whichever is greater 250 Umin at TV of 2 L within ±10% of reading or ±15 Umin, whichever is greater
zero to 14
Resistance and Back Pressure
Test Signal 3-L Cal Syringe
Less than 1.5 cm H20/L/s Less than 1.5 cm H2O/L/s
24 standard Waveforms 3-L Cal Syringe 24 standard Waveforms
Back extrapolation
± 14 Same as FEV1
zero to 15 14 ± 14
Same as FEV1
15 Same as FEV1
± 3%
12 to 15
26 flow standard waveforms 24 standard waveforms Proof from manufacturer
Pressure less than Sine ±10 cm wave H2O at 2-L pump TV at 2.0 Hz
The basic components of Mobispiro are a microcontroller, the communication add-in, the airflow sensor the digital display and the algorithms for the measurements processing. For the implementation of the Mobispiro the Wavecom [3] Wireless CPU Q2686 microcontroller was used as the main microcontroller of the device. This microcontroller has a USB 2.0 and two UART outputs. Digital control can be IFMBE Proceedings Vol. 29
500
E.J. Sakka, P. Aggelidis, and M. Psimarnou
obtained through two INT, two SPI, an I2C and a 5 x 5 keyboard. The advanced feature of this microcontroller is that it supports an embedded GSM/GPRS communication protocol covering GSM bands worldwide i.e. 800/900/1800/1900 MHz. With the appropriate libraries installed the microcontroller gives the opportunity to connect, register to the cellular network and have all the available functionality of the GSM networks. The microcontroller was programmed on the Deployment Kit Q26 of Wavecom, whereas for the source code the Microsoft Visual Studio .NET 2003 [4] and the development platform Open AT v1.08.02 were used. The airflow sensor is based on MEMS technology and was developed by THEON Sensors [5]. It is developed to comply with the ATS standards. It is a brand new sensor developed for Mobispiro spirometer and the operating principle is the hot film anemometer. Basically, it is an ohmic sensor which is warmed by an electric signal and exposed to the airflow. The sensor has a digital UART output so as for the measurement process to be more efficient and easier.
• •
•
•
• •
VC (vital capacity), that is the maximum volume of air which can be exhaled or inspired during either a forced (FVC) or a slow (VC) manoeuvre, FEV1 – (Forced Expiratory Volume in 1 Second), that is the volume expired in the first second of maximal expiration after a maximal inspiration and is a useful measure of how quickly full lungs can be emptied. FEV3 – (Forced Expiratory Volume in 3 Seconds), that is the volume expired in the first three seconds of maximal expiration after a maximal inspiration and is a useful measure of how quickly full lungs can be emptied. FEV1/FVC (FEV1%), that is the FEV1 expressed as a percentage of the VC or FVC (whichever volume is larger) and gives a clinically useful index of airflow limitation. PEF (Peak Expiratory Flow), that is the maximal expiratory flow rate achieved and occurs very early in the forced expiratory manoeuvre. and FEF 25-75% or 25-50% (Forced Expiratory Flow 2575% or 25-50%) that is the average expired flow over the middle half of the FVC manoeuvre and is regarded as a more sensitive measure of small airways narrowing than FEV1.
In order to evaluate the device special tests were made by pneumonologists. In these tests the reliability of the device was verified. During these tests several patients ware asked to take a spirometry test before and after a bronchodilator. These measurements were compared to measurements taken by spirometers available in market. C. Mobispiro Service
Fig. 2 MEMS Air flow Sensor The LCD display that was used is Crystalfontz [6] CFA635-TFE-KU1. The display is connected through a UART port to the microcontroller and its overall size is 142x37 mm, having 82.95x27.5 mm displaying surface. It has, also, four LED lights and six buttons. The last component of the Mobispiro spirometer is the software that implements sensor’s data handling, data processing, display of the measurements on the LCD, communication with the base station and, finally, transmission of data to the server. The Mobispiro software was implemented on Microsoft .NET Visual Studio 2003. The algorithms for data processing and calculations for the pulmonary function evaluation are based on the sum of squares formula. The following parameters are being calculated and send to the Mobispiro service centre:
Mobispiro spirometer can also be used as part of an integrated patient telemonitoring service that includes a centralized database and a web based application for the presentation of the measurements. The Mobispiro Service is user-oriented, aiming at the collection and transmission of pulmonary function parameters with minimum user intervention. The user only needs to blow to the mouthpiece of the medical device; then, by pressing a button the measurement is transmitted to a web server, for centralized storage. A web based application allows user friendly retrieval, visualization and process of the medical data by specialized healthcare professionals. For each user/patient a comprehensive Electronic Health Record can be created, including the patient demographics and her medical profile. The distinctive diagnostic value of the Mobispiro Service lies on the fact that medical data recording is feasible at the time when the patient feels pain, discomfort or any other symptom, not necessarily present when and while being in a hospital.
IFMBE Proceedings Vol. 29
Mobispiro: A Novel Spirometer
501
for diagnostic spiromenters and enables recording and transmission of vital signs over GSM/GPRS networks and supports patient-centric, continuous monitoring models of healthcare provision. Mobispiro is a prototype and the future steps include the amelioration of user friendliness of the device as well as the improvement of the software bugs that were revealed during the verification tests.
ACKNOWLEDGMENT Mobispiro was implemented during the incubation period at Diogenes Business Incubator [7] University of Cyprus Ltd, in Nicosia, Cyprus.
Fig. 3 The MobiSpiro service Overall, Mobispiro enables the provision of costeffective telemonitoring services to chronic patients with pulmonary diseases (e.g. asthma, COPD, etc). The novel spirometer along with the web-based application constitute a unique m-health solution, bringing the point of care closer to the patient, and enabling personalized treatment plans and efficient health monitoring.
III. CONCLUSIONS The monitoring and the examination of patients suffering from pulmonary diseases usually involve examination with spirometers, for the recording of two principal parameters: the total air volume that the lungs can inhale and exhale, and the peek of the exhaled air flow. Mobispiro is a novel spirometer that complies with the ATS recommendations
REFERENCES 1. 2. 3. 4.
WHO Chronic obstructive pulmonary disease (COPD) at www.who.int American Thoracic Society, Standardization of Spirometry, 1994 Wavecom at http://www.wavecom.com Microsoft .NET Visual Studio at http://msdn.microsoft.com/en-us/vstudio/default.aspx 5. Theon Sensors at http://www.theon.com/en/homepage_en.php 6. Crystalfrontz at http://www.crystalfontz.com/ 7. Diogenes Business Incubator at http://www.diogenes.com.cy/ Corresponding author: Author: Institute: Street: City: Country: Email:
IFMBE Proceedings Vol. 29
Eleni J. Sakka Ρ.Α. Mobihealth LTD 91 Aglandjia Avenue 1678, Nicosia Cyprus
[email protected].
A computer program for the functional assessment of the rotational vestibulo-ocular reflex (VOR) A. B¨ohler1 , M. Mandal´a2 and S. Ramat3 1
Medical Device Technology, Upper Austria University for Applied Sciences, Linz, Austria. 2 Dipartimento di Scienze Anestesiologiche e Chirurgiche, University of Verona, Italy. 3 Dipartimento di Informatica e Sistemistica, University of Pavia, Pavia, Italy.
Abstract— The vestibulo-ocular reflex (VOR) uses head angular acceleration information transduced by the semicircular canals in the inner ear in order to drive eye movements that compensate for head rotations, and thus stabilize the visual scene on the retina. Peripheral and central vestibular pathologies may impair the function of the VOR so that compensation becomes incomplete, making clear vision during head movement impossible. The clinical assessment of vestibular function is made difficult by the adaptive processes activated by the central nervous system of the patient, which quickly learns to use residual vestibular information or information provided through other senses to supplement the deficient VOR, especially for slow head movements. Clinical assessment may still be made using the head impulse test. A compensatory saccade at the end of the head movement is the clinical sign of a vestibular deficit. Here we propose a new computerized technique for assessing vestibular function at different head angular accelerations, based on evaluating the ability of the patient in reading a character briefly displayed on a computer screen while the head is being rotated. Keywords— inertial sensor, vestibular system, clinical testing, VOR
I. I NTRODUCTION The vestibular system contributes to the control of balance by transducing head accelerations and triggering reflex responses aimed at stabilizing gaze and body position in space. The VOR is responsible for the stabilization of gaze, and without an effective VOR vision would be impaired every time the head moves. The life-long prevalence of dizziness is about 30% in the general population and this figure is even larger in the elderly since aging affects the function of the vestibular system. The assessment of dizziness needs to evaluate the vestibular system, yet such evaluation may prove difficult since the central nervous system tries to compensate vestibular dysfunction by adaptation and substitution processes. Such mechanisms are especially efficient in response to low acceleration rotational stimuli, yet most diagnostic tools also investigate movements
in the same stimulus range. The head thrust test [1] is generally accepted as the clinical test of reference for high acceleration stimuli: the patient is asked to fix upon a target (typically the examiner’s nose) while the examiner briskly rotates the head. A normally working VOR will hold gaze steady, otherwise a corrective saccade will be needed at the end of the head movement to bring the image of the target back to the fovea. Such saccade represents the clinical sign of dysfunction of the semicircular canal towards which the head has been rotated. The detection of the corrective saccade may be difficult at the bedside and performing the test therefore requires an experienced clinician. In the few laboratories equipped with a magnetic search coil system it is possible to implement a quantitative, yet invasive, version of the test by simultaneously recording both the eye and the head movement [2], but such approach clearly cannot be of widespread use. Previous work has suggested a measure of vestibular function in terms of dynamic visual acuity, i.e. the assessment of the visual acuity of a subject during head movement [3, 4, 5]. With such testing technique the head is either actively [6] or passively [5] rotated and its angular velocity is recorded so that an optotype is presented on a computer display when a fixed threshold velocity is exceeded. Current studies aim at assessing the dynamic visual acuity so that the resulting measurement is expressed in terms of the logarithm of the minimum angle resolvable (logMAR), which typically results in a lower visual acuity for dynamic vs. static conditions. The DVA may be used for diagnostic assessment either by comparing the decrease in visual acuity with a normative database or by comparing the performance in one direction of rotation with that in the other. Predictable head motion, such as self generated or sinusoidal head rotations, has been shown to cause low sensitivity of DVA as a diagnostic tool [4, 7]. Our approach differs from that of DVA assessment as we want to test vestibular function in order to gain information on which parameters of the head movement, and not those of the visual stimulus, affect a subjects ability to stabilize gaze in space. With such information we will be able to improve the sensitivity and the specificity of the test while understand-
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 502–505, 2010. www.springerlink.com
A Computer Program for the Functional Assessment of the Rotational Vestibulo-Ocular Reflex (VOR)
ing which natural activities may be impairing vision for each patient and should therefore be considered as potentially dangerous for himself and the people around him. In the following we describe a software program implementing a new technique for the assessment of vestibular function at different head angular accelerations. The subject’s ability to stabilize gaze during head movement is challenged only by the intensity of head angular acceleration. The size of the optotype being displayed is normalized to the individual visual acuity and enlarged enough for readability not to be an issue.
503
soon as the specified number of data points are captured. In order to reduce the effect of sampling noise, raw data is captured at 10000Hz, but then smoothed to 100Hz (100 samples are captured at 100Hz and averaged to provide one angular velocity or linear acceleration data point).
II. M ETHOD Based on the head impulse test rationale, we have developed a Python software program for acquiring angular velocity and linear accelerations from a head mounted sensor and displaying an optotype on a computer monitor when the imposed head angular acceleration exceeds a user-defined threshold. The software simultaneously verifies that the head rotational stimulus is correctly delivered. The sensing device consists of two parts: the sensor attached to the patient’s head and the acquisition system, attached to a computer.
Fig. 1: Flowchart
The inertial sensor is made up of a gyroscope (ADXRS300) and a 3-axis accelerometer (ADXL330), both manufactured by Analog Devices. The gyroscope gives accurate information about the rotation of the head within a range of ±300◦ /s, whereas the accelerometer allows the verification of the movement within a range of ±3g. The sensors are packed together on a 2cm by 2cm circuit board, weighing only a few grams. This assembly is then mounted with an elastic band to the subject’s head, thus allowing natural head movements.
The CalcThread does the acutal data processing: It first converts Volts to ◦ /s and then calculates the derivative (acceleration). When the threshold is reached, the DisplayThread is notified about the change and asked to display the optotype on screen. The CalcThread continues to capture data for one second and then dispatches the dataset to the EvalThread. The EvalThread in turn does an offline analysis of the dataset: it verifies the accelerometer data to assess the correctness of the stimulus (see section Head Movement Criteria) and assigns the trial to the correct head acceleration bin, based on the maximum angular acceleration reached. The DisplayThread itself is responsible for the presentation of the letter, after a user-selectable delay the letter and for a user-modifiable number of video frames. After presenting the letter the DisplayThread signals the GUI to ask the user for the letter displayed.
B. Data Acquisition and Data Processing
C. Head Movement Criteria
The analog outputs of the sensors are captured by data acquisition hardware from National Instruments. We successfully tested our Software on various USB-based cards, even some bus-powered ones, allowing the portable use of the system. We developed a simple wrapper around the C-library provided by NI using Python’s ctypes library. The code is based on sample code from the scipy.org Cookbook as well as NI’s C example code. The entire system is GUI-driven with a GTK+ interface designed using Glade-3. The Flowchart in Figure 1 shows how the communication with the hardware runs in its own thread (ListenThread) and calls a function in the calculation thread (CalcThread), as
In order to ensure that the test is correctly performed, our software looks for characteristics of the imposed head movement which may undermine the reliability of the test, and excludes trials presenting these patterns from those considered in the assessment of vestibular functionality. Two main concerns are addressed: 1- the variability of the axis of head rotation and 2- the presence of a translational movement of the head. In order to correctly stimulate one vestibular canal at a time, head rotation needs to occur in the plane of that canal. Horizontal semicircular canals, for instance, are tilted about 20◦ up with respect to the gravitational horizontal. Therefore,
A. Sensor System
IFMBE Proceedings Vol. 29
504
A. Böhler, M. Mandalá, and S. Ramat
in order to properly stimulate that canal pair with a horizontal head rotation, the head needs to be tilted nose-down. As a general rule of thumb the head should be pitched forward by an angle between 20 and 30 ◦ . Once the subject is being positioned accordingly, our software ensures that during the rotational impulse the orientation of the head is not varied beyond a predefined tolerance (e.g. 5 deg). When the threshold tolerance is exceeded the trial is discarded and a message detailing the error type is displayed. Second, the software monitors the presence of translational components in the delivered stimulus, a condition that may frequently occur when the experimenter attempts to deliver higher acceleration stimuli. The compensation of a head translation requires the intervention of the translational VOR (tVOR), which depends on the otolith organs, and whose performance is known to be less than compensatory even in normal subjects [8]. Therefore, introducing a translational component in the stimulus would reduce the subjects possibilities of reading the displayed optotype and make the results of the test unreliable. To avoid such biasing factor, our software verifies that the linear acceleration relative to the plane in which the head rotation occurs does not exceed that expected by the amount of instantaneous head angular velocity. Ideally our head movement sensor should be positioned on the axis of head rotation, so that the only acceleration sensed by the three axis accelerometer will be that of gravity in both static and dynamic conditions. If the sensor is displaced with respect to the axis of head rotation by a distance r, the accelerometer will instead pick up two components of head acceleration: a centripetal (directed radially, ar ) and a tangential acceleration (at ), as shown in Eq. 1. ar = −ω 2 · r at = r ·
dω dt
III. P ERFORMANCE & V ERIFICATION A. Timing The use of the Python language has both advantages and disadvantages with respect to performance. Although Python is a very efficient language, it is still interpreted. Normally timing is a problematic issue, but fortunately the acquisition hardware by National Instruments offers hardware-timed sampling, allowing to avoid software-based timers. B. Verification In order to verify the timing of the system, a crucial performance for the accuracy of the test, we attached a photodiode to the screen that detects a white square that pops up together with the letter. The output of the photodiode is also captured via the data acquisition system and fed to the analysis software. The software can save the raw dataset in an HDF5 file for later analysis. The photodiode itself is a generic RS-Components diode, BPW21, RS part no. 303-719. It has a typical rise-time of 1μs, which was therefore neglected.
(1) (2)
Prior to beginning the test we therefore perform some example head rotations while monitoring head angular acceleration and we reposition the sensor in order to minimize the linear accelerations transduced by the accelerometer along the radial and tangential directions. This implies reducing the distance r in Eq. 1, thus improving the positioning of the sensor so that it is closer to the axis of head rotation. During the test of individual semicircular canal function, we can then use the angular velocity and acceleration data acquired through the gyroscope to verify that the radial and tangential accelerations match those expected based on Eq. 1. Trials with lateral head accelerations greater than an adaptive threshold (Eq. 2) will be rejected. max(4, 1.1 · αT /1000)
where αT is the angular acceleration threshold in ◦ /s2 .
Fig. 2: Raw dataset and timing verification To verify the timing of the visual stimulus we developed a simple Matlab software measuring real-world delays based on the known sampling rate. A typical raw dataset is shown in Figure 2. The threshold was set at 2000◦ /s2 (marked with a circle), the diode threshold was about 3.5V (marked with an asterisk). The delay, that is the time from overcoming the threshold until the diode detects the optotype, is 34ms. The
(3) IFMBE Proceedings Vol. 29
A Computer Program for the Functional Assessment of the Rotational Vestibulo-Ocular Reflex (VOR)
505
time on screen, that is the time the diode’s value is high, is 16ms corresponding to one frame at 60Hz. The artificial delays before displaying the letter and the display time of the letter are controlled via a call to the sleep() function. The accuracy of this approach was similarly verified. Figure 2 also shows how, with the current prototype and test setup, the letter is displayed around the time when maximum head acceleration is reached. C. Operating System Influence
Fig. 3: Left: Healthy subject; Right: Patient with deficits
Our tests showed severe differences between Windows XP and Windows Seven. On Windows XP, the optotype display is about 25ms faster than on Windows Seven (on the same hardware), see Table 1 for details. The first test was done using Windows Seven 64bit (but 32bit Python!), the second one using Windows XP Professional Tablet PC Edition 2005, 32bit. The program was set with no artificial delay and a character display time of 16ms. W7 Delay 65.7
W7 Disp.Time XP Delay 32.7 41.5 Table 1: Timing Comparison
XP Disp.Time 34.7
IV. R EPORTS AND R ESULTS The analysis software currently supports four different modes of representing results. The “Detailed Results” page lists every trial, the letter it asked for, the users’s response and direction as well as rate and (max) acceleration. For better readability correct results are colored green, false answers are colored red. The “Bins” page puts all results within a range of e.g. 1000◦ /s2 into one bin and lists the percentage and number of correct answers in every direction. The “Vestibulogram” graphically shows the percentage of correct answers per bin with respect to the rotation direction. Finally, the “Error-Plot” represents the number of errors per bin, also divided into clockwise and counterclockwise. Figure 3 compares the results of two recordings. While the left-hand picture is from a healthy subject, the picture on the right are answers from a patient with bilateral deficits. As expected, a normal subject can read nearly 100% of the presented letters while the vestibular patient has severe problems identifying the letter even at lower accelerations and his performance drops dramatically at higher accelerations.
V. D ISCUSSION We have developed a software program for performing a test of the functionality of the individual semicircular canals of the vestibular system. The test evaluates the ability of the vestibulo-ocular reflex to maintain stable vision during passive, high acceleration impulsive head rotations of a range of intensities. The software displays a letter optotype on the test screen for a predefined number of video frames, when the acceleration of the head overcomes an adjustable threshold. We have verified the timing of the visual stimulus with respect to the head acceleration recordings and confirmed that the display occurs when the acceleration reaches its peak values. The software ensures that the stimuli delivered to the head are correct by discarding trials presenting either changes of pitch angle or spurious translations. Four different pages summarize the test results and provide detailed diagnostic information about the performance of the subject. Head movement and test performance data are saved to disk for further analysis.
R EFERENCES 1. G. M. Halmagyi and I. S. Curthoys, “A clinical sign of canal paresis,” Arch.Neurol., vol. 45, no. 7. pp.737-739, July, 1988. 2. S. T. Aw, T. Haslwanter, G. M. Halmagyi et al., “Three-dimensional vector analysis of the human vestibuloocular reflex in response to high-acceleration head rotations. I. Responses in normal subjects,” J.Neurophysiol., vol. 76, no. 6. pp.4009-4020, Dec., 1996. 3. S. J. Herdman, “Role of vestibular adaptation in vestibular rehabilitation,” Otolaryngol.Head Neck Surg., vol. 119, no. 1. pp.49-54, July, 1998. 4. J. R. Tian, I. Shubayev, and J. L. Demer, “Dynamic visual acuity during yaw rotation in normal and unilaterally vestibulopathic humans,” Ann.N.Y.Acad.Sci., vol. 942. pp.501-504, Oct., 2001. 5. M. C. Schubert, A. A. Migliaccio, and C. C. la Santina, “Dynamic visual acuity during passive head thrusts in canal planes,” J.Assoc.Res.Otolaryngol., vol. 7, no. 4. pp.329-338, Dec., 2006. 6. S. J. Herdman, R. J. Tusa, P. Blatt et al., “Computerized dynamic visual acuity test in the assessment of vestibular deficits,” Am.J.Otol., vol. 19, no. 6. pp.790-796, Nov., 1998. 7. M. C. Schubert, S. J. Herdman, and R. J. Tusa, “Vertical dynamic visual acuity in normal subjects and patients with vestibular hypofunction,” Otol.Neurotol., vol. 23, no. 3. pp.372-377, May, 2002. 8. S. Ramat and D. S. Zee, “Ocular motor responses to abrupt interaural head translation in normal humans,” J.Neurophysiol., vol. 90, no. 2. pp.887-902, Aug., 2003.
IFMBE Proceedings Vol. 29
New Application for Automatic Hemifield Damage Identification in Humphrey Field Analyzer (HFA) Visual Fields A. Salonikiou, V. Kilintzis, A. Antoniadis, and F. Topouzis AUTH, Laboratory of Research and Clinical Applications in Ophthalmology, AHEPA Hospital, Thessaloniki, Greece
Abstract— Glaucoma is a disease affecting the optic nerve. Structural as well as functional changes occur as the disease progresses. Glaucomatous optic discs appear with certain patterns of structural damage, and the disease’s main consequence is patient’s visual field damage. Taking therapeutic decisions for glaucomatous patients requires visual field examination which until today is the cornerstone of glaucomatous patients’ management. Humphrey Field Analyzer (HFA) is the most commonly used device for visual field examination. There are certain demographic and anatomic risk factors contributing to glaucoma occurrence and progression. Vascular risk factors as well as ocular blood flow have been studied also, but their role remains unclear. Studying the correlation between various risk factors and structural and functional damage in glaucoma is one of the most interesting research fields lately. The purpose of this project was the development of software that reads HFA data from large databases. Then it categorizes visual fields according to hemifield damage. The results can be used for the correlation of hemifield damage with differences in ocular blood between the upper and lower half of the patients’ retina, as measured with the Heidelberg Retina Flowmeter (HRF). The software can be also used for the calculation of the binocular visual field, after the proper modifications. The software was developed using php and the interface was designed so as to offer a user friendly environment and a summary as well as a descriptive results display. Keywords— glaucoma, Humphrey Field Analyzer, visual field, hemifield, structure-function correlation.
I. INTRODUCTION Glaucoma is the leading cause of irreversible blindness that could be prevented in the world [1]. It is a complex disease [2, 3] with unknown etiology and includes a group of different clinical entities. It affects the optic nerve causing characteristic damage to the optic disc and the retinal nerve fiber layer (RNFL). Visual field (VF) damage is its consequence if it remains untreated. Its course is long lasting and does not cause symptoms in the majority of the patients until the visual field is severely damaged and the central vision is affected.
Risk factors for glaucoma occurrence and progression have been determined and include age, central corneal thickness, family history of glaucoma and elevated intraocular pressure [4-7]. The latter is the only modifiable risk factor. Vascular risk factors are also being studied, but their role remains unclear [8-11]. Diagnosis and management is based firstly on clinical examination. VF examination is necessary for the determination of the stage of the disease and is the cornerstone of further management. Automated perimetry is the method to examine patients’ visual fields, and Humphrey Field Analyzer (HFA) is one of the most frequently used devices. Nowadays new laser imaging technologies have been developed for the diagnosis and further management of glaucoma. This fact offers the opportunity to study different structural aspects of the disease and provides with quantified data enabling us to study correlation between structure and function. Correlation of visual field damage and blood flow has been done using various methods. Hemifield damage has been correlated to mean blood flow measured with the Heidelberg Retina Flowmeter (HRF) [12], while VF indices like the mean deviation (MD) or Corrected Pattern Standard Deviation (CPSD) have been correlated with blood flow calculated using 10x10 pixel window analysis of the HRF [13] or laser speckle flowgraphy (LSFG) [14]. Finally, other studies have correlated VF damage to RNFL measurements with the Optical Coherence Tomography (OCT) [15,16] or the Scanning laser polarimetry (GDx) [17]. The purpose of this project was the development of software that reads HFA data from large databases. Then it identifies hemifield with the more extended VF defect. Current HFA software does not provide this information and identification of the hemifield with the more extended damage is currently done by clinicians when they evaluate the printout result of the VF test. Clinicians evaluating printouts would be impractical when it has to do with large databases. This software will provide with a useful tool in order to automatically obtain the information of the hemifield with the more extended damage in large VF databases. This can be used firstly for the correlation of hemifield damage with differences in ocular blood flow between the
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 506–509, 2010. www.springerlink.com
New Application for Automatic Hemifield Damage Identification in Humphrey Field Analyzer (HFA) Visual Fields
507
upper and lower half of the patients’ retina, as measured with the Humphrey Retina Flowmeter (HRF) using the pixel-by-pixel technique [18]. Further, other potential applications include hemifield structure-function correlation with Heidelberg Retina Tomograph (HRT) measurements. Moreover, after the proper modifications the software could be used i) to evaluate the hemifield damage not only with regards to how extended it is but also according to how deep it is and ii) to calculate the binocular visual field.
II. METHODS A. HFA Operation and Description of the Printout HFA records the sensitivity of certain test points in the visual field to known intensity light targets. The sensitivity is measured in decibels (dB). HFA has a range of suprathreshold and full threshold strategies, with the 30-2 and 24-2 SITA standard program being the most commonly used [19]. The 24-2 program tests 54 points while 30-2 tests 76 points. At the end of the examination, the HFA gives the measured sensitivity of each test point in a printout (Fig.1) containing displays including the greyscale, the numerical, the total and standard deviation. The total deviation display represents the deviation of the patient’s results from that of age-matched controls, while pattern deviation is adjusted for any generalized depression in the overall field which might be caused by other factors such as lens opacities or miosis [19]. There are also indices summarizing test results in a single number. The more often used is Mean Deviation (MD), which is a measure of the overall field loss [19]. Important piece of information on the printout are the reliability indices fixation losses, false positives and false negatives. Values >33% for fixation losses or >20% for false positives and false negatives render the test unreliable, according to the European Glaucoma Society (EGS) Guidelines [20]. This means that the test results may not represent the true visual field status. However, a value of more than 33% for false negative answers may be a sign of disease severity [19]. There are studies that consider as acceptable false positive and false negative values of up to 33% [21]. What should be mentioned at this point is the fact that there is no open source software for the extraction and processing of HFA data up to date.
Fig. 1 Visual Field printout randomized clinical trial [22]. The investigators of this study have developed quantitative methods to assess the test reliability and measure the severity of glaucomatous visual field defects with the 24-2 threshold program of the Humphrey Visual Field Analyzer. More specifically, the scoring of visual field defects is based on the following: • • • •
•
Defects may occur in the upper or lower hemifield, or in the nasal field. The total deviation plot is used. Test locations above and below the center of the physiologic blind spot are excluded from scoring since they are not reported on the total deviation plot. For a defect in the hemifields to be considered, three or more adjacent sites within the hemifield must be affected. Two locations are adjacent if they are side by side either horizontally, vertically or obliquely. Three or more locations in a cluster of sites are adjacent if each location in the cluster is adjacent to at least one other in the cluster. To be considered defective, the depression of a patient’s threshold at a test site must be sufficiently large, compared with age-adjusted normal values, as to be unlikely due to spontaneous intra-test fluctuation [23]. The defect should be caused by glaucoma and not by other ocular diseases. This is based on the clinical ophthalmic examination.
B. VF Defect Scoring
•
Evaluation of a VF test examination is based on certain rules, since not all defects are attributed to glaucoma. Most commonly used criteria are those suggested by the Advanced Glaucoma Intervention Study (AGIS), a multicenter,
The amount of depression that renders a test site defective varies with its location, as shown in Fig.2.
IFMBE Proceedings Vol. 29
508
A. Salonikiou et al.
A cluster of three or more depressed sites in a hemifield constitutes a hemifield defect.
•
Fig. 2 Minimum
amount of depression, in decibels, that identifies test locations as defective in Humphrey Field Analyzer threshold 24-2 test total deviation plot of the STATPAC-2 printout. Array shown is for right eye; left eye is a mirror image. Sites above and below the center of the blindspot are not counted [22]
•
C. Software Development The developed web based application incorporates the advantages of a user friendly interface and remote access capability. The development language was PHP5 and the service was supported by Apache Web Server, both commonly used open source tools. We used VF data from our Laboratory‘s HFA database. The conversion from the sensitivity map to the deviation map was done with an external commercial program linked to the HFA, producing a comma separated value (csv) file. • •
•
•
At first, a script that parses the csv file was implemented. Two 4x9 arrays were defined to hold 24-2 VF data or two 5x10 arrays for 30-2 VFs. Each one of the two 4x9 or 5x10 arrays represents the upper and lower hemifield respectively. Peripheral points with no values in the VF were given the value 99 so as to be ignored during the calculations in the next steps. The same value was given to the points corresponding to the “blind spot” (the points corresponding to the optic nerve head). With the use of the appropriate loops, data from the csv file were read and test point values were attributed to the proper array positions. Data in the file is stored in a different order for right and left eyes, so the parsing script manipulates data differently, according to eye. The result of this procedure was the reconstruction of the total deviation plot. Then the deviation from the sensitivity thresholds was calculated, using the AGIS array as shown in Fig.2. This was done by algebraically adding the minimum amount of depression to each test point’s sensitivity
•
value. Since the array contains thresholds only for 24-2 VFs, the peripheral values of 30-2 VFs were ignored and these specific VFs were handled as if they were 242. As soon as there are respective values in the literature for these test points, they will be incorporated in the program. If the number resulting from the above procedure had a negative sign, this meant that the specific test point had a reduction in sensitivity outside the normal range. All test points with a negative value were detected and their neighboring test points were checked for a possible negative value, with properly designed loops. Each time a defected neighbor was detected, a counter increased its value by one. At the end of the procedure, a new same dimension array as the initial was produced, containing values representing the number of defected neighbors for each specific point. Then, clusters of defected test points were detected. When the sum of the values of two adjacent points was >2, meaning that the specific point and one of its neighbors had at least one more defected neighbor, this meant the existence of a cluster. A counter held the number of hemifield test points belonging to clusters. The comparison of the number of points forming clusters for the upper and lower hemifield revealed the hemifield with the most extended defect.
Reliability of VFs was checked according to the aforementioned criteria [20]. If a VF test was found unreliable, then it was displayed in red, allowing exclusion of the specific test from further investigations. The interface is designed in a user friendly way. The user chooses the local csv file to be analysed. A notification appears if a wrong file type is selected. There is the option to see the results either in a list format or in a more descriptive display. The results can be saved as .txt files for further processing. Application home page is as illustrated in Fig.3.
IFMBE Proceedings Vol. 29
Fig. 3 Application home page
New Application for Automatic Hemifield Damage Identification in Humphrey Field Analyzer (HFA) Visual Fields
III. CONCLUSIONS This software will offer the opportunity to study large VF databases, allowing pointwise VF data processing. For the time being, our program runs on a local environment. However, the fact that it is a web based application allows it to be remotely used and therefore serve, for example, databases from large population based multicentric studies. Hemifield-based analysis of glaucomatous VFs along with measurements using imaging technologies may reveal interesting information on structure-function correlation in glaucoma. Furthermore, after the proper adjustments in the program, reconstruction of the binocular visual field will be useful for the evaluation of the consequences of VF damage in glaucomatous patients’ every day living.
ACKNOWLEDGMENT We would like to thank the scientific as well as the technical staff of the Laboratory of Research and Clinical Applications in Ophthalmology for their valuable advice and help during the realization of this project.
REFERENCES 1. Quigley HA. (1996) Number of people with glaucoma worldwide. Future of health insurance. Br J Ophthalmol. 80:389-393 2. Copin B, Brézin AP, Valtot F et al. (2002) Apolipoprotein EPromoter Single-Nucleotide Polymorphisms affect the phenotype of primary open-angle Glaucoma and demonstrate interaction with the myocilin gene. Am J Hum Genet..70:1575-1581 3. Wiggs JL, Allingham RR, Hossain A et al. (2000) Genome-wide scan for adult onset primary open angle glaucoma. Hum Mol Genet.9:1109-1117 4. Leske MC, Wu SY, Hennis A et al. (2008) Risk Factors for Incident Open-angle Glaucoma The Barbados Eye Studies. Ophthalmology 115:85–93 5. Leske MC et al. (2004) Factors for progression and glaucoma treatment: the Early Manifest Glaucoma Trial. Curr Opin Ophthalmol.15:102-6 6. Leske MC, Heijl A, Hussein M et al. (2003) Factors for glaucoma progression and the effect of treatment: the early manifest glaucoma trial. Arch Ophthalmol;121:48 –56 7. Leske MC, Heijl A, Hyman L et al. Predictors of Long-term Progression in the Early Manifest Glaucoma Trial. Ophthalmology 2007;114:1965–1972 8. Wilson MR, Hertzemark E, Walker AM et al. (1987) A case-control study of risk factors in open angle glaucoma. Arch Ophthalmol 105: 1066-1071
509
9. McLeod SD, West SK, Quigley HA, Forzard JL. (1990) A longitudinal study of the relationship between intraocular and blood pressures. Invest Ophthalmol Vis Sci 31: 2361-2366 10. Gasser P, Flammer J. (1991) Blood cell velocity in the nailfold capillaries in patients with normal-tension and high-tension glaucoma. Am J Ohpthalmol 111: 585-588 11. Drance SM, Douglas GR, Wisjman K et al. (1988) Response of blood flow to warm and cold in normal and low-tension glaucoma patients. Am J Ophthalmol 105: 35-39 12. Sato EA, Ohtake Y, Shinoda K et al. (2006) Decreased blood flow at neuroretinal rim of optic nerve head corresponds with visual field deficit in eyes with normal tension glaucoma. Graefes Arch Clin Exp Ophthalmol.244:795-801 13. Ciancaglini M, Carpineto P, Costagliola C et al. (2001) Perfusion of the optic nerve head and visual field damage in glaucomatous patients. Graefes Arch Clin Exp Ophthalmol.239:549-55 14. Yaoeda K, Shirakashi M, Fukushima A et al. (2003) Relationship between optic nerve head microcirculation and visual field loss in glaucoma. Acta Ophthalmol Scand.81:253-9 15. Harwerth RS, Vilupuru AS, Rangaswamy NV et al. (2007) The relationship between nerve fiber layer and perimetry measurements.Invest Ophthalmol Vis Sci.48:763-73 16. Hood DC, Kardon RH. (2007) A framework for comparing structural and functional measures of glaucomatous damage. Prog Retin Eye Res.26:688–710 17. Mai TA, Reus NJ, Lemij HG. (2007) Structure-function relationship is stronger with enhanced corneal compensation than with variable corneal compensation in scanning laser polarimetry. Invest Ophthalmol Vis Sci.48:1651-8 18. Mavroudis L, Harris A, Topouzis F et al. (2008) Reproducibility of pixel-by-pixel analysis of Heidelberg retinal flowmetry images: the Thessaloniki Eye Study. Acta Ophthalmol.:86: 81–86 19. Kanski J (2003) Clinical Ophthalmology A systematic approach. Butterworth Heinemann, Elsevier, USA 20. European Glaucoma Society. Terminology and Guidelines for Glaucoma. 3rd edition 2008 21. Topouzis F, Wilson MR, Harris A et al. (2007) Prevalence of openangle glaucoma in Greece: the Thessaloniki Eye Study. Am J Ophthalmol.144:511-9 22. The Advanced Glaucoma Intervention Study Investigators. (1994) Advanced Glaucoma Intervention Study 2. Visual Field Test Scoring and Reliability. Ophthalmology101:1445-1455 23. Jampel HD, Vitale S, Ding Y et al. (2006) Test-Retest variability in structural and functional parameters of glaucoma damage in the Glaucoma Imaging Longitudinal Study. J Glaucoma15:152-157
Author: Angeliki Salonikiou Institute: Aristotle University of Thessaloniki, Laboratory of Research and Clinical Applications in Ophtalmology, A’ Department of Ophthalmology, AHEPA Hospital Street: Stilponos Kyriakidi 1 City: Thessaloniki Country: Greece Email:
[email protected] IFMBE Proceedings Vol. 29
The Effect of Mechano– and Magnetochemically Synthesized Magnetosensitive Nanocomplex and Electromagnetic Irradiation on Animal Tumor V.E. Orel1, A.V. Romanov1, I.I. Dzyatkovska1, M.O. Nikolov1, Yu.G. Mel'nik1, N.M. Dzyatkovska1 and I.B. Shchepotin2 1 National Cancer Institute/Medical Physics & Bioengineering Laboratory, Kyiv, Ukraine National Cancer Institute/Department of Tumors of Abdominal Cavity and Retroperitoneal Space, Kyiv, Ukraine
2
Abstract—The research of animals with Guerin carcinoma was shown, that mechano– and magnetochemically synthesized magnetosensitive nanocomplex (MNC) on the basis of nanoparticles of Fe3O4, KCl and doxorubicin had greater antitumor effect than conventional doxorubicin and similar mechanochemically synthesized MNC at next local electromagnetic irradiation and mild hyperthermia of animal tumors. Survival rate of animals with tumors was maximal in experiments after introduction of mechanochemically or mechano– and magnetochemically synthesized MNC and next irradiation of animal tumors. Keywords— mechano– and magnetochemically synthesized magnetosensitive complex, magnetic nanoparticles, electromagnetic irradiation, mild hyperthermia, tumor. I. INTRODUCTION
The problem of cancer therapy is, that often the dose of systemically applied chemotherapeutics needed to annihilate all tumor cells would also end the life of cancer patient. The use of magnetic nanoparticles is one attempt to over-come this dilemma. Direct injection of magnetic particles themselves for inducing local hyperthermia are currently under investigation. Magnetic drug targeting is another promising attempt for treating malignancies. This method makes use of superparamagnetic nanoparticles bound to chemotherapeutics, focused by a strong external magnetic field to the tumor region. This leads to higher doses of the chemotherapeutic agent in the region of the malignancy, even if the overall dose is reduced [1]. The use of spatially inhomogeneous electromagnetic field in local inductive hyperthermia at physiological temperatures increased antitumor effect of antitumor drug doxorubicin (DR) for transplanted DR-resistant Guerin's carcinoma and accompanied by the change of thermodynamical entropy. [2]. We suppose that the magnetosensitive spin-dependent reaction between structural defects initiated by mechanochemical activation [3] probably will increase antitumor effect of DR. In this study we are focusing on the effect of mechano– and magnetochemically synthesized magnetosensitive
nanocomplex (MNC) and spatially inhomogeneous electromagnetic irradiation (EI) on nonlinear dynamics of the growth for Guerin carcinoma. II. MATERIALS AND METHODS
A. Experimental animals and tumor transplantation. In the study 56 male rats weighing 100 ± 15g bred in the vivarium of National Cancer Institute were used. The transplantation of Guerin carcinoma was performed according to the established procedure. All animal procedures were carried out according to the rules of the regional ethic committee. Animals were housed in 7 groups: 1 – control (no treatment); 2 – treatment by DR; 3 – DR + ȿI; 4 – treatment by mechanochemically synthesized ɆNC; 5 – mechano– and magnetochemically synthesized ɆNC; 6 – mechanochemically synthesized ɆNC + ȿI; group 7 –mechano– and magnetochemically synthesized ɆNC+EI. B. Mechano- and magnetochemical synthesis. Electromagnetic irradiation. MNC was mechano– and/or magnetochemically synthesized from Fe3O4 nanoparticles, KCl and DR by laboratorial magnetic-resonance high-precision tribogenerator. Mechanical processing was performed at a frequency 35 Hz, amplitude 9 mm for 5 min using an input mechanical energy of 20W/g and 27.7MHz EI with an initial power of 100 W. Mean diameter of Fe3O4 nanoparticles was 20–40 nm. First prototype of the device for EI called “Magnetotherm” (Radmir, Ukraine) was used (Nikolov et al., 2008). The frequency of EI was 40 MHz with an initial power of 100 W. The animal tumors were irradiated locally by inductive coaxial applicators that had spatial inhomogeneity of electromagnetic field and initiated mild hyperthermia (37.9qC) in tumors [2]. Experimental animals were treated by MNC: DR (Pharmacia & Upjohn) in the dose 1.5 mg/kg, Fe3O4 + KCl in the dose 3 mg/kg. Weight percentage of KCl was 3%.The treatment was performed three times by drug and EI from 3 day
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 510–512, 2010. www.springerlink.com
The Effect of Mechano– and Magnetochemically Synthesized Magnetosensitive Nanocomplex and Electromagnetic Irradiation
511
day after tumor transplantation every other two days. In an area of tumors was disposed a permanent magnet with H = 1990 A/m for localization of MNC within tumor. C. The analysis of nonlinear kinetics of tumor volume. Nonlinear kinetics of tumor volume was evaluated by growth factor M according to autocatalytic equation and the braking ratio [4]. Statistical processing of numerical results was carried out using Statistica 6.0 (© StatSoft, Inc. 1984–2001) computer program with parametric Student’s t-test. III. RESULT AND DISCUSSION.
The growth kinetics of animal tumors is shown in Table 1. The growth kinetics for 7th group had minimal response under the influence of mechano– and magnetochemically synthesized ɆNC and EI. Table 1 The growth kinetics of Guerin carcinoma from 7 to 24 days after transplantation N
Parameters
Treatment
M, day-1 (M ± m) 1 4 2 6 3 5 7
Control (without DR, MNC and ȿI) Mechanochemically synthesized ɆNC DR Mechanochemically synthesized ɆNC + ȿI DR + ȿI Mechano– and magnetochemically synthesized ɆNC Mechano– and magnetochemically synthesized ɆNC+EI
N
0.31 r 0.04 0.28 r 0.04 0.18 r 0.01* 0.16 r 0.01* 0.16 r 0.02*
1 1.08 1.66 1.94 1.89
0.16 r 0.01*
1.88
0.13 r 0.03*
2.43
* Statistically significant difference from control group
Greatest survival rate was observed for animals from 6th and 7th groups (Fig. 1). It exceeded survival rate for animals from a control group on 77%, and animals from a 4th group with injected mechanochemically synthesized MNC on 50%. For animals from 5th group, when MNC before introduction was synthesized as mechano- as magnetochemically, the survival rate increased on 39% in comparison with 4th group, but it was lesser on 20% compare to 6th and 7th group of animals.
Fig. 1 The survival rate of animals with Guerin carcinoma: 1– control (without DR, MNC and ȿI); 2 – DR; 3 – DR + ȿI ; 4–mechanochemically synthesized ɆNC; 5 – mechano– and magnetochemically synthesized ɆNC; 6 –mechanochemically synthesized ɆNC + ȿI ; 7 – mechano– and magnetochemically synthesized ɆNC+EI Our co-operation studies with Dr. A.P. Burlaka and Dr. S.N. Lukin shown that electron spin resonance spectra of mechano– and magnetochemically synthesized MNC had broad peak with g –factors in range 4.25 – 6.0. That is typical g –factors for the iron-transport proteins methemoglobin and transferrin [5]. Nanoparticles Fe3O4 have g–factor 2.0839 and 2.18838 [6]. For conventional and mechanochemically synthesized DR g-factor equal to 2.005; 2.003 and 1.97 [7]. We purpose, that effect increases of antitumor MNC was result of spin conversion in radical electron pair during mechano– and magnetochemically synthesized ɆNC and EI of animal tumor. IV. CONCLUSION
The research of animals with Guerin carcinoma was shown, that mechano– and magnetochemically synthesized MNC on the basis of nanoparticles of Fe3O4, KCl and doxorubicin had greater antitumor effect than conventional doxorubicin and similar mechanochemically synthesized MNC at next local electromagnetic irradiation and mild hyperthermia of animal tumors. Survival rate of animals with tumors was maximal in experiments after introduction of mechanochemically or mechano– and magnetochemically synthesized MNC and next irradiation of animal tumors.
REFERENCES 1.
Peng X, Qian X, Mao H et al (2008) Targeted magnetic iron oxide nanoparticles for tumor imaging and therapy. Int J Nanomedicine 3: 311–321
IFMBE Proceedings Vol. 29
512 2.
3. 4. 5. 6. 7.
V.E. Orel et al. Orel V, Romanov A (2010) The Effect of Spatially Inhomogeneous Electromagnetic Field and Local Inductive Hyperthermia on Nonlinear Dynamics of the Growth for Transplanted Animal Tumors. In: Nonlinear Dynamics (Ed. Todd Evans). INTECH, Croatia Golovin Y (2004) Mechanochemical reaction between structural defects in magnetics fields. J Mat Sci 39: 5129–5134 Emanuel N (1977) Kinetics of experimental tumor processes, Nauka, Moscow (in Russian) Saifutdinov R, Larina L, Vakulskaya T et al (2001) Electron paramagnetic resonance in biochemistry and medicine. Kluwer Academic, Pleunim Publisher, New York. Köseoglu Y, Ysildiz F, Kim D et al (2004) EPR studies on Na–oleate coated Fe3O4 nanoparticles. physica status solidi. Physica status solidi 12: 3511–3515 Orel V, Kudryavets Y, Bezdenezhnih N et al (2005) Mechanochemically activated doxorubicin nanoparticles in combination with 40 MHz frequency irradiation on A–549 lung carcinoma cells. Drug Delivery 12: 171–178 Author: Institute: Street: City: Country: Email:
Valerii E. Orel National Cancer Institute 33/43 Lomonosova Str Kyiv Ukraine
[email protected] IFMBE Proceedings Vol. 29
Verification of Measuring System for Automation Intra – Abdominal Pressure Measurement T. Tóth, M. Michalíková, L. Bednarčíková, M. Petrík, and J. Živčák Technical University of Košice, Faculty of Mechanical Engineering, Department of Biomedical Engineering, Automation and Measurement, Košice, Slovakia Abstract— With the newest medical researches and studies increase equipment‘s rate of health centers. By contrast to it, any parts of medicine in SR and abroad don‘t utilize all actual knowledge’s and possibilities. One of this is the measurement of intra-abdominal pressure in critical ill patients. From the results of analysis the most often intra-abdominal pressure measurement is via bladder. This is not invasive method with high significant weight. The measurement techniques what are in present time usually used does not meet criteria for modern diagnostic methods. This paper describes the verification of measuring system for intra – abdominal pressure measurement. The measuring system is the basic part of proposal device for automated measurement of intra – abdominal pressure. Keywords— intra-abdominal syndrome, measurement.
pressure,
•
the high abdominal pressure decrease the vessel perfusion, and may cause the death.
The intra – abdominal compartment syndrome was described by Kron, Harman and Nolan in 1984 and Feistam was the first who use the term “abdominal compartment syndrome” in 1989. [8] The standardization of terms was beginning in the Second World Conference of Abdominal Compartment Syndrome and the final report was publishing in 2004.
II. BASIC THERMS
compartment
I. INTRODUCTION Accidents in abdominal area and polytraumas cause that the number of patient with intra – abdominal hypertension and abdominal compartment syndrome increasing. By one of methods which allow preventing complications due high abdominal pressure is his measurement. The untreated intra – abdominal hypertension have high death rate. Intra – abdominal pressure (IAP) was first described in a work of Marey in 1863. In the year 1865 Braune register the first measurement of intra – abdominal pressure. The measurement was performed via rectum. In the next 25 years was documented the results of treatment and measurement of intra – abdominal pressure. Heinricius from Germany (1890) determine, that the pressure between 27 and 45cmH2O (approx. 20 – 33mmHg, 2,64 – 4,41kPa) is lethal for the animals with breath disease whereby decrease the blood pressure and heart diastolic distension. Approximately in the year 1911 Haven Emerson issue the work about intra – abdominal pressure. The results from this work are: • the contraction of diaphragm during the breath in is the main factor of intra – abdominal pressure raising, • the anesthesia and muscle paralysis reduce the pressure in abdomen,
The abdomen can be considered a closed box with walls either rigid (costal arch, spine, and pelvis) or flexible (abdominal wall and diaphragm). The elasticity of the walls and the character of its contents determine the pressure within the abdomen at any given time. Since the abdomen and its contents can be considered as relatively noncompressive and primarily fluid in character, behaving in accordance to Pascal’s law, the IAP measured at one point may be assumed to represent the IAP throughout the abdomen [1, 3]. Intra-Abdominal Pressure (IAP) IAP is defined as a stable pressure into the abdominal cavity. In the inspiration the IAP increasing (diaphragm contraction) and in the expiration decreasing (diaphragm relaxation). It is directly dependent on the volume of organs, presence of diseases and with limitations of abdominal wall expansion. [2, 4, 6] In the Figure 2 is illustrated the intra – abdominal hypertension, her value was depend on clinical scenario. Intra-Abdominal Hypertension (IAH) Pathological IAP is a continuum ranging from mild IAP elevations without clinically significant adverse effects to substantial increases in IAP with grave consequences to virtually all organ systems in the body. [2, 4, 6]
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 513–516, 2010. www.springerlink.com
514
T. Tóth et al.
25mmHg. The bag is connected to the bottom of container with Velcro. The second part is the sensing system. For level detection are used two glass pipes as a level gauges. The first gauge is connected via reduction to the saline bag and measure the pressure (water column above the bag surface) in the bag. The second gauge directly measure the level of water column. The gauges inputs have the same distance from the bottom of the container. The sensing system is connected to reduction through the hose – pipe with 4mm inner diameter. [10] The measurement was performed with following approach: Fig. 1 Description of the pressure status in abdomen and organ dysfunction depending up the intra – abdominal pressure Abdominal Compartment Syndrome (ACS) Critical IAP in the majority of patients, as outlined above, appears to reside somewhere between 10 and 15 mmHg. It is at this pressure that reductions in microcirculatory blood flow occur, and the initial development of organ dysfunction and failure is first witnessed. ACS is the natural progression of these pressure-induced end-organ changes and develops if IAH is not recognized and treated in a timely manner. [2, 4, 6] The recognition of significance of IAP monitoring at IAH, IAP diagnostic and management starting the development of direct (invasive) and indirect (noninvasive) measurement methods. In medical praxis the most used method is the indirect measurement via bladder. In the Table 1 are correlation coefficients for intra – abdominal pressure measured via bladder. [3, 5] Table 1
• • • • • • •
saline bag was filled with 100ml of water creation of 5mmHg pressure through the water column, the value was read from the level gauge, stabilization of the water level (15 - 20)s, measuring process, increasing the pressure up to 25mmHg with 5mmHg steps, measuring after each increment, decreasing the pressure after the reaching 25mmHg with 5mmHg steps, measuring after each decrement, with this approach was obtained 20 measurement packs, in each pack contain 5 levels (5 – 25mmHg with 5mmHg step). [9]
Pressure dependency of bladder pressure and intra – peritoneal
Author
Publicize in
Ridings Johna Fusco Davis Risin Schachtrupp
J Trauma CC forum J Trauma Int Care Med Am J Surg Crit Care Med
Year 1995 1999 2001 2005 2006 2006
Correlation Coefficient 0,98 0,92 0,88 0,95 0,96 0,95
III. VERIFICATION OF MEASURING SYSTEM The testing device for pressure sensor verification is set up from two parts (Figure 2). The first part represents the abdomen model which is made from 250ml saline bag (replacement of bladder). The saline bag is set in the bottom of 35L container, which allow create the pressure up to
Fig. 2 Schematic representation of proposal for the experimental verification of sensing system The one measurement contains 50 values with 100ms pause between two values. The pressure sensor have analog output (0 – 5V), which is processed in PIC microprocessor. The program in PIC was designed for data reading from sensor, A/D conversion (10-bit) and data sending to PC. The values from A/D converter are recalculated to the pressure value (Table 2).
