UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Metabolite concentrations and myelin water fraction : correlations and trends within white matter measured… Moll, Rachel Francesca 2002

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

Item Metadata

Download

Media
831-ubc_2002-0502.pdf [ 5.93MB ]
Metadata
JSON: 831-1.0090620.json
JSON-LD: 831-1.0090620-ld.json
RDF/XML (Pretty): 831-1.0090620-rdf.xml
RDF/JSON: 831-1.0090620-rdf.json
Turtle: 831-1.0090620-turtle.txt
N-Triples: 831-1.0090620-rdf-ntriples.txt
Original Record: 831-1.0090620-source.json
Full Text
831-1.0090620-fulltext.txt
Citation
831-1.0090620.ris

Full Text

Metabolite Concentrations and Myelin Water Fraction: Correlations and Trends within White Matter measured with Magnetic Resonance Spectroscopy and T Relaxation Techniques 2  Rachel Francesca M o l l B. Sc. (Honours Physics) Trent University, 1999 B . Ed. Queen's University, 2000  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF T H E REQUIREMENTS FOR T H E D E G R E E OF M A S T E R OF SCIENCE  in  T H E F A C U L T Y OF G R A D U A T E STUDIES D E P A R T M E N T OF PHYSICS A N D A S T R O N O M Y  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA  September 2002  © Rachel Francesca Moll  I  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department  or by his or  her  representatives.  It is  understood  that  copying or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department of  Physic£ asnrJ As£/vs?0 s?rcj  The University of British Columbia Vancouver, Canada  •ate  DE-6 (2/88)  HrA  S^/ZOOX  Abstract  Absolute concentrations of in vivo human brain metabolites, N-acetyl-aspartate (NAA), choline (Cho) and creatine (Cre) were measured using single voxel spectroscopy. Myelin water fractions (MWF) were measured with a 48 echo T relaxation pulse sequence. 2  N A A is a possible neuronal marker, choline may be involved in myelin lipid synthesis and M W F represents the amount of water trapped between layers of myelin. Regional distributions and trends of measurements were examined and compared to similar measurements  made  with  magnetic  resonance  spectroscopic imaging (MRSI).  Measurements were made in four white matter regions in normal human brain, frontal white matter (FW), occipital white matter (OW), posterior internal capsules (PIC) and the splenium of the corpus callosum (SP). In the single voxel spectroscopy study the regional distribution of M W F and absolute concentrations of N A A , Cho and Cre were consistent with previous studies. Metabolite concentrations varied significantly between white matter structures. In some cases, variation was partially attributed to the amount of white matter in the voxel. In all white matter, trends between metabolite concentrations and M W F were weakly significant for N A A and creatine. Within structures weakly significant trends were also observed and varied between structures.  In the MRSI study M W F trends reproduced  previous work but spectroscopy measurements, which could not be absolutely quantified, produced no significant trends between structures or any consistent results between the three volunteers. Thus it remains unclear whether myelin water fraction and metabolite concentrations such as N A A or choline should.be related. The current literature provides conflicting results on both the function of N A A and the regional distribution between white and grey matter of metabolite concentrations. This study, the first to examine trends within white matter, contributes to a body of literature which aims to determine the function and importance of magnetic resonance spectroscopy visible metabolites and their possible contributions to clinical medicine.  ii  Table of Contents  Abstract  ii  List of Tables  vi  List of Figures  vii  Glossary  x  Acknowledgements  xii  1. Introduction  1  1.1 The Human Brain  1  1.2 Myelin and Signal Propagation  2  1.3 Brain Metabolites  3  1.3.1 N-acetyl-aspartate  4  1.3.2 Creatine  8  1.3.3 Choline  9  1.4 Brain White Matter Structures  10  1.5 Relevant Previous M R I and Spectroscopy Studies  12  1.6 Motivation  18  2. Basic NMR Theory  20  2.1 Introduction  20  2.2 Precession and Relaxation  21  2.3 N M R Signal Detection and Measurements  24  2.3.1 T Measurement 2  25  iii  2.3.2 T i Measurement  27  2.4 M R I Spatial Localization  28  2.5 Magnetic Resonance Spectroscopy  31  2.5.1 Water Suppression  32  2.5.2 Spectroscopy Localization Sequences  33  2.5.3 Magnetic Field Homogeneity and Shimming  35  3. Materials and Methods  37  3.1 Introduction  37  3.2 Single Voxel Study: Spectroscopy Acquisition and Analysis  37  3.2.1 Subjects and Voxel Placement  38  3.2.2 Pulse Sequence and Parameters  39  3.2.3 Spectroscopy Analysis: In Vivo spectral fitting  39  3.2.4 Spectroscopy Analysis: Quantification  41  3.3 Single Voxel Study: T Relaxation Acquisition and Analysis 2  3.3.1 Pulse Sequence and Parameters  45 46  3.3.2 T2 Relaxation Analysis: Non-Negative Least Squares Algorithm  46  3.3.3 T2 Relaxation Analysis: Segmentation  49  3.4 Spectroscopic Imaging Study: Acquisition and Analysis  50  3.4.1 Subjects and Voxel Placement  51  3.4.2 Pulse Sequence and Parameters  51  3.4.3 Spectroscopy Analysis: In Vivo spectral fitting  51  3.4.4 Spectroscopy Analysis: Quantification  52  iv  3.4.5 T Relaxation Analysis: N N L S and Segmentation 2  3.5 Statistical Analysis 4. Single Voxel Spectroscopy Study Results  53 53 56  4.1 Introduction  56  4.2 Metabolite Concentrations  56  4.3 Myelin Water Fraction  59  4.4 Correlations between M W F and Metabolite Concentrations  63  4.5 Correlations between Grey and White Matter Fractions and Metabolite Concentrations  70  4.6 Discussion  72  4.7 Summary  79  5. Spectroscopic Imaging Study Results  81  5.1 Introduction  81  5.2 Metabolite Concentrations and Myelin Water Fractions  81  5.3 Correlations between M W F and Metabolite Concentrations  85  5.4 Correlations between Grey and White Matter Fractions and Metabolite Concentrations  93  5.5 Discussion  95  5.6 Summary  98  6. Conclusions  100  6.1 Conclusions  100  6.2 Future Work  103  Bibliography  105  v  List of Tables  4.1 Comparison of CSF fractions calculated with two methods  57  4.2 Mean absolute and relative metabolite concentrations  58  4.3 Significant differences in metabolite concentrations between white matter regions  58  4.4 Mean myelin water fractions  63  4.5 Mean white matter fraction  70  4.6 Comparison of mean M W F in white matter between 1997 study [41] and the current study 4.7 Comparison of mean absolute metabolite concentrations between the current study and a similar study [32].  73  5.1 Mean myelin water fractions and number of good voxels for three M R S I subjects  73  84  vi  List of Figures  1.1  A diagram of a typical neuron  2  1.2  A myelinated axon  3  1.3  Chemical structure of N A A  4  1.4  Chemical structure of creatine  8  1.5  Chemical structure of choline  9  1.6  Lobes and structures of the human brain  10  1.7  A ) Tracts through the internal capsule B) Structures in right cerebral hemisphere  11  1.8  A typical T distribution with three compartments  13  1.9  Spin echo images and corresponding myelin maps for a normal and an  2  M S brain  14  2.1  Time evolution of magnetization vector M  23  2.2  Decay of T and T *  25  2.3  Spin echo timing sequence  26  2.4  C P M G pulse sequence with spatial localization  31  2.5  A typical human in vivo brain spectrum  31  2.6  CHESS water suppression pulse sequence  33  2.7  PRESS localization pulse sequence  34  2.8  S T E A M localization pulse sequence  3.1  Placement of spectroscopy voxels in white matter structures  38  3.2  LCModel output for single voxel spectroscopy study  40  2  2  "'  34  vii  3.3  T2 relaxation signal multi-exponential decay curve  45  3.4  Regularized and non-regularized N N L S solutions  49  3.5  LCModel output for MRSI study  52  4.1  Myelin map from current study  60  4.2  Bar graph showing comparison of mean M W F measured with 6 methods  61  4.3  Typical M W F histograms for white matter voxels  62  4.4  N A A concentration vs myelin water fraction for all white matter  65  4.5  N A A concentration vs myelin water fraction for individual structures  65  4.6  Choline concentration vs myelin water fraction for all white matter  66  4.7  Choline concentration vs myelin water fraction for individual structures  66  4.8  Creatine concentration vs myelin water fraction for all white matter  67  4.9  Creatine concentration vs myelin water fraction for individual structures  67  4.10 NAA/Cre vs myelin water fraction for all white matter  68  4.11 NAA/Cre vs myelin water fraction for individual structures  68  4.12 Cho/Cre vs myelin water fraction for all white matter  69  4.13 Cho/Cre vs myelin water fraction for individual structures  69  4.14 N A A concentration vs grey/white matter fraction  71  4.15 Choline concentration vs grey/white matter fraction  71  4.16 Creatine concentration vs grey/white matter fraction  72  5.1  Grid of M R S I spectra superimposed onto a C P M G image  82  5.2  Plots of mean metabolite concentrations of all regions within each subject  85  5.3  N A A concentration vs myelin water fraction for three subjects  87  5.4  Choline concentration vs myelin water fraction for three subjects  87  viii  5.5  Creatine concentration vs myelin water fraction for three subjects  88  5.6  [NAA] vs myelin water fraction for individual structures for subject A  88  5.7  [NAA] vs myelin water fraction for individual structures for subject B  89  5.8  [NAA] vs myelin water fraction for individual structures for subject C  89  5.9  [Choline] vs myelin water fraction for individual structures for subject A  90  5.10 [Choline] vs myelin water fraction for individual structures for subject B  90  5.11 [Choline] vs myelin water fraction for individual structures for subject C  91  5.12 [Creatine] vs myelin water fraction for individual structures for subject A  91  5.13 [Creatine] vs myelin water fraction for individual structures for subject B  92  5.14 [Creatine] vs myelin water fraction for individual structures for subject C  92  5.15 N A A concentration vs grey/white matter fraction for subject C  93  5.16 Choline concentration vs grey/white matter fraction for subject A  94  5.17 Creatine concentration vs grey/white matter fraction for subject B  94  ix  Glossary  Abbreviation  Meaning  CHESS  CHEmical Shift Selective Spectroscopy Pulse Sequence  Cho  Choline  CNS  Central Nervous System  CPMG  Carr-Purcell-Meiboom-Gill Pulse Sequence  Cre  Creatine  CSF  Cerebrospinal Fluid  CSI  Chemical Shift Imaging  FID  Free Induction Decay  FOV  Field Of View  FW  Frontal White Matter  GM  Grey Matter  MRI  Magnetic Resonance Imaging  MRS  Magnetic Resonance Spectroscopy  MRSI  Magnetic Resonance Spectroscopic Imaging  MWF  Myelin Water Fraction  NAA  N-Acetyl-Aspartate  [NAA]  N-Acetyl-Aspartate Concentration  NAA/Cre  Ratio between N A A and Creatine concentrations  NAAG  N-Acetyl-Aspartyl-Glutamate  NMR  Nuclear Magnetic Resonance  NNLS  Non Negative Least Squares  OW  Occipital White Matter  PIC  Posterior Internal Capsules  PRESS  Point RESolved Spectroscopy Pulse Sequence  rf  Radio Frequency  ROI  Region Of Interest  x  SNR  Signal to Noise Ratio  SP  Splenium of the Corpus Callosum  STEAM  Stimulated Echo Acquisition Mode Pulse Sequence  SVS  Single Voxel Spectroscopy Study  Ti  Spin-Lattice Relaxation Time  T2  Spin-Spin Relaxation Time  TE  Echo Time  TR  Repeat Time  WM  White Matter  WMF  White Matter Fraction  Acknowledgements  Attending graduate school in the Physics and Astronomy department at U B C has been a tremendous experience for me. I owe this to the talented colleagues, friends and mentors that I have encountered both during my time here and throughout my student career. First of all, I would like to thank my supervisor Alex MacKay for welcoming me into his research group. When I was hesitant and undecided about attending grad school it was Alex's encouraging words which helped me make the decision. His continued optimism, encouragement, wisdom and insight were paramount to the completion of this thesis. Secondly I must thank Ken Whittall whose opinion I hold in the highest regard and to whom I must credit most of my analysis tools. Elana Briefs previous spectroscopy research and the spectroscopy analysis tools she developed were also absolutely essential to the current work. Finally I would like to thank the rest of the research group, David, Burkhard, Craig, Corree, Irene, Matt, Evan, and Tony for their support and input throughout my degree. To my editors, Alex MacKay, Irene Vavasour and Ken Whittall, thank you for your detailed and thoughtful comments.  Thanks to my second reader,  Qing-San Xiang, for agreeing to read this thesis. Beyond my colleagues in the M R I research group, I would also like to thank those who helped make my masters a positive experience. To the crew in Room 100: Phil, Matt G., Matt S., Evan, Sumia, and Charmaine, thanks for always keeping spirits up in the room where the sun doesn't shine. A special thanks to Janis McKenna who is a pillar of support for all the students in this department but who also nominated me a prestigious award. The entire Physics and Astronomy Department has helped me become a better teacher and scientist by striving to provide many opportunities beyond academics and research. N S E R C and U B C provided the financial support for this work. On a personal note, I am proud to thank my incredible parents who inspire me with their success and support me with their love and most of all, who encouraged me to take advantage of this opportunity. I am in debt to all my friends and family, both near and far, who have helped me learn to work hard, to laugh often and to love.  xii  Chapter 1  Introduction  1.1 The Human Brain Metabolism and structure of the normal adult human brain are the focus of the current study. Specifically the myelin sheath of neurons and concentrations of metabolites N acetyl-aspartate (NAA), choline and creatine were studied with magnetic resonance imaging (MRT) and spectroscopy (MRS). This chapter will attempt to present what is currently known and understood about myelin and brain metabolites. Although highly complex and specialized, the brain is composed of only two types of cells, neurons and glial cells.  Neurons are the basic structural units of the  nervous system, designed to carry out the brain's activities. Glial cells support neurons and are 5 times more abundant [1]. Neurons vary widely in size and shape and are composed of a cell body, dendrites and an axon. See Figure 1.1. The cell body closely resembles that of other cells, containing a nucleus, organelles and cytoplasm.  Dendrites  are branched processes that extend from the cell body and respond to stimuli by conducting impulses to the cell body. The axon is a long, cylindrical process extending and carrying impulses away from the cell body.  Axons vary in length from a few  millimeters to over a meter in some parts of the spinal cord.  Axons can be  unmyelinated or myelinated. A myelinated axon is surrounded by a myelin sheath which aids propagation of impulses and will be described in more detail in the next section. Glial cells, so named from the Greek word for glue, provide structural and functional support to neurons. There are six categories of glial cells, among them oligodendrocytes, which form myelin layers around axons in the central nervous system (CNS) and astrocytes, which provide structural support by regulating the passage of molecules from the blood to the brain.  1  Figure 1.1: A diagram of a typical neuron showing its main components.  