Magnetic Resonance Neuroimaging: Methods and Protocols (Methods in Molecular Biology 711)

ME T H O D S IN MO L E C U L A R BI O L O G Y Series Editor John M. Walker School of Life Sciences University of Her...

Author: Michel Modo | Jeff W.M. Bulte

23 downloads 600 Views 17MB Size Report

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

DOWNLOAD PDF

ME T H O D S

IN

MO L E C U L A R BI O L O G Y

Series Editor John M. Walker School of Life Sciences University of Hertfordshire Hatfield, Hertfordshire, AL10 9AB, UK

For other titles published in this series, go to www.springer.com/series/7651

TM

Magnetic Resonance Neuroimaging Methods and Protocols

Edited by

Michel Modo Department of Neuroscience, Institute of Psychiatry, King’s College London, UK

Jeff W.M. Bulte Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA

Editors Michel Modo Department of Neuroscience Institute of Psychiatry King’s College London London, SE5 9NU, UK [email protected]

Jeff W.M. Bulte Russell H. Morgan Department of Radiology and Radiological Science Division of MR Research Cellular Imaging Section Institute for Cell Engineering The Johns Hopkins University School of Medicine Baltimore, MD, 21205, USA [email protected]

ISSN 1064-3745 e-ISSN 1940-6029 ISBN 978-1-61737-991-8 e-ISBN 978-1-61737-992-5 DOI 10.1007/978-1-61737-992-5 Springer New York Dordrecht Heidelberg London © Springer Science+Business Media, LLC 2011 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Humana Press, c/o Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. While the advice and information in this book are believed to be true and accurate at the date of going to press, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein. Printed on acid-free paper Humana Press is part of Springer Science+Business Media (www.springer.com)

Preface To visualize the inside of a living human brain has been the goal of physicians since ancient times. The advent of noninvasive imaging technology, such as magnetic resonance imaging (MRI), during the latter half of the twentieth century has allowed for the opening of new vistas of the inner workings of the brain to biologists and clinicians on a daily basis. Great strides in unraveling the secrets of the brain have been achieved since the widespread implementation of imaging protocols in universities and hospitals. The gradual merging of molecular biology and imaging techniques at the beginning of the twenty-first century now affords a detailed investigation of the molecular underpinning of a working brain. The 30 chapters in this book contain experimental MRI protocols that can be used to noninvasively interrogate the healthy and diseased brain. The protocols are divided into general techniques (e.g., measuring relaxivity, magnetic resonance spectroscopy, diffusion tensor imaging, MR reporter genes) and specific applications in brain imaging (e.g., phenotyping transgenic animals, detecting amyloid plaques, fMRI in psychiatry). Most of these methods can be applied to both animal and human studies and may therefore provide a great resource for translational efforts. Clinical neurologists, psychiatrists, and radiologists will find these protocols useful, as will basic scientists working in the field of neuroscience. Michel Modo Jeff W.M. Bulte

v

Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

v

Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xi

SECTION I 1.

INTRODUCTION

From Molecules to Man: The Dawn of a Vitreous Man . . . . . . . . . . . . . . Michel Modo and Jeff W.M. Bulte

SECTION II

3

GENERAL TECHNIQUES

2.

Magnetic Resonance Safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrew Simmons and Kristina Hakansson

17

3.

Measuring the Absolute Water Content of the Brain Using Quantitative MRI . . . Nadim Joni Shah, Veronika Ermer, and Ana-Maria Oros-Peusquens

29

4.

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain . . . Sean C.L. Deoni

65

5.

Magnetic Resonance Brain Image Processing and Arithmetic with FSL . . . . . . 109 William R. Crum

6.

Diffusion Tensor Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 Derek K. Jones and Alexander Leemans

7.

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI) . . . . . . . . . . 145 Cynthia A. Massaad and Robia G. Pautler

8.

Sodium MRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 Ronald Ouwerkerk

9.

MR Spectroscopy and Spectroscopic Imaging of the Brain . . . . . . . . . . . . . 203 He Zhu and Peter B. Barker

10. Amide Proton Transfer Imaging of the Human Brain . . . . . . . . . . . . . . . 227 Jinyuan Zhou 11. High-Field MRI of Brain Iron . . . . . . . . . . . . . . . . . . . . . . . . . . . 239 Jozef H. Duyn 12. Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications . . . 251 Manisha Aggarwal, Jiangyang Zhang, and Susumu Mori 13. CEST MRI Reporter Genes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 Guanshu Liu, Jeff W.M. Bulte, and Assaf A. Gilad

vii

viii

Contents

14. Longitudinal Functional Magnetic Resonance Imaging in Animal Models . . . . . 281 Afonso C. Silva, Junjie V. Liu, Yoshiyuki Hirano, Renata F. Leoni, Hellmut Merkle, Julie B. Mackel, Xian Feng Zhang, George C. Nascimento, and Bojana Stefanovic 15. Combining EEG and fMRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303 Karen Mullinger and Richard Bowtell 16. MR Angiography and Arterial Spin Labelling . . . . . . . . . . . . . . . . . . . 327 David Thomas and Jack Wells SECTION III SPECIFIC APPLICATIONS 17. MRI Phenotyping of Genetically Altered Mice . . . . . . . . . . . . . . . . . . . 349 Jason P. Lerch, John G. Sled, and R. Mark Henkelman 18. Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications . . . . . . . 363 Philip K. Liu and Christina H. Liu 19. Molecular MRI Approaches to the Detection of CNS Inflammation . . . . . . . . 379 Nicola R. Sibson, Daniel C. Anthony, Sander van Kasteren, Alex Dickens, Francisco Perez-Balderas, Martina A. McAteer, Robin P. Choudhury, and Benjamin G. Davis 20. Brain Redox Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397 Ken-ichiro Matsumoto, Fuminori Hyodo, Kazunori Anzai, Hideo Utsumi, James B. Mitchell, and Murali C. Krishna 21. Systems Biology Approach to Imaging of Neural Stem Cells . . . . . . . . . . . . 421 Li Hua Ma, Yao Li, Petar M. Djurić, and Mirjana Maletić-Savatić 22. MRI of Transplanted Neural Stem Cells . . . . . . . . . . . . . . . . . . . . . . 435 Stacey M. Cromer Berman, Piotr Walczak, and Jeff W.M. Bulte 23. MRI of Experimental Gliomas . . . . . . . . . . . . . . . . . . . . . . . . . . . 451 Frits Thorsen 24. MRI in Experimental Stroke . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473 Timothy Q. Duong 25. Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of Parkinson’s Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487 Anthony C. Vernon and Michel Modo 26. Detecting Amyloid-β Plaques in Alzheimer’s Disease . . . . . . . . . . . . . . . . 511 Christof Baltes, Felicitas Princz-Kranz, Markus Rudin, and Thomas Mueggler 27. Assessing Subtle Structural Changes in Alzheimer’s Disease Patients . . . . . . . . 535 Jennifer L. Whitwell and Prashanthi Vemuri 28. Pharmacological Application of fMRI . . . . . . . . . . . . . . . . . . . . . . . 551 Mitul A. Mehta and Owen G. O’Daly

Contents

ix

29. MRI of Neuronal Plasticity in Rodent Models . . . . . . . . . . . . . . . . . . . 567 Galit Pelled 30. MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579 Yuexi Huang and Kullervo Hynynen Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595

Contributors MANISHA AGGARWAL • The Russell H. Morgan Department of Radiology and Radiological Science, Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, USA DANIEL C. ANTHONY • Department of Pharmacology, University of Oxford, Oxford, UK KAZUNORI ANZAI • Radiation Modifier Research Team, Heavy-Ion Radiobiology Research Group, National Institute of Radiological Sciences, Research Center for Charged Particle Therapy, Chiba, Japan CHRISTOF BALTES • Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland PETER B. BARKER • The Kennedy Krieger Institute, Baltimore, MD, USA; Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, USA RICHARD BOWTELL • School of Physics and Astronomy, Sir Peter Mansfield Magnetic Resonance Centre, University of Nottingham, Nottingham, UK JEFF W.M. BULTE • Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA ROBIN P. CHOUDHURY • Department of Cardiovascular Medicine, University of Oxford, Oxford, UK STACEY M. CROMER BERMAN • Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA WILLIAM R. CRUM • King’s College London, Institute of Psychiatry, Centre for Neuroimaging Sciences, London, UK BENJAMIN G. DAVIS • Department of Chemistry, University of Oxford, Oxford, UK SEAN C.L. DEONI • Division of Engineering, Brown University, Providence, RI, USA ALEX DICKENS • Department of Pharmacology, Department of Chemistry, CR-UK/MRC Gray Institute for Radiation Oncology and Biology, University of Oxford, Oxford, UK PETAR M. DJURI C´ • Department of Computer and Electrical Engineering, Stony Brook University, Stony Brook, NY, USA TIMOTHY Q. DUONG • Research Imaging Institute, University of Texas Health Science Center at San Antonio, San Antonio, TX, USA JOZEF H. DUYN • Advanced MRI Section, National Institute for Neurological Disease and Stroke, National Institute of Health, Bethesda, MD, USA VERONIKA ERMER • Institute of Neuroscience, Medicine (INM-4), Research Centre Juelich, Juelich, Germany ASSAF A. GILAD • Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA

xi

xii

Contributors

KRISTINA HAKANSSON • King’s College London, Institute of Psychiatry, Centre for Neuroimaging Sciences, London, UK R. MARK HENKELMAN • The Mouse Imaging Centre, Hospital for Sick Children, Toronto, ON, Canada; Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada YOSHIYUKI HIRANO • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA YUEXI HUANG • Sunnybrook Health Sciences Centre, Toronto, ON, Canada KULLERVO HYNYNEN • Sunnybrook Health Sciences Centre, Toronto, ON, Canada; Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada FUMINORI HYODO • Innovation Center for Medical Redox Navigation, Kyushu University, Fukuoka, Japan DEREK K. JONES • School of Psychology, Cardiff University Brain Research Imaging Centre (CUBRIC), Cardiff University, Cardiff, UK MURALI C. KRISHNA • Radiation Biology Branch, National Cancer Institute, National Institutes of Health, Center for Cancer Research, Bethesda, MD, USA ALEXANDER LEEMANS • Image Sciences Institute, University Medical Center Utrecht, Utrecht, The Netherlands RENATA F. LEONI • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA JASON P. LERCH • The Mouse Imaging Centre, Hospital for Sick Children, Toronto, ON, Canada; Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada YAO LI • Department of Computer and Electrical Engineering, Stony Brook University, Stony Brook, NY, USA JUNJIE V. LIU • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA PHILIP K. LIU • Department of Radiology, AA Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, USA CHRISTINA H. LIU • Department of Radiology, AA Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, USA GUANSHU LIU • Kennedy Krieger Institute, F.M. Kirby Research Center for Functional Brain Imaging, The Johns Hopkins University School of Medicine, Baltimore, MD, USA LI HUA MA • Department of Pediatrics, Baylor College of Medicine, Houston, TX, USA JULIE B. MACKEL • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA MIRJANA MALETI C´ -SAVATI C´ • Department of Pediatrics, Baylor College of Medicine, Houston, TX, USA CYNTHIA A. MASSAAD • Department of Molecular Physiology and Biophysics, Baylor College of Medicine, Houston, TX, USA KEN-ICHIRO MATSUMOTO • Radiation Modifier Research Team, Heavy-Ion Radiobiology Research Group, National Institute of Radiological Sciences, Research Center for Charged Particle Therapy, Chiba, Japan

Contributors

xiii

MARTINA A. MCATEER • Department of Cardiovascular Medicine, University of Oxford, Oxford, UK MITUL A. MEHTA • Institute of Psychiatry at King’s College London, Centre for Neuroimaging Sciences (PO89), London, SE5 8AF, UK; Medical Research Council Clinical Sciences Centre, Imperial College London, Hammersmith Hospital, London, UK HELLMUT MERKLE • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA JAMES B. MITCHELL • Radiation Biology Branch, National Cancer Institute, National Institutes of Health, Center for Cancer Research, Bethesda, MD, USA MICHEL MODO • Department of Neuroscience, Kings College London, Institute of Psychiatry, Centre for the Cellular Basis of Behaviour, The James Black Centre, London, UK SUSUMU MORI • The Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, USA THOMAS MUEGGLER • Pharmaceuticals Division, F. Hoffmann-La Roche Ltd, Basel, Switzerland KAREN MULLINGER • School of Physics and Astronomy, Sir Peter Mansfield Magnetic Resonance Centre, University of Nottingham, Nottingham, UK GEORGE C. NASCIMENTO • Department of Biomedical Engineering, Federal University of Rio Grande do Norte, Campus Universitario, Natal, RN, Brazil OWEN G. O’DALY • Institute of Psychiatry at King’s College London, Centre for Neuroimaging Sciences (PO89), London, UK ANA-MARIA OROS-PEUSQUENS • Institute of Neuroscience, Medicine (INM-4), Research Centre Juelich, Juelich, German; King’s College London, Institute of Psychiatry, Centre for Neuroimaging Sciences, London, UK RONALD OUWERKERK • Cardiovascular Imaging, National Institute of Diabetes and Digestive and Kidney Disease, National Institute of Health, Bethesda, MD, USA ROBIA G. PAUTLER • Department of Molecular Physiology and Biophysics, Department of Neuroscience, Department of Radiology, Baylor College of Medicine, Houston, TX, USA GALIT PELLED • Department of Radiology, Kennedy Krieger Institute, Johns Hopkins School of Medicine, Baltimore, MD, USA FRANCISCO PEREZ-BALDERAS • CR-UK/MRC Gray Institute for Radiation Oncology and Biology, University of Oxford, Oxford, UK FELICITAS PRINCZ-KRANZ • Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland MARKUS RUDIN • Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland; Institute of Pharmacology and Toxicology, University of Zurich, Zurich, Switzerland NADIM JONI SHAH • Institute of Neuroscience and Medicine (INM-4), Research Centre Juelich, Juelich, Germany; RWTH Aachen University, Aachen, Germany NICOLA R. SIBSON • CR-UK/MRC Gray Institute for Radiation Oncology and Biology, University of Oxford, Oxford, UK AFONSO C. SILVA • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA ANDREW SIMMONS • King’s College London, Institute of Psychiatry, Centre for Neuroimaging Sciences, London, UK

xiv

Contributors

JOHN G. SLED • Department of Medical Biophysics, The Mouse Imaging Centre, Hospital for Sick Children, Toronto, ON, Canada; University of Toronto, Toronto, ON, Canada BOJANA STEFANOVIC • Imaging Research, Sunnybrook Health Science Centre, Toronto, ON, Canada DAVID THOMAS • Department of Medical Physics and Bioengineering, University College London, London, UK FRITS THORSEN • Department of Biomedicine, University of Bergen, Bergen, Norway HIDEO UTSUMI • Innovation Center for Medical Redox Navigation, Kyushu University, Fukuoka, Japan SANDER VAN KASTEREN • Department of Chemistry, University of Oxford, Oxford, UK PRASHANTHI VEMURI • Department of Radiology, Mayo Clinic, Rochester, MN, USA ANTHONY C. VERNON • Department of Neuroscience, Institute of Psychiatry, Centre for the cellular basis of behaviour, The James Black Centre, Kings College London, London, UK PIOTR WALCZAK • Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA JACK WELLS • Centre for Advanced Biomedical Imaging, University College London, London, UK JENNIFER L. WHITWELL • Department of Radiology, Mayo Clinic, Rochester, MN, USA XIAN FENG ZHANG • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA JIANGYANG ZHANG • The Russell H. Morgan Department of Radiology and Radiological Science, F.M. Kirby Research Center for Functional Brain Imaging, Johns Hopkins University School of Medicine, Baltimore, MD, USA JINYUAN ZHOU • Department of Radiology, Johns Hopkins University, Baltimore, MD, USA HE ZHU • Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, USA

Section I Introduction

Chapter 1 From Molecules to Man: The Dawn of a Vitreous Man Michel Modo and Jeff W.M. Bulte Abstract One of the greatest challenges to study the structure, function, and molecules in the living brain is that it is enclosed within the skull and difficult to access. Although biopsies are feasible, they are invasive, could lead to functional impairments, and in any case will only provide a small regional sample that is not necessarily reflecting the entire brain. Since the beginning of the twentieth century, in vivo imaging has gradually, and steadily, matured into non-invasive techniques that enable the repeated investigation of the structural, functional, cellular, and molecular composition of the brain. Not only is this information of great importance to scientists aiming to understand how the brain works, but these techniques are also essential to physicians who use imaging to diagnose and treat disease. The current book is a collection of 29 cutting-edge methods and protocols that are used in the current field of neuroimaging. Key words: Neuroimaging, magnetic resonance imaging, computer tomography, positron emission tomography.

1. Looking Inside the Living Brain: A Historical Primer

Correct functioning of the brain is central to our everyday lives. Developmental problems or damage to the brain can interfere with someone’s ability to take care of basic live functions. However, the study of the brain is hampered by it being enclosed by the skull that prevents us from seeing or studying the brain directly. Initial theories and studies of the brain were based on ill-conceived ideas, but these have gradually paved the way for a thorough scientific study of the brain and mind. Since the beginning of modern medicine, these studies have been highly dependent on technological developments.

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_1, © Springer Science+Business Media, LLC 2011

3

4

Modo and Bulte

Arguably, the eighteenth-century pseudo-scientific ideas of Franz Gall’s cranioscopy, i.e., the appearance (scopos in Greek) of the skull (cranium), are at the origin of modern brain science. Gall professed that the external appearance of the skull mirrors the enclosed brain (1). Importantly, phrenology (phrenos for mind and logos for study) further developed this concept and implied that these external anatomical characteristics are indicative of particular behaviors or personality traits (2). Although even during its day doubts about this approach arose and were considered spurious by some (3), phrenology retained a large following and in some cases was used to justify prejudice and political agendas (4). However, in the second half of the nineteenth century, linking damage of particular brain areas with functional impairments surpassed the scientifically unfounded rhetoric of phrenology. Notably, the seminal studies of Phineas Gage by John Harlow (5) and Tan’s aphasia by Paul Broca (6) ushered in a new era of science that attempted to directly link an anatomical brain region with its contribution to behavior. Ever since these seminal studies, postmortem neuropathology has been the foundation of brain science against which in vivo imaging has been compared. The cellular and molecular compositions of the brain are the gold standard of evidence that damage or aberrations occurred in that region. Although postmortem histopathology can be very informative about the locale of damage or abnormalities, it only provides a means to study conditions after someone has passed away. It is hence of no diagnostic value. In contrast, being able to visualize aspects of pathology in vivo not only allows a more rigorous study of the temporal and spatial progression of these pathologies but potentially also provides a means to diagnose particular conditions and establish a differential diagnosis for an appropriate treatment. At present, both specialties of psychiatry and neurology depend on the study of the brain in vivo to elucidate the underlying causes of behavioral dysfunction. The first step in achieving this in vivo visualization of the brain has been taken at the beginning of the twentieth century with elementary X-ray-based imaging techniques, such as pneumoencephalography (PEG) (7, 8). Table 1.1 provides a time line of milestones in neuroimaging. Although these early techniques provided insights into the living brain, they were also often causing damage to the patient’s brain (e.g., injection of air into the lateral ventricles to provide contrast). Gradually technological developments, such as tomography, where X-rays are rotated around the patient to record (graphens) the signal on single sections (tomos) (9, 10), heralded new innovations that are still in use today. Already in the 1920s, Edgar Moniz used imaging to noninvasively visualize blood vessels in the brain to identify the location of brain tumors (11). Despite these early pioneering advances,

From Molecules to Man: The Dawn of a Vitreous Man

5

Table 1.1 Time line of technological and methodological milestones in (neuro) imaging. MRI milestones are in bold Year

Researcher

Milestone

1895

Roentgen (43)

X-ray image of skull

1910

Bachem and Gunther (44)

First use of contrast media

1916

Dandy (7)

Pneumoencephalography

1924

Hevesy (45)

Radiotracer use in animals

1927

Moniz (11)

Angiography

1931

Vallebona (22)

Stratigraphic imaging

1935

Grossman (9, 10)

Tomographic imaging

1936

Gorter (46)

Paramagnetic relaxation

1938

Rabi (47)

Nuclear magnetic resonance

1942

Bloch (48) and Purcell (49)

Measured NMR signal

1953

Brownell and Sweet (50)

Positron imaging in brain tumors

1956

Kuhl (51)

Recorder for radionuclide scanning

1958

Anger (52)

Scintillation camera

1962

Rankowitz and Robertson (53)

PET transverse section instrument

1963

Kuhl (54)

Emission reconstruction tomography

1965

Harper and Lathrup (55)

Tc-99m radiotracer for brain

1971

Damadian (56)

Hydrogen density in tumors measured by NMR

1973

Hounsfield (13) and Cormack (12)

Computer tomography (CT)

1973

Mansfield (57) and Lauterbur (58)

Magnetic resonance imaging (MRI)

1974

Budinger and Gullberg (59)

SPECT

1974

Hoult (60)

Magnetic (MRS)

1975

Ter-Pogossian (61) and Phelps (62)

Positron emission tomography (PET)

1975

Kuhl (63)

First quantitative cerebral blood volume measurement

1975

Ernst (64)

Phase encoding for MRI

1977

Jaszcak (65)

First head SPECT

1977

Ido and Alavi (67)

FDG-PET

1977

Damadian (66)

First MRI scan of patient

1977

Mansfield (68)

Echo planar imaging (EPI)

1980

Redpath (69)

Spin-warp technique for MRI

1981

Bydder (70)

MR contrast agent

resonance

spectroscopy

1983

Wagner (14)

First neuroreceptor imaging using PET

1984

Weinmann (71)

Gd-DTPA

1986

Nishimura (72)

MR angiography

6

Modo and Bulte

Table 1.1 (continued) Year

Researcher

Milestone

1987

Kornguth (30)

T-cell tracking by MRI

1989

Friston (15)

Statistical parametric mapping

1989

Koretsky (36)

Creatine kinase reporter gene for 1 P NMR

1990

Kalender (73)

Spiral CT

1990

Wolff and Balaban (32)

Magnetization transfer

1990

Weissleder (74)

USPIO

1992

Ogawa (17)

fMRI

1994

LeBihan (23)

Diffusion tensor imaging

1995

Tjujavev (75)

Thymidine kinase reporter gene for PET

1995

Wright (76)

Voxel-based morphometry

1997

Cherry (42)

PET/MRI

1998

Kinahan (77)

PET/CT

1999

Pruessman (78)

Parallel imaging

2000

Ward and Balaban (79)

CEST

2000

Louie (31)

Gene expression by MRI

2003

Dronkers (26)

Voxel lesion-symptom mapping

2005

Gleich and Weizenecker (40)

Magnetic particle imaging (MPI)

it was only in the late 1970s that the advent of computers heralded new technological developments in noninvasive brain imaging that resulted in the spread of scanners beyond a few dedicated academic research centers. The advent of X-ray-based computer tomography (CT) (12, 13) undoubtedly provided the single most important step forward in neuroimaging. CT rapidly became a core assessment tool in neurology during the 1980s and has remained the most commonly used imaging modality in a day-to-day clinical setting. As the importance of clinical neuroimaging grew, the importance of technological developments and its impact on clinical practice and science increased as well. Developments and implementations of novel techniques, such as positron emission tomography (PET) and magnetic resonance imaging (MRI), further increased the diagnostic accuracy. Especially the sensitivity of PET for specific receptors opened up the opportunity to determine particular neurochemical deficits in vivo (14). Sophisticated computational models were required though to generate accurate maps of neuroreceptor binding of radioligands (15). The models could also be used to investigate regional brain activity as


7

18-fluorodeoxyglucose (FDG), a glucose analogue used in PET imaging, is only taken up by metabolically active cells. Functional activity was hence measurable and localizable using FDG-PET. Although this was an important step forward for scientists interested in regional activation during particular behaviors, the low temporal resolution of the technique makes it more suitable as a diagnostic tool for brain tumors. As of today, cost, subject throughput, and the required resources (such as a cycloctron, which is needed to generate 18-FDG) still limit the widespread use of PET imaging. As PET does not provide detailed anatomical information about the exact substructural brain area of interest, additional scans using CT or MRI are often necessary.

2. The Anatomy of Behavior: The Birth of Functional Neuropathology

Developments in nuclear magnetic resonance (NMR) during the 1980s rapidly developed diverse applications ranging from structural and spectroscopic imaging to angiography and mapping of cerebral blood flow. Although CT and PET were initially most commonly employed, since the start of the 1990s MRI has increasingly become the modality of choice to study the brain (Fig. 1.1a). Most importantly in this predominance of MRI is the use of “functional” MRI (Fig. 1.1b). In 1990, Seji Ogawa (16) described the blood oxygen level-dependent (BOLD) contrast that takes advantage of the paramagnetic relaxivity of deoxygenated blood. As areas in the brain consume oxygen during increased activity, the regional concentration of deoxygenated blood increases proportionally, with a concomitant decrease of MRI signal. This basic principle has revolutionized the study of the relationship between brain activity and common behaviors by using functional MRI (fMRI) (17). Although the specific neural correlates of this effect remain under investigation and a matter of a scientific debate (18), the underlying assumption that it is possible to localize some aspects of brain activity using BOLD has mostly remained unchallenged. Concerns have, nevertheless, been raised that regions of activity are often referred to as, for instance, “the region” of happiness (19). This has led some to argue that care must be taken not to commit the same misleading pseudo-scientific interpretations of localized brain activity than those associated with phrenology (20–22). It is important to acknowledge that brain areas are interconnected and multiple regions may be involved in one particular behavior. Regions of activity on fMRI scans merely indicate what area is mostly correlate during a specific paradigm.

8

Modo and Bulte

Fig. 1.1. A chronological comparison of neuroimaging publications. a The three main imaging modalities revolutionized brain imaging during the 1980s. However, only a limited number of publications exist for each year (~20). At the start of the 1990s, during the NIH-proclaimed “decade of the brain,” the advent of neuroimaging exhibited a dramatic tenfold increase. Although the use of PET and CT remained fairly consistent from thereon, there is a linear increase in papers using MRI to study the brain. b By comparing the different MRI techniques, it becomes obvious that the increase in MRI since the mid-1990s is predominantly due to functional MRI (fMRI). In 2009, fMRI accounted for over 60% of all papers using MRI to study the brain. Other techniques, such as structural MRI (sMRI), diffusion tensor imaging (DTI), cellular/molecular MRI, and magnetic resonance spectroscopy (MRS), also steadily increased, albeit at a slower rate than fMRI. In contrast, reports using magnetic resonance angiography (MRA) have not increased over almost 20 years. The diversity of techniques and ease of use of MRI compared to CT and PET increasingly advocate MRI as the modality of choice for scientific and clinical studies.


9

To uncover multiple regions of activity involved in a given behavior (e.g., speech), a systematic modification of task paradigms and their effect on brain activity need to be conducted. Apart of brain activity that is sited in the gray matter of the brain, it is also important to uncover the connections, i.e., white matter fiber tracts, between these sites of activity. Taking advantage of the anisotropic movement of water in white matter, diffusion tensor images can be generated to indicate the presence and direction of fiber tracts (23). This brain hodology (i.e., the study of pathways) allows scientists and clinicians now to go beyond finding one particular area that is involved in a given behavior but afford the construction of brain networks that are interconnected in solving a particular behavioral task. For instance, it is now possible to revisit the early conclusions by Paul Broca to uncover a network of brain areas that are involved in the understanding and generation of speech, with damage to each subcomponent leading to a specific neuropsychological deficit (24). Advances in MRI are no longer just dependent on technological advances in scanner hardware or acquisition, but also increasingly depend on sophisticated image analysis. Early studies merely inspected images to find particular hallmarks that were apparent (e.g., hyperintensity on T2-weighted images of patients with stroke), but ever more subtle differences were revealed by steadily refining these methods to measure regions of interests (ROIs) and to statistically compare individual voxels. For instance, functional MRI often detects signal changes by 2–5% difference using dedicated statistical image analysis. Similar subtle differences can also be found in structural images that are taken over a long time span. For instance, in patients with Alzheimer’s disease, statistical comparisons of serial images (i.e., voxel-based morphometry) can reveal which brain regions shrink or enlarge (25). The degree of change in tissue can be compared using deformation-based morphometry (DBM) that can be further integrated with a subject’s behavioral performance in so-called voxel lesion-symptom mapping (VLSM) (26) analyses to indicate which structural changes are actually associated with a given change in behavior.

3. Molecular and Cellular Imaging: Back to the Future?

Although conventional human brain imaging studies can unravel the regional connectivity and functional contributions to behavior, such investigations do not uncover the molecular and cellular underpinnings of these. Although magnetic resonance spectroscopy (MRS) is frequently used for studying brain tumors, its requirement for acquiring signal from large voxels of interest

10

Modo and Bulte

has limited its application to link metabolic changes to behavior. Nevertheless, in some instances, MRS can provide longitudinal data on the molecular/cellular composition of a region of interest. For instance, neuronal loss or cell death may be assessed using MRS (27). Some groups have even suggested that it can be applied to assess neurogenesis due to specific lipid profiles of neural stem cells (28), although this has been questioned (29). To reduce sampling bias, chemical shift imaging (CSI) is anticipated to provide detailed spatial information about metabolite changes in the living brain. However, MRS and CSI can only detect a limited range of molecules, and consequently additional approaches are required to visualize the true spatial distribution of particular molecules and cells. Analogous to PET and SPECT imaging, for MRI this may be achieved by applying contrast agents that selectively recognize specific (targeted) molecules. To achieve this, contrast agents act by changing the relaxation of nearby protons. One way of targeting an MR contrast agent to depict a molecule of interest is conjugating it to a monoclonal antibody (30). “Smart” contrast agents can be engineered to only change relaxivity in the presence of a particular molecule (31). By applying specific radiofrequency pulses, it is also possible to employ a variety of nonmetallic contrast agents using their chemical exchange saturation transfer (CEST) properties (32). This approach potentially allows “multicolor” imaging (33) that could visualize the presence of more than one targeted molecule within the same region, akin to multiple fluorescent dyes in microscopy. It is hoped that more than one type of cell could be tracked simultaneously in vivo. Tracking of cells using MRI has been of growing interest to determine the infiltration of immune cells (30), the cellular contribution to organogenesis (34), and more recently the localization of cell transplants (35). One drawback of this approach is that contrast agents also produce contrast if cells die. Developments of MRI reporters that only produce contrast if cells are alive can overcome this problem (36), but effects on cellular functions need to be further investigated. One issue with modifying proton contrast is that many pathological measurements depend on this contrast and hence cells that modify this contrast can interfere with the detection of pathology. MRI can also detect other nuclei, for instance, 19 F, and 19 F-MRI has been used to detect cells (37) and the presence of fluorinated compounds (38). However, sensitivity remains a major issue with cellular and molecular MRI, given that only a few molecules or cells of interest may be located within the brain. To overcome this, micron-sized particles of iron oxide (MPIOs) (39) or magnetic particle imaging (MPI) (40) might be alternative approaches to explore. To overcome practical limitations of the different imaging modalities, multimodal imaging is gradually working its way into


11

clinical implementation. For instance, in cancer PET/CT imaging, FDG-PET images of tumor metabolism are superimposed on high-resolution anatomical CT images in order to improve tumor resection (41). As patients only require a single scanning session, information flow can be optimized and provide a substantial benefit as compared to using two independent machines. Similar benefits can be expected from PET/MRI (42), especially in areas such as acute stroke. For instance, PET can provide information regarding penumbral tissue metabolism, whereas anatomical MRI and MR angiography can be used to visualize tissue and blood vessel injury. As pharmacological therapies to limit stroke are expected to have a short time window, obtaining this information quickly within a single scanning session will improve the delivery of acute treatments that can have a maximum impact on the degree of damage caused to the brain. Hence, a wide variety of brain abnormalities can now be studied within a single scanning session. Eventually, this will generate a comprehensive picture of the brain that will increasingly form the basis for a differential diagnosis determining the most appropriate action in treating brain disease.

4. Conclusion: The Dawn of a Vitreous Man

With its humble beginnings in “pseudoscience,” the study of the living brain – which controls our behavior – has rapidly developed into a systematic experimental science, where novel frontiers are mostly driven by technological and methodological innovations. Novel and improved technology protocols will be the fundamental catalyst in the interdisciplinary field of brain imaging. To support the implementation and innovation of these approaches, this book provides a collection of methods and protocols that are state of the art in MR neuroimaging. The last 20 years in MR neuroimaging have already allowed us to visualize in ever-greater detail the structure, function, connectivity, and molecular composition of the brain. Further integration of these methods will eventually allow us to unravel ever-greater mysteries of the brain. Indeed, the dawn of a vitreous man is likely to be upon us soon.

References 1. Gall, F. J. Cranologie, ou decouvertes nouvelles concernant le cerveau, le crâne et les organes. Paris: F: H. Nicolle; 1807. 2. Spurzheim, J. G. Manuel de phrenologie. Paris: F: Porthann; 1832.

3. Magendie, F. An Elementary Treatise on Human Physiology. New York, NY: Harper; 1843. 4. Gould, S. J. The Mismeasure of Man. London: Penguin Books; 1981.

12

Modo and Bulte

5. Harlow, J. M. Passage of an iron rod through the head. Boston Med Surg J 1848;39: 389–393. 6. Broca, P. Perte de la parole, ramolissement chroniqe et destruction partielle du lobe anterieure du cerveau. Bulletin De La Societe D’anthropologie 1861;2:235–238. 7. Dandy, W. E. Ventriculography following injection of air into the cerebral ventricles. Ann Surg 1918;68:5–11. 8. Dandy, W. E. Rontgenography of the brain after injection of air into the spinal canal. Ann Surg 1919;70:397–403. 9. Grossman, G. Tomography I. RöFO – Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren. 1935;51:61–80. 10. Grossman, G. Tomography II. RöFO – Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren. 1935;51:191–209. 11. Moniz, E. L’encephalographie arterielle, son importance dans la localization des tumeurs cerebrales. Rev Neurol 1927;34:72–90. 12. Cormack, A. M. Reconstruction of densities from their projections with applications in radiological physics. Phys Med Biol 1973;18:195–207. 13. Ambrose, J., Hounsfield, G. Computerized transverse axial scanning (tomography). Br J Radiogr 1973;46:1016–1022. 14. Wagner, H. N., Jr., Burns, H. D., Dannals, R. F. et al. Imaging dopamine receptors in the human brain by positron tomography. Science 1983;221:1264–1266. 15. Friston, K. J., Passingham, R. E., Nutt, J. G., Heather, J. D., Sawle, G. V., Frackowiak, R. S. Localisation in PET images: Direct fitting of the intercommissural (ACPC) line. J Cereb Blood Flow Metab 1989;9: 690–695. 16. Ogawa, S., Lee, T. M., Kay, A. R., Tank, D. W. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci USA 1990;7: 9868–9872. 17. Ogawa, S., Tank, D. W., Menon, R. et al. Intrinsic signal changes accompanying sensory stimulation: Functional brain mapping with magnetic resonance imaging. Proc Natl Acad Sci USA 1992;89:5951–5955. 18. Mangia, S., Giove, F., Tkac, I. et al. Metabolic and hemodynamic events after changes in neuronal activity: Current hypotheses, theoretical predictions and in vivo NMR experimental findings. J Cereb Blood Flow Metab 2009;29:441–463. 19. Breiter, H. C., Etcoff, N. L., Whalen, P. J. et al. Response and habituation of the human

20.

21. 22. 23. 24. 25.

26. 27.

28.

29.

30.

31.

32. 33.

amygdala during visual processing of facial expression. Neuron 1996;17:875–887. Vul, E., Harris, C., Winkielman, P., Pashler, H. Puzzling high correlations in fMRI studies of emotion, personality, and social cognition. Perspect Psychol Sci 2009;4:274–290. Kennedy, D. Neuroimaging: Revolutionary research tool or a post-modern phrenology? Am J Bioeth 2005;5:19, discussion W3–4. Raichle, M. E. Modern phrenology: Maps of human cortical function. Ann N Y Acad Sci 1999;882:107–118, discussion 28–34. Basser, P. J., Mattiello, J., LeBihan, D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994;66:259–267. Catani, M., ffytche, D. H. The rises and falls of disconnection syndromes. Brain 2005;128:2224–2239. Good, C. D., Scahill, R. I., Fox, N. C. et al. Automatic differentiation of anatomical patterns in the human brain: Validation with studies of degenerative dementias. Neuroimage 2002;17:29–46. Bates, E., Wilson, S. M., Saygin, A. P. et al. Voxel-based lesion-symptom mapping. Nat Neurosci 2003;6:448–450. Griffith, H. R., Stewart, C. C., den Hollander, J. A. Proton magnetic resonance spectroscopy in dementias and mild cognitive impairment. Int Rev Neurobiol 2009;84:105–131. Manganas, L. N., Zhang, X., Li, Y. et al. Magnetic resonance spectroscopy identifies neural progenitor cells in the live human brain. Science 2007;318:980–985. Jansen, J. F., Gearhart, J. D., Bulte, J. W. Comment on “magnetic resonance spectroscopy identifies neural progenitor cells in the live human brain”. Science 2008;321:640. Kornguth, S. E., Turski, P. A., Perman, W. H. et al. Magnetic resonance imaging of gadolinium-labeled monoclonal antibody polymers directed at human T lymphocytes implanted in canine brain. J Neurosurg 1987;66:898–906. Louie, A. Y., Huber, M. M., Ahrens, E. T. et al. In vivo visualization of gene expression using magnetic resonance imaging. Nat Biotechnol 2000;18:321–325. Wolff, S. D., Balaban, R. S. Nmr imaging of labile proton-exchange. J Magn Reson 1990;86:164–169. McMahon, M. T., Gilad, A. A., DeLiso, M. A., Berman, S. M., Bulte, J. W., van Zijl, P. C. New “multicolor” polypeptide diamagnetic chemical exchange saturation transfer (DIACEST) contrast agents for MRI. Magn Reson Med 2008;60:803–812.

From Molecules to Man: The Dawn of a Vitreous Man 34. Jacobs, R. E., Fraser, S. E. Magnetic resonance microscopy of embryonic cell lineages and movements. Science 1994;263:681–684. 35. Bulte, J. W., Zhang, S., van Gelderen, P. et al. Neurotransplantation of magnetically labeled oligodendrocyte progenitors: Magnetic resonance tracking of cell migration and myelination. Proc Natl Acad Sci USA 1999;96:15256–15261. 36. Koretsky, A. P., Traxler, B. A. The B isozyme of creatine kinase is active as a fusion protein in escherichia coli: In vivo detection by 31p NMR. FEBS Lett 1989;243:8–12. 37. Ratner, A. V., Hurd, R., Muller, H. H. et al. 19f magnetic resonance imaging of the reticuloendothelial system. Magn Reson Med 1987;5:548–554. 38. Burt, C. T., Moore, R. R., Roberts, M. F., Brady, T. J. The fluorinated anesthetic halothane as a potential NMR biologic probe. Biochim Biophys Acta 1984;805:375–381. 39. Hinds, K. A., Hill, J. M., Shapiro, E. M. et al. Highly efficient endosomal labeling of progenitor and stem cells with large magnetic particles allows magnetic resonance imaging of single cells. Blood 2003;102:867–872. 40. Gleich, B., Weizenecker, J. Tomographic imaging using the nonlinear response of magnetic particles. Nature 2005;435:1214–1217. 41. Yang, S., Zhang, C., Zhu, T. et al. Resection of gliomas using positron emission tomography/computed tomography neuronavigation. Neurol Med Chir (Tokyo) 2007;47:397–401, discussion 2. 42. Garlick, P. B., Marsden, P. K., Cave, A. C. et al. PET and NMR dual acquisition (PANDA): Applications to isolated, perfused rat hearts. NMR Biomed 1997;10:138–142. 43. Rontgen, W. C. 1895. Eine neure Art von Strahlen. Sitzungsberichte der Physikalishmedizinischen Gesellschaft Zu Wurzburg 1895. 44. Bachem, C., Gunther, H. Z. Bariumsulfat also schattenbioldendes kontrastmittel bei rontgenuntersuchungen. Zeitschrift Fur Rontgenkunde Und Radiumforschung 1910;12:369–376. 45. Christiansen, I. A., Hevesy, G., Lomholt, S. Chimie physiologique. Recherches, par une methode radiochimique, sur la circulation du bismuth dans l’organism. Compte Rendu De L’academie Des Sciences 1924;178:1324–1326. 46. Gorter, C. J. Paramagnetic relaxation. Physica 1936;3:503–514. 47. Rabi, I. I., Zacharias, J. R., Millman, S., Kusch, P. A new method of measur-

48. 49. 50. 51.

52. 53.

54. 55.

56. 57.

58.

59.

60.

61.

62.

63.

13

ing nuclear magnetic moment. Phys Rev 1938;53:526–535. Bloch, F., Rabi, I. I. Atoms in variable magnetic fields. Rev Mod Phys 1945;17:237–244. Purcell, E. M., Torrey, H. C., Pound, R. V. Resonance absorption by nuclear magnetic moments in a solid. Phys Rev 1946;69:37–38. Brownell, G. L., Sweet, W. H. Localization of brain tumors with positron emitters. Nucleonics 1953;11:40–45. Kuhl, D. E., Chamber, R. H., Hale, J., Gorson, R. O. A high-contrast photographic recorder for scintillation counter scanning. Radiology 1956;66:730–739. Anger, H. O. Scintillation camera. Rev Sci Instrum 1958;29:27–33. Rankowitz, S., Robertson, J. S., Higinbotham, W. A., Rosenblum, M. J. Positron scanner for locating brain tumors. Proc Inst Radio Eng Int Conv Rec 1962;9:49–56. Kuhl, D. E., Edwards, R. Q. Image separation radioisotoe scanning. Radiology 1963;80:653–662. Harper, P. V., Lathrop, K. A., Jiminez, F., Fink, R., Gottschalk, A. Technetium 99m as a scanning agent. Radiology 1965;85:101–109. Damadian, R. Tumor detection by nuclear magnetic resonance. Science 1971;171:1151–1153. Garroway, A., Grannell, P. K., Mansfield, P. Image formation in NMR by a selective irradiative process. J Phys Part C Solid State Phys 1974;7:L457–L462. Lauterbu, P. C. Image formation by induced local interactions – Examples employing nuclear magnetic-resonance. Nature 1973;242:190–191. Budinger, T. F., Gullberg, G. T. Letter: Three-dimensional reconstruction of isotope distributions. Phys Med Biol 1974;19:387–389. Hoult, D. I., Busby, S. J., Gadian, D. G., Radda, G. K., Richards, R. E., Seeley, P. J. Observation of tissue metabolites using 31p nuclear magnetic resonance. Nature 1974;252:285–287. Ter-Pogossian, M. M., Phelps, M. E., Hoffman, E. J., Mullani, N. A. A positron-emission transaxial tomograph for nuclear imaging (PETT). Radiology 1975;114:89–98. Phelps, M. E., Hoffman, E. J., Mullani, N. A., Ter-Pogossian, M. M. Application of annihilation coincidence detection to transaxial reconstruction tomography. J Nucl Med 1975;16:210–224. Kuhl, D. E., Reivich, M., Alavi, A., Nyary, I., Staum, M. M. Local cerebral blood vol-

14

64. 65.

66.

67.

68. 69.

70.

71.

Modo and Bulte ume determined by three-dimensional reconstruction of radionuclide scan data. Circ Res 1975;36:610–619. Kumar, A., Welti, D., Ernst, R. R. NMR fourier zeugmatography. J Magn Reson 1975;18:69–83. Jaszczak, R. J., Murphy, P. H., Huard, D., Burdine, J. A. Radionuclide emission computed tomography of the head with 99mCc and a scintillation camera. J Nucl Med 1977;18:373–380. Damadian, R., Goldsmith, M., Minkoff, L. NMR in cancer: XVI. FONAR image of the live human body. Physiol Chem Phys 1977;9:97–100. Reivich, M., Kuhl, D., Wolf, A. et al. Measurement of local cerebral glucose metabolism in man with 18f-2-fluoro-2deoxy-d-glucose. Acta Neurol Scand Suppl 1977;64:190–191. Mansfield, P., Maudsley, A. A. Medical imaging by NMR. Br J Radiol 1977;50:188–194. Edelstein, W. A., Hutchison, J. M., Johnson, G., Redpath, T. Spin warp NMR imaging and applications to human whole-body imaging. Phys Med Biol 1980;25:751–756. Young, I. R., Clarke, G. J., Bailes, D. R., Pennock, J. M., Doyle, F. H., Bydder, G. M. Enhancement of relaxation rate with paramagnetic contrast agents in NMR imaging. J Comput Tomogr 1981;5:543–547. Weinmann, H. J., Brasch, R. C., Press, W. R., Wesbey, G. E. Characteristics of gadoliniumDTPA complex: A potential NMR contrast agent. AJR Am J Roentgenol 1984;142: 619–624.

72. Nishimura, D. G., Macovski, A., Pauly, J. M. Magnetic resonance angiography. IEEE Trans Med Imaging 1986;5:140–151. 73. Kalender, W. A., Seissler, W., Klotz, E., Vock, P. Spiral volumetric CT with singlebreath-hold technique, continuous transport, and continuous scanner rotation. Radiology 1990;176:181–183. 74. Weissleder, R., Elizondo, G., Wittenberg, J., Rabito, C. A., Bengele, H. H., Josephson, L. Ultrasmall superparamagnetic iron oxide: Characterization of a new class of contrast agents for MR imaging. Radiology 1990;175:489–493. 75. Tjuvajev, J. G., Stockhammer, G., Desai, R. et al. Imaging the expression of transfected genes in vivo. Cancer Res 1995;55:6126–6132. 76. Wright, I. C., McGuire, P. K., Poline, J. B. et al. A voxel-based method for the statistical analysis of gray and white matter density applied to schizophrenia. Neuroimage 1995;2:244–252. 77. Kinahan, P. E., Townsend, D. W., Beyer, T., Sashin, D. Attenuation correction for a combined 3d PET/CT scanner. Med Phys 1998;25:2046–2053. 78. Pruessmann, K. P., Weiger, M., Scheidegger, M. B., Boesiger, P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med 1999;42:952–962. 79. Ward, K. M., Aletras, A. H., Balaban, R. S. A new class of contrast agents for MRI based on proton chemical exchange dependent saturation transfer (CEST). J Magn Reson 2000;143:79–87.

Section II General Techniques

Chapter 2 Magnetic Resonance Safety Andrew Simmons and Kristina Hakansson Abstract The safe operation of both clinical and pre-clinical MR systems is critical. There are a wide range of potential MR hazards. This chapter covers both the theoretical background to issues of MR safety and the guidance on more practical issues. The main sources of information on national and international MR safety guidance and advice are discussed, as well as local safety policies which are required for all MR installations. The projectile effect and other MR safety issues due to static and time-varying magnetic fields are considered, such as peripheral nerve stimulation, tissue heating and RF burns. Finally, contrast agents, auditory effects and medical implants and devices are discussed, as well as the less thought about issue of biological safety of clinical and pre-clinical MR systems. Key words: MR safety, MRI safety, MR safe, projectile effect, quench, SAR.

1. Legislation, Guidance and Best Practice

Legislation, guidance and best practice are all continually evolving within the field of MR safety, and there are often multiple sets of documents applicable within a single country (Table 2.1). This section cannot, therefore, be an exhaustive guide, and the reader is advised to consult local and national experts for the latest information. Good sources of up-to-date information include the International Society of Magnetic Resonance in Medicine and European Society of Magnetic Resonance in Medicine and Biology web sites (www.ismrm.org/mr_sites.htm#Spotlight and www.esmrmb.org) and www.mrisafety.com maintained by Professor Shellock. Regular MR safety updates are given at the ISMRM and ESMRMB annual conferences amongst others.

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_2, © Springer Science+Business Media, LLC 2011

17

18

Simmons and Hakansson

Table 2.1 Main hazards of MR imaging MRI hazards Static magnetic field

• Projectiles • Medical devices • Rotational forces • Lenz effect • Cryogenic liquids (magnet quenches)

Field gradients

• Peripheral nerve and muscle stimulation • Acoustic noise

Radiofrequency pulses

• Thermal heating • Contact burns • Induced current burns

Contrast agents

• Gadolinium side effects

Biological hazards

• Transmission of bacteria and viruses

National and international radiology organisations also provide guidelines and advice. At the time of writing, current international advice on exposure limits for patients and volunteers is given in the ICNIRP (International Commission on NonIonizing Radiation Protection) publications on MR procedures (1–5), as well as the IEC (International Electrotechnical Commission) International Standard 60601-2-33 Edition 2.1, published in 2006 (6). At a national level, the UK Health Protection Agency’s (HPA) published advice on “Protection of Patients and Volunteers Undergoing MRI Procedures” in 2008 (7), for example, is based on ICNIRP’s recommendations and focuses on their application in the UK. MHRA (Medicines and Healthcare products Regulatory Agency) has also produced guidelines based on UK and international MR safety advice (8). In the USA, the American College of Radiology produces guidance documents, the most recent one in 2007 (9). In addition, the FDA (Food and Drug Administration) publication entitled “Criteria for Significant Risk Investigations of Magnetic Resonance Diagnostic Devices” contains advice on exposure limits (10). Within the European Union EU Directive 2004/40/EC (11) contains a set of limits and action values affecting different frequencies of electromagnetic fields which initially stated that all EU member states should bring into force laws and regulations to comply with the directive by 30 April 2008. However, in an amendment to the directive (Directive 2008/46/EC (12)) this date was changed to 30 April 2012.

Magnetic Resonance Safety

19

2. Local Safety Policies All MR installations must have detailed local safety policies which are regularly reviewed and updated to reflect changes in working practices, updated safety guidance and national legislation. The requirements may differ slightly depending on whether the unit has been established for health care, human research or preclinical imaging, but the approaches to safety and the risks are the same. Local safety policies need to be written to cover the work of departmental staff, occasional visitors, cleaners, maintenance and security staff as well as the emergency services. Training should be provided to each group of staff who are likely to need access to the MR scanning unit, tailored to their specific needs. 2.1. Controlled Area

Access to the MR scanning suite is typically defined with respect to controlled areas. Two levels of controlled areas are used to avoid accidents involving pacemakers and projectiles. Both of these areas are three-dimensional in nature, since the magnetic field in the vertical direction needs to be considered to include floors above and below the MRI suite. • The MR controlled area includes all accessible areas where the magnetic field is above 5 G (0.5 mT). All entrances to the controlled area must have warning signs, and access should be restricted to authorised staff and screened patients (and their escorts). Other screened staff and screened visitors may also enter the controlled area if accompanied by an authorised member of staff. Persons with pacemakers or other medical implants must not enter this area. • The inner MR controlled area is defined by the 30-G (3 mT) field contour, which is inside the scanner room. This is used to reduce the risk of projectiles. No ferromagnetic objects should be brought into this area.

2.2. Classification of Persons

The overall responsibility for the safety of all staff and members of the public lies with the employer. In an MRI department, this responsibility is delegated to the responsible person. The responsible person updates the operational and safety policies, ensures adequate training and is responsible for the maintenance of safety facilities. The employer typically also appoints an MR safety adviser, who provides specialist advice on the scientific and technical issues relating to MR safety. Authorised persons are members of staff who have completed an MR safety induction and are authorised to enter the controlled

20


area. The authorised person who is in control of the MR system at a given time is called the MR operator. 2.3. Local Rules

All MR departments should have a written set of local rules, which contain information about controlled areas, contingency plans, patient and equipment management and the names of the department’s responsible and authorised persons. The local rules are issued by the MR responsible person after full consultation with the MR safety advisor and representatives of all MR authorised personnel and should be reviewed and updated at regular intervals.

2.4. MR Safety Classification of Equipment

Items used in or near MR environments can be classified as MR safe, MR unsafe or MR conditional. The symbols and definitions of these terms are shown in Fig. 2.1.

Fig. 2.1. MR safe, MR unsafe and MR conditional markings. The MR environment is defined as the volume within the 5-G field contour.

Marking items with the relevant symbol is an efficient way of reducing the risk of accidents. For example, a defibrillator may be marked as MR unsafe and an aluminium patient trolley with no ferromagnetic parts as MR safe. 2.5. Safety in Practice

Classification of areas, persons and equipment makes it easier to control access to MR areas and prevents accidents from happening. The controlled area for one MR unit is shown in Fig. 2.2. All rooms inside the bold outer line are part of the controlled area.

3. Safety Screening All MR centres must have in place clear policies and procedures for screening anyone who may enter the MR environment.


21

Fig. 2.2. Plan of MRI suite including control room, scanner room and equipment room with field lines.

Figure 2.3 gives an example of a form used to screen patients, research volunteers or staff to be scanned in a whole-body MR scanner. The form includes both questions designed to elucidate any safety queries and several questions designed to investigate the subject’s likely tolerance for the MR scan. In most instances, the form will be explained to the subject who will then fill in the form. An experienced member of staff will then go through the form with the subject, following up on any areas of concern and erring on the side of caution when deciding whether the subject can either enter the controlled areas or be scanned as appropriate.

4. MagnetRelated Safety Issues 4.1. Static Magnetic Field

The strength of the static magnetic field of a scanner is expressed in Tesla (T). One Tesla equals 10,000 G, and 1 G equals 0.1 mT. The earth’s magnetic field is approximately 0.05 mT (0.5 G). There are three principal types of MR magnets in use today. • The most widely used are superconducting magnets which rely on liquid helium to cool the specially constructed coil windings to extremely low temperatures close to absolute zero.

22


Fig. 2.3. Example MR screening form.

• Permanent magnets are similar in concept to bar magnets in that they do not require cooling or an electrical power source to operate. • Resistive electromagnets require a permanent electrical source to operate. The magnetic field will cease once the power is turned off.


23

Typically most MR magnets (superconducting and permanent magnets) are permanently on, even in the event of a power failure. Permanent and resistive magnets are much less common than superconducting magnets. One of the major MR safety issues relating to static magnetic fields is the projectile effect which happens when ferromagnetic materials are attracted by the main magnetic field. Figure 2.2 shows the fringe magnetic field for a whole-body MR system. The magnetic field increases rapidly close to the magnet in a nonlinear manner dependent on the magnet design. Ferromagnetic materials should therefore not be brought into the MR scanner room. Magnetophosphenes are flashes of light experienced by people in high magnetic fields which are thought to represent direct stimulation of the optic nerve and/or retina. They are generated by time-varying magnetic fields caused by movements of the head in a high magnetic field. They are not reported below 2.0 T but are experienced frequently at 4.0 T and above. Another effect experienced in high magnetic fields is the generation of a metallic taste which again has been reported to be generated by movement in a magnetic field. To date, there has been no conclusive evidence for irreversible or hazardous bioeffects of static magnetic fields, so the projectile effect for ferromagnetic materials remains the main safety concern for static magnetic fields. 4.2. Magnet Quenches

Although some low-field MR systems utilise permanent or resistive magnets, most high-field MR systems (1.5 T and above) use superconducting magnets. The magnet windings are cooled using liquid nitrogen or liquid helium in order to reach the low temperatures needed for superconductivity. Over time the amount of cryogens will gradually decrease due to low levels of boil off. If the amount of cryogens drops too low or the magnet begins to heat then an uncontrollable release of freezing gases termed a quench can occur. Modern MR systems should be fitted with an extraction system for these gases via an external piping system. The MRI suite should be fitted with a number of oxygen monitors at critical locations in order to detect any increase in nitrogen or helium caused by a quench. The volume of gas given off by an uncontrolled quench can be extremely large, and there is the potential for a patient to be lying on the scanner bed when a quench starts. Staff must be familiar with the procedures for evacuating a patient from the MR room full of freezing gases and should practise evacuation periodically. It is possible for high pressures to build up in MR rooms making it impossible to open the scanner door, so it may be necessary to break the viewing window to release the pressure. It is important to remember that the only remaining sign of a quench occurring overnight or at weekends may be an oxygen

24


alarm ringing when staff start work the next day, indicating low levels of oxygen in the MR suite. MR scanners are typically provided with an emergency quench button in order to turn off the magnetic field in the case of an emergency. Quenching the magnet in this way should only be considered in extreme cases, such as a person trapped against the side of the magnet by a ferromagnetic item.

5. Time-Varying Gradient Fields Time-varying gradient fields are a key and integral part of MRI. Gradients are turned on and off rapidly, for varying durations and with varying maximum strength, depending on the pulse sequence and scanner under consideration. This has two main effects: peripheral nerve and muscle stimulation and acoustic noise. The induced current density J in a circular loop of radius r placed in a time-varying field B is given by J =σ×

dB r × 2 dt

where σ is the conductivity of the material, in this case the type of tissue carrying the current. Faraday’s law of induction means that a time-varying magnetic field induces a voltage in a conductor and the induced voltage can lead to an induced current. The rate of change of the gradient field is termed dB/dt, and this can vary greatly depending on the scanner, pulse sequence and application. The time-varying gradient fields have negligible thermal effects, but strong time-varying gradient fields could potentially lead to seizures, magnetophosphenes, changes in nerve conduction, peripheral nerve stimulation, cardiac arrhythmias or cardiac arrest. Peripheral nerve stimulation (PNS) and muscle stimulation can occur when a time-varying magnetic field induces electric currents in nerve and muscle cells. Peripheral nerve stimulation occurs at up to 5 kHz. At frequencies of about 10–100 Hz, cardiac muscle stimulation may lead to ventricular fibrillation. The threshold current density for this is about 1.2 A/m2 , so it can be avoided by keeping the current densities below 0.4 A/m2 . The maximum gradient strength has increased substantially over the last two decades with improvements in engineering and applications such as echo planar imaging and diffusion imaging which both rely on strong rapidly switching gradients. The


25

threshold for nerve excitation in the human body varies greatly with the lowest threshold being for retinal neurones, then largediameter peripheral nerves, small-diameter peripheral nerves and finally cardiac muscle. The strongest concern focuses around cardiac excitation in the impaired patient. Clinical scanners are designed with restrictions to ensure that only peripheral nerve excitation, if at all, is possible. With the patient’s nose at isocentre, large-diameter peripheral nerve excitation tends to occur in the lower back, while with the patient’s naval at isocentre, excitation is mostly likely to occur at the shoulder. The y-gradient tends to be most effective in producing excitation, and the shape of the radiofrequency pulses used is important (13).