IFMBE Proceedings Vol. 29
Verification of Measuring System for Automation Intra – Abdominal Pressure Measurement
The results of measurements are effected by measuring methodic. The biggest problem is the connection of saline bag to the container. The bag is connected with Velcro only in his centre line. Table 2
Table of the recalculated values for 20 measurement packs p(5) 5,413 5,368 5,408 5,377 5,411 5,430 5,430 5,403 5,368 5,443 5,422 5,421 5,361 5,371 5,431 5,491 5,504 5,415 5,472 5,443 5,419 0,040
Pressure [mmHg] p(10) p(15) p(20) 10,326 15,180 20,084 10,194 15,072 19,870 10,267 15,154 19,950 10,232 15,113 19,978 10,284 15,157 20,001 10,249 15,114 19,944 10,273 15,255 19,985 10,214 15,075 19,911 10,267 15,075 19,933 10,236 15,092 19,762 10,271 15,114 19,982 10,226 15,041 19,889 10,227 15,095 19,960 10,254 15,051 19,905 10,249 15,063 19,971 10,309 15,097 19,997 10,352 15,166 20,006 10,283 15,107 20,026 10,279 15,157 20,015 10,315 15,135 20,006 10,265 15,116 19,959 0,040 0,052 0,069
p(25) 24,889 24,834 24,845 24,821 24,840 24,886 24,817 24,837 24,817 24,824 24,804 24,839 24,786 24,826 24,783 24,905 24,853 24,865 24,837 24,861 24,838 0,032
p + 3s p
5,538
10,384
15,270
20,166
24,934
p − 3s p
5,300
10,146
14,961
19,752
24,743
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Mean p Standard Deviation sp
The dependency of measured values on expected value is linear with correlation coefficient from 0,98 to 1. The total error of measurement is given by sum of sensor εs and A/D converter εc maximum error. [7]
ε = ε s + ε c = 0,375 + 0,035 = 0,41 mmHg
IV. CONCLUSION The proposing of measuring devices for medical applications must satisfy terms for safety and reliable using. By one of these terms is the sterilization of all parts which come into the touch of body liquids (urine). This condition has had the basic effect at the sensor selection. This measurement has the character of pilot measurement for acquiring the basic parameters of sensor and measuring string elements.
515
From the measurement result, that the first problem is the saline bag fixation to the container. One of the possibilities is fixation of the bag border. For decreasing of the total error of measurement is possible used the stand alone 16 – bit A/D converter. The total error will be in this case equal to the sensor error. After the application of proposal changes are necessary additional testing including long time testing of sensor stability.
ACKNOWLEDGMENT This research has been supported by the research project 1/0829/08 VEGA - Correlation of Input Parameters Changes and Thermogram Results in Infrared Thermographic Diagnostic.
REFERENCES 1. Malbrain, ML., Cheatham, ML., Kirkpatrick, A., Sugrue, M., Parr, M., De Waele, J., Balogh, Z., Leppäniemi, A., Olvera, C., Ivatury, R., D'Amours, S., Wendon, J., Hillman, K., Johansson, K., Kolkman, K., Wilmer, A.: Results from the International Conference of Experts on Intra-abdominal Hypertension and Abdominal Compartment Syndrome. I. Definitions, Intensive Care Med. 2006 Nov;32(11):1722-32. Epub 2006 Sep 12 2. Malbrain, ML., Cheatham, ML., Kirkpatrick, A., Sugrue, M., Parr, M., De Waele, J., Balogh, Z., Leppäniemi, A., Olvera, C., Ivatury, R., D'Amours, S., Wendon, J., Hillman, K., Wilmer, A.: Results from the International Conference of Experts on Intra-abdominal Hypertension and Abdominal Compartment Syndrome. II. Recommendations. Intensive Care Med. 2007 Jun;33(6):951-62. Epub 2007 Mar 22 3. Malbrain, ML., Deeren DH.: Effect of bladder volume on measured intravesical pressure: a prospective cohort study, Critical Care 2006, 10:R98, http://ccforum.com/content/10/4/R98 4. Efstathiou, E., Zaka, M., et al.: "Intra-abdominal pressure monitoring in septic patients." Intensive Care Medicine 31, 2005, Supplement 1(131): S183, Abstract 703 5. Kinball, EJ.: IAP measurement: Bladder techniques, WCACS,, Antwerp, 2007 6. Malbrain ML, Cheatham ML, Kirkpatrick A, Sugrue M, De Waele J, Ivatury R.: Abdominal compartment syndrome: it's time to pay attention!, Intensive Care Medicine, Volume 32, Number 11, November 2006 , pp. 1912-1914(3) 7. Kozlíková, K.: Základy spracovania biomedicínskych meraní I, Askepios Bratislava, 2003, ISBN 80-7167-064-2 8. Ivatury, R., Cheatham M., Malbrain, M., Sugrue, M.: Abdominal Compartment Syndrome, Landes Biosciences, ISBN 978-1-58706196-7 9. Toth, T.: Návrh zariadenia na meranie intra – abdominálneho tlaku, Doktorandská dizertačná práca, košice, 2009 10. Tóth, T., et al.: Meranie intra-abdominálneho tlaku, Automatizácia a riadenie v teórii a praxi, ARTEP 2009 : Workshop odborníkov z univerzít, vysokých škôl a praxe v oblasti automatizácie a riadenia : Zborník príspevkov : 4.3. - 6.3.2009, Stará Lesná, SR. Košice : TU, 2009. s. 68-1-68-7. ISBN 978-80-553-0146-4
IFMBE Proceedings Vol. 29
516
T. Tóth et al. Author: Teodor Tóth Institute: Technical University of Košice, Faculty of Mechanical Engineering, Department of Biomedical Engineering, Automation and Measure-ment Street: Letná 9 City: Košice Country: Slovakia Email:
[email protected] IFMBE Proceedings Vol. 29
Evolution in Bladder Pressure Measuring Implants Developed at K.U.Leuven P. Jourand1 , J. Coosemans1,2 and R. Puers1 1
Katholieke Universiteit Leuven, Departement Elektrotechniek, ESAT-MICAS, Leuven, Belgium 2 now with Zenso, Heverlee, Belgium
Abstract— Bladder pressure monitoring devices have been a topic of great interest for the past two decades. Three devices developed at ESAT-MICAS in this time are reviewed, showing the evolution of these devices. Two of these devices are diagnostic tools small enough to be inserted into the bladder cavity through minimal invasive cystoscopy. The third device is a long term bladder pressure monitoring implant that, if used to drive an artificial sphincter muscle, forms a urological pacemaker that could diminish or rule out urinary incontinence. After a summary of the three devices, recent results are presented from one of the diagnostic tools. Keywords— Bladder pressure, Capacitive sensor, Battery operated, Silicone embedding
Fig. 1: A long term bladder pressure monitoring device [5] developed and produced at ESAT-MICAS
I. I NTRODUCTION The study of urodynamics allows for the direct assessment of possible lower urinary tract dysfunctions [1]. After recording micturation patterns and performing a free flow study, a cystometry is often needed to obtain a correct diagnosis or a visual inspection of the bladder. The standard procedure (which is considered minimal invasive) for a urological investigation is to introduce a small sized catheter directly into the bladder cavity through the natural opening. This allows the study of pressure variations under filling and voiding conditions. If a visual inspection of the bladder is needed, a cystoscope is used instead of a catheter. The catheter or cystoscope remains in place during the procedure, which introduces discomfort, pain and elevated risk of infections for the patient. Furthermore, these clinical investigations are far from ”normal life” conditions since the patient has to remain in an uncomfortable position in the inspection room. The only viable solution to overcome the discomfort and to improve the quality of the recorded signals is the use of a wireless embedded device that either communicates through telemetry [2, 3], or logs the data [4]. The typical technical specifications, recommended for the recording of bladder pressure are [1]: • ±1 cm H2 O or ± ∼ 0.98 mbar pressure resolution. • Ranges of 0 − 250 cm H2 O or ∼ 0 − 245 mbar (relative to atmospheric pressure). • Measurement frequency of 10 Hz.
II. D EVELOPMENTS IN UROLOGICAL TOOLS At the ESAT-MICAS labs of the K.U.Leuven, investigations on bladder pressure measurements have been ongoing since the mid 80s [2, 3]. Two approaches are taken in this research, depending on the term of implantation. Diagnostic tools are used to investigate pathologies under normal life conditions for a short term. For obvious reasons, the use of such tools require a non or minimal invasive procedure. On the other hand, long term implants intend to realize the ultimate dream of the urological pacemaker where an invasive procedure can be justified. A. Long term implantation: towards a urologic pacemaker A proper sensing of bladder pressure is the first step in the idea of a full bladder control system, introduced in 1987 [6]. An autonomous and reliable bladder pressure measurement can ensure adequate stimulation of the detrusor and bladder sphincter muscles. In this way, the muscles are not overstimulated and the lifetime of the ”urological pacemaker” is prolonged. This idea can even be extended one step further by adding a bidirectional telemetric link: • The device signalling the patient when the pressure buildup has reached a critical point, alerting the need to void. • The patient instructing the urological pacemaker when he or she is ready to commence voiding, stopping the stimulation of the sphincter muscles.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 517–520, 2010. www.springerlink.com
518
P. Jourand, J. Coosemans, and R. Puers
An invasive approach was chosen for the development of such a long term bladder pressure monitoring system [5] which is depicted in Fig. 1. Bladder pressure is indirectly measured by placing a pressure transducer (Fig. 1 right) on the outside of the bladder wall inside the abdomen. An inductive link (Fig. 1 left) is used for both powering [7, 8] and communication [9], and is placed right beneath the skin to get a good coupling with the external coil. Both parts are interconnected with flexible tracks. The implant coil has an outer diameter of 23 mm and is 1.5 mm thick. All electronics needed for communication, processing of the results and power conversion, are placed within this coil on a single sided flex. A combination of Pulse Position Modulation (PPM) and Binary Phase Shift Keying (BPSK) is used to reduce the ON time of the load modulation for the downlink communication. A PIC16F88-ML from Microchip is used to digitise the pressure readings and for communication. By using an external clock of 132 kHz, extracted from the RF carrier frequency instead of the internal 4 MHz clock, the current drain of the device drops from 0.61 mA at 2 V to 0.20 mA at 1.8 V. The device is embedded with Nusil Med4210, a biomedical grade polydimethylsiloxane (PDMS) designed for encapsulating medical devices. The PDMS is poured after mixing the two components. Any entrapped air is removed by exposing the device to a vacuum, followed by a curing for 4 hours at 60 ◦ C. Where this inductively powered device introduces restrictions on patient movement, should the bidirectional communication link be used to drive an artificial sphincter, a urological pacemaker with unlimited lifetime (power-wise) is created. B. Short term: diagnostic tools Diagnostic tools require non or minimal invasive procedures. A completely non-invasive procedure focuses on the externally applied pressure, required to interrupt the urinary flow. Using a penile cuff, the internal bladder pressure is assessed by the applied pressure on the cuff. The results are fair but yield limited information. Furthermore, the procedure restricts patient movement and is considered slightly uncomfortable [10, 11]. A minimal invasive approach is reported in [12] allowing some movement to the patient, yet the catheter tube of this device remains in place during the procedure elevating both risk of infections and discomfort. Creating a diagnostic tool while placing all electronics inside the bladder cavity is the most challenging method, yet the only way to create a truly imperceptible bladder pressure measurement system. Research on this method started as soon as 1984 [2] and has been ongoing ever since [4, 13, 14].
Fig. 2: A short term bladder pressure monitoring device developed and produced at ESAT-MICAS
The following approach is used: the system is introduced into the bladder cavity by inserting it through a cystoscope, rendering the procedure minimal invasive. Powering such devices can either be achieved by wireless transfer [7, 8] or by incorporating batteries. Data must either be logged on a memory module and read out after retrieval, or transmitted wirelessly. Using such a cystoscope as an introduction tool extremely restricts the size of the application. The diameters of catheters and cystoscopes used in urology [15] vary depending on both procedure and patient using the French scale to indicate the size. Campbell and Walsh’s bible on urology [16] suggests F8 to F12 and F16 to F25 cystoscopes, for paediatric and adult cystometry respectively. Taking into account that the internal diameter has typical values of 85-90% of the outer diameter, using an F20 cystoscope limits the diameter of the bladder pressure device to 5 mm. Two devices were developed at ESAT-MICAS over the years: the first was developed in 1984 and is depicted in Fig. 2. The device measured pressure values using a resistive Honeywell-Philips pressure sensor, at a sampling rate of 10 Hz. The results were transmitted using a small inductor. The complete hybrid system, operated on a Leclanch´e SR33 5 mAh mercury battery with a nominal voltage of 1.45 V, a diameter of 3.3 mm and a length of 3.4 mm. The complete structure sized 4 mm in diameter by 40 mm length. Although lifetime was increased significantly by switching the pressure sensor, it was still rather limited. Recent technological achievements with capacitive pressure sensors enabled the creation of the second device, depicted in Fig. 3. The device is built-up with off-the-shelf discrete components placed on a Kapton foil and embedded in a protective silicone encapsulation. This results in a low cost flexible bladder pressure measurement system. Because the device is battery operated and it logs and stores the pressure data on an EEPROM module, it is imperceptible for the patient, since no external devices are mandatory during recording. The device, is powered by a BR316 lithium cell with a capacity of 13 mAh. Its power can be switched off magnetically with a reed switch to increase shelf-life after assembly. The E1.3N capac-
IFMBE Proceedings Vol. 29
Evolution in Bladder Pressure Measuring Implants Developed at K.U. Leuven
Range and Accuracy Test of an Unembedded Sample 1300 Calculated Pressure (mbar)
itive pressure sensor from MicroFab was chosen for its operating range (0.5 to 1.3 bar absolute pressure) and size. Within normal bladder pressures (1 to 1.3 bar absolute pressure), the capacitance value varies between 6.0026 and 6.260 pF. These values are digitized by an AD7153 capacitance-to-digital converter (CDC) from Analog Devices to a 12 bit value. The pressure data is reduced to 5 bits by a PIC10F206 microcontroller from Microchip and further combined with 11 bits
519
1250 1200 1150 1100 1050 1000
0
100
200
300
400
500 Time (s)
600
700
800
900
Detail of Accuracy Test
Calculated Pressure (mbar)
1037
1036
1035
1034
1033
1032
810
Fig. 3: The latest short term bladder pressure monitoring device developed at ESAT-MICAS and produced by CMST-Gent
of timing data counting the number of unchanged samples. These 16 bit data samples are stored on a 64 kbit EEPROM through I2 C communication. The devices were tested with a Druck DPI 600 reference pressure, showing a resolution of 30 mbar with the 5 bit pressure result, while consuming 320μA on average at 3 V. Having a diameter of 5 mm and a length of 40 mm, the device can be inserted using an F20 cystoscope therein logging and storing over 24 hours of pressure data on an EEPROM module. The system specifications can be summarized as: • Sizing 40 mm in length and 5 mm in diameter allowing insertion through an F20 cystoscope. • Detecting absolute pressures between 1000 and 1300 mbar. Without a connection to the outside world, the bladder cavity lacks a reference pressure, explaining the need for an absolute pressure sensor. • Operating on battery power for over 24 hours. The diameter of the battery must not exceed 4 mm. • Logging pressure data on a 64kbit EEPROM module. • Containing an on/off switch for extended shelf life. • Low-cost implementation and production.
III. M EASUREMENTS AND RESULTS The most recent device was tested in its non-encapsulated form inside a dedicated pressure chamber with a Druck DPI
815
820
825
830
835 840 Time (s)
845
850
855
860
865
Fig. 4: Accuracy and resolution test of an unembedded prototype digital reference pressure. After sensor calibration, the pressure curve P(x) was found to be: P = 1.70 · 10−7 · x3 − 0.94 · 10−3 · x2 + 2.20 · x + 470 (1) In this equation x is the digital CDC read out obtained either in its 12 bit form from serial communication, or its 5 bit read out from the EEPROM. The first is only available during testing, needing a wired connection, which will not be available in the final implementation. Nevertheless, these 12 bit results are used to obtain a clear view of the potential pressure resolution. Using equation 1, the system was put through the following test:
1. The sample was kept at a pressure of ∼ 1016 mbar for a duration of 100 seconds. 2. Pressure was increased to ∼ 1300 mbar in 180 seconds. 3. Pressure was decreased to atmospheric pressure (∼ 1005 mbar) in 70 seconds by opening a valve. 4. Pressure was increased to ∼ 1036 mbar. 5. Pressure was decreased in steps of 0.1 mbar by moving a membrane within the Druck DPI 600 to accurately assess the resolution of the sample.
IFMBE Proceedings Vol. 29
520
P. Jourand, J. Coosemans, and R. Puers
The result is shown in Fig. 4. During this test, the current drain was monitored and found to be 320μA. The test reveals a resolution of 0.7 mbar. This resolution however is not further exploited in order to save memory space. Where only a window of five out of the twelve bits is used to indicate the pressure, the resolution of the final device can than be calculated by a multiplication with 25 since the 5 least significant bits and the 2 most significant bits, will be discarded. This yields a resolution of 22mbar, which is acceptable for the envisioned tests. An encapsulated sample was also tested, revealing a decrease in sensitivity of ∼ 40% resulting in a resolution of 1 mbar and 32 mbar for respectively the 12 bit and 5 bit results.
IV. C ONCLUSION Three devices produced at ESAT-MICAS, have been reviewed showing a clear evolution in functionality. One of the devices presents a major step forward towards the idea of a ”urological pacemaker”, a long term bladder pressure monitoring device with inductive powering and bidirectional communication. The other two devices are diagnostic tools for short term implementation. The first one was developed over 20 years ago. The second takes advantage of capacitive based pressure sensors and uses a silicone embedding as a protective encapsulation. Using 5 bit pressure values at a sampling rate of 2 Hz to conserve both power consumption as well as memory, a pressure resolution of ∼ 32 mbar is achieved. Where this does not meet the standards set for urological tools yet, should the available power and memory be increased, both can be augmented by reprogramming the microcontroller. In such a case, the full 12 bit pressure readings are available presenting an accuracy of 0.7 mbar. With the current parameters, the device consumes on avarage 320μA at 3 V, logging pressure data for a period of over 24 hours.
R EFERENCES 1. Sh¨afer W, Abrams P, Liao L, et al. Good Urodynamic Practices: Uroflowmetry, Filling Cystometry, and Pressure-Flow Studies Neurology and Urodynamics. 2002;21:261–274. 2. Puers R, Sansen W, Vereecken R. Development Considerations of a Micropower Control Chip and Ultraminiature Hybrid for Bladder Pressure Telemetry in Biotelemetry;VIII(Dubrovnik, Yugoslavia):328–332 1984. 3. Puers R, Sansen W, Vereecken R. Realisation of a telemetry capsule for cystometry in in IEEE Frontiers of Engeneering and Computing in Health Care 1984:711–714 1984. 4. Coosemans J, Puers R.. An Autonomous Bladder Pressure Monitoring System Sensors and Actuators A:Physical. 2005;123–124:155–161. 5. Coosemans J. Wireless and Battery-less Medical Monitoring Devices. Katholieke Universiteit Leuven 2008. 6. Sansen W, Vereecken R, Puers R, Folens G, Nuland T Van. A closed loop system to control the bladder function in In Proceedings of the Ninth Annual Conference of the IEEE Engineering in Medicine and Biology Society(Boston, USA):1149–1150 1987. 7. Lenaerts B, Puers R. An inductive power link for a wireless endoscope Biosensors & Bioelectronics. 2007;22:1390–1395. 8. Carta R, Tortora G, Thon´e J, et al. Wireless powering for a selfpropelled and steerable endoscopic capsule for stomach inspection Biosensors & Bioelectronics. 2009;25:845–851. 9. Carta R, Jourand P, Hermans B, et al. Design and implementation of advanced systems in a flexible-stretchable technology for biomedical applications Sensors and Actuators A: Physical. 2009;156:79–87. 10. Blake C, Abrams P. Non invasive techniques for the measurement of isovolumetric bladder pressure Journal of Urology. 2004;171:12–19. 11. Harding C K, Robson W, Drinnan M J, Ramsden P D, Griffiths C, Pickard R S. Variation in Invasive and Noninvasive Measurements of Isovolumetric Bladder Pressure and Categorization of Obstruction According to Bladder Volume Journal of Urology. 2006;176:172–176. 12. Tan R, McClure T, Lin C K, et al. Development of a fully implantable wireless pressure monitoring system Biomedical Microdevices. 2009;11:259–264. 13. Wang C-C, Huang C-C, Liou J-S, et al. A Mini-Invasive Long-Term Bladder Urine Pressure Measurement ASIC and System IEEE Transactions on Biomedical Circuits and Systems. 2008;2:44–49. 14. Jourand P, Puers R. An Autonomous, Capacitive Sensor Based and Battery Powered Internal Bladder Pressure Monitoring System in Proceedings of the Eurosensors XXIII Conference, Procedia Chemistry;1(Lausanne, Switserland):1263–1266 2009. 15. ApexMed at http://www.apexmed.eu/ (February 2010) 16. Wein A J, Kavoussi L R, Novick A C, Partin A W, Peters C A. Campbell-Walsh Urology, Edition 9. Saunders 2009.
ACKNOWLEDGEMENTS This research has been developed in the frame of BIOFLEX, an IWT funded project, contract number IWT040101. Special thanks to Michel De Cooman for producing the flex prints used in the testing of the devices, the Centre for Microsystems and Technology Gent for moulding the devices and Alexander Thomas from ESAT-VISICS for editing the ifmbe LATEX style-file to handle multiple affiliations.
Author: Philippe Jourand Institute: Katholieke Universiteit Leuven, ESAT-MICAS Street: Kasteelpark Arenberg 10 City: B-3000 Leuven Country: Belgium Email:
[email protected] IFMBE Proceedings Vol. 29
Including the effect of the thermal wave in theoretical modeling for radiofrequency ablation J.A. López Molina1, M.J. Rivera1, M. Trujillo1, V. Romero-García2 and E.J. Berjano3 1
Departamento de Matemática Aplicada, Instituto de Matemática Pura y Aplicada, Universidad Politécnica de Valencia, Valencia, Spain 2 Centro de Tecnologías Físicas: Acústica, Universidad Politécnica de Valencia, Valencia, Spain 3 Departamento de Ingeniería Electrónica, Universidad Politécnica de Valencia, Valencia, Spain
Abstract— In this paper we outline our main findings about the differences between the use of the Bioheat Equation and the Hyperbolic Bioheat Equation in theoretical models for RF ablation. At the moment, we have been working on the analytical approach to solve both equations, but more recently, we have considered numerical models based on the Finite Element Method (FEM). As a first step to use FEM, we conducted a comparative study between the temperature profiles obtained from the analytical solutions and those obtained from FEM. Keywords— Ablation, COMSOL, Finite Element Method, theoretical model, radiofrequency ablation. I. INTRODUCTION
Radiofrequency (RF) heating of biological tissues is currently employed in many surgical and therapeutic procedures such as the elimination of cardiac arrhythmias, the destruction of tumors, the treatment of gastroesophageal reflux disease, and the heating of the cornea for refractive surgery. In order to investigate and develop new RF ablation techniques, besides understanding the complex electrical and thermal phenomena involved in the heating process, numerous theoretical models have been employed [1]. To date, all these models have employed the Bioheat Equation (BE) proposed by Pennes [2], in which the heat conduction term is based on Fourier’s theory (i.e. they have employed a parabolic heat transfer equation). Therefore, it related to r heat flux ( q ) in the following way:
r r r r q (r , t ) = −k∇T (r , t )
(1) r where k is the thermal conductivity and T (r , t ) the tem-
r
perature at point r at time t. This approach assumes an infinite thermal energy propagation speed, and although it might be suitable for most RF ablation procedures, it has been suggested that under certain conditions (such as very short heating times), a non-Fourier model should be considered by means of the Hyperbolic Bioheat Equation (HBE), i.e. considering a thermal relaxation time (τ) for the tissue ≠0 [3]. It is known that heat is always found to propagate at
a finite speed [4], and in fact Cattaneo [5] and Vernotte [6] simultaneously suggested a modified heat flux model in the form: r r r r (2) q (r , t + τ ) = −k∇T (r , t ) where τ is the thermal relaxation time of the biological tissue. Equation (2) assumes that the effect (heat flux) and the cause (temperature gradient) occur at different times and that the delay between heat flux and temperature gradient is τ. The particular case of considering τ = 0 obviously corresponds to the BE. In order to study how the temperature profiles could be altered when HBE is considered in place of BE, we have conducted different theoretical studies based on onedimensional analytical models [7-9]. In this models, we solved both BE and HBE under different circumstances. Obviously, since that the analytical approach does not allow easily to consider complex geometries or to solve non-linear equations, recently we are using a complementary approach based on numerical techniques, specifically the Finite Element Method (FEM). In this paper we summarize the main findings using the analytical approach and present new results about the use of COMSOL Multiphysics to solve the HBE in models for RF ablation. II. ANALITYCAL APPROACH
For this approach, we have used model geometry very simple. Briefly, we considered a r0 radius spherical electrode completely imbedded in and in close contact with the biological tissue (see Fig. 1), which had an infinite dimension. This model presented radial symmetry and a onedimensional approach was possible. Regarding the electrical problem, we always modeled a constant-power protocol, i.e. the source term for the BE and HBE (i.e. the Joule heat produced per unit volume of tissue, Q(r,t)) was always: P ⋅ r0 (3) Q (r , t ) = H (t ) 4 ⋅π ⋅ r 4
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 521–524, 2010. www.springerlink.com
522
J.A.L. Molina et al.
where P is the total applied power (W), r0 the electrode radius, and H(t) is the Heaviside function. Although this temporal function have not been included in the previous study by Erez and Shitzer [10], later it was crucial to study the pulsed protocol in RF ablation [11].
To set the boundary condition in r = r0 we adopted a simplification assuming the thermal conductivity of the electrode to be much larger than that of the tissue (i.e. assuming that the boundary condition at the interface between electrode and tissue is mainly governed by the thermal inertia of the electrode). This obviously modeled a dry electrode. Other thermal boundary conditions should be considered for the case of internally cooled electrodes [13]. III. NUMERICAL APPROACH
Fig. 1 Schematic diagram of the model geometry. A spherical electrode (grey circle) of radius r0 is completely imbedded and in close contact with the biological tissue, which has an infinite dimension. As a result, the model presented a radial symmetry, and a one-dimensional approach is possible (dimensional variable is r).
The HBE was obtained by combining the energy equation:
∂T (r , t ) (4) ∂t where ρ is the density and c the specific heat, with the heat transfer model proposed by Özişik and Tzou’s [12]: ∂q (r , t ) (5) q(r , t ) + τ = −k∇T (r , t ) ∂t The result was: 1 ∂T ( r , t ) ∂ 2T ( r , t ) )= − ∆T ( r , t ) + ( +τ α ∂t ∂ 2t (6) 1 ∂Q(r , t ) (Q(r , t ) + τ ) k ∂t where α is the thermal diffusivity. Finally, we combined (3) and (6) to obtain the HBE: ∂ 2T (r , t ) 2 ∂T (r , t ) ∂T (r , t ) −α( + )+ r ∂r ∂t ∂r 2 (7) Pα r0 ∂T (r , t ) +τ = ( H (t ) + τ δ (t )) ∂t 4π k r 4 where δ(t) is Dirac’s function. It is important to emphasize that in all these cases both BE and HBE did not consider the blood perfusion term, and hence they are only useful to model RF ablation in non-perfused tissue (e.g. cornea) or in those tissue where this term has been suggested to be negligible (e.g. cardiac tissue far from large vessels). − ∇q ( r , t ) + Q ( r , t ) = ρ c
The majority of heat transfer problems of real situations involve complex geometries, are non-linear problems or their initial and boundary conditions lead us to use numerical methods to solve them. This is absolutely true in RF ablation. Some widespread numerical methods to solve this kind of problems are the Finite Element Method (FEM) and the Finite Differences Method. There is abundant available software for building models, solving them by the mentioned methods and post-processing the results. We have chosen COMSOL Multiphysics (Burlington, MA, USA), which has been broadly employed in the study of the RF ablation of biological tissues. However, all of those previous studies considered the BE. In this respect, our recent objective has focused on the validation of COMSOL Multiphysics for using the HBE in obtaining the temperature distribution during RF ablation. This issue is especially important by taking into account the cuspidal-type singularities found in the analytical solutions of the HBE, which are materialized as a temperature peak traveling through the medium at a finite speed [7]. In other words, it is necessary to know if this behavior will be accurately modeled by numerical methods in general, and by COMSOL in particular. For this reason, we build with COMSOL the same onedimensional model previously solved by analytical methods, and then we obtained the numerical solution. Our idea was to validate COMSOL by comparing the numerical and analytical solution. We used COMSOL Multiphysics software version 3.2b, which can virtually model and solve any physical phenomenon which can be described with Partial Differential Equations (PDE) using the FEM. COMSOL presents several models to solve a wide range of PDEs. In a previous modeling study we tried to validate COMSOL by using a 2D model, but we found important differences (up to 13ºC after 60 s) between analytical and numerical solutions [14]. Now, we have chosen a one-dimensional problem. We used the automatic mesh generated by COMSOL, and for this reason we conducted a sensibility analysis to check that a more refinished meshes do not produce results closer to the analytical ones. The control parameter used to
IFMBE Proceedings Vol. 29
Including the Effect of the Thermal Wave in Theoretical Modeling for Radiofrequency Ablation
conduct this sensibility analysis was the maximum temperature reached at the interface electrode-tissue after 60 s. In order to compare analytical and numerical solution we plotted the progress of temperature from each solution. In the case of the analytical solution we used the software Mathematica 7.0 software (Wolfram Research, Champaign, IL, USA). To make graphics of the numerical solution we used the post-processing option of COMSOL. In order to plot the results we particularized the solutions for a specific case. As biological tissue we chose the liver with the following characteristics: density ρ of 1060 kg/m3, specific heat c of 3600 J/kg⋅K and thermal conductivity k of 0.502 W/m⋅K. The electrode characteristics were the density of 21500 kg/m3 and the specific heat c of 132 J/kg⋅K. The initial temperature of tissue was 37ºC. The applied power was of P=1 W. Moreover, we included the term of blood perfusion both in BE and HBE. These numerical solutions were compared to those obtained analytically (data non published yet).
523
explained due to the fact that when using HBE a period of time is needed for heat to travel to a particular location inside the tissue. When these conclusions were particularized for specific tissues, once more the differences between BE and HBE temperature profiles were greater for lower times and shorter distances. For this reason, our results suggested that the HBE should be considered in the case of RF heating of the cornea (heating time 0.6 s), and for short time ablation in cardiac tissue (less than 30 s) [8]. When BE and HBE were analytically solved for a pulsed application of RF power, we found three typical waveforms for the temperature progress depending on the relations between the duration of the RF pulse and the λ ( ρ − 1) being λ the dimensionless thermal relaxation time and ρ the dimensionless position. In BHE solution we also observed that the temperature at any location is the result of the overlapping of different heat sources delayed different durations (each heat source being produced by an RF pulse of limitless duration).
IV. RESULTS AND DISCUSSION
Regarding the analytical solutions, we found, from a mathematical point of view, that the HBE solution shows cuspidal-type singularities in the form of a temperature peak traveling through the medium at finite speed (see Fig. 2). This peak arises at the electrode surface, and clearly reflects the wave nature of the thermal problem. In [11] we tried to provide an explanation about this behavior which is based on the interaction of forward and reverse thermal waves.
Fig. 2 Dimensionless temperature progress during the RF heating of the biological tissue at three normalized locations: on the electrode surface ρ=1, and at ρ=1.7 and 2.7. The thermal relaxation time of the biological tissue was 16 s. Two solutions are shown from different equations: BE (dashed line) and HBE (solid line).
At the beginning of heating (i.e. when the considered time was comparable to or shorter than the thermal relaxation time), BE provided temperature values lower than those provided by HBE. In general, the speed of temperature change in the case of HBE was slower than BE. This can be
Fig. 3. Progress of dimensionless temperature of the HBE for three conditions. (A) Case in which duration of the RF pulse is higher than λ ( ρ − 1) . (B) Case for an RF pulse lower than
λ ( ρ − 1) . (C) Transitional case λ ( ρ − 1) . The solid lines correspond with the temperature from HBE and the dashed line with the temperature from BE.
where the duration of the RF pulse is equal to
IFMBE Proceedings Vol. 29
524
J.A.L. Molina et al.
Regarding the numerical results, Figure 4 shows the temperature progress for HBE and for two values of thermal relaxation time (1 and 16 s) and for two blood perfusion conditions (without perfusion ω=0, and perfusion ω=0.01 1/s). These results were almost coincident with those obtained from analytical approach, both using BE and HBE, which suggests that COMSOL can be a suitable tool to model the heating of biological tissues using BE and HBE. Now, future work will be conducted to implement theoretical models based on FEM (COMSOL) with more realistic geometries.
ACKNOWLEDGMENT This work received financial support from the Spanish “Plan Nacional de Investigation Científica, Desarrollo e Innovación Tecnológica del Ministerio de Educación y Ciencia” (TEC2008-01369/TEC) and FEDER Projects MTM2007-64222 and MAT2009-09438.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8.
Fig. 4. Temperature progress obtained from COMSOL during 60 s of RF ablation using the HBE and for two values of thermal relaxation time (1 and 16 s) and for two blood perfusion conditions (without perfusion ω=0, and perfusion ω=0.01 1/s). The plots correspond with a location r=2r0.
V. CONCLUSION
9. 10. 11.
In this paper, we have outlined our main findings about the differences between the BE and HBE models for RF ablation. These differences encourage the use of the HBE approach for processes in which great amounts of heat are transferred to any material in very short times, e.g. RF heating in the cornea. At the moment, we have been working on the analytical modeling of the HBE, but more recently, we have considered numerical models based on FEM. As the first step to use FEM should be the validation, we have conducted a comparative study between the temperature profiles obtained from the analytical solutions and those obtained from FEM.
12. 13.
14.
Berjano E J (2006) Theoretical modeling for radiofrequency ablation: state-of-the-art and challenges for the future. Biomed Eng Online 5 24 Pennes H H (1998) Analysis of tissue and arterial blood temperatures in the resting human forearm. 1948. J Appl Physiol 85:5–34 Liu J, Chen X, Xu L X (1999) New thermal wave aspects on burn evaluation of skin subjected to instantaneous heating. IEEE Trans Biomed Eng 46:420–428 Hader MA, Al-Nimr MA, Abu Nabah BA (2002) The dual-phase-lag heat conduction model in thin slabs under a fluctuating volumetric thermal disturbance. Int J Thermophysics. 23:1669–1680 Catteneo C (1958) Sur une forme de l’équation de la chaleur éliminant le paradoxe d’une propagation instantaneé. Compes Rendus 247:431–433 Vernotte P (1958) Les paradoxes de la théorie continue de l´équation de la chaleur. Comptes Rendus 246: 3154–3155 López-Molina JA, Rivera MJ, Trujillo M et al (2008) Effect of the thermal wave in radiofrequency ablation modeling: an analytical study. Phys Med Biol 53:1447–1462 López-Molina JA, Rivera MJ, Trujillo M et al (2008) Assessment of hyperbolic heat transfer equation in theoretical modeling for radiofrequency heating techniques. Open Biomed Eng J 10:2:22–27 Tung MM, Trujillo M, López-Molina JA et al (2009) Modeling the heating of biological tissue based on the hyperbolic heat transfer equation. Mathematical and Computer Modelling 50:665–672 Erez A, Shitzer A (1980) Controlled destruction and temperature distributions in biological tissues subjected to monoactive electrocoagulation. J Biomech Eng 102:42–49 López Molina JA, Rivera MJ, Trujillo M et al (2009) Thermal modeling for pulsed radiofrequency ablation: analytical study based on hyperbolic heat conduction. Med Phys 36:1112–1119 Özişik MN, Tzou DT (1994) On the wave theory in heat conduction. ASME J Heat Transfer 116:526–535 Rivera MJ, Molina JA, Trujillo M et al (2009). Theoretical modeling of RF ablation with internally cooled electrodes: comparative study of different thermal boundary conditions at the electrode-tissue interface. Math Biosci Eng 6:611–627 Romero-García V, Trujillo M, Rivera MJ et al (2009) Hyperbolic Heat Transfer Equation for Radiofrequency Heating: Comparison between Analytical and COMSOL solutions. Proceedings of the COMSOL Conference 2009 Milan. Author: Enrique J Berjano Institute: Departamento de Ingeniería Electrónica (7F) Street: Camino de Vera s/n City: Valencia 46022 Country: spain Email:
[email protected] IFMBE Proceedings Vol. 29
Textile Integrated Monitoring System for Breathing Rhythm of Infants H. De Clercq1 , P. Jourand1 and R. Puers1 1
Katholieke Universiteit Leuven, ESAT-MICAS, Kasteelpark Arenberg 10, 3001 Heverlee, Belgium
Abstract— Monitoring the breathing rhythm of infants during sleep can be life saving. But today, most monitoring systems lack patient comfort. In this paper an innovative biomedical monitoring system with textile integrated sensors is developed and tested. Monitoring breathing activity is used as a case study, yet, the platform is extendable by an architecture that can contain up to twenty modular sensor channels, divided in several sensor islands. It is therefore useful for all kinds of (biomedical) applications. Flexible carriers for the electronic circuit lead to better textile integration and more wearing comfort. Quantification of breathing rhythm and volume is performed by accelerometers. These breathing signals are calculated on the basis of the technique that the angle of the gravitation vector in the coordinate systems of the accelerometers changes, because of movement of the abdomen. Differential use of two accelerometers makes this measurement insensitive for movement and posture. The comparison of these signals with a spirometer yields promising results. Keywords— Textile integration, Breathing measurement, Accelerometers, Infant monitoring, Home monitoring
I. I NTRODUCTION In the era of ubiquitous miniature intelligent systems, an ever increasing amount of portable electronics is carried about by people. Textile has the potential to integrate most of these voluminous devices into one coherent wearable system. For biomedical applications in particular, measurement devices can ”seamlessly” be integrated into textile to enhance both patient comfort and ease of use in everyday monitoring. Special precautions need to be taken during the design of such systems, prerequisite being of course absolute patient safety. This requirement includes guaranteed electrical, allergen and toxic safety, but also avoiding manual configuration during set-up to reduce human error. Additionally, the patient should not be hampered during his/her normal activities by neither cables, nor rigid or voluminous electronics. Just-in-time processing of signals ought to provide the user and/or physician with the appropriate information at the right moment, using an easily interpretable interface. An emerging market for textileintegrated electronics is the monitoring of vital parameters during sleep. Particularly Sudden Infant Death Syndrome (SIDS), going together with deceleration of breathing and low oxy-
genation of body tissues, can severely threaten infants’ lives. Although deceases because of SIDS have consistently been decreasing during the past decennia through improved knowledge about its causes, an important risk group, as well as major concern among infants’ parents, remains. A low-cost, reliable and easy-to-use system for everyday monitoring is designed, to measure breathing activity. After all, continuous monitoring seems to be the most reliable detection technique for the symptoms of SIDS. The whole system is powered by a flexible lithium-polymer battery (3.7V ). Integration with an existing inductive coupling [1] is envisaged as a next step.
II. ACCELEROMETER - BASED ESTIMATION OF RESPIRATORY ACTIVITY
An accurate measure of breathing rate and, if possible, breathing volume is required for e.g. assessment of the symptoms of SIDS in infants. In clinical environment, this respiratory activity is mostly measured with a spirometer. This device consists of a mouthpiece through which the patient has to breath. Inside a propeller rotates in function of the air flow. This technique is very precise in both volume as rhythm measurement and is therefore used as reference. A spirometer is however not usable for long term measurements, nor is it very comfortable for the patient. Hence other monitoring techniques, more suitable for home monitoring were introduced, such as impedance variation measurement [2] and respiratory inductive plethysmography (RIP) [3]. Although these methods are more comfortable, they still deal with disadvantageous like skin irritation and short life span. This paper discusses a recent technique with more promising results regarding these issues. A. Two techniques based upon accelerometers A technique based upon accelerometers to estimate the breathing waveform was refined from [4] for increased robustness to motion and changes of posture. Dual-axis accelerometers (ADIS16003) were placed in the transversal body plane by bending the flexible substrate 90 degrees up from the abdomen wall (Fig 1) to register only relevant signals, since the third axis would normally be horizontal during sleep. Table 1 contains the most important specifications of these accelerometers.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 525–528, 2010. www.springerlink.com
526
H. De Clercq, P. Jourand, and R. Puers x2
x2 x1 y1
y2
x1
α2
X
α1 Δα
y2
y1
Y (=g)
Y (=g)
Fig. 2: Inclination variation technique for registration of respiration using two accelerometers.
Fig. 1: The accelerometers mounted on a flexible pcb. Table 1: Specifications of accelerometer ADIS16003
Measure range Cross sensitivity Noise density Resolution Non-linearity Offset Power
Value ±1, 7 5 110 10 < 2, 5 140 5, 55
x (t)2 + y (t)2
with: magnitude (t) =
Unit g % √ μg/ Hz rms bits % LSB mW (typ.@3, 7V )
B. Results & Discussion Results for both assumptions, together with the spirometer signal, are shown (Fig 3). inclination [degrees]
Parameter
magnitude was derived as the modulus of the sensed accelerations in the x- and y-direction. The same placement of accelerometers was used, summing both magnitude variations to increase sensitivity. The magnitude is derived
-15 -25 -35
The second technique is formulated using the rhythmical motion of the abdomen wall due to breathing. This causes the accelerometer to sense magnitude variations, added to the dominant DC signal due to gravitation. The
20
30
40
50
60
70
80
90
100
110
80
90
100
110
80
90
100
110
time [s]
0.5
0.3 10
spirometer signal
The first technique is based upon inclination variations when the accelerometers are placed more laterally (about 8 cm left/right from the umbilicus), since the abdomen, during inhalation, expands outwards, slightly tilting both y-axes inwards (Fig 2). The inclination of the accelerometers was derived from the orientation of the gravitation vector in the xy-plane of the accelerometer. One accelerometer was placed on each side of the umbilicus, resulting in a common-mode signal (sum of the sensed inclination corners) because of the patient’s posture, and a differential signal (difference of the sensed inclination corners) because of the patient’s breathing movements. The angle is derived with: angle (t) = y(t) 180 π (sign (x (t)) − sign (|x (t)|) + 1) arctan − π 2 x(t)
magnitude [g]
10
Two techniques were tested and compared to a Jaeger PulmAssist spirometer,used as a reference for the breathing measurements, evaluating their ability to extract the breathing rhythm and amplitude in an accurate and reliable way.
20
30
40
50
60
70
time [s] 20 0 -20 10
20
30
40
50
60
70
time [s]
Fig. 3: A comparison of 3 methods for measuring respiration activity including detection of breathing cycles.
Since during relaxed breathing, respiration is mostly coordinated by the diaphragm, measurements are performed on the abdomen. Necessary information is extracted to relate symptoms of SIDS to the patient’s posture and movements, for it is well known that prone sleep position leads to increased risk for SIDS [5]. To extract the breathing rhythm and amplitude, the robust, yet simple respiration rate estimation algorithm proposed by Lukocius [6] was adopted. A differential moving window is scanning the signal for changes in slope, after which maxima and minima in between are extracted (Fig 4). A threshold determines whether the detected peak originates from respiration, and is updated adaptively, based upon the information of the last 8 respiration cycles. From the respective location and amplitude of the extrema, breathing rate and
IFMBE Proceedings Vol. 29
Textile Integrated Monitoring System for Breathing Rhythm of Infants
38
breathing rhythm (bpm)
volume were calculated and plotted in real-time. The same processing sequence was applied to all respiration signals.
527
34
30
26
22 0
Mean relative error = 6.8% 20
40
60
80
100
120
time(s)
Fig. 6: Comparison of the respiration frequency between spirometer(red) and magnitude method(blue). 2.6
Fig. 4: Peak detection algorithm proposed by Lukocius [6]. Both techniques for the accelerometer were tested and compared vis-`a-vis the spirometer, during two minute tests where the patient was asked to vary his/her breathing amplitude and frequency in a random way. In supine lying position, the mean respiration rate error for the inclination technique was 3.8% (Fig 5), while the magnitude assumption showed an error of 6.8% (Fig 6). On the other hand, the respective breathing amplitudes were compared in a scatter plot. Here, the R2 -value of the correlation again shows better results for the first assumption (0.85 (Fig 7) compared to 0.46 for the second assumption (Fig 8)). With the patient lying on the side, the accuracy of rate detection and correlation of the amplitudes remained quite constant for the first assumption, but declined significantly for the second.
amplitude accelerometer (normalised)
2.4 2.2 2 1.8 1.6 1.4 1.2 1 0.8
trend line: y = 0.37 + 0.64 R2-value = 0.85
0.5
1
1.5
2
2.5
3
3.5
amplitude spirometer (normalised)
Fig. 7: Correlation between amplitude of angle method and amplitude of
2.4 2.2
30
25
Mean relative error = 3.8% 20 0
20
40
60
80
100
120
time [s]
Fig. 5: Comparison of the respiration frequency between spirometer(red) and angle method(blue).
amplitude accelerometer (normalised)
breathing rhythm [bpm]
spirometer.
2 1.8 1.6 1.4 1.2 1 0.8 0.6
During change of lying position, the peak detection algorithm showed a robust detection, although movements during transition were not fully canceled. Results are shown in Fig. 9, where the patient’s posture (delayed due to processing) was added.
2
R -value = 0.46 0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
amplitude accelerometer (normalised)
Fig. 8: Correlation between amplitude of magnitude method and amplitude
IFMBE Proceedings Vol. 29
of spirometer.
H. De Clercq, P. Jourand, and R. Puers
breathing signal (degrees)
80 60 40 20 0 -20
ECG
4000
3600
60
0
10
20
30 time (s)
40
50
60
Fig. 9: Respiration and posture signals derived from the accelerometers. The patient changes sides after 20 and 40 seconds.
III. A POLYVALENT ARCHITECTURE Because of the wide range of applications in textile integrated electronics [7], the system consists of modular sensor islands, which can easily be changed with other islands according to the application (Fig 10).
Sensor island 3
SPI
SPI
Sensor island 2
Battery 3.7 V
Wireless data transmission Nordic
Master island
SPI
Receive island
500
1000
1500
2000
time (ms)
2500
breathing frequency
50
ï
ï
ï 0
500
1000
1500
2000
time (ms)
80 60 40
0
500
1000
0
500
1000
0
500
1000
1500
2000
2500
1500
2000
2500
1500
2000
2500
time (ms)
32 30 28 26
2500
time (ms)
1 0.8
1905
pulse oximeter
40
magnitude respiration
30
free space
posture (degrees)
20
80 60 40 20 0
Sensor island 1
100
3800
3400
10
heart rrhythm
528
0.6
1900
0.4
1895
0.2 0
0
500
1000
1500
2000
time (ms)
2500
1890
time (ms)
Fig. 11: Real-time visualization of measured signals. Detected signal peaks are indicated with vertical lines.
IV. C ONCLUSION Two methods with accelerometers for breathing monitoring are compared. The inclination method gives noticeable better results, especially concerning posture and movement dependency. A comparision is made with a spirometer, resulting in a very small error. Also the comparision of breathing volume extraction is promising. Finally, a polyvalent system architecture is described.
RS232
Sensor island 4
SPI
SPI
R EFERENCES Sensor island 5
Fig. 10: The system architecture that can process up to 20 sensor signals, divided on 5 islands, and communicate wirelessly with a processing unit. A slave island contains up to four sensors, sensor processing electronics and an A/D-converter (MCP3204). The obtained digital signals (twelve bit quantization) are sent to the master island, using SPI. These bits give a resolution of 0.9mV , which also stipulates the maximum sensitivity of the analog sensor outputs. The master island (Fig 1) contains a microcontroller (PIC16F687) and a wireless transceiver (nRF24L01), that sends the data through wireless transmission to a receive island preceding the processing unit. The bandwidth of 1000 Hz for each sensor is more than satisfactory for most biomedical signals, but can be improved if necessary by making the system dedicated to one application and converting the processing unit into an ASIC. A real-time visualisation of some measured signals is shown (Fig 11). In this case these signals are an ECG, a respiration signal, measured by the accelerometes and a pulse oxymeter signal.