Tissue in the human brain can be classified into white matter (WM), grey matter (GM) and cerebrospinal fluid (CSF). The white matter contains the pathways throughout the CNS and is composed of groups of dendrites, myelinated axons of neurons and associated neuroglia. Grey matter contains deep grey matter structures and cerebral grey matter which surrounds white matter in an outer convoluted layer. It consists of either nerve cell bodies and dendrites or bundles of unmyelinated axons and neuroglia. The entire CNS is encased in a bony structure called the cranium and is bathed in cerebrospinal fluid that circulates in the hollow ventricles of the brain. CSF aids the brain's complex metabolic needs by acting as a medium for exchange of nutrients and waste products between the blood and the nervous tissue.  1.2 Myelin and Signal Propagation  Neurons respond to stimuli and react by sending impulses to other neurons in the form of action potentials. The nerve fiber is polarized by an abundance of sodium ions outside the axon's membrane creating a difference in electrical charge, the resting potential. When the nerve fiber is stimulated, its membrane depolarizes as sodium ions diffuse into the axon triggering an action potential along the axon. The action potential is a pulse of current that briefly depolarizes the neuron's plasma membrane. As one region of the membrane depolarizes it triggers depolarization in neighbouring regions and thus the  2  action potential moves along the axon. This process, called passive depolarization, occurs in non-myelinated axons and is relatively slow. The myelin sheath of the axon promotes more rapid propagation of the action potential. The myelin sheath is a lipid protein membrane which surrounds the axon in sections separated by small gaps called the nodes of Ranvier.  In the CNS,  oligodendrocytes myelinate axons by surrounding a small portion of the axon and creating myelin. One oligodendrocyte can have several extensions and can myelinate multiple axons. Myelin is deposited in tightly wound layers, as shown in Figure 1.2, and by insulating the axon, accelerates the depolarization process. When the action potential reaches a node it is amplified. The nodes of Ranvier contain many voltage-gated N a  +  channels where the action potential is rapidly depolarized and is regenerated, so that action potentials travel along the axon very quickly, jumping from node to node. Disruptions in this process lead to neurological dysfunction.  Figure 1.2: The axon is surrounded by sections of tightly wound layers of myelin separated by nodes of Ranvier.  1.3 Brain Metabolites  Magnetic resonance spectroscopy is sensitive to several brain metabolites including N A A , creatine, choline containing compounds, myo-inositol, glutamine and glutamate. There are many other metabolites, including alanine, G A B A , glucose, N-acetyl-aspartylglutamate (NAAG) and taurine, which make up the signal shown in Figure 3.2 but their  3  concentrations and signal to noise ratios (SNR) are so low that it is difficult to resolve their peaks from the baseline. The metabolites with the highest concentrations, N A A , Cho and Cre are quantifiable and were those studied in this work. They are described in detail in the next sections.  1.3.1  N-Acetyl-Aspartate  The highest peak in the in vivo H brain spectrum, N A A has the highest concentration of l  M R S visible metabolites and although it is most widely studied, its function remains unclear. The structure of N A A is shown in Figure 1.3. CH I 0 = C  3  NH x  C- CH - C- C 2  HO  /  I H  \ OH  Figure 1.3: Chemical structure of NAA. C H is the most prominent peak at 2.01 ppm, smaller peaks arise from C H protons. 3  2  Several independent forms of evidence support that N A A is located only within neurons. N A A is hardly found outside the nervous system [2]. hrimunocytochemical techniques, which stain for N A A , show it to be predominantly localized in neurons, specifically the axons and dendrites within the CNS, and do not stain glial cells [3]. It has been shown that there is no N A A in tumours derived from different types of glial cells [4]. Neuron specific toxins deplete levels of N A A [5]. Also studies of diseases which involve neuronal and/or axonal loss such as multiple sclerosis lesions and seizure foci, have all shown reduced N A A and clinically a decrease in N A A is used as an index of neuronal viability or integrity [6]. N A A levels have even been correlated with intelligence measures [7]. Therefore N A A is generally understood to be a neuronal marker and N A A levels have been equated with neuronal density. Recently evidence to the contrary has been presented. In 1992, Urenjak et al. [8] confirmed that N A A was localized in neurons using purified cell populations and high resolution M R S . They also showed that N A A  4  was  present  in  oligodendrocyte-type  2  astrocyte  progenitors  and  immature  oligodendrocytes, but not in mature oligodendrocytes. Mast cells have also been shown to contain levels of N A A [9] and finally N A A was measured in mature oligodendrocytes in 1999 [10]. These results are in vitro and do not necessarily represent in vivo properties of N A A . However there are conflicting clinical cases which also suggest that N A A may not be specific for neuronal processes. Examples include a large elevation of N A A in Canavan's disease due to a deficiency in aspartoacylase, an enzyme that degrades N A A to acetate and aspartate [11]. In these cases high N A A concentration does not indicate a sudden increase in neuronal density.  Incidences of reversible N A A losses suggest that  the initial decrease in N A A was not due to neuronal or axonal loss. Some examples of reversible N A A losses include multiple sclerosis, temporal lobe epilepsy and acquired immunodeficiency syndrome [12]. Most recently however, is a case of a 3 year old child with neurodevelopmental retardation and aberrant myelination for whom MRS has shown a global absence of N A A [13]. Clearly this child has some neurons, thus the function of N A A and the significance of its measurement by MRS is being discussed and questioned. In response, Bjartmar et al. [14] studied N A A in vivo in rat optic nerves undergoing axonal degeneration where it has been shown that oligodendrocyte progenitor cells and oligodendrocytes remain relatively stable while the myelinated axons degenerate rapidly. They found that axonal degeneration correlated with reduced N A A and axonal loss with undetectable N A A .  They believe this shows that N A A is a monitor of axon pathology  and axonal loss in white matter as it is a specific marker of myelinated axons in vivo [14]. Since N A A is synthesized in the mitochondria, N A A levels may also vary locally due to factors affecting mitochondrial activity, numbers associated with increased metabolic demand and /or neuronal size [14]. Besides being a neuronal marker the function of N A A is not well understood. Early studies hypothesized that N A A 1) is involved in lipid synthesis in the production of myelin [15], 2) acts as a regulator of protein synthesis [16], 3) is a breakdown product of another compound such as N-acetyl-aspartyl-glutamic acid (NAAG) or 4) serves as a storage form for aspartate [17]. A recent detailed study into the metabolism of N A A and the related N A A G may explain why these metabolites are so hard to study and understand.  It was well understood that both N A A and N A A G are synthesized in  5  neuronal mitochondria. This in itself is curious, as the neuron cannot actually use or hydrolyze these compounds. N A A does not enter into neuronal protein metabolism or into any other neuronal metabolic pathway with the exception of the few neurons that convert N A A to N A A G [18]. N A A is dynamic and turns over within a few days while cycling between neurons and oligodendrocytes where N A A is synthesized.  It was also  shown that N A A G cycles between neurons and astrocytes because its catabolic enzyme is only present in astrocytes. The N A A G process is strongly linked to N A A therefore the entire cycle is tri-compartmental and may be the only metabolic sequence which requires all three of the brain's major cell types. This metabolic sequence contains all the elements required for a signaling device that coordinates the complex physical and metabolic interactions of these cells. This leads to the conclusion that the primary role of N A A and N A A G could be neuronal-glial cell specific signaling and communication [18]. Thus these substances could control many aspects of development and interaction between brain cells and the maintenance of the nervous system. Another recent study points to a different possible function of N A A . Based on the fact that the acetyl group of N A A can be incorporated into brain fatty acids [15] this study sought to determine whether N A A could contribute acetyl groups for myelin lipid synthesis [19]. They found that purified myelin contained high levels of aspartoacylase an enzyme which releases acetyl groups from N A A , and that liberated acetyl groups within myelin could be incorporated into myelin lipids.  It was concluded that this evidence could point toward a strong link  between N A A and myelination as a possible function for the metabolite. The most recent study into the function of N A A hypothesized that N A A could act as a molecular water pump in myelinated neurons [20]. Molecular water pumps exist at the compartmental boundary of living cells and can actively pump water against its gradient. Myelinated neurons may require a special mechanism for metabolic transport of water since large portions of the cell surface are insulated from extracellular fluid by myelin.  The N A A  tri-compartmental cycle has all the characteristics of a molecular water pump and when the cycle is disrupted, as in Canavan's disease, the pathology is primarily associated with a water imbalance. Consequences of this study may be that N A A turnover would be a better measure of neuronal integrity and brain function. The confusion surrounding the function of N A A makes it difficult to interpret results for clinical applications.  6  Despite the debate, N A A has important clinical applications, some of which have been alluded to in the previous discussion.  High levels of N A A are measured in  individuals with Canavan's disease [6]. This is due to a lack of aminoacylase required to breakdown N A A .  The underlying pathology is unclear and consequences include  impaired myelin synthesis (perhaps due to N A A ' s role in donating acetyl groups for lipid synthesis.), a disruption of the N A A / N A A G metabolic sequence and spongy pathology due to a build up of fluid. As previously stated, N A A is decreased in many pathologies associated with neuronal damage or loss. Examples include, infarcted brain tissue after a stroke [21], chronic hypoxic-ischemic encephalopathy [22], and Alzheimer's disease [23]. In primary degenerative dementia, reduction is observed prior to significant brain atrophy and in the early stages of HIV infection, reductions in N A A occur before any structural changes in brain tissue are observed with M R I [24]. In many of these diseases N A A continues to decrease and can be used to monitor progress of neurological manifestations of the disease.  In multiple sclerosis (MS) reversible N A A was measured  in a large plaque which resolved, without treatment, and returned to normal, indicating that re-myelination correlated with N A A levels [25]. In another case, reduction of N A A was observed when histological examination showed inflammation with no evidence of demyelination or neuronal loss [26].  Thus changes in N A A may indicate neuronal  dysfunction in some cases instead of neuronal loss.  Significant changes in normal  appearing white and grey matter have also been observed with M R S suggesting it is more sensitive to M S pathologies than magnetic resonance imaging. M R S is also being used to study psychiatric disorders including social phobia, schizophrenia and bipolar disorder [6]. This discussion of the clinical applications of spectroscopy measurements of N A A is not meant to be exhaustive but to demonstrate the ways in which N A A is being used to examine disease clinically and to show that more research into the physiological function of N A A will have significant effects on the interpretation of N A A results and the use of N A A clinically.  7  1.3.2  Creatine  The creatine (Cre) signal is comprised of two chemicals, creatine (Figure 1.4) and phosphocreatine, which contribute overlapping resonances at 3.0 and 3.9 ppm (Figure 3.2). Phosphocreatine is essential for energy production in cellular processes. When A T P is depleted from a sudden demand for energy, the phosphocreatine reservoir replenishes A T P much faster than through catabolic pathways. The creatine kinase enzyme is the catalyst for this reaction which catalyzes phosphocreatine and A D P into creatine and A T P [27]. Normally most of the brain's energy is derived from oxidation of glucose but A T P can also be produced with the creatine kinase reaction in the brain [28]. Phosphocreatine is much more abundant in skeletal muscle than in the brain but when oxygen levels are low, increases in creatine/phosphocreatine are observed.  N H - C — N C H — CH — C — OH 3  Figure 1.4: Chemical structure of creatine.  The creatine signal is sometimes assumed to be constant and is used as a reference peak [29,30]. Regional variations of creatine have been shown in several studies and creatine concentration in grey matter is significantly higher than in white matter [31,32]. In vitro studies show that creatine is more related to astrocytes than to neurons [33]. Studies have also shed light on some functions of creatine beyond energetics. Creatine in the brain is controlled by distant liver and kidney enzymes, shown by reduced creatine in chronic liver disease and recovery after liver transplantation. and fall to maintain osmotic equilibrium.  Creatine levels also rise  This effect is observed in late hypoxic-  encephalopathy where creatine increases to maintain creatine kinase equilibrium in a residual cell population. Other clinical applications of creatine measurements include an increase in creatine in brains exhibiting trauma and hyperosmolarity. Creatine decreases in hypoxia, stroke, and tumors [34].  8  1.3.3  Choline  The choline (Cho) signal located at 3.24 ppm (Figure 3.2) arises from the N(CH3)3 groups of several water soluble choline containing compounds including glycerophosphocholine (GPC),  phosphocholine (PC) and small amounts of free choline. GPC and PC are  compounds involved in membrane metabolism, synthesis and degradation. The myelin sheath is a multilamellar membrane system containing a high proportion of lipids which insulates the neuronal axon. Choline (Figure 1.5) is a component of lipids and is thus theoretically associated with myelin although the choline head groups of the phosphatidyl-choline in myelin don't contribute much to the signal because they are bound together tightly to form the membrane. Some studies have shown that choline concentration is essentially constant through white matter and grey matter [35]. On the other hand, others have shown that choline shows the most regional variability in the brain and is significantly higher in white matter than in grey matter which correlates with myelin [36]. Measurements of GPC + PC concentrations post mortem confirm the spectroscopy results of about 1.5 m M [34]. CH  3  HOCH CH — N — CH +  2  2  CH  3  3  Figure 1.5: Chemical structure of choline  Similar to creatine, choline concentration can be affected by osmotic events. Links to biosynthesis and hormonal influences from organs outside the brain such as the liver are still being investigated.  Changes in the choline signal are interpreted as  disruptions in the membrane metabolism and myelin production. Elevated choline has been observed in cases of trauma, diabetes and tumors [34], for increased numbers of glial cells [37] and during active demyelination where myelin phospholipids are degraded by G P C [38]. Decreases in choline occur in hepatic encephalopathy [39], liver disease and stroke [34].  