6. Radiofrequency Effects

The radiofrequency (RF) fields used to manipulate the magnetisation lead to induced currents in the body, which in turn causes power dissipation, i.e. heating. The amount of heating that is acceptable for any particular organ depends on its blood flow, since blood carries the heat away and spreads it through the body. In general, human tissues can tolerate a rise of about 1◦ C. Concern here is most focused on compromised patients and on organs without thermoregulation, such as the eyes, or those that are particularly heat sensitive, such as the reproductive organs. The quantity used to measure RF exposure is the specific absorption rate (SAR), defined as SAR =

σ × E2 2ρ

where σ is the conductivity, E is the induced electric field and ρ is the density of the tissue. The factor of 1/2 comes from averaging over time for an alternating field. SAR is measured in units of W/kg. There are limits on whole-body SAR for different operating modes of the scanner, which are calculated from limits on temperature rises (see Table 2.2). For patients with metallic implants, heating is more of a concern because the metal will absorb more energy (since they have higher conductivity than tissue), and this heat will spread to surrounding tissues. Another heating effect of RF pulses is RF burns. These burns are caused by highly concentrated absorption of RF energy at a single point, resulting in local increases in temperature and (if

26


Table 2.2 SAR values averaged over 6 min Operating mode

Limit on core temperature Whole-body SAR rise (◦ C) limit (W/kg)

Normal

0.7

2

First-level controlled

1

4

Second-level controlled

>1

>4

the temperature is high enough) tissue burning. RF burns occur when there is a conductive loop. The risk can be minimised by avoiding loops of conductor, e.g. keeping ECG leads from forming a loop. Loops in the patient’s body should also be avoided, such as clasped hands. Padding can be used to position the patient correctly.

7. Use-Related Safety Concerns 7.1. Contrast Agents

Contrast agents are often used for clinical MRI imaging and are also used less frequently for some research applications, such as dynamic susceptibility contrast MRI to measure perfusion and contrast-enhanced MR angiography. For example, gadoliniumbased contrast agents are widely used for brain and spine imaging, as well as for contrast-enhanced MR angiography. There are also a variety of organ-specific contrast agents, such as liver contrast agents. Side effects from MR contrast agents are generally low, for example often showing no significant difference between an injection of a gadolinium-based contrast agent and a placebo injection of saline. Side effects in small numbers of patients may include nausea, headache, vomiting and hives. Anaphylactic reactions may occur in 1 in 500,000 subjects, and the MR unit must be prepared for the possibility of this. One major side effect of some gadolinium-based contrast agents is nephrogenic systemic fibrosis (NSF) which can affect patients with kidney disease. This leads to swelling and tightening of the skin with large areas of hardened skin. Glomerular filtration rate should be measured for all patients with kidney disease prior to deciding on the use of gadolinium contrast agents.

7.2. Auditory Effects

Sound is produced by Lorentz forces acting on the MR gradient coils. Noise levels as high as 135 dB have been measured in


27

MR systems. Fast pulse sequences and higher field MR systems in particular can lead to higher levels of noise in MRI systems. Sound levels are a safety issue for both patients and staff. Wearing ear plugs typically reduces noise by 10–20 dB and MRcompatible headphones by more than this. Staff should always wear headphones when in the MR scanner room while the scanner is operating, and patients should be provided with effective ear protection matched to the application in hand. 7.3. Implants and Devices

Some patients or staff will have an implant or device, such as an aneurysm clip, cardiac pacemaker or metal screw, used to fix a broken bone. A particularly good resource is a book by Frank G. Shellock entitled Reference Manual for MR Safety, Implants and Devices which is updated annually, the latest version at the time of writing being the 2009 edition (14). Over 2,300 objects have been tested in the MR environment and are reported in the book, typically at a field strength of 1.5 T. Approximately 900 of these have additionally been tested at 3 T. Manufacturers of implants will often produce many types of an implant, such as an aneurysm clip over time, some of which may be MR unsafe, while others may be MR safe. MR conditional implants may be safe with a particular combination of field strength, maximum spatial gradient and RF coil position but unsafe with other combinations. The static magnetic field can exert a rotational force on nonspherical ferromagnetic objects. Some implanted clips, such as aneurysm clips, can twist inside the patient causing injury or death. The static magnetic field can also interact with implanted medical devices, such as pacemakers and defibrillators. Even at low fields of about 1 mT, the magnetic field can alter the operating mode of some pacemakers.

7.4. Biological Safety

There is often a strong emphasis on MR-specific safety issues in relation to MR scanners and peripheral equipment, such as MRcompatible patient monitors. It is important, however, to ensure that biological safety is also considered for MR installations. Like any other area of a hospital or pre-clinical facility, there is the potential for the transmission of bacteria, viruses and other biological hazards. The scanner bore, control panels, MR door handles and the scanner room must be cleaned regularly according to facility guidelines, using appropriate cleaning techniques and MR safe cleaning equipment.

7.5. Physiological Monitoring

Physiological monitoring is important for three key reasons in MRI. First, respiratory and cardiac/peripheral monitoring is a requirement for some MR applications. Second, physiological monitoring is necessary for some research applications. Lastly, physiological monitoring is required for some patients,

28


particularly those who are impaired in some way or are under general anaesthetic. Specific MR-compatible physiological monitoring equipment may be required for these applications.

References 1. ICNIRP. Statement on medical magnetic resonance (MR) procedures: Protection of patients. Health Phys 2004;87(2): 197–216. 2. Amendment to the ICNIRP. Statement on medical magnetic resonance (MR) procedures: Protection of patients. Health Phys 2009;97(3):259–261. 3. ICNIRP. Guidelines on limits of exposure to static magnetic fields. Health Phys 2009;96(4):504–514. 4. ICNIRP. Guidelines for limiting exposure to time-varying electric, magnetic, and electromagnetic fields (up to 300 GHz). Health Phys 1998;74(4):494–522. 5. ICNIRP. Statement on the “guidelines for limiting exposure to time-varying electric, magnetic and electromagnetic fields (up to 300 GHz)”. Health Phys 2009;97(3): 257–259. 6. IEC 60601-2-33 Consol. ed2.1 (incl. am1): Medical electrical equipment – Part 2: Particular requirements for the safety of magnetic

7. 8. 9. 10.

11. 12. 13. 14.

resonance equipment for medical diagnosis, 2006. HPA Protection of patients and volunteers undergoing MRI procedures: Advice from the Health Protection Agency, 2008. MHRA Device Bulletin: Safety Guidelines for Magnetic Resonance Imaging Equipment in Clinical Use (DB2007(03)), 2007. ACR. Guidance document for MR safe practices. AJR 2007;188:1–27. FDA Guidance for Industry and FDA Staff: Criteria for Significant Risk Investigations of Magnetic Resonance Diagnostic Devices, 2003. EU Directive 2004/40/EC EU Directive 2008/46/EC Abart, J. et al. Peripheral nerve stimulation by time-varying magnetic fields. J Comput Assist Tomogr 1997;21(4):532–538. Shellock, F. G. Reference Manual for Magnetic Resonance Safety, Implants, and Devices 2009. Los Angeles, CA: Biomedical Research Publishing Company; 2009.

Chapter 3 Measuring the Absolute Water Content of the Brain Using Quantitative MRI Nadim Joni Shah, Veronika Ermer, and Ana-Maria Oros-Peusquens Abstract Methods for quantitative imaging of the brain are presented and compared. Highly precise and accurate mapping of the absolute water content and distribution, as presented here, requires a significant number of corrections and also involves mapping of other MR parameters. Here, either T1 and T2 ∗ or T2 is mapped, and several corrections involving the measurement of temperature, transmit and receive B1 inhomogeneities and signal extrapolation to zero TE are applied. Information about the water content of the whole brain can be acquired in clinically acceptable measurement times (10 or 20 min). Since water content is highly regulated in the healthy brain, pathological changes can be easily identified and their evolution or correlation with other manifestations of the disease investigated. In addition to voxelbased total water content, information about the different environments of water can be gleaned from qMRI. The myelin water fraction can be extracted from the fit of very high-SNR multiple-echo T2 decay curves with a superposition of a large number of exponentials. Diseases involving de- or dysmyelination can be investigated and lead to novel observations regarding the water compartmentalisation in tissue, despite the limited spatial coverage. In conclusion, quantitative MRI is emerging as an unparalleled tool for the study of the normal and diseased brain, replacing the customary time–space environment of the sequential mixed-contrast MRI with a multi-NMR-parametric space in which tissue microscopy is increasingly revealed. Key words: Quantitative imaging, water mapping, T1 mapping, T2 ∗ mapping, T2 mapping, brain imaging.

1. Introduction In the last few years, the novel approach of quantitative MRI (qMRI) has gained importance. In comparison to conventional qualitative MRI, qMRI is an unbiased measurement method and M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_3, © Springer Science+Business Media, LLC 2011

29

30

Shah, Ermer, and Oros-Peusquens

allows for more straightforward statistical modelling. Especially for human brain imaging, it provides an attractive method to study changes in the brain caused by disease. Moreover, qMRI facilitates large national or international multi-centre studies since the data can be more directly compared, as well as longitudinal experiments for precise disease evaluation. Degenerative diseases, such as multiple sclerosis (MS), can be more accurately monitored and investigated in order to better understand, for instance, the correlation of lesion burden and physical disabilities, and especially their long-term behaviour. Quantitative measures of relaxation parameters, such as T1 and/or T2 , provide a good basis for the detection of cerebral abnormalities in many diseases. The detection of disease can potentially occur at an earlier stage than allowed by standard imaging methods, as significant confounding effects present in conventional approaches are explicitly corrected in qMRI. A number of methods for mapping of tissue relaxation parameters and water content have been published (1–10). We will address the methods developed in our group in more detail below and present a brief overview of the few other existing methods. Recently, Warntjes et al. (11) presented a novel method for the simultaneous quantification of T1 , T2∗ , proton density and B1 field. Using a multi-echo acquisition of a saturation recovery with a turbo spin-echo readout approach, the authors were able to acquire full brain water content maps with a resolution of 0.8 × 0.8 × 5 mm3 at 1.5 T in 5 min. The published water maps, however, show a rather inhomogeneous distribution reminiscent of B1 effects. Preibisch et al. (12) suggested the use of exponential excitation pulses for a more accurate estimation of S0 in 2D acquisitions, which forms the basis for water content mapping in a similar manner to that presented later in this chapter. The water content of the brain is a very sensitive measure for a variety of pathologies, such as MS, hepatic encephalopathy (HE) or brain tumours. The non-invasive determination of the local or global increase or decrease of brain water content enables an exact differentiation of, for instance, tumour and oedema. This differentiation may avoid a possible lethal side effect for the patient arising from elevated pressure in the skull cavity. The human body consists of up to 90% water distributed either within (intracellular) or outside the cells (extracellular). To actively participate in biological processes, water interacts with biological macromolecules to form a hydration layer (“bound water”), which is involved in processes such as metabolism or biosynthesis, or acts as a solvent and/or transport medium throughout the biological system (“free water”). In the first place, the direct influence of the bound and intracellular water pools on relaxation properties enables the differentiation of tissues by

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

31

NMR/MRI (13), since the free and extracellular water might be expected to have more homogeneous properties across tissue. There are several characteristics of hydration water which play a crucial role in this context, such as restricted motion and partial orientation of water molecules, distribution of correlation times, multiple frequencies of motion, rotation and anisotropic diffusion. These properties of water near the surface of macromolecules result in time-dependent, local magnetic field fluctuations which influence the relaxation mechanisms and therefore provide valuable information about the molecular states and dynamic structure of water at macroscopic and microscopic levels (13). Numerous biological processes, both normal and pathological, can modify the state of water in vivo. Possible reasons for a change in the relaxation times, as well as in the water content, are different phases of cell cycle and cell growth, changes in the physiological pH value, age and maturation or metabolism (13). For instance, it has been found that faster growing cells demonstrate an increased T1 relaxation time. Such fast-growing cells are possible indicators for tumours (13). Physical conditions, such as the state of nutrition, alcohol or drug abuse, or stress, might influence the water balance and therefore also the relaxation times. Three major aspects should be investigated to address the water characteristics in vivo: the total water content, the microscopic (intra/extracellular distribution) and macroscopic (anatomy-related) distribution of water and macromolecular– water interactions. Changing one or more of these factors can alter the tissue-specific water balance in vivo. Abnormal values of water content and distribution, of T1 and T2 , will follow and can be characterised by MRI. The NMR relaxation times are field dependent, and relaxation time-based comparisons between diseased and normal states should use data acquired at the same field strength. The relaxation times are also temperature dependent and can change, e.g. due to fever or hyperthermia treatment. These aspects should be carefully considered in a proper experimental design (7, 13). The MRI-measured water content, on the other hand, is in principle field independent and could be used for comparison of data acquired at different field strengths. However, a careful assessment of possible changes in the MR-visible water compartment with field must be performed and is not available at present. Over the past few years, a fast, precise and accurate method for quantitative mapping of the absolute water content of the human brain in vivo has been developed in Juelich. Accurate information on cerebral water content is of high relevance for many diseases and has been specifically used for the study of hepatic encephalopathy (14).

32

Shah, Ermer, and Oros-Peusquens

The first part of this chapter describes two different methods to quantitatively map the free water compartment in the brain in vivo. Both strategies have been developed at the authors’ Institute and are based on the combination of fast multi-slice and multitime point MR sequences. The common basis of both approaches is the QUTE (quantitative T2∗ image) sequence1 which is suitable for rapidly mapping the T2∗ relaxation time (15, 16) and is used to extrapolate the signal to echo time, TE = 0. The major difference between the two procedures is the way in which the longitudinal relaxation time is mapped. That is, the first approach uses an additional MR sequence, called TAPIR (T1 mapping with partial inversion recovery), to obtain a series of T1 -weighted images and produce T1 maps (1, 2, 17). This method, which is based on the Look–Locker (18) approach with FLASH readout and an innovative interleaving of slices and time points, provides highresolution T1 maps in clinically applicable times. The strength of TAPIR is that it enables the acquisition of multiple time points and thereby enables accurate fitting of the resulting data. In addition, multi-component T1 relaxation can be investigated. The most significant disadvantage of FLASH-based Look–Locker methods, the long acquisition time, can easily be ameliorated through the use of saturation instead of inversion recovery and the acquisition of more than one phase-encoded echo (FLASHEPI readout) for a given slice and inversion time. The use of sliceselective inversion pulses (19) can further speed up the method. The TAPIR-QUTE method for water content mapping has the significant advantage of delivering accurate T1 and T2∗ maps, providing a three-quantitative-parameter description of the brain. An alternative solution to determine the longitudinal relaxation time is to acquire a second set of images with the QUTE sequence with a different TR and/or different flip angle (20, 21). The method can provide a speed-up of the data acquisition but usually involves a decrease of the accuracy of T1 mapping. Further details are described in Section 2.3. The different combinations of QUTE and TAPIR have led to two established methods which are both capable of quantitatively mapping the absolute water content in vivo and result in both accurate and precise maps of the free water content of the brain. The second part of this chapter gives a current literature review of techniques for myelin water mapping and its applications.

1

This sequence has recently become available on Siemens scanners as the multiecho variant of the standard multi-slice gradient echo. Other manufacturers also have similar product sequences.

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

33

An important question related to the mapping of the brain water content with MRI is whether all mobile water protons have been detected. If not, then the degree of mobility and whether other protons have contributed to the signal are also of importance. In addition to the water protons, MR signals can be generated by protons bound in molecules, such as lipids, proteins and nucleic acids – non-aqueous protons. However, this signal decays to zero in less than 100 μs, due to dipolar interactions of the nonaqueous protons with the other nuclei bound in the molecule. In contrast, the signal from water protons in tissue has T2 times of longer than 10 ms (22–24). Conventional MRI therefore detects the signal from brain water with no contamination from the fast decaying non-aqueous tissue. The relaxation properties of water protons are sensitive to the microscopic water environment and thus are expected to vary over dimensions much smaller than the voxel size. It has been recognised over the years that the MRI signal from water in the healthy human brain is generated from three different components: a long T2 component (~2 s), an intermediate component (~80 ms) and a short T2 component (~20 ms) (22, 24, 25). The long component can be easily attributed to CSF and the intermediate one to intra/extracellular water. The short component is thought to be due to myelin water. Myelin is a fatty sheath that surrounds neurons and allows for faster conduction of electrical impulses. It consists of numerous windings (lamellae) with a spacing of the order of 150 Å. According to Brownstein and Tarr (26, 27), the observed T2 of water in a confined compartment can be described as a function of the bulk T2 (T2,bulk ), the radii of the compartment along the x, y and z direction (Rx,y,z ) and the rate of wall relaxation or surface sink strength density (H): 1/T2,obs = H (1/Rx + 1/Ry + 1/Rz ) + 1/T2,bulk. The shortest T2 component can therefore be attributed to water in the most restricted environment. However, observation of the fast relaxing species is hindered by the relatively small contribution this compartment makes to the overall measured signal. For a reliable fit of the fast relaxing component, the short TE (> T1 ), the exp(−TR/T1 ) term is omitted from Eq. [4]. Robust estimation of these three parameters requires fitting the data to more than three signal measurements and, in general, at least 7–8 points along the T1 recovery curve are usually sampled (Fig. 4.9). Example multiple TI inversion recovery brain data is shown in Fig. 4.10, along with corresponding calculated T1 , ρ, and β maps.

Fig. 4.9. Measurement of T1 with an IR experiment. Following an inversion pulse and delay (TI), the longitudinal magnetization is sampled (black circles) and then allowed to fully recover to equilibrium. This process is repeated with varied TI to characterize the recovery curve. Theoretically, the magnetization recovers along the dashed curve. However, in practice, we measure the magnitude of the signal (solid curve). T1 is calculated by fitting the IR signal expressions to the measured data.

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

75

Fig. 4.10. Example IR T1 mapping experiment data comprising images acquired with different inversion times and the corresponding calculated T1 , ρ, and inversion efficiency maps.

Sample Protocol (3.0 T): Axial oriented 25 × 25 cm field of view (FOV); 256 × 128 matrix; 5 mm slice thickness; TE = 10 ms; TR = 6,000 ms; TI = {50, 100, 150, 200, 400, 800, 1,600, 3,200} ms; Flip angle = 90◦ ; Receiver bandwidth (BW) = 488 Hz/voxel; Acquisition time = ~13 min per TI image. 3.1.1.2. T2 Measurement

T2 is customarily measured using a Carr–Purcell–Meiboom–Gill (CPMG) spin-echo sequence (14, 15) consisting of a 90◦ pulse followed by a series of equally spaced 180◦ refocusing pulses (separated by echo time TE) with the magnetization measured at the mid-point between 180◦ pulses. The 180◦ pulses eliminate the effect of macroscopic field inhomogeneities and other non-motion-related sources of T2 dephasing (15). If we consider a pair of neighboring stationary protons with aligned magnetic moments, but each experiencing slightly different magnetic fields due to a macroscopic inhomogeneity (B0 and B0 + B). After time TE/2, the individual magnetic moments will be separated by angle B × TE/2. The 180◦ pulse ‘flips’ the spin system, inverting this angle from B × TE/2

76

Deoni

to – B × TE/2. After a second TE/2 interval, the total angle difference between the two protons is – B × TE/2 + B × TE/2 = 0. This is graphically illustrated in Fig. 4.11 and is also commonly explained using the running race example, where each proton is a runner in a race. At the half-way point (TE/2), the runners turn around and run back to the start line and are once again realigned in spite of the different speeds.

Fig. 4.11. The basics of a spin echo. Following an RF pulse (time = 0), all the proton moments are aligned in phase. Over time, these moments disperse or fan out. At time TE/2 a 180◦ pulse is applied which ‘flips’ the protons and the moments begin to re-phase, forming an echo at time TE.

However, it is important to note that random magnetic field variations caused by proton thermal motion will not be re-phased. Thus, over a series of spin echoes, the magnetization measured at each echo decays according to T2 (Fig. 4.12). To reconstruct the T2 decay curve (Fig. 4.13), 7–8 spinechoes are acquired. T2 is estimated either by a non-linear or linear (after log transforming the data) fit of SCPMG = ρe −TE/T2

[5]

to the data for T2 and ρ. While the linear property of Eq. [5] allows estimation of T2 from as few as two points, within heterogeneous tissues such as brain white and gray matter, two point measures are seldom enough to accurately characterize the relaxation curve and the estimates will be sensitive to the chosen TE values (16). An example spin echo data set is shown in Fig. 4.14, along with corresponding calculated T2 and ρ maps. Sample Protocol (3.0 T): Axial oriented 25 × 25 cm field of view (FOV);


77

Fig. 4.12. Magnetization envelope over a series of spin echoes. During the free induction decay (FID) the signal decays by T2 ∗ . The 180◦ spin echo at time TE/2 causes an echo to form at time TE. However, the overall amplitude of the echoes decreases by T2 .

Fig. 4.13. Measurement of T2 with a CPMG experiment. Following a 90◦ saturation pulse, an echo is formed at time TE, the transverse magnetization sampled, and then allowed to recover back to equilibrium. This process is repeated for different TE to characterize the decay curve and T2 is calculated by fitting the SE signal expressions to the measured data.

256 × 128 matrix; 5 mm slice thickness; TR = 10,000 ms; TE = {10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160} ms; Flip angle = 90◦ ;

78

Deoni

Fig. 4.14. An example of spin echo T2 mapping experiment consisting of eight spin echo images, along with corresponding calculated T2 and ρ maps.

Receiver bandwidth (BW) = 488 Hz/voxel; Acquisition time = ∼21 min per slice. 3.1.1.3. T2 ∗ Measurement

To measure T2 ∗ , the 180◦ (spin-echo) pulses are replaced with a pair of balanced, but opposing magnetic field gradients (creating a gradient echo). Considering our neighboring stationary protons again, if a field gradient is applied across them, then after time TE/2, their individual magnetic moments will be separated by B × TE/2 + δB. Here B is the macroscopic field inhomogeneity and δB is the difference in the field due to the applied gradient. At time TE/2, the applied gradient is reversed, so that over a further TE/2 interval, the angle between the protons is B × TE/2 – δB. Summing these results gives the overall angle at time TE of B × TE. Unlike the spin echo case, where the 180◦ pulse perfectly corrected for B, in the gradient echo case this factor is not corrected and the signal is a function of T2 ∗ rather than T2 . Measurement of T2 ∗ is, therefore, performed in a similar manner as T2 but using a gradient recalled echo sequence (GRE) (Fig. 4.15), with the signal is modeled as ∗

SGRE = ρe −TE/T2 .

[6]


79

Fig. 4.15. Magnetization envelope over a series of gradient echoes. Application of a gradient causes the magnetization to de-phase. Reversing the gradient, re-phases the spins to form an echo. However, macroscopic magnetic field inhomogenieties are not re-phased like in a spin echo.

3.1.2. Equation Model Fitting

Derivation of T1 , T2 , or T2 ∗ estimates from the acquired data requires accurate and precise fitting of the governing signal expressions (Eqs. [4], [5], and [6], respectively). Almost exclusively, this is performed through a least squares minimization approach, in which the sum-of-squares residuals between the model and acquired data are minimized (Fig. 4.16). Accepted methods for performing this minimization differ in their computational complexity, speed, and sensitivity to local minima. For the well-behaved functions described by Eqs. [4], [5], and [6], usual fitting approaches include Powell’s method (17), the simplex approach of Nelder and Mead (18), and gradient descent techniques, such as that of Levenburg and Marquardt (19). Each of these approaches begins from some initial estimation (guess) of T1 , T2 , T2 ∗ , ρ, and β and iterates to a solution.

Fig. 4.16. Least squares minimization curve fitting. a Relaxation time estimates are derived by minimizing the sum of squared residuals between the model and the acquired data. b Fitting algorithms iterate on a solution to provide the best fit of the model to the data.

80

Deoni

While these fitting routines afford rapid convergence times, they can suffer sensitivity to local minima and depend upon the initial estimate. More global search techniques, such as genetic algorithms (20), simulated annealing (21), swarming (22), or region contraction (23), can alleviate these local minima concerns, but at the expense of computation time. 3.1.3. Rapid Techniques for Measuring T1 and T2

Although inversion recovery and CPMG approaches are considered the ‘gold standards’ for T1 and T2 estimation, they suffer lengthy acquisition times making them unsuitable for most clinical applications. Consequently, a number of accelerated techniques have been proposed. Of these, we will briefly describe the more commonly employed Look–Locker and variable flip angle spoiled gradient (DESPOT1) method for measuring T1, the variable flip angle steady-state-free precession (DESPOT2) technique for measuring T2, and the inversion-prepared steady-state-free precession technique for simultaneous measurement of T1 and T2 . Many of the alternative techniques proposed may be considered as variants of these or the IR and CPMG techniques. For example, echo planar imaging (EPI) or spiral readout trains in combination with the IR approach can significantly shorten the acquisition time associated with IR.

3.1.3.1. T1 Measurement with the Look–Locker Method

Originally proposed by Look and Locker in 1970 (24), and later evolving into the TOMROP (T One by Multiple ReadOut Pulses) (25), this technique offers a subtle but important distinction to the conventional IR approach. With IR, the recovering longitudinal magnetization is sampled by tipping it back into the transverse plane (90◦ pulse) and then forming a spin echo. The 90◦ pulse saturates the magnetization, substantially disrupting the longitudinal recovery process and necessitating a lengthy delay to allow the spin system to recover fully. However, if the 90◦ pulse is replaced by a much smaller RF pulse and the spin-echo measurement by a gradient-echo measurement, the longitudinal recovery is only moderately disrupted and can be sampled continuously without requiring full recovery to equilibrium. As a result, only a single inversion pulse is needed, following by a train of small angle α pulses (Fig. 4.17) to measure the T1 relaxation curve. While conceptually simple, this technique can be confounded by corruption of the T1 signal by residual transverse magnetization and stimulated echoes. The small portion of the magnetization that is tipped into the transverse plane by each α pulse decays with rate R2 . If the time between successive α pulses is less than T2 , a portion of this magnetization will still be present when the next α pulse is applied. The magnetization in the transverse plane will, therefore, contain the residual magnetization from the preceding pulse, plus the amount tipped by the current pulse,


81

Fig. 4.17. In the method of Look and Locker, the longitudinal magnetization recovery is sampled with small angle RF pulses following a single inversion pulse. This eliminates the need to apply multiple inversion pulses and wait for the magnetization to fully recovery, as required by the IR approach.

corrupting our measurement. It is essential, therefore, that any magnetization in the transverse plane must be eliminated between each pulse. This can either be accomplished by spacing the α pulses more than 5 × T2 apart (i.e., long enough that less than 1% of the magnetization remains) or by spoiling the transverse magnetization. By applying a magnetic field gradient across the voxel, we can de-phase the protons, spreading their individual magnetic moments about 360◦ (or some other increment of PI) and eliminating the residual coherent transverse magnetization within the voxel (Fig. 4.18). Such a process is termed gradient spoiling and is a common feature of most rapid T1 -weighted imaging sequences, including SPGR and spoiled FLASH. In the above description, we assumed that the magnetization recovery is not disturbed by the train of small angle α pulses. In practice, however, even a very small α pulse (less than 5◦ ) will sufficiently disturb the longitudinal magnetization recovery depending on the T1 of the sample. The effect of this continued perturbation is to drive the recovery to equilibrium via an effective T1 , T1 ∗ , related to T1 and equal to T1∗ =

T1 . 1 − T1 TRln(cos α)

[7]

In practical application, a T1 map image is created by fitting for T1 ∗ from the acquired data and then converting to T1 . Despite offering substantive time savings over the conventional IR approach, for large three-dimensional volumes the

82

Deoni

Fig. 4.18. Gradient spoiling of the transverse magnetization. Applying a gradient across each voxel causes the proton moments to de-phase, eliminating, or ‘spoiling’ the transverse magnetization.

Look–Locker technique can still require several hours since the inversion pulse – α pulse train is repeated for each phase-encode step. Accelerated three-dimensional variants have, therefore, been presented requiring only minutes for reasonable spatial resolution (26, 27), or EPI readouts are used to reduce the number of phase encodes (28). 3.1.3.2. The Method of Variable Flip Angles: Driven Equilibrium Single Pulse Observation of T1 (DESPOT)

The principal rate-limiting factors in inversion recovery are (1) the inversion pulse and subsequent inversion time delay, which increases for each measurement; (2) the 90–180 spin-echo measurement, which perturbs the longitudinal magnetization recovery; and (3) the lengthy delay required to allow the spin system to return to equilibrium before repeating the process. The Look– Locker approach addressed the latter two points, replacing the spin-echo measurement with a train of small tip angle pulses and gradient-echo measurements and eliminating the length recovery time. What if the inversion pulse and delay were also eliminated, leaving only a train of small angle, swiftly spaced, RF pulses? Curiously, if the transverse magnetization is adequately spoiled between pulses, this rapid application of α pulses drives the longitudinal magnetization to a dynamic equilibrium that is governed by the sample ρ, T1 , T2 ∗ , and the magnitude of the RF pulse. The described pulse sequence, referred to as SPGR (SPoiled Gradient Recalled echo) or spoiled FLASH (Fast Low Angle Single sHot), is a commonly used clinical sequence on account of its rapid acquisition time and high T1 -weighted signal, described by the expression SSPGR = ρ

1 − e −TR/T1 sin α

1 − e −TR/T1

cos α

∗

e −TE/T2 .

[8]


83

Provided TE is kept constant, or short relative to T2 ∗ ; TR kept constant; and ρ remains unchanged, SSPGR becomes a function of only T1 and α. By measuring SSPGR over a range of flip angles, an SSPGR vs. α curve like that shown in Fig. 4.19 can be obtained (the peak of which occurs at the Ernst angle, αERNST = α cos e −TR/T1

[9]

Fig. 4.19. The SPGR signal curve. T1 can be calculated either by fitting the signal function to SPGR data acquired over a range of flip angles or by calculating where the peak of the signal curve occurs (Ernst angle).

The use of this signal curve to calculate T1 was first described by Christensen in 1974 in the context of NMR spectroscopy (29). Using the Ernst angle definition, Christensen showed that T1 could be calculated directly if the peak of the SSPGR curve is known. This concept was further refined by Homer and Beevers (30), who reasoned that only a single measure of SSPGR was necessary to calculate T1 (provided it was collected at the Ernst angle), and termed the method DESPOT – Driven Equilibrium Single Pulse Observation of T1 . However, this is a bit of a misnomer. To calculate T1 from a single point requires a priori knowledge of the Ernst angle, which is only known if T1 is also known a priori. In practice, a number of signal measurements are required in order to determine the curve peak and, subsequently, T1 . If several measures of SSPGR are acquired, however, it is more appropriate to fit Eq. [7] for T1 and ρe−TE/T 2 ∗ (31, 32) by exploiting the linearization property of the signal model. Rewriting the expression in the linear Y = slope × X + intercept form as

84

Deoni

SSPGR SSPGR ∗ = e −TR/T1 + ρe −TE/T2 tan α sin α

[10]

allows T1 to be calculated directly from the slope of the line. Further, since calculation of the slope of a line requires only two points, T1 may be calculated from just two SSPGR measurements. Despite the use of more than one point, this multi-point approach is still commonly referred to as DESPOT. Since the original description of this approach, a number of refinements and optimizations have been presented (33–35), making DESPOT one of the most common accelerated T1 measurement methods in use. 3.1.3.3. Driven Equilibrium Single Pulse Observation of T2 (DESPOT2)

In both Look–Locker and DESPOT techniques, magnetization spoiling was required to eliminate T2 effects from the T1 measurements. Without spoiling, the freely evolving signal depends upon ρ, T1 , T2 , T2 ∗ , RF pulse number, and flip angle in a complex manner. This freely evolving signal can, however, be modified by recycling (or refocusing) the transverse magnetization between RF pulses. In a typical gradient echo, the transverse magnetization is dephased, re-phased, and de-phased over the course of the applied gradient (Fig. 4.20). To recycle the magnetization, an additional re-phasing gradient lobe is added. The described sequence, termed SSFP (Steady-State Free Precession), balanced SSFP, or true-FISP (Fast Imaging with Steady Precession), is one of the oldest sequences in NMR (dating back to the 1950s (36)) and was the basis of the one of the very first NMR imaging methods (37). Under ideal conditions, the resultant SSFP signal is well described by the expression SSSFP

1 − e−TR/T1 e−TE/T2 sin α

. =ρ 1 − e−TR/T1 e−TR/T2 + e−TR/T1 − e−TR/T2 cos α [11]

If TR is kept constant and ρ remains unchanged; SSFP becomes a function of T1 , T2 , and flip angle. If T1 is known a priori, SSFP may be used to calculate T2 in an analogous fashion as to how T1 is calculated using SPGR. Similar to Eq. [7], the SSFP signal expression can be rewritten in the Y = slope × X + intercept form as −TR/T

1 − e−TR/T2 S e ρ 1 − e−TR/T1 e−TE/T2 SSSFP SSFP = + tan α 1 − e−TR/T1 e−TR/T2 sin α 1 − e−TR/T1 e−TR/T2 [12] with T2 calculated from the slope of the line. This linear property allows T2 to be calculated from as few as two SSFP measurements.


85

Fig. 4.20. Gradient refocusing in SSFP. Balancing the gradients around the echo refocuses the transverse magnetization just before the next RF pulse.

The similarity of this technique with the DESPOT approach led to its being dubbed DESPOT2 – Driven Equilibrium Single Pulse Observation of T2 . The combination of DESPOT and DESPOT2 allows combined calculation of T1 and T2 from as few as four rapidly acquired images. 3.1.3.4. Simultaneous T1 and T2 Measurement with IR-SSFP

Given the mixed contribution of T1 and T2 to the SSFP signal, is it necessary to calculate T1 separately to T2 ? Indeed, cannot both be calculated directly from the SSFP signal? Scheffler and Hennig (38) first demonstrated the ability to quantitatively measure T1 using an inversion prepared (IR-) SSFP sequence. The inversion pulse was added to increase the T1 dependence of the signal (97). This technique is similar to the Look–Locker approach described above, but with an SSFP readout used in place of the spoiled FLASH readout, and like the Look–Locker signal, the longitudinal magnetization in IR-SSFP is driven to equilibrium via an effective T1 , which depends on the flip angle and sample T1 and T2 , as T1∗

=

1 1 2 α 2 α cos /2 + sin /2 . T1 T2

[13]

86

Deoni

Performing a more thorough examination of the IR-SSPP signal, Schmitt et al. (40) used the effective T1 recovery rate, with combined T1 and T2 influence, to simultaneously measure both relaxation times. The rapid acquisition times associated with this technique make it a useful alternative to the combined DESPOT and DESPOT2 techniques. 3.1.4. Errors in T1 and T2 Measurements

Though straightforward in concept, accurate and precise measurement of T1 and T2 using the described techniques requires careful consideration of the potential sources of error. Over the following section, we outline some of the more egregious pitfalls and describe techniques for avoiding or correcting for them.

3.1.4.1. Flip Angle Inhomogeneity

The accelerated measurement techniques described above rely on small tip angle RF pulses to sample the magnetization and incorporate these flip angles values into the T1 calculation itself. Thus, accurate knowledge of the applied flip angle is essential to correct T1 and T2 estimations. Deviations of the transmitted flip angle from the intended or ‘prescribed’ value arise from two main sources: RF pulse profile errors and RF attenuation and tissue dielectric effects. Ideally, the excitation profile of the RF pulse has a square shape, providing the desired flip angle value across the excited volume or slice and zero elsewhere (Fig. 4.21). Unfortunately, this is rarely the case. Instead, the pulse profile is more apt to be Gaussian shaped with the transmitted flip angle varying across the volume or slice. For three-dimensional (3D) volumetric acquisitions, this profile affect can be tolerated if the anatomy of interest is within the center portion of the excited slab, where the flip angle is approximately uniform and of the desired value. For single and multiple 2D slice applications, this profile effect will yield variation through the image slice, with the measured signal becoming an integrated function of flip angle (41).

Fig. 4.21. Imperfections in the RF pulse profile can lead to significant variations in signal across the image volume.

RF coil inhomogeneities and RF attenuation and dielectric resonance effects can also produce significant deviations from the intended flip angle throughout the image volume. Asymmetric RF coils have non-uniform RF power profiles that cause the transmitted flip angle to vary with distance from the coil. Dielectric


87

resonance, or RF penetration, effects are perhaps best understood through consideration of the RF wavelength relative to object size. When the wavelength is large relative to the object, the transmitted power is approximately equal throughout the volume. However, when the RF wavelength and object size are of the same order, the RF power varies throughout the object, resulting in variations in the transmitted flip angle. These effects become particularly pronounced at high magnetic field strengths where the RF wavelength decreases. Minimization of RF pulse profile and attenuation and dielectric effects can be achieved through improved RF pulse design (42), use of B1 insensitive pulses (43, 44), numerical modeling (41), or calibration of the transmitted flip angle (45). Through optimized RF pulse design, such as SLR pulses (42), the RF pulse shape can be made as box like as possible, minimizing the fall-off regions at the edges of the slab. In single-slice applications, were pulse design is much less of a factor, Parker et al. (41) showed how the flip angle could be numerically modeled and accounted for in T1 measurements. Composite or fast passage adiabatic pulses, which are less sensitive to RF attenuation and dielectric effects, are perhaps the most obvious approach to eliminating flip angle errors. However, these pulses suffer lengthy pulse durations (potentially doubling imaging time) and can have high energy deposition (limiting their use at high field strengths or in pediatric populations). Quantitative measurement, or calibration, of the transmitted flip angle has received increased attention with the move to higher magnetic field strengths and has benefited from a proliferation of rapid techniques. The most common of these techniques, the double-angle approach (45) acquires two spin-echo images with flip angles α and 2α and through a trigonometric relationship determines α through the ratio of the signal intensities. Rapid volumetric approaches, such as (46–49), also provide robust calculation of the flip angle field and are fast enough to be used as a calibration step before a T1 or T2 experiment. A less common approach improving flip angle spatial uniformity is parallel transmission (50). Here the RF power of the independent elements of modern multi-element RF coils is tuned and optimized to produce an overall uniform pulse profile throughout the object. This approach has the added advantage of reducing the overall RF power deposition, an important consideration at higher magnetic field strengths or in certain clinical populations. 3.1.4.2. Residual or Incoherent Transverse Magnetization

When the time between successive RF pulses (TR) is greater than 5 × T2 and T2 ∗ , there is little coherence among the proton spins in the transverse plane (i.e., the transverse magnetization is naturally spoiled or de-phased). However, in accelerated T1 measurement methods, TR is generally much less than T2 and T2 ∗ . Thus,

88

Deoni

there is residual transverse magnetization, which left unchecked, will introduce unwanted T2 weighting into the signal and result in significant errors in the T1 estimates (51). To eliminate this residual transverse magnetization, a combination of RF and gradient spoiling is necessary and is an essential feature of SPGR and spoiled FLASH sequences. Though spoiling is not used in SSFP, the related phenomenon of incoherent transverse magnetization is a significant confound. The simplified balanced SSFP signal model provided by Freeman (52) (Eq. [10]) assumes the transverse magnetization is perfectly recycled at the end of the TR interval. Unexpected precession of the magnetization, caused by unbalanced gradient errors or, more commonly, susceptibility-induced off-resonance, results in deviations of the signal from this theoretical value. In areas of susceptibility-induced field gradients, a well-known banding artifact can appear (Fig. 4.21), corrupting T1 and T2 estimates made with IR-SSFP or DESPOT2. A common approach to deal with this artifact, presented by Zur et al., is RF phase cycling (53). An RF pulse acts to rotate the magnetic moment around an axis. A 90◦ pulse, for example, rotates the longitudinal magnetization around the y-axis, yielding a magnetic moment oriented along the x-axis (Fig. 4.22a). While we may assume that each RF pulse in an imaging sequence is identical, rotating the magnetization around the same axis of rotation, this is generally not the case. Rather, in practice the axis of rotation is incremented around the XY-plane (Fig. 4.22b) by either a constant or a variable angle and the angle between the axis of rotation and the x-axis is termed the phase ( ) of the RF pulse. An RF pulse that rotates the magnetization around the y-axis, for example, has 0◦ phase, while an RF pulse that rotates the magnetization around the x-axis has a phase of 90◦ . In the SSFP sequence, each RF pulse is applied with a constant increment added to the RF phase (such as 180◦ ). Changing this increment has an unexpected effect on the acquired image, shifting the spatial location of the SSFP banding artifact in the image. An artifact-free SSFP image can, therefore, be obtained by combining two images acquired with different RF phases, usually 0 and 180◦ (Fig. 4.23). Deoni et al. (54) has shown how this technique can also be used to calculate artifact-free T2 maps using DESPOT2. 3.1.4.3. Ensuring Steady State

Most accelerated T1 and T2 relaxation measurement techniques utilize steady-state imaging sequences, where the magnetization is driven to a dynamic equilibrium that depends on the time between RF pulses, flip angle, and T1 and T2 . Care must be taken to ensure this dynamic steady-state condition is established before data acquisition, a process that can take several seconds (generally at least a time equal to T1 ). In practice, this condition


89

Fig. 4.22. Rotation of the magnetization by an RF pulse around an arbitrary axis of rotation. Conventionally, we assume a single constant axis of rotation (such as the Y-axis) that each RF pulse tilts the magnetization around (a). In practice, the axis of rotation is incremented around the XY-plane on each pulse (b) by some phase angle, . In SSFP, the RF pulse is commonly incremented by = 180◦ , so that the tipped magnetization alternates between pointing in the +X and –X directions (c).

Fig. 4.23. Illustration of RF pulse phase cycling in SSFP. Incrementing the phase of each RF pulse shifts the spatial location of signal bands (yellow arrows). The maximum intensity projection of two SSFP images acquired with different phase-cycling patterns (phase angles) can produce an artifact-free image (right panel).

90

Deoni

is achieved by playing out a series of ‘dummy’ pulses, without signal acquisition, before imaging. Though this process may add several seconds to the acquisition, it is essential to correct T1 or T2 estimation. 3.1.4.4. Movement and Flow

Beyond bulk motion artifacts (i.e., ghosting, blurring), which can have devastating effects on image appearance and the derived T1 and T2 measures, physiological motion can produce more subtle effects. Introduced in the above section, most accelerated techniques require the establishment of a steady state. In the case of T1 , thorough spoiling of the transverse magnetization is also necessary. For moving tissues, such as flowing blood, these conditions may be violated. Depending on flow rate and the extent of the excited image volume, flowing blood may have exited the volume before it has reached steady state. Further, the movement of blood protons through the applied spoiling gradients may result in incomplete spoiling or worse, refocused magnetization. In either case, the estimated T1 values will be erroneous. In large volume three-dimensional applications, this effect may be subtle, as the flowing magnetization will evolve into steady state as it navigates the image volume, leaving only subtle artifact near the edge of the volume where flow initiates. Application of a large spoiling gradient can further assist in eliminating unwanted residual transverse magnetization, even in the presence of rapid flow. In single and multiple slice 2D applications, this artifact can be far more insidious, requiring the use of saturation bands, which null the flowing signal immediately outside the slice of interest. It is also unlikely flowing blood will achieve adequate steady state before exiting the slice, making 2D approaches ill suited to quantifying blood T1 or T2 .

3.2. Exchange and Multiple Component Relaxation

Throughout this chapter, we have assumed the relaxation of signal in each image voxel is characterized by a single T1 and T2 . This is analogous to assuming only a single water environment within each image voxel. In tissue, water is compartmentalized into multiple distinct micro-anatomical environments, each with unique biophysical and biochemical properties and, therefore, distinct T1 and T2 characteristics. Further, if the boundaries between these compartments are permeable to water, protons may easily exchange between them. When averaged over the spatial dimensions of a voxel, this exchange and differential relaxation result a complex signal decay. ill-described by a single T1 or T2 value. Such values are more correctly termed effective relaxation times and represent a weighted average of the compartmental T1 and T2 values and exchange rate (16). The distinct relaxation properties of each compartment, however, provide a potential means for discriminating and


91

isolating the individual signal. Within brain and spinal cord, multicomponent analysis of T2 and, more recently, T1 and T2 ∗ relaxation data provides evidence of at least two distinct and reproducible water environments (55–59). These are broadly attributed to (1) the intra- and extra-cellular water and (2) water trapped between the hydrophobic lipid bilayers of the myelin sheath (55). Quantification of this latter component provides a non-invasive means of measuring and monitoring myelin content and has been applied to the study of known and suspected de- and dys-myelinating white matter disorders (such as multiple sclerosis (60) and schizophrenia (61)). An important consideration in multicomponent analysis is the temporal timescales of relaxation and the magnetization exchange rate between water compartments. Zimmerman and Britten coined the concept of exchange regime (62) to describe this phenomenon. In the fast exchange regime, exchange is rapid relative to the relaxation timescale. Due to the quick mixing of the environments, an averaged, mono-exponential relaxation curve is produced, masking the information from each compartment. In contrast, in the slow exchange regime where the exchange time is slow relative to the T1 and T2 of the individual components, a multi-exponential relaxation curve is observed and the individual compartments can be interrogated. It is conventionally assumed that in brain tissue, exchange is fast relative to T1 , but slow relative to T2 and T2 ∗ . 3.2.1. Measurement of Multicomponent Relaxation

The gold standard approach to measuring multicomponent relaxation remains the multi-echo T2 approach championed by McKay, Whittall, and colleagues (58). Assuming the proton and magnetization exchange time between water compartments is slow relative to T2 , the spin-echo signal from a multicomponent system is a general expansion of Eq. [5]: S=ρ

N

fi e−TE/T2 ,

[14]

i=1

where fi and T2,i are the volume fraction and T2 is the relaxation time of the ith water component. To derive estimates of each components volume fraction and T2 , a Poon–Henckelman CPMG sequence (63) is used to sample upward of 32 uniformly spaced echo times between 10 and 3,200 ms with TR kept long to mitigate T1 effects. Discrete two, three, or N (i.e., continuous distribution) component models are fit to this data to derive the desired volume fraction and T2 values (Fig. 4.24). In the most commonly presented implementation, single-slice data can be acquired with scan times on the order of 16 min. Current implementations, however, can provide 8–12 contiguous

92

Deoni

Fig. 4.24. Multicomponent relaxation theory and practice. A simple model of brain tissue contains two water components: free intra- and extra-cellular water (blue) and water trapped between the lipid bilayers of the myelin sheath (green). The measured MR signal contains contributions from each of these water pools. MCR aims to reconstruct these individual contributions and quantify the volume of the myelin water pool (right).

2D slices in a similar time-frame. More recently, Du et al. (64) have proposed the alternative use of a multi-echo gradient echo sequence to derive similar information based on component T2 ∗ differences. Finally, building on the multicomponent T1 work of Kreis et al. (65) and Spencer and Fishbein (66), Deoni et al. (67) have recently presented a combined multicomponent relaxation technique that models both T1 and T2 effects, taking into account proton exchange between the water compartments. 3.3. Analysis of Relaxation Time Data

The qualitative nature of conventional T1 -and T2 -weighted images makes it difficult to perform direct quantitative comparisons between data acquired of different subject groups (i.e., patient and healthy control groups); longitudinally acquired data (for example, of a single subject group when establishing treatment efficacy); or across multiple imaging centers (in order to obtain sufficient data to study a rare disorder or phenotype).


93

Differences in scanner hardware, scanner drift, software upgrades, pulse sequence implementation differences, coil geometry, and placement differences all but make it impossible to directly compare qualitative T1 -and T2 -weighted data. A common approach to side-step this obstacle is to derive quantitative metrics from the inherently qualitative data and to perform comparative studies using these derivative metrics. The usual approach, referred to as voxel-based morphometry (68), is to segment the weighted image into maps of white matter and gray matter percentages (also termed white and gray matter density maps) to investigate changes in gray or white matter volume with pathology. Derivation and comparison of cortical thickness, through cortical gray matter segmentation, are also increasingly used (69). However, while these approaches may highlight regions of difference, they provide little information regarding the underlying basis of identified changes. Further, results are inherently sensitive to tissue contrast, which depends nonlinearly upon ρ, T1 , and T2 (and other non-tissue affects). For example, an increase in white matter T1 along a gray/white matter boundary could present as a corresponding increase in white matter density, despite there being no actual change in the amount of white matter. Alternatively, opposing changes in ρ and T1 (as seen in edema or inflammation) mask each other in a conventional T1 -weighted scan and may not present as a change in white matter density, despite there being a very real tissue change. An alternative to these indirect assessments of tissue structure, volume and density are direct comparison of the T1 and T2 relaxation times. Depending on the spatial extent and resolution of the acquired maps, comparisons can be at the whole brain, hemispheric, regional, white matter tract, or voxel-wise levels. In addition to group comparisons, a powerful attribute of relaxation data is the ability to perform single subject comparisons against population norms without requiring correction for scanner hardware, acquisition strategy, etc. (70). 3.3.1. Group-Wise Comparisons

The overwhelming majority of clinical and research structural neuroimaging studies involve normal vs. pathological comparisons to determine (1) if there is a difference in brain structure associated with the condition; (2) where in the brain those differences are manifested; and (3) how identified differences correlate with degree or severity of pathology. To address these basic questions, a number of analysis approaches have been developed and refined. Here we cover the more commonly used approaches, describing them in order to increasing spatial specificity.

3.3.2. Histogram-Based Comparisons

Histogram-based analysis offers a straightforward means of addressing the most basic question: is there a difference between

94

Deoni

normal and pathological tissue? (71) The method is advantageous in that it requires no spatial normalization (alignment of images from each participant) and it does not require an a priori hypothesis as to where changes might be expected. Indeed, no spatial information is provided. Further, histograms provide an intuitive and tangible medium for visualizing group differences. Calculation of T1 or T2 histograms, simply the frequency of binned values, is straightforward. Correction for brain volume is accomplished by normalizing the bin frequencies by the total number of voxels included in the histogram (or area under the histogram curve). Averaged patient and control histograms can be visually and statistically compared, with standard metrics of comparison including mean, median, mode, skewness, kurtosis, peak height, peak location, among others. An example of histogrambased comparison of white matter T1 and T2 in healthy adolescents and adolescents with autism is shown in Fig. 4.25.

Fig. 4.25. Histogram-based comparison of whole-brain white matter T1 (left) and T2 (right) in healthy young adults and those with autism. The T1 histogram reveals a global increase in T1 in autism.

The lack of any spatial information, however, is the primary disadvantage of histogram-based analysis. Further, the approach may be ill posed in cases where subtle, small scale, or regional differences are expected. Subtle or focal differences may unfortunately be obscured when included with whole-brain data. 3.3.3. Voxel-Based Comparisons

If the primary disadvantage of histogram analysis is the lack of spatial information, region of interest analysis provides an effective alternative. Such analysis may take the form of either using pre-defined regions, or more generally, to treat each voxel as an independent region of interest and to perform voxel-wise comparisons (68). Following linear or nonlinear spatial normalization of patient and control image data (68, 72), voxel-wise t-tests (or a nonparameter equivalent) with appropriate correction for multiple comparisons (58, 73) can identify regions of group difference. The ability to examine the whole brain, without requiring consideration of a priori hypotheses, offers tremendous potential for


95

speculative or exploratory studies, where affected regions may not be known before hand, or when the spatial progression of the pathology is the topic of study. An example of voxel-based comparison of T1 and T2 in schizotypy is shown in Fig. 4.26, demonstrating hemispheric differences in the relaxation times in normal vs. high schizotypy groups.

Fig. 4.26. Voxel-based T1 comparison of medium and high schizotypy patients. Voxels showing a significant difference following multiple comparison correction are shown in the far right panel (1–p-value map).

Potential pitfalls of voxel-based comparisons, however, include mis-registration or poor spatial alignment, too little or too much spatial smoothing, and inadequate correction for multiple comparisons (or a lack thereof). Each of these effects combines to make reproducibility across different research groups challenging [Ref Derek]. 3.3.4. Tract-Based Comparisons

A classic theme among neuroimaging researchers is the use of a connectionist approach to understand neurological or psychiatric disorders (74). This approach implies consideration of the white matter tracts which connect the disparate brain regions that comprise the integrated brain systems and networks. While voxel-based approaches can elucidate regions of differences, these regions may contain multiple independent white matter pathways connecting different gray matter regions. Thus, additional information may be gleaned by considering the T1 and T2 characteristics along specific tracts of interest (75). Two common approaches for isolating specific white matter pathways are the use of digitized atlases (76, 77) or the combined acquisition of relaxation and diffusion tensor imaging (DTI) data (78). As will be discussed later, diffusion imaging provides estimates of local fiber orientation (78). By stitching these independent orientation measures together (tractography), three-dimensional representations of the white matter paths may be reconstructed (79). Using either atlas or tractography data to supply regions of interest, T1 and T2 values within these regions can be isolated

96

Deoni

Fig. 4.27. Tract-specific comparison of T1 and T2 in the right and left superior longitudinal fasciculi (shown as the green volume rendering superimposed on the anatomical images) in healthy young adults and those with autism, respectively. Histograms of values along these tracts demonstrate substantive alteration in T1 and T2 .

and statistical comparisons made. As an example, Fig. 4.27 shows comparison of T1 and T2 histogram data for the left and right superior longitudinal fasiculi in patients with autism and healthy age, sex, and IQ-matched controls. 3.3.5. Comparisons with Population Norms

In a variety of disorders, group-wise comparisons are difficult or ill suited. For example, multiple sclerosis is characterized by acute focal white and gray matter lesions occurring throughout the brain and spinal cord (80). From a disease monitoring, prognosis or predictive perspective, it is not necessarily the lesions themselves that are of interest, but the surrounding ‘normal appearing’ white matter (81). Voxel-wise comparisons across groups are not appropriate given the near-random location of lesions. While histogram-based approaches are useful, they are void of the spatial information necessary to pin-point affected ‘normal appearing’ brain regions. Here, then, it is preferable to perform subjectspecific analysis, comparing each subject with a matched population normal (or average) template to identify affected areas. Similar to group-wise comparisons, a population template (comprising both the mean and the variance) can be determined by spatially normalizing healthy participant data (matched for age, sex, handedness, etc. as required) and calculating the voxel-wise mean and standard deviation. Identification of voxels or regions that differ substantively from the population norm can be determined through straightforward voxel-wise z-tests. Illustrated in Fig. 4.28 is a comparison of T1 data from an MS patient against a matched population average, demonstrating how this form of analysis can reveal affected white matter regions that appear normal on conventional clinical T1 - and T2 -weighted scans.


97

Fig. 4.28. Single subject T1 analysis of an MS (relapsing remitting, EDSS score of 4) patient. The population average and variance were calculated from 18 healthy age-matched controls. Z-score analysis reveals significant distribution in cerebral white matter that appears normal in the patient’s clinical FLAIR image.

4. Applications of T1 and T2 Quantification

4.1. Improving Tissue Contrast

The fundamental relationship linking tissue biochemistry and micro-structure with T1 and T2 underpins the use of T1 and T2 to identify, investigate, diagnose, and monitor pathology. However, while T1 and T2 are exquisitely sensitive to tissue alteration, they are notoriously non-specific, requiring careful interpretation. Over the following section, a summary of prominent relaxation time measurement applications in neuroimaging is provided, highlighting some of the more robust neurological and psychiatric findings. A central goal in neuroimaging research is improved visualization of subtle structure. This goal can be achieved through a combination of increased spatial resolution and enhanced tissue contrast. Increases in spatial resolution have been made possible through the proliferation of high field strength imaging systems (i.e., 3 T), coupled with parallel reception techniques (82). However, the relationship between T1 , field strength, and T1 -weighted image contrast often means a trade-off between spatial resolution and contrast (Fig. 4.29). As T1 increases, T1 -weighted contrast decreases since the signal is related to exp(−TR/T1 ). Acquisition of T1 maps can circumvent this issue, providing near optimum contrast that increases with field strength, since the T1 difference between tissues also increases with field strength. Figure 4.30 shows a comparison of T1 -weighted and T1 maps acquired at 1.5 and 3 T using similar acquisition strategies. Variations in T1 , T2 , and ρ should not be considered independently as these parameters are inherently inter-related. However, these relationships can vary on a tissue-by-tissue and pathologyspecific basis. For example, the formation of the myelin sheath in white matter causes not only a decrease in T1 but also the displacement of free water decreases ρ. Associated changes in relaxation

98

Deoni

Fig. 4.29. Comparison of 1.5 and 3 T T1 -weighted and T1 map images acquired with similar imaging parameters. While image contrast is slightly worse in the 3 T-weighted image, the increased gray–white matter T1 difference provides improved contrast in the map image, relative to 1.5 T.

Fig. 4.30. Comparison of a T1 -weighted, T1 , and ρ map of the cerebellum. The increased T1 and ρ of the dentate nucleus (purple arrows) relative to the surrounding cerebellar white matter obscures the structure in the weighted image. However, the structure is easily visualized on the map images.

times and proton density can reduce T1 - and T2 -weighted image contrast and obscure tissue differences and boundaries. To illustrate this, Fig. 4.30 contains a high SNR T1 -weighted image of the cerebellum alongside corresponding T1 and ρ maps. While the deep cerebellar dentate nucleus is clearly visible on the map images (83), it cannot be distinguished from the surrounding cerebellar white matter in the weighted image. This masking is due to the prolonged T1 and increased proton density of the dentate


99

nucleus relative to the surrounding white matter. These associated changes effectively ‘cancel’ each other and the measured T1 weighted signal is identical to that of the white matter. In addition to the cerebellum, T1 maps show improved tissue contrast throughout the brain, revealing subtle myelo and cyto-architectural differences unseen on conventional T1 - and T2 weighted scans, as illustrated in Fig. 4.31.

Fig. 4.31. High spatial resolution T1 map images of different brain regions, demonstrating visualization of subtle myelo and cyto-architecture. Identified structures include the internal and external globus pallidus (GPi, GPe), anterior ventral lateral nucleus (AVL), ventral anterior nucleus (VA), ventral lateral nucleus (VL), medial dorsal nucleus (MD), sub-thalamic nucleus, substantia nigra (SN), line of Gennari, and cortical striations of the hippocampus.

4.2. Neurological Disorders 4.2.1. Multiple Sclerosis

Multiple sclerosis (MS) is a neurodegenerative and neuroinflammatory disorder characterized by focal white and gray matter lesions in the brain and spinal cord. The lipid myelin sheath within and surrounding these lesions has been damaged, reduced, or lost (80, 81, 84). Though gray matter lesions are also present, their role in the disorder is less well established, and until the recent proliferation of high-field (i.e., 7 T) scanners, have been detectable only through histological and histochemical approaches (86). The destruction of the myelin sheath, and replacement by inflammatory cells, free water, and other proteins, leads to substantive changes in T1 and T2 . Lesion areas commonly present hypo-intensely on T1 or hyper-intense on T2 -weighted scans (80). Based on this association between pathology and relaxation times, T1 and T2 have been proposed as surrogate markers of disease activity for monitoring and therapeutic trial outcome purposes.