1. J. Coosemans, B. Hermans, and R. Puers, “Integrating wireless ecg monitoring in textiles,” Sensors and Actuators A, vol. 130-131, pp. 48–53, 2006. 2. C. S. Poon, Y. C. Chung, T. T. C. Choy, and J. Pang, “Evaluation of two noninvasive techniques for exercise ventilatory measurements,” Engineering in Medicine and Biology Society, 1988. 3. M. N. Fiamma, Z. Samara, T. S. P. Baconnier, and C. Straus, “Respiratory inductive plethysmography to assess respiratory variability and complexity in humans,” Respiratory Physiology and Neurobiology, vol. 156(2), pp. 234–239, May 2007. 4. P. D. Hung, S. Bonnet, R. Guillemaud, E. Castelli, and P. Yen, “Estimation of respiratory waveform using an accelerometer,” in Biomedical Imaging: From Nano to Macro, vol. 5, June 2008, pp. 1493–1496. 5. M. Willinger, H. J. Hoffman, and R. B. Hartford, “Infant sleep position and risk for sudden infant death syndrome,” National Institutes of Health, vol. 22, p. 42, November 1993. 6. R. Lukocius, J. A. Virbalis, J. Daunoras, and A. Vegys, “The respiration rate estimation method based on the signal maximums and minimums detection and the signal amplitude evaluation,” Electronics and Electrical Engineering, vol. 8, pp. 51–54, 2008. 7. R. Carta, P. Jourand, B. Hermans, J. Thone, D. Brosteaux, T. Vervust, F. Bossuyt, F. Axisa, J. Vanfleteren, and R. Puers, “Design and implementation of advanced systems in a flexible-stretchable technology for biomedical applications,” Sensors and Actuators A, vol. 156, pp. 79–87, November 2006.
Email:
[email protected] IFMBE Proceedings Vol. 29
Comparison between VHDL-AMS and PSPICE modeling of ultrasound measurement system for biological medium N. Aouzale1, A. Chitnalah 1, H. Jakjoud 1, D. Kourtiche 2, M. Nadi 2 1
2
L.S.E.T, Université CADI AYYAD FST BP 549, 40000 Gueliz Marrakech Maroc. L.I.E.N, Nancy Université, Faculté des Sciences et techniques, BP 70239, 54506 Vandœuvre, France
Abstract— Piezoelectric materials are commonly used in many applications. Different approaches were developped to predict the piezoelectric transducer behaviour. Among them, the resolution of piezoelectric equations by numerical methods is currently used. Another method is based on the equivalent electrical circuit simulation : Pspice or VHDL– AMS tools. This paper proposes a comparison between VHDL-AMS and Pspice models for a pulse echo ultrasonic system. The simulation is based on the Redwood model and its parameters are deduced from the transducer acoustical characteristics. The electrical behaviour of the proposed model is in very good agreement with the real system behaviour. Keywords— Ultrasound, VHDL-AMS, Modeling, Redwood, Measurement I. INTRODUCTION
Ultrasound systems are widely used. They find many applications in engineering, medicine, biology, and other areas [1]. Modelling and simulation of such systems is a difficult task due to the presence of multi-physics effects and their interactions. VHDL-AMS language is appropriate for ultrasonic systems methodology conception because it could take into account all the transducer environment including microelectronic stimulation and acoustic load. The use of behavioural models in simulation simplify physics and explore interactions between different domains in a reasonable amount of time . This paper presents a method for multi-domain behavioural modelling of ultrasound measurement system. We validated this methodology through a study cases in linear ultrasound measurement in which the ultrasound transducer model takes great importance. To perform the implementation, a virtual-prototyping environment, ADVance® MS (ADMS) tool from Mentor Graphics is used. This environment provides multi-level model integration required for real systems design and analysis. The obtained results are compared to those obtained with Pspice simulation and by measurement.
II. TRANSDUCER MODELLING
Many electrical equivalent circuits have been made to represent ultrasonic transducer. The model of Mason represents the transducer in the form of an electrical equivalent circuit where the transducer acoustic port is represented by localized elements connected to the electrical port of the transducer by an ideal transformer. This model presents some disadvantages such as the negative capacitor and it can not model multilayered transducers. Redwood improved this electromechanical model by incorporating a transmission line, making possible to extract useful information on the temporal response of the piezoelectric component [2, 3]. Figure 1 present the transducer and his Redwood’s transmission line version of Mason’s equivalent circuit. The multiple parameters appearing in this model are as follows : Q1 and Q2 are the acoustic particles velocities at the front and the back faces of the disc, F1 and F2 are the acoustic forces at the transducer faces. T is an ideal electro-acoustic transformer with a ratio h33.C0 with h33, a piezoelectric stiffness constant for the ceramic and C0, the clamped capacitance. The Redwood model is divided in two parts, first is the electrical port which is composed by the capacitors C0 and –C0 that represent the capacitance motional effect, the electrical port is connected to a resistance R and a voltage source noted V , the second part is composed by the two acoustic ports,. Piezoceramic layer is assimilated to a propagation line characterized by its thickness e, acoustic impedance of the ceramic Zt = U. c0 A, with c0 the particle velocity, ȡ the material density and A the area of the ceramic. One branch of the piezoceramic layer is in contact with the back medium (Zback) and the other is in contact with the propagation medium (Zfront). Transducer modelling with VHDL-AMS language is based on writing of the different equations of the Redwood scheme elements. The multiple parameters appearing in this model are defined below : v1 and v2 (m/s) are the acoustic particles velocities at the front and the back faces of the disk, the parameter k is
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 529–532, 2010. www.springerlink.com
530
N. Aouzale et al.
the wave number for the piezoelectric ceramic, k = Z /vD, where Z the angular frequency (rad/s), vD (m/s) is the wave speed of compressional waves in the D C 33 / U ҏ in terms of piezoelectric plate given by Q D the elastic constant of the piezelectric ceramic, C33D (N/m2) , at constant electric flux density, and Uҏ (kg/m2), the density of the ceramic. h33 (V/m) , is the piezoelectric stiffness constant for the ceramic, and C0, the clamped capacitance of the plate. C0 is given by C0 A/ E33s d where A (m2) is the area of the ceramic and E 33s (m/F) is the dielectric impermeability of the ceramic at constant strain, and d (m) is the ceramic thickness. The quantity Zt = A U vD (Rayl) is the plane wave acoustic impedance of the piezoelectric ceramic, while ZB (Rayl) ҏis the corresponding acoustic impedance of the backing, which is a function of frequency [4,5].
III. ULTRASOUND PULSE-ECHO SYSTEM
The most common configuration of an ultrasonic system widely used for acoustical measurements is shown on the figure 3. Transmitter-receiver device
Propagating medium
Reflector
d/2 Oscilloscope display
Transducer
Fig. 3 Bloc diagram of experimental setup
d
Area A
F1
F2
Q1
Q2 z
I3 V3 Fig.1 Model (1-D) for electrical and acoustical parameters for piezoelectric ceramic and its representation as a three port system
Q1 F1
Q2
ZA
ZA
F2
Z I -C0
C0
It involves the generation, propagation and reception of the signal. The system operates in pulse echo mode. The ultrasonic waves generated by the transducer propagate through the medium and the received echo is converted by the same transducer to electrical signal. The ultrasonic transducer is a bi-physical device that transforms electrical signal to acoustical wave and vice versa. To obtain a VHDL-AMS description of this pulse echo measurement system, one must give: i)- a description of the ultrasonic transducer by using equivalent circuit of Mason as adapted by Redwood. The mechanical part of the piezoelectric transducer is easily represented using a transmission line model, two parameters are sufficient to entirely define the mechanical part of the transducer, the impedance and the sound propagation delay through the transducer. ii)- a description of the propagation medium by means of a transmission line. This model corresponds to the electric equivalent circuit of Branin [5]. The transmission line parameters are calculated according to the characteristics of the biological medium. IV. VHDL-AMS MODELLING OF THE EXPERIMENTAL SETUP
I3 V3
Fig. 2 Mason's equivalent circuit model of three port system ZA = -j Zt tan(k d/), Z = -j Zt sin (k d), I = h33 C0
The global schema with the pulse echo transducer is implemented with Redwood’s model Fig. 4. The inferior part is the transducer in reception mode and its connected to a charge (Cscope, Rscope) which represent the input impedance of the electric measurement tool. The piezoceramic layer equivalent circuit is shown as a linear propagation diagram. The superior part correspond to the transducer in emission mode which is connected to the
IFMBE Proceedings Vol. 29
Comparison between VHDL-AMS and PSPICE Modeling of Ultrasound Measurement System for Biological Medium
electric source. The associated test-bench with VHDLAMS language of the experimental setup is presented in Fig 5. The frequential transducers response study is essential to predict the sensitivity of the system for the various analyzed biological mediums. The studied transducers are produced with PZT ceramic of P188 type (Quartz et Silice) with characteristics are recalled in table 1. TABLE1 TRANSDUCERS ACOUSTIC CHARACTERISTIC Parameters F0 A e Zt co
Value Type A
Quantity Resonance Frequency Area Thickness Acoustic impedance Acoustic velocity Capacitor of the ceramic disc Dielectric constant Thickness coupling factor Piezoelectric Constant
Co
Ǽ33 kt h33
2.25 MHz 132.73 mm² 1 mm 34.9 MRayls 4530 m/s 1109.8 pf 650.0 0.49
531
to the electrical input. The simulation code used to obtain simulation is presented in Fig 5. The simulation results with VHDL-AMS and PSPICE are compared with a transducer in vitro measurement. The pulse voltage is – 100 Volt during 0.222 μs and we consider a 50.0 Ohms resistor between the electrical transducer input and the voltage source. The time 0.222 μs corresponds to 1/(2uF0) where F0 is the resonance frequency of the transducer. The study is done on time response and its spectral frequency analysis. The results are presented in figures 6 and 7. The simulated voltage is V3 : the electrical transducer port. The simulations show good agreement with the measurement obtained with a real transducer. The pick voltage obtained with VHDL-AMS (-4.20 V) is more important than the one obtained with PSPICE (-1.30 V) and thus nearest to the measured one (-3.85 V). The signal wave form obtained with VHDL-AMS is not disturbed by irregularities contrary to the PSPICE simulation.
1.49.10+9 ENTITY Measure_cell IS END Measure_cell; ARCHITECTURE structure OF Measure_cell IS
Piezoceramic layer
linear Medium
v2
F2
v1
Zback
F1 Co
R
I3
T Emitter
V3 Co
TERMINAL n1,n2,n3,n4,n5,n6,n8,n9 : ELECTRICAL; TERMINAL Tb,Tb2,Tf,Tf2 : kinematic_v; CONSTANT A : real := 132.73e-3 CONSTANT e: real := 1.0e-3; CONSTANT Co: real := 1109.8e-12; CONSTANT Va : real := 4530.0; CONSTANT kt : real := 0.49; CONSTANT epsi0 :real:= 8.8542e-12; CONSTANT epsi33 :real:= 650.0; CONSTANT ro :real:= 3300.0; CONSTANT h :real:= kt*Va*sqrt(ro/(epsi0*epsi33)); CONSTANT K : real := h*Co CONSTANT ZT : real := 34.9e6; CONSTANT Zfront : real := 1.5e6; CONSTANT Zback : real := 445.0; QUANTITY vinput across ie through n1 electrical_ground;
to
BEGIN If now > 0.0 and now < 222.2e-9 USE vinput == -100.0; Else vinput == 0.0; End USE;
Piezoceramic layer v2
F2
v1
Z back
F1 Co
C Scope
R Scope
Receiver
T
Co
Fig. 4 Measurement cell schema to be implemented in VHDL-AMS
V. RESULTS
The tools used for simulation are: ADMS v3.0.2.1 of Mentor Graphics company for VHDL-AMS simulation and OrCAD software for PSPICE simulation. To perform the transducer frequency analysis, we used a VHDLAMS Testbench where a frequential source is connected
R1 : entity Resistor(bhv) generic map (50.0) port map ( n1, n2); T1 : entity Redwood(bhv) generic map( Co, K, A*ZT, e/Va) port map( n3, kinematic_v_ground, n4, kinematic_v_ground, n2, electrical_ground); Medium : entity linearMedium(bhv) generic map (1.5e6, 20e-9 ,fo ,1500.0 ,5.0 ,0.9 ,0.045 ,1.0e-9) port map( n3, kinematic_v_ground, n5, kinematic_v_ground); back : entity Resiskinematic(bhv) generic map ( A*Zback ) port map ( n4, kinematic_v_ground ); back2 : entity Resiskinematic(bhv) generic map ( A*Zback ) port map ( n8, kinematic_v_ground ); T1 : entity Redwood(bhv) generic map( Co, Kt, A*ZT, e/Va) port map( n3, kinematic_v_ground, n8, kinematic_v_ground, n9, electrical_ground ); RScope : entity Resistor (bhv) generic map ( 1.0e6) port map ( n9, ground); Cscope : entity Capacitor (bhv) generic map ( 13.0 e12) port map ( n9, ground); END ARCHITECTURE structure;
IFMBE Proceedings Vol. 29
Fig.5 VHDL-AMS code of the measurement cell
532
N. Aouzale et al. VI. CONCLUSIONS
4 3
In this paper, based on previous works an approach of ultrasonic transducer modeling system is presented. The use of VHDL-AMS language shows the advantage to combine multiphysical domains. The approach can be readily used in current electronic design flow to include distributed physics effects into modeling and simulation process with VHDL-AMS. The transducer is simulated by the Redwood model, while the medium of pro pagation is represented by a transmission line which supposes the plane wave theory. The model allows development of further optimisation with respect to electrical matching and transmitted waveform. It also could be extended to include other phenomena like diffraction and distortion of the acoustic wave propagation in the biological medium under test. Usual medium modelling are based on transmission line theory. To perform measurements sensitivity, we can easily adjust in simulation transducer acoustic parameters and also take into account the best parameter for the transducer conception. The transducer response obtained in simulation shows a good correlation with measurement and indicate that the simulation of an ultrasound sensing device, including both electronics and transducers (electromechanical) is possible using VHDL-AMS. 4 3
(a) Amplitude (Volt) Measure
2
VHDL-AMS
1 0 -1 -2 -3 -4 -5 0.0
0.3
1.0
t (10-6.s)
2.0
3.0
(b) Magnitude (V².Hz-1)
0.25 0.2
(a) Amplitude (Volt)
Measure PSPICE
2 1 0 -1 -2 -3 -4 -5 0.0
1.0
2.0
3.0
t (10 -6 .s) 0.3
(b) Magnitude (V².Hz-1)
0.25
FFT_PSPICE
0.2
FFT_Measure
0.15 0.1 0.05 0 0.0
2.0
4.0
6.0
8.0
10.0
F (MHz)
. Fig. 7 Comparison between the measure time response and the PSPICE simulation (a) and their Fourier transform (b).
REFERENCES 1- Mason W.P. (1942) Electromechanical transducers and wave filters. 2nd ed., New York : Van Nostrand, 2- Redwood M. (1961), “Transcient performance of a piezoelectric transducer,” 33, J. Acoust. Soc. Amer, 527536. 3- Morris S. A., Hutchens C.G. (1986) “Implementation of Mason’s model on circuit analysis programs,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr 33, 295-298. 4- Leach, W.(1994) “Controlled-source analogous circuits and SPICE models for piezoelectric transducers,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr, 41, 60-66 5- Guelaz, R. and al. Modelling and Simulation of Ultrasound Non Linearities Measurement for Biological Mediums, 11th Mediterranean Conference on Medical and Biomedical Engineering and Computing 2007 Springer Berlin H, 16, 377-380
FFT_VHDL-AMS FFT_Measure
0.15
Author:
0.1 0.05 0 0.0
2.0
4.0 6.0 F (MHz)
8.0
1.0
Fig. 6 Comparison between the measure time response and VHDLAMS simulation (a) and their Fourier transform (b)
Djilali KOURTICHE
Institute: LIEN, Nancy Université, Faculté des sciences et technologie Street: BP 70239, Boulevard des Aiguillettes City: Vandoeuvre, 54506 Country: France Email:
[email protected] IFMBE Proceedings Vol. 29
Stimulation Parameter Testing and Verification during Pacing Martin Augustynek1, Marek Penhaker1, Pavel Sazel1, and David Korpas2 1
VSB - Technical University of Ostrava / Department of Measurement and Control, Ostrava, Czech Republic 2 Palacký University / Faculty of Medicine, Olomouc, Czech Republic
Abstract— The main object of the work is measurement and verification of the adjusted values of the cardiostimulator’s parameters. For measuring this parameters we used the devices Impulse 7000D from Fluke company. On a dual-chamber pacemaker we will measure the pulse width, amplitude and impedance on the out of pacemaker.
The CONTAK RENEWAL® TR 2 cardiac resynchronization therapy pacemaker (CRT-P), Model H145 is meant to provide cardiac resynchronization therapy (CRT). Cardiac resynchronization therapy is for the treatment of heart failure (HF) and uses biventricular electrical stimulation to synchronize ventricular contractions.
Keywords— transform, pacemaker, measurement methods, stimulation voltage, detection.
I. INTRODUCTION The cardio stimulator is an electronic device assuming the primary function of myocardial muscle stimulation by generating of electric impulses in patients with sinus node dysfunction or cardiac conduction system dysfunction. Stimulation can be divided from various standpoints: into indirect stimulation (stimulation through surrounding tissues) and direct stimulation (stimulation is performed in the heart cavity), or according to the duration of the cardio stimulator application into temporary (with the stimulator outside the patient’s body) or permanent (the stimulator placed under the skin), according to the dependence on the heart action into asynchronous or synchronous stimulation, according to point of the stimulation into single-chamber and dual-chamber stimulation.
Fig. 1 CONTAK RENEWAL® TR 2, Model H145 The ZOOM® LATITUDE™ Programming System Model 3120 Programmer/Recorder/Monitor (PRM) is intended to be used as a complete system to communicate with Guidant or Boston Scientific implantable pulse generators. The software in use controls all communication functions for the pulse generator. For detailed software application instructions, refer to the System Guide for the Guidant or Boston Scientific pulse generator being interrogated.
Table 1 The pacing mode is designated most often by a three-digit NGB code Stimulated chamber
Sensed chamber
Response mode
A - atrium
A - atrium
V - ventricle
V - ventricle
T - triggering I - inhibition
D - booth
D - booth
D - booth
O - none
O - none
O - none
II. MATHERIALS AND METHODS The device evaluated is the pacemaker Contak Renewal TR 2 CRT-P (model H145, type DDDR) with attached electrodes. The next component is the ZOOM® LATITUDE™ Programming System Model 3120 PRM and system for parameters testing Impulse 7000D from Fluke Company.
Fig. 2 The ZOOM® LATITUDE™ Programming System Model 3120
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 533–536, 2010. www.springerlink.com
534
M. Augustynek et al.
The Impulse 7000DP Defibrillator/Transcutaneous Pacer Analyzer Test Systems are rugged, portable precision test instruments that ensure proper operation and ultimate performance of critical life-support cardiac-resuscitation equipment. The Impulse 7000DP test capabilities encompass the spectrum of worldwide-established pulse shapes, showcase breakthrough AED technology compatibility, and outperform in accuracy and standards. Additionally, the Impulse7000DP incorporates the tests and the extensive range of test loads and measurement algorithms needed to test external transcutaneous pacemakers.
500
538
38
1,10
550
589
39
0,10
600
648
48
8,90
650
701
51
11,90
700
739
39
0,10
750
803
53
13,90
800
852
52
12,90
850
879
29
10,10
900
938
38
1,10
950
971
21
18,10
1000
1006
6
33,10
1050
1085
35
4,10
1100
1164
64
24,90
1150
1215
65
25,90
1200
1256
56
16,90
1250
1256
6
33,10 24,10
1300
1315
15
1350
1380
30
9,10
1400
1433
33
6,10
1450 1500
1511 1511
61 11
21,90 28,10
Fig. 3 The Impulse 7000DP Defibrillator/Transcutaneous Pacer Analyzer A. Impedance Measuring by Dual-Chamber Pacemaker For this measure we used only one ventricular electrode. This electrode was strip off and then we connected this electrode to Impulse 7000DP. On this device was a setting the mode for pacemakers testing. In the next table we can see the results by the impedance measuring. Table 2 The results of impedance measuring Settings on the FLUKE [Ω]
Measured [Ω]
Deviation [Ω]
Absolute Departure
50
97%) with a concomitant absence of CD45, CD31, CD117 and CD34 (= r0), containing – at the considered resolution – the central atom and its closest neighbors. It continues with sets of eight octagonal (or six hexagonal) segments of the circular lanes delimited by the family of circles: r0 2 L LC ⎝ ⎠
ing of a capacitor, with an initial voltage U0, to the coil is: I = U 0 ωL ⋅ sinh (ωt )exp(− αt )
(3)
where, α = R /( 2 L) , ω = α 2 − 1 / LC , C is the capacitance, and R and L are the resistance and inductance of the coil, respectively. The inductivity of a circular coil, having the radius r, the number of turns N and the radius of the wire conductor rw, is given by the following formula: ⎛ ⎛8⋅r L = μ 0 ⋅ r ⋅ N 2 ⋅ ⎜ ln ⎜⎜ ⎜ ⎝ ⎝ rw
⎞ ⎞ ⎟⎟ − 1, 75 ⎟ ⎟ ⎠ ⎠
(4)
where Vm is the transmembrane voltage, Ex the axial component of the induced electric field, λ the space constant of the cable and τ the time constant. The term on the right of equation (6) represents the activation function and is computed using the method described in paragraph A. While the passive cable model provides the way of the interaction between the induced electric field and the nerve, it does not completely describe the dynamics of nerve stimulation. In order to study the stimulation and propagation of action potentials, we must consider an active membrane model. We use the Hodgkin-Huxley model to represent the nerve membrane [5]. To implement this model, we modify the initial passive cable model. The extracellular potential produced by the fiber’s own activity is negligible. This assumption is valid because the extracellular potential produced by an action potential propagating along a single nerve axon lying in a large extracellular volume conductor is less than 1 mV. The resistance per unit length of the fiber ri can be expressed in terms of the fiber radius a and the resistivity of the axoplasm Ri, as: ri= Ri/π a2. The membrane current per unit length im is related to the membrane current density Jm by the expression: im=2πa Jm; similarly the membrane capacitance per unit length cm is related to the capacitance per unit area Cm by: cm=2πaCm. Finally we replace the membrane resistance per unit length rm by an active model of time and voltage dependent sodium, potassium and leakage channels. With these changes, the cable equation becomes: a ∂ 2Vm − g Na m 3 h(Vm − E Na ) + g K n 4 (Vm − E K ) + g S (Vm − E S ) 2 Ri ∂x 2 ∂V a ∂E x ( x, t ) = Cm m + 2 Ri ∂x ∂t
(
)
(7)
where gNa, gK and gS are the peak sodium, potassium and leakage membrane conductance’s per unit area, and ENa, EK IFMBE Proceedings Vol. 29
Active Behavior of Peripheral Nerves during Magnetic Stimulation
735
and ES are the sodium, potassium and leakage Nerst potentials. The gating variables m, n, h are dimensionless functions of time and voltage which vary between zero and one. We assumed that the resting potential is -65mV, Vm is measured in mV, α and β in ms-1. The values of model parameters used in our computations are given in Table 1:
For the following simulations, the initial voltage on the circuit’s capacitor is set to U0=30V. Figure 4 shows the induced electric field gradient as a function of time and distance along the fiber.
Table 1 ENa
Sodium Nerst potential
50 mV
EK
Potassium Nerst potential
-77 mV
ES
Leakage Nerst potential
-54.387 mV
gNa
Sodium conductance
120 mΩ/ cm2
gK
Potassium conductance
36 mΩ/ cm2
gS
Leakage conductance
0.3 mΩ/ cm2
Cm Ri a
Membrane capacitance Resistivity of axoplasm Fiber radius
1 μF/ cm2 0.0354 kΩּcm 0.0238 cm
a)
Fig. 4 The activation function evaluated along the length of the nerve fiber a) The coil is parallel with a displacement of 25 mm with the cylinder’s axis. b) The coil is perpendicular on the cylinder’s axis
III. RESULTS AND DISCUSSIONS The magnetic coil considered in our simulation has 30 turns, a radius of 25mm and the wire’s radius is 1mm. The computed inductance of this coil is 0.165mH and its resistance is 3Ω. The coil is part of a magnetic stimulator that also comprises a capacitance, C=200 μF. Two cases were studied, for different positions of the coil relative to the arm. Figure 3 (a, b) shows the two analyzed case; first the coil is perpendicular on the cylinder’s axis and in the second case the coil is parallel with the tissue, but with 25mm displacement with respect to the cylinder axis. In order to obtain the transmembrane potential (the response of the nerve fibber to stimulation) as a function of distance and time, first we modulate the electric field gradient in time ( ∂E z ( z , t ) / ∂z ). The electric field gradient represents the activation function and is calculated along the cylinder (the arm), on a line with y=0mm and x= 25–6.25 = 18.75mm, that is on a depth of 6.25mm in the tissue, below the edge of the coil.
Fig. 3 Geometry of the problem
b)
For describing the subthreshold behavior of the nerve fiber inside the tissue, we employ the cable equation (equation 6) to model the passive properties of the nerve fiber. Figure 5 shows the three–dimensional plot of the transmembrane potential Vm, as a function of distance along the fiber and elapsed time since the pulse is applied. The stimulus strength is below threshold, so the fiber behaves like a passive cable and the induced voltage is dissipated.
a)
b)
Fig. 5 The transmembrane potential for the subthreshold behavior of the cylindrical conductor (arm). a) The coil is parallel with the tissue but with 25mm displacement with respect to the cylinder’s axis; b) The coil is perpendicular on the cylinder’s axis Direct comparison of Figure 4 and Figure 5 shows that the resulting transmembrane potential has a time course that resembles the time course of the activation function, although the response of the nerve is somewhat delayed due to the time required for the accumulation of charge on the membrane. The transmembrane potential Vm(x,t) and the three gating parameters m(x,t), n(x,t) and h(x,t) are computed using the method of finite differences, implemented in Matlab with an
IFMBE Proceedings Vol. 29
736
M. Cretu, R. Ciupa, and L. Darabant
iterative algorithm (we compute the value of each parameter knowing its value for the previous time step - 0.1 ms). The space discretization uses a step of 5 mm. It is assumed that the membrane is initially at rest: ∂V m = ∂m = ∂h = ∂n = 0 for t = 0 ∂t ∂t ∂t ∂t
(8)
The boundary conditions of the problem, applied for x = ± L , far from the region where the stimulus strength is large, are that the axial gradients in the transmembrane potential and the three gating parameters vanish. ∂V m = ∂m = ∂h = ∂n = 0 ∂x ∂x ∂x ∂x
(9)
The model is used to determine the response of the nerve membrane to the applied electric field, for different values of the initial voltage on the capacitor of the stimulation circuit. For the case when the coil is parallel with a displacement of 25mm compared to cylinder’ axis, an action potential is evoked at U0 = 40V after a latency period of about 2ms. When the coil is perpendicular, the action potential appears at U0 = 140V, after a latency period of 2.25ms. After the latency periods the transmembrane potential rises rapidly to the value of about 50 V.
Fig. 7 A three-dimensional plot of the response of the nerve fiber to electromagnetic stimulation – the action potential - above threshold. The vertical axis is the action potential, and the horizontal axes represents the distance along the fiber x and the time after capacitor is discharged, t
IV. CONCLUSIONS In our paper we have computed the response of the nerve fiber to magnetic stimulation. If the stimulus strength is below threshold, the nerve fiber behaves like a passive cable and the induced voltage is dissipated. If the stimulus strength is slightly larger, an action potential is evoked. The position of the coil with respect to the tissue that is going to be stimulated influences the response of the nerve. When the coil is parallel to the tissue (with a displacement of 25 mm) the stimulation appears at a smaller voltage (U0 = 40V) compared with the case when the coil is perpendicular on the nerve (U0 = 140V). If the coil is parallel then the latency period is shorter, and the action potential is initiated at x=0 mm which corresponds to the position of the maximum of −∂E x / ∂x .
ACKNOWLEDGMENT This work has been supported within the research programme PNII_IDEI, No. 1078/2007 - “Neural Magnetic Stimulation: Research Concerning the Improvement of the Electrical Equipment Performances and Clinical Efficiency in Diagnostic and Treatment”.
Fig. 6 Variation
of the transmembrane potential and the three gating parameters in time for the two analyzed cases. When the coil is parallel to the tissue the latency period is 1ms, when the coil is perpendicular the latency is about 1.55ms
In figure 6 we plotted the action potential for these two positions of the coil, for different values of U0. Figure 7 (a, b) shows the three-dimensional plot of the response of the nerve fiber to electromagnetic stimulation above threshold. This figure clearly indicates that the depolarized portion of the nerve has been stimulated, while the hyperpolarized portion is not.
REFERENCES 1. Thielscher A., Kammer T. (2004) Electric field properties of two commercial figure-8 coils in TMS: calculation of focality and efficiency. Clin. Neurophysiol. 2004 Jul;115(7):1697-708 2. V. Lin, I. Hsiao, V. Dhaka (2000) Magnetic Coil Design Considerations for Functional Magnetic Stimulation. IEEE Transactions on Biomedical Engineering, vol. 47, no. 5 3. Ruohonen J., Ravazzani P., Nilsson J., Panizza M. (1996) A volumeconduction analysis of magnetic stimulation of peripheral nerves. IEEE Trans. Biomed. Eng. Jul;43(7):669-78 4. Plesa M., Darabant L., Ciupa R., (2008) A Medical Application of Electromagnetic Fields: The Magnetic Stimulation of Nerve Fibers Inside a Cylindrical Tissue. OPTIM 2008, , IEEE Catalog Number: 08EX1996C, ISBN: 2007905111, pp.87-92 5. Roth B.J., Basser P.J. (1990) A Model of the Stimulation of a Nerve Fiber by Electromagnetic Induction. IEEE Transactions on Biomedical Engineering, vol. 37, no. 6
IFMBE Proceedings Vol. 29
Preparations’ methodology for the introduction of Information Systems in Hospitals J. Sarivougioukas1 A. Vagelatos2 and Ch. Kalamara1 1
General Hospital of Athens “G. Gennimatas”, Athens, Greece 2 R.A. Computer Technology Institute, Athens, Greece
Abstract— In the early stages of the introduction of an Information System in a Hospital and before the initiation of the procedures for the actual installation, numerous preparatory actions must be concluded as part of the project. The set of these actions comprise a certain preparatory phase which must be considered part of the project and it has to be well suited in the initial plans. Omitting or limiting the preparatory phase will have as a consequence noticeable delays and in some cases the failure of modules or even endanger the entire project. The successful completion of the preparatory actions rely on a systematic approach which ensures that all the influencing factors have been considered and all interacting conflicts have been resolved prior to the project’s commencement; this can be realized by employing a methodological framework such as the 4-by-3 structure with shafts, which is proposed in the present paper. The considered preparations provide a clear picture and a feedback of the existing organizational readiness and the necessary existing IT cultural status, thus allow taking in advance, the appropriate and adequate measures. Keywords— Hospital Information Systems, Health Informatics.
I. INTRODUCTION
On a regular basis, when Healthcare authorities consider that it is about to look in the market for an Information System (I.S.), the concerns focus on the preparations of a document adequate for a tender. The tender’s document contains descriptions requiring the latest state of the art technology to be applied in an implied way at the intended healthcare organization to achieve a set of business goals. However, each of the desired information systems requires the properly trained personnel following well established internal procedures, applying the appropriate set of data along with the suitably set and configured hardware in order to succeed in promoting the healthcare organization’s aims. In a tender’s process, most of the time period is consumed examining the available market’s solutions visualizing each vendor’s offer for the particular healthcare organization’s purposes or even visiting other installations of the same vendor to evaluate the applied solution. Also, another concern is about the employment of the latest technology through offers without assuring the satisfaction of the prerequisites related to the compatibility with the existing tech-
nological environment and the ability of the personnel to be promptly familiarized and successfully adapted in the application requirements. Hence, the healthcare organization’s business needs, the available personnel, the so far followed procedures, the current workload, and the organizational capacity issues are all considered to be fulfilled or to be satisfied by technology which most of the times turns out to be incapable of satisfying the expectations. From the past experience earned with public tenders concerning the procurement of Hospital Information Systems in Greece [8, 9], typically in a tender’s document, the interest is focused on the description of the carried procedures followed by a set of documents describing with the greatest possible accuracy the technical specifications of the latest technology of both hardware and software items required. Especially, the state healthcare organizations go over an exhaustive analysis of the technical specifications of each and every of the items included on the requesting list according to which a tender is going to take place. The situation is exaggerated more in the cases when the elapsed time between the technical specifications’ writing and the actual tender time turns to be quite long (e.g. more than a year) and new hardware models (or even software revisions) have been announced in the market. The work flow of the healthcare organization is rarely placed on a detailed diagram; it is usually included in specifications describing either abstract or specialized works that it is intended to be escalated by the employed technology lacking parameters like trained personnel, time, consumables, and resources. In the past, we faced all these obstacles during the introduction of IS in “G. Gennimatas” hospital [8, 9]. More specifically, after reviewing that project’s outcomes, it was realized, that most of its complications emerge due to the lack of a preparatory phase that was needed in order to take the appropriate actions and measures for the smooth realization of the actual project. In this paper, the preparative procedures prior to the introduction of a hospital information system are discussed from a methodological point of view. More specifically in the following paragraphs, a literature review is presented. Then the appropriate preparation steps are proposed. Next follows the discussion on the impact of lack of such preparation. And finally the conclusions are presented.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 737–740, 2010. www.springerlink.com
738
J. Sarivougioukas, A. Vagelatos, and Ch. Kalamara
II.
A BRIEF LITERATURE REVIEW
The preparation phase of a project for the implementation of any information system in a hospital is most of the times implicitly addressed in the relevant literature. Thus one has to look for “lessons learned”, “barriers to success” or “factors of failure” in order to collect the appropriate material for his research needs. Nevertheless, the common root of all the announced in the relative literature problems is the lack of a well defined plan of preparations at the early stages of the project’s schedule to meet the requirements of the chosen for implementation system. In the following, some of the key findings of the relevant literature are described concerning information systems for hospitals in general (that is even modules of a hospital information system). The authors in [1] identify a number of four crucial parameters as barriers towards the adoption of Computer Physician Order Entry (CPOE) – a significant module of any hospital I.S. - which have to be considered ahead of the implementation time. The first parameter is concerned with the physicians’ work practice including the working conditions at the wards, the office or at the emergency rooms. The second parameter refers to the current level of technology and the concessions that the medical professionals are able to agree on. The third parameter points at the status of the existing commercial systems and the involved reliability and standardization. The forth parameter is the most complicated since it involves incentives of various origins spanning from financial to scientific motivations. In [2] the authors, while examining the systemic dimension of CPOE argue that CPOE should be avoided to be the first IT module introduced in the wards or in clinics due its high complexity that might develop an even higher degree of resistance to the system’s acceptance by the discouraged personnel. In [4] the authors claim that a systematic information plan (SIM) must be developed prior to the actual introduction of any new information system in the hospital, predicting and satisfying all the future needs. On the other hand, the introduction of CPOE influences the quality of the delivered health care in the hospital with respect to both the administrative and clinical operations too [3]. The successful implementation of CPOE requires the proactive preparations to ensure the personnel’s participation in the introduction of the new system, high level of leadership from the influential staff, specified commitment in the project’s framework for the personnel, and wide communication of the project’s issues. Recorded opinions in the relative literature state that the lack of a well known and standardized application which drive the respective market gets very important of the commonly developed perception [5]. With this in mind, all ef-
forts focus on the provision and the inclusion of the most advanced technology in HIS applications that inevitably will face the raised issues. Other opinions emphasize the prime importance of personnel during the preparation and the actions to be followed for education and training of staffing at all levels [6]. In particular, in [6] new positions are proposed as a result of the lack of continuous knowledge and simultaneous expertise in the clinical, the administrative, the informatics, and the managerial duties. There are success parameters that have to be considered ahead of the introduction of IS to hospitals, referring to the locally holding conditions [7]. The influence of the social culture of both patients and staff too, the local perception and rules for the professional practice of hospitals operations, the abilities of the available computers’ users and the design of an adequate training program, the complexity of the clinical and administrative procedures, the duration of the project, and the experience from similar projects constitute some of the most essential reasons for failing to introduce a new system.
III.
METHODOLOGICAL STRATEGY AND PROPOSED PREPARATION STEPS
From a systemic point of view hospital information systems are comprised by four major parameters: the technological parameter (H/W and S/W) which refers to both the hardware and the software too, the information parameter (data) implying the type and the kind of data to be processed, the procedural parameter (procedures) which is concerned with the flows of works, and the human parameter (personnel) that corresponds to people who both use the system and get benefits from it too. Each of those parameters can be further analyzed into categories and subcategories. All parameters are pair wised interacting provoking the complexity level of Hospital Information System (HIS) as a whole. Among the aims of the introduction of an I.S. in a hospital, is the application of Good Practices (GP) at all levels of a hospital’s activities (see “e-Health Impact project”: http://www.ehealth-impact.org/). In spite of the fact that there is a limited number of exceptional paradigms, GP itself provides two key components: standardization and localization. The application of GP standards seems straight forward process, like a recipe, hides less pitfalls but it demands for localization which requires adaptation, adjustment, etc. Therefore, the inclusion of GPs must be applied to all four major system’s parameters mentioned above. Hospitals provide medical and social services to the local society and they can not be considered as being isolated. In fact, the locally holding conditions in the government, in the legal system, in the market, and in the society in general, affect the operation of the hospital and consequently the
IFMBE Proceedings Vol. 29
Preparations’ Methodology for the Introduction of Information Systems in Hospitals
requirements from HIS. In other words, the four major parameters mentioned above, they must be examined against each and every one of the influencing external parameters. Organizing the preparation for the introduction of an Information System in a hospital, a set of actions must be performed based on an organizational structure consisted of three layers and four coordinates. As it was described above, the highest layer is dedicated for the interactions with all external to the hospital parties, the intermediate layer concerns the decision to apply GPs, and the lowest layer refers to the system itself as it was analyzed above. The described structure provides the framework within which all preparation considerations and actions will take place, in a sound and complete manner. Additional layers may be added in the direction to increase the level of detailed description of the preparation steps. For reasons of convenience, the organizational structure mentioned above will be referred by the 4-by-3 structure (see Table 1.). Table 1 Functions as Preparatory Actions Business Requirements Systemic Parameters
Hospital specific operation (so)
Good Practices (gp)
External to the Hospital (ex)
Software and Hardware (sh) Procedures (pr) Personnel (pe) Data (da)
F1(sh, ex)
F2(sh, ex)
F3(sh, ex)
H1(pr, ex) G1(pe, ex) R1(da, ex)
H2(pr, ex) G2(pe, ex) R2(da, ex)
H3(pr, ex) G3(pe, ex) R3(da, ex)
The preparations and the actions prior to the attempts to introduce a HIS depend primarily on the decided quantified goals that will be achieved by the operation of HIS, on the organizational readiness of the hospital, on the availability of time and resources to perform the preparations, and on the understanding of the functionality of the new system. As it happens with ordinary system analysis, there are available two approaches, top-down and bottom-up. Thus, given the 4-by-3 structure, the analysis can either start with the lowest or the upper most layer. In either case, the 4-by-3 structure must be penetrated and traversed from either side until the opposite last layer is reached. Each penetration’s routing corresponds to a single preparatory functional action which is called, for convenience as a shaft. Therefore, visualizing the setting, shafts are traversing the 4-by-3 structure and each shaft corresponds to preparatory actions. As an application example let us address the issue of considering as candidate the codification systems, the selection criteria and the preparations required prior to its usage. In this example consider the necessity for all HIS stakeholders to share the same meanings over the employed coding systems as it is required by the software vendor. Following a bottom-up approach, the firstly examined dimension refers to the information coordinate or the data. In
739
this case we consult the vendor’s requests for all codification used in the application software ending with a set of codes. Continuing the process, each codification is examined against the second dimension which is the procedural coordinate, the applied procedures, trying to locate the corresponding workflow where the coding is used. On the next step, the coordinate of people, personnel, is inspected against all the findings of the previous coordinates testing the needs of the various personnel’s classes involved in the usage of each coding, as professionals classes are assigned in this coordinate, revealing the required characteristics of each coding along with the needs of each of the people’s classes. The medical professionals’ needs for each coding formulate the educational and training requirements. The training requirements span from the formal training to the concepts included in the coding to the extent of applying the coding in the execution of the everyday duties with the application software. The previous step ends the works placing shafts at the lowest level of 4-by-3 structure but the shafts continue at the intermediate level where the GP is applied. All decisions taken at the previous level are examined against the known GPs. Once a GP is chosen, for example, the directions of the World Health Organization about the coding of diseases, it must, then, be examined the limits and the completeness of the chosen codification, e.g. for ICD-10 (http://www.who.int/classifications/icd/en/), and the appropriate actions have to be taken on all previously decided items. Moving forward the shaft on the intermediate level, the localization attributes have to be revealed and adequate measures to be taken on the initial plan.
IV.
THE IMPACT OF MISSING THE PREPARATION
There are many failure stories in the introduction of an information system to a healthcare organization [9, 10]. There have been stated a number of serious causes in each such case without succeeding to address the similarities and differences of the observed cases. In this paper, the lack of the preparation phase is considered among the major factors that influence a project’s evolution. The preparation phase must be considered as part of the project itself since a number of important decisions have to be accounted and addressed accordingly. The time required in performing contacts with the stakeholders inside and outside of the healthcare organization settling contractual agreements, usually, turn out to be in excess compared to the rest of the project’s duration. In the past, in Greece, many of the projects announced for the introduction of an information system to a healthcare organization, stopped in the development process to make up decisions related to issues such as the unavailability of pro-
IFMBE Proceedings Vol. 29
740
J. Sarivougioukas, A. Vagelatos, and Ch. Kalamara
cedures, lack of personnel, codifications, incomplete training, insufficient support, just to mentioned a few. The time frame for the project’s completion of the works execution is substantially less than the time required for the preparation phase due to the complexity internal to the organization. The order of the project’s deliverables may have to be altered if the preparation stage is omitted. Some deliverables require either to meet decisions in place or organizational adaptations prior to the installation of certain items, for example, the store of the pharmacy have to wait until a coding system and the actual descriptions of the handled materials have been entered into the system’s database. Most developed projects dealing with the introduction of an I.S. to a healthcare organization, have evidently showed that the most expensive part of the project is related to services. The lack of controlling a project’s schedule, it may turn out to be devastating for the future and the success of the entire project. It is impossible to let specialized personnel to hold waiting for either data or instructions. The extended intervals of time in the project’s duration causes the personnel to sense minimal progress which is discouraging and it turns even harder in the future to trust the project’s management, get motivated, and participating again. Hence, the promises must be controlled and given with confidence meeting timely the requirements. The introduction of a Hospital Information System during the late years of last decade in the General Hospital of Athens, “G. Gennimatas”, Greece, provided fruitful conclusions and evidence in the writing of this paper [8]. At that time, the project lasted two years, with rather discouraging operational efficiency and results. To overcome the difficulties, outsourced services had been hired from the manufacturer who provided skilful operators [10]. A decade later, the General Hospital of Nicosia, Nicosia, Cyprus, faced analogous, but in a smaller scale, problems related mostly to the organizational readiness of the Hospital. The schedule of the project was initially planned to last one year which was considered very optimistic and finally it was proved to be so, since the project lasted three years converging gradually to the initial set quality targets. In the case of the “G. Gennimatas” State Hospital of Athens, the parameters of the developed model of 3-by-4 with shafts were identified leading to the development of a managerial approach while in the case of the State Hospital of Nicosia the same approach has been tested and verified with minor improvements. V. CONCLUSIONS
The preparatory phase of a project that declares the ambitions to introduce an information system in a hospital is equally important, if not more precious, than the actual
implementation since it secures both investments and expectations. The various preparations described above aim to assist the IT systems supplier to easily adopt the proposed hospital information system design. In order to achieve the intended purposes, a systematic approach must be applied based on a method which provides procedural steps exhausting all sources of influence that include the principal of feedback at every task. Standardizing the preparation stages of the project is still difficult to be defined and for the time being, it is limited to the discrete measures as it was described above with the employment of shafts in the 4-by-3 structure. Following such directions, it is certain that the involved costs both, with respect to, time and resources will be drastically reduced since the project’s implementing team has available every required information and the procedural set up have been addressed.
REFERENCES 1.
Doolan D, Bates D, Computerized Physician Order Entry Systems in Hospitals: Mandates and Incentives, Health Affairs, vol. 21, N 4 (2002), 180-188. 2. Kuperman G, Gibson R, Computer Physician Order Entry: Benefits, Costs, and Issues. Annals of internal medicine, Vol. 139, N 1 (2003). 3. Upperman J, Staley P, Friend K, Benes J, Dailey J, Neches W, Wiener W, The introduction of computerized physician order entry and change management in a tertiary pediatric hospital, Pediatrics, vol. 116, N 5 ( 2005), 634-642. 4. Brigl B, Ammenwerth E, Dujat C, Graber S, Grobe A, Haber A, Jostes C, Winter A, Preparing strategic information management plans for hospitals: a practical guideline SIM plans for hospitals: a guideline, Int. journal of Medical Informatics, vol. 74 (2005), 51-65. 5. Giuse D, Kuhn K, Health information systems challenges: the Heidelberg conference and the future, Int. journal of Medical Informatics, vol. 69 (2003), 105-114. 6. Ash J, Stavri P, Dykstra R, Fournier L, Implementing computerized physician order entry: the importance of special people, Int. journal of Medical Informatics, vol. 69 (2003), 235-250. 7. Littlejohns P, Wyatt J, Garvican L, Evaluating computerized health information systems: hard lessons still to be learnt, BMJ, vol. 326, (2003), 860-863. 8. Sarivougioukas J, Vagelatos A, Introduction of a Clinical Information System in a Regional General State Hospital of Athens, Greece, Proceeding of Medical Informatics in Europe Conference (MIE2000), Hannover, Germany (2000). 9. Vagelatos A, Sarivougioukas J: Critical Success Factors for the Introduction of a Clinical Information System, Proceedings of IX Mediterranean Conference on Medical and Biological Engineering and Computing, Pula, Croatia (2001). 10. Sarivougioukas J, Vagelatos A: IT outsourcing in the Healthcare sector: The case of a state general hospital, SIGCPR 2002 Conference, Kristiansand, Norway, May 2002. Author: Institute: Street: City: Country: Email:
IFMBE Proceedings Vol. 29
John Sarivougioukas “G. Gennimatas” Hospital Mesogeion 137 Athens Greece
[email protected] Supervised and Unsupervised Finger Vein Segmentation in Infrared Images Using KNN and NNCA Clustering Algorithms M. Vlachos and E. Dermatas Department of Electrical Engineering and Computer Technology, University of Patras, Patras, Greece Abstract— In this paper, two new methods to segment infrared images of finger in order to perform the finger vein pattern extraction task are presented. In the first, the widespread known and used K nearest neighbor (KNN) classifier, which is a very effective supervised method for clustering data sets, is used. In the second, a novel clustering algorithm named nearest neighbor clustering algorithm (NNCA), which is unsupervised and has been recently proposed for retinal vessel segmentation, is used. As feature vectors for the classification process in both cases two features are used: the multidirectional response of a matched filter and the minimum eigenvalue of the Hessian matrix. The response of the multidirectional filter is essential for robust classification because offers a distinction between vein-like and edgelike structures while Hessian based approaches cannot offer this. The two algorithms, as the experimental results show, perform well with the NNCA has the advantage that is unsupervised and thus can be used for full automatic finger vein pattern extraction. It is also worth to note that the proposed vector composed only of two features is the simplest feature set which has proposed in the literature until now and results in a performance comparable with others that use a vector with much larger size (31 features). NNCA also quantitatively evaluated on a database which contains artificial images of finger and achieved the segmentation rates: 0.88 sensitivity, 0.80 specificity and 0.82 accuracy. Keywords— KNN, NNCA, Vein pattern, Hessian matrix, matched filter, morphological postprocessing.
I. INTRODUCTION An application specific processor for vein pattern extraction, and its application to a biometric identification system, is proposed in [1]. The conventional vein-pattern-recognition algorithm consists of an original image grab part, a preprocessing part, and a recognition part, and the last two parts take most of the processing time. The preprocessing part consists of a Gaussian low-pass filter (which works iteratively), a high-pass filter, and a modified median filter. Consequently the conventional algorithm [1, 2, and 4] consists of low pass spatial filtering for noise removal, high pass spatial filtering for emphasizing vascular patterns and thresholding. An improved vein pattern extracting algorithm is proposed in [3] which compensate the loss of vein patterns in the edge area, gives more enhanced and stabilized vein pattern information, and shows better performance than the existing algorithm. In [5], a direction-based vascular pattern extraction algorithm based on the directional information of vascular patterns is
presented for biometric applications. It applies two different filters: row vascular pattern extraction filter for abscissa vascular pattern extraction, and column vascular pattern extraction filter for effective extraction of the ordinate vascular patterns. The combined output of both filters produces the final hand vascular patterns. Unlike the conventional hand vascular pattern extraction algorithm, the directional extraction approach prevents loss of the vascular pattern connectivity. In [6-7] a method for personal identification based on finger-vein patterns is presented and evaluated using line tracking starting from various positions. Local dark lines are identified and line tracking is executed by moving along the lines pixel by pixel. When a dark line is not detectable, a new tracking operation starts at another position. This procedure executes repeatedly, so the dark lines that tracked a lot of times have a great probability to be veins. An algorithm for finger vein pattern extraction in infrared images is proposed in [8]. The low contrast images, due to the light scattering effect, are enhanced and the fingerprint lines are removed using 2D discrete wavelet filtering. Kernel filtering produces multiple images by rotating the kernel in six different directions, focus into the expected directions of the vein patterns. The maximum of all images is transformed into a binary image. Further improvement is achieved by a two-level morphological process: a majority filter smoothes the contours and removes some of the misclassified isolated pixels, and a reconstruction procedure removes the remaining misclassified regions. The final image has segmented into two regions, the vein and the tissue. In [9] a certification system that compares vein images for low-cost, high speed and high precision certification is proposed. The equipment for authentication consists of a near infrared light source and a monochrome CCD to produce contrast enhanced images of the subcutaneous veins. As recognition algorithm used only phase correlation and template matching. In [10], the theoretical foundation and difficulties of hand vein recognition are introduced at first. Then, the threshold segmentation method and thinning method of hand vein image are deeply studied and a new threshold segmentation method and an improved conditional thinning method are proposed. An initial work for localizing surface veins via near-infrared (NIR) imaging and structured light ranging is presented in [11]. The eventual goal of the system is to serve as the guidance for a fully automatic (i.e., robotic) catheterization device. The proposed system is based
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 741–744, 2010. www.springerlink.com
742
M. Vlachos and E. Dermatas
upon near-infrared (NIR) imaging, which has previously been shown effective in enhancing the visibility of surface veins. In [12, 13] a Vein Contrast Enhancer (VCE) has been constructed to make vein access easier by capturing an infrared image of veins, enhancing the contrast using software, and projecting the vein image back onto the skin. On the other hand supervised methods have been used mainly for pixel classification and retinal vessel segmentation. In [14] a vessel segmentation method based on pixel classification has been proposed. For each pixel in the image, a feature vector is constructed and a classifier is trained with these feature vectors. Features are extracted only from the green plane of the retinal image. They conducted experiments using three classifiers (KNN, linear and quadratic) and they concluded that KNN classifier was superior for all experiments. The feature vector consisted of 31 features including Gaussian and its derivatives up to order 2 at different scales. In [15] the authors proposed the use of three features instead of 31 in conjunction with the KNN classifier to classify the pixels in retinal image as vessels or not. They also claimed that the processing time decreased something which is logical due to the smaller feature set used. Finally, in [16] the same authors as in [15] introduced a novel clustering algorithm named nearest neighbor clustering algorithm (NNCA) which is unsupervised and they segmented retinal images using the same feature set as in [15].
