Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Myelin water imaging : development at 3.0T, application to the study of multiple sclerosis, and comparison… Kolind, Shannon Heather 2008

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2009_spring_kolind_shannon.pdf [ 3.41MB ]
Metadata
JSON: 24-1.0066854.json
JSON-LD: 24-1.0066854-ld.json
RDF/XML (Pretty): 24-1.0066854-rdf.xml
RDF/JSON: 24-1.0066854-rdf.json
Turtle: 24-1.0066854-turtle.txt
N-Triples: 24-1.0066854-rdf-ntriples.txt
Original Record: 24-1.0066854-source.json
Full Text
24-1.0066854-fulltext.txt
Citation
24-1.0066854.ris

Full Text

  MYELIN WATER IMAGING: DEVELOPMENT AT 3.0 T, APPLICATION TO THE STUDY OF MULTIPLE SCLEROSIS, AND COMPARISON TO DIFFUSION TENSOR IMAGING  by  Shannon Heather Kolind  B. Sc., The University of Victoria, 2001 M. Sc., The University of British Columbia, 2003    A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY  in  THE FACULTY OF GRADUATE STUDIES  (Physics)    THE UNIVERSITY OF BRITISH COLUMBIA   (Vancouver)    December 2008     © Shannon Heather Kolind, 2008  ii 0BABSTRACT  T2 relaxation imaging can be used to measure signal from water trapped between myelin bilayers; the ratio of myelin water signal to total water is termed the myelin water fraction (MWF).  First, results from multi-component T2 relaxation and diffusion tensor imaging (DTI) were compared for 19 multiple sclerosis (MS) subjects at 1.5 T to better understand how each measure is affected by pathology.  In particular, it was determined that the detection of a long-T2 signal within an MS lesion may be indicative of a different underlying pathology than is present in lesions without long-T2 signal.  Next, the single-slice T2 relaxation measurement was implemented, refined, and validated at 3.0 T.  Scan parameters were varied for phantoms and in-vivo brain, and changes in multi-exponential fit residuals and T2 distribution-derived parameters such as MWF were monitored to determine which scan parameters minimized artifacts.  Measurements were compared between 1.5 T and 3.0 T for 10 healthy volunteers.  MWF maps were qualitatively similar between field strengths.  MWFs were significantly higher at 3.0 T than at 1.5 T, but with a strong correlation between measurements at the different field strengths.  Due to long acquisition times, multi-component T2 relaxation has thus far been clinically infeasible.  The next study aimed to validate a new 3D multi-component T2 relaxation imaging technique against the 2D single-slice technique most commonly used.  Ten  iii healthy volunteers were scanned with both the 2D single-slice and 3D techniques.  MWF maps were qualitatively similar between scans.  MWF values were highly correlated between the acquisition methods.  Although MWF values were generally lower using the 3D technique, they were only significantly so for peripheral brain structures, likely due to increased sensitivity of slab-selective refocusing pulses used for the 3D approach.  The 3D T2 relaxation sequence was then applied to the study of MS to take advantage of the increased brain coverage.  Thirteen MS subjects and 11 controls underwent T2 relaxation and DTI examinations to produce histograms of MWF and several DTI- derived metrics.  MS MWF histograms differed considerably from those of controls, and differences in MS MWF histograms did not mirror differences in DTI histograms relative to matched controls.  iv TABLE OF CONTENTS  ABSTRACT........................................................................................................................ ii TABLE OF CONTENTS................................................................................................... iv LIST OF TABLES............................................................................................................ vii LIST OF FIGURES ......................................................................................................... viii GLOSSARY ....................................................................................................................... x ACKNOWLEDGEMENTS.............................................................................................. xii CO-AUTHORSHIP STATEMENT.................................................................................. xv 1 INTRODUCTION ...................................................................................................... 1 1.1 BRAIN ................................................................................................................ 1 1.1.1 White and grey matter................................................................................. 1 1.1.2 Myelin ......................................................................................................... 2 1.1.2.1 Myelin structure ...................................................................................... 2 1.1.2.2 Myelin function....................................................................................... 4 1.2 MULTIPLE SCLEROSIS................................................................................... 4 1.2.1 MS pathology.............................................................................................. 6 1.2.1.1 Lesions .................................................................................................... 7 1.2.1.2 Normal appearing white matter (NAWM).............................................. 8 1.3 MAGNETIC RESONANCE IMAGING............................................................ 8 1.3.1 Measuring T2 in brain................................................................................ 10 1.3.1.1 T2 decay curve acquisition .................................................................... 12 1.3.1.2 T2 decay curve analysis......................................................................... 17 1.3.2 Measuring the diffusion tensor in brain .................................................... 20 1.3.2.1 Diffusion MRI acquisition .................................................................... 21 1.3.2.2 Diffusion analysis ................................................................................. 23 1.4 OVERVIEW OF THESIS................................................................................. 25 1.5 REFERENCES ................................................................................................. 28 2 ROI ANALYSIS COMPARING MULTI-COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 1.5T ................................................................................................................................... 32 2.1 INTRODUCTION ............................................................................................ 32 2.2 METHODS ....................................................................................................... 37 2.2.1 Subject information................................................................................... 37 2.2.2 MR data acquisition .................................................................................. 37 2.2.3 MR data analysis....................................................................................... 38 2.3 RESULTS ......................................................................................................... 41 2.4 DISCUSSION................................................................................................... 48 2.4.1 Comparison between healthy white matter and MS cNAWM ................. 48 2.4.2 Comparison between cNAWM and lesions with and without long-T2 components ............................................................................................................... 49 2.4.3 Correlations between MWC and diffusion metrics in lesions with long-T2 components ............................................................................................................... 50 2.4.4 Difference between lesions with and without long-T2 components.......... 51 2.5 ACKNOWLEDGEMENTS.............................................................................. 53  v 2.6 REFERENCES ................................................................................................. 54 3 IMPLEMENTATION AND DEVELOPMENT OF MULTI-COMPONENT T2 RELAXATION IMAGING AT 3.0T, AND VALIDATION AGAINST 1.5T MEASUREMENTS.......................................................................................................... 60 3.1 INTRODUCTION............................................................................................... 60 3.2 MATERIALS AND METHODS ........................................................................ 63 3.2.1 Pulse sequence development at 3.0T ............................................................ 63 3.2.2 MR data acquisition ...................................................................................... 64 3.2.3 MR data analysis........................................................................................... 66 3.3 RESULTS............................................................................................................ 69 3.3.1 3.0T multi-exponential T2 relaxation pulse sequence refinement................. 69 3.3.2 Comparison between multi-exponential T2 relaxation at 1.5T and 3.0T ...... 75 3.4 DISCUSSION...................................................................................................... 79 3.4.1 3.0T multi-exponential T2 relaxation pulse sequence development-residuals of the multi-exponential fit ....................................................................................... 79 3.4.2 3.0T multi-exponential T2 relaxation pulse sequence development-MWF, 2T and IE peak width ..................................................................................................... 82 3.4.3 Comparison between multi-exponential T2 relaxation at 1.5T and 3.0T ...... 84 3.5 ACKNOWLEDGMENTS................................................................................... 86 3.6 REFERENCES .................................................................................................... 87 4 VALIDATION OF 3D MULTI-COMPONENT T2 RELAXATION IMAGING AGAINST THE 2D SINGLE-SLICE TECHNIQUE AT 3.0T........................................ 90 4.1 INTRODUCTION............................................................................................... 90 4.2 MATERIALS AND METHODS ........................................................................ 92 4.2.1 MR data acquisition ...................................................................................... 92 4.2.2 analysis.......................................................................................................... 95 4.3 RESULTS............................................................................................................ 97 4.3.1 MWF values: 2D single-slice compared to 3D measurements ..................... 98 4.3.2 2T  values: 2D single-slice compared to 3D measurements ........................ 101 4.3.3 Standard deviation of residuals: 2D single-slice compared to 3D measurements.......................................................................................................... 102 4.3.4 Qualitative comparison of parameter maps ................................................ 102 4.3.5 NNLS fit residuals ...................................................................................... 104 4.4 DISCUSSION.................................................................................................... 106 4.4.1 Differences in MWF and 2T  values............................................................ 106 4.4.2 Differences in standard deviation of residuals............................................ 108 4.4.3 Comparison to multi-slice multi-component T2 relaxation techniques....... 109 4.4.4 Concluding remarks .................................................................................... 111 4.5 ACKNOWLEDGMENTS................................................................................. 111 4.6 REFERENCES .................................................................................................. 112 5 VOXEL-WISE (HISTOGRAM) ANALYSIS COMPARING 3D MULTI- COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 3.0T........................................................................... 115 5.1 INTRODUCTION............................................................................................. 115 5.2 MATERIALS AND METHODS ...................................................................... 118 5.2.1 Subject information..................................................................................... 118  vi 5.2.2 MR data acquisition .................................................................................... 119 5.2.3 MR data analysis......................................................................................... 120 5.2.4 Lesion and NAWM identification .............................................................. 121 5.2.5 Histogram analysis...................................................................................... 123 5.2.6 Statistical analysis....................................................................................... 124 5.3 RESULTS.......................................................................................................... 124 5.4 DISCUSSION.................................................................................................... 134 5.4.1 Histograms across subjects (Figure 5.2 and Figure 5.3)............................. 137 5.4.2 Average histograms and parameter values (Figure 5.4 and Table 5.2) ...... 138 5.4.3 Histogram differences between subjects (Figure 5.5 and Figure 5.6) ........ 140 5.4.4 Correlations with EDSS and disease duration (Table 5.3) ......................... 141 5.4.5 Correlations between MR-derived metrics (Table 5.3) .............................. 142 5.5 CONCLUSION ................................................................................................. 143 5.6 ACKNOWLEDGMENTS................................................................................. 143 5.7 REFERENCES .................................................................................................. 144 6 CONCLUSIONS..................................................................................................... 149 6.1 ROI ANALYSIS COMPARING MULTI-COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 1.5T....................................................................................................................... 150 6.1.1 Hypotheses.............................................................................................. 150 6.1.2 Significance............................................................................................. 150 6.1.3 Strengths and weaknesses ....................................................................... 151 6.1.4 Potential applications and future work ................................................... 151 6.2 IMPLEMENTATION AND DEVELOPMENT OF MULTI-COMPONENT T2 RELAXATION IMAGING AT 3.0T, AND VALIDATION AGAINST 1.5T MEASUREMENTS.................................................................................................... 151 6.2.1 Hypotheses.............................................................................................. 152 6.2.2 Significance............................................................................................. 153 6.2.3 Strengths and weaknesses ....................................................................... 153 6.2.4 Potential applications and future work ................................................... 154 6.3 VALIDATION OF 3D MULTI-COMPONENT T2 RELAXATION IMAGING AGAINST THE 2D SINGLE-SLICE TECHNIQUE AT 3.0T.................................. 155 6.3.1 Hypotheses.............................................................................................. 155 6.3.2 Significance............................................................................................. 155 6.3.3 Strengths and weaknesses ....................................................................... 156 6.3.4 Potential applications and future work ................................................... 156 6.4 VOXEL-WISE (HISTOGRAM) ANALYSIS COMPARING 3D MULTI- COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 3.0T.................................................... 157 6.4.1 Hypotheses.............................................................................................. 157 6.4.2 Significance............................................................................................. 158 6.4.3 Strengths and weaknesses ....................................................................... 158 6.4.4 Potential applications and future work ................................................... 159 6.5 RELATED WORK ......................................................................................... 159 6.6 REFERENCES ............................................................................................... 162 APPENDIX A: ETHICS CERTIFICATES.................................................................... 164  vii LIST OF TABLES  Table 2.1: Mean value of MWC, FA, <D>, λ | , and λ|| for cNAWM and all lesions ........ 42 Table 2.2: Mean value of MWC, FA, <D>, λ | , and λ|| for lesions with and without long-T2 signal ......................................................................................................................... 43 Table 3.1: 2T  and peak width for the IE peak for 10 volunteers using 1.5T and 3.0T ..... 78 Table 3.2: SNR and standard deviation of residuals for 10 volunteers using 1.5T and 3.0T ................................................................................................................................... 78 Table 4.1: MWF, 2T  and standard deviation of residuals for the 2D and 3D relaxation techniques ................................................................................................................. 98 Table 5.1: Clinical data as well as analyzed NAWM and lesion volume for MS subjects, and NWM volume for control subjects................................................................... 125 Table 5.2: Histogram measures for MWF, 2T , FA, λ | , λ||, and <D>, averaged across all MS subjects or all controls...................................................................................... 129 Table 5.3: Significant correlation coefficients for EDSS and disease duration, MWF in NAWM and lesion, and 2T  in lesion, with various diffusion histogram metrics. .. 133   viii LIST OF FIGURES  Figure 1.1: Schematic of a neuron ...................................................................................... 2 Figure 1.2: Electron micrograph of a transaction of a myelinated axon............................. 3 Figure 1.3: Disease progression as a function of time of the phenotypes of MS ............... 6 Figure 1.4: A typical T2 distribution for healthy brain showing the different water components ............................................................................................................... 11 Figure 1.5: Phase diagram for a CPMG pulse sequence................................................... 12 Figure 1.6: CPMG T2 decay.............................................................................................. 13 Figure 1.7: First portion of a Poon-Henkelman T2 relaxation pulse sequence. ................ 16 Figure 1.8: Stejskal-Tanner pulsed field gradient diffusion MRI pulse sequence............ 21 Figure 2.1: Correlations between MWC and FA, <D>, λ | , and λ|| for cNAWM.............. 44 Figure 2.2: Correlations between MWC and FA, <D>, λ | , and λ|| for the subset of lesions with no long-T2 signal, and lesions exhibiting long-T2 signal. ................................. 45 Figure 2.3: Correlations between LT2F and FA, <D>, λ | , and λ|| for lesions exhibiting long-T2 signal ............................................................................................................ 46 Figure 2.4: First echo image, MWC map, LT2F map, <D> map, λ |  map, and λ|| map, for one MS patient. ......................................................................................................... 47 Figure 3.1: Standard deviation of residuals for water-based phantoms, fixed brain, and in- vivo brain. ................................................................................................................. 70 Figure 3.2: MWF values for fixed brain, and in-vivo brain.............................................. 71 Figure 3.3: 2T  for water-based phantoms, fixed brain and in-vivo brain ......................... 72 Figure 3.4: Peak widths for fixed brain and in-vivo brain................................................ 74 Figure 3.5: Residuals of the multi-exponential fit for fixed brain and in-vivo brain........ 75 Figure 3.6: MWF maps at 1.5T and 3.0T for one volunteer............................................. 76 Figure 3.7: Correlations between average MWF at 1.5T and 3.0T .................................. 77 Figure 4.1: First portion of the 2D single-slice and 3D pulse sequences ......................... 93 Figure 4.2: Correlation between 3D and 2D MWF measurements .................................. 99 Figure 4.3: Bland-Altman plots for 3D and 2D measurements of MWF and 2T ........... 100 Figure 4.4: Correlation between 3D and 2D 2T  measurements...................................... 101 Figure 4.5: Maps of MWF, 2T , and standard deviation of residuals using the 2D and the 3D technique ........................................................................................................... 103 Figure 4.6: Map of transmitted field sensitivity and difference between the 2T  calculated from the 3D and 2D T2 data .................................................................................... 104 Figure 4.7: Residuals of the multi-exponential fit for 2 brain structures using the 2D and 3D techniques.......................................................................................................... 105 Figure 5.1: Registered FLAIR, 3D T1 TFE, NAWM mask, MWF map, <D> map and FA map for one MS subject .......................................................................................... 123 Figure 5.2: Histograms of MWF and 2T  for NWM for all healthy controls and NAWM for all MS subjects .................................................................................................. 126 Figure 5.3: Histograms of FA, λ | , λ||, and <D> for NWM for all healthy controls and NAWM for all MS subjects .................................................................................... 127  ix Figure 5.4: Average histograms for NAWM and lesion across all MS subjects, and NWM across the control subjects for MWF, 2T , FA, λ | , λ||, and <D> .............................. 128 Figure 5.5: Histograms of MWF, 2T , FA, λ | , λ||, and <D> for 3 of the MS subjects and their respective age and gender matched controls .................................................. 131 Figure 5.6: White matter masks for 3 MS subjects and their respective age and gender matched controls. .................................................................................................... 132    x 1BGLOSSARY  ANOVA Analysis of variance B0 Main magnetic field strength B1 Radio frequency field strength B1+ B1 transmitted field b Factor summarizing influence of gradients on DWI BW Bandwidth cNAWM Contralateral normal appearing white matter CLEAR Constant level appearance algorithm CNS Central nervous system CPMG Carr-Purcell-Meiboom-Gill (spin-echo pulse sequence) CSF Cerebrospinal fluid D Translational diffusion co-efficient <D> Mean diffusivity DTI Diffusion tensor imaging DWI Diffusion weighted imaging EDSS Expanded Disability Status Scale EPI Echo planar imaging FA Fractional anisotropy FLAIR  Fluid attenuated inversion recovery FOV Field of view GM Grey matter IE Intra and extra-cellular λ|| Parallel diffusivity λ U | U  Perpendicular diffusivity LT2F Long-T2 fraction MRI Magnetic resonance imaging MS Multiple sclerosis MWC Myelin water content MWF Myelin water fraction NAGM Normal appearing grey matter NAWM Normal appearing white matter NGM Normal grey matter NNLS Non negative least squares NWM Normal white matter PD Proton density PNS Peripheral nervous system RF Radio frequency ROI Region of interest SAR Specific absorption rate SE Standard error SENSE Sensitivity encoding SNR Signal-to-noise ratio T1 Spin-lattice relaxation time T2 Spin-spin relaxation time  xi 2T  Geometric mean T2 TE Echo time TFE Turbo field echo TI Inversion time TR Repetition time T/R Transmit/receive WM White matter   xii 2BACKNOWLEDGEMENTS  Huge thanks to Alex MacKay, the best supervisor I could ask for, I have learned so much under your guidance.   I would also like to thank the rest of my supervisory committee, David Li, Carl Michal, and San Xiang.  I have been very lucky to have such varied points of view and academic backgrounds, and have benefited greatly from your advice. Special thanks to Alex MacKay, David Li and San Xiang for helping me apply for funding repeatedly, which has set me up not only for this project but the next.  I would also like to recognize the work and contributions of my university examiners Stefan Reinsberg and Robert Rohling as well as my external examiner Andrew Alexander.  Thanks also to my examining committee chair John O’Kusky.  Burkhard Mädler also played a large role in the work that went into this thesis, thanks for all the hard work and invaluable feedback.  Thanks also to Corree Laule and Irene Vavasour for endless help and humour.  I have also had support from several students at various levels: Thor Bjarnason, Nicole Fichtner, Saeed Kalantari, Matt Lam, Joseph Lee, Erin MacMillan, Sandra Meyers, Kirstie Whitaker, and Eugene Yip.  I have had so much fun and learned a lot from working with you.  Particular gratitude to Alex’s angels (CL, IV and EM) for putting up with me as a roommate at various meetings.   xiii I’ve also had fantastic advice on a variety of topics from Tony Traboulsee, Wayne Moore, Stefan Fischer, Roger Tam, Andrew Riddehough, Donna Lang, Ken Whittall, Alex Rauscher, Piotr Kozlowski, Craig Jones, and Trevor Andrews.  Particular thanks to Dr Moore for assistance with the introductory pathology section of this thesis.  Thanks to Agnes d’Entremont for creating the coolest MRI music video ever, all the work with metal artifact, and repeatedly saving my life at ISMRM.  My life and the group in general would fall apart without Linda Chandler.  We have amazing MR technologists, thanks Trudy Harris, Paula Coutts, Paul Hamill, Jennifer McCord, Sylvia Renneberg, Linda Zimmer, Karen Smith, Lesley Costley and Linda James.  I can’t believe you kept letting me use the scanner after I crashed it so many times.  Thanks for teaching me so much about the scanner, and about patience.  Sincere thanks to all of our volunteers, especially to the MS patients, who participated in these studies despite many difficulties.  In the physics department at UBC, I would particularly like to thank Fran Bates, Dave Axen, Doug Bonn, Joseph O’Connor, Domenic Di Tomaso, Janis McKenna, Lydia Wong, Oliva Dela Cruz-Cordero, and Lori Boerma.   xiv This work could not have been done without the generous support of NSERC, CIHR, the Killam Trusts (special thanks to Jeffrey Hsu), and the MS Society of Canada (special thanks to Jackie Munroe).  I will always be grateful for the opportunities I have been provided with.  Finally, the biggest thanks of all go to my family.  I have you to truly thank for who I am today, and will be tomorrow.  Thank you for all the support and being my best friends. My immediate family: Richard Stackhouse, Karl and Debbie Kolind, Chelsea Kleckner and Kris Kolind and your own families!  My extended family: Linda, Jesse and Ron, Sandra, TJ and Kelly, and Farfar and Farmor Kolind and their families.  And of particular importance, Rena and Ron Yeats (always my Nana and Papa).  Nana, you are dearly missed.  xv 3BCO-AUTHORSHIP STATEMENT  ROI ANALYSIS COMPARING MULTI-COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 1.5 T Identification and design of research program I participated in group meetings planning the data acquisition, and identified the need to compare diffusion tensor imaging (DTI) results with myelin water fraction (MWF).  I also proposed separating the lesions based on the presence of a long-T2 component for this analysis.  Performing the research I attended many of the actual data acquisition sessions, though magnetic resonance imaging (MRI) scans were performed by technologists.  Data analyses I performed all diffusion analysis and aided in the T2 relaxation analysis.  I designed and wrote a program to transfer ROIs to the diffusion data.  I performed all statistical tests.  Manuscript preparation I prepared the manuscript and figures and did the original writing.  The co-authors provided guidance and proof-reading for the manuscript.   xvi IMPLEMENTATION AND DEVELOPMENT OF MULTI- COMPONENT T2 RELAXATION IMAGING AT 3.0 T, AND VALIDATION AGAINST 1.5 T MEASUREMENTS Identification and design of research program I volunteered to learn to pulse program the technique on the new 3.0 T MRI scanner and designed the experiments for refinement of scan parameters as well as for comparison between field strengths.  I also devised the method of comparing results (using the standard deviation of fit residuals), based on a parameter proposed by a former group member.  Performing the research Initial pulse programming was conducted with much assistance from co-authors, and completed by myself.  I performed all of the scans.  Data analyses I performed all data analysis.  Manuscript preparation I prepared the manuscript and figures and did the original writing.  The co-authors provided guidance and proof-reading for the manuscript.   xvii VALIDATION OF 3D MULTI-COMPONENT T2 RELAXATION IMAGING AGAINST THE 2D SINGLE-SLICE TECHNIQUE AT 3.0 T Identification and design of research program I decided in cooperation with the co-authors that this comparison should be completed. The 3D sequence was developed by one of the co-authors (Dr Mädler).  I designed the examination protocol for this study with assistance from co-authors.  Performing the research I performed all of the scans.  Data analyses I performed all data analysis.  Manuscript preparation I prepared the manuscript and figures and did the original writing.  The co-authors provided guidance and proof-reading for the manuscript.  VOXEL-WISE (HISTOGRAM) ANALYSIS COMPARING 3D MULTI- COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 3.0 T Identification and design of research program  xviii I participated in group meetings planning the data acquisition, and identified the need to compare DTI results with MWF for a larger section of brain with more consistent brain regions than previously examined.  I performed much of the preliminary work for the study, including reproducibility studies for several of the scans, developing other methods, and test runs of the protocol.   I proposed doing histogram analysis for the data set and wrote the programs to perform this particular analysis.  I also tested several programs for data registration and DTI analysis and determined the optimal analysis procedure for this data, with assistance from 3 of the co-authors.  Performing the research I attended many of the actual data acquisition sessions, though MRI scans were performed by MRI technologists.  Data analyses I had assistance from co-authors in running registration and T2 relaxation analyses.  I performed all DTI analysis, segmentation, and histogram analysis as well as statistical tests.  Manuscript preparation I prepared the manuscript and figures and did the original writing.  The co-authors provided guidance and proof-reading for the manuscript. Chapter 1  1 INTRODUCTION  Magnetic resonance imaging (MRI) has become an indispensable tool for studying demyelinating diseases such as multiple sclerosis (MS).  While conventional MRI techniques are not specific to changes in myelination, several techniques have been developed in recent years that, though not capable of assessing the myelin bilayers explicitly, can be used to study myelin indirectly.  This thesis describes the implementation and validation of one such technique, multi-component T2 relaxation imaging, at high magnetic field (3.0 T), and in 3 dimensions.  It also aims to provide a better understanding of T2 relaxation imaging measurements through comparison with another MRI technique, diffusion tensor imaging (DTI), in MS brain.  The introduction reviews myelin structure and function, as well as how it is affected by MS.  It also provides an overview of T2 relaxation imaging and diffusion tensor imaging.  1.1 BRAIN 1.1.1 White and grey matter Neurons (nerve cells) are the core components of the central and peripheral nervous system (CNS and PNS) and provide a means of transmission and processing of electrical signals within the nervous system.  The neuron is made up of the soma or cell body (which contains the nucleus), the axon (which transmits nerve impulses between 1 neurons), the axon terminals (which pass signals to other neurons), and dendrites (which receive signals), as depicted in Figure 1.1.  In the CNS, the cell bodies of neurons tend to be grouped together; areas made up primarily of neuron cell bodies are called grey matter.  Grey matter also contains glial cells (non-neuronal cells providing support, nutrition and aid in signal transmission) and capillaries.  Areas composed mainly of axons are known as white matter, the white colour being due to the presence of myelin, a lipid-protein lamellar membranous structure which surrounds axons. axon terminals dendrites grey matter white matter soma (cell body) axon myelin  Figure 1.1: Schematic of a neuron showing the soma (cell body), dendrites, myelin, axon, and axon terminals. The axon terminals from one neuron connect to dendrites of other neurons. The inset shows a cross-section of the myelinated axon. Schematic is not to scale.  1.1.2 Myelin 1.1.2.1 Myelin structure CNS myelin is created by the oligodendrocyte, a type of glial cell, the process of which ensheaths the axon and forms the insulating myelin sheath by compaction and fusion of its cell membranes (Martenson 1992; Morell et al., 1989; Raine 1984).  There are gaps in the myelin along an axon, called nodes of Ranvier, and the myelin sheath between nodes 2 of Ranvier is referred to as an internode, with multiple myelin internodes being formed by a single oligodendrocyte.  These internodes are thinner and shorter for axons of smaller diameter (Trapp and Kidd 2004).  At the end of each internode are uncompacted oligodendrocyte processes called paranodal loops which allow ion exchange, while the remaining internode consists of compact myelin which inhibits ion exchange during nerve conduction.  This compact myelin consists of bilayers comprising roughly 80% lipid and 20% protein, which represent the fusion of the cell membranes of the oligodendrocyte process and comprise the major dense line of the myelin sheath. Between the major dense lines there is extracellular space (Kirschner et al., 1984), referred to as the intraperiod line, which is approximately 30 Å wide, contains water, and is the key to myelin water imaging.  Typically, an internode in healthy human CNS white matter will have between 50-100 lipid bilayers.  Figure 1.2 shows an example of a cross section of a myelinated axon.  Figure 1.2: Electron micrograph of a transaction of a myelinated axon. The bar in the bottom right corner represents 0.1 μm x 150000 (Raine 1984). 3 1.1.2.2 Myelin function The primary purpose of myelin is to increase the speed of conduction of electrical signal along the axon.  This signal takes the form of an action potential (also known as a nerve impulse or spike), which is a pulse-like wave of electrical depolarization across the axon cell membrane, mediated predominately by the movement of sodium ions into the axon via specific channels located in the node of Ranvier.  A certain critical voltage threshold must be exceeded for the spike to occur; once initiated, the action potential propagates along the axonal membrane.  The nodes of Ranvier allow saltatory conduction along the nerve fibre.  Essentially, the electrical potential difference between a depolarized node of Ranvier and the next downstream node of Ranvier causes the depolarization wave to jump to the latter.  The high electrical resistance and low capacitance of the compact myelin sheath inhibit charge leakage, allowing this saltatory conduction of the action potential to occur.  Saltatory conduction speeds up the transmission by 1 – 2 orders of magnitude above continuous propagation along an unmyelinated axon, and represents a major evolutionary advance; it is one of the factors underlying the highly complex integrative functions of the human brain.  