100

Deoni

Though the principal presentations of MS on MR images are focal lesions, correlations between lesion number and extent with disease severity (EDSS) and activity have, unfortunately, been week (86). Recent interest has, therefore, turned to investigating the surrounding ‘normal appearing’ white matter (NAWM) as well as diffusely abnormal ‘dirty appearing’ white matter (DAWM). Investigations of NAWM and DAWM have revealed global alterations in T1 (81) and T2 (87) suggestive of wide-scale tissue disruption and have again suggested T1 and T2 may be suitable biomarkers. As MS is inherently a disease of myelin, direct investigation of myelin content is desirable (Fig. 4.32). The use of MCR, therefore, has played a significant albeit, to date, research-only role in characterizing myelin damage in MS (88). In addition to expected focal reductions in myelin water fraction within lesions, MCR has revealed significant alterations within NAWM (89) and shown potential to discriminate between acute, chronic and active, and chronic and inactive lesion subtypes (90).

Fig. 4.32. Example MWF image in MS compared with a conventional clinical T2 -weighted FLAIR image.

4.2.2. Stroke

Diffusion-weighted imaging is the de facto standard in clinical stroke imaging, providing clear visualization of stroke extent. However, the diffusion signal is unable to disambiguate salvageable from unsalvageable tissue, an important clinical distinction. The significant alteration of tissue biochemistry and microstructure immediately following stroke results in substantial increases in both T1 and T2 (primarily stemming from the influx of free water into the region). The magnitude of T1 and T2 prolongation has been shown to differentiate infarcted regions from the surrounding affected tissue (91) and that T1 is superior to either T2 or diffusion in detecting early ischemic change.


101

The time course of relaxation time changes is also informative of underlying tissue pathology and of diagnostic utility. The early and rapid T1 and T2 increases are followed by gradual declines to a non-normative equilibrium 2–3 days later (92). Measurement of these early changes has been demonstrated to accurately predict final infarct area (93). 4.2.3. Epilepsy

Atrophy of the hippocampus (hippocampal sclerosis, HS) is the most common cause of temporal lobe epilepsy. Though HS is commonly associated with an increased T2 -weighted signal, due to a prolongation of the T2 relaxation time, the ambiguous nature of T2 -weighted signal changes make definitive diagnosis challenging (94). As an adjunct to conventional spin-echo acquisitions, quantitative estimates of T2 in normal and pathologic hippocampal tissue are an effective method for the detection and monitoring of hippocampal structural changes (95). Evidence of T2 differentiation and prolongation in epilepsy is shown in Fig. 4.33.

Fig. 4.33. Comparison of hippocampal T2 maps from healthy (top) and epileptic (bottom) individuals showing the prolonged T2 of the HS patient.

In addition to T2 alterations, temporal lobe T1 values have also been shown to be significantly longer in epileptic patients compared with healthy controls (96). Further, T1 values throughout the hemisphere containing the seizure focus were also prolonged, intimating a potential role for quantitative T1 mapping in conjunction with EEG for identifying seizure foci. 4.3. Psychiatric Disorders: Dementia and Alzheimer’s Disease

T1 and T2 measurements have been studied in most dementia subtypes, including vascular dementia, dementia with Lewy bodies, and Alzheimer’s disease (AD). The classic neuropathological hallmark of AD is the presence of iron-containing beta amyloid plaque deposits. The iron composition of these plaques has been

102

Deoni

suggested as a means of permitting their visualization; however, to date, direct plaque visualization has only been possible in animal models or in vitro specimens (97) or at ultrahigh field strengths (i.e., 7 T). Measurement of more diffuse changes in T2 within the hippocampus, and basal ganglia, caused by aggregate accumulation of plaques, however, may reduce the need for direct plaque visualization (98). In addition to plaque deposits, white matter hyper-intensities are also commonly observed in AD (99). While the exact mechanisms behind these white matter changes remain unknown (though they may be associated with vascular changes), a recent disease model forwarded by Bartzokis (100) suggests white matter and demyelination may play an underlying role in this traditionally gray matter-centric disorder. Indirect support for this myelin hypothesis has been the observation of T2 increases (presumably due to decreased myelin and increased free water content) throughout the white matter of subjects reporting memory loss and confirmed AD patients (101). 4.4. Neurodevelopment

Abnormal brain maturation and neurodevelopment is a hypothesized substrate in a number of neurological and psychiatric disorders, including autism (102). Neurodevelopment is marked by a wide range of biophysical and biochemical changes, including axonal sprouting, synaptic pruning, and the establishment of the lipid myelin layer around axons (myelination). The influx of lipids, proteins, and other macromolecules associated with these processes can be indirectly monitored through T1 and T2 measurements. The influence of myelin precursory proteins and the sheath itself on T1 has been well established (7), and acquisition of T1 maps throughout neurodevelopment can provide a non-invasive window into the dynamic myelination process. Reduction of T1 throughout the white matter tracts broadly mirrors the histologically established time course of myelination. Alongside conventional (single-component) T1 measurement, multicomponent relaxometry can provide a more direct and quantitative assessment of myelination throughout neurodevelopment. Figure 4.34 shows an example of myelin water fraction maps obtained from six healthy infants from 3 through 8

Fig. 4.34. Myelin development throughout infancy. MWF maps acquired throughout the first 8 months of infancy were high pass filtered at the level of 3.5. A surface was fit to the remaining voxels.


103

months of age. Figure 4.34 provides a powerful yet intuitive view of white matter development during infancy and corresponds to spatiotemporal pattern derived from prior histological studies. While alternative imaging methods, including diffusion tensor and magnetization transfer imaging, have been proposed as surrogate markers of myelination, these techniques provide information that is only related to myelin content and not specific to it.

5. Toward a More Complete Picture of Tissue: Combining Relaxation Data with Other Forms of Imaging Information

The development of non-invasive imaging, including computed tomography (CT), positron emission tomography (PET), and MRI, has made possible the significant gains in our understanding of brain development, function, and the pathology that affects them. Within the field of MRI, different imaging and acquisition techniques provide complementary information that, when fused, can provide a more holistic view of tissue microstructure. Diffusion tensor imaging provides voxel-wise information regarding local fiber orientation, axonal density, and axonal size. T1 and T2 provide indirect measures of myelin and free water content, as well as lipid, protein, macromolecule, and paramagnetic material concentrations. Multi-component relaxometry provides a direct assessment of myelin content. Additional techniques, such as magnetic resonance spectroscopy, can provide further information related to metabolism, function, and integrity. Cumulatively, these data are informative of the principal facets of brain tissue microstructure, integrity, and function. An imaging protocol combining these elements would be well positioned to address crucial questions related to tissue alteration in pathology. The continued gains in imaging technology, including ever increasing magnetic field strengths, continued reductions in acquisition time through improvements in gradient technology and parallel transmission and reception, and the development of novel imaging pulse sequences, have brought the realization of such a multi-parametric, quantitative imaging protocol ever closer. Based on current state of the art, a combined whole-brain diffusion tensor/quantitative multi-component relaxometry protocol requires approx. 30 min. Single-slice chemical shift spectroscopic imaging can be completed in a further 10 min. Many functional MRI paradigms require less than 5 min with good statistical power. Thus, a single 45 min protocol could provide a new mechanism for investigating structure–function relationships and associated alterations in pathology. An example of a combined relaxometry/diffusion study is shown in Fig. 4.35. Though technically more challenging than conventional T1 - or T2 -weighted imaging, quantitative relaxation time

104

Deoni

Fig. 4.35. Combination of diffusion tensor tractography and relaxation time measurement. The colormap along the tracts shows the corresponding T1 and T2 values.

measurement affords a number of advantages. As outlined throughout this chapter, diagnostic observations in weighted images are non-linear and complex functions of the underlying relaxation times, acquisition strategy, and scanner hardware. Relaxation time measurement cleanly separates these individual contributors, providing a standardized basis for comparison. This not only provides an ideal basis for large, multi-center, and longitudinal research studies but also has clinical utility in diagnosis and treatment monitoring. Comparisons with population-based norms, as shown in Fig. 4.28, may be a crucial next step in bridging the gap between research studies and clinical adoption, allowing not only disease progression or treatment response to be monitored and quantified on an individual basis but also providing a more sensitive diagnostic tool.

Acknowledgments A special thanks is extended to those who provided clinical examples and imaging data used herein: Prof. Derek Jones, Dr. Shannon Kolind, Dr. Janneke Zinkstok, Dr. Marco Catani, Dr. Emma Burkus, Dr. Mark Richardson, Katrina McMullin, Catherine Traynor, and Sarah Kwan. A debt of gratitude is owed to Dr. David Lythgoe, Dr. Fernando Zelya, Astrid Pauls, and Katrina McMullen for proof reading and comments. References 1. Blinkov, S. M., Glezer, I. I. The Human Brain in Figures and Tables. New York, NY: A Quantitative Handbook. Plenum Press; 1968. 2. Damadian, R. V. Tumor detection by nuclear magnetic resonance. Science 1971;171:1151. 3. Rabi, I. I., Zacharias, J. R., Millman, S., Kusch, P. A new method of measuring

the nuclear magnetic moment. Phys Rev 1938;53:318. 4. Bloembergen, N., Purcell, E. M., Pound, R. V. Relaxation effects in nuclear magnetic resonance absorption. Phys Rev 1948;73:679–715. 5. Stanisz, G. J., Odrobina, E. E., Pun, J., Escaravage, M., Graham, S. J. et al. T1,T2


6.

7.

8.

9.

10.

11.

12.

13.

14.

15. 16.

relaxation and magnetization transfer in tissue at 3t. Magn Reson Med 2005;54: 507–512. Bottomley, P. A., Hardy, C. J., Argersinger, R. E., Allen-Moore, G. A review of 1 H nuclear magnetic resonance relaxation in pathology: Are T1 and T2 diagnostic? Med Phys 1987;14:1–37. Paus, T., Collins, D. L., Evans, A. C., Leonard, G., Pike, B., Zijdenbos, A. Maturation of white matter in the human brain: A review of magnetic resonance studies. Brain Res Bull 2001;54:255–266. Gelman, N., Ewing, J. R., Gorell, J. M., Spickler, E. M., Solomon, E. G. Interregional variation of longitudinal relaxation rates in human brain at 3.0t: Relation to estimated iron and water contents. Magn Reson Med 2001;45:71–79. Gelman, N., Gorell, J. M., Barker, P. B., Savage, R. M., Spickler, E. M. et al. MR imaging of the human brain at 3.0t: Preliminary report on transverse relaxation rates and relation to estimated iron content. Radiology 1999;210:759–767. Naoko, S., Sakai, O., Ozonoff, A., Jara, H. Relaxo-volumetric multispectral quantitative magnetic resonance imaging of the brain over the human lifespan: Global and regional aging patterns. Mag Reson Imaging 2009;27:895–906. Williams, L. A., Gelman, N., Picot, P. A., Lee, D. S., Ewing, J. R. et al. Neonatal brain: regional variability of in vivo MR imaging relaxation rates at 3.0t – initial experience. Radiology 2005;235:595–603. Hoque, R., Ledbetter, C., Gonzalez-Toledo, E., Misra, V., Menon, V. et al. The role of quantitative neuroimaging indices in the differentiation of ischemia from demyelination: An analytical study with case presentation. Int Rev Neurobiol 2007;79:491–519. Li, Y., Srinivasan, R., Ratiney, H., Lu, Y., Chang, S. M., Nelson, S. J. Comparison of T1 and T2 metabolite relaxation times in glioma and normal brain at 3t. J Magn Reson Imaging 2008;28:342–350. Carr, H. Y., Purcell, E. M. Effects of diffusion on free precession in nuclear magnetic resonance experiments. Phys Rev 1954;94: 630–638. Meiboom, S., Gill, D. Modified spin-echo method for measuring nuclear relaxation times. Rev Sci Instrum 1958;29:688–691. Whittall, K. P., MacKay, A. L., Li, D. K. B. Are mono-exponential fits to a few echoes sufficient to determine T2 relaxation for in vivo human brain? Magn Reson Med 1999;43:1255–1257.

105

17. Powell, M. J. D. A tolerant algorithm for linearly constrained optimization calculations. Math Programming 1989;45:547–566. 18. Nelder, J. A., Mead, R. A simplex method for function minimization. Comput J 1965;7:308–313. 19. Levenberg, K. A method for the solution of certain non-linear problems in least squares. Appl Math 1944;2:164–168. 20. Holland, J. H. Adaptation in Natural and Artificial Systems. Ann Arbor, MI: The University of Michigan Press; 1975. 21. Kirkpatrick, S., Gelatt, C. D., Vecchi, M. P. Optimization by simulated annealing. Science 1983;220:671–680. 22. Clerc, M., Kennedy, J. The particle swarmexplosion, stability, and convergence in a multidimensional complex space. IEEE Trans Evol Comput 2002;6:58–73. 23. Berger, M. F., Silverman, H. F. Microphone array optimization by stochastic region contraction. IEEE Trans Signal Proc 1991;38:2377–2386. 24. Look, D. C., Locker, D. R. Time saving in measurement of NMR and EPR relaxation times. Rev Sci Instrum 1970;41: 250–251. 25. Brix, G., Schad, L. R., Deimling, M., Lorenz, W. J. Fast and precise T1 imaging using a TOMROP sequence. Magn Reson Imaging 1990;8:351–356. 26. Henderson, E., McKinnon, G., Lee, T. Y., Rutt, B. K. A fast 3d look-locker method for volumetric T1 mapping. Magn Reson Imaging 1999;17:1163–1171. 27. Gai, N. D., Butman, J. A. Modulated repetition time look-locker (MORTLL): A method for rapid high resolution threedimensional T1 mapping. J Magn Reson Imaging 2009;30:640–648. 28. Freeman, A. J., Gowland, P. A., Mansfield, P. Optimization of the ultrafast look-locker echo-planar imaging T1 mapping sequence. Magn Reson Med 1998;16:765–772. 29. Christensen, K. A., Grand, D. M., Schulman, E. M., Walling, C. Optimal determination of relaxation times of Fourier transform nuclear magnetic resonance. Determination of spin-lattive relaxation times in chemically polarized species. J Phys Chem 1974;78: 1971–1977. 30. Homer, J., Beevers, M. S. A re-evaluation of a rapid ‘new’ method for determining NMR spin-lattice relaxation times. J Magn Reson 1985;63:287–297. 31. Wang, H. Z., Riederer, S. J., Lee, J. N. Optimizing the precision in T1 relaxation estimation using limited flip angles. Magn Reson Med 1987;5:399–416.

106

Deoni

32. Homer, J., Roberts, J. K. Routine evaluation of mo ratios and T1 values from driven equilibrium NMR specta. J Magn Reson 1990;87:265–272. 33. Deoni, S. C. L., Rutt, B. K., Peters, T. M. Rapid combined T1 and T2 mapping using gradient recalled acquisition in the steady state. Magn Reson Med 2003;49: 515–526. 34. Deoni, S. C. L., Peters, T. M., Rutt, B. K. Determination of optimal angles for variable nutation proton magnetic spin-lattice, T1, and spin-spin, T2, relaxation times measurement. Magn Reson Med 2004;51: 194–199. 35. Chang, L. C., Koay, C. G., Basser, P. J., Pierpaoli, C. Linear least-squares method for unbiased estimation of T1 from SPGR signals. Magn Reson Med 2008;60:496–501. 36. Carr, H. Y. Steady-state free precession in nuclear magnetic resonance. Phys Rev 1958;112:1693–1701. 37. Hinshaw, W. S. Spin mapping: The application of moving gradients to NMR. Phys Lett A 1974;48:87–88. 38. Scheffler, K., Hennig, J. T1 quantification with inversion recovery truefisp. Magn Reson Med 2001;45:720–723. 39. Deimling, M., Heid, O. (1994). Magnetization Prepared True FISP Imaging. In: Proceedings of the 2nd annual meeting of the ISMRM. San Fransico, p. 495. 40. Schmitt, P., Griswold, M. A., Jakob, P. M., Kotas, M., Gulani, V. et al. Inversion recovery truefisp: Quantification of T1, T2 and spin density. Magn Reson Med 2004;51: 661–667. 41. Parker, G. J., Barker, G. J., Tofts, P. S. Accurate multislice gradient echo T1 measurement in the presence of non-ideal RF pulse shape and RF field nonuniformity. Magn Reson Med 2001;45:838–845. 42. Pauly, P., Le Roux, P., Nishimura, D., Macovski, A. Parameter relations for the shinnar-le roux selective excitation pulse design algorithm. IEEE Trans Med Imaging 1991;10:53–65. 43. Madhuranthakam, A. J., Busse, R. F., Brittain, J. H., Rofsky, N. M., Alsop, D. C. B1-insensitive fast spin-echo using adiabatic square wave enabling of the echo train (SWEET) excitation. Magn Reson Med 2008;59:1386–1393. 44. Garwood, M., Ugurbil, K., Rath, A. R., Bendall, M. R., Ross, B. D., Mitchell, S. L., Merkle, H. Magnetic resonance imaging with adiabatic pulses using a single surface coil for RF transmission and signal detection. Magn Reson Med 1989;9:25–34.

45. Insko, E. K., Bolinger, L. Mapping of the radiofrequency field. J Magn Reson Ser A 1993;103:82–85. 46. Jiru, F., Klose, U. Fast 3D radiofrequency field mapping using echo-planar imaging. Magn Reson Med 2006;56:1375–1379. 47. Morrell, G. R. A phase-sensitive method of flip angle mapping. Magn Reson Med 2008;60:889–894. 48. Cunningham, C. H., Pauly, J. M., Nayak, K. S. Saturated double-angle method for rapid B1+ mapping. Magn Reson Med 2006;55:1326–1333. 49. Deoni, S. C. L. High resolution T1 mapping of the brain at 3T with driven equilibrium single pulse observation of T1 with highspeed incorporation of RF field inhomogeneities (DESPOT1-HIFI). J Magn Reson Imaging 2007;26:1106–1111. 50. Katscher, U., Bornert, P. Parallel RF transmission in MRI. Nmr Biomed 2006;19:393–400. 51. Preibisch, C., Deichmann, R. Influence of RF spoiling on the stability and accuracy of T1 mapping based on spoiled FLASH with varying flip angles. Magn Reson Med 2009;61:125–135. 52. Freeman, R., Hill, H. D. W. Phase and intensity anomalies in Fourier transform NMR. J Magn Reson 1971;4:366–383. 53. Zur, Y., Wood, M. L., Neuringer, L. J. Motion-insensitive steady-state free precession imaging. Magn Reson Med 1990;16:444–459. 54. Deoni, S. C. L. Transverse relaxation time (T2) mapping in the brain with off-resonance correction using phase-cycled steady-state free precession imaging. J Magn Reson Imaging 2009;30:411–417. 55. MacKay, A., Laule, C., Vavasour, I., Bjarnason, T., Kolind, S., Madler, B. Insights into brain microstructure from the T2 distribution. Magn Reson Imaging 2006;24:515–525. 56. Kroeker, R. M., Henkelman, R. M. Analysis of biological NMR relaxation data with continuous distributions of relaxation times. J Magn Reson 1986;69:218–235. 57. Menon, R. S., Rusinko, M. S., Allen, P. S. Multiexponential proton relaxation in model cellular systems. Magn Reson Med 1991;20:196–213. 58. Whittal, K. P., MacKay, A. L., Graeb, D. A., Nugent, R. A., Li, D. K., Paty, D. W. In vivo measurement of T2 distributions and water contents in normal human brain. Magn Reson Med 1997;37:34–43. 59. Cheng, K. H. In vivo tissue characterization of human brain by chisquares


60.

61.

62.

63. 64.

65.

66.

67.

68. 69.

70.

parameter maps: Multiparameter proton T2relaxation analysis. Magn Reson Imaging 1994;12:1099–1109. Laule, C., Leung, E., Lis, D. K., Traboulsee, A. L., Paty, D. W. et al. Myelin water imaging in multiple sclerosis: Quantitative comparison with histopathology. Mult Scler 2006;12:747–753. Flynn, S. W., Lang, D. J., MacKay, A. L., Goghari, V., Vavasoir, I. M. et al. Abnormalities of myelination in schizophrenia detected in vivo with MRI, and post-mortem with analysis of oligodendrocyte proteins. Mol Psychiatry 2003;8:811–820. Zimmerman, J. R., Britten, W. E. Nuclear magnetic resonance studies in multiple phase systems: Lifetime of a water molecule in an adsorbing phase on a silica gel. J Phys Chem 1957;61:1328–1333. Poon, C. S., Henkelman, R. M. Practical T2 quantitation for clinical applications. J Magn Reson Imaging 1992;2:541–553. Du, Y. P., Chu, R., Hwang, D., Brown, M. S., Kleinschmidt-DeMasters, B. K. et al. Fast multislice mapping of the myelin water fraction using multicomponent analysis of T2∗ decay at 3t: A preliminary postmortem study. Magn Reson Med 2007;58:865–870. Kreis, R., Fusch, C., Boesch, C. (1992). In vivo characterization of three water compartments in human white matter using a single voxel technique with short TE. In: Proceedings of the 11th Annual Meeting of SMRM, Berlin, Germany, p. 1963. Spencer, R. G. S., Fishbein, K. W. Measurement of spin-lattice relaxation times and concentrations in systems with chemical exchange using the one-pulse sequence: Breakdown of the Ernst model for partial saturation in nuclear magnetic resonance spectroscopy. J Magn Reson 2000;142:120–135. Deoni, S. C. L., Rutt, B. K., Arun, T., Pierpaoli, C., Jones, D. K. Gleaning multicomponent T1 and T2 information from steady-state imaging data. Magn Reson Med 2008;60:1372–1387. Ashburner, J., Friston, K. J. Voxel-based morphometry – the methods. Neuroimage 2000;11:805–821. Tosun, D., Rettmann, M. E., Han, X., Tao, X., Xu, C. et al. Cortical surface segmentation and mapping. Neuroimage 2004;23(S1):S108–S118. Deoni, S. C. L., Williams, S. C. R., Jezzard, P., Suckling, J., Murphy, D. G., Jones, D. K. Standardized structural magnetic resonance imaging in multicentre studies using quantitative T1 and T2 imaging at 1.5T. Neuroimage 2008;40:662–671.

107

71. van Buchem, M. A., McGowan, J. C., Grossman, R. I. Magnetization transfer histogram methodology: Its clinical and neuropsychological correlates. Neurology 1999;53: S23–S28. 72. Jenkinson, M., Bannister, P. R., Brady, J. M. Improved optimization for the robust and fast linear registration and motion correction of brain images. Neuroimage 2002;17: 825–841. 73. Voormolem, E. H., Wei, C., Chow, E. W., Bassett, A. S., Mikulis, D. J., Crawley, A. P. Voxel-based morphometry and automated lobar volumetry: The trade-off between spatial scale and statistical correction. Neuroimage 2009;18 (ahead of print). 74. Catani, M., ffytche, D. H. The rises and falls of disconnection syndromes. Brain 2005;128:2224–2239. 75. Kanaan, R. A., Shergill, S. S., Barker, G. K., Catani, M., Ng, V. W. et al. Tract-specific anisotropy measurements in diffusion tensor imaging. Psychiatry Res 2006;146:73–82. 76. More, S., Oishi, K., Jiang, L., Li, X., Akhter, K. et al. Stereotactic white matter atlas based on diffusion tensor imaging in an ICBM template. Neuroimage 2008;40:570–582. 77. Catani, M., Thiebaut de Schotten, M. A diffusion tensor imaging tractography atlas for virtual in vivo dissections. Cortex 2009;44:1105–1132. 78. Basser, P. J. Inferring microstructural features and the physiological state of tissues from diffusion-weighted images. Nmr Biomed 1995;8:333–344. 79. Jones, D. K., Simmons, A., Williams, S. C., Horsfield, M. A. Non-invasive assessment of axonal fiber connectivity in the human brain via diffusion tensor MRI. Magn Reson Med 1999;42:37–41. 80. Bakshi, R., Thompson, A. J., Rocca, M. A., Pelletier, D., Dousset, V. et al. MRI in multiple sclerosis: Current status and future prospects. Lancet Neurol 2008;7:615–625. 81. Manfredonia, F., Ciccarelli, O., Khaleeli, Z., Tozer, D. J., Saste-Garriga, J. et al. Normal appearing brain T1 relaxation time predicts disability in early primary progressive multiple sclerosis. Arch Neurol 2007;64: 411–415. 82. Wang, Yi. Description of parallel imaging in MRI using multiple coils. Magn Reson Med 2000;44:495–499. 83. Deoni, S. C. L., Catani, M. Visualization of the deep cerebellar nuclei using quantitative T1 and rho magnetic resonance imaging at 3 Tesla. Neuroimage 2007;37:1260–1266. 84. Ffrench-Constant, C. Pathogenesis of multiple sclerosis. Lancet 1994;29:271–275.

108

Deoni

85. Be, L. The histopathology of grey matter demyelination in multiple sclerosis. Acta Neurol Scand Suppl 2009;189:51–57. 86. Gasperini, C., Horsfield, M. A., Thorpe, J. W., Kidd, D., Barkr, G. J. et al. Macroscopic and microscopic assessments of disease burden by MRI in multiple sclerosis: Relationship to clinical parameters. J Magn Reson Imaging 1996;6:580–584. 87. Ropele, S., Strasser-Fuchs, S., Augustin, M. et al. A comparison of magnetization transfer ratio, magnetization transfer rate, and the native relaxation time of water protons related to relapsing-remitting multiple sclerosis. AJNR Am J Neuroradiol 2000;21:1885–1889. 88. Laule, C., Vavasour, I. M., Moore, G. R., Oger, J., Li, D. K., Paty, D. W., MacKay, A. L. Water content and myelin water fraction in multiple sclerosis. A T2 relaxation study. J Neurol 2004;251:284–293. 89. Laule, C., Vavasour, I. M., Kolind, S. H., Traboulsee, A. L., Moore, G. R. et al. Long T2 water in multiple sclerosis: What else can we learn from multi-echo T2 relaxation? J Neurol 2007;254:1579–1587. 90. Kolind, S. H., Laule, C., Vavasour, I. M., Li, D. K., Traboulsee, A. L. et al. Complementary information from multi-exponential T2 relaxation and diffusion tensor imaging reveals differences between multiple sclerosis lesions. Neuroimage 2008;40:77–85. 91. DeWitt, L. D., Kistler, J. P., Miller, D. C., Richardson, E. P., Buonanno, F. S. NMRneuropathologic correlation in stroke. Stroke 1987;18:342–351. 92. Kaur, J., Tuor, U. I., Zhao, Z., Peterson, J., Jin, A. Y., Baber, P. A. Quantified T1 as an adjunct to apparent diffusion coefficient for early infarct detection: A high-field magnetic resonance study in a rat stroke model. Int J Stroke 2009;4:159–168. 93. Bernarding, J., Braud, J., Hohmann, J., Mansmann, U., Hoehn-Berlage, M. et al. Histogram-based characterization of healthy and ischemic brain tissues using multiparametric MR imaging including apparent diffusion coefficient maps and relaxometry. Magn Reson Med 2000;43:52–61.

94. Reutens, D. C., Stevens, J. M., Kingsley, D., Kendall, B., Mose- ley, I. et al. Reliability of visual inspection for the detection of volumetric hippocampal asymmetry. Neuroradiology 1996;38:221–225. 95. Jackson, G. D., Connelly, A., Duncan, J. S., Gruenewald, R. A., Gadian, D. G. Detection of hippocampal pathology in intractable partial epilepsy: Increased sensitivity with qualitative magnetic resonance T2 relaxometry. Neurology 1993;43:1793–1799. 96. Conlon, P., Trimble, M. R., Rogers, D., Callicot, C. Magnetic resonance imaging in epilepsy: A controlled study. Epilepsy Reson 1988;2:37–43. 97. Jack, C. R., Jr., Wengenack, T. M., Reyes, D. A., Garwood, M., Curran, G. L. et al. In vivo magnetic resonance microimaging of individual amyloid plaques in Alzheimer’s transgenic mice. J Neurosci 2005;25:10041–10048. 98. Schenck, J. F., Zimmerman, E. A. Highfield magnetic resonance imaging of brain iron: Birth of a biomarker? NMR Biomed 2004;17:433–445. 99. Duan, J. H., Wang, H. Q., Xu, J., Lin, X., Chen, S. Q. et al. White matter damage of patients with Alzheimer’s disease correlated with decreased cognitive function. Surg Radiol Anat 2006;28:150–156. 100. Barkzokis, G., Lu, P. H., Mintz, J. Human brain myelination and amyloid beta deposition in Alzheimer’s disease. Alzheimers Dement 2007;3:122–125. 101. House, M. J., St. Pierre, T. G., Foster, J. K., Matrins, R. N., Clarnette, R. Quantitative MR imaging R2 relaxometry in elderly participants reporting memory loss. AJNR 2006;27:430–439. 102. Akshoomoff, N., Pierce, K., Courchesne, E. The neurobiological basis of autism from a developmental perspective. Dev Psychol 2002;14:613–634. 103. Saito, N., Sakai, O., Ozonoff, A., Jara, H. Relaxo-volumetric multispectral quantitative magnetic resonance imaging of the brain over the human lifespan: Global and regional aging patterns. Magn Reson Imaging 2009;27:895–906.

Chapter 5 Magnetic Resonance Brain Image Processing and Arithmetic with FSL William R. Crum Abstract Medical imaging has been transformed by a move from qualitative to quantitative approaches where image processing is used to enhance visual information and image analysis is used to derive structural and functional measurements. The ideal quantitative analysis methods are automatic and require no user intervention, and so-called image analysis pipelines exist for some applications. However, in the majority of cases automatic methods seldom live up to their name, may fail when prior assumptions are not met, and may not exist at all for new applications. The identification and careful use of well-known image processing and analysis techniques is a vital part of imaging and invaluable when problems arise with automatic methods. Here a number of key image analysis tasks in brain imaging are presented with particular reference to the freely available FMRIB Software Library. Key words: Medical image analysis, FSL, image segmentation, image registration, image arithmetic.

1. Introduction Medical image analysis has changed many aspects of clinical research and is finding application in clinical practice. Acquisition and analysis of single images in isolation are useful for answering very specific questions, e.g. “How big is this tumour?” “What is the lesion load?” “How big is the hippocampus?” Derived measurements such as lengths, areas, volumes, shape measures, and texture (1) can be analysed within and across groups and have been the basis of many successful scientific studies. Computer processing-based applications in neuroimaging, particularly in group-based neuroscience research studies, M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_5, © Springer Science+Business Media, LLC 2011

109

110

Crum

are amongst the best developed and fastest evolving uses of image analysis in medicine. Examples include Statistical Parametric Mapping (2) (http://www.fil.ion.ucl.ac.uk/spm), FreeSurfer (http://surfer.nmr.mgh.harvard.edu) (3), and the FMRIB Software Library (http://www.fmrib.ox.ac.uk/fsl) (4). With the growth in methods and applications comes a potentially bewildering collection of algorithms and tools to choose from. Some common analysis tasks benefit from automated approaches, particularly in studies involving large numbers of scans. However, automated analysis techniques can often fail when assumptions are broken (for instance when an atypical MR imaging sequence is used or when a study group has unusual brain structure or function) or image preparation (preprocessing) is not performed adequately. Specific applications or research studies can require preprocessing, measurements, or analysis steps not anticipated in automated pipelines. Therefore there remains a fundamental need for user-driven image processing, and analysis algorithms which are accessible to the non-specialist can be tuned for new applications and which when used effectively can ensure the maximum use is made of precious images. In this chapter we describe how to successfully apply some of the most common and useful brain image processing analysis steps using freely available software. We will separately consider noise reduction, automated skull and scalp removal, brain tissue classification, brain image registration, and image arithmetic. Finally we will show how simple techniques can be combined into sophisticated analysis pipelines by showing how to implement the Brain Boundary Shift Integral algorithm for brain atrophy measurement.

2. Materials 2.1. Software

There are many software packages available which can perform some or all of the tasks described in this chapter. We focus particularly on the FMRIB Software Library (FSL) (4, 5) as it is freely available for non-commercial use, extremely well supported via a well-supported email discussion forum (www.jiscmail.ac.uk/lists/fsl.html), and does not depend on other non-free software typically used as a platform for medical image analysis (e.g. MATLAB). Readers should note that FSL has many more tools providing much more functionality than we can cover in this chapter.

2.1.1. FSL System Requirements

FSL (the FMRIB Software Library) is developed and made available (http://www.fmrib.ox.ac.uk/fsl/) by the Oxford Centre

Magnetic Resonance Brain Image Processing and Arithmetic with FSL

111

for Functional MRI of the Brain (http://www.fmrib.ox.ac.uk/) licensed (http://www.fmrib.ox.ac.uk/fsl/fsl/licence.html) via Isis Innovation (http://www.isis-innovation.com/). It is a collection of computer programs and scripts which provide access to stand-alone processing tools and more sophisticated analysis pipelines. FSL is hardware compatible with PC, Mac, Sun, Silicon Graphics. Memory requirements will vary depending on the analysis and image and group data sizes, but 512 MB minimum available RAM is required (see Note 1). Swap space of at least 2 GB is recommended and should always at least equal the available RAM. Disk space of 10 times the size of the images to be analysed is recommended. The recommended operating system is Linux. Precompiled binaries are available for Centos (www.centos.org) and Debian (www.debian.org) Linux flavours. On Windows PC XP/Vista, Linux FSL can be run using VMware (www.vmware.com) by following the instructions here (http://www.fmrib.ox.ac.uk/fsl/fsl/windows.html). Source code is available for compiling on other operating systems. To validate the installation, an associated test suite of data and scripts, the FSL Evaluation and Example Data Suite (FEEDS) (http://www.fmrib.ox.ac.uk/fsl/feeds/doc/), should be downloaded and applied after a new installation. 2.1.2. FSL Tools

susan – non-linear noise reduction bet – brain extraction tool fast – brain tissue classification flirt – FMRIB’s Linear Image Registration Tool fslmaths – image arithmetic and operations fslstats – report summary intensity statistics All are part of the standard installation.

2.2. Example Image Data

We use the FSL Evaluation and Example Data Suite (FEEDS) (http://www.fmrib.ox.ac.uk/fsl/feeds/doc/) for most of the examples in this chapter (see Note 2).

3. Methods Analysis of MRI brain images involves some standard tasks which are independent of scanner and sequence type such as noise reduction, brain extraction, tissue classification, image registration, and image statistics. Generic algorithms have been developed which successfully tackle these problems. However, most of

112

Crum

these algorithms require parameters to be specified. In some cases default parameters are suitable for a wide range of images and applications. In other cases, careful tuning of parameters on a set of test data will be required. 3.1. Software Usage

The methods in this chapter use the command-line versions of the software which are invoked by typing commands into a terminal window. This form facilitates batch processing by including a series of commands in a shell script in Linux or other Unix variant. To see the format and allowed parameters for each command simply type its name without any supplied arguments or parameters. Where the software can also be launched with a graphical user interface which allows user interaction with a computer mouse this is indicated in the text.

3.2. Noise Reduction with Susan 3.2.1. Basic Usage and Parameters

The task is to reduce the appearance of noise in the image while preserving intensity gradients and region boundaries. See also Note 3. susan <use_median> = image to be de-noised = brightness threshold = spatial scale of de-noising (mm) = planar (2) or volumetric (3) de-noising <use_median> = do (0) or don’t (1) apply median filter to point noise = 0 = de-noised image

3.2.2. Example Usage

To denoise the image called structural.nii.gz (see Notes 4 and 5) in 3D with brightness threshold = 2,000, spatial scale = 2.0 mm, without using the median filter and producing a new de-noised image called structural-denoised.nii.gz use susan structural.nii.gz 10 2.0 3 0 0 structural-denoised.nii.gz

3.2.3. Parameter Setting

Figure 5.1 shows an example calibration run where the brightness threshold and spatial scale parameters were varied for different amounts of added Rician noise (6) (Note 6). See also Note 7 regarding computational requirements.

3.2.4. GUI Version

Susan


113

Fig. 5.1. Examples of applying susan for noise reduction of the image structural.nii.gz while varying the brightness threshold, bt, and the spatial scale, dt. Clockwise from top left: original, best result (bt = 2,000, dt = 2), brightness threshold too high (bt = 4,000, dt = 2), brightness threshold too high and spatial scale too large (bt = 4,000, dt = 4).

3.3. Brain Extraction Using Bet 3.3.1. Basic Usage

The task is to start with a brain scan which includes whole head and neck and produce a brain image with non-brain tissues such as neck, scalp, eyes, etc., removed (7). bet = original image = original image with non-brain tissues removed

3.3.2. Example Usage

To remove non-brain tissues from structural.nii.gz and generate a new image called structural_bet.nii.gz use bet structural.nii.gz structural_bet.nii.gz To also generate a brain mask called structural_bet_mask. nii.gz suitable for image registration use bet structural.nii.gz structural_bet.nii.gz -m

3.3.3. Advanced Usage

There are several parameters which are used to customise bet for non-standard images. bet -c -f <x y z > -g -r -c <x y z> = coordinates of “centre of gravity” (in voxels) of initial brain mesh

114

Crum

-f = fractional intensity threshold (range [0, 1], default = 0.5). Smaller values of f result in larger brain outline estimates -g = vertical gradient in fractional intensity threshold (range [–1, +1], default = 0). Positive (negative) values give larger (smaller) superior brain outline and smaller (larger) inferior brain outline. -r = head radius (in mm) which is used to set the size of the initial brain mesh 3.3.4. Example Advanced Usage

To remove non-brain tissues from structural.nii.gz and generate a new image called structural_bet.nii.gz while encouraging a larger brain outline with large initial estimate of head radius use bet structural.nii.gz structural_bet.nii.gz –f 0.7 -r 250.0

3.3.5. Parameter Setting

Figure 5.2 shows the use of bet and the different representations of the extracted brain.

Fig. 5.2. The different output options available for brain extraction of structural.nii.gz using bet. Clockwise from top left: original image, default BET output, original overlaid with binary mask, overlaid BET outline.

3.3.6. GUI Version

Bet

3.4. Brain Tissue Classification Using Fast 3.4.1. Basic Usage and Prerequisites

The task is to produce a brain image where voxels which are predominantly grey matter (GM), white matter (WM) and cerebrospinal fluid (CSF) are set to indicative values (background = 0,


115

CSF = 1, GM = 2, WM = 3) (8). To ensure non-brain voxels with similar intensities are not classified as brain tissue, brain extraction (Section 3.3) should be applied before brain tissue classification. Brain tissue classification is by default integrated with a correction for MRI intensity inhomogeneity (9). fast --nopve -o --nopve prevents additional partial volume estimation being performed = brain-extracted image (see Note 8) = base-name for output classification images 3.4.2. Example Usage

To create a simple brain tissue classification called structural_fast_seg.nii.gz from structural_bet.nii.gz use fast --nopve -o structural_fast structural_bet.nii.gz To additionally create a partial volume estimated brain tissue classification called structural_fast_pveseg.nii.gz (see also Note 9) use fast -o structural_fast structural_bet.nii.gz


There are several parameters which can be used to generate more detailed information. fast -B -o -B additionally outputs an intensity inhomogeneity-corrected image called _restore.nii.gz Multi-channel data can also be classified provided it is preregistered (see Note 10). fast -S -o . . . -S = number of supplied image channels (default = 1) The tissue classification can be made spatially smoother by increasing the -H and -R parameters. fast -H -R -o -H = initial segmentation spatial smoothness (default = 0.1) -R = PVE spatial smoothness (default = 0.3)

3.4.4. Example Advanced Usage

To perform tissue classification using partial volume estimation on registered T1-weighted and T2-weighted brain-extracted image volumes, outputting the intensity inhomogeneity-corrected image and forcing a smooth PVE segmentation use fast -S 2 -B -H 0.4 -o image_fast image_T1_bet.nii.gz image_ T2_bet.nii.gz

116

Crum

The resulting images will be called image_fast_seg.nii.gz, image_fast_pveseg.nii.gz, image_fast_pve_0.nii.gz, image_ fast_pve_1.nii.gz, image_fast_pve_2.nii.gz and image_fast_ restore.nii.gz. 3.4.5. Classification Types

Figure 5.3 shows examples of the images generated by fast.

Fig. 5.3. The different output options available for brain tissue classification with fast. Clockwise from top left: original structural image, standard fast segmentation overlaid on original, partial volume estimation (PVE) fast segmentation, overlaid on original, the estimated intensity bias field, the mixel map showing the different tissue compartments used in the analysis, the original image corrected for intensity bias.

3.4.6. GUI Version

Fast

3.5. Brain Image Registration Using Flirt 3.5.1. Basic Usage and Prerequisites

The task is to determine and apply an affine coordinate transformation which maps corresponding features (10) from an input brain scan onto a reference scan (11) (see Note 11).


117

flirt –in -ref -out = brain scan to be transformed = reference brain scan = transformed image 3.5.2. Example Usage

To register image.nii.gz with reference.nii.gz optimising the match over the reference brain (-refweight) (Note 12) and outputting a copy of the transformation matrix (-omat) use flirt -refweight reference_weight.nii.gz -in image.nii.gz -ref reference.nii.gz -out image_reg –omat matrix.mat To correct only for positional differences (three rotations and three scalings) between the brains use flirt -dof 6 -refweight reference_bet_mask.nii.gz -in image.nii.gz -ref reference.nii.gz -out image_reg –omat matrix.mat To correct for positional differences and global scalings (Note 13) between the brains use flirt -dof 9 -refweight reference_bet_mask.nii.gz -in image.nii.gz -ref reference.nii.gz -out image_reg –omat matrix.mat To apply an existing registration result to an additional image, image_add.nii.gz, use flirt -in image_add.nii.gz -ref reference.nii.gz -init matrix.mat -applyxfm -out image_add_reg.nii.gz


There are several parameters which affect the way the registration is optimised. -nosearch switch off the global rotational optimisation -init matrix_init.mat = supply an initial estimate of the transformation matrix -searchrx <min_angle> <max_angle> = specify angular search range around the x-axis in degrees (default is –90 90). Also –searchry and –searchrz. -coarsesearch <delta_angle> = specify coarse angular search increment in degrees (default is 60) -finesearch <delta_angle> = specify coarse angular search increment in degrees (default is 18) There are several parameters which affect the way the registration is assessed. -cost costfn = specify function used to assess registration, (default is corratio) -searchcost costfn (default is corratio)

118

Crum

In both cases, costfn should be one of (mutualinfo, corratio, normcorr, normmi, leastsq, labeldiff) (see Note 14). 3.5.4. Example Advanced Usage

To register two images with weighting images applied to both using normalised mutual information as the cost function and restricting the angular search around the z-axis use flirt -dof 6 –searchz 0 60 –cost normmi -searchcost normmi –inweight iweight.nii.gz -refweight rweight.nii.gz –in input.nii.gz -ref reference.nii.gz -out output.nii.gz -omat output.mat The resulting image will be called output.nii.gz with an associated rigid body transformation matrix called output.mat.

3.5.5. Registration Example

Figure 5.4 shows examples of images registered using flirt.

Fig. 5.4. Example of affine registration of T1 and T2 axial images with flirt. Top row: left = T1-weighted image, middle = misregistered T2-weighted image, right = registered T2-weighted image. The registration is relatively subtle with some in-plane positional displacement and some through-plane misregistration visible around the left medial cortex. Bottom row: left = overlay of misregistered T2-weighted and T1-weighted images, right = overlay of registered T2-weighted and T1-weighted images. Misregistration is particularly evident around the ventricles where light and dark boundaries show the regions of mismatch and around the cortex. These misregistrations are clearly resolved on the overlaid registered images.

3.5.6. GUI Version

Flirt

3.6. Image Operations, Statistics and Arithmetic 3.6.1. Basic Usage and Prerequisites

The FMRIB Software Library provides some general purpose utilities such as fslmaths (image arithmetic and other operations) and fslstats (image statistics) which, in conjunction with bet, fast,


119

flirt, enable sophisticated image analysis to be performed. This is demonstrated in Section 3 by showing how fslmaths and fslstats can be used to implement the Brain Boundary Shift Integral (12, 13) algorithm for measuring brain atrophy in longitudinal imaging. First some common analysis steps involving brain tissue masks, arithmetic operations on pairs of images and image intensity histogram construction are detailed. fslmaths [operations] [operations] are typically of the form where data is a number or another image fslstats [options] [options] are typically single letter arguments, e.g. “-m” outputs the mean intensity 3.6.2. Example Usage

To create a binary GM mask (where non-GM voxels have intensity 0 and GM voxels have intensity 1) from the output of fast (see Notes 15–17) use fslmaths structural_fast_pveseg.nii.gz -thr 2 -uthr 2 –div 2 structural_gm_mask.nii.gz To create a binary brain mask (where non-brain voxels have intensity 0 and brain voxels have intensity 1) from the output of fast (see Note 15, 16) use fslmaths structural_fast_pveseg.nii.gz -thr 2 -min 1 structural_brain_mask.nii.gz To create a subtraction image, image12_diff.nii,gz, highlighting differences between two images, image1.nii.gz and image2.nii.gz (see Notes 18, 19), use fslmaths image1.nii.gz diff.nii,gz

-sub

image2.nii.gz

image12_

To compute the entries of a 64-bin intensity histogram for structural.nii.gz (Fig. 5.5) use fslstats structural.nii.gz -h 64 3.6.3. Advanced Usage

Computing the Brain Boundary Shift Integral, a measure of brain volume change which has occurred in the time between two scans image1.nii.gz and image2.nii.gz being acquired. See Fig. 5.6 for a flow chart of the main image processing operations. First perform brain extraction using bet (Section 3.3) generating image1_bet.nii.gz, image1_bet_mask.nii.gz, image2_bet. nii.gz, image2_bet_mask.nii.gz. Perform tissue classification using fast (Section 3.4) on image1_bet.nii.gz and image2_bet.nii.gz. The images generated are image1_fast_pveseg.nii.gz and image2_fast_pveseg.nii.gz.

120

Crum

Fig. 5.5. A graph showing the intensity histogram with 64 bins for the example image structural.nii.gz. The data were generated from the command fslstats structural.nii.gz –h 64. Note that for illustrative purposes the log of the entries in each intensity bin is plotted.

Generate a binary brain mask from each pveseg image generated in step 3b using fslmaths (Section 3.6, Section 2b) producing image1_brain_mask.nii.gz and image2_brain_mask.nii.gz. Generate a combined dilated (expanded) brain and CSF mask of the first scan for registration with foreground/background intensity ratio = 250/1. fslmaths image1_bet_mask.nii.gz –dilM –mul 249 –add 1 image1_bet_mask_weight.nii.gz Register image2.nii.gz to image1.nii.gz as in Section 3.5 (using the option -refweight image1_brain_mask_weight.nii.gz) to produce image2_reg.nii.gz and matrix.mat. Use flirt with the -interp nearestneighbour option to apply the registration result to the corresponding brain mask to produce image2_brain_mask_reg.nii.gz (Section 3.5.2) (Note 20). Compute the intersection of the two registered brain masks (Note 21). fslmaths image1_brain_mask.nii.gz –mul fslmaths image2_ brain_mask_reg.nii.gz brain_mask_int.nii.gz Compute the union of the two registered brain masks (Note 22). fslmaths image1_brain_mask.nii.gz -add fslmaths image2_ brain_mask_reg_thresh.nii.gz –min 1 brain_mask_uni. nii.gz Apply binary erosion to the intersection brain mask and binary dilation to the union brain mask:


121

Fig. 5.6. A schematic showing the image processing steps which generate the brain– CSF boundary mask for the Brain Boundary Shift Integral calculation. (i) Scan 2 is registered to scan 1 and transformed to give scan 2 , (ii) the brain is extracted from scan 1 and scan 2 using bet, (iii) brain tissue classification is performed on scan 1 and scan 2 using fast, (iv) the fast tissue maps are manipulated using fslmaths to produce brain masks, and the brain mask on scan 2 is transformed using the registration result giving mask 2 , (v) the registered brain masks are combined using fslmaths to produce the intersection (inner) and union (outer) brain masks, (vi) the inner brain mask is eroded and the outer brain mask is dilated, both using fslmaths, (vii) the brain boundary mask is the difference between the dilated outer mask and the eroded inner mask, again computed using fslmaths. See text for a full description of all these steps.

fslmaths brain_mask_uni.nii.gz -dilM brain_mask_uni_ d.nii.gz fslmaths brain_mask_int.nii.gz -ero brain_mask_int_e.nii.gz Generate a mask covering the boundary of brain and CSF by computing the difference between the dilated union and the eroded intersection images (Note 23). fslmaths brain_mask_uni_d.nii.gz -sub brain_mask_int_e. nii.gz image_csf_bnd.nii.gz Normalise the original images over the internal region so the mean intensity is 100: #!/bin/csh # See Note 24. # First, compute and store the mean brain intensity

122

Crum

set mean1 = `fslstats image1.nii.gz -k brain_mask_int_e.nii.gz –m` set mean2 = `fslstats image2_reg.nii.gz -k brain_mask_int_ e.nii.gz –m` # Then normalise the image intensities to have average intensity = 100 fslmaths image1.nii.gz –div ${mean1} -mul 100 image1_ norm.nii.gz fslmaths image2_reg.nii.gz –div ${mean2} -mul 100 image2_ reg_norm.nii.gz Clip the maximum and minimum normalised intensities to an appropriate window (Note 25) and compute the difference image. fslmaths image1_norm.nii.gz -max 25 -min 75 image1_ norm_clip.nii.gz fslmaths image2_reg_norm.nii.gz -max 25 -min 75 image2_ reg_norm_clip.nii.gz fslmaths image1_norm.nii.gz –sub image2_reg_norm.nii.gz image12_sub.nii.gz Finally compute the boundary shift integral by summing the clipped intensity difference across the brain boundary (see Fig. 5.7).

Fig. 5.7. The final stages of the Brain Boundary Shift Integral calculation. Left: the normalised subtraction image of two scans of a subject showing bright areas of atrophy in brain. Middle: a mask of the brain–CSF boundary. Right: the mask overlaid on the subtraction image showing the parts of the subtraction image which contribute to the integral.

#!/bin/csh set bbsi = `fslstats image12_sub.nii.gz -k brain_csf_bnd.nii.gz –V -M` set pix = `fslinfo image12_sub.nii.gz | egrep pixdim` # See Note 26 concerning then next line set bbsiml = ècho scale=2\; $$bbsi[3] ∗ $bbsi[1] ∗ $pix[2] ∗ $pix[4] ∗ $pix[6] $/$1000 ∗ 50 $ | bc` echo BBSI is ${bbsiml} ml


123

4. Notes 1. A “512 MB computer” will have significantly less than 512 MB RAM available for analysis because of memory required by the operating system and other processes. 2. Images may be viewed using another FSL program called fslview. The general use is fslview image1.nii.gz image2.nii.gz. . . imagen.nii.gz. 3. Susan is most useful for visualisation or for subsequent manually driven measurements (e.g. manual segmentation). Many automated analysis techniques incorporate a noise model, and susan should then not be applied as it may change the assumed noise properties of the image. 4. The image called structural.nii.gz is a three-dimensional structural MRI brain volume supplied with the FEEDS testing and evaluation suite. 5. We assume images are stored in the NIfTI-1format (http://nifti.nimh.nih.gov) and then compressed using the common Unix tool gzip. Thus structural.nii indicates an image called structural stored in the NIfTI format and structural.nii.gz is obtained by issuing the command gzip structural.nii on Unix/Linux systems. Note that the FSL software reads compressed files transparently, so there is usually no need for the user to compress or uncompress (using gunzip) images by hand. 6. The figures show single slices for illustrative purposes, but it should be remembered that MR images are usually comprised of multiple slices which either are collected to be approximately contiguous or are reconstructed to be contiguous from a 3D data volume. The image analysis techniques in this book operate on all slices in the volume unless otherwise specified. 7. The computational time for noise reduction with susan rises significantly with increasing values of the spatial scale parameter . 8. The brain-extracted image supplied to fast is assumed to be T1 weighted. When a T2 or proton density-weighted image is used the –t option specifies the image type. fast –t 1 –o image_T1_fast image_T1_bet.nii.gz fast –t 2 –o image_T2_fast image_T2_bet.nii.gz fast –t 3 –o image_PD_fast image_PD_bet.nii.gz 9. By default, as well as generating a partial volume segmented image, fast also outputs individual partial volume estimates of each tissue class. For an output base name specified as

124

Crum

-o image_fast, the default output images are image_fast_ seg.nii.gz, image_fast_pveseg.nii.gz, image_fast_pve_0.nii. gz, image_fast_pve_1.nii.gz and image_fast_pve_2.nii.gz. In principle these individual pve images, which contain fractional estimates of tissue occupancy per voxel, can be used for more accurate tissue volume estimation. 10. Multi-channel data can produce more robust tissue classification. The classic MRI example is using T1-weighted images in conjunction with T2-weighted images. However, post acquisition, these images will not generally be aligned and usually have different voxel dimensions and/or fields of view. Therefore a registration and resampling procedure must be performed to ensure that the T2-weighted brain image is aligned with the T1-weighted image and has the same voxel dimensions; flirt can be used for this purpose. 11. Brain image registration is almost never performed without other options because by default, no distinction is made between foreground and background voxels or between brain and non-brain voxels. The optimum registration computed for a T1-weighted volume including air, neck, scalp, etc., in addition to brain is highly unlikely to produce a good alignment of the brain, even when the scans are of the same person, because of differences in shape and position of the different tissues across scans. 12. When using weight images, it is tempting to set the background weight to 0, e.g. by simply using the binary mask output by bet as a weighting image. However, this can destabilise the registration by removing too much information about surrounding tissues. It is far better to use a weights image with a high foreground value and a low background value. The simplest way to accomplish this starting with a binary brain mask is to use fslmaths to scale and add a small constant value (e.g. 1) to every point in the input mask (binary_mask.nii.gz) to produce a suitable weights_image.nii.gz: fslmaths binary_mask.nii.gz –mul 99 –add 1 weights_image.nii.gz. 13. “dof” stands for degrees of freedom and specifies the number of independent parameters in the affine transformation matrix. The default is 12 corresponding to 3 translations, 3 rotations, 3 scalings and 3 skews. Other commonly used dofs are six (three translations, three rotations, often called “rigid body”) and nine (three translations, three rotations and three scalings). 14. When the image intensity characteristics are not well characterised (the usual scenario unfortunately) it is safest to choose costfn as normmi (normalised mutual information) or corratio (correlation ratio) which assume an


125

unknown probabilistic or functional relationship between voxel intensities in the two images, respectively. When the images to be registered are known to have very similar intensity characteristics (i.e. they are structurally similar and were collected in the same way on the same scanner) then costfn can be normcorr (normalised correlation) which assumes a linear relationship between voxel intensities; this can be more robust but is also more susceptible to deviations from those assumptions. 15. The standard segmentation of a T1-weighted volume output by fast has voxel intensities 0 (background), 1 (CSF), 2 (GM) and 3 (WM). 16. Operations are executed in order. –thr 2 sets any voxels of intensity less than 2 to 0. –uthr 2 sets any voxels of intensity greater than 2 to 0. –div 2 divides each voxel intensity by 2. The result is to leave voxels which were originally of intensity 2 (corresponding to GM) set to 1 and all other voxels set to 0. Other operations include –min x (replace each voxel with the lower of its original intensity or x) and –max x (replace each voxel with the higher of its original intensity or x). 17. Substitute “3” for “2” to get a WM mask instead of a GM mask. 18. In brain imaging scans would typically first be aligned using flirt and then intensity normalised as in the previous example before creating the difference image. 19. Images can also be added, multiplied or divided by replacing –sub with –add, –mul or –div, respectively. 20. During registration and image transformation it is necessary to interpolate one image to match the voxel boundaries of another. The –interp option allows the intensity interpolation method to be specified. In order of increasing computing time and accuracy the options are nearestneighbour, trilinear and sinc; trilinear is sufficient unless very high accuracy is required and nearestneighbour is most often used to interpolate binary region masks. 21. The intersection of two binary voxels is equivalent to a logical AND (i.e. both voxels must be set). Therefore a simple implementation is to multiply voxels together – if either is 0 the result will be 0. 22. The union of two binary voxels is equivalent to a logical OR (i.e. either voxel is set). Therefore a simple implementation is to add voxels together and then reset positive results to 1. 23. To expand the region, add more –dilM and –ero terms, respectively, to the first two fslmaths commands.

126

Crum

24. These commands make use of Unix csh (“c-shell”) syntax. They should be put in a data file called normalise-script and saved to the current directory. Then do chmod +x ./normalise-script to make the script executable. Then do ./normalise-script to run the commands. 25. The intensity window limits of 25 and 75 follow the original reference with width = centre = 50 for intensities scaled by a factor of 100. The optimum values will be application dependent and can be determined using procedures outlined in (12, 13). 26. This line converts the Boundary Shift Integral into sensible units using the inbuilt Unix calculator bc to two decimal places (scale = 2). It computes the raw integral equal to mean masked intensity ($bbsi[3]) times the number of voxels ($bbsi[1]) converted to mm3 by multiplying by the voxel dimensions $pixdim[1], $pixdim[2], $pixdim[3]. The integral must be normalised by dividing by the intensity window width (50) and converted to ml (1,000). References 1. Crum, W. R. Shape and texture. Quantitative MRI of the Brain: Measuring Changes Caused by Disease, P. Tofts, (ed.) Chichester: Wiley; 2004, pp. 559–579. 2. Friston, K. J., Ashburner, J. T., Kiebel, S. J., Nichols, T. E., Penny, W. D., (eds.). Statistical Parametric Mapping: The Analysis of Functional Brain Images. Amsterdam: Academic Press; 2007. 3. Fischl, B., Liu, A., Dale, A. M. Automated manifold surgery: Constructing geometrically accurate and topologically correct models of the human cerebral cortex. IEEE Trans Med Imaging 2001;20(1):70–80. 4. Smith, S. M., Jenkinson, M., Woolrich, M. W., Beckmann, C. F., Behrens, T. E. J., Johansen-Berg, H., Bannister, P. R., De Luca, M., Drobnjak, I., Flitney, D. E., Niazy, R., Saunders, J., Vickers, J., Zhang, Y., De Stefano, N., Brady, J. M., Matthews, P. M. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage 2004;23(S1):208–219. 5. Woolrich, M. W., Jbabdi, S., Patenaude, B., Chappell, M., Makni, S., Behrens, T., Beckmann, C., Jenkinson, M., Smith, S. M. Bayesian analysis of neuroimaging data in FSL. Neuroimage 2009;45:S173–S186. 6. Gudbjartsson, H., Patz, S. The Rician distribution of noisy MRI data. Magn Reson Med 1995;34(6):910–914.