II. ALGORITHM OVERVIEW A. Image Acquisition The original image acquired under infrared light using an inexpensive CCD camera. The finger was placed between the camera and the light source which consists of a row of infrared leds (five elements) with adjustable illumination.
a
b
Fig. 1 a. Original image b. ROI image Due to the fact that haemoglobin has strong absorption in the infrared light the veins are shown in the image darker than the other human tissues. So, the goal of our study is to extract these dark regions, corresponding to veins, from the background, corresponding to the other human parts (tissue). The original image which acquired as described above is shown in fig. 1a.
B. Preprocessing ROI extraction and Intensity Normalization. From the original image a region of interest is extracted for further processing in order to isolate the part of the image which contains the maximum extent of useful information and to eliminate the background. The ROI image is shown in fig. 1b. In many cases, the acquired image may have not the desirable contrast or may have been distorted by noise. Thus, a preprocessing step is inevitable in order to give in the next stage an image with the desirable characteristics. A local normalization procedure is proposed which uniformizes the local mean and variance of an image. This is especially useful for correcting non-uniform illumination or shading artifacts. The local normalization of acquired image is computed as follows (and uses two smoothing operators) f ( x, y ) − m f ( x, y ) , where f(x,y) is the original image, g ( x, y ) =
σ f ( x, y )
mf(x,y) is an estimation of a local mean of f(x,y), σf(x,y) is an estimation of the local standard deviation and g(x,y) is the output image. C. Feature Extraction In the case of finger vein pattern extraction the veins oriented mainly along the x-axis and their diameter does not vary enough. So, it can be done the assumption of a fix diameter, something that simplifies the problem and does not require multiscale analysis. In our study features which can effectively segment the image in two regions: vein and no vein are seek. The two appropriate features could be the response of a multidirectional matched filter and the minimum eigenvalue of the Hessian matrix. The first feature, the maximum response of a multidirectional matched filter is selected in order to perform a distinction between vein-like and edge-like structures. The second feature is selected in order to account for the fact that veins are tubular piecewise linear structures recognized by extracting centerlines (image ridges). In a next subsection also the selection of these two features is evaluated based on the estimation of the covariance matrix. Maximum multidirectional response of a matched filter. A filter which represents a kind of a cross sectional profile is designed. It is a symmetrical filter kernel with odd number of coefficients. The centre element has the minimum negative value and the first adjacently elements has this value increment by one, the second adjacently elements has this value increment by two and so on until the two final terminally elements. These elements have the appropriate value that ensures that the sum of the values of all elements that belongs to the same row of the kernel is zero. The number of the negative elements represents the width of the cross sectional profile. All the rows of the kernel are identical. This attribute orientates the filter at a unique
IFMBE Proceedings Vol. 29
Supervised and Unsupervised Finger Vein Segmentation in Infrared Images Using KNN and NNCA Clustering Algorithms
direction. The filter spectrum has this shape because in order to detect veins that have darker pixel values than the background. For our process of detection of finger veins a filter kernel of size 11x11 is used because the average vein diameter has almost six pixels width and does not vary too much along the finger. This filter kernel is applied to whole image by convolution, in six different directions from 0 to 180 degrees (0, 30 … 150) to detect finger veins in these directions. This is accomplished by rotating sequentially the filter kernel by the appropriate amount of degrees and convolves it with the preprocessed image. Then the maximum of all six responses is estimated in a pixel by pixel basis and is selected as the first element of the two dimensional feature vector of each pixel. Minimum eigenvalue of Hessian matrix. The second directional derivatives describe the variation in the gradient of intensity in the neighborhood of a point. A ridge is located on top of a line structure where the gradient undergoes large changes and occurs where there is a local maximum in one direction. Therefore it must have a negative second directional derivative and also a zero first directional derivative in the direction across the ridge [l7]. In order to obtain useful information from the partial derivatives, a 2x2 Hessian matrix for each pixel is constructed. The Hessian matrix is symmetrical with real eigenvalues and orthogonal eigenvectors which are rotation invariant [18]. The eigenvalues of the Hessian matrix measure convexity and concavity in the corresponding eigendirections. A ridge is a region where λ1 ≈ 0 and λ2 0) stands for the affine distortion matrix and v = (vx , vy )T represents the simple translation vector. The model describes accurately 3-D rotation and translation of a planar object, which is orthographically projected to the image plane [1], or perspectively projected given that the object’s viewing distance is relatively large. Using singular values decomposition, the positive-definite affine matrix is analyzed as follows: 1 0 cos(φ ) sin(φ ) cos(τ ) − sin(τ ) [A] = · s1 · 0 s2 · , − sin(φ ) cos(φ ) sin(τ ) cos(τ ) s1 (2) where s21 and s22 are the eigenvalues of [A][A]T . Based on equation (2), the affine distortion can be considered as a sequence of separate operations: a) Rotation through an angle τ (tilt), b) Non-uniform scaling in only one direction (slant), c) Uniform scaling and d) further rotation through φ (swing). B. The motion model in the frequency domain The affine transform leads to a similar transform of the 2D FT amplitude, while the shift v affects only the phase of the 2-D FT. Mathematically: T f2 (x)= f1 [A]−1 (x − v) → F2 (u)= det[A] · F1 [A]T u · e ju v , (3) where u = (ux , uy )T is the vector of spatial-frequency coordinates. In correspondence to (1) and (2), the affine transform in the frequency domain is described by: |F2 (ua )| = det[A] · |F1 (u)|,
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 749–752, 2010. www.springerlink.com
(4)
750
D.S. Alexiadis and G.D. Sergiadis
tion are separated, since translations do not affect the amplitude of the 2-D FT. An example of linear coordinates transform in the intensity and in the spatial-frequency domain is given Fig. 1(a),(b), where it is clear that the amplitude of the 2-D FT undergoes a similar transformation.
III. T HE
PROPOSED METHODOLOGY
The transformation of the spatial-frequency coordinates in (5) can be rewritten as: 1 cos(φ ) − sin(φ ) cos(τ ) − sin(τ ) s1 0 sin(φ ) cos(φ ) · ua = 0 s1 · sin(τ ) cos(τ ) · u.
uy
uy
(a) f 1 (x) and f2 (x)
2
ux
(6) Consider at this point that the rotation angles φ (swing) and τ (tilt) are known. Equation (6) states that the rotated version of the spectrum |F1 (u)|, through an angle φ , and the rotated version of |F2 (u)|, through an angle τ , are identical up to a non-uniform scaling. Based on this fact, we sample the 2D spectrum |F2 (u)| with respect to two orthogonal directions φ and φ + π /2, with logarithmic manner, i.e.:
ux
ω1 (c) N11 =N1 (τi = τ )
ω2
ω2
ω2
(b) F1 (u) and F2 (u)
ω1
ω1 = log(|ux cos φ − uy sin φ |) ω2 = log(|ux sin φ + uy cos φ |),
ω1
(d) N21 =N2 (φi = φ )
(e) N23 =N2 (φi = φ )
(g) PC {N11 ,N23 }
Fig. 1: (a) Two consecutive CT slices of a fetus skull. The second slice is affine distorted. The affine distortion parameters are τ = −2o (tilt), s1 = 1.1, s2 = 0.95 and φ = 3o (swing). (b) The corresponding 2-D FTs. (c) The spectral representation N1 (ω1 , ω2 ; τi = τ ) with τi being equal to the actual one. (d)-(e) The spectral representations N2 (ω1 , ω2 ; φi ) for φi = {−3,3}. (f)-(g) The corresponding PC functions. Notice that N23 matches well with N11 and consequently the PC method produces a sharp peak. where 1 cos(τ ) − sin(τ ) cos(φ ) sin(φ ) s1 0 · ua = − sin(φ ) cos(φ ) · sin(τ ) cos(τ ) · u. 0 s1 2
(5) The above equations reveal the importance of studying the linear parametric motion in the frequency domain. The problems of estimating the affine distortion matrix and the transla-
(7b)
which leads to a 2-D spectral representation N2 (ω1 , ω2 ; φ ). Similarly, sampling |F1 (u)| with respect to the orthogonal directions τ and τ + π /2, we obtain the spectral representationN1(ω1 , ω2 ; τ ). Then, the representations N1 (ω1 , ω2 ; τ ) and N2 (ω1 , ω2 ; φ ) are related to each other: N2 (ω1 , ω2 ; φ ) = det([A])N1 (ω1 − ds1, ω2 − ds2; τ ),
(f) PC {N11 ,N21 }
(7a)
(8)
where ds1 = log(s1 ),
ds2 = log(s2 ).
(9)
Consequently, if the rotation angles φ and τ are a-priori known, then the 2-D spectral representations N1 (ω1 , ω2 ; τ ) and N2 (ω1 , ω2 ; φ ) are simply translated versions of each other (up to the scale factor det([A])). This is shown in Fig. 1. The translation depends only on the scaling factors s1 and s2 . For the estimation of the shift between N1 (ω1 , ω2 ; τ ) and N2 (ω1 , ω2 ; φ ) we propose the application of the PC methodology [7, 12], which presents immutability to multiplication factors, such as det([A]). Namely, the following function is calculated: PC(d1 , d2 ; τ , φ ) = |PC {N1 (ω1 , ω2 ; τ ), N2 (ω1 , ω2 ; φ )}|, (10) which is presented directly below. The shifting parameters ds1 and ds2 are given by the position of the maximum in the
IFMBE Proceedings Vol. 29
Estimation of Linear Parametric Distortions and Motions in the Frequency Domain
751
A. The detailed algorithm 0.08
The proposed algorithm for the estimation of the affine distortion parameters is formulated as follows: A LGORITHM 1: PART A
m(τ , φ )
0.06 0.04 0.02 5
φ
5
0 0 −5
−5
τ
(a)
(b)
Fig. 2: The actual parameters are τ = −2o (tilt), s1 = 1.1, s2 = 0.95 and φ = 3o (swing), like in Fig. 1. (a) The similarity measure (12) for (τi , φi ) ∈ T × Φ. It is maximized for (τi , φi ) = (τ , φ ). From the corresponding PC function, the scaling factors estimates are: sˆ1 = 1.05, sˆ2 = 0.97. (b) The compensated image f2 (x) = f2 ([A]−1 x).
PC function. The corresponding scaling factors s1 , s2 are then calculated from (9). The phase-correlation (PC) function: For two images g1 (x) and g2 , the PC function [7, 12] is calculated from: r(u) , (11) PC {g1 (x), g2 (x)} = F T −1 |r(u)| where r(u) = F T {g1 (x)} · F T {g2 (x)} is the 2-D cross power spectrum of g1 (x) and g2 (x). If the images are shifted versions of each other, the inverse 2-D FT produces a strong peak at the position corresponding to the shift between the images. Due to the normalization (division by |r(u)|) the PC method is robust against illumination changes. A proposed similarity measure: The peak’s sharpness in the PC function PC(d1 , d2 ; τ , φ ) consists a criterion for the similarity between N1 (ω1 , ω2 ; τ ) and N2 (ω1 , ω2 ; φ ). Consequently, it is an indirect measure of the closeness of the used rotation angles τ and φ to the actual ones. Furthermore, it is a confidence measure for the estimation of the scaling factors s1 and s2 . Thus, the use of the following measure is proposed: m(τ , φ ) =
Pmax (τ , φ ) P(τ , φ ) − Pmax (τ , φ )
(12)
where Pmax (τ , φ ) = max{PC(d1, d2 ; τ , φ )}2 and P(τ , φ ) =
∑ PC(d1 , d2; τ , φ )2 ,
(13)
1. Compute the 2-D FTs of f1 (x) and f2 (x). 2. For both FTs, apply a 2-D high-pass filter H(u) (in the frequency domain by multiplication), for example: H(u) = 1 − cos(ux /2) cos(uy /2), where the spatial frequency (ux , uy )T is considered normalized in the interval [−π , π ) × [−π , π ). For natural images the energy is mainly concentrated in low spatial frequencies. Therefore, the amplification of the high-pass components is helpful. Furthermore, compute the logarithm of the 2D FT’s amplitudes. Let those be denoted as |F1 (u)| and |F2 (u)|. Our experiments showed that the use of amplitude’s logarithm leads to enhanced results. 3. For different angles τi and φi on a grid (τi , φi ) ∈ T × Φ: • Calculate the PC function PC(d1 , d2 ; τi , φi ) between the representations N1 (ω1 , ω2 ; τi ) and N2 (ω1 , ω2 ; φi ). • Determine the corresponding similarity measure from (12). PART B 4. Find the angles τˆ and φˆ for which the similarity measure is maximized. 5. From the position of the maximum in the PC function PC(d1 , d2 ; τˆ , φˆ ), estimate the scaling factors s1 and s2 , using (9). As shown in Fig. 2, the application of the algorithm to the images of Fig. 1 lead to the correct estimation of the affine parameters. B. Algorithm for multiple consecutive frames Let us now consider Nt consecutive frames f (x;t), each one related to the previous one by a linear coordinates transform, which is fixed with time, i.e: x(t) = [A] · x(t − 1) + v. In order to render our methodology more robust, one could exploit the information of all Nt consecutive frames. Let N(ω1 , ω2 ;t; τ ) and N(ω1 , ω2 ;t + 1; φ ) denote the spectral representations of f (x;t) and f (x;t + 1). Furthermore, the PC function between them is denoted as:
(14)
d1 ,d2
The proposed measure is the energy of the peak, divided by the energy in the remaining area of the PC function. For the example of Fig. 1, the calculated similarity measure is depicted in Fig. 2. The similarity measure is maximized for the actual values of τ and φ .
PC(d1 , d2 ;t; τ , φ )=|PC {N(ω1 , ω2 ;t; τ ), N(ω1 , ω2 ;t +1; φ )}|. (15) Finally, let m(t; τ , φ ) stand for the corresponding similarity measure. We define the mean similarity measure for all Nt −2 consecutive frames: m(τ , φ ) = Nt1−1 ∑tN1t=0 m(t; τ , φ ).
IFMBE Proceedings Vol. 29
752
D.S. Alexiadis and G.D. Sergiadis
(a) f (x;t),
3. Find the pair of rotation angles τi , φ j for which the mean similarity measure is maximized. These angles constitute the estimates of the actual tilt and swing values. Let these be denoted as τˆ and φˆ , respectively. 4. For each time instant t ∈ [0, Nt − 2]: • From the position of the maximum in the PC function PC(d1 , d2 ;t; τˆ , φˆ ) find the scaling factors sˆ1 (t) and sˆ2 (t), using (9). 5. Find the medians of the sequences sˆ1 (t) and sˆ2 (t).
t = 0,1,2,3
−3
x 10
(τ , φ )
10 8
Results from the application of the extended algorithm are given in Fig. 3. Despite the existence of the static background, the algorithm correctly determined the dominant motion.
6 4 −2
τ
0
2
5
φ
(b) m(τ , φ )
0
−5
IV. C ONCLUSION
(c) PC(d1 ,d2 ;t = 0; τˆ , φˆ )
, (d) COMPt=0 f (x;t); [A]
Studying the linear parametric motion model in the spatialfrequency domain, we presented a frequency-domain algorithm for the registration of two affine distorted images. The idea extends existed PC techniques, e.g. [7]. Also, the algorithm was modified in order to handle multiple consecutive frames for the case of the motion estimation problem. The experimental results verify the correctness of our arguments
t = 0,1,2,3
Fig. 3: “Sunset & seagull” sequence: (a) The frames #1, #2, #3 and #4. (b) The mean similarity measure m(τ , φ ), which is maximized for τˆ = −1 and φˆ = 3, i.e. for angles equal to the actual rotation angles of the “seagull”-object, despite the existence of the background. (c) The PC function PC(d1 ,d2 ;t; τˆ , φˆ ) for t = 0. It contains a strong peak at the appropriate location, which corresponds nearly to the actual scaling factors. (d) The motion compensated frames, with respect to the time instant t = 0, One can notice that the “seagull”-object based on the estimation [A]. remains almost unchanged. The estimation of the rotation angles τ and φ can be based on the maximization of the mean measure m(τ , φ ). Having estimated the actual τ and φ , the scaling factors s1 and s2 are calculated for each pair of consecutive frames f (x;t) and f (x;t + 1). Let those be denoted as sˆ1 (t) and sˆ2 (t). The final estimation of the actual scaling factors can be based on the mean or the median (for the rejection of the outlier values) of the sequences sˆ1 (t) and sˆ2 (t),t ∈ [0, T − 1]. The overall algorithm for the estimation of the parametric motion A = {[A], v}, is summarized as follows: E XTENDED A LGORITHM : 1. For each time instant t ∈ [0, Nt − 2], with f1 (x) = f (x;t) and f2 (x) = f (x;t + 1): • Apply PART A of A LGORITHM 1 and calculate the similarity measures m(t; τi , φ j ) for (τi , φi ) ∈ T × Φ. 2. Calculate the mean similarity measures m(τi , φ j ), for (τi , φi ) ∈ T × Φ.
R EFERENCES 1. Y. Wang, J. Ostermann, Y. Zhang, Video Processing and Communications. New Jersey: Prentice Hall, 2002. 2. J. L. Barron, D. J. Fleet, and S. S. Beauchemin, “Systems and experiment: Performance of optical flow techniques,” Intern. J. of Comput. Vision, vol. 12, no. 1, pp. 43-47, 1994. 3. B. K. Horn and B. G. Schunck, “Determining optical flow,” Artif. Intell., vol. 17, pp. 185-203, 1981. 4. M. J. Black and P. Anandan, “The robust estimation of multiple motions: Parametric and piecewise-smooth flow-fields,” Computer Vision and Image Understanding, vol. 63, pp. 75-104, Jan. 1996. 5. J. Shi and C. Tomasi “Good Features to Track,” IEEE Conference on Computer Vision and Pattern Recognition, Seattle , Jun. 1994. 6. D. S. Alexiadis, and G. D. Sergiadis, “Estimation of Multiple, TimeVarying Motions using Time-Frequency Representations and MovingObjects Segmentation,” IEEE Trans. Image Processing, vol. 7, pp. 982990, Jun. 2008. 7. B. S. Reddy and B. N. Chatterji, “An FFT-based technique for translation, rotation and scale-invariant image registration,” IEEE Trans. Image Processing, vol. 5, pp. 1266-1271, 1996. 8. P. Burlina and R. Chellappa, “Analyzing looming motion components from their spatiotemporal spectral signature,” IEEE Trans. Pattern Anal. Machine Intell., vol. 18, pp. 1029-1033, Oct. 1996. 9. J. Ben-Arie and Z. Wang, “Pictorial Recognition of Objects Employing Affine Invariance in the Frequency Domain,” IEEE Trans. Pattern Anal. Machine Intell., vol. 20, pp. 604-618, Jun. 1998. 10. E. H. Adelson and H. R. Bergen, “Spatiotemporal energy models for the perception of motion,” J. Opt Soc. Am. A, vol. 2, pp. 284-299, 1985. 11. B. A. Watson and A. J. Ahumada, “Model of human visual-motion sensing,” J. Opt Soc. Am. A, vol. 2, pp. 322-342, 1985. 12. H. Foroosh, J. B. Zerubia and M. Berthod, “Extension of Phase Correlation to Subpixel Registration,” IEEE Trans. Image Processing, vol. 11, pp. 188-200, Mar. 2002.
IFMBE Proceedings Vol. 29
Ex Vivo and In Vivo Regulation of Arginase in Response to Wall Shear Stress R.F. da Silva1,2, V.C. Olivon3, D. Segers4, R. de Crom4, R. Krams5, and N. Stergiopuloss1 1
Laboratory of Hemodynamics and Cardiovascular Technology, Swiss Federal Institute of Technology. Lausanne, Switzerland 2 Department of Fundamental Neurosciences, University of Geneva, Geneva, Switzerland 3 Department of Pharmacology, University of São Paulo, Ribeirão Preto, Brazil 4 Department of Cardiology and Genetics, Erasmus Medical Center, Rotterdam, The Netherlands 5 Department of Bioengineering, Imperial College London, London, United Kingdom
Abstract— Background: Alterations of wall shear stress can predispose the endothelium to the development of atherosclerotic plaques. We evaluated the modulation of arginase by wall shear stress. Material and methods: We perfused isolated carotid arterial segments to either unidirectional high mean shear stress (HSS) or low mean and oscillating shear stress (OSS) for 3 days. Vascular function was analyzed by diameter measurement, arginase expression and localization by western blot and immunohistochemistry, respectively. These effects were also evaluated in right carotid artery of apolipoprotein E (apoE-/-) deficient mice, fed with high cholesterol diet, which was exposed to HSS, LSS and OSS flow conditions by the placement of a shear stress modifier for 9 weeks. ApoE-/- mice received either the arginase inhibitor nor-Noha (20mg/kg, 5 days/week) or placebo for 9 weeks. Plaque size and I/M ratio were determined by histology. Results: Our data from ex vivo perfusion showed that exposure of carotid segments to both low and oscillatory flow conditions significantly increase arginase II protein expression and activity as compared to high shear stress athero-protective flow condition. Long-term treatment with nor-Noha effectively decreased arginase activity at LSS and OSS regions, which in turn was accompanied by a decreased I/M ratio and the size of atherosclerotic lesion. In the lesion, inhibition of arginase decreased the number of CD68 positive cells at LSS and OSS zones. Exposure of carotid artery to OSS induced a more pronounced activation of arginase as compared to HSS. Conclusions: Arginase is modulated by patterns of wall shear stress. Long-term treatment of apoE/- mice with arginase inhibitor decreased carotid I/M ratio and atherosclerotic lesion at LSS and OSS regions. Therefore, inhibition of arginase by nor-Noha may emerge as a distinct way to target atherosclerosis disease.
in carotid artery perfused ex vivo. The contribution of arginase to the process of shear stress-induced atherogenesis in vivo will also be discussed in the present paper.
II. MATERIAL AND METHODS To evaluate the effects of wall shear stress in a wellcontrolled hemodynamic environment and without neuroendocrine factors, carotid arterial segments were perfused ex vivo[1] under either HSS or OSS for 3 days. After 3 days of flow exposure, vascular function was analyzed by diameter measurement, arginase expression and localization by western blot and immunohistochemistry, respectively. These effects were also evaluated in an innovative mice model of shear stress-induced atherogenesis developed by the group of R. Krams [2-4]. Carotid arteries of apolipoprotein E deficient (apoE-/-) mice fed with high cholesterol diet were exposed to different hemodynamic by the placement of shear stress modifier device (referred as a cast, Fig. 1A) in the mice carotid for 9 weeks. The vessel diameter and velocity of blood flow were then measured by Doppler analysis in order to precisely derive the shear stress values The cast is able to impose a fixed geometry on the vessel wall, thereby causing a gradual stenosis, resulting in increased shear stress in the vessel segment inside the cast, a decrease in blood flow and consequently a lowered shear stress region upstream the cast, and a vortex upstream from the cast (oscillatory shear stress) (Fig. 1B).
Keywords— Atherosclerosis, hemodynamics, wall shear stress, vulnerable plaque, arginase pathway.
I. INTRODUCTION Alterations of wall shear stress, the frictional force generated by blood flow, can predispose the endothelium to the development of atherosclerotic plaques. In parallel, growing evidence suggests a key role for vascular arginase II in the pathophysiology of atherosclerosis disease. We are currently investigating the modulation of arginase by wall shear stress P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 753–756, 2010. www.springerlink.com
Fig. 1
754
R.F. da Silva et al.
In our study design protocol, apoE-/- mice (n=20 per group) were randomly assigned to receive either the N-ωHydroxy-nor-L-arginine (nor-Noha, 20mg/kg, 5 days/week) or placebo for 9 weeks. Arginase activity in carotid tissue lysates was measured by colorimetric determination of urea formed from L-arginine. Plaque size and intima/media (I/M) ratio were determined by immunohistochemistry analysis after classical eosin-hematoxilin coloration. Collagen fibers were stained with 0.5% picrus sirus red solution. Collagen type I and III were visualized by polarized light Olympus microscopy. Lipid deposition was determined by Sudan IV coloration and analyzed by the later microscopy. Systolic blood pressure (SBP) and heart rate were measured in conscious mice two days before sacrifice by using the indirect tail-cuff method method (BP2000, Visitech System, Apex, North Carolina, USA). After mice were deprived of food for at least 12 hours, blood was collected by cardiac puncture into heparin-coated tubes. Plasma was obtained through centrifugation of blood for 15 minutes at 4500g /4°C and stored at –80°C until each assay was performed. Total cholesterol, triglyceride, LDL, HDL and creatinine levels were determined using an automated clinical chemistry analyzer. Data were analysed using one-way ANOVA and Bonferroni's post-test or unpaired ‘t' test (Graph Pad Prism 5.0; GraphPad Software Inc., San Diego, CA). Value of P ≤ 0.05 was considered significant.
Fig. 2
III. RESULTS Only arginase II (ArgII) isoform was detected on isolated porcine carotid endothelial cells and on carotid artery. Exposure of arteries to OSS increased ArgII expression and activity as compared to HSS in a time-dependent manner (Fig. 2). Inhibition of arginase by nor-Noha improved NOdependent endothelial function and decreased total vascular ROS formation in arteries submitted to OSS. For in vivo experiments, arginase activity was significant increased on contra-lateral carotid artery of apoE-/- mice as compared to WT. Nor-Noha treatment decrease arginase activity in the carotid of apoE-/- mice. In arteries containing the shear stress-modifier, vascular arginase activity was significantly increased at both low and oscillatory shear stress regions as compared to HSS region. Long-term treatment with nor-Noha effectively decreased arginase activity at these regions, which in turn was accompanied by a decrease I/M ratio and % of atherosclerotic lesion (Fig.3 A and B). In the lesion, inhibition of arginase decreased the % CD68 positive cells at LSS and OSS zones (Fig.3C and D), with no significant changes on the % of smooth muscle cells and lipid content.
Fig. 3
IV. DISCUSSION ArgI and II isoforms have very distinct subcellular localization and tissue distribution. In the vascular system, arginase can be either constitutively expressed or activated upon stimuli. Moreover, changes on vascular arginase appear to
IFMBE Proceedings Vol. 29
Ex Vivo and In Vivo Regulation of Arginase in Response to Wall Shear Stress
be specie-dependent. The present study demonstrated that only ArgII isoform was detectable in carotid artery tissue. Emerging evidence indicates that increase in arginase expression and/or activity correlates with several risk factors for cardiovascular diseases including atherosclerosis [5-7]. In particular, the group of Berkowitz elegantly showed that inhibition of ArgII decreased the formation of atherosclerotic plaque in apoE-/- mice [8]. It is well recognized that alterations of wall shear stress plays an important role as a local risk factor for the development of atherosclerosis disease [3]. In the present study, we investigated the regulation of arginase in response to wall shear stress in a wellcontrolled hemodynamic environment. Our results showed that exposure of carotid arteries to HSS values for 3 days significantly increased the expression of ArgII as compared to fresh artery. Oscillatory flow, a hemodynamic condition known to favour plaque development, increased the expression of vascular ArgII already 1 day after flow. Interestingly, after 3 days OSS significantly increased the expression and activity of ArgII as compared to HSS. These results indicated that ArgII is stimulated by wall shear stress depending on duration and type of flow pattern. Vascular arginase compete with eNOS for the same substrate L-arginine, thus lowering the formation of NO [9]. For the above reasons and also due to the pronounced increase of ArgII in the endothelium, we investigated whether changes in arginase expression/activity by OSS would affect NO-dependent endothelial function. Our results demonstrated that OSS significantly decreased the vasorelaxation capacity in response to bradykinin and that the arginase inhibitor nor-Noha restored the NO-dependent endothelial function. High levels of vascular ROS can react with NO to form peroxynitrite and to lower NO bioavailability. In our ex vivo model, perfusion of arteries under OSS increased total vascular ROS formation as evidenced by DHE staining. Several key mediators of atherothrombosis such as thrombin [10], inflammatory cytokines [11] and oxidized LDL [5] have been shown to up-regulate the activity of ArgII isoform in EC. More recently, vascular increase in arginase activity has been shown to contribute to the mechanism of endothelial dysfunction present in atherosclerotic vessel [12]. Whether arginase mediates the endothelial dysfunction induced by OSS is not yet known. Our results showed that arginase expression and activity is modulated by patterns of wall shear stress in vivo. In arteries containing the shear stress-modifier, vascular arginase activity was significantly increased at both low and oscillatory shear stress regions as compared to HSS region. Long-term treatment with nor-Noha effectively decreased arginase activity
755
at these regions, which in turn was accompanied by a decrease I/M ratio and % of atherosclerotic lesion. In the lesion, inhibition of arginase decreased the % CD68 positive cells at LSS and OSS zones, with no significant changes on the % of smooth muscle cells and lipid content.
V. CONCLUSIONS Exposure of carotid artery to oscillatory flow induced a more pronounced activation of arginase as compared to HSS. Inhibition of arginase in arteries exposed to OSS improved NO-dependent endothelial function and decrease smooth muscle cell proliferation rate, both processes that are important for the focal development of atherosclerotic plaque. Vascular expression and activity of ArgII is regulated by patterns of wall shear stress in vivo by decreasing plaque size and I/M thickness. Inhibition of arginase emerges as a distinct way to target atherosclerosis disease by acting via NO-dependent and independent pathways, which in turn could respectively improve endothelial function and arterial remodelling. Nevertheless, it is important to note that so far commercially available inhibitors of arginase are not selective to vascular ArgII isoform and therefore it may induce adverse effects through inhibition of ArgII and/or ArgI in other tissues.
ACKNOWLEDGMENT We would like to thank Elodie Deluz for her excellent technical assistance. This work was supported by the Swiss National Science Foundation (grant numbers 3100A0103823 to N.S., PIOIB117205/1 to RF.S.), by Olga Mayenfisch Stiftung and Novartis Consumer Health Foundation (to RF.S.).
REFERENCES 1. Gambillara, V., et al. 2006. Plaque-prone hemodynamics impair endothelial function in pig carotid arteries. Am J Physiol Heart Circ Physiol. 290: H2320-2328. 2. Cheng, C., et al. 2007. Shear stress-induced changes in atherosclerotic plaque composition are modulated by chemokines. J Clin Invest. 3. Cheng, C., et al. 2006. Atherosclerotic lesion size and vulnerability are determined by patterns of fluid shear stress. Circulation. 113: 2744-2753. 4. Cheng, C., et al. 2005. Shear stress affects the intracellular distribution of eNOS: direct demonstration by a novel in vivo technique. Blood. 106: 3691-3698. 5. Ryoo, S., et al. 2006. Oxidized low-density lipoprotein-dependent endothelial arginase II activation contributes to impaired nitric oxide signaling. Circ Res. 99: 951-960.
IFMBE Proceedings Vol. 29
756
R.F. da Silva et al.
6. White, A. R., et al. 2006. Knockdown of arginase I restores NO signaling in the vasculature of old rats. Hypertension. 47: 245-251. 7. Zhang, C., et al. 2004. Upregulation of vascular arginase in hypertension decreases nitric oxide-mediated dilation of coronary arterioles. Hypertension. 44: 935-943. 8. Ryoo, S., et al. 2008. Endothelial arginase II: a novel target for the treatment of atherosclerosis. Circ Res. 102: 923-932. 9. Ash, D. E. 2004. Structure and function of arginases. J Nutr. 134: 2760S-2764S; discussion 2765S-2767S.
10. Ming, X. F., et al. 2004. Thrombin stimulates human endothelial arginase enzymatic activity via RhoA/ROCK pathway: implications for atherosclerotic endothelial dysfunction. Circulation. 110: 3708-3714. 11. Satriano, J. 2004. Arginine pathways and the inflammatory response: interregulation of nitric oxide and polyamines: review article. Amino acids. 26: 321-329. 12. Ryoo, S., et al. 2008. Endothelial arginase II: a novel target for the treatment of atherosclerosis. Circ Res. 102: 923-932.
IFMBE Proceedings Vol. 29
Process choreography for Interaction simulation in Ambient Assisted Living environments C. Fern´andez-Llatas1 , J.B. Mochol´ı1 , C. S´anchez1 , P. Sala1 and J.C. Naranjo1 1
TSB-ITACA, Universidad Polit´ecnica de Valencia, Spain
Abstract— AAL solutions and Universal Design are more and more used to empower individuals to manage his needs by using ICTs. Nevertheless, the use of design-for-all principles in ICTs is a very difficult task. Products that may seem great ideas in design time can be useless for people with limitations. The use of systems that can simulate the interaction between the user and AAL services and devices can help designers to make fewer design errors and create more useful products. In this paper, a simulator of AAL processes and services is presented. This simulator, based on process choreography principles, is intended to be used in all stages of design process. For that the system allows not only the simulation of single processes, devices and its relationship but also the connection with real devices to test the interaction between real and simulated processes and devices. Keywords— Choreography, Orchestration, Workflow, processes, AAL, Simulation
I. I NTRODUCTION With increase life expectancy, the quantity of people with age-specific disorders are also growing more and more, thus increasing the number of people who will require assistance and care in the future. Inside Ambient Intelligence paradigm, AAL [1] (Ambient Assisted Living) is the solution that appears in the horizon to empower individuals to manage his needs by using ICT. According to AAL thesis, technologies should be fully adapted to human needs and cognition, and totally accessible regardless of the specific condition of the user (Universal Design). AAL puts the technology in the hands of any person in a way that is transparent to him and specifically developed to manage his needs. Hence, seniors will have all the advantages of the AAL services and will avoid all the disadvantages of using technology. However, in many cases, the complexity and lack of accessibility and usability of ICT is a major barrier. There exist several R&D projects currently dealing with this challenge, mainstreaming accessibility in goods and services, as well as developing up-to-date assistive technologies. ASK-IT project [2] uses the design-for-all principles to develop services based on ICT that allow mobility impaired people to live and travel more independently. eAbilities [3]
project aims at developing a framework for current and future actions in research, education and technology transfer in the field of ICT accessibility in the home, vehicle and working environments in Europe. The MAPPED [4] project takes into consideration all the accessibility needs of disabled users to provide them with the ability to plan excursions from point to point, at any time, using public transport, their own vehicle, walking, or using a wheelchair. Although accessibility and usability are crucial concepts in current society and the seven principles of the Universal Design [5] or Design for All are well-known and applicable to a wide variety of domains, its use on real systems is very limited. These ideas are usually taken into account in research and development activities but present significant reservations in production and deployment phases. This is because business stakeholders are still reticent to use them in fact. The main cause of that reticence is the high cost of validating the services and devices in accessibility and usability terms. To solve that it is necessary to develop validation tools and methodologies that support the designers along the whole development cycle, for example, allowing them to simulate the interaction of AAL components during the design phase to detect conceptual errors in early stages of the development process. In order to simulate the interaction among the system components in design phase it is necessary to describe the components execution flow in an formal way. To solve that, the use of workflow technology [6] is usually the best choice. Workflow technology allows the specification and simulation of processes by non programming experts. This allows the accessibility and usability experts define the AAL component process without knowing programming languages. Nevertheless, although the workflows can describe processes flow, they are not able to simulate interaction among different processes. Workflows are designed to be orchestrated. The orchestration assumes that the flow is managed by a central component that decides the next step in the process. In an interaction among processes the flow is decided in a distributed way. In this case, the next step is decided by the combination of the flows of the interacting components. That kind of execution of processes is usually known as process choreog-
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 757–760, 2010. www.springerlink.com
758
C. Fernández-Llatas et al.
raphy [7]. Choreography assumes that the processes are able to exchange data to execute processes in a distributed way. In this paper, a computer-aided supporting tool to simulate AAL components interaction by using choreography and orchestration principles is presented. The solution is intended to be used by Interaction Designers and Usability Engineers at all development stages. This tool was developed in the framework of VAALID european project [8].
II. M ATERIALS AND M ETHODS In order to simulate the interaction between AAL components, we need to describe two levels of simulation: • The first level is the simulation of AAL components. Those components are the services and the devices that will be executed in the system. This simulation requires a formalism that allow experts to describe the flow of each component to be simulated. • The second simulation level is the interaction among the components processes. This simulation requires an environment that allows the message exchange among the simulated processes. Workflow technology can deal with first level of simulation. A wide-accepted definition of workflow defines it as the formal specification of a process. It defines actions, which are performed by humans or by computerized systems, and the set of allowed transitions among them that defines the possible paths that a custom process can follow. More specifically, the Workflow Management Coalition [6] defines a workflow as the automation of a business process, in whole or part, during which documents, information or tasks are passed from one participant to another for action, according to a set of procedural rules. The use of workflows allows experts to formally describe AAL components by using easy to use graphical tools. Workflows can be executed using Workflow Engines. A Workflow Engine is a software system that can understand and perform the actions described in a workflow in the specified order. There are several Workflow systems available in the literature depending on the final system needs: jBPM [9], the java open source Workflow Engine from jBOSS, the Windows Workflow Foundation [10] from Microsoft, the proprietary Staffware [11] or the created in the scientific community YAWL [12] inter alia. The second simulation level cannot be executed using workflows. The workflow technology is designed to execute exhaustively described processes where the workflow engine (orchestrator) is who takes the decisions of passing from one state to other. This paradigm is called orchestration. The simulation of interactions requires that the decision of changing
from one state to other has to be taken in a distributed way. In that case, there is no central module that takes the decision, but the transition among states occurs by the combination of individual flows for each of the system components. This paradigm is called choreography [7]. Multi-Agents technology can be used to simulate the interaction among components. A multi-agent system (MAS) is a system composed of multiple interacting intelligent agents. Intelligent agents (IA) are autonomous entities which observe and act within an environment. There are frameworks like JADE [13] to build MAS systems and rich MAS communication protocols like FIPA [14] that allows agents exchange all the data necessary to interact. However MAS is not a good solution for simulation due to its inefficiency. Multi-Agent systems require too much resources, a problem in tasks that involve high number of simulation elements, and in certain situations they take a long time before providing a response. A more efficient alternative is the Enterprise Service Buses (ESB) [15]. ESB are general-purpose interconnection software systems based on Service Oriented Architecture (SOA) [16]. In that paradigm, the ESB are thought to interconnect different endpoints. Typically, the ESBs are implemented using event-driven messaging in a central engine who actuates as the message dispatcher among the choreography components. ESB promote flexibility in the transport layer and enable loose coupling and easy connection between services. There are powerful ESBs like OpenESB [17] and jBossESB [18] that require containers to be executed or lightweight ESB like Mule [19]. ESB are specifically appropriate to interconnect existing applications. For that, these systems are mainly focused in using the maximum quantity of inputs and outputs formats. This problem is not a critical issue in simulating interactions because the data exchange format can be previously defined. Other alternative to create SOA systems by using efficient technologies is OSGi [20]. OSGi framework is a dynamic modular system written for Java technology that allows building applications from small, reusable and collaborative components. These components can be registered in OSGI framework and interact with others using the provided mechanism. Using OSGi primitives is possible to subscribe/unsubscribe services and define interfaces to interconnect those services. In addition, OSGI framework is providing more and more components in order to ensure the security, provide communication protocols, connection to handheld devices... etc. OSGi is becoming one of the most popular technology for SOA systems, even jBOSS is planning to program its ESB using OSGi framework. Comparing the models, while ESB provides a closed bus for data exchange, OSGi provides a framework that allows us
IFMBE Proceedings Vol. 29
Process Choreography for Interaction Simulation in Ambient Assisted Living Environments
to create a tailored choreographer designed to simulate interactions in a simple way, but it is necessary to develop it.
III. R ESULTS In our problem, although MAS systems provide a very rich exchange data format and very powerful model to represent AAL components as agents, Workflows are more efficient to represent the services and devices, and are more readable to be used by interaction experts. In case of service choreography, ESB systems are more efficient than MAS Systems. ESBs are more focused on system interoperability and are usually very closed and difficult to be tailored. The VAALID project election was the creation of a tailored processes choreographer to simulate interaction using OSGi framework.
Fig. 1: Choreographer Architecture In Figure 1 it is shown the general architecture of the VAALID Choreographer presented in this paper. The choreographer core is a OSGi module that allows the interaction among other OSGi modules (Bundles). The choreographer has implemented a listener that automatically connects OSGI bundles that are registered in the framework using the specific software interface (IService). Once the choreographer gets a service, the service is started and can be publically called from other services connected to choreographer. The choreographer dispatches the messages among the modules using an specific XML message protocol called XMSG based on the combination of FIPA and SOAP protocols. The classic FIPA protocol, defined for communication in MAS, allows sharing knowledge using different protocols. XMSG in based on FIPA headers to route and characterize the messages. The content in XMSG is SOAP protocol. SOAP is a well known and widely used protocol to perform service calls. The XMSG protocol allows broadcast, multicast, and P2P message calls using wild symbols in the destiny address. In addition to the choreographer, some general purpose modules are implemented to perform the interaction simulation:
759
• A workflow engine, based on jBPM, which executes the workflows defined for the different devices and systems. • A rule engine, based on Drools, that allows the execution of deductive rules to simulate human decisions or intelligent systems. • A wrapper to connect to 3D scenes, based on Instant Reality engine, to show the interaction simulation effects. • A wrapper to connect with real devices by using the domotic bus KNX • Some wrappers to interconnect the Choreographer with other Choreographers, real local or remote objects (written in Java or DotNet platforms) or 3D simulation environments. Using these components the choreographer is able to interconnect Workflow orchestrations as well as 3D representations and real modules. Workflow orchestration allows the formal representation and simulation of the flow of devices and services in a simple way directly defined by interaction engineers. The interconnection of the choreographer with real modules allows the test of the interaction of individual real systems in combination with simulated modules in simulated environments. This permits the use of the system with real devices and services after they have been implemented. The use of 3D environments to show the interaction simulation makes easier the evaluation of the real interaction problems. In addition, a .NET wrapper module has been written in order to probe the interconnection capabilities with other programming platforms. In this case, it is possible to connect modules written in Java with Microsoft .NET platform modules. This interoperability allows the simulation of services implemented on different platforms. In practice, these modules are intended to be used from design to testing phases of the AAL product cycle of life. On design phase, interaction designers can describe the behaviour of the new product by using workflows and the look and feel using 3D tools. The interaction of the product that is being developed can be simulated in some predefined environments with other real or simulated products with real or simulated users in order to detect design problems in early stages. Once the product is implemented the system is useful too. The simulation of the interaction of the real product in simulated environments allows the engineers detecting implementation errors before it’s used in real environments.
IV. C ONCLUSION In this paper a new process choreographer for simulating interactions was presented. The service choreography can be
IFMBE Proceedings Vol. 29
760
C. Fernández-Llatas et al.
used to simulate the interaction among AAL services and devices. This solution can empower AAL designers to create better defined products to be introduced on real markets. The simulator implemented has the capability of simulating the interaction of processes defined in different design stages (in functional specification stage using workflows or in testing stage using the real module) and in different platforms (Microsoft .NET and Java) testing the use of the simulator with modules implemented in different platforms. This makes the choreographer useful independently of the final implementation language chosen. The simulator is connected to a workflow engine (jBPM) to provide a formal way to execute workflows; it uses a rule engine (Drools) to allow the executions of business rules that permits the simulation of complex decisions and intelligent systems; it is connected to a domotic bus (KNX) that allows the communication with real hardware; and finally it uses a 3D visualization Module (Instant Reality) allowing the visual feedback to designers using 3D environments. The choreographer is being tested in VAALID project using simulated modules and real devices in a real living lab.
ACKNOWLEDGMENTS The authors wish to thank the European Commission for the project funding and the VAALID consortium for their support.
R EFERENCES 1. European Communities Commission. Ageing well in the information society: An i2010 initiative tech. rep.European Comision 2007. 2. Consortium ASKIT. ASK-IT Project: Ambient Intelligence System of Agents for Knowledge-based and Integrated Services http://www.askit.org/ 2003-2008. 3. Consortium . A virtual platform to enhance and organise the coordination among centres for accessibility resources and support http://www.eabilities-eu.org/ 2006-2009.
4. Consortium MAPPED. Mobilisation and Accessibility Planning for People with Disabilities http://services.txt.it/MAPPED/ 2004-2007. 5. Story Molly Follette, Mueller James L., Mace Ronald L.. THE UNIVERSAL DESIGN FILE Designing for People of All Ages and Abilities. The Center for Universal Design 1998. 6. WfMC . Workflow Management Coalition Terminology Glossary. WFMC-TC-1011, Document Status Issue 3.0 1999. 7. Zimmermann Olaf, Doubrovski Vadim, Grundler Jonas, Hogg Kerard. Service-oriented architecture and business process choreography in an order management scenario: rationale, concepts, lessons learned in OOPSLA ’05: Companion to the 20th annual ACM SIGPLAN conference on Object-oriented programming, systems, languages, and applications(New York, NY, USA):301–312ACM 2005. 8. Consortium VAALID. VAALID Project: Accessibility and Usability Validation Framework for AAL Interaction Design Process http://www.vaalid-project.org/default.aspx 2008-2010. 9. Division JBoss Red Hat. JBoss jBPM White Paper: http://jboss.com/pdf/jbpm whitepaper.pdf 10. Corporation Microsoft. Introducing Microsoft Windows Workflow Foundation: http://msdn2.microsoft.com/en-us/library/Aa480215.aspx 11. Inc TIBCO Software. StaffWare ProcessSuite WhitePaper: http://about.reuters.com/partnerships/tibco/material/Staffware whitepaper.pdf 12. Aalst W., Hofstede A.. YAWL: Yet Another Workflow Language Information Systems. 2005;30:245-275. 13. Bellifemine Fabio Luigi, Caire Giovanni, Greenwood Dominic. Developing Multi-Agent Systems with JADE. Wiley 2007. 14. O’Brien P. D., Nicol R. C.. FIPA Towards a Standard for Software Agents BT Technology Journal. 1998;3:51-59. 15. Chappell David. Enterprise Service Bus. O’Reilly Media, Inc. 2004. 16. Erl Thomas. Service-Oriented Architecture: Concepts, Technology, and Design. Upper Saddle River, NJ, USA: Prentice Hall PTR 2005. 17. Microsistems Sun. OpenESB: The Open Source ESB for SOA & Integration https://open-esb.dev.java.net/. 18. Community. JBoss. JBoss Enterprise Service Bus http://www.jboss.org/jbossesb/. 19. Community Mulesoft. Mule Enterprise Service Bus http://www.mulesoft.org/display/COMMUNITY/Home. 20. Alliance Osgi. Osgi Service Platform, Release 3. IOS Press, Inc. 2003.
Author: Carlos Fern´andez-Llatas Institute: ITACA-Universidad Polit´ecnica de Valencia Street: Camino de Vera S/N City: Valencia Country: Spain Email: cfl
[email protected] IFMBE Proceedings Vol. 29
Measuring Device for Determination of Forearm Force P. Hlavoˇn1 , J. Krejsa2 , and M. Zezula1 1 2
Brno University of Technology/Faculty of Mechanical Engineering, Brno, Czech Republic Institute of Thermomechanics AS CR, v. v. i., Academy of Sciences of the Czech Republic
Abstract— Maximum achievable force at the end of person’s forearm depends on elbow joint angle. Specialized measuring device was designed to identify this relationship. The force is measured during continuous movement of forearm from one limit position to the other. This movement is forced by electric motor so it is not influenced by the subject. The force at the end of forearm and the elbow joint angle are captured using electric transducers connected to the Spider8 data acquisition system. Keywords— Spider8, arm, forearm, elbow.
I. I NTRODUCTION The idea of designing the measuring device arised during the development of experimental mobile motorized orthosis, that acts against patient’s muscles, and must be able to overcome their force. We need to determine this force to design the orthosis correctly. Moreover, the magnitude of this force heavily depends on elbow joint angle (among other factors, such as patient’s actual condition). If we determine this relationship, we can further optimize the driving mechanism of the orthosis and focus its maximum force to certain range of elbow joint angle only.
Fig. 1 Basic measuring method Basic measuring method is illustrated in figure 1. The forearm is set to desired position (which can be expressed using angle ϕ ) and the force gauge is mounted between frame and wristband. Subject exerts a force F and peak magnitude of the force is captured. This procedure can be repeated for different values of ϕ to get full relationship.
To get sufficient amount of data to evaluate the relationship, we need to measure in at least 6 positions. This implies that the method is time-consuming, because the force gauge and its mounting must be manually repositioned between individual measurements. Moreover, the magnitude of the force depends on actual mood and will of the subject, which can change in time. This results in bad repeability of measurement. This problem can be solved by using single continuous measurement from one limit position to the other instead of performing individual discrete measurements. Actual values of F and ϕ are captured during the motion, therefore we obtain both variables as functions of time. These functions can be further processed to obtain desired relationship.