1.2 MULTIPLE SCLEROSIS The name multiple sclerosis, meaning “many scars”, describes the multiple areas of discolouration and shrinkage throughout the brain and spinal cord (lesions), with the definitive clinical description published in the early 19th century (Charcot 1868).  MS is a demyelinating disease, and is also an inflammatory disorder.  While the exact mechanism of MS is controversial, it is widely believed that MS is a disorder with an immunologic 4 pathogenesis directed against myelin.  It is generally accepted that the relationship between lesions and parenchymal blood vessels is close, and that there is an increase in the permeability of the blood-brain barrier.  The course of the disease varies greatly between individuals, but has 4 phenotypes which are pictorially displayed in Figure 1.3 and are outlined below (Lublin and Reingold 1996): • Relapsing remitting: 85-90% of cases begin with a relapsing remitting disease course, which consists of isolated clinical attacks during which neurological function deteriorates, interspersed with periods with no disease progression. Attacks may be followed by full or partial recovery. • Primary progressive: This disease course involves gradual, continual deterioration of neurological function, and affects roughly 10% of people with MS.  While there are no distinct remissions, plateaus and small recoveries do sometimes occur. • Secondary progressive: Relapsing remitting patients often progress into this disease course, during which recovery is reduced or non-existent between attacks and the condition progressively worsens. • Progressive relapsing: 5-6% of patients are affected by this type of MS, characterized by steady progression of the disease between remission periods. 5 D is ea se  P ro gr es sio n Time Relapsing remitting Primary progressive Secondary progressive Progressive relapsing D is ea se  P ro gr es sio n  Figure 1.3: Illustration of the disease progression as a function of time of the phenotypes of MS, adapted from (Lublin and Reingold 1996).  1.2.1 MS pathology Underlying pathologic processes occurring in MS include chronic inflammation, edema, demyelination, gliosis, oligodendrocyte loss, and axonal degeneration (Lassmann 2004). Chronic inflammation is characterized by infiltration of lymphocytes, plasma cells, macrophages and monocytes.  Macrophages engulf and digest cellular debris and pathogens.  In MS, macrophages attack and engulf myelin sheaths, resulting in a loss of myelin from the axon, a process referred to as demyelination.  The chronic inflammatory 6 infiltrates, which are particularly prominent in blood vessel walls, are thought to be responsible for damage to the endothelial lining of the blood vessel, resulting in blood- brain barrier breakdown.  The ensuing extravasation of fluid into the extracellular space in the CNS produces edema, defined as swelling of the tissue.  Gliosis is a response to the tissue destruction and is characterized by enlargement and proliferation of astrocytes (glial cells with numerous intracellular filaments), which produce a firm or “sclerotic” scar.  The loss of oligodendrocytes, the glial cells that maintain myelin internodes, may be a result of the demyelination, but there is probably a primary attack on oligodendrocytes as well.  Axonal degeneration, which can consist of atrophy (shrinkage) or complete loss of the axon, is thought to be a side effect of the inflammatory infiltrates in the MS lesion and can be manifest away from the point of injury as Wallerian degeneration.  1.2.1.1 Lesions The term “lesion” is a general pathologic term referring to a focal area of abnormal tissue.  In the case of MS, the lesions of focal demyelination with the attendant pathologic features noted above are known as plaques.  Plaques, which are generally centered on blood vessels, can be classified as acute (which are temporally acute and contain an abundance of inflammatory cells as well as large astrocytes with on-going active demyelination throughout their extent), chronic active (with active demyelination confined to the expanding edge and less inflammation in the centre of the lesion, which now shows advanced gliosis, the astrocytes becoming longer and thinner), or chronic silent (with little active demyelination or inflammation and intense fibrillary gliosis). 7 Lesions can remyelinate; remyelinated axons have thinner myelin and shortened (but probably functional) internodes.  1.2.1.2 Normal appearing white matter (NAWM) Lesions are not the only regions in MS brain that show abnormalities compared to healthy brain.  It has been shown that white matter which appears normal on conventional MRI can contain many of the same pathological features of lesions including inflammation, demyelination, proliferation of astrocytes and axonal loss (Allen et al., 2001; Evangelou et al., 2000a; Evangelou et al., 2000b; Moore 1998).  1.3 MAGNETIC RESONANCE IMAGING MRI is a very powerful tool for studying CNS tissue (among many other applications!). A large external magnetic field (B0) aligns protons along the direction of the magnetic field, either parallel or anti-parallel, which have slightly different energies.  A net magnetization results from a small excess of protons in the lower energy state.  When tilted away from B0 (by application of an additional nonparallel magnetic field), the protons, or spins, precess around the B0 vector at the Larmor frequency,  00 Bγω =  (1.1) where γ is the gyromagnetic ratio.  Smaller magnetic field gradients are applied along various axes to spatially encode the protons by changing the magnetic field and therefore the precession frequency.  If an orthogonal oscillating magnetic field with magnitude B1 8 is applied at the Larmor frequency, the pulse’s energy will be absorbed by protons in the sample and the net magnetization will tip by an angle  ptB1γθ =  (1.2) where tp is the length of time that the oscillating field (which is usually in the radio frequency (RF) range) is applied.  An RF coil (either the same coil as used for excitation or a separate coil) detects signal from the sample, which is reconstructed into an image using the spatial encoding information.  The signal intensity for each location, or voxel, in the image depends on the density of protons (ρ), and the relaxation times T1 and T2. T1, called the longitudinal relaxation time or spin-lattice relaxation time, describes the recovery of the magnetization vector in the B0 direction, and involves an exchange of energy between the protons and surrounding tissue/material.  T2, called the transverse relaxation time or spin-spin relaxation time, characterizes the decay of the net signal in the transverse plane due to interactions between protons causing irreversible loss of phase coherence.  The magnetization has been described by Felix Bloch as a function of time (Bloch 1946), and if B0 is assumed to be along the z-axis and the entire initial magnetization M0 is aligned in the transverse plane (after a 90º pulse), then   (1.3) tiTt xy eeMtM 02 / 0)( ω−=   (1.4) ).1()( 1/ 0 Tt z eMtM −−=   If a reference frame that is rotating at the Larmor frequency is used, then the precession frequency can be demodulated out of the transverse magnetization equation to give   (1.5) .)( 2/ 0 Tt xy eMtM −= 9 The measured signal intensity is a function of the magnitude of Mxy, after digitization and filtering.  For the remainder of this thesis, the rotating reference frame (rotating at the Larmor frequency about the z-axis) will be assumed.  1.3.1 Measuring T2 in brain The signal from protons attached to macromolecules and membrane lipids decays too quickly to contribute directly to MRI measurements.  However, T2 times for brain water protons are influenced by interactions between water and tissue components such as cell membranes and cytoplasmic proteins, and are detectable by careful MR measurement. Though T2 varies throughout the brain, it is generally on the order of 100 times shorter than pure water T2.   Because CNS tissue is a complex arrangement of microscopic cellular structures, a single MRI voxel with a typical volume on the order of 1-10 mm3 contains water molecules in several different environments, thus T2 relaxation for an MRI voxel of brain tissue typically cannot be characterized by a single exponential decay.  In addition, the transverse magnetization decays faster than described by T2 due to magnetic field inhomogeneities (which are unvarying in time) within the sample, so that spins dephase and the net magnetization in the xy plane is reduced.  This decay (termed T2' decay) is reversible as described below, while signal loss due to T2 is irreversible as it relates to thermal agitation of the spins (which varies in space and time).  10 In healthy human brain, T2 relaxation imaging has revealed that the water signal can be separated into components from cerebrospinal fluid (CSF) (the longest T2 compartment, T2 > 2 s), intra and extra-cellular (IE) water (with intermediate T2 times, T2 ~ 100 ms) and the water trapped between the myelin bilayers, termed myelin water (the shortest T2 component, T2 ~ 20 ms), as shown in Figure 1.4 (MacKay et al., 1994; Menon and Allen 1991; Vasilescu et al., 1978; Whittall et al., 1997).  The size and T2 time of the IE peak can be used to probe the intra and extra-cellular water environments (Stanisz et al., 2005; Stewart et al., 1993; Whittall et al., 2002). The ratio of the myelin water T2 component to the total water signal is called the myelin water fraction (MWF), and there is strong evidence that MWF can be used as a marker for myelin (Laule et al., 2008; Laule et al., 2006; Webb et al., 2003). 10 100 1000 T2 (ms) A m pl itu de Intra/extracellular water 40-200ms Myelin water 0-40ms Myelin water fraction (MWF) + = + Cerebrospinal fluid >1000ms A m pl itu de  Figure 1.4: A typical T2 distribution for healthy brain showing the different water components. The area under each peak is proportional to the number of protons in that environment.  11 1.3.1.1 T2 decay curve acquisition An MRI pulse sequence is described by magnetic field gradients along 3 orthogonal axes (Gx, Gy and Gz), the RF pulse train, and the signal acquisition (echoes).  T2 decay curve data can be acquired with a multi-echo spin-echo pulse sequence, such as a Carr-Purcell- Meiboom-Gill (CPMG) sequence (Carr and Purcell 1954; Meiboom and Gill 1958) shown in Figure 1.5, consisting of a 90º excitation pulse followed by a train of equally spaced 180º refocusing pulses.  The 90º pulse rotates the magnetization onto the y-axis and the spins dephase due to field inhomogeneities for a time τ at which point the first 180º pulse is applied about the y-axis.  The spins change phase in the opposite manner to before the 180º pulse, so after a time τ the spins have rephased and form what is known as a spin echo. z x yy z x z x y z x y z x y t=0 t=τ- t=τ t=τ+ t=2τ 90º x 180º y  Figure 1.5: Phase diagram for a CPMG pulse sequence. If a 90º pulse occurs about the x-axis at time 0, and a 180º pulse occurs about the y-axis at time τ, an echo will occur at time 2τ.  This process is repeated for a series of refocusing pulses with spacing 2τ, and the signal from each echo decays with a time constant of T2, as shown in Figure 1.6. 12 2τ 90º 180º 180º 180º T2 decay 2τ 2τ  Figure 1.6: After a 90º pulse, a series of 180º pulses with spacing 2τ results in echoes whose amplitudes decay with time constant T2.  Challenges for collecting and analyzing T2 decay curves include random noise, stimulated echoes, out-of slice signal, eddy currents, magnetization transfer, flow, patient motion, and magnetic field (both B0 and B1) inhomogeneities.  It has been reported that a high SNR (at least 100:1 at the shortest echo time) is required for accurate extraction of the multi-exponential components of the T2 relaxation, hence random noise must be minimized (Graham et al., 1996).  Stimulated echoes are spurious contributions to the signal, and occur when the magnetization is not perfectly refocused.  Imperfect RF pulses lead to part of the magnetization rephasing as expected, another part continuing to dephase, and a portion stored along the z-axis and hence not dephasing at all (Hennig 1988; Woessner 1961). Subsequent imperfect refocusing pulses will superimpose these stimulated echoes on the primary echoes.  13 Out-of-slice signal contamination occurs because the slice profile of the RF refocusing pulse is not perfect, and this contamination (or extra signal) is overlayed on top of signal from the prescribed slice.  As it won’t experience all of the same magnetic fields as the desired slice of data, the decay curves can be affected.  An eddy current is a circulating flow of electrons which occurs in conductors exposed to changing magnetic fields.  According to Lenz’s law (Lenz 1834), the changing field creates an electromagnet with magnetic field opposing the change in magnetic field. Spatial localization and spurious signal crushing in MRI are accomplished using changes in magnetic field gradients, and hence cause eddy currents in the structure of the magnet which can distort the MR images.  Most scanners employ eddy current compensation but it is optimized for default gradient slew rates, so changing the slew rate may result in a considerable increase in eddy current artifact.  The magnetic field gradient applied along the z-axis for slice selection can cause off- resonance RF pulses for other slices, and it has been shown that off-resonance pulses (or magnetization transfer (MT) pulses) preferentially reduce the signal from water trapped between myelin bilayers due to direct saturation (Vavasour et al., 2000).  The MT effect has made simple multi-slice T2 relaxation imaging unfeasible.  CSF contains protons spins which may be flowing quickly out of the excitation slice and hence artifactually show a short T2.  General patient motion (head movements or even breathing) can have the same effect, and in very long scans, patient motion is a large 14 concern.  Motion causes blurring and a pattern of ghosts along the phase-encoding direction.  If water diffuses through a magnetic susceptibility gradient in the time between 180º refocusing pulses, it will not be completely refocused.  Also, the presence of paramagnetic or supermagnetic (e.g. iron) particles can alter the magnetic field on a microscopic scale that may not be completely compensated by a CPMG pulse sequence (Stanisz et al., 2005).  Jensen et al. (Jensen et al., 2001) found that a multi-echo spin-echo pulse sequence is dependent on echo spacing due to brain iron, with iron significantly shortening observed T2 times for longer echo spacings, an effect that is expected to become more pronounced at higher magnetic field.  The gold standard approach used to collect quantitative T2 relaxation data, developed by Poon and Henkelman (Poon and Henkelman 1992), is based on a single-slice CPMG pulse sequence with slice-selection along the z-axis, frequency encoding along the x-axis and phase encoding along the y-axis (Figure 1.7).  It should be noted that if a phase- encoding gradient is applied before the refocusing pulses, the phase relationship between the net magnetization and the refocusing pulses required by the CPMG scheme is not met, thus it is not a CPMG sequence.  For correct T2 value measurements, accurate 180º refocusing pulses are critical to avoid stimulated echoes.  For this purpose, composite block refocusing pulses (90ox-180oy-90ox) are used to achieve the most accurate 180º rotations possible, as they have been shown to be robust in the presence of inhomogeneous B1 and B0 fields (Levitt and Freeman 1981; Poon and Henkelman 1992). 15 In addition, balanced crusher gradients are applied along the z-axis on either side of each refocusing pulse.  As the gradient on either side of the refocusing pulse is identical in shape and size, protons that have undergone a 180º rotation will be rephased, but protons not experiencing a 180º rotation will be dephased.  It was shown that a pattern of crusher gradients that alternate in polarity and descend in amplitude along the echo train is the most effective for reduction of stimulated echoes and out-of-slice signal (Poon and Henkelman 1992). Signal GX GY GZ RF 90º 180º 180º 180º  Figure 1.7: First portion of a Poon-Henkelman T2 relaxation pulse sequence.  16 Typical in-vivo sequence parameters for T2 relaxation imaging using the Poon- Henkelman approach include a long repetition time (TR) to minimize T1 weighting (TR 2500-3000 ms), echo spacing of 10 ms, 5 mm slice thickness, 220 mm field of view (FOV), a matrix of 256x128, and 4 averages.  1.3.1.2 T2 decay curve analysis While there are several techniques available for fitting multi-exponential relaxation decay (Fenrich et al., 2001; Graham et al., 1996; Henkelman 1985; Kroeker and Henkelman 1986; Provencher 1982; Whittall and MacKay 1989), the Non Negative Least Squares Algorithm (NNLS) (Lawson and Hanson 1974) is often used in the field of CNS and PNS T2 relaxation (Beaulieu et al., 1998; Gareau et al., 2000; Laule et al., 2004; MacKay et al., 1994; Menon et al., 1992; Oh et al., 2006; Tozer et al., 2005; Webb et al., 2003; Whittall et al., 1997) because convergence is guaranteed, and it does not require initial guesses of the solution or a priori information regarding the number of contributing T2 components.  The NNLS algorithm determines the set of amplitudes sj minimizing the least squares misfit given the decay curve data and the T2 partition.  The echo decay curve intensities can be written as:   (1.6) ,,...2,1),/exp( 1 2∑ = =−= M j jiji NiTtsy where ti are the echo times used, sj are the unknown relative amplitudes for each partitioned relaxation time T2j , N is the number of echoes acquired, and M is the number of elements in the pre-specified T2 partition.  Typically for in vivo brain, N is between 32 17 and 48 and M is on the order of 100 relaxation times equally spaced on a logarithmic scale from 15 ms to 2 s.  The NNLS routine minimizes both χ2 and an energy constraint, or regularizer, which provides more robust fits in the presence of noise.  While there are many choices of regularizer, the sum of the squares of the solution amplitudes (called the small model) (Whittall and MacKay 1989) is often used, for which the expression to be minimized is:  .  (1.7) 0, 1 22 ≥+ ∑ = μμχ M j js The larger the value of μ, the more the results will satisfy the constraints of the small model while increasing the χ2 misfit.  Generally, we wish to adjust μ until the χ2 misfit approaches a value of N (the number of data points), which is the value expected for χ2. If χ2 is much smaller than N, the data are fit too closely, and artifact will be apparent in the T2 distribution, while if χ2 is much greater N, the data are not fit closely enough and information will be lost.  The relationship between χ2 and the fit depends on having an accurate value for the variance, which is difficult to obtain.  Thus, regularized smooth T2 distributions are created by minimizing the equation 1.7 with a typical energy constraint of 1.02χ2min ≤ χ2 ≤ 1.025χ2min. Essentially, the amplitudes sj are found using the NNLS routine for μ = 0 to give χ2min, and then solved again for a range of μ values (typically logarithmically spaced from 10-5 to 10-1), and the new χ2 (expression (1.7)) is also calculated for each μ value.  The value of μ resulting in a χ2 value at the midpoint between the constraints is selected (Jones 2003).   With only a finite number of measurements, there is an inherent non-uniqueness to the fit; a range of distributions can 18 fit the data equally well.  For regularized solutions, the parameter μ picks out a single solution from the infinity otherwise available for underdetermined problems.  In this case, this solution has the least energy for the given misfit (Lawson and Hanson 1974; Whittall and MacKay 1989).  Component fractions are calculated as the fraction of the total sj that falls within the ranges expected for each water compartment:  ∑∑ == = M j j M Mj j ssFractionComponent 1 max min  (1.8) where Mmin and Mmax define the range within the T2 distribution assigned to the water compartment in question.  In this thesis, water compartments examined were myelin water (T2 range of 15 ms-40 ms), IE water (T2 range of 40-200 ms) and long-T2 water (T2 range of 200-800 ms).  The geometric mean T2 ( 2T ), analogous to the amplitude-weighted mean on a logarithmic scale (Whittall et al., 1997), can be calculated as follows:  ⎥⎦ ⎤⎢⎣ ⎡= ∑∑ == max min max min )log(exp 22 M Mj j M Mj jj sTsT  (1.9) and the peak width (analogous to the variance but on a logarithmic scale, and providing information regarding the homogeneity of the T2 times of the water compartment), is given by (Whittall et al., 1997):  .1logexp max min max min 2 2 2 −⎥⎥⎦ ⎤ ⎢⎢⎣ ⎡ ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛ ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛= ∑∑ == M Mj j M Mj j j sT T sWidthPeak  (1.10) 19  1.3.2 Measuring the diffusion tensor in brain Diffusion of water molecules in the brain can be studied to gain information regarding the environment of the water.  Diffusion of water in a homogeneous medium follows a random walk pattern.  The Einstein relation gives the distance traveled by a water molecule:   (1.11) nDTR 2 2 = where R2 is the mean square distance traveled in time T, n is the number of dimensions, and D is the translational diffusion coefficient.  If R2 is plotted against T, the slope will depend directly on D.  In an anisotropic medium, the slopes along different axes will differ.  During the diffusion time of a typical MRI diffusion experiment (10-100 ms), most water molecules in brain travel between 5 and 20 μm, which is larger than the dimension of most cellular organelles or other microbiological obstacles, thus diffusion-weighted images allow (indirectly) the study of barriers causing restriction such as myelin sheaths, axonal membranes, etc.  In grey matter, D is roughly 2.5 times smaller than in pure water due to obstructions such as intracellular organelles and macromolecules (Le Bihan and Basser 1995; Le Bihan et al., 1995).  Although the diffusion in grey matter is restricted or hindered, it still has little anisotropy compared to white matter, where preferred water directions have long been known to exist (Moseley et al., 1990; Moseley et al., 1991).  Diffusion anisotropy in the 20 CNS is thought to be caused by the water molecules preferentially diffusing along the length of the axon because diffusion perpendicular to the axon is hindered by barriers such as the myelin sheath (Moseley et al., 1990; Thomsen et al., 1987).  While it has been shown that myelin contributes to diffusion anisotropy, it is by no means the sole cause of anisotropic water diffusion in brain (Beaulieu and Allen 1994a; Beaulieu and Allen 1994b; Gulani et al., 2001; Prayer et al., 1997; Wimberger et al., 1995).  1.3.2.1 Diffusion MRI acquisition Diffusion weighted imaging (DWI) can be accomplished using the Stejskal-Tanner pulsed field gradient diffusion experiment (Stejskal 1965; Stejskal and Tanner 1965), which uses the pulse sequence illustrated in Figure 1.8.  Figure 1.8: Stejskal-Tanner pulsed field gradient diffusion MRI pulse sequence.  The diffusion gradients (yellow) can be applied in various directions; in this case they are applied along the y- (phase-encoding) axis.  δ is the duration of the diffusion gradient, Δ is the time between the beginnings of each diffusion gradient, and g is the diffusion gradient amplitude.  21 The probability of a molecule moving a given distance in time T is determined by the propagator:  ),4/exp()4(),( 22 3 DTDTTP RR −= −π  (1.12) where R is the relative spin displacement.  Based on the propagator, the following equation can be derived to give (roughly) the resulting magnetization M from the Stejskal-Tanner experiment:  )  (1.13) exp(0 bDMM −= where M0 is the magnetization obtained for diffusion gradient strength g = 0, and b is a constant that depends on the properties of the phase encoding gradient (δ, Δ, and g in Figure 1.8):   (1.14) )3/( 222 δδγ −Δ= gb where δ is the duration of the diffusion gradient and Δ is the time between the beginnings of each diffusion gradient.  From equations 1.12 through 1.14 there are only two unknowns, M0 and D.  By taking two measurements of the signal with different b-values, D can be determined using standard linear regression once the natural logarithm of the spin echo signal has been taken.  DWI is limited to describing isotropic Gaussian diffusion, therefore diffusion tensor imaging (DTI) is needed to study anisotropic diffusion as found in regularly ordered microstructures.  For anisotropic diffusion, D can be represented as a second order tensor: 22   (1.15) . ⎥⎥ ⎥ ⎦ ⎤ ⎢⎢ ⎢ ⎣ ⎡ = zzzyzx yzyyyx xzxyxx DDD DDD DDD D In the principle axis system, D can be diagonalized to   (1.16) . 00 00 00 3 2 1 ⎥⎥ ⎥ ⎦ ⎤ ⎢⎢ ⎢ ⎣ ⎡ λ λ λ Since the diffusion tensor has 6 independent elements, at least 6 non-collinear gradient orientations must be used.  Typical acquisitions acquire on the order of 10 gradient orientations in addition to a b = 0 (non-diffusion weighted) image, and b-values are usually chosen to be between 800 and 1000 s/mm2.  1.3.2.2 Diffusion analysis For an anisotropic sample, the system of equations that can be used to solve for the diffusion tensor is given by: CDA =  (1.17) where [ Tyzxzxyzzyyxx DDDDDD=D ]  (1.18) and for k diffusion directions, A is a k by 6 matrix (where g represents the diffusion sensitizing gradient vector): ⎥⎥ ⎥⎥ ⎥ ⎦ ⎤ ⎢⎢ ⎢⎢ ⎢ ⎣ ⎡ = kzkykzkxkykxkzkykx zyzxyxzyx zyzxyxzyx ggggggggg ggggggggg ggggggggg 222 222 222 || 1 222 222222 2 2 2 2 2 2 111111 2 1 2 1 2 1 MMMMMMgA  (1.19) 23 and T k k b MM b MM b MM ⎥⎦ ⎤⎢⎣ ⎡= )/ln()/ln()/ln( 0 2 02 1 01 LC  (1.20) where Mk is the signal resulting from applying diffusion gradient gk.  If k is greater than 6, then the system of equations is over-determined.  In that case, a solution is sought that best fits all of the equations, usually by minimizing the least-squares difference between the fit and the data (Jiang et al., 2006).  Then the solution is found by multiplying the C vector by the pseudo-inverse of A, which is often computed using singular value decomposition (Press et al., 1992).  The mean diffusivity <D> is given by the average of the three eigenvalues:  3 321 λλλ ++=D  (1.21) and a commonly used measure of anisotropy which is orientationally independent is the fractional anisotropy, FA, calculated as:  2 3 2 2 2 1 2 13 2 32 2 21 )()()( 2 1 λλλ λλλλλλ ++ −+−+−=FA  (1.22) which by definition is a number between 0 and 1.  Since changes in FA can be an ambiguous measure (for instance, a decrease in FA could be caused by either an increase in λ1 or a decrease in λ2 or λ3), the diffusion tensor eigenvalues themselves are often quoted.  The eigenvalues can be summarized as axial (or parallel) diffusivity (λ|| = λ1,the largest eigenvalue, along the primary diffusion direction) and radial (or perpendicular) diffusivity (λ |  = (λ2 + λ3)/2, the average of the two smaller eigenvalues, perpendicular to the dominant direction of diffusion). 24  1.4 OVERVIEW OF THESIS This thesis focuses on multi-component T2 relaxation of in vivo brain, both implementing and validating the technique at higher magnetic field and extending it to 3 dimensions, as well as understanding results in multiple sclerosis brain.  First, MS pathology was studied using both single-slice multi-echo T2 relaxation imaging and DTI at 1.5 T. Hypotheses: • T2 relaxation imaging metrics such as MWF will relate to DTI metrics such as FA as both have been linked to myelin content. • T2 relaxation imaging metrics will provide additional information regarding MS pathology in vivo. Specific Aims: • Compare MWF and long-T2 fraction to FA, <D>, λ|| and λ |  in MS NAWM and lesion using region of interest (ROI) analysis. • Study lesions with long-T2 components to see how they differ from other MS lesions.  Second, multi-component T2 relaxation imaging was implemented at 3.0 T and sequence parameters were adjusted to minimize artifact in phantoms, fixed-brain, and healthy volunteers.  Final in vivo results were compared with results obtained at 1.5 T. Hypotheses: 25 • Implementation of the Poon-Henkelman T2 relaxation technique will be feasible at 3.0 T. • The higher field strength and reduced B1 homogeneity will result in more artifact. Specific aims: • Understand sources of artifact in T2 relaxation images at 3.0 T and minimize them where possible. • Validate results by showing consistency with results at 1.5 T.  Third, a 3-dimensional (3D) version of the T2 relaxation imaging sequence was tested and validated at 3.0 T by scanning 10 normal volunteers with the gold-standard single-slice technique and the new 3D technique. Hypotheses: • 3D MWF measurements will be consistent with single-slice measurements with explainable differences. • Multi-component T2 relaxation imaging will be more efficient using the 3D technique, with insignificant increases in artifact and noise. Specific aims: • Investigate any differences in MWF, geometric mean T2 of the IE peak, and multi-exponential fit, between the single-slice and 3D T2 relaxation techniques. • Validate 3D T2 relaxation results by showing consistency with single-slice T2 relaxation results.  26 Finally, MS subject and healthy control brain were studied using both 3D multi- component T2 relaxation imaging and DTI at 3.0 T. Hypotheses: • The increase in brain coverage will make histogram analysis of T2 relaxation metrics such as MWF and geometric mean T2 possible, enabling the assessment of more subtle abnormalities of MS brain tissue that can be missed using ROI analysis. • Relationships between MWF and diffusion anisotropy will be stronger due to a larger variation in MWF and anisotropy values in the greater range of tissue sampled. • Level of disability and disease duration will correlate with histogram metrics. Specific aims: • Compare MWF and geometric mean T2 of the IE peak to FA, <D>, λ||  and λ |  in MS NAWM and lesion as well as control normal white matter using histogram analysis. • Investigate differences in DTI and T2 relaxation imaging histograms for various states of disease.  The concluding chapter discusses the overall findings of the manuscript chapters, including the significance of the thesis research, strengths and weaknesses of each study, potential applications and future directions.  27 1.5 REFERENCES Allen I.V., McQuaid S., Mirakhur M., Nevin G., 2001. Pathological abnormalities in the normal-appearing white matter in multiple sclerosis. Neurological Sciences 22, 141-4. Beaulieu C., Allen P.S., 1994a. Determinants of anisotropic water diffusion in nerves. Magn Reson Med 31, 394-400. Beaulieu C., Allen P.S., 1994b. Water diffusion in the giant axon of the squid: implications for diffusion-weighted MRI of the nervous system. Magn Reson Med 32, 579-83. Beaulieu C., Fenrich F.R., Allen P.S., 1998. Multicomponent water proton transverse relaxation and T2-discriminated water diffusion in myelinated and nonmyelinated nerve. Magn Reson Imaging 16, 1201-10. Bloch F., 1946. Nuclear Induction. Physical Review 70, 460-474. Carr H.Y., Purcell E.M., 1954. Effects of Diffusion on Free Precession in Nuclear Magnetic Resonance Experiments. Physical Review 94, 630-638. Charcot J.M., 1868. Histologie de la sclérose en plaques. Gaz Hopit Civils Milit 41, 554,557,566. Evangelou N., Esiri M.M., Smith S., Palace J., Matthews P.M., 2000a. Quantitative pathological evidence for axonal loss in normal appearing white matter in multiple sclerosis. Annals of Neurology. 47, 391-5. Evangelou N., Konz D., Esiri M.M., Smith S., Palace J., Matthews P.M., 2000b. Regional axonal loss in the corpus callosum correlates with cerebral white matter lesion volume and distribution in multiple sclerosis. Brain. 123, 1845-9. Fenrich F.R., Beaulieu C., Allen P.S., 2001. Relaxation times and microstructures. NMR Biomed. 14, 133-9. Gareau P.J., Rutt B.K., Karlik S.J., Mitchell J.R., 2000. Magnetization transfer and multicomponent T2 relaxation measurements with histopathologic correlation in an experimental model of MS. Journal of Magnetic Resonance Imaging 11, 586- 95. Graham S.J., Stanchev P.L., Bronskill M.J., 1996. Criteria for analysis of multicomponent tissue T2 relaxation data. Magnetic Resonance in Medicine 35, 370-8. Gulani V., Webb A.G., Duncan I.D., Lauterbur P.C., 2001. Apparent diffusion tensor measurements in myelin-deficient rat spinal cords. Magnetic Resonance in Medicine 45, 191-195. Henkelman R.M., 1985. Measurement of signal intensities in the presence of noise in MR images. Med Phys 12, 232-3. Hennig J., 1988. Multiecho imaging sequences with low refocusing flip angles. J Magn Reson 78, 397-407. Jensen J.H., Chandra R., Yu H., 2001. Quantitative model for the interecho time dependence of the CPMG relaxation rate in iron-rich gray matter. Magn Reson Med 46, 159-65. Jiang H., van Zijl P.C., Kim J., Pearlson G.D., Mori S., 2006. DtiStudio: resource program for diffusion tensor computation and fiber bundle tracking. Comput Methods Programs Biomed 81, 106-16. 28 Jones C.K., 2003. T2 Decay Curve Acquisition and Anaysis in MRI; Noise Considerations, Short T2, and B1 Field Encoding. Vancouver: University of British Columbia. 162 p. Kirschner D.A., Ganser A.L., Caspar D.L.D., 1984. Diffraction studies of molecular organization and membrane interactions in myelin. In: Morell P., editor. Myelin. 2nd Edition ed. New York: Plenum Press. Kroeker R.M., Henkelman R.M., 1986. Analysis of biological NMR relaxation data with continuous distributions of relaxation times. Journal of Magnetic Resonance 69, 218-235. Lassmann H., 2004. Cellular Damage and Repair in Multiple Sclerosis. In: R L., J G., H L., K N., R M., BD T., editors. Myelin Biology and Disorders 2. London: Elsevier Academic Press. Laule C., Kozlowski P., Leung E., Li D.K.B., MacKay A.L., Moore G.R.W., 2008. Myelin water imaging of multiple sclerosis at 7 T: Correlations with histopathology. Neuroimage 40, 1575-1580. Laule C., Leung E., Li D.K., Traboulsee A.L., Paty D.W., MacKay A.L., Moore G.R., 2006. Myelin water imaging in multiple sclerosis: quantitative correlations with histopathology. Mult Scler. 12, 747-53. Laule C., Vavasour I.M., Moore G.R.W., Oger J., Li D.K.B., Paty D.W., MacKay A.L., 2004. Water content and myelin water fraction in multiple sclerosis: A T2 relaxation study. Journal of Neurology 251, 284-293. Lawson C.L., Hanson R.J., 1974. Solving Least Squares Problems. Englewood Cliffs, NJ: Prentice-Hall,. Le Bihan D., Basser P.J., 1995. Molecular diffusion and nuclear magnetic resonance. In: Le Bihan D., editor. Diffusion and perfusion magnetic resonance imaging. New York: Raven Press. p 5-17. Le Bihan D., Turner R., N. P., 1995. Diffusion MR imaging in normal brain and in brain tumors. In: Le Bihan D., editor. Diffusion and perfusion magnetic resonance imaging. New York: Raven Press. p 134-40. Lenz E., 1834. Ueber die Bestimmung der Richtung der durch elektrodynamische Vertheilung erregten galvanischen Ströme. Annalen der Physik 107, 483-494. Levitt M.H., Freeman R., 1981. Compensation for pulse imperfections in NMR spin-echo experiments. J. Magn. Reson 43, 65-80. Lublin F.D., Reingold S.C., 1996. Defining the clinical course of multiple sclerosis: results of an international survey. National Multiple Sclerosis Society (USA) Advisory Committee on Clinical Trials of New Agents in Multiple Sclerosis. Neurology 46, 907-11. MacKay A., Whittall K., Adler J., Li D., Paty D., Graeb D., 1994. In vivo visualization of myelin water in brain by magnetic resonance. Magnetic Resonance in Medicine 31, 673-7. Martenson R.E., editor. 1992. Myelin: biology and chemistry. Boca raton: CRC Press. Meiboom S., Gill D., 1958. Modified spin-echo method for measuring nuclear relaxation times, Re V. Sci. Instrum 29, 688–91. Menon R.S., Allen P.S., 1991. Application of continuous relaxation time distributions to the fitting of data from model systems and excised tissue. Magnetic Resonance in Medicine. 20, 214-27. 29 Menon R.S., Rusinko M.S., Allen P.S., 1992. Proton relaxation studies of water compartmentalization in a model neurological system. Magn Reson Med. 28, 264- 74. Moore G.R.W., 1998. Neuropathology and Pathophysiology of the Multiple Sclerosis Lesion. In: Paty D., Ebers G.C., editors. Multiple Sclerosis. Philadelphia: F.A. Davis. Morell P., Quarles R.H., Norton W., 1989. Formation, Structure, and Biochemistry of Myelin. In: Siegel J., editor. Basic Neurochemistry: Molecular, Cellular, and Medical Aspects. 4th ed. New York: Raven Press Ltd. p 109-136. Moseley M.E., Cohen Y., Kucharczyk J., Mintorovitch J., Asgari H.S., Wendland M.F., Tsuruda J., Norman D., 1990. Diffusion-weighted MR imaging of anisotropic water diffusion in cat central nervous system. Radiology 176, 439-45. Moseley M.E., Kucharczyk J., Asgari H.S., Norman D., 1991. Anisotropy in diffusion- weighted MRI. Magnetic Resonance in Medicine. 