9  1.4 Brain White Matter Structures  Myelin and metabolite measurements were conducted in four different white matter brain structures shown in Figure 3.1. Results vary between structures therefore characteristics of each, frontal white matter (FW), occipital white matter (OW), splenium of the corpus callosum (SP) and the posterior internal capsules (PIC) will be described in detail. A l l four structures are located within the cerebrum, the largest part of the brain.  The  cerebrum is divided into 5 paired lobes within two hemispheres. The four lobes that can be observed on the surface are shown in Figure 1.6. Two hemispheres are connected by the corpus callosum. The corpus callosum is a large tract of white matter containing the genu of the corpus callosum at the anterior end and the splenium of the corpus callosum at the posterior end. The splenium is a large white matter bundle which connects the posterior regions of the two hemispheres. It includes all three types of white matter fiber tracts: association fibers which conduct impulses between neurons within the cerebral hemisphere, commissural fibers which connect neurons and gyri of one hemisphere with those of the other and projection fibers which ascend and descend to other parts of the brain including the spinal cord. A compact bundle of fibers from the splenium sweeps back into the occipital lobe to form the major forceps.  Figure 1.6: Lobes of the human brain.  The principal function of the occipital lobe is vision as it integrates eye movements by directing and focusing the eye. Visual association is also located in the occipital lobe. Lesions in the major forceps of the occipital lobe can lead to sensory difficulties and impaired reading ability.  The left occipital lobe and major forceps  10  transfer input from the right visual field to Wernicke's area on the left which is the language center where spoken and written words are interpreted. Frontal white matter is located in the frontal lobe, the anterior portion of the cerebrum. The functions of the frontal lobe include initiating voluntary motor impulses for the movement of skeletal muscles, analyzing sensory experiences and responses to personality,  memory,  communication.  emotions,  reasoning,  judgment,  planning  and  verbal  The minor forceps are located within the frontal white matter and  similar to the major forceps, are bundles of fibers from the genu of the corpus callosum extending into the frontal lobe. Frontal white matter contains association fibers running from the frontal cortex to multiple areas. Tracts also run between lobes forming long association fibers including a large white matter bundle at the edge of the corpus callosum, which fans out connecting the frontal, parietal and occipital lobes parallel to the corpus callosum.  Thalamo-  Figure 1.7: A) Tracts through the internal capsule B) Structures in the right cerebral hemisphere including internal capsule (IC) and surrounding structures.  The final white matter structure to be described is the internal capsule (IC). See Figure 1.7. The internal capsule is composed largely of white matter fibers running in many directions, connecting to the frontal region in the anterior limb and reaching to the  11  cerebral cortex from the posterior limb. (Figure 1.7A) Tracts contain corticospinal fibers which run from the cortex to the spine and are involved in motion and motor control. Other tracts include thalamic radiations beginning in the cortex and passing through the pons to the contralateral side of the brain or body and auditory radiations and corticopontine fibers from the temporal and parieto-occipital lobes.  The posterior  internal capsules are located between two grey matter structures, the putamen and the thalamus, which introduces partial voluming into the voxel. (Figure 1.7B)  1.5 Relevant Previous MRI and Spectroscopy Studies  The current study uses a magnetic resonance imaging in vivo T2 relaxation pulse sequence, developed by Dr. Alex MacKay's research group at the University of British Columbia, which allows the separation of three different water pools in the brain based on the distribution of T2 times in normal brain tissue [40,41]. Shown in Figure 1.8, the three water pools are 1) a short (10-50 ms) T2 component assigned to "myelin water" and  associated with the water compartmentalized within myelin bilayers, 2) an  intermediate component, between 70 and 95 ms, arising from intra and extra cellular water and 3) a long component with T2 greater than Is from cerebrospinal fluid. The short T2 component has been observed in other in vitro (eg. [42,43,44]) and in vivo (eg. [45]) studies of CNS tissue and in myelinated peripheral nervous system tissue (eg. [46]). A number of in vivo relaxation studies have failed to observe the short T2 component in white matter (eg. [47,48,49]) and most likely the reasons lie in the pulse sequences used [41].  The ability to resolve the fast relaxing myelin water component of the T2  distribution allows the quantification of myelin through calculation of the myelin water fraction, defined as the ratio of the signal from the short T2 component to the signal from the entire T distribution. The myelin water fraction is thought to be directly proportional 2  to the amount of myelin present. Several studies have been conducted to validate the myelin water fraction. T2 relaxation performed on guinea pigs with experimental allergic encephalomyelitis showed that demyelinated lesions correlated well with reduced myelin content [50]. Myelin maps of normal brain and diseased brain, shown in Figure 1.9, show that in white matter, where the amount of myelin is high, the signal from myelin  12  water is also high compared to grey matter and that in diseased brain demyelinated plaques appear as regions without myelin water [51]. Qualitative correlations between myelin maps and histological stains for myelin on fixed brains are very close [51,52]. Myelin water fractions are higher in white matter structures than grey matter structures and these results are very reproducible [41]. Specifically the internal capsules and the splenium of the corpus callosum have the highest myelin water fraction measurements (-12-16%) and grey matter structures such as the insular cortex and the cingulate gyrus have the lowest myelin water fraction measurements (~2 %) [41]. The total water content was also measured and was shown to be higher in grey matter than in white matter. Several studies have attempted to correlate other M R I measurements with myelin water content. Diffusion tensor fractional anisotropy measurements, sensitive to white matter nerve fiber tracts, vary for different brain matter structures in a similar manner to myelin water content [53].  M  y  Intra-cellular and Extracellular Water  e l i n  W a t o r  10  100  1000  10000  T2 Time (ms) Figure 1.8: Typical T distribution with three components and the water compartments to which they are assigned. 2  Magnetisation transfer ratio (MTR) has been shown to correlate with demyelination in a primate model [54]. When M T R was correlated with myelin water fraction in white matter a poor correlation was found throughout all white matter but when individual structures were examined correlations improved and varied between structures [55]. The current study attempts to find correlations between myelin water content and magnetic  13  resonance spectroscopy measures and to observe trends between the two measures throughout white matter.  C  D  Figure 1.9: A) Conventional spin echo image of a normal brain and B) its corresponding myelin map showing high myelin content in white matter areas. C) Conventional spin echo image of MS patient and D) its corresponding myelin map showing dark demyelinated areas where lesions appear on the spin echo image.  Magnetic resonance spectroscopy measures the concentration of M R S visible metabolites using two techniques.  Single Voxel Spectroscopy (SVS) measures one  spectrum from a localized voxel placed by the user in a region of interest. Magnetic Resonance Spectroscopic Imaging (MRSI) or Chemical Shift Imaging (CSI) is a combination of imaging and M R S where the entire field of view is divided into a matrix of voxels and for each a spectrum is recorded. Both techniques have been used to study regional variation of metabolites throughout the brain but the results are inconclusive. In 1999 Noworolski completed an excellent review of spectroscopy techniques and their application in the study of regional differences of metabolite concentrations [56]. The most consistent results occur in the measurement of choline which most studies find has higher levels in white matter than in grey matter [39,31,57,58,29,59,32,60]. This result is generally attributed to choline's role in lipid synthesis and myelin. Although Pouwels and Frahm [36] found choline to exhibit the most regional variation of the metabolites, other studies have found that it depends on the location and that in cortical tissue no  14  differences are seen between grey and white matter [35,56]. This result is also consistent with extract studies [5,61]. Wiedermann et al. found that measurements of choline with short echo time M R S I had low reliability compared to N A A and creatine [62]. In aging studies, choline is found to be highest in infants and remains constant as the brain ages from adulthood [63,59]. This is consistent with myelination during childhood. Most studies have found creatine concentration higher in grey matter than in white matter [62,39,31,64,59,36,60]. This result confirms biopsy studies [65]. It has been hypothesized that high concentrations of creatine in grey matter are due to increased demands for phosphocreatine due to the higher metabolic rate [64]. It was shown that creatine concentration varies throughout grey matter with higher concentrations in parietal than frontal grey matter [59].  Some studies found creatine to be constant  between grey and white matter [57,35] and one study found creatine concentrations were increased in white matter when compared to grey matter [58]. These results call into question the standard practice of calculating metabolite ratios relative to creatine. In general most studies agree quite closely for absolute concentrations and regional variation of choline and creatine. This is not the case for N A A . Studies which agree closely on choline and creatine results, often disagree on measurements of N A A concentrations. Many studies have measured N A A to be higher in grey matter than in white matter [56,66,30,67]. These results are supported by extract studies [61,5,65] and by the theory that as a neuronal marker, N A A should have a higher signal in grey matter than in white matter, where neurons have a higher density since axons actually constitute a small fraction of white matter space [58]. A comparable number of studies, including some recent ones, have found the opposite result, significantly more N A A in white matter than grey matter [35,57,58,39,59,62]. Schuff et al. [59] emphasize that the magnitude of the difference between white matter and grey matter for N A A is on the order of the measurement accuracy for N A A . A few studies found no grey/white differences in N A A [31,60]. Although the patterns between grey and white matter have yet to be firmly established, patterns within grey and white matter have also been studied. Studies using single voxel spectroscopy found that of all the metabolites and regions studied (usually frontal, parietal and occipital tissue), the only significant difference was elevated N A A in  15  the occipital lobe [36,68,69]. In these studies the voxels were large (8cc or greater) so they were not classified as either white matter or grey matter.  These results were  confirmed with MRSI where relative concentrations of NAA/Cho were significantly higher in the occipital lobe than in the frontal lobe [56]. It was also found that choline levels were very low in the occipital lobe [56,36]. In two separate studies, Tedeschi et al. studied N A A variations within grey matter and white matter [58,70].  Significant  differences between two white matter structures (centrum semiovale and the genu of the corpus callosum) were observed for N A A and no significant differences were found throughout cortical grey matter. One other study did find significant differences in N A A between frontal and parietal grey matter [62]. Differences between grey and white matter N A A were recently studied in a diseased state, hi relapsing remitting multiple sclerosis differences were seen in disease related changes of metabolite concentrations suggesting that pathological processes differ in the two tissues. For example, reductions in N A A were as strong in grey matter as in white matter but increased myo-inositol in white matter suggests that the degree of gliosis differs between the two tissue types [71]. Other relevant results from these studies include a failure to observe any effects due to gender or to asymmetry [36], although recently gender effects were demonstrated when Rogers et al. found that women had elevated levels of N A A in frontal white and grey matter [72]. Despite the fact that N A A is the highest peak in the in vivo spectrum it still seems to be the hardest metabolite to quantify and to understand. The range of concentrations reported in the literature for choline and creatine is much narrower than that for N A A and reported trends in creatine and choline throughout the brain conflict less than N A A trends. For example, comparing Kreis et al. [39], Michaelis et al. [31] and Hetherington et al. [35], choline concentrations range from 1.6 to 1.9 mmol, creatine concentrations range from 6.1 to 6.3 mmol and N A A concentrations range from 8.8 to 11.2 mmol in white matter.  Thus although different techniques were used, creatine and choline  concentrations agree well between studies where N A A results vary. Different acquisition and quantification techniques are the largest source of variation between the studies referenced in this review. The results of studies investigating regional differences of N A A may provide clues leading to the discovery of the function of N A A and its true pathological  16  significance. Currently the function of N A A is unknown thus changes in its signal are not well understood in terms of pathology and are of limited clinical use. The belief that N A A is a neuronal marker is supported by findings of high N A A concentration in grey matter where neuronal density is high. Links between N A A and myelin lipid synthesis are supported by high concentrations of N A A in white matter. Aging studies also show links between N A A and myelin.  In a study examining subjects from infancy to  adulthood an increase in N A A was observed in the first few years of life [63]. Since the number of neurons decreases shortly after birth, an N A A increase cannot be explained by N A A being a neuronal marker. N A A is not only present in cell bodies, but in axons, dendrites and synaptic terminals. A n increase in N A A during development may reflect the formation of dendritic arborizations and synaptic connections and would imply that N A A is a marker of functioning neurons. Another study found that N A A of white matter increased with age, this could be explained not by an increase of neurons, but by a decrease in intra-extra cellular water leading to an increase in axonal density [59]. Recently studies are being designed which aim to learn more about M R S quantities by correlating them to quantitative M R I measurements.  