II. D ESIGN OF M EASURING D EVICE A. Concept The movement must not be controlled by the subject but it must be enforced by a drive unit. The duration of continuous measurement (and resulting speed of movement) must be chosen correctly. If the speed is too high, the dynamic effects can significantly influence measured value of the force. If the speed is too low, the subject can get tired and this can influence the force at the end of the measurement. These requirements are contradictory, the duration of 5 seconds was chosen as a compromise. Because the typical range of ϕ is between 0 and 150 degrees, this implies the speed of movement to be 30 degrees per second. Rapid change of angular position and force requires application of electric transducers. The direction of force on the forearm varies as the forearm moves, but the force vector remains perpendicular to the forearm. This implies the force gauge must change its orientation during measurement or the force must be transferred into the place where the force retains its orientation. The latter approach was used to eliminate potential problems with mobile connection to the transducer. It is necessary to fix both the arm and the forearm (while enabling the rotation in elbow joint) to ensure good repeability of the measurement. Main operational dimensions of fixtures must be adjustable. It is beneficial if the fixtures enable quick mounting and unmounting of subject’s upper limb.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 761–763, 2010. www.springerlink.com
762
P. Hlavoň, J. Krejsa, and M. Zezula
Fig. 2 Kinematic scheme Previously mentioned requirements resulted in kinematic design displayed in figure 2. There is main pulley next to the arm of the subject, the axis of pulley is coincident with the axis of elbow joint. Forearm of the subject is fixed to the pulley near the wrist. A cable is running over the pulley and it is attached to it in one point, so the pulley acts as a one-turn cable drum. The cable runs over another smaller pulley. Axis of smaller pulley is mounted to the base via force transducer, so the transducer indirectly measures the tension in the cable, thet is proportional to the force on the forearm of the subject. The other end of the cable is mounted on the nut, that moves on the screw. The screw is directly driven by electric motor. B. Realization Figure 3 shows CAD model of measuring device, reference [1] covers more design details. The device is mounted on the metal plate with T-grooves. The frame of a tripod shape is built from thin-walled steel profiles. Main pulley is carried in ball bearings mounted on the frame. Arm and forearm fixtures are made of wooden board. All fixtures are open in direction towards subject, who sits next to the device. Such design allows to fix arm and forearm easily, the subject just inserts his/her arm and forearm between the boards. The device is driven by induction motor. Generally, the rotation velocity of induction motor is only little affected by its load. This keeps the steady speed of movement during the measurement without the need of closed-loop speed regulator. The range of motion is limited by two limit switches mounted above opposite ends of the driving screw. The limit switches are triggered by the nut, that moves on the screw. The nose on the nut fits into the T-groove in the metal plate and prevents the rotation of the nut. All control circuitry is placed in the control panel near the motor. Circuitry includes
Fig. 3 CAD model motor reverse switch, start and stop buttons and braking circuit (the motor is braked by powering its windings with DC). The device is capable to measure both left and right arm. To measure the other arm, main pulley bearing house has to be repositioned to the opposite side of the frame and the fixtures have to be adjusted. The device can also measure in both directions (forearm moving up or down) depending on the way the cable is looped around the main pulley. Force transducer U9A by HBM (range 0 to 10 kN, full strain-gauge bridge) was used. The angle transducer is custom made (resolution 1 deg, quadrature encoder) and it is located above the bearing house. The encoder wheel is mounted directly on the main pulley. Both transducers are connected to the Spider8 data acquisition system [2]. During initial tests it was discovered that it is very difficult to switch on the motor and trigger the capturing of data exactly at the moment when the subject starts to exert the force. This was solved using Spider8 internal triggering feature (the capturing starts when some measured value crosses preset threshold). Moreover, Spider8 provides two status signals (MSR and RDY – measure and ready) on its I/O socket. Status signals were used to start the motor and the motor control circuitry was modified accordingly. This further simplifies the whole measurement: subject just sits on the chair next to the device and inserts his/hers arm and forearm between the fixtures. Once the subject starts to exert the force, the rise of the force triggers data capturing and the motor is turned on. When the nut reaches the limit switch, the motor is stopped. The data capturing is stopped by timeout.
IFMBE Proceedings Vol. 29
Measuring Device for Determination of Forearm Force
763 160
III. R ESULTS
140 120 100 F [N]
Obtained data are processed after the measurement. The signal from force transducer is filtered using low-pass filter to remove the noise. The signal from the angle transducer is interpolated using shape-preserving piecewise cubic Hermite interpolant. Because the change of the angle is monotonic, this results in virtually increased resolution of angle transducer. Finally, both courses are matched to get desired relationship between the force and elbow joint angle.
80 60 40 20 0
0
45
90
135 150
phi [deg]
160
Fig. 5 Force/angle relationship for “down” direction
140
F [N]
120 100
IV. C ONCLUSION
80
Figures 4 and 5 shows that maximum force was obtained at ϕ = 80 deg in the direction “up” and at about ϕ = 100 deg in the direction “down”. This corresponds to our expectations and previous preliminary results obtained by the basic measuring method. Measurement results were utilized in mobile motorized orthosis design. The presented measuring device proved its usability for determination of the relationship between the force at the end of forearm and forearm position. The device can be used to compare this relationship between right and left upper limb, or among different people.
60 40 20 0
0
45
90
135 150
phi [deg]
Fig. 4 Force/angle relationship for “up” direction
To illustrate the general course of force/angle relationship, figure 4 shows results from 18 individual measurements. Their arithmetic average is shown in bold. These results were obtained from 3 subjects. Only the right limb was measured. In all measurements, the forearm was moving up, towards the body of subject. Individual courses does not span entire range from 0 to 150 degrees, because boundary parts (where the device accelerates and decelerates) were cropped. Although individual courses are imperfect and contain some peaks, they all exhibit similar trend. Figure 5 shows the results of another 18 measurements with their average shown in bold. In this case, the forearm was moving down, away from the body of the subject, so different muscles were loaded. This explains why the courses are slightly different. The results were obtained from 3 subjects, only the right limb was measured as in the previous case. One course lies away from the cloud of others. This course belongs to an unsuccessful measurement and is an example of the outlier automatically excluded from further processing.
A CKNOWLEDGEMENTS This paper was written with the support of the research project AV0Z20760514 and the project CZ.1.07/2.2.00/07.0406 “Problem Based Learning”.
R EFERENCES 1. Zezula M. Design and realization of measuring device for determination of human arm force depending on elbow joint angle. Brno University of Technology, Faculty of Mechanical Engineering 2009. 2. Hottinger Baldwin Messtechnik GmbH. Spider8 Spider8-30 and Spider8-01 : PC measurement electronics Operating Manual [2006]. Available from: “http://www.hbm.pl/pdf/b0405.pdf”.
Author: Institute: Street: City: Country: Email:
IFMBE Proceedings Vol. 29
Ing. Pavel Hlavoˇn, Ph.D. BUT, Faculty of Mechanical Engineering Technick´a 2896/2 Brno Czech Republic
[email protected] Subclavian Steal Syndrome – A Computer Model Approach C. Manopoulos and S. Tsangaris National Technical University of Athens/School of Mechanical Engineering, Laboratory of Biofluidmechanics & Biomedical Engineering, Greece Abstract— Subclavian steal syndrome is a constellation of signs and symptoms that arise from retrograde (reversed) flow of blood in the vertebral artery or the internal thoracic artery, due to a proximal stenosis (narrowing) and/or occlusion of the subclavian artery. The present study aim to describe this syndrome in patients with subclavian steal steno-occlusive disease based on a computer lumped parameter model. It is studied the cerebral circulation starting from the aortic arch till the arterial branches leaving the circle of Willis. The governing equations of the model are based on the conservation of mass on each node of arteries and on conservation of energy in every loop consisted of artery branches. In addition, the energy loss equation is employed for every artery branch. The system of the equations describing the model is solved with MATLAB. The results show that when the short low resistance path (along the subclavian artery) becomes a high resistance path (due to narrowing) blood flows around the narrowing via the arteries that supply the brain (left and right vertebral artery, left and right internal carotid artery). Keywords— Subclavian Steal Syndrome, Subclavian Steal Steno-Occlusive Disease, Arterial Stenoses, Cerebral Circulation, Circle of Willis.
I. INTRODUCTION The subclavian steal syndrome (SSS) occurs when there is stenosis or occlusion of the subclavian artery proximal to the origin of the vertebral artery. This may cause flow reversal in the ipsilateral vertebral artery as blood is 'stolen' from the circular vertebro-basilar system, to supply the distal territory of the occluded or stenosed artery. Retrograde flow in the vertebral artery, associated with a subclavian or innominate (brachiocephalic) artery stenosis, can be an incidental finding during Doppler ultrasound examination of the cerebral supply. Contorni first described retrograde flow in the vertebral artery in 1960 [1]. Reivich in 1961 first recognized the association between this phenomenon and neurologic symptoms [2]. The same year Fisher dubbed this combination of retrograde vertebral flow and neurologic symptoms subclavian steal syndrome, suggesting that blood is stolen by the ipsilateral vertebral artery from the contra lateral vertebral artery [3]. It was later suggested that such "steal" may cause brainstem ischemia and stroke, either continuously or secondary to arm exercise.
Lord et al. evaluated statistically in vivo the arteriograms and other findings of 42 patients with the subclavian steal syndrome to determine which factors predisposed to vertebrobasilar ischemia [4]. In vitro studies on the syndrome were done by Rodkiewicz et al. in 1992-93 [5] [6]. They used an experimental model of the arterial system in order to show that, for some specific large occlusions magnitude in the left or right subclavian, or in the brachiocephalic artery the blood flow reverses its direction in the left or the right vertebral or right common carotid, or the right internal carotid arteries. It is shown that besides the known single steal syndrome there exists also a double steal syndrome, i.e., blood reverses its flow direction simultaneously in two arteries, both on the right side of the arterial system. This blood is taken from the circle of Willis, which at the same time is significantly supplemented by the increased blood flow through the other arteries leading into the circle of Willis.
II. MATERIALS AND METHODS A. Cerebral Circulation The blood supply of the brain is achieved through a network of blood vessels (arteries and veins) consisting the cerebral circulation. The arteries deliver oxygenated blood, glucose and other nutrients to the brain and the veins carry deoxygenated blood back to the heart, removing carbon dioxide, lactic acid, and other metabolic products. Since the brain is very vulnerable to compromises in its blood supply, the cerebral circulatory system has many safeguards. Failure of these safeguards results in cerebrovascular accidents, commonly known as strokes. Blood is supplied to the human brain by two pairs of arteries: the internal carotid arteries (right & left) and the vertebral arteries (right & left). They have called the supra-aortic arteries. The arteries of the brain originate from the two internal carotid arteries and from the basilar artery, which is formed by the union of the two vertebral arteries (Fig. 1). The intracranial circulation assumes the form of a polygon, called the circle of Willis [7]. The posterior cerebral artery is the anatomical and functional junction between the anterior circulation (carotid system) and the posterior one (vertebro-basilar system) of the circle of Willis.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 764–767, 2010. www.springerlink.com
Subclavian Steal Syndrome – A Computer Model Approach
765
B. The Theoretical Model Figure 2 shows a lumped parameter model focused on cerebral circulation. This model assumes symmetric circulation routes via the brain vessels (with non-flexible walls) from the aorta to the right heart atrium. The model shown in Figure 2 is equivalent to the circuit shown in Figure 1. For simplicity reasons, the local resistances of the individual vessels are expressed in lumped parameters.
Fig. 1 Heart to brain main artery circuit. Are depicted: (1) Circle of WillisCoW, (2) arch of aorta, (3) ascending aorta, (4) descending aorta, (5) right & left main coronary arteries, (6) brachiocephalic trunk or innominate artery-IA, (7) left subclavian artery-SA, (8) right subclavian artery-SA, (9) left [right] common carotid artery-CCA, (10) left [right] external carotid artery-ECA, (11) left [right] internal carotid artery- ICA, (12) left [right] vertebral arteryVA, (13) basilar artery-BA, (14) left [right] posterior cerebral artery-PCA, (15) left [right] posterior communicating artery-PCoA, (16) left [right] middle cerebral artery-MCA, (17) left [right] anterior cerebral artery-ACA, (18) anterior communicating artery-ACoA, (a) right [left] posterior inferior cerebellar artery, (b) anterior spinal artery, (c) right [left] anterior inferior cerebellar artery, (d) pontine arteries, (e) right [left] superior cerebellar artery, (f) right [left] anterior choroidal artery, (g) right [left] ophthalmic artery, (h) anteromedial central (perforating) arteries, (i) Heubner's recurrent artery, (j) right [left] labyrinthine (internal acoustic) artery The blood supply to the brain in a given time defined as cerebral blood flow (CBF). In an adult, CBF is typically 750 ml/min or 15% of the cardiac output. This equates to 50 to 54 ml of blood per 100 grams of brain tissue per minute [8]. CBF is tightly regulated to meet the brain's metabolic demands. The arrangement of the brain's arteries into the circle of Willis (Fig. 1) creates redundancies in the cerebral circulation. If one part of the circle becomes blocked or narrowed (stenosed) or one of the arteries supplying the circle is blocked or narrowed, blood flow from the other blood vessels can often preserve the cerebral perfusion well enough to avoid the symptoms of ischemia [9].
Fig. 2 Lumped-parameter
model of the cerebral circulation shows extra stenosis resistance (Rs) for the right subclavian artery
Assuming the pathological condition where the right subclavian artery demonstrates stenosis equal to a percentage α %≥50 % which is defined as follows: α =ˆ
A0 − As × 100 % A0
(1)
where Α0 is the internal cross-section of the healthy proximal part of the right subclavian artery and Αs the minimum cross-section of the narrowed part of right subclavian artery due to the development of atheromatous plaque. In this circuit, for each side, R1 is the total resistances due to the heart cardiovascular system, R0 is the total resistance after the aortic arch up to the right atrium, R2 is the total of resistances from the end of subclavian artery up to the right atrium, Rs is the resistance that causes stenosis to the right subclavian artery, R3 is the total of resistances from the end of common carotid artery up to the right atrium and R4, R5 and R6 are the total of resistances from
IFMBE Proceedings Vol. 29
766
C. Manopoulos and S. Tsangaris
the end of posterior, middle and anterior cerebral artery, respectively, up to the right atrium. Considering steady blood flow and assuming that the pressure gradient between aorta and right atrium is 80 mmHg with mean blood supply equal to 6.1 l/min (corresponding to the normal total cardiac output) is supposed that a normal reference blood flow distribution is achieved as described by reference [6] from the aortic valve up to the circle of Willis. By this assumption are determined the unknown resistances and the normal distribution of cerebral blood flow of the model taking into account the blood flow symmetry in the brain and the ratio R6/R5=2 according to reference [10]. The resistances of the normal (unoccluded) arteries are calculated by the equation resulting from the Poiseuille formula: Ri−j =
8μL i − j πri4− j
8πμL s 4.8 πμ ρ + + 2 Qs A s2 As A 3s
Re =
4A 0 ρQ s
(3)
which represents the stenosis as a sudden contraction followed by a sudden expansion [11]. In equation (3), Ls=10 mm is the length of the stenosis in the right subclavian artery showing the development area size of atheromatous plaque, ρ=1.055 gr/cm3 is the blood density and Qs is the blood flow-rate in the right subclavian artery. As the stenosis percentage increases the Reynolds number becomes higher due to the turbulence and inertia phenomena
(4)
πμA 0
III. RESULTS AND DISCUSSION The quantitative analysis of the results concerning the blood flow-rates for different degrees of stenosis is shown in the Table 1. The numerical order of the nodes in the first column of the Table 1 shows the direction of normal blood flow when α=0%. Table
1 Dimensionless blood flow-rates Qi-j/(Qi-j)n for several stenosis percentages α,[(Qi-j)n is the normal blood flow of each branch when α=0%]
(2)
where μ=0.0365 poise is the dynamic viscosity of the blood, Li-j and ri-j is the length and the radius of the artery, respectively starting from the i node and ending to the j one. The length and diameter of each arterial segment have been taken from the references [6] and [10]. The solution of the circuit of Figure 2 for the non pathological condition when α=0% is achieved by way of calculating all the blood flow-rates (Qi-j)n of the branches and the unknown resistances. The equations used are based on the Continuity of mass on each node of arteries and on Conservation of energy in every loop consisted of artery branches. In addition, the energy loss equation is employed for every artery branch. The non-linear system of the equations describing the model is solved with MATLAB. After the normal distribution blood flow assessment of the model, the solution of the circuit of Figure 2 is achieved calculating all flow-rates of the branches when the stenosis percentages equal to α=50, 60, 70, 80, 90 and 100 %. The resistance caused by the stenosis in the right subclavian artery is calculated according to the following equation: Rs =
occur especially in the blood outlet from the stenosis point. The Reynolds number is calculated as follows:
Branch (i-j) 2-6 6-17 2-5 2-3 3-4 4-9 5-9 3-7 6-0 7-0 4-0 5-0 9-11 7-14 11-12 11-13 13-10 12-8 17-13 14-12 17-18 14-21 14-15 17-16 15-20 16-19
α 50%
60%
70%
80%
90%
100%
1.0953
1.1376
1.2139
1.3651
1.6870
2.1290
1.1333
1.1926
1.2993
1.5108
1.9612
2.5796
1.0786
1.1151
1.1807
1.3109
1.5882
1.9688
0.9116
0.8716
0.7996
0.6570
0.3532
-0.0638
0.9194
0.8817
0.8139
0.6795
0.3931
0.0000
0.8122
0.7243
0.5663
0.2529
-0.4146
-1.3308
1.1975
1.2893
1.4544
1.7817
2.4789
3.4359
0.9059
0.8643
0.7894
0.6408
0.3245
-0.1097
0.9997
0.9996
0.9994
0.9990
0.9981
0.9969
1.0003
1.0004
1.0007
1.0012
1.0022
1.0037
0.9914
0.9874
0.9801
0.9658
0.9353
0.8934
0.9999
0.9999
0.9998
0.9997
0.9994
0.9990
1.0033
1.0046
1.0068
1.0113
1.0207
1.0336
0.8680
0.8096
0.7045
0.4961
0.0524
-0.5567
1.0144
1.0206
1.0316
1.0535
1.1003
1.1644
0.9923
0.9886
0.9820
0.9689
0.9411
0.9028
0.9954
0.9933
0.9895
0.9820
0.9659
0.9439
0.9953
0.9931
0.9892
0.9814
0.9648
0.9420
1.1085
1.1621
1.2584
1.4495
1.8564
2.4150
0.2916
-0.0173
-0.5729
-1.6747
-4.0211
-7.2423
0.9973
0.9961
0.9939
0.9895
0.9803
0.967
1.0027
1.0039
1.0061
1.0104
1.0196
1.0323
0.6141
0.4433
0.1358
-0.4738
-1.7720
-3.5542
1.4028
1.5818
1.9037
2.5423
3.9020
5.7687
1.0075
1.0108
1.0167
1.0286
1.0537
1.0883
0.9922
0.9887
0.9825
0.9701
0.9437
0.9075
Observing the blood flow-rate values (dimensionless form) of the Table 1 the following behavior of the blood
IFMBE Proceedings Vol. 29
Subclavian Steal Syndrome – A Computer Model Approach
flow distribution is noticed for stenosis percentages α>50%. As the degree of stenosis in the right subclavian artery increases, the flow-rates to the left side increase too in comparison with the ones to the right side, where the flow-rates decrease. Namely, flow-rates from the aortic arch up to the CoW of LSA, LCCA, LICA and LVA increase significantly, while the corresponding ones of RSA, RCCA, RICA and RVA decrease significantly, with the last three reversing their directions for high degrees of stenosis. Figure 3 presents the dimensionless flow-rates of some vessels, along with the increment stenosis in the RSA. 3.5
LVA
3
LICA
2.5 2 1.5
Q/Qn
LCCA LSA
increment area
1
reduction area
0.5 0
RSA RCCA
reversed area
-0.5
767
significantly, while Q11-12 rises a little. Moreover, flow-rate of ACoA is directed from the left to the right (16 to 15) and increases, as the degree of stenosis does the same.
IV. CONCLUSIONS A simple model has been developed in order to describe the subclavian steal syndrome, when the heart to brain main artery circuit is occluded at one location on the right side (e.g. beginning of RSA). It is considered steady blood flow in a circuit with non flexible vessels. It has been shown that when the short low resistance path (along the subclavian artery) becomes a high resistance path (due to narrowing) blood flows around the narrowing via the arteries that supply the brain (left and right vertebral artery, left and right internal carotid artery). Three areas of blood flow are distinguished as the RSA is narrowed. The area of blood flow increment that is on the left side vessels, the reduction and the reversed areas of blood flow that are on the right side. However, further investigation is needed to show the blood flow influence due to more occluded locations (e.g. in the IA and/or LSA), and an unsteady blood flow consideration should be taken into account, in order to achieve more realistic results.
RICA
-1
REFERENCES
RVA
-1.5 0
10
20
30
40
50
α (%)
60
70
80
90
100
Fig. 3 Dimensionless blood flow-rates Q /Qn along with stenosis percentages α, Qn is the normal blood flow of each branch when α=0%
The reversed flow is mentioned to the IA too, since it lies on the right side. Α slight blood flow-rate increment is observed in BA. This increment helps the blood to move (together with the rise of the LICA flow-rate) through CoW around stenosis. In this way the blood circulation is maintained. Due to the high values of resistances R2 and R3 the blood flow-rates to the branches 4-0, 5-0, 6-0 and 7-0 are kept almost constant. Slight variations of blood flow are mentioned in the brain, as the percentage of the stenosis increases. Namely, a small blood flow reduction happens in the LPCA, RPCA, LMCA and LACA, while a small increment takes place in the RMCA and RACA. Finally, in the CoW the blood flow reverses in the RPCoA, as soon as α=60%. More stenosis increment, up to 75%, also makes the blood flow-rate Q14-15 to reverse. Flow-rates Q17-13 of LPCoA and Q17-16 rise
1. Contorni L (1960) The vertebro-vertebral collateral circulation in obliteration of the subclavian artery at its origin. Minerva Chir 15:268-271. 2. Reivich M, Holling H E, Roberts B, Toole J F (1961) Reversal of blood flow through the vertebral artery and its effect on cerebral circulation. N Engl J Med 265:878-885. 3. Fisher C M (1961) A new vascular syndrome: "The subclavian steal". N Engl J Med 265:912-913. 4. Lord R S A, Adar R, Stein R L (1969) Contribution of the Circle of Willis to the Subelavian Steal Syndrome. Circulation XL:871-878. 5. Rodkiewicz C M, Centkowski J, Zajac S (1992) On the subclavian steal syndrome. In vitro studies. J Biomech Eng 114:527-532. 6. Rodkiewicz C M, Centkowski J, Zajac S (1993) On the subclaviancarotid transportation for the subclavian steal syndrome. In vitro studies. J Biomech Eng 115:205-206. 7. Feneis H, Dauber W (2000) Pocket Atlas of Human Anatomy. Thieme, Stuttgart - New York. 8. Nuffield Department of Anaesthetics at: http://www.nda.ox.ac.uk/wfsa/html/u08/u08_013.htm 9. De Boorder M J, Van der Grond J, Van Dongen A J, Klijn C J M, Kappelle J L, Van Rijk P P, Hendrikse J (2006) SPECT measurements of regional cerebral perfusion and carbondioxide reactivity: Correlation with cerebral collaterals in internal carotid artery occlusive disease. J Neurol 253:1285–1291. 10. Hillen B, Hoogstraten H W, Post L (1986) A mathematical model of the flow in the circle of Willis. J Biomech 19:187-194. 11. May A G, De Weese J A, Rob C G (1963) Hemodynamic effects of arterial stenosis. Surgery 53:513-524.
IFMBE Proceedings Vol. 29
Mullins effect in human aorta described with limiting extensibility evolution L. Horny1, E. Gultova1, H. Chlup1,2, R. Sedlacek1, J. Kronek1, J. Vesely1 and R. Zitny3 1
Laboratory of Biomechanics, Faculty of Mechanical Engineering, Czech Technical University in Prague, Prague, Czech Republic 2 Institute of Thermomechanics, Academy of Sciences of the Czech Republic, Prague, Czech Republic 3 Department of Process Engineering, Faculty of Mechanical Engineering Czech Technical University in Prague, Prague, Czech Republic Abstract— Cyclic uni-axial tensile tests with samples of human aorta were performed with an aim to obtain data describing the Mullins effect of arterial tissue. According to presumed anisotropy, reported widely, both samples oriented longitudinally and circumferentially were tested. Each of tested samples underwent cyclic tension up to a particular value of a stretch four times, consecutively maximum limit of reached stretch was increased and subsequent four cycles were performed. Significant stress softening of aortic tissue and residual strains were confirmed. An idealization was made in such a way that reloading and unloading curves are coincident. It was hypothesized that the stress softening observed within reloading of previously loaded tissue may be described by an evolution of material parameters. These parameters should be related to an alternation of internal structure. We proposed a model based on changes in limiting fiber extensibility of fibrillar component of the aortic wall, primarily represented by a collagen. The arterial wall was assumed to be hyperelastic transversely isotropic material with different response under primary loading and unloading. A stored energy function was additively split into isotropic and anisotropic part. Preferred direction in continuum, defined in referential configuration, was assumed to be unchanged with cyclic loading. Every straining level in the cyclic test had its own value of fiber extensibility related to strain maximum previously reached. The isotropic matrix response was modeled using Neo-Hooke term with shear modulus values different under primary loading and reloading, however all reloading values were held the same. The predictions of the model described above were in good agreement with observations. Keywords— aorta, damage, limiting fiber extensibility, Mullins effect, stress softening. I. INTRODUCTION
Significant progress has been made in an area of blood vessels constitutive modeling in last decade. Since 2000 when Holzapfel et al. [1] have published their model of an artery where anisotropy arises from helically arranged bundles of collagenous fibrils, this approach seems to be dominant. Successful applications in computational analyses were reported for example by Cacho et al. [2] in a coronary artery bypass surgery simulation or by Holzapfel et al. [3] in an artery-stent interaction. This model has recently been modified to account for distributed collagen fibrils orientations [4].
A strain energy function based on exponential terms originates from Y. C. Fung and his research in seventies of the last century. This mathematical form was many times validated to be able to capture a material nonlinearity of biological tissues. In 2005 Horgan and Saccomandi [5] suggested a model of the strain energy function motivated by the idea of limiting chain extensibility which has been successfully used in polymer mechanics [6,7]. In [5] this approach was modified to limiting fiber extensibility suitable for composite materials with progressive large strain stiffening. Horny et al. [8] used this model to describe mechanical response of a coronary artery bypass graft and in a constitutive modeling of human aorta [9]. Models mentioned above are capable to describe elastic arterial response. However, it is well known that blood vessels show some inelastic effects [1]. A visco-elastic behavior (dumping, creep, relaxation) can be captured using an internal variables approach [10]. Another inelastic phenomenon in vitro observable in arteries is a stresssoftening similar to the Mullins effect well known in polymer physics. This kind of strain-induced softening is named after Leonard Mullins due to his extensive research on this theme within the last century [11,12]. The Mullins effect in idealized form is described as follows. When previously non-loaded (so-called virgin) material is loaded to a particular value of the deformation, a stress-strain curve is usually called as primary loading curve. Within unloading process the stress-strain curve does not coincide with primary loading and a stress softening is attained. Subsequent loading curve coincide with unloading related to previous cycle. When the value of previous maximum deformation is reached, then the stress-strain curve returns to previous stress maximum and primary loading continues. Described behavior is depicted in Fig. 1. Except Mullins effect soft tissues exhibit further softening within so-called preconditioning that is defined as deformation process necessary to obtain repeatable mechanical response under cyclic loading. The Mullins effect in elastomeric materials is also related with a presence of residual strains and induced anisotropy [13,14]. Many attempts were made to find suitable constitutive description of the Mullins effect [13]. However, the best choice is still in question. There are two main theories in
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 768–771, 2010. www.springerlink.com
Mullins Effect in Human Aorta Described with Limiting Extensibility Evolution
Fig. 1 Typical stress-strain curve recorded for cyclic uniaxial tension of the human thoracic aorta (the strip oriented circumferentially is depicted). Each level of maximum stretch (λ=1.1;1.2;1.3;1.4 and 1.5) was cycled four times. Unloading and reloading curves do not coincide exactly and significant hysteresis remains after four cycles. However, maximum of the stress-softening phenomenon is realized between primary loading and the first unloading.
use. The first approach is based on Continuum Damage Mechanics which operates with a damage parameter considered as an internal variable. The method was proposed by Simo [15] and applied for example by Guo and Sluys [16] or Gracia et al. [17]. Ogden and Roxburgh [18] proposed a pseudo-elastic model to describe the Mullins effect. This was further modified with Dorfmann and Ogden [19] to capture residual strains. Both continuum damage theory and theory of pseudo-elasticity are similar, however the first explicitly involves Clausius-Planck inequality. Above mentioned models are purely phenomenological. Second main approach is physically motivated and several considerations for internal structure of a material are being made [13]. The leading is so called network alternation theory, see e.g. Marckmann et al. [20]. Similar idea may be traced back to Mullins and Tobin [12] who supposed that rubber is a two phase continuum in which stiff phase is transformed to compliant one depending on a deformation. Modern network alternation incorporates knowledge about macromolecular structure of a rubber. Mullins effect is considered to be a consequence of a breakage of links and increasing length of chains, Chagnon et al. [21], thus softening is based on the deformation depending network evolution. The softening effect and pseudo-elastic mechanical response of arteries have been known for a long time [22]. However, only few attempts have been made to develop new theories. Especially in last two years a publication activity has grown. Pena and Doblare [23] used anisotropic pseudo-elastic approach to reproduce the Mullins effect
769
observed in an uniaxial tension of a vena cava. Damage mechanics was employed by Pena et al. [24] in modeling of aortic uniaxial tension. A generalized model based on internal variables applicable to arteries was also proposed by Ehret and Itskov [25]. This model can take account for a preconditioning behavior. The main goal of this paper is to present hypothesis that the Mullins effect observed in uniaxil tension of an artery can be captured by methods of pseudo-hyperelasticity using limiting fiber extensibility depending on maximum previous deformation. There are supposed two ways of material response both governed by hyperelastic description. The first is primary loading and second is unloading/reloading described with changed limiting fiber extensibility parameter. To account for changed response of an isotropic matrix shear modulus values different under primary loading and reloading are also supposed, however constant value for every unloading/reloading is held. II. METHODS
Cyclic uniaxial tensile tests with four samples of human thoracic aorta were performed with MTS Mini Bionix testing machine (MTS, Eden Prairie, USA). Samples were obtained from cadaveric donors with the approval of the ethic committee in the University Hospital Na Kralovskych Vinohradech in Prague. All experiments were performed within 48 hours after the death. Two samples were tested in the direction aligned circumferentially with respect to the natural configuration of the artery and other two were aligned longitudinally. A referential geometry was obtained via digital image analysis performed in NIS-Elements software (Nikon Instruments Inc., Melville, USA). An extension and loading force were measured by MTS testing machine. The cyclic loading was applied as follows: five levels of maximum deformation were performed according to stretch λ=1.1; 1.2; 1.3; 1.4 and 1.5, where λ is the ratio between current length l and referential length L. Each level was cycled four times. It means that for instance after four cycles limited to λ=1.1 the maximum of the deformation was increased to 1.2 and new four cycles performed. Employing incompressibility assumption loading stresses σ were obtained as σ = λ FS −1 (1) In the equation (1) F denotes applied force and S is the cross-section in the reference configuration.
IFMBE Proceedings Vol. 29
770
L. Horny et al. III. MODEL
Arteries were modeled as an incompressible hyperelastic continuum in which the anisotropy rises from the reinforcement by one family of fibers. In such a case constitutive equations can be read in the form of (2). ∂W σ i = λi (2) − p i = 1, 2, 3 ∂λi Here p is the Lagrangian multiplicator determined from boundary conditions. W denotes stored energy function. The model based on limiting fiber extensibility, originally proposed in [5] was incorporated in the form published in [8]. It is expressed in the equation (3). c W = ( I1 − 3) 2 2 λ12 cos 2 β + λ22 sin 2 β − 1 ⎞ (3) μ J f ⎛⎜ ⎟ − ln 1 − ⎜⎜ ⎟⎟ 2 J 2f ⎝ ⎠ Here I1 denotes first invariant of right Cauchy-Green strain tensor, c and μ are stress-like material parameters, and Jf is so-called limiting fiber extensibility parameter (dimension-less). β denotes declination of fibers relative to circumferential direction in a natural configuration of an artery. λ1 is the stretch ration of the strip aligned with circumferential direction of the natural (tubular) configuration of an artery (λ2 is aligned longitudinally). The model (3) is called limiting extensibility model due to the existence of a finite value of a fiber stretch λf (λf2 = λ12cos2β +λ22sin2β) in which the stored energy approaches infinity. Shear strains/stresses were not considered in the model.
(
primary loading curve and μ12 for unloading curves. The limiting extensibility parameter was being varied with successive loading as follows: Jf0 – primary loading; Jf1 – cycle limited to λ=1.1; Jf2 – cycle limited to λ=1.2. Thus, used material parameters can be summarized this way: primary loading parameters – c, μ0, Jf0 and β ; unloading/reloading limited to λ=1.1 – c, μ12, Jf1 and β ; unloading/reloading limited to λ=1.2 – c, μ12, Jf2 and β. Table 1 Material parameters c
μ0
μ12
[kPa]
[kPa]
[kPa]
[1]
[1]
[1]
[°]
83.4
1877
36.9
0.3415
0.0226
0.0917
50.94
Jf0
Jf1
Jf2
)
According to (2) σ1 and σ2 can be computed (σ3=0 to determine p). Eq. (3) generally involves three stretch ratios, reduced to two assuming material incompressibility. However, only displacements in loading directions were recorded (due to the testing machine design). Thus we incorporated additional boundary condition which restricts the value of the transversal stress to be zero (lateral surfaces of the strip are non-loaded). IV. RESULTS
Primary loading curve and the fourth unloading curve for
λ=1.1 and λ=1.2 were included in regression analysis (the idealization assumes unloading=reloading). Material parameters were estimated using weighted least square method in Maple (Maplesoft, Wareloo, Canada). They are summarized in Tab. 1. Parameters c and β were held the same for all data. Shear modulus μ0 was considered for
Fig. 2 Typical stress-strain curves and model fitting.
IFMBE Proceedings Vol. 29
β
Mullins Effect in Human Aorta Described with Limiting Extensibility Evolution V. CONCLUSIONS
It was found that the stored energy density function (3) based on limiting fiber extensibility is capable to describe stress-strain curves obtained from cyclic uniaxial tensile test. Varying two parameters (μ and Jf) was sufficient for the description of the stress softening like Mullins effect. It was suggested that changes in these parameters were straininduced, probably due to an alternation in internal structure of the material. This is only a pilot study. However, results suggest that appropriate description of the Mullins effect in arteries would be obtained if suitable mathematical expression of the dependence of Jf and μ on previous maximum stretch (or deformation history) was found.
ACKNOWLEDGMENT This work has been supported by Czech Ministry of Education project MSM 68 40 77 00 12 and Czech Science Foundation GACR 106/08/0557.
REFERENCES 1.
2.
3.
4.
5.
6. 7.
8.
9.
Holzapfel GA, Gasser TC, Ogden RW (2000) A new constitutive framework for arterial wall mechanics and a comparative study of material models. J Elast 61:1–48 Cacho F, Doblaré M, Holzapfel GA (2007) A procedure to simulate coronary artery bypass graft surgery. Med Bio Eng Comput 45:819– 827 Holzapfel GA, Stadler M, Gasser TC (2005) Changes in the mechanical environment of stenotic arteries during interaction with stents: Computational assessment of parametric stent design. J Biomech Eng Trans ASME 127:166–180 Gasser TC, Ogden RW, Holzapfel GA (2006) Hyperelastic modelling of arterial layers with distributed collagen fiber orientations. J R Soc Interface 3:15–35 Horgan CO, Saccomandi G (2005) A new constitutive theory for fiber-reinforced incompressible nonlinearly elastic solids. J Mech Phys Solids 53:1985–2015 Gent AN (1996) A new constitutive relation for rubber. Rubber Chem Technol 69:59–61 Horgan CO, Saccomandi G (2003) A description of arterial wall mechanics using limiting chain extensibility constitutive models. Biomechan Model Mechanobiol 1:251–266 Horny L, Chlup H, Zitny R, Konvickova S, Adamek T (2009) Constitutive behavior of coronary artery bypass graft, IFMBE Proc. vol. 25/IV, World Congress on Med. Phys. & Biomed. Eng., Munich, Germany, 2009, pp 181–184 Horny L, Chlup H, Zitny R (2008) Strain energy function for arterial walls based on limiting fiber extensibility, IFMBE Proc. vol. 22, 4th European Conference of the International Federation for Medical and Biological Engineering, Antwerp, Belgium, 2008, pp 1910–1913
771
10. Holzapfel GA, Gasser TC, Stadler M (2002) A structural model for the viscoelastic behavior of arterial walls: Continuum formulation and finite element analysis. Eur J Mech A-Solids 21:441–463 11. Mullins L (1969) Softening of rubber by deformation. Rubber Chem Technol 42:339–361 12. Mullins L, Tobin N (1957) Theoretical model for the elastic behavior of filled-reinforced vulcanized rubbers. 30:551–571 13. Diani J, Fayolle B, Gilormini P (2009) A review on the Mullins effect. Eur Polym J 45:601–612 14. Diani J, Brieu M, Vacherand JM (2006) A damage directional constitutive model for Mullins effect with permanent set and induced anisotropy. Eur J Mech A-Solids 25:483–496 15. Simo JC (1987) On a fully three-dimensional finite-strain viscoelastic damage model: Formulation and computational aspects. Comput Methods Appl Mech Eng 60:153–173 16. Guo Z, Sluys LJ (2006) Computational modelling of the stresssoftening phenomenon of rubber-like materials under cyclic loading. Eur J Mech A-Solids 25:877–896 17. Gracia LA, Pena E, Royo JM, Pelegay JL, Calvo B (2009) A comparison between pseudo-elastic and damage models for modelling the Mullins effect in industrial rubber components. Mech Res Commun 36:769–776 18. Ogden RW, Roxburgh DG (1999) A pseudo-elastic model for the Mullins effect in filled rubber. Proc R Soc London A 455:2861–2877 19. Dorfamnn A, Ogden RW (2004) A constitutive model for the Mullins effect with permanent set in particle reinforced rubber. Int J Solids Struct 41:1855–1878 20. Marckmann G, Verron E, Gornet L, Chagnon G, Charrier P, Fort P (2002) A theory of network alternation for the Mullins effect. J Mech Phys Solids 50:2011–2028 21. Chagnon G, Verron E, Marckmann G, Gornet L (2006) Development of new constitutive equations for the Mullins effect in rubber using the network alternation theory. Int J Solids Struct 43:6817–6831 22. Fung YC, Fronek K, Patitucci P (1979) Pseudoelasticity of arteries and the choice of its mathematical expression. Am J Physiol 237:H620–H631 23. Pena E, Doblare M (2009) An anisotropic pseudo-elastic approach for modelling Mullins effect in fibrous biological materials. Mech Res Commun 36:784–790 24. Pena E, Pena JA, Doblare M (2009) On the Mullins effect and hysteresis of fibered biological materials: A comparison between continuous and discontinuous damage models. Int J Solids Struct 46:1727–1735 25. Ehret AE, Itskov M (2009) Modeling of anisotropic softening phenomena: Application to soft biological tissues. Int J Plast 25:901– 919 Author: Institute: Street: City: Country: Email:
IFMBE Proceedings Vol. 29
Lukas Horny Fac. of Mech. Eng. Czech Technical University in Prague Technicka 4 Prague Czech Republic
[email protected] A distribution of collagen fiber orientations in aortic histological section L. Horny1, J. Kronek1, H. Chlup1,2, R. Zitny3 and M. Hulan1 1
Laboratory of Biomechanics, Faculty of Mechanical Engineering, Czech Technical University in Prague, Prague, Czech Republic 2 Institute of Thermomechanics, Academy of Sciences of the Czech Republic, Prague, Czech Republic 3 Department of Process Engineering, Faculty of Mechanical Engineering Czech Technical University in Prague, Prague, Czech Republic Abstract–Distributions of collagen fibrils and smooth muscle cells nuclei (SMC) orientations were investigated in histological sections obtained from medial layer of human thoracic aorta. The sections were stained with van Gieson. Digital image of the sections was converted to binary pixel map with target component represented in white (logical unity). Selected image was indicated to elucidate the sensitivity to threshold conditions and three different binary conversions were performed. Consecutively images were processed by in house developed software BinaryDirections which uses an algorithm of the rotation line segment to determine significant directions in digital images. The algorithm operates in the way that in each target pixel a line segment is rotated step by step to explore neighborhood of the pixel. Exploring the neighborhood the number of unity pixels in each rotating step is determined. The distribution of orientations in the entire image is gained after normalization either as averaged density distribution from all pixels or as an histogram of the most abundant directions in the image. It was found that both collagen fibrils and SMC nuclei analyses give significant peak in distributions. Its value ranges between 45° - 65° (defined as declination from longitudinal axis of an artery in a tubular configuration) depending on the method. It implies that preferred direction in aortic medial layer was oriented circumferentially rather than longitudinally. This conclusion was almost independent of the threshold setup. Results suggest that the orientations of SMC nuclei and collagen fibrils are mutually correlated and determination of collagen fibril orientations, which may be stained in insignificant manner, could be supported with SMC nuclei orientations to obtain more realistic models. Keywords– aorta, anisotropy, collagen, fiber distribution, histology, probability density. I. INTRODUCTION
Innovative biomedical engineering moves towards patient-specific (bio-)artificial implants and computational simulations. It involves a development of sophisticated constitutive models for biological tissues. These models have to account for complex material structure. Biological tissues comprise large number of different cells, matrix proteins and bonding elements. Moreover, there are many complex interrelations between these constituents. If an arterial wall is considered, there are three main layers usually distinguished, Holzapfel et al. [1]. Each layer has different function and a composition. Briefly, innermost
layer called tunica intima consists of endothelial cells which rest on basal lamina. There is also a subendothelial layer dominated with collagen and smooth muscle cells (SMC). Tunica intima and subendothelial layer are very thin and compliant in physiological state; however they may become mechanically considerable in pathological situation, see Holzapfel et al. [3]. Tunica media, middle layer of an artery, has distinctive three-dimensional network structure. It is arranged into so-called musculoelastic fascicles which consist of SMC, bundles of collagen fibrils and elastin. They are separated with fenestrated elastin laminae and repeat through medial layer [2]. Directional arrangement of fibrillar components in media is helical with small pitch (almost circumferential orientation) [1,2]. Tunica adventitia is the outermost layer where especially fibroblasts, fibrocytes and collagen fibers are presented. Collagens form two helically arranged families of fibers with significantly dispersed individual fibers. Above mentioned layers are separated by elastic laminae. However, described structure is valid only in the case of elastic artery. Histological structure of muscular arteries may differ significantly. For more details see [1] and [2] and references therein. Models of internal structure are being implemented in constitutive theories. Nowadays constitutive theories based on continuum mechanics can incorporate information about fibrillar structure to predict appropriate anisotropic behavior of a material. The anisotropy may rise from either finite number of preferred directions or their continuous distribution (for finite number of directions see Holzapfel et al. [1,6] or Lanir [4]; and for continuous distribution Gasser et al. [2], Lanir [4] or Driessen et al. [5]). It is also possible to build up a model with composite structure (mixture theory) which will be intrinsically multiphase. Well known is the fact that fibrillar components of biological tissues are usually crimped. Models accounting for an uncrimping and successive straining within a loading process have also been proposed; see Lanir [4], Freed and Doehring [7] or Cacho et al. [8]. This paper deals with internal structure of arterial wall based on observations made within an inspection of histological sections. There are different ways to obtain such information. Analysis of histological sections and employing transmitted light microscopy (TLM) is probably the oldest and simplest one. Nevertheless any drawbacks
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 772–775, 2010. www.springerlink.com
A Distribution of Collagen Fiber Orientations in Aortic Histological Section
exist. First of all a section is mechanically loaded during sectioning and chemical fixation (in formalin) may also change internal structure. Moreover this method does not seem to be suitable for an investigation of rearrangements in internal structure under a loading, what is of cardinal importance, because of complicated load fixation. However, such papers exist; see Sokolis et al. [9], they reported straightening of elastic lamellae under uniaxial tension. The polarized light microscopy (PLM) may be used in the same way because of the collagen and SMC birefringence, Canham et al. [10], [11]. Both studies revealed approximately circumferential orientations of collagen fibrils and SMC in coronary arteries and internal mammary artery. Small-angle scattering methods may be used as an alternative to the optical microscopy. Small-angle X-ray scattering (SAXS) seems to be capable to reveal a relation between fibers orientation and external load in a nano-scale, Schmid et al. [12]. Small-angle light scattering (SALS) method may also be employed rather than for nano-scale in micro-scale investigations of collagen organization in a tissue, Sacks et al. [13]. The main goal of this paper is to present findings obtained within investigation of histological sections of human abdominal aorta using digitalized images. In house software BinaryDirections with implemented algorithm of the rotating line segment (RLS), see Horny et al. [14], was used to quantify an organization of collagen fibrils and nuclei of SMC. As first RLS algorithm is described and obtained results follow. II. METHODS
Histological sections were obtained from abdominal aorta of 36-years-old male donor. They were stained with van Gieson and digitalized under 10 x magnifications. Four images were the same as in [14] and were used to confirm previous results. It was due to the fact that the algorithm in BinaryDirections (RLS) has been improved in 2009. Additional image from medial layer was indicated to examine if the correlation between collagen fibers orientation and SMC nuclei exists. This image was also used in the analysis of threshold sensitivity to elucidate an influence of a binary conversion on results; threshold of RGB filter which transforms stained collagen to white (logical unity) pixels and non-collagen components to black (logical zero) pixels. III. RLS ALGHORITHM
Exact mathematical formulation of the rotating line segment (RLS) was described in details in [14]. Here only
773
key ideas will be repeated with emphasis on improvements made in 2009. The RLS processes binary pixel maps as follows. In non-zero pixel (target pixel) of the processed image the line segment is step by step rotated in order to evaluate the number of non-zero pixels in the target pixel neighborhood. The evaluation is based on so-called matching coefficient, C(α), which is normalized number of non-zero pixels shared with the line segment and neighborhood of target pixel at given rotating step α. Thus RLS is two-parametric algorithm which depends on given α and positive integer N. N is the linear dimension of the neighborhood (square with side length equal to N). The neighborhood and the line segment are represented as matrices with elements equal to zero or one depending on filtered pixels’ values. The rotation of the line segment is performed as a creation of new matrix with different nonzero elements. The number of non-zero pixels (collagen) in given direction (rotating step) is obtained as a product between elements in neighborhood matrix and line segment matrix, where corresponding positions in matrices are multiplied and the products are summed. This number is then normalized with respect to dimensions of selected neighborhood and the line segment, thus matching coefficient C(α) is obtained. There are two ways to reveal relevant information about directional frequency of non-zero (collagen) pixels in the neighborhood of target pixel. First, one can reduce information to the most abundant angle (rotation step) only, and create an histogram over entire image. A dominant direction distribution is then obtained. If full information is stored then average distribution of collagen pixels frequency over entire image can be computed after normalizing to unit surface. If multiple maxima in the matching coefficient are obtained, they are added to their corresponding orientations weighted with multiplicative inverse of their multiplicity (newly added). The second improvement implemented in BinaryDirections software is a penalty function which suppresses contribution of almost uniform distributions in averaging procedure. IV. RESULTS
Selected results for processed images are shown in Fig. 4–6. Fig. 1 shows original digital image of used histological section (abdominal aortic media). Filtering this image to collagen bundles is in Fig. 2; smooth muscle cell nuclei, easy distinguishable in Fig. 1, are filtered in Fig. 3. Fig. 2 and Fig. 3 are overlaid with orange grid spaced every 100 pixels to highlight length scales. A distribution of orientations is build up using information from a pixel neighborhood, thus its dimension (N) should be chosen with respect to dimensions of image texture.
IFMBE Proceedings Vol. 29
774
L. Horny et al.
Fig 1 Additional histological section obtained from abdominal aortic media; stained with van Gieson. Bundles of collagen fibrils are stained in red. Black nuclei of smooth muscle cells are also apparent.
Fig. 4 Empirical probability density distribution of collagen fiber orientations obtained from the binary map in Fig. 2. Sensitivity to neighborhood dimension is shown.
Fig. 5 Comparison between empirical probability density distributions of Fig. 2 Binary pixel map of the section from Fig. 1 filtered with respect to collagen (white). Orange grid spacing is 100 pixels; 0.64μm/pix.
Fig. 3 Binary pixel map of the section from Fig. 1 filtered with respect to SMC nuclei (white). Orange grid spacing is 100 pixels; 0.64μm/pix.
collagen fibers orientations obtained form Fig. 2 with different filter thresholds and empirical probability density distribution of smooth muscle cell nuclei in Fig. 3.
Fig. 6 Comparison between empirical probability density distribution for collagen orientation and smooth muscle nuclei orientation for medial section published in [14].
IFMBE Proceedings Vol. 29
A Distribution of Collagen Fiber Orientations in Aortic Histological Section
Fig. 4 shows small sensitivity of the texture (Fig. 2) on N. Density peak is located approximately at 55°. Filtering to target component is not unambiguous process and may depend on individual skills of an operator. To find out this influence the section (Fig. 1) was binary converted three times with small perturbations in collagen threshold. Fig. 5 shows that all images gave similar results, the peak at ≈55°. Due to existence of so-called musculoelastic fascicles in aortic media it was hypothesized that the correlation between collagen fibers and SMC orientations may exist. In the case of histological section in Fig. 1 (and its binary conversions above mentioned) this hypothesis was confirmed (Fig. 5). Both methods revealed density maximum between 45°-65°. Fig. 6 shows that the section previously analyzed in [14] suggests correlation between SMC and collagen orientations. However, additional analyses of further sections did not provide this result in every case.
REFERENCES 1.
2.
3.
4. 5.
6.
7. 8.
V. CONCLUSIONS
The results suggest that the rotation line segment may be powerful instrument in analyses of arterial architecture. It was shown that analyzed histological image from human aortic media (Fig. 1) gives continuous density distribution with maximum at app. 55° for the collagen (the angle is defined with respect to longitudinal axis in cylindrical configuration). Similar result was obtained within the analysis of smooth muscle cell nuclei orientations (continuous distribution with one peak located between 45°65°). This may not be a rule, but may serve as confirmation for purely collagen-based analyses. The correlation between SMC nuclei and collagen fibrils orientations may be expected especially within medial layer where so-called musculoelastic fascicles are presented. Prospective mismatches between SMC and collagen distributions may be attributed to finite thickness of histological section (5 μm in this study) which may contain more that one fascicle. Results show that preferred direction is oriented circumferentially rather than longitudinally, however the orientations density maximum does not meet exactly circumferential direction as in [6]. Within analysis of histological sections published in [14] with improved algorithm, obtained results were similar in character with reported in [14].
775
9.
10.
11.
12.
13.
14.