19, 321-6. Oh J., Han E.T., Pelletier D., Nelson S.J., 2006. Measurement of in vivo multi- component T2 relaxation times for brain tissue using multi-slice T2 prep at 1.5 and 3 T. Magn Reson Imaging 24, 33-43. Poon C.S., Henkelman R.M., 1992. Practical T2 quantitation for clinical applications. Journal of Magnetic Resonance Imaging. 2, 541-53. Prayer D., Roberts T., Barkovich A.J., Prayer L., Kucharczyk J., Moseley M., Arieff A., 1997. Diffusion-weighted MRI of myelination in the rat brain following treatment with gonadal hormones. Neuroradiology 39, 320-5. Press W.H., Teukolsky S.A., Vetterling W.T., Flannery B.P., 1992. Numerical Recipes in C: The Art of Scientific Computing. . Cambridge University Press, Cambridge. Provencher S.W., 1982. A constrained regularization method for inverting data represented by linear algebraic or integral equations. Raine C.S., 1984. Morphology of myelin and myelination. In: Morell P., editor. Myelin. 2nd Edition ed. New York: Plenum Press. Stanisz G.J., Odrobina E.E., Pun J., Escaravage M., Graham S.J., Bronskill M.J., Henkelman R.M., 2005. T1, T2 relaxation and magnetization transfer in tissue at 3T. Magn Reson Med 54, 507-12. Stejskal E.O., 1965. Use of Spin Echoes in a Pulsed Magnetic-Field Gradient to Study Anisotropic, Restricted Diffusion and Flow. The Journal of Chemical Physics 43, 3597. Stejskal E.O., Tanner J.E., 1965. Spin Diffusion Measurements: Spin Echoes in the Presence of a Time-Dependent Field Gradient. The Journal of Chemical Physics 42, 288. Stewart W.A., MacKay A.L., Whittall K.P., Moore G.R., Paty D.W., 1993. Spin-spin relaxation in experimental allergic encephalomyelitis. Analysis of CPMG data using a non-linear least squares method and linear inverse theory. Magnetic Resonance in Medicine. 29, 767-75. Thomsen C., Henriksen O., Ring P., 1987. In vivo measurement of water self diffusion in the human brain by magnetic resonance imaging. Acta Radiol 28, 353-61. Tozer D.J., Davies G.R., Altmann D.R., Miller D.H., Tofts P.S., 2005. Correlation of apparent myelin measures obtained in multiple sclerosis patients and controls 30 31 from magnetization transfer and multicompartmental T2 analysis. Magn Reson Med. 53, 1415-22. Trapp B., Kidd G., 2004. Structure of the Myelinated Axon. In: R L., J G., H L., K N., R M., BD T., editors. Myelin Biology and Disorders 1. London: Elsevier Academic Press. p 3-27. Vasilescu V., Katona E., Simplaceanu V., Demco D., 1978. Water compartments in the myelinated nerve. III. Pulsed NMR results. Experientia. 34, 1443-4. Vavasour I.M., Whittall K.P., Li D.K., MacKay A.L., 2000. Different magnetization transfer effects exhibited by the short and long T(2) components in human brain. Magnetic Resonance in Medicine 44, 860-6. Webb S., Munro C.A., Midha R., Stanisz G.J., 2003. Is multicomponent T2 a good measure of myelin content in peripheral nerve? Magnetic Resonance in Medicine. 49, 638-45. Whittall K.P., MacKay A.L., 1989. Quantitative interpretation of nmr relaxation data. Journal of Magnetic Resonance 84, 64-71. Whittall K.P., MacKay A.L., Graeb D.A., Nugent R.A., Li D.K., Paty D.W., 1997. In vivo measurement of T2 distributions and water contents in normal human brain. Magnetic Resonance in Medicine 37, 34-43. Whittall K.P., MacKay A.L., Li D.K., Vavasour I.M., Jones C.K., Paty D.W., 2002. Normal-appearing white matter in multiple sclerosis has heterogeneous, diffusely prolonged T(2). Magnetic Resonance in Medicine 47, 403-8. Wimberger D.M., Roberts T.P., Barkovich A.J., Prayer L.M., Moseley M.E., Kucharczyk J., 1995. Identification of" Premyelination" by Diffusion-Weighted MRI. Journal of Computer Assisted Tomography 19, 28. Woessner D.E., 1961. Effects of diffusion in nuclear magnetic resonance spin-echo experiments. Journal of Chemical Physics 34, 2057-61.    Chapter 2  2 ROI ANALYSIS COMPARING MULTI-COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 1.5 T*  2.1  INTRODUCTION Multiple sclerosis (MS) is a disease affecting the central nervous system (CNS) characterized by edema, inflammation, demyelination, axonal loss and gliosis (Keegan and Noseworthy 2002).  Magnetic resonance imaging (MRI) is a powerful tool for diagnosing and monitoring MS as it provides a non-invasive method to visualize CNS tissue.  Commonly used primary outcome measures of disease activity in clinical trials include lesion volume on T2 weighted images (burden of disease) and the number of gadolinium enhancing lesions (Group 1995). However, only weak correlations have been found between such measures and clinical disability (Gasperini et al., 1996; Group 1995; Jacobs et al., 1986; Koopmans et al., 1989; Thompson et al., 1992).  Thus, it is likely that there are pathological processes taking place that are not visible on conventional MR images; histopathological studies of white matter have shown inflammation, demyelination and astrocyte proliferation where no lesion or other macroscopically visible anomaly is present (Allen et al., 2001; Alling et al., 1971; Itoyama et al., 1980; Moore 1998), as well as axonal loss attributed to Wallerian degeneration (Evangelou et  * A version of this chapter has been published.  Kolind SH, Laule C, Vavasour IM, Li DK, Traboulsee AL, Mädler B, Moore GR, and Mackay AL. (2008) Complementary information from multi-exponential T2 relaxation and diffusion tensor imaging reveals differences between multiple sclerosis lesions. Neuroimage. 40(1):77-85.  32 al., 2000a; Evangelou et al., 2000b).  In order to better understand the pathogenesis of both (macroscopically) normal appearing white matter (NAWM) and MS lesions, techniques are required that are more sensitive and specific to white matter abnormalities than conventional MRI measures.  Two MRI techniques used to probe deeper into MS pathology are T2 relaxation imaging and diffusion tensor imaging (DTI).  In normal human white matter, water protons can be separated into different reservoirs based on their T2 relaxation time using multi-echo T2 relaxation.  The shortest T2 component (~20 ms) is attributed to water trapped between the myelin bilayers and an intermediate T2 component (~80 ms) is thought to arise from intra/extracellular (IE) water (MacKay et al., 1994; Whittall et al., 1997).  The ratio of the short T2 signal (< 40 ms) to the total signal in the T2 distribution gives a measure of the myelin water.  Results from several validation studies strongly support using myelin water as an in vivo marker for myelin content.  In guinea pig (Gareau et al., 1999; Gareau et al., 2000) and rat (Webb et al., 2003) models, white matter T2 distributions were shown to be multi-compartmental and the size of the shortest T2 component was found to correlate with histological measures of myelin.  Studies from formalin-fixed human brain yielded T2 distributions very similar to those found in vivo, and histopathological studies have shown both a strong qualitative and quantitative correlation between myelin water and the optical density of luxol fast blue staining for myelin (Laule et al., 2006; Moore et al., 2000). Studies specific to MS have shown decreases in myelin water within both NAWM and lesion (Laule et al., 2004; Oh et al., 2006; Tozer et al., 2005; Vavasour 1998; Wu et al., 2006).  A longer T2 signal (> 200 ms) has been detected in brains of individuals with MS,  33 Alzheimer’s disease (AD), and phenylketonuria (PKU) that is not present in healthy volunteers (Armspach et al., 1991; Helms 2001; Larsson et al., 1989; Laule et al., 2007b; Menon et al., 1992; Oakden et al., 2006; Rumbach et al., 1991; Sirrs et al., 2007).   It has been shown that MS lesions with long-T2 signal tend to have significantly longer T1, longer IE T2 time, higher water content, and a lower magnetization transfer ratio than lesions without long-T2 signal (Laule et al., 2007b).  They are also more likely to be T1 black holes (Vavasour et al., 2007) (T1 black holes are hypointense on T1-weighted images and hyperintense on T2-weighted images, and classified as acute when occurring coincidently with enhancement, and as persisting when present after 6 months post- enhancement; they are thought to represent areas of major tissue destruction in MS patients (Bitsch et al., 2001; Bruck et al., 1997; van Waesberghe et al., 1999; van Walderveen et al., 1998)).  All of these clues tend to indicate inflammation and/or severe damage.  Thus, while the pathological cause of long-T2 signal in MS lesions is unknown, suggestions include inflammation, extensive destruction with increased extracellular water, and severe axonal damage.  DTI is also thought to offer insight into myelin and axonal integrity.  Although a voxel or region of interest (ROI) in white matter generally contains many fibers running in different directions, the diffusion of water in this complicated architecture can be approximated using a diffusion tensor (Basser et al., 1994).  This tensor can be diagonalized to obtain the diffusion tensor eigenvectors which allow diffusion to be described along three mutually perpendicular directions.  The diffusion tensor eigenvalues are the magnitudes of the eigenvectors and are also referred to as  34 diffusivities.  The largest eigenvalue, λ||, is the diffusivity parallel to the dominant diffusion direction, and the average of the two smaller eigenvalues, λ | , is the diffusivity perpendicular to the prevailing direction of diffusion (Basser 1995; Basser et al., 2000; Xue et al., 1999).  Diffusion anisotropy in the CNS is thought to occur due to the water molecules preferentially diffusing along the length of the axons and being hindered by barriers such as the myelin sheath when diffusing perpendicular to the axons (Moseley et al., 1990; Thomsen et al., 1987).  Thus λ|| has been linked to axonal integrity, while a change in λ |  is thought to reflect myelin pathology (Beaulieu et al., 1996; Biton et al., 2006; Kim et al., 2006; Song et al., 2003; Song et al., 2002; Song et al., 2005; Sun et al., 2006a; Sun et al., 2006b; Thomalla et al., 2004).  <D> is the average of the three eigenvalues, and represents the mean diffusivity.  The value of <D> is larger when there are fewer microscopic impediments to diffusion; for example, the ventricles have a high <D> due to relatively unrestricted diffusion.  Fractional anisotropy (FA) is a scalar invariant related to the degree of diffusional anisotropy.  The FA value is high when diffusion is strongly favoured along one direction, and is thus affected by the presence of myelin sheaths or axonal membranes which restrict directionality of diffusion.  However, FA is sensitive to other factors and is very heterogeneous throughout the brain; in particular, FA may appear low in regions of crossing fibers or where there is no preferred fiber bundle direction dominating the region.  Studies employing DTI to investigate MS have reported increased <D> and reduced FA in NAWM (Bammer et al., 2000; Cercignani et al., 2001; Cercignani et al., 2000; Christiansen et al., 1993; Ciccarelli et al., 2001; Iannucci et al., 2001; Mainero et al., 2001; Nusbaum et al., 2000; Werring et al., 1999), but these results were not specific to a particular pathology or structural change.  35 In terms of the eigenvalues, diffusion studies in MS NAWM have shown increased λ | and little or no change in λ||, which could be consistent with demyelination (Henry et al., 2003; Lin et al., 2007; Lowe et al., 2006).  Since both myelin water and diffusion metrics such as FA and λ |  have been linked to myelin, it could be hypothesized that these measures would correlate with each other. Two studies have compared myelin water to DTI metrics in healthy volunteers; Mädler et al. (Mädler et al., 2002) found that FA was correlated with myelin water (R2 = 0.55, p < 0.0001), and Bells et al. (Bells et al., 2007) showed that myelin water was strongly correlated with FA and λ |  (R2 = 0.86 and 0.85 respectively), somewhat correlated with <D> (R2 = 0.60) and not at all correlated with λ|| (R2 = 0.05).  However, in normal volunteers it is difficult to determine whether the dominant factor affecting each metric is the presence of myelin sheaths or axonal membranes as they are highly correlated in healthy brain.  It is also important to know how each measure is relatively affected by inflammation, edema, myelin or axonal loss, or other pathology.  The goal of this study was to investigate the relationship between measures obtained from T2 relaxation and DTI in MS lesions and NAWM. In particular, we wish to better characterize MS lesions with a long-T2 signal, including establishing whether the presence of this signal indicates a different underlying pathology than is present in lesions not exhibiting long-T2 signal.   36 2.2 METHODS 2.2.1 Subject information 19 subjects with clinically definite MS (14 female, 5 male; 14 relapsing-remitting, 5 secondary-progressive; median Expanded Disability Status Scale (EDSS) = 2.5 (range 1.0-8.0); mean age = 38 yrs (range 23-54 yrs); mean disease duration = 10.5 yrs (range 1- 35 yrs)) were studied. None of the patients were receiving steroid treatment.  For all subjects, informed written consent as approved by the Clinical Research Ethics Board of our institution was obtained.  2.2.2 MR data acquisition MR images were obtained with a 1.5 T GE Signa MR scanner (General Electric Medical Systems, Milwaukee, WI) operating at the Epic 5.7 software level, using a transmit/receive head coil.  Six MnCl2-doped agarose water reference phantoms with various T1 and T2 values were placed within the volume of interest (Mitchell et al., 1986). After a sagittal localizer (TR = 350 ms, TE = 14 ms), an axial FLAIR (fluid attenuated inversion recovery (Hajnal et al., 1992)) (TR = 10000 ms, TE = 145 ms, TI = 2200 ms, 28 slices) was acquired.  T1 relaxation was measured using a single slice fast gradient echo inversion-recovery prepared sequence (TE = 8 ms, 1 average, matrix size = 256×128, 14 inversion times ranging from 0.1 – 2.5 s).  T2 relaxation was measured using a single-slice multi-echo experiment, consisting of a 90o slice selective pulse followed by 48 composite block pulses (90ox-180oy-90ox) flanked by crusher gradient pulses along the z-axis designed to minimize stimulated echoes (Poon and Henkelman 1992) (TR =  37 2120-3800 ms, echo spacing for the first 32 echoes = 10 ms, for last 16 echoes = 50 ms, matrix size = 256×128, averages = 4).  The TR was 3800 ms for the central line of k- space, and decreased linearly to 2120 ms for the largest positive and negative k values (Laule et al., 2007a).  This centric encoding shortened the acquisition time by 8 minutes at the cost of slight image blurring, resulting in a total scan time of 25.4 minutes.  The slice location for the relaxation experiments was adjusted for each patient, chosen to include the maximal amount of lesional white matter.  DTI was performed with a single shot pulsed-field gradient EPI sequence using 3 b-values between 0 and 1600 s/mm2 in 7 non-colinear directions (Mädler et al., 2000), and was RR-gated (cardiac gated to the RR interval, the time between identical points in the heart cycle) using a finger-pulse- oximeter (TR was chosen to be 3-4 RR-intervals such that the TR was at least 3 s, TE = 85 ms, δ = 25 ms, Δ = 38 ms, matrix size = 128x128, averages = 4).  DTI was measured over 4 slices, one of which matched the slice for which T2 relaxation was measured.  A conventional spin echo PD/T2 scan was also performed (TR = 2500 ms, TE = 30/90 ms, 28 slices, matrix size = 256×192).  Finally, a post-contrast T1-weighted spin echo scan (TR = 550 ms, TE = 8 ms, 28 slices, matrix 256×192) was collected 5 minutes after the injection of gadolinium-DTPA (Gad) at a dose of 0.1 mmol/kg.  The field of view for all exams was 22 cm and the slice thickness was 5 mm.  2.2.3 MR data analysis ROIs were drawn around lesions (identified by a trained observer) and contralateral NAWM (cNAWM) on the first echo image (echo time = 10ms) of the single-slice T2 relaxation data set.  Only cNAWM regions that did not appear abnormal on the PD/T2  38 images, including the slices immediately above and below the ROI, were included in the analysis.  In-house software was created for rigid registration of the ROIs to the non- diffusion weighted (b = 0) image from the DTI data set for the corresponding slice by user-defined single pixel shifts to the right or left, up or down.  The T2 decay curves were analyzed using a regularized non-negative least squares (NNLS) method (Lawson and Hanson 1974; Whittall et al., 1991; Whittall and MacKay 1989).  For each decay curve, the algorithm determines the set of amplitudes sj that minimizes the least-squares misfit given the T2 partition.  The echo decay curve intensities are given by:   (2.1) NiTtsy M j jiji ,...,2,1,)/exp( 1 2 =−= ∑ = where ti are the echo times used, sj are the unknown relative amplitudes for each partitioned relaxation time T2j, N is the number of echoes acquired (N = 48 for this study), and M = 120 relaxation times equally spaced on a logarithmic scale from 15 ms to 2 s.  The plot of sj versus T2 is known as the T2 distribution.  Component fractions were calculated as the fraction of the total signal within the specified ranges divided by the signal for the entire T2 distribution.  The ranges used were 15 ms – 40 ms for the myelin water fraction, and 200 ms – 800 ms for the long-T2 fraction (LT2F).  Myelin water content (MWC) was found by multiplying the myelin water fraction by the total water content, which was defined as the total area under the T2 distribution normalized to the water standards, corrected for temperature (Kolind et al., 2007; Lin et al., 2000; Neeb et al., 2006; Tofts 2003) and T1 relaxation (T1 values were  39 calculated from the T1 relaxation experiment, results of which were fit to a mono- exponential function).  The diffusion tensor D was calculated by solving:  ∑∑ = = −= 3 1 3 1 , )0( )(ln i j ijij DbA bA  (2.2) where A(b) is the signal intensity for the corresponding b value, bij is the ijth element of the b matrix, and Dij is the ijth element of the diffusion tensor.  The eigenvalues λ1, λ2, and λ3 (from largest to smallest, respectively) were obtained from the resulting diffusion tensor D.  <D> is given by the average of the three eigenvalues.  FA was defined as:  , )()()( 2 1 2 3 2 2 2 1 2 13 2 32 2 21 λλλ λλλλλλ ++ −+−+−=FA  (2.3) and the diffusivities were given by λ|| = λ1 and λ |  = (λ2 + λ3)/2.  Maps of MWC, LT2F, <D>, λ | , and λ|| were created by displaying the value at each pixel in the image.  The total burden of disease (representing the amount of lesion present in the PD/T2 multi- echo scan) was determined as follows (Jones et al., 2002): A 2-dimensional PD/T2 histogram was obtained (Jones and Wong 2002), from which a set of basis points representing grey matter, white matter, and cerebrospinal fluid were calculated automatically (Jones et al., 2002).  A radiologist placed a seed point in each lesion present in the PD/T2 scans, and then lesions were grown by classifying the four voxels neighbouring the seed point as lesion, grey matter, white matter, or cerebrospinal fluid  40 points.  Any voxels classified as lesion were added to the region representing the full lesion, and subsequently these voxels were used as seed points for the next iteration of the growing operation.  The lesion was grown outward until no more bordering voxels classified as lesion were found.  Statistical analysis was carried out using a two-tailed Student’s t test with a p-value of < 0.05 considered significant.  All errors are expressed as standard errors (SE). The Pearson correlation coefficient (R2) was calculated between pairs of MR measurements.  2.3 RESULTS One hundred and seven lesions (33 from secondary-progressive patients) and 83 cNAWM ROIs (14 from secondary-progressive patients) from 19 MS patients were analyzed.  It was not possible to obtain a corresponding cNAWM ROI for 24 of the lesions because the tissue contralateral to those lesions was visibly affected by the disease.  Due to known regional differences in MWC and diffusion metrics, it would not be appropriate to use an ROI that was not directly contralateral to the lesion.  The average burden of disease was 14,000 mm3 (range 125 – 93,500 mm3, standard deviation 21,500 mm3).  27 of the lesions (25%) exhibited long-T2 signal.  Six enhancing lesions were identified in 3 MS patients, 4 of which had a long-T2 component.  No enhancing lesions were detected in any of the other MS patients.  10 of the 19 MS patients had lesions with a long-T2 component.  The average volume of the lesions without a long-T2 component was 110 mm3, and the average volume of lesions with long-T2 signal was 330 mm3. While not all big lesions had a long-T2 component, within a patient, the largest lesions  41 were apt to be the ones with long-T2 signal.  The extent of the long-T2 signal tended to roughly match the size and shape of the lesion as seen on the first echo of the T2 relaxation data, although the long-T2 signal area was sometimes smaller.  In particular, any part of the lesion that appeared diffuse or less bright than the bulk of the lesion on the first echo image usually did not have long-T2 signal.  Average values for MWC, FA, <D>, λ | , and λ|| in lesions were all found to be significantly different (p < 0.0001) from average values in cNAWM (see Table 2.1). Excluding lesions without a corresponding cNAWM ROI did not affect this result.  MWC (g/100ml) FA <D> (μm2/ms) λ |  (μm2/ms) λ|| (μm2/ms) cNAWM 5.7 (0.4) 0.43 (0.01) 0.69 (0.01) 0.55 (0.01) 1.03 (0.02) All lesions 3.3 (0.2) 0.31 (0.01) 0.97 (0.03) 0.83 (0.03) 1.27 (0.02)  Table 2.1: Mean value (SE) of MWC, FA, <D>, λ | , and λ|| for cNAWM and all lesions.  The difference between the two groups was highly significant (p < 0.0001) for all metrics.  Separation of the lesions into two groups based on the presence of a long-T2 signal revealed that all of the diffusion metrics were significantly different (p < 0.0008) in lesions with long-T2 signal than in those without (see Table 2.2), while MWC was not found to be significantly lower in lesions with long-T2 signal (p = 0.08).  42   MWC (g/100ml) FA <D> (μm2/ms) λ |  (μm2/ms) λ|| (μm2/ms) Lesions without long T2 signal 3.5 (0.3) 0.34 (0.02) 0.89 (0.02) 0.74 (0.02) 1.21 (0.02) Lesions with long T2 signal 2.6 (0.5) 0.23 (0.03) 1.21 (0.06) 1.08 (0.07) 1.46 (0.06) p-value 0.08 0.0008 0.0001 <0.0001 0.0005 Table 2.2: Mean value (SE) of MWC, FA, <D>, λ | , and λ|| for lesions with and without long-T2 signal.  The p-value resulting from a Student’s t-test between the two groups of lesions is indicated in bold where significant (p < 0.05).  No significant correlation was found between MWC and any of the diffusion metrics (FA, <D>, λ | , or λ||) in NAWM (p > 0.07, see Figure 2.1).  Lesions with no long-T2 signal had weak correlations (R2 ranging from 0.05 to 0.14, p < 0.04) between MWC and all of the diffusion metrics (see Figure 2.2).  In contrast, <D>, λ | , and λ|| all had strong negative correlations with MWC in lesions exhibiting a long-T2 component (R2 = 0.61, 0.54 and 0.60 respectively, p < 0.0001) while FA and MWC were not significantly correlated in this lesion subtype (R2 = 0.13, p = 0.07).  43 R2 = 0.00 p = 0.6 0.0 0.2 0.4 0.6 0.8 R2 = 0.04 p = 0.07 0.0 0.5 1.0 1.5 2.0 R2 = 0.02 p = 0.2 0.0 0.5 1.0 1.5 2.0 R2 = 0.02 p = 0.2 0.0 0.5 1.0 1.5 2.0 λ || (μ m 2 /m s) λ | (μ m 2 /m s) <D > (μ m 2 /m s) FA MWC (g/100ml) 0 5 10 15 20 λ || (μ m 2 /m s) λ | (μ m 2 /m s) <D > (μ m 2 /m s) FA  Figure 2.1: Correlations between MWC and FA, <D>, λ | , and λ|| for cNAWM. R2 and the p-value are indicated on the plot.  None of the correlations were found to be significant (p < 0.05).  44 R2 = 0.13 p = 0.07 0.0 0.2 0.4 0.6 0.8 R2 = 0.61 p < 0.0001 0.0 0.5 1.0 1.5 2.0 2.5 R2 = 0.54 p < 0.0001 0.0 0.5 1.0 1.5 2.0 2.5 R2 = 0.60 p = 0.0001 0.0 0.5 1.0 1.5 2.0 2.5 R2 = 0.07 p = 0.02 0.0 0.2 0.4 0.6 0.8 R2 = 0.14 p = 0.0008 0.0 0.5 1.0 1.5 2.0 2.5 R2 = 0.13 p = 0.001 0.0 0.5 1.0 1.5 2.0 2.5 R2 = 0.05 p = 0.04 0.0 0.5 1.0 1.5 2.0 2.5 λ || (μ m 2 /m s) λ | (μ m 2 /m s) <D > (μ m 2 /m s) FA MWC (g/100ml) 0 5 10 MWC (g/100ml) 0 5 Lesions with no long-T2 signal Lesions with long-T2 signal 10 λ || (μ m 2 /m s) λ | (μ m 2 /m s) <D > (μ m 2 /m s) FA  Figure 2.2: Correlations between MWC and FA, <D>, λ | , and λ|| for (left) the subset of lesions with no long-T2 signal, and (right) lesions exhibiting long-T2 signal.  For each correlation, R2 and the p-value are indicated on the plot.  The p-value is bold for significant correlations (p < 0.05).  The LT2F (for the 27 lesions with a long-T2 component) also showed a strong negative correlation with <D>, λ | , and λ||  (R2 = 0.64, 0.69, and 0.42 respectively, p < 0.0003) and a modest positive relationship between FA and LT2F (R2 = 0.27, p = 0.006) (see Figure 2.3).  45 R2 = 0.27 p = 0.006 0.0 0.2 0.4 0.6 0.8 R2 = 0.64 p < 0.0001 0.0 0.5 1.0 1.5 2.0 2.5 R2 = 0.69 p < 0.0001 0.0 0.5 1.0 1.5 2.0 2.5 R2 = 0.42 p = 0.0003 0.0 0.5 1.0 1.5 2.0 2.5 λ || (μ m 2 /m s) λ | (μ m 2 /m s) <D > (μ m 2 /m s) FA LT2F (%) 0 25 50 75 100 λ || (μ m 2 /m s) λ | (μ m 2 /m s) <D > (μ m 2 /m s) FA  Figure 2.3: Correlations between LT2F and FA, <D>, λ | , and λ|| for lesions exhibiting long-T2 signal.  R2 and the p-value are indicated on the plots.  All correlations were found to be significant (p < 0.05).  Figure 2.4 depicts the first echo image (echo time = 10 ms) of the single-slice T2 relaxation data set as well as maps of MWC, LT2F, <D>, λ | , and λ|| for one MS subject. Note that the images created from T2 relaxation data offer information complementary to that provided by the diffusion tensor imaging maps.  There are 14 lesions evident on the first echo image; on the MWC map these lesions are apparent but while most show some  46 remaining myelin water, 4 are nearly completely dark.  One such lesion is indicated in Figure 2.4 (lesion a), where the MWC is nearly zero.  There are also regions on the MWC map that are darker than surrounding NAWM that do not stand out on the first echo image (such as region b specified in Figure 2.4), possibly as a result of diffuse demyelination or Wallerian degeneration.  Region b is not very conspicuous in the maps of the diffusion metrics except faintly in the map of λ | .  Only 5 of the lesions are visible on the LT2F map.  These long-T2 lesions (such as lesion c in Figure 2.4) seem to be more evident on the diffusion metric maps, particularly the maps of <D> and λ||.  It is interesting that while lesion c has a large LT2F, its MWC is still relatively high. aa λ λ First echo MWC LT2F <D> a a a a bb b b b b c c c c c c  Figure 2.4: First echo image (echo time = 10 ms) of the single-slice T2 relaxation data set, MWC map, LT2F map, <D> map, λ |  map, and λ|| map, for one MS patient.  Region a outlines a lesion with little remaining myelin water, region b indicates a region of decreased myelin water content that is not apparent on the first echo image, and region c outlines a lesion with long-T2 signal.  47 2.4 DISCUSSION 2.4.1 Comparison between healthy white matter and MS cNAWM In healthy controls, myelin water has previously been shown to be strongly correlated with FA when both white matter and grey matter ROIs are included (R2 = 0.55, p<0.0001) (Mädler et al., 2002).  However, results in normal white matter or grey matter alone show much weaker correlations.  Since FA is a ratio and depends on both parallel and perpendicular diffusion, it can be ambiguous; therefore, it is more informative to look directly at the eigenvalues.  In healthy brain, Bells et al. (Bells et al., 2007) found that λ | was strongly correlated with myelin water whether or not grey matter was included (R2 = 0.85 with grey matter, or 0.54 without).  λ|| has been shown to be affected by the presence of axons, and since myelin content is highly linked to axons in healthy brain, a correlation between λ|| and myelin water could be anticipated.  However, Bells et al. found that λ|| was not correlated with myelin water (R2 = 0.05 including grey matter ROIs, or 0.03 without).  While the authors did not speculate on this lack of correlation, it could be because λ|| did not vary much in healthy brain tissue and a relationship would only become clear in cases of axonal damage causing changes in λ||.  Conversely, in this study, none of the diffusion metrics were found to be significantly correlated with MWC in cNAWM (see Figure 2.1), which may be because different brain regions were examined than in studies of healthy controls.  For the cNAWM ROIs (largely above and around the ventricles), the variation in diffusion metrics, with standard deviation on the order of 20% of the average, was much smaller than the variation in MWC with standard deviation approximately 65% of the average. The standard deviation  48 for the diffusion metrics may have been artificially smaller in this study due to the use of 5mm slices which results in partial volume effects due to overlapping fibre tracks in the slice (Oouchi et al., 2007).  However the variation in MWC from NAWM measured here is representative of that measured in normal brain (Whittall et al., 1997) and much larger than the  reproducibility of MWC which is on the order of 18% of the average (Vavasour et al., 2006).  2.4.2 Comparison between cNAWM and lesions with and without long-T2 components As seen in Table 2.1, <D> was higher while both MWC and FA were lower in lesions than in cNAWM, in agreement with the literature (Bammer et al., 2000; Cercignani et al., 2001; Cercignani et al., 2000; Christiansen et al., 1993; Ciccarelli et al., 2001; Iannucci et al., 2001; Mainero et al., 2001; Nusbaum et al., 2000; Werring et al., 1999).  The increase in <D> was the result of an increase in both λ |  (by an average of 51%) and in λ|| (by an average of 23%).  Thus, while diffusion increased both along the primary direction of the fibers and perpendicular to it, FA decreased overall because there was a greater increase in λ |  than in λ||.  In lesions with no long-T2 component, weak correlations between MWC and diffusion measures were observed that were not present in cNAWM; however, looking at only the lesions with long-T2 signal, a stronger correlation between diffusion measures and MWC became apparent (see Figure 2.2).  This suggests that there is some fundamental difference in pathology between these two groups of lesions.  49  2.4.3 Correlations between MWC and diffusion metrics in lesions with long- T2 components In lesions with long-T2 signal, <D> increased with decreasing MWC as expected, since a reduction in myelin should result in less hindrance to diffusion, and thus a larger average diffusivity.  There was a strong correlation between λ |  and MWC in lesions with a long-T2 component, with λ |  higher for lower values of MWC.  This result is consistent with findings in animal models indicating that changes in λ |  are linked to myelin integrity (Beaulieu et al., 1996; Biton et al., 2006; Song et al., 2003; Song et al., 2002; Song et al., 2005; Sun et al., 2006b).  It is unclear why this relationship was less apparent in lesions without long-T2 signal.  MWC was also strongly related to λ|| in lesions exhibiting long-T2 signal; lesions with lower MWC values yielded higher values for λ||.  Indeed, λ|| may be influenced by myelin content; Biton et al. (Biton et al., 2006) observed that λ|| increased significantly in NAWM within excised myelin deficient rat spinal cord compared to age-matched controls, and Wu et al. (Wu et al., 2007) found a similar result in shaking pup brain (a canine mutant with a profound paucity of myelin without axonal loss).  While Song et al. (Song et al., 2002) did not find a significant difference in λ|| between shiverer mice (which have a near complete absence of myelin in the CNS) and controls, Tyszka et al. (Tyszka et al., 2006) did find shiverer mice to have significantly increased λ|| compared to  50 control mice in several brain regions, and suggested decreased extra-axonal diffusion restrictions due either to an absence of compact myelin or an increase in the partial volume of cell bodies within the ROI as possible mechanisms.  Each of these studies reported increases in λ |  compared to controls.  2.4.4 Difference between lesions with and without long-T2 components Lesions with a long-T2 component show many traits which suggest greater tissue destruction with axonal damage and severe myelin damage, such as increased T1, T2 and water content values, and decreased magnetization transfer ratio (Laule et al., 2007b). Subjects with lesions exhibiting long-T2 signal also tend to have a longer disease duration (Laule et al., 2007b).  While the additional signal from the long-T2 component may be due to inflammation or edema for acute lesions, for the more destructive lesions studied it more likely represents increased extracellular water, replacing the now absent myelin or axons. The increased water should result in a spreading apart and less packing of the fibers, with fewer impediments to diffusion in between the fibers; this is supported by the increase in <D> in lesions with long-T2 signal and its strong correlation with LT2F (see Figure 2.3), which are both consistent with an increase in free water.  Unlike histopathological studies (Song et al., 2003; Song et al., 2005; Sun et al., 2006a; Sun et al., 2006b), where a decrease in λ|| occurred with axonal damage, the average values of λ|| were higher in those lesions with a long-T2 component (expected to have greater axonal damage) than in those without.  Thus the source of changes in λ|| in this study is probably not the same as in those animal model studies.  Histopathological  51 studies showing reductions in λ|| with axonal damage generally use models of acute axonal damage (such as retinal ischemia or cuprizone treatment) where the presence of axonal debris is believed to restrict parallel diffusivity; for more chronic axonal damage where much of the debris may have been cleared away, λ|| should increase.  Thus, the factors that contribute to increases or decreases in λ|| require further elucidation before it can be relied on as a MRI marker of axonal integrity.  We present a possible interpretation for the observation of increased λ|| (and λ | ) in long- T2 lesions.  If long-T2 lesions are chronically and substantially demyelinated but retain a sufficient number of oriented and unmyelinated axons, there is a high probability of isomorphic fibrillary gliosis (Ellison et al., 2004).  Early in the development of a MS lesion, astrocytes become quite broad and very large and synthesize many glial fibrillary acidic protein (GFAP) filaments, which at that point are arranged more or less in a random fashion in their cell bodies.  With time (i.e. months to years) the astrocyte cell bodies become smaller and sinuous with the GFAP filaments now oriented primarily longitudinally parallel to the cell membranes of the long processes in which they reside. If this occurs in an environment where there are uninterrupted demyelinated axons, these elongated astrocytic processes line up against the axons in a very regular fashion; this is known as isomorphic gliosis.  Essentially the former myelin sheaths are replaced by long fibrillary astrocytic processes which contain intracellular fluid.  In lesions with isomorphic gliosis, one would expect an increase in parallel diffusivity due to diffusion along these elongated astrocytic processes in addition to the diffusion along the surviving  52 (but demyelinated) axons. More research is required to better define the relationship between the reported MR metrics and MS pathology.  In conclusion, we have shown that the presence of a long-T2 signal in MS lesion indicates a different underlying pathology than is present in lesions not exhibiting this phenomenon; we report a remarkable relationship between diffusion metrics and MWC which does not exist in NAWM or other lesions.  More studies are required to better characterize the underlying pathology present in this lesion type.  Multi-echo T2 relaxation is of great benefit to the study of MS as the MWC and long-T2 signal offer information that is not available through more commonly used techniques.  2.5 ACKNOWLEDGEMENTS Sincere thanks to the MS patients and the MR technologists at UBC Hospital.  The authors also wish to thank Andrew Riddehough, Erick B. Wong, Dr. Guo Jun Zhao and Melinda Medina for their assistance in calculating the burden of disease.  This study was funded by the Multiple Sclerosis Society of Canada.  Shannon Kolind was the recipient of a Killam Predoctoral Fellowship.   53 2.6 REFERENCES Allen I.V., McQuaid S., Mirakhur M., Nevin G., 2001. Pathological abnormalities in the normal-appearing white matter in multiple sclerosis. Neurological Sciences 22, 141-4. Alling C., Vanier M.T., Svennerholm L., 1971. Lipid alterations in apparently normal white matter in multiple sclerosis. Brain Research. 35, 325-36. Armspach J.P., Gounot D., Rumbach L., Chambron J., 1991. In vivo determination of multiexponential T2 relaxation in the brain of patients with multiple sclerosis. Magn Reson Imaging. 