7. Smith, S. M. Fast robust automated brain extraction. Hum Brain Mapp 2002;17(3):143–155. 8. Zhang, Y., Brady, M., Smith, S. Segmentation of brain MR images through a hidden Markov random field model and the expectation maximization algorithm. IEEE Trans Med Imaging 2001;20(1): 45–57. 9. Vovk, U., Pernus, F., Likar, B. A review of methods for correction of intensity inhomogeneity in MRI. IEEE Trans Image Process 2007;26(3):405–421. 10. Crum, W. R., Hartkens, T., Hill, D. L. G. Non-rigid registration, theory and practice. Br J Radiol 2004;77: S140–S153. 11. Jenkinson, M., Smith, S. M. A global optimisation method for robust affine registration of brain images. Med Image Anal 2001;5(2):143–156. 12. Fox, N. C., Freeborough, P. A. Brain atrophy progression measured from registered serial MRI: Validation and application to Alzheimer’s disease. J Magn Reson Imaging 1997;7:1069–1075. 13. Freeborough, P. A., Fox, N. C. The boundary shift integral: An accurate and robust measure of cerebral volume changes from registered repeat MRI. IEEE Trans Med Imaging 1997;16(5):623–629.

Chapter 6 Diffusion Tensor Imaging Derek K. Jones and Alexander Leemans Abstract Diffusion tensor MRI (DT-MRI) is the only non-invasive method for characterising the microstructural organization of tissue in vivo. Generating parametric maps that help to visualise different aspects of the tissue microstructure (mean diffusivity, tissue anisotropy and dominant fibre orientation) involves a number of steps from deciding on the optimal acquisition parameters on the scanner, collecting the data, pre-processing the data and fitting the model to generating final parametric maps for entry into statistical data analysis. Here, we describe an entire protocol that we have used on over 400 subjects with great success in our laboratory. In the ‘Notes’ section, we justify our choice of the various parameters/choices along the way so that the reader may adapt/modify the protocol to their own time/hardware constraints. Key words: Diffusion tensor, MRI, sampling schemes, pulse sequence, optimal, data quality.

1. Introduction Diffusion tensor MRI (DT-MRI), developed in the early- to mid1990s (1, 2), provides a means for non-invasively characterising the properties of soft tissue on a microstructural scale. It works by sensitising the MRI signal to the random molecular motion of water molecules (diffusion) by addition of ‘diffusionencoding gradients’ to a standard MR pulse sequence (3). At the typical resolution of a DT-MRI experiment (2–3 mm voxel sizes), in the grey matter and cerebro-spinal fluid, the diffusionweighted signal is independent of the direction in which the gradients are applied, and the diffusion appears to be isotropic (2). In white matter, water molecules diffuse more freely along the dominant fibre orientation than across them (4). This anisotropy of diffusion provides insights into the microstructural organisation M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_6, © Springer Science+Business Media, LLC 2011

127

128

Jones and Leemans

of the white matter. The simplest model that encapsulates this anisotropic diffusion is the diffusion tensor (1). By applying diffusion-encoding gradients in at least six non-collinear and noncoplanar orientations, one can estimate the six unknown elements of the diffusion tensor (1) and thus characterise the anisotropy. Further, the direction in which the diffusion-weighted signal has the greatest attenuation gives an indication of the dominant fibre orientation – which can be used to create voxel-wise maps of fibre orientation (e.g. 5) or pieced together to reconstruct continuous trajectories throughout the white matter (i.e. ‘tractography’) (e.g. 6–10). Such information has previously been unavailable in vivo – and so it is understandable that the technique has attracted huge interest from, and has enjoyed rapid uptake by, clinical and neuroscientific communities. Here we present a protocol that begins by bringing the subject to the scanner room and preparing them for the diffusion tensor imaging data acquisition, through the acquisition of the data and pre-processing, through to estimation of the diffusion tensor in each voxel and subsequent computation of quantitative parametric maps. We stress the importance of checking the data at each stage of the pipeline to ensure that the data are robust. It should be noted that what is ‘optimal’ for one laboratory – which can only allocate 10 min to DT-MRI on a 1.5 T system made by manufacturer X, for example, will be of little interest to the group that can allocate 30 min on a 3 T system made by manufacturer Y. Therefore, in the Notes section, details will be provided on how to proceed in choosing what is optimal for a given set of circumstances.

2. Materials MRI Scanner: General Electric HDx 3.0 T system (see Note 1). a. Gradients: Twin-speed gradient system with gradient strength = 40 mT/m and maximum slew rate = 150 T/m/s (see Note 2). b. RF Coils: Whole-body birdcage coil used for RF transmit; eight-channel head coil (made by MRI Devices Corp.) used for RF receive (see Note 3). c. Scanner Software Capability: Software to provide diffusion tensor imaging capability (see Note 4). d. Peripherals: Adequate padding for the head (wedge cushions, etc.); hearing protection (ear plugs); a peripheral pulse-oximeter; a squeeze-bulb (for the participant to communicate to the operator).

Diffusion Tensor Imaging

129

3. Methods 1. Preparing the Participant: After successful completion of standard MR screening and appropriate informed consent forms, the participant is led to the MR magnet room. They are given earplugs before being placed onto the scanner bed in the supine position. The pulse-oximeter is then placed onto the subject’s forefinger, and they are given the squeezebulb. We take special care to warn the participant that the diffusion tensor imaging part of the protocol is ‘louder than the other scans’ and that they ‘can expect the bed to vibrate quite a lot’ (see Note 5). We also warn the participant that ‘there will be irregularly timed knocking noises – and these will appear to move about as the scan progresses. This is fully expected’ (see Note 5). 2. Scanning: The integrated laser alignment system is used to landmark on the nasion, and the participant slid into the magnet, taking particular care not to trap the squeeze-bulb/pulse-oximeter leads during the process. As an optional extra, we provide the participant with the option of watching a subtitled movie of their choice in the scanner via a rear projection onto a periscope mounted on the head coil (see Note 6). The sequence is a twice-refocused spin-echo EPI sequence (11) (see Note 7), with a parallel imaging (ASSET) factor of 2 (see Note 8). Sixty axially oriented slices are prescribed to cover the entire head (see Note 9). The field of view is 230 mm, with an acquisition matrix of 96 × 96 and a slice thickness of 2.4 mm (see Note 10). A total of 66 images are acquired (see Note 11) at each of 60 slice locations. Six images are acquired with no diffusion-weighting gradients applied, and 60 diffusion-weighted images are acquired at a b-value of 1,000 s/mm2 (see Note 12). The diffusion-weighted images are acquired with encoding gradients applied along 60 non-collinear directions (see Table 6.1 for the configurations – and see Note 13). The echo time is 87 ms (see Note 14), and the sequence is triggered to the cardiac cycle via a pulse-oximeter placed on the participant’s forefinger (see Note 15). We select a minimal trigger delay (see Note 16), and the effective TR that is set depends on the heart rate estimated from the pulse-oximeter trace. Each image is initially stored in DICOM format. We then convert the separate DICOM images into a 4D data set (with ‘time’ or ‘diffusion-weighted measurement’ as the fourth dimension) in the NIFTI imaging format.

130

Jones and Leemans

Table 6.1 60 Electrostatically arranged sampling vectors (21) optimally ordered according to Cook et al. (24) 0.1706

0.3255

0.9300

0.3933

0.9154

−0.0853

−0.6263

−0.2862

0.7252

−0.6058

0.7763

0.1742

0.9204

−0.0040

0.3911

0.2903

−0.6194

0.7294

−0.7022

0.3532

0.6182

0.3712

0.7702

0.5187

−0.9738

−0.0801

0.2126

0.2158

−0.9371

0.2743

−0.0604

−0.2236

0.9728

0.8147

0.5611

0.1462

−0.2150

0.7211

0.6587

0.7166

−0.5846

0.3804

0.5354

−0.1672

0.8279

−0.2157

−0.7327

0.6454

0.6549

0.3557

0.6668

−0.0601

0.9726

0.2244

−0.7389

−0.5195

0.4290

−0.9432

0.3042

0.1338

−0.4428

0.2032

0.8733

−0.1629

−0.9173

0.3634

0.0988

0.6380

0.7637

0.7879

−0.2745

0.5513

−0.3122

−0.4225

0.8509

3. Initial Quality Check: The data are loaded into FSLview (part of the FSL software package) (www.fmrib.ox.ac.uk/fsl) and viewed in all three planes simultaneously – in ‘cine’ mode – looping through the separate volumes to check for any obvious artefacts in the data (see Fig. 6.1a and Note 17). 4. Correcting Motion/Distortion: We register the 2nd through 66th diffusion-weighted volumes to the 1st diffusionweighted volume using a global, 12 degrees of freedom affine deformation, with normalised mutual information as the cost function (see Note 18). We extract the rotational component of the transformation for each volume and apply this rotation to the gradient encoding tables (see Note 19).


131

Fig. 6.1. Example of the data quality assessment through visual inspection. a Visualisation of the raw diffusion-weighted data. This is done as a cine loop to quickly view all diffusion-weighted volumes. Abnormal data values, such as signal dropouts (i) or hyperintensities (ii), are easily detectable on the planes orthogonal to the acquisition plane (axial), i.e. the coronal and sagittal image views. b Looking at the data residuals to the diffusion tensor fit may reveal other data abnormalities, for instance, in the form of hyperintensities around the rim of the brain (iii), which can also be observed in the FA maps (c). This high anisotropy rims suggest image misalignment across the different diffusion-weighted images due to subject motion or geometric distortions. In (d), the FA map is shown after performing this correction procedure.

132

Jones and Leemans

5. Re-inspecting the Data: We re-inspect the data in three orthogonal planes in a cine format to ensure that the motion/distortion correction has been performed correctly and that no additional artefacts have been introduced into the data. 6. Fitting the Diffusion Tensor: We perform an initial ordinary least squares fit of a single Gaussian diffusion tensor to the data to generate a starting estimate for subsequent entry into a nonlinear least squares (Levenberg–Marquardt) algorithm to estimate the tensor in each voxel (see Note 20). 7. Further Inspection of the Data: We generate a 4D data set of the residuals to the tensor fit (Fig. 6.1b) and again view these in three orthogonal planes as a movie to look for obvious outliers/artefacts (see Note 21). 8. Computation of Parametric Maps: The tensor in each voxel is diagonalised to derive the eigenvectors and eigenvalues, and parametric maps of the mean diffusivity and fractional anisotropy (2) are computed (see Note 22). We also compute the directionally encoded colour (DEC) map showing fibre orientation (5). 9. Final Inspection of the Data: Before passing any maps into further analyses, we visualise the FA, MD and DEC maps in three orthogonal orientations to check for any obvious artefacts, such as rims of high anisotropy at the edge of the brain (see Fig. 6.1c and Note 23).

4. Notes 1. In most research environments for human imaging, the field strength is 1.5 or 3.0 T with a few exceptions. The advantage of higher field is higher signal-to-noise ratio (SNR) per unit time – allowing higher resolution for fixed scan time, shorter scan time for the same resolution or higher precision in the data if all scan parameters are kept the same. The disadvantage is that the standard approach to DT-MRI (1, 2) uses echo-planar imaging (EPI) – which renders the images to be very sensitive to differences in magnetic susceptibility (such as at air–tissue interfaces) – leading to either a stretch or a compression of the image along the phase encode direction (usually aligned with the anterior–posterior axis of the head), and these distortions become worse at higher field strength. 2. The imaging gradients should also ideally have the capability of producing gradient amplitudes of above 20 mT/m. The stronger the gradient amplitudes, the better for


133

diffusion MRI. The key factors in determining the amount of diffusion weighting, characterised by the ‘b-factor’, b, are the amplitude (G), the duration (δ) and the temporal separation ( ), given by the so-called Stejskal–Tanner equation (3): δ , − 3

2

2 2

b=γ G δ

where γ is the gyromagnetic ratio. Increasing the gradient amplitude means that smaller values of δ and can be employed, which in turn means shorter echo times and therefore increased SNR. 3. A multi-channel head coil is preferable for improved SNR and the possibility to employ parallel imaging, which in turn helps to reduce the EPI-based distortions (12). Eightchannel head coils are prevalent in neuroimaging research centres, but again – more channels, if available, are preferred with 12-channel and even 32-channel coils becoming purchasable options. Again – these will boost the SNR – which is most definitely beneficial for DT-MRI. Further, more channels permit the use of higher speed-up factors in parallel imaging acquisition strategies, which is advantageous for DT-MRI (12). 4. The scanner should provide the capability of applying diffusion-encoding gradients in at least six different orientations (1), although more is better (see Note 13). The capability to choose one’s own encoding directions (normally facilitated under a research agreement with the manufacturer) is preferred, particularly if the sampling orientations are to be optimally ordered (see Paragraph 2 of Note 13). 5. Past experience has shown that it is beneficial to alert the participant to the fact that the vibrations can be quite severe. This is particularly true with the twice-refocused pulse sequence – as reported in (13). Moreover, it is useful to warn the participant that there will be spatially varying irregularly spaced knocking which refers to the fact that the TR will be non-uniform due to the cardiac gating, and the knocking is the gradients being played out in different orientations. 6. The use of a video projection system, at least anecdotally, appears to reduce head motion as it engages the participant and reduces the likelihood of them looking around (which may result in additional head motion) or being distracted by other thoughts over what can be a considerable duration of scan (sometimes up to 30 min). It is not expected that there will be functionally dependent changes in the signal

134

Jones and Leemans

acquired with a standard DT-MRI protocol that will impact on standard DT-MRI analyses. 7. The most commonly used acquisition scheme for DT-MRI is now the twice-refocused spin-echo EPI sequence (11) and is provided by most manufacturers as part of the DT-MRI package that they sell. The twicerefocused sequence will markedly reduce the effects of eddy current-induced distortions resulting from the rapid switching of the diffusion-encoding gradients. 8. For EPI-based DT-MRI acquisitions, there is a clear benefit to the use of parallel imaging (12) using approaches such as the image domain SENSE approach (14) or k-space domain approaches such as GRAPPA (15). The choice of which to use normally depends on availability and/or software provided by the manufacturer. A parallel imaging factor of 2 seems to provide a reasonable compromise between SNR, distortion reduction and speed-up factor. 9. Pure axial orientation gives the best quality data in terms of ghosts and distortions. 10. For most purposes, it is highly desirable to have isotropic imaging voxels for DTI so that there is no preferential averaging of fibre orientations along a particular axis. This is particularly important for tractography applications. The limitations to consider are (1) The slice profile: On some scanners (e.g. General Electric), the issue of the fat–water frequency shift, particularly problematic in EPI, is addressed through the use of a spatially and spectrally selective pulse – which leaves the fat (in the scalp, for example) unexcited. While this is an alternative to implicitly turning on ‘FatSat’, it does tend to limit the minimal slice thickness that can be achieved (around 2.5 mm is typical). (2) The signal-to-noise ratio: It is worth bearing in mind that the average SNR in the diffusion-weighted image is on the order of 30% of that in the non-diffusion weighted image. If the non-diffusion weighted signal intensity is I0 then the diffusion-weighted signal is I0 exp(-bD), where D is the apparent diffusion coefficient along the axis of the applied encoding gradient. As discussed below, the b-value is typically on the order of the reciprocal of D so that I = I0 exp(–1) = 0.33I0 . For isotropic media, all diffusion-encoding directions will have the same attenuation. It is important that the SNR in the diffusion-weighted images does not go below approximately 3:1. This is the domain where the Rician-distributed data begin to look non-Gaussian (16, 17), and one encounters problems with the rectified noise floor that cause underestimation of dif-


135

fusivities (18) and corrupt estimates of diffusion anisotropy (19), among other problems. For anisotropic media, say with a fractional anisotropy (20) of 0.7, the largest eigenvalue will be twice the mean diffusivity. This, in turn, means that the signal attenuation will be equal to exp(–2bD) = 0.13. Therefore to void the rectified noise floor issue, the SNR in the non-diffusion weighted image should be at least 3:0.13∼ = 22:1. It is therefore advisable, when setting up the protocol, to make some SNR measurements (with the coil, parallel imaging factor, etc., for the experiment already chosen and in place) before settling on the final resolution. In summary, one should go for the highest resolution possible that achieves isotropic resolution but ensures that the SNR in white matter in the non-diffusion image is greater than 20. The field of view should be sufficient to cover the entire head and selected in consideration of the image acquisition matrix and slice thickness to ensure isotropic voxels. 11. The more measurements are made, the more accurate and precise the parameters derived from DTI will be. The total number of measurements will therefore be dictated by the time allocated in the protocol. The ratio of measurements made at the higher b-value to those with no diffusion weighting should be around 8–10:1 (21); it is highly recommended that the images with b = 1,000 s/mm2 are acquired with gradients distributed as uniformly in space as possible (21, 22 – see Note 13). If time permits, 30 directions (with two or three non-diffusion weighted scans) will give measures that are statistically rotationally invariant (22) – but if time permits even more measurements to be made, then one should take advantage and acquire more points. 12. It is recommended that just two different diffusion weightings are used: one close to zero and one close to b = 1,000 s/mm2 , the latter being a compromise between ensuring sufficient attenuation for precise estimation of diffusivities and balancing this against the penalty of increasing the echo time to accommodate the diffusion-encoding gradients, which in turn leads to signal loss through T2 relaxation (21). 13. For estimating the tensor, it is beneficial to space out the sampling gradients as uniformly in space as possible (21, 22). This helps to reduce the dependence of the reproducibility (variance) of measurements on the orientation of the material being imaged (22). The de facto standard, adopted by most manufacturers and research groups, is now to use an ‘electrostatic repulsion’ algorithm (21) to

136

Jones and Leemans

space out the gradients. This is done by simulating what happens if the gradient sampling vectors are all rods passing through the origin (so that each rod points in the direction of a sampling vector and its antipode) and if there are point charges placed at the ends of each rod. The orientations of the rods are then changed until the sum of repulsive forces between all possible pairs of charges is minimised (this is effectively what happens in the formation of crystals in nature – so it is no surprise that, for appropriate numbers of sampling vectors, the orientations that are produced by this algorithm produce the regular polyhedric arrangements – such as the tetrahedral arrangement of sp3 hybrid orbitals seen in crystals (23)). If the scan is run to completion, then the order in which the different sampling orientations are played out is irrelevant – as the complete data set is available and all points go into the estimation of the tensor. However, in subjects where it is anticipated that there may be a curtailment (for example, due to excessive motion/claustrophobia or general non-compliance), stopping a randomly ordered acquisition before completion could lead to, for example, lots of measurements being made along similar orientations – with large portions of the sampling space left out (which is sub-optimal). To counter this, Cook et al. (24) and Dubois et al. (25) have independently proposed ways of re-ordering the directions derived from the electrostatic repulsion algorithm so that, if (for example) a 30-direction sampling scheme is interrupted half-way through, the 15 directions that have been collected are still pretty much uniformly distributed. This seems an eminently sensible strategy and, if the full data set is collected anyway, has no impact on the data since, as stated before, the ordering is immaterial. Table 6.1 lists the example of 60 electrostatically arranged sampling vectors that have been ordered according to this principle and are used in our protocol. For the sake of completeness, in Tables 6.2 and 6.3, we tabulate the optimal point ordering for 18 and 30 unique sampling orientations, respectively. These were obtained with the Camino package (26), which could be used to generate other numbers of ordered point sets. 14. One is always battling with SNR issues in DT-MRI (27). Therefore, one should strive to minimise additional attenuation – consequently, whenever possible, having selected the b-value on the scanner console, one should select the ‘minimum TE’ setting. This can force the acquisition into a partial k-space acquisition, with, for example, 8 lines of k-space acquired before the echo and 48 lines acquired


137

Table 6.2 18 Electrostatically arranged and optimally ordered sampling vectors 0.7371

−0.5680

0.3662

0.7958

0.4311

0.4253

−0.8225

0.3677

0.4339

0.0006

0.9856

0.1692

0.2290

0.1508

0.9617

−0.4124

−0.7535

0.5120

−0.3586

0.2328

0.9040

−0.8912

−0.4176

0.1768

0.3199

−0.4987

0.8056

0.3099

0.6677

0.6769

0.5797

−0.8070

−0.1124

−0.2096

−0.3585

0.9097

0.9907

−0.1123

0.0774

0.1533

−0.9033

0.4008

0.5302

0.8454

0.0651

−0.2829

0.7167

0.6374

0.7201

−0.0527

0.6919

−0.7339

−0.1786

0.6554

afterwards. In some cases, it may be necessary to increase the number of lines of k-space before the centre of k to avoid corruption of the data due to excessive vibration. 15. It has been shown in several studies that diffusion MRI measurements can be severely corrupted by the effect of cardiac pulsation (28–31). As the heart beats – the pressure wave is carried through to the brain – and one gets both local deformation of the tissue (32, 33), which results in local misregistration between successive images, and intra-voxel dephasing due to diffusion (30). The latter will be interpreted as increased diffusion measured at the point a particular encoding gradient is applied – which, in turn, will lead to biases in the estimate of the diffusion tensor, resulting in inaccurate estimates of both anisotropy and fibre orientation (31). To avoid this, it is recommended that the acquisition be timed so as to avoid acquiring data during the time that the brain tissue is susceptible to this pulsation. 16. We have previously found that corruption of data due to cardiac pulsation can be avoided by waiting until at least

138

Jones and Leemans

Table 6.3 30 Electrostatically arranged and optimally ordered sampling vectors 0.0559

−0.9920

−0.1134

−0.6669

−0.6780

0.3091

0.1637

0.5330

0.8301

−0.4320

−0.0898

0.8974

0.8865

0.2190

0.4076

0.7671

−0.6107

0.1964

0.3840

−0.2940

0.8753

−0.5464

0.6192

0.5639

−0.9500

0.0308

0.3108

0.0973

−0.7119

0.6955

−0.7308

−0.6541

−0.1950

0.0777

0.0950

0.9924

−0.3115

0.9086

−0.2781

0.2961

0.8430

0.4491

−0.7523

−0.3014

0.5858

0.7882

−0.2137

0.5771

0.8458

−0.4786

−0.2358

−0.1722

−0.9086

0.3806

−0.2838

0.3816

0.8797

0.5428

0.1339

0.8291

−0.3989

−0.6000

0.6934

0.5072

−0.8486

−0.1505

−0.9375

−0.3410

0.0689

0.1467

−0.8120

−0.5649

−0.9757

0.1655

−0.1434

220 ms after the onset of the R-wave in the ECG trace recorded from chest leads (30). In turn, we have also found that the arrival of the pulse wave on a peripheral pulse-oximeter placed on the finger is 249 ± 17 ms (mean ± SD) after the R-wave of the ECG (30). Consequently, the delay needed after the peak on a peripheral oximeter trace to avoid DT-MRI data corruption is minimal. It is therefore beneficial to use a peripheral oximeter – as there is less ‘dead time’ in each pulse-to-pulse interval, and it is undoubtedly more convenient for the participant. One point to note is that since the repetition time will then vary due to natural variations in the cardiac cycle, it is important that the effective TR is at least five times the T1 of the tissue


139

of interest (to avoid partial recovery effects). However, in practice – this condition is almost always satisfied – as one can rarely squeeze in more than three or four slices into an R–R interval and, for a typical 60-slice acquisition, this means an effective TR of 15 or 20 R–R intervals. We use a look-up table to rapidly determine the optimal effective TR (Table 6.4).

Table 6.4 Prescribed effective TR as a function of participant heart rate Participant heart rate

Effective TR

95 bpm

60 R–R intervals

17. The first and most important thing that must be done is to examine the raw data. With pressures to get studies completed quickly, this is an often overlooked step in many laboratories. However, given that one is going to take multiple images and use them to compute the diffusion tensor at each voxel (1) – it is important to check that there are no corruptions in any of the individual data points. It can be extremely informative to view the data in three orthogonal planes simultaneously (see Fig. 6.1a). There are multiple tools available for doing this (for example, the popular FSLview from the FSL software library). This allows for rapid identification of slice-to-slice intensity variations (for an axially acquired data set, these will be visible on the coronal and sagittal planes). With the data set stored in a 4D format (the fourth dimension being the number of the diffusion-weighted scan), viewing the data set as a looping movie is a very efficient way of checking the data for unexpected signal dropouts (which will be visible as unusually dark horizontal bands on the sagittal and coronal slices) or other artefacts (see Fig. 6.1a). 18. It is extremely important to correct the data for subject motion and eddy current-induced distortions (e.g. 34, 35). Although the twice-refocused spin-echo sequence will ameliorate much of the eddy currents (11), there may still be residual distortions that need to be taken care of and it is unlikely that the participant will have remained perfectly still. The ‘industry standard’ approach is to use a global affine registration (12 degrees of freedom – with translation, rotation, magnification and shear along each of the

140

Jones and Leemans

principal axes) to register each diffusion-weighted volume to the volumes collected without any diffusion-weighting applied (the ‘b = 0 s/mm2 images’). Given that the image contrast is so different between the b = 0 image and those acquired at b = 1,000 s/mm2 , cost functions for the registration such as cross-correlation tend to fail and much better results are obtained with entropy-based metrics, such as the mutual information index (and its normalised version) (36, 37). Again, there are many software packages that cater to such global affine-based registration requirements, allowing one to specify the cost function. 19. Estimation of the tensor matrix requires exact knowledge of the orientation of the diffusion-encoding gradient with respect to the participant’s head (1). If the participant moves their head during the acquisition – this can be corrected with image realignment methods – as just discussed. However, simple naïve application of ‘off-the-shelf’ registration software will not account for the fact that, during such a rotation, the angle between the participant’s head and the pre-selected gradient sampling vectors will change. Failing to account for this can lead to substantial errors in estimates of anisotropy and of fibre orientation (38). Therefore, it is desirable that, when available, the gradient table that is used as input into the tensor estimation routine is modified accordingly. The amount of diffusion encoding for a particular combination of gradients is characterised by the b-matrix (1). The effect of rotation can be handled by deriving the rotational part of the transformation required to realign the images and subsequently applying this rotation to the b-matrix prior to estimation of the tensor. Figure 6.2 shows the effect of neglecting to perform this step on estimates of anisotropy and fibre orientation. 20. There are three widely used approaches to estimating the diffusion tensor from the b-matrix and the diffusionweighted data: ordinary linear least squares (OLLS), weighted linear least squares (WLLS) (1) and nonlinear least squares (NLLS) (39). (Note that there are lots more approaches in the literature – but these are the most common.) It is outside the scope of this chapter to go into the fine detail of these different approaches – but standard mathematical processing packages provide access to all of these. For the first two approaches, i.e. linear fitting, the diffusion-weighted signal intensities are first log transformed (1). In OLLS, each observation contributes equally to the fit – and thus a set of simultaneous equations relating the log of the signal to the unknown elements of the tensor are set up – and a simple matrix


141

Fig. 6.2. Directionally encoded fibre orientation maps without (a) and with (b) the required b-matrix rotation prior to estimating the diffusion tensor. The differences in orientation (colour) and fractional anisotropy (intensity) between the images is clearly visible, as shown, for instance, with the enlarged view of the posterior region of the corpus callosum.

inversion yields the unknown elements of the tensor. Given the rapid nature of this approach, it is extremely popular and is employed in several popular software packages (e.g. FSL). The consequence, however, of taking the log transformation is that, although the noise/random errors in the signal prior to the log transformation are uniform (i.e. homoscedastic) – after the log transformation, the variance in the (log-transformed) signal becomes a function of the signal itself (1, 40) and so the errors are heteroscedastic. To properly address this, a weighted linear regression (WLLS) is required, where one has to compute a covariance matrix – deriving the relationship between the variance in the logtransformed and non-log transformed data – and include it in the regression step (1). Although this makes computation longer, the results are far more robust and WLLS is to be preferred over OLLS (39). In NLLS, on the other hand, there is no log transformation of the signal – and thus the errors remain homoscedastic, so the covariance matrix is a multiple of the identity matrix and can effectively be factored out of the analysis, which is an advantage over the linear framework approaches. While NLLS is attractive in that it fits the model to the data directly, producing results that are superior to WLLS (and therefore to OLLS) (39), the computational time is considerably longer – and special care has to be taken that the fitting algorithm has not been trapped in a local extremum. 21. Once one has fitted the model, it is expected that any differences between the observed signal and those predicted by the model (i.e. the ‘residuals’) should be random and

142

Jones and Leemans

should not contain any structure. Deviations will rapidly identify points that are corrupt (41) (see Fig. 6.1b). Therefore, before proceeding with further analyses, we recommend inspecting a map of the residuals to ensure that there is nothing unexpected in the data. 22. The final stage in a standard DT-MRI pipeline is to derive parameters of interest from the diffusion tensor. These are invariably the mean diffusivity (or trace of the diffusion tensor) (2), a measure of anisotropy (27) – and the principal diffusion orientation, which can be used for creation of directionally encoded colour (DEC) maps (5) or for fibre-tracking analyses (6–10). With regard to the choice of anisotropy index, there is a plethora of indices in the literature (20). However, by far the most popular is the fractional anisotropy, derived from the three eigenvalues of the diffusion tensor. One can show that this has a better signalto-noise ratio characteristic than other popular indices such as relative anisotropy (42). 23. It is always prudent to look carefully at the resultant parametric maps before inputting them into any form of analysis. Values of anisotropy greater than 1 are physically nonmeaningful, since they are designed to lie between 0 and 1 – and will result when one or more of the eigenvalues is negative. Again, this is non-physically meaningful – but can occur when the diffusion-weighted signal is higher in intensity than the non-diffusion weighted signal. In turn, this may arise in regions that are particularly noisy – or where there is insufficiently corrected misregistration. A similar tell-tale sign for artefact is to examine the rim of the anisotropy maps. A high anisotropy around the rim of the adult mammalian brain is not expected in practice (Fig. 6.1c). Given that high anisotropy means that the DW signal varies rapidly with the direction of the diffusionencoding gradient, the appearance of such a bright rim would be consistent with misregistration of, for instance, CSF and gray matter with either white matter or perhaps the edge of the brain. References 1. Basser, P. J., Mattiello, J., LeBihan, D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994;66: 259–267. 2. Pierpaoli, C., Jezzard, P., Basser, P. J., Barnett, A., Di Chiro, G. Diffusion tensor MR imaging of the human brain. Radiology 1996;201:637–648.

3. Stejskal, E. O., Tanner, J. E. Spin diffusion measurements: Spin echoes in the presence of a time-dependent field gradient. J Chem Phys 1965;42:288–292. 4. Moseley, M. E., Cohen, Y. C., Kucharczyk, J., Asgari, H. S., Wendland, M. F., Tsuruda, J., Norman, D. Diffusion-weighted MR imaging of anisotropic water diffusion


5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

in cat central nervous system. Radiology 1990;176:439–445. Pajevic, S., Pierpaoli, C. Color schemes to represent the orientation of anisotropic tissues from diffusion tensor data: Application to white matter fiber tract mapping in the human brain. Magn Reson Med 1999;42:526–540 (Erratum in: Magn Reson Med 2000 43:921). Mori, S., Crain, B. J., Chacko, V. P., van Zijl, P. C. Three-dimensional tracking of axonal projections in the brain by magnetic resonance imaging. Ann Neurol 1999;45: 265–269. Jones, D. K., Simmons, A., Williams, S. C., Horsfield, M. A. Non-invasive assessment of axonal fiber connectivity in the human brain via diffusion tensor MRI. Magn Reson Med 1999;42:37–41. Conturo, T. E., Lori, N. F., Cull, T. S., Akbudak, E., Snyder, A. Z., Shimony, J. S., McKinstry, R. C., Burton, H., Raichle, M. E. Tracking neuronal fiber pathways in the living human brain. Proc Natl Acad Sci USA 1999;96:10422–10427. Basser, P. J., Pajevic, S., Pierpaoli, C., Duda, J., Aldroubi, A. In vivo fiber tractography using DT-MRI data. Magn Reson Med 2000;44:625–632. Parker, G. J. M., Haroon, H. A. et al. A framework for a streamline-based probabilistic index of connectivity (PICo) using a structural interpretation of MRI diffusion measurements. J Magn Reson Imag 2003;18:242–254. Reese, T. G., Heid, O., Weisskoff, R. M., Wedeen, V. J. Reduction of eddy-currentinduced distortion in diffusion MRI using a twice-refocused spin echo. Magn Reson Med 2003;49:177–182. Bammer, R., Auer, M., Keeling, S. L., Augustin, M., Stables, L. A., Prokesch, R. W., Stollberger, R., Moseley, M. E., Fazekas, F. Diffusion tensor imaging using single-shot SENSE-EPI. Magn Reson Med 2002;48:128–136. Hiltunen, J., Hari, R., Jousmäki, V., Müller, K., Sepponen, R., Joensuu, R. Quantification of mechanical vibration during diffusion tensor imaging at 3 T. Neuroimage 2006;32: 93–103. Pruessmann, K. P., Weiger, M., Scheidegger, M. B., Boesiger, P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med 1999;42:952–962. Griswold, M. A., Jakob, P. M., Heidemann, R. M., Nittka, M., Jellus, V., Wang, K., Kiefer, B., Haase, A. Generalized autocalibrating partially parallel

16.

17. 18.

19.

20.

21.

22.

23.

24.

25.

26.

27. 28.

143

acquisitions (GRAPPA). Magn Reson Med 2002;47:1202–1210. Edelstein, W. A., Bottomley, P. A., Pfeifer, L. M. A signal-to-noise calibration procedure for NMR imaging systems. Med Phys 1984;11:180–185. Gudbjartsson, H., Patz, S. The Rician distribution of noisy MRI data. Magn Reson Med 1995;34:910–914. Dietrich, O., Heiland, S., Sartor, K. Noise correction for the exact determination of apparent diffusion coefficients at low SNR. Magn Reson Med 2001;45:448–453. Jones, D. K., Basser, P. J. “Squashing peanuts and smashing pumpkins”: How noise distorts diffusion-weighted MR data. Magn Reson Med 2004;52:979–993. Basser, P. J., Pierpaoli, C. Microstructural and physiological features of tissue elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B 1996;111:209–219. Jones, D. K., Horsfield, M. A., Simmons, A. Optimal strategies for measuring diffusion in anisotropic systems by magnetic resonance imaging. Magn Reson Med 1999;42: 515–525. Jones, D. K. The effect of gradient sampling schemes on measures derived from diffusion tensor MRI: A Monte Carlo study. Magn Reson Med 2004;51:807–815. Conturo, T. E., McKinstry, R. C., Akbudak, E., Robinson, B. H. Encoding of anisotropic diffusion with tetrahedral gradients: A general mathematical diffusion formalism and experimental results. Magn Reson Med 1996;35:399–412. Cook, P. A., Symms, M., Boulby, P. A., Alexander, D. C. Optimal acquisition orders of diffusion-weighted MRI measurements. J Magn Reson Imaging 2007;25: 51–58. Dubois, J., Poupon, C., Lethimonnier, F., Le Bihan, D. Optimized diffusion gradient orientation schemes for corrupted clinical DTI data sets. Magma 2006;19: 134–143. Cook, P. A., Bai, Y., Nedjati-Gilani, S., Seunarine, K. K., Hall, M. G., Parker, G. J., Alexander, D. C. (2006) Camino: OpenSource Diffusion-MRI Reconstruction and Processing, 14th Scientific Meeting of the International Society for Magnetic Resonance in Medicine, Seattle, WA, USA, p. 2759, May 2006. Pierpaoli, C., Basser, P. J. Toward a quantitative assessment of diffusion anisotropy. Magn Reson Med 1996;36:893–906. Turner, R., Le Bihan, D., Maier, J., Vavrek, R., Hedges, L. K., Pekar, J. Echo-planar

144

29.

30.

31.

32.

33.

34.

35.

Jones and Leemans imaging of intravoxel incoherent motions. Radiology 1990;177:407–414. Skare, S., Andersson, J. L. On the effects of gating in diffusion imaging of the brain using single shot EPI. Magn Reson Imaging 2001;19:1125–1128. Pierpaoli, C., Marenco, S., Rohde, G., Jones, D. K., Barnett, A. S. (2003). Analyzing the contribution of cardiac pulsation to the variability of quantities derived from the diffusion tensor in “Proc. ISMRM 11th Annual Meeting, Toronto” p. 70. Jones, D. K., Pierpaoli, C. (2005) The contribution of cardiac pulsation to variability in tractography results. In “Proc. ISMRM 13th Annual Meeting, Miami”, p. 225. Poncelet, B. P., Wedeen, V. J., Weisskoff, R. M., Cohen, M. S. Brain parenchyma motion: Measurement with cine echoplanar MR imaging. Radiology 1992;185: 645–651. Enzmann, D. R., Pelc, N. J. Brain motion: Measurement with phase-contrast MR imaging. Radiology 1992;185: 653–660. Rohde, G. K., Barnett, A. S., Basser, P. J., Marenco, S., Pierpaoli, C. Comprehensive approach for correction of motion and distortion in diffusion-weighted MRI. Magn Reson Med 2004;51:103–114. Andersson, J. L., Skare, S. A modelbased method for retrospective correction of geometric distortions in diffusionweighted EPI. Neuroimage 2002;16: 177–199.

36. Maes, F., Collignon, A., Vandermeulen, D., Marchal, G., Suetens, P. Multimodality image registration by maximization of mutual information. IEEE Trans Med Imaging 1997;16:187–198. 37. Studholme, C., Constable, R. T., Duncan, J. S. Accurate alignment of functional EPI data to anatomical MRI using a physics-based distortion model. IEEE Trans Med Imaging 2000;19:1115–1127. 38. Leemans, A., Jones, D. K. The B-matrix must be rotated when motion correcting diffusion tensor imaging data. Magn Reson Med 2009;61:1336–1349. 39. Koay, C. G., Chang, L. C., Carew, J. D., Pierpaoli, C., Basser, P. J. A unifying theoretical and algorithmic framework for least squares methods of estimation in diffusion tensor imaging. J Magn Reson 2006;182: 115–125. 40. Bevington, P. R., Robinson, D. K. Data Reduction and Error Analysis for the Physical Sciences, 2nd ed. New York, NY: McGrawHill; 1992. 41. Leemans, A., Evans, C. J., Jones, D. K. (2008). Quality assessment through analysis of residuals of diffusion image fitting. In “Proc. ISMRM 16th Annual Meeting, Toronto”. p. 3300. 42. Hasan, K. M., Alexander, A. L., Narayana, P. A. Does fractional anisotropy have better noise immunity characteristics than relative anisotropy in diffusion tensor MRI? An analytical approach. Magn Reson Med 2004;51:413–417.

Chapter 7 Manganese-Enhanced Magnetic Resonance Imaging (MEMRI) Cynthia A. Massaad and Robia G. Pautler Abstract The use of manganese ions (Mn2+ ) as an MRI contrast agent was introduced over 20 years ago in studies of Mn2+ toxicity in anesthetized rats (1). Manganese-enhanced MRI (MEMRI) evolved in the late nineties when Koretsky and associates pioneered the use of MEMRI for brain activity measurements (2) as well as neuronal tract tracing (3). Currently, MEMRI has three primary applications in biological systems: (1) contrast enhancement for anatomical detail, (2) activity-dependent assessment and (3) tracing of neuronal connections or tract tracing. MEMRI relies upon the following three main properties of Mn2+ : (1) it is a paramagnetic ion that shortens the spin lattice relaxation time constant (T1 ) of tissues, where it accumulates and hence functions as an excellent T1 contrast agent; (2) it is a calcium (Ca2+ ) analog that can enter excitable cells, such as neurons and cardiac cells via voltage-gated Ca2+ channels; and (3) once in the cells Mn2+ can be transported along axons by microtubule-dependent axonal transport and can also cross synapses trans-synaptically to neighboring neurons. This chapter will emphasize the methodological approaches towards the use of MEMRI in biological systems. Key words: MEMRI, rodents, manganese, central nervous system, contrast agent, MRI.

1. Introduction Mn2+ is a trace element essential for normal body function and development throughout the lifespan of mammals (4). Most notably Mn2+ is an essential cofactor for several enzymes responsible for a wide variety of physiological body functions (4). Such enzymes include manganese superoxide dismutase (5) which is essential for oxidative stress prevention, pyruvate carboxylase (6) which plays a critical role in gluconeogenesis, arginase (7) which is involved in urea production by the liver, and glutamine synthetase M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_7, © Springer Science+Business Media, LLC 2011

145

146

Massaad and Pautler

(8), an astrocyte-specific enzyme regulated by about 80% of brain Mn2+ . The importance of Mn2+ is illustrated by studies linking disruption of Mn2+ homeostasis to disease occurrence (5, 9). Mn2+ deficiency has been linked to deficient bone metabolism in rats (10), as well as skin lesions, bone malformation, epileptic seizures, and increased Ca2+ and phosphorous levels in humans (11). Although Mn2+ deficiency is clearly associated with adverse effects, the aforementioned studies were achieved with an artificially induced Mn2+ -deficient diet (11). No naturally occurring Mn2+ -deficiency related diseases have been observed. However, Mn2+ is more frequently of toxicological concern. Although it is considered the least toxic of all essential elements (12), excessive exposure to the metal leads to central nervous system toxicity (4). It has been shown that Mn2+ can enter the central nervous system either directly via the olfactory receptor neurons or through the blood brain barrier by diffusion or active transport (13, 14). Once in the nervous system, Mn2+ is transported along neurons by microtubule-dependent axonal transport (15, 16) and can traverse synapses to accumulate in neighboring neurons (17, 18). The resulting neurotoxicity preferentially targets the striatum leading to Parkinson’s disease-like symptoms, including generalized bradykinesia, widespread rigidity, tremors, hallucinations, and memory loss (4, 19, 20). In addition to multiple roles in normal physiology, Mn2+ is also a Ca2+ analog and can enter excitable cells via several types of Ca2+ channels such as voltage-gated Ca2+ channels and the Na+ /Ca2+ exchanger (21–24). Mn2+ also accumulates in mitochondria via the mitochondrial Ca2+ uniporter (25, 26). The analogy of Mn2+ with Ca2+ resulted in the use of Mn2+ as a fura-2 quencher and hence Ca2+ indicator in biological systems by fluorescence microscopy (27–31). Another very important feature of Mn2+ is that it is paramagnetic and produces MR contrast by causing a strong reduction in the T1 relaxation times of water (32–35). Positive contrast is detected in T1 -weighted images of tissues where Mn2+ accumulates (32–35). The combined physical and biological properties of Mn2+ make it a useful contrast agent for anatomical and functional imaging in multiple systems. Indeed, manganese-enhanced MRI (MEMRI) has been gaining growing interest in the past few years (2, 3, 36, 37) and currently has three main applications for biological systems. First, owing to its contrast-enhancing properties, systemic Mn2+ injections are used for enhancement of the brain cytoarchitecture for anatomical studies (38–44). This technique has been used in adult, as well as in young developing organisms. Its use has further been extended to studying the development of embryos in utero (45). Second, given that Mn2+ can enter cells via voltage-gated Ca2+ channels, it is used as a

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

147

marker of activity in specific protocols that promote its accumulation in active brain areas (2, 46–52). This use of MEMRI is termed activation-induced MEMRI or AIM-MRI. AIM-MRI also has applications in the heart because of the high concentration of Ca2+ channels (53). This protocol, however, will emphasize MEMRI applications in the nervous system. The third and last application of MEMRI is tract tracing; given that Mn2+ is transported by microtubule-dependent axonal transport and can cross synapses to reach post-synaptic neurons, MEMRI has been used as a neuronal tract tracer for several neuronal pathways including the visual, olfactory, and somatosensory pathways, in a variety of animal models, such as mice, rats, monkeys, and birds (3, 54–63). This review will focus on MEMRI applications in rodents. The versatility of MEMRI is also demonstrated by the development of methods for dynamic Mn2+ transport imaging, which are proving as useful markers of disease and related therapy (64, 65). The following chapter will expand upon each of the three applications of MEMRI with special emphasis on techniques related to each application.

2. Materials 2.1. Anatomical Contrast Enhancement 2.1.1. Intravenous MnCl2

1. MnCl2 as a source of Mn2+ 2. Sterile water 3. Sterile saline 4. Beaker of warm water for tail warming and dilation of the tail vein 5. 27- or 30-gauge needle 6. Forceps 7. 1-ml syringes 8. Tubing suitable to attach to a 27-gauge needle 9. Pre-anesthetic (e.g., glycopyrrolate) 10. Anesthetic (e.g., isoflurane) 11. Analgesic (e.g., bupivicaine) 12. Sterile saline 13. Tape 14. 4-gauge nylon suture 15. Syringe pump

148

Massaad and Pautler

16. Warming blanket 17. Small animal-monitoring system complete with rectal temperature probe and respiration sensor 18. Neuromuscular blocking agent (pancuronium bromide or gallamine triethiodide) 19. Ventilator 2.2. ActivationInduced MRI 2.2.1. Intravenous MnCl2

see Section 2.1.1.

2.2.2. Blood Brain Barrier (BBB) Disruption 2.2.2.1. Hyperosmolar Mannitol Infusion:

2.2.2.2. Hyperosmolar Mannitol Injection Through the External Carotid Artery:

• This requires the infusion of mannitol through the tail vein (or femoral vein). Materials will be identical to Section 2.1 with the exception of using a 5–10% mannitol solution instead of the MnCl2 solution (66–69). 1. Anesthesia (e.g., isoflurane, urethane or α-chloralose) 2. Tape 3. Hair clipper (Note 1) 4. Surgical tools (blade with holder, hemostat, forceps, scissors etc.) 5. Microvascular clips 6. Disinfecting solutions (e.g., betadine, chlorhexidine, alcohol) 7. Polyethylene tubing PE-50, thinned to an outer diameter of ∼0.4 mm (Note 2) 8. PE90 tube attached to the hub of a needle 9. 6-0 nylon suture 10. Mannitol solution (20%) 11. Small metal laryngoscope 12. Muscular blocking agent (pancuronium bromide or gallamine triethiodide) 13. Ventilator

2.2.3. Intraperitoneal MnCl2 – Visual and Auditory Activation

1. MnCl2 as a source of Mn2+ (66 mg/kg in saline – for intraperitoneal injections) (48, 70). 2. 1-ml syringe. 3. For auditory activation studies: auditory isolation box enabled for auditory stimulation with the addition of a sound synthesizer, audio amplifier, and speakers.


149

4. For visual activation studies: visual stimulation box consisting of four walls made up of 14-inch computer screens. Remaining areas consist of black-painted wood protected by aluminum mesh. 5. Anesthesia (isoflurane or urethane). 2.2.4. Intranasal MnCl2 – Olfactory System Activation

1. MnCl2 as a source of Mn2+ (10 mM in H2 O – for intranasal administration of 7 μl/naris or 1.5 M in H2 O for aerosolized Mn2+ administration with a vaporizer) (52, 61) 2. Pipette – 10 μl 3. Anesthesia (e.g., isoflurane, urethane) 4. Heating pad 5. Odorant for olfactory stimulation (e.g., 1:100 amyl acetate, 1:10 octanal, 1:10 carvone etc.) 6. Vaporizer 7. Fume hood

2.3. Tract Tracing 2.3.1. Tract Tracing – Visual System

1. MnCl2 as a source of Mn2+ (1 M in H2 O – for intravitreal injections) (36, 62) 2. 27-gauge needle 3. Polyethylene tubing (0.4 mm diameter) 4. A 5-μl Hamilton syringe 5. Anesthesia (isoflurane or ketamine/xylazine combination or pentobarbital sodium, see Note 11) 6. Heating pad 7. Dissecting microscope

2.3.2. Tract Tracing – Olfactory System

1. MnCl2 as a source of Mn2+ (3.79 M in H2 O – for intranasal administration) (61, 65) 2. Pipette – 2 μl or 10 μl 3. Anesthesia (e.g., isoflurane) 4. Heating pad

2.3.3. Tract Tracing – Deep Brain Structures

1. MnCl2 as a source of Mn2+ (5 mM in H2 O – for intracranial injections) (36, 63, 71, 72) 2. Anesthesia (e.g., ketamine/xylazine combination as a preoperative followed by 2% isoflurane for maintenance) 3. Surgical tools (blade with holder, hemostat, forceps, scissors etc.) 4. Disinfecting solution (chlorhexidine, betadine, alcohol) 5. Small sharp scissors and/or rodent hair clipper

150

Massaad and Pautler

6. Small drill with associated bits (similar to a dental drill) 7. 6-0 nylon suture 8. Mouse/rat brain atlas 9. Capillary tube puller 10. Quartz capillary tubes with filament 11. Surgical area including stereotaxic holder, dissecting microscope and gaseous anesthesia line 12. Surgical tool sterilizer (e.g., glass beads electric sterilizer) 13. Picospritzer with holder and push/pull options (to fill injection needles and subsequently inject solution out of them) 14. Heating pad 15. Sterile cotton swabs 16. Eye ointment 17. Leveling tool (small fork-shaped metallic tool that can be used to ascertain 2D horizontal leveling of the mouse/rat head in the stereotaxic holder) 18. Calibrated volume gauge (Note 3)

3. Methods 3.1. Anatomical Contrast Enhancement

Anatomical contrast enhancement by systemic Mn2+ injection has been studied in rodents (38–40, 43, 44), birds (57, 58), and primates (54, 55). The methods presented here are specifically designed for rat brain visualization based upon work developed in Koretsky’s laboratory (44). The following methods can be adapted for use with any organism, provided reasonable optimization is conducted on the organism of interest as well as magnetic field strength.

3.1.1. Preparation of the MnCl2 Solution

Different concentrations of MnCl2 can be used in systemic injections for positive contrast in T1 -weighted images (37, 44). Optimization, with regards to the animal model used, as well as available MRI hardware, should be performed for the best results. Also, when preparing MnCl2 , care should be taken as to the tonicity and pH of the final solution. The body fluid has an osmolarity of 300 mOsm/l. One mole of MnCl2 is equivalent to 3 Osm. Therefore, concentrations in the range of 100 mM should be used to insure proper tonicity when large amounts of MnCl2 are to be infused to the animals. When adjusting the pH of MnCl2 solution to a physiological pH of 7.4, bicine buffer, equilibrated with


151

NaOH, is a good buffer. Following are guidelines for the preparation of 100 mM MnCl2 at pH 7.4, suitable for imaging of rat cytoarchitecture by systemic MnCl2 injection. 1. Dissolve 1.63 g bicine (FW=163.17) in 100 ml water to obtain a 100 mM bicine solution. 2. Bring solution to pH 7.4 using NaOH. 3. Sterilize solution (either by autoclaving or by filtering). 4. Dissolve 98.95 mg of MnCl2 .4H2 0 (FW=197.91) into 5 ml of sterile bicine solution. Depending on the weight of the animals used, 2–4 ml will be enough for imaging one animal; increase the volume of solution according to the number of animals to be imaged. 3.1.2. Intraperitoneal MnCl2 Injection

3.1.3. Setting Up a Tail Vein Line

Systemic administration of MnCl2 by intraperitoneal injection consists of one injection of 100 mM MnCl2 at a dose of 66 mg/kg. Imaging can be performed as early as 3 h and up to 24 h post-injection. 1. Anesthetize the rat with 4% isoflurane in O2 initially and then keep it anesthetized with 1.5–2% isoflurane using a facemask. 2. Using forceps, break the metallic part of the 27-gauge needle away from its plastic base connector. Take care to avoid causing the needle to get blocked. Connect the metallic part of the needle to its plastic base using a piece of tubing long enough to allow you to comfortably place the connected syringe onto the pump. 3. Fill 1-ml syringe with sterile physiological saline and attach it to the needle/tubing combination. 4. Immerse the tail in warm water to dilate the tail vein. 5. Insert the tip of the needle into the vein; proper insertion is confirmed by the backflow of blood from the tail into the saline-filled attached tubing. 6. To fix the needle in proper place, first use tape over the tubing to loosely hold everything in place. Then, using the nylon suture, tie a knot around the tail/metallic part of the needle. The suture will not go through the skin. 7. Carefully remove the saline-filled syringe and replace it with a MnCl2 pre-filled syringe. 8. Place the MnCl2 syringe into the holder of the syringe pump and set the infusion rate to 1.8 ml/h. Do not start the infusion yet.

3.1.4. MnCl2 Infusion

1. Inject 0.01 mg/kg glycopyrrolate intramuscularly. Glycopyrrolate is a muscarinic cholinergic blocker used as a

152

Massaad and Pautler

pre-anesthetic medication to diminish the risk of vagal inhibition to the heart. 2. Insert a rectal probe into the rat, and maintain the temperature at 37.5◦ C during the infusion using an animal-heating system (e.g., warming blanket or heated air). 3. Keep the anesthesia light during the infusion (0.5–1% isoflurane). 4. The goal MnCl2 concentration is 175 mg/kg, which amounts to approximately 2 ml of total volume per animal (Note 4). 5. To avoid dehydration, inject sterile saline subcutaneously (6.7 ml/100 g) immediately and 6 h after the MnCl2 infusion. 6. Keep the animals under controlled temperature for up to 24 h post-infusion. It is normal for the animals to display lethargic behavior at the end of the MnCl2 infusion. Their behavior will gradually improve to normal by 24 h postinfusion. 3.1.5. Animal Preparation for MRI

1. Anesthetize rats with 4% isoflurane initially. 2. Intubate the animals and keep them ventilated with 1.5% isoflurane in O2 (see Section 3.2.1.1. for detailed intubation protocol). 3. Maintain body temperature at 37.5◦ C using an animalheating system. 4. Monitor temperature, blood pressure, and respiration rate with a small animal physiological monitoring system. 5. Inject the animals with pancuronium bromide (2.5 mg/kg) intraperitoneally to suppress motion (an alternative neuromuscular blocking agent is gallamine triethiodide 80 mg/kg i.v.).

3.1.6. Imaging Parameters

It should be noted that imaging protocols and parameters will vary considerably depending on the field strength to be used. Reported below are the optimal imaging parameters for proton MRI on an 11.7 T magnet based upon the work done by Aoki et al (44). The following parameters can be used as a starting point; however, further optimization should be performed for different field strengths and imaging protocols. Two-dimensional multi-slice multi-echo (MSME-2D) Repetition time (TR) = 300 ms Echo time (TE) = 10.5 ms Matrix size = 256 × 256


153

Field of view (FOV) = 25.6 × 25.6 mm Slice thickness (ST) = 1 mm Number of averages (NEX) = 8 An inversion recovery sequence could be used to acquire T1 weighted images as well. The parameters are as follows: Inversion time = 1,100 ms TR = 4,000 ms TE = 11.2 ms Matrix size = 512 × 256 FOV = 38.4 × 19.2 mm ST = 1 mm NEX = 1 Three-dimensional spin echo (SE-3D) TR = 250 ms TE = 7.3 ms Matrix size = 256 × 256 × 128 FOV = 19.2 × 19.2 × 9.6 mm NEX = 2 Total acquisition time = 273 min 3.1.7. Expected Results

The expected results are an increase in positive contrast enhancement in the central nervous system (Fig. 7.1). The pattern of enhancement can be obtained over a wide range of MnCl2 concentrations, with regions of the brain lacking a blood brain barrier (BBB), such as the pituitary gland, exhibiting stronger enhancement at low doses. Regions with an intact BBB, such as the hippocampus and cortex show a dose-dependent increase in contrast enhancement. Higher doses may even allow the detection of finer details of the neuroarchitecture, such as cortical laminae structure.

3.2. AIM-MRI

Activation-induced MRI (AIM) is essentially a Mn2+ -based functional MRI paradigm. It was introduced in 1997 by Lin and Koretsky as a blood-flow independent alternative to functional MARI (fMRI) (2). This method is based on the following two essential properties of Mn2+ : (1) Mn2+ can enter the brain parenchyma from the blood via a disrupted (or leaky) blood brain barrier (BBB) (66) and (2) Mn2+ is a Ca2+ analog that can enter neurons via voltage-gated Ca2+ channels and accumulates in neurons in an activity-dependent manner (30, 73–75). Mn2+ ions cannot efficiently enter the brain parenchyma through an intact BBB (66). Some diffusion may occur at the blood/CSF interface in choroid plexuses, but the amount of Mn2+ entering the brain is minimal compared to the cases where the BBB is disrupted

154

Massaad and Pautler

Fig. 7.1. T1 -weighted MRI after systemic MnCl2 administration in the rat. T1 -weighted MRI of a control rat (column A) and a rat 1 day after IV infusion of MnCl2 solution (column B). Top row shows transverse slices at the level of the olfactory bulb (OB, Bregma: +7 mm). The middle row shows horizontal slices including the hippocampal formation (Bregma: –6 mm). The bottom row shows sagittal slices. The signal intensity of the T1 weighted MRI was enhanced prominently 1 day after systemic administration of MnCl2 in the rat. There were characteristic signal enhancements that were large in the olfactory bulb (OB), hippocampus, cerebellum, and pituitary. Reprinted from Aoki et al. (44), copyright 2004, with permission from Elsevier.

(1, 76, 77). As a result, most AIM studies to date were performed in conjunction with BBB disruption (66). Some studies on the activation of the auditory (48, 78, 79) and visual pathways (70, 80) following auditory and visual stimulation respectively were performed in mice without disruption of the BBB. Also, a subset of functional studies in the olfactory system, capitalized on the active entry of Mn2+ through Ca2+ channels and its trans-synaptic


155

transport (61). These studies were a spin-off from tract-tracing studies and were also performed without BBB disruption. Several groups have used this technique since 1997 with certain variations regarding the route of administration of MnCl2 as well as the paradigm followed to break the BBB. Protocols are presented here for intravenous or intraperitoneal MnCl2 administration with hyperosmolar mannitol injections and infusions to break the BBB. Protocols involving olfactory, visual, or auditory activation without BBB disruption are also presented. 3.2.1. Intravenous Infusion of MnCl2 with BBB Disruption

3.2.1.1. Direct Oral/Tracheal Intubation for Artificial Ventilation (Note 5)

The protocol for preparing MnCl2 and setting up a tail vein line for intravenous MnCl2 infusion was described in Sections 3.1.1– 3.1.3. Here, a protocol for disrupting the BBB in conjunction with MnCl2 infusion is described. 1. Anesthetize the mouse/rat with an intraperitoneal injection of α-chloralose and urethane combination (Note 6). 2. Upon lack of toe pinch reflex, place the animal in dorsal position on a pre-heated warming pad. 3. Tape the limbs down. 4. Pull the head back by placing 4-0 nylon behind the upper incisors. 5. Use a small metal laryngoscope to pull the lower jaw down and expose the tracheal opening. 6. Insert a ∼2-cm long PE-90 tubing attached to the hub of a needle about 3 mm into the trachea. The tip of the tube should be beveled in the direction of the natural bend of the tubing to avoid any tissue damage during insertion (Note 7). 7. Attach the needle hub to a ventilator and set it to 80–100 breaths/min. (Do not turn the ventilator on yet; this will occur when imaging begins.)