Carmanos et al.  presented correlations between NAA/Cre and diffusion measures of structural integrity of normal appearing white matter in multiple sclerosis. Strong correlations between the diffusion tensor imaging measurements of fractional anisotropy and NAA/Cre relative concentrations were found, also between NAA/Cre and measures of clinical disability and cerebro-functional compensation [73]. Quantitative measures of magnetization transfer (MT) as a measure of demyelination were compared to relative concentrations of N A A and no correlation was seen, perhaps indicating that demyelination may not lead to axonal damage [73].  In the same study a good correlation was achieved between  NAA/Cre and lesion volume weighted fractional size of restricted pool, said to represent myelin, which is speculated to reflect Wallerian degeneration of axons transected within lesions. NAA/Cre has also been correlated strongly with volumetric M R I measurements of atrophy. M S patients show an increase in atrophy during the progression of disease. Strong correlations between atrophy and NAA/Cre suggest that they measure the same or at least dependent effects and that axonal injury (possibly measured by MRS) coincides with a loss of periventricular white matter and cognitive function (as results also  17  correlated with cognitive tests) [74].  Volume changes in the thalamus of multiple  sclerosis patients have also been found to correspond to similar decreases in N A A concentration [75]. The current study endeavours to establish correlations between N A A , Cre and Cho concentrations and the quantitative M R I measure myelin water fraction.  1.6 Motivation  From the literature several facts seem clear: N A A is thought to be a neuronal marker or a marker of neuronal integrity, but its function is not well understood. Past studies of the regional distribution of metabolite concentration have failed to produce reliable, reproducible results, especially in white matter. Few studies have been conducted which correlate M R S measures to quantitative M R I measures. Therefore the objectives of the current study are:  •  To use established single voxel spectroscopy pulse sequences and quantification methods to measure N A A , choline and creatine absolute concentrations in four white matter structures.  •  To determine i f N A A , choline and creatine absolute concentrations vary throughout white matter.  •  To use established T2 pulse sequences and analysis techniques to measure the myelin water fraction in the same regions where SVS was conducted.  •  To compare trends between myelin water fraction and absolute N A A , Cho and Cre concentrations in all white matter and in individual structures.  •  To compare SVS results in white matter to magnetic resonance spectroscopic imaging.  •  To use MRSI to examine trends in both white matter, and individual brain white matter structures with more voxels and better spatial resolution.  Ultimately the results of these studies will help improve understanding of the anatomy of the brain through the study of neurons. The aim is to establish quantitative relationships between two types of measurements which each examine properties of neurons. Results  18  of these studies will indicate whether N A A , Cho and Cre concentrations are related to myelin water fraction measurements and whether these measurements examine similar or complimentary neuronal properties. They will also contribute additional information to the literature about regional distribution of metabolites and spectroscopy quantification techniques which will hopefully add clues to the continuing debate surrounding N A A , its location, its function and ultimately its use clinically.  19  Chapter 2  Basic NMR Theory  2.1 Introduction LL Rabi first measured the nuclear magnetic moment in 1938 [76]. His work was the beginning of nuclear magnetic resonance (NMR) experiments and was recognized with a Nobel Prize in physics in 1944. Nuclear magnetic resonance is a phenomenon which describes the interaction between a nucleus and an external magnetic field.  NMR  sensitive nuclei must possess a magnetic moment as a result of angular momentum. Nuclei with an odd number of nucleons have a non-zero total angular momentum.  This  can be explained by looking at atomic structure in detail. The total angular momentum, / , of the nucleus is determined by the balance of the angular momenta of its composite nucleons.  Due to the Pauli exclusion principle,  nucleons experience a pairing force which favours the coupling of nucleons so that their orbital angular momentum and spin angular momentum each add to zero. Each nucleon is described by three quantum numbers, I, s, and j, corresponding to the orbital, spin and total angular momentum of the nucleon. For a particle with spin s = Vi its projection onto the Z axis is m = ± A. In nuclei with an even number of protons and neutrons the l  s  nucleons form spin zero pairs where a nucleon with m = + A pairs up with a nucleon with l  s  m = - Vi. Thus the total angular momentum is zero. The nucleus of hydrogen ( H) !  s  contains one proton (1=0, s = A). In this case the total angular momentum 1= l + s = A. l  l  The magnetic moment of a nucleus is related to the angular momentum by  \i = y[  (2.1)  20  where y is the gyromagnetic ratio. The gyromagnetic ratio is unique for each nucleus and for ' H , y ~ 2.675 x 10 s^T" . 8  1  According to Equation 2.1, when the total angular  momentum of a nucleus is zero it has no magnetic moment. In order for a nucleus to be suitable for N M R is must have a high natural abundance to create a detectable signal. Nuclei that are commonly used for biological studies include ' H with a natural abundance of 99.985%, C (1.11%) and P (-100%). In a given sample the natural abundance is 1 3  3 1  the percentage of the particular isotope. Studies in this thesis used the hydrogen nucleus which is very useful in both clinical applications and research because of its high sensitivity due to a high Larmor frequency, high signal to noise and high natural abundance.  2.2 Precession and Relaxation  N M R describes the effects of a nucleus with non-zero angular momentum in an external magnetic field. The equations of motion of a nucleus in an external magnetic field can be derived using both quantum and classical physics. The final equations of motion do not contain the Planck constant implying that the classical explanation will be sufficient. In classical mechanics the rate of change of the total angular momentum, / , equals the total torque, T, exerted on the system:  (2.2)  When a sample is placed in an external magnetic field some protons assume the low energy configuration and align with the field, others against.  Boltzmann statistics show  that the probabilities of the nuclei being in each energy state are not equal. There is a small excess of spins in the lower energy state, aligned with the magnetic field, which gives rise to a net magnetization of the sample, M. The net magnetization is very small, for two million protons at room temperature in a 1.5T magnetic field there are nine more nuclei in the lower energy state!  In an external magnetic field, Bo, the sample  experiences a torque between the field and the total magnetization such that  21  (2.3)  Substituting Equation 2.1 into Equation 2.3 gives the following, which describes the precession of the magnetization of a sample around the external magnetic field.  ^ - = (MxB ) at r  (2.4)  0  The precession of M in an external magnetic field is almost exactly analogous to the precession of a spinning top in a gravitational field. The rate at which M precesses is called the Larmor frequency, co . and is given by the Larmor equation. 0  a =yB 0  (2.5)  0  When a radio frequency (rf) pulse Bj is applied perpendicular to the static field Bo at the Larmor frequency, the energy will be absorbed by the sample and the net magnetization vector M will tip away from B by an angle a where 0  a = yB{u  (2.6)  and T is the length of time the rf pulse is applied. Once M has been tipped away from Bo the magnetization will theoretically precess forever according to Equation 2.4. In reality the magnetization will relax in two directions back to its original position. This motion gives rise to two important N M R measures, Ti and T . 2  The magnetization relaxes  longitudinally (along Bo and towards Mo) as precession is affected by interactions between the nucleus and the lattice environment.  The rate of the magnetization's  longitudinal relaxation is characterized by Ti or the spin-lattice relaxation time.  The  magnetization also relaxes transversely (orthogonal to Bo) and is the result of variations of the Larmor frequencies of the composite nuclei caused by both molecular interactions  22  and field inhomogeneity. This rate is characterized by T and is also called the spin-spin 2  relaxation time. Equation 2.4 can be modified to describe these effects.  ^  = r(MxB )- ^ M  dt  +  0  T  M  ^ -  (  2  M  z  T  M  °  )  (2.7)  S  x  This more complete equation of motion is called the Bloch equation where x, y and z are unit vectors in the X , Y and Z directions, M , M and M are the x  y  z  components of M in these directions, M is the total magnetization at equilibrium, and B 0  0  is the external magnetic field in the Z-direction. The time evolution of M is demonstrated in Figure 2.1.  (a)  (b)  Figure 2.1: Time evolution of the magnetization vector M (a) M , My, and M as functions of time, (b) x  z  The motion of M vector in 3D. The M vector undergoes a Larmor precession with coo = y Bo  a s w  e  u  as T i  and T2 relaxations, and eventually reaches its thermal equilibrium state MQ. (reproduced with permission from S. Xiang)  23  The graphical representations of the magnetization over time show that initially M (0) = x  0, M (0)= M ° y  xy  , and M (0)=M  and at a time t after the rf pulse is applied the  0  z  z  magnetization can be described by the following expressions  M ^ ( 0 = M^sin(ffl 0exp(-//7' )  (2.8)  M (t) = Mly cos(a> f) exp(-r / T )  (2.9)  M (t)  (2.10)  0  0  Y  2  = M +(M° -M )ex (-t/T )  z  From these equations T i and T  2  0  2  z  0  V  l  of the system can be described in more detail.  Expressions 2.8 - 2.10 describe the magnetization of a simple spin system undergoing isotropic motional narrowing which occurs when the translation motion of the nucleus allows the spin to sample a range of magnetic fields, thus narrowing the resonance [77]. Equation 2.10 shows that the magnetization approaches its maximum value when t »  Tj  when all the magnetization has equilibrated with the lattice environment. It also shows that T] relaxation takes place along the Z axis, parallel to the external field BoEquations 2.8 and 2.9 show that T relaxation occurs in the X Y plane, orthogonal to the 2  external field. After a time t the magnetization in this plane decreases as inhomogeneous spins gradually become out of phase. Eventually as t  —>oo,  M x and M y will approach  zero. The transverse motion of the magnetization vector in an external field, as described by the Bloch equation, is key to detecting a signal from a sample in N M R .  2.3 NMR Signal Detection and Measurements  To detect an N M R signal, a sample with a net magnetization is placed in an external magnetic field and pulsed with and rf pulse at the Larmor frequency which, for typical clinical magnets of 1.5T, is in the radio frequency range.  This rf pulse tips the net  magnetization by an angle a given by Equation 2.6. When a = 90° the magnetization is tipped into the X Y plane with no magnetization along the Z axis.  After the rf pulse is  turned off the magnetization will continue to precess and will decay transversely by T relaxation.  2  According to Faraday's Law the time varying magnetization induces a  24  current in the receiver coil of the magnet. This signal is called the free induction decay (FID).  2.3.1 T2 Measurement  The spins at different locations precess at slightly different rates, thus the Larmor frequency of the sample is not unique. This causes the spins to dephase as they rotate in the X Y plane and reduces the magnetization and FID. Magnetic field inhomogeneities cause the transverse magnetization of a sample to decay faster than the predicted exp(-t/T2) which only takes molecular interactions into account.  Thus the actual  transverse relaxation time is called T* and is a combination of T2 effects from both molecular interactions (pure T 2 effect) and field inhomogeneity (inhomogeneous T 2 effect). Figure 2.2 demonstrates the relationship between T and T *. 2  2  Signal •  exp(-t/T ) 2  > Time  Figure 2.2: Individual echoes decay at a rate characterized by T * (dotted line) where the signal over time decays according to true T (dashed line). 2  2  The loss in signal can be avoided using spin echo experiments. In 1950, Hahn discovered that multiple rf pulses could be used to rephase and recover lost magnetization [78]. Hahn used two 90° pulses separated by time x to detect a strong signal from an echo at time 2x. Carr and Purcell modified the Hahn sequence to create the commonly used CP experiment [79]. They used a 90° pulse in the X direction to tip the spins into the X Y plane pointing in the Y direction. See Figure 2.3. As the spins relax they also dephase due to field inhomogeneity. After a time x an 180° pulse in the X direction is applied. This  25  pulse rotates the spins about the X axis, thus changing their phase by n radians. They continue to precess at the same angular frequency and eventually after a time 2x the spins will refocus in the - Y direction creating an echo. The time at which an echo occurs is called the echo time (TE).  90° R F  I  180° T  — H  H  T — + •  Signal  t  1  FID  Echo  Figure 2.3: Spin echo timing sequence. After time 2T spins are refocused and an echo produces a strong induction signal. For CP 180° pulse is in X direction, for CPMG 180° pulse is in Y direction.  Meiboom and Gill modified the CP sequence to create the C P M G pulse sequence [80].  In the C P M G sequence the 180° pulse at time x is applied in the Y direction.  Therefore the spins are flipped about the Y axis and refocus in the positive Y direction. This improves on the CP experiment because it corrects for errors arising from imperfect 180° pulses. When the spins are flipped about the X axis in the CP experiment the errors would accumulate with each flip. When the pulse is applied about the Y axis these errors cancel each other out, every other echo. The CP and C P M G sequences can be extended by repeating 180° pulses. When the 180° pulse is applied at times x, 3x, 5x ... then echoes are detected at times 2x, 4x, 6x .... The signal intensity of the echoes decays according to true T where 2  2M T  S(2«r) = S(0)exp(  -)  (2.11)  ^2  Therefore spin echo experiments can be used to measure the true T2 of a sample without being affected by field inhomogeneities. It should be mentioned that T2 relaxation is also  26  affected by the diffusion rates of the nuclei. The CP sequence can minimize the effects of diffusion since the relative contribution of diffusion compared to the T term depends 2  on x. It is possible to vary and minimize diffusion without affecting T [77]. 2  2.3.2 T i Measurement  The longitudinal relaxation time T i characterizes the interactions between the spin nuclei and their lattice environment. It reflects the rate at which the longitudinal component of the magnetization returns to thermal equilibrium after being perturbed by the rf pulse. A saturation recovery experiment measures Ti using a series of 90° pulses separated by a repetition time (TR). After each pulse the corresponding FID is detected. If TR is much larger than T , then the transverse magnetization is assumed to be fully decayed between 2  rf pulses.  Therefore any change in magnetization between excitations is due to  longitudinal T i relaxation according to the following equation, arising from the solution of the Bloch equation (Equation 2.7).  