9, 107-13. Bammer R., Augustin M., Strasser-Fuchs S., Seifert T., Kapeller P., Stollberger R., Ebner F., Hartung H.P., Fazekas F., 2000. Magnetic resonance diffusion tensor imaging for characterizing diffuse and focal white matter abnormalities in multiple sclerosis. Magnetic Resonance in Medicine 44, 583-91. Basser P.J., 1995. Inferring microstructural features and the physiological state of tissues from diffusion-weighted images. NMR in Biomedicine 8, 333-44. Basser P.J., Mattiello J., LeBihan D., 1994. Estimation of the effective self-diffusion tensor from the NMR spin echo. J Magn Reson B 103, 247-54. Basser P.J., Pajevic S., Pierpaoli C., Duda J., Aldroubi A., 2000. In vivo fiber tractography using DT-MRI data. Magnetic Resonance in Medicine 44, 625-632. Beaulieu C., Does M.D., Snyder R.E., Allen P.S., 1996. Changes in water diffusion due to Wallerian degeneration in peripheral nerve. Magn Reson Med 36, 627-31. Bells S., Morris D., Vidarsson L., Comparison of linear combination filtering to DTI and MTR in whole brain myelin-water imaging. 15th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2007; Berlin, GER. p 1606. Biton I.E., Duncan I.D., Cohen Y., 2006. High b-value q-space diffusion MRI in myelin- deficient rat spinal cords. Magn Reson Imaging. 24, 161-6. Epub 2005 Dec 27. Bitsch A., Kuhlmann T., Stadelmann C., Lassmann H., Lucchinetti C., Bruck W., 2001. A longitudinal MRI study of histopathologically defined hypointense multiple sclerosis lesions. Ann Neurol. 49, 793-6. Bruck W., Bitsch A., Kolenda H., Bruck Y., Stiefel M., Lassmann H., 1997. Inflammatory central nervous system demyelination: correlation of magnetic resonance imaging findings with lesion pathology. Annals of Neurology. 42, 783- 93. Cercignani M., Bozzali M., Iannucci G., Comi G., Filippi M., 2001. Magnetisation transfer ratio and mean diffusivity of normal appearing white and grey matter from patients with multiple sclerosis. Journal of Neurology, Neurosurgery & Psychiatry 70, 311-7. Cercignani M., Iannucci G., Rocca M.A., Comi G., Horsfield M.A., Filippi M., 2000. Pathologic damage in MS assessed by diffusion-weighted and magnetization transfer MRI. Neurology 54, 1139-44. Christiansen P., Gideon P., Thomsen C., Stubgaard M., Henriksen O., Larsson H.B., 1993. Increased water self-diffusion in chronic plaques and in apparently normal white matter in patients with multiple sclerosis. Acta Neurologica Scandinavica. 87, 195-9.  54 Ciccarelli O., Werring D.J., Wheeler-Kingshott C.A.M., Barker G.J., Parker G.J.M., Thompson A.J., Miller D.H., 2001. Investigation of MS normal-appearing brain using diffusion tensor MRI with clinical correlations. Neurology 56, 926-933. Ellison D., Love S., Chimelli L., Harding B., Lowe J.S., Vinters H., 2004. Neuropathology, -A Reference Text of CNS Pathology. Edinburgh: Mosby. Evangelou N., Esiri M.M., Smith S., Palace J., Matthews P.M., 2000a. Quantitative pathological evidence for axonal loss in normal appearing white matter in multiple sclerosis. Annals of Neurology. 47, 391-5. Evangelou N., Konz D., Esiri M.M., Smith S., Palace J., Matthews P.M., 2000b. Regional axonal loss in the corpus callosum correlates with cerebral white matter lesion volume and distribution in multiple sclerosis. Brain. 123, 1845-9. Gareau P.J., Rutt B.K., Bowen C.V., Karlik S.J., Mitchell J.R., 1999. In vivo measurements of multi-component T2 relaxation behaviour in guinea pig brain. Magnetic Resonance Imaging. 17, 1319-25. Gareau P.J., Rutt B.K., Karlik S.J., Mitchell J.R., 2000. Magnetization transfer and multicomponent T2 relaxation measurements with histopathologic correlation in an experimental model of MS. Journal of Magnetic Resonance Imaging 11, 586- 95. Gasperini C., Horsfield M.A., Thorpe J.W., Kidd D., Barker G.J., Tofts P.S., MacManus D.G., Thompson A.J., Miller D.H., McDonald W.I., 1996. Macroscopic and microscopic assessments of disease burden by MRI in multiple sclerosis: relationship to clinical parameters. Journal of Magnetic Resonance Imaging 6, 580-4. Group I.M.S.S., 1995. Interferon beta-1b in the treatment of multiple sclerosis: final outcome of the randomized controlled trial. The IFNB Multiple Sclerosis Study Group and The University of British Columbia MS/MRI Analysis Group. Neurology 45, 1277-85. Hajnal J.V., Bryant D.J., Kasuboski L., Pattany P.M., De Coene B., Lewis P.D., Pennock J.M., Oatridge A., Young I.R., Bydder G.M., 1992. Use of fluid attenuated inversion recovery (FLAIR) pulse sequences in MRI of the brain. J Comput Assist Tomogr 16, 841-4. Helms G., 2001. Volume correction for edema in single-volume proton MR spectroscopy of contrast-enhancing multiple sclerosis lesions. Magnetic Resonance in Medicine 46, 256-63. Henry R.G., Oh J., Nelson S.J., Pelletier D., 2003. Directional diffusion in relapsing- remitting multiple sclerosis: a possible in vivo signature of Wallerian degeneration. J Magn Reson Imaging. 18, 420-6. Iannucci G., Rovaris M., Giacomotti L., Comi G., Filippi M., 2001. Correlation of multiple sclerosis measures derived from T2-weighted, T1-weighted, magnetization transfer, and diffusion tensor MR imaging. AJNR Am J Neuroradiol 22, 1462-7. Itoyama Y., Sternberger N.H., Webster H.D., Quarles R.H., Cohen S.R., Richardson E.P., Jr., 1980. Immunocytochemical observations on the distribution of myelin- associated glycoprotein and myelin basic protein in multiple sclerosis lesions. Annals of Neurology 7, 167-77.  55 Jacobs L., Kinkel W.R., Polachini I., Kinkel R.P., 1986. Correlations of nuclear magnetic resonance imaging, computerized tomography, and clinical profiles in multiple sclerosis. Neurology. 36, 27-34. Jones C.K., Wong E.B., Multi-scale application of the N3 method for intensity correction of MR images. In: Fitzpatrick M.S.a.J.M., editor. SPIE Medical Imaging 2002: Image Processing; 2002; San Diego, USA. p 1123-1129. Jones C.K., Wong E.B., Riddehough A., Li D.K.B., Paty D.W., Semi-automatic segmentation of MS lesions compared to manual tracing. 10th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2002; Honolulu, USA. p 352. Keegan B.M., Noseworthy J.H., 2002. Multiple sclerosis. Annual Review of Medicine 53, 285-302. Kim J.H., Budde M.D., Liang H.F., Klein R.S., Russell J.H., Cross A.H., Song S.K., 2006. Detecting axon damage in spinal cord from a mouse model of multiple sclerosis. Neurobiol Dis. 21, 626-32. Epub 2005 Nov 17. Kolind S.H., Mädler B., Vavasour I.M., Laule C., Yip E., MacKay A.L., Quantitative in- vivo brain water content measurements at 3.0T. 15th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2007; Berlin, GER. p 2259. Koopmans R.A., Li D.K., Grochowski E., Cutler P.J., Paty D.W., 1989. Benign versus chronic progressive multiple sclerosis: magnetic resonance imaging features. Annals of Neurology. 25, 74-81. Larsson H.B., Frederiksen J., Petersen J., Nordenbo A., Zeeberg I., Henriksen O., Olesen J., 1989. Assessment of demyelination, edema, and gliosis by in vivo determination of T1 and T2 in the brain of patients with acute attack of multiple sclerosis. Magnetic Resonance in Medicine. 11, 337-48. Laule C., Kolind S.H., Bjarnason T.A., Li D.K., Mackay A.L., 2007a. In vivo multiecho T(2) relaxation measurements using variable TR to decrease scan time. Magn Reson Imaging 25, 834-9. Laule C., Leung E., Li D.K., Traboulsee A.L., Paty D.W., MacKay A.L., Moore G.R., 2006. Myelin water imaging in multiple sclerosis: quantitative correlations with histopathology. Mult Scler. 12, 747-53. Laule C., Vavasour I.M., Madler B., Kolind S.H., Sirrs S.M., Brief E.E., Traboulsee A.L., Moore G.R., Li D.K., Mackay A.L., 2007b. MR evidence of long T(2) water in pathological white matter. J Magn Reson Imaging 26, 1117-21. Laule C., Vavasour I.M., Moore G.R.W., Oger J., Li D.K.B., Paty D.W., MacKay A.L., 2004. Water content and myelin water fraction in multiple sclerosis: A T2 relaxation study. Journal of Neurology 251, 284-293. Lawson C.L., Hanson R.J., 1974. Solving Least Squares Problems. Englewood Cliffs, NJ: Prentice-Hall,. Lin F., Yu C., Jiang T., Li K., Chan P., 2007. Diffusion Tensor Tractography-Based Group Mapping of the Pyramidal Tract in Relapsing-Remitting Multiple Sclerosis Patients. AJNR Am J Neuroradiol. 28, 278-282. Lin W., Venkatesan R., Gurleyik K., He Y.Y., Powers W.J., Hsu C.Y., 2000. An absolute measurement of brain water content using magnetic resonance imaging in two focal cerebral ischemic rat models. J Cereb Blood Flow Metab 20, 37-44.  56 Lowe M.J., Horenstein C., Hirsch J.G., Marrie R.A., Stone L., Bhattacharyya P.K., Gass A., Phillips M.D., 2006. Functional pathway-defined MRI diffusion measures reveal increased transverse diffusivity of water in multiple sclerosis. Neuroimage 32, 1127-33. MacKay A., Whittall K., Adler J., Li D., Paty D., Graeb D., 1994. In vivo visualization of myelin water in brain by magnetic resonance. Magnetic Resonance in Medicine 31, 673-7. Mädler B., Whittall K., MacKay A., Multiple b-value Aquisition of in-vivo Diffusion Tensor Images in Human White Matter Brain. European Experimental NMR Conference; 2000; Leipzig. Mädler B., Whittall K.P., MacKay A.L., Correlation of multicomponent T2- relaxation data with diffusion tensor anisotropy measures in human brain. International Society of Magnetic Resonance in Medicine; 2002; Honolulu, Hawaii. p 1188. Mainero C., De Stefano N., Iannucci G., Sormani M.P., Guidi L., Federico A., Bartolozzi M.L., Comi G., Filippi M., 2001. Correlates of MS disability assessed in vivo using aggregates of MR quantities. Neurology 56, 1331-1334. Menon R.S., Rusinko M.S., Allen P.S., 1992. Proton relaxation studies of water compartmentalization in a model neurological system. Magn Reson Med. 28, 264- 74. Mitchell M., Kundel H., Axel L., Joseph P.M., 1986. Agarose as tissue equivalent phantom material for NMR imaging. Magnetic Resonance Imaging 4, 263-266. Moore G.R.W., 1998. Neuropathology and Pathophysiology of the Multiple Sclerosis Lesion. In: Paty D., Ebers G.C., editors. Multiple Sclerosis. Philadelphia: F.A. Davis. Moore G.R.W., Leung E., MacKay A.L., Vavasour I.M., Whittall K.P., Cover K.S., Li D.K., Hashimoto S.A., Oger J., Sprinkle T.J. and others, 2000. A pathology-MRI study of the short-T2 component in formalin-fixed multiple sclerosis brain. Neurology 55, 1506-10. Moseley M.E., Cohen Y., Kucharczyk J., Mintorovitch J., Asgari H.S., Wendland M.F., Tsuruda J., Norman D., 1990. Diffusion-weighted MR imaging of anisotropic water diffusion in cat central nervous system. Radiology 176, 439-45. Neeb H., Zilles K., Shah N.J., 2006. A new method for fast quantitative mapping of absolute water content in vivo. Neuroimage 31, 1156-68. Nusbaum A.O., Tang C.Y., Wei T.C., Buchsbaum M.S., Atlas S.W., 2000. Whole-brain diffusion MR histograms differ between MS subtypes. Neurology 54, 1421-1427. Oakden W., Lobaugh N.J., Black S.E., Stanisz G.J., T2 hyperintensity does not always mean demyelination (and neither does decreased fractional anisotropy). Imaging Myelin: Formation, Destruction and Repair; 2006; Vancouver, Canada. p P41. Oh J., Han E.T., Pelletier D., Nelson S.J., 2006. Measurement of in vivo multi- component T2 relaxation times for brain tissue using multi-slice T2 prep at 1.5 and 3 T. Magn Reson Imaging. 24, 33-43. Epub 2005 Dec 19. Oouchi H., Yamada K., Sakai K., Kizu O., Kubota T., Ito H., Nishimura T., 2007. Diffusion anisotropy measurement of brain white matter is affected by voxel size: underestimation occurs in areas with crossing fibers. AJNR Am J Neuroradiol 28, 1102-6.  57 Poon C.S., Henkelman R.M., 1992. Practical T2 quantitation for clinical applications. Journal of Magnetic Resonance Imaging. 2, 541-53. Rumbach L., Armspach J.P., Gounot D., Namer I.J., Chambron J., Warter J.M., Collard M., 1991. Nuclear magnetic resonance T2 relaxation times in multiple sclerosis. J Neurol Sci. 104, 176-81. Sirrs S.M., Laule C., Maedler B., Brief E.E., Tahir S.A., Bishop C., MacKay A.L., 2007. Normal appearing white matter in subjects with phenylketonuria: water content, myelin water fraction, and metabolite concentrations. Radiology 242, 236-243. Song S.K., Sun S.W., Ju W.K., Lin S.J., Cross A.H., Neufeld A.H., 2003. Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage. 20, 1714-22. Song S.K., Sun S.W., Ramsbottom M.J., Chang C., Russell J., Cross A.H., 2002. Dysmyelination revealed through MRI as increased radial (but unchanged axial) diffusion of water. Neuroimage. 17, 1429-36. Song S.K., Yoshino J., Le T.Q., Lin S.J., Sun S.W., Cross A.H., Armstrong R.C., 2005. Demyelination increases radial diffusivity in corpus callosum of mouse brain. Neuroimage. 26, 132-40. Sun S.W., Liang H.F., Le T.Q., Armstrong R.C., Cross A.H., Song S.K., 2006a. Differential sensitivity of in vivo and ex vivo diffusion tensor imaging to evolving optic nerve injury in mice with retinal ischemia. Neuroimage. 32, 1195-204. Epub 2006 Jun 22. Sun S.W., Liang H.F., Trinkaus K., Cross A.H., Armstrong R.C., Song S.K., 2006b. Noninvasive detection of cuprizone induced axonal damage and demyelination in the mouse corpus callosum. Magn Reson Med. 55, 302-8. Thomalla G., Glauche V., Koch M.A., Beaulieu C., Weiller C., Rother J., 2004. Diffusion tensor imaging detects early Wallerian degeneration of the pyramidal tract after ischemic stroke. Neuroimage 22, 1767-74. Thompson A.J., Miller D., Youl B., MacManus D., Moore S., Kingsley D., Kendall B., Feinstein A., McDonald W.I., 1992. Serial gadolinium-enhanced MRI in relapsing/remitting multiple sclerosis of varying disease duration. Neurology 42, 60-3. Thomsen C., Henriksen O., Ring P., 1987. In vivo measurement of water self diffusion in the human brain by magnetic resonance imaging. Acta Radiol 28, 353-61. Tofts P.S., editor. 2003. Quantitative MRI of the brain : measuring changes caused by disease. Chapter 9, page 316. Chichester, West Sussex ; Hoboken, NJ: Wiley. Tozer D.J., Davies G.R., Altmann D.R., Miller D.H., Tofts P.S., 2005. Correlation of apparent myelin measures obtained in multiple sclerosis patients and controls from magnetization transfer and multicompartmental T2 analysis. Magn Reson Med. 53, 1415-22. Tyszka J.M., Readhead C., Bearer E.L., Pautler R.G., Jacobs R.E., 2006. Statistical diffusion tensor histology reveals regional dysmyelination effects in the shiverer mouse mutant. Neuroimage 29, 1058-65. van Waesberghe J.H., Kamphorst W., De Groot C.J., van Walderveen M.A., Castelijns J.A., Ravid R., Lycklama a Nijeholt G.J., van der Valk P., Polman C.H., Thompson A.J. and others, 1999. Axonal loss in multiple sclerosis lesions:  58  59 magnetic resonance imaging insights into substrates of disability. Annals of Neurology. 46, 747-54. van Walderveen M.A., Kamphorst W., Scheltens P., van Waesberghe J.H., Ravid R., Valk J., Polman C.H., Barkhof F., 1998. Histopathologic correlate of hypointense lesions on T1-weighted spin-echo MRI in multiple sclerosis. Neurology. 50, 1282-8. Vavasour I.M., 1998. Magnetic Resonance of Human and Bovine Brain [Doctoral]. Vancouver: University of British Columbia. 120 p. Vavasour I.M., Clark C.M., Li D.K., Mackay A.L., 2006. Reproducibility and reliability of MR measurements in white matter: clinical implications. Neuroimage. 32, 637- 42. Epub 2006 May 3. Vavasour I.M., Laule C., Kolind S.H., Li D.K.B., Traboulsee A.L., MacKay A.L., Using relaxation, magnetization transfer and diffusion to characterise multiple sclerosis lesion pathology. 15th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2007; Berlin, Germany. p 272. Webb S., Munro C.A., Midha R., Stanisz G.J., 2003. Is multicomponent T2 a good measure of myelin content in peripheral nerve? Magnetic Resonance in Medicine. 49, 638-45. Werring D.J., Clark C.A., Barker G.J., Thompson A.J., Miller D.H., 1999. Diffusion tensor imaging of lesions and normal-appearing white matter in multiple sclerosis. Neurology 52, 1626-1626. Whittall K.P., Bronskill M.J., Henkelman R.M., 1991. Investigation of analysis techniques for complicated NMR relaxation data. Journal of magnetic resonance 95, 221-234. Whittall K.P., MacKay A.L., 1989. Quantitative interpretation of nmr relaxation data. Journal of Magnetic Resonance 84, 64-71. Whittall K.P., MacKay A.L., Graeb D.A., Nugent R.A., Li D.K., Paty D.W., 1997. In vivo measurement of T2 distributions and water contents in normal human brain. Magnetic Resonance in Medicine 37, 34-43. Wu Y.-C., Alexander A.L., Duncan I.D., Field A.S., Hybrid Diffusion Imaging in a Brain Model of Dysmyelination. 15th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2007; Berlin, GER. p 318. Wu Y., Alexander A.L., Fleming J.O., Duncan I.D., Field A.S., 2006. Myelin water fraction in human cervical spinal cord in vivo. J Comput Assist Tomogr. 30, 304- 6. Xue R., van Zijl P.C.M., Crain B.J., Solaiyappan M., Mori S., 1999. In vivo three- dimensional reconstruction of rat brain axonal projections by diffusion tensor imaging. Magnetic Resonance in Medicine 42, 1123-1127.    Chapter 3 3 IMPLEMENTATION AND DEVELOPMENT OF MULTI-COMPONENT T2 RELAXATION IMAGING AT 3.0 T, AND VALIDATION AGAINST 1.5 T MEASUREMENTS*  3.1 INTRODUCTION It has long been a goal of magnetic resonance imaging (MRI) to obtain a quantitative measure of myelin in order to better understand the natural development of the central nervous system (CNS) as well as to study an array of neurological diseases affecting myelin.  While there are several MR techniques used to investigate myelin content or integrity, so far no MR measure has been shown to relate explicitly to myelin content. For instance, while diffusion tensor imaging measures reflect changes in myelination, they are affected by the degree of fiber tract orientational order and packing properties of macroscopically large fiber bundles.  Magnetization transfer measures are very sensitive to tissue damage due to myelin loss, but are highly influenced by other factors such as inflammation and axonal loss; and proton magnetic resonance spectroscopy is capable of detecting active demyelination but is not able to assess intact myelin.  A technique for studying myelin not yet in common usage but rapidly gaining in popularity is multi- component T2 relaxation, which is capable of resolving the various water reservoirs present within an image voxel based on the respective T2 relaxation times of the water  * A version of this chapter has been submitted for publication. Kolind SH, Mädler B, Fischer S, Li DKB, MacKay AL. Myelin water imaging: Implementation and development at 3.0 T, and validation against 1.5 T measurements.  60 protons in different microscopic environments (compartments).  In normal human white matter, the shortest T2 component (~20 ms) is attributed to water trapped between the myelin bilayers and an intermediate T2 component (~80 ms) is thought to arise from intra/extracellular (IE) water (MacKay et al., 1994; Whittall et al., 1997).  The myelin water fraction (MWF) is the ratio of the short T2 (or myelin water) signal to the total signal in the T2 distribution.  MWF has been shown to be very highly correlated with histological measures of myelin in rats (Odrobina et al., 2005; Pun et al., 2005; Stanisz et al., 2004; Webb et al., 2003), guinea pigs (Gareau et al., 1999; Gareau et al., 2000; Stewart et al., 1993), and formalin-fixed human brains (Laule et al., 2008; Laule et al., 2006; Moore et al., 2000).  Other measures that can be extracted from multi-component T2 relaxation include the geometric mean T2 ( 2T , analogous to the amplitude-weighted mean, but on a logarithmic scale) and peak width of the IE peak.  Changes in the size of the IE peak and its 2T  time indicate changes in the intra and extracellular water environments, and consideration of both MWF and 2T  has been demonstrated to aid in distinguishing between demyelination and inflammation (Stanisz et al., 2004).  The width of the IE peak provides information regarding the homogeneity of the intra and extracellular environments (Stewart et al., 1993).  The total signal in the T2 distribution is proportional to the total water content in the tissue and, if a water reference standard is included in the image, the total signal can be used to estimate the absolute water content (Whittall et al., 1997).   61 The approach most often used to collect multi-component T2 relaxation data was initially developed by Poon and Henkelman (Poon and Henkelman 1992), and consists of a single slice multi-echo pulse sequence, utilizing large gradient crushers of alternating polarity and composite radiofrequency (RF) block pulses.  It is critical to achieve accurate 180º rotations (in the presence of inhomogeneous B1 and B0 fields), and to minimize any contribution from stimulated echoes and signal excited outside of the selected slice. Some further challenges in implementing this technique include the need for a high signal-to-noise ratio (SNR) and short echo spacing.  In recent years, a number of groups have published results of myelin water imaging in- vivo (Chia et al., 2006; Du et al., 2007; Dula et al., 2006; MacKay et al., 1994; Oakden et al., 2006; Oh et al., 2006; Tozer et al., 2005; Vermathen et al., 2007; Vidarsson et al., 2005; Wu et al., 2006); however, the majority of these studies were carried out at 1.5 T. Higher field strength provides increased SNR, and it has been reported (Graham et al., 1996) that a high SNR (noise standard deviation less than 1% of the signal strength at the shortest echo time) is required for accurate extraction of the multi-exponential components of the T2 relaxation.  Also, for multi-component T2 relaxation to be more widely applied clinically it must be feasible at higher magnetic field strengths as more and more sites move to higher field MRI scanners.  Unfortunately, T2 relaxation imaging has additional challenges at higher field strength, such as increased radio frequency field inhomogeneity, as well as reduced observed T2 times.   62 The goal of this study was to implement and refine a multi-echo T2 relaxation pulse sequence at 3.0 T based on the Poon-Henkelman technique, and to validate results against accepted values in healthy in-vivo human brain at 1.5 T.  3.2 MATERIALS AND METHODS 3.2.1 Pulse sequence development at 3.0 T The Poon-Henkelman multi-echo T2 pulse sequence was implemented on a Philips Intera 3.0 T MRI scanner (Best, The Netherlands).  This was achieved by modifying a single- slice multi-echo Carr-Purcell-Meiboom-Gill (CPMG) (Carr and Purcell 1954; Meiboom and Gill 1958) pulse sequence to (1) include 32 echoes in the echo train, (2) use composite block pulses (90ox-180oy-90ox) for the refocusing pulses following the 90o slice-selective pulse, and (3) apply z-axis gradient crushers on either side of the refocusing pulses which decrease linearly in strength and alternate in direction.  A number of sequence parameters were varied from default values to look for potential sequence improvements.  Parameter changes that required pulse programming were maximizing the area under the crushers (including changing the slew rate), and changing the position of the phase-encoding gradients from before the first refocusing pulse to after it (then rewinding and reapplying it for each refocusing pulse). Parameters that could be altered from the user-interface of the scanner included image bandwidth, echo spacing (TE), matrix size and repetition time (TR).   63 3.2.2 MR data acquisition Initial tests of the multi-component T2 relaxation technique at 3.0 T were performed on 5 MnCl2-doped agarose water reference phantoms with various T1 and T2 values (Mitchell et al., 1986), as well as (in a separate scanning session) on a formalin-fixed brain sample hemisected along the midsagittal plane from a patient with clinically definite multiple sclerosis (MS).  After sagittal and axial localizer scans, T2 relaxation measurements were acquired for a single axial slice through the centre of the water phantoms, and through the base of the genu and splenium of the corpus callosum for the fixed brain.  Each scan used a slice thickness of 5 mm, a field of view (FOV) of 22 cm, 2 averages, and a birdcage design, quadrature transmit/receive (T/R) head coil.  Parameters that were varied individually (while other parameters were held constant) were: gradient crusher area (between 0% and 100% of maximum, with the maximum being 21 mT/m for a duration of 1.95 ms), echo spacing (between 7.5 ms and 30 ms), positioning of the phase encoding gradients (fixed-brain only), crusher slew rate (default of 42 mT/m/ms and maximum of 100 mT/m/ms), and image bandwidth (between 57 kHz and 132 kHz).  The default scan was also performed twice in order to obtain a measure of scan-rescan reproducibility.  The above tests were next conducted in-vivo; T2 relaxation measurements were performed for a single transverse slice through the base of the genu and splenium for 7 healthy volunteers (4 female, 3 male, mean age 29 yrs (range 21 yrs – 37 yrs)) using the same constant parameters that were used for the phantom and formalin-fixed brain tests. For each subject, one parameter was varied while others were held constant.  The parameters that were varied were the same as for the phantom studies, plus the  64 acquisition matrix size and TR.  Five more healthy volunteers (3 female, 2 male, mean age 24 yrs (range 20 yrs – 29 yrs)) were subsequently scanned twice with the identical sequence in a single scanning session to characterize reproducibility within a single person.  In order to validate the technique at 3.0 T, 10 healthy volunteers (4 female, 6 male, mean age 23 yrs (range 19 yrs – 29 yrs)) were scanned first on a GE Signa 1.5 T MRI scanner operating at the Epic 5.7 software level (Whittall et al., 1997), and then on a Philips Intera 3.0 T scanner operating at software level 10.4.  On both systems, after sagittal and axial localizer scans, T2 relaxation measurements were acquired for a single transverse slice through the base of the genu and splenium of the corpus callosum using a 32-echo sequence (BW = 69 kHz, TR = 3000 ms, 256x128 matrix, 10 ms echo spacing, slice thickness = 5 mm, FOV = 22 cm, 4 averages, T/R head coil, crusher gradient amplitude 21 mT/m and duration 1.95 ms with a slew rate of 42 mT/m/ms and phase encoding gradients before the refocusing train).  While the images were not registered to each other because they were only single-slice data sets, care was taken in positioning on the 3.0 T scanner to match the slice scanned on the 1.5 T scanner as closely as possible.  10 additional healthy volunteers (5 female, 5 male, mean age 27 yrs (range 20 yrs – 58 yrs)) were also scanned on the Philips 3.0 T scanner with a 6-element phased-array coil instead of the T/R head coil, and the constant level appearance (CLEAR) algorithm for signal intensity homogeneity correction was utilized to compensate for inhomogeneous receiver sensitivities.  65  3.2.3 MR data analysis For the water phantoms, regions of interest (ROIs) were drawn in each bottle on the first echo image from the T2 relaxation data set.  For the fixed brain and in-vivo data, ROIs were outlined on the first echo image from the T2 relaxation data set in grey matter (GM): cingulate gyrus, insular cortex, putamen, head of the caudate nucleus and thalamus, and white matter (WM): minor forceps, major forceps, genu and splenium of the corpus callosum, and internal capsules (in-vivo ROIs were outlined in both hemispheres and results were averaged).  The decay curves resulting from the measured multi-echo relaxation signal could be described using the following general equation (Whittall and MacKay 1989):   (2.1) NiTtsy M j jiji ,...,2,1,)/exp( 1 2 =−= ∑ = where ti are the measured times, M = 120 logarithmically spaced T2 times within the range of 15 ms to 2 s (except where the echo spacing was 8 ms, in which case the range of T2 times was from 12 ms to 2 s), N = 32 is the total number of data points in the decay curve (number of echoes), and sj is the relative amplitude for each partitioned T2 time.  A non-negative least squares (NNLS) algorithm was used to minimize both χ2 and an energy constraint that smoothes the T2 distribution, sj(T2j ), providing better, consistent fits in the presence of noise (Fenrich et al., 2001; Lawson and Hanson 1974; Whittall and MacKay 1989).  The following expression was minimized:   (2.2) ,0, 1 22 ≥+ ∑ = μμχ M j js  66 where the larger the parameter μ becomes, the more the T2 distribution is smoothed at the cost of misfit.  The χ2min fit results for the case that μ = 0.  Regularized smooth T2 distributions were created by minimizing equation 2.2 with the energy constraint of 1.02χ2min ≤ χ2 ≤ 1.025χ2min.  It should be noted that the parameter μ affects peak width, but the same energy constraint was used for all analyses in this study; therefore peak width results can be safely compared within this paper, but caution should be used for comparison to other studies.  The peak assigned to myelin water was defined as having 15 ms < T2 < 50 ms for data obtained at 1.5 T and 15 ms < T2 < 40 ms at 3.0 T, and the IE water peak was defined as having 50 ms < T2 < 200 ms and 40 ms < T2 < 200 ms for 1.5 T and 3.0 T respectively.  The shift of the T2 boundaries for both myelin and IE water at the higher field strength is due to apparent shortening of T2 values in brain as a result of increased sensitivity to spatial variations in magnetic susceptibility; boundary values were determined by examining the T2 distributions to establish where the IE peak began, and the change was in accordance with expected changes in observed T2 times for in-vivo brain between 1.5 T and 3.0 T (Gelman et al., 1999; Pell et al., 2004; Sled and Pike 2001; Stanisz et al., 2005).  MWF was defined as the ratio between myelin water peak area to total signal in the T2 distribution.   For the IE peak, the geometric mean T2 (given by:  ,)log(exp max min max min 22 ⎥⎦ ⎤⎢⎣ ⎡= ∑∑ == M Mj j M Mj jj sTsT  (2.3) where Mmin and Mmax represent the range used for the IE T2 component) and the peak width (analogous to the variance, but on a logarithmic scale, providing information regarding the homogeneity of the T2 times of the IE water compartment), given by:  67  1logexp max min max min 2 2 2 −⎥⎥⎦ ⎤ ⎢⎢⎣ ⎡ ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛ ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛= ∑∑ == M Mj j M Mj j j sT T sWidth , (2.4) were calculated (Whittall et al., 1997).  For the water bottles, 2T  was calculated using a range of T2 values from 15 ms to 2 s.  The SNR was calculated as the extrapolated signal intensity at TE = 0 ms (from NNLS) divided by the standard deviation of the signal from air in the background that is not affected by spurious signal from flow or ghosting after dividing by 0.66 to obtain the Gaussian-noise standard deviation (Constantinides et al., 1997; Henkelman 1985).  The size of the ROI drawn in the background of the image was matched to the size of the ROI in brain.  This analysis becomes complicated for magnitude images from a phased-array coil using CLEAR because the reconstructed component coil signals that contribute to the magnitude sum do not have identical noise power; a rigorous analysis of the statistical properties of the noise for such images does not fit the purpose of this report.  A measure of noise more appropriate for T2 relaxation data is the standard deviation of the residuals of the multi-component T2 NNLS fit, which was normalized by dividing by the theoretical amplitude at TE  = 0 ms.  The coefficient of variation for the standard deviation of residuals and MWF values was calculated between scans with identical parameters.  For the in-vivo data, the coefficient of variation was averaged over 5 volunteers.  MWF maps were created by calculating the MWF value at each pixel in the image.   68 Statistical analysis was carried out using a two-tailed Student’s t test with a p-value of < 0.05 considered significant.  All errors are expressed as standard errors (SE). The Pearson correlation coefficient (R2) was calculated between different data sets.  3.3 RESULTS 3.3.1 3.0 T multi-exponential T2 relaxation pulse sequence refinement The standard deviation of residuals (as a percentage of the initial signal), a measure of how well the data can be fit with a multi-exponential function, is shown in Figure 3.1 for each of the scans in which a single parameter was varied for the water-based phantoms, fixed brain, and in-vivo with error bars indicating standard deviation.  Results are grouped with respect to different variable parameters.  For the water-based phantoms, the average value in all 5 phantoms was used.  In the case of the fixed brain, the WM and GM values were averaged over all 5 ROIs in each of white and grey matter.  The same approach was used for the in-vivo data, except that the values were averaged for both brain hemispheres.  Scans for which the percentage difference in the standard deviation of residuals from the default scan was greater than twice the coefficient of variation for the scan-rescan data are indicated with an asterisk.  In the water-based phantoms, the crusher area had no significant effect on the standard deviation of residuals, but a larger bandwidth or smaller echo spacing increased the standard deviation of residuals.  For fixed brain, decreasing the crusher area increased the standard deviation of residuals in most cases, as did decreasing the echo spacing or applying phase rewinding.  In-vivo, the only parameters to measurably increase the standard deviation of residuals were phase rewinding, decreasing the echo spacing and increasing the slew rate.  69 0.0% 0.2% 0.4% 0.6% 0.8% BW  4 2k Hz BW  6 9k Hz Cr us he r A re a 10 0% Cr us he r A re a 72 % Cr us he r A re a 1% M at rix  2 56 x1 28 M at rix  1 28 x1 28 Ph as e En co de  B ef or e Ph as e En co de  A fte r De fa ul t S le w Ra te M ax  S le w Ra te Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 8m s TR  2 13 9m s TR  3 00 0m s TR  4 00 0m s TR  5 00 0m s 0.0% 0.1% 0.2% 0.3% 0.4% BW  4 2k Hz BW  6 9k Hz Cr us he r A re a 10 0% Cr us he r A re a 73 % Cr us he r A re a 60 % Cr us he r A re a 33 % Cr us he r A re a 0% Ph as e En co de  B ef or e Ph as e En co de  A fte r De fa ult  S le w Ra te M ax  S lew  R at e Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 8m s 0.0% 0.1% 0.2% 0.3% 0.4% BW  5 7k Hz BW  6 5k Hz BW  1 32 kH z Cr us he r A re a 10 0% Cr us he r A re a 90 % Cr us he r A re a 80 % Cr us he r A re a 75 % Cr us he r A re a 50 % Cr us he r A re a 25 % Cr us he r A re a 10 % Cr us he r A re a 0% De fa ult  S lew  R at e M ax  S lew  R at e Ec ho  S pa cin g 30 m s Ec ho  S pa cin g 20 m s Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 9m s Ec ho  S pa cin g 8m s Ec ho  S pa cin g 7. 5m s White Matter Grey Matter * * * * In -v iv o B ra in  R es id ua ls W at er -b as ed  P ha nt om  R es id ua ls * * * * * White Matter Grey Matter * * * * * * * * * Fi xe d B ra in  R es id ua ls In -v iv o B ra in  R es id ua ls W at er -b as ed  P ha nt om  R es id ua ls Fi xe d B ra in  R es id ua ls  Figure 3.1: Standard deviation of residuals (as a percentage of initial signal) for multi-component T2 relaxation data taken at 3.0 T for scans in which a single parameter was varied for water-based phantoms with various T1 and T2 times (top), fixed brain (middle), and in-vivo (bottom). Error bars indicate standard deviation. Asterisks indicate scans for which the percentage difference from the default scan was larger than twice the scan-rescan coefficient of variation.  70  Figure 3.2 shows the MWF for fixed brain and in-vivo data for the various changes in scan parameters with error bars indicating standard deviation.  Asterisks indicate scans with MWF values different by more than twice the scan-rescan MWF coefficient of variation.  While most parameters had a large effect on MWF values in fixed brain, only phase rewinding had a measurable impact on MWF values in-vivo. Fi xe d Br ai n M W F In -v iv o B ra in  M W F 0% 5% 10% 15% 20% 25% BW  4 2k Hz BW  6 9k Hz Cr us he r A re a 10 0% Cr us he r A re a 72 % Cr us he r A re a 1% M at rix  2 56 x1 28 M at rix  1 28 x1 28 Ph as e En co de  B ef or e Ph as e En co de  A fte r De fa ul t S le w Ra te M ax  S le w Ra te Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 8m s TR  2 13 9m s TR  3 00 0m s TR  4 00 0m s TR  5 00 0m s White Matter Grey Matter * * 0% 20% 40% 60% 80% BW  4 2k Hz BW  6 9k Hz Cr us he r A re a 10 0% Cr us he r A re a 73 % Cr us he r A re a 60 % Cr us he r A re a 33 % Cr us he r A re a 0% Ph as e En co de  B ef or e Ph as e En co de  A fte r De fa ult  S le w Ra te M ax  S lew  R at e Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 8m s White Matter Grey Matter * * * * * * * Fi xe d Br ai n M W F In -v iv o B ra in  M W F  Figure 3.2: MWF values at 3.0 T for scans in which a single parameter was varied for fixed brain (top), and in-vivo (bottom). Error bars indicate standard deviation. Asterisks indicate scans for which the percentage difference from the default scan was larger than twice the scan-rescan coefficient of variation.   