3.2.1.2. Catheterization of the External Carotid Artery for Hyperosmolar Mannitol Injection

1. While the animal is still in dorsal position from the previous step, shave the neck area with a hair clipper (for rats) or sharp small scissors (for mice) (Note 1). 2. Disinfect the operating field with betadine and 75% alcohol. 3. Make a vertical 5-cm incision and expose the arteries. 4. Temporarily clamp the right common and internal carotids using the microvascular clamps. 5. Carefully insert the PE50 tubing into the external carotid artery through a small puncture in the retrograde direction (Note 8).

156

Massaad and Pautler

6. Secure the tubing in place with 6-0 nylon suture. Tighten the sutures around the tube very well to prevent any further bleeding from the artery. 7. Remove the microvascular clamps. 8. Sew the wound back roughly keeping the tubes well in place. 9. The mouse is now ready to be placed in the magnet. 3.2.1.3. Animal Preparation for MRI

1. Place the animal in a suitable holder with bite bars. 2. Set up a tail vein line for MnCl2 infusion as described in Section 3.1.2 (Note 9). 3. Extend all tubing outside of the magnet room. 4. Maintain anesthesia with 2% isoflurane. 5. Inject the animal with pancuronium bromide (2.5 mg/kg) intraperitoneally to suppress motion (an alternative neuromuscular blocking agent is gallamine triethiodide 80 mg/kg i.v.). 6. Start the ventilator for artificial breathing (80–100 breaths/min).

3.2.1.4. Experimental Protocol

Keeping in mind that there are several different variations of experimental protocols, a typical setting for AIM experiments is given below. 1. Acquire a series of baseline scans. 2. Start the MnCl2 infusion (typically infusions last about 1 h). Care should be taken in selecting the concentration of MnCl2 for AIM studies. MnCl2 causes toxicity to the heart due to Mn2+ homology to Ca2+ . A very common initial effect of MnCl2 is a drop in blood pressure that is typically recovered to normal within 10 min. Therefore, the concentration of MnCl2 may not have drastic effects in long-term studies, but it does affect short-term studies such as AIM. A concentration of 0.2 mmol/kg infused over an hour has been shown to work well. 3. About halfway through the infusion, inject a bolus of 25% mannitol (5–7 ml/kg) through the external carotid catheter. Keep the room temperature higher than 25◦ C and pre-warm all tubing and attachments related to the mannitol injection. This is necessary to prevent mannitol recrystallization and subsequent micro-infarcts (Note 10). Unilateral injection usually results in the opening of the BBB on the side of the injection, with the contralateral receiving less Mn2+ . The inhomogeneous opening of the BBB is not very well understood, but modulating some factors such as dose of mannitol, injection rate, distance of the injection site from the artery, age of the animal etc. plays an important role


157

in obtaining sufficient BBB disruption for suitable contrast achievement. 4. At the completion of the MnCl2 infusion, administer the activating signal specific for the AIM study at hand (pharmacological as well as behavioral stimuli have been described). 5. Begin T1 -weighted image acquisitions at different time intervals to measure activation of specific brain areas involved in your study. 3.2.1.5. Imaging Parameters

Different imaging paradigms are possible for nervous system activity measurements using MEMRI in conjunction with BBB disruption. The imaging parameters will depend on the area of interest, organism used, as well as magnetic field strength. Optimization with regards to all of these variables is required for best results. Following is an example of imaging parameters, excerpted from Weng et al (50), for imaging cortical activity in rats following whisker stimulation at 3T. Multi-slice spin-echo sequence TR = 500 ms TE = 10 ms In-plane resolution = 187 μm ST = 1.5 mm

3.2.1.6. Expected Results

Depending upon the system under study, one can expect to detect increased signal enhancement in areas of the brain involved in the activity studied. For example, following whisker stimulation, Weng et al. observed signal enhancement in the right cortical barrels of the rat brain (Fig. 7.2) (50).

Fig. 7.2. Three consecutive slices of the averaged Mn2+ -enhanced T1 WIs under urethane anesthesia a. Mn2+ enhancement was observed in right cortical barrels. The color maps of the averaged Mn2+ -enhanced T1 WIs b. Reprinted from Weng et al. (50), copyright 2007, with permission from Elsevier.

158

Massaad and Pautler

3.2.2. Activation Studies Without BBB Disruption – Visual and Auditory Systems 3.2.2.1. Activation Studies in the Auditory System 3.2.2.1.1. MnCl2 Administration

1. Administer 66 mg/kg MnCl2 as an intraperitoneal injection. 2. Place the animals in normal isolation cages with free access to food and water. 3. Place the isolation cage (with the animal in it) inside an auditory isolation box. 4. Subject the animals to different sound stimulation protocols over a period of 24 h. 5. Upon completion of the auditory stimulation paradigm, initially anesthetize the animals with 5% isoflurane followed by maintenance with 2% isoflurane and acquire T1 -weighted images. 6. A target region of interest for MEMRI quantification is in the auditory nuclei.

3.2.2.1.2. Imaging Parameters

Different imaging paradigms are possible for activity measurements from the auditory system using MEMRI. Following are optimal imaging parameters for the acquisition of T1 -weighted images of the mouse brain at 7T. These parameters are adapted from the work of Daniel Turnbull and associates (48) studying nerve activity in the auditory pathway of mice and are meant to be a starting guide. Further optimization is required depending upon the animal model used as well as the magnetic field strength. 3D gradient echo pulse sequence TR = 50 ms TE = 4 ms Flip angle = 65◦ Total imaging time = 1 h 50 min per mouse This sequence yields a volumetric image set covering the whole brain, with an isotropic resolution of 100 μm

3.2.2.1.3. Expected Results

With the described auditory stimulation protocol combined with a systemic MnCl2 administration, one can expect to observe significant Mn2+ enhancement in structures of the auditory system, including the auditory nuclei in the brainstem and the thalamus (Fig. 7.3). Although the auditory cortex would also be expected to show enhancement, this particular protocol does not lead to


159

Fig. 7.3. MEMRI enhancement in brainstem auditory nuclei was altered in mice with conductive hearing loss (CHL). Comparisons of individual mice with bilateral CHL (bi-CHL) (a), mice with unilateral CHL (uni-CHL) (b), and normal mice (c) demonstrated marked differences in MEMRI signals in the cochlear nucleus (CN) (arrow heads) and inferior colliculus (IC) (arrows), but not in non-auditory caudate putamen (CPu). Adapted by permission from (48), copyright 2005.

signal enhancement in the auditory cortex, perhaps due to less Mn2+ reaching its remote location (48). Nonetheless, this technique is a very useful approach for studying neural activation in the auditory system and can be applied to different animal models of diseases involving the auditory system (e.g. Fig. 7.3). 3.2.2.2. Activation Studies in the Visual System 3.2.2.2.1. MnCl2 Administration

1. Prior to the start of the procedure, house the animals in darkness for 8–12 h. All subsequent procedures and animal handling should be performed in darkness or under very dim red light. 2. Administer 66 mg/kg MnCl2 as an intraperitoneal injection. 3. Place the animal in the stimulation chamber for 8 h. Stimulation consists of a moving square wave (as opposed to constant diffuse light) to avoid habituation. 4. Upon completion of the stimulation paradigm, anesthetize the animals with isoflurane or urethane and acquire T1 weighted MR images.

3.2.2.2.2. Imaging Parameters

Different imaging paradigms are possible for activity measurements from the visual system using MEMRI. Following are two sets of optimal imaging parameters for the acquisition of T1 weighted images of the mouse or rat brain at 4.7T. These parameters are adapted from the work of Bruce Berkowitz and associates (70, 80) studying visual system activity in awake animals and are meant to be a starting guide. Further optimization is required depending on the animal model used as well as the magnetic field strength.

160

Massaad and Pautler

3.2.2.2.3. Mouse Brain

Adiabatic spin echo sequence TR = 350 s TE = 16.7 ms Number of acquisitions = 16 Sweep width = 61,728 Hz Matrix size = 512 × 512 ST = 620 μm FOV = 12 × 12 mm2

3.2.2.2.4. Rat Brain

Rapid Acquisition with Relaxation Enhancement (RARE) sequence TR = 330 ms TE = 16.6 ms RARE factor = 8 Number of acquisitions = 2 Matrix size = 256 × 256 × 173 FOV = 3.84 × 3.84 cm2 ST = 150 μm Time = 80 min/image

3.2.2.2.5. Expected Results

With the aforementioned protocol, one can expect to detect Mn2+ enhancement consistent with layer-specific visual cortex activity in awake and free-moving animals. Layers of a given cortical region respond differently to sensory stimulation and this MEMRI protocol appears to be sensitive enough to detect subtle changes in layer-specific activity (70, 80).

3.2.3. Activation Studies Without BBB Disruption – Olfactory System 3.2.3.1. MnCl2 Administration

Two current Mn2+ exposure modalities exist for activation studies from the olfactory system: Intranasal administration of MnCl2 1. Anesthetize the animal with 5% isoflurane. 2. Pipet 7 μl of a 10 mM MnCl2 solution in each naris. 3. Allow the animal to recover on a heating pad. 4. Place the animal in a clean cage and drop 7 μl of odorant solution in each of the four corners of the cage. 5. Allow odorant exposure for 20 min. 6. Anesthetize the animal with 5% isoflurane for imaging. Exposure to aerosolized MnCl2


161

1. Prepare a 1.5 M solution of MnCl2 in H2 O. 2. Place the solution in the heating chamber of a humidifier or vaporizer. For experiments involving the exposure to an odor, mix the odor within the MnCl2 solution. Use a different humidifier for every odor used to avoid crosscontamination between experiments. 3. Place the vaporizer inside a fume hood to avoid exposure to Mn2+ vapors. 4. Animal exposure to the aerosolized MnCl2 with or without odor is performed on either awake or anesthetized animals and is done as follows: • For awake animals: place the animal in the same plastic box housing the humidifier in the hood. Turn the humidifier on for 30 min. Keep the mouse in the box for 1.5 h after the humidifier has been turned off. It is important not to open the box during that time because of possible exposure to the aerosolized Mn2+ still present. • For anesthetized animals: anesthetize the animal with 20 mg/kg urethane and secure it on top of the humidifier with a restraining device. The exposure paradigm involves two sequences of 5-min on and 5-min off, and then the animal is kept for 1.5 h in the chamber with the humidifier off. Again, it is important not to open the box during that time because of possible exposure to the aerosolized Mn2+ still present. 3.2.3.2. Imaging Parameters

Different field strengths will dictate different imaging parameters for best results. The following parameters are adapted from the work of Alan Koretsky and associates (52) for studying olfactory activation in mice at 11.7T. These imaging parameters are meant to be a starting guide; further optimization with regards to field strength and organism used is required. T1 -weighted images acquired by a 3D RARE sequence TR = 300 ms TE = 10 ms Matrix size = 128 × 128 × 64 RARE factor = 2 Isotropic spatial resolution = 100 μm


The expected results from the activity-dependent olfactory tract tracing are a gradual increase in signal enhancement ranging from the olfactory epithelium to the olfactory bulbs. Signal enhancement will follow a region-specific enhancement depending on the stimulating odorant used (Fig. 7.4).

162

Massaad and Pautler

Fig. 7.4. Detecting odor-dependent Mn2+ enhancement in mouse olfactory bulb by MRI. MEMRI maps after stimulation by acetophenone, carvone, octanal, and control in four mice, respectively, show distributed enhancement in the glomerular layer with each odorant having its own distinct spatial pattern. High signal change at the interface between the olfactory nerve layer and olfactory turbinates (arrow) indicates where Mn2+ flowed in. Scale bars represent 1 mm. Reprinted from Chuang et al. (52), copyright 2009, with permission from Elsevier.

3.3. Tract Tracing

Tract tracing takes advantage of the following two properties of Mn2+ : (1) it is transported along neurons by microtubuledependent axonal transport and (2) it can traverse synapses and reach second-order neurons leading to contrast enhancement of the whole neuronal system in question. Tract-tracing studies have been performed in several systems such as the visual and olfactory systems as well as from deep brain structures such as the hippocampus and amygdala.

3.3.1. Tracing the Visual Pathway

3.3.1.1. MnCl2 Administration

1. Anesthetize the animal initially with 5% isoflurane and then maintain it with 2% isoflurane. 2. Place the animal in the prone position on a heating pad to maintain body temperature 3. Gently detach the metallic piece of the 27-gauge needle from its plastic hub using forceps. Connect the metallic portion of the needle to the Hamilton syringe via a small piece of polyethylene tubing. 4. Insert the tip of the needle into the vitreous with the aid of a microscope. A good injection site is about 2 mm posterior to the dorsal limbus. 5. Inject 0.1 μl of the MnCl2 solution over 5 min. The volume injected can be gauged by the advancement of the meniscus in the polyethylene tube using the scale of the Hamilton syringe. 6. Leave the injection needle in the eye for at least 15 min and then withdraw it very slowly to minimize the loss of MnCl2 through leakage from the injection site. This waiting time is necessary to insure homogenous distribution of the MnCl2


163

inside the eye as well as for the intraocular pressure to reach equilibrium. 7. Terminate anesthesia and return the animals to their cages. 8. Following the MnCl2 injection and prior to imaging, it is advised to check the integrity of the eyes. Successful injection can be ascertained by a bright-looking vitreal humor on a T1 -weighted image. 3.3.1.2. Imaging Parameters

Imaging parameters will vary depending on the animal model used as well as on the magnetic field strength. Following are the optimal imaging parameters used by Watanabe et al. to trace the visual pathway of rats at 2.35T (62). These parameters should be used as a starting guide and further optimization with regards to field strength performed for best results. T1 -weighted 3D FLASH gradient echo sequence TR = 15 ms TE = 4.2 ms Flip angle = 25◦ FOV = 50 × 50 × 16 mm Matrix size = 256 × 256 × 128 NEX = 8 Acquisition time = 65.5 min.


Signal enhancement is expected to be seen in the entire visual pathway, starting from the eye and extending to the superior colliculus. An example of such enhancement is illustrated in Fig. 7.5 (62).

3.3.2. Tracing the Olfactory Pathway 3.3.2.1. MnCl2 Administration

1. Anesthetize the animal with 5% isoflurane. 2. Hold animal in a vertical position by slightly pinching the hair in the back of the head. 3. Using a 10-μl pipette, administer 2 μl of a 3.9 M MnCl2 solution to each naris. The 2 μl can be either administered as 2 × 1 μl or at once. It is normal to observe some bubbling from the nose following the nasal lavage. 4. Place the animal on a heating pad to accelerate recovery. Usually it only takes a few minutes for the animal to regain consciousness. 5. Return the animal to the housing cage. 6. Proper lavage can be ascertained by very dark-looking turbinates on a T1 -weighted image (due to T2 effects of the concentrated Mn2+ solution).

164

Massaad and Pautler

Fig. 7.5. Signal enhancement of the rat visual pathway (24 h after Mn2+ -injection into the left eye) in oblique sections 235◦ (top left), 210◦ (top right), 15◦ (bottom left), and 137.5◦ (bottom right) relative to a transverse reference plane. Enhanced structures are (1) left retina, (2) left optic nerve, (3) optic chiasm, (4) right optic tract, (5) right lateral geniculate nucleus, (6) right brachium of the superior colliculus, (7) right pretectal region, and (8) right superior colliculus. Reprinted from (62), copyright 2001, with permission from John Wiley & Sons, Inc.

3.3.2.2. Imaging Parameters

Based upon the work of Pautler et al (3), the optimal imaging parameters for tracing the olfactory system of the mouse at 7T are as follows: T1 -weighted multi-slice spin echo sequence TR = 307 ms TE = 12.7 ms FOV = 2.5 cm ST = 1 mm Matrix size = 128 × 128 Higher resolution 3D scans can be acquired using the following parameters: TR = 300 ms TE = 8.7 ms


165

FOV = 2.5 × 2.5 × 2.5 cm Matrix size = 128 × 128 × 128 3.3.2.3. Expected Results

The expected results are a positive contrast enhancement in the olfactory bulbs as well as the primary olfactory cortex as illustrated in Fig. 7.6 (3).

Fig. 7.6. a Three sagittal slices of a mouse treated with Mn2+ in the naris from a representative 3D T1 -weighted MRI sequence. Note the highlighting of the olfactory bulb as well as the primary olfactory cortex leading from the bulbs. b Four axial slices from the same mouse treated with Mn2+ in the naris from a 3D T1 -weighted MRI sequence. Note the highlighting of the outer layers of the olfactory bulbs where the olfactory glomeruli are located. In addition, the enhanced contrast continues caudally into the primary olfactory cortex. Due to the length of the scan, mice were sacrificed before 3D imaging. Reprinted from (3), copyright 1998, with permission from John Wiley & Sons, Inc.

3.3.3. Tracing of Deep Brain Structures

3.3.3.1. Injection Site Coordinates

1. Identify the brain region you wish to inject.

3.3.3.2. Preparing the Injection Needle

1. Using a pipet puller and a quartz capillary tube with filament, pull the tube to create the injection needle. A long, fine-tip needle is needed. Micropipette pullers use a

2. Utilizing a stereotaxic brain atlas, determine the stereotaxic coordinates of the region of interest. This region will most likely encompass multiple slices. Be sure to choose the coordinates that correspond with the largest region in the structure of interest if possible. Additionally, it should be noted that the stereotaxic coordinates will vary based upon sex, age, and animal strain.

166

Massaad and Pautler

combination of heat (laser output power), filament (scan size) velocity (determines the point at which the heat is turned off), delay (the time between deactivation of the laser and the application of a hard pull), and pull (final hard pull applied on the capillary tube) to create needles of different shapes and lengths. Keeping in mind that different pullers operate under different settings (same pullers may even differ depending on the type of filament they have), the following parameters are good starting points: Heat=700; Filament=3; Velocity=60; Delay=140; Pull=175 2. Using fine forceps, gently break the tip of the needle to open it (the needle comes out sealed from the puller). 3. Place the needle in the picospritzer holder. 4. Set the picospritzer to the “pull” option and slowly fill the needle with the MnCl2 solution. 5. Set the prepared needle aside and proceed to preparing the animal for surgery. 6. It is advisable to pull and fill several needles, in case one breaks during the surgery. Store the filled needles in a humidified chamber to prevent crystallization of the MnCl2 solution. 3.3.3.3. Surgery

1. Anesthetize the animal with pentobarbital sodium or ketamine/xylazine combination (Note 11). 2. Upon lack of toe pinch reflex, place the animal on a warming blanket and clip the hair on the back of the head (the area extending from between the ears to the start of the back) (see Note 1). 3. Fix the animal’s head in a stereotaxic holder complete with a bite bar and cheek/ear bars. 4. Maintain anesthesia with 2% isoflurane. 5. Clean the operating field with the disinfecting solution chlorhexidine alternating with sterile water (three times). 6. Make a vertical incision extending from the nose to the start of the back; hold the skin open with hemostats. 7. With the help of the leveling tool and the microscope, make sure that the head is leveled properly both in the longitudinal and horizontal directions. 8. Using the stereotaxic device, determine the coordinates of your animal’s Bregma. 9. Calculate the placement of your region of interest with regards to the Bregma coordinates. For example, if your region of interest mesolateral position was –4.2 and your


167

animal’s Bregma was located at the mesolateral coordinate of 34.6, then you will need to place your holder at the 30.4 position (34.6–4.2). The same type of calculation applies to the other coordinates. 10. Mark the 2D location of the injection site by making a small scrape with the tip of a 27-gauge needle. 11. Begin drilling carefully at that location. Be sure to only drill a hole into the skull (e.g., mouse skull is less than 1 mm thick). The drill bit should not go through the dura or brain tissue. 12. Attach the volume gauge to the pre-filled needle with a small piece of modeling clay. 13. Lower the filled needle into the drilled hole. Carefully let the tip of the needle pierce the dura. Keep lowering until you reach the pre-calculated depth coordinate of your region of interest. 14. Adjust the microscope focus on the gauge so that the injected volume can be monitored. 15. Set the injection pressure to 20 psi approximately and begin with an injection time of 5 ms. Gradually increase the injection time until you see the meniscus slightly move within the needle. 16. Inject the full volume using this setting. Typically 10– 20 nl is suitable for tracing from deep brain structures (see Note 3). 17. Leave the injection needle in for at least 5 min and then withdraw it very slowly. This step is necessary to avoid the backflow of MnCl2 through the injection canal. 18. Place a few drops of analgesic such as bupivicaine just underneath the scalp and away from the drill hole and then suture the wound with 6-0 nylon suture. 3.3.3.4. Imaging Parameters

Different imaging paradigms are possible for tracing from deep brain structures following setereotaxic Mn2+ injections. Both 2D and 3D protocols can be used (3D recommended). Following are optimal imaging parameters for the acquisition of T1 -weighted images of the guinea pig brain at 3T. These parameters are adapted from the work of Lee et al. for tracing the auditory pathway in guinea pigs (63) and are meant to be a starting guide. Further optimization is required depending on the organism used as well as on the magnetic field strength. Two-dimensional spin echo sequence TR = 450 ms TE = 13 ms

168

Massaad and Pautler

Matrix size = 256 × 256 FOV = 50 × 50 mm ST = 1.5 mm (with a slice gap of 0.1 mm) NEX = 10 Three-dimensional gradient echo sequence TR = 10.2 ms TE = 2.5 ms Flip angle = 30◦ Matrix size = 256 × 256 × 128 FOV = 50 × 50 × 50 mm NEX = 7 3.3.3.5. Expected Results

The expected results are a multi-synaptic tract tracing to all structures involved in the system that is peripherally injected. For example, following injection of Mn2+ to the cochlea (63), signal enhancement can be observed in the entire auditory pathway, including the cochlear nucleus, the lateral lemniscus, the inferior colliculus, the medial geniculate nucleus, and the trigeminal tract (Fig. 7.7).

Fig. 7.7. T1 -weighted, 2D spin-echo MR image (A) before MnCl2 administration and T1 -weighted, 3D gradient-echo image (B) after 12 h of MnCl2 administration at the left cochlea in the guinea pig. The images’ orientation was obtained at the coronal section, and the voxel resolution was 195×195×200 μm (3). The post-injection image shows signal enhancement of the auditory pathway. Enhanced structures are as follows: (a) cochlear nucleus (CN), (b) lateral lemniscus (LL), (c) inferior colliculus (IC), (d) medial geniculate nucleus (MGN), and (e) trigeminal tract (TT). Reprinted from Lee et al. (63), copyright 2007, with permission from Elsevier.

3.4. Concluding Remarks

As delineated by the multitude of techniques described in this chapter, MEMRI is undoubtedly a very useful technique for the study of the brain anatomy and activity. Perhaps the most


169

important aspect of MEMRI is that it is minimally invasive and offers the possibility of longitudinal studies. This is of utmost importance for efficient diagnosis and understanding of disease states. Although MEMRI has been quite developed and refined in several organ systems (37, 53) of various animal models such as rodents (3, 44, 81, 82), song-birds (56–58), and non-human primates (55, 83), its use in humans is still almost non-existent. To date, Mn2+ has been approved for clinical imaging only in its chelated form (84). This is primarily due to the toxicity associated with a high concentration of free Mn2+ ions. High concentrations of Mn2+ have been shown to cause acute cardiovascular depression (85) as well as neurodegenerative damage to the nervous system (4). Many efforts are currently focused on developing Mn2+ contrast agents lacking the traditional side effects of Mn2+ . One such agent, available from Eagle Vision Pharmaceuticals, consists of free Mn2+ ions formulated with Ca2+ to override the transient effect of Mn2+ as a Ca2+ competitive inhibitor. This agent is currently used in dogs and pigs for cardiac (86, 87) as well as for vascular imaging (88). The development of such agent shows promise for the imminent use of Mn2+ as a clinical contrast agent for cardiac and brain imaging.

4. Notes 1. For fur trimming in the rat, a conventional rodent hair clipper can be used. For the mouse, it is recommended to use small sharp scissors. Slightly pull on the skin in the opposite direction of the fur growth and then cut the hair as close as possible to the skin, taking care not to injure the skin in the process. 2. It is necessary to use a “thinned out” PE50 tube as opposed to a smaller tube such as PE10. Use of PE10 tubing may not allow successful contrast agent injection because of high back pressure. 3. To construct a calibrated volume gauge, use Photoshop (or an equivalent drawing software) to draw a vertical line and add horizontal graduations to it that are separated by 1 pixel. Print the pattern on clear plastic (such as transparencies), cut it down to its proper size, and attach it to the injection syringe by means of a small piece of modeling clay. On such a scale and using 1-mm quartz capillary tubes, each graduation will correspond to 10 nl of fluid. 4. Several different concentrations of Mn2+ have been used in systemic administration studies. Doses ranged from

170

Massaad and Pautler

6.6 mg/kg to 175 mg/kg. With the higher doses, temporary side effects may occur; however, those effects resolve completely within an hour of MnCl2 administration. Many experimental factors such as the concentration of MnCl2 , the rate of infusion, the route of administration as well as the type, and level of anesthesia play a critical role in a successful MnCl2 administration. 5. An alternative to the direct oral intubation for artificial ventilation is the performance of a tracheotomy. This is however not recommended for most experimental paradigms as it adds more surgical trauma to the animal. 6. α-Chloralose/urethane combination dose: 25 mg/kg αchloralose with 450 mg/kg urethane intraperitoneal injection. 7. Once the tube is properly inserted into the tracheal opening, it is relatively stable. To avoid any possible dislodging of the tube, it can be secured with 6-0 surgical nylon sutures to the front incisors of the animal 8. When the catheter is introduced in the external carotid artery, care should be taken to introduce it in the direction of the common carotid artery, so that blood flow to the inferior carotid artery remains undisturbed. 9. When administering MnCl2 systemically through the venous system, an alternative to the tail vein catheterization is femoral vein catheterization. Both techniques have been shown to be equally effective. 10. Mannitol can recrystallize in solution and cause microinfarcts to the animal. To minimize the occurrence of re-crystallization, use pre-warmed tools such as syringes, saline, tubing etc. at 45◦ C. Flush the catheter with warm saline prior to mannitol administration and keep the room temperature above 25◦ C. Additionally, a 0.22-μm filter can be used and should be connected as close as possible to the external carotid artery. 11. Sodium pentobarbital dose: 50 mg/kg intraperitoneal injection. Ketamine/xylazine combination dose: 7.5 mg/kg ketamine with 5 mg/kg xylazine.

References 1. London, R. E., Toney, G., Gabel, S. A., Funk, A. Magnetic resonance imaging studies of the brains of anesthetized rats treated with manganese chloride. Brain Res Bull 1989;23:229–235.

2. Lin, Y. J., Koretsky, A. P. Manganese ion enhances T1-weighted MRI during brain activation: An approach to direct imaging of brain function. Magn Reson Med 1997;38:378–388.

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI) 3. Pautler, R. G., Silva, A. C., Koretsky, A. P. In vivo neuronal tract tracing using manganese-enhanced magnetic resonance imaging. Magn Reson Med 1998;40(5): 740–748. 4. Santamaria, A. B. Manganese exposure, essentiality & toxicity. Indian J Med Res 2008;128:484–500. 5. Hurley, L. S., Keen, C. L. Manganese. In: Trace Elements in Human Health and Animal Nutrition, Underwood, E., Mertz, W. (eds.) New York, NY: Academic Press; 1987, pp 185–225. 6. Zwingmann, C., Leibfritz, D., Hazell, A. S. Brain energy metabolism in a subacute rat model of manganese neurotoxicity: An ex vivo nuclear magnetic resonance study using [1–13 c]glucose. Neurotoxicology 2004;25:573–587. 7. Patchett, M. L., Daniel, R. M., Morgan, H. W. Characterisation of arginase from the extreme thermophile ‘Bacillus caldovelox’. Biochim Biophys Acta 1991;1077: 291–298. 8. Wedler, F. C., Denman, R. B. Glutamine synthetase: The major Mn(II) enzyme in mammalian brain. Curr Top Cell Regul 1984;24:153–169. 9. Keen, C. Nutritional and toxicological aspects of manganese intake: an overview. In: Risk Assessment of Essential Elements, Mertz W., Abernathy C., Olin S. (eds.) Washington, DC: ILSI Press; 1994. 10. Strause, L. G., Hegenauer, J., Saltman, P., Cone, R., Resnick, D. Effects of longterm dietary manganese and copper deficiency on rat skeleton. J Nutr 1986;116: 135–141. 11. Friedman, B. J., Freeland-Graves, J. H., Bales, C. W., Behmardi, F., Shorey-Kutschke, R. L., Willis, R. A., Crosby, J. B., Trickett, P. C., Houston, S. D. Manganese balance and clinical observations in young men fed a manganese-deficient diet. J Nutr 1987;117:133–143. 12. Venugopal, B., Luckey, T. Metal Toxicity in Mammals. Chemical Toxicity of Metals and Metalloids. New York, NY: Plenum Press; 1978. 13. Aschner, M., Guilarte, T. R., Schneider, J. S., Zheng, W. Manganese: Recent advances in understanding its transport and neurotoxicity. Toxicol Appl Pharmacol 2007;221: 131–147. 14. Crossgrove, J., Zheng, W. Manganese toxicity upon overexposure. NMR Biomed 2004;17:544–553. 15. Sloot, W. N., Gramsbergen, J. B. Axonal transport of manganese and its relevance to

16.

17.

18. 19.

20.

21.

22.

23.

24.

25. 26. 27.

28.

29.

171

selective neurotoxicity in the rat basal ganglia. Brain Res 1994;657:124–132. Takeda, A., Kodama, Y., Ishiwatari, S., Okada, S. Manganese transport in the neural circuit of rat CNS. Brain Res Bull 1998;45:149–152. Tjälve, H., Mejàre, C., Borg-Neczak, K. Uptake and transport of manganese in primary and secondary olfactory neurones in pike. Pharmacol Toxicol 1995;77:23–31. Takeda, A., Ishiwatari, S., Okada, S. In vivo stimulation-induced release of manganese in rat amygdala. Brain Res 1998;811:147–151. Barbeau, A. Manganese and extrapyramidal disorders (a critical review and tribute to dr. George C. Cotzias). Neurotoxicology 1984;5:13–35. Mena, I., Marin, O., Fuenzalida, S., Cotzias, G. C. Chronic manganese poisoning. Clinical picture and manganese turnover. Neurology 1967;17:128–136. Finley, J. W. Manganese uptake and release by cultured human hepato-carcinoma (hepG2) cells. Biol Trace Elem Res 1998;64: 101–118. Kerper, L. E., Hinkle, P. M. Cellular uptake of lead is activated by depletion of intracellular calcium stores. J Biol Chem 1997;272:8346–8352. Mason, M. J., Mayer, B., Hymel, L. J. Inhibition of Ca2+ transport pathways in thymic lymphocytes by econazole, miconazole, and SKF 96365. Am J Physiol 1993;264: C654–C662. Murphy, V. A., Smith, Q. R., Rapoport, S. I. Saturable transport of ca into CSF in chronic hypo- and hypercalcemia. J Neurosci Res 1991;30:421–426. Takeda, A. Manganese action in brain function. Brain Res Brain Res Rev 2003;41:79–87. Chance, B. The energy-linked reaction of calcium with mitochondria. J Biol Chem 1965;240:2729–2748. Huang, C., Cheng, H., Lin, K., Cheng, J., Tsai, J., Liao, W., Fang, Y., Jan, C. Tamoxifen-induced [Ca2+ ]i rise and apoptosis in corneal epithelial cells. Toxicology 2009;255:58–64. Tas, P. W. L., Stössel, C., Roewer, N. Inhibition of the histamine-induced Ca2+ influx in primary human endothelial cells (HUVEC) by volatile anaesthetics. Eur J Anaesthesiol 2008;25:976–985. Merritt, J. E., Jacob, R., Hallam, T. J. Use of manganese to discriminate between calcium influx and mobilization from internal stores in stimulated human neutrophils. J Biol Chem 1989;264:1522–1527.

172

Massaad and Pautler

30. Narita, K., Kawasaki, F., Kita, H. Mn and Mg influxes through Ca channels of motor nerve terminals are prevented by verapamil in frogs. Brain Res 1990;510:289–295. 31. Simpson, P. B., Challiss, R. A., Nahorski, S. R. Divalent cation entry in cultured rat cerebellar granule cells measured using Mn2+ quench of fura 2 fluorescence. Eur J Neurosci 1995;7:831–840. 32. Cory, D. A., Schwartzentruber, D. J., Mock, B. H. Ingested manganese chloride as a contrast agent for magnetic resonance imaging. Magn Reson Imaging 1987;5:65–70. 33. Geraldes, C. F., Sherry, A. D., Brown, R. D., Koenig, S. H. Magnetic field dependence of solvent proton relaxation rates induced by Gd3+ and Mn2+ complexes of various polyaza macrocyclic ligands: implications for NMR imaging. Magn Reson Med 1986;3:242–250. 34. Fornasiero, D., Bellen, J. C., Baker, R. J., Chatterton, B. E. Paramagnetic complexes of manganese(II), iron(III), and gadolinium(III) as contrast agents for magnetic resonance imaging. The influence of stability constants on the biodistribution of radioactive aminopolycarboxylate complexes. Invest Radiol 1987;22:322–327. 35. Mendonça-Dias, M. H., Gaggelli, E., Lauterbur, P. C. Paramagnetic contrast agents in nuclear magnetic resonance medical imaging. Semin Nucl Med 1983;13:364–376. 36. Pautler, R. G. Biological applications of manganese-enhanced magnetic resonance imaging. Methods Mol Med 2006;124: 365–386. 37. Silva, A. C., Lee, J. H., Aoki, I., Koretsky, A. P. Manganese-enhanced magnetic resonance imaging (MEMRI): Methodological and practical considerations. NMR Biomed 2004;17:532–543. 38. Natt, O., Watanabe, T., Boretius, S., Radulovic, J., Frahm, J., Michaelis, T. Highresolution 3D MRI of mouse brain reveals small cerebral structures in vivo. J. Neurosci Methods 2002;120:203–209. 39. Watanabe, T., Natt, O., Boretius, S., Frahm, J., Michaelis, T. In vivo 3D MRI staining of mouse brain after subcutaneous application of MnCl2. Magn Reson Med 2002;48: 852–859. 40. Watanabe, T., Radulovic, J., Spiess, J., Natt, O., Boretius, S., Frahm, J., Michaelis, T. In vivo 3D MRI staining of the mouse hippocampal system using intracerebral injection of MnCl2. Neuroimage 2004;22:860–867. 41. Watanabe, T., Radulovic, J., Boretius, S., Frahm, J., Michaelis, T. Mapping of the habenulo-interpeduncular pathway in living

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

mice using manganese-enhanced 3D MRI. Magn Reson Imaging 2006;24:209–215. Bock, N. A., Paiva, F. F., Nascimento, G. C., Newman, J. D., Silva, A. C. Cerebrospinal fluid to brain transport of manganese in a non-human primate revealed by MRI. Brain Res 2008;1198:160–170. Silva, A. C., Lee, J. H., Wu, C. W., Tucciarone, J., Pelled, G., Aoki, I., Koretsky, A. P. Detection of cortical laminar architecture using manganese-enhanced MRI. J Neurosci Methods 2008;167:246–257. Aoki, I., Wu, Y. L., Silva, A. C., Lynch, R. M., Koretsky, A. P. In vivo detection of neuroarchitecture in the rodent brain using manganese-enhanced MRI. Neuroimage 2004;22:1046–1059. Deans, A. E., Wadghiri, Y. Z., Berrios-Otero, C. A., Turnbull, D. H. Mn enhancement and respiratory gating for in utero MRI of the embryonic mouse central nervous system. Magn Reson Med 2008;59:1320–1328. Duong, T. Q., Silva, A. C., Lee, S. P., Kim, S. G. Functional MRI of calcium-dependent synaptic activity: Cross correlation with CBF and BOLD measurements. Magn Reson Med 2000;43:383–392. Aoki, I., Tanaka, C., Takegami, T., Ebisu, T., Umeda, M., Fukunaga, M., Fukuda, K., Silva, A. C., Koretsky, A. P., Naruse, S. Dynamic activity-induced manganesedependent contrast magnetic resonance imaging (DAIM MRI). Magn Reson Med 2002;48:927–933. Yu, X., Wadghiri, Y. Z., Sanes, D. H., Turnbull, D. H. In vivo auditory brain mapping in mice with Mn-enhanced MRI. Nat Neurosci 2005;8(7):961–968. Parkinson, J. R. C., Chaudhri, O. B., Bell, J. D. Imaging appetite-regulating pathways in the central nervous system using manganeseenhanced magnetic resonance imaging. Neuroendocrinology 2009;89:121–130. Weng, J., Chen, J., Yang, P., Tseng, W. I. Functional mapping of rat barrel activation following whisker stimulation using activity-induced manganese-dependent contrast. Neuroimage 2007;36:1179–1188. Lu, H., Xi, Z., Gitajn, L., Rea, W., Yang, Y., Stein, E. A. Cocaine-induced brain activation detected by dynamic manganese-enhanced magnetic resonance imaging (MEMRI). Proc Natl Acad Sci USA 2007;104: 2489–2494. Chuang, K., Lee, J. H., Silva, A. C., Belluscio, L., Koretsky, A. P. Manganese enhanced MRI reveals functional circuitry in response to odorant stimuli. Neuroimage 2009;44:363–372.

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI) 53. Wendland, M. F. Applications of manganeseenhanced magnetic resonance imaging (MEMRI) to imaging of the heart. NMR Biomed 2004;17:581–594. 54. Murayama, Y., Weber, B., Saleem, K. S., Augath, M., Logothetis, N. K. Tracing neural circuits in vivo with Mn-enhanced MRI. Magn Reson Imaging 2006;24: 349–358. 55. Saleem, K. S., Pauls, J. M., Augath, M., Trinath, T., Prause, B. A., Hashikawa, T., Logothetis, N. K. Magnetic resonance imaging of neuronal connections in the macaque monkey. Neuron 2002;34:685–700. 56. Tindemans, I., Verhoye, M., Balthazart, J., Van Der Linden, A. In vivo dynamic MEMRI reveals differential functional responses of RA- and area X-projecting neurons in the HVC of canaries exposed to conspecific song. Eur J Neurosci 2003;18: 3352–3360. 57. Van der Linden, A., Verhoye, M., Van Meir, V., Tindemans, I., Eens, M., Absil, P., Balthazart, J. In vivo manganese-enhanced magnetic resonance imaging reveals connections and functional properties of the songbird vocal control system. Neuroscience 2002;112:467–474. 58. Van der Linden, A., Van Meir, V., Tindemans, I., Verhoye, M., Balthazart, J. Applications of manganese-enhanced magnetic resonance imaging (MEMRI) to image brain plasticity in song birds. NMR Biomed 2004;17:602–612. 59. Van Meir, V., Verhoye, M., Absil, P., Eens, M., Balthazart, J., Van der Linden, A. Differential effects of testosterone on neuronal populations and their connections in a sensorimotor brain nucleus controlling song production in songbirds: A manganese enhanced-magnetic resonance imaging study. Neuroimage 2004;21:914–923. 60. Pautler, R. G. In vivo, trans-synaptic tracttracing utilizing manganese-enhanced magnetic resonance imaging (MEMRI). NMR Biomed 2004;17:595–601. 61. Pautler, R. G., Koretsky, A. P. Tracing odor-induced activation in the olfactory bulbs of mice using manganese-enhanced magnetic resonance imaging. Neuroimage 2002;16:441–448. 62. Watanabe, T., Michaelis, T., Frahm, J. Mapping of retinal projections in the living rat using high-resolution 3D gradient-echo MRI with Mn2+ -induced contrast. Magn Reson Med 2001;46(3):424–429. 63. Lee, J., Park, J., Lee, J., Bae, S., Lee, S., Jung, J., Kim, M., Lee, J., Woo, S., Chang, Y. Manganese-enhanced auditory tract-tracing

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

173

MRI with cochlear injection. Magn Reson Imaging 2007;25:652–656. Serrano, F., Deshazer, M., Smith, K. D. B., Ananta, J. S., Wilson, L. J., Pautler, R. G. Assessing transneuronal dysfunction utilizing manganese-enhanced MRI (MEMRI). Magn Reson Med 2008;60:169–175. Smith, K. D. B., Kallhoff, V., Zheng, H., Pautler, R. G. In vivo axonal transport rates decrease in a mouse model of alzheimer’s disease. Neuroimage 2007;35:1401–1408. Aoki, I., Naruse, S., Tanaka, C. Manganeseenhanced magnetic resonance imaging (MEMRI) of brain activity and applications to early detection of brain ischemia. NMR Biomed 2004;17:569–580. Kidoguchi, K., Tamaki, M., Mizobe, T., Koyama, J., Kondoh, T., Kohmura, E., Sakurai, T., Yokono, K., Umetani, K. In vivo X-ray angiography in the mouse brain using synchrotron radiation. Stroke 2006;37: 1856–1861. Zhao, X., Wu, N., Deng, M., Yin, Y., Zhou, J., Fang, Y., Huang, L. An improved method of left ventricular catheterization in rats. Physiol Meas 2006;27:N27–N33. Brown, R. H., Walters, D. M., Greenberg, R. S., Mitzner, W. A method of endotracheal intubation and pulmonary functional assessment for repeated studies in mice. J Appl Physiol 1999;87:2362–2365. Bissig, D., Berkowitz, B. A. Manganeseenhanced MRI of layer-specific activity in the visual cortex from awake and free-moving rats. Neuroimage 2009;44:627–635. Pautler, R. G., Mongeau, R., Jacobs, R. E. In vivo trans-synaptic tract tracing from the murine striatum and amygdala utilizing manganese enhanced MRI (MEMRI). Magn Reson Med 2003;50:33–39. Watanabe, T., Frahm, J., Michaelis, T. Functional mapping of neural pathways in rodent brain in vivo using manganese-enhanced three-dimensional magnetic resonance imaging. NMR Biomed 2004;17:554–568. Burnett, K. R., Goldstein, E. J., Wolf, G. L., Sen, S., Mamourian, A. C. The oral administration of MnCl2: A potential alternative to IV injection for tissue contrast enhancement in magnetic resonance imaging. Magn Reson Imaging 1984;2:307–314. Kita, H., Narita, K., Van der Kloot, W. Tetanic stimulation increases the frequency of miniature end-plate potentials at the frog neuromuscular junction in Mn2+ -, CO2+ -, and Ni2+ -saline solutions. Brain Res 1981;205:111–121. Drapeau, P., Nachshen, D. A. Manganese fluxes and manganese-dependent neurotrans-

174

76.

77.

78.

79.

80.

81.

82.

83.

Massaad and Pautler mitter release in presynaptic nerve endings isolated from rat brain. J Physiol (Lond) 1984;348:493–510. Rabin, O., Hegedus, L., Bourre, J. M., Smith, Q. R. Rapid brain uptake of manganese(II) across the blood-brain barrier. J Neurochem 1993;61:509–517. Murphy, V. A., Rosenberg, J. M., Smith, Q. R., Rapoport, S. I. Elevation of brain manganese in calcium-deficient rats. Neurotoxicology 1991;12:255–263. Yu, X., Sanes, D. H., Aristizabal, O., Wadghiri, Y. Z., Turnbull, D. H. Largescale reorganization of the tonotopic map in mouse auditory midbrain revealed by MRI. Proc Natl Acad Sci USA 2007;104: 12193–12198. Yu, X., Zou, J., Babb, J. S., Johnson, G., Sanes, D. H., Turnbull, D. H. Statistical mapping of sound-evoked activity in the mouse auditory midbrain using Mn-enhanced MRI. Neuroimage 2008;39: 223–230. Berkowitz, B. A., Gradianu, M., Bissig, D., Kern, T. S., Roberts, R. Retinal ion regulation in a mouse model of diabetic retinopathy: Natural history and the effect of Cu/Zn superoxide dismutase overexpression. Invest Ophthalmol Vis Sci 2009;50:2351–2358. Hu, T. C., Pautler, R. G., MacGowan, G. A., Koretsky, A. P. Manganese-enhanced MRI of mouse heart during changes in inotropy. Magn Reson Med 2001;46:884–890. Lee, J. H., Silva, A. C., Merkle, H., Koretsky, A. P. Manganese-enhanced magnetic resonance imaging of mouse brain after systemic administration of MnCl2: Dose-dependent and temporal evolution of T1 contrast. Magn Reson Med 2005;53:640–648. Bock, N. A., Kocharyan, A., Silva, A. C. Manganese-enhanced MRI visualizes V1 in the non-human primate visual cortex. NMR Biomed 2009, Available at:

84.

85.

86.

87.

88.

http://www.ncbi.nlm.nih.gov/pubmed/ 19322808 [Accessed July 21, 2009]. Federle, M. P., Chezmar, J. L., Rubin, D. L., Weinreb, J. C., Freeny, P. C., Semelka, R. C., Brown, J. J., Borello, J. A., Lee, J. K., Mattrey, R., Dachman, A. H., Saini, S., Harmon, B., Fenstermacher, M., Pelsang, R. E., Harms, S. E., Mitchell, D. G., Halford, H. H., Anderson, M. W., Johnson, C. D., Francis, I. R., Bova, J. G., Kenney, P. J., Klippenstein, D. L., Foster, G. S., Turner, D. A. Safety and efficacy of mangafodipir trisodium (MnDPDP) injection for hepatic MRI in adults: Results of the U.S. multicenter phase III clinical trials (safety). J Magn Reson Imaging 2000;12: 186–197. Wolf, G. L., Baum, L. Cardiovascular toxicity and tissue proton T1 response to manganese injection in the dog and rabbit. AJR Am J Roentgenol 1983;141:193–197. Storey, P., Chen, Q., Li, W., Seoane, P. R., Harnish, P. P., Fogelson, L., Harris, K. R., Prasad, P. V. Magnetic resonance imaging of myocardial infarction using a manganesebased contrast agent (EVP 1001-1): Preliminary results in a dog model. J Magn Reson Imaging 2006;23:228–234. Storey, P., Danias, P. G., Post, M., Li, W., Seoane, P. R., Harnish, P. P., Edelman, R. R., Prasad, P. V. Preliminary evaluation of EVP 1001-1: A new cardiac-specific magnetic resonance contrast agent with kinetics suitable for steady-state imaging of the ischemic heart. Invest Radiol 2003;38: 642–652. Zuo, C. S., Seoane, P., Lanigan, T., Harnish, P., Prasad, P. V., Storey, P., Li, W., Rofsky, N. M. T1 efficacy of EVP-ABD: A potential manganese-based MR contrast agent for hepatic vascular and tissue phase imaging. J Magn Reson Imaging 2002;16: 668–675.

Chapter 8 Sodium MRI Ronald Ouwerkerk Abstract Sodium (23 Na) imaging has a place somewhere between 1 H-MRI and MR spectroscopy (MRS). Like MRS it potentially provides information on metabolic processes, but only one single resonance of ionic 23 Na is observed. Therefore pulse sequences do not need to code for a chemical shift dimension, allowing 23 Na images to be obtained at high resolutions as compared to MRS. In this chapter the biological significance of sodium in the brain will be discussed, as well as methods for observing it with 23 NaMRI. Many vital cellular processes and interactions in excitable tissues depend on the maintenance of a low intracellular and high extracellular sodium concentration. Healthy cells maintain this concentration gradient at the cost of energy. Leaky cell membranes or an impaired energy metabolism immediately leads to an increase in cytosolic total tissue sodium. This makes sodium a biomarker for ischemia, cancer, excessive tissue activation, or tissue damage as might be caused by ablation therapy. Special techniques allow quantification of tissue sodium for the monitoring of disease or therapy in longitudinal studies or preferential observation of the intracellular component of the tissue sodium. New methods and high-field magnet technology provide new opportunities for 23 Na-MRI in clinical and biomedical research. Key words: MRI, sodium, 23 Na, other nuclei, cancer, stroke, brain.

1. Introduction The most exciting aspect of 23 Na-MRI is that the tissue sodium concentrations are very sensitive to changes in the metabolic state of tissues and the integrity of the cell membrane. Cells in healthy tissue actively maintain a large Na concentration gradient across the cell membrane, and almost any impairment of energy metabolism or insult to the cell membrane integrity leads to an increase in intracellular sodium (1). This leads to very significant 23 Na signal intensity changes in cancer, stroke, or myocardial M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_8, © Springer Science+Business Media, LLC 2011

175

176

Ouwerkerk

infarction. In excitable tissues, such as cardiac (1, 2), skeletal muscle (3), and brain, even normal activity can lead to transient changes in sodium (4).

2. The Physiology of Brain Tissue Sodium Content

One limitation of 23 Na-MR is that the sodium MR signal stems from both intracellular and extracellular compartments. There are some techniques that, at the cost of signal to noise or spatial resolution, do provide a weighting of the signal that favors the metabolically more interesting intracellular signal. The combined tissue sodium signal, however, can also provide a lot of information due to the fact that, through tissue perfusion, the extracellular sodium concentration [Na+ ]ex is constant and even in disease rarely deviates from the plasma sodium concentration of about 140 mmol/l. The intracellular sodium concentration [Na+ ]in is much lower, typically 10–15 mM, and this is maintained by active pumping of the Na+ /K+ ATPase, which is powered by ATP. The high difference in concentration of sodium (and potassium) between the intracellular and the extracellular compartments creates a potential that is used for transmitting nerve impulses and for pumping protons and small molecules across the cell membrane. Potassium leaves and sodium enters the cell both during depolarization as a result of a nerve impulse and when for instance protons are pumped out of the cell by the Na+ /H+ exchanger. Normally these ion fluxes are compensated by the exchange of intracellular sodium with extracellular potassium by Na+ /K+ ATPase using hydrolysis of high-energy ATP. In the brain, sodium is not only transported across membranes for the recovery of resting membrane potential but is even more solidly linked to the energetics of brain function through the pivotal role that the transmembrane sodium concentration gradient plays in the uptake of the neurotransmitter glutamate. Astrocytes, wrapped around synaptic contact sites, possess sodium glutamate co-transporters (5, 6), which clear glutamate from the extracellular space. These transporters are powered by the electrochemical gradient of Na+ . The sodium that enters the astrocyte with the glutamate is again pumped out by Na+ /K+ ATPase. Although the mechanisms also involve many other factors, such as Ca2+ (which is also linked to intracellular and extracellular sodium levels), the role of sodium as a link between glutamate and astrocytic energy metabolism is firmly established (7). In cultured fetal mouse astrocytes, it was found that under non-stimulated conditions, the Na+ /K+ ATPase consumes about 20% of astrocytic

Sodium MRI

177

ATP production (8), and sodium levels are maintained at around 10 mmol/l. Just 1 mmol/l extracellular glutamate can evoke a transient threefold increase in [Na+ ]in in primary cultured mouse astrocytes (9) lasting more than 2 min. Clearly, on continued or repeated activation the intracellular sodium level can rise considerably.

3. Measuring Sodium In Vivo The question is whether 23 Na-MRI can be used to observe these changes in vivo. Although the intracellular changes are substantial, the sodium concentration in the extracellular compartment is much higher and this may mask changes in the tissue sodium concentration due to an increase in intracellular sodium. Normally, however, the extracellular space is relatively small, and roughly half of the total observed tissue sodium is from intracellular sodium. Whereas, in spite of this, the normal transient increases in intracellular sodium may be too small and short lived to be observed with the current MR technology, abnormal circumstances or prolonged activation may lead to clearly observable changes in tissue sodium. When the demand for ATP exceeds the production of ATP, the ATP supply for the Na+ /K+ ATPase will be insufficient to maintain or quickly restore the low intracellular sodium concentration and thus a sustained increase of tissue sodium concentration (TSC) can be observed. This phenomenon has been observed in sustained exercise in muscle (3, 10, 11) and when energy supply is interrupted as in an ischemic heart muscle in animal models (12, 13). Due to the large sodium concentration a gradient is maintained across the cell membrane of resting healthy tissue and the influx of sodium can very rapidly more than double the intracellular sodium content. Even if that intracellular change is only half reflected in the tissue sodium content, it is still readily observable by 23 Na-MRI (13). Additionally, various factors, such as the concomitant potassium efflux and acidification, can trigger vasodilation in some tissues and thus lead to an increase in extracellular space and consequently in observed TSC. An increase in intracellular sodium paired with an increase in extracellular partial volume precludes a detailed physiological interpretation of changes in TSC without additional measurements. A 23 Na-MR method that is sensitive to either only intracellular sodium or any MR method that can be used to quantify the intracellular versus extracellular volumes would resolve this issue. Because neither goal can be achieved without introducing new

178

Ouwerkerk

ambiguities, or large signal to noise losses, or agents that cannot be used on human subjects, we still do not have a method to reliably measure [Na+ ]in in human disease. Fortunately, the intracellular and extracellular changes generally do not cancel each other, and thus, the TSC generally turns out to be a good indicator of tissue health or cancer malignancy.

4. 23 Na-MRI from Past to Present The feasibility of 23 Na-MRI on human subjects has been demonstrated as far back as the late 1980s (14), and the image contrast obtainable in cancer (15), edema, and stroke (16, 17) has also been demonstrated for quite some time. Despite this, 23 Na-MRI has yet to find wider acceptance as a possible clinical MR tool. At present, the increased availability of high field MRI systems appears to lead to a revival of 23 Na-MRI evidenced by a recent surge not only in methodological papers on 23 Na-MRI of the brain (18–21) but also in new clinical MR research papers, where 23 Na-MRI is used to study disease in humans (17, 22–26). Reduced scan times now allow the addition of a 23 Na-MRI to existing 1 H-MRI protocols. It is through the merging with a comprehensive 1 H-MRI protocol that the potential of 23 Na-MRI as a diagnostic tool can best be developed. A scanner with higher field strength will offer a better signal to noise, which can be traded for higher resolution or shorter experiment time. A few recent and early results are listed in Table 8.1. The technical advances since the early 1990s have paid off in an increase in signal-to-noise ratio (SNR) and reduction of voxel size and total scan time. From this table, it is also clear that 23 Na-MRI results are better in terms of SNR when scan methods with a short echo time (TE) are used. Brain 23 Na at 3 T with TPI has been reported to yield the expected twofold SNR increase over the same method at 1.5 T. Thus, it should be possible to acquire 450 μ

~22

(18)

2007

1.5 T

10

4×4×4

64

Projection imaging

200 μs

3

(19)

2007

1.5 T

10

4×8×10

320

FLASH

2.7 ms

19

(19)

2003

4T

14:20

3×6×11

194

3D-GRE

1.6 ms

>20 WM

(20)

2007

9.4 T

5:56

3×3×3

27

TPI

260 μs

Unknown (21)

2006

4T

30

7.5×7.5×7.5 423

SPRITE

20

(36)

∼1–2

(74)

1.5

(75)

220

1987

1.5 T

16:20

4×4×20

280

Multi-echo GRE

13 ms∗

1985

1.3 T

10:30

5×5×10

273

3D FT (3D-GRE)

14 ms

TPI, Twisted projection imaging; GRE, Gradient-recalled-echo; FLASH, Fast low-angle shot; ∗ TE series 13, 26 and 39 ms SPRITE, Single-point ramped imaging with T1 enhancement.

reached, where 23 Na-MRI scan times have been reduced to the point where the addition of a 23 Na-MRI to an existing 1 H-MRI protocol has become a viable option for studies on patients rather than just extremely tolerant volunteers.

5. MR Properties of 23 Na In Vivo The SNR of 23 Na-MRI is much lower than that of 1 H-MRI, mainly because of the difference in abundance between water protons and sodium ions in the body and also because the gyromagnetic ratio is a factor four lower. This lower SNR is partially offset by the short longitudinal relaxation time, T1 , on the order of 25–40 ms, which allows signal averaging with fairly rapid repetition rates, TR. Even when the TR is relatively long to allow quantification of TSC without T1 relaxation corrections acquisition rates (TR > three times T1 ) of about 10 excitations per second are still possible. However, the transverse relaxation time, T2 , of sodium is not helping at all. The sodium T2 is very fast and is bi-exponential in many biological tissues and in gels. This dual exponential decay is a consequence of the spin number of 23 Na, which is 3/2. Protons have a spin number 1/2 and can have two distinct energy

180

Ouwerkerk

levels. Nuclei with spin number 3/2 can have four distinct energy levels with spin number m = −3/2, −1/2, 1/2, and 3/2. There are therefore more possible transitions between the energy levels. Without going into quantum physics too much, we can understand that the energy level in a nuclear spin can change between two adjacent levels, called a single quantum transition, but it can also be transferred to other levels with double or triple quantum transitions. We normally see only single quantum transitions, and in a watery solution all the possible single quantum transitions have the same energy, so we see a simple single line spectrum. In highly ordered restricting environments, such as cartilage or in gels, the gaps between the energy levels for 23 Na start to differ. This is due to anisotropy. When the sodium comes into close contact with the ordered molecules, and the system has a distinct directionality on the molecular level, this can lead to a spectrum with a central peak and two satellites separated from the central peak by the quadrupole frequency. Also the relaxation times associated with the different transitions will differ. The possibility of distinct transitions will lead to a bi-exponential transverse signal decay with a short T2 , on the order of 1–2 ms for 60% of the signal and a longer T2 on the order of 20–30 ms for 40% of the signal. The relaxation time of the satellite transitions is shorter, and therefore these lines in the MR spectrum will be much broader, sometimes so broad that they disappear into the baseline. In the past, this has led to speculations that part of the MR signal in vivo is invisible. Unless the acquisition delay is short this is indeed the case, and even if observed, the short T2 is bound to broaden the points spread function in 23 Na imaging. The images will be a superposition of a sharp and a fuzzy image. Images recorded with TE that are long compared to this short TE may look sharper than short TE images with the same nominal resolution, but this could be remedied by time domain filtering of the short TE images. In most biological tissues, not all of the many different environments in which sodium is present (cytosol, endoplasmatic reticulum, mitochondria, extracellular matrix, blood plasma, cerebro-spinal fluid, etc.) will have the same density of ordered binding sites that lead to bi-exponential relaxation. Thus, except for a homogeneous, highly ordered tissue, such as cartilage, the fraction of the fast relaxing component is much less than 60%. Still, substantial signal losses can occur as a result of T2 relaxation. The T2 for 23 Na in the brain has been determined with a single exponential model and by some even with a bi-exponential model. The results, summarized in Table 8.2, give some idea of what sequence parameters to use for 23 Na-MRI of the brain, and it is clear that CSF and crystalline saline solutions have a long T2 , whereas white matter and gray matter have (slow) T2 s in the order of 17–25 ms. For quantification of TSC, short TE, ideally on the order of 0.2–0.4 ms, is required to reduce those T2 losses

Sodium MRI

Table 8.2 Literature values for transverse animal tumor models

23 Na-MR

Tumor

GM

WM

T2 fast ms (42%)

3.6a

-

-

T2 slow ms (58%)

18a

17–18

17–18

CSF

relaxation rates in human brain and in VH

Saline

57

59

Whole brain

2.1 ± 0.3

T2 slow ms (40%)

20 ± 2.3 4.6 ± 1.8b

T2 slow ms (48%)

23.3 ± 7.6b

T2 ms

Reference

Perman (76)

T2 ast ms (60%)

T2 ast ms (52%)

181

Constantinides (77)

Summers (78) 11

T2 ms

12 32

T2 ms

47

58

53

42

60

21

Winkler (79) Feinberg (75)

57

Perman (80)

GM, gray matter; WM, white matter; CSF, cerebro-spinal fluid; VH, Vitreous humor. a Rabbit VX2 carcinoma. b IMR-5 neuroblastoma in nude mice.

to less than 5–10% of the total signal. In the next section, a few techniques are discussed that will enable the acquisition of 23 Na with minimal signal losses due to T2 relaxation.