Holzapfel GA, Gasser TC, Ogden RW (2000) A new constitutive framework for arterial wall mechanics and a comparative study of material models. J Elast 61:1–48 Gasser TC, Ogden RW, Holzapfel GA (2006) Hyperelastic modelling of arterial layers with distributed collagen fiber orientations. J R Soc Interface 3:15–35 Holzapfel GA, Sommer G, Gasser CT, Regitnig P (2005) Determination of layer-specific mechanical properties of human coronary arteries with nonatherosclerotic intimal thickening and relative constitutive modeling. Am J Physiol Heart Circ Physiol 289:2048–2058 Lanir Y (1983) Constitutive equations for fibrous connective tissues. J Biomech 16:1–12 Driessen NJB, Bouten CVC, Baaijens FPT (2005) A structural constitutive model for collagenous cardiovascular tissue incorporating the angular fiber distribution. J Biomech Eng Trans ASME 127:494– 503 Holzapfel GA, Gasser TC, Stadler M (2002) A structural model for the viscoelastic behavior of arterial walls: Continuum formulation and finite element analysis. Eur J Mech A-Solids 21:441–463 Freed AD, Doehring TC (2005) Elastic model for crimped collagen fibrils. J Biomech Eng Trans ASME 127:587–593 Cacho F, Elbischger PJ, Rodríguez JF, Doblaré M, Holzapfel GA (2007) A constitutive model for fibrous tissues considering collagen fiber crimp. Int J Non-Linear Mech 42:391–402 Sokolis DP, Kefaloyannis EM, Kouloukoussa M, Marinos E, Boudoulas H, Karayannacos PE (2006) A structural basis for the aortic stress-strain relation in uniaxial tension. J Biomech 39:1651– 1662 Canham PB, Finlay HM, Dixon JG, Boughner DR, Chen A (1989) Measurements from light and polarised light microscopy of human coronary arteries fixed at distending pressure. Cardiovasc Res 23:973–982 Canham PB, Finlay HM, Boughner DR (1997) Contrasting structure of the saphenous vein and internal mammary artery used as coronary bypass vessels. Cardiovasc Res 34:557–567 Schmid F, Sommer G, Rappolt M., Schulze-Bauer CAJ, Regitnig P, Holzapfel GA, Laggner P, Amenitsch H (2005) In situ tensile testing of human aortas by time-resolved small-angle X-ray scattering. J Synchrot Radiat 12:727–733 Sacks MS, Smith DB, Hiester ED (1997) Small angle light scattering device for planar connective tissue microstructural analysis. Ann Biomed Eng 25:678–689 Horny L, Hulan M, Zitny R, Chlup H, Konvickova S, Adamek T (2009) Computer-Aided Analysis of Arterial Wall Architecture. IFMBE Proc. vol. 25/4, World Congress on Med. Phys. & Biomed. Eng., Munich, Germany, 2009, pp 1494–1497
Author: Institute: Street: City: Country: Email:
ACKNOWLEDGMENT This work has been supported by Czech Ministry of Education project MSM 68 40 77 00 12 and Czech Science Foundation GACR 106/08/0557.
IFMBE Proceedings Vol. 29
Lukas Horny Fac. of Mech. Eng. Czech Technical University in Prague Technicka 4 Prague Czech Republic
[email protected] An Innovative Approach for Right Ventricular Volume Calculation during Right Catheterization P. Toumpaniaris1, I. Skalkidis1, S. Markatis2, and D. Koutsouris1 1
National Technical University of Athens / School of Electrical and Computers Engineering, Biomedical Engineering Laboratory, Athens, Greece 2 National Technical University of Athens / School of Applied Mathematics and Physical Sciences, Athens, Greece
Abstract— In this paper, an innovative approach is presented for the right ventricular volume calculation during right catheterization. The right catheterization is really important in critically ill patient management. However, there is a concern about the sufficiency of outcomes from pulmonary artery catheter (PAC) which is used for right catheterization. In order to improve the reliability of the particular catheterization, the additional parameter which should be used is the volume of right ventricle chamber and it is proposed to be measured with the aid of ultrasonic technology. In order to compute the volume, five circular groups of six ultrasonic sensors are mounted along the catheter surface. Each sensor measures the distance from the corresponding vertical point on the ventricular wall forming five hexagons in five different planes. Between any two of them a polyhedron of volume V is created . The sum of the polyhedra which are formed in the cavity is equivalent to the volume of the right ventricle chamber. As this method is dependent only on the distances measured by the ultrasonic sensors and the fixed separation distance between the groups of sensors, it is expected to be a reliable method in volume calculation during catheterization. Keywords— Right Catheterization, Right Ventricular Volume.
I. INTRODUCTION Heart failure, sepsis and multiple organ failure are the main factors for the use of right catheterization. The pulmonary artery catheter is generally used in the intensive care unit for the estimation of hemodynamic condition in critically ill patients. It is a powerful diagnostic and monitoring tool that has been used extensively in the assessment of cardiovascular physiology [1]. The pulmonary artery catheter allows direct, simultaneous measurement of pressures in the right atrium, right ventricle, pulmonary artery, and the filling pressure ("wedge" pressure) of the left atrium. So far, the right catheterization of heart evaluates only pressures. Hemodynamic monitoring is essential for treatment of critical ill patients. However the existing techniques have a lot of restrictions in their basic technical parameters [2]-[4].
The problem can be solved if this method is enriched with the precise calculation of right ventricle volume. Right Ventricle (RV) volume could be calculated with high accuracy with method of magnetic resonance (MRI). However, with this method of measurement, the volume will be recorded at a specific instance. Repeated volume measurements of the right ventricle for a time period of a few hours in a hemodynamic laboratory or in an intensive care unit, could contribute in efficient management of heart failure patients, since potential changes of right ventricle’s volume can cause alteration of parallel conductance volume and thus affect the left ventricle’s mechanical parameters. Unlike the left ventricle, the shape of which is close to ellipsoid and not difficult to evaluate it, the right ventricle shape is mostly a mixture of crescent and conical shape which easily could be regarded as a complex nongeometrical shape making its evaluation more difficult [5]. There are many studies presented in the literature for volume measurement of cardiac ventricles [6]-[8]. Most of them use methods related with measurements from outer surface, such as MRIs and ultrasounds on the top of the skin. However there are a few works where the volume measurements are taken from its inner space [8] but no one known to the authors, related with real time RV volume measurement using PAC. This paper will introduce a novel method of right ventricle volume measurement using pulmonary artery catheter.
II. ARCHITECTURE The PAC is a catheter with length usually at 110cm and external diameter of 7French (or 2.33mm). In order to access the right heart it enters from a large vein which often is the internal jugular or the subclavian vein and though the superior vena cava enters the right atrium as shown in Figure 1. Through the tricuspid valve is inserted to the right ventricle cavity and then curves in order to enter through the pulmonary valve to the pulmonary artery [9].
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 776–779, 2010. www.springerlink.com
An Innovative Approach for Right Ventricular Volume Calculation during Right Catheterization
777
Fig. 1 Right Catheterization procedure Fig. 2 Ultrasound Beam from PAC The curvature of the catheter in the right ventricle chamber has the form of an arc of a circle and could be considered to be the same in any heart as the location and the distance between the valves is approximately unchanged. Therefore, as a result from the above, if a number of ultrasonic sensors are mounted on the catheter surface with certain geometry, the angle of ultrasound beam from the sensor to the ventricular wall will be constant. Each ultrasonic sensor measures the distance from the sensor to a point of the wall of the cavity as shown in Figure 2. In fact, it measures the time that the ultrasound beam takes to reach the cavity and return to the point of emission. Since the speed of sound is known it is easy to calculate the distance d by the formula:
d=
ct 2
Where c is the velocity of sound in the medium (in blood= 1566m/s) and t the sound travel time (from transmission of sound to echo reception). We assume 5 circular dissections on the tube of the catheter. Let the total length of the catheter that is inside the RV be which can be considered as a piece of circular arc. We will appoint the 5 sections where the ultrasound transducers (UsT) will be mounted with a separation of /5 between them along the catheter, so that the first point will be just after the tricuspid valve (the valve from which the catheter enters to right ventricle chamber) and the 5th point is just before from the pulmonary valve (the valve from which the catheter leaves the ventricle). In each of the 5 points, 6 UsT will lay rotatively on the surface of the catheter with 60o separation.
III. METHODOLOGY The catheter inside the RV is considered to be of circular shape, of length , K is considered to be the center of the circle and ρ the radius. So ρ=ΚΟ1 =ΚΟ2 =ΚΟ3=ΚΟ4=ΚΟ5 The length of the arc (Ο1Ο5)= and (Ο1Ο2)= (Ο2Ο3)= (Ο3Ο4)= (Ο4Ο5)= /5 The angle The planes u1, u2, u3, u4, u5, that the UsT’s define in each position are perpendicular to the catheter and all of them pass through the line which is perpendicular to the plane of the circle that crosses through the center K. In every plane of transducers there exist 6 UsT crosssectional lines in pairs of two, with 60o angle between them. In the planes u1, u3, u5, the UsT pair Ai∆i is in the radius ΚΟi while the planes u2, u4, the UsT BiEi is in the perpendicular to the radius ΚΟi, considering that Α, Β, Γ, ∆, Ε, Ζ are the points where the ultrasound beam reached the wall of the RV and O is the point where the UsT is, as shown in Figure 3. These are true since UsTs from plane to plane are twisted by 30o.
IFMBE Proceedings Vol. 29
Fig. 3 Hexagons created by consecutive UsT beams
778
P. Toumpaniaris et al.
The solid volume is defined by the planes for example u1, u2 and has the shape that is shown in Figure 4
Fig. 5 Height calculation
Fig. 4 Solid volume between consecutive planes The volume V12 between the planes u1, u2 is comprised by the sum of the following: V12= Vpyramid(Ο2Α1Β1Γ1∆1Ε1Ζ1) + Vtetrahedrons(Ο2Ζ2Ζ1Α1 + Ο2Ζ2Α2Α1 + Ο2Α2Α1Β1 + Ο2Α2Β2Β1+ Ο2Β2Β1Γ1+ Ο2Β2Γ2Γ1+ Ο2Γ2Γ1∆1+ Ο2Γ2∆2∆1+ Ο2∆2∆1Ε1+ Ο2∆2Ε2Ε1+ Ο2Ε2Ε1Ζ1+ Ο2Ε2Ζ2Ζ1) where Vpyramid(Ο2Α1Β1Γ1∆1Ε1Ζ1) = Area(base)·height = ε·h ε = Area (Α1Β1Γ1∆1Ε1Ζ1) = Area(Ο2Α1Β1) + Area(Ο2 Β1Γ1) + Area(Ο2 Γ1∆1) + Area(Ο2 ∆1Ε1) + Area(Ο2 Ε1Ζ1) + Area(Ο2Z1A1) = (Ο2Α1)( Ο2 Β1)·sin 60 +… = rarb·sin 60 +… where ra is the distance that the UsT Ai has calculated and rb is the distance that the UsT Bi has calculated. The height h is the length of the projection of O2 to the plane u1 = O2Θ. In the triangle Ο2ΚΘ (where Θ is a right angle) sin =ρ · sin So h= Using the above the volume of the pyramid is calculated. The volume of the tetrahedron Ο2A2B2B1 is Ο2Α2B2·height(B1). The area of Ο2Α2B2 is computed similar to the base of the pyramid. The height(B1) O1β = O1B1cos60 = rB1cos60 βb = (Kβ)·sinφ = (KO1+O1β)·sinφ = (ρ+Ο1B1cos60)·sinφ = (ρ+rB1cos60)·sinφ
So it can be calculated and the same procedure is followed for Ζ1, Α1, Γ1, ∆1, Ε1. For the points Γ1, ∆1, Ε1 it should be noted that –rΓ1 should be used and for A1 because the angle is 90 the cos90 will be 1. So the volume for the tetrahedrons Ο2Ζ2Α2Α1, Ο2Α2Β2Β1, Ο2Β2Γ2Γ1, Ο2Γ2∆2∆1, Ο2∆2Ε2Ε1, Ο2Ε2Ζ2Ζ1 can be calculated. To calculate the tetrahedrons Ο2Ζ2Ζ1Α1, Ο2Α2Α1Β1, Ο2Β2Β1Γ1, Ο2Γ2Γ1∆1, Ο2∆2∆1Ε1, Ο2Ε2Ε1Ζ1 we consider a coordinate system Kxyz. The coordinates are A1(ρ+rA1,0,0) Z1(ρ+rZ1cos(-60), rZ1sin(-60),0) O2(ρcosφ, 0, ρsinφ) Z2(ΚΞcosφ, rZ2sin(-30), ΚΞsinφ)=((ρ+rZ2cos(-30))cosφ, rZ2sin(-30),(ρ+rZ1cos(-30))sinφ) The volume Ο2Ζ2Ζ1Α1 is the absolute value of the determinant. 1 1 VolumeΟ2Ζ2Ζ1Α1= = 1 1 ሺɏ ͳǡͲǡͲሻ ͳ ͳ ሺɏ ͳ
ሺെͲሻǡ ͳሺെͲሻǡͲሻ ͳ ተ ተ ሺɏ
ɔǡ Ͳǡ ɏɔሻ ͳ ሺሺɏ ʹ
ሺെ͵Ͳሻሻ
ɔǡ ʹሺെ͵Ͳሻǡ ሺɏ ͳ
ሺെ͵Ͳሻሻɔሻ ͳ
So the total volume of the solid that is created between two consecutive planes can be calculated and the total volume of the RV will derived by summing up the 4 solid volumes created by the 5 groups of sensors.
IV. DISCUSSION The volume measurement of right ventricle during right catheterization is an important factor for the estimation of
IFMBE Proceedings Vol. 29
An Innovative Approach for Right Ventricular Volume Calculation during Right Catheterization
hemodynamics of critically ill patients. Therefore the accuracy in measurement is compulsory. A thorough effort is been paid to the appropriate layout geometry of the ultrasonic sensors as they will build along the pulmonary artery catheter’s surface, so that the RV volume could be measured. Because of the complex shape of right ventricle it is difficult to define a method that fulfils all the requirements and limitations that arise in order to measure its volume, as the volume should be measured by the distances taken from ultrasonic sensors mounted along the PAC. In a theoretical aspect, the particular method seems to meet those requirements to a large extend. As the MRI method is considered to be a gold standard in ventricles volume measurement, a calibration of this proposed method is suggested so that a number of right ventricles volumes are to be measured both with MRI method and the proposed method. The mean difference between two methods could be regarded as an offset, the algebraic value of which should be taken into account in the new method for calibration.
REFERENCES 1. Takala J, (2006) The pulmonary artery catheter: the tool versus treatments based on the tool. Critical Care 10:162 2. Harvey S, Harrison DA, Singer M et al. (2005) Assessment of the clinical effectiveness of pulmonary artery catheters in management of patients in intensive care (PAC-Man): a randomised controlled trial. Lancet 366(9484):472-7
779
3. Shah MR, Hasselblad V, Stevenson LW et al. (2005) Impact of the pulmonary artery catheter in critically ill patients: meta-analysis of randomized clinical trials. JAMA 294(13):1664-70 4. Binanay C, Califf RM, Hasselblad V et al. (2005) Evaluation study of congestive heart failure and pulmonary artery catheterization effectiveness: the ESCAPE trial. JAMA 294(13):1625-33 5. Czegledy F, Aebischer N, Tarnari A et al. (1999) A new mathematical model for right ventricular geometry. Annual Intemational Conference of the IEEE Engineering in Medicine and Biology Society 12:4:18131814 6. Fritz D, Rinck D, Unterhinninghofen R et al. (2005) Automatic segmentation of the left ventricle and computation of diagnostic parameters using regiongrowing and a statistical model. SPIE Medical Imaging 1844–1854 7. Fritz D, Rinck D, Dillmann R et al. (2006) Segmentation of the left and right cardiac ventricle using a combined bi-temporal statistical model” Medical Imaging 2006: Visualization, Image-Guided Procedures, and Display. SPIE 6141:605-614 8. Vazquez de Prada JA, Jose A, Chen MH et al. (1996) Intracardiac echocardiography: in vitro and in vivo validation for right ventricular volume and function. The American heart journal 131:2:320-328 9. Headley JM. (1989) Invasive hemodynamic monitoring: Physiological principles and clinical applications. Baxter Healthcare
Author: P. Toumpaniaris Institute: NTUA – Biomedical Engineering Laboratory Street: Iroon Polytechniou 9, Zografou City: Athens Country: Greece Email:
[email protected] IFMBE Proceedings Vol. 29
Service composition to support ambient assisted living solutions for the elderly V.Moumtzi1, C.Wills2 and A. Koumpis1 1
2
ALTEC Software S.A./ Research Programmes Division, Thessaloniki, Greece Kingston University/Faculty of Computing, Information Systems & Mathematics, London, UK
Abstract— Recent studies show that over the last decade electronic systems designed, created and continuously developed with the common goal to help the elderly live a more independent life. Each of the existing electronic systems offer services that cover different needs of ageing people living, such as self-determined life at home, social interaction, etc. In this paper we propose a set of service components that constitute a model for an e-service architecture capable to support the autonomy of older people. Furthermore, we argue that by exploiting existing operational electronic systems and integrate them according to the services they offer we can possibly converge to an ideal system that enhances assisted living of elderly people. Keywords— Ambient Assisted Living (AAL), elderly, eservices, Service Oriented Architecture, autonomy.
I. INTRODUCTION
The older population is on thethreshold of a boom [1]. According to U.S. Census Bureau projections, a substantial increase in the number of older people will occur during the 2010 to 2030 period. As a result of the aforementioned projection, there will be a dramatically increasing need to support the independent living of this globally growing elderly population. The fact that physical and mental routines are getting more difficult influences their whole lives. Institutions and medical facilities support old people needs, especially those caused by illness, however, there have been many efforts helping assisted people living . Technological solutions seem to be the way out of the expanding costs of health, help and support. [2] Taking into account the increasing number of elderly population in Europe and the identification of its subsequent social and financial consequences, national and European research efforts have focused on such independent living solutions. Apart from e-Health solutions, the field of Ambient-Assisted Living (AAL) has been developed, aiming on alleviating the difficulties of every day life for the elderly or people with disabilities in general. Senior person notification in case of an emergency, sensors monitor movements, person location, cognitive and emotional processes and comfort services are some of the main features implemented separately from different such systems.
The challenge is to take advantage of the many possibilities that existing Ambient Assisted Living technology systems entail but that take account of the fears that may arise amongst users when introducing new forms of AAL and integrate them appropriately. Thus the impact of the research will be to stimulate ICT innovation in favour of the elderly. II. THE ELDERLY NEEDS
It is very important to clarify what is it that the elderly require to satisfy their needs to enable them to enjoy a high level of quality of life. In this respect, needs mean a lack of something. These needs influence behavior as well as cognitive and emotional processes. Often there is a separation between primary needs, e.g. biological / physiological deficiency such as hunger, and secondary needs, e.g. all desires and wishes that determine man’s satisfaction on a personal, intellectual, (ir-) rational, social or cultural level. [3] We consider emergency treatment to be the kernel of any assistive living system. It aims at the early prediction of and recovery from critical conditions that might result in an emergency situation and the safe detection and alert activation in emergency situations such as sudden falls, heart attacks, strokes, etc. Autonomy enhancement services is the term used to denote all services that make it possible to supplant or replace previous direct care administered by medical and social care personnel or relatives, and replace it by appropriate system support. Comfort services cover all areas that do not fall into the above categories. Examples of comfort services are social contact assistance, logistic assistance, etc. [4] Therefore the types of services are categorised into three types of services which should provide the systems set up to support the needs of elderly: (a) Emergency treatment (b) Autonomy enhancement (c) Comfort III. THE CURRENT SOLUTIONS
Substantial advances have been made over recent years in applying technology to meet the needs of older people. In
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 780–783, 2010. www.springerlink.com
Service Composition to Support Ambient Assisted Living Solutions for the Elderly
the field of Ambient-Assisted Living (AAL) has been developed. This offers a sense of safety and reassurance to the elders themselves and their relatives that they will receive the care required in a time of need, without having to be succumbed to intensive care. A. Emergency treatment solutions AttentiaNet and Seniority comprise two already completed projects which aim at improving the quality of assistance and hence quality of life of elder people in Europe by utilizing advanced technologies for telemonitoring and telecommunications. A different approach is followed by MobilAlarm which enables older people to initiate an alarm call whenever and wherever they need to do so (using GPS; mobile telephony; body-worn alarm devices; service centres; geographic localisation and alerting software). More recently, a number of solutions have been proposed that utilize various sensor networks, from audio and movement to micro- and nano- sensors hand, to detect common elderly accidents, like falls. Projects Netcarity, INHOME, EMERGE and OLDES fall under this category. Finally, the R&D project called Companionable, which is still in its initial stages, will try to synergistically combine the strengths of a mobile robotic companion with the advantages of a stationary smart home, improving the elder person’s interaction with the system as well as the care itself. B. Autonomy enhancement solutions Besides AAL systems, a second dimension has been introduced in the field of ICT solutions for the elderly, comprising of the systems that aim on compensating for their cognitive decline. In accordance with mainstream eInclusion targets, this approach’s objective is to retain elderly people socially active and more self-reliant for a wider period of time. An example of such projects is VM (Vital Mind), which provides cognitive training by using related psychology, a TV-set and advanced ICT. The reasoning of VM is to enable elders to exercise actively and autonomously in front of the familiar to them television medium. On the other hand, the FP7 HERMES project aims at providing an integrated approach to cognitive care, based on assistive technology that reduces age-related decline of cognitive capabilities. HERMES offers cognitive training through games, while also supporting them in indoor as well as outdoor environments, when necessary. Support for elderly people with cognitive disabilities, and especially mild dementia and Alzheimer’s disease, is provided by COGKNOW which aims to develop a cognitive prosthetic device which will help elder “navigate through their day”.
781
C. Comfort solutions T-Seniority is an innovative ICT-solution especially fitted in areas of public interest that is directly addressing the main themes of the i2010 initiative, especially focusing in ICT for accessibility by the elderly and their social integration, reinforcing both European and existing national initiatives. T-Seniority Services are designed to be accessible to the greatest number of people and thus seeks to minimize designs that exclude the elderly from the benefits of Information Society, by thinking beyond the conventional media and creating a better overall experience throughout TV. [6] VITAL project proposes a combination of advanced information and communication technologies that uses a familiar device like the TV as the main vehicle for the delivery of services to elderly users in home environments. IV. THE DEVICE LIMITATIONS
It is highly recommended that technology systems which facilitate independent living of ageing people should be accessible through technological devices that elderly can use easily. In the following we list specific requirements that AAL devices should fulfill. [5] The main Design requirements directly related to the ability of handling the device are physical ones. Regarding the portable devices apart from the decreasing the size and weight of the device itself we should keep acceptable size of buttons and displays. More specifically, elderly people require having buttons as large as their finger, definitely not smaller. Pressing a button needs to give a sound alert so that the user can feel and hear that the action is done. Easy operation of device is tightly connected with the number of control elements. In case of simple inputs, two or three buttons (Yes, No, Quit) with obvious meaning can be implemented by the elderly without discomforting him. However a little bit more complex devices can be also used by elderly if there is a minimum balance among the buttons such as Enter and Escape buttons, which is enough to navigate through menus. Regarding the supply, wireless devices connected with the host system should have a power- off button in case of danger or malfunction the user is given a chance to say stop. In addition battery powered devices should be autorecharging showing a signal (like a light or sound alert) to remind the elderly recharge batteries by himself. On the other hand, devices which are wired connected with the host system should have one cable. The designer of a device for elderly should take into account that (s)he is not capable of solving installation conflicts (missing drivers, port numbers). The designer should
IFMBE Proceedings Vol. 29
782
V. Moumtzi, C. Wills, and A. Koumpis
be careful that the devices will not occurrence of any trouble discourages user from using relatively good device. V. THE PROPOSED SOLUTION
Our research aims to propose a well-defined and complete set of services that an integrated home based ICT solution should provide to facilitate the needs of elderly people we addressed above, emergency treatment, autonomy enhancement and comfort. Firstly, the recommended solution should provide cognitive and physical training for elderly people inside the framework and safety of an assisted living environment. The service should be installed in homes and institutions and ensure the accident-free, personalized and monitored corporal and mental training of its users. Meanwhile, users should be able to take advantage of the features of an independent living solution. This can be accomplished by home automations that compensate for the disabilities of people with cognitive problems or mild dementia during their daily activities. Finally, an elaborate distributed sensor network would guarantee immediate response in case of an emergency, by calling for help through public telephone lines. In this respect, the proposed solution aims to offer emergency treatment needs. Extensive research on this topic has proven that sensory training is one of the most effective countermeasures against cognitive deterioration and mild dementia. Furthermore, corporal activation is a well acknowledged factor of healthy living, having significant positive effects in the physical well-being of people. As regards to elderly autonomy enhancement, assistance services at home are necessary for this kind of system. Featuring typical AAL solution characteristics should be deployed, like accident detection, support in daily activities and third party notification in case of emergency, and combining them with preemptive measures like training. Finally, the ICT solution we suggest to be appropriate for various groups of elderly people also according to their individual social needs to guarantee safe, socially rich independent life. To this end, the elderly should be able to choose among many different options of public or personalized services: communicate with their relatives, friends and colleagues; ask for shopping, repairs, appointments, on-line banking, etc. The ICT solution should offer a flexible combination of general public services and personalized services demonstrating the versatility of its technological platform according to the user’s preferred or available ICT media. To accommodate the deliverance of these services the suggested solution should utilize state-of-the-art hardware
and software technology. In particular: x
User Interfaces: touch screens or simple screens for interaction with the users. All system functionalities, including home environment management, cognitive training and physical exercising performance monitoring, are displayed and set from the Local User Interface, which is a touch screen. x Sensors: to monitor movements inside the house. They are deployed as a distributed network of wirelessly connected sensors that identify moving patterns and detect deviations from those patterns or falls. In these cases they notify the user’s relatives or caretakers via the Remote User Interface. x Facility to connect Instrumented Power Outlets: which are sensors measuring voltage and current (power) fed into appliances. These sensors detect of forgotten switched on electrical appliances. x Both aforementioned sensor components guarantee the safe living of the elderly inside their home environments without need for exclusive intensive care. x Facility to connect Actuators: which facilitate acts like opening windows, doors and blinds and are remotely operated with the Local User Interface. x Processing units: an embedded processor and a general purpose PC which are used for coordinating and management of the AAL environment, for executing the cognitive training software and for storing and processing the physical training performance information. x Use digital TV as the most widely available and preferred channel for info-marginated sectors, helping to reach difficult-to-reach audiences, such as “disabled people getting older”, who may have less access to other forms of digital technology, improving current situation and affording the demands of a growing elderly population. The concept of the proposed integrated ICT solution comprises a cluster of different kind of users, needs and devices that should be totally replace their vertical integration into a virtual one. SOA is an architectural style in which software applications are organized as a set of loosely coupled services. The service-orientation paradigm advocates building functionality from services that align with functional processes. It is a process and a path to achieving diverse business goals. SOA is being adopted by many enterprises today because it results in agile IT systems that can be more easily adapted in response to change. The SOA paradigm enables the linking of business and computational resources - mainly organizations, applications and data - on demand. It is seen to be essential for delivering business agility and IT flexibility [8][9].
IFMBE Proceedings Vol. 29
Service Composition to Support Ambient Assisted Living Solutions for the Elderly
The proposed architecture of the AAL integrated ICT solution is shown in Fig. 1. The central point of user interaction is the application and media layer which should aggregate the content available via enabling technologies and composite services. The suggested components which comprise the integrated system are the following: x Integration technologies: Data store and history of collaborators which should be access via a web service. x Reasoning technologies: the nature of technologies which support elderly autonomy. x Service packaging: a software module that handles different components. x Portal: A portal from which the user has access to the system. x Web x: Web applications that can transfer the information that the user wants to share and access on the World Wide Web. x Windows x: Software operating environment which is an interface between hardware and user that acts as a host for computing applications run on the machine. x Devices and people: People insert service alerts to devices to interact with the ICT system. x Emergency treatment, autonomy enhancement and comfort: services provided to the final user.
initiative, today eInclusion is an issue affecting all European countries. Even if the technology system towards the improvement of people’s lives is growing rapidly unfortunately it has been unknown and unexploited by a significant amount of people, especially from elderly. The wider uptake of existing ICT technology as a means to facilitate the assisted living of senior citizens constitutes one of the most significant targets of our solution. To sum up, the proposed solution presented in this paper contributes towards a number of problems that arise all over Europe. It covers the technological gap between elderly. It offers them the possibility to monitor their health status, it motivates them to follow a personalized training program, thus strengthening their cognitive and physical capabilities, it helps them to use interactive e-services and finally reduces the marginalization chances against people of the Third Age.
REFERENCES 1. 2.
3.
4.
5.
6.
Fig. 1 Scalable Service Oriented Architecture for integrated AAL ICT solutions
7.
8.
VI. CONCLUSIONS
In this paper we tried to list a set of services focused on the needs of elderly people that are necessary for their independent living and in relation to these, getting benefit from the existing ICT technology, we proposed a solution which accommodates independent living and provide useful, familiar and comprehensive content. Our vision is to improve the current situation and meet the demands of a growing elderly population. As it has been demonstrated by the European e-Inclusion
783
9.
He W, Sengupta M, et al (2005) Current Population Reports, Series P23-209 65+ in the United States: 2005, December 2005 Fuchsberger V. (2008) Ambient assisted living: elderly people's needs and how to face them. 1st ACM international workshop on Semantic ambient media experiences, Vancouver, British Columbia, Canada, 2008, pp21-24 Dickinson A, Mick J, Gregor S. (2006) Designing a portal for older users: A case study of an industrial/academic collaboration. ACM Transactions on Computer-Human Interaction (TOCHI), September 2006, ISSN:1073-0516, 13(3), pp 347 - 375 Nehmer J, et al. (2006) Living assistance systems: an ambient intelligence approach. The 28th international conference on Software engineering, Shanghai, China May 20 - 28, 2006, ISBN:1-59593-375-1, pp 43-50. Smrz J. (2009) More than user-friendly I/O devices. The 2nd International Conference on PErvsive Technologies Related to Assistive Environments, Corfu, Greece 09 - 13 June 2009. ISBN:978-1-60558409-6. ICT PSP - 224988 Project T-Seniority (2008-2011), Expanding the benefits of Information Society to Older People through digital TV channels Schoeni, et al, (2001), Persistent, Consistent, Widespread, and Robust? Another Look in Old-Age Disability, Journals of Gerontology Series B: Psychological Sciences and Social Sciences, Vol. 56B, No. 4, pp. S206–S218. Harding C. An Open Marketplace for Services (2007) DM Review, vol. 17, September, 2007, 20-39. Tews R, Beyond IT: The business value of SOA (2007), AIIM EDOC, vol. 21, 14-17, Sept/Oct. 2007
Author: Institute: Street: City: Country: Email:
IFMBE Proceedings Vol. 29
Vasiliki Moumtzi ALTEC Software S.A 6, M.Kalou str. 546 29 Thessaloniki Greece
[email protected] Oscillations in subthalamic nucleus measured by multi electrode arrays J. Stegenga1, T. Heida1 1
MIRA institute, dept EEMCS, BSS-group, University of Twente, Enschede, The Netherlands
Abstract— The subthalamic nucleus (STN) of the basal ganglia, is involved in the generation of Parkinsonian symptoms and forms one of the main targets for Deep Brain Stimulation (DBS). Effective frequencies of DBS are around 130 Hz. The effect of such stimuli in the STN is largely unknown but has been hypothesized to result in neuronal block, interrupting the pathophysiological oscillatory behavior which is observed in the Parkinsionian basal ganglia. Modelling studies suggest that synchronized oscillation at tremor (4-8 Hz) or beta (14-30 Hz) frequencies may occur. To study synchronicity of the STN in detail, we record actionpotential activity from rat brain slices using multi electrode arrays (MEAs). These arrays consist of 60 recording sites and thus allow the study of spatio-temporal activity patterns. Here we show the characteristics of spike trains which we recorded in the STN.
may self-perpetuate through mediation by the Ttype calcium current. This membrane current causes cells to depolarize after prolonged inhibition, leading to bursts of action potentials when released from inhibition. The possibility of such a mechanism in the basal ganglia has been demonstrated using organotypic culture preparations [4]. 1.4 mm
Keywords— Parkinson’s disease, multi electrode arrays, brain slices, oscillatory activity I. INTRODUCTION
The origin of the three cardinal symptoms of Parkinson’s Disease (tremor, rigidity and bradykinesia) is speculated to lie in the basal ganglia. Specifically, it is in the substantia nigra pars compacta of the basal ganglia that deterioration of dopaminergic neurons occurs. The loss of subsequent dopaminergic innervation to the basal ganglia, the striatum and also the motor cortex ultimately leads to PD symptoms. From a number of studies it has been shown that oscillatory activity in the lower frequency bands ( 0.5 [17,18]. In addition, the differences in correlation coefficients of each Q with different components (factor loadings) should be > 0.20 [18,19]. For exploring the possible correlation of the four aforementioned scales with various socio-demographic variables, and self-perceived health status, Spearman rank correlation coefficient was utilized. Descriptive statistics were calculated for the sociodemographic variables. In the univariate analysis, the relationships among these variables, prior admissions, and survey completion logistics variables with the four satisfaction scales were studied. Moreover, the Student’s t test, the analysis of variance (ANOVA), and the Kruskal-Wallis test were utilized for continuous variables, and the Chi-square test or Fisher's exact probability test for categorical variables. The threshold p value for statistical significance (2-sided) was set to 0.05. Statistical analysis was performed by Statistical Package for Social Sciences (SPSS v. 16.0).
III. RESULTS A. Patient Characteristics The mean age of participants was 58.9 years old; 53.4% were men, 73.0% married, 51.5% came from the city of Alexandroupolis, 33.7% had a university degree, and 44.2% had never been hospitalized in this hospital before. B. Principal Component Analysis (PCA) Two components were found which constructed two summated (multi-item) scales. The two components (C1 and C2) explained 83.479% of total variance. Q22 and Q29 gave similar factor loadings and were excluded from further analysis. Q17-21 (satisfaction mainly from medical staff) seemed to relate to C2 with a Cronbach’s coefficient of 0.829. Removal of Q17 increased the coefficient’s value to 0.947. On the other hand, Q24-28 and Q30-32 (satisfaction from accommodation and administration) seemed to relate to C1 with a Cronbach’s coefficient of 0.949. By removing Q24, internal consistency improved (coefficient=0.966). The responses in Q18-21, Q24-28, Q30-32 (and not Q17 and Q24,) of the two multi-item scales were grouped and two new summated scales were constructed; satisfaction from accommodation/administration and medical services. The mean satisfaction from medical services (88.880±1.103) was larger in comparison with satisfaction computed for accommodation/administration (74.167±1.423). The first one also presented larger minimum values (50% vs. 17.86%). Mean satisfaction from nursing services was
IFMBE Proceedings Vol. 29
794
G.K. Matis et al.
slightly smaller as compared to the physician one (84.259±1.265). The global satisfaction was even lower (73.313±1.494). Females, Greek citizens, patients in the 19-35 age group, and those with elementary education and with only one prior admission presented the best scores. The biggest reported mean medical satisfaction score was also found in married subjects living in semi-urban areas. Divorced and patients working in public sector reported the biggest satisfaction score with respect to the nursing services. High satisfaction derived from accommodation-administration and global satisfaction was found mainly in unmarried brain cancer patients living in other than Alexandroupolis urban centers, and having Pronoia (poor patients) as their insurance provider. C. Correlations of Four Summated Scales with Socio-demographic Variables No variable was found to relate to Q33 (global satisfaction). Age was related only to the nursing satisfaction (rs=0.181, p=0.021) (older patients were more satisfied). Citizenship related only to accommodation/ administration satisfaction (rs=0.158, p=0.044) (Greeks more satisfied). Self-perceived health status related to medical (rs=-0.157, p=0.045) and nursing (rs=-0.168, p=0.032) satisfaction at admission (patients with better selfperceived health were more dissatisfied). No such correlation was found for medical (rs=-0.060, p=0.446) and nursing satisfaction (rs=-0.056, p=0.483) at discharge. Finally, gender, location, insurance provider, level of education, marital status, and number of prior admissions did not seem to relate to any satisfaction scales.
IV. DISCUSSION The majority of the variables presented in this paper have been previously studied in other Greek satisfaction surveys [10,20]. However, to our knowledge, there is no published study implementing all these variables, and by no means is there a report of hospitalised brain cancer patients’ satisfaction investigation in the region of Thrace, Greece. The questionnaire completion after hospitalization seems to limit positive bias, since patients are no longer “hostages” of the hospital system [21]. Our data suggest that the four dimensions of satisfaction were affected by age, citizenship, and self-perceived level of health at admission. In our patients global satisfaction was found to be 73.313% similar to other Greek studies carried out in a general patient population; Polyzos et al (2005) reported a rate of 3.98 (in a scale 0-5) [22]. No variable was statistically related to this dimension. Other researchers
linked older age and lower education with high satisfaction [10,23,24]. However, satisfaction in patients with an age of > 80 years seems to decline [25]. In contrast to our findings, it has been documented that women [26] and married patients [27] are more satisfied with hospital care. However, Quintana et al (2006) reported higher levels of satisfaction in men and in the single/divorced subgroup [28]. Satisfaction from medical services was 88.88% (in other Greek hospitals ranged from 80.4% to 92.8%) [10,20,22] and satisfaction from nursing services was 84.259% (in other Greek hospitals ranged from 52.4% to 90.0%) [10,20,22]. Both of these dimensions were correlated with poor self-perceived health status at admission. Interestingly, patients with prior admissions were not found to be more demanding, as published research states [28], in which the bigger the number of prior admissions, the bigger the satisfaction scores. In addition, other studies linked good self-perceived health status at admission to a higher satisfaction score [23,29]. It should be stressed that patients generally give bigger rates to physicians, since they do not fully understand their services, and that the lack of nursing personnel plays a crucial role in acquired responses [30]. The accommodation/administrative dimension was graded with 74.167%. Surveys conducted in Greece reported a satisfaction rate for administrative services ranging from 75% to 96.2% [20,22], while Niakas et al (2004) reported a satisfaction score for accommodation services of 75.9% [10]. No association between length of stay or number of prior admissions and satisfaction on domains such as cleanliness was found, even though this is frequently the case in many international studies [23,28]. Certain limitations of this study should be underlined. Firstly, Q18 and Q21 (professional efficiency of physicians and nurses) could be problematic since patients rarely possess the necessary knowledge to judge on this matter. Secondly, all questions were considered as ratio scales. However, other researchers treat them as ordinal data [31] or even as interval scales [32]. Thirdly, the investigation of patient satisfaction is a dynamic and evolving process which demands not only the participation of large number of patients from different hospitals at various points in time, but also the management of lack of sensitivity of questionnaires to changes in patients’ satisfaction as well [10,24,33]. Finally, it has been documented that satisfied patients tend to answer more often [34]. Similar studies in American and European countries highlight the importance of health personnel participation in educational activities such as seminars and postgraduate courses [35]. Of major importance seems to be the adoption of a patient-centred communication, a notion incorporating patient understanding through his/her own values and special psychosocial environment [36,37].
IFMBE Proceedings Vol. 29
Satisfaction Survey of Greek Inpatients with Brain Cancer
V. CONCLUSION These results were in line with similar Greek studies, demonstrating a uniformity in patients’ responses throughout the Greek territory, even though our sample was limited to patients diagnosed with brain cancer [10,20,22,38,39].
REFERENCES 1. Andrzejewski N, Lagua RT (1997) Use of a customer satisfaction survey by health care regulators: a tool for total quality management. Public Health Rep 112:206-210 2. Cafferky ME (1997) Managed care & You. Health Information Press, Los Angeles, pp 167-176 3. Donabedian A (2003) An introduction to quality assurance in health care. Oxford University Press, New York, pp 46-57, 139-143 4. Enthoven AC, Vorhaus CB (1997) A vision of quality in health care delivery. Health Affairs 16:44-57 5. Cunningham L (1991) The quality connection in health care. Integrating patient satisfaction and risk management. Jossey-Bass Inc, Publishers, San Francisco, pp 105-144, 148-155 6. Urden LD (2002) Patient Satisfaction Measurement. Current Issues and Implications. Outcomes Manage 6:125-131 7. Bittner MJ, Booms BH, Tetreault MS (1990) The service encounter: diagnosing favorable and unfavorable incidents. J Mark 54:69-82 8. Zeithaml VA, Berry LL, Parasuraman A (1995) The nature and determinants of customer expectations of service. In: Bateson JEG (ed). Managing services marketing. Text and readings. The Dryden Press, Fortworth, pp 7-56, 133-147, 233-248 9. Bar-Tal Y, Barnoy S, Zisser B (2005) Whose informational needs are considered? A comparison between cancer patients and their spouses’ perceptions of their own and their partners knowledge and informational needs. Soc Sci Med 60:1459-1465 10. Emanuel LL, Alpert HR, Baldwin DC et al (2000) What terminally ill patients care about: toward a validated construct of patients’ perspectives. J Palliat Med 3:419-431 11. Lipsman N, Skanda A, Kimmelman J et al (2007) The attitudes of brain cancer patients and their caregivers towards death and dying: a qualitative study. BMC Palliat Care 6:7 12. Gonzalez N, Quintana JM, Bilbao A et al (2005) Development and validation of an in-patient satisfaction questionnaire. Int J Qual Health Care 17:465-472 13. Likert R (1932) A technique for the measurement of attitudes. Arch Psychol 140:1-55 14. Kline P (1994) An Easy Guide to Factor Analysis. Routledge, London, pp 28-41, 56-79 15. Kaiser HF (1970) A second-generation Little Jiffy. Psychometrika 35:401-415 16. Nunnally JC, Bernstein ICH (1994) Psychometric Theory. 3rd ed. McGraw-Hill, New York, pp 25-58 17. Westaway MS, Rheeder P, Van Zyl DG et al (2003) Interpersonal and organizational dimensions of patient satisfaction: the moderating effects of health status. Int J Qual Health Care 15:337-344 18. Labarere J, Francois P, Bertrand D et al (1999) Outpatient satisfaction: Validation of a French-language questionnaire: Data quality and identification of associated factors. Clin Perform Qual Health Care 7:63-69 19. Cronbach LJ (1951) Coefficient alpha and the internal structure of tests. Psychometrika 16:297-334
795 20. Niakas D, Gnardellis C (2000). Inpatient satisfaction in a regional hospital of Athens. Iatriki 77:464-470 (in greek) 21. Ford RC, Bach SA, Fottler MD (1997) Methods of measuring patient satisfaction in health care organizations. Health Care Manage Rev 22:74-89 22. Polysos N, Bartsokas D, Pierrakos G et al (2005) Comparative surveys for patient satisfaction between hospitals in Athens. Arch Hell Med 22:284-295 (in greek) 23. Nguyen Thi PL, Briancon S, Empereur F et al (2002) Factors determining inpatient satisfaction with care. Soc Sci Med 54:493-504 24. Campbell SM, Roland MO, Buetow SA (2000) Defining quality of care. Soc Sci Med 51:1611-1625 25. Jaipaul KC, Rosenthal GE (2003) Are older patients more satisfied with hospital care than younger patients? J Gen Intern Med 18:23-30 26. Hargraves JL, Wilson IB, Zaslavsky A et al (2001) Adjusting for patient characteristics when analyzing reports from patients about hospital care. Med Care 39:635-641 27. Hall JA, Dornan MC (1990) Patient sociodemographic characteristics as predictors of satisfaction with medical care: a meta-analysis. Soc Sci Med 30:811-818 28. Quintana JM, González N, Bilbao A et al (2006) Predictors of patient satisfaction with hospital health care. BMC Health Serv Res 6:102110 29. Jackson JL, Chamberlin J, Kroenke K (2001) Predictors of patient satisfaction. Soc Sci Med 52:609-620 30. Hyrkãs K, Paunonen M, Laippala P (2000) Patient satisfaction and research-related problems (part 1). Problems while using a questionnaire and the possibility to solve them by using different methods of analysis. J Nurs Manage 8:227-236 31. Finkelstein BS, Singh J, Silvers JB et al (1998) Patient hospital characteristics associated with patient assessments of hospital obstetrical care. Med Care 36:AS68-78 32. Krowinski WJ, Steiber SR (1996) Measuring and managing patient satisfaction. American Hospital Publishing, Illinois, pp 40-49 33. Garratt AM (2007) Parent experiences of paediatric care (PEPC) questionnaire: reliability and validity following a national survey. Acta Paediatr 96:246-252 34. Mazor KM, Clauser BE, Field T et al (2002) A Demonstration of the Impact of Response Bias on Results of Patient Satisfaction Surveys. HSR 37:1403-1417 35. Weiner JS, Cole SA (2004) Three principles to improve clinician communication for advance care planning: overcoming emotional, cognitive, and skill barriers. J Palliat Med 7:817-829 36. Van den Brink-Muinen Α (2002) The role of gender in healthcare communication. Patient Educ Couns 48:199-200 37. Vieder JN, Krafchick MA, Kovach AC et al (2002) Physician-patient interaction: What do elders want? JAOA 102:73-78 38. Gnardellis C, Niakas D (2005) Factors influencing inpatient satisfaction. An analysis based on the Greek National Health System. Int J Healthcare Technol Manage 6:307-320 39. Labiris G, Niakas D (2005) Patient satisfaction surveys as a marketing tool for Greek NHS hospitals. J Med Mark 5:324-330
Author: Georgios K. Matis Institute: Democritus University of Thrace Street: Dragana City: Alexandroupolis Country: Greece Email:
[email protected] IFMBE Proceedings Vol. 29
End stage renal disease patients’ projections using Markov Chain Monte Carlo simulation A.Rodina, K. Bliznakova, N. Pallikarakis University of Patras, Department of Medical Physics, BITU Patras, Greece
Abstract— End stage renal disease (ESRD) patients’ number continuous increase remains one of the major problems of this century. In 2007 about 11300 ESRD patients were in need of renal replacement therapy (RRT) in Greece. Same year Greece have been on the 8th place internationally in the prevalence per million population (pmp) in RRT and on the 4th place in the incidence pmp of ESRD patients, dropping only behind United States, Japan and Germany. ESRD treatment methods are considered to be one of the most expensive procedures for chronic conditions worldwide. Predicting future patients’ number on RRT is essential for health care providers in order to achieve more effective resource management. In this work we present a Markov Chain Monte Carlo simulation for predicting the future number of ESRD patients for the period 2008-2012 in Greece. Continuous increase in incidence and prevalence was projected, resulting in 26% prevalence increase in 2012 compared to 2007. Keywords— End-stage renal disease, Markov model, Monte Carlo simulation. I. INTRODUCTION
End stage renal disease (ESRD) remains one of the major problems of the 21st century, despite the progress in medical technologies and disease preventing programs. It is a chronic disease related to complete or almost complete failure of kidney function. During the last decade, Greece has been among the countries with the highest number of incident and prevalent ESRD patients [1, 2]. Since 2002, Greece has never been lower than the 5th place in incidence and the 10th place in prevalence internationally. In 2007, there were about 11300 patients in Greece in need of renal replacement therapy (RRT). There are three major types of RRT available: hemodialysis (HD), peritoneal dialysis (PD) and transplant (TX). Each therapy affects the patients’ quality of life in a different degree. TX and home HD are considered to provide better quality of life of RRT patients compare to other therapies [3-5]. An investigation on cost analysis of the available RRT in Greece, performed in 2002, revealed that organ transplantation, telematic HD and home HD, besides improving patients’ quality of life, could significantly reduce treatment costs [6, 7]. Future prognosis of RRT patients is
essential for the health care authorities and policy makers to reveal the best scenarios for better resource allocation and technology assessment as well as predicting the future demands of the RRT in Greece. We have previously reported on a deterministic approach using Markov model to predict the number of ESRD patients per age group and therapy for Greece for the period 2007-2012 [8]. In this model, the transition probabilities are constant and they were derived from several assumptions due to limited data. In order to evaluate the results of the model and therefore the model itself, a sensitivity analysis should be performed. This analysis will set the variations of the model outputs as a function of the uncertainties of the input model parameters such as transition probabilities and treatment assignment probabilities. This can be achieved in several ways. Probabilistic sensitivity analysis by means of Monte Carlo techniques is among the techniques that enable easy implementation and simultaneously testing the uncertainties of the input parameters. The purpose of this work is to extend the available deterministic Markov chain model to involve Monte Carlo techniques in predicting the future number of ESRD patients in Greece for the period 2008-2012. The next step of this work is to introduce probabilistic sensitivity analysis into created model. II. METHODS
The Markov chain Monte Carlo (MCMC) comprises Monte Carlo techniques used to sample from probability distributions of the constructed Markov chain. The Markov chain consists of three mutually exclusive treatment states: hemodialysis (HD), peritoneal dialysis (PD), kidney transplantation (TX) and death, that is the absorbing state (figure 1). The possible transitions between the treatments are shown by the arrows. The backward bending arrows show that it is possible for patients to remain in the state they were in during the previous cycle.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 796–799, 2010. www.springerlink.com
End Stage Renal Disease Patients’ Projections Using Markov Chain Monte Carlo Simulation
HD
Death
PD
Tx
797
The incidence for the age groups 45-65 and >65 was modeled by a linear regression model based on data available from 1998-2007. It was observed that during the period 1998 to 2007 the number of incident patients less than 45 years of age fluctuated between 173 and 203. Therefore, the incidence for this age group was represented by the averaged number of incident patients for the period 1998 to 2007.
Fig. 1 The possible transitions of the patients between the states of the Markov chain model. Y = start year NP = prevalence (Y)
The future number of prevalent patients was calculated by the following equation: P(t+1) = P(t)*TM + I(t+1)*Id
(1)
Where t is a time, TM is the transition matrix, I is an incident patients’ number and Id is incident patients’ initial assignment probability to RRT. Monte Carlo techniques are used to sample the probabilities of the patient’s initial therapy assignment and the probability to change current therapy or to die. Figure 2 shows the realization of the MCMC model, while figure 3a and 3b show in details the implementation of Monte Carlo techniques. The simulation starts with tracing the prevalence patients from the initial year, i.e. 1998. Each patient is individually followed for the period 1998 – 2012 until the final year is reached or the patient enters the death state. For each year, the algorithm simulates the probability a patient to change the current therapy. This is depicted in figure 3a. All prevalent patients in the starting year are followed in this way. Further on, each subsequent year, there are incident patients in need to start RRT. Their initial assignment to treatment, i.e. HD, PD or TX is based on sampling from probability distributions obtained from available data for the period 1998 to 2007 (figure 3b). Each set of probabilities consists of three treatments resulting in 100% overall probability [8]. One set of the initial assignment probabilities is chosen randomly out of ten available. The statistical data used in this study were extracted from the European Renal Association - European Dialysis and Transplant annual reports covering the period from 1998 to 2007. Patients were classified by age into three groups: 65. The model incorporates 48 possible transitions made up of three 4x4 transition matrices. Transition matrices for each age group were predetermined based on the work of Boletis [9] and Ioannidis et al [10] and in accordance with associate nephrologists.
Trace incident patients m
Y=m+1
Y=m NP1
Y > end year YES
Assign patient to therapy (3a)
YES DEATH NO
NO NI = incidence (Y)
NI1 Define the treatment (3b)
Y=Y+1 Save patients in the stack
NO
Y > end year
YES
NIi = Ni + 1
NPi = NPi + 1
NPi > NI NO
NPi > NP
NO
YES NP = NI(Y)
YES END
Fig. 2 MCMC model for patients on RRT. NP is the number of prevalent patients, NI is the number of incident patients and Y is a year.