71 025 50 75 100 BW  4 2k Hz BW  6 9k Hz Cr us he r A re a 10 0% Cr us he r A re a 72 % Cr us he r A re a 1% M at rix  2 56 x1 28 M at rix  1 28 x1 28 Ph as e En co de  B ef or e Ph as e En co de  A fte r De fa ult  S lew  R at e M ax  S lew  R at e Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 8m s TR  2 13 9m s TR  3 00 0m s TR  4 00 0m s TR  5 00 0m s 0 25 50 75 100 BW  4 2k Hz BW  6 9k Hz Cr us he r A re a 10 0% Cr us he r A re a 73 % Cr us he r A re a 60 % Cr us he r A re a 33 % Cr us he r A re a 0% Ph as e En co de  B ef or e Ph as e En co de  A fte r De fa ult  S le w Ra te M ax  S le w Ra te Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 8m s 0 25 50 75 100 BW  5 7k Hz BW  6 5k Hz BW  1 32 kH z Cr us he r A re a 10 0% Cr us he r A re a 90 % Cr us he r A re a 80 % Cr us he r A re a 75 % Cr us he r A re a 50 % Cr us he r A re a 25 % Cr us he r A re a 10 % Cr us he r A re a 0% De fa ul t S le w Ra te M ax  S le w Ra te Ec ho  S pa cin g 30 m s Ec ho  S pa cin g 20 m s Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 9m s Ec ho  S pa cin g 8m s Ec ho  S pa cin g 7. 5m s White Matter Grey Matter* In -v iv o B ra in  T 2 (m s) W at er -b as ed  P ha nt om  T 2 (m s) White Matter Grey Matter Fi xe d B ra in  T 2 (m s) In -v iv o B ra in  T 2 (m s) W at er -b as ed  P ha nt om  T 2 (m s) Fi xe d B ra in  T 2 (m s)  Figure 3.3: 2T  for multi-component T2 relaxation data taken at 3.0 T for scans in which a single parameter was varied for water-based phantoms (top), fixed brain (middle), and in-vivo (bottom). Error bars indicate standard deviation. Asterisks indicate scans for which the percentage difference from the default scan was larger than twice the scan-rescan coefficient of variation.  72 The 2T  values for one of the water bottle phantoms, fixed brain and in-vivo brain are displayed in Figure 3.3.  None of the parameters had a significant effect on the measured 2T  values for the water-based phantom or fixed brain, but the smallest crushers led to larger measured 2T  values for GM in vivo. Figure 3.4 depicts the peak width of the IE peak for fixed brain and in-vivo brain.  The peak widths were quite variable, and no significant differences were found between peak widths for the in-vivo data.  In fixed brain, increasing the bandwidth, decreasing the gradient crusher area, increasing the gradient slew rate or applying phase-rewinding all decreased the peak width.  73 025 50 75 100 125 150 BW  4 2k Hz BW  6 9k Hz Cr us he r A re a 10 0% Cr us he r A re a 72 % Cr us he r A re a 1% M at rix  2 56 x1 28 M at rix  1 28 x1 28 Ph as e En co de  B ef or e Ph as e En co de  A fte r De fa ul t S lew  R at e M ax  S lew  R at e Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 8m s TR  2 13 9m s TR  3 00 0m s TR  4 00 0m s TR  5 00 0m s 0 50 100 150 200 250 300 BW  4 2k Hz BW  6 9k Hz Cr us he r A re a 10 0% Cr us he r A re a 73 % Cr us he r A re a 60 % Cr us he r A re a 33 % Cr us he r A re a 0% Ph as e En co de  B ef or e Ph as e En co de  A fte r De fa ul t S le w Ra te M ax  S le w Ra te Ec ho  S pa cin g 10 m s Ec ho  S pa cin g 8m s Fi xe d B ra in  P ea k W id th  (m s) In -v iv o B ra in  P ea k W id th  (m s) White Matter Grey Matter White Matter Grey Matter * * * * * * * * Fi xe d B ra in  P ea k W id th  (m s) In -v iv o B ra in  P ea k W id th  (m s)  Figure 3.4: Peak widths for the IE peak at 3.0 T for scans in which a single parameter was varied for fixed brain (top), and in-vivo (bottom). Error bars are standard deviation. Asterisks indicate scans for which the percentage difference from the default scan was larger than twice the scan-rescan coefficient of variation.  Examples of residuals of the multi-exponential fit to the decay curves (expressed as a percentage of the initial signal) are shown in Figure 3.5.  Examples are given for large and small gradient crushers (100% and 33% of the maximum gradient crusher area) in the same ROI for fixed brain, with larger crusher gradients resulting in much smaller residuals for the first few echoes, and for two different ROIs (the minor forceps and  74 internal capsules) for in-vivo brain, where the residuals in the minor forceps decreased rapidly while those in the internal capsules were large throughout the echo train. -1.0% -0.5% 0.0% 0.5% 1.0% 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 Time (s) M ul ti- ex po ne nt ia l f it re si du al  (p er ce nt ag e of  in iti al  si gn al ) Crusher Area 100% Crusher Area 33% Changing gradient crusher area in fixed brain -1.0% -0.5% 0.0% 0.5% 1.0% 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 Time (s) M ul ti- ex po ne nt ia l f it re si du al (p er ce nt ag e of  in iti al  si gn al ) Internal Capsules Minor Forceps Different brain regions in-vivo M ul ti- ex po ne nt ia l f it re si du al  (p er ce nt ag e of  in iti al  si gn al ) M ul ti- ex po ne nt ia l f it re si du al (p er ce nt ag e of  in iti al  si gn al ) M ul ti- ex po ne nt ia l f it re si du al (p er ce nt ag e of  in iti al  si gn al )  Figure 3.5: Residuals of the multi-exponential fit expressed in percentage of the initial signal along the decay train for (above) fixed brain using a crusher area of 100% and 33% of the maximum, and (below) in- vivo brain in the posterior internal capsules and minor forceps.  3.3.2 Comparison between multi-exponential T2 relaxation at 1.5 T and 3.0 T Representative MWF maps for one volunteer at 1.5 T and 3.0 T are demonstrated in Figure 3.6, along with the corresponding proton-density weighted image.  The maps show good qualitative agreement.  75 First echo (TE=10ms) of T2 relaxation data set MWF map at 1.5T MWF map at 3.0T  Figure 3.6: MWF maps at 1.5 T and 3.0 T for one volunteer.  The proton-density weighted image (first echo image of the T2 relaxation data set, TE=10ms) is also shown.  Figure 3.7 shows the correlations between MWF at 1.5 T and 3.0 T, both with the T/R head coil and the phased-array head coil.  Both correlations were highly significant (R2 = 0.96, p < 0.0005 for T/R coil and R2 = 0.92, p < 0.0005 for phased-array coil).  While the relationship with 1.5 T data was closer to unity for the 3.0 T phased-array coil data (slope = 1.06, intercept = 2.1%) than for the 3.0 T T/R coil (slope = 1.10, intercept = 4.9%), R2 was slightly smaller (R2 = 0.92 for the phased-array coil and R2 = 0.96 for the T/R coil). Correlations were still strong when only WM ROIs were included (R2 = 0.92, p = 0.008 for T/R coil and R2 = 0.87, p = 0.02 for phased-array coil).  A bar chart of the MWF values at 1.5 T and 3.0 T (with both head coils) is also shown, indicating the brain structures where MWF was not significantly different between the two head coils at 3.0 T, which consisted of most of the WM structures.  76 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 0.0% 3.0% 6.0% 9.0% 12.0% 15.0% 1.5T MWF 3. 0T  M W F us in g T /R  c oi l 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 0.0% 3.0% 6.0% 9.0% 12.0% 15.0% 1.5T MWF 3. 0T  M W F us in g ph as ed -a rr ay  c oi l Cingulate Gyrus Putamen Caudate Insular Cortex Thalamus Minor Forceps Major Forceps Genu Splenium Internal Capsules 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% Ci ng ula te Gy rus Pu tam en Ca ud ate Ins ula r C ort ex Th ala mu s Mi no r F orc ep s Ma jor  Fo rce ps Ge nu Sp len ium Int ern al Ca psu les M W F 1.5T 3.0T using phased-array coil 3.0T using T/R coil 3.0T MWF = 1.10•(1.5T MWF) + 4.9% R2 = 0.96, p < 0.0005 3.0T MWF = 1.06•(1.5T MWF) + 2.1% R2 = 0.92, p < 0.0005 * * * * * 3. 0T  M W F us in g T /R  c oi l 3. 0T  M W F us in g ph as ed -a rr ay  c oi l M W F  Figure 3.7: Correlations between average MWF at 1.5 T and 3.0 T using the T/R head coil (top left) or the phased-array head coil (top right).  The equation of the line, R2 and p-value are indicated on the graphs.  On the bottom left is a bar chart of MWF in each brain structure at 1.5 T and 3.0 T using both head coils. MWF values were significantly different between all 3 cases (p<0.005), except for the brain regions indicated with a * where the values obtained at 3.0 T were not significantly different between the phased-array coil and the T/R coil.  Error bars represent standard error.  Table 3.1 gives the average 2T  and peak width values for all 10 volunteers using 1.5 T (with a T/R head coil), 3.0 T with the T/R head coil, and 3.0 T with the phased-array head coil.  Significant differences (p<0.05) are indicated.  2T  was significantly different for all structures between field strengths.  Peak widths were generally larger at 3.0 T, particularly in GM, though not significantly so for most structures.  77 Cingulate Gyrus 95.4  (1.1) 73.4  (1.8) 75.5  (1.1) a b 22  (5) 53  (15) 41  (9) a b Putamen 78.2  (0.6) 56.6  (0.9) 59.1  (1.4) a b c 7  (3) 8  (3) 7  (2) Caudate 84.6  (0.5) 63.2  (1.0) 66.5  (0.9) a b c 14  (6) 32  (9) 14  (3) a c Insular Cortex 95.2  (1.0) 70.0  (1.4) 75.0  (1.0) a b c 21  (11) 35  (11) 22  (8) Thalamus 79.7  (0.6) 58.8  (1.0) 64.3  (0.8) a b c 8  (3) 17  (4) 9  (4) a c Minor Forceps 76.0  (0.7) 61.3  (1.3) 64.5  (0.7) a b c 19  (6) 32  (9) 27  (4) Major Forceps 85.2  (0.9) 66.7  (1.2) 71.5  (0.7) a b c 22  (6) 48  (13) 56  (12) a b Genu 73.4  (0.6) 59.2  (0.8) 62.9  (0.9) a b c 31  (5) 46  (11) 30  (8) Splenium 86.8  (1.7) 67.0  (1.6) 73.1  (1.1) a b c 55  (10) 63  (14) 41  (14) Internal Capsules 87.7  (1.8) 66.8  (1.6) 73.8  (1.5) a b c 64  (13) 67  (11) 61  (11) ANOVA ANOVA T 2 (ms) Peak Width (ms) 1.5T 3.0T 3.0T T/R phased-array T/R phased-array 1.5T 3.0T 3.0T  Table 3.1: 2T  and peak width for the IE peak for 10 volunteers using 1.5 T, 3.0 T with a T/R head coil and 3.0 T with a phased-array head coil.  Standard error indicated in parentheses.  a = significant difference between 1.5 T and 3.0 T using the T/R coil, b = significant difference between 1.5 T and 3.0 T using the phased-array coil, and c = significant difference between 3.0 T using the T/R coil and 3.0 T using the phased-array coil, with the level of significance taken as p < 0.05.  ANOVA Cingulate Gyrus 88  (7) 152  (16) a 0.15  (0.01) 0.20  (0.03) 0.23  (0.04) b Putamen 92  (7) 169  (18) a 0.11  (0.02) 0.21  (0.03) 0.37  (0.05) a b c Caudate 92  (7) 166  (19) a 0.13  (0.01) 0.27  (0.04) 0.32  (0.06) a b Insular Cortex 87  (7) 158  (17) a 0.14  (0.01) 0.22  (0.03) 0.29  (0.04) a b c Thalamus 89  (7) 168  (17) a 0.14  (0.02) 0.21  (0.04) 0.29  (0.04) a b Minor Forceps 80  (6) 138  (15) a 0.07  (0.01) 0.08  (0.02) 0.20  (0.02) b c Major Forceps 78  (6) 138  (14) a 0.06  (0.01) 0.15  (0.03) 0.20  (0.03) a b Genu 81  (6) 150  (16) a 0.11  (0.01) 0.18  (0.03) 0.24  (0.03) a b Splenium 80  (6) 150  (15) a 0.11  (0.01) 0.22  (0.04) 0.22  (0.03) a b Internal Capsules 82  (6) 156  (16) a 0.12  (0.01) 0.19  (0.03) 0.24  (0.04) a b T/R phased-array Standard Deviation of Residuals (%) 1.5T 3.0T 3.0T ANOVA SNR 1.5T 3.0T T/R  Table 3.2: SNR and standard deviation of residuals (as a percentage of initial signal) for 10 volunteers; SNR is only indicated for data taken with a T/R head coil.  Standard error indicated in parentheses.  a = significant difference between 1.5 T and 3.0 T using the T/R coil, b = significant difference between 1.5 T and 3.0 T using the phased-array coil, and c = significant difference between 3.0 T using the T/R coil and 3.0 T using the phased-array coil, with the level of significance taken as p < 0.05.  The average SNR and standard deviation of residuals for all 10 controls are given in Table 3.2 for both field strengths.  The SNR is only given for scans with the T/R head coil.  While the SNR was nearly double at 3.0 T, this increase in signal did not translate  78 into a better multi-exponential fit of the data with NNLS; the standard deviation of residuals was mostly significantly higher at 3.0 T than at 1.5 T, with little difference between the two head coils.  3.4 DISCUSSION 3.4.1 3.0 T multi-exponential T2 relaxation pulse sequence development – residuals of the multi-exponential fit One of the most common difficulties in short-T2 component measurement is avoiding fluctuations in the echo amplitudes, particularly for the first few echoes (Vold et al., 1973).  Typical causes of a poor NNLS fit include stimulated echoes, out-of-slice signal, random noise, eddy currents, and flow.  Stimulated echoes tend to cause large fluctuations in the first few points (Hennig et al., 2003; Woessner 1961), while flow artifact or random noise will affect the entire echo train.  Figure 3.5 illustrates the effects of stimulated echoes in fixed brain (larger crushers reduced stimulated echoes resulting in smaller initial oscillations in the residuals of the first few echoes) and flow in-vivo (the posterior internal capsules (near the ventricles) showed residuals alternating in polarity along the entire echo train while the minor forceps (far from the ventricles) had small residuals for the later echoes).  For development of the multi-echo T2 relaxation sequence at 3.0 T, the standard deviation of the residuals was used as an indicator of whether or not changing a parameter decreased such noise in the echo train and thus improved our ability to fit the data to a multi-exponential function.  Examination of decay curves by means of their residuals throughout the echo train was helpful for confirming the source of the misfit.  79  Larger crusher gradients can decrease signal (increasing the residuals normalized to initial signal intensity), however, they are essential for minimizing out-of-slice signal and contributions from stimulated echoes (decreasing residuals).  For the water-based phantoms, out-of-slice signal should not negatively impact the image as the phantoms were uniform, and indeed no effect on the standard deviation of residuals was observed with crusher area in the phantoms.  In fixed-brain, decreasing the crusher area increased the standard deviation of residuals (primarily in WM), and the decay curves had the largest variations for the first few echoes in an oscillating pattern, consistent with out-of- slice signal or stimulated echoes (Poon and Henkelman 1992).  Due to the lower SNR in- vivo compared to fixed brain, no detectable effect of crusher area was observed in-vivo.  Since a larger bandwidth increases the noise level, the standard deviation of residuals (normalized to initial signal) was expected to increase with increasing bandwidth, which was the case for the water-based phantoms (see Figure 3.1).  In fixed and in-vivo brain, this effect was not large enough to detect a difference.  Decreasing the matrix size (thus increasing the size of the voxel for a constant FOV) increases the signal per voxel, which could be expected to decrease the relative amplitude of residuals, but no change was evident for in-vivo brain.  Increasing TR should decrease T1-weighting and thus could be expected to decrease the normalized standard deviation of residuals, but only a very small opposite effect was observed in-vivo.  Thus for in-vivo T2 studies at 3.0 T, decay curve fluctuations are dominated by systematic rather than random noise, and one can take  80 advantage of greater resolution (wider bandwidth and larger matrix size) and shorter scan times (shorter TR).  Applying phase rewinding (changing the timing of the phase-encoding gradient pulses from before to after the first refocusing pulse) effectively phase-encodes signal from outside of the slice, increasing misfits due to out-of-slice signal, but also phase-encodes stimulated echoes from signal within the slice, decreasing misfits due to stimulated echoes (Poon and Henkelman 1992).  Phase rewinding increased the standard deviation of residuals in WM and GM considerably in both fixed and in-vivo brain.  This implies that stimulated echoes arising from signal within the slice were less of a problem than uncrushed out-of-slice signal, and hence phase-encoding should take place before refocusing.  Decreasing the echo spacing increased the standard deviation of residuals in most cases, most likely due to eddy currents.  An echo spacing of 10 ms was found to be sufficiently long to minimize counter-productive effects on the residuals.  Increasing the gradient slew rate led to smaller standard deviations of residuals for the water-based phantoms and fixed brain, possibly because the crusher length was held constant thus the smaller crusher area may have increased signal.  In-vivo, the standard deviation of residuals increased for the maximum slew rate (100 mT/m/ms), presumably due to increased eddy current artifacts, so the default slew rate (42 mT/m/ms) was preferred.   81 The standard deviation of residuals was lowest in-vivo in regions less likely to be contaminated by flow artifact, i.e. regions that were not in-line with the ventricles along the phase-encoding direction.   Decay curves from structures near the ventricles, e.g. the posterior internal capsules, thalamus and putamen, tended to have residuals that alternated in polarity and continued throughout the echo train, characteristic of even-echo rephasing of flowing magnetization.  For fixed brain, the standard deviation of residuals was lowest for interior structures, far from the periphery of the brain, where there was less signal drop-off due to B1 inhomogeneity.  3.4.2 3.0 T multi-exponential T2 relaxation pulse sequence development – MWF, 2T  and IE peak width The MWF values for fixed brain were much higher than in-vivo MWF values (see Figure 3.2).  This apparent increase is likely due to dehydration in the fixed brain, which is expected to occur preferentially for the intra and extracellular compartments as more energy is required to extract the water trapped between the myelin bilayers.  Evidence of dehydration is also apparent from the lower normalized standard deviations of residuals (see Figure 3.1) as the RF coil was better able to tune with less water present.  For fixed brain, greatly reducing the gradient crushers resulted in smaller measured MWF values, likely due to increased oscillations in the T2 decay curve caused by out-of-slice signal and stimulated echoes.  Moving the phase-encoding gradients to after the refocusing pulse, thus phase-encoding out-of-slice signal, had the same effect, and was also apparent in-vivo.  Eddy current contributions also decreased the MWF values as  82 signal in the first few echoes was most affected when parameters that affect eddy currents were varied (e.g. slew rate).  MWF decreased with increased slew rate in fixed brain and for small echo spacings in-vivo.  2T  values were very robust in the water-based phantom and fixed brain, and the only significant difference in-vivo was for very small gradient crushes (leading to an increased measured 2T  in GM, consistent with increased stimulated echoes).  The peak widths had a wide range of values, and several significant differences for changes of parameters were found for fixed brain.  However, none of the changes in-vivo was significant.  Thus although the residuals change with scan parameters (and hence signal, noise, or artifact change in such a way that the data does not fit as closely to our model), the quantities of interest (MWF, 2T  and peak width) generally do not vary with scan parameter changes beyond the extent expected from reproducibility analysis; the larger residuals do not necessarily translate into a significant change in the fitted T2 distribution.  Since within a control subject, MWF, 2T  and peak width values were quite robust with respect to parameter changes, one may choose sequence parameters that minimize the residuals.  Parameter choices for the comparison study between 1.5 T and 3T included larger bandwidth (69 kHz), larger crusher area, larger matrix size (256x128), phase- encoding gradients placed before the refocusing pulse, 10 ms echo spacing, and shorter TR (3000 ms).   83 3.4.3 Comparison between multi-exponential T2 relaxation at 1.5 T and 3.0 T The observed reduction of 2T  at 3.0 T (Table 3.1) was presumably due to increased sensitivity to magnetic susceptibility at higher field strength.  If water diffuses through a magnetic susceptibility gradient in the time between 180º refocusing pulses, it will not be completely refocused.  Also, the presence of paramagnetic or supermagnetic (e.g. iron) particles can alter the magnetic field on a microscopic scale that may not be completely compensated by a CPMG pulse sequence (Stanisz et al., 2005).  As can be seen in Figure 3.6, MWF maps created from 1.5 T and 3.0 T data for the same volunteer are qualitatively similar, with white matter as identified in the proton-density weighted image appearing brighter on both MWF maps.  More signal was apparent on the 3.0 T map, although the region around the internal capsules appeared noisy (likely due to flow artifact).  The relationship between 1.5 T MWF values and 3.0 T MWF values obtained with a phased-array coil was closer to unity than the relationship between 1.5 T MWF and 3.0 T T/R coil MWF values (see Figure 3.7) although R2 was slightly smaller, possibly because different individuals were scanned using the phased-array coil than those scanned at 1.5 T.  The values in GM were nearly all significantly higher using the T/R coil than for the phased-array coil, with the lower values matching better with 1.5 T MWF values.  The discrepancies may have been due to an increased sensitivity to B1 field inhomogeneity using the T/R head coil.   84 Though there was a strong significant correlation between MWF values at 1.5 T and 3.0 T, MWF values were significantly higher at 3.0 T (see Figure 3.7).  Other groups that measured MWF at 3.0 T in-vivo also observed elevated MWF values; although they used a distinctly different acquisition technique, Oh et al. (Oh et al., 2006) reported very similar MWF values to this study at 3.0 T with a similar proportional increase in values over those measured at 1.5 T.  They attributed their MWF differences to increased dielectric effects causing greater B1 inhomogeneity resulting in echo train artifacts.  Our results showed that the standard deviation of residuals, which can be affected by B1 and measures the ability of NNLS to fit the data, was higher at 3.0 T than at 1.5 T (see Table 3.2).  Local B1 effects due to the iron content distribution, which is higher in GM than in WM (as illustrated by the increased effect on 2T  and peak width in GM compared to WM at 3.0 T, see Table 3.1) may also be a factor.  In conclusion, we have successfully adapted the single-slice Poon-Henkelman technique for measuring multi-component T2 relaxation from 1.5 T to 3.0 T.  For in-vivo studies a higher bandwidth, larger matrix, shorter TR, and moderate echo spacing (~10 ms) are acceptable, and default gradient slew rate, large gradient crushers, and phase-encoding before refocusing are preferred.  While the many benefits of higher magnetic field (such as increased SNR) were somewhat offset by the adverse effects of B1 inhomogeneity, MWF values at 3.0 T to 1.5 T were similar (although higher), following the same trend and resulting in qualitatively similar MWF maps.   85 3.5 ACKNOWLEDGMENTS Sincere thanks to our volunteers and the technologists of UBC hospital.  We are also grateful to Philips Medical Systems for assistance in the development of pulse programs, and Cornelia Laule for useful discussions.   The research was supported by the Canadian Institutes of Health Research, the Killam Trusts, the Natural Science and Engineering Research Council of Canada and the Multiple Sclerosis Society of Canada.   86 3.6 REFERENCES Carr H.Y., Purcell E.M., 1954. Effects of Diffusion on Free Precession in Nuclear Magnetic Resonance Experiments. Physical Review 94, 630-638. Chia C.L.L., Bjarnason T.A., Mackay A.L., Pike G.B., Cross-site reproducibility of myelin water estimates. 14th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2006; Seattle, USA. p 2520. Constantinides C.D., Atalar E., McVeigh E.R., 1997. Signal-to-noise measurements in magnitude images from NMR phased arrays. Magn Reson Med 38, 852-7. Du Y.P., Chu R., Hwang D., Brown M.S., Kleinschmidt-DeMasters B.K., Singel D., Simon J.H., 2007. Fast multislice mapping of the myelin water fraction using multicompartment analysis of T2* decay at 3T: a preliminary postmortem study. Magn Reson Med 58, 865-70. Dula A.N., Does M.D., Gochberg D.F., Imaging of Myelin in Ferret Brain using Magnetic Resonance Multi-component T2 and qMT. Imaging Myelin: Formation, Destruction and Repair; 2006; Vancouver, BC. p P39. Fenrich F.R., Beaulieu C., Allen P.S., 2001. Relaxation times and microstructures. NMR Biomed. 14, 133-9. Gareau P.J., Rutt B.K., Bowen C.V., Karlik S.J., Mitchell J.R., 1999. In vivo measurements of multi-component T2 relaxation behaviour in guinea pig brain. Magnetic Resonance Imaging. 17, 1319-25. Gareau P.J., Rutt B.K., Karlik S.J., Mitchell J.R., 2000. Magnetization transfer and multicomponent T2 relaxation measurements with histopathologic correlation in an experimental model of MS. Journal of Magnetic Resonance Imaging 11, 586- 95. Gelman N., Gorell J.M., Barker P.B., Savage R.M., Spickler E.M., Windham J.P., Knight R.A., 1999. MR imaging of human brain at 3.0 T: preliminary report on transverse relaxation rates and relation to estimated iron content. Radiology 210, 759-67. Graham S.J., Stanchev P.L., Bronskill M.J., 1996. Criteria for analysis of multicomponent tissue T2 relaxation data. Magnetic Resonance in Medicine 35, 370-8. Henkelman R.M., 1985. Measurement of signal intensities in the presence of noise in MR images. Med Phys 12, 232-3. Hennig J., Weigel M., Scheffler K., 2003. Multiecho sequences with variable refocusing flip angles: optimization of signal behavior using smooth transitions between pseudo steady states (TRAPS). Magn Reson Med 49, 527-35. Laule C., Kozlowski P., Leung E., Li D.K.B., MacKay A.L., Moore G.R.W., 2008. Myelin water imaging of multiple sclerosis at 7 T: Correlations with histopathology. Neuroimage 40, 1575-1580. Laule C., Leung E., Li D.K., Traboulsee A.L., Paty D.W., MacKay A.L., Moore G.R., 2006. Myelin water imaging in multiple sclerosis: quantitative correlations with histopathology. Mult Scler. 12, 747-53. Lawson C.L., Hanson R.J., 1974. Solving Least Squares Problems. Englewood Cliffs, NJ: Prentice-Hall,.  87 MacKay A., Whittall K., Adler J., Li D., Paty D., Graeb D., 1994. In vivo visualization of myelin water in brain by magnetic resonance. Magnetic Resonance in Medicine 31, 673-7. Meiboom S., Gill D., 1958. Modified spin-echo method for measuring nuclear relaxation times, Re V. Sci. Instrum 29, 688–91. Mitchell M., Kundel H., Axel L., Joseph P.M., 1986. Agarose as tissue equivalent phantom material for NMR imaging. Magnetic Resonance Imaging 4, 263-266. Moore G.R.W., Leung E., MacKay A.L., Vavasour I.M., Whittall K.P., Cover K.S., Li D.K., Hashimoto S.A., Oger J., Sprinkle T.J. and others, 2000. A pathology-MRI study of the short-T2 component in formalin-fixed multiple sclerosis brain. Neurology 55, 1506-10. Oakden W., Lobaugh N.J., Black S.E., Stanisz G.J., T2 hyperintensity does not always mean demyelination (and neither does decreased fractional anisotropy). Imaging Myelin: Formation, Destruction and Repair; 2006; Vancouver, Canada. p P41. Odrobina E.E., Lam T.Y., Pun T., Midha R., Stanisz G.J., 2005. MR properties of excised neural tissue following experimentally induced demyelination. NMR Biomed. 18, 277-84. Oh J., Han E.T., Pelletier D., Nelson S.J., 2006. Measurement of in vivo multi- component T2 relaxation times for brain tissue using multi-slice T2 prep at 1.5 and 3 T. Magn Reson Imaging 24, 33-43. Pell G.S., Briellmann R.S., Waites A.B., Abbott D.F., Jackson G.D., 2004. Voxel-based relaxometry: a new approach for analysis of T2 relaxometry changes in epilepsy. Neuroimage 21, 707-13. Poon C.S., Henkelman R.M., 1992. Practical T2 quantitation for clinical applications. Journal of Magnetic Resonance Imaging. 2, 541-53. Pun T.W., Odrobina E., Xu Q.G., Lam T.Y., Munro C.A., Midha R., Stanisz G.J., 2005. Histological and magnetic resonance analysis of sciatic nerves in the tellurium model of neuropathy. Journal of the Peripheral Nervous System 10, 38-46. Sled J.G., Pike G.B., 2001. Quantitative imaging of magnetization transfer exchange and relaxation properties in vivo using MRI. Magn Reson Med 46, 923-31. Stanisz G.J., Odrobina E.E., Pun J., Escaravage M., Graham S.J., Bronskill M.J., Henkelman R.M., 2005. T1, T2 relaxation and magnetization transfer in tissue at 3T. Magn Reson Med 54, 507-12. Stanisz G.J., Webb S., Munro C.A., Pun T., Midha R., 2004. MR properties of excised neural tissue following experimentally induced neuroinflammation. Magnetic Resonance in Medicine 51, 473-479. Stewart W.A., MacKay A.L., Whittall K.P., Moore G.R., Paty D.W., 1993. Spin-spin relaxation in experimental allergic encephalomyelitis. Analysis of CPMG data using a non-linear least squares method and linear inverse theory. Magnetic Resonance in Medicine. 29, 767-75. Tozer D.J., Davies G.R., Altmann D.R., Miller D.H., Tofts P.S., 2005. Correlation of apparent myelin measures obtained in multiple sclerosis patients and controls from magnetization transfer and multicompartmental T2 analysis. Magn Reson Med. 53, 1415-22. Vermathen P., Robert-Tissot L., Pietz J., Lutz T., Boesch C., Kreis R., 2007. Characterization of white matter alterations in phenylketonuria by magnetic  88  89 resonance relaxometry and diffusion tensor imaging. Magn Reson Med 58, 1145- 56. Vidarsson L., Conolly S.M., Lim K.O., Gold G.E., Pauly J.M., 2005. Echo time optimization for linear combination myelin imaging. Magn Reson Med. 53, 398- 407. Vold R.L., Vold R.R., Simon H.E., 1973. Errors in measurements of transverse relaxation rates. Journal of Magnetic Resonance 11, 283-298. Webb S., Munro C.A., Midha R., Stanisz G.J., 2003. Is multicomponent T2 a good measure of myelin content in peripheral nerve? Magnetic Resonance in Medicine. 49, 638-45. Whittall K.P., MacKay A.L., 1989. Quantitative interpretation of nmr relaxation data. Journal of Magnetic Resonance 84, 64-71. Whittall K.P., MacKay A.L., Graeb D.A., Nugent R.A., Li D.K., Paty D.W., 1997. In vivo measurement of T2 distributions and water contents in normal human brain. Magnetic Resonance in Medicine 37, 34-43. Woessner D.E., 1961. Effects of diffusion in nuclear magnetic resonance spin-echo experiments. Journal of Chemical Physics 34, 2057-61. Wu Y., Alexander A.L., Fleming J.O., Duncan I.D., Field A.S., 2006. Myelin water fraction in human cervical spinal cord in vivo. J Comput Assist Tomogr. 30, 304- 6.    Chapter 4 4 VALIDATION OF 3D MULTI-COMPONENT T2 RELAXATION IMAGING AGAINST THE 2D SINGLE- SLICE TECHNIQUE AT 3.0 T*  4.1 INTRODUCTION Development of sensitive, clinically relevant imaging measures for diagnosis, prognosis and treatment monitoring of neurological disease is a significant challenge in magnetic resonance imaging (MRI) research.  In particular, there is great need for a non-invasive measure that is sensitive and specific to myelin content in order to study white matter disease, injury, and development, as well as therapies that are targeted towards promoting remyelination.  A technique showing great promise in this area is multi-component T2 relaxation, which allows different reservoirs to be resolved based on T2 relaxation times. In normal human white matter, the shortest T2 component (~20 ms) is attributed to water trapped between the myelin bilayers (labeled myelin water) and an intermediate T2 component (~80 ms) is thought to arise from intra/extracellular (IE) water (MacKay et al., 1994; Menon et al., 1992; Whittall et al., 1997).  In certain pathological tissue, an additional T2 component with longer T2 (> 200 ms) is sometimes detected (Armspach et al., 1991; Helms 2001; Kolind et al., 2008; Larsson et al., 1989; Laule et al., 2007a; Laule et al., 2007b; Menon et al., 1992; Oakden et al., 2006; Rumbach et al., 1991; Sirrs et al., 2007).  The ratio of the short- T2 signal (T2 < 40 ms) to the total signal in the T2  * A version of this chapter will be submitted for publication. Kolind SH, Mädler B, MacKay AL. More efficient myelin water imaging in vivo: Validation of 3D multi-component T2-relaxation measurements. .  90 distribution gives the myelin water fraction (MWF, the fraction of water trapped between the myelin bilayers), which has been shown to be very highly correlated with histological measures of myelin in rat (Odrobina et al., 2005; Pun et al., 2005; Stanisz et al., 2004; Webb et al., 2003), guinea pig (Gareau et al., 1999; Gareau et al., 2000; Stewart et al., 1993), and formalin-fixed human brain (Laule et al., 2008; Laule et al., 2006; Moore et al., 2000).  In addition to this very sensitive and specific marker for myelin, T2 relaxation can also be used to obtain several other measures sensitive to changes in the human brain. With a water standard included in the image and the collection of data to measure accurate T1 relaxation times, the total signal can be used to determine the absolute water content (Whittall et al., 1997), which can be used to study global or focal increases in water content such as those which occur in multiple sclerosis (MS), tumours, and stroke or general ageing effects (Laule et al., 2004; Lin et al., 1997; Neeb et al., 2006; Whittall et al., 1997; Wick and Kuker 2004).  The geometric mean T2 ( 2T , analogous to the amplitude-weighted mean on a logarithmic scale) and peak width of the IE peak help probe the intra and extracellular water environments (Stewart et al., 1993), and aid in distinguishing between demyelination and inflammation (Stanisz et al., 2004).  The long- T2 fraction (the ratio of the long-T2 signal to total signal) has been shown to distinguish between different types of lesion in multiple sclerosis (MS) (Kolind et al., 2008), and has been linked to edema, gliosis, or vacuolation in the brains of subjects with MS, Alzheimer’s disease (AD), and phenylketonuria (PKU) (Armspach et al., 1991; Barnes et al., 1987; Helms 2001; Kolind et al., 2008; Larsson et al., 1989; Laule et al., 2007a; Laule et al., 2007b; Menon et al., 1992; Oakden et al., 2006; Rumbach et al., 1991; Sirrs et al., 2007; Vermathen et al., 2007).  91  Unfortunately, the most commonly used technique for acquiring multi-component T2 relaxation data, which consists of a single-slice multi-echo pulse sequence utilizing large gradient crushers and composite radiofrequency refocusing block pulses (Poon and Henkelman 1992), is difficult to implement and has a very long acquisition time (on the order of 25 min per slice), making it clinically unfeasible.  A rapid 3D multi-component T2 relaxation sequence was recently introduced by Mädler et al. (Mädler and MacKay 2006), which is capable of acquiring 7 slices in roughly the same amount of time required to scan one slice using the 2D single-slice pulse sequence.  While other multi-slice multi- component T2 techniques have been proposed (Oh et al., 2006; Vidarsson et al., 2005), to our knowledge none have been quantitatively compared to standard 2D single-slice multi-component T2 results.  The goal of this study was to validate the new 3D multi- component T2 relaxation imaging technique introduced by Mädler et al. (Mädler and MacKay 2006) against the 2D technique which is the current standard.  4.2 MATERIALS AND METHODS 4.2.1 MR data acquisition 10 healthy controls (7 female, 3 male; mean age: 33 years (range: 21-59 years)) were scanned on a Philips Achieva 3.0 T MRI scanner (Best, The Netherlands) using a multi- coil array eight-element, six channel receive head coil.  For all subjects, informed written consent as approved by the Clinical Research Ethics Board of our institution was obtained.   92 2D single-slice multi-component T2 relaxation measurements were performed on a transverse slice through the base of the genu and splenium of the corpus callosum (32 echoes, BW = ±32 kHz, TR = 2500 ms, slice thickness = 5 mm, FOV = 24x20 cm, 256x128 matrix, 10 ms echo spacing, 4 averages, scan duration = 18.3 min) (MacKay et al., 1994).  The 2D single-slice sequence used a slice–selective 90o pulse, 32 composite (90x-180y-90x) block refocusing pulses with a series of z-axis crusher gradient pulses of alternating sign with descending amplitude flanking the refocusing pulse to eliminate contributions from stimulated echoes (MacKay et al., 1994; Poon and Henkelman 1992). Gx Gy Gz RF 2D single-slice pulse sequence Gx Gy Gz RF 3D pulse sequence  Figure 4.1: First portion of the 2D single-slice (above) and 3D (below) pulse sequences. Note that overlapping gradients are summed in practice.  93 3D T2 relaxation measurements were acquired such that the centre slice was aligned with the location of the 2D single-slice  T2 measurement (7 slices, 32 echoes, BW = ±42 kHz, TR = 1200 ms, slice thickness = 5 mm, FOV = 24x20 cm, 256x128 matrix, 10 ms echo spacing, 1 average, scan duration = 19.8 min) (Mädler et al., 2008; Mädler and MacKay 2006).  Modifications to the 2D sequence to achieve 3D measurement included the addition of phase-encoding gradients along Gz for each echo, the use of smaller non- alternating-descending z-axis crusher gradient pulses, the replacement of the initial slice- selective 90° pulse with a 90° slab-selective pulse, and replacing the composite rectangular refocusing pulses by slab-selective pulses.  