6. 23 Na-MR Techniques The gradient-recalled echo technique (GRE) is a good initial choice for 23 Na-MRI, because it is relatively simple to implement. The sequence can be set up by a few simple adaptations from a 1 H-MRI sequence that is available on any scanner platform. The requirement for a short TE, together with the need for small receiver bandwidths (RBW) for better SNR, makes it difficult to use gradient-recalled echo sequences with small T2 losses and good SNR. When using GRE as a method for 23 Na-MRI it is advisable to optimize the SNR by recording a series of images with different RBW, all at the minimum achievable TE for that RBW to find the best compromise in terms of SNR. Typically, for a 64-point readout, the optimum at 1.5 or 3 T will be around 4–16 kHz RBW with a TE of about 5 ms for the lowest RBW and 2 ms for 16 kHz RBW. The actual optimal numbers are very

182

Ouwerkerk

scanner specific, as faster gradients will allow a shorter TE for a given RBW. A sample 3D 23 Na GRE image of a healthy volunteer is shown in Fig. 8.1. This scan (after obtaining informed written consent to an IRB-approved protocol) was performed on a 1.5 T GE Signa scanner (GE Medical Systems, Milwaukee, WI) using a sequence that was only minimally modified from the product 1 Hfast GRE sequence. Modifications included the replacement of

Fig. 8.1. Fast GRE 23 Na image of the brain of a healthy volunteer at 1.5 T. Non-selective excitation was used with an otherwise standard 3D-GRE sequence on a GE Signa scanner (GE Medical Systems, Milwaukee, WI), recording 64 points at 8 kHz RBW with a TR/TE = 30.9/5 ms, 32 slices (not all shown), and 12 averages for 7.5×7.5×15 mm voxel size in a total scan time of 80% hematocrit suspension of red blood cells with the extracellular component (e) shifted upfield by Dy(PPPi)2 and the sodium in the reference sample (r) shifted downfield by Dy(TTHA). a Control spectrum and b after 1 h exposure to 1 mM ouabain. The intracellular signal (i) increased and the potassium leaving the cells reduced the shift of the extracellular component by competitive binding to the Dy(PPPi)2 shift reagent. Figure modified from (77).

animal model may help to understand the changes observed with TQF better. Even so, in non-human primates some really dramatic changes in TQF sodium signals have been demonstrated occlusion and subsequent reperfusion. In the same study, a SQ 23 NaMRI only 12 min prior to the TQF images shows only small changes (relative to the >300% increase in TQF signals) (54). Clearly, in spite of ambiguities caused by the T2 dependency of the TQ filter the TQF appears to be far more sensitive to the changes in the brain after ischemia and reperfusion. Unfortunately, the visible changes observed with SQ 23 Na were not quantified in this study.

Sodium MRI

189

Fig. 8.4. Maximum signal intensities obtainable with single, double, or triple quantum-filtered (SQF, DQF, TQF, respectively) 23 Na-MR detected at an acquisition delay of 0.4 ms calculated according to Eqs. [2], [3], and [4] in (50). a The maximum signal intensities observable with 23 Na-MR at TE = 0.4 ms, as a function of the fast T2 with triple quantum filtering (solid line, slow T2 = 30 ms, and long dashed line, slow T2 = 20 ms) and without multiple quantum filter (short dashed line, slow T2 = 20 ms). b The same data as in (a) now expressed as ratio of the signal over the single SQF signal. The dot dash line is the signal ratio for a double quantum filter. All data in (a) and (b) assume an optimized multiple quantum filter. c The signal as a function of the preparation time τ as for three different fast T2 with the slow T2 fixed at 20 ms and a TE of 0.1 ms. d As in (c), but the fast T2 fixed at 1 ms and the slow T2 set at three values 10, 20, and 30 ms.

In light of all the shortcomings of MQF 23 Na-MR, the search for a method to distinguish intracellular and extracellular signals must go on. A relatively low-tech solution is inversion recovery (IR) T1 weighting (12). This approach has been used to study chemotherapy response in a mouse–human prostate cancer xenograft (55) to selectively observe cartilage tissue (56) and in the brain to suppress the CSF signal (57). The brain application looks promising even though the IR method reduces the SNR of normal brain tissues and the selective suppression of CSF appears to work as long as B0 field homogeneity is good. Also, the IR pulse works on differences in both T1 and T2 , because with the species with shorter T2 , the inversion pulses are less effective and the resulting incomplete inversion leads to a quicker signal

190

Ouwerkerk

recovery. The IR method precludes quantitative 23 Na-MRI, and it is unclear how T1 might change in disease. Perhaps where the MR properties of 23 Na fail to give us a handle on intracellular and extracellular volume, we should instead be looking at the much higher natural abundance and SNR available in 1 H spins to quantify intracellular and extracellular compartment volumes. At present, without a reliable means to exclusively quantify intracellular sodium, we can still use 23 NaMR imaging of the total tissue sodium to our advantage, but only in combination with an optimized, comprehensive 1 H-MRI protocol. To appreciate what type of information 23 Na-MRI adds to an MRI examination, we can be pragmatic and study TSC in disease and see whether 23 Na is a good biomarker. At the same time, we need to continue studies on perfused organs and whole-animal models in order to understand the origin of the disease-specific contrast delivered by measurement of TSC with 23 Na-MRI.

10. Brain Tissue Sodium and Cancer

Proliferating cells have an abnormally high sodium content (58), because the intracellular sodium concentration is elevated as a result of altered Na+ /H+ transport kinetics (59–62) and pH. Outside the cells, continuous perfusion of living tissue will ensure a constant sodium concentration of approximately 140 mmol/l. Thus, an increase in the extracellular partial tissue volume through the increased vascularization (angiogenesis) and the increased interstitial space in tumors (63, 64) will also lead to increases in tissue sodium concentration (TSC) in tumors. Again, as in active or energetically challenged excitable tissues, both intracellular sodium concentration and extracellular volume are contributing to an increase in tissue sodium. Using a quantitative 23 Na-MRI technique, a 50% increase in TSC, relative to non-involved contralateral tissues, was found in malignant brain tumors (36). An example study of a patient with an astrocytoma is shown in Fig. 8.5. Likewise, in malignant breast tumors, an increase of 50% in TSC was found relative to noninvolved glandular tissue and benign lesions. A small study of 13 subjects with suspected low-grade glioma 23 Na-MRI, combined with 1 H-MRS, revealed a doubling of the sodium signal compared with contralateral white matter (65). A negative correlation was found between sodium signal and N-acetylaspartate (NAA). In this study, the ratio of NAA over the 23 Na signal was shown to be the best metric for separating glioma from healthy tissue. The 23 Na-MRI data in this study were recorded with a gradient-recalled echo (GRE) sequence with a

Sodium MRI

191

Fig. 8.5. 1 H- and 23 Na images of the brain of a patient (36-year-old male) with an astocytoma grade III. a T1 -weighted post-contrast (Gd–). b T2 -weighted fluid-attenuated inversion recovery (FLAIR). c Quantitative 23 Na-twisted projection image with 100 ms TR and 0.4 ms TE. Two reference sample tubes with 2% agarose gel containing known [Na+ ] were placed on either side of the head for quantification of brain TSC and to facilitate registration of 1 H and 23 Na images. The 23 Na images were recorded at 1.5 T with a separate 23 Na birdcage coil that replaced the 1 H birdcage coil after the 1 H-MRI protocol without moving the patient’s head or reference phantoms. The 23 Na image (c) is an oblique slice from the isotropic 3D 23 Na data set, interpolated to register with the 1 H-MR images. Level contours drawn on the post-Gd T1 -weighted image (a) were copied to the 23 Na image (c) to show the location of the Gd enhancement.

relatively long TE of 3.8 ms. The results were corrected for T2∗ relaxation assuming an 18 ms (single exponential) T2∗ relaxation time in normal tissue versus 27 ms in the low-grade glioma. The latter value was measured in only one subject, but it is a substantial difference with the normal value. If this increase in T2∗ holds true for all brain tumors, it makes the detection of cancer-related increases in TSC with single quantum 23 Na-MR less dependent on the choice of MR sequence timing.

11. Sodium Changes in Stroke 23 Na-MRI

has been used to study stroke in humans as far back as 1993 (16). With the limited SNR no significant increase in sodium was seen in the first 13 h post-infarction. After that initial increase, 23 Na was seen to peak at 45–82 h. In animal models, the sequence of events could be studied more accurately. In a rat model of stroke induced by tandem occlusion of the right middle cerebral artery (MCAO) and common carotid arteries (CCAs), sodium as determined from punch samples with flame spectrometry appeared to increase linearly with time up to at least about 450 min after occlusion at a rate of about 1 mmol/kg dry weight/min against a staring concentration of 240–250 mmol/kg dry weight (66). In animal models, these findings have been confirmed with 23 Na-MRI (67, 68). Whether measuring TSC or using special techniques to be more sensitive to [Na+ ]in , the complex sequence of changes in the brain after transient ischemia cannot be fully understood by

192

Ouwerkerk

looking at the changes in sodium alone. It is important to know what happens to cell volume and the relative volume of the extracellular space in order to know what a change in observed sodium signal means. The observed change can be due to actual influx of sodium into the cell (a sign of compromised membrane integrity or a shortfall in energy metabolism) or shrinkage of the cell. In TQF 23 Na, this would lead to a more restricted environment, which would change relaxation parameters and probably increase the portion of intracellular sodium experiencing anisotropic environments that lead to a TQF signal. Although a study that combined 1 H-MRI and 23 Na-MRI (69) failed to show that any of these techniques can unambiguously predict the potential infarct size, non-human primate studies and human studies using combined 1 H- and 23 Na-MRI showed encouraging results (17).

12. 23 Na-MRI to Monitor Therapy Lack of substrate and oxygen can cause quite significant changes in tissue sodium content. As soon as the energy-dependent Na+ /K+ -ATPase stops pumping sodium out of the cell, passive sodium influx from the extracellular environment will rapidly raise the intracellular levels of sodium several fold. This effect is exacerbated when stress on the cells increases the permeability of the cell wall for sodium ions. The acute effect of a therapy that causes cell death or membrane rupture on an appreciable scale should, therefore, be easy to monitor with 23 Na-MRI. This was confirmed by the observation of an increase in sodium 24 h after administration of taxotere in a xenograft animal model of prostate cancer (55). In the long run, the elevated sodium found in cancer should again decline if the therapy is successful. In a limited number of patients with breast cancer, who were undergoing preoperative systemic chemotherapy, the effect of the therapy in responders was demonstrated as a decline in the tumor TSC along with a decline in lesion size (70, 71). In this application of 23 Na-MRI, it is important to have a reference or a quantification method by which to compare the results of baseline images and multiple post-chemo therapy images. Of course therapies other than chemotherapy can also be monitored. Surgery in the brain will leave a void that is filled with CSF. The high signal of CSF could obscure high TSC in the adjacent residual tissue. Thus, for postoperative evaluations the method of Stobbe et al. to suppress the CSF signal (57) might be interesting. Ablation methods should lead to an easily detectable increase in TSC. The effect of the highly focused ultrasound (HIFUS) ablation of human uterine fibroids could be

Sodium MRI

193

seen on 23 Na-MRI images obtained within 24 h of therapy (72). The long-term effects of such therapies may be more difficult to predict, because this depends on what type of tissue replaces the damaged tissue in the ablated areas. Whereas most of the aforementioned therapies typically are designed to induce quick cell death in a lesion, one study actually implies that 23 Na can also be used to detect cell death in a slow progressive disease, such as mild Alzheimer’s disease. A small, but significant, increase in hippocampal sodium signal as measured in five patients with a 23 Na GRE sequence at 3 T was seen to inversely correlate with hippocampal volume (73). It is encouraging to see that such subtle changes can also be detected, even with a relatively easily implemented sequence, such as GRE.

13. How to Perform a 23 Na-MRI Study of the Brain

All the preceding information on 23 Na methods is vital knowledge for gathering and publishing biomedical 3 Na research data. Even so, given the right hardware and research tools on the MR scanner, the task of measuring a 23 Na-MRI data set does not have to be daunting. A relatively simple method, such as 3D-GRE, straight projection imaging or UTE-CSI, can be implemented relatively easily. In the following recipe, the 3D-GRE method was selected for simplicity and universal availability. Stepwise, here is what to do: 1. Choose the best scanner for the protocol. The highest possible field is good for 23 Na, but the availability of a comprehensive 1 H-MRI protocol for the disease to be studied is just as important. The scanner must be equipped with the hardware required for measuring other nuclei. If not present, a 23 Na pre-amplifier must be acquired. These are relatively inexpensive, and it is advisable to have a spare. 2. Obtain a suitable MR sequence or get access to the scanner software and create one of the sequences above. Test the sequence in a simulator to see whether literature values for RBW, resolution, and minimum TE can be reproduced. For a GRE sequence, the selective RF pulse can be replaced by a hard (rectangular) RF pulse or by an adiabatic half-passage (AHP) pulse (Fig. 8.6). The latter ensures that even if the coil homogeneity is not perfect, the whole brain is excited with a 90◦ flip angle. Because of the short T1 of 23 Na, a 90◦ flip angle can yield optimum SNR per unit time, with a relatively short TR. Assuming a brain 23 Na T1 is about 30 ms the optimum TR in terms of optimum SNR

194

Ouwerkerk

Fig. 8.6. A 3D-GRE sequence for 23 Na-MRI. a The standard sequence as present on most scanners for 1 H-MRI with phase encoding in the Y and Z directions. The slice is relatively thick and subdivided by phase encoding in the Z direction superimposed on the slice-selective refocusing gradient lobe. b The modified sequence for 23 Na-MRI. The RF pulse has been replaced by a non-selective AHP. The dephasing (X) and phase-encoding (Y, Z) gradient lobes are made as short as possible, a partial echo readout is used, and the RF pulse is a non-selective tanh/tan-modulated AHP pulse. A relatively large field of view and low resolution keep all minimum gradient times short, but a (desirable) low RBW will increase the TE.

per unit time for the 90◦ flip angle is just under 40 ms, or 1.25 × T1 . 3. Figure 8.6 shows a comparison between a standard 3DGRE sequence and a version suitable for 23 Na-MRI. In the optimize sequence of Fig. 8.6b, the slice-selective pulse is replaced by a tan/tanh modulated AHP. Not only does this allow the use of coil with a less homogeneous B1 field (which could allow a better filling factor and thus SNR), but it also allows the elimination of the slice-selective

Sodium MRI

195

gradient and the refocusing lobe. Because the symmetry point of the AHP is at the end of the pulse, the nominal TE is not increased by an increase of the pulse duration, but there will be some T2 relaxation during the pulse for the fast T2 component of the signal (37). The limiting factor on achieving minimal TE for this sequence is still the length of the dephasing gradient and the RBW. Without resorting to the aforementioned UTE-CSI or projection imaging methods, this cannot be avoided. 4. Obtain a 23 Na head coil that can be swapped for a clinicalgrade 1 H-MR head coil without moving the subjects’s head (36) or get a dual-tuned 23 Na-1 H head coil that has a clinical-grade performance on the 1 H channels(s) without sacrificing too much of the 23 Na performance (74). For brain, the 23 Na coil must be transmit and receive in quadrature. This requires a splitter for 23 Na to separate transmit RF from the receiver input and at the other end split the RF signals into two orthogonal phase quadrature signals. This component should typically be included in the coil, and the pre-amplifier mentioned in point 1 could be integrated as well. The 1 H can be receive only and possibly multi-channel. 5. Create phantoms to test the coil. One phantom should be large and loaded with 100–150 mmol/l NaCl and doped with ca. 2 g/l CuSO4 . It is advisable to create gel phantoms containing known concentrations of sodium ranging from 20 to 150 mmol/l for calibration and for use as feducials and reference standards. The use of 2–4% (weight/volume) agarose and about 2 g/l CuSO4 will reduce the T2 and T1 of the samples. 6. Measure 23 Na-MRI images with different flip angles spanning at least a factor two variation. The actual local flip angles can be determined from a sine fit of the image pixel intensities as a function of nominal flip angle. If AHP pulses are used, the B1 must be varied to determine the threshold B1 for adiabaticity. Determine the minimum B1 at which the signal is B1 invariant for all locations. For actual experiments on the brain, use a B1 that is at least 50% higher than the threshold for as much as patient safety limitations of the specific absorption rate (SAR) will allow. Note that the patient’s weight entered in phantom studies is important in the SAR calculations. 7. Measure the SNR with different RBW and the minimum TE at each RBW. Determine the optimum settings. 8. Measure the 23 Na T1 and T2 (or T2∗ ) of the concentration and reference phantoms. If the 23 Na-MR sequence does

196

Ouwerkerk

not allow this, it can be done non-localized with a spectroscopy sequence one sample at a time. This approach is tedious, but precise. The agarose gel and CuSO4 should bring the relaxation rates of the sodium in the phantom close to physiological values, but this must be verified and recorded for corrections of reference phantom data collected with in vivo studies. Create a concentration calibration plot (Fig. 8.7). Repeat this with the concentration test phantoms in different positions.

Fig. 8.7. Example of a single concentration calibration series. Seven 50-ml tubes were filled with 2% w/vol aqueous agarose gel with 2 g/l CuSO4 and 30, 45, 60, or 75 mmol/l NaCl. A 1,240 projection TPI sequence with TR/TE = 100/0.4 ms was used to record an image with 6 mm isotropic actual resolution in 10 min using a 3-turn solenoid breast coil. Images were reconstructed by gridding to 64 × 64 × 64 points and FFT to a 22-cm isotropic field of view image. The SNR was determined from mean pixel intensities in circular regions in the tubes and regions devoid of meaningfull signal of 1–2 thousand pixels. SNR was calculated as the ratio of the signal minus the mean noise divided by the noise standard deviation. The SNR correlated with concentration with r2 > 0.98.

9. Test the sequence on a healthy volunteer and measure SNR in a relatively CSF-free part of the brain and if possible determine gray and white matter SNR from a segmented brain image created with a registered 1 H-MRI. Place two or more agarose gel samples with known sodium concentrations against the skull (Fig. 8.1). These can serve as concentration reference and as feducials. The agarose and CuSO4 drastically improves visibility on T2 -weighted 1 H images. From these tests, determine the minimum scan time that yields the desired resolution and estimated contrast to noise. 10. Determine the estimated TSC. If possible, measure T2∗ and T1 in the brain. Even when not aiming for quantitation of TSC, the sequence should be set up to minimize signal losses as a result of relaxation, because we cannot know how T2∗ and T1 might change in disease. Ideally, signal changes reflect changes in TSC.

Sodium MRI

197

11. Test the entire 1 H-23 Na protocol on healthy volunteers. 12. Scan patients, analyze the data, and publish! Of course, the more complex techniques, such as TPI or SWIFT, could possibly yield better SNR or more precise quantification of TSC. Whether this is worth the extra effort depends on your long-term research goals, technical prowess, and endurance. The above recipe is designed to provide a quick and solid introduction into 23 Na-MRI.

14. Quantification of TSC Apart from the reduced signal losses, the very short TE methods have the added advantage that quantification can be performed with minimal or no correction for T2 losses. By extending the TR to three times the T1 (or less of smaller flip angles are used) the correction for signal saturation can also be bypassed. This slower TR reduces the efficiency in terms of SNR per unit time, but it removes the ambiguity associated with the use of relaxation corrections when we cannot be sure that these remain the same in disease. Unless the SNR for brain tissues is really high (>100) sodium concentrations are best calculated from mean intensities in regions of interest (ROI). A convenient way to do is to trace ROI on co-registered with 1 H images for guidance and copy the regions to the 3D 23 Na images. The two or three sodium concentration reference tubes placed close to the head provide a feducial for co-registering images. For 23 Na imaging of the brain with a quadrature head coil we can assume homogeneous receive sensitivity and use the external reference samples or use a B1 map, as measured in point 5 of the list above, to correct for differences in receive sensitivity and flip angles (unless AHP pulse were used). Alternatively, the signal of CSF or even ocular fluids can be used as an internal standard. These fluids have a long T2 > 40 ms and also an even longer T1 . The latter needs to be determined for necessary T1 corrections; the T2 corrections are less of a problem.

15. Conclusion To get the most out of 23 Na-MRI, it is advisable to complement the TSC-related information gained from 23 Na-MR with 1 H-MR techniques that provide additional information of the

198

Ouwerkerk

physiological environment of sodium in the tissue. Although the water distribution between intracellular and extracellular is different and no technique gives direct information about the relative compartment volumes some well-established 1 H-MR techniques, such as contrast-enhanced (CE)-MRI or diffusionweighted (DW)-MRI, are sensitive to changes in cell volume, extracellular space, and membrane permeability. The best protocol will be different for each particular application, and in each application, we have to examine whether 23 NaMRI has enough added value to justify the extra scan time and effort. But in doing so, we must not forget that practically all MRI exams consist of a series of scans and only the whole set will yield the desired diagnostic sensitivity and specificity. We cannot expect that 23 Na-MRI can compete with clinical 1 H-MRI protocols when implemented in an exam consisting of only a scout image and a 23 Na-MRI, but a well-designed protocol including 23 Na-MRI can yield exiting results. References 1. Murphy, E., Eisner, D. A. Regulation of intracellular and mitochondrial sodium in health and disease. Circ Res 2009;104: 292–303. 2. Hilgemann, D. W., Yaradanakul, A., Wang, Y., Fuster, D. Molecular control of cardiac sodium homeostasis in health and disease. J Cardiovasc Electrophysiol 2006;17(Suppl 1):S47–S56. 3. Ouwerkerk, R., Lee, R. F., Bottomley, P. A. Dynamic changes in sodium levels in human exercising muscle measured with 23 Na-MRI. Proc Lntl Soc Mag Reson Med 1999;7: p 1530. 4. Clausen, T. The sodium pump keeps us going. Ann N Y Acad Sci 2003;986: 595–602. 5. Jakovcevic, D., Harder, D. R. Role of astrocytes in matching blood flow to neuronal activity. Curr Top Dev iol 2007;79:75–97. 6. Magistretti, P. J., Pellerin, L. Cellular mechanisms of brain energy metabolism and their relevance to functional brain imaging. Philos Trans R Soc Lond iol Sci 1999;354: 1155–1163. 7. Magistretti, P. J. Neuron-glia metabolic coupling and plasticity. J Exp iol 2006;209: 2304–2311. 8. Silver, I. A., Erecinska, M. Energetic demands of the Na+ /K+ ATPase in mammalian astrocytes. Glia 1997;21:35–45. 9. Chatton, J. Y., Marquet, P., Magistretti, P. J. A quantitative analysis of L-glutamateregulated Na+ dynamics in mouse cor-

10.

11.

12.

13.

14.

15.

tical astrocytes: Implications for cellular bioenergetics. Eur J Neurosci 2000;12: 3843–3853. Constantinides, C. D., Gillen, J. S., Boada, F. E., Pomper, M. G., Bottomley, P. A. Human skeletal muscle: Sodium MR imaging and quantification-potential applications in exercise and disease. Radiology 2000;216: 559–568. Weber, M. A., Nielles-Vallespin, S., Huttner, H. B. et al. Evaluation of patients with paramyotonia at 23 Na MR imaging during cold-induced weakness. Radiology 2006;240:489–500. Hotta, Y., Ando, H., Takeya, K., Sakakibara, J. Direct measurement of increased myocardial cellular 23 Na NMR signals in perfused guinea-pig heart induced by dihydroouabain and grayanotoxin-I. Mol Cell iochem 1994;139:59–70. Constantinides, C. D., Kraitchman, D. L., O‘Brien, K. O., Boada, F. E., Gillen, J., Bottomley, P. A. Noninvasive quantification of total sodium concentrations in acute reperfused myocardial infarction using Na23 MRI. Magn Reson Med Sci 2001;46: 1144–1151. Igarashi, H. Studies of metabolic changes during and following cerebral ischemia in gerbils by in vivo multi nuclear magnetic resonance spectroscopy. Nippon Ika Daigaku Zasshi 1989;56:339–348. Perman, W. H., Heiberg, E. V., Hayes, C. E., Dunphy, T. 3D sodium-23 imaging of

Sodium MRI

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

breast lesions. Proc Int Soc Magn Reson Med Sci 1995;3:p 1196. Shimizu, T., Naritomi, H., Sawada, T. 23 Na-MRI Sequential changes on after cerebral infarction. Neuroradiology 1993;35:416–419. Thulborn, K. R., Davis, D., Snyder, J., Yonas, H., Kassam, A. Sodium MR imaging of acute and subacute stroke for assessment of tissue viability. Neuroimaging Clin N Am 2005;15:639–653, xi–xii. Stobbe, R., Beaulieu, C. Advantage of sampling density weighted apodization over postacquisition filtering apodization for sodium MRI of the human brain. Magn Reson Med 2008;60:981–986. Nielles-Vallespin, S., Weber, M. A., Bock, M. et al. 3D radial projection technique with ultrashort echo times for sodium MRI: Clinical applications in human brain and skeletal muscle. Magn Reson Med 2007;57:74–81. Clayton, D. B., Lenkinski, R. E. MR imaging of sodium in the human brain with a fast three-dimensional gradient-recalled-echo sequence at 4 T. Acad Radiol 2003;10: 358–365. Atkinson, I. C., Renteria, L., Burd, H., Pliskin, N. H., Thulborn, K. R. Safety of human MRI at static fields above the FDA ST guideline: sodium imaging at 9.4T does not affect vital signs or cognitive ability. J Magn Reson Imaging 2007;26:1222–1227. Sandstede, J. J. W., Hillenbrand, H., Beer, M. et al. Time course of Na-23 signal intensity after myocardial infarction in humans. Magn Reson Med 2004;52:545–551. Jacobs, M. A., Ouwerkerk, R., Wolff, A. C. et al. Multiparametric and multinuclear magnetic resonance imaging of human breast cancer: Current applications. Technol Cancer Res Treat 2004;3:543–550. Jacobs, M. A., Ouwerkerk, R., Kamel, I., Bottomley, P. A., Bluemke, D. A., Kim, H. S. Proton, diffusion-weighted imaging, and sodium (Na-23) MRI of uterine leiomyomata after MR-guided highintensity focused ultrasound: A preliminary study. J Magn Reson Imaging 2009;29: 649–656. Ouwerkerk, R., Bleich, K. B., Gillen, J. S., Pomper, M. G., Bottomley, P. A. Tissue sodium concentration in human brain tumors as measured with 23Na MR imaging. Radiology 2003; 227(2):529–537. Ouwerkerk, R., Jacobs, M. A., Macura, K. J. et al. Elevated tissue sodium concentration in malignant breast lesions detected with noninvasive Na-23 MRI. Breast Cancer Res Treat 2007;106:151–160.

199

27. Robson, M. D., Tyler, D. J., Neubauer, S. Ultrashort TE chemical shift imaging (UTECSI). Magn Reson Med 2005;53:267–274. 28. Kennedy, C. B., Balcom, B. J., Mastikhin, I. V. Three-dimensional magnetic resonance imaging of rigid polymeric materials using single-point ramped imaging with T-1 enhancement (SPRITE). Can J ChemRevue Canadienne De Chimie 1998;76: 1753–1765. 29. Romanzetti, S., Halse, M., Kaffanke, J., Zilles, K., Balcom, B. J., Shah, N. J. A comparison of three SPRITE techniques for the quantitative 3D imaging of the 23 Na spin density on a 4T whole-body machine. J Magn Reson 2006;179:64–72. 30. Idiyatullin, D., Corum, C., Park, J. Y., Garwood, M. Fast and quiet MRI using a swept radiofrequency. J Magn Reson 2006;181:342–349. 31. Ra, J. B., Hilal, S. K., Oh, C. H. An algorithm for MR imaging of the short T2 fraction of sodium using the FID signal. J Comput Assist Tomogr 1989;13: 302–309. 32. Jerecic, R., Bock, M., Wacker, C., Bauer, W., Schad, L. R. 23 Na-MRI of the human heart using a 3D radial projection technique. Biomed Tech (erl) 2002;47(Suppl 1 Pt 1):458–459. 33. Jerecic, R., Bock, M., Nielles-Vallespin, S., Wacker, C., Bauer, W., Schad, L. R. ECG-gated 23 Na-MRI of the human heart using a 3D-radial projection technique with ultra-short echo times. MAGMA 2004;16: 297–302. 34. Boada, F. E., Gillen, J. S., Shen, G. X., Chang, S. Y., Thulborn, K. R. Fast three dimensional sodium imaging. Magn Reson Med 1997;37:706–715. 35. Thulborn, K. R., Davis, D., Adams, H., Gindin, T., Zhou, J. Quantitative tissue sodium concentration mapping of the growth of focal cerebral tumors with sodium magnetic resonance imaging. Magn Reson Med 1999;41:351–359. 36. Ouwerkerk, R., Bleich, K. B., Gillen, J. S., Pomper, M. G., Bottomley, P. A. Tissue sodium concentration in human brain tumors as measured with 23 Na MR imaging. Radiology 2003;227:529–537. 37. Ouwerkerk, R., Weiss, R. G., Bottomley, P. A. Measuring human cardiac tissue sodium concentrations using surface coils, adiabatic excitation, and twisted projection imaging with minimal T2 losses. J Magn Reson Imaging 2005;21:546–555. 38. Gurney, P. T., Hargreaves, B. A., Nishimura, D. G. Design and analysis of a practi-

200

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

Ouwerkerk cal 3D cones trajectory. Magn Reson Med 2006;55:575–582. Ouwerkerk, R.. Using Heading and Speed to Design Gradient Waveforms with Lower Slew Rates. In: ISMRM workshop on nonCartesian MRI; 2007 25–28 Feb 2007; Sedona, AZ; 2007. Qian, Y. X., Boada, F. E. Acquisitionweighted stack of spirals for fast highresolution three-dimensional ultra-short echo time MR imaging. Magn Reson Med 2008;60:135–145. Du, J., Hamilton, G., Takahashi, A., Bydder, M., Chung, C. B. Ultrashort echo time spectroscopic imaging (UTESI) of cortical bone. Magn Reson Med 2007;58:1001–1009. Rahmer, J., Bornert, P., Dries, S. P. Assessment of anterior cruciate ligament reconstruction using 3D ultrashort echotime MR imaging. J Magn Reson Imaging 2009;29:443–448. Liu, W., Dahnke, H., Rahmer, J., Jordan, E. K., Frank, J. A. Ultrashort T2 ∗ relaxometry for quantitation of highly concentrated superparamagnetic iron oxide (SPIO) nanoparticle labeled cells. Magn Reson Med 2009;61:761–766. Jackson, J. I., Meyer, C. H., Nishimura, D. G., Macovski, A. Selection of a convolution function for Fourier inversion using gridding. IEEE Trans Med Imaging 1991;10: 473–478. Ouwerkerk, R., Vanechteld, C. J. A., Staal, G. E. J., Rijksen, G. Erythrocyte Na+ /K+ ATPase activity measured with Na-23 NMR. Magn Reson Medicine 1989;12:164–171. Knubovets, T., Shinar, H., Navon, G. Quantification of the contribution of extracellular sodium to 23 Na multiple-quantum-filtered NMR spectra of suspensions of human red blood cells. J Magn Reson 1998;131: 92–96. Winter, P. M., Poptani, H., Bansal, N. Effects of chemotherapy by 1,3-bis(2-chloroethyl)1-nitrosourea on single-quantum- and triple-quantum-filtered 23 Na and 31P nuclear magnetic resonance of the subcutaneously implanted 9L glioma. Cancer Res 2001;61:2002–2007. Dizon, J. M., Tauskela, J. S., Wise, D., Burkhoff, D., Cannon, P. J., Katz, J. Evaluation of triple-quantum-filtered 23 Na NMR in monitoring of intracellular Na content in the perfused rat heart: Comparison of intra- and extracellular transverse relaxation and spectral amplitudes. Magn Reson Med 1996;35: 336–345. Jelicks, L. A., Gupta, R. K. On the extracellular contribution to multiple quantum fil-

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

60. 61.

tered 23 Na NMR of perfused rat heart. Magn Reson Med 1993;29:130–133. Navon, G. Complete elimination of the extracellular 23 Na NMR signal in triple quantum filtered spectra of rat hearts in the presence of shift reagents. Magn Reson Med 1993;30:503–506. Boada, F. E., Tanase, C., Davis, D. et al. Non-invasive assessment of tumor proliferation using triple quantum filtered 23/Na-MRI: Technical challenges and solutions. Conf Proc IEEE Eng Med iol Soc 2004;7:5238–5241. Ooms, K. J., Cannella, M., Vega, A. J., Marcolongo, M., Polenova, T. Na23 TQF NMR imaging for the study of spinal disc tissue. J Magn Reson 2008;195: 112–115. Borthakur, A., Hancu, I., Boada, F. E., Shen, G. X., Shapiro, E. M., Reddy, R. In vivo triple quantum filtered twisted projection sodium MRI of human articular cartilage. J Magn Reson 1999;141:286–290. LaVerde, G., Nemoto, E., Jungreis, C. A., Tanase, C., Boada, F. E. Serial triple quantum sodium MRI during non-human primate focal brain ischemia. Magn Reson Med 2007;57:201–205. Kline, R. P., Wu, E. X., Petrylak, D. P. et al. Rapid in vivo monitoring of chemotherapeutic response using weighted sodium magnetic resonance imaging. Clin Cancer Res 2000;6:2146–2156. Rong, P., Regatte, R. R., Jerschow, A. Clean demarcation of cartilage tissue 23 Na by inversion recovery. J Magn Reson 2008;193: 207–209. Stobbe, R., Beaulieu, C. In vivo sodium magnetic resonance imaging of the human brain using soft inversion recovery fluid attenuation. Magn Reson Med 2005;54: 1305–1310. Cameron, I. L., Smith, N. K., Pool, T. B., Sparks, R. L. Intracellular concentration of sodium and other elements as related to mitogenesis and oncogenesis in vivo. Cancer Res 1980;40:1493–1500. Rotin, D., Steele-Norwood, D., Grinstein, S., Tannock, I. Requirement of the Na+ /H+ exchanger for tumor growth. Cancer Res 1989;49:205–211. Lagarde, A. E., Pouyssegur, J. M. The Na+ :H+ antiport in cancer. Cancer iochem iophys 1986;9:1–14. Nagy, I., Lustyik, G., Lukacs, G., Nagy, V., Balazs, G. Correlation of malignancy with the intracellular Na+ :K+ ratio in human thyroid tumors. Cancer Res 1983;43: 5395–5402.

Sodium MRI 62. Reshkin, S. J., Bellizzi, A., Caldeira, S. et al. Na+ /H+ exchanger-dependent intracellular alkalinization is an early event in malignant transformation and plays an essential role in the development of subsequent transformation-associated phenotypes. FASEB J 2000;14:2185–2197. 63. Patchett, A. A., Cordes, E. H. The design and properties of N-carboxyalkyldipeptide inhibitors of angiotensin-converting enzyme. Adv Enzymol Relat Areas Mol iol 1985;57: 1–84. 64. Hayes, C., Padhani, A. R., Leach, M. O. Assessing changes in tumour vascular function using dynamic contrast-enhanced magnetic resonance imaging. NMR iomed 2002;15:154–163. 65. Bartha, R., Megyesi, J. F., Watling, C. J. Low-grade glioma: correlation of short echo time 1H-MR spectroscopy with 23 Na MR imaging. AJNR Am J Neuroradiol 2008;29:464–470. 66. Wang, Y., Hu, W. X., Perez-Trepichio, A. D. et al. Brain tissue sodium is a ticking clock telling time after arterial occlusion in rat focal cerebral ischemia. Stroke 2000;31:1386–1391. 67. Jones, S. C., Kharlamov, A., Yanovski, B. et al. Stroke onset time using sodium MRI in rat focal cerebral ischemia. Stroke 2006;37:883–888. 68. Yushmanov, V. E., Yanovski, B., Kharlamov, A., LaVerde, G., Boada, F. E., Jones, S. C. Sodium mapping in focal cerebral ischemia in the rat by quantitative Na23 MRI. J Magn Reson Imaging 2009;29: 962–966. 69. Lin, S. P., Song, S. K., Miller, J. P., Ackerman, J. J. H., Neil, J. J. Direct, longitudinal comparison of H-1 and Na-23 MRI after transient focal cerebral ischemia. Stroke 2001;32:925–932. 70. Jacobs, M. A., Ouwekerk, R., Wolff, A. C. et al. Diffusion Weighted Imaging, ADC Mapping, and Sodium MR Imaging of Oper-

71.

72.

73.

74.

75.

76.

77.

201

able Breast Cancer After Neoadjuvant Therapy: Preliminary Results. In: Proceedings of the Seventeenth Meeting of the International Society of Magnetic Resonance in Medicine; 2009 May; Honolulu, HI, USA; 2009, p 2117. Jacobs, M. A., Ouwekerk, R., Wolff, A. C. et al. Multinuclear and Multiparametric MR imaging as an early treatment response biomarker for preoperative systemic therapy in breast cancer: Preliminary Results. In: Proceedings of the Seventeenth Meeting of the International Society of Magnetic Resonance in Medicine; 2009 May; Honolulu, HI, USA; 2009, p 2220. Jacobs, M. A., Ouwerkerk, R., Kamel, I., Bottomley, P. A., Bluemke, D. A., Kim, H. S. Proton, diffusion-weighted imaging, and sodium (23 Na) MRI of uterine leiomyomata after MR-guided high-intensity focused ultrasound: A preliminary study. J Magn Reson Imaging 2009;29:649–656. Mellon, E. A., Pilkinton, D. T., Clark, C. M. et al. Sodium MR imaging detection of mild Alzheimer disease: Preliminary study. Am J Neuroradiol 2009;30:978–984. Shen, G. X., Boada, F. E., Thulborn, K. R. Dual-frequency, dual-quadrature, birdcage RF coil design with identical B1 pattern for sodium and proton imaging of the human brain at 1.5 T. Magn Reson Med 1997;38:717–725. Tannus, A., Garwood, M. Improved performance of frequency-swept pulses using offset-independent adiabaticity. J Magn Reson Ser A 1996;120:133–137. Garwood, M., Yong, K. Symmetrical pulses to induce arbitrary flip angles with compensation for RF inhomogeneity and resonance offsets. J Magn Reson 1991;94: 511–525. Ouwerkerk, R., van Echteld, C. J., Staal, G. E., Rijksen, G. Erythrocyte Na+ /K+ ATPase activity measured with 23 Na NMR. Magn Reson Med 1989;12:164–171.

wwwwwww

Chapter 9 MR Spectroscopy and Spectroscopic Imaging of the Brain He Zhu and Peter B. Barker Abstract Magnetic resonance spectroscopy (MRS) and the related technique of magnetic resonance spectroscopic imaging (MRSI) are widely used in both clinical and preclinical research for the non-invasive evaluation of brain metabolism. They are also used in medical practice, although their ultimate clinical value continues to be a source of discussion. This chapter reviews the general information content of brain spectra and commonly used protocols for both MRS and MRSI and also touches on data analysis methods and quantitation. The main focus is on proton MRS for application in humans, but many of the methods are also applicable to other nuclei and studies of animal models as well. Key words: Brain, magnetic resonance spectroscopy, spectroscopic imaging, spatial localization, metabolites.

1. Introduction In vivo magnetic resonance spectroscopy (MRS) of the human brain has developed rapidly since its first observation in the 1980s (1, 2). Early studies in both humans and animals focused on the 31 P nucleus which allowed the measurement of energy metabolites such as phosphocreatine and ATP, as well as inorganic phosphate and phosphoesters (1). With the development of improved techniques for spatial localization and water suppression, proton MRS became more prevalent in the 1990s because of its higher sensitivity and greater convenience (since it can be performed without hardware modification on most MRI machines, unlike MRS of other nuclei) (3). While interest remains, particularly at high magnetic field strengths, in nuclei such as 31 P, 23 Na, and M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_9, © Springer Science+Business Media, LLC 2011

203

204

Zhu and Barker 13 C

(particularly for isotopically labeled and/or hyper-polarized molecules (4)), the vast majority of brain MRS studies in vivo use the proton. The remainder of this article therefore focuses on protocols for 1 H-MRS.

2. Information Content of Proton MR Spectra of the Brain

Because of its relatively low sensitivity, only small, mobile molecules which are present in millimolar quantities are generally detectable in an in vivo MR spectrum. At commonly used field strengths such as 1.5 or 3.0 T, only signals from choline (Cho), creatine (Cr), and N-acetylaspartate (NAA) are observed in normal brain at long echo times (e.g., 140 or 280 ms) (Fig. 9.1a), while compounds such as lactate, alanine, or others may be detectable in pathological conditions which increase their concentration (5–7). At short echo times (e.g., 35 ms or less) other compounds such as glutamate, glutamine, myo-inositol, as well as lipids and macromolecular resonances (Fig. 9.1b), are detectable. A summary of all compounds that have been detected in the human brain by proton MRS is given in Table 9.1, and a complete list of metabolite structures and their spectra can be found in (8). The biological significance of the major compounds is discussed below.

(a)

(b)

Fig. 9.1. 3 T PRESS brain spectra recorded from a 2-year-old boy with TE 135 ms (a) and TE 30 ms (b) at the level of the centrum semiovale with a nominal voxel size of 1.5 cm3 . In the long TE spectrum, signals are present from choline (Cho), creatine (Cr), and N-acetylaspartate (NAA), while in the short TE spectrum additional signals from myo-inositol (mI), glutamate and glutamine (Glx), and lipids (Lip) are present.

MR Spectroscopy and Spectroscopic Imaging of the Brain

205

Table 9.1 Compounds detected by proton MRS in the human brain

Compounds normally present

Compounds which may be detected under pathological or other abnormal conditions

Large signals at long TE

Long TE

N-Acetylaspartate (NAA) Creatine (Cr) and phosphocreatine (PCr) Cholines (Cho): Glycerophosphocholine (GPC) Phosphocholine (PC), free choline (Cho)

Lactate (Lac) β-Hydroxy-butyrate, acetone

Large signals at short TE

Short TE

Glutamate (Glu) Glutamine (Gln) myo-Inositol (mI)

Lipids Macromolecules Phenylalanine Galactitol

Small signals (short or long TE)

Exogenous compounds (short or long TE)

N-Acetylaspartylglutamate (NAAG), aspartate Taurine, betaine, scyllo-inositol, ethanolamine Threonine Glucose, glycogen Purine nucleotides Histidine

Propan-1,2-diol Mannitol Ethanol Methylsulfonylmethane (MSM)

Succinate, pyruvate Alanine Glycine

Small signals that can be detected with the use of spectral editing techniques γ-Amino-butyric acid (GABA) Homocarnosine Glutathione Threonine Vitamin C (ascorbic acid)

2.1. N-Acetylaspartate

NAA is the largest signal in the normal adult brain spectrum, resonating at 2.01 ppm, with a small and usually unresolved contribution from N-acetylaspartylglutamate (NAAG) at 2.04 ppm (9, 10). NAA is one of the most abundant amino acids in the central nervous system. It has been speculated to be a source of acetyl groups for lipid synthesis, a regulator of protein synthesis, a storage form of acetyl-CoA or aspartate, a breakdown product of NAAG (which, unlike NAA, is a neurotransmitter), or an osmolyte (11). NAA is synthesized in neuronal mitochondria, from aspartate and acetyl-coA. NAA is often referred to as a “neuronal marker,” since immunocytochemical studies have suggested that NAA is predominantly restricted to neurons, axons,

206

Zhu and Barker

and dendrites within the central nervous system (12). However, other studies have suggested that NAA may be found in nonneuronal cells, such as mast cells or isolated oligodendrocyte preparations (13–15). Overall, NAA does appear to be a good surrogate marker of neuronal health, but (as with all surrogate markers) it may sometimes change independent of neuron cell density or function. 2.2. Choline

The “choline” signal (“Cho,” 3.20 ppm) is a composite peak consisting of contributions from the trimethylamine groups of glycerophosphocholine (GPC), phosphocholine (PC), and a small amount of free choline itself (16). These compounds are involved in membrane synthesis and degradation, and they are often elevated in disease states where increased membrane turnover is involved (e.g., tumors). Glial cells have also been reported to have high levels of Cho (17, 18). Other pathological processes which lead to Cho elevation include active demyelination (19), either resulting from the degradation of myelin phospholipids primarily to GPC or perhaps due to inflammation (20). Low brain Cho has been observed in hepatic encephalopathy (21), and there is also some evidence to suggest that dietary intake of choline can modulate cerebral Cho levels (22).

2.3. Creatine

The “creatine” methyl resonance (“Cr,” 3.03 ppm) is a composite peak consisting of both creatine and phosphocreatine, compounds that are involved in energy metabolism via the creatine kinase reaction, generating ATP. A resonance from the CH2 of creatine can also be observed at 3.91 ppm. In vitro, glial cells contain a two- to fourfold higher concentration of creatine than do neurons (23). Creatine also shows quite large regional variations, with lower levels in white matter than gray matter in normal brain, as well as very high levels of Cr in the cerebellum compared to supratentorial regions (24).

2.4. Lactate

The lactate resonance (a doublet with a 7 Hz coupling constant centered at 1.31 ppm) is usually not detectable in the brain under normal conditions. However, lactate is often detected by MRS in pathological conditions such as acute hypoxic (25) or ischemic (5, 26) injury, or in brain tumors (27) or mitochondrial diseases (7, 28).

2.5. myo-Inositol

One of the larger signals in short echo time spectra occurs from myo-inositol (mI) at 3.5–3.6 ppm. mI is a pentose sugar, which is part of the inositol triphosphate intracellular second messenger system. Glial cells in vitro have been shown to contain higher levels of mI than neurons (29, 30). mI has been reported to be reduced in hepatic encephalopathy (31), and increased in Alzheimer’s dementia (32) and demyelinating diseases (33).


207

2.6. Glutamate and Glutamine

Glutamate (Glu) is the most abundant amino acid in the brain and is the dominant neurotransmitter (34). At 1.5 T, there is almost complete overlap of Glu and glutamine (Gln), and they are detected as a composite “Glx” peak (21). At higher fields (3.0 T and above), Glu and Gln become better resolved and can be quantified individually with good accuracy using appropriate spectral analysis techniques (35). Glu has been found to be elevated in MS plaques (36), and elevated cerebral Gln is commonly observed in patients with liver failure (for example, hepatic encephalopathy (31) and Reye’s syndrome (37)).

2.7. Less Commonly Detected Compounds

Approximately 25 additional compounds have been detected in proton spectra of the human brain (Table 9.1). Some of these compounds are present in the normal human brain, but are difficult to detect routinely because they are very small and/or have overlapping peaks. Some examples of these compounds include NAAG, aspartate, taurine, scyllo-inositol, betaine, ethanolamine, purine nucleotides, histidine, glucose, and glycogen (38). Other compounds are yet more difficult to detect and require the use of “spectral editing” techniques (see later), because in conventional spectra they overlap and are obscured by much larger signals. Examples of compounds requiring spectral editing to be measured include γ-amino-butyric acid (GABA) and glutathione (GSH) (39, 40). Some compounds are only detected under disease or other abnormal conditions. Examples include the ketone bodies βhydroxy-butyrate and acetone (41, 42) in patients who are ketotic and other compounds such as phenylalanine (in phenylketonurea (43)), galactitol, ribitol, arabitol in “polyol disease” (44), and succinate, pyruvate, alanine, glycine, and threonine in various disorders. Exogenous compounds which are able to cross the blood– brain barrier may be detected by proton MRS; examples include the drug delivery vehicle propan-1,2-diol (45), ethanol (46), and methylsulfonylmethane (MSM) (47). Histidine, homocarnosine, and the amide resonance of NAA are low signal intensity compounds downfield from water which can be detected by the use of short echo times, appropriate water suppression methods, and high magnetic field strengths. Using oral loading of histidine, Vermathen et al. were able to estimate brain pH from the chemical shift difference of the C2 and C4 resonances of the imidazole side chain of histidine (48); similarly, Rothman et al. were able to use the imidazole resonances of homocarnosine to estimate brain pH in epilepsy patients who were receiving vigabatrin (49). The rate of exchange of the NAA amide protons with water is also pH sensitive and can be used to estimate brain pH (50).

208

Zhu and Barker

3. Spatial Localization Techniques 3.1. Single-Voxel Techniques

Nearly all single-voxel localization techniques use three orthogonal slice-selective pulses to select a signal from the region (“voxel”) where they intersect (Fig. 9.2a). Signals from outside the voxel are removed by the use of “crusher” field gradient

Fig. 9.2. Single-voxel localization techniques: (a) spatial localization is achieved by collecting signals from the intersection of three slice-selective RF pulses applied in orthogonal directions; (b) the STEAM sequence, consisting of three 90◦ slice-selective pulses; (c) the PRESS sequence, consisting of a slice-selective 90◦ excitation pulse and two 180◦ sliceselective refocusing pulses; (d) the LASER sequence, which uses a non-slice-selective adiabatic half-passage excitation pulse, followed by three pairs of hyperbolic secant 180◦ refocusing pulses; (e) the semi-LASER sequence, which uses a slice-selective 90◦ excitation pulse and two pairs of hyperbolic secant 180◦ refocusing pulses; (f) the “SPECIAL” pulse sequence, which uses an alternating slice-selective 180◦ inversion pulse (every second average) in combination with a 90◦ –180◦ bar-selective spin echo.


209

pulses, alternating the phases of the slice-selective pulses and receiver (‘phase-cycling’), and the use of outer-volume suppression pulses (51, 52) (53, 54). Typical voxel sizes for human brain spectroscopy are 4–8 cm3 . The “STEAM” sequence (52, 55) (Fig. 9.2b) uses three 90◦ pulses to form a “stimulated echo,” while the “PRESS” sequence (Fig. 9.2c) uses one 90◦ and two 180◦ refocusing pulses to create a spin echo. STEAM and PRESS have been compared in detail (56); perhaps the biggest difference is that the spin-echobased PRESS sequence has twice the signal compared to STEAM and is therefore often preferred. However, advantages of STEAM include better slice profiles and higher bandwidth of the 90◦ pulses, lower RF power requirements, and the ability to obtain shorter echo times. In this regard, STEAM may be particularly advantageous for brain MRS at high field strengths (e.g., above 3 T (35)). Short TE STEAM may be preferable for observing resonances with shorter T2 s (57), while long TE PRESS (with its superior SNR) should generally be used for resonances with longer T2 s (such as Cho, Cr, NAA, and lactate). In vivo MRS performed at high field strengths (e.g., 3 T or higher) is associated with additional technical challenges. For instance, uniform RF transmit (B1 ) fields become difficult to achieve because of wavelength effects in volume RF coils (58) or when using inhomogeneous surface coils for excitation. In either case, it may be difficult to achieve the desired flip angles in PRESS or STEAM, and the flip angles may vary inside the voxel, resulting in signal loss. To address these problems, adiabatic excitation or refocusing pulses have been implemented in techniques such as “LASER” (localization by adiabatic selective refocusing; 59, 60) (Fig. 9.2d) or its simplified version “semi-LASER” (61, 62) (Fig. 9.2e). The LASER sequence consists of a non-slice-selective adiabatic half-passage 90◦ pulse for excitation and three pairs of hyperbolic secant (HS) refocusing pulses in three directions for localization. Since a single HS 180◦ pulse with a slice-selection gradient produces a large first-order phase variation across the spectrum, two consecutive HS pulses are needed to cancel it out (63, 64). The LASER sequence produces a more uniform excitation profile and takes advantage of the large bandwidths of the adiabatic HS pulses to reduce chemical shift displacement errors. However, the large number of RF pulses used in LASER results in higher RF power requirement and longer TE compared to conventional localization sequences. The semi-LASER sequence consists of a non-adiabatic 90◦ slice-selective pulse and two pairs of adiabatic HS pulses for refocusing as in LASER; while some insensitivity to B1 inhomogeneity is lost, this sequence does have reduced RF power and can achieve shorter TE than LASER. At very high fields for human MRS such as 7 T, apparent metabolite T2 relaxation times are significantly shorter than at

210

Zhu and Barker

lower field strengths, and it is therefore desirable to minimize the TE of the localization sequence as much as possible. A localization technique dubbed “SPECIAL” (spin-echo full-intensity acquired localized spectroscopy; 65) () combines desirable features such as the full signal intensity of PRESS and the shorter TE of STEAM. The sequence consists of a slice-selective inversion pulse followed by a spin-echo sequence, with each pulse applied in a different direction (Fig. 9.2f). The sequence collects full-intensity signal from a 1D strip-like volume defined by the intersection of the selected slices of the 90◦ and 180◦ pulses. The slice-selective inversion pulse is applied to every other TR (similar to what is used in the “ISIS” experiment (66)) so that a minimum of two scans are required to achieve full spatial localization. This sequence is a promising technique to investigate compounds with short T2 s in vivo. 3.2. Multiple-Voxel (MRSI) Techniques

While single-voxel MRS can be performed quickly and easily in most parts of the human brain, it provides no information on the spatial variations of metabolites and is generally limited to one or two brain regions in most clinical studies. In contrast, MRSI is usually more time-consuming but can be used to measure multiple-voxel locations simultaneously. Most often, MRSI is based on signal excitation of a restricted region using the PRESS sequence in combination with phaseencoding in two directions (Fig. 9.3) (55). This allows B0 field homogeneity to be optimized on the desired region of interest, limits the number of phase-encoding steps needed for a given spatial resolution, and avoids exciting lipid signals from the scalp. Other methods for lipid suppression are discussed later.

Fig. 9.3. 2D-PRESS-MRSI pulse sequence. A PRESS sequence is used to excite a large volume of brain tissue while excluding signal from lipid in the scalp and/or regions of poor field homogeneity, and then phase-encoding gradients (in blue, GY and GZ ) are used to localize spectra from regions within the excited region. A CHESS prepulse and crusher gradient are applied for water suppression. Crusher gradients applied around the 180◦ refocusing pulses are also shown.


211

Fig. 9.4. 2D-PRESS-MRSI scan (3 T, TR/TE 1,700/135 ms, nominal voxel size 1.5 cm3 ) in the axial plane of a 3-year-old girl with developmental delay. The central 6×6 spectra are shown (indicated in red on the localizer T2 -weighted MRI).

Figure 9.4 shows the results of this sequence from a 3-year-old girl with an idiopathic developmental delay. However, PRESS-MRSI also has shortcomings, including unreliable spectra at the edges of the PRESS box due to imperfect slice profiles of the 180◦ pulses, the inability to perform multislice acquisitions (although 3D is possible), and the difficulty in covering to the edges of the brain because of the rectangular shape of the PRESS excitation. An alternative approach is a sliceselective spin-echo sequence that excites a whole transverse slice (Fig. 9.5) and which can be used in a multi-slice mode (53). Preceding the spin-echo sequence there are usually multiple, carefully placed OVS pulses to suppress the lipid signals from the scalp (53), as well as the usual water suppression pulses. Figure 9.6 shows an example of one slice from a multi-slice 2D MRSI data set from a normal volunteer. To minimize artifacts due to residual water and lipid and also field inhomogeneity, MRSI with large spatial coverage is typically performed at long echo time (e.g., 140 or 280 ms). The multi-slice technique can allow a sufficient number of slices to cover the whole brain but the resulting scan time can be too long with conventional phase-encoding techniques. Specifically, the length of the pulse sequence for each slice is in the range of 0.5–1.0 s including all RF pulses and data acquisition window, which needs to be long enough to gain enough spectral resolution. Four or five such slices interleaved can result in a TR prohibitively long that causes long scan times (67). 3D-PRESSMRSI, on the other hand, can also lead to very long scan times if large brain coverage is prescribed. The number of phase-encoding steps (N) is equal to the field of view (FOV) divided by the desired

212

Zhu and Barker

Fig. 9.5. A multi-slice 2D-MRSI pulse sequence. A slice-selective spin echo is preceded by an optimized water and lipid suppression scheme (“HGDB”), including outer-volume suppression (OVS) pulses (8, indicated in red) for spatial suppression of scalp signals (103). 2D phase-encoding gradients (blue) are applied on GY and GZ . In this example, three slices are collected within one repetition time (TR). Gradients associated with the water and lipid suppression are omitted for clarity.

spatial resolution (N = FOV/ ). Therefore, in order to minimize the scan time (i.e., minimize N) without reducing a desired spatial resolution, it is important to prescribe a FOV as small as possible constrained only by the dimensions of the object to be imaged. In the case of brain imaging, the left–right FOV should be smaller than the anterior–posterior, since the brain (usually of an oval shape) is smaller in this dimension (68). In general, scan time can be reduced by an additional 25–30% with a reduced FOV in the left–right direction. Generally, if large FOVs and high resolutions are sought in all three dimensions, fast MRSI techniques are required to maintain clinically reasonable scan times. 3.3. Fast MRSI Techniques

A number of different approaches for fast MRSI have been developed and reviewed previously (69). Some of the more frequently used methods are discussed here.