77? M (77?) = M [ l - e x p ( - — ) ] z  0  (2.12)  where M (TR) is the magnetization just before the next rf excitation and Mo is the initial Z  magnetization. When this experiment is repeated at several TRs the dependence on Ti can be calculated. The inversion recovery (IR) experiment is a modification of the saturation recovery experiment. It begins with a 180° pulse so that the magnetization is inverted into the - Z direction, opposite to the direction of its thermal equilibrium value. After an inversion time (TI) a 90° pulse is applied and an FID is detected. The IR experiment is more accurate because the magnetization moves through a greater range.  27  2.4 MRI Spatial Localization  The T i , T and proton density weighted signal intensities of a sample can be displayed as 2  two dimensional images using three basic signal localization techniques, slice selection, frequency encoding and phase encoding. These techniques form the basis of magnetic resonance imaging (MRI). Spatial localization in MRI is made possible by superimposing a magnetic field gradient G along the external magnetic field Bo. The magnetic field gradient perturbs the field linearly in each direction according to the following equation:  Afl = r • G = xG + yG + zG x  y  2  (2.13)  where G = dB /dx assuming B is along the Z axis. When the gradient is combined with x  z  0  the static external magnetic field the following non uniform magnetic field is created.  B(r) =  z(B +vG) 0  (2.14)  If a magnetic field gradient G is applied in a certain direction then the nuclei in the plane perpendicular to G will have the same Larmor frequency. B y combining Equations 2.5 and 2.14 and specifying the Z direction, the relationship between the frequency, slice thickness and gradient magnitude can be determined.  a> = y(B  0  +zG ) 2  (2.15)  When a gradient G is applied, each slice of the sample (the location of which is z  determined by z) will have a unique center frequency and its thickness (Az) can be written in terms of the frequency bandwidth, Aco, of the rf pulse and the magnitude of the gradient.  Az =  Aco 7G  (2.16)  2  28  Therefore while the gradient is applied to a sample, an rf pulse can be designed to excite a slice of spins with the appropriate center Larmor frequency which matches the center frequency of the rf pulse and the thickness of the slice corresponds to a narrow range of frequencies determined by the bandwidth of the rf pulse. Applying a time modulated rf pulse with a sine shaped time dependence corresponds to a finite rectangular slice in the frequency spectrum according to the Fourier relationship between time and frequency. The magnetic field gradient will create phase inhomogeneity throughout the slice therefore a gradient lobe in the opposite direction of G is applied to rephase the spins. See G in Figure 2.4. z  The next step in spatial localization uses two more gradients in the X and Y directions. G is the frequency encoding gradient and G is the phase encoding gradient. x  y  G varies the Larmor frequencies of the spins throughout the slice depending on their X x  locations. Applying G for a fixed time creates phase differences between spins which y  depends on their Y locations. The spatial information is contained in the 2D Fourier transform (FT) of the signal which contains both frequency and phase information. The phase encoding gradient, G , is repeated and the amplitude is varied a number of times y  depending on the size of the data array in 2 dimensions, usually 64, 128 or 256 times. This creates a matrix of vectoral projections along the Y axis where for each position along the phase direction, Y , of the image the signal is multiplied by a phase factor:  e> iK  (2.17)  yt  where T is the fixed time the gradient was applied. Phase encoding is followed by single gradient, G , which frequency encodes, along the X direction, each set of phase x  dispersions in the Y direction. Therefore the signal for each position along the frequency direction (X) of the image becomes:  S(x,t) = S (x)e- ° > ir(B  0  +8  x)l  (2.18)  29  Combining phase and frequency gradients creates a signal which can be solved with a 2D Fourier transform.  S(t) = S(x,y)e- ''e' ^ i)s  i  (2.19)  yr  Defining k-space as k =yg t and k =Yg x the M R signal in k-space can be described by x  x  y  y  the following equation  S(k ,k )= x  y  $S(x, y)e~ '  dxdy  Kk x+kfy)  (2.20)  which can be Fourier transformed in two dimensions to create the M R image.  S(x,y) = \S(k ,k )e ^ dk dk iik  x  y  (2.21)  y)  x  y  T relaxation of a sample is studied by measuring the decay of the signal acquired 2  after multiple spin echoes. After the initial 90° excitation pulse, a 180°, refocusing pulse is applied a time x later producing an echo at 2x. A second 180° is applied at 3x, which produces a second echo at 4x. This is repeated for n echoes so that a pulse is applied at (2n - l ) x times producing echoes at (2n)x. The time between echoes T E is typically very short which allows a better measurement of T . 2  The entire sequence is repeated after a  time TR. To minimize longitudinal relaxation effects TR is much larger than T i .  This  allows Ti relaxation to occur almost completely so that the refocusing pulse tips the entire magnetization into the transverse plane at each repetition.  30  TE  90°  rf Gz  J  x  180°  180°  y  y  L  f  H  LJ  Gy Gx Signal  Figure 2.4: CPMG pulse sequence with spatial localization. The rf pulse creates echo signal. Gz is slice selective pulse, Gy is phase encoding pulse (applied multiple times), Gx is frequency encoding pulse.  2.5 Magnetic Resonance Spectroscopy  In spectroscopy the FID created by precessing magnetization is Fourier transformed to create a spectrum of overlapping peaks like that shown in Figure 2.5. Each peak, or group of peaks corresponds to the signal from a metabolite. Several characteristics of the spectrum give information about the sample. The type of peak (singlet, doublet etc...) is a characteristic of the quantum coupling between the protons in the metabolite.  The  amplitude is proportional to the amount of metabolite that produces a signal at that frequency.  NAA  Figure 2.5: A typical human in vivo brain spectrum composed of overlapping metabolite signals some of which include A) Creatine, B) Choline, C) Inositol, D) NAA E) Glu + Gin  31  Peaks in the spectrum are located within a range of frequencies due to a phenomenon called chemical shift. Hydrogen nuclei in different molecules and in different parts of a molecule experience a different static magnetic field Bo because they are shielded from the field by varying degrees depending on the shape and distribution of electron density surrounding each nucleus. From Equation 2.22, a nucleus with greater shielding will produce a signal at a lower frequency than a nucleus that is experiencing a higher field.  G)  =  yB (\-CT) 0  (2.22)  Chemical shift is measured in parts per million (ppm) and is referenced to a specific material. For example in P spectroscopy the signal from creatine phosphate is located 3 ,  at 0 ppm (and 0 Hz) and in *H spectroscopy in vivo water is at 4.7 ppm.  observed frequency (Hz) - reference frequency (Hz) ppm =  (2.23) Larmor frequency (MHz)  2.5.1 Water Suppression  To detect a signal from metabolites in vivo, the water signal from the sample must be suppressed because it is 10 000 times more abundant than metabolites, creating a huge signal which obscures weaker metabolite signals. The fundamental concept in suppression of a particular resonance is in the exploitation of property differences between the protons in the suppressed molecule and those which one wants to observe. Properties that are commonly used include magnetic properties such as chemical shifts, scalar coupling, and relaxation as well as diffusion and exchange [81].  The most  frequently used property is chemical shift where water is selectively excited, using its Larmor frequency, or other resonances are excited without including water. Therefore only this technique will be discussed. In order to select a specific frequency, long duration (>19 ms) "soft" frequency selective pulses are used. These pulses must be optimized for a particular frequency  32  response in order to excite only the required spectral range.  There are several pulse  shapes that can achieve this, including multi-lobe sine and Gaussian pulses.  The pulse  sequence includes a soft rf pulse, a gradient pulse and a delay before the spectroscopy localization sequence begins. When the soft rf pulse is 90°, the pulse sequence is called CHESS (CHEmical Shift Selective) [82].  The 90° rf pulse tips only the protons  associated with water into the transverse plane. The gradient that follows dephases the magnetization so that there is no signal from water when the spectroscopy sequence begins. The gradient should follow the rf pulse as soon as possible to avoid any T i relaxation. The efficiency of suppression depends on Bo and B] homogeneity. In cases of optimum field homogeneity one CHESS pulse will suffice, in reality a series of three CHESS pulses, shown in Figure 2.6, gives much better results.  90° rf pulses  Gradients  Figure 2.6: CHESS pulse sequence with three rf pulses precedes every spectroscopy sequence.  2.5.2 Spectroscopy Localization Sequences  As in MRI, the spectroscopy voxel of interest must be spatially localized using a pulse sequence. The most common localization sequences are called PRESS (Point REsolved Spectroscopy Sequence) and S T E A M (STimulated Echo Acquisition Mode).  PRESS  acquires a signal from a spin echo whereas S T E A M acquires a signal from a stimulated echo. The spin echo in the PRESS sequence has twice the signal of a stimulated echo therefore PRESS is generally favoured over S T E A M .  Both sequences spatially localize  by applying three rf pulses which each select one of three orthogonal planes. Only the nuclei in the volume at the intersection of the three planes will produce a signal. S T E A M uses three 90° pulses where PRESS uses a 90° pulse followed by two 180° pulses. Pulse sequences are illustrated in Figures 2.7 and 2.8. Nuclei that are not located in the voxel  33  but have been excited by one of the rf pulses are dephased by large crusher gradients on either side of the rf pulse.  TE  1  I  slice selective  crushers  Figure 2.7: PRESS pulse sequence.  I  slice selective  I  I  I  crashers  Figure 2.8: STEAM pulse sequence.  Each pulse sequence has its advantages and disadvantages. mentioned, PRESS has twice the signal to noise ratio (SNR) of S T E A M .  As previously Yet PRESS is  more sensitive to field inhomogeneity as it relies on two perfect 180° pulses. Historically S T E A M could operate at lower TEs than PRESS for the detection of more complicated spin systems. Recently PRESS has been improved so that TE=30 ms is possible.  S T E A M can still be altered to achieve lower TEs than PRESS for specific  applications. Short echo time spectra contain signals from more compounds and have better SNRs but have worse water and lipid contamination. Long echo times lead to  34  better resolved spectra with flat baselines but have significant T weighting. 2  For this  thesis, short echo time TE = 30 ms PRESS was used because of its improved SNR. PRESS and S T E A M sequences can be used for localization in both single voxel spectroscopy and spectroscopic imaging.  In single voxel spectroscopy (SVS) three  intersecting slices are chosen so that resulting signal only arises from a specific voxel or region of interest (ROI). The major challenge in SVS is to eliminate the signal from the area surrounding the voxel. Spectroscopic imaging (MRSI), also called chemical shift imaging (CSI), produces a number of spectra from a large volume, each one spatially localized to a volume element.  MRSI is acquired by adding a spectral dimension to  conventional M R I data. A time dimension is added to the spatial k-space dimensions so that for 2D single slice MRSI the data set is three dimensional and for 3D volumetric MRSI the data set is four dimensional. Commonly the spectroscopic time evolution is sampled directly in the absence of applied magnetic field gradients [83].  MRSI  techniques offer better spatial resolution but lower SNR for each spectrum when compared to single voxel techniques. MRSI is more complex to acquire and analyze but much more information is gathered than with SVS in similar time frames. In this thesis both SVS and M R S I techniques were used and the results will be compared.  2.5.3 Magnetic Field Homogeneity and Shimming  Spectroscopic techniques such as CHESS water suppression and PRESS localization rely heavily on field homogeneity. A n important element of any spectroscopic pulse sequence is shimming. Shimming is a process whereby the field inhomogeneity is detected and adjustments are made to the field by running currents through the shim coils. Most scanners use the X , Y and Z gradient coils for shimming and incorporate automated shimming procedures in auto prescan functions. Field inhomogeneity arises from a variety of factors. Magnets are designed so that their magnetic field is as uniform as possible. When objects are introduced into a magnetic field there is an interaction between the object and the magnetic field.  This  interaction is called magnetic susceptibility and in vivo it arises from different magnetic permeabilities of air and tissue [84]. For example, the permeability difference between  35  water and air is 9.5xl0" and creates field shifts of many ppm at air/tissue interfaces [85]. 6  When a human being is placed in an M R I scanner significant Bo inhomogeneities are created which are specific to each person. Therefore before each scan automatic in vivo shimming techniques acquire a Bo field map and adjustments are made using appropriate currents for each available shim coils, computed using a least squared minimization procedure. Good shimming is essential for water suppression. If a magnet is well shimmed then only one CHESS pulse is required to tip the water magnetization and crush it. In an inhomogeneous field the 90° pulse does not select all the water and is repeated several times to ensure good suppression. Shimming in MRSI is more difficult than for SVS as the voxels are much smaller and field inhomogeneity will have a larger effect. Shimming techniques for SVS are usually localized to the voxel but in MRSI the entire field of view (FOV) is globally shimmed. PRESS and S T E A M sequences can be used in M R S I to localize and shim a large voxel inside the FOV so that spectra of interest inside the head (for brain MRS) are resolved.  Issues of shimming and field inhomogeneity bring up  interesting questions as researchers push for higher field magnets. High field magnets (>1.5 T) mean greater signal and improved SNR by a factor of B . 0  Unfortunately  magnetic susceptibility increases with field strength, thus any technique which relies on good field homogeneity, such as spectroscopy, will require an improvement in shimming techniques to be useful at high fields.  36  Chapter 3  Materials and Methods  3.1 Introduction  To study the trends of metabolite concentrations and myelin water fractions in white matter, single voxel spectroscopy (SVS) and spectroscopic imaging (MRSI) experiments were compared to T relaxation experiments in two separate studies. In the single voxel 2  study, SVS was used to measure metabolite concentrations in voxels of interest in four white matter structures. T relaxation measurements were used to calculate myelin water 2  fractions (MWF) in the same voxels of interest. The M W F was plotted against N A A , choline and creatine concentrations for all white matter and for each structure to examine if significant trends exist and to determine i f trends vary from structure to structure. In the M R S I study, spectroscopic imaging and T relaxation experiments were carried out 2  on the same slice, trends between metabolite concentrations and myelin water fractions were examined in many locations in a single brain. In both studies trends between metabolite concentration and white matter fraction will also be examined. In this chapter, the details of the pulse sequences and required analysis will be outlined. The statistics used to quantify relationships between variables will also be described. 