While the composite refocusing pulses would result in a better B1 profile than the slab-selective refocusing pulses employed, the SAR deposition would have been much higher, requiring a large increase in TR, resulting in an unfeasible scan time.  Saturation bands could not be used to limit signal outside of the volume from imperfect refocusing pulses for either the 2D or 3D sequence due to resulting magnetization transfer effects which have been shown to preferentially reduce the signal from water trapped between myelin bilayers due to direct saturation (Vavasour et al., 2000).  Pulse sequence diagrams for both the 2D and 3D sequences are illustrated in Figure 4.1.  CLEAR (Constant LEvel AppeaRance, utilized for compensation of inhomogeneous receiver sensitivities) was employed for both the 2D and 3D multi-echo acquisitions.  Six of the subjects had both the 2D single-slice and 3D scans in the same session, while 4 subjects had the two scans on separate days.  For six of the subjects, images were also acquired in order to calculate the transmitted magnetic field sensitivity (B1+), using the method introduced by Wang et al. (Wang et al.,  94 2005), which requires two spin echo images (TR = 10000 ms, TE = 31 ms, slice thickness = 5 mm, FOV = 24x24 cm, 112x79 matrix, 1 average, flip angle 60º/120º, refocusing angle 120º/240º).  4.2.2 Analysis In all cases, the 3D T2 data set was registered to the 2D T2 data using in-house registration software based on the maximization of mutual information (Pluim et al., 2003).  The registration method was a multi-scale technique that used the Shannon entropy computed from the joint histogram of the two images as the similarity measure.  Image interpolation was performed using a Blackman windowed sinc filter.  For the 3D data set, only the slice corresponding to the 2D single slice was analyzed.  The measured multi-echo signal (yi) for each echo (i from 1 to 32) from both the 2D single-slice and 3D T2 relaxation experiments could be described as follows (Whittall and MacKay 1989):   (4.1) 32,...,2,1,)/exp( 120 1 2 =−= ∑ = iTtsy j jiji where ti are the measured times for each of the 32 echoes, and sj is the relative amplitude for each of the 120 logarithmically spaced partitioned T2 times within the range of 15 ms to 2 s.  A non-negative least squares (NNLS) algorithm was used to minimize both χ2 and an energy constraint that smoothes the T2 distribution, sj(T2j ), providing better, consistent fits in the presence of noise (Fenrich et al., 2001; Lawson and Hanson 1974; Whittall and MacKay 1989).  The expression to be minimized was:  95  , (4.2) 0, 1 22 ≥+ ∑ = μμχ M j js where the T2 distribution is increasingly smoothed at the cost of misfit for increasing values of μ.  This approach provides a measure of robustness against noise variation in the data.  Regularized smooth T2 distributions were created by minimizing equation 4.2 with the energy constraint of 1.02χ2min ≤ χ2 ≤ 1.025χ2min.  The peak assigned to myelin water was defined as having 15 ms < T2 < 40 ms, and the intra/extracellular (IE) water peak was defined as having 40 ms < T2 < 200 ms.  Maps of the following parameters were created by calculating the value for each voxel in the image: MWF (the ratio of myelin water peak area to total signal in the T2 distribution), geometric mean T2 ( 2T , analogous to the amplitude-weighted mean on a logarithmic scale (Whittall et al., 1997)) of the IE pool, and the standard deviation of residuals expressed as a percentage of the theoretical amplitude at TE = 0 ms (providing a measure of the difficulty fitting the data) (Whittall et al., 1997).  Five grey matter (GM) (cingulate gyrus, cortical grey, putamen, head of the caudate nucleus, and thalamus) and five white matter (WM) (minor forceps, major forceps, genu and splenium of the corpus callosum, and posterior internal capsules) structures were outlined on the first echo (TE=10 ms) image of the 2D single-slice relaxation data.  The ROIs were then mapped onto the various parameter maps and the values within the ROI were averaged and standard errors were calculated.   96 B1+ maps were calculated from the ratio of the scans with excitation/refocusing flip angles of 60º/120º and 120º/240º, respectively, using the method described in equation [5] from Wang et al. (Wang et al., 2005):  3/11 )8/arccos( 1 λγτ= +B  (4.3) where γ is the gyromagnetic ratio, τ is the duration of the radiofrequency pulse, and λ is given by:  , )(sin )(sin )( )( 1 3 2 3 1 2 x x xSI xSI α αλ ==  (4.4) with SIi representing the signal intensity of the image with flip/refocusing angles of 60º/120º for i = 1 and 120º/240º for i = 2.  Linear regression and Pearson correlations were calculated using all of the ROI results, and a two-tailed paired Student’s t-test was used to compare results for each brain structure between the two methods.  Statistical significance was taken as p ≤ 0.05.  A Bland-Altman plot (the difference between results plotted against the average of the results) (Bland and Altman 1986) was calculated to compare values obtained using 2D or 3D measurements.  4.3 RESULTS Table 4.1 gives the average values of MWF, 2T , and standard deviation of residuals normalized to the initial signal (standard error) for the 2D single-slice data and registered 3D data, with significant differences (p < 0.05) indicated.  97 Cingulate Gyrus 3.2 (0.4) 0.7 (0.2) 66.8 (0.8) 81.0 (0.7) 0.34 (0.02) 0.34 (0.03) Cortical Gray 3.6 (0.5) 0.5 (0.1) 61.3 (0.8) 75.7 (0.6) 0.39 (0.02) 0.44 (0.03) Putamen 3.3 (0.8) 3.9 (0.7) 51.5 (1.0) 63.2 (0.7) 0.46 (0.03) 0.31 (0.02) Caudate 2.5 (0.4) 1.8 (0.3) 56.1 (0.5) 66.5 (0.6) 0.44 (0.03) 0.37 (0.02) Thalamus 4.4 (0.6) 4.9 (0.5) 56.6 (0.4) 65.6 (0.5) 0.38 (0.02) 0.35 (0.02) Minor Forceps 9.3 (0.7) 4.4 (0.4) 56.5 (0.4) 67.2 (0.3) 0.34 (0.02) 0.31 (0.03) Major Forceps 10.9 (0.6) 9.3 (0.5) 62.7 (0.6) 74.8 (0.6) 0.33 (0.02) 0.31 (0.01) Genu 13.1 (0.6) 11.6 (0.6) 55.7 (0.4) 64.7 (0.5) 0.33 (0.02) 0.24 (0.01) Splenium 15.1 (0.4) 16.1 (0.4) 64.4 (0.6) 75.6 (0.7) 0.36 (0.02) 0.34 (0.01) Internal Capsules 16.5 (0.6) 18.0 (0.5) 68.0 (0.9) 78.8 (0.5) 0.39 (0.02) 0.32 (0.02) MWF (%) Standard Deviation GMT2 (ms) 2D 3D of Residuals (%) 2D 3D 2D 3D  Table 4.1: MWF, 2T  and standard deviation of residuals (standard error) for the 2D single-slice and 3D relaxation techniques.  Significant differences (p < 0.05) between techniques are indicated in bold.  4.3.1 MWF values: 2D single-slice compared to 3D measurements MWF values were in good agreement between the 2D and 3D measurements, though 3D MWF values were found to be significantly lower for the cingulate gyrus, cortical grey matter, and minor forceps (Table 4.1).  Plots of the correlation between MWF using 2D single-slice and 3D T2 relaxation for the average values in each brain structure, as well as for all 200 ROIs (20 per subject) are shown in Figure 4.2 with error bars indicating standard error.  The relationship was linear and highly significant (p < 0.0001) in both cases, with R2 = 0.73 using all 200 ROIs and R2 = 0.91 for the averages for each brain structure.  The slope was near unity with a small, negative intercept in each case indicating that both measures follow the same trend with slightly higher MWF values resulting from the 2D technique.  It is apparent that the minor forceps deviate the most from the relationship; excluding them from the analysis yields a linear fit of 3D MWF = 1.00*(2D MWF) – 0.5%, R2 = 0.79, p < 0.0001 using all 200 ROIs.  With the correlations calculated separately for each individual, the mean R2 was 0.74 (median R2 = 0.78, range  98 0.51-0.90) with all p-values < 0.0005.  From the Bland-Altman plot for all MWF values (Figure 4.3, top), there was a small bias which fitted the equation: 3D MWF – 2D MWF = 0.11*(3D MWF + 2D MWF)/2 + 1.9%; the average difference was -1.05%. 3D MWF = 1.11*(2D MWF) - 2.0% R2 = 0.91, p < 0.0001 0 5 10 15 20 0 5 10 15 20 2D MWF (%) 3D  M W F (% ) Cingulate Gyrus Cortical Gray Putamen Caudate Thalamus Minor Forceps Major Forceps Genu Splenium Internal Capsules 3D MWF =  0.95*(2D MWF) - 0.6% R2 = 0.73, p < 0.0001 0 5 10 15 20 25 0 5 10 15 20 25 2D MWF (%) 3D  M W F (% ) Cingulate Gyrus Cortical Gray Putamen Caudate Thalamus Minor Forceps Major Forceps Genu Splenium Internal Capsules 3D  M W F (% ) 3D  M W F (% )  Figure 4.2: Correlation between 3D and 2D MWF measurements using (top) mean MWF for all 10 volunteers for each brain structure and (bottom) all 22 ROIs for each subject. Error bars are standard error.  99 y = 0.13x + 3.0 ms R2 = 0.06 0 5 10 15 20 25 50 60 70 80 90 3D T 2 + 2D T 2 (ms) 3D  T 2 -  2 D  T 2 ( m s) 2 y = 0.11x - 1.9% R2 = 0.03 -15 -10 -5 0 5 10 15 0 5 10 15 20 25 3D MWF + 2D MWF (%) 3D  M W F - 2 D  M W F (% ) 2 3D  T 2 -  2 D  T 2 ( m s) 3D  T 2 -  2 D  T 2 ( m s) 3D  M W F - 2 D  M W F (% )  Figure 4.3: Bland-Altman plots of the difference between 3D and 2D measurements of MWF (top) and 2T (bottom) against the mean of the two measurements.  The dashed line indicates the mean difference, and the dotted lines denote +/- 1.96 times the standard deviation of the difference.   100 4.3.2 2T  values: 2D single-slice compared to 3D measurements 3D T2 = 1.16*(2D T2) + 1.9ms R2 = 0.94, p < 0.0001 45 55 65 75 85 45 55 65 75 85 2D T2 (ms) 3D  T 2 ( m s) Cingulate Gyrus Cortical Gray Putamen Caudate Thalamus Minor Forceps Major Forceps Genu Splenium Internal Capsules 3D T2 = 0.99*(2D T2) + 12ms R2 = 0.77, p < 0.0001 40 50 60 70 80 90 100 40 60 80 100 2D T2 (ms) 3D  T 2 ( m s) Cingulate Gyrus Cortical Gray Putamen Caudate Thalamus Minor Forceps Major Forceps Genu Splenium Internal Capsules 3D  T 2 ( m s) 3D  T 2 ( m s)  Figure 4.4: Correlation between 3D and 2D 2T  measurements using (top) mean value for all 10 volunteers for each brain structure and (bottom) all 22 ROIs for each subject. Error bars are standard error.  2T  was significantly higher using the 3D technique for all brain structures (on average, 15% longer, see Table 4.1 and Figure 4.4).  From the Bland-Altman plot (Figure 4.3, bottom), there was a small bias which fitted the equation: 3D 2T  - 2D 2T  = 0.13*(3D 2T + 2D 2T )/2 + 3.0 ms; the average difference was 11.3 ms.  101  4.3.3 Standard deviation of residuals: 2D single-slice compared to 3D measurements The standard deviation of residuals normalized to initial signal was typically lower for the 3D technique, with significantly lower values found in the putamen, genu and internal capsules (Table 4.1).  Seventy-eight percent of the 200 ROIs had a lower standard deviation of residuals with the 3D technique, and no structure had a significantly higher standard deviation of residuals using the 3D acquisition.  The standard deviation of residuals was not correlated between the 2D single-slice and 3D techniques (R2 = 0.18, p = 0.2).  4.3.4 Qualitative comparison of parameter maps Maps of the MWF, 2T  and standard deviation of residuals for both 2D and 3D acquisitions are shown for one volunteer in Figure 4.5.  While qualitatively similar, the MWF maps show the same differences as are apparent in Table 4.1: the MWF values appeared to be lower for the 3D technique, particularly for external structures such as the minor forceps and cortical grey matter.  The 2T  maps were very similar for the two techniques, but brighter (higher 2T  values) using the 3D technique, which is consistent with Table 4.1, and with more visible flow artifact for the 2D single-slice scan.  The standard deviation of residuals had a very different distribution for the two techniques: the 2D technique yielded more uniform values except in regions where flow artifact is most likely (due to the ventricles) whereas the 3D map exhibited a ring of low values  102 with higher values for external and central brain.  The left minor forceps showed particularly high values of standard deviation of residuals for the 3D technique. 2D single-slice MWF T2 Standard deviation of residuals 3D 0% 30% 20% 10% 40ms 200ms 80ms 120ms 160ms 0% 1.5% 1.0% 0.5% 2T  Figure 4.5: Maps of MWF (top), 2T  (middle), and standard deviation of residuals (bottom) using the 2D single-slice technique (left) and the 3D technique (right).  The B1+ field map and the difference between 2T  values obtained using the 3D and 2D techniques are shown in Figure 4.6 for one volunteer.  The B1+ map was bluest  103 (indicating flip angles much smaller than prescribed) for the top left of the brain, which corresponds to the brightest region on the image of the difference between 2T  values (indicating the largest difference), as well as the brightest region on the map of 3D standard deviation of residuals (Figure 4.5).  Large differences in 2T  were also observed adjacent to the ventricles (bright in Figure 4.6, right), where flow artifact is apparent on the 2D 2T  map (Figure 4.5). B1 map 3D T2 value – 2D T2 value 1.17 1.00 0.83 0.67  Figure 4.6: For the same volunteer as in Figure 4.5, map of transmitted field sensitivity (B1+), with the ratio of actual to prescribed flip angle indicated in the legend (left), and difference between the 2T  calculated from the 3D multi-component T2 data and the 2D single-slice T2 data, brighter for larger differences (right).  4.3.5 NNLS fit residuals Figure 4.7 shows the residuals of the multi-exponential fit for the posterior internal capsules and cortical grey matter for one volunteer from both the 2D single-slice and 3D acquisitions.  In the internal capsules, which had (on average) a lower standard deviation of residuals for the 3D acquisition (see Table 4.1), both techniques showed similar initial oscillations but the residuals at later time points were larger for the 2D single-slice  104 acquisition.  For cortical grey matter, where a higher standard deviation of residuals was found for the 3D technique on average (see Table 4.1), the 3D residuals had much larger initial oscillations. -2.0% -1.5% -1.0% -0.5% 0.0% 0.5% 1.0% 1.5% 2.0% 0 50 100 150 200 250 300 350 Time (ms) M ul ti- ex po ne nt ia l f it re sid ua l (p er ce nt ag e of  in iti al  si gn al ) 2D Single-Slice Technique, SNRNNLS = 450 3D Technique, SNRNNLS = 197 -2.0% -1.5% -1.0% -0.5% 0.0% 0.5% 1.0% 1.5% 2.0% 0 50 100 150 200 250 300 350 Time (ms) M ul ti- ex po ne nt ia l f it re sid ua l (p er ce nt ag e of  in iti al  si gn al ) 2D Single-Slice Technique, SNRNNLS = 516 3D Technique, SNRNNLS = 523 Internal Capsules Cortical Grey 3D tec i , stdev of residuals = 0.19% 2D technique, stdev of residuals = 0.19% 3D technique, stdev of residuals = 0.51% 2D technique, stdev of residuals = 0.22% M ul ti- ex po ne nt ia l f it re sid ua l (p er ce nt ag e of  in iti al  si gn al ) M ul ti- ex po ne nt ia l f it re sid ua l (p er ce nt ag e of  in iti al  si gn al )  Figure 4.7: Residuals of the multi-exponential fit expressed in percentage of the initial signal along the decay train for (above) the posterior internal capsules and (below) cortical grey matter for the same volunteer as Figure 4.5.  Results from the 2D technique are shown as black squares and for the 3D technique as white circles.  The standard deviation of residuals (expressed as a percentage of initial signal) for these particular cases are given in the legends.  Visual inspection of images for various echo times for each sequence confirmed the presence of artifact characteristic of stimulated echoes (ghosting) on early echoes which were more dramatic for the 3D sequence, and flow artifact (in line with the ventricles in  105 the phase-encoding direction) for all echoes with the 2D single-slice technique.  Relative signal levels tended to be higher for the 3D sequence throughout the echo train.  4.4 DISCUSSION 4.4.1 Differences in MWF and 2T  values The linear relationship with slope near unity between MWF measured using each technique as well as the lack of significant differences between measurements in most brain regions or of a significant systematic bias from the Bland-Altman plot, demonstrates that in terms of MWF, the 3D multi-component relaxation technique and the standard 2D single-slice methods are measuring the same tissue property.  The most external structures measured (the minor forceps, cortical grey matter and cingulate gyrus) showed the greatest deviations, with lower MWFs and longer 2T s using the 3D technique (see Figure 4.2 and Table 4.1).  This difference in measurements is most likely due to the increased sensitivity of the 3D technique to B1 inhomogeneity; the regions with the greatest deviations correspond with the regions of the B1+ map in Figure 4.6 showing the largest discrepancy from the prescribed flip angle.  The 2D Poon-Henkelman sequence uses composite (90x-180y-90x) block refocusing pulses which have been shown to be less sensitive to variations in the B1 and B0 field than slice-selective pulses (Poon and Henkelman 1992), and large alternating-descending crushers flanking each refocusing pulse to reduce the effects of out-of-slice signal and stimulated echoes from signal generated by the refocusing pulses.  The 3D technique utilized slab-selective refocusing pulses in place of the block pulses, for signal  106 cancellation of out-of-slab signal and reduction of the potentially high specific absorption rate (SAR) at high field (3.0 T).  While these slab-selective pulses were less sensitive to B0 inhomogeneity than the slice-selective pulses tested by Poon and Henkelman (and rejected in favour of block composite pulses in that study), the B1 variations common at higher field led to less accurate measurements around the periphery of the brain resulting in lower measured MWF values in those regions (Jones 2003).  The sensitivity to B1 and smaller, non-alternating-descending crusher gradients (< 1% of the crusher area used for the first refocusing pulse of the 2D single-slice pulse sequence used in this study) could also lead to more stimulated echoes, which was demonstrated to be minimized using the 2D single-slice technique (Poon and Henkelman 1992).  Spins that do not receive a perfect refocusing pulse create magnetization along the longitudinal axis, which is stored along the z-axis only to be later flipped back to the transverse plane by successive imperfect pulses (Hennig 1988).  The stimulated echo component decay rate depends on T1 and T2 as well as the storage time.  Larger ratios of T1 to T2 give the largest stimulated echo contributions.  Because the net magnetization vector spends more time out of the transverse plane, the stimulated echoes lead to more signal for later echoes in the echo train, which gives rise to an apparently slower T2-decay for the intermediate (IE water) T2 peak.  This is apparent from the longer 2T  values that were observed for the 3D technique (see Table 4.1), particularly in external brain regions where B1 inhomogeneity is expected to have the greatest impact (see Figure 4.6), and in GM which has a larger ratio of T1 to T2 than WM.   107 4.4.2 Differences in standard deviation of residuals Even with presumably greater contribution from stimulated echoes, the standard deviation of residuals was generally lower using the 3D technique, significantly so for internal structures such as the putamen, genu and internal capsules (see Table 4.1).  In Figure 4.5, the standard deviation of residuals for the 3D technique appeared lower for a ring somewhere between the centre and periphery of the brain, most likely related to the distribution of the B1 inhomogeneity (where flip angles the closest to the prescribed angle were found in the B1+ map, see Figure 4.6).  The left minor forceps had particularly high standard deviations of residuals for the 3D technique, and this phenomenon was also observed for the other 9 volunteers, with the standard deviation of residuals always higher for the left minor forceps than for the right (on average 44% higher (range: 8– 66%), p = 0.002).  This may be a characteristic of the coil or the constant level appearance (CLEAR) algorithm for signal intensity homogeneity correction utilized to compensate for inhomogeneous receiver sensitivities; CLEAR uses a reference scan which provides a sensitivity map of the phased-array coil, enabling the system to calculate the exact signal contribution of each pixel to the image.  The reference scan performs separate scans using the body coil and then the phased-array coil to obtain sensitivity profiles for each coil, then uses the information to match the sensitivity profile of the phased-array coil to that of the body coil.  CLEAR was used for both the 3D and 2D single-slice techniques, but the 3D technique was more sensitive to B1 inhomogeneities.  Indeed, the areas of high standard deviation of residuals for the 3D technique corresponded roughly to regions of inaccurate flip angles indicated on the B1+ map (Figure 4.6).  The 2D map of standard deviation of residuals showed higher values  108 in regions typically contaminated by flow artifact (for the 3D sequence, phase encoding is applied for each echo, as opposed to the 2D technique where phase encoding is applied only at the beginning of the echo train, hence the 2D sequence may be more sensitive to flow artifact), but was otherwise quite uniform.  The residuals to the NNLS fit shown in Figure 4.7 for the posterior internal capsules demonstrate that the majority of the error dies away in the first few echoes for the 3D technique, typical of stimulated echoes, while oscillations are still apparent at later time points for the 2D acquisition, which is expected in the presence of flow artifact.  In cortical grey matter, the initial oscillations from the 3D technique were much larger than for the 2D single-slice technique, still showing the characteristic alternation of stimulated echoes.  4.4.3 Comparison to multi-slice multi-component T2 relaxation techniques Another imaging technique used to collect MWF data rapidly is linear combination myelin imaging (Jones et al., 2004; Vidarsson et al., 2005), and a multi-slice implementation has been presented (Vidarsson et al., 2005).  For this 3-echo linear combination technique, the echo times and weights are chosen to maximize the signal-to- noise ratio (SNR) of the short-T2 signal and suppress the signal from IE water and cerebrospinal fluid (CSF).  MWF values using the 3D technique had fair agreement with values quoted using linear combination myelin imaging (Vidarsson et al., 2005), although the linear combination technique yielded lower values in GM than the 3D technique used in this study; GM MWF values from the linear combination myelin imaging did not agree as closely with the Poon-Henkelman 2D single-slice technique.  The advantage of the 3- echo linear combination myelin imaging technique is that several slices can be acquired  109 in less than 5 minutes.  Disadvantages include the necessity of choosing a model (which may break down in pathology), the tradeoff between short-T2 SNR efficiency and suppression of the other water signal, and the limitation to sampling only one water compartment since the signal from water in other compartments is suppressed.  Oh et al. (Oh et al., 2006) also found similar values for MWF  employing a novel spiral acquisition multi-slice technique that includes T2 prep (Foltz et al., 2003) using composite 180º block refocusing pulses) with a radiofrequency cycling scheme which mitigates the effects of T1 recovery between the T2 prep and readout (Wright et al., 1996), allowing multi-slice acquisition without significant magnetization transfer weighting.  They were able to acquire 16 slices in 10 minutes, with voxel resolution 2x2x10 mm3.  However, the values observed by Oh et al. in GM were higher than those observed with either the 2D single-slice or 3D technique in the present study, and certain WM structures had lower MWF values than expected using the 2D standard technique.  It was suggested that these differences were a result of the 3.0 T field strength, where signal loss due to the imperfect refocusing train should be larger than at 1.5 T due to the increased effective B1 inhomogeneity, or that the increased SNR could better detect the short-T2 component resulting in higher (truer) MWF values.   Since the study by Oh et al. used much larger voxels (2x2x10 mm3) than the present study (0.9x1.9x5 mm3), the differences in MWF values could simply be an effect of partial voluming.   110 4.4.4 Concluding remarks The use of 3D multi-component T2-relaxation allows for greatly improved brain coverage in clinically feasible times.  For the 3D technique, the standard deviation of residuals was typically lower than for the standard 2D single-slice technique and MWF was consistent with gold-standard results for nearly all brain structures examined.  Exceptions occurred in peripheral brain regions due to the increased sensitivity of the slab-selective refocusing pulses used by the 3D T2-relaxation technique to B1 inhomogeneity.  4.5 ACKNOWLEDGMENTS We sincerely thank our volunteers and the technologists of UBC hospital.  We are also grateful to Cornelia Laule and Piotr Kozlowski for useful discussions, and Roger Tam for assistance with registration.  The research was supported by the Canadian Institutes of Health Research, the Killam Trusts, the Natural Science and Engineering Research Council of Canada and the Multiple Sclerosis Society of Canada.   111 4.6 REFERENCES Armspach J.P., Gounot D., Rumbach L., Chambron J., 1991. In vivo determination of multiexponential T2 relaxation in the brain of patients with multiple sclerosis. Magn Reson Imaging. 9, 107-13. Barnes D., McDonald W.I., Johnson G., Tofts P.S., Landon D.N., 1987. Quantitative nuclear magnetic resonance imaging: characterisation of experimental cerebral oedema. Journal of Neurology, Neurosurgery & Psychiatry 50, 125-133. Bland J.M., Altman D.G., 1986. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1, 307-310. Fenrich F.R., Beaulieu C., Allen P.S., 2001. Relaxation times and microstructures. NMR Biomed. 14, 133-9. Foltz W.D., Al-Kwifi O., Sussman M.S., Stainsby J.A., Wright G.A., 2003. Optimized spiral imaging for measurement of myocardial T2 relaxation. Magn Reson Med. 49, 1089-97. Gareau P.J., Rutt B.K., Bowen C.V., Karlik S.J., Mitchell J.R., 1999. In vivo measurements of multi-component T2 relaxation behaviour in guinea pig brain. Magnetic Resonance Imaging. 17, 1319-25. Gareau P.J., Rutt B.K., Karlik S.J., Mitchell J.R., 2000. Magnetization transfer and multicomponent T2 relaxation measurements with histopathologic correlation in an experimental model of MS. Journal of Magnetic Resonance Imaging 11, 586- 95. Helms G., 2001. Volume correction for edema in single-volume proton MR spectroscopy of contrast-enhancing multiple sclerosis lesions. Magnetic Resonance in Medicine 46, 256-63. Hennig J., 1988. Multiecho imaging sequences with low refocusing flip angles. J Magn Reson 78, 397-407. Jones C.K., 2003. T2 Decay Curve Acquisition and Anaysis in MRI; Noise Considerations, Short T2, and B1 Field Encoding. Vancouver: University of British Columbia. 162 p. Jones C.K., Xiang Q.S., Whittall K.P., MacKay A.L., 2004. Linear combination of multiecho data: short T2 component selection. Magn Reson Med. 51, 495-502. Kolind S.H., Laule C., Vavasour I.M., Li D.K., Traboulsee A.L., Mädler B., Moore G.R., Mackay A.L., 2008. Complementary information from multi-exponential T(2) relaxation and diffusion tensor imaging reveals differences between multiple sclerosis lesions. Neuroimage 40, 77-85. Larsson H.B., Frederiksen J., Petersen J., Nordenbo A., Zeeberg I., Henriksen O., Olesen J., 1989. Assessment of demyelination, edema, and gliosis by in vivo determination of T1 and T2 in the brain of patients with acute attack of multiple sclerosis. Magnetic Resonance in Medicine. 11, 337-48. Laule C., Kozlowski P., Leung E., Li D.K., Mackay A.L., Moore G.R., 2008. Myelin water imaging of multiple sclerosis at 7 T: Correlations with histopathology. Neuroimage [Epub ahead of print] Laule C., Leung E., Li D.K., Traboulsee A.L., Paty D.W., MacKay A.L., Moore G.R., 2006. Myelin water imaging in multiple sclerosis: quantitative correlations with histopathology. Mult Scler. 12, 747-53.  112 Laule C., Vavasour I.M., Kolind S.H., Traboulsee A.L., Moore G.R.W., Li D.K.B., MacKay A.L., 2007a. Long T2 Water in Multiple Sclerosis: What else can we learn from multi-echo T2 Relaxation? Journal of Neurology 254, 1579-87. Laule C., Vavasour I.M., Madler B., Kolind S.H., Sirrs S.M., Brief E.E., Traboulsee A.L., Moore G.R., Li D.K., Mackay A.L., 2007b. MR evidence of long T(2) water in pathological white matter. J Magn Reson Imaging 26, 1117-21. Laule C., Vavasour I.M., Moore G.R.W., Oger J., Li D.K.B., Paty D.W., MacKay A.L., 2004. Water content and myelin water fraction in multiple sclerosis: A T2 relaxation study. Journal of Neurology 251, 284-293. Lawson C.L., Hanson R.J., 1974. Solving Least Squares Problems. Englewood Cliffs, NJ: Prentice-Hall,. Lin W., Paczynski R.P., Venkatesan R., He Y.Y., Powers W.J., Hsu C.Y., Haacke E.M., 1997. Quantitative regional brain water measurement with magnetic resonance imaging in a focal ischemia model. Magn Reson Med 38, 303-10. MacKay A., Whittall K., Adler J., Li D., Paty D., Graeb D., 1994. In vivo visualization of myelin water in brain by magnetic resonance. Magnetic Resonance in Medicine 31, 673-7. Mädler B., Drabycz S.A., Kolind S.H., Whittall K.P., MacKay A.L., 2008. Is diffusion anisotropy an accurate monitor of myelination? Correlation of multicomponent T2 relaxation and diffusion tensor anisotropy in human brain. Magnetic Resonance Imaging June 3, 2008 (Epub ahead of print), Mädler B., MacKay A.L., In-vivo 3D Multi-component T2-Relaxation Measurements for Quantitative Myelin Imaging at 3T. 14th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2006; Seattle, USA. p 2112. Menon R.S., Rusinko M.S., Allen P.S., 1992. Proton relaxation studies of water compartmentalization in a model neurological system. Magn Reson Med. 28, 264- 74. Moore G.R.W., Leung E., MacKay A.L., Vavasour I.M., Whittall K.P., Cover K.S., Li D.K., Hashimoto S.A., Oger J., Sprinkle T.J. and others, 2000. A pathology-MRI study of the short-T2 component in formalin-fixed multiple sclerosis brain. Neurology 55, 1506-10. Neeb H., Zilles K., Shah N.J., 2006. Fully-automated detection of cerebral water content changes: study of age- and gender-related H2O patterns with quantitative MRI. Neuroimage 29, 910-22. Oakden W., Lobaugh N.J., Black S.E., Stanisz G.J., T2 hyperintensity does not always mean demyelination (and neither does decreased fractional anisotropy). Imaging Myelin: Formation, Destruction and Repair; 2006; Vancouver, Canada. p P41. Odrobina E.E., Lam T.Y., Pun T., Midha R., Stanisz G.J., 2005. MR properties of excised neural tissue following experimentally induced demyelination. NMR Biomed. 18, 277-84. Oh J., Han E.T., Pelletier D., Nelson S.J., 2006. Measurement of in vivo multi- component T2 relaxation times for brain tissue using multi-slice T2 prep at 1.5 and 3 T. Magn Reson Imaging 24, 33-43. Pluim J.P., Maintz J.B., Viergever M.A., 2003. Mutual-information-based registration of medical images: a survey. IEEE Trans Med Imaging 22, 986-1004.  113  114 Poon C.S., Henkelman R.M., 1992. Practical T2 quantitation for clinical applications. Journal of Magnetic Resonance Imaging. 2, 541-53. Pun T.W., Odrobina E., Xu Q.G., Lam T.Y., Munro C.A., Midha R., Stanisz G.J., 2005. Histological and magnetic resonance analysis of sciatic nerves in the tellurium model of neuropathy. Journal of the Peripheral Nervous System 10, 38-46. Rumbach L., Armspach J.P., Gounot D., Namer I.J., Chambron J., Warter J.M., Collard M., 1991. Nuclear magnetic resonance T2 relaxation times in multiple sclerosis. J Neurol Sci. 104, 176-81. Sirrs S.M., Laule C., Maedler B., Brief E.E., Tahir S.A., Bishop C., MacKay A.L., 2007. Normal appearing white matter in subjects with phenylketonuria: water content, myelin water fraction, and metabolite concentrations. Radiology 242, 236-243. Stanisz G.J., Webb S., Munro C.A., Pun T., Midha R., 2004. MR properties of excised neural tissue following experimentally induced neuroinflammation. Magnetic Resonance in Medicine 51, 473-479. Stewart W.A., MacKay A.L., Whittall K.P., Moore G.R., Paty D.W., 1993. Spin-spin relaxation in experimental allergic encephalomyelitis. Analysis of CPMG data using a non-linear least squares method and linear inverse theory. Magnetic Resonance in Medicine. 29, 767-75. Vavasour I.M., Whittall K.P., Li D.K., MacKay A.L., 2000. Different magnetization transfer effects exhibited by the short and long T(2) components in human brain. Magnetic Resonance in Medicine 44, 860-6. Vermathen P., Robert-Tissot L., Pietz J., Lutz T., Boesch C., Kreis R., 2007. Characterization of white matter alterations in phenylketonuria by magnetic resonance relaxometry and diffusion tensor imaging. Magn Reson Med 58, 1145- 56. Vidarsson L., Conolly S.M., Lim K.O., Gold G.E., Pauly J.M., 2005. Echo time optimization for linear combination myelin imaging. Magn Reson Med. 53, 398- 407. Wang J., Qiu M., Yang Q.X., Smith M.B., Constable R.T., 2005. Measurement and correction of transmitter and receiver induced nonuniformities in vivo. Magn Reson Med. 53, 408-17. Webb S., Munro C.A., Midha R., Stanisz G.J., 2003. Is multicomponent T2 a good measure of myelin content in peripheral nerve? Magnetic Resonance in Medicine. 49, 638-45. Whittall K.P., MacKay A.L., 1989. Quantitative interpretation of nmr relaxation data. Journal of Magnetic Resonance 84, 64-71. Whittall K.P., MacKay A.L., Graeb D.A., Nugent R.A., Li D.K., Paty D.W., 1997. In vivo measurement of T2 distributions and water contents in normal human brain. Magnetic Resonance in Medicine 37, 34-43. Wick W., Kuker W., 2004. Brain edema in neurooncology: radiological assessment and management. Onkologie 27, 261-6. Wright G.A., Brittain J.H., Stainsby J.A., Preserving T1 or T2 contrast in magnetization preparation sequences. 4th Annual Meeting of the International Society for Magnetic Resonance in Medicine; 1996; New York, New York. p 1474.   Chapter 5 5 VOXEL-WISE (HISTOGRAM) ANALYSIS COMPARING 3D MULTI-COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 3.0 T*  5.1 INTRODUCTION Multiple sclerosis (MS) is a complex disease of the central nervous system (CNS), and the mechanisms underlying clinical progression are poorly understood.  Pathological aspects of MS include inflammation, edema, demyelination, gliosis, oligodendrocyte loss, and axonal degeneration (Keegan and Noseworthy 2002).  While MS can be studied non- invasively using magnetic resonance imaging (MRI), conventional MRI techniques are not specific to the diverse features of MS and are often insensitive to subtle changes in normal appearing white matter (NAWM) (Rashid and Miller 2008; Zivadinov et al., 2008).  Newer techniques such as multi-component T2 relaxation imaging and diffusion tensor imaging (DTI) have shown differences between NAWM and control normal white matter (NWM) (for a review, see (Laule et al., 2007a)).  From the decay curves obtained using a multi-echo T2 relaxation imaging sequence, the relative contributions from water in different environments can be resolved based on their  * A version of this chapter will be submitted for publication. Kolind SH, Laule C, Vavasour IV, Mädler B, Rauscher A, Devonshire V, Hooge J, Oger J, Smythe P, Traboulsee AL, Moore GRW, Li DKB, MacKay AL. Histograms from multi-component T2 relaxation imaging in multiple sclerosis: Characterization and comparison with histograms from diffusion tensor imaging.  115 T2 relaxation times, with the shortest component (~20 ms) arising from water trapped between myelin bilayers (myelin water), and an intermediate component (~80 ms) attributed to intra/extracellular (IE) water (MacKay et al., 1994; Whittall et al., 1997). The ratio of the myelin water T2 component (15 ms < T2 < 40 ms) to the total signal in the T2 distribution is called the myelin water fraction (MWF), and there is strong evidence that MWF can be used as a marker for myelin (Laule et al., 2008; Laule et al., 2006; Webb et al., 2003).  The geometric mean T2 ( 2T , analogous to the amplitude-weighted mean on a logarithmic scale) of the intermediate T2 component can be used to probe the intra and extracellular water environments (Stewart et al., 1993) and may aid in distinguishing demyelination from inflammation (Stanisz et al., 2004).  Previously, multi-component T2 relaxation imaging has been constrained to a lengthy single-slice acquisition (on the order of 25 minutes per slice), limiting the quantity of data too severely for histogram analysis.  Very recently, a few techniques have overcome this limitation (Oh et al., 2006; Vermathen et al., 2007; Vidarsson et al., 2005); this study makes use of the use of a rapid 3D multi-echo T2 relaxation imaging sequence developed by Mädler et al.(Mädler and MacKay 2006) which allows the acquisition of 7 slices in roughly the same amount of time needed for one slice using the traditional multi-echo T2 relaxation sequence.  Previous studies in MS have found that, compared to healthy brain, MWF is reduced in both NAWM and lesion (Laule et al., 2004; Oh et al., 2007; Tozer et al., 2005; Vavasour et al., 1998; Wu et al., 2006).  