3.3.1. “Turbo” or Fast Spin-Echo MRSI

One of the earliest approaches to fast MRSI was to use a multipleecho acquisition, with each echo having its own phase-encoding gradient (70) (Fig. 9.7a). Typically 3 or 4 echoes are used, reducing scan time by a factor of 3 or 4 if the same repetition time (TR) is used as in a single-echo experiment. This approach does suffer from several limitations, however: the spectral resolution is limited by having to keep each echo readout short, while the later echoes suffer from reduced signal due to T2 relaxation. Compounds with short T2 relaxation times cannot be observed with this technique. For multiple echoes, the minimum TR is generally


213

Fig. 9.6. Example data from one slice (at the level of the lateral ventricles) of a multi-slice 2D-MRSI data from a normal human subject recorded at 3.0 T using the pulse sequence of Fig. 9.5. TR 2.5 s, TE 140 ms, nominal voxel size 0.65 cm3 . In addition to the T1 -weighted MRI scan, spectroscopic images of choline, creatine, and N-acetylaspartate and selected spectra from the left and right hemispheres are shown.

longer than for a single-echo acquisition, so that time-savings may be less than expected. For these reasons, multi-echo MRSI has only seen limited adoption in practice, although it has been applied to studies of brain tumors (71). 3.3.2. Echo-Planar (EPSI) and Spiral-MRSI

In EPSI, an oscillating read gradient is applied during data acquisition so that both spectral and spatial information are collected simultaneously (72) (73). The oscillating read gradient can be viewed as repeatedly collecting one line of k-space at different time points. Conventional phase-encoding is then applied in the other one or two directions to extend the experiment to either two or three spatial dimensions, respectively (Fig. 9.7b). The EPSI readout reduces the number of phase-encoding steps by an order of magnitude compared to conventional MRSI, thereby achieving a large scan time reduction. EPSI-encoding can be

214

Zhu and Barker

Fig. 9.7. Different readout strategies for fast MRSI pulse sequences. In the examples shown here, all three sequences use spin-echo excitation preceded by CHESS water suppression and OVS lipid suppression, although other excitation and suppression sequences can be used. (a) In fast spin-echo (“turbo”) MRSI, multiple spin echoes are acquired, each one with its own phase-encoding gradient; (b) in echo-planar spectroscopic imaging (EPSI), an oscillating read gradient is applied during data acquisition; and (c) in spiral-MRSI, two oscillating read gradients are applied during data acquisition. For full 3D-encoding, conventional phase-encoding gradients can be applied in the remaining directions.

implemented as 2D multi-slice with spin-echo excitation, or 3D with a PRESS or spin-echo thick slab excitation (74, 75). EPSI is one of the fastest acquisition techniques for MRSI, but does have some limitations. High-performance gradient systems are required, and any imbalances between positive and negative gradient lobes can lead to “ghost” artifacts in the metabolic images. This problem can be addressed by processing the readout lines of two opposite directions separately, with subsequent combination of the two data sets after spatial transformation. This solution, however, reduces the available spectral bandwidth, which may already be quite low in EPSI (76). SNR may also be slightly lower than conventional MRSI recorded in the same scan time depending on the readout gradient waveform and whether ramp sampling is used or not. Because of its speed and much improved post-processing methods, EPSI (also called “PEPSI” (proton EPSI)) has been gaining popularity in whole-brain 3D MRSI (75) and has also been used to study various brain pathologies (77, 78). Spiral-MRSI is similar to EPSI in that read gradients are applied during data acquisition (Fig. 9.7c). In spiral-MRSI, however, gradient waveforms in 2D are applied so that k-space data


215

are sampled from the center to the edge along a spiral trajectory. These gradient waveforms are repeated several times concurrent with the evolution of the readout time in one TR. The k,t-space can be filled using multiple shots to satisfy the desired FOV and spatial and spectral resolutions (79). Post-processing of spiral-MRSI data usually starts with a “regridding” procedure (80) where raw data are resampled onto a Cartesian k-space grid by interpolation, after which conventional MRSI processing by Fourier transformation can be done. Spiral-MRSI has some unique advantages over EPSI, including the ability to manipulate the point-spread function, scan time, and SNR by varying the sample spacing in k-space with a variable density trajectory (81, 82). In addition, if the center of k-space is collected at the beginning of every spiral readout, it is possible to correct errors in phase or frequency associated with motion or other processes (83). Like EPSI, spiral-MRSI applies a read gradient during data acquisition, so it shares a similar level of dependence on the gradient system performance. EPSI is being applied to clinical applications, but is partly hampered due to lack of commercial availability and the need for dedicated reconstruction software. However, spiral-MRSI has been used to map metabolic abnormalities in patients with multiple sclerosis (84). 3.3.3. Parallel Encoded-MRSI

Parallel imaging techniques originally developed for speeding up MRI can also be adopted for MRSI (85, 86). The basic principle is to use the different sensitivity profiles of multiple, phasedarray receiver coils to encode spatial information, so that fewer phase-encoding steps are required, thereby reducing scan time. In the “SENSE” approach, Fourier transformation of undersampled k-space data leads to “aliased” spectroscopic images from each channel, which can then be unfolded and reconstructed using each coil’s sensitivity profile to produce a single spectroscopic image with uniform sensitivity. Alternatively, algorithms such as “SMASH” and “GRAPPA” (87–89) can be implemented to interpolate the missing k-space data points, which are then Fourier transformed as in conventional MRSI. Both “SENSE” (90) and “GRAPPA” (91) MRSI have been successfully implemented in humans. Reducing scan time using parallel encoding is an attractive option since it can be performed with any existing MRSI pulse sequence. 2D- and 3D-MRSI involve phase-encoding in multiple directions, so SENSE-encoding can also be performed in 2D or 3D (provided that receive arrays with appropriate geometry are available) leading to large scan time reductions (90). An example of a multi-slice MRSI scan with a SENSE factor of 6=3×2, in AP and RL directions, applied using a 32-channel head coil, is shown in a patient with a brain tumor in Fig. 9.8. This scan took 5:05 min; with conventional phase-encoding, scan

216

Zhu and Barker

Fig. 9.8. An example of 2D-SENSE-MRSI in a patient with a high-grade left frontal glioma recorded post-treatment at 3.0 T. TR 2.5 s, TE 140 ms, nominal voxel size 0.65 cm3 , SENSE acceleration factor = 6, scan time 5:05 min. FLAIR and noncontrast T1 -weighted (“MP-RAGE”) MR images, spectroscopic images (choline, creatine, NAA), and selected spectra are shown. The core of the lesion is characterized by reduced levels of NAA and other metabolites, consistent with necrotic tissue, while the T2 hyperintense perilesional areas demonstrate elevated Cho compared to the contralateral hemisphere, consistent with residual or recurrent tumor.

time would have been more than 30 min. Scan times on the order of 1–2 min can be achieved by combining SENSE-MRSI with other fast MRSI techniques, such as Turbo MRSI (92) or EPSI (93), as long as sufficient SNR is available. Parallel MRSI also has some potential problems. Errors in the coil sensitivity profiles and/or use of too high SENSE factors will lead to incomplete unfolding of MRSI data and the presence of artifacts. Unfolding of strong peri-cranial lipid signals is particularly challenging; successful SENSE-MRSI requires the application of efficient lipid suppression techniques (94) (see later). 3.4. Other Approaches to MRSI of the Brain

Other approaches to MRSI of the brain exist that are based on a t1 evolution period (as in high-resolution 2D NMR spectroscopy) to encode spectral information, with a fast imaging readout to determine spatial information. Scan times for these methods depend on the number of t1 values required to obtain sufficient spatial resolution and may be relatively short compared to conventional MRSI. These methods show promise in small animal studies, but for the most part have not been applied in humans. The sensitivity may be somewhat lower than in conventional MRSI because of the T2 signal decay that occurs during t1 time period. However, they do offer some unique advantages, such as using a


217

constant evolution time with a sliding refocusing pulse, to produce “homonuclear decoupled” proton brain spectra (95).

4. Water and Lipid Suppression

Brain metabolites observed by MRS are in the millimolar concentration range, while brain water is approximately 80 M. Lipids in scalp tissue are also present in very high concentrations. Therefore, efficient water suppression (and lipid suppression for MRSI) is vital for the reliable observation and measurement of brain metabolites. The most common method for water suppression is to presaturate the water signal using frequency-selective saturation pulses applied prior to the localization sequence (“CHESS”) (96). Multiple CHESS pulses with optimized flip angles and delays can be used to give good suppression factors over a range of transmit B1 values and water T1 relaxation times (which is important for the suppression of both brain water and CSF) (97, 98). For example, the “WET” scheme employs up to four Gaussian pulses (98), while the “VAPOR” scheme consists of seven pulses (57). Occasionally, it may be desirable to perform water suppression during the localization sequence, either opposed, or in addition, to presaturation. For instance, a water suppression pulse can be placed in the STEAM sequence between the second and third 90◦ pulses because the magnetization of the stimulated echo pathway is stored along the Z-axis during this time period. This can be used to improve suppression compared to presaturation only. It is also possible to include frequency-selective refocusing pulses inside a localization sequence to improve water suppression (e.g., the “MEGA” or “BASING” sequences) (99, 100). Lipid suppression is commonly performed in three different ways. One method, to suppress lipid signals in the scalp, is to use spatial outer-volume suppression (OVS) pulses (Fig. 9.5) (53). Alternatively, an inversion recovery scheme can be used, taking advantage of the difference in T1 values between lipid (typically 300 ms at 1.5 T) and metabolites (1,000–2,000 ms) (101). At 1.5 T, an inversion time of 200 ms (= T1 ∗ ln [2]) will selectively null the lipid signal, while most of the metabolite magnetization remains inverted. This method can suppress lipid signals anywhere in the brain because no assumption is made about the spatial distribution of the lipid, but may somewhat reduce metabolite SNR. Finally, lipid suppression may also be performed using frequencyselective saturation pulses, similar to frequency-selective water suppression techniques such as CHESS (94). Recently, methods

218

Zhu and Barker

have been designed which combine both water and lipid suppression into a single, dual band approach (94, 102, 103).

5. Spectral Editing Techniques

As mentioned above, certain molecules, such as GABA or glutathione, are almost totally obscured in conventional brain MRS by signals from other compounds which are present at much higher concentrations. “Spectral editing” techniques are required in order to detect these molecules while suppressing the signal from the unwanted compounds. A commonly used sequence for this purpose is the so-called “MEGA-PRESS” sequence (Fig. 9.9). Spectral editing makes use of molecules which contain “coupled” spin systems – the presence of coupling (J, measured in Hz) between functional groups allows the signal on one group to be modulated by applying a selective radiofrequency pulse on the other. For instance, for GABA, setting the frequency of the selective editing pulse to 1.9 ppm resonance of GABA will refocus the outer two peaks of the 3.02 ppm GABA pseudo-triplet. A second scan is performed without the selective pulse, and with TE=1/J (68 ms), the unaffected modulation results in two inverted peak (the outer two lines of the triplet) at 3.02 ppm. By subtracting the first scan from the second, the 3.02 GABA resonance can be selected (Fig. 9.10a) (99). For glutathione, the resonance at 2.95 ppm can be observed with the selective editing pulse set to 4.56 ppm and TE of 130 ms (Fig. 9.10b). A similar approach

Fig. 9.9. Pulse sequence for spectral editing (“MEGA-PRESS”). Frequency-selective editing pulses are added (blue) to the conventional PRESS sequence on alternating scans. The TE (= TE1 +TE2 ) is set equal to 1/J for doublets. Subtraction of alternating scans causes cancellation of all signals not effected by the editing pulses, leaving only the target edited molecules.


219

Fig. 9.10. Examples of (a) GABA- and (b) glutathione-edited spectra using the MEGAPRESS pulse sequence. TR 2 s, voxel size 3.5×3.5×3.5 = ∼43 cm3 , centered on the anterior cingulate gyrus, with TE 68 ms for GABA and 130 ms for GSH. Scan time was 8 min 32 s for GABA and 17 min 4 s for GSH. Note that in (a) glutamate/glutamine (Glx) coedit with GABA, and in (b) the aspartyl resonances of NAA coedit with GSH.

(called the “BASING” sequence) can be used to selectively detect brain lactate without contamination from lipids, making use of the coupled lactate resonances at 1.3 and 4.1 ppm (104).

6. Data Analysis and Quantification The concentration of a metabolite is linearly proportional to its spectral peak area. However, peak area measurements in in vivo spectroscopy are complicated by resonance overlap, baseline distortions, and non-ideal lineshapes and will also depend on factors such as relaxation times, pulse sequence used, and scanner hardware (e.g., receiver gain, coil loading). Various methods have been used to measure peak areas, ranging from simple integration, to fitting algorithms in the time- or frequency domains

220

Zhu and Barker

Fig. 9.11. An example of the “LCModel” analysis method. The experimental data are fit as a linear combination of spectra of pure compounds recorded under the same experimental conditions as the in vivo spectrum. Automated baseline and phase correction is performed, and an estimate of metabolite concentrations provided relative to the brain water signal. In this example of a 2×2×2 cm PRESS spectrum recorded at 3.0 T from a normal control subject (TR/TE/number of averages = 2,000/35/128), the difference between the original data and the curve-fit (red) is shown in the top trace. Metabolite concentrations in blue correspond to those with an estimated uncertainty of less than 20%.

(105, 106). One of the more widely used methods for spectral quantitation in recent years is the linear combination model (“LCModel”) method developed by Provencher et al. (107) (Fig. 9.11). The LCModel fits the in vivo spectrum as a combination of pure, model spectra from each of the expected compounds in the brain (107). The model also includes automatic phase correction and baseline correction, or the baseline may also be modeled as a combination of macromolecular resonances. Provided that each scanner is properly calibrated with the appropriate model solutions, the program returns metabolite concentrations (relative to an unsuppressed water signal) as well as estimates of uncertainty (e.g., Cramer–Rao lower bounds). Quantification methods based on internal or external standards have been extensively developed and tested for single-voxel spectroscopy (108). With care, it is also possible to quantify spectroscopic images (109). Many studies, however, do not attempt to quantify metabolite concentrations, but rather report relative amounts (ratios) of each metabolite, often using Cr as a reference. While ratios have some inherent advantages, for instance to account for partial volume effects or to enhance spectroscopic “contrast” in conditions where metabolites may change in opposite directions (e.g., Cho increases, NAA decreases), they also may be misleading if all metabolites are changing simultaneously. In


221

particular, Cr shows quite marked regional variations (24), and often changes in pathology, so caution should be used when interpreting ratios of metabolites to Cr.

7. Conclusions MRS and MRSI are mature techniques that are very commonly used for research studies in both humans and animal models. The protocols described in this chapter represent the most widely used and validated techniques currently used, but are by no means comprehensive. MRS protocol development, particularly for high-field applications, hyper-polarization, and fast MRSI techniques continue to be an active area of research investigation.

Acknowledgments Supported in part by National Institutes of Health grant P41 RR015241. References 1. Cady, E. B., Costello, A. M., Dawson, M. J., Delpy, D. T., Hope, P. L., Reynolds, E. O., Tofts, P. S., Wilkie, D. R. Noninvasive investigation of cerebral metabolism in newborn infants by phosphorus nuclear magnetic resonance spectroscopy. Lancet 1983;1(8333):1059–1062. 2. Bottomley, P. A., Edelstein, W. A., Foster, T. H., Adams, W. A. In vivo solventsuppressed localized hydrogen nuclear magnetic resonance spectroscopy: A window to metabolism? Proc Natl Acad Sci USA 1985;82(7):2148–2152. 3. Barker, P. B., Lin, D. D. In vivo proton MR spectroscopy of the human brain. Prog NMR Spect 2006;49:99–128. 4. Goldman, M., Johannesson, H., Axelsson, O., Karlsson, M. Hyperpolarization of 13C through order transfer from parahydrogen: A new contrast agent for MRI. Magn Reson Imaging 2005;23(2):153–157. 5. Barker, P. B., Gillard, J. H., van Zijl, P. C., Soher, B. J., Hanley, D. F., Agildere, A. M., Oppenheimer, S. M., Bryan, R. N.

6.

7.

8.

9.

Acute stroke: Evaluation with serial proton MR spectroscopic imaging. Radiology 1994;192(3):723–732. Remy, C., Grand, S., Lai, E. S., Belle, V., Hoffmann, D., Berger, F., Esteve, F., Ziegler, A., Le Bas, J. F., Benabid, A. L., Decorps, M., Segebarth, C. M. 1H MRS of human brain abscesses in vivo and in vitro. Magn Reson Med 1995;34(4):508–514. Lin, D. D., Crawford, T. O., Barker, P. B. Proton MR spectroscopy in the diagnostic evaluation of suspected mitochondrial disease. AJNR Am J Neuroradiol 2003;24(1):33–41. Govindaraju, V., Young, K., Maudsley, A. A. Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed 2000;13(3):129–153. Frahm, J., Michaelis, T., Merboldt, K. D., Hanicke, W., Gyngell, M. L., Bruhn, H. On the N-acetyl methyl resonance in localized 1H NMR spectra of human brain in vivo. NMR Biomed 1991;4(4): 201–204.

222

Zhu and Barker

10. Pouwels, P. J., Frahm, J. Differential distribution of NAA and NAAG in human brain as determined by quantitative localized proton MRS. NMR Biomed 1997;10(2):73–78. 11. Barker, P. B. N-acetyl aspartate – a neuronal marker? Ann Neurol 2001;49(4):423–424. 12. Simmons, M. L., Frondoza, C. G., Coyle, J. T. Immunocytochemical localization of Nacetyl-aspartate with monoclonal antibodies. Neuroscience 1991;45(1):37–45. 13. Bhakoo, K. K., Pearce, D. In vitro expression of N-acetyl aspartate by oligodendrocytes: Implications for proton magnetic resonance spectroscopy signal in vivo. J Neurochem 2000;74(1):254–262. 14. Burlina, A. P., Ferrari, V., Facci, L., Skaper, S. D., Burlina, A. B. Mast cells contain large quantities of secretagogue-sensitive Nacetylaspartate. J Neurochem 1997;69(3): 1314–1317. 15. Urenjak, J., Williams, S. R., Gadian, D. G., Noble, M. Specific expression of Nacetylaspartate in neurons, oligodendrocytetype-2 astrocyte progenitors, and immature oligodendrocytes in vitro. J Neurochem 1992;59(1):55–61. 16. Barker, P., Breiter, S., Soher, B., Chatham, J., Forder, J., Samphilipo, M., Magee, C., Anderson, J. Quantitative proton spectroscopy of canine brain: In vivo and in vitro correlations. Magn Reson Med 1994;32: 157–163. 17. Gill, S. S., Small, R. K., Thomas, D. G., Patel, P., Porteous, R., van Bruggen, N., Gadian, D. G., Kauppinen, R. A., Williams, S. R. Brain metabolites as 1H NMR markers of neuronal and glial disorders. NMR Biomed 1989;2(5–6):196–200. 18. Gill, S. S., Thomas, D. G., Van, B. N., Gadian, D. G., Peden, C. J., Bell, J. D., Cox, I. J., Menon, D. K., Iles, R. A., Bryant, D. J. Proton MR spectroscopy of intracranial tumours: In vivo and in vitro studies. J Comput Assist Tomogr 1990;14(4):497–504. 19. Davie, C. A., Hawkins, C. P., Barker, G. J., Brennan, A., Tofts, P. S., Miller, D. H., McDonald, W. I. Detection of myelin breakdown products by proton magnetic resonance spectroscopy. Lancet 1993;341(8845):630–631. 20. Brenner, R. E., Munro, P. M., Williams, S. C., Bell, J. D., Barker, G. J., Hawkins, C. P., Landon, D. N., McDonald, W. I. The proton NMR spectrum in acute EAE: The significance of the change in the Cho:Cr ratio. Magn Reson Med 1993;29(6):737–745. 21. Kreis, R., Ross, B. D., Farrow, N. A., Ackerman, Z. Metabolic disorders of the brain in chronic hepatic encephalopathy

22.

23.

24.

25.

26.

27.

28.

29.

30.

31. 32.

detected with H-1 MR spectroscopy. Radiology 1992;182(1):19–27. Stoll, A. L., Renshaw, P. F., De Micheli, E., Wurtman, R., Pillay, S. S., Cohen, B. M. Choline ingestion increases the resonance of choline-containing compounds in human brain: An in vivo proton magnetic resonance study. Biol Psychiatry 1995;37(3):170–174. Urenjak, J., Williams, S. R., Gadian, D. G., Noble, M. Proton nuclear magnetic resonance spectroscopy unambiguously identifies different neural cell types. J Neurosci 1993;13:981–989. Jacobs, M. A., Horska, A., van Zijl, P. C., Barker, P. B. Quantitative proton MR spectroscopic imaging of normal human cerebellum and brain stem. Magn Reson Med 2001;46(4):699–705. Penrice, J., Cady, E. B., Lorek, A., Wylezinska, M., Amess, P. N., Aldridge, R. F., Stewart, A., Wyatt, J. S., Reynolds, E. O. Proton magnetic resonance spectroscopy of the brain in normal preterm and term infants, and early changes after perinatal hypoxiaischemia. Pediatr Res 1996;40(1):6–14. Petroff, O. A., Graham, G. D., Blamire, A. M., al-Rayess, M., Rothman, D. L., Fayad, P. B., Brass, L. M., Shulman, R. G., Prichard, J. W. Spectroscopic imaging of stroke in humans: Histopathology correlates of spectral changes. Neurology 1992;42(7): 1349–1354. Alger, J. R., Frank, J. A., Bizzi, A., Fulham, M. J., DeSouza, B. X., Duhaney, M. O., Inscoe, S. W., Black, J. L., van Zijl, P. C., Moonen, C. T. Metabolism of human gliomas: Assessment with H-1 MR spectroscopy and F-18 fluorodeoxyglucose PET. Radiology 1990;177(3):633–641. Mathews, P. M., Andermann, F., Silver, K., Karpati, G., Arnold, D. L. Proton MR spectroscopic characterization of differences in regional brain metabolic abnormalities in mitochondrial encephalomyopathies. Neurology 1993;43(12):2484–2490. Brand, A., Richter-Landsberg, C., Leibfritz, D. Multinuclear NMR studies on the energy metabolism of glial and neuronal cells. Dev Neurosci 1993;15(3–5):289–298. Flogel, U., Willker, W., Leibfritz, D. Regulation of intracellular pH in neuronal and glial tumour cells, studied by multinuclear NMR spectroscopy. NMR Biomed 1994;7(4): 157–166. Ross, B. D., Michaelis, T. Clinical applications of magnetic resonance spectroscopy. Magn Reson Q 1994;10:191–247. Shonk, T. K., Moats, R. A., Gifford, P., Michaelis, T., Mandigo, J. C., Izumi, J.,


33.

34. 35.

36.

37.

38.

39.

40.

41.

42.

43.

Ross, B. D. Probable Alzheimer disease: Diagnosis with proton MR spectroscopy. Radiology 1995;195(1):65–72. Kruse, B., Hanefeld, F., Christen, H. J., Bruhn, H., Michaelis, T., Hanicke, W., Frahm, J. Alterations of brain metabolites in metachromatic leukodystrophy as detected by localized proton magnetic resonance spectroscopy in vivo. J Neurol 1993;241(2): 68–74. Magistretti, P. J., Pellerin, L., Rothman, D. L., Shulman, R. G. Energy on demand. Science 1999;283(5401):496–497. Tkac, I., Andersen, P., Adriany, G., Merkle, H., Ugurbil, K., Gruetter, R. In vivo 1H NMR spectroscopy of the human brain at 7 T. Magn Reson Med 2001;46(3):451–456. Srinivasan, R., Sailasuta, N., Hurd, R., Nelson, S., Pelletier, D. Evidence of elevated glutamate in multiple sclerosis using magnetic resonance spectroscopy at 3 T. Brain 2005;128(Pt 5):1016–1025. Kreis, R., Pfenninger, J., Herschkowitz, N., Boesch, C. In vivo proton magnetic resonance spectroscopy in a case of Reye’s syndrome. Intensive Care Med 1995;21(3): 266–269. van Zijl, P. C. M., Barker, P. B. Magnetic Resonance Spectroscopy and Spectroscopic Imaging for the Study of Brain Metabolism. Imaging Brain Structure and Function, Vol. 820. New York, NY: Annals of the New York Academy of Sciences; 1997, pp 75–96. Rothman, D. L., Petroff, O. A., Behar, K. L., Mattson, R. H. Localized 1H NMR measurements of gamma-aminobutyric acid in human brain in vivo. Proc Natl Acad Sci USA 1993;90(12):5662–5666. Terpstra, M., Henry, P. G., Gruetter, R. Measurement of reduced glutathione (GSH) in human brain using LCModel analysis of difference-edited spectra. Magn Reson Med 2003;50(1):19–23. Pan, J. W., Telang, F. W., Lee, J. H., de Graaf, R. A., Rothman, D. L., Stein, D. T., Hetherington, H. P. Measurement of betahydroxybutyrate in acute hyperketonemia in human brain. J Neurochem 2001;79(3): 539–544. Seymour, K. J., Bluml, S., Sutherling, J., Sutherling, W., Ross, B. D. Identification of cerebral acetone by 1H-MRS in patients with epilepsy controlled by ketogenic diet. Magma 1999;8(1):33–42. Kreis, R., Pietz, J., Penzien, J., Herschkowitz, N., Boesch, C. Identification and quantitation of phenylalanine in the brain of patients with phenylketonuria by means of localized in vivo 1H magnetic-

44.

45.

46.

47.

48.

49.

50.

51. 52.

53.

54. 55.

223

resonance spectroscopy. J Magn Reson B 1995;107(3):242–251. van der Knaap, M. S., Wevers, R. A., Struys, E. A., Verhoeven, N. M., Pouwels, P. J., Engelke, U. F., Feikema, W., Valk, J., Jakobs, C. Leukoencephalopathy associated with a disturbance in the metabolism of polyols. Ann Neurol 1999;46(6):925–928. Cady, E. B., Lorek, A., Penrice, J., Reynolds, E. O., Iles, R. A., Burns, S. P., Coutts, G. A., Cowan, F. M. Detection of propan-1,2diol in neonatal brain by in vivo proton magnetic resonance spectroscopy. Magn Reson Med 1994;32(6):764–767. Hanstock, C. C., Rothman, D. L., Shulman, R. G., Novotny, E. J., Jr., Petroff, O. A., Prichard, J. W. Measurement of ethanol in the human brain using NMR spectroscopy. J Stud Alcohol 1990;51(2):104–107. Rose, S. E., Chalk, J. B., Galloway, G. J., Doddrell, D. M. Detection of dimethyl sulfone in the human brain by in vivo proton magnetic resonance spectroscopy. Magn Reson Imaging 2000;18(1):95–98. Vermathen, P., Capizzano, A. A., Maudsley, A. A. Administration and (1)H MRS detection of histidine in human brain: Application to in vivo pH measurement. Magn Reson Med 2000;43(5):665–675. Rothman, D. L., Behar, K. L., Prichard, J. W., Petroff, O. A. Homocarnosine and the measurement of neuronal pH in patients with epilepsy. Magn Reson Med 1997;38(6): 924–929. Mori, S., Eleff, S. M., Pilatus, U., Mori, N., van Zijl, P. C. Proton NMR spectroscopy of solvent-saturable resonances: A new approach to study pH effects in situ. Magn Reson Med 1998;40(1):36–42. Frahm, J. Localized proton spectroscopy using stimulated echoes. J Magn Reson 1987;72(3):502–508. van Zijl, P. C., Moonen, C. T., Alger, J. R., Cohen, J. S., Chesnick, S. A. High field localized proton spectroscopy in small volumes: Greatly improved localization and shimming using shielded strong gradients. Magn Reson Med 1989;10(2):256–265. Duyn, J. H., Gillen, J., Sobering, G., van Zijl, P. C., Moonen, C. T. Multisection proton MR spectroscopic imaging of the brain. Radiology 1993;188(1):277–282. Ordidge, R. J. Random noise selective excitation pulses. Magn Reson Med 1987;5(1): 93–98. Moonen, C. T. W., Sobering, G., van Zijl, P. C. M., Gillen, J., von Kienlin, M., Bizzi, A. Proton spectroscopic imaging of human brain. J Magn Reson 1992;98(3):556–575.

224

Zhu and Barker

56. Moonen, C. T., von Kienlin, M., van Zijl, P. C., Cohen, J., Gillen, J., Daly, P., Wolf, G. Comparison of single-shot localization methods (STEAM and PRESS) for in vivo proton NMR spectroscopy. NMR Biomed 1989;2(5–6):201–208. 57. Tkac, I., Starcuk, Z., Choi, I. Y., Gruetter, R. In vivo 1H NMR spectroscopy of rat brain at 1 ms echo time. Magn Reson Med 1999;41(4):649–656. 58. Collins, C. M., Liu, W., Schreiber, W., Yang, Q. X., Smith, M. B. Central brightening due to constructive interference with, without, and despite dielectric resonance. J Magn Reson Imaging 2005;21(2):192–196. 59. Slotboom, J., Mehlkopf, A. F., Bovee, W. M. M. J. A single-shot localization pulse sequence suited for coils with inhomogeneous RF fields using adiabatic slice-selective RF pulses. J Magn Reson 1991;95:396–404. 60. Garwood, M., DelaBarre, L. The return of the frequency sweep: Designing adiabatic pulses for contemporary NMR. J Magn Reson 2001;153(2):155–177. 61. Scheenen, T. W., Heerschap, A., Klomp, D. W. Towards 1H-MRSI of the human brain at 7T with slice-selective adiabatic refocusing pulses. Magma 2008;21(1–2):95–101. 62. Scheenen, T. W., Klomp, D. W., Wijnen, J. P., Heerschap, A. Short echo time 1HMRSI of the human brain at 3T with minimal chemical shift displacement errors using adiabatic refocusing pulses. Magn Reson Med 2008;59(1):1–6. 63. Hwang, T. -L., Shaka, A. J. Water suppression that works. Excitation sculpting using arbitrary waveforms and pulsed field gradients. J Magn Reson Ser A 1995;112: 275–279. 64. Levitt, M. H., Freeman, R. Compensation for pulse imperfections in NMR spin echo experiments. J Magn Reson 1981;43:65–80. 65. Mekle, R., Mlynarik, V., Gambarota, G., Hergt, M., Krueger, G., Gruetter, R. MR spectroscopy of the human brain with enhanced signal intensity at ultrashort echo times on a clinical platform at 3T and 7T. Magn Reson Med 2009;61(6):1279–1285. 66. Ordidge, R. J., Connelly, A., Lohman, J. A. B. Image-selected in vivo spectroscopy (ISIS) – a new technique for spatially selective NMR-spectroscopy. J Magn Reson 1986;66(2):283–294. 67. Ernst, R. R., Bodenhausen, G., Wokaun, A. Principles of Nuclear Magnetic Resonance in One and Two Dimensions. New York, NY: Oxford University Press; 1990, 640 p. 68. Golay, X., Gillen, J., van Zijl, P. C., Barker, P. B. Scan time reduction in proton

69.

70.

71.

72. 73.

74.

75.

76.

77.

78.

79.

80. 81.

magnetic resonance spectroscopic imaging of the human brain. Magn Reson Med 2002;47(2):384–387. Pohmann, R., von Kienlin, M., Haase, A. Theoretical evaluation and comparison of fast chemical shift imaging methods. J Magn Reson 1997;129(2):145–160. Duyn, J. H., Moonen, C. T. Fast proton spectroscopic imaging of human brain using multiple spin-echoes. Magn Reson Med 1993;30(4):409–414. Martin, A. J., Liu, H., Hall, W. A., Truwit, C. L. Preliminary assessment of turbo spectroscopic imaging for targeting in brain biopsy. AJNR Am J Neuroradiol 2001;22(5): 959–968. Mansfield, P. Spatial mapping of chemical shift in NMR. Magn Reson Med 1984;1: 370–386. Posse, S., Tedeschi, G., Risinger, R., Ogg, R., Le Bihan, D. High speed 1H spectroscopic imaging in human brain by echo planar spatial-spectral encoding. Magn Reson Med 1995;33(1):34–40. Posse, S., DeCarli, C., Le Bihan, D. Threedimensional echo-planar MR spectroscopic imaging at short echo times in the human brain. Radiology 1994;192(3):733–738. Ebel, A., Soher, B. J., Maudsley, A. A. Assessment of 3D proton MR echo-planar spectroscopic imaging using automated spectral analysis. Magn Reson Med 2001;46(6): 1072–1078. Ebel, A., Maudsley, A. A., Weiner, M. W., Schuff, N. Achieving sufficient spectral bandwidth for volumetric 1H echo-planar spectroscopic imaging at 4 Tesla. Magn Reson Med 2005;54(3):697–701. Pelletier, D., Nelson, S. J., Grenier, D., Lu, Y., Genain, C., Goodkin, D. E. 3-D echo planar (1)HMRS imaging in MS: Metabolite comparison from supratentorial vs. central brain. Magn Reson Imaging 2002;20(8):599–606. Govindaraju, V., Gauger, G. E., Manley, G. T., Ebel, A., Meeker, M., Maudsley, A. A. Volumetric proton spectroscopic imaging of mild traumatic brain injury. AJNR Am J Neuroradiol 2004;25(5):730–737. Adalsteinsson, E., Irarrazabal, P., Topp, S., Meyer, C., Macovski, A., Spielman, D. M. Volumetric spectroscopic imaging with spiral-based K-space trajectories. Magn Reson Med 1998;39(6):889–898. Block, K. T., Frahm, J. Spiral imaging: A critical appraisal. J Magn Reson Imaging 2005;21(6):657–668. Adalsteinsson, E., Star-Lack, J., Meyer, C. H., Spielman, D. M. Reduced spatial side


82.

83.

84.

85.

86.

87.

88.

89.

90.

91.

92.

93.

lobes in chemical-shift imaging. Magn Reson Med 1999;42(2):314–323. Sarkar, S., Heberlein, K., Hu, X. Truncation artifact reduction in spectroscopic imaging using a dual-density spiral K-space trajectory. Magn Reson Imaging 2002;20(10): 743–757. Kim, D. H., Adalsteinsson, E., Spielman, D. M. Spiral readout gradients for the reduction of motion artifacts in chemical shift imaging. Magn Reson Med 2004;51(3):458–463. Adalsteinsson, E., Langer-Gould, A., Homer, R. J., Rao, A., Sullivan, E. V., Lima, C. A., Pfefferbaum, A., Atlas, S. W. Gray matter N-acetyl aspartate deficits in secondary progressive but not relapsing-remitting multiple sclerosis. AJNR Am J Neuroradiol 2003;24(10):1941–1945. Pruessmann, K. P., Weiger, M., Scheidegger, M. B., Boesiger, P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med 1999;42:952–962. Sodickson, D. K., Manning, W. J. Simultaneous acquisition of spatial harmonics (SMASH): Fast imaging with radiofrequency coil arrays. Magn Reson Med 1997;38(4):591–603. Jakob, P. M., Griswold, M. A., Edelman, R. R., Sodickson, D. K. AUTO-SMASH: A selfcalibrating technique for SMASH imaging. Simultaneous acquisition of spatial harmonics. Magma 1998;7(1):42–54. McKenzie, C. A., Yeh, E. N., Ohliger, M. A., Price, M. D., Sodickson, D. K. Selfcalibrating parallel imaging with automatic coil sensitivity extraction. Magn Reson Med 2002;47(3):529–538. Griswold, M. A., Jakob, P. M., Heidemann, R. M., Nittka, M., Jellus, V., Wang, J., Kiefer, B., Haase, A. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn Reson Med 2002;47(6):1202–1210. Dydak, U., Weiger, M., Pruessmann, K. P., Meier, D., Boesiger, P. Sensitivity-encoded spectroscopic imaging. Magn Reson Med 2001;46(4):713–722. Tsai, S. Y., Otazo, R., Posse, S., Lin, Y. R., Chung, H. W., Wald, L. L., Wiggins, G. C., Lin, F. H. Accelerated proton echo planar spectroscopic imaging (PEPSI) using GRAPPA with a 32-channel phased-array coil. Magn Reson Med 2008;59(5):989–998. Dydak, U., Pruessmann, K. P., Weiger, M., Tsao, J., Meier, D., Boesiger, P. Parallel spectroscopic imaging with spin-echo trains. Magn Reson Med 2003;50(1):196–200. Lin, F. H., Tsai, S. Y., Otazo, R., Caprihan, A., Wald, L. L., Belliveau, J. W., Posse, S. Sensitivity-encoded (SENSE)

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

104.

105.

106.

225

proton echo-planar spectroscopic imaging (PEPSI) in the human brain. Magn Reson Med 2007;57(2):249–257. Smith, M. A., Gillen, J., McMahon, M. T., Barker, P. B., Golay, X. Simultaneous water and lipid suppression for in vivo brain spectroscopy in humans. Magn Reson Med 2005;54(3):691–696. Dreher, W., Leibfritz, D. Detection of homonuclear decoupled in vivo proton NMR spectra using constant time chemical shift encoding: CT-PRESS. Magn Reson Imaging 1999;17(1):141–150. Haase, A., Frahm, J., Hanicke, W., Matthei, D. 1 H NMR chemical shift selective imaging. Phys Med Biol 1985;30(4):341–344. Moonen, C. T. W., van Zijl, P. C. M. Highly efficient water suppression for in vivo proton NMR spectroscopy. J Magn Reson 1990;88:28–41. Ogg, R. J. WET a T1- and B1-insensitive water-suppression method for in vivo localized 1H NMR spectroscopy. J Magn Reson B 1994;104:1–10. Mescher, M., Merkle, H., Kirsch, J., Garwood, M., Gruetter, R. Simultaneous in vivo spectral editing and water suppression. NMR Biomed 1998;11(6):266–272. Star-Lack, J., Nelson, S. J., Kurhanewicz, J., Huang, L. R., Vigneron, D. B. Improved water and lipid suppression for 3D PRESS CSI using RF band selective inversion with gradient dephasing (BASING). Magn Reson Med 1997;38(2):311–321. Spielman, D. M., Pauly, J. M., Macovski, A., Glover, G. H., Enzmann, D. R. Lipidsuppressed single- and multisection proton spectroscopic imaging of the human brain. J Magn Reson Imaging 1992;2(3):253–262. Gu, M., Spielman, D. M. B1 and T1 insensitive water and lipid suppression using optimized multiple frequency-selective preparation pulses for whole-brain 1H spectroscopic imaging at 3T. Magn Reson Med 2009;61(2):462–466. Zhu, H., Ouwerkerk, R., Barker, P. B. Dual band water and lipid suppression for MR spectroscopic imaging at 3 Tesla. Magn Reson Med 2009;(accepted for publication). Kelley, D. A., Wald, L. L., Star-Lack, J. M. Lactate detection at 3T: Compensating J coupling effects with BASING. J Magn Reson Imaging 1999;9(5):732–737. Raphael, C. In vivo NMR spectral parameter estimation: A comparison between time and frequency domain methods. Magn Reson Med 1991;18:358–370. de Beer, R., van den Boogaart, A., van Ormondt, D., Pijnappel, W. W., den

226

Zhu and Barker

Hollander, J. A., Marien, A. J., Luyten, P. 108. Henriksen, O. In vivo quantitation of metabolite concentrations in the brain by R. Application of time-domain fitting in the means of proton MRS. NMR Biomed quantification of in vivo 1H spectroscopic 1995;8(4):139–148. imaging data sets. NMR Biomed 1992;5(4): 109. Soher, B. J., van Zijl, P. C., Duyn, 171–178. J. H., Barker, P. B. Quantitative pro107. Provencher, S. W. Estimation of metaboton MR spectroscopic imaging of the lite concentrations from localized in vivo human brain. Magn Reson Med 1996;35(3): proton NMR spectra. Magn Reson Med 356–363. 1993;30(6):672–679.

Chapter 10 Amide Proton Transfer Imaging of the Human Brain Jinyuan Zhou Abstract Amide proton transfer (APT) imaging is a new MRI technique that detects endogenous mobile proteins and peptides in tissue via saturation of the amide protons in the peptide bonds. Initial studies have shown promise in detecting tumor and stroke, but this technique was hampered by magnetic field inhomogeneity and a low signal-to-noise ratio. Several important prerequisites for performing APT imaging experiments include designing an effective APT imaging pulse sequence based on the hardware capability, optimizing the experimental protocol for the best clinical imaging quality, and developing data-processing approaches for effective image assessment. In this chapter, technical issues, such as pulse sequence design and optimization, magnetic field inhomogeneity correction, specific absorption rate minimization, and scan duration, are addressed. Key words: APT, magnetization transfer, protein, brain tumor, field inhomogeneity, MRI.

1. Introduction Proteins constitute 18% of the total mass of a typical mammalian cell. From an MRI point of view, these cellular proteins can be divided into two broad types: bound proteins, which possess solid-like properties and have protons with short T2 (∼ μs), and mobile proteins and peptides, which rotate rapidly and whose protons have relatively long T2 (∼ tens of ms). Solid-like macromolecules can be detected by conventional magnetization transfer (MT) (1, 2). It was recently demonstrated that it is possible to produce endogenous mobile protein- and peptide-based MRI contrast (3) using a chemical exchange saturation transfer (CEST) enhancement scheme (4). This approach, called amide proton transfer (APT) imaging (3), was shown to be sensitive to M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_10, © Springer Science+Business Media, LLC 2011

227

228

Zhou

pH changes in stroke (3, 5, 6) due to the effect of pH on proton exchange, and to be able to provide brain tumor contrast (7–9), based on the increased cellular content of proteins and peptides in malignant tumors (10, 11). APT imaging has the potential to expand the range of molecular MRI techniques to the endogenous protein and peptide level. It is a safe, non-invasive technology that can be easily implemented using existing hardware for clinical neuroimaging of brain tumors, stroke, and other neurologic disorders. The APT approach is in early development, and the technique is far from optimal. In this chapter, the initial experience and the prerequisites for performing an effective APT imaging experiment on the human brain at 3T are provided, using imaging of a human glioma as an example.

2. Materials 2.1. Hardware and Software

1. A Philips 3T MRI scanner (Philips Medical Systems, Best, The Netherlands). 2. An eight-channel phased-array head coil. 3. Two earplugs, a foam head holder, and a headband. 4. A personal computer loaded with the Philips pulseprogramming environment (PPE) and interactive data language (IDL, Research Systems, Inc., Boulder, CO, USA).

2.2. Phantoms, Human Subjects, and Other Materials

1. A bottle of liquid egg whites from the supermarket (see Note 1). 2. Several large bottles of water (3–5 L). 3. Healthy human subjects. 4. Patients with brain tumors. 5. Gadolinium contrast agents (ProHance, Bracco Diagnostic Inc., Princeton, NJ, USA).

3. Methods The effect of APT is associated with a low-concentration proton pool of mobile proteins and peptides. Therefore, the first step in conducting a successful APT imaging experiment is to design an effective imaging pulse sequence for the maximal APT enhancement. Because APT occurs in a small offset range around the water resonance frequency, a weak RF field should be applied to

Amide Proton Transfer Imaging of the Human Brain

229

avoid too much water signal intensity attenuation due to direct water saturation and conventional MT. After the pulse sequence is programmed, it must be tested and optimized for the best image and z-spectrum (see Note 2) quality. APT imaging is confounded by local magnetic field (B0 ) inhomogeneity (see Note 3). The initial human studies used a two-offset approach, symmetric around the global water center frequency, causing significant field-inhomogeneity-based image artifacts in many regions, especially in the frontal lobe and near the skull (8, 9). To solve this problem, a practical, six-offset, multi-acquisition method, combined with a single-acquisition zspectrum, may be used (see Fig. 10.1). As a compromise between optimum signal-to-noise ratio (SNR) and the ability for some correction, this method can acquire high-SNR APT images with B0 inhomogeneity correction within an acceptable scanning time (a few minutes).

Fig. 10.1. Schemes for APT-image acquisition. a Standard two-offset APT scan (+3.5 ppm for label, –3.5 ppm for reference). b Six-offset APT scan (±3, ±3.5, ±4 ppm). The effects of conventional MT and direct water saturation reduce the water signal intensities at all offsets (±3, ±3.5, ±4 ppm), and the existence of APT causes an extra reduction around the offset of 3.5 ppm (Reproduced from Ref. 9 with permission from Wiley-Liss, Inc.).

Finally, it is important to develop an APT data-processing approach for effective image assessment. This includes the following two steps: determination of the water-center-frequency shifts for each voxel and B0 inhomogeneity correction for z-spectra or APT images. The flow chart for APT-image data processing is shown in Fig. 10.2. 3.1. Design of an APT Imaging Pulse Sequence

1. These instructions assume the use of a Philips 3T MRI scanner, together with a body coil for RF transmission and an eight-channel phased-array coil for reception. Adjustments should be made according to the capability and limitations of the scanner hardware, particularly the duty cycle of radiofrequency (RF) power amplifiers (see Note 4). 2. Add a low-power long block pulse for proton saturation (up to 4 μT and 500 ms). The weak continuous-wave (CW) RF

230

Zhou

Fig. 10.2. Flow chart of APT-image data processing. The procedure is divided into two steps: (1) the generation of a water-frequency shift map and (2) the correction of APTimage data using the obtained shift map. Both steps are performed on a voxel-by-voxel basis.

saturation scheme has been widely used for various CEST imaging experiments. The RF saturation power and time are the most important pulse sequence parameters that must be optimized. 3. Use the turbo spin-echo (TSE) for imaging readout. A single slice is acquired (see Note 5). Single-shot acquisition should be used. Sensitivity encoding (SENSE) is used to reduce the TSE factor and specific absorption rate (SAR). 4. Add lipid suppression, such as selective partial inversion recovery (SPIR). 5. Adjust all pulse sequence parameters. This should also include the repetition time (TR), echo time (TE), field of view (FOV), image matrix, and slice thickness. The SAR should be kept below the U.S. Food and Drug Administration (FDA) limit for head (3.0 W/kg). 3.2. Test on Phantoms

1. Buy a bottle of egg whites from the supermarket. Shake the phantom for 5 min and put the phantom into the center of the magnet. To increase the loading, add several large bottles of water around the phantom. 2. Acquire localizer images and a SENSE reference scan. 3. Perform a z-spectrum experiment using an offset range of 8 to –8 ppm with an interval of 0.5 ppm (one signal average). One image without RF saturation is acquired for


231

normalization. To protect the scanner, conservative MRI parameters should be used for the first test: saturation time 100 ms, saturation power 1 μT, and TR 10 s. Higher order of (up to second order) volume shimming should be used. 4. Improve image quality and remove any image artifacts by modifying the imaging parameters, including the SENSE factor. 5. Repeat the z-spectrum experiments using the other pulse sequence parameters: saturation time 500 ms, saturation power 2–4 μT, and TR 3 s. 6. Examine the z-spectrum characteristics from a small region of interest (ROI). An ideal z-spectrum should be very smooth across all frequency offsets, with the lowest signal intensity at the water frequency, but the z-spectrum is generally shifted by the B0 inhomogeneity. 7. Perform the B0 inhomogeneity correction on a voxelby-voxel basis for each z-spectrum experiment (see Section 3.6). 8. Plot the z-spectrum and MTRasym -spectrum (see Note 2) for a small ROI, and compare the curves at the different power levels. The maximal APT effect at 3.5 ppm is about 5–8%. 3.3. Optimization on Healthy Normal Subjects

1. Complete the initial screening and the consent form. 2. Put the subject into the center of the magnet. The head of the subject should be restrained to avoid motion artifacts. This is accomplished using a very comfortable foam head holder and a headband. 3. Acquire localizer images and a SENSE reference scan. 4. Perform a z-spectrum experiment using an offset range of 8 to –8 ppm with an interval of 0.5 ppm (saturation time 500 ms, power 1 μT, FOV 212×212 mm2 , matrix 128×64, slice thickness 5 mm, TR 3 s, TE 11 ms, one signal average). One image without RF saturation is acquired for normalization. Higher order of (up to second order) volume shimming should be used. 5. Improve image quality and remove any image artifacts by modifying the imaging parameters, including the SENSE factor. 6. Repeat the z-spectrum experiments with higher saturation power levels at 2–4 μT. 7. Move the subject out of the magnet. 8. Examine the z-spectrum characteristics from a small ROI. An ideal z-spectrum should be very smooth across all

232

Zhou

frequency offsets with the lowest signal intensity at the water frequency, but the z-spectrum may be shifted by the B0 inhomogeneity. 9. Perform the B0 inhomogeneity correction on a voxelby-voxel basis for each z-spectrum experiment (see Section 3.6). 10. Plot the z-spectra and MTRasym -spectra (see Note 2) from gray matter, white matter, and cerebrospinal fluid (CSF). Compare the curves at the different power levels. Determine a characteristic RF saturation power at which MTRasym (3.5 ppm) is approximately zero for the whole brain (see Note 6). 3.4. Two- and Six-Offset Acquisition Scheme for APT Images

1. These instructions assume that the APT pulse sequence has been optimized with the z-spectrum experiments (see Section 3.3). The characteristic RF saturation power is 3 μT. The other imaging parameters are as follows: saturation time 500 ms, TR 3 s, FOV 212×212 mm2 , matrix 128×64, slice thickness 5 mm, TR 3 s, and TE 11 ms. 2. Acquire the saturation images at +3.5 ppm twice (eight signal averages). 3. Determine the signal intensity from a small ROI in one saturation image. Subtract the two saturation images and determine the noise from the same ROI in the difference image. Calculate the SNR. The SNR should be 100:1 or better to see the APT effect that is a few percent of the water intensity. 4. Perform an APT-image experiment using a standard twooffset acquisition scheme (+3.5 ppm for label and –3.5 ppm for reference) and eight signal averages. One image without RF saturation is acquired for normalization. 5. Calculate the APT image (see Note 2) and examine the effect of the B0 inhomogeneity on the APT images. 6. Acquire APT-image data using six frequency offsets (±3, ±3.5, ±4 ppm) and eight signal averages. One image without RF saturation is acquired for normalization. In this practical acquisition scheme, four extra offsets around ±3.5 ppm are acquired in the high SNR scan, and it is possible to correct for the artifacts on the APT image caused by B0 inhomogeneity. 7. Acquire a z-spectrum (33 offsets from 8 to –8 ppm with intervals of 0.5 ppm, one average) as an extra scan (see Note 7). One image without RF saturation is acquired for normalization.


233

8. Calculate the corrected APT image (see Section 3.6). The APT image is shown in color. For the healthy subject, a typical APT image is quite homogenous over the whole slice. 3.5. Scanning on Patients

1. The optimized z-spectrum and APT-image scans are added to the conventional MRI protocol. The total scan time for each subject should be limited to 1 h to maximize patient comfort. 2. Complete the initial screening and the consent form. 3. Put the subject into the center of the magnet. The head of the subject should be restrained to avoid motion artifacts. This is accomplished using standard configurations with a very comfortable foam head holder and a headband. 4. Acquire localizer images and a SENSE reference scan. 5. Acquire T2 -weighted images using the dual-echo TSE sequence. 6. Acquire fluid attenuated inversion recovery (FLAIR) images. 7. Identify the location of the lesion from the T2 -weighted and FLAIR images, and put the APT-image slice(s) on the lesion. 8. Acquire the six-offset APT-image scan. 9. Turn off the prescan to avoid changes in shim and frequency offset settings. 10. Acquire the z-spectrum scan. The same slice localization as that used for the APT-image scan should be used. 11. Acquire the T1 -weighted images using the magnetizationprepared rapid gradient-echo (MPRAGE) sequence. 12. Inject the gadolinium contrast agent into the patient (0.2 mL/kg, i.v.). 13. Acquire the gadolinium-enhanced T1 -weighted images using the MPRAGE sequence. 14. Move the subject out of the magnet. 15. Perform the B0 inhomogeneity correction on a voxel-byvoxel basis for each z-spectrum experiment (see Section 3.6). Plot the z-spectra and MTRasym -spectra (see Note 2) from the lesion and other ROIs. 16. Calculate the corrected APT image (see Section 3.6). The APT image is displayed in color. An example of the results produced is shown in Fig. 10.3.

3.6. Data Processing

1. These instructions assume the use of IDL. They are easily adaptable to other languages, such as MATLAB.

234

Zhou

Fig. 10.3. MR images of a patient with an anaplastic astrocytoma. Elevated APT signal can be seen in Gd-enhanced tumor core (red arrow), potentially providing unique information about the presence and grade of brain tumors, without the injection of exogenous contrast agents. The hyperintense APT area is comparable in size to the lesion identified on FLAIR but larger than that on the Gd-T1 w image (Reproduced from Ref. 9 with permission from Wiley-Liss, Inc.).

2. Fit the full z-spectrum through all 33 offsets using a 12thorder polynomial (the maximum order available with IDL) on a voxel-by-voxel basis (see Fig. 10.2). 3. Interpolate the fitted curve using an offset resolution of 1 Hz (2,049 points). 4. The actual water resonance should be at the frequency with the lowest signal intensity. The deviation of the water frequency in Hertz forms a map of water-center-frequency shifts. 5. To correct for the field inhomogeneity effects on zspectra, the measured z-spectrum for each voxel is interpolated to 2,049 points and shifted along the direction of the offset axis to correspond to 0 ppm at its lowest intensity. 6. The realigned z-spectra are interpolated back to 33 points for visual purposes. 7. Plot the z-spectrum and MTRasym -spectrum for a small ROI. The outermost points of ±7.5 and ±8 ppm are excluded in the display. 8. To correct for the field inhomogeneity effects on APT images (see Fig. 10.2), the acquired APT data for offsets +4, +3.5, and +3 ppm (or +512, +448, +384 Hz) for each voxel are interpolated to 257 points over the range from +4.5 to +2.5 ppm (or +576, +575, . . ., +320 Hz). 9. Realign the APT-image data using the fitted z-spectrum central frequency shift for the same voxel.


235

10. Perform the B0 inhomogeneity correction on a voxelby-voxel basis for the negative-offset data (–3, –3.5, –4 ppm). 11. Calculate the corrected APT image using the shiftcorrected data at the two offsets ±3.5 ppm. 12. The calculated APT image is thresholded based on the signal intensity of the S0 image to remove voxels outside the brain and displayed in color.

4. Notes 1. Many polymers, such as poly-L-lysine, dendrimers, and histone (12, 13), have a strong APT effect. However, these chemicals are very expensive; thus, they are not cost-effective for making large phantoms for use in human scanners. Egg whites in a bottle (protein 10%) from the supermarket are a straightforward protein phantom for APT imaging studies. In addition, some fruits, such as cantaloupe, are very useful phantoms with which to test imaging pulse sequences. Using a cantaloupe, up to about 25% of the CEST effect can be observed around a 1–2 ppm offset, which is due to various sugars in fruits. 2. In MT-type imaging (1, 2), water saturation is often measured as a function of transmitter frequency, producing the “z-spectra (14).” Such spectra are dominated by large direct water saturation around the water frequency at about 4.7 ppm and other saturation effects, such as conventional MT based on semi-solid tissue structures. The CEST effect is generally identified by asymmetry analysis with respect to this water signal (3), which is generally assigned to a reference frequency of 0 ppm. Quantitatively, the MT ratio (MTR) is defined as follows: MTR = 1 − Ssat /S0 , where Ssat and S0 are the signal intensities with and without RF irradiation, respectively. The MTR asymmetry (MTRasym ) parameter with respect to the water frequency is defined as follows: MTRasym = MTR( + offset) − MTR(−offset) = Ssat ( − offset)/S0 − Ssat ( + offset)/S0 .

236

Zhou

The APT image is quantified by the MTRasym parameter at ±3.5 ppm as follows: MTRasym (3.5 ppm) = Ssat ( − 3.5 ppm)/S0 − Ssat ( + 3.5ppm)/S0

= MTRasym (3.5ppm) + APTR .

3. Under higher order of slice shimming, as seen in the shift in the center of the z-spectrum, the B0 inhomogeneity is typically less than 20 Hz over most of the slice, but it could be as large as 60–80 Hz in the sinus and ear areas. A small B0 inhomogeneity can easily cause a few percentage point change in the MT asymmetry data, thus resulting in large artifacts on APT images. 4. The magnitude of the APT effect increases exponentially with the RF saturation time, and several seconds of saturation time are required to maximize the measurements. The use of weak CW saturation pulses is feasible on animal MRI scanners. It may also be feasible on human MRI systems if a transmit/receive (T/R) head coil is used. When body coil excitation with a phased-array coil receive is used, the RF saturation pulse is restricted. In the Philips 3T MRI scanner with the tube amplifier, as used in this study, the saturation pulse duration is limited to 500 ms. 5. The clinical application of APT imaging to data remains limited to single-slice. Multi-slice or whole-brain imaging would be feasible, but many technical issues, such as long scan times, APT contrast loss between slices, magnetic field inhomogeneity, and SAR must be resolved properly. 6. As the applied RF saturation power increases, MTRasym (3.5 ppm) for gray matter and white matter increases initially and then decreases back to zero, while that for CSF remains at around zero. At the characteristic RF saturation power, MTRasym (3.5 ppm) is approximately zero for both normal brain tissue and CSF. In contrast, MTRasym (3.5 ppm) would be hyperintensive in tumors and hypointensive in stroke regions. 7. If the z-spectrum measurement has the same saturation parameters as the APT-image scan, the z-spectrum data can be used for both identification of the APT effect and fitting of the B0 inhomogeneity map. If the z-spectrum is used to fit only the B0 inhomogeneity map, then lower saturation power and shorter saturation time may be used. Such a scan, called water saturation shift referencing (WASSR) (15), would provide a narrower z-spectrum; thus, the B0 inhomogeneity map can be fitted more accurately.