3.2 Single Voxel Study: Spectroscopy Acquisition and Analysis Spectra were acquired in four white matter structures from twenty four volunteers using single voxel spectroscopy. Concentrations of N A A , choline and creatine were calculated using LCModel [86] and were absolutely quantified using external water standards and CSF, T i and T corrections [32]. 2  37  3.2.1 Subjects and Voxel Placement  Twenty four normal healthy volunteers were scanned, eleven men and thirteen women, mean age 27 years, ranging from 21 to 57 years. Volunteers provided informed consent prior to participation in the study. Single voxel spectra were acquired from four white matter regions: frontal white matter (FW), occipital white matter (OW), posterior internal capsules (PIC) and the splenium of the corpus callosum (SP). These regions were chosen because previous studies have shown that some of these structures have high myelin water contents [41] such as the splenium and posterior internal capsules or because large voxels could be placed inside the structure to get maximum signal to noise without a lot of partial voluming which is true for frontal and occipital white matter. Eleven spectra were acquired in each structure. During the scan, two spectra from two different regions were acquired from each volunteer. In some volunteers only one spectrum could be used, therefore there are 24 volunteers, not 22! Voxels were placed in white matter structures as shown in Figure 3.1. Frontal white, occipital white and posterior internal capsule voxels were placed on either the right or left side of the brain depending on which side exhibited more white matter in the voxel. The slice location of the voxel was chosen by looking at consecutive five mm thick proton density axial localizer slices and by deciding which two had the largest white matter region in the area of the white matter structure. The slice was generally located in an axial plane through the genu and splenium of the corpus callosum where all four structures are visible.  Frontal White Matter  Occipital White " Matter  Posterior Internal Capsules  |  Splenium  Figure 3.1: Placement of voxels in white matter structures.  38  3.2.2  Pulse Sequence and Parameters  Spectra were collected on a 1.5 T General Electric Horizon Signa scanner using a single voxel PRESS sequence with T E = 30 ms, T R = 5000 ms, 96 averages, and maximum phase cycling.  For spectra in frontal and occipital white matter the voxel was an  18.5x18.5 mm square placed on a 10 mm slice for a total volume of 3.42 cm . For the 2  3  posterior internal capsules and splenium the voxel was rectangular, to better match the shape of the structure, with dimensions 9.9x33.9x10 mm and 3.32 cm total volume. When two spectra were collected in the same volunteer each spectrum was auto prescanned for shimming before collection.  Water suppression was achieved using  CHESS pulses.  3.2.3 Spectroscopy Analysis: In vivo spectral fitting  The in vivo brain spectra were fitted with LCModel (version 5.2-1) [86]. This program performs  an eddy current correction with the unsuppressed water signal and  automatically phases the spectrum.  LCModel also calculates the concentration of a  metabolite by comparing the in vivo spectrum to a set of templates of representative metabolites at known concentrations and by finding the linear combination of templates that best fit the in vivo spectrum.  The area of each template included in the fit is  numerically integrated by LCModel to obtain the area which represents each metabolite peak. The known template concentrations and the fractional area of the fit are used by LCModel to determine the concentration of each in vivo metabolite signal,  A tme  LCModel uses scanner tuning values to normalize the in vivo signal areas to the template areas for absolute concentrations.  This function was disabled and external water  standards were used for normalization instead (see Section 3.2.4). The templates used by LCModel were created in house. They were collected with the same parameters used to acquire the in vivo spectra. The following templates were created: alanine, creatine ( C H and C H 3 peaks), 2  choline, G A B A , glucose, glutamine, glutamate, glycerophosphorylcholine, myo-inositol,  39  lactate, N-acetyl-aspartate, N-acetyl-aspartyl-glutarnate and taurine.  The two creatine  peaks, CH2 and CH3, were fit with two independent singlet acetate peaks. Solutions of individual metabolites were carefully prepared with 100 m M concentration and titrated to pH 7.2. They were scanned within days of preparation at 21°C in a 300 mL spherical glass flask at the center of the headcoil. A n example output from LCModel is shown in Figure 3.2. The program returns a metabolite concentration relative to creatine, an absolute concentration and, a percent standard deviation (%SD). Twice the %SD is used to get the rough 95% confidence interval. Although other measures of the reliability of concentration estimates, such as the full width half maximum of the peak and the SNR, are provided by LCModel, the %SD is the best indicator because it accounts simultaneously for both resolution and noise level[87]. When a metabolite's concentration returns a %SD of less than 20% then it was well fit by the templates and the data can be included in the analysis. A plot of the best fit spectrum is plotted on top of the raw data and the residuals can also be displayed.  Figure 3.2: LCModel output including measured spectrum, fitted spectrum, residuals and estimates and %SD of metabolite concentrations.  40  3.2.4  Spectroscopy Analysis: Quantification  The goal of quantification is to obtain the absolute concentrations of metabolites from the area under the given signal.  The acquisition of in vivo spectra includes a prescan  procedure which optimizes scanner tuning for the 90° pulse in each subject. Therefore a signal area for one subject does not represent the same concentration as the same signal area in a different subject. Signal areas are also reduced by Ti and T relaxation for a 2  given T E and TR. and  Accurate quantification requires T i , T and water content corrections 2  results must be normalized for scanner tuning variations.  The absolute  concentrations of N A A , Choline (Cho) and Creatine (Cre) were calculated. The quantification process in this study included all these steps. The metabolite concentrations were Ti and T corrected with Ti and T values obtained in rigorous 2  2  studies that examined a large range of TR and TE to determine the accurate values [88,32]. The metabolite areas, A t , and the unsuppressed water signal of the voxel, me  A ater, were both T i and T 2 corrected. The unsuppressed water signal was measured by W  numerical integration after an automatic eddy current correction.  T i corrections were  made using  A n c o r r  where A  m e a  1-expKra/rj  v  '  J  s is the measured signal area of either the metabolite or of unsuppressed water.  Aiicorr is the signal area in the limit of TR—>ao where Ti weighting is at its minimum. T R  is equal to 5000 ms. A previous study has shown that metabolite Tis vary throughout white matter [88]; in F W voxels Ti values for frontal white matter were used (Ti = 1.34 s for Cho, 1.71 s for Cre, 1.59 s for N A A and 0.76 s for water [88]). For OW, PIC and SP metabolite Tis have not been investigated and measured with the same method. The T i values for parietal white matter were used to approximate their Tis (Ti = 1.19 s for Cho, 1.50 s for Cre, 1.35 s for N A A and 0.76 s for water [88]) and the same approximation was made for T corrections because the T s of metabolites have only been studied 2  2  rigorously in parietal white matter [89].  41  T corrections for metabolites were made using 2  A A  —  C3 2)  MEAS  exp[-TE/T ] 2  where  AT2corrmet  is the singlet signal area at T E = 0 ms (where there is no T weighting), 2  TE = 30 ms and T = 0.257 s for Cho, 0.161 s for Cre, 0.326 s for N A A (measured in 2  parietal W M [89]). The T corrected water signal, 2  spectroscopy water signal  A30TEwater  A-ncon-wat,  was found by comparing the  measured at T E = 30 ms to the T relaxation water 2  signal measured using the C P M G 48 echo pulse sequence. The T relaxation water signal 2  from the spectroscopy voxel was calculated at both T E = 0 ms (ACPMGOTE) and T E = 30 ms  (ACPMG30TE)-  The correction is equal to  A  ^Tlcorrwater  _  A  A  ~ A  ^•iOTEwater  CPMGOTE  <3\  CPMG30TE  Therefore both the unsuppressed water signal and the in vivo metabolite signal were T i and T corrected to give the final corrected signal in equations 3.4 and 3.5 respectively. 2  •^watercorr ~ ^Tlcorrwat ^Tlcorrwat  (3-4)  ^metcorr ~ ^T\corrmet ^T2corrmet  (3-5)  J  J  The absolute water content of the spectroscopy voxel was found by comparing the C P M G water signal at T E = 0 ms from the spectroscopy voxel to the C P M G water signal from six external water standards scanned in the same slice as the spectroscopy voxel. The C P M G water signal at TE = 0 ms from each of the six water standards were found by examining the signal produced from an ROI drawn within the water phantom image. Since the water bottles were on the edge of the image the Bj field inhomogeneity is significant and some of the fits produced by NNLS (see Section 3.3.2) were very poor with high x 2  These fit values were omitted and on average only four out of six water  bottles, those with shorter T s, were averaged to make a water signal at TE = 0 ms for the 2  42  external water standards. To normalize, the water signal for T E = 0 ms for the external standard was divided into the water signal at TE = 0 ms of the spectroscopy voxel.  A —  -^CPMGTEOvoxel  g\  *CPMGTEOwaterbottles  Equation 3.6 gives the absolute water content (WC) of the spectroscopy voxel in g/ml, which served as a standard to normalize concentrations between subjects thus eliminating variations due to scanner tuning. In a reproducibility study conducted with similar water phantoms, the standard deviation of the water content after averaging all the water bottles, was less than the standard deviation of the metabolite signal area [32,90]. Therefore averaging the signals for several water bottles to create one external standard is a valid technique. The concentrations of metabolites were calculated using  C  met  =  A m e , c o r r  x  n P w a , e r  xWCx 550000mM  (3.7)  watercorr  Where  A tcorr me  and A  wat  ercorr  are the corrected signals for metabolites and water, W C is the  water content, np is the numbers of protons contributing to the singlet signal (for N A A np = 3, Cho np = 9, Cre np = 3) and 55000mM is the concentration of water in the voxel i f the W C were unity. T analysis showed that the fraction of cerebrospinal fluid (CSF) in white matter 2  voxels was, for some, significant, therefore a CSF correction for the metabolite concentration was applied. Specifically voxels located in the occipital white matter and the splenium of the corpus callosum had significant CSF fractions due to their proximity to the ventricles.  The CSF fraction was calculated using a segmentation routine  described in Section 3.3.3. The CSF correction accounts for the fact that less of the volume of the voxel contains brain tissue therefore measurements of concentrations for brain tissue should be scaled up. The correction also accounts for the fact that CSF has a  43  different T i than the water in white matter and thus corrects the water content and the water signal according to the following equations:  WC  corr  =  JVC  WC — WMF/T  =  WMF  corr  (3.8) K  XcorrWU  WMF/T  CSF/T  UorrWM+  XcorrCSF  or WC  corr  WC (1 - exp(-7K / T )) * WMF  =  —  v  iCSF  (- exp(-77? / T  XCSF  ) + exp(-7H / T  XWM  )) * WMF +1 - exp(-77? /T  WM  (3 9) ' J  )  where W M F is the fraction of white matter in the voxel (WMF = 1 - CSF fraction) and TicorrWM  is the I i correction for white matter which according to Equation 3.1 is carried  out by dividing by a factor of (l-exp(-TR/TiwM) • The water content was T i corrected using T R = 2.68 s which was the T R used in the T relaxation pulse sequence and T i = 2  3.0 s for CSF and T i = 0.76 s for white matter [41]. The water signal A rcorr, which has wate  already been T i and T corrected, was CSF corrected in a similar manner using: 2  A  A wmercorr  water  ~ WMF  corr  A ~  water  / o i r\\  WMFIT +CSFIT  XcorrWM  WMFIT  UorrWM  XcorrCSF  The Tj correction for the unsuppressed water signal was made with T R = 5 s which was the T R used in the single voxel spectroscopy sequence and T i = 3.0 s for CSF and T i = 0.9 s for the water in brain tissue.  The final concentrations were calculated using  Equation 3.7 with CSF corrected values for both W C and water signal, A  w a t e r c 0  n--  The  standard deviations returned by LCModel were used as estimates of the error in the measurements of absolute concentrations of metabolites.  44  3.3  Single Voxel Study: T2 Relaxation Acquisition and Analysis  For each subject a T2 relaxation pulse sequence collected data from the 10 mm slice containing the spectroscopy voxel. The data from the 48 echo C P M G pulse sequence was analysed to extract the T2 components of the different water compartments of the sample. A typical T 2 relaxation decay curve produced by measuring the decaying signal at different T E times is shown in Figure 3.3. If the curve were mono-exponential it would be easily described by Equation 3.11 and would be characterized by a single T2.  S(TE) = S(0)exp(  (3.11)  Since the curve in Figure 3.3 is multi-exponential it cannot be described by a single T2 but by the sum of several exponentials each with different T2S, which is shown graphically by several lines which each fit the signal differently.  The multi-exponential  character of the decay in vivo implies that the water in brain is compartmentalized. The curve fitting algorithm used in this study, N N L S [91], resolves at least three T 2 exponentials corresponding to different compartments of water, which have been assigned to myelin water, intra and extra cellular fluid and cerebrospinal fluid.  See  Figure 1.8. 1000  1 0  4 0  .  .  .  .  .  .  .  rJ  40  80  120  160  200  240  280  320  T E (ms)  Figure 3.3: Plot of T relaxation decay curve measured at 32 decay times where TE = 10 ms. The signal's multi-exponential decay is a function of T . 2  2  45  3.3.1  Pulse Sequence and Parameters  T2 relaxation measurements were made using a modified 48-echo C P M G pulse sequence [79,80,91,41].  A single 10 mm slice was imaged to match the slice containing the  spectroscopy voxels. A dual echo spacing pulse sequence with T E = 10 ms for the first 32 echoes and T E = 50 ms for the last 16 echoes was employed. This modification to the pulse sequence uses extended echo times to improve the resolution of peaks at longer T 2 times without compromising T information and significantly changing values such as the 2  myelin water fraction [92].  A variable T R modification to the pulse sequence was also  used. The T R was ramped up from 2.12 s to 3.80 s in the first 64 lines of k space and back down to 2.12 s in the last 64 lines of k space. It has been shown that collecting some regions of k space at shorter TR values does not significantly affect quantitative measures such as T 2 but saves valuable scan time [93]. The data matrix was 128x128 and was collected with a 22 cm F O V and 4 N E X . The bandwidth was 16 Hz. The total acquisition time was about 25 minutes, including spectroscopy and localizers; the total scan time was about 1 hour. Six water phantoms with a range of T]S and T s were placed alongside the head to 2  act as external standards for quantification of metabolite concentrations. The T2S of the water bottles ranged from 25 ms T2 to 110 ms T2. They were doped with NiCL. because its T relaxation has been shown to be relatively independent of temperature and field 2  strength.  