IFMBE Proceedings Vol. 29
798
A. Rodina, K. Bliznakova, and N. Pallikarakis
The model adequacy was tested by the data splitting approach. Data from period 1998-2002 were used to project the year-end prevalence for the period 2003-2007. The highest absolute error value was less than 5%.
Incident patients in need of RRT
Read patient’s state S
(3b)
(3a) Generate a uniformly distributed random variable n
Generate a uniformly distributed random variable n
YES
n < PXĺHD
HD
YES
patients’ incidence projection by age group. 1998-2007: real data, 2008-2012: projection. Linear trend was used to project the incidence for the age groups 45-65 and >65.
n < PHD NO
NO n (PSĺHD ) && n 65). In the rest two groups, 45. Linear trends in incidence have been reported in other countries as well. This fact has been explained by population ageing and development of medical technology being effective in co-morbidities and chronic disease treatment [11]. The predicted and experienced average annual growth of about 5-6% for both prevalence and incidence is comparable with the other studies [12]. Taking into account the evidence of RRT patients’ number constant increase, the prediction of the future trends is considered to be a valuable tool for healthcare decision makers in order to achieve a more efficient and reasonable resources allocation and management. At the moment home HD, which is reported to be more cost-effective compare to in-center HD and PD, is not well introduced in Greece [13]. Similarly there is insufficient number of transplantations resulting in long waiting lists. The reasons for the low number of transplants are reported to be lack of organization on the transplant delivery and cultural-religious points of view [6]. Further to, the transplantation is known to be the most cost-effective treatment with the most potent effect on patients’ quality of life [3-6]. Introducing new technologies
1. 2. 3. 4. 5. 6.
7.
8.
9. 10. 11. 12. 13.
European Renal Association and European Dialysis and Transplant Association at http://www.era-edta.org/ Finnish registry of Kidney disease at http://www.musili.fi/fin/munuaistautirekisteri/finnish_registry_for_ki dney_diseases/ Roberts D. S, Maxwell D. R, Gross L.T (1980) Cost-Effective Care of End-Stage Renal Disease: A Billion Dollar Question 92 (2 (Part 1)): p. 243-248. Mallik, N (1997) The cost of renal services in Britain. Nephrol Dial Transplant 12 (Suppl 1), p. 25-28 Hakkarainen P, Kapanen S, Honkanen E et al. (2005) Long slow night hemodialysis and quality of life. Hemodialysis international 9(1), pp. 97 Kaitelidou D, Ziroyanis P, Maniadakis N et al. (2005) Economic evaluation of hemodialysis: implications for technology assessment in Greece. International journal of technology assessment in health care, 21(1):40-6 Stavrianou K, Pallikarakis N (2007) Quality of life of end-stage renal disease patients and study on the implementation of nocturnal home hemodialysis in Greece. Hemodialysis International Volume 11 Issue 2, Pages 204 – 209 Rodina A, Bliznakova K (2009) Prevalence prognosis of the end stage renal disease patients in Greece. IFMBE Proceedings WC 2009 World Congress on Medical Physics and Biomedical Engineering. Vol. 25, 2009, Munich, Germany Ioannidis G, Papadaki O, Tsakiris D (2002) Statistical and epidemiological data of the renal replacement therapy in Greece. Report of the Hellenic Renal Registry, Hellenic Nephrology 14 pp. 525–548 Boletis J (2001) Renal transplantation in Greece Nephrology Dialysis Transplantation, Volume 16, Supplement 6, 25 September, pp. 137139(3) Apostolou, T (2007) Quality of life in the elderly patients on dialysis. Int Urol Nephrol 39(2): p. 679-83. Shaubel E.D, Morrison H.I, Fenton S.S (2000) Projecting renal replacement therapy-specific end-stage renal disease prevalence using registry data. Kidney Int 57(Suppl. 74): p. S-49-S54. Kontodimopoulos N, Niakas D (2006) A 12-year Analysis of Malmquist Total Factor Productivity in Dialysis Facilities. J Med Syst 30:333-342 Corresponding author: Author: Anastassia Rodina Institute: University of Patras Street: 26500 Rio City: Patras Country: Greece Email:
[email protected] IFMBE Proceedings Vol. 29
Influence of Bioimplant Surface Electrical Potential on Osteoblast Behavior and Bone Tissue Formation Yu. Dekhtyar1, I. Khlusov2, N. Polyaka1, R. Sammons3, and F. Tyulkin1 2
1 Riga Technical University, Riga, Latvia Tomsk Branch of Russian Ilizarov Scientific Centre «Restorative Traumatology and Orthopaedics»,, Tomsk, Russia 3 University of Birmingham, School of Dentistry,, Birmingham, United Kingdom
Abstract— Following the general adhesion theory the bone cells attachment to the bioimplant could be controlled by its surface electrical potential. The latter was adjusted both by deposition of the electrical charge and the surface microgeometry. It was demonstrated that a relief of the surface had an influence on the potential at the nanoscale and could be engineered to promote cell adhesion. Both deposition of the charge and engineering of the potential demonstrated influence on human osteoblast attachment and bone tissue formation. The surface areas of the electrical potential demonstrated a correlation length effecting fabrication of the bone tissue. Keywords— biomaterial, bioimplant, electrical potential, surface, bone cell, bone tissue.
I. INTRODUCTION In spite of the high success in understanding of human cells interaction with bioimplants positioned in the human organism there are still biocompatibly problems. These often are connected with incapability of the eligible human cells to be attached to the implant surface. Following the general adhesion theory [1] attachment of the cell to the bioimplant could be controlled by an electrostatic force contributing interaction between the cell and implant [1,2]. Generally the electrical communication could be engineered owing to a surface electrical potential of the implant. The potential could be supplied by the both external sources and the surface itself. The article is directed to demonstrate both possibilities.
II. METHODS AND MATERIALS To demonstrate a possibility to engineer the potential from the external source the hydroxyapatite (HAP) based implant model were selected. This was stipulated due to availability of the electrical charge deposition non contact novel 3D technology invented for HAP [3].
The technologies that are typically in use [4,5] supply the electrical charge to HAP specimens because of their polarization in an electrical field [4] or due to provision of charged particles radiation [5]. In both cases the opposite surfaces of the HAP based implant acquire opposite (in sign) charges. This provides an inadequate influence to osteoinduction of the bone tissue. The cells, induced by one side of the charged implant surface could be uninduced by the other contrary charged side of the implant. This restricts implementation of the above technologies for medical applications. HAP has the hydrogen connected to the oxygen: Ca5(PO4)3OH. Disposition of the hydrogen in respect to the oxygen has an influence on the HAP surface charge, HAP surface electrical charge strongly depends on a density of the protons [6] coupled to the oxygen. Shift of the hydrogen gives an opportunity to affect the charge. For instance, external highly pressured hydrogen gas could induce proton disposition at the HAP surface layer [7]. This could be provided because of the pressure induced diffusion/migration [8] of the protons. As the result, the HAP surface charge will be influenced. The HAP ceramic specimens that had a diameter 5 mm and thickness 1 mm were supplied from the EC project PERCERAMICS. The porosity of the specimens was equal to zero and their surface morphology had a irregularity ≤0.5 μm. The specimens were processed by a 6 MPa of the hydrogen gas during 1 hour. An alteration of the surface charge was estimated because of the photoelectron emission electron work function (ϕ) measurement. The value of ϕ characterizes an energy that is necessary to supply to the electron to escape it from the solid. From the energy conservation law the emitted electron is influenced by an electrical field of the surface charge, that contributes to ϕ. To measure ϕ photoelectron emission current (I) of the electrons from the specimen was induced by ultraviolet photons in a range 3-6 eV. The value of ϕ was identified
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 800–803, 2010. www.springerlink.com
Influence of Bioimplant Surface Electrical Potential on Osteoblast Behavior and Bone Tissue Formation
because of the dependence of I to the photon energy. Particularly, the energy of the photons, when I=0 was assumed as ϕ. Measurements were provided in vacuum conditions (10-1 Pa) applying a hand made spectrometer [9]. Attachment and proliferation of the osteoblast cells were explored by the following way. SAOS-2 human osteoblasts (ATCC Cat No. HTB-85™; LGC Standards, Teddington, UK) were cultured in McCoy’s 5a medium (Gibco, Paisley, UK).containing 10% foetal calf serum (Sigma, UK), 2.5% Hepes (Gibco, Paisley, UK) and 1% Penicillin/Streptomycin (Gibco, Paisley, UK) until confluent, harvested using trypsin-EDTA (Gibco, Paisley, UK) and resuspended in the same medium at a concentration of 1 x 105 cells/ml. The cells were allowed to recover from the enzyme for 1h at 37º C and 1 ml of suspension was then added to each specimen in separate wells of a 24-well plate at 37º C in an incubator containing 5% CO2, for 30min [10]. Experiments were terminated by removal of the cell suspension and washing three times in phosphate buffered saline to remove non-attached cells. Attached cells were then fixed for 1 h in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.3, dehydrated in alcohol and hexamethyldisilizane and further gold-sputter coated for scanning electron microscopy (SEM) using a Jeol lv3500 microscope at an accelerating voltage of 15kV, working distance 13mm. Images of five non-overlapping fields of view were captured using the SemAfore 4.0 programme (JEOL/Skandinaviska, Sweden) at a magnification of 150x (image area = approximately 800 x 600 µm). The individual cells were counted in each image and the average number of cells was determined. To test proliferation of the cells, the HAP samples were sterilised in 70% ethanol for 15 min and then dried in air for 10 min. Test samples in 24 well plates were covered with 1 ml of a suspension of SAOS-2 cells, prepared as above, at a density of 5 x 104 cells/ml. The cells were allowed to proliferate for 7 days. Culture medium was then removed and the samples were washed in PBS, fixed and prepared for SEM as above. As the cells were not confluent and individual cells could be identified, the cells were counted in 5 separate images at 150x magnification, as described above. Average cell number was estimated. In both microbiological tests the observed specimens were statistically selected on a base of the Smirnov criteria (significance level 0.05). After the microbiological experiment the implant models were tested with animals.
801
Different physical-chemical properties of calcium phosphate (CP) materials (crystallinity and porosity degree, solubility, surface roughness etc.) have diverse ability for osteogenesis promotion. A key combination of their structure, thickness and dissolution rate for realization of osteogenous potential of mesenchymal stromal cell pull (MSCP) is not found till now. Calcium phosphate (CP) coatings were provided on titanium discs by anodic spark technology using HAP particles (0.03-70 μm). Roughness of all coatings was characterized with the micro relief 5.13-24.16 μm. The prepared specimens passed the above hydrogenation procedure. Studies in situ were implemented by the method of ectopic bone formation when treated samples with bone marrow were implanted in mice for 45 days without additional injection of growth factors. To demonstrate a possibility to engineer the potential due to provision of the surface roughness the above specimens were coated with CP due to different technologies to reach dissimilar surface microirregularity: 0.7-14.4 µm. After that the biological experiment with mice on the above scheme was supplied, the probability on bone formation being estimated. The surface of the specimens was explored due to atomic force microscopy measurements: morphology and voltage (Kelvine probe) distribution were measured. For this 3 areas of each specimen were scanned with 3 scans at least. These data were in use to calculate average values. Alongside the values of ϕ were detected as above.
III. RESULTS AND DISCUSSION The hydrogenation of the HAP ceramic specimens increased their ϕ. This means that the electrical charge became more negative in contrast with the native specimens. The number of the attached cells correlated to the increment of ϕ (Fig. 1). The correlation coefficient was equal to 0.67 that was acceptable at the 0.05 level of significance. The number of the attached cells increased 10 times, when the increment of ϕ succeeded +0.18 eV. The result evidences that the osteoblast cells were more successfully attached to the surface, if electrical charge was shifted to the negative value. The shift of the charge to the negative value enhanced also proliferation of the cells. Proliferation increased in 1.6 time when the values of ϕ increased to +0.1 eV. The achieved results demonstrate that deposition on the surface of the negative charge enhances both attaching and proliferation capacity by the osteoblastic cells.
IFMBE Proceedings Vol. 29
802
Yu. Dekhtyar et al.
Fig. 1 Correlation of the number of the attached cells to ϕ Results of the experiments with the animals showed that due to hydrogenation of the CP surfaces their ϕ increased on ~0.1-1 eV. This had an effect on the directions of mesenchymal stromal cell pull (MSCP) differentiation. For example, connective tissue growth was improved. Probability of following ossification with growth of membrane reticulated bone improved on ≥ 20%. The surface microscopy measurements evidenced that the values of the surface voltage correlated to ϕ (Fig. 2). This taking into account the above results gave a hope to meet the correlations between the roughness of the surface, voltage and bone tissue development.
voltage. However the standard deviation of the surface voltage over the surface (SDV) correlated on SDSI (Fig. 4). On the other hand the voltage distribution should be characterized with the screening length (l). To estimate it, the auto correlation function K(V, x) was calculated for the voltage scans using the measured voltage data in dependence on the scan coordinate (x). The value of l was assumed at x, corresponding to the zero value of K(V, x). If voltage has the influence on bone fabrication, the correlation between l (that is actually the correlation length) and probability to produce tissue should be arisen. Fig. 5 demonstrates such the correlation. Because the correlation length for the surface geometrical irregularity did not demonstrate any correlations on tissue development one could assume the voltage has the major influence on bone tissue growth. The value of l related to the voltage correlates on SDSI (Fig.6). This means that l has a possibility to be engineered by SDSI that in turn could be reached owing to eligible technological processing.
Fig. 3 Probability to fabricate bone tissue in depend on standard deviation of the surface irregularity
IV. CONCLUSIONS 1.
2. Fig. 2 Correlation of the surface voltage to ϕ The best relation was provided between the probability to fabricate bone tissue and the standard deviation of the surface irregularity (SDSI) at the nanoscale (Fig.3). This result could arise a confuse on the searched influence of the
3.
Deposition on the surface of HAP of the negative charge enhances attaching and proliferation capacity by the osteoblastic cells as well as fabrication of bone tissue. The electrical potential of the surface at the nanoscale has an influence on bone formation. Influence of the potential is characterized with the correlation length. The correlation length of the electrical potential of the surface could be provided by its geometrical irregularity standard deviation at the nano scale.
IFMBE Proceedings Vol. 29
Influence of Bioimplant Surface Electrical Potential on Osteoblast Behavior and Bone Tissue Formation
803
ACKNOWLEDGMENT The authors are deeply indebted to the:
EC project PERCERAMICS NMP3-CT-2003504937 for the HAP specimens delivering; OlainFram company (Latvia) for the autoclave provision for hydrogenation.
REFERENCES
Fig. 4 SDV correlation to SDSI
Fig. 5 Correlation of probability to fabricate bone tissue to l of the voltage
1. Derjaguin B.V., Landau L.D. (1941) Theory of the stability of strongly charged lyophobic sols and of the adhesion of strongly charged particles in solution of electrolytes. Acta Physic Chimica, USSR, 14, 633-662. 2. Ratner B., Hoffman A., Schoen F., et al. (1996) Biomaterial science. London. 3. Dekhtyar Yu., Polyaka N., Sammons R (2008). Electrically charged hydrohyapatite enhances immobilization and proliferation of osteoblasts. . IFMBE proc. V20, 357-360. 4. Ito S., Shinomiya K., Nakamura S., Kobayyashi T., Nakamura M/. Yamashita K. (2006) Effect of electrical polarization of hydroxyapatite ceramics on new bone formation. Journal of the Japanese BioElectrical Research Society, 20: 23-27 5. Aronov D., Molotskii M., Rosenman G. (2007) Charge-induced wettability modification. Applied Phys. Let. 90: 104104-1-104104-1 6. Bystrov V., Bystrova N., Dekhtyar Yu/., Filippov S., Karlov A., Katashev A., Meissner C., Paramonova E., Patmalnieks A., Polyaka N., Sapronova A. (2006) Size depended electrical properties of hydroxyapatite nanoparticles IFMBE proc. 14.:3149-3150. 7. Dekhtyar Yu., Līvdane A., Poļaka N., Timofeeva A. The method to process insulator layer of the bio prostheses (2006). The patent of the Republic of Latvia LV13522. 8. Geguzin Ya. E. Diffusion zone.(1979). Moscow (In Russian). 9. Akmene R.J., Balodis A.J., Dekhtyar Yu.D., Markelova G.N., Matvejevs J.V., Rozenfelds L.B., Sagaloviąs G.L., Smirnovs J.S., Tolkaąovs A.A., Upmiņš A.I.(1993) Exoelectron emission specrometre complete set of surface local investigation. Poverhnost, Fizika, Himija, Mehanika (Russian Journal), 8:125.-128. 10. Lumbikanonda, N, & Sammons, R.L. (2001) Bone cell Attachment to Dental Implants of Different Surface characteristics. International Journal of Oral and Maxillofacial Implants. 16: 627-636. The address of the corresponding author: Author:
Institute: Street: City: Country: Email:
Fig. 6 Correlation l to SDSI
IFMBE Proceedings Vol. 29
Yu. Dekhtyar Riga Technical University, Riga, Latvia 1 Kalku Riga Latvia
[email protected] Nano-Sized Drug Carrier for Cancer Therapy: Dose-Toxicity Relationship of PEGPCL-PEG Polymeric Micelle on ICR Mice 2
J.L. Jiang 1, N.V. Cuong1, S.C. Jwo and M.F. Hsieh1 1
Department of Biomedical Engineering, Chung Yuan Christian University, 200, Chung Pei Rd., Chung Li, Taiwan 32023, ROC. 2 Division of General Surgery, Chang Gung Memorial Hospital, Keelung, Taiwan 20401, ROC.
Abstract—Amphiphilic polymeric nanoparticles garnered attention in pharmaceutical and medical fields because of its benefits of the sustained release, the extended circulation, the uptake by reticuloendothelial system in human body when used as the carrier of anti-cancer drugs. In the present study, we aimed to evaluate the biocompatibility of biodegradable triblock copolymer (PEG-PCL-PEG) used as the carrier of anti-cancer drugs. The synthesized PEG-PCL-PEG was synthesized and formed nano-sized micellar nanoparticles spontaneously in aqueous solution. The particle size was in the range of 40.2-85.3 nm. In-vitro hemolysis test indicated that 2.0 mg/mL micelle without drug loading caused no damage to red blood cells. In-vivo study on ICR mice displayed minor pathological response in liver and kidney of tested mice administrated intravenously 71.43 mg/kg nanoparticles. When the dose of nanoparticles increased to 91.95 mg/kg, both renal and liver toxicity was observed. This finding brings us further design of in-vivo study for loading anti-cancer drug, such as doxorubicin in the PEG-PCL-PEG carrier within a safe administration dose. Keywords— Amphiphilic polymer, Nanoparticles, Hemolysis, In-vivo Toxicity, Poly(εε-caprolactone), Polyethylene glycol. I. INTRODUCTION
Since the discovery of the enhanced permeation and retention effect in 1986 [1], a great variety of nano-sized dosage forms have been developing. Anti-cancer drugloaded nanoparticles as classified a kind of nano-sized dosage form, are functional materials that can undergo surface modification for targeting to cancer cells of cellular membrane, cytoplasmic or nuclei. Furthermore, the encapsulation of anti-cancer drugs in the nanoparticles can prevent the multi-drug resistance and prolong the circulation half life of drugs in the bloodstream. Amphiphilic copolymers, a kind of polymeric surfactant, can undergo self-assembly into spherical nano-sized micelles. There are two regions of a copolymeric micelle. The hydrophobic micelle core acts as a drug reservoir. The hydrophilic corona provides a protective interface between the core and the aqueous environment. The use of block copolymer micelles for the targeted delivery of drugs has proven to be a promising approach for improving the therapeutic efficacy of drugs, prolonging the circulation time and reducing the side-effect. The waterinsoluble drug, such as doxorubicin is encapsulated into the core of a polymeric micelle by chemical conjugation or physical entrapment.
The development of nano-sized dosage form for anti-cancer drug is divided into several stages including in-vitro and invivo experiments. Aside from the research and development of active pharmaceutical agents, the toxicity of drug carrier, also named excipient, is considered a key factor toward an efficacious dosage form. For example, the toxicity elucidated by excipients may lead to complex therapeutic outcomes along with the administration of drugs in the dosage form. Therefore, the in-vivo toxicity of the excipient is of critical and prerequisite for the development of nano-sized dosage form of anti-cancer drugs. Poly(ε-caprolactone) (PCL) is a biodegradable, biocompatible polyester. Moreover, PCL is highly hydrophobic polyester with degradation time could be tailored to fulfill targeted therapeutic applications [2-3]. Ring-opening polymerization of ε-caprolactone (CL) with monomethoxy polyethylene glycol (PEG) as an initiator and stannous octoate as catalyst, respectively, has been intensively investigated [4-5]. Herein we report the evaluation of the toxicity of micellar PEG-PCL-PEG nanoparticles in hemolysis test to which the damage of red blood cells by the nanopaticles was tested. In the experiment of the inflammatory reactions inducible by the nanoparticles, ICR mice were used and histological sections were taken from liver and kidney of the mice. II.
MATERIALS AND METHODS
A. Materials Monomethoxy poly(ethylene glycol) (PEG, Mn= 5000), ε-caprolactone, 2-diphenyl-1,3,5-hexatriene (DPH) and dimethylsulfoxide (DMSO) were purchased from SigmaAldrich (St.Louis, MO, USA). Stannous 2-ethyl hexanoate (stannous octoate, SnOct) was obtained from MP Biomedicals, Inc. The mPEG was purified by re-crystallization on the dichloromethane/diethyl ether system, and εcaprolactone was dried using CaH2 and distilled under a reduced pressure. Triethylamine (TEA), 1,4-dioxane, dichloromethane (DCM), diethyl ether, tetrahydrofuran (THF) and other chemicals were all reagent-grade and were used without further purification. B. Preparation of PEG-PCL-PEG nanoparticles PEG-PCL diblock copolymer, reported elsewhere [6] was synthesized by ring opening polymerization of CL in the presence of PEG as a macro initiator with SnOct as
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 804–807, 2010. www.springerlink.com
Nano-Sized Drug Carrier for Cancer Therapy: Dose-Toxicity Relationship of PEG-PCL-PEG Polymeric Micelle on ICR Mice
catalyst at 130oC for 24 h. The final polymer, PEG-PCLPEG was synthesized by coupling two PEG-PCL polymers using carbodiimide chemistry. The micelle solution of PEG-PCL-PEG was prepared by dissolving 20 mg of triblock PEG-PCL-PEG copolymer in 2.0 mL of THF, and then 1.0 mL of double distilled water was added under stirring. The resulting solution was placed at room temperature for 12 h, then was transferred to a dialysis bag and dialyzed against double distilled water for 24 h (MWCO: 50,000 Da, Spectrum Laboratories, USA). The particle size and its polydispersity were determined by dynamic light scattering (DLS) at 25 oC using a Zetasizer 3000HSA (Malvern Instruments Ltd, UK) with an excitation of 633 nm illuminated at a fixed angle of 90o. The micellar solutions were diluted into a final concentration of 0.2 mg/mL and then filtered through a Millipore 0.2 µm filter prior to measurements. The average diameter was calculated by a CONTTIN analytical method. C. The measurement of critical association concentration (CAC) of PEG-PCL-PEG micelles The critical association concentration of copolymers was determined by UV-VIS spectroscopy using DPH as the fluorescent probe. Samples for UV-VIS measurement were prepared according to the literature [7] and the concentration of aqueous copolymers solution ranged from 0.001 wt% to 1 wt% and concentration of DPH was 4×10-6 M. Note that CAC is also referred the Critical Micelle Concentration (CMC) in many research groups in this research topic. D. Hemolysis test of PEG-PCL-PEG nanoparticles The experimental procedure is based on colorimetric detection of Drabkin’s solution[8]. To a 0.7 mL of micellar solution at various concentrations (0.1 mg/mL to 2.0 mg/mL) was incubated in 0.1 mL of rabbit blood at 37 oC and for 3 h. Following incubation, the solution was centrifuged at 3800 rpm for 15 min. To determine the supernatant hemoglobin, 0.5 mL of Drabkin’s solution was added to 0.5 mL of supernatant and the sample was allowed to stand for 15 min. The amount of cyanmethemoglobin in the supernatant was measured by spectrophotometer at a wavelength 540 nm and, then, compared to a standard curve (hemoglobin concentrations ranging from 0.003 to 1.2 mg/mL). The percent hemolysis refers to the hemoglobin concentration in the supernatant of a blood sample not treated with micelles to the obtained percentage of particle-induced hemolysis. Finally, phosphate buffered solution (PBS) and double distilled water were used as negative and positive control, respectively.
805
E. In-vivo test: the toxicity of micellar PEG-PCL-PEG nanoparticles This test was conducted in accordance with Guidelines for Animal Care and Use Committee of Chung Yuan Christian University. Nine female ICR mice with average body weight of 27 g were feed in the animal facility where room temperature and humidity were maintained at 25 oC and RH 60 %. The mice were anesthetized with Rompum(R) solution and then the micellar solutions were injected intravenously to the mice at two doses of 71.43 mg/kg and 91.95 mg/kg. PBS was used as placebo. For multiple dosing tests, 71.43 mg/kg micellar solution was injected at 0, 3rd and 6th days of the test. The body weight of the mice was then monitored for a test period of 14 days post injection. To evaluate the toxicity to the liver and kidney where the most drugs are metabolized and excreted, the tissue sections with a thickness of 5 μm were taken after the mice were sacrificed. Tissue specimens were fixed in 10% formalin and embedded in paraffin. Afterward, these sections were stained with hematoxylin and eosin for microscopic examination. To quantitate the pathological tissues induced by the injection of the micellar solutions, the whole kidney and liver sections were divided into 950 and 1400 square grids observed under a upright microscope (Eclipse 50i, Nikon, Japan), then abnormal fibrotic tissue and loosen morphology were counted as adverse reaction due to the injection of micellar nanoparticles. III.
RESULTS AND DISCUSSION
A. Basic properties of PEG-PCL-PEG triblock copolymers and the micellar nanoparticles The molecular structure of the triblock copolymer used in this study is displayed in Fig. 1. These polymers were prepared with various feeding ratios of CL/PEG so as to adjust the block length of PCL. In the present work, the molecular weight (Mw) of PCL was designed to 3.6k, 8k and 22k Da, respectively. The properties of the synthesized PEG-PCLPEG are listed in Table 1. Since the molecular weight of PEG was fixed, the measured Mw reflected the length of PCL increased from 4.5k Da to 25k Da. With the increased hydrophobicity (PCL), it can be seen that the particle size increased with the length of PCL segment from 40.2 nm to 85.3 nm, while CAC of the micellar solutions decreased from 5.6 x 10-3 mg/mL to 5.1 x 10-4 mg/mL. From a view point of practical clinical application, a lowest CAC for micellar nanoparticles is desirable provided that the dilution effect may disintegrate the structure of the micelles leading to burst release of loaded drugs in the core of micelles when the micellar solution is injected into blood pool. Therefore, we chose a polymer (EC220E) having an Mw of 25k Da for the subsequent experiments.
IFMBE Proceedings Vol. 29
806
J.L. Jiang et al.
C. In-vivo toxicity test of micellar nanoparticles
Fig. 1 The molecular structure of PEG-PCL-PEG synthesized by coupling two PEG-PCL block using carbodiimide chemistry. Table 1 The molecular weights of PEG-PCL-PEG measured by gel permeation chromatography and the properties of nanoparticles Sample
Mw (mole/g)
Micellar size (nm)
CMC (mg/mL)
EC36E
14,500
40 ± 1
5.6×10-3
EC80E
19,000
56 ± 3
1.2×10-3
EC220E
35,000
85 ± 2
5.1×10-4
B. The hemolytic index of micellar nanoparticles Figure 2 shows the hemolytic index of EC220E micelle solutions having the concentrations from 0.1 mg/mL to 2.0 mg/mL. It’s evident that the concentrations were several orders of magnitude higher than CMC of EC220E. According to ASTM F-756 protocol (assessment of hemolytic properties of materials), an index of 2% or less is non-hemolytic to the testing material. In this study, the micellar solution with a concentration less than 0.1 mg/mL is non-hemolytic. However, the dosage form of this kind of polymer prepared in this concentration range may carry low drug payload. Hence, we compromised the payload and hemolytic index for the subsequent toxicity test in-vivo. Hemolytic index 100
Hemolytic index (%)
8
6
4
The body weights of ICR mice receiving single (71.43 mg/kg, 91.95 mg/kg) or multiple injections (3 x 71.43 mg/kg) of the nanoparticles in a period of 14 days increased and showed no difference to the control in which mice were injected with same volume of PBS. Meanwhile, no mouse died in the test period. So PEG-PCL-PEG nanoparticles prepared could be classified non-lethal to ICR mice receiving the tested doses. Figure 3 displays the histological sections of liver and kidney of ICR mice for single and multiple injections of micellar PEG-PCL-PEG nanoparticles. Red arrows indicate Kupffer cells and mesangial cells in the tissue sections and they stands for the activation of immune response to foreign body. We found that the single injection at a dose of 71.43 mg/kg result in no obvious immune reaction in liver and kidney, irrespective of single or multiple injections. However, when the dose increased to 91.95 mg/kg, we found intense immune cells and loosen structure of liver and kidney indicating a maximum dose was reached. In addition to qualitative observation, the quantitative analysis of the whole organ sections gave us a sign that a dose of 91.95 mg/kg is upper dose region for micellar PEGPCL-PEG nanoparticle. In a total grid of 1400 in whole liver section, 35.15% grids were marked as inflammation or fibrotic tissue. In conjunction with liver section, 18.21 % grids in whole kidney section were classified abnormal. From this in-vivo experiment, we succeeded in searching a safe dose of micellar PEG-PCL-PEG for intravenous injection. IV. CONCLUSIONS
The present study has developed a dosage form of micellar nanoparticles composed of self-assembled PEG-PCLPEG solution. This study tested the blank solution, e.g. no drug was loaded into the nanoparticles. A safe in-vitro toxicity of the dosage form to red blood cells was found to be 0.1-0.5 mg/mL. A higher and safe dose of 2 mg/mL, corresponding to 71.43 mg/kg body weight was reached for invivo toxicity in ICR mice. When the dose was elevated to 91.95 mg/kg a toxic response occurred. Therefore, we concluded that 71.43 mg/kg is a safe dose for intravenous injection of micellar PEG-PCL-PEG nanoparticles in ICR mice.
2
ACKNOWLEDGMENT
0 2
1
0.5
0.1
PBS
ddH2O
concentration (mg/mL)
Fig. 2 hemolystic indices of micellar PEG-PCL-PEG nanoparticles in the concentration range of 0.1 mg/mL to 2.0 mg/mL. The positive and negative controls were distilled water and phosphate buffered solution, respectively.
The authors would like to thank National Science Council of Republic of China for financial support under the grant number NSC98-2221-E-033-072. The gratitude is extended to Laboratory Animal Center of National Taiwan University Hospital for providing ICR mice.
IFMBE Proceedings Vol. 29
Nano-Sized Drug Carrier for Cancer Therapy: Dose-Toxicity Relationship of PEG-PCL-PEG Polymeric Micelle on ICR Mice
REFERENCES 1.
2. 3. 4. 5.
6.
7. 8.
Matsumura Y, Maeda H (1986) A new concept for macromolecular therapeutics in cancer-chemotherapy - mechanism of tumoritropic accumulation of proteins and the antitumor agent smancs. Cancer Res:6387-6392. Panyam J, Labhasetwar V (2003) Biodegradable nanoparticles for drug and gene delivery to cells and tissue. Adv Drug Deliv Rev: 329347. Sinha V R, Bansal K, Kaushik R et al. (2004) Poly-epsiloncaprolactone microspheres and nanospheres: an overview. Int J Pharm: 1-23. Hsieh M F, Lin T Y, Gau R J et al. (2005) Biodegradable polymeric nanoparticles bearing stealth PEG shell and lipophilic polyester core. Chin Inst Chem Engrs:609-615 Du Z X, Xu J T, Yang Y et al. (2007) Synthesis and characterization of poly(epsilon-caprolactone)-b-poly(ethylene glycol) block copolymers prepared by a salicylaldimine-aluminum complex. J Appl Polym Sci: 771-776. Hsieh M F, Cuong N V, Chen C H et al. (2008) Nano-sized micelles of block copolymers of methoxy poly(ethylene glycol)-poly(epsiloncaprolactone)-graft-2-hydroxyethyl cellulose for doxorubicin delivery. J Nanosci Nanotechnol:2362-2368. Hwang M J, Suh J M, Bae Y H et al. (2005) Caprolactonic poloxamer analog: PEG-PCL-PEG. Biomacromolecules:885-890. E. Beutler (1971), In: Grune and Stratton (Eds.), Red Cell Metabolism. A Manual of Biochemical Methods, New York, 1971. Author:
Ming-Fa Hsieh
Institute: Department of Biomedical Engineering Chung Yuan Christian University Street: No. 200, Chung Pei Rd. City: Chung Li Country: Taiwan, R.O.C. Email:
[email protected] Fig. 3 (continue)
Fig. 3 H&E staining of liver sections (a, c) of ICR mice receiving single injection of PBS and micellar solution (71.43 mg/kg), respectively. Kidney sections are displayed for (b) ICR mice with PBS injection and (d) ICR mice with 71.43 mg/kg injection of micellar solution. Scale bar = 100 μm.
IFMBE Proceedings Vol. 29
807
“Internet of Things”, an RFID – IPv6 Scenario in a Healthcare Environment H. Tsirbas, K. Giokas, and D. Koutsouris Biomedical Engineering Laboratory, School of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece Abstract— In recent years, increasing costs in Health Care have directed efforts for more effective resource management in hospitals. “Internet of Things” is a very important step towards an increased networking state of hospital inventory. The use of RFID and IPv6 allows for a low cost, low maintenance solution. The above combination also helps to develop a transparent information infrastructure that serves the health professional and the patient alike. Apart from the obvious advantages of IPv6 such as larger address space (128 bits long), better header format, provision for extension, resource allocation support (Flow Label) and security features and the RFID’s no line of sight requirement, long read range, real-time tracking, multiple tag read/write and database portability, the combination of the two and our proposed Virtual MAC Address Generator can provide a unique tracking, low-cost, low power tool in a multi-sensor hospital environment. Keywords— Internet of things, healthcare, RFID, IPv6.
I. INTRODUCTION In recent years, almost all countries of the world have been increasing their allocated resources for health care. The market in developed countries is based primarily on the middle-aged and older people group. This trend has lead to a greater demand for health care related to health services and greater competition among healthcare providers [1]. The effective management of a hospital is a key target for developed countries. In order to achieve this they need to understand the cost structure of hospitals and the inefficient use of resources. This will help to form policies on health care and budgetary decisions. The cost of hospitalization could be controlled more effectively resulting to health care being more accessible to all. Although the healthcare resources in the last years have increased, there should be a concentrated effort by information technology to reduce the cost of care in the western world. In order to achieve reduced costs, the use of intelligent mobile systems will play important role. Intelligent systems will provide the ability to the hospital staff to intelligently process personalized information that will come from patients. New diseases are appearing while the need for care increases, but the cost, quality and delivery of health care is not improving. The management system of health care is far behind compared to other industries.
In the last two decades the average life expectancy has raised leading to an increased concern since healthcare needs to be geared towards older people. The technology of mobile sensors can give us information in real time including vital signs and other physiological indicators of health and fitness of an individual. Such a system could find wide use in hospitals, in monitoring systems at home, office and health research studies [7]. The Internet is a technology that applies to health care because they it can provide telemedicine services to remote areas. The problem using the Internet in health services is mainly focused on how to integrate its capabilities and interconnectivity in a growing number of very small mobile sensors with lower costs and better energy savings. The application of these principles can be facilitated by the use of “The Internet of Things”. “The Internet of things” is a network of billions or trillions of machines communicating with one another. It is a major or dominant theme for the evolution of information and communications over the next few decades. The Internet of Things cannot be reduced to one specific technology or application. Radio Frequency Identification (RFID), Near Field Communication (NFC), ZigBee and Bluetooth are among the key technologies currently in the “sphere” of the Internet of things. Its simplest form is already here. There are 1.3 billion radiofrequency identification tags (RFIDs) and two billion mobile service users worldwide [2]. RFID allows each object to have its own unique identifier (rather than one identity number for each product type) which can be read at a distance. This allows automatic, real time identification and tracking of individual objects. Wireless sensor technologies allow objects to provide information about their environment and context. Smart technologies such as robotics and wearable computing will enable everyday objects to think and communicate. Nanotechnology and energy scavenging technologies are packing more processing power in less space, so that networked computing can be woven into the fabric of things around us. These technologies will progressively create an almost invisible infrastructure, with far-reaching capabilities organized into global systems that serve society. In this paper, “Internet of Things”, an RFID – IPv6 scenario in a Healthcare Environment, we propose an
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 808–811, 2010. www.springerlink.com
“Internet of Things”, an RFID – IPv6 Scenario in a Healthcare Environment
RFID - IPv6 networking configuration for use in a healthcare scenario. A network using a Virtual MAC Address Generator receives an RFID tag ID of various length and generates an IP address. Also, the information of RFID tag ID and IP address is mapped and is stored in the Virtual MAC Address Generator. By using this networking configuration in healthcare any Thing/Sensor, which is RFID tagged is able to be accessible from anywhere and from any services through an IP network.
II RELATED WORKS A. RFID In the ubiquitous network environment RFID is an automatic identification method which uses radio waves to automatically identify tagged items. It is similar to a bar code but uses a microchip and radio waves. An RFID tag is made up of an RFID chip attached to an antenna. Tags can be read from several meters away and beyond the line of sight of the reader. The most common method of identification is to store a serial number that identifies a person or an object, but an RFID tag can also store some additional information depending on the size of its memory. A tag is attached physically to the object to be identified. The tag is an electrical device designed to receive a specific signal and automatically transmit a specific reply. Tags can be passive, semi-passive or active, based on their power source and the way they are used, and can be read-only, read/write or read/write/re-write, depending on how their data is encoded. Passive RFID tags take the energy from the electro-magnetic field emitted by readers. Tags use the transmitting frequency in kilohertz, megahertz, and gigahertz ranges [6]. Table 1 Tags and Features Internal Power Source Signal by backscattering the carrier wave from the reader Response Size Cost Range Sensors
Tags and Features
Passive Tag
Active Tag
Semi Passive Tag
No
Yes
Yes
Yes
No
Yes
Weaker Small Less expensive 10 centimeters to few meters No
Stronger Big More expensive Hundreds of meters Yes
Stronger Medium Less Hundreds of meters Yes
An RFID reader or interrogator is a hardware device that is used to read the transmitted data from the tag. For reading passive RFID tags, the reader has to use extra power in
809
electromagnetic wave form. The reader hardware consists of a bipolar antenna and a microprocessor chip. The bipolar antenna is used to transmit the signals and power to the tag. It is also sensitive to read the reflected signals from the tag. The microprocessor controls and runs all the reader related processes. The reader can be a dedicated handheld device or embedded in a mobile device. An RFID tag is represented through an EPC (Electronic Product Code) that is currently managed by EPCglobal. EPC has various lengths 64 bit 96 bit and 256 bit in order to identify a product. The RFID Tag ID cannot perform the IP networking by itself because of being only an object-person identifier. B. IPv6 Internet Protocol version 6 (IPv6) is the next-generation Internet Protocol version designated as the successor to IPv4. The IPV6 increases the size of the IP address from 32 to 128 bits. IPv6 defines both a stateful and stateless address autoconfiguration mechanism. Stateless autoconfiguration requires no manual configuration of hosts, minimal (if any) configuration of routers, and no additional servers. The stateless mechanism allows a host to generate its own addresses using a combination of locally available information and information advertised by routers. Routers advertise prefixes that identify the subnet(s) associated with a link, while hosts generate an "interface identifier" that uniquely identifies an interface on a subnet. An address is formed by combining the two. In the absence of routers, a host can only generate link-local addresses. However, link-local addresses are sufficient for allowing communication among nodes attached to the same link [5]. In the stateful autoconfiguration model, hosts obtain interface addresses and/or configuration information and parameters from a server. Servers maintain a database that keeps track of which addresses have been assigned to which hosts. The stateful autoconfiguration protocol allows hosts to obtain addresses, other configuration information or both from a server. Stateless and stateful autoconfiguration complement each other. For example, a host can use stateless autoconfiguration to configure its own addresses, but uses stateful autoconfiguration to obtain other information [5]. C. Dhcpv6 The DHCP (Dynamic Host Configuration Protocol) was proposed for establishing static or dynamic addresses. DHCP provides the configuration parameter to the internet host. The DHCP protocol significantly reduces the system administration workload as the network devices can be added to the network with little or no change in the device configuration.
IFMBE Proceedings Vol. 29
810
H. Tsirbas, K. Giokas, and D. Koutsouris
DHCP also allows network parameter assignment at a single DHCP server or a group of such server located across the network. The dynamic host configuration is made possible with the automatic assignment of IP addresses, default gateway, subnet masks and other IP parameters. On connecting to a network, a DHCP configured node sends a broadcast query to the DHCP server requesting for necessary information. Upon receipt of a valid request, the DHCP server assigns an IP address from its pool of IP addresses and other TCP/IP configuration parameters such as the default gateway and subnet mask [4]. DHCP allocates IP addresses to the network devices in three different modes: dynamic mode, manual mode and static mode. In the static mode, the DHCP server statically binds the IP address and physical address and maintains the managed static database. In the case of the static allocation, the DHCP server assigns the persistent IP address to a client. In the dynamic mode, firstly the DHCP server confirms whether the physical address of a client is in the static database or not. If the requested physical address exists, the permanent IP address is assigned. If not, the IP address is assigned to a client for a limited time. The address assigned by the IP address pool is temporary. The DHCP server leases the IP address during the connection time. If the lease time is up, the client releases the IP address or has to be assigned a new one. In the manual mode, the client IP address is allocated by the network operator. DHCP is used in order to deliver the address assigned to a client.
Address Generator, the RFID Reader and the RFID Tag. The role of each one is explained below [3]. The role of the main network is to deliver the packets to each destination through routers. The Internet element represents all external users outside of main network. The DHCP server dynamically assigns the IP addresses. The RFID Reader reads the RFID Tag or writes data in the RFID Tag. Also the Virtual Mac Address Generator is a server where the RFID Tag ID is delivered and creates the virtual MAC address. The generator creates the virtual physical address of forty-eight bits. The RFID Tag, the virtual address and the assigned IP at the DHCP server are mapped and stored in the generator. B. Procedures Figure 2 depicts the IP generation sequence diagram which presents how an IP is generated from the RFID tag ID.
III. RFID-IPV6 NETWORK CONFIGURATION A. Network Topology The network topology shown in the above figure consists of the Main Network, the DHCP server, the Virtual Mac
Fig. 2 IP Address Generation - sequence diagram
Fig. 1 Network diagram
The RFID tag is read from a RFID reader. The RFID reader transmits the tag ID to the Virtual MAC Address Generator. The VMAG generates a virtual MAC address using the tag ID and transmits it to DHCP server. The DHCP server assigns the IP address to the virtual MAC address and sends it back to the generator. Finally the IFMBE Proceedings Vol. 29
“Internet of Things”, an RFID – IPv6 Scenario in a Healthcare Environment
Virtual MAC Address Generator binds the IP address, the virtual MAC address and the tag ID and stores them.
IV. RFID-IPV6 NETWORKING SCENARIO IN HEALTHCARE In the proposed network topology the generated IP address is mapped with the RFID tag ID and the virtual MAC address in the Virtual MAC Address Generator which has its own unique IP address. Figure 3 shows the RFID-IPV6 Networking scenario. This networking scenario is designed to meet the requirements of a basic healthcare network. Figure 3 is comprised of the main network and the Internet. The main network routers, DHCP Server, Virtual MAC Address Generator, RFID Reader and local end-users ‘Health Professionals’ are connected. All external end-users are connected through the Internet. In case a local Health Professional wants to read an RFID tag or write it, transmits a packet with the IP address of the RFID tag ID. At this time a packet is encapsulated with the IP address of the Virtual MAC Address Generator and is transmitted. A packet transmitted through the main network arrives at the generator. The generator decapsulates the received packet and searches for the mapped RFID Tag ID by using the destination IP address. If the RFID Tag ID exists in the storage (VMAG), a packet is delivered to the RFID Reader and the reader transmits the packet to the RFID Tag having this ID.
811
the packet to Virtual MAC Address Generator. The generator handles the practical processing for the packet transmission, so it transmits a packet to the IP address of the Health Professional using the destination and source address of the previous transmission. It encapsulates a packet to the IP address of the generator. The Health Professional, finally decapsulates the received packet. In case an external Health Professional wants to read an RFID tag or write it through the Internet, he/she should follow the same procedure as above. The only difference is that the packets are transmitted from the end user to the RFID tag through the Internet and the main network and vice versa.
V. CONCLUSION In our paper we are presenting a networking scenario in a healthcare environment using an overall “Internet of Things” solution. In reality the combination of RFID and IPv6 characteristics along with the proposed Virtual MAC Address Generator constitutes among other things, an internal address generation tool for the easier and more effective management of nodes (RFID Tags therefore tagged items). Such a solution has high scalability and lowers the cost of inventory maintenance in a hospital.
REFERENCES 1. F. A. Correa, M.J. Gil, and L.B. Redín (2005) Benefits of Connecting RFID and Lean Principles in Health Care. Business Economics Series 10, Madrid 2. Gérald Santucci (2009) From Internet of Data to Internet of Things. International Conference on Future Trends of the Internet, Luxembourg 3. Dong Geun Yoon, Dong Hyeon Lee, Chang Ho Seo, Seong Gon Choi (2008) RFID Networking Mechanism Using Address Management Agent. Fourth International Conference on Networked Computing and Advanced Information Management, Gyeongju 4. R. Droms, Ed (2001) Dynamic Host Configuration Protocol for IPv6 (DHCPv6). IETF RFC3646, USA 5. S. Thomson, T. Narten (1998) IPv6 Stateless Address Autoconfiguration. IETF RFC2462, USA 6. The EPCglobal Architecture Framework Version 1.2 at http://www.epcglobalinc.org 7. Opportunities for Use in Wearable Devices for Health Monitoring at http://www.sensorsmag.com/
Fig. 3 RFID-IPV6 Networking Scenario diagram Also if the Tag is active, it can transmit data to the Health Professional. In this case the RFID reader transmits
Author: Haris Tsirbas Institute: Biomedical Engineering Laboratory Street: Iroon Polytechniou 9 City: Athens Country: Greece Email:
[email protected] IFMBE Proceedings Vol. 29
Development of software tool for quantitative gait assessment in Parkinsonian patients with and without Mild Cognitive Impairment L. Iuppariello1, R. Tranfaglia1,3, M. Amboni2,3,4, L. Lista3 and M.Sansone1 1
University Federico II of Naples, Department of Biomedical, Electronic and Telecommunication Engineer , Naples, Italy 2 IDC Hermitage, Naples, Italy 3 University of Naples Parthenope, Naples, Italy 4 University of Naples Federico II, Department of Neurological Sciences
Abstract— The construct of Mild Cognitive Impairment (MCI) is considered as a transition state between normal aging and dementia. Originally it has been conceptualized as the transitional state between normalcy and Alzheimer’s disease (AD), more recently, it has been applied to patients with Parkinson’s disease (PD) as pre-dementia state. The relationship between cognition and gait disturbances has received increasing attention and is well established. In order to perform a quantitative assessment of gait in PD patients with or without MCI, an operating software tool has been developed. Ittakes advantage of the potentialities of the software Qualisys Track Manager and of the Matlab and allows the quantitative evaluation of the cinematic parameters of the human gait. Keywords— Motion Analysis, Gait Analysis, Mild Cognitive Impairment (MCI), Parkinson’s disease (PD)
I. INTRODUCTION
Analysis of human posture and movement is now a fast growing biomedical field of great interest from the clinical point of view, because the posture and movement are the result of the interaction of three major physiological systems: the nervous system, the muscle-skeletal system and the sensory system. Human movement, specifically, is the result of a complex process of signal processing carried out under the control of the central nervous system [1]. The objective of this work is the development of a multifactorial analysis software of neuromuscular and biomechanical parameters. The cinematic analysis of the movement offers a fundamental contribute for evaluation, Thanks to the preventive and therapeutic purposes. possibility to evidence associate motor strategies with the onset of pathologies of Central Nervous system and of Muscular Skeletal Apparatus, it allows to explore the residual motor abilities and to monitor with objective pointers the results of the participations put into effect, affording an appraisal of the relationships cost-benefits and orienting the choice of the various options through criteria of effectiveness and efficiency [2].
II. MATERIALS AND METHODS
A. The software tool The quantitative analysis of the movement is carried out in the laboratory of Neuro-Mechanic of the private hospital “Hermitage” in Naples. The opto-electronic system present in the laboratory, aimed to capture passive markers, consists of two main subsets: an acquisition structure (television cameras, illuminators, cards of acquisition) and an structure elaboration software. Specifically, the lab has installed a set of infrared cameras ProReflex MCU (Motion Capture Unit), made with CCD technology spread over an area of 4.5 x 6 m. The MCU uses a CCD sensor with low noise and high speed and a built-in microprocessor that allows them to achieve enhanced performance. The cameras support a infrared system that projects the special IR light that is reflected by markers and captured by optics of the cameras. Qualisys Track Manager is a Windows-based data acquisition software with an interface that allows the user to perform 2D and 3D motion capture. QTM is designed to provide both advanced features required by technically advanced users and a simple method of application for the inexperienced user. The subject of which we want to acquire the movement has put on himself some markers in strategic points , usually near the joints. To the aims of our observation, for the positioning of the markers we have been used to change the protocol of Davis [3] compared to his standard, by placing markers in the following points of anatomical landmark: Anterior Superior Iliac Spine, greater trochanter, thigh,lateral epicondyle of femur, lateral malleolus, calcaneus, V metatarsal head (Figure 1).
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 812–814, 2010. www.springerlink.com
Development of Software Tool for Quantitative Gait Assessment in Parkinsonian Patients with and without Mild Cognitive Impairment
813
Measuring the temporal and spatial distances between the minimums of the trajectories, it is possible to go back to the classics parameters of the human gait: Gait Cycle, Gait Stride, Stance Phase, Phase Swing, Support Double, Single Support, Step Length, Cadence, Velocity. B. The clinical study
Fig. 1 Markers set The QTM has been used in order to acquire and to reconstruct the trajectories of the markers. Successively the data have been exported in txt format for the elaboration in Matlab. A Matlab routine has been developed, that includes various algorithms for the calculation of the main cinematic gait parameters[4]. Txt file is a matrix containing the 3D coordinates of markers placed on the subject, sampled to 240 Hz. L '(?)M-files created in MATLAB provides to open read-only file. Txt. It extracted, from the complete matrix, the columns of the coordinates z of markers that are located on calcaneus and on metatarsal heads. The matlab routine calculates the minimums and the variations of trajectories function. It has been assumed that each minimum of the calcaneus and metatarsal heads represent the principal moment of the support to the land (fig. 2). The analysis of the variation of the function gives the stand and swing periods.