237

Acknowledgments The author thanks the team at the F.M. Kirby Research Center for Functional Brain Imaging for helpful discussions and technical assistance. This work was supported in part by grants from NIH (EB002634, EB005252, EB009112, EB009731, and RR015241) and the Dana Foundation. References 1. Wolff, S. D., Balaban, R. S. Magnetization transfer contrast (MTC) and tissue water proton relaxation in vivo. Magn Reson Med 1989;10:135–144. 2. Henkelman, R. M., Stanisz, G. J., Graham, S. J. Magnetization transfer in MRI: A review. NMR Biomed 2001;14:57–64. 3. Zhou, J., Payen, J., Wilson, D. A., Traystman, R. J., van Zijl, P. C. M. Using the amide proton signals of intracellular proteins and peptides to detect pH effects in MRI. Nat Med 2003;9:1085–1090. 4. Ward, K. M., Aletras, A. H., Balaban, R. S. A new class of contrast agents for MRI based on proton chemical exchange dependent saturation transfer (CEST). J Magn Reson 2000;143:79–87. 5. Sun, P. Z., Zhou, J., Sun, W., Huang, J., van Zijl, P. C. M. Delineating the boundary between the ischemic penumbra and regions of oligaemia using pHweighted magnetic resonance imaging (pHWI). J Cereb Blood Flow Metab 2007;27: 1129–1136. 6. Jokivarsi, K. T., Grohn, H. I., Grohn, O. H., Kauppinen, R. A. Proton transfer ratio, lactate, and intracellular pH in acute cerebral ischemia. Magn Reson Med 2007;57:647–653. 7. Zhou, J., Lal, B., Wilson, D. A., Laterra, J., van Zijl, P. C. M. Amide proton transfer (APT) contrast for imaging of brain tumors. Magn Reson Med 2003;50: 1120–1126. 8. Jones, C. K., Schlosser, M. J., van Zijl, P. C., Pomper, M. G., Golay, X., Zhou, J. Amide proton transfer imaging of human brain tumors at 3T. Magn Reson Med 2006;56:585–592.

9. Zhou, J., Blakeley, J. O., Hua, J. et al. Practical data acquisition method for human brain tumor amide proton transfer (APT) imaging. Magn Reson Med 2008;60: 842–849. 10. Hobbs, S. K., Shi, G., Homer, R., Harsh, G., Altlas, S. W., Bednarski, M. D. Magnetic resonance imaging-guided proteomics of human glioblastoma multiforme. J Magn Reson Imaging 2003;18:530–536. 11. Howe, F. A., Barton, S. J., Cudlip, S. A. et al. Metabolic profiles of human brain tumors using quantitative in vivo 1 H magnetic resonance spectroscopy. Magn Reson Med 2003;49:223–232. 12. Goffeney, N., Bulte, J. W. M., Duyn, J., Bryant, L. H., van Zijl, P. C. M. Sensitive NMR detection of cationic-polymer-based gene delivery systems using saturation transfer via proton exchange. J Am Chem Soc 2001;123:8628–8629. 13. McMahon, M. T., Gilad, A. A., Zhou, J., Sun, P. Z., Bulte, J. W. M., van Zijl, P. C. M. Quantifying exchange rates in chemical exchange saturation transfer agents using the saturation time and saturation power dependencies of the magnetization transfer effect on the magnetic resonance imaging signal (QUEST and QUESP): pH calibration for poly-L-lysine and a starburst dendrimer. Magn Reson Med 2006;55:836–847. 14. Bryant, R. G. The dynamics of water-protein interactions. Annu Rev Biophys Biomol Struct 1996;25:29–53. 15. Kim, M., Gillen, J., Landman, B. A., Zhou, J., van Zijl, P. C. M. Water saturation shift referencing (WASSR) for chemical exchange saturation transfer (CEST) experiments. Magn Reson Med 2009;61:1441–1450.

wwwwwww

Chapter 11 High-Field MRI of Brain Iron Jozef H. Duyn Abstract Recent developments in high-field MRI have provided opportunities to detect iron in human brain with much improved sensitivity. The combination of increased magnetic field strength with multi-channel detectors has made it possible to routinely obtain images at about 300 μm resolution. These images can be sensitized to tissue iron by exploiting the improved magnetic susceptibility contrast at high field. Together, these techniques have the potential to map the fine scale distribution of iron in human brain at the level of fiber bundles and cortical laminae, and may aid in the understanding of the role and transport of iron in normal brain and in disease. In this chapter, we will look at these techniques in detail and present some examples of high-field MRI data of human brain. Key words: Iron, ferritin, MRI, brain, magnetic susceptibility.

1. Introduction Cellular iron has important roles in brain development and function, and abnormal concentrations may lead to pathological conditions (1–3). Since the beginning of MRI, attempts have been made to map in vivo brain iron distributions under normal and pathological conditions (4–6). MRI is a versatile technique that is able to generate a variety of contrasts, a number of which reflect tissue iron content. For example, iron can affect the MRI signal through several of its major contrast parameters including T1 , T2 , and T2 ∗ (6). All of these have been used to study brain iron, and each has its advantages and disadvantages. The recent proliferation of high-field scanners of 7 T and above has reinvigorated the study of brain iron with MRI. Major factors in this have been the increased sensitivity (signal-to-noise M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_11, © Springer Science+Business Media, LLC 2011

239

240

Duyn

ratio or SNR) available with the new scanners, in particular when employing magnetic susceptibility contrast. This increased sensitivity is due to the paramagnetic properties of iron in most of its forms that are present in brain. Magnetic susceptibility contrast is reflected in T2 ∗ and resonance frequency, both of which may convey information about tissue iron content (6–8). In the following, we will discuss some of the methodological aspects of high-field MRI sensitized to brain iron, including the equipment and acquisition techniques involved, as well as the analysis and interpretation of the data.

2. Materials 2.1. MRI Hardware 2.1.1. Magnets

Advances in magnet technology have led to the availability of stronger magnets and higher sensitivity for human MRI. While early clinical MRI systems employed magnets of 1 T field strength or below, today’s scanners are at 1.5 T and 3.0 T, and at even 7.0 T. Field strengths continue to increase, and currently experimental 9.4 T and 11.7 T systems are in use or being developed. The increased sensitivity of these systems can be exploited to improve spatial resolution. In addition, magnetic susceptibility contrast is increased, providing particular advantages for the detection of iron. These increases are dependent on the echo time (see Methods) and for typical conditions are around threefold for both T2 ∗ and resonance frequency contrast (Fig. 11.1).

2.1.2. RF Detectors

An additional increase in sensitivity has come from the development of multi-channel detectors that has played out over the last decade. Multi-channel detectors improve signal detection by allowing the individual detector elements to be placed closer to the object under study and, at the same time, limit the noise received from the sample. Together, these advantages result in sensitivity improvements that can average around two- to threefold over the brain for 32 channel detectors that are currently widely available. A second advantage of multi-channel detectors is the fact that they enable image acceleration through parallel imaging techniques (9, 10). This is a particularly useful feature to achieve high resolution, which requires more data to be acquired and therefore longer scan times.

2.1.3. Respiratory Compensation

The SNR and CNR increases at high field are accompanied by an increased sensitivity to physiological variations and motion that can compromise image quality. In order to fully exploit the

High-Field MRI of Brain Iron

241

Fig. 11.1. Simulated contrast-to-noise ratio (CNR) for susceptibility weighted MRI at 3 T and 7 T. The CNR gain at high field is dependent of echo time and ranges from 2 to 4 for echo time in the commonly used range of 10–60 ms. The simulation ignored T1 effects and assumed identical acquisition bandwidth, a linear sensitivity increase with field strength, and tissue R2 ∗ values of 20–1 and 30 s–1 at 3 T and 7 T, respectively.

potential of high-field MRI, these unwanted confounds need to be dealt with. For example, the respiratory cycle can induce subtle magnetic field fluctuations in the brain that lead to ghosting artifacts in susceptibility weighted MRI (11). This is particularly noticeable at high field as the amplitude of these fluctuations increases linearly with field strength. Currently, prototype hardware exists to compensate for these field fluctuations and improve image quality (Fig. 11.2). Major components are a pressure belt to register chest motion, a computer to calculate field (shim)

Fig. 11.2. Compensation of respiration-induced magnetic field fluctuations. The patient’s chest position, registered with a stretch sensor placed around the chest, is used to estimate adjustments to RF frequency, magnetic field gradients, and magnet field shims. This is done with a personal computer, which sends the adjustment values to the MRI system electronics.

242

Duyn

corrections, and shim coils driven with fast current sources to allow rapid field adjustments. 2.1.4. Head Motion Correction

Sudden or even slow head motion during MRI scanning can compromise image quality, in particular when scanning at high resolution. A number of techniques have been proposed to compensate for this motion (12–15). Recent implementations employ video cameras to track head motion and feed this information back to the scanner to make adjustments to the image acquisition process (14, 15). An alternative approach measures the local, position-dependent magnetic field induced by the MRI gradient system through the use of small coils placed around the head (16). Although these techniques are quite effective, they have not been yet fully developed for widespread use.

3. Methods 3.1. Acquisition Methods

Magnetic susceptibility inclusions in brain tissues, for example local areas of increased iron content, lead to magnetic field shifts that are generally inhomogeneous over the scale of an image voxel. This results in incoherent phase accumulation and therefore to T2 ∗ reduction and signal loss in gradient echo imaging (GRE). In addition, these inclusions may lead to a net frequency shift of the voxel-averaged signal, which manifests itself as a voxel phase shift. This effect has been recently exploited at 7 T to improve visualization of subtle anatomical variations in grey and white matter of human brain. (8)

3.1.1. Resolution

The choice of image resolution is important as it directly affects scan times and image quality. Too low a resolution may lead to partial volume effects and affect the conspicuity of small anatomical variations. Too high a resolution may increase the sensitivity to patient motion and lead to low signal-to-noise ratio (SNR). This SNR can only partially be recovered with spatial averaging during image reconstruction. Using 32 channel detectors at 7 T, image resolutions of 0.2 × 0.2 × 1 mm or 0.3 × 0.3 × 0.3 mm are feasible within scan times of about 10 min.

3.1.2. Echo Time and Bandwidth

In choosing optimal echo time (TE) and bandwidth, one needs to take into account the contrast-to-noise ratio (CNR). While increasing echo time (TE) leads to increased percentage signal loss and absolute phase shift, it also reduces image SNR. It turns out that one can optimize the CNR of both signal amplitude and phase by choosing TE equal to the average T2 ∗ values of the brain structures involved (8). At 7 T, this means TE needs to be chosen


243

in the range of 10–30 ms, which covers much of the range of T2 ∗ values found in (normal) human brain. Furthermore, CNR optimization requires minimization of the acquisition bandwidth, i.e., capture as much of the signal decay curve by maximizing the duration of the data acquisition window (TACQ) (Fig. 11.3). One caveat with this is that increasing TACQ increases image blurring and off-resonance related pixel shifts, some of which can be corrected in post-processing. Typical TACQ values at 7 T range from 10 to 30 ms.

Fig. 11.3. Optimization of sensitivity (SNR). Maximum SNR is obtained when the acquisition duration (TACQ) is maximized. As TACQ is generally centered around the gradient echo time (TE), TACQ < 2∗ TE.

3.1.3. Multi-slice 2D Versus 3D Techniques

GRE MRI can be performed either in a multi-slice or true 3D fashion. The latter may have a significant SNR advantage if a large area of interest (or the entire brain) needs to be imaged. 3D techniques excite the entire slab of interest with each RF pulse rather than sub-sections in a sequential fashion. The SNR advantage comes about when the time to run through all the sub-sections in a multi-slice scan exceeds the longitudinal relaxation time of the tissue. Under this condition, multi-slice techniques become rather inefficient (in term of SNR per square root of total scan time) compared to true 3D techniques. A caveat with 3D techniques is that the generated MRI signal may have a larger dynamic range and therefore put increased demands on the MRI acquisition hardware. Further, 3D techniques may be less robust in the presence of head motion.

3.1.4. Image Acceleration with Parallel Imaging

Higher resolution MRI requires the acquisition of large data matrices, leading to long scan times. A typical 20-slice highresolution 2D acquisition with image matrix of 1,024 × 768 requires 15,360 repeated RF excitations. With T2 ∗ of up to 30 ms, most of the signal decay curve can be sampled in about 50 ms. Assuming each repeated excitation to last 50 ms, the

244

Duyn

scan time for the entire acquisition would be close to 13 min. Increasing volume coverage beyond the 20 slice example given above may lead to prohibitively long scan times, necessitating the use of image acceleration approaches such as parallel imaging. Methods, such as SENSE (10) and SMASH (17), allow some of the acquisition matrix element to be estimated from the spatial information contained in the sensitivity profiles of the detector elements. The saving in acquisition time resulting from this is ultimately limited by the number of detector elements. With 32channel detectors and acceleration in one dimension, acceleration rates of 3–4 (i.e., scan time reduction of 66–75%) are feasible without significant degradation of image quality. 3.1.5. Multi-echo Techniques

The image intensity in GRE MRI data is dependent not only on tissue T2 ∗ values but also on other MR parameters, such as spin density and T1 , and sequence parameters, such as TR and flip angle. This complicates extraction of quantitative and reproducible measures. To overcome this, one can acquire multiple sequential echoes, each of which can be reconstructed into a separate image. The varying T2 ∗ weighting of each image can be used to extract quantitative T2 ∗ values. The generation of multiple echoes can be effectuated by repeated reversal of the read gradient (Fig. 11.4).

Fig. 11.4. Multi-echo acquisition. Multiple reversals of the read gradient are used to generate a number of echo signals with increasing T2 ∗ -weighting.

The added benefit of the multi-echo approach is that each echo is acquired in a shorter time (higher bandwidth) and therefore is less affected by off-resonance artifacts and T2 ∗ blurring (18). Furthermore, there is no significant SNR penalty in doing this, as the multiple echo data can be recombined into a single image with SNR similar to that of the low bandwidth T2 ∗ weighted image of the conventional single echo approach.


245

Drawbacks are the increased stress on the gradient system, the increased acoustic noise, and the increased sensitivity to motion. The latter originates from the fact that it becomes more difficult (and less efficient) to employ motion compensation strategies that rely on gradient moment nulling. (19) 3.2. Reconstruction Methods

A number of processing steps are required to convert the raw data into interpretable images that can provide a measure of iron content. These include combining of the coil signals with or without parallel imaging reconstruction and calculation of frequency (or phase) maps. Additional steps can include the calculation of magnetic susceptibility and T2 ∗ maps. We will discuss each of these briefly.

3.2.1. Coil Combining

Combination of coil signals can be performed with the standard SENSE reconstruction, as described previously (20). This is appropriate for both standard non-accelerated acquisitions (R=1) and accelerated acquisitions (R>1). It may be beneficial to use subject-specific coil sensitivity data, which can be used to generate the required coil sensitivity reference maps (20). For this purpose, a fast, low resolution scan can be performed using the same slice locations as the high-resolution data. Preferably, a scan with minimal T2 ∗ contrast is used. This can be achieved by using short TE.

3.2.2. Calculation of Frequency Maps

The complex image data generated with the SENSE reconstruction can be converted into both magnitude (i.e., signal amplitude) and phase maps, both of which are sensitive to tissue iron content. The phase maps are then further processed to remove unwanted spatial variations associated with large-scale bulk susceptibility effects at, e.g., air-tissue interfaces. This can be effectively achieved with spatial high-pass filtering through homodyne methods or polynomial fitting (8, 21, 22). These methods also allow convenient removal of any phase jumps (at boundaries of the [–π, π] phase range) that may be present in the raw data. The remaining signal phase can be attributed to off-resonance effects that reflect the underlying tissue properties, including the local iron content. The amplitude of this effect (in Hertz) can be calculated by dividing the local phase shift (in cycles) by the echo time (in seconds).

3.2.3. Susceptibility Maps

Although phase/frequency images have been used to directly estimate local iron content (7), one caveat is that the two are only indirectly related. One important confound is that local resonance frequency is dependent in a complicated manner on geometry and orientation of both local and surrounding distribution of iron inclusions (6, 23, 24). A number of research groups are

246

Duyn

addressing this problem and are attempting to reconstruct susceptibility maps from 3D phase distributions (25–28). The former would not be dependent on orientation and geometry and more directly represent the local tissue composition. Preliminary experience in brain suggests that susceptibility calculation is indeed possible (28); however, there are still unresolved issues that impact the quality of the susceptibility maps. These issues include the presence of streaking artifacts due to focal areas of ill-defined phase (e.g,. in vessels or near air-tissue interfaces) and noise amplification for structures that are at the magic angle relative to the main magnetic field (28). It is anticipated that these issues will be resolved, at least partly, in coming years. The data available with multi-echo techniques allow calculation of quantitative T2 ∗ values (or R2 ∗ values; R2 ∗ = 1/T2 ∗ ), which may supplement susceptibility information for the study of brain iron content. T2 ∗ values can be derived for multi-echo data by simple T2 ∗ fitting of the signal decay with increasing echo time. When both positive and negative echoes in a GRE echo train are used, correction of off-resonance related distortions may be required prior to fitting. This can be done based on Bo maps that can be derived from the phase data. Sample R2 ∗ and T2 ∗ maps are shown in Fig. 11.5.

Fig. 11.5. Example of various contrasts that can be derived from a multi-echo data acquisition. Shown are magnitude signal of the first echo (primarily proton-density weighted) (a), frequency (b), T2 ∗ (c), and R2 ∗ (d).


3.3. Data Interpretation

247

Although it has been over two decades since the first MRI study of brain iron distribution, the development of a quantitative method to estimate brain iron content from MRI-derived measures is still a work in progress. The primary reason for this is that MRI contrast mechanisms are generally complex, and this is certainly the case for the phase shifts and T2 ∗ values derived from GRE data at high field. For example, the T2 ∗ relaxation caused by intra-voxel phase dispersion may originate from a number of sources in addition to iron, including exchange effects with amide protons (29), and inhomogeneous fields generated by susceptibility inclusions, such as proteins, myelin, and deoxyhemoglobin (8, 30, 31). These effects may have a geometry and an orientation dependence. This is also the case for MRI frequency maps, which are affected by many of the same contributors. Nevertheless, in regions where iron dominates the contrast, T2 ∗ and susceptibility values may have a relatively well-defined dependence on iron content. In these regions, these measures may provide reasonable estimates of tissue iron content (7, 22). However, it is expected that the relative contribution of the sources contributing to contrast in susceptibility weighted MRI will vary across brain regions. For example, in white matter, some of the paramagnetic susceptibility of iron may be cancelled out by diamagnetic contributions of myelin. This could explain the absence of a paramagnetic phase shift in WM (relative to cerebrospinal fluid) (8) and a remaining diamagnetic shift after iron extraction (32). Also, because of the generally highly ordered microscopic structure of WM, the orientation of this structure may affect the MRI susceptibility measures (30, 33). Accurate analysis of brain iron content with MRI will likely require a comprehensive understanding of the mechanisms and relative importance of contributing compounds to the various MRI contrast parameters. It is likely that a combined analysis of the various MRI contrasts will contribute to this understanding. For example, a combined analysis of T2 ∗ and phase data may be helpful in quantification of myelin and iron content, as these compounds differentially contribute to the two contrasts.

Acknowledgments My colleagues in the laboratory of Advanced MRI are acknowledged for their contributions to this work. This research was supported by the Intramural Research Program of NIH, NINDS.

248

Duyn

References 1. Morris, C. M. et al. Brain iron homeostasis (Translated from Eng). J Inorg Biochem 1992;47(3–4):257–265 (in Eng). 2. Burdo, J. R., Connor, J. R. Brain iron uptake and homeostatic mechanisms: An overview (Translated from Eng). Biometals 2003;16(1):63–75 (in Eng). 3. Zecca, L., Youdim, M. B., Riederer, P., Connor, J. R., Crichton, R. R. Iron, brain ageing and neurodegenerative disorders (Translated from Eng). Nat Rev Neurosci 2004;5(11):863–873 (in Eng). 4. Drayer, B. et al. MRI of brain iron. Am J Roentgenol 1986;147(1):103–110. 5. Schenck, J. F. Magnetic resonance imaging of brain iron. J Neurol Sci 2003;207(1–2): 99–102. 6. Haacke, E. M. et al. Imaging iron stores in the brain using magnetic resonance imaging. Magn Reson Imaging 2005;23(1):1–25. 7. Ogg, R. J., Langston, J. W., Haacke, E. M., Steen, R. G., Taylor, J. S. The correlation between phase shifts in gradient-echo MR images and regional brain iron concentration. Magn Reson Imaging 1999;17(8): 1141–1148. 8. Duyn, J. H. et al. High-field MRI of brain cortical substructure based on signal phase. Proc Natl Acad Sci USA 2007;104(28):11796–11801. 9. Sodickson, D. K., Manning, W. J. Simultaneous acquisition of spatial harmonics (SMASH): Fast imaging with radiofrequency coil arrays. Magn Reson Med 1997;38(4):591–603. 10. Pruessmann, K. P., Weiger, M., Scheidegger, M. B., Boesiger, P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med 1999;42(5):952–962. 11. van Gelderen, P., de Zwart, J. A., Starewicz, P., Hinks, R. S., Duyn, J. H. Real-time shimming to compensate for respirationinduced B0 fluctuations. Magn Reson Med 2007;57(2):362–368. 12. Derbyshire, J. A., Wright, G. A., Henkelman, R. M., Hinks, R. S. Dynamic scanplane tracking using MR position monitoring (Translated from Eng). J Magn Reson Imaging 1998;8(4):924–932 (in Eng). 13. Tremblay, M., Tam, F., Graham, S. J. Retrospective coregistration of functional magnetic resonance imaging data using external monitoring. Magn Reson Med 2005;53(1): 141–149. 14. Zaitsev, M., Dold, C., Sakas, G., Hennig, J., Speck, O. Magnetic resonance imaging of freely moving objects: Prospective

15.

16.

17. 18.

19.

20.

21.

22.

23.

24.

25.

real-time motion correction using an external optical motion tracking system (Translated from Eng). Neuroimage 2006;31(3): 1038–1050 (in Eng). Qin, L. et al. Prospective head-movement correction for high-resolution MRI using an in-bore optical tracking system (Translated from Eng). Magn Reson Med 2009;62: 924–934 (in Eng). Ooi, M. B., Krueger, S., Thomas, W. J., Swaminathan, S. V., Brown, T. R. Prospective real-time correction for arbitrary head motion using active markers (Translated from Eng). Magn Reson Med 2009;62(4): 943–954 (in Eng). Sodickson, D. K., Griswold, M. A., Jakob, P. M. SMASH imaging. Magn Reson Imaging Clin N Am 1999;7(2):237–254, vii–viii. Fischl, B. et al. Sequence-independent segmentation of magnetic resonance images (Translated from Eng). Neuroimage 2004;23(Suppl 1):S69–S84 (in Eng). Haacke, E. M., Lenz, G. W. Improving MR image quality in the presence of motion by using rephasing gradients (Translated from Eng). AJR Am J Roentgenol 1987;148(6):1251–1258 (in Eng). de Zwart, J. A., Ledden, P. J., Kellman, P., van Gelderen, P., Duyn, J. H. Design of a SENSE-optimized high-sensitivity MRI receive coil for brain imaging. Magn Reson Med 2002;47(6):1218–1227. Abduljalil, A. M., Schmalbrock, P., Novak, V., Chakeres, D. W. Enhanced gray and white matter contrast of phase susceptibilityweighted images in ultra-high-field magnetic resonance imaging. J Magn Reson Imaging 2003;18(3):284–290. Yao, B. et al. Susceptibility contrast in high field MRI of human brain as a function of tissue iron content (Translated from Eng). Neuroimage 2009;44(4):1259–1266 (in Eng). Salomir, R., de Senneville, B. D., Moonen, C. T. W. A fast calculation method for magnetic field inhomogeneity Due To an arbitrary distribution of bulk susceptibility. Concepts Magn Reson B 2003;19B(1):26–34. Marques, J. P., Bowtell, R. W. Using forward calculations of the magnetic field perturbation Due To a realistic vascular model to explore the BOLD effect. NMR Biomed 2008;21(6):553–565. de Rochefort, L., Brown, R., Prince, M. R., Wang, Y. Quantitative MR susceptibility mapping using piece-wise constant regularized inversion of the magnetic field


26.

27.

28.

29.

(Translated from Eng). Magn Reson Med 2008;60(4):1003–1009 (in Eng). Kressler, B. et al. Nonlinear regularization for per voxel estimation of magnetic susceptibility distributions from MRI field maps (Translated from Eng). IEEE Trans Med Imaging 2010;29(2):273–281 (in Eng). Liu, T., Spincemaille, P., de Rochefort, L., Kressler, B., Wang, Y. Calculation of susceptibility through multiple orientation sampling (COSMOS): A method for conditioning the inverse problem from measured magnetic field map to susceptibility source image in MRI (Translated from Eng). Magn Reson Med 2009;61(1):196–204 (in Eng). Shmueli, K., Li, J., Duyn, J. H. Magnetic susceptibility mapping of brain tissue in-vivo using MRI phase data (Translated from Eng). Magn Reson Med 2009;62(6):1510–1522 (in Eng). Zhong, K., Leupold, J., von Elverfeldt, D., Speck, O. The molecular basis for gray and

30.

31.

32.

33.

249

white matter contrast in phase imaging. Neuroimage 2008;40(4):1561–1566. He, X., Yablonskiy, D. A. Biophysical mechanisms of phase contrast in gradient echo MRI (Translated from Eng). Proc Natl Acad Sci USA 2009;106(32):13558–13563 (in Eng). Lee, D., Hirano, Y., Fukunaga, M., Silva, A. C., Duyn, J. H. On the contribution of deoxy-hemoglobin to MRI gray-white matter contrast at high field. Neuroimage 2010;49(1):193–198. Fukunaga, M. et al. Layer-specific variation of iron content in cerebral cortex as a source of MRI contrast (Translated from Eng). Proc Natl Acad Sci USA 2009;107(8):3834–3839 (in Eng). Lee, J. et al. Sensitivity of MRI resonance frequency to the orientation of brain tissue microstructure (Translated from Eng). Proc Natl Acad Sci USA 2010;107(11):5130– 5135 (in Eng).

wwwwwww

Chapter 12 Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications Manisha Aggarwal, Jiangyang Zhang, and Susumu Mori Abstract In this chapter, we introduce modern magnetic resonance imaging (MRI)-based mouse brain atlases. Although unable to match the resolution and specificity of their histology-based counterparts, MRI-based mouse brain atlases feature higher anatomical fidelity and can facilitate high-throughput computerassisted analysis of certain brain phenotypes. This chapter discusses several technical aspects of MRI-based mouse brain atlases, which are important to understand the usefulness as well as limitations of existing atlases. We focus on a novel MRI technique, diffusion tensor imaging (DTI), which provides rich tissue contrasts and is uniquely suited for studying white matter structures and immature mouse brains. The chapter then demonstrates several applications of MRI-based mouse brain atlases in anatomical phenotyping and guiding stereotaxic operations. Key words: Mouse brain atlas, magnetic resonance imaging, brain morphology, diffusion tensor imaging.

1. Introduction The laboratory mouse is widely used in neuroscience research, because it shares many genes, physiological processes, and disease loci with humans and is relatively easy to handle. With modern gene technology and the availability of the mouse genome database, it is relatively easy to generate genetically modified mouse strains in order to investigate the mechanisms of genetic controls of the brain. Numerous mouse models of human diseases have been established in the last few decades, and they have played essential roles in advancing our knowledge on the mechanisms M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_12, © Springer Science+Business Media, LLC 2011

251

252

Aggarwal, Zhang, and Mori

of diseases in the brain and developing new therapeutic treatments. Atlases of the mouse brain are important tools in examining changes in the mouse brain due to genetic mutations or pathological conditions. A brain atlas presents neuroanatomical information in the forms of anatomical images and structural delineations, from which the morphological properties of brain structures and their spatial relationships can be appreciated. The information can be used to guide surgical operations and target-specific delivery of drugs or cells. A brain atlas can serve as the central platform for data analysis and reporting. For instance, data on cellular and molecular events in various parts of the brain can be conveniently incorporated in a brain atlas to study their distributions and structural differences using advanced bioinformatics tools (1, 2). There are several well-established mouse brain atlases based on histology (3–11), which provides rich cellular and molecular information at high resolution. The staining methods used in these atlases reveal the biochemical composition and microstructural information of brain structures. The contrasts presented in the stained histological sections are used to delineate detailed structures in the mouse brain, e.g., the thalamic nuclei, and have been widely regarded as the gold standard for distinguishing brain structures. One limitation of histology based brain atlases is that they have limited anatomical fidelity. Histology is performed on ex vivo brain specimens. The sectioning and embedding processes can cause significant tissue injury and deformation, which can be more severe for embryonic or neonatal brain samples, since the immature brain tissue is relatively soft and easily deformed compared to the adult brain. Furthermore, the sectioning process can disrupt the spatial relationships between structures, which are difficult to recover even with complex reconstruction algorithms. Lack of anatomical fidelity limits the application of histology based atlases. For example, the precision of stereotaxic operations strongly depends on whether the atlas in use to guide such operations is an accurate representation of the target brain anatomy. This point will be discussed in detail in a later section. One solution to ensure high anatomical fidelity is to construct mouse brain atlases based on images acquired using three dimensional (3D) non-destructive imaging techniques, which do not require sectioning to acquire a stack of cross-sectional images, and therefore preserve high anatomical fidelity. 3D imaging techniques also have several unique advantages. Images can often be acquired from live animals, and therefore the anatomical information presented in such images is the least perturbed by the imaging procedure. The results of 3D imaging techniques can be viewed in any required oblique orientation and can be processed and analyzed directly without complex 3D reconstruction procedures. This property potentially allows direct quantitative

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

253

analyses of anatomical variability in the brain and facilitates the generation of population-averaged atlases. There are several 3D imaging techniques currently available. Micro-computed tomography (CT) can generate moderate to high-resolution images (1–100 μm) depending on the instrument and available X-ray source. It can be used to study bone and skull features, and, when used together with certain contrast agents, can reveal vascular information in the mouse brain and other organs (12, 13). However, micro-CT often fails to generate satisfactory contrast in the brain for structural delineation. Optical projection tomography (OPT) (14) is a recently developed technique, which provides high spatial resolution (∼5 μm) and imaging speed. However, the image contrast in the brain provided by OPT is limited because it is difficult to stain intact specimens. Compared to micro-CT and OPT, magnetic resonance imaging (MRI) provides far richer soft tissue contrast in the brain with modest spatial resolution (up to 10–20 μm). There are several unique tissue MRI contrasts that can be used to create an MRI-based atlas, including T1 , T2 , T2 ∗ , diffusion, and magnetization transfer. These MRI contrasts reflect the physical and chemical microenvironment of tissue water molecules, for example, water and myelin contents. Figure 12.1 shows examples of in vivo and ex vivo mouse brain MR images acquired with different contrast techniques. These MRI contrasts have been widely used to study normal brain anatomy and physiology, as well as various pathological conditions in the brain, in diseases, such as multiple sclerosis (15–17) and stroke (18, 19). Although the sensitivity and specificity of MRI contrasts cannot compete with the contrasts provided by histology, the ability to monitor anatomical and physiological changes

Fig. 12.1. Appearances of live and postmortem mouse brains with different MR contrasts. The in vivo images were acquired with a resolution of 0.1 × 0.1 × 0.4 mm3 . The ex vivo images were acquired with a resolution of 0.125 × 0.125 × 0.125 mm3 . DW, diffusion weighted; FA, fractional anisotropy; MT, magnetization transfer. The FA images represent one of the different contrasts derived from diffusion tensor imaging. The in vivo MT image is not coregistered with other in vivo images.

254


in vivo makes MRI the ideal imaging technique for many studies, and therefore it is imperative to have MRI-based atlases of mouse brains to guide and facilitate such studies. Given the advantages of MRI, it is not surprising that several mouse brain atlases already exist. Modern MRI-based mouse brain atlases contain high-resolution population-averaged brain images with detailed structural segmentations and annotations. An excellent overview of currently existing MRI-based mouse brain atlases can be found in Dorr et al. (20). The major applications of MRI-based mouse brain atlases include visualization of mouse brain anatomy, mapping and analyses of experimental data, and anatomical phenotyping in the mouse brain. In the following sections, we have presented the details involved in creating an MRI-based mouse brain atlas and have outlined the usefulness of such atlases through several applications.

2. Considerations in Creating an MRI-Based Mouse Brain Atlas

2.1. MR Images: The Foundations of MRI-Based Mouse Brain Atlases

An MRI-based mouse brain atlas may contain the following several components: MR images with one or multiple contrasts illustrating the mouse brain anatomy; detailed structural segmentations; and along with a user interface and image analysis tools. Two key features that are often used to compare existing MRIbased atlases are the spatial resolution and contrasts of the MR images in each atlas. These two important and interlocking factors ultimately determine the usefulness of an MRI-based atlas. Satisfactory image contrast and high spatial resolution are essential for resolving miniature structures in the brain. While it is possible to generate MRI-based atlases with multiple image contrasts, doing so will inevitably prolong the time needed to acquire a complete set of high-resolution data, which may become impractical due to degradation of the specimen and instrument instability. The choice of image contrast therefore often depends on the intended applications of the atlas. For example, if the intended applications of an atlas are analyses of T2 images of the mouse brain, then it is necessary to have T2 images in the atlas. Most current MRIbased mouse brain atlases are based on T2 or T2 ∗ MRI, because they are widely used to study mouse brain anatomy and pathology and provide satisfactory tissue contrast for structural delineation in the adult mouse brain. As for resolution is considered, high resolution requires longer imaging time and is not always practical. Clinically, anatomical images of the human brain can be routinely acquired at a resolution of 1 × 1 × 1 mm. Considering


255

that the human brain is approximately 3,000 times larger than the mouse brain in terms of total brain volume (21), a comparable resolution in the mouse brain would be 0.07 × 0.07 × 0.07 mm. Several existing adult mouse brain atlases provide resolution in the range of 0.03 × 0.03 × 0.03–0.05 × 0.05 × 0.05 mm (13, 22–24). Development of new imaging techniques that can improve image contrast and resolution is an important area of research. Recently, the use of “active staining” techniques (25) has also been introduced in the mouse brain that use contrast agents in conjunction with perfusion fixatives for increasing the tissue contrast and signal-to-noise ratio in MR images of the postmortem mouse brain (25, 26). Multi-spectral MR acquisition with enhanced T2 contrasts in the actively stained images has been shown to reveal more detailed morphological aspects of the mouse brain compared to conventional T2 imaging. With partial k-space acquisition and contrast agents that shorten tissue T1 (26), high-resolution mouse brain imaging with up to 21.5 μm isotropic resolution has been achieved (27). In the last decade, diffusion tensor imaging (DTI) has emerged as a novel MR technique that can reveal tissue microstructure with endogenous contrasts (28, 29). Tissue water diffusion can be characterized by diffusion constants along six different orientations. The distribution of diffusion constants along each direction is under the influence of local tissue microstructure. DTI measures these diffusion constants and fits them into a tensor model, from which the diffusion anisotropy and principal diffusion orientation are calculated. These parameters have been shown to have inherent sensitivity to tissue architecture and physiological conditions. The degree of anisotropy is sensitive to the existence of axonal projections and the degree of myelination (30). The principal diffusion orientation provides an additional white matter tissue contrast that is based on the trajectory of white matter tracts. The orientation and diffusion anisotropy data can be used to reconstruct the trajectories of white matter tracts in 3D, using the so-called fiber-tracking technique (31), as shown in Fig. 12.2b. Applications of DTI on human and animal brains have shown great potential for elucidating the complex white matter structures. Readers can find more information on this technique in detailed review articles (32–34). One important advantage of DTI over T1 and T2 MRIs is that it can provide superior contrasts to delineate anatomical structures in premature mouse brains (35, 36). Figure 12.3 shows ex vivo high-resolution T2 and diffusion tensor images of developing mouse brains from postnatal day 0 (p0, at birth) to postnatal day 80 (p80). Because the myelination process starts at approximately p7 in the mouse and is not completed until late postnatal stages and since tissue T2 is heavily influenced by myelin content, T2

256


Fig. 12.2. Diffusion tensor imaging of an adult mouse brain. a Nissl-stained histology (left) and diffusion tensor images of a perfusion-fixed adult mouse brain. b Reconstructed white matter tracts from the DTI data. 2n, optic nerve; ac, anterior commissure; cc, corpus callosum; cp, cerebral peduncle; DG, dentate gyrus; ec, external capsule; f, fornix; fi, fimbria; H, hippocampus; ml, medial lemniscus; opt, optic tract; py, pyramidal tract; sm, stria medularis. The scale bar represents 1 mm. The color arrows indicate the color coding used for diffusion anisotropy orientation. Red, green, and blue represent rostral-caudal, medial-lateral, and dorsal-ventral orientations, respectively.

images of early postnatal mouse brains show limited contrast for white matter structures. In comparison, diffusion tensor images provide superior white matter contrast, and the contrast is consistent from p0 to p80 and later stages. White matter structures, for example, the cerebral peduncle (cp) and optic tract (opt), can be easily identified in diffusion tensor images, but not in T2 images. DTI can also reveal anatomical structures in embryonic mouse brains, for example, the cortical plate and subventricular zone in the embryonic mouse cortex. DTI is therefore an important MR contrast for MRI-based atlases of embryonic and neonatal mouse brains. 2.2. Structural Segmentation

As described in the introduction, an atlas needs both highresolution anatomical images with rich anatomical contrasts and detailed structural delineations to be useful. In several histology based mouse brain atlases, numerous structures are manually delineated on images of stained tissue sections based on


257

Fig. 12.3. T2 and diffusion tensor color map images of postnatal mouse brains. In the color map images, we have used R, G, B colors to visualize white matter orientation. Red indicates that local tissue orientation is perpendicular to the plane, green indicates horizontal orientation, and blue indicates vertical orientation. The orange arrows indicate the locations of forceps major of the corpus callosum during p0–p10 or the locations of splenium of the corpus callosum during p20–p80. The yellow arrows indicate the locations of medial lemniscus, and the pink arrows indicate the locations of fasciculus retroflexus.

expert knowledge of the unique cellular and molecular markers of each structure and their spatial relationships. In MRI-based brain atlases, because the images are stored in 3D format, structural segmentation in the 3D images, in which voxels that belong to a particular structure are selected and classified as a 3D entity, is the common form of structural delineation. Segmentations in mouse brain MR images have been done either manually (20, 37) or semi-automatically (21, 38). Figure 12.4 shows examples of structural segmentations in images of an adult mouse brain based on both manual segmentation and fiber tracking for white matter structures. The numbers of structural segmentations in existing MRI-based mouse brain atlases vary approximately from 20 to 70 and are limited compared to hundreds of structures defined in histology based mouse brain atlases. In addition to structural segmentations, a user-friendly interface also adds to the value of a brain atlas. A user interface should allow users to browse through the anatomical images and structural annotations and can be located in users’ computers or on the web. The lower panel of Fig. 12.4 demonstrates the interface of our current MRI/DTI-based mouse brain atlas. It allows users to navigate through multiple 3D MR images with different tissue contrasts, read coordinates, display segmented structures and view overlays of 2D and 3D structural definitions, and to visualize the reconstructed structures in 3D.

258


Fig. 12.4. Structural segmentations and user interface of an MRI-based mouse brain atlas. Upper panels: coronal images of an adult mouse brain from our MRI/DTI-based mouse brain atlas. The top panel shows the T2 -weighted (T2 ), fractional anisotropy (FA), and direction-encoded color map (DEC) images, and the bottom panel shows the images overlaid with reconstructed white matter and gray matter structures currently included in the atlas. The color schemes for gray matter structures in T2 and FA images are as follows: white, neocortex; blue, hippocampus; purple, striatum. The color schemes for white matter structures are as follows: yellow: fimbria (fi); pink: corpus callosum (cc); magenta: fasciculus retroflexus (fr); red: stria terminalis (st); light green: fornix (fx); dark green: the trigeminal nerve (5n); light blue: cerebral peduncle (cp); dark blue: stria medularis (sm); light purple: mammillothalamic tract (mt); purple: optic tract (opt). Bottom panel: user interface of our data-viewing software, “AtlasView.” The software allows users to navigate through 3D MRI multicontrast data of a mouse brain. It overlays 2D and 3D structural definitions and visualizes reconstructed structures in 3D.

2.3. In Vivo and Ex Vivo Mouse Brain Atlases

Most current MRI-based atlases are based on images acquired from postmortem samples because it is relatively easy to acquire high-resolution images from postmortem specimens than from


259

live animals. Postmortem specimens, mostly formaldehyde fixed, can be trimmed to fit into small and sensitive MR coils, which, when combined with a high-performance gradient system, can generate high-resolution images with superior quality. Such combinations of optimized imaging hardware are often not available for in vivo MRI due to the added requirement of animal monitoring and support system. In addition, ex vivo MRI can last for 12 h or even longer, which can further improve the image quality by increasing signal averaging, while in vivo imaging may be susceptible to artifacts due to subject motion or the stability of animals, and is mostly limited to 2–3 h. All these factors have made ex vivo imaging the ideal choice for acquiring high-resolution and highquality images of the mouse brain. In comparison, in vivo MRI is limited by the sensitivity of the imaging instruments and the time that animals can stay relatively stable in the magnet under anesthesia. Even with various triggering techniques, e.g., ECG and respiratory triggers, it is still difficult to completely eliminate the effects of motion on the acquired images. The residual motion will cause a certain degree of degradation in the resolution and contrast of the images, as shown in Fig. 12.1. Even with the aforementioned disadvantages of in vivo MRI compared to ex vivo MRI, atlases based on in vivo MRI are still useful if the intended application of the atlas is to analyze in vivo MR images of the mouse brain. It has been reported that tissue contrasts in in vivo and ex vivo images are not entirely the same. The fixation process can alter the local physical and chemical environments experienced by water molecules due to the crosslinking of fixatives with macromolecules, which alters the appearance or contrast of the brain structures in ex vivo MRI compared to in vivo MRI. Another consideration is that there are noticeable morphological changes between the brains in live animals and in perfusion-fixed brain specimens. The lateral ventricles in postmortem specimens often have significantly reduced volumes or are completely collapsed due to removal of the cerebrospinal fluid (CSF) pressure. Also depending on the osmotic pressure of the fixation solution, the brain tissue may show slight enlargement and/or shrinkage. In Table 12.1, we have compared the structural volumes of the same mouse brains measured in vivo and ex vivo. The drastic changes in the volumes of the lateral ventricles can be appreciated. Although more detailed analyses are still needed, the results here suggest that ex vivo MRI-based atlases may not represent in vivo brain morphology accurately, necessitating the generation of in vivo MRI-based atlases (39). 2.4. Single-Subject and PopulationAveraged Brain Atlases

While histology based atlases are mostly based on one or a few specimens, the 3D images generated by MRI and the recent advances in computational techniques make it possible to perform spatial normalization and averaging of images from multiple

260


Table 12.1 Comparisons of regional brain volumes (mm3 ) measured using in vivo MRI and ex vivo MRI from perfusion-fixed mouse brains. Data are expressed as mean ± SE (n=3). Brain region

In vivo MRI

Ex vivo MRI

Whole brain

486.95±9.32

465.65±2.12

4.4

Striatum

29.87±0.94

23.37±0.07

21.8

Hippocampus

28.45±2.27

26.34±0.84

7.4

119.02±2.09

110.74±1.92

6.9

64.05±1.98

59.30±1.81

7.4

Cortex Cerebellum Lateral ventricle

4.88±1.27

0.05±0.001

% shrinkage

99.0

specimens. It is also possible to use these techniques to generate so-called minimal deformation atlases that approximate the geometrical average of normal brains and reflect the average morphological characteristics of the sample population. Detailed procedures for generating unbiased population-averaged brain images can be found in (37, 40, 41). An additional benefit of group averaging is that the averaged images have higher signal-to-noise ratio than the individual images, and therefore, facilitate structural segmentation and visualization (20). While it is beneficial to have population-averaged brain atlases, the usefulness of these atlases can be constrained by the image normalization techniques adopted. High accuracy of the image normalization step is critical for atlases based on population averaging. Since the spatial normalization of images is based on similarity of their pixel intensity values, the subject and template images should have similar contrast patterns for the mapping to be accurate. Ex vivo images have high spatial resolution and strong structural contrasts, which makes it relatively easy to construct mappings between ex vivo images in order to create population-averaged atlases. In vivo images, on the other hand, have limited structural contrasts as explained before. It is therefore more challenging to achieve accurate mapping between in vivo images. Averaging of inaccurately normalized images can cause further degradation of the tissue contrast and result in poor quality in registration of subsequent in vivo subject images to the population-averaged brain image. Figure 12.5 A shows our current population-averaged ex vivo mouse brain atlas with averaged T2 and DTI data and structural segmentation. With the recent advances in fast in vivo brain imaging, we now demonstrate that it is possible to construct in vivo population-averaged brain atlases using DTI. Images of 2month-old female C57BL/6 mouse brains were acquired with a


261

Fig. 12.5. In vivo and ex vivo group-averaged mouse brain atlases based on T2 MRI and DTI. a Group-averaged ex vivo T2 and diffusion tensor images of adult mouse brains (C57BL/6, 2 month old, n = 10) and structural segmentations (Atlas). b Group-averaged in vivo mouse brain atlas with preliminary structural segmentations (C57BL/6, 2 month old, n = 9). c Mid-sagittal view of the population-averaged in vivo images and structural segmentations. In the direction-encoded colormap (DEC) images, red, green, and blue represent the medial-lateral, anterior-posterior, and superior-inferior axes, respectively.

spatial resolution of 0.1 × 0.1 × 0.4 mm3 and total imaging time of 2–3 h. After iterative spatial normalization and averaging, the population-averaged images from nine mouse brains were generated (Fig. 12.5b). The rich tissue contrasts provided by DTI and high spatial resolution provided by fast 3D imaging techniques enable accurate spatial normalization, as can be appreciated from the sharp structural boundaries in the averaged images. Because the images were acquired in 3D fashion, the data can be viewed and manipulated in 3D (Fig. 12.5c), which is important for examining 3D structural properties. We have performed initial segmentation of 58 structures in the averaged brain images (Fig. 12.5 Atlas) based on our ex vivo atlas. In general, ex vivo T2 images have richer tissue contrasts than in vivo T2 images, probably due to tissue fixation and higher spatial resolution. The volume of the lateral ventricles in the in vivo images is significantly larger than the ex vivo images due to the lack of CSF pressure in postmortem samples. In comparison, the DTI contrast patterns are relatively undisturbed from in vivo to ex vivo, a finding that has also been reported by other groups (42, 43).

3. Applications of MRI-Based Mouse Brain Atlas 3.1. Morphological Studies

Measuring changes in brain morphology (shapes and volumes of specific anatomical areas) is important for many studies, but not always straightforward even with 3D MRI data. Manual delineation of structures in serial MR images is time-consuming and the results from different operators and different laboratories may

262


not be compatible due to differences in definition of anatomical boundaries and inter-rater variations. With the availability of brain atlases and image registration and mapping techniques, we now can characterize brain morphology efficiently and consistently. In this section, an application of atlas-based analysis in quantitative characterization of neurodegenerative atrophy in the R6/2 mouse brain is demonstrated. The R6/2 transgenic mouse is a widely used model of Huntington’s disease (HD) in pre-clinical therapeutic trials. It has progressive HD-like gross atrophy in the brain and especially in the striatum (44). We collected high resolution 3D T2 -weighted images of R6/2 mice (n = 7) and wild-type littermates (n = 8) longitudinally from 3 to 12 weeks after birth. To analyze the differences in structural volumes between R6/2 and wild-type mice, a single-subject mouse brain atlas was first created. The mouse brain images used in the atlas were selected from a set of in vivo T2 -weighted MR images of C57BL/6 mice (n = 10, female, 12 weeks old), one of the background strains of the R6/2 strain. The atlas image has whole brain and ventricular volumes (obtained via manual segmentation) close to the median values of the 10 mice. The image was then manually adjusted to the orientation defined in the Paxinos’ atlas (3) and resampled to an isotropic resolution of 0.1 mm × 0.1 mm × 0.1 mm per pixel. This atlas contains manual segmentation of 10 brain structures that follow closely the definition by Paxinos (3). The brain was segmented from the rostral ends of the olfactory bulbs to the caudal end of the cerebellum; the cortex was defined by the corpus callosum and external capsule, with the ventral boundary by rhinal fissure (we excluded the part ventral to the rhinal fissure in this study); the striatum was defined by the corpus callosum, external capsule, and anterior commissure; the hippocampus was defined by the external capsule, lateral and third ventricle, and thalamus; and the ventricle was defined by intense signal from CSF. The boundary of striatum and accumbens is often difficult to identify even with histology slides, but borders of the corpus callosum, external capsule, anterior commissure, and globus pallidus are clear in MRI. We used clearly identifiable anatomical landmarks and the Paxinos’ atlas to define the boundary as reproducibly as possible. Linear and nonlinear image transformations were used for spatial normalization of R6/2 and wild-type mouse brain images. Figure 12.6 shows the images from three R6/2 and three wildtype mouse brains after linear affine transformation and nonlinear registration based on the LDDMM technique (45). Affine transformation can adjust the overall size and orientation of each image to the atlas, but to normalize detailed morphology, e.g., the shape of the lateral ventricles, to the atlas requires accurate nonlinear transformation. LDDMM can accurately transform both wild-type and R6/2 mouse brain images to the atlas so that the


263

Fig. 12.6. A diagram that illustrates atlas-based analysis of brain atrophy in the R6/2 mouse, a model of Huntington’s disease. The atlas is a single-subject atlas with embedded structural delineations. For purpose of illustration, we have used the yellow dashed line along the brain surface, the blue curve on the corpus callosum and external capsule, and the solid green line outlining the ventricles to represent the structural delineations in the atlas. Images from R6/2 and wild-type mice (shown here with three mice in each group) were spatially normalized to the atlas image using affine transformation and LDDMM. After spatial normalization, the structure delineation defined in the atlas image can be transferred and overlaid on the transformed subject images.

structural delineations embedded in the atlas can be directly transferred to the normalized images. Using inverse transformations that deform the atlas image and structural delineations to the subject images, automated segmentation of major gray matter structures (neocortex, striatum, hippocampus, cerebellum, etc) can be achieved. The accuracy of this segmentation approach was examined using expert manual segmentation as the gold standard. Strong correlations between the manual and atlas-based segmentation results (Table 12.2) indicate the accuracy of the atlas-based approach. Figure 12.7 further demonstrates the quality of atlas-based automated segmentation of p21, p42, and p84 mouse brain images with LDDMM. The transformations generated by LDDMM carry structural segmentation in the atlas to each subject image. The segmented structures are outlined and overlaid on in vivo MR images, and the reconstructed surfaces were visualized in 3D. Volumes of major brain structures were obtained via automated segmentation followed by manual correction (Fig. 12.7 b–e). The R6/2 mice show significantly reduced whole brain volume and hippocampal volume at 4 weeks of age (p28). Differences in the striatal volume and lateral ventricles become significant at 5 weeks of

264


Table 12.2 R2 values of correlation analysis of the manual segmentation and LDDMM-based automated segmentation results at 4, 6, and 12 weeks of age in both the wild-type and R6/2 (within the parentheses) mice. Structures

4 weeks

6 weeks

12 weeks

Striatum

0.841 (0.997)

0.959 (0.925)

0.956 (0.571)

Hippocampus

0.985 (0.985)

01.977 (0.653)

0.899 (0.704)

Lateral ventricles

0.968 (0.969)

0.942 (0.933)

0.952 (0.919)

Cortex

0.966 (0.942)

0.618 (0.985)

0.771 (0.871)

age (p35). The cerebellar volume showed no significant change. These results show that MRI can detect atrophy in R6/2 mice as early as p28, and the atrophy progresses afterwards. These results show the usefulness of the atlas for detecting atrophy in mouse brains. 3.2. Combined Micro-CT and DTI-Based Atlas for Guidance of Stereotaxic Operations

The precision of targeting structures during stereotaxic surgery in the mouse brain depends on the accuracy of the brain atlas used to guide such operations. 2D histology based atlases that provide stereotaxic coordinates of brain sections relative to skull landmarks are conventionally used for guidance of stereotaxis in the mouse brain. MRI-based mouse brain atlases can provide anatomical information of the brain, but lack bone tissue contrasts for identification of cranial landmarks, which are essential for stereotaxis. Recently, 3D stereotaxic mouse brain atlases based on combining MR images with micro-CT have been developed (37, 46). To construct single-subject stereotaxic atlases, C57BL/6 mouse head specimens at different stages of postnatal development were scanned with MRI and subsequent micro-CT. To incorporate the brain images in skull-based stereotaxic coordinates, the 3D MR images were coregistered to micro-CT images of the same subject, that provide fine bone tissue contrasts for delineation of skull surface landmarks (Fig. 12.8). The coregistered CT-MRI images were oriented with the lambda and bregma landmarks in the same horizontal plane, which is the standard orientation used in most histology based stereotaxic atlases of murine brains (3, 47). The atlases contain CT and multiple MR contrasts (T2 -weighted, diffusion-weighted, fractional anisotropy and diffusion orientation). As mentioned previously, DTI contrasts are particularly useful for delineation of anatomical structures during early postnatal stages, when incomplete myelination


265

Fig. 12.7. a Automated segmentation of R6/2 mouse brain images. Pre-defined segmentation of major gray matter structures (caudate putamen: yellow; hippocampus: green; cerebellum: purple; lateral ventricle: blue) in the atlas mouse brain images was transformed to individual images. b–e MR-based volume measurements of the brain, lateral ventricles, striatum, and hippocampus. ∗ Significant (p

Magnetic Resonance Neuroimaging: Methods and Protocols (Methods in Molecular Biology 711)

Surface Plasmon Resonance: Methods and Protocols (Methods in Molecular Biology)

Yeast Protocols (Methods in Molecular Biology) (Methods in Molecular Biology)

Molecular Imaging: Methods and Protocols (Methods in Molecular Biology)

Molecular Chaperones: Methods and Protocols (Methods in Molecular Biology, v787)

Molecular Chaperones: Methods and Protocols (Methods in Molecular Biology 787)

Molecular Profiling: Methods and Protocols (Methods in Molecular Biology, 823)

Molecular Toxicology Protocols (Methods in Molecular Biology)

Bioinformatics. Methods and Protocols in Molecular Biology

Microchip Capillary Electrophoresis: Methods And Protocols (Methods in Molecular Biology) (Methods in Molecular Biology)

Epidermal Cells: Methods and Protocols (Methods in Molecular Biology) (Methods in Molecular Biology Series)

Cytoskeleton Methods and Protocols (Methods in Molecular Biology Vol 161)

Human Embryogenesis: Methods and Protocols (Methods in Molecular Biology)

Immunological Tolerance: Methods and Protocols (Methods in Molecular Biology)

Clostridium difficile: Methods and Protocols (Methods in Molecular Biology, 646)

Hepatocytes: Methods and Protocols (Methods in Molecular Biology 640)

Biofuels: Methods and Protocols (Methods in Molecular Biology 581)

Bioconjugation Protocols: Strategies and Methods (Methods in Molecular Biology)

Riboswitches: Methods and Protocols (Methods in Molecular Biology)

Bioinformatics Methods and Protocols (Methods in Molecular Biology)

Argonaute Proteins: Methods and Protocols (Methods in Molecular Biology 725)

Neuropeptides: Methods and Protocols (Methods in Molecular Biology 789)

Microbial Toxins: Methods and Protocols (Methods in Molecular Biology 739)

Retinoids: Methods and Protocols (Methods in Molecular Biology)

Human Papillomaviruses: Methods and Protocols (Methods in Molecular Biology)

Tumor Biomarker Discovery: Methods and Protocols (Methods in Molecular Biology)

Potassium Channels: Methods and Protocols (Methods in Molecular Biology)

Metabolomics: Methods and Protocols (Methods in Molecular Biology Vol 358)

Mast Cells: Methods and Protocols (Methods in Molecular Biology)

Tissue Microarrays: Methods and Protocols (Methods in Molecular Biology)

Nanostructure Design: Methods and Protocols (Methods in Molecular Biology)

Magnetic Resonance Neuroimaging: Methods and Protocols (Methods in Molecular Biology 711)

Surface Plasmon Resonance: Methods and Protocols (Methods in Molecular Biology)

Yeast Protocols (Methods in Molecular Biology) (Methods in Molecular Biology)

Molecular Imaging: Methods and Protocols (Methods in Molecular Biology)

Molecular Chaperones: Methods and Protocols (Methods in Molecular Biology, v787)

Molecular Chaperones: Methods and Protocols (Methods in Molecular Biology 787)

Molecular Profiling: Methods and Protocols (Methods in Molecular Biology, 823)

Molecular Toxicology Protocols (Methods in Molecular Biology)

Bioinformatics. Methods and Protocols in Molecular Biology

Microchip Capillary Electrophoresis: Methods And Protocols (Methods in Molecular Biology) (Methods in Molecular Biology)

Epidermal Cells: Methods and Protocols (Methods in Molecular Biology) (Methods in Molecular Biology Series)

Cytoskeleton Methods and Protocols (Methods in Molecular Biology Vol 161)

Human Embryogenesis: Methods and Protocols (Methods in Molecular Biology)

Immunological Tolerance: Methods and Protocols (Methods in Molecular Biology)

Clostridium difficile: Methods and Protocols (Methods in Molecular Biology, 646)

Hepatocytes: Methods and Protocols (Methods in Molecular Biology 640)

Biofuels: Methods and Protocols (Methods in Molecular Biology 581)

Bioconjugation Protocols: Strategies and Methods (Methods in Molecular Biology)

Riboswitches: Methods and Protocols (Methods in Molecular Biology)

Bioinformatics Methods and Protocols (Methods in Molecular Biology)

Argonaute Proteins: Methods and Protocols (Methods in Molecular Biology 725)

Neuropeptides: Methods and Protocols (Methods in Molecular Biology 789)

Microbial Toxins: Methods and Protocols (Methods in Molecular Biology 739)

Retinoids: Methods and Protocols (Methods in Molecular Biology)

Human Papillomaviruses: Methods and Protocols (Methods in Molecular Biology)

Tumor Biomarker Discovery: Methods and Protocols (Methods in Molecular Biology)

Potassium Channels: Methods and Protocols (Methods in Molecular Biology)

Metabolomics: Methods and Protocols (Methods in Molecular Biology Vol 358)

Mast Cells: Methods and Protocols (Methods in Molecular Biology)

Tissue Microarrays: Methods and Protocols (Methods in Molecular Biology)

Nanostructure Design: Methods and Protocols (Methods in Molecular Biology)

Recommend Documents