2T  of the IE peak in NAWM has been shown to increase  116 diffusely across all white matter (Whittall et al., 2002) and new  MS lesions showed increases in 2T  which returned toward pre-lesion values at later months, though never reaching NAWM values (Vavasour et al., 1999).  Diffusion imaging measures the mobility of water molecules, which can be restricted by barriers in biological tissue such as myelin sheaths, axonal membranes, etc.  The diffusion tensor gives a more complete description of the barriers of interest, and can be described by 3 eigenvalues, also referred to as diffusivities.  Invariant scalar indices have been derived from the diffusion tensor, including the mean diffusivity (<D>, equal to the average of the three eigenvalues) and fractional anisotropy (FA), which reflects the integrity and degree of alignment of structures.  FA can be an ambiguous measure: a decrease in FA could be caused by either a reduction in restriction perpendicular to the fibre tracts or simply an increase in diffusion parallel to fibre tracts.  The diffusion tensor eigenvalues themselves have been proposed as more specific measures of pathology (Basser et al., 2000; Le Bihan and Basser 1995; Xue et al., 1999).  The largest eigenvalue, λ||, is believed to be along the direction of the axon, and thereby linked to axonal integrity, while the average of the two smaller eigenvalues, λ | , is the diffusivity perpendicular to the dominant direction of diffusion, so changes in λ |  are thought to be related to reflect myelin pathology (Beaulieu et al., 1996; Biton et al., 2006; Kim et al., 2006; Song et al., 2003; Song et al., 2002; Song et al., 2005; Sun et al., 2006a; Sun et al., 2006b; Thomalla et al., 2004).  DTI studies of MS have found increased <D> and reduced FA in NAWM (Bammer et al., 2000; Cercignani et al., 2001a; Christiansen et al., 1993; Ciccarelli et al., 2001; Lin et al., 2007; Mainero et al., 2001; Werring et al., 1999)  117 and increased λ |  with little or no change in λ|| (Henry et al., 2003; Kolind et al., 2008; Lin et al., 2007; Lowe et al., 2006; Oh et al., 2004).  Volumetric histograms of diffusion metrics have been used to study both visible and invisible damage in MS brain (Cercignani et al., 2000; Cercignani et al., 2001b; Henry et al., 2003; Iannucci et al., 2001; Nusbaum et al., 2000; Oh et al., 2004; Pulizzi et al., 2007; Rovaris et al., 2002; Wilson et al., 2001; Yu et al., 2007).  Given that we now have the capability to measure multi-echo T2 relaxation data across multiple slices, the goal of this study was to examine the MWF and 2T  histograms, alongside DTI histograms, and assess the more subtle abnormalities of MS brain tissue that can be missed using region-of-interest analysis.  Also, we wished to investigate the magnitude of the correlations between the various T2 relaxation and DTI metrics, as well as correlations between these MRI-derived measures and both level of disability and disease duration.  5.2 MATERIALS AND METHODS 5.2.1 Subject information Thirteen subjects with clinically definite relapsing-remitting MS (10 female, 3 male; median Expanded Disability Status Scale (EDSS) = 2.5 (range 1.0-6.0); mean age = 40 yrs (range 28-57 yrs); mean disease duration = 8.5 yrs (range 0.5-27 yrs), see Table 5.1), and 11 age and gender matched controls underwent an MR examination.  None of the MS subjects were receiving steroid treatment or immune modulating drugs at the time of  118 imaging.  For all subjects, informed written consent as approved by the Clinical Research Ethics Board of our institution was obtained.  5.2.2 MR data acquisition MR images were obtained with a Philips Achieva 3.0 T MRI scanner (Best, The Netherlands), using a phased-array head coil and Dual Nova gradients.  A quick T1- weighted survey for patient positioning was followed by a true-midline T1 inversion recovery (IR) sagittal scan (1 slice, TR = 1900 ms, TI = 800 ms, TE = 10 ms, slice thickness = 4 mm) and T2-weighted axial (28 slices, TR = 2500 ms, TE = 90 ms, slice thickness = 5 mm) localizer to establish location and angulations for the slab of interest for the multi-echo T2 relaxation and diffusion tensor imaging scans.  The centre slice of the transverse slab was located just superior to the ventricles, such that the slice was parallel to the base of the genu and splenium of the corpus callosum.  The 3D multi-echo T2 relaxation sequence utilized a 90º slab-selective excitation pulse followed by 32 slab- selective refocusing pulses flanked by z-axis gradient crusher pulses (7 slices (note only 5 were analyzed due to aliasing artifact in the outermost slices), 32 echoes, BW = ±42 kHz, TR = 1200 ms, slice thickness = 5 mm, FOV = 24 cm, 256x128 matrix, 10 ms echo spacing, 1 average, scan duration = 19.8 min) (Mädler et al., 2008; Mädler and MacKay 2006).  To reduce energy deposition, a flip-angle sweep approach (Hennig and Scheffler 2001; Hennig et al., 2003) was employed: the flip angles were reduced as a sweep along the echo train starting with high flip angles (180º) to a minimum of 170º, resulting in a reduction in the RF power.  The DTI data was acquired for 13 slices, centered at the same location as the T2 relaxation scan such that the middle 7 slices were matched to the 7 T2  119 relaxation slices, using a single-shot Echoplanar Imaging (EPI) sequence with unipolar pulsed field gradients for diffusion weighting (TR = 2000 ms, TE = 55 ms, slice thickness = 5 mm, FOV = 24 cm, 128x96 matrix, SENSE factor = 2.0, δ = 13.2 ms, Δ = 27.4 ms, 2 b-values (0, 1000 s/mm2), 16 non-colinear gradient directions including one b = 0 image, 2 averages, scan duration = 3.1 min).  Additional scans included a 3D T1-weighted turbo field echo (TFE) scan for segmentation (120 slices, TR = 10 ms, TE = 6 ms, matrix = 192 x 163, slice thickness = 1.1 mm), and an axial FLAIR (fluid attenuated inversion recovery (Hajnal et al., 1992)) for lesion detection (28 slices, TR = 10000 ms, TE = 125 ms, TI = 2800 ms, matrix = 256 x 203, slice thickness = 5 mm).  5.2.3 MR data analysis The T2 relaxation data was analyzed using a regularized non-negative least squares (NNLS) approach (Lawson and Hanson 1974; Whittall et al., 1991; Whittall and MacKay 1989) to give a T2 distribution.  The peak assigned to myelin water was defined as having 15 ms < T2 < 40 ms, and the intra/extracellular (IE) water peak was defined as having 40 ms < T2 < 200 ms.  MWF was defined as the ratio between myelin water peak area to the total signal in the T2 distribution.   For the IE peak, the geometric mean T2 ( 2T , analogous to the amplitude-weighted mean on a logarithmic scale (Whittall et al., 1997)) was calculated.  The diffusion weighted images were realigned to correct for misregistration due to eddy currents using eddycorrect (from FDT: FMRIB’s Diffusion Toolbox, FMRIB Software Library, www.fmrib.ox.ac.uk), and the b = 0 image was registered to the T2 data using  120 FLIRT (FMRIB's Linear Image Registration Tool, FMRIB Software Library, www.fmrib.ox.ac.uk) with mutual information and 12 degrees of freedom (translation, rotation, scaling, and shearing in 3 dimension each).  Due to the distortions present in the EPI images (particularly susceptibility-induced distortions), 12 degrees of freedom were found to provide the best match to the T2 data.  The transformation matrix was applied to the remaining diffusion-weighted images.  The diffusion tensor was fit using dtifit (from FDT, FMRIB Software Library, www.fmrib.ox.ac.uk), yielding the diffusion eigenvalues λ1, λ2 and λ3 (from largest to smallest, respectively). <D> was given by the average of the three eigenvalues.  FA was defined as:  2 3 2 2 2 1 2 13 2 32 2 21 )()()( 2 1 λλλ λλλλλλ ++ −+−+−=FA . (5.1) The parallel and perpendicular diffusivities were given by λ|| = λ1 and λ |   = (λ2 + λ3)/2, respectively.  Maps of MWF, 2T , FA, λ | , λ|| and <D> were created by displaying the value at each pixel in the image for the slices corresponding to the centre 5 slices of the multi-echo T2 relaxation acquisition (the outer slices were rejected due to 3D fold-over artifact).  The FLAIR, T2-weighted, and 3D T1 TFE images were also registered to the T2 relaxation data set using FLIRT.  5.2.4 Lesion and NAWM identification Regions of interest (ROIs) were drawn around MS lesions (identified by a trained observer) on the FLAIR images, using the T2-weighted images to increase confidence in lesion identification.  The lesion ROIs were mapped onto the T2-weighted images of the control subjects.  If a mapped ROI did not fall in the same brain region as the MS subject  121 due to differences in brain morphology, the location of the ROI was adjusted by user- defined pixel shifts to the left, right, up, or down, so that the size and shape of the ROI was maintained.  The lesion ROIs were masked on the 3D T1 TFE images, and the extracerebral tissue was removed using BET (Brain Extraction Tool, FMRIB Software Library, www.fmrib.ox.ac.uk).  The remaining data was segmented into white matter (NAWM for MS, NWM for controls), grey matter, and cerebrospinal fluid (CSF) using FAST (FMRIB’s Automated Segmentation Tool, FMRIB Software Library, www.fmrib.ox.ac.uk).  In-house software was used to erode the NAWM and NWM masks by 2 voxels in order to minimize potential partial volume error.  This large amount of erosion may be considered overly conservative and resulted in significantly less data being included in the analysis, but was chosen in order to avoid any contamination of the WM masks with GM, CSF or lesion particularly in case of any remaining misalignments in the DTI data due to misregistration or distortion.  Thus we had great confidence that the data that was included was from the expected tissue type.  The eroded NAWM and NWM masks and the lesion ROI masks were visually checked against the registered maps of MWF, FA and <D> for each subject and control to confirm that the same brain regions were covered in all cases.  An example of the registered FLAIR, 3D T1 TFE, NAWM mask, MWF, <D> and FA maps for the centre slice for one MS subject is demonstrated in Figure 5.1, with a lesion outlined in white.  122 a b c d e f  Figure 5.1: An example of registered (a) FLAIR, (b) 3D T1 TFE, (c) NAWM mask, (d) MWF map, (e) <D> map and (f) FA map for one MS subject.  A lesion is outlined in white.  5.2.5 Histogram analysis The lesion ROI masks and NAWM/NWM masks were then transferred to the registered maps of MWF, 2T , FA, λ | , λ|| and <D>, and histogram analysis was conducted separately for NAWM/NWM and lesion voxels.  To account for differences in brain volume, each histogram was normalized by dividing the height of each bin count by the total number of voxels in the histogram.  The relative peak height, peak location, and mean value were calculated for each histogram.  A large proportion of voxels (particularly for MS subjects) had MWF values of exactly zero, resulting in a large peak in the lowest  123 histogram bin.  Thus for calculation of peak location, the first histogram bin (MWF = 0 – 0.01%) was excluded for the MWF histograms.  The percentage of voxels in the NAWM/NWM or lesion mask with MWF values of zero was also reported.  Average histograms for all MS subjects and for all controls were created by averaging the normalized histograms from each subject so that each subject was equally weighted, regardless of the actual number of voxels in NAWM/NWM or lesion.  5.2.6 Statistical analysis Spearman rank correlation coefficients (R) were used to assess correlations between the different MR-derived measures and EDSS or disease duration.  Group comparisons were evaluated using a two-tailed Student’s t-test.  MS lesion histogram values were compared to matched lesion ROIs in controls for group comparisons.  p values of less than 0.05 were considered statistically significant.  5.3 RESULTS The NAWM and lesion volumes analyzed for MS subjects and the NWM volumes analyzed for control subjects are given in Table 5.1.  The control NWM volumes were significantly larger than the NAWM volumes for MS subjects.  124 Disease Analyzed Lesion Analyzed NAWM Analyzed Control NWM Subject Age (years) Sex EDSS Duration (years) Volume (cm3) Volume (cm3) Volume (cm 3 ) S01 37 F 2.5 12 8.8 24 54 S02 28 M 1 1 5.5 53 72 S03 52 F 3.5 10 8.6 27 71 S04 42 F 3.5 11 12.1 16 56 S05 33 F 1 2 0.9 57 77 S06 34 M 3.5 8 4.3 50 80 S07 46 F 1.5 7 0.1 69 S08 30 F 3.5 9 2.0 30 67 S09 43 F 1 0.5 1.9 54 57 S10 28 M 6 5.5 2.0 33 59 S11 52 F 2.5 14 0.4 56 65 S12 36 F 1 4 0.6 68 S13 57 F 3.5 27 5.1 35 61 Average 40 2.5 8.5 4.0 44 65 (SD) (10) (1.5) (7.0) (3.8) (17) (9) Table 5.1: Clinical data as well as analyzed NAWM and lesion volume for MS subjects, and NWM volume for control subjects (italics). Standard deviations given in parentheses.  Histograms of MWF and 2T  are plotted for NWM for each control and NAWM for each MS subject in Figure 5.2.  2T  histograms were very consistent among normal subjects, while the histograms in MS showed more variation between subjects; some were simply shifted to longer 2T  values (example S10) while others had a shoulder on the long 2T  side of the main peak (example S04).  While MWF histograms of NWM for healthy controls varied in peak location and mean, the widths and peak heights were remarkably similar (standard deviations and peak heights were all within 0.8% and 2% of each other, respectively).  Conversely, the NAWM MWF histograms from MS subjects varied largely in shape and location.  125 0% 5% 10% 15% 20% 25% 40 60 80 100 120 140 T2 (ms) 0% 2% 4% 6% 8% 10% 12% 0% 5% 10% 15% 20% 25% 30% MWF C01 C02 C03 C04 C05 C06 C08 C09 C10 C11 C13 0% 5% 10% 15% 20% 25% 40 60 80 100 120 140 T2 (ms) 0% 2% 4% 6% 8% 10% 12% 0% 5% 10% 15% 20% 25% 30% MWF S01 S02 S03 S04 S05 S06 S07 S08 S09 S10 S11 S12 S13 C on tro l S ub je ct s M S Su bj ec ts MWF 2T C on tro l S ub je ct s M S Su bj ec ts  Figure 5.2: Normalized histograms of MWF (left) and 2T  (right) for NWM for all 11 healthy controls (top) and NAWM for all 13 MS subjects (bottom).  Note that for the MWF histograms the vertical scales are limited to a maximum of 12% for better visibility of detail, resulting in an intensity cutoff for the lowest histogram bin.  DTI-derived metric (FA, λ | , λ|| ,and <D>) histograms of NWM for each control and NAWM for each MS subject are plotted in Figure 5.3.  The NAWM FA curves were not smooth, which obscured intuitive results; the λ |  and λ|| histograms were smoother.  λ |  and <D> histograms were very reproducible across healthy controls, while in MS subjects the curves were shifted to larger values and were more variable in terms of peak location, width and height.  FA and λ|| histograms were less consistent across controls than λ |  or <D>.  126 0% 5% 10% 15% 20% 25% 30% 35% 40% 0 0.5 1 1.5 2 2 <D> (μm2/ms) .5 0% 2% 4% 6% 8% 10% 12% 14% 16% 0 0.5 1 1.5 2 2.5 λ  (μm2/ms) 0% 5% 10% 15% 20% 25% 30% 0 0.5 1 1.5 2 2.5 λ  (μm2/ms) 0% 5% 10% 15% 20% 25% 30% 0 0.5 1 1.5 2 2.5 λ  (μm2/ms) 0% 2% 4% 6% 8% 10% 12% 0 0.2 0.4 0.6 0.8 1 FA S01 S02 S03 S04 S05 S06 S07 S08 S09 S10 S11 S12 S13 0% 2% 4% 6% 8% 10% 12% 14% 16% 0 0.5 1 1.5 2 2.5 λ  (μm2/ms) 0% 5% 10% 15% 20% 25% 30% 35% 40% 0 0.5 1 1.5 2 2. <D> (μm2/ms) C on tro l S ub je ct s M S Su bj ec ts FA λ | , 5 λ|| <D> 0% 2% 4% 6% 8% 10% 12% 0 0.2 0.4 0.6 0.8 1 FA C01 C02 C03 C04 C05 C06 C08 C09 C10 C11 C13 C on tro l S ub je ct s M S Su bj ec ts  Figure 5.3: Normalized histograms of (from left to right) FA, λ | , λ||, and <D> for NWM for all 11 healthy controls (top) and NAWM for all 13 MS subjects (bottom).  The average histogram for NAWM and lesion across all 13 MS subjects, and for NWM across the 11 control subjects, are plotted for MWF, 2T , FA, λ | , λ||, and <D> in Figure 5.4.  The mean peak height, peak location, and average value for each parameter (standard deviation) are given in Table 5.2 for NAWM and lesion for MS subjects, and NWM and matched lesion ROIs (NWM) for healthy controls.  The percentage of voxels with MWF equal to zero is also given.  Significant differences between MS and control data (p < 0.05) are indicated in bold.  127 0% 5% 10% 15% 20% 40 60 80 100 120 140 T2 (ms) N or m al iz ed  V ox el  C ou nt MS NAWM Control NWM MS Lesion 0% 5% 10% 15% 20% 25% 30% 35% 0 0.5 1 1.5 2 2.5 <D> (μm2/ms) N or m al iz ed  V ox el  C ou nt MS NAWM Control NWM MS Lesion 0% 2% 4% 6% 8% 10% 12% 14% 0 0.5 1 1.5 2 2.5 λ  (μm2/ms) N or m al iz ed  V ox el  C ou nt MS NAWM Control NWM MS Lesion 0% 5% 10% 15% 20% 25% 0 0.5 1 1.5 2 2.5λ  (μm2/ms) N or m al iz ed  V ox el  C ou nt MS NAWM Control NWM MS Lesion 0% 2% 4% 6% 8% 10% 0% 5% 10% 15% 20% 25% 30% MWF N or m al iz ed  V ox el  C ou nt MS NAWM Control NWM MS Lesion 0% 2% 4% 6% 8% 10% 0 0.2 0.4 0.6 0.8 1 FA N or m al iz ed  V ox el  C ou nt MS NAWM Control NWM MS Lesion  Figure 5.4: Normalized average histograms for NAWM (solid blue line) and lesion (solid red line) across all 13 MS subjects, and for NWM (dashed blue line) across the 11 control subjects, for T2 relaxation- derived metrics on the left (MWF, top, and 2T , bottom) and DTI-derived metrics on the right (from top to bottom: FA, λ | , λ||, and <D>).   128 MS NAWM 7.8% (0.7%) 18% (2%) 8.7% (0.8%) 19% (3%) 13% (1%) 27% (5%) Peak MS Lesion 11% (3%) 12% (5%) 12% (2%) 14% (5%) 14% (5%) 16% (7%) Height Control NWM 8.6% (0.4%) 20% (1%) 8.5% (0.6%) 21% (1%) 13% (1%) 32% (3%) Control Matched Lesion ROIs 11% (2%) 24% (7%) 13% (3%) 29% (8%) 17% (6%) 43% (12%) MS NAWM 7% (2%) 73 (3) 0.36 (0.04) 0.68 (0.04) 1.20 (0.04) 0.87 (0.03) Peak MS Lesion 6% (3%) 92 (16) 0.26 (0.07) 0.8 (0.2) 1.4 (0.1) 1.0 (0.2) Location Control NWM 9% (1%) 71 (2) 0.37 (0.03) 0.66 (0.03) 1.17 (0.04) 0.83 (0.03) Control Matched Lesion ROIs 9% (2%) 74 (2) 0.40 (0.03) 0.67 (0.04) 1.2 (0.1) 0.85 (0.04) MS NAWM 7% (2%) 74 (2) 0.37 (0.02) 0.71 (0.04) 1.27 (0.04) 0.90 (0.03) Average MS Lesion 5% (2%) 93 (6) 0.30 (0.05) 0.9 (0.1) 1.4 (0.1) 1.1 (0.1) Control NWM 9% (1%) 73 (1) 0.39 (0.02) 0.66 (0.02) 1.23 (0.03) 0.85 (0.02) Control Matched Lesion ROIs 9% (2%) 75 (1) 0.41 (0.04) 0.66 (0.04) 1.24 (0.04) 0.85 (0.03) MS NAWM 10% (4%) % zeros MS Lesion 19% (13%) Control NWM 4% (3%) Control Matched Lesion ROIs 2% (2%) λ   (μm2/ms) <D > (μm2/ms)MWF T 2 (ms) FA λ   (μm2/ms)  Table 5.2: Histogram measures for MWF, 2T , FA, λ | , λ||, and <D>, averaged across all 13 MS subjects or all 11 controls, with standard deviation indicated in parentheses.  Significant differences between MS and control data are indicated in bold. MS lesion data was compared to matched lesion ROIs in controls for examination of significant differences.  Comparing NWM to NAWM, the percentage of data for which the histograms overlapped, from highest to lowest, was 94% for FA, 89% for 2T , 86% for λ||, 84% for λ | , 83% for MWF and 79% for <D>.  The NAWM MWF histograms had significantly lower peak heights (p = 0.02) and both NAWM and lesion MWF were significantly shifted to lower MWF peak locations (p = 0.04 and p = 0.005 respectively) and mean values (p = 0.01 and p < 0.001 respectively). Significantly more voxels had MWF values of 0 for NAWM (p = 0.001) and lesion (p = 0.001) than in NWM.  The average 2T  histogram in NAWM was shifted to significantly longer 2T  than NWM (p = 0.02).  Lesion 2T  histograms had significantly smaller peak heights (p < 0.001) and larger peak positions (p = 0.002) and average 2T  values (p < 0.001) than matched NWM in controls.  129  Of the reported diffusion metrics, <D> (peak height, locatin and average), λ |  (peak height and average) and λ|| (average) separated NAWM from NWM at the p < 0.05 level.  The differences apparent in the NAWM FA and <D> histograms from NWM were due to changes in λ | , since λ|| curves were quite similar between MS and control WM as seen in Figure 5.4 and Table 5.2.  Lesional WM was successfully separated from the matched NWM ROIs in controls by the peak location and average value of all diffusion metrics, as well as peak height for λ |  and <D>.  Figure 5.5 illustrates differences in histograms between 3 MS subjects and their respective age and gender matched controls, particularly between MWF and FA histograms (both thought to reflect changes in myelin).  MS subject S01 had NAWM and lesion MWF histograms that were clearly distinct from the NWM MWF histogram of the matched control C01; however, there was not a large shift in the lesion or NAWM histograms for FA.  MS subject S03 also had distinctly different MWF histograms from C03, with the NAWM and lesion histograms being quite similar to each other, but the lesion FA histogram was shifted considerably to lower FA values with little shift in the NAWM FA histogram.  Finally, the NAWM and lesion MWF histograms for MS subject S08 were much more similar to the histograms for C08, despite the fact that the lesion FA histogram was shifted to much lower FA values and the FA NAWM histogram was also shifted to lower FA.  130 0% 5% 10% 15% 20% 40 60 80 100 120 140 T2 0% 5% 10% 15% 20% 40 60 80 100 120 140 T2 0% 5% 10% 15% 20% 40 60 80 100 120 140 T2 0% 5% 10% 15% 20% 25% 30% 35% 0 0.5 1 1.5 2 2.5 <D> (μm2/ms) 0% 5% 10% 15% 20% 25% 30% 35% 0 0.5 1 1.5 2 2.5 <D> (μm2/ms) 0% 5% 10% 15% 20% 25% 30% 35% 0 0.5 1 1.5 2 2.5 <D> (μm2/ms) 0% 5% 10% 15% 0 0.5 1 1.5 2 2.5λ  (μm2/ms) 0% 5% 10% 15% 0 0.5 1 1.5 2 2.5λ  (μm2/ms) 0% 5% 10% 15% 0 0.5 1 1.5 2 2.5λ  (μm2/ms) 0% 5% 10% 15% 20% 25% 0 0.5 1 1.5 2 2.5λ  (μm2/ms) 0% 5% 10% 15% 20% 25% 0 0.5 1 1.5 2 2.5λ  (μm2/ms) 0% 5% 10% 15% 20% 25% 0 0.5 1 1.5 2 2.5λ  (μm2/ms) 0% 2% 4% 6% 8% 10% 12% 14% 0 0.2 0.4 0.6 0.8 1 FA 0% 2% 4% 6% 8% 10% 12% 14% 0 0.2 0.4 0.6 0.8 1 FA 0% 2% 4% 6% 8% 10% 12% 14% 0 0.2 0.4 0.6 0.8 1 FA 0% 2% 4% 6% 8% 10% 12% 0% 10% 20% 30% MWF 0% 2% 4% 6% 8% 10% 12% 0% 10% 20% 30% MWF 0% 2% 4% 6% 8% 10% 12% 0% 10% 20% 30% MWF S03 (EDSS 3.5) S08 (EDSS 3.5)S01 (EDSS 2.5) MS NAWM Control NWM MS Lesion  Figure 5.5: Normalized histograms of (from top to bottom) MWF, 2T , FA, λ | , λ||, and <D> for 3 of the MS subjects (one per column, solid lines) and their respective age and gender matched controls (dashed lines). NAWM and NWM are in blue, lesions are in red.   131 Maps of the NAWM/NWM masks of one slice for the same subjects from Figure 5.5 are shown in Figure 5.6, with grey representing non-zero MWF values and white indicating a zero value for the MWF.  It is interesting to note how many more zero MWF values are observed in the MS subjects than their age and gender matched controls, and that for S01 and S03 (with zero MWF percentages relative to the total number of voxels in their NAWM masks of 15% and 12%, respectively), the zeros tend to fall in the frontal and posterior white matter.  S08 (with a zero MWF percentage of 7%) had zero MWFs in similar regions to C08. Control Subjects MS Subjects C01 S01 C08C03 S03 C08 S08  Figure 5.6: White matter masks for three MS subjects (right) and their respective age and gender matched controls (left), with voxels with non-zero MWF values in grey and zero MWF values in white.   132 EDSS was significantly correlated with MWF NAWM average value (p = 0.02), as well as the percentage of zero MWF values (p = 0.04).  Disease duration was significantly correlated with peak height for <D> NAWM (p < 0.001) and λ |  NAWM (p = 0.001), and average λ|| (p = 0.01) as seen in Table 5.3. EDSS NAWM MWF Average -0.62 % zeros 0.58 NAWM λ Peak Height NAWM λ Average NAWM <D > Peak Height NAWM λ Peak Location Lesion λ Peak Location Peak Height Peak Location Average Lesion FA Peak Height -0.71 Peak Location 0.62 -0.57 Average 0.77 Lesion λ Peak Height 0.80 -0.70 Peak Location -0.92 0.60 0.74 Average -0.78 0.67 Lesion λ Peak Height 0.69 -0.68 -0.74 Peak Location 0.66 Average 0.66 Lesion <D> Peak Height 0.80 -0.74 Peak Location -0.73 0.67 Average -0.80 0.71 Lesion T 2 Lesion MWF Average -0.57 -0.86 NAWM MWF Average -0.58 Disease Duration -0.78 0.68  Table 5.3: Spearman correlation coefficients for significant correlations (p < 0.05) for EDSS and disease duration (top), MWF in NAWM and lesion (middle), and 2T  in lesion (bottom) with various diffusion histogram metrics.  None of the histogram metrics for MWF or 2T  were significantly correlated with any of the diffusion histogram metrics in NWM, and in MS only MWF average values and λ |  133 peak location were significantly correlated (see Table 5.3) (p = 0.04 for both NAWM and lesion).  No correlation was found between 2T  and diffusion histogram metrics in NWM, but several significant relationships were found in lesion (see Table 5.3).  5.4 DISCUSSION While conventional MRI is a powerful tool for diagnosing and monitoring MS, more sensitive and specific approaches to studying both MS lesions and NAWM are needed to gain insight into the pathogenesis of the disease.  Volumetric histogram analysis of <D> and FA provide a monitor for the extent of both visible (focal) and invisible (normal appearing) damage in MS brain.  <D> histograms of total MS brain tissue and NAWM have shown increases in peak location and average <D> as well as decreases in peak height compared to control brain tissue (Cercignani et al., 2000; Cercignani et al., 2001b; Iannucci et al., 2001; Nusbaum et al., 2000; Pulizzi et al., 2007; Rovaris et al., 2002; Wilson et al., 2001; Yu et al., 2007).  FA histograms of MS brain tissue and NAWM showed decreased peak location and average FA with increased peak height (Cercignani et al., 2001b; Pulizzi et al., 2007; Rovaris et al., 2002; Yu et al., 2007).  Diffusion eigenvalue histograms of MS brain are much less common in the literature, but mean λ |  has been found to be significantly increased in central brain normal appearing brain tissue (Henry et al., 2003; Oh et al., 2004).  When normal appearing brain tissue was divided into subsets of low, medium or high anisotropy, <D> and λ |  were significantly increased in normal appearing brain tissue compared to controls in all three subsets, but λ|| was only significantly increased in regions of low or medium  134 anisotropy, and FA was only significantly decreased in regions of high anisotropy (Henry et al., 2003).  Preliminary results showed increased 3rd quartiles in NAWM for the largest and second largest eigenvalues compared to controls (Chen et al., 2007).  <D> may well be sensitive to myelin or axonal loss, however it is also affected by inflammatory processes, gliosis, and the presence of debris.  FA is also influenced by many factors, and though values can be modulated by myelination, FA is highly sensitive to the degree of fibre orientation; for instance, FA is very low in regions of crossing fibres regardless of myelination.  λ |  and λ|| may provide more specificity than FA but are similarly affected by crossing fibres, and have not yet been extensively used in the study of MS, and thus further study is required before their relationship with pathological processes is well understood.  Multi-component T2 relaxation measurement yields metrics which provide information that has been shown to be complementary to that from DTI (Kolind et al., 2008; Vermathen et al., 2007). Although FA, λ | , and MWF have all been linked to myelin integrity, MWF has only been shown to correlate with FA or λ |  in certain situations.  In healthy CNS tissue, MWF was significantly correlated with both FA and λ |  (Bells et al., 2007; Mädler et al., 2008), although the correlation coefficients became much smaller if only WM regions were considered.  In NAWM, MWF was not found to correlate significantly with FA or λ |  (Kolind et al., 2008), but in that analysis, the NAWM ROIs were chosen to be contralateral to lesions and thus came from a small subset of brain structures, showing little variation in FA or λ |  values.  In MS lesion, a significant  135 correlation was only found between MWF and FA or λ |  for lesions with long T2 fractions (Kolind et al., 2008), thought to be lesions with extensive damage (Laule et al., 2007b). Thus MWF may provide a better understanding of what causes changes in diffusion parameters, and ultimately a better understanding of the pathogenesis of MS.  The previous single-slice acquisition restriction of the multi-echo T2 relaxation technique has historically prevented examination of the MWF or 2T  over a sizeable portion of the brain, thus limiting the extent of comparisons to other techniques and characterization of MWF and 2T  in MS brain.  In this study, the 5-fold increase in coverage allowed histogram analysis and more extensive comparisons to DTI metrics.  Unfortunately, due to time constraints in the MRI examination, the DTI sequence employed in this study was not optimal for a direct comparison with T2 relaxation imaging.  While nearly 20 minutes were dedicated to collecting T2 relaxation data, less than 4 minutes were spent acquiring DTI data.  While DTI is able to be acquired much faster, and thus should not necessarily be granted an equal acquisition time, the current acquisition suffered in terms of spatial resolution (with a matrix of only 128x96 as opposed to 256x128 for the T2 relaxation data).  Also, a slice thickness of 5 mm resulted in greater partial volume effects and crossing fibers than can be achieved with narrower slices, however this thickness was chosen to match the anatomy covered by the T2 relaxation sequence.  The large voxel size for the DTI data is expected to result in an underestimation of anisotropy due to the inclusion of more crossing fibers, and also less heterogeneity between DTI-derived metrics between voxels due to averaging, although it should be noted that decreased SNR from smaller voxels is known to underestimate  136 anisotropy and diffusivity as well (Oouchi et al., 2007).  In any case, this difference in resolution should be considered when comparing values, and may have contributed to the lack of correlation between DTI metrics and MWF.  5.4.1 Histograms across subjects (Figure 5.2 and Figure 5.3) From Figure 5.3, it is apparent that <D> was robust in normal controls, consistent with the literature (Cercignani et al., 2000; Cercignani et al., 2001b; Iannucci et al., 2001; Nusbaum et al., 2000; Pulizzi et al., 2007; Rovaris et al., 2002; Steens et al., 2004; Wilson et al., 2001; Yu et al., 2007).  λ |  was also very stable across control subjects.  FA and λ|| were slightly less robust, showing greater ranges of peak positions and shapes. MWF peak location and average were more variable for controls than the diffusion metric histograms, but the shape, height and width were very consistent.  As expected, there was more variation in MS histograms of λ | , <D>, 2T , and MWF than control histograms due to different disease states (FA histograms were too variable in controls to detect a difference from MS histograms , and λ|| is not expected to change significantly in MS (Henry et al., 2003; Kolind et al., 2008; Lin et al., 2007; Lowe et al., 2006)). Although MWF was the most variable, it has been shown (Vavasour et al., 2006) to have a high reliability coefficient (an estimate of the consistency of the subject’s ranking over time for a given measurement), and could reflect true biological differences between subjects, however larger sample sizes are needed to detect pathological changes.   137 5.4.2 Average histograms and parameter values (Figure 5.4 and Table 5.2) Both MWF NAWM and lesion histograms were shifted to significantly lower MWF values than NWM histograms (p = 0.01 for NAWM and p < 0.001 for lesion), presumably due to demyelination.  The NAWM and lesion histograms were both wider (larger standard deviation) than corresponding control histograms, indicating increased heterogeneity not only in acute lesions but also in NAWM.  Whittall et al. found that 2T  histograms in NAWM ROIs were displaced to longer 2T times, but maintained the same shape and peak height as corresponding control NWM ROIs, and concluded that changes in 2T  resulted from diffuse abnormalities such as axonal loss, demyelination, and inflammation, as opposed to small focal abnormalities with only a few affected voxels within the ROI (Whittall et al., 2002).  It was suggested that if the increased 2T  was caused by small focal lesions throughout the NAWM, the bulk of the NAWM voxels would have the same 2T  values as in NWM, with only a few voxels showing higher 2T  values, manifested as a bump on the histogram peak on the right side (higher 2T  values) with the main peak at the same position as for NWM but with lower peak height, as opposed to a general shift of the peak to higher 2T  values with the same shape and height as NWM.  In this study, the peak location of the average NAWM 2T  histogram was shifted to a significantly higher 2T  value (p = 0.02), but the peak height was also significantly reduced (p = 0.006) due to a small amount of skewness to higher 2T  values.  Looking at the individual 2T  histograms in Figure 5.2, histograms for some subjects appeared to be displaced without a change in shape while others had  138 distinctly lower peak heights.  When the locations of the voxels with longer 2T  values (more than 3 standard deviations from the mean for NAWM) were examined, they were found to be adjacent to regions of lesional white matter (but still within WM appearing normal on conventional MRI scans).  Therefore the voxels with long 2T  values were not scattered throughout the NAWM mask but concentrated along the borders of lesions.  If these voxels were excluded from the analysis of NAWM (due to association with known lesions), the 2T  peaks in NAWM were simply shifted to longer 2T  times with the same shape as for NWM, supporting the hypothesis of diffuse changes throughout the brain and not the suggestion of small focal lesions.  On average, the NAWM λ|| histogram was very similar to the NWM histogram, while the NAWM λ |  histogram was widened with a larger average than the NWM histogram, which resulted in the observed shift of the NAWM FA histogram toward lower FA values.  Both the λ |  and λ|| lesion histograms were shifted to higher values (p = 0.006 and p = 0.003, respectively), but the relative shift and change in λ |  histogram width were greater, leading to a significant shift of the lesion FA histogram to lower FA values (p < 0.001).  The <D> histograms for NWM, NAWM and lesion closely resembled those for 2T  in shape and difference between MS subjects and controls.  <D> histograms were consistent with those previously seen in the literature (Cercignani et al., 2000; Cercignani et al., 2001b; Iannucci et al., 2001; Nusbaum et al., 2000; Pulizzi et al., 2007; Rashid et al., 2004; Rovaris et al., 2002), although there is a large range of values due to differences in the amount of brain tissue included in the analysis (most included all brain  139 tissue as opposed to purely NAWM or lesion, and covered a greater portion of the brain) and the subject group characteristics.  5.4.3 Histogram differences between subjects (Figure 5.5 and Figure 5.6) While DTI histogram parameters, particularly for λ |  and <D>, tend to separate disease from control more efficiently, the true merit of T2 relaxation histograms becomes clear from specific case studies, as illustrated in Figure 5.5.  The information from the MWF and 2T  histograms is clearly complementary to that from the DTI histograms.  Although subjects S01 and S03 had similar analyzed lesion volumes, great differences are seen in the DTI histograms between the subjects, with the DTI histograms for S01 showing relatively small changes compared to C01 (particularly FA) and larger changes for S03 relative to C03, while the MWF histograms reveal considerable changes in NAWM and lesion myelin for both patients.  Since MWF and λ |  have both been linked to myelin integrity, it could be expected that the histograms would show similar changes for MS subjects compared to their age and gender matched controls, but λ |  showed much larger shifts in NAWM and lesion histograms toward larger values for S03 than S01, while the MWF changes were similar for both.  <D> was also more affected in S03 lesion and NAWM, and more voxels had longer 2T  values for S03 lesion, which implies the presence of more free water for S03. As λ |  is also expected to increase in the presence of additional water, this could explain the disparity between MWF and λ |  results.  It is also interesting to note that λ|| is expected  140 to decrease in the presence of axonal damage, to increase with myelin damage (unless it is an acute injury leading to more diffusion barriers), and to increase with free water, thus no axonal damage was detectable with any of the histograms, and the greater increase in lesion λ|| for S03 than S01 again could indicate more free water.  Conversely, for S08 (with the same EDSS as S03 but a much smaller analyzed lesion volume) the MWF histogram indicated less change in myelin for NAWM or lesions compared to NWM for C08 than S03/C03, while the NAWM and lesion DTI histograms were visibly affected.  This could indicate that much of the myelin in the NAWM and lesions for S08 remained but other pathological changes such as patchy edema were taking place.  The percentage of zero MWF values was also much lower for S08 than S01 or S03, and as can be seen in Figure 5.6, the eroded WM map and zero MWF voxels for S08 were much more similar to the control (C08) than for S01/C01 or S03/C03.  