The glass tubes (diameter 1 cm and length 16 cm) were vacuum sealed for  maximum stability over time.  3.3.2  T Relaxation Analysis: Non-Negative Least Squares Algorithm 2  A l l T relaxation data was analyzed using a Non-Negative Least Squares (NNLS) fitting 2  routine [91]. N N L S decomposes decay curves into an arbitrary number of exponential terms. It is not necessary for the signal to be multi-exponential and if N N L S is applied to a mono-exponential system it will produce a solution with a single exponential that can be solved for a single T2.  Multi exponential systems can be described by the following  integral  46  y(TE )= t  jsiTJexpi-TE^TJdTt  i = 1, 2, . . . N  (3.12)  ^2 min  Therefore the signal, a function of TE, is measured at N different T E times. For the 48 echo pulse sequence used in this study N = 48 and s(T ) is the unknown amplitude of the 2  component which has a certain relaxation time T . The true T distribution can be 2  2  approximated as a sum of M delta functions and Equation 3.12 becomes a sum of M discrete exponential terms  M  yiTEJ^siT^exvi-TEt/T,.)  i = l,2,...N  (3.13)  7=1  In the N N L S algorithm M is set to a large number so that the solution is unbiased. The analysis in this study used a set of M=100 delta functions spaced logarithmically between 10 ms and 10 s. For in vivo imaging the measured data y(TEj) are contaminated with noise which is taken into account by considering a noise term to Equation 3.13 when it is fit. The misfit x is given by the difference between the measured data and the predicted 2  data as in  * =i>,'-J',) /<r, 2  2  (3-14)  2  where yf are the predicted data, y are the measured data and o\ are the standard deviations of the i  t h  data point. Ideally, x ~ N and each data point is fit within one 2  standard deviation. When x » N then the data has not been fit well and i f x « N the 2  2  fit is too accurate and it is likely that extra components have been introduced to fit the noise component of the data. Non-regularized N N L S , the simplest application of the algorithm, solves Equation 3.13 with a noise term by minimizing x • The non-regularized N N L S solution is composed of a discrete number of delta functions. This algorithm is fast and simple but is sensitive to noisy data and can produce delta functions that are not  47  true time constants of the system and do not represent a separate water compartment. The non-regularized N N L S algorithm can be improved upon by incorporating an additional term, a regularizer, into the fit which minimizes both the misfit  and a second  quantity such as the roughness of the solution [91]. Where the smoothness is defined as  M  S=J>(r )  3.15  2  2 y  7=1  Thus regularized N N L S produces a smooth spectrum which is probably more representative of brain tissue. . Examples of non-regularized and regularized T  2  distributions are shown in Figure 3.4. In this study the non-regularized solutions were used because they are fast and convenient and proved to be robust enough with the SNR acquired in 10 mm slices T relaxation data. 2  Initially the T relaxation data in the spectroscopy voxel or the region of interest 2  (ROI) was analysed by averaging the data from each voxel in the ROI to create a single plot from which non-regularized N N L S produced a single T distribution as a solution. 2  From the amplitude and location of the peaks in the T distribution the amount of myelin 2  water, intra and extracellular fluid and cerebrospinal fluid in the spectroscopy voxel can be determined. The myelin water fraction (MWF) is defined as the sum of amplitudes of all the peaks between T = 1 ms and T = 40 ms divided by the total amplitude of all the 2  2  peaks in the distribution. Peaks up to T = 50 ms can be used in the myelin water fraction 2  but the more conservative window was used in this study. The CSF fraction is the sum of the amplitude of the peaks between T = 500 ms and T = 10 s divided by the total sum. 2  2  The sum of the peaks between T = 40 ms and T = 500 ms divided by the total sum is 2  2  assigned to the fraction of intra and extra cellular fluid [91]. Thus using N N L S , both the myelin water fraction and the CSF fraction can be estimated. This method of applying N N L S , to a single averaged decay from the ROI, does not produce an estimate of the measurement error for myelin water fraction. Also in some ROIs the CSF fraction was significantly high. Therefore the ROIs were segmented so that only white matter voxels in the ROI were included in the M W F calculation.  48  Figure 3.4: A) Spiky non regularized solution to NNLS algorithm B) smoothed regularized solution  Myelin maps were made for each subject. Myelin maps were created by applying non-regularized N N L S to each voxel in the image and calculating a myelin water fraction for each.  Once segmentation was complete and the white matter voxels in the  spectroscopy ROI were identified the corresponding myelin water fractions for these voxels were determined from the myelin map. The myelin water fractions of the white matter voxels in the spectroscopy ROI were statistically analysed to determine the mean, median, standard deviation and standard error of the myelin water fraction in the ROI. Histograms of the myelin water fractions were also created.  3.3.3  T Relaxation Analysis: Segmentation 2  Myelin water fractions are significantly higher in white matter structures when compared to grey matter structures [41]. Therefore an ROI with a significant amount of grey matter or CSF will have an artificially low M W F .  To correct myelin water fraction for partial  voluming, the ROIs were segmented into white matter, grey matter and CSF. Metabolite concentrations could not be similarly corrected although the T i correction takes into account the CSF fraction. This was done by drawing ROIs in regions judged by the user to be pure white matter, grey matter and CSF on the T E = 10 ms C P M G image. The same user completed segmentation for each subject. The white matter ROI was always  49  drawn in frontal white matter on the right side of the brain, the grey matter ROI was always drawn in the head of the caudate nucleus on the right side of the brain. The CSF ROI was usually drawn in the anterior lateral ventricle i f enough CSF was present, otherwise another suitable region was chosen. The segmentation routine averaged the signal intensity over all 48 echo images to determine the decay curve for 'pure' white and grey matter and CSF. Each voxel in the ROI was segmented by determining the linear combination of these three decay curves which best described the signal intensity in that voxel.  The white matter fraction of the ROI was determined by averaging the white  matter fraction over the 144 voxels in the ROI. The CSF fraction returned by the segmentation algorithm for each spectroscopy voxel was used to perform the CSF correction described in Section 3.2.4 for absolute quantification. Voxels in the ROI that were at least 80% white matter were accepted as pure white matter voxels.  A l l voxels  were analyzed with N N L S and a myelin water fraction was calculated for each and displayed as a myelin water map (See Figure 4.1) where voxels with high myelin water fractions are bright (mainly white matter areas). Within the spectroscopy voxel only the voxels with 80% white matter were included in the average myelin water fraction calculated for the entire voxel.  3.4  Spectroscopic Imaging Study: Acquisition and Analysis  Several subjects were examined using spectroscopic imaging (MRSI) to explore the trends of myelin water fraction and metabolite concentrations throughout one slice of the brain. 32x32 spectra were acquired for each subject and non-normalized absolute N A A , choline and creatine concentrations were calculated with LCModel. segmented and classified into white matter structures.  ROIs were  Myelin water fractions were  calculated for each ROI and trends between M W F and metabolite concentrations were examined in white matter and within four white matter structures. These results, for each of three subjects, will be compared to the results of the single voxel study obtained across many subjects.  50  3.4.1  Subjects and Voxel Placement  Three normal healthy volunteers, two males and one female, with mean age of 26 years, ranging from 23 to 29 years were scanned with a MRSI pulse sequence and a 48 echo T relaxation pulse sequence.  2  The MRSI and T relaxation were performed on the same 2  slice of the brain. The slice was 10 mm thick and located in an axial plane through the genu and splenium of the corpus callosum. The goal was to include the same structures that were studied in the single voxel study so that the results could be compared.  3.4.2  Pulse Sequence and Parameters  The M R S I pulse sequence divided the F O V into 32x32 voxels or ROIs and measured a spectrum for each. The spectral width was 1000 Hz with 512 points, TE = 30 ms, and T R = 1000 ms. The acquisition matrix was 256x128 with 22 cm FOV.  A large rectangular  voxel including the entire brain but excluding the skull or any fat near the skull was chosen on an axial localizer. The PRESS sequence and auto pre-scan function localized and shimmed this area so that the spectra from these ROIs had good data. The water was suppressed using CHESS pulses.  The unsuppressed water signal was acquired by  repeating the MRSI sequence with only 16x16 voxels and by turning off the water suppression pulses. The T relaxation pulse sequence was identical to that used in the 2  single voxel study. Water bottles were not included in the scan since the metabolite concentrations were not absolutely quantified.  3.4.3  Spectroscopy Analysis: In Vivo Spectral Fitting  LCModel was also used to fit the MRSI data. Before applying LCModel to the entire data set a mask was used to filter out ROIs which were outside the head. This reduced analysis time because the number of spectra was reduced to about 25x25 from 32x32. Even with the mask the LCModel analysis took more than one day! A similar basis function was used for MRSI as T E = 30 ms and PRESS was used as in the single voxel study. The difference in TR between the basis function (TR = 5000 ms) and in vivo  51  acquisition (TR = 1000 ms) did not affect the results significantly. LCModel produced an output similar to that for single voxel data including a plot of the best fit spectrum, absolute and relative concentrations with %SD. For MRSI the %SD were much higher than in single voxel analysis due to the reduced signal to noise in the smaller voxels. A n example M R S I spectrum from white matter is displayed in Figure 3.5.  0S0KHthi«_PaS68S_13.18  O. OOO »»M. o. ooo i *y%\ :aae'.'22»- t7% :-IM. '«H * « fwp'.tm. is* o.ooo > » i •**ii.''c« '; iii t  :  :  !  in  zu  m  tJ'<W«V<t-i l.MICtt o. two ow»* Z. 1>7 OXC d.'HI O U l.llttM 1. I L I M A  im>. ai%  'ijjijlt*:-.  Mnu.ii J W**tt": WC-  I ;ta*«  *r*«i  i  1: t a l o MCXOCUJU p j » . '..t/n •.= '» . i i t i ' ' W « t - » o:VMi:!'ji(pai". • i t o*ypp« Vat » 4*9 :  ;  r r u t z o » * / . . p e r t 'ttoaa/LCH>i  :  Figure 3.5: LCModel analysis of an MRSI spectrum from white matter. Observe that the resolution is much lower than a single voxel spectrum and that %SD are very high.  3.4.4  Spectroscopy Analysis: Quantification  In the M R S I study many voxels from throughout the head were acquired.  Since  metabolite T] varies throughout the brain, absolute quantification of concentrations would be very challenging and time consuming [88]. The Tj of each metabolite, in each region, would have to be accurately measured.  Therefore LCModel non normalized  concentrations were used for analysis of metabolite trends throughout the brain.  For  N A A spectra with %SD of less than 20% were considered reliable and for choline and creatine spectra with %SD < 40% were included. Choline and creatine are weaker signals than N A A and thus have larger %SD. Spectra with SNR greater than 1 were considered  52  reliable by inspection. Trends between myelin water fraction and metabolite relative concentrations were examined within each subject because enough data were acquired for each person. Comparisons between subjects were not possible because concentrations have not been absolutely quantified to eliminate variations in scanner tuning, T i and T . 2  3.4.5  T Relaxation Analysis: NNLS and Segmentation 2  In the MRSI study there are 32x32 voxels or ROIs for each subject. The myelin water fraction was calculated for each ROI using non-regularized N N L S with a myelin water window of T = 1 ms to T = 40 ms. One M W F for each ROI was determined by 2  2  averaging the signal from the voxels in the ROI. No estimate of the error of this measurement was made.  The same N N L S parameters, such as the number of delta  functions used to approximate the true T distribution (M=100), were used as in the 2  single voxel study T relaxation analysis. 2  In order to examine trends in metabolite concentrations and myelin water fraction in white matter structures, it was necessary to classify each ROI as white matter, grey matter or CSF. Instead of qualitatively observing spectra superimposed on an image, each ROI was segmented into white matter, grey matter and CSF using a segmentation algorithm similar to that implemented in the single voxel study. Only ROIs with a minimum of 70% white matter were classified as pure white matter ROIs.  A lower white matter  threshold was used for MRSI because signal to noise is low in the small voxels. White matter ROIs were grouped into white matter structures by observing a plot of the ROIs superimposed onto the TE = 10 ms C P M G image as in Figure 5.1.  3.5 Statistical Analysis  In the single voxel study the mean myelin water fraction and N A A , choline and creatine concentrations were calculated over 44 white matter measurements and over 11 measurements for individual white matter structures.  The standard deviation of each  sample, which represents the distribution of the sample, was calculated by taking the positive square root of the variance as in:  53  £* -(2»  n  2  2  (3.16)  n{n -1)  The standard error of the mean, which represents the distribution of the mean, was also calculated  (3.17)  For myelin water fraction in the single voxel study the mean across subjects is actually a mean of means. In the MRSI study the mean M W F and metabolite concentrations for all white matter and for several white matter structures were calculated for each subject. Significant differences between two means were tested using a two-tailed, unequal variance Student's t-test. In some cases the data was collected from the same group of volunteers under the same conditions so a paired t-test would have been appropriate. However this was not always true so the most conservative test (two-tailed and unequal variance test) was chosen for comparing all means. This test assumes that the data in the two samples arise from normal distributions. This assumption is fair because there were enough data points in each set (N>10). The value returned from the test is a probability, p, of the two means arising from the same normal distribution. A p-value of less than 0.05 represents a significant difference between the two means.  