The developed tool has been used and tested in several neuro-mechanical studies carried out both on healthy and ill people. In particular , it has been of remarkable support to a clinical study titled: "Quantitative gait assessment in Parkinsonian patients with and without Mild Cognitive Impairment(MCI)”, which is currently underway and conducted in collaboration with a multidisciplinary team consisting of neurologists, bioengineers and experts in movement disorders. To our knowledge, this is the first study aimed to investigate the relationship between specific gait parameters anomalies in PD patients with MCI and the possible subsequent development of dementia. The target of this work is, at first, the evaluation of gait in PD patients with and without MCI in order to identify any differences in the gait parameters; successively, after two years, the revaluation of the patients will be carried out, with the objective to find a possible correlation between early specific anomalies of the gait and the development of dementia. To date, fourteen PD patients were investigated. All patients were neither demented nor depressed. The clinical assessment included clinical data collection, H&Y stage, UPDRS I, II and IV at on state, UPDRS III both at on and off state and an extensive neuropsychological assessment in order to classify the patients according to MCI criteria [5]. Quantitative gait analysis was performed by means of motion capture system (Qualysis, Sweden) and was applied during the following conditions both at off and on state: 1) normal gait (Gait-off and Gait-on); 2) motor dual task (carrying a tray with two glasses filled with water,while walking, Mot-off and Mot-on); 3) cognitive dual task (serially subtracting 7’s, starting from 100, while walking, Cog-off and Cog-on). Statistical analysis was carried out by means of MannWhitney test. III. RESULTS
Fig. 2
Representative graphs of the z coordinates of markers of metatarsal head and calcaneus; the minimums of the function are circled in red
Based on neuropsychological testing, seven subjects were classified as patients with MCI (MCI+) and seven subjects were classified as patients with normal cognition (MCI-). There were no significant differences in age, disease duration, H&Y stage, UPDRS and MMSE scores
IFMBE Proceedings Vol. 29
814
L. Iuppariello et al.
between the two groups. The following gait parameters significantly differed between the two groups: 1) double support time was longer in MCI+ vs. MCI- in Gait-off (p=0,007), Gait-on (p=0,026), Cog-on (p=0,026); 2) velocity was reduced in MCI+ vs. MCI- in Mot-on (p=0,038) and in Cog-off (p=0,011); 3) cadence was reduced in MCI+ vs. MCI- in Cog-on (p=0,038). In comparison with MCI- PD patients MCI+ display specific gait features: slower velocity due to the reduction of cadence and impairment of dynamic stability as revealed by the longer double support time.
Moreover the authors thanks the medical staff of Hermitage, that has supported the bio-engineering study.Format the Acknowledgment and References headlines without numbering.
REFERENCES 1.
2.
3.
IV. CONCLUSIONS
In this work we wanted to highlight the growing importance of the multifactorial integrated analysis of the movement in the biomedical field. As evidence of this, many facilities and hospitals are going to be equipped with laboratories designed for this type of analysis. Quantitative assessement of motion systems origin from the need to break free from the limitations of qualitative assessment that, based on sensory perceptions, results imprecise and depending on the observer. The tool developed in MATLAB has enabled us to reach an objective understanding of the motor pattern of each patient examined. Measurable and repeatable results have been obtained in this way, demonstrating that the software tool is able to perform efficiently the mission for which it was designed.
ACKNOWLEDGMENT
4. 5.
6.
7.
Cappello A, Cappozzo A ,Pietro Enrico di Prampero. XXII scuola annuale di Bioingegneria della postura e del movimento; Bressanone 2003: 63-65. Sheldon R. Quantification of human motion: gait analysis—benefits and limitations to its application to clinical problems. Journal of Biomechanics 2004; 37:1869–1880. Roy B. Davis III, Dunckley, Sylvia Ounpuu, Dennis Tyburski , James R. Gage. A gait analysis data collection and reduction technique. Human Movement Science 10(1991) 575-587. Perry J. Gait analysis: Normal and pathological functions. Hardcover 1992. Caviness J, Dunckley, Connor D, Sabbagh M, Hentz J, Noble B, .Evidente B, Shill H, Adler C. Defining Mild Cognitive Impairment in Parkinson’s Disease. Movement Disorders 2007; 22 (9) : 1272– 1277 Ronald C. Petersen, PhD, MD; Glenn E. Smith, PhD; Stephen C. Waring, DVM, Phd; Robert J. Ivnik, PhD; Eric G. Tangalos, MD; Emre Kokmen, MD . Mild Cognitive Impairment, Clinical Characterization and Outcome . Arch Neurol. 1999; 56: 303-308 Ronald C. Petersen, Ph.D., M.D. . Mild Cognitive Impairment, Current Research and Clinical Implications . Semin Neurol. 2007; 27: 22-31.
Author: Luigi Iuppariello Institute: Department of Biomedical, Electronic and Telecommunication Engineer, University of Naples Federico II Street: Via Claudio 21 City: Naples Country: Italy Email:
[email protected] The authors thank the private hospital “Hermitage” in Naples, that has shared the laboratory and the necessary instrumentation for the study.
IFMBE Proceedings Vol. 29
Tensile Stress Analysis of the Ceramic Head Endoprosthesis with different Micro Shape Deviations of the Contact Areas V. Fuis1 and M. Koukal2 1
Centre of Mechatronics – Institute of Thermomechanics AS CR and Institute of Solid Mechanics, Mechatronics and Biomechanics, Faculty of Mechanical Engineering, Brno University of Technology, Brno, Czech Republic 2 Institute of Solid Mechanics, Mechatronics and Biomechanics, Faculty of Mechanical Engineering, Brno University of Technology, Brno, Czech Republic Abstract— The paper deals with the problems of ceramic head of hip joint endoprosthesis destructions in vivo, and with assessing the impact of shape deflections of conical surfaces on the tensile stress in the head. Concerned are shape deviations from the ideal conical surfaces of the stem and the head of the endoprosthesis. The shape deviations may be modelled at the macro-level - this concerns model shape inaccuracies such as deviation from the nominal degree of taper, at the micro-level when the stochastic distribution of unevenness on the contact areas is respected. The problem of stress in ceramic heads was solved using the finite element method – system ANSYS under ISO 7206-5 loading. There are presented and analysed the results of solution of the macro shape deviations and micro shape deviations, obtained from measurements made on the cones of stems and heads. There are analysed two different groups of micro shape deviations. The first is three variants of the sizes of the measured micro shape deviations (measured, doubled and halved). The second group of the analysed shapes deviations contained modeled deviations with the linear transformation of the measured deviations along the cone depth.
II. METHODS
The computational modelling of stress has been made on a system consisting of a testing steel stem and a ceramic head. The head has been put on the cone of the testing stem and the load of this system has been in compliance with ISO 7206-5 – Fig. 1, which is the standard for determining static strength of ceramic heads for hip joint endoprostheses. With view to geometrical inputs, the deviations from ideal shapes can be divided into two groups - global (macro) deviations, i.e. the deviations from the stem's or head's nominal coneshape - angle D in Fig. 1. Maximum allowed difference of the head's and stem's cones is D = 10’ and this value is assumed in the computational modelling. D = 32 mm y
Keywords— Hip joint endoprosthesis, ceramic head, shape deviations of contact areas, tensile stress
F
I. INTRODUCTION
HEAD
H = 19 mm
0 coordinate of the cone depth [mm]
The failure of cohesion of ceramic heads of total hip joint prostheses has been stated in a not negligible number of patients in the Czech Republic. The implant’s failure of the ceramic head has always traumatic consequences for the patient, since a part of or even the whole endoprosthesis has to be re-operated. Hence, it is desired to reduce the number of implant re-operations to the minimum. Therefore the computational modelling (using FEM) of the stress and the failure probability (based on Weibull weakest link theory [1]) of the ceramics head was realised. In this case the influence of the micro shape deviations of the stem and the head contact cone were analysed in detail. The values of the deviations and its lay out along the cone depth are important for the tensile stress distribution in the ceramic head.
STEM Ds Dh
D/2
Fig. 1 Analysed system under ISO 7206-5 loading, macro shape deviation P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 815–818, 2010. www.springerlink.com
816
V. Fuis and M. Koukal
The second group is local (micro) deviations which were measured using IMS UMPIRE device. These deviations are represented by micro shape deviations from the ideal cone are shown in Fig. 2. The deviations from the ideal cone shape are represented in the developed section of the cone in Fig. 2a – measured head cone micro deviations (from -3.9 Pm to +4.12 Pm), Fig. 2b – stem cone micro deviations (from -2.27 Pm to +2.28 Pm).
deviations on the largest cone diameter. These modelled variants (VAR. 0-1 and VAR. 1-0) are used for stem (Fig. 3) and head (Fig. 4) cone micro deviations.
Fig. 3 Modelled variants of the stem cone shape deviations
Fig. 2 Measured micro shape deviation of the stem and head cones The computational modeling was realised for three values of the micro shape deviations – the measured head and stem deviations – VAR. 1x, the double of the measured deviations – VAR. 2x and the half of the measured deviations – VAR. 0.5x. The micro shape deviations are shown in Fig. 2 is the same for all three variants, only the scale is changed (doubled or halved). The rest assumed variants of the micro shape deviations are the linear transformation of the measured deviation along the cone depth – 1-0 or 0-1. VAR. 1-0 means the real deviations on the smallest cone diameter and zero deviations on the highest cone diameter. VAR. 0-1 means the zero deviations on the smallest cone diameter and real zero
Fig. 4 Modelled variants of the head cone shape deviations
IFMBE Proceedings Vol. 29
Tensile Stress Analysis of the Ceramic Head Endoprosthesis with different Micro Shape Deviations of the Contact Areas
The modeled shape deviations (VAR. S/H 1-0 or 0-1) is connected to the measured counterpart – for example – VAR. S 1-0 – stem shape deviations are linearly transformed (the smallest diameter has got measured deviations and the highest cone diameter has got zero deviation), is connected to the measured head deviations VAR. 1x (Fig. 2a). The state of stress of the ceramic head of endoprosthesis can be strongly affected by the process of endoprosthesis implantation, when the surgeon fits the head on the cone of the stem. The fitting of the head on the cone of the stem is a random process, in view of the mutual position of the head to the stem (in sense of the head's slight turning around the axis y, defined by angle E - Fig. 5). Therefore a series of computations will be made for each pair and for various positions of the head towards the stem.
817
The distribution of maximum values of tensile stresses in the ceramic head (Vmax), in dependence on the value of the head's loading, is shown in Figs. 6, 7 and 8. The values Vmax for various positions of the head to the stem (various value of angle E) form in Figs. 6 – 8 a belt of curves. The wide of the curve’s belt responds the dispersion of the Vmaxin the ceramic head.
Fig. 6 Maximum tensile stress in the head for different values of the micro shape deviations – VAR. 2x – doubled deviations, VAR. 0.5x – halved deviations
Fig. 5 Representation of various head to stem positions defined by angle E The head is modelled as a linear isotropic continuum with the following constants – Eh = 390 GPa and Ph = 0.23. The stem is modelled as a linear isotropic continuum (elastic) too, material parameters are the following Es = 210 GPa and Ph = 0.3. The coefficient of friction between the head and the stem is f = 0.15 [2]. FEM ANSYS system has been used for modelling the stress in the system. From the viewpoint of the bonds of the system, it may be stated that besides the head - stem contact also the shift of the stem's lower part has been prevented. To ensure convergence, the head's rotation on the stem has also to be prevented. The load of the system corresponds to ISO 7206 - 5 - the force acts as linear load on a on a circle with a radius of 10 mm – Figs. 1 and 5. III. RESULTS
In a ceramic head a 3D state of stress sets in under loading. In view of the reliability of the head, which is made of brittle material, the most important are the tensile stresses [3]. For this reason only extreme tensile stresses in the ceramic head will further be analysed.
The Fig. 6 shows the influence of the sizes of the head’s and stem’s micro shape deviations of the contact cone surface. There are assumed three variants – VAR. 1x – measured micro shape deviations, VAR. 2x – doubled sizes of the measured values and VAR. 0.5x - halved sizes of the measured values. The dispersions of the Vmax curves of the VAR. 1x and VAR. 2x are nearly same, only the Vmax values of the VAR. 2x are higher. Micro shape deviations for VAR. 0.5x caused the reduction of the Vmax values and dispersion too. The Fig. 7 shows the maximum tensile stress in the head for different lay out of the micro shape deviations – deviations on the smallest cone diameter are measured, on the largest diameter are zero. For comparison there is the measured variant shown too (VAR. 1x). The tensile stress curves for all three displayed variants created very similar belt, so the shape of the micro deviations on the smallest diameter determinate the value of the maximum tensile stresses in the head. The rest regions are not so important (see bellow). The Fig. 8 shows again the maximum tensile stress in the head for different lay out of the micro shape deviations – deviations on the highest cone diameter are measured, on the smallest diameter are zero. Measured variant (VAR. 1x) is shown in Fig. 8 too due to the comparison of the curves belts. The curves belts in Fig. 8 are different from curves in
IFMBE Proceedings Vol. 29
818
V. Fuis and M. Koukal
Fig. 7 – variants with modified micro shape deviations (VAR. 0-1) have got wider curve’s belt. In the other hand the narrowest belt is for VAR. S 0-1 (yellow color in Fig. 8) and highest tensile head’s stress reduction is for VAR. H 01 (light blue color in Fig. 8). The wide of the VAR. H 0-1 is smaller than for VAR. 1x. The computational modeling results show that the most important location of the cone in front of view is the smallest cone diameter. The reduction of the micro shape deviations on this location causes the reduction of the tensile stress in the head.
IV. CONCLUSIONS
By computational modelling (using FEM) it has been proved that the location of the micro shape deviations on the stem’s and head’s cone can reduce the maximum tensile stresses in the ceramic head (VAR. H 0-1 or VAR. S 0-1). The decreasing of the micro shape deviations on the smallest diameter of the stem and the head cone caused the tensile stress reduction. The values of the sizes of the micro shape deviations which are added to the macro shape deviation, significantly influence the tensile stress in the head and the value of the dispersion the of Vmax curves (different position of the head on the stem come). The doubled micro deviations (VAR. 2x) caused increasing tensile stress state and the dispersion of the stress curves too. The halved micro deviation (VAR. 0.5x) caused the stress reduction. The tensile stress in the ceramic head cause the brittle fracture and therefore the reduction of the tensile stress (location of the micro deviations (VAR. 0-1) or halved the micro shape deviations (VAR. 0.5x)) causes the decreasing of the head’s failure probability which is based on the Weibull weakest link theory [1].
ACKNOWLEDGMENT Fig. 7 Maximum tensile stress in the head for different lay out of the micro shape deviations – VAR. 1x – measured deviations, VAR. S or H 1-0 (S – stem, H- head) - deviations on the smallest cone diameter are measured, on the largest diameter are zero – see Figs. 3 a 4
This paper was prepared as a part of the projects AV0Z20760514 and CZ.1.07/2.3.00/09.0228 „Complex System for Attracting, Education and Continuing Involment of Talented Individuals to Research Centers of AS CR and FME BUT“, which is cofinanced by the European Social Fund and the state budget of the Czech Republic.
REFERENCES 1. 2. 3.
Fig. 8 Maximum tensile stress in the head for different lay out of the micro shape deviations – VAR. 1x – measured deviations, VAR. S or H 0-1 (S – stem, H- head) - deviations on the smallest cone diameter are zero, on the largest diameter are measured (real) – see Figs. 3 a 4
McLean A F, Hartsock D L (1991) Engineered materials handbook, Vol. 4, Ceramics and Glasses. ASM International 1991, 676-689 Fuis V, Janíþek P (2002) Stress and reliability analyses of damaged ceramic femoral heads. Damage & Fracture Mechanics VII. WIT Press, 475-486 Fuis V, Návrat T, HlavoĖ P, Koukal M, Houfek M (2007) Analysis of contact pressure between the parts of total hip joint endoprosthesis with shape deviations. Journal of Biomechanics, Vol. 40, Suppl. 2 Author: Vladimir Fuis Institute: Centre of Mechatronics – Institute of Thermomechanics AS CR and Institute of Solid Mechanics, Mechatronics and Biomechanics, Faculty of Mechanical Engineering, Brno University of Technology Street: Technicka 2, 616 69 City: Brno Country: Czech Republic Email:
[email protected] IFMBE Proceedings Vol. 29
Modelling of Cancer Dynamics and Comparison of Methods for Survival Time Estimation Tomas Zdrazil1 and Jiri Holcik1,2
2
1 Institute of Biostatistics and Analyses, Masaryk University, Brno, Czech Republic Institute of Measurement Sciences, Slovak Academy of Sciences, Bratislava, Slovakia
Abstract— The paper deals with an epidemiological compartmental model to describe cancer dynamics in a given population. Three methods including Kaplan-Meier algorithm and Weibull and exponential distribution fitting were used for estimation of the mean survival time for four TNM stages of prostate cancer described by the 1977-2007 data from the National Oncological Register of the Czech Republic. In comparison to Kaplan-Meier estimation both the distribution fitting methods resulted in overestimated length of probable survival. However, such overestimation can be corrected using demography mortality data. Keywords— Prostate cancer, survival analysis, KaplanMeier estimator, right-censored data.
I. INTRODUCTION There are two basic ways to analyse oncological data. The first deals with modelling a growth or behaviour of a single tumour at the cellular level, while the second describes a global cancer dynamics in a population. One approach to model a cancer dynamics in a population assumes that the disease progresses through four stages according to TNM classification [1]. The number of people in the stages can be described by state-space variables of compartmental model. To determine transition rates between the compartments we need to estimate a mean survival time for particular stages, which is mostly complicated by the fact that the real data are right-censored. There are several approaches to solve this problem. While Kaplan-Meier algorithm [2] uses a nonparametric maximum likelihood estimate of the survival function, Cox regression [3] evaluates the effect of several variables upon the time a specified event takes to happen. Another method is based on linearization of given distribution and estimation of missing fractiles by linear regression [4].
II. MATERIALS AND METHODS We assumed the disease progresses through four stages [1], so we chose a model of an epidemiological type, described by the system of following differential equations
x1' = c0 n − c1 x1 x2' = c1 x1 − c2 x2 x3' = c2 x2 − c3 x3
(1)
x4' = c3 x3 − c4 x4 where n is a number of healthy men in the population, which is assumed to be constant, x1, x2, x3, x4 are numbers of patients in the four disease stages 1, 2, 3 and/or 4 resp. according to the TNM classification, and c1, c2, c3, c4 are parameters referring to transition rates between particular compartments. The disease behaviour and health care services effectiveness changes with time. That is why we created groups of patients based on the year of diagnosis and calculated parameters for each of them. The data source, the National Oncological Register (NOR) of the Czech Republic contains a huge amount of data about patients diagnosed with cancer in years 19772007. For each patient NOR contains date of birth, date of diagnosis, stage of the disease at the time of diagnosis and eventually date of death. Based on these data we first computed c0, which represents a proportion of healthy men diagnosed per year, as a product of a reciprocal value of mean age at the time of diagnosis and probability of disease onset during life. In order to compute c1, c2, c3, c4 we had to estimate mean survival time. For the most of cases some patients were still alive at the end of observation time (end of year 2007), thus we obtained just only a part of a distribution function of a survival time. This is usually called right-censored data problem. Percentages of missing data are introduced in Table 1. We used Kaplan-Meier algorithm, that is a widely used approach to correct the estimation of the incomplete data series, and two developed methods based on parameter estimates of given distribution function by nonlinear regression method.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 819–822, 2010. www.springerlink.com
820
T. Zdrazil and J. Holcik
Table 1 Censored data percentages year 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
stage 1 0 0 0 20 0 11 0 0 9 17 0 17 0 33 21 35 12 35 54 54 53 59 66 71 84 81 86 88
stage 2 0 0 0 0 0 29 0 0 14 14 0 0 25 18 0 18 19 21 23 43 49 54 57 68 75 81 86 93 94 97 99
stage 3 10 0 100 0 0 0 0 0 0 0 10 0 0 33 9 0 27 17 21 24 32 43 50 48 58 65 66 84 89 95 98
a) Weibull Distribution Two parametric distribution function of Weibull distribution is given as
stage 4 0 0 0 0 0 0 0 0 0 0 14 0 8 0 13 5 2 6 5 9 12 14 15 16 19 24 30 40 51 63 85
b
F ( x) = 1 − e
⎛x⎞ −⎜ ⎟ ⎝a⎠
(3)
Choice of this distribution was based on a typical character of majority of experimental data. Great part of histograms, especially for stages 1 and 2, seemed to fit the density of the Weibull distribution for the value of parameter b around 1.5, such as histogram for stage 2 in 1996. The other part showed a shape of exponential distribution, that is a special case of the Weibull distribution for b=1.
A. Kaplan-Meier Algorithm Kaplan-Meier algorithm is based on nonparametric maximum likelihood estimate of a survival function. Probability S(t) of having life time greater than t is estimated by statistic
n − di Sˆ (t ) = ∏ i ni ti < t
Fig. 1 Weibull distribution-shaped histogram In case of the two-parametric Weibull distribution we calculated its parameters using constrained optimization based on a condition
∫
(2)
where ni is a number of persons at risk and di stands for a number of deaths at the time ti . B. Distribution Fitting This method is based on estimation of parameters of a chosen distribution function. According to standard approaches to survival data modelling and in relation to the shape of survival time histograms we chose the Weibull and exponential distribution and applied nonlinear regression to estimate their parameters.
xα
0
F ' ( x) = α
(4)
That means that the value of F(x) in maximum survival time xα is equal to the number of proportion of patients that died before the end of observation. In other words, the distribution function goes through its α-fractile. This condition results in an equation describing a relationship between parameters a and b b=
ln[− ln(1 − α )] . x ln( α )
(5)
α
After parameter estimation the mean of the Weibull distribution was calculated as
IFMBE Proceedings Vol. 29
Modelling of Cancer Dynamics and Comparison of Methods for Survival Time Estimation
stage 3
stage 4
2007
2005
2003
2001
1999
1997
1995
1993
1991
1989
1987
1977
Most data, especially for stages 3 and 4 seemed to fit this distribution.
stage 2
1985
(7)
stage 1 18000 16000 14000 12000 10000 8000 6000 4000 2000 0 1983
F ( x) = 1 − e
⎛ x⎞ −⎜ ⎟ ⎝a⎠
Prevalence of prostate cancer in stages 1,2,3,4
1981
b) Exponential Distribution Because the Weibull distribution did not always describe experimental data well, we also tested the exponential distribution. The distribution function of exponential distribution is given as
The data used for the analysis came from the National Oncological Register. We applied the three described methods on C61 diagnosis, the prostate cancer.
1979
(6)
number of patients
⎛1 ⎞ E ( X ) = aΓ ⎜ + 1 ⎟ , ⎝b ⎠ where Γ represents Γ function.
821
year
Fig. 3 Prevalence of prostate cancer for stages 1, 2, 3, 4
III. RESULTS As a result of application of the three described methods we obtained time series of mean survival time for stages 1, 2, 3 and/or 4 in years 1977-2007. Because of the variability of the results we used moving average to smooth the data trends. All analyses were calculated by SPSS software. Kaplan-Meier mean survival time s tage 1
stage 2
stage 3
s tage 4
sur vival tim e (days)
5000,000
Fig. 2 Exponential distribution-shaped histogram
4000,000 3000,000 2000,000 1000,000
2005
2007
2007
2003
2001
1999
1997
1995
1993
1991
1989
1987
1985
Estimated mean of Weibull distribution stage 1
stage 2
stage 3
stage 4
(9)
survival time (days)
25000 20000 15000 10000 5000
year
1 c1 = t1 − t2
Fig. 5 Weibull distribution mean estimates IFMBE Proceedings Vol. 29
2003
2001
1999
1997
1995
1993
1991
1989
1987
1985
1983
1979
0 1977
1 c2 = t2 − t3
1983
Fig. 4 Kaplan-Meier mean estimates
30000
1 t4
1 . c3 = t3 − t 4
2005
(8)
Now let us assign survival times for stages 1, 2, 3, 4 as t1, t2, t3, and t4. Then parameters c1, c2, c3, and c4 of model (1) were calculated as c4 =
1981
1977
year
1981
E( X ) = a
1979
,000
The mean of the exponential distribution is equal to the value of parameter a.
822
T. Zdrazil and J. Holcik
overestimation of the values by the distribution fitting methods is caused by the fact that neither Weibull nor exponential distribution limits the survival time, which can theoretically go to infinity. We believe that this problem can be fixed by adding proper weights decreasing with time to both the Weibull and exponential density functions, derived from the Czech mortality tables.
Estimated mean of Exponential distribution stage 2
stage 3
stage 4
V. CONCLUSION
2007
2005
2003
2001
1999
1997
1995
1993
1991
1989
1987
1985
1983
1981
1979
1977
mean survival time (days)
stage 1 20000,000 18000,000 16000,000 14000,000 12000,000 10000,000 8000,000 6000,000 4000,000 2000,000 ,000
year
Fig. 6 Exponential distribution mean estimates
IV. DISCUSSION In the Czech Republic, the average age of men at the time of prostate cancer diagnosis is about 70 years. According to the mortality tables, the mean survival time for 70 years old man is 12 years. That is why we consider means of survival derived from the distribution fitting overestimated. On the other hand, as stated in [5], the Kaplan-Meier methodology underestimates the true mean for rightcensored data, which gets worse as the censoring increases. Fig.4 confirms this fact. As a result of these inaccuracies, the number of patients predicted by model (1) does not match the real numbers for any of the three used methods. We found that the data in the NOR shows a slowly increasing linear trend of survival time for all stages, which corresponds to the assumption that health care services are becoming more successful in time. We suppose that the
On the basis of compartmental model of cancer dynamics we applied three methods for the mean survival time estimation. While the generally accepted Kaplan-Meier algorithm underestimates mean survival time for right-censored data, method of parameter estimation for the Weibull and exponential distribution done by nonlinear regression clearly overestimates it. Our suggestion to improve the distribution fitting methods was to add proper weights to the density functions of given distributions, possibly calculated from mortality tables.
REFERENCES 1. 2. 3. 4. 5.
http://en.wikipedia.org/wiki/Prostate_cancer_staging Kaplan E. L., Meier P, (1958), Nonparametric estimation from incomplete observations, Journal of the American Statistical Association, Vol. 53, 457-81 Cox, D.R.., (1972), Regression model and life tables, Journal of the Royal Statistical Society, Series B 34, 187-220 Kupka, K., (2002), Reliability, Durability, Failure rate and their modeling, Automa Zhong M., Kenneth R. H., (2009), Mean Survival Time from The Right Censored Data, Cobra preprint series
IFMBE Proceedings Vol. 29
Implications of Data Quality Problems within Hospital Administrative Databases J.A. Freitas1,2, T. Silva-Costa1,2, B. Marques1,2, and A. Costa-Pereira1,2 1
2
Department of Biostatistics and Medical Informatics, Faculdade de Medicina, Universidade do Porto, Portugal CINTESIS - Center for Research in Health Technologies and Information Systems, Universidade do Porto, Portugal
Abstract— Administrative data, even with some coding irregularities, is easily accessible, inexpensive, and includes episodes from several years for large populations. This routine data can be used to study hospital performance and to screen potential problems in healthcare. In this context, the quality of data is a fundamental issue if we want to achieve reliable results. In fact, the problem of poor data quality cannot be ignored and should more and more be considered. In this paper we present some issues about quality problems, expose some examples of poor data in administrative databases and discuss some potential implications of these problems. Keywords — data quality problems, hospital administrative data, diagnosis-related groups.
I. INTRODUCTION Data Quality has become increasingly important to many organizations as they build data warehouses and focuses more on customer relationship. In particular, in the health care field, cost pressures and the need to improve patient care impel efforts to integrate and clean organizational data. Over the past years, the number of medical registries has greatly increased. Their value strongly depends on the quality of data contained in the registry [1]. With the development of informatics technology, medical databases tend to be more reliable. Issues regarding data quality are now even more relevant as the utilization of these databases increase in magnitude and importance. For health care organizations, data is vital to an effective health care and to financial survival. Data regarding effectiveness of treatments, accuracy of diagnosis and practices of health care providers is crucial to organizations that do their best to maintain and improve health care delivery. Hospital inpatient databases, commonly referred as Administrative Data, are on the most relevant medical repositories. Administrative Data is present in all hospitals in Portugal and in the majority of hospitals around the world. There are many studies, researches and analyses based on administrative data but there are also many problems in the use of this type of data.
A. Administrative Data Administrative data was created primarily to monitor utilization, to determine the consumption of resources, or to find out the capacity to supply a service. Administrative data was generated as part of standard hospital discharge coding procedures. In these procedures, certified coding clerks’ record diagnoses and procedures in standard ICD-9CM format and abstract information from hospital medical records. Administrative data is normally defined as: “large, computerized data files generally compiled in billing for healthcare services such as hospitalization” [2], and should only be used to identify areas for further study [3], whereas they can have significant influence in the improvement of healthcare quality [4]. The core data elements of the administrative data system are admission date, discharge date and status, principal and secondary ICD-9-CM diagnoses, procedures, external causes of injury and some demographic information. This data is often available in compiled research databases from federal agencies, state health departments, health plans, and private data institutions [5]. In Portugal, administrative data contain almost the same variables that exist in other countries. There is common database in all hospitals of the National Health Service (NHS), containing variables of administrative data along with some others variables related to Diagnosis Related Groups (DRG). This database has been used since 1990 [6] with a positive impact on the productivity and technical efficiency of some diagnostic technologies [7]. The use of administrative data for health care quality reporting involves balancing between a number of advantages and disadvantages. This type of data is informative about major processes of healthcare; however, they are primarily collected for payment purposes. For quality assessment such data is often inaccurate and offer reduced clinical detail. Death, length-of-stay and readmission usually are consistent descriptors, but accuracy of diagnosis coding can be inconsistent. Nevertheless, this type of databases can be easily obtained, include millions of episodes and covers large areas.
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 823–826, 2010. www.springerlink.com
824
J.A. Freitas et al.
B. Taxonomy of Data Quality Problems
⋅ Incorrect reference [IR], when the referential integrity is respected but the foreign key contains a value, which is not the correct one; ⋅ Heterogeneity of syntaxes [HS], the existence of different representation syntaxes in related attributes; ⋅ Heterogeneity of measure units [HMU], the use of different measure units in related attributes; ⋅ Heterogeneity of representation [HR], the use of different sets of values to code the same real-world property; ⋅ Existence of homonyms [EH], the use of syntactically equal values with different meanings, among related attributes from multiple data sources.
II. EXAMPLES AND IMPLICATIONS OF DATA QUALITY PROBLEMS IN HOSPITAL ADMINISTRATIVE DATA
Fig. 1 Organization model of relational data As described by Oliveira et al. [8], data quality problem can be classified as (Fig. 1): ⋅ Missing values [MV], when a required attribute is not filled; ⋅ Syntax violation [SV], when an attribute value violates the predefined syntax; ⋅ Domain violation [DV], when an attribute value violates the domain of valid values; ⋅ Incorrect value [IV], when an attribute contains a value which is not the correct one, but the domain of valid values is not violated; ⋅ Violation of business rule [VBR], a problem that can happens at all granularity levels, when a given business domain rule is violated; ⋅ Uniqueness violation [UV], when two (or more) tuples have the same value in a unique value attribute; ⋅ Existence of synonyms [ES], the use of syntactically different values with the same meaning; ⋅ Violation of functional dependency [VFD], when the value of a tuple violates an existing functional dependency among two or more attributes; ⋅ Approximate duplicate tuples [ADT], the same realworld entity is represented (equally or with minor differences) in more than one tuple; ⋅ Inconsistent duplicate tuples [IDT], when the same realworld entity is represented in more than one tuple but with inconsistencies between attributes values; ⋅ Referential integrity violation [RIV], when a value in a foreign key attribute does not exist in the related relation as a primary key value;
Next we present examples and discuss possible implications of some problem in administrative data. The database associated with the Portuguese resource allocation system was the main source of data for these analyses. This database, with 9.098.628 episodes between 2000 and 2007, includes data from inpatient and outpatient discharges from the majority of the public acute care hospitals of the National Health Service (NHS). The access to the data was provided by ACSS, I. P. (Administração Central do Sistema de Saúde, I. P.), the Ministry of Health’s Central Administration for the Health System. Clear Identification of Type of Care (Missing Values) Missing values occurs when a required attribute is not filled, that is, there is absence of value in a mandatory attribute. The variable Type of care distinguishes between the different types of care the hospital provides, namely indicating if it is an outpatient or an inpatient episode. In the described database we may find 2.411.772 (26.5%) episodes with missing Type of Care. Possible implications: These missing values can have negative impact in many situations, including in the calculation of the hospital Case-Mix Index and consequently possibly affecting hospital budget. Even for research, and for many episodes, it is not either easy or clear, how to use other variables to split data. So, this is a fundamental variable for the majority of data analysis performed using these databases. Length-of-Stay (Domain Violation) We found 232 problems in the variable Length-of-stay, including a few cases with negative values. Possible implications: these errors are uncommon but can have implication in the hospital budget (these cases will be excluded). Clearly these errors should also be considered
IFMBE Proceedings Vol. 29
Implications of Data Quality Problems within Hospital Administrative Databases
in any statistical analysis or in any process of calculation of performance or quality indicators. Incorrect Value for Age The values of variable Age can be checked using two other variables: Birth date of the patient and the Admission date present on the registry. We recalculated this variable and detected a difference of one year of age in 290.479 episodes. We suspect that the problem is in the rule used by the system in some hospitals to calculate the age of patients. These systems might have only used the year as base of calculation while in our rule we used the entire date (DDMM-YYYY). Possible implications: a difference of one year in age may originate a different Diagnosis-Related Group (DRG). In fact, the age is one of the variables used to group episodes into DRGs. For instance, in AP-DRG (All Patient DRGs) version 21, age is a fundamental factor in the assignment of DRGs 185 and 186. According to the Ministry of Health, DRG 185, for age > 17 years, is priced 1.056€ while DRG 186, for age < 18 years, is priced 659€. An episode of an 18 years-old patient with age incorrectly assigned to 17 years will erroneously lead to the attribution of DRG 186 (with direct implications in the hospital budget). This situation can occur in many other DRGs. This wrong attribution of age can also have implications in many other analyses where age is one the factors. Hospital Transfers (Violation of Business Rule) We studied inter-hospitals transfers and found important problems in the national database. In general, the number of patient transfers from hospital A does not correspond to the number of patients received in other hospitals originated from hospital A. Specifically, we found a total of 263.639 patient transfers, and 556.710 patient receptions (ideally, if all data was correctly coded, this value should be equal). We also verified that not all transferred patients have a correspondence in the group of received patients. Possible implications: this data problem has evident implications in hospital reimbursement. A hospital that transfers a patient to other hospital due to lack of resources, can receive at most 50% of the value of the DRG. On the other side, if the patient is transferred to continue to receive hospital care; the hospital that receives the patient is limited to a restricted group of DRGs. This problem may also impose limitations to any analyses of patient movement’s and to studies about the efficiency of the NHS regarding the fulfillment of patient treatments.
825
Duplicated Episodes By analyzing the entire tuple of attributes, we detected 30.932 duplicate cases. This means that patients have different registries in the same date, at the same hospital, and matching all clinical information (e.g., diagnosis and procedures). This is a clear example of inconsistent duplicate tuples. Possible implications: if not considered, these repetitions can adulterate any calculation of indicators or any other statistical analysis. Principal Diagnosis (Referential Integrity Violation) We studied the variable Principal diagnosis and found 565 episodes with diagnosis code without respective reference in the ICD-9-CM lookup table (invalid code). According to the briefly described taxonomy, this situation can be classified under the problem of referential integrity violation. Possible implications: episodes with this problem are classified with DRG 470 (ungroupable), and the hospital will not be reimbursed for them (claims grouped in this DRG will have payment denied). Hospital Department Codification We detected different representations for the variable that represents the hospital department code (the first department for each hospital stay). As an example for this heterogeneity we considered the department of Pneumology, and found presentations such as “Pneum” (2550 cases), “Pneumo” (531 cases) and the code “210” (with 22cases). Possible implications: This heterogeneity of representations can have several implications. In can, for instance, complicate any study comparing medical specialties among hospitals or any study related with departmental transfers. Repeated ICD-9-CM Codes Another problem was found when studying diagnosis and procedures. We found repeated codes inside one episode, i.e., the same ICD-9-CM code was found, for instance, in the principal diagnosis and in one (or more) secondary diagnosis. In this situation we found 99.995 repetitions, representing 1.1% of total episodes. Considering procedures, we found 121.850 repetitions (1.3% of all episodes have repeated procedures). Possible implications: Under these circumstances we need to be careful when using the number of secondary diagnoses as a proxy for severity of illness, or if we want to study the evolution in the number of procedures or secondary diagnosis coded.
IFMBE Proceedings Vol. 29
826
J.A. Freitas et al.
III. CONCLUSIONS In this paper we present some problems and some possible implications. Some of these problems can be corrected, and other can not. Nevertheless, they should always be considered when interpreting results from analysis of administrative databases. Administrative data can contain inaccurate and unstable data, but they are readily available, relatively inexpensive and are widely used [2, 9]. But in some situations they can be the only source of information to look for a clinical question. Despite some existing problems [10-13], administrative data can, for instance, be used in the production of quality indicators, or for providing benchmarks for hospital activity [5, 11, 14, 15]. However calculating these indicators using only this data may bias quality reports. This analysis must be done carefully in order to avoid these problems. Health care organization and consumers are concerned on how health care can be improved and costs reduced thought the use of information technology. Improving decision-making with better data can make a significant contribution. In this context it is crucial to understand data and to give steps toward a continuous increase of data quality.
ACKNOWLEDGMENT The authors with like to thank the support given by the research project HR-QoD - Quality of data (outliers, inconsistencies and errors) in hospital inpatient databases: methods and implications for data modeling, cleansing and analysis (project PTDC/SAU-ESA/75660/2006).
REFERENCES 1. Arts D, Keizer N, Scheffer G-J. Defining and Improving Data Quality in Medical Registries: A Literature Review Case Study, and Generic Framework. J Am Med Inform Assoc l2002;9: 600-611. 2. Iezzoni LI. Risk Adjustment for Measuring Health Care Outcomes. Health Administrative Press; Chicago, IL. Foundation of the American College of Executives; 1997.
3. Iezzoni LI. Assessing Quality Using Administrative Data. Ann Intern Med l1997;127: 666-673. 4. Price J, Estrada CA, Thompson D. Administrative Data Versus Corrected Administrative Data. Am J Med Qual l2003;19: 38-44. 5. Zhan C, Miller MR. Administrative data based patient safety research: a critical review. Qual Saf Health Care l2003;12: 58-63. 6. Bentes ME, Urbano JA, Carvalho MdC, Tranquada MS. Using DRGs to Fund Hospitals in Portugal: An Evaluation of the Experience. In: Diagnosis Related Groups in Europe: Uses and perspectives. M. Casas, M. M. Wiley (Eds.). Springer-Verla ,Berlin; 1993. 7. Dismuke C, Sena V. Has DRG Payment Influenced The Technical Efficiency and Productivity Of Diagnostic Technologies In Portuguese Public Hospitals? An Empirical Analysis. In: The V European Workshop on Efficiency and Productivity Analysis. Copenhagen, Denmark; 1997. 8. Oliveira P, Rodrigues F, Henriques P, Galhardas H. A Taxonomy of Data Quality Problems. In: 2nd International Workshop on Data and Information Quality. Porto, Portugal; 2005. 9. Torchiana D, Meyer G. Use of administrative data for clinical quality measurement. J Thorac Cardiovasc Surg l2005;129: 1222-4. 10. Powell AE, Davies HTO, Thomson RG. Using routine comparative data to assess the quality of health care: understanding and avoiding common pitfalls. Qual Saf Health Care l2003;12: 122–128. 11. Williams J, Mann R. Hospital episode statistics: time for clinicians to get involved? Clin Med l2002;2: 34–37. 12. Calle J, Saturno P, Parra P, Rodenas J, Perez M, Eustaquio F, Aguinaga E. Quality of the information contained in the minimum basic data set: results from an evaluation in eight hospitals. European Journal of Epidemiology l2000;16: 1073-80. 13. Sutherland JM, Botz CK. The effect of misclassification errors on case mix measurement. In: Health Policy; 2006. 14. Jarman B, Gault S, Alves B, Hider A, Dolan S, Cook A, Hurwitz B, Iezzoni LI. Explaining differences in English hospital death rates using routinely collected data. BMJ l1999;318: 1515-20. 15. Scott I, Youlden D, Coory M. Are diagnosis specific outcome indicators based on administrative data useful in assessing quality of hospital care? Qual Saf Health Care l2004;13: 32–39.
Corresponding author: Author: Alberto Freitas Institute: Department of Biostatistics and Medical Informatics, Faculdade de Medicina, Universidade do Porto, Portugal Street: Alameda Hernani Monteiro City: 4200 Porto Country: Portugal Email:
[email protected] IFMBE Proceedings Vol. 29
Can the EEG Indicate the FiO2 Flow of a Mechanical Ventilator in ICU Patients with Respiratory Failure? E.G. Peranonti1, M.A. Klados2, C.L. Papadelis3, D.G. Kontotasiou2, C. Kourtidou-Papadeli3, and P.D. Bamidis1 2
1 Greek Aerospace Medical Association and Space Research, Thessaloniki, Greece Aristotle University of Thessaloniki, School Of Medicine, Laboratory of Medical Informatics, P.O. Box 323 54124, Thessaloniki, Greece 3 Center for Brain/Mind Sciences (CIMEC), University of Trento, Mattarello, (TN) Italy
Abstract— The aim of this paper is to show that the brain activity of patients with acute respiratory failure hospitalized in Intensive Care Units (ICUs) can provide useful medical information, which is directly related to neurological rehabilitation. It also aims to show that the entropy and kurtosis, widely used indices of the electroencephalographic (EEG) signals, are able to identify EEG changes associated with cerebral hypoxia. EEG signals were recorded from eight adult patients with acute respiratory failure admitted to the ICU. The measurements were recorded in five stages, with FiO2 at 40%, 100%, 60%, 20% and 0% (T-piece) respectively. Total time of recordings was 50min (10 min. for each stage). The EEG signals were filtered and further cleaned from ocular and muscular artifacts as well as from the artifacts introduced by other external devices, electrodes movements and electrode’s bad tangencies. Afterwards the 10-min EEG signals of each stage were segmented in ten epochs with one minute fixed length. Then Kurtosis and Shannon’s Entropy were calculated in each segment. One-Way ANOVA verified the assumption that there are statistically significant differences between the various stages of our protocol, while the Scheffe Post-Hoc tests revealed the homogeneous subsets compiled by the aforementioned stage. The results suggest that the EEG is directly connected with the mechanical ventilator’s changes, so in the future, clinicians could probably use the EEG as particularly useful and time-critical information, especially during the weaning procedure from the mechanical ventilator. Keywords— ICU, EEG, Kurtosis, Entropy, Respiratory Failure.
I. INTRODUCTION Weaning from mechanical ventilation is an essential and universal element in the care of critically ill patients intubated in mechanical ventilation. It covers the entire process of liberating the patient from mechanical support and from the endotracheal tube, including relevant aspects of terminal care. There is an uncertainty about the best methods for conducting this process, which will generally require the cooperation of the patient during the phase of recovery from
critical illness. This makes weaning an important clinical issue for patients and clinicians [1]. Vallverdu et al [2] reported that weaning failure occurred in 61% of chronic obstructive pulmonary disease patients, in 41% of neurological patients and in 38% of hypoxaemic patients. Patients with difficult weaning represent 10% of ICU admissions and consume a significant amount of the overall ICU patient-days and 50% of financial resources [3]. Therefore, strategies to improve the weaning process are needed. The international consensus conference [2] identified six stages within the weaning process as illustrated in: 1. 2. 3. 4. 5. 6.
Treatment of acute respiratory failure (ARF). Suspicion that weaning may be possible. Assessment of readiness to wean. Spontaneous breathing trial (SBT). Extubation and possibly. Reintubation.
Weaning from mechanical ventilation usually implies two separate but closely related aspects of care, discontinuation of mechanical ventilation and removal of any artificial airway. The first problem that clinician faces, is how to determine when a patient is ready to resume ventilation on his or her own. Several studies [4,5] have shown that a direct method of assessing readiness to maintain spontaneous breathing is simply to initiate a trial of unassisted breathing. Once a patient is able to sustain spontaneous breathing, a second judgment must be made regarding whether the artificial airway can be removed. This decision is made on the basis of the patient's mental status, airway protective mechanisms, ability to cough and character of secretions. If the patient has an adequate sensorium with intact airway protection mechanisms, and is without excessive secretions, it is reasonable to extubate the trachea [6]. Early prediction of neurological rehabilitation of the patient would be very important, providing the clinician with useful information for the optimal time initiating the weaning procedure from the mechanical ventilation. The electroencephalogram (EEG) allows us to continuously monitor and record the brain activity of the patient. Although the
P.D. Bamidis and N. Pallikarakis (Eds.): MEDICON 2010, IFMBE Proceedings 29, pp. 827–830, 2010. www.springerlink.com
828
E.G. Peranonti et al.
EEG provides a particularly useful and time-critical information, it is not widely used [7]. To our best knowledge, there is a significant lack of studies related to the EEG recordings during the weaning procedure. This work comes to investigate if the EEG is capable to measure the brain responses during the weaning procedure. It also shows that kurtosis and Shannon’s entropy, widely used EEG parameters, are useful indicators of neurological recovery of patients with acute respiratory failure, who are hospitalized in ICUs.
II. MATERIALS AND METHODS A. Subjects Eight adult patients with acute respiratory failure admitted to the ICU of Saint Paul General Hospital, Thessaloniki, Greece, from April 2007 to December 2007, were enrolled in this study. B. Protocol At the phase 1 of the study, all the patients reaches the criteria to be disconnected from mechanical ventilation: Vital capacity-10 mL/ kg, Tidal volume > 4 mL/ kg , Minute ventilation 22 mL/cm H2O, Static compliance > 33 mL/cm H2O, Pa O2/PAO2 ratio > 0.35, Peak negative pressure 20-30 cm H2O, Dead space to tidal volume ratio 0.6, Frequency/tidal volume ratio < 105 breaths/min/L. At phase 2 BISpectral index (BIS) was used in order to evaluate the depth of anaesthesia [8]. Afterwards a neurologic deficit score from a specialist neurologist or the intensivist took place and the BIS device was then disconnect in order to apply the INVOS, for oxygen saturation measurement, ECG and EEG. Here it has to be mentioned that the INVOS® System provides a direct, non-invasive measurement of changes in regional brain blood oxygen status. Is a real-time guide to therapeutic intervention and it measures changes in oxygen levels beneath sensors in order to protect the brain and vital organ areas from hypoxia. At the third phase of the protocol, the FiO2 was initialized at 40% (Stage1) for 10 min, and then the FiO2 was increased in 100% (Stage2) for also 10 min. Afterwards we have decreased the FiO2 to 60% for 10 min (stage3) and finally return to 40% again for 10 min (stage4) also. At the phase 4, a 2 min recording took place when patient is in at the process of disconnection from mechanical ventilation, while at phase 5, a 10 min recording when patient is in T piece with oxygen administration (stage5).
C. Data Acquisition and Analysis Multichannel EEG measurements took place during the aforementioned protocol. Nineteen scalp electrodes were placed according to the 10-20 International-System. The earlobes were used as reference points [9]. The signals were digitized at a rate of 500 Hz and were further filtered (band pass filter at 0.5-40 Hz and notch filter at 50Hz). Electroocculographic (EOG) recordings took place simultaneously with the EEG. Two electrodes where placed above and below the left eye, while another two at the outer canthi of each eye. From these electrodes, two bipolar signals were obtained, namely vertical-EOG (VEOG), which is equal to upper minus lower electrode values and horizontalEOG (HEOG) which is equal to left minus right electrode values. Each EOG signal was further filtered with a bandpass filter at 0.5-5 Hz and with a notch filter at 50 Hz for power line noise extraction. In order to obtain more accurate results, EOG and EMG artifacts were rejected. For EOG artifact rejection, REGICA technique was applied [10] in the filtered EEG data, while the EMG artifacts were rejected according to [11]. Afterwards three independent EEG observers marked the contaminated segments by external noise like electrode movements or electrodes’ bad tangencies. The next linear formula was used in order to fill these corrupted segments: a
EEG (i, x) b = EEG (i, a) ⋅
b− x x−a + EEG (i, b) ⋅ b−a b−a
where a, b are the starting and the ending sample point of the corrupted segment respectively, x is inside the set [a, b] and denotes the sample point which is under estimation, while EEG and i defines the EEG waveform in i th electrode. D. Kurtosis In probability theory and statistics, kurtosis (from the Greek word kurtos, meaning bulging) is a measure of the "peakedness" of the probability distribution of a real-valued random variable. Higher kurtosis means more of the variance is due to infrequent extreme deviations, as opposed to frequent modestly-sized deviations. Data sets with high kurtosis tend to have a distinct peak near the mean, decline rather rapidly, and have heavy tails. Data sets with low kurtosis tend to have a flat top near the mean rather than a sharp peak. A uniform distribution would be the extreme case. Kurtosis was computed by the next formula:
IFMBE Proceedings Vol. 29
K = m4 − 3m22
Can the EEG Indicate the FiO2 Flow of a Mechanical Ventilator in ICU Patients with Respiratory Failure?
Table 2 Kurtosis
where m n is the n th central moment:
mn = E
{ (x − ~x ) }. n
Stage
E. Shannon’s Entropy The Shannon entropy was first introduced by Shannon [12]. Entropy is a function of the Probability Distribution Function (PDF) p(x) , and is sometime written as:
H ( p( x1 ), p( x2 ),..., p( xn )) It has to be noted that the entropy of X does not depend on the actual values of X , it only depends on p ( x) . The definition of Shannon's entropy can be written as the above expectation value:
829 Mean Values
kurtosis (mean ±SD)
Stage 1
0,6681±2.08985
Stage 2
1.4914±8.47761
Stage 3
0,3226±0,64772
Stage 4
0,2154±0,38658
Stage 5
0,2312±0,39702
Statistics
F(102.087)= 6.305 p