5.4.4 Correlations with EDSS and disease duration (Table 5.3) Several previous studies have failed to find a correlation between EDSS or disease duration and DTI histogram parameters (Cercignani et al., 2000; Iannucci et al., 2001; Nusbaum et al., 2000; Oh et al., 2004; Rashid et al., 2004; Rovaris et al., 2002), although others did find some significant but often weak correlations between EDSS or disease duration and <D>  (Cercignani et al., 2001b; Pulizzi et al., 2007; Wilson et al., 2001; Yu et al., 2007) and FA (Pulizzi et al., 2007) histogram parameters.  The inability to find significant correlations between diffusion metrics and EDSS may be due to the score’s bias toward locomotor disability, which is likely more dependent on spinal cord  141 involvement than changes in brain.  Also, studies that did find correlations tended to have MS cohorts with a wider range of clinical disability scores.  The NAWM average MWF decreased significantly with increasing EDSS (p = 0.02), while the percentage of zero MWF values increased (p = 0.04), implying that there was less myelin remaining for subjects with more severe disease symptoms.  A larger range of EDSS scores may have revealed more significant correlations.  The disease duration correlation with <D> peak height (p < 0.001), also detected by Pulizzi et al. (Pulizzi et al., 2007), and with λ |  peak height (p = 0.001) and λ|| average values (p = 0.01) (not previously reported), indicate that subjects with a longer disease duration have a larger spread of <D> and λ |  values and greater diffusion along the fibre direction, likely due to more brain regions with inflammation or axonal or myelin damage.  5.4.5 Correlations between MR-derived metrics (Table 5.3) Despite the fact that MWF, FA and λ |  have all been linked to changes in myelin, previous work in MS brain tissue has not revealed a significant relationship between MWF and FA or λ |  (Kolind et al., 2008).  We hypothesized that if data were acquired over a larger, more heterogeneous brain volume, relationships could be found to match those reported in normal brain tissue (Bells et al., 2007; Mädler et al., 2008).  The lack of correlations with MWF found here not only for NAWM but for NWM may be due to the relatively narrow range of the diffusion metrics over this brain region.  Past studies in  142 normal brain have focused on data from more inferior brain regions, such as the genu and splenium of the corpus callosum, major and minor forceps, internal capsules, head of the caudate nucleus, putamen, and thalamus, which as a group, have more variation in both T2 relaxation and DTI-derived metrics.  2T  was related to many DTI histogram parameters but only in MS lesion, where there is expected to be increased water, emphasizing the dependence on all of these metrics on water content.  5.5 CONCLUSION This study demonstrated that MWF and 2T  histograms derived from multi-component T2 relaxation imaging were different for multiple sclerosis subjects compared with age and gender matched controls, suggesting that these histograms could be a useful tool for observing subtle changes in myelination in NAWM. Histograms of diffusion tensor imaging-derived metrics were also shown to differ significantly from MWF histograms, implying that the application of multiple MR techniques with different sensitivities to the numerous pathological features of MS might lead to increased insight into MS pathophysiology.  5.6 ACKNOWLEDGMENTS Sincere thanks to the MS patients, control volunteers and the MR technologists at UBC Hospital. The research was supported by the Killam Trusts and the Multiple Sclerosis Society of Canada.  143 5.7 REFERENCES Bammer R., Augustin M., Strasser-Fuchs S., Seifert T., Kapeller P., Stollberger R., Ebner F., Hartung H.P., Fazekas F., 2000. Magnetic resonance diffusion tensor imaging for characterizing diffuse and focal white matter abnormalities in multiple sclerosis. Magnetic Resonance in Medicine 44, 583-91. Basser P.J., Pajevic S., Pierpaoli C., Duda J., Aldroubi A., 2000. In vivo fiber tractography using DT-MRI data. Magnetic Resonance in Medicine 44, 625-632. Beaulieu C., Does M.D., Snyder R.E., Allen P.S., 1996. Changes in water diffusion due to Wallerian degeneration in peripheral nerve. Magn Reson Med 36, 627-31. Bells S., Morris D., Vidarsson L., Comparison of linear combination filtering to DTI and MTR in whole brain myelin-water imaging. 15th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2007; Berlin, GER. p 1606. Biton I.E., Duncan I.D., Cohen Y., 2006. High b-value q-space diffusion MRI in myelin- deficient rat spinal cords. Magn Reson Imaging. 24, 161-6. Epub 2005 Dec 27. Cercignani M., Bozzali M., Iannucci G., Comi G., Filippi M., 2001a. Magnetisation transfer ratio and mean diffusivity of normal appearing white and grey matter from patients with multiple sclerosis. Journal of Neurology, Neurosurgery & Psychiatry 70, 311-7. Cercignani M., Iannucci G., Rocca M.A., Comi G., Horsfield M.A., Filippi M., 2000. Pathologic damage in MS assessed by diffusion-weighted and magnetization transfer MRI. Neurology 54, 1139-44. Cercignani M., Inglese M., Pagani E., Comi G., Filippi M., 2001b. Mean diffusivity and fractional anisotropy histograms of patients with multiple sclerosis. Ajnr: American Journal of Neuroradiology 22, 952-8. Chen Y., Jewells V., Sayed C., Markovic-Plese S., An H., Bullitt E., Lin W., A Spatial normalized study of diffusion tensor imaging in clinically isolated syndrome and multiple sclerosis patients. 15th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2007; Berlin, GER. p 2179. Christiansen P., Gideon P., Thomsen C., Stubgaard M., Henriksen O., Larsson H.B., 1993. Increased water self-diffusion in chronic plaques and in apparently normal white matter in patients with multiple sclerosis. Acta Neurologica Scandinavica. 87, 195-9. Ciccarelli O., Werring D.J., Wheeler-Kingshott C.A.M., Barker G.J., Parker G.J.M., Thompson A.J., Miller D.H., 2001. Investigation of MS normal-appearing brain using diffusion tensor MRI with clinical correlations. Neurology 56, 926-933. Hajnal J.V., Bryant D.J., Kasuboski L., Pattany P.M., De Coene B., Lewis P.D., Pennock J.M., Oatridge A., Young I.R., Bydder G.M., 1992. Use of fluid attenuated inversion recovery (FLAIR) pulse sequences in MRI of the brain. J Comput Assist Tomogr 16, 841-4. Hennig J., Scheffler K., 2001. Hyperechoes. Magn Reson Med 46, 6-12. Hennig J., Weigel M., Scheffler K., 2003. Multiecho sequences with variable refocusing flip angles: optimization of signal behavior using smooth transitions between pseudo steady states (TRAPS). Magn Reson Med 49, 527-35.  144 Henry R.G., Oh J., Nelson S.J., Pelletier D., 2003. Directional diffusion in relapsing- remitting multiple sclerosis: a possible in vivo signature of Wallerian degeneration. J Magn Reson Imaging. 18, 420-6. Iannucci G., Rovaris M., Giacomotti L., Comi G., Filippi M., 2001. Correlation of multiple sclerosis measures derived from T2-weighted, T1-weighted, magnetization transfer, and diffusion tensor MR imaging. AJNR Am J Neuroradiol 22, 1462-7. Keegan B.M., Noseworthy J.H., 2002. Multiple sclerosis. Annual Review of Medicine 53, 285-302. Kim J.H., Budde M.D., Liang H.F., Klein R.S., Russell J.H., Cross A.H., Song S.K., 2006. Detecting axon damage in spinal cord from a mouse model of multiple sclerosis. Neurobiol Dis. 21, 626-32. Epub 2005 Nov 17. Kolind S.H., Laule C., Vavasour I.M., Li D.K., Traboulsee A.L., Mädler B., Moore G.R., Mackay A.L., 2008. Complementary information from multi-exponential T(2) relaxation and diffusion tensor imaging reveals differences between multiple sclerosis lesions. Neuroimage 40, 77-85. Laule C., Kozlowski P., Leung E., Li D.K.B., MacKay A.L., Moore G.R.W., 2008. Myelin water imaging of multiple sclerosis at 7 T: Correlations with histopathology. Neuroimage 40, 1575-1580. Laule C., Leung E., Li D.K., Traboulsee A.L., Paty D.W., MacKay A.L., Moore G.R., 2006. Myelin water imaging in multiple sclerosis: quantitative correlations with histopathology. Mult Scler. 12, 747-53. Laule C., Vavasour I.M., Kolind S.H., Li D.K., Traboulsee T.L., Moore G.R., MacKay A.L., 2007a. Magnetic resonance imaging of myelin. Neurotherapeutics 4, 460- 84. Laule C., Vavasour I.M., Kolind S.H., Traboulsee A.L., Moore G.R.W., Li D.K.B., MacKay A.L., 2007b. Long T2 Water in Multiple Sclerosis: What else can we learn from multi-echo T2 Relaxation? Journal of Neurology 254, 1579-87. Laule C., Vavasour I.M., Moore G.R.W., Oger J., Li D.K.B., Paty D.W., MacKay A.L., 2004. Water content and myelin water fraction in multiple sclerosis: A T2 relaxation study. Journal of Neurology 251, 284-293. Lawson C.L., Hanson R.J., 1974. Solving Least Squares Problems. Englewood Cliffs, NJ: Prentice-Hall,. Le Bihan D., Basser P.J., 1995. Molecular diffusion and nuclear magnetic resonance. In: Le Bihan D., editor. Diffusion and perfusion magnetic resonance imaging. New York: Raven Press. p 5-17. Lin F., Yu C., Jiang T., Li K., Chan P., 2007. Diffusion Tensor Tractography-Based Group Mapping of the Pyramidal Tract in Relapsing-Remitting Multiple Sclerosis Patients. AJNR Am J Neuroradiol. 28, 278-282. Lowe M.J., Horenstein C., Hirsch J.G., Marrie R.A., Stone L., Bhattacharyya P.K., Gass A., Phillips M.D., 2006. Functional pathway-defined MRI diffusion measures reveal increased transverse diffusivity of water in multiple sclerosis. Neuroimage 32, 1127-33. MacKay A., Whittall K., Adler J., Li D., Paty D., Graeb D., 1994. In vivo visualization of myelin water in brain by magnetic resonance. Magnetic Resonance in Medicine 31, 673-7.  145 Mädler B., Drabycz S.A., Kolind S.H., Whittall K.P., MacKay A.L., 2008. Is diffusion anisotropy an accurate monitor of myelination? Correlation of multicomponent T2 relaxation and diffusion tensor anisotropy in human brain. Magnetic Resonance Imaging June 3, 2008 (Epub ahead of print), Mädler B., MacKay A.L., In-vivo 3D Multi-component T2-Relaxation Measurements for Quantitative Myelin Imaging at 3T. 14th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2006; Seattle, USA. p 2112. Mainero C., De Stefano N., Iannucci G., Sormani M.P., Guidi L., Federico A., Bartolozzi M.L., Comi G., Filippi M., 2001. Correlates of MS disability assessed in vivo using aggregates of MR quantities. Neurology 56, 1331-1334. Nusbaum A.O., Tang C.Y., Wei T.C., Buchsbaum M.S., Atlas S.W., 2000. Whole-brain diffusion MR histograms differ between MS subtypes. Neurology 54, 1421-1427. Oh J., Han E.T., Lee M.C., Nelson S.J., Pelletier D., 2007. Multislice brain myelin water fractions at 3T in multiple sclerosis. J Neuroimaging 17, 156-63. Oh J., Han E.T., Pelletier D., Nelson S.J., 2006. Measurement of in vivo multi- component T2 relaxation times for brain tissue using multi-slice T2 prep at 1.5 and 3 T. Magn Reson Imaging 24, 33-43. Oh J., Henry R.G., Genain C., Nelson S.J., Pelletier D., 2004. Mechanisms of normal appearing corpus callosum injury related to pericallosal T1 lesions in multiple sclerosis using directional diffusion tensor and 1H MRS imaging. J Neurol Neurosurg Psychiatry 75, 1281-6. Oouchi H., Yamada K., Sakai K., Kizu O., Kubota T., Ito H., Nishimura T., 2007. Diffusion anisotropy measurement of brain white matter is affected by voxel size: underestimation occurs in areas with crossing fibers. AJNR Am J Neuroradiol 28, 1102-6. Pulizzi A., Rovaris M., Judica E., Sormani M.P., Martinelli V., Comi G., Filippi M., 2007. Determinants of disability in multiple sclerosis at various disease stages: a multiparametric magnetic resonance study. Arch Neurol 64, 1163-8. Rashid W., Hadjiprocopis A., Griffin C.M., Chard D.T., Davies G.R., Barker G.J., Tofts P.S., Thompson A.J., Miller D.H., 2004. Diffusion tensor imaging of early relapsing-remitting multiple sclerosis with histogram analysis using automated segmentation and brain volume correction. Mult Scler 10, 9-15. Rashid W., Miller D.H., 2008. Recent advances in neuroimaging of multiple sclerosis. Semin Neurol 28, 46-55. Rovaris M., Bozzali M., Iannucci G., Ghezzi A., Caputo D., Montanari E., Bertolotto A., Bergamaschi R., Capra R., Mancardi G.L. and others, 2002. Assessment of normal-appearing white and gray matter in patients with primary progressive multiple sclerosis: a diffusion-tensor magnetic resonance imaging study. Arch Neurol 59, 1406-12. Song S.K., Sun S.W., Ju W.K., Lin S.J., Cross A.H., Neufeld A.H., 2003. Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage. 20, 1714-22. Song S.K., Sun S.W., Ramsbottom M.J., Chang C., Russell J., Cross A.H., 2002. Dysmyelination revealed through MRI as increased radial (but unchanged axial) diffusion of water. Neuroimage. 17, 1429-36.  146 Song S.K., Yoshino J., Le T.Q., Lin S.J., Sun S.W., Cross A.H., Armstrong R.C., 2005. Demyelination increases radial diffusivity in corpus callosum of mouse brain. Neuroimage. 26, 132-40. Stanisz G.J., Webb S., Munro C.A., Pun T., Midha R., 2004. MR properties of excised neural tissue following experimentally induced neuroinflammation. Magnetic Resonance in Medicine 51, 473-479. Steens S.C., Admiraal-Behloul F., Schaap J.A., Hoogenraad F.G., Wheeler-Kingshott C.A., le Cessie S., Tofts P.S., van Buchem M.A., 2004. Reproducibility of brain ADC histograms. Eur Radiol 14, 425-30. Stewart W.A., MacKay A.L., Whittall K.P., Moore G.R., Paty D.W., 1993. Spin-spin relaxation in experimental allergic encephalomyelitis. Analysis of CPMG data using a non-linear least squares method and linear inverse theory. Magnetic Resonance in Medicine. 29, 767-75. Sun S.W., Liang H.F., Le T.Q., Armstrong R.C., Cross A.H., Song S.K., 2006a. Differential sensitivity of in vivo and ex vivo diffusion tensor imaging to evolving optic nerve injury in mice with retinal ischemia. Neuroimage. 32, 1195-204. Epub 2006 Jun 22. Sun S.W., Liang H.F., Trinkaus K., Cross A.H., Armstrong R.C., Song S.K., 2006b. Noninvasive detection of cuprizone induced axonal damage and demyelination in the mouse corpus callosum. Magn Reson Med. 55, 302-8. Thomalla G., Glauche V., Koch M.A., Beaulieu C., Weiller C., Rother J., 2004. Diffusion tensor imaging detects early Wallerian degeneration of the pyramidal tract after ischemic stroke. Neuroimage 22, 1767-74. Tozer D.J., Davies G.R., Altmann D.R., Miller D.H., Tofts P.S., 2005. Correlation of apparent myelin measures obtained in multiple sclerosis patients and controls from magnetization transfer and multicompartmental T2 analysis. Magn Reson Med. 53, 1415-22. Vavasour I.M., Clark C.M., Li D.K., Mackay A.L., 2006. Reproducibility and reliability of MR measurements in white matter: clinical implications. Neuroimage. 32, 637- 42. Epub 2006 May 3. Vavasour I.M., MacKay A.L., Whittall K.P., Li D.K.B., Oger J., Paty D.W., A Serial Magnetic Resonance Study of Multiple Sclerosis: T2 Relaxation, T1 Relaxation and Magnetization Transfer. 51st Annual Meeting of the American Academy of Neurology; 1999; Toronto, Canada. p A360. Vavasour I.M., Whittall K.P., MacKay A.L., Li D.K., Vorobeychik G., Paty D.W., 1998. A comparison between magnetization transfer ratios and myelin water percentages in normals and multiple sclerosis patients. Magnetic Resonance in Medicine. 40, 763-8. Vermathen P., Robert-Tissot L., Pietz J., Lutz T., Boesch C., Kreis R., 2007. Characterization of white matter alterations in phenylketonuria by magnetic resonance relaxometry and diffusion tensor imaging. Magn Reson Med 58, 1145- 56. Vidarsson L., Conolly S.M., Lim K.O., Gold G.E., Pauly J.M., 2005. Echo time optimization for linear combination myelin imaging. Magn Reson Med. 53, 398- 407.  147  148 Webb S., Munro C.A., Midha R., Stanisz G.J., 2003. Is multicomponent T2 a good measure of myelin content in peripheral nerve? Magnetic Resonance in Medicine. 49, 638-45. Werring D.J., Clark C.A., Barker G.J., Thompson A.J., Miller D.H., 1999. Diffusion tensor imaging of lesions and normal-appearing white matter in multiple sclerosis. Neurology 52, 1626-1626. Whittall K.P., Bronskill M.J., Henkelman R.M., 1991. Investigation of analysis techniques for complicated NMR relaxation data. Journal of magnetic resonance 95, 221-234. Whittall K.P., MacKay A.L., 1989. Quantitative interpretation of nmr relaxation data. Journal of Magnetic Resonance 84, 64-71. Whittall K.P., MacKay A.L., Graeb D.A., Nugent R.A., Li D.K., Paty D.W., 1997. In vivo measurement of T2 distributions and water contents in normal human brain. Magnetic Resonance in Medicine 37, 34-43. Whittall K.P., MacKay A.L., Li D.K., Vavasour I.M., Jones C.K., Paty D.W., 2002. Normal-appearing white matter in multiple sclerosis has heterogeneous, diffusely prolonged T(2). Magnetic Resonance in Medicine 47, 403-8. Wilson M., Morgan P.S., Lin X., Turner B.P., Blumhardt L.D., 2001. Quantitative diffusion weighted magnetic resonance imaging, cerebral atrophy, and disability in multiple sclerosis. J Neurol Neurosurg Psychiatry 70, 318-22. Wu Y., Alexander A.L., Fleming J.O., Duncan I.D., Field A.S., 2006. Myelin water fraction in human cervical spinal cord in vivo. J Comput Assist Tomogr. 30, 304- 6. Xue R., van Zijl P.C.M., Crain B.J., Solaiyappan M., Mori S., 1999. In vivo three- dimensional reconstruction of rat brain axonal projections by diffusion tensor imaging. Magnetic Resonance in Medicine 42, 1123-1127. Yu C.S., Lin F.C., Liu Y., Duan Y., Lei H., Li K.C., 2007. Histogram analysis of diffusion measures in clinically isolated syndromes and relapsing-remitting multiple sclerosis. Eur J Radiol Zivadinov R., Stosic M., Cox J.L., Ramasamy D.P., Dwyer M.G., 2008. The place of conventional MRI and newly emerging MRI techniques in monitoring different aspects of treatment outcome. J Neurol 255 Suppl 1, 61-74.    Chapter 6  6 CONCLUSIONS  In this thesis, understanding of multi-component T2 relaxation imaging has been increased both in terms of interpretation of results (chapters 2 and 5) and in terms of the artifacts that affect images (chapters 3 and 4).  While the first manuscript chapter (chapter 2) provided insight into the meaning of multi-component T2 relaxation measurements (MWF and long-T2 fraction) in MS brain through comparison with DTI in various types of MS pathology, it was limited to a single slice of data for each patient.  The next chapter (chapter 3) took a step toward improving our ability to collect multi-echo T2 relaxation data through implementation at higher magnetic field strength.  The process of programming and developing the pulse sequence led to a deeper comprehension of the technical aspects of the data acquisition.  Chapter 4 made an even larger step toward improved data collection by validating a 3D version of the multi-component T2 relaxation technique, allowing much greater coverage within feasible scan times, while also providing further insight into the underlying physics of the technique.  Finally, chapter 5 returned to practical application in MS using the advancements made in the previous two chapters, particularly the increase in brain coverage (making histogram analysis possible and giving a greater range of brain tissue from which to interpret T2 relaxation findings).    149 6.1 ROI ANALYSIS COMPARING MULTI-COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 1.5 T 6.1.1 Hypotheses While it was thought that MWF and diffusion anisotropy (FA and λ | ) would be correlated in MS pathology since correlations have been seen in healthy controls (Bells et al., 2007; Mädler et al., 2008; Mädler et al., 2002), there were no significant relationships found in NAWM or lesions without a long-T2 component.  This result could suggest that the relationship breaks down in pathology, or simply that there was not a large enough range of values to detect a strong correlation.  Lesions with long-T2 components not only showed correlations between MWF and diffusion derived metrics, but were also significantly different in terms of each studied diffusion metric from lesions without long-T2 components, revealing a new class of lesion.  6.1.2 Significance The fact that DTI and multi-component T2 relaxation measures were not strongly related proved that they are not redundant measures, but provide important, complementary information.  The differences seen in lesions with long-T2 components suggested that there is a disparity in pathology for these lesions which may provide insight into disease progression.  150  6.1.3 Strengths and weaknesses A strength of this study was that the difference between results for NAWM and lesions without a long-T2 component versus lesions with a long-T2 component were very striking; correlations coefficients increased by factors ranging from 4.5 to 30.  Weaknesses included the limited coverage in the brain (a single slice), low field strength (and hence lower SNR than could be achieved at higher magnetic field), and inconsistent brain regions for various subjects.  Also, it is very difficult to interpret MRI results in terms of a specific pathology, as many different factors can affect measured values.  6.1.4 Potential applications and future work The source of the long-T2 signal will need to be investigated; histopathological studies may provide the key to this problem.  By scanning fixed brain with a multi-component T2 relaxation sequence and then matching areas of long-T2 signal to histological stains, it may be determined whether the cause is axonal damage, edema, or fibrillary gliosis, or some combination of all three.  Once the source of the long-T2 signal is determined, it may be useful to establish an MR technique capable of determining the presence of a long-T2 component which is less time consuming and analysis intensive for clinical use.   151 Applications include the use of long-T2 characteristics of lesions for MS research and determining prognoses, and the complementary use of multi-component T2 relaxation imaging and DTI for a more complete characterization of MS cases.  6.2 IMPLEMENTATION AND DEVELOPMENT OF MULTI- COMPONENT T2 RELAXATION IMAGING AT 3.0 T, AND VALIDATION AGAINST 1.5 T MEASUREMENTS 6.2.1 Hypotheses Implementation of the Poon-Henkelman multi-component T2 relaxation imaging sequence was indeed feasible at higher magnetic field, although pulse programming was required for the Philips Intera/Achieva MRI scanner.  While the variation of several scan parameters in water-based phantoms and fixed brain had significant effects on the measured MWF, 2T  and IE peak width values as well as residuals in the multi- exponential fit, very few significantly affected in vivo measurements.  The changes were explainable in all cases, increasing understanding of the artifacts that can occur using this technique.  While there was evidence that results were more affected by differences in magnetic field susceptibility (such as the decreased 2T values) and B1 inhomogeneity (such as the increased standard deviation of fit residuals in peripheral brain structures) at the higher field strength, results were well correlated between 1.5 T and 3.0 T.   152 6.2.2 Significance It was proved that the Poon-Henkelman T2 relaxation imaging technique could be successfully migrated to higher magnetic field, despite increased sensitivity to magnetic field susceptibility and decreased B1 homogeneity.  Higher SNR was achieved, and the technique can now be employed on a much larger number of scanners (those with magnetic field greater than 1.5 T).  Although there were not many advantages to moving to higher magnetic field for the T2 relaxation technique in particular, there are great benefits for other types of MRI scans (such as spectroscopy, susceptibility weighted imaging, or functional MRI) which provide additional information; multi-component T2 relaxation can now be included in examinations conducted at higher magnetic field for the benefit of other sequences.  6.2.3 Strengths and weaknesses The high correlation coefficient and low p-values with slope near 1 between MWF values at the two field strengths gives great confidence in 3T MWF values.  Very few other centres have the expertise to support this development as much of the work with this technique at 1.5 T was pioneered by the UBC group.  In fact, several other groups have attempted implementation at 3.0 T without success.  The major weakness of this work was that the experiments in vivo were not repeated for more volunteers, particularly during development of the 3.0 T sequence.  Though they were very patient, with the inherently long scan time of the sequence trying more parameter variations in a single session was not practical for live volunteers.  Also  153 matching the brain regions scanned was very challenging due to the limitation of acquiring only one slice, rendering registration difficult.  6.2.4 Potential applications and future work The pulse programming modifications made for the Philips scanner can now be shared with other users of Philips scanners at the same software level, hence the technique can be applied in more centres.  It would be beneficial to determine the consistency of results between centres using the same technique to study the reproducibility.  The availability could also allow this sequence to be compared directly to newer T2 relaxation acquisition techniques, such as those presented by Oh et al. (Oh et al., 2006), Vidarsson et al. (Vidarsson et al., 2005), Vermathen et al. (Vermathen et al., 2007), and Deoni et al. (Deoni et al., 2008).  Future work will include implementing and validating variable echo spacing for the pulse sequence so that later echoes can be acquired at considerably later times, allowing for the detection of long-T2 signals (Skinner et al., 2007).  The sequence could also be modified to employ variable TR in order to speed up acquisition (Laule et al., 2007a).   154 6.3 VALIDATION OF 3D MULTI-COMPONENT T2 RELAXATION IMAGING AGAINST THE 2D SINGLE-SLICE TECHNIQUE AT 3.0 T 6.3.1 Hypotheses The 3D version of the multi-component T2 relaxation imaging sequence resulted in qualitatively similar MWF maps to the 2D single-slice technique.  MWF values were highly correlated with a slope near one, and no significant bias from the Bland-Altman plot, giving confidence that MWF values obtained using the much more time efficient 3D sequence are consistent with 2D values.  Differences were observed primarily in peripheral brain structures, likely due to the poorer B1 homogeneity of the slab-selective refocusing pulses used for the 3D approach.  The 3D technique resulted in less flow artifact than the 2D single-slice technique, covered more brain in similar times, and typically had lower standard deviations of fit residuals.  6.3.2 Significance This manuscript is a significant advance for multi-component T2 relaxation imaging as it justifies the use of this much faster data acquisition, making the technique usable for many more applications.  No one has previously validated a rapid multi-component T2 relaxation imaging sequence directly against the gold-standard Poon-Henkelman single- slice technique.  The manuscript also provides insight into any potential pitfalls of using the 3D technique so that the sequence can be applied in an optimal fashion for each application, and results can be properly interpreted.  155  6.3.3 Strengths and weaknesses The significant correlations observed between MWF values obtained using each pulse sequence strongly validate the 3D technique, as do the lack of biases for the Bland- Altman plots.  Another strength of this study was that due to lack of availability of the Poon-Henkelman single-slice pulse sequence, very few other groups are able to compare results directly to single-slice results.  The work could be improved upon through validation in disease, as pathological cases may not show the same level of agreement between techniques as was found in healthy CNS tissue.  6.3.4 Potential applications and future work This sequence can now be easily used at other sites with similar MRI scanners, making the technique much more widely available.  In fact, it has already been shared with several centres.  Applications are copious, as there is great interest in the study of myelin for dozens of neurological diseases as well as for myelin development in general.  There has also been interest in using this technique for drug trials.  For multi-centre trials, a comparison between results at different sites would be useful. Other future work will include validation of variations of this technique, including  156 variable echo spacing, built-in B1 corrections, and a version employing a hybrid of spin echoes and gradient echoes.  6.4 VOXEL-WISE (HISTOGRAM) ANALYSIS COMPARING 3D MULTI-COMPONENT T2 RELAXATION IMAGING WITH DIFFUSION TENSOR IMAGING IN MULTIPLE SCLEROSIS AT 3.0 T 6.4.1 Hypotheses This manuscript illustrated the utility of 3D multi-component T2 relaxation imaging in practice through histogram analysis of NAWM and lesion in MS brain and NWM in control brain.  The MWF and T2 histograms provided a wealth of knowledge, with some common histogram features between subjects and yet large changes in other histogram parameters.  Despite the larger brain coverage, few correlations between diffusion anisotropy and MWF or 2T  were apparent in NAWM or lesion, and none were significant in NWM. This confirmed once again that T2 relaxation imaging and DTI provide complementary information, and the combined use leads to increased insight into the pathological features being imaged.  Few correlations were found between histogram metrics and level of disease, possibly due to the small range of disability of patients enrolled in the study.  157  6.4.2 Significance The increased variation in MWF and 2T  histograms in MS NAWM and lesion compared to NWM, and the fact that the information from T2 relaxation histograms did not simply mirror information from DTI histograms, prove that employing 3D multi-component T2 relaxation imaging in vivo to study disease or development is worthwhile.  It also provides a deeper understanding of what is represented by changes in DTI-derived histograms.  Changes in MWF and 2T  were consistent with literature, which continues the validation of the 3D multi-component T2 relaxation technique.  6.4.3 Strengths and weaknesses The much larger brain coverage made available by the 3D technique provided an abundance of data for each subject.  Also, with consistent brain regions studied, and age and gender matched controls, comparisons were more compelling between individual MS subjects, between the MS group and control group, or between individuals and their respective controls.  A larger cohort would be beneficial for strengthening any statistical tests.  Inclusion of a measure of total water content would aid in resolving the root cause of changes in DTI measures or MWF and 2T .  The DTI sequence had a limited resolution which may have  158 contributed to lack of observed correlations, and data registration is always a weak point, particularly with distortions common to DTI imaging.  6.4.4 Potential applications and future work Results from this study will facilitate informed interpretation of MRI findings.  The report of average differences observed between the MWF and 2T  histograms for a large portion of the brain of MS subjects and controls can be used as a benchmark for future observations.  Future work will include studying more patients, following the MWF and 2T  histograms for the MS patients over time to observe changes as the disease progresses, and investigating different types of lesion (enhancing, T1-hyperintense, T1-hypointense, etc). Histogram metrics will also be compared to traditional measures of disease such as T2 and T1 lesion volume and brain volume.  Results will also be compared to findings from other types of MRI scan.  6.5 RELATED WORK During the course of my PhD studies, I have co-authored several publications that relate (directly or indirectly) to the work presented in this thesis.  These publications are detailed below. • (MacKay et al., 2006): This manuscript provides a review of T2 relaxation imaging in brain.  159 • (Skinner et al., 2007): This study examined the effect of using variable echo spacing in T2 relaxation imaging to capture the decay of long-T2 components.  It was determined that using an extended echo spacing for the acquisition of the final echoes allows characterization of the long-T2 fraction without losing valuable short-T2 component information. • (Laule et al., 2007a): In this work, it was demonstrated that the acquisition time for T2 relaxation imaging can be reduced by varying TR across k-space, resulting in minimal loss of image resolution and no significant affect on proton density, MWF or 2T . • (Laule et al., 2007b): This manuscript provides a review of MRI of myelin. • (Laule et al., 2007d): This study described the observation of long-T2 signal in MS and phenylketonuria (PKU), supporting the usefulness of using a variable echo sequence as introduced by Skinner et al. (Skinner et al., 2007). • (Laule et al., 2007c): In this manuscript, the long-T2 component was further characterized in MS lesion and NAWM, including comparison to the magnetization transfer ratio, total water content, MWF, T1 and 2T . • (Mädler et al., 2008): This work carried out a comparison of MWF and diffusion anisotropy in healthy controls, demonstrating systematic differences between these measures (both thought to relate to myelin) in particular brain structures. • (Whitaker et al., 2008): This study employed a novel approach to short-T2 component analysis: examining the number of highly-myelinated voxels instead of the average MWF.  Using this approach, a strong correlation was reported between myelination of the corpus callosum and verbal intelligence quotient  160 scores in children.  This work also introduced a filter based on fit residuals to remove voxels not accurately fitting the modeled multi-exponential decay.   161 6.6 REFERENCES Bells S., Morris D., Vidarsson L., Comparison of linear combination filtering to DTI and MTR in whole brain myelin-water imaging. 15th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2007; Berlin, GER. p 1606. Deoni S.C., Rutt B.K., Arun T., Pierpaoli C., Jones D.K., Gleaning Multi-Component T1 and T2 Information from Steady-State Imaging Data. 16th Annual Meeting of the International Society of Magnetic Resonance in Medicine; 2008; Toronto, Canada. p 240. Laule C., Kolind S.H., Bjarnason T.A., Li D.K., Mackay A.L., 2007a. In vivo multiecho T(2) relaxation measurements using variable TR to decrease scan time. Magn Reson Imaging 25, 834-9. Laule C., Vavasour I.M., Kolind S.H., Li D.K., Traboulsee T.L., Moore G.R., MacKay A.L., 2007b. Magnetic resonance imaging of myelin. Neurotherapeutics 4, 460- 84. Laule C., Vavasour I.M., Kolind S.H., Traboulsee A.L., Moore G.R.W., Li D.K.B., MacKay A.L., 2007c. Long T2 Water in Multiple Sclerosis: What else can we learn from multi-echo T2 Relaxation? Journal of Neurology 254, 1579-87. Laule C., Vavasour I.M., Madler B., Kolind S.H., Sirrs S.M., Brief E.E., Traboulsee A.L., Moore G.R., Li D.K., Mackay A.L., 2007d. MR evidence of long T(2) water in pathological white matter. J Magn Reson Imaging 26, 1117-21. MacKay A., Laule C., Vavasour I., Bjarnason T., Kolind S., Madler B., 2006. Insights into brain microstructure from the T2 distribution. Magn Reson Imaging. 24, 515- 25. Epub 2006 Mar 20. Mädler B., Drabycz S.A., Kolind S.H., Whittall K.P., MacKay A.L., 2008. Is diffusion anisotropy an accurate monitor of myelination? Correlation of multicomponent T2 relaxation and diffusion tensor anisotropy in human brain. Magnetic Resonance Imaging June 3, 2008 (Epub ahead of print), Mädler B., Whittall K.P., MacKay A.L., Correlation of multicomponent T2- relaxation data with diffusion tensor anisotropy measures in human brain. International Society of Magnetic Resonance in Medicine; 2002; Honolulu, Hawaii. p 1188. Oh J., Han E.T., Pelletier D., Nelson S.J., 2006. Measurement of in vivo multi- component T2 relaxation times for brain tissue using multi-slice T2 prep at 1.5 and 3 T. Magn Reson Imaging 24, 33-43. Skinner M.G., Kolind S.H., MacKay A.L., 2007. The effect of varying echo spacing within a multiecho acquisition: better characterization of long T2 components. Magn Reson Imaging 25, 840-7. Vermathen P., Robert-Tissot L., Pietz J., Lutz T., Boesch C., Kreis R., 2007. Characterization of white matter alterations in phenylketonuria by magnetic resonance relaxometry and diffusion tensor imaging. Magn Reson Med 58, 1145- 56. Vidarsson L., Conolly S.M., Lim K.O., Gold G.E., Pauly J.M., 2005. Echo time optimization for linear combination myelin imaging. Magn Reson Med. 53, 398- 407.  162  163 Whitaker K., Kolind S.H., Mackay A.L., Clark C.M., 2008. Quantifying development: Investigating highly myelinated voxels in preadolescent corpus callosum. Neuroimage Jul 29, [Epub ahead of print].    APPENDIX A: ETHICS CERTIFICATES  This appendix contains copies of the UBC Research Ethics Board's Certificates of Approval for all research reported in this thesis.  164   165   166   167

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0066854/manifest

Comment

Related Items