Mean MWFs were  compared between structures to see i f some structures had significantly higher myelin water fractions as predicted in the literature. Similarly, mean metabolite concentrations were compared to see i f they differed from structure to structure. Means for individual structures were also compared to the average over all white matter. In both studies mean and median M W F and metabolite concentration were plotted against each other in all white matter and in each white matter structure. Linear trends • • 2 2 • between two variables were evaluated using the coefficient of determination R . R is a value between 0 and 1 which measures the degree of correlation between two variables. A high value for R indicates a good correlation. R is calculated by comparing the ratio 2  2  54  between the residual sum of squares and the total sum of squares. The residual sum of squares is the sum over all points of the squared difference between the estimated y value and the actual y value of a point. The total sum of squares is the sum of the squared differences between the actual y values and the average of the y values. If the residual sum of squares is small compared to the total sum of squares then R is high. The significance of a linear trend was evaluated using the F-distribution. The F-distribution is the ratio of two independent distributions, each divided by their degrees of freedom [94]. The F-statistic gives the probability that the two variances of the two distributions are the same. A n F-statistic which is higher than the F-critical value for degrees of freedom u i and t>2 and p= 0.05 shows a significantly low probability that the trend between two variables occurred by chance.  55  Chapter 4  Single Voxel Study Results  4.1 Introduction  In the single voxel study, single voxel spectroscopy (SVS) was used to measure metabolite concentrations in four white matter (WM) regions. Eleven measurements were taken in each of four regions, frontal white matter (FW), occipital white matter (OW), posterior internal capsules (PIC) and the splenium of the corpus callosum (SP), for a total of 44 measurements in W M . Magnetic resonance imaging was used to measure myelin water fraction in the same regions. In this chapter the results will be presented in four sections. First metabolite concentrations, means and significant differences between regions will be presented (Section 4.2). Myelin water fraction means and distributions will be illustrated (Section 4.3), correlations between the two measures will be plotted and discussed (Section 4.4) and the trend between metabolite concentrations and brain tissue fractions will also be examined (Section 4.5).  4.2 Metabolite Concentrations  Metabolite concentrations were absolutely quantified according to the procedure described in Section 3.2.4. A n important step in the quantification process was the CSF correction which corrects the water content and the unsuppressed water signal of the spectroscopy voxel for the volume of CSF in the voxel and for effects introduced by differences between the Ti of CSF and the Ti of water in W M . The CSF fraction of each spectroscopy voxel was calculated using two methods.  In the first method, N N L S  calculated a CSF fraction from the long T component of the T distribution of the voxel 2  2  56  (Section 3.3.2) and in the second a tissue segmentation algorithm divided each voxel into white matter, grey matter and CSF fractions (Section 3.3.3). A comparison was made between the two methods of calculating the CSF fraction and the results are displayed in Table 4.1. The table shows the average CSF fraction calculated with each method over all 44 white matter measurements and over 11 measurements in each white matter region. The CSF fraction calculated by the segmentation routine was on average higher than the CSF fraction calculated by N N L S . Both methods show that the splenium voxels had the highest CSF fractions and that the frontal white matter voxels had the lowest CSF fractions.  The ratio between the averages of each method should be close to one,  unfortunately this was only achieved in F W , in other white matter regions and when averaged over all white matter, the segmented CSF fraction was almost double the N N L S CSF fraction. The segmented CSF fraction was used for quantification because it did not underestimate the effect of CSF in the voxel and thus gave the most conservative estimate of metabolite concentration. Also, the segmentation algorithm was used to estimate the white matter fraction for myelin water calculations so it was sensible to use the same method to estimate CSF fractions for metabolite concentration corrections.  Region A l l White Matter Frontal W M Occipital W M Posterior IC Splenium  Average CSF Fraction NNLS (%±%SE) 3.81±0.77 0.757±0.063 2.12±0.034 1.13±0.25 11.2±1.6  Average CSF Fraction Segmented (%±%SE) 7.78±1.6 0.92±0.20 4.62±0.93 2.48±0.56 23.1±2.9  averageCSF(NNLS) averageCSF (Seg) 0.490 0.824 0.459 0.456 0.485  Table 4.1: Comparison of CSF fractions with standard errors (SE) calculated with two methods, NNLS and with a segmentation algorithm. CSF fractions were averaged over 44 white matter (WM) measurements and over 11 measurements in each WM region. The correspondence between the two is expressed as a ratio.  Absolute metabolite concentrations were calculated for N-acetyl aspartate (NAA), choline (Cho) and creatine (Cre). LCModel calculated concentrations for N A A , choline and  creatine and concentrations relative to creatine.  These concentrations were  normalized using water bottles as external standards, and T i , T and CSF corrections 2  57  (Section 3.2.4). For each white matter region, means and standard errors of absolute and relative metabolite concentrations are presented in Table 4.2. The two-tailed Student's ttest with unequal variance was used to compare the variation of metabolite concentrations throughout white matter.  For [NAA] the average concentration over all white matter  voxels was found to be 10.1±0.2 mmol. [NAA] in frontal white matter was significantly lower (p<0.05) than the white matter average. In fact, [NAA] in frontal white matter was significantly lower than [NAA] in all other white matter regions (p<0.05 for OW, SP and p<0.01 for PIC). The standard error in [NAA] varied the most between structures and the highest standard error occurred in frontal white matter where most of the significant differences were found for [NAA].  Region A11WM Frontal W M Occipital W M Posterior IC Splenium  [NAA]±SE (mmol) 10.U0.2 9.2±0.4 10.1±0.1 10.6±0.2 10.3±0.3  NAA/Cre ±SE 2.9±0.1 2.45±0.08 2.9±0.1 2.35±0.07 4.0±0.2  [Cho]±SE (mmol) 1.93±0.05 2.1±0.1 1.88±0.06 1.98±0.06 1.73±0.08  Cho/Cre ±SE 1.37±0.05 1.4±0.1 1.30±0.08 1.10±0.05 1.66±0.05  [Cre]±SE (mmol) 7.1±0.2 7.5±0.2 6.9±0.2 8.9±0.2 5.1±0.2  Table 4.2: Mean absolute and relative metabolite concentrations with standard errors.  Significance  [NAA]  NAA/Cre  [Cho]  Cho/Cre PIC/SP PIC/WM SP/WM  pO.OOl  None  SP/all regions PIC/WM  None  p<0.01  FW/PIC  PIC/OW  None  SP/OW  p<0.05  FW/WM FW/OW FW/SP  FW/WM FW/OW  SP/WM SP/FW SP/PIC  PIC/OW PIC/FW  [Crel SP/all regions PIC/all regions  Table 4.3: Significant differences between white matter regions (all white matter (WM), frontal white matter (FW), occipital white matter (OW), posterior internal capsules (PIC) and splenium (SP)) for absolute and relative concentrations.  Absolute concentration of choline [Cho] in all white matter was calculated to be 1.93±0.05 mmol. Choline concentrations in all white matter and within each structure  58  were compared and significant differences were found between splenium and all other structures (p<0.05) except for occipital white matter where no significant difference was observed. Creatine concentration averaged over all white matter was 7.1±0.2 mmol. [Cre] in the posterior internal capsule was significantly higher than all other white matter structures (p<0.001) and [Cre] in the splenium was significantly lower than all other white matter structures (p<0.001). Similar comparisons were made for relative metabolite concentrations calculated by LCModel, NAA/Cre and Cho/Cre, to see whether trends in metabolite concentration distribution throughout  white matter were similar between uncorrected relative  concentrations and T i , T2, and CSF corrected absolute concentrations.  NAA/Cre was  significantly higher in the splenium when compared to all other white matter regions (p<0.001) thus in SP significant differences in relative N A A were much stronger than with [NAA]. In frontal white the same significances for relative N A A and [NAA] were observed with O W and averaged W M but no significant difference was observed when compared to PIC. Strong significant differences in NAA/Cre were also observed between PIC and averaged W M (p<0.001). Where few significant differences were observed with absolute choline concentrations, Cho/Cre concentrations showed more variation between structures.  Significant differences in Cho/Cre between SP and O W (p<0.01), and W M ,  PIC (p<0.001) were strong compared to the significance calculated for absolute choline concentration (p<0.05). Also Cho/Cre was significantly lower in the PIC, in some cases highly significant (pO.OOl for SP, W M and p<0.05 for OW, FW).  Significant  differences are summarized in Table 4.3.  4.3 Myelin Water Fraction As discussed in Section 3.3.2 the myelin water fraction was calculated from T 2 distributions using the non-regularized N N L S  algorithm [91]. Two methods of  calculating the myelin water fraction were used. The first calculated a single myelin water fraction from an averaged signal from the entire spectroscopy voxel. The second method used myelin water maps (See Figure 4.1) and a segmentation algorithm to correct the voxel for partial voluming with grey matter and CSF. The myelin water map  59  calculated a myelin water fraction for each of the 144 voxels inside the spectroscopy voxel. These voxels were segmented and only those with at least 80% white matter were used to calculate a mean and median myelin water fraction for the spectroscopy voxel. The myelin water fraction is defined as the sum of the amplitudes of peaks in the short T range of the T distribution divided by the total water content. Previous studies 2  2  have shown that the T of the myelin peak can range from 15 to 40 ms and thus the T 2  window for myelin was set to 10 to 50 ms [40].  2  Since the previous study, pulse  sequences have been improved (from 32 echo to 48 echo for example) and the range of myelin water T s has narrowed making it possible to use a more conservative range of T 2  2  times for the calculation of myelin water fraction.  Figure 4.1: An example of a myelin water map produced using the pulse sequence and analysis used in this study. Areas of high intensity are areas with large myelin water fractions.  In the current study, both 10 to 40 ms and 10 to 50 ms myelin water fractions were calculated for each voxel. In 6 voxels out of 44 the myelin water fraction calculated for the 10 - 50 ms window was much higher than the 10 - 40 ms window indicating that in these voxels, a large peak between 40 and 50 ms was present in the T distribution. 2  When the myelin water fraction was calculated from a single averaged relaxation decay curve from the spectroscopy voxel and compared to the mean M W F calculated from the segmented myelin water map it was those with large peaks between 40 and 50 ms which gave very different results between the two methods. Thus, one method had to be chosen and to do so the methods were systematically compared. The segmented myelin map  60  method produced both a mean and median M W F for each voxel therefore six measurements of M W F were calculated and compared. The mean myelin water fraction in all white matter and in each white matter structure was calculated for each method and for each T window (10—40 ms and 10-50 ms). See Figure 4.2. 2  From Figure 4.2 some trends are clear. The highest M W F was observed in the splenium and the lowest in frontal white matter.  The MWFs calculated from the  segmented myelin maps were higher because grey matter (which has a lower M W F [41]) was segmented out of the voxel. In the posterior internal capsule the greatest difference between the two T windows was observed because most of the voxels with large peaks 2  between 40 and 50 ms were located in the PIC. Figure 4.2 shows that the variation between structures for all measurements is larger than the variation between methods of measurement within each structure. The unequal variance two-tailed Student's t-test was used to compare the six measures of M W F in all white matter for significant differences and only one was found between the myelin water map mean for 10 - 50 ms window and the single averaged decay curve method for the 10 - 40 ms window.  Also trends  between M W F and metabolite concentrations, presented in the next section, were similar with each set of myelin water fraction measurements. 18 16  AIIWM  Frontal W M  Occipital W M  Posterior IC  Splenium  Figure 4.2: Plot of mean myelin water fractions, with standard errors, calculated in each region with six different methods. For each region, thefirsttwo bars are means calculated with the single averaged decay curve method (10-40 and 10-50 ms windows). The next two are mean and median MWF calculated with segmented myelin maps (10-40 ms window), the last two are mean and median MWF calculated with the same method but with 10-50 ms window.  61  Finally significant differences of myelin water fraction between regions were exactly the same, no matter which data set was used. The segmented myelin map mean and median myelin water fractions theoretically measure M W F in white matter more precisely since grey matter and CSF voxels do not contribute to the calculation. These measurements also provide an estimate of the error in M W F since multiple measurements within a spectroscopy voxel were made. Histograms of white matter myelin water fractions for each spectroscopy voxel were examined to determine whether mean or median myelin water fractions should be used. Some histograms, but not all, exhibited a normal distribution which a mean could describe well. Examples of histograms are shown in Figure 4.3.  B  i  1L_  C  i—i  U_L  D  Figure 4.3: Examples of histograms showing the distribution of myelin water fraction of white matter voxels in a spectroscopy voxel. Histograms A and B are nice histograms approximating normal distributions of MWF (for A, n=69 white matter voxels and for B, n= 86). Histograms C and D are not normally distributed (for C, n=87 and for D, n = 46).  Since the histograms show a wide variation in the shape of myelin water fraction distributions, it was unclear whether mean or median calculations of M W F would more accurately describe the data. No significant differences were found between methods therefore the choice of measurement method would not significantly affect the final outcome of the results. The mean M W F from segmented myelin maps with 10-40ms T  2  window is the conservative estimate of the M W F but is not the most accurate because the myelin water fractions are calculated from within small voxels which don't have much  62  signal compared to the FID method. This data are displayed in Table 4.4 and were used to examine regional distribution of myelin water fraction and correlations between M W F and metabolite concentrations. Regional distribution of myelin water fraction within white matter was tested with an unequal variance two-tailed Student's t-test. A l l regions of white matter were found to be significantly different from each other except between frontal and occipital white matter.  M W F in SP was significantly higher than PIC (p<0.05), OW (pO.OOl), F W  (pO.OOl) and all white matter (p<0.01). In PIC highly significant differences (pO.OOl) were found between it and F W and OW. The M W F in F W (pO.OOl), O W (pO.Ol) and SP (p<0.01) were significantly different than the average myelin water fraction in white matter, 7.21±0.65%. Therefore the only significant differences which weren't observed were between F W and OW, and between PIC and all white matter.  Region A l l White Matter Frontal W M Occipital W M Posterior IC Splenium  Mean Myelin Water Fraction (%±%SE) 7.21±0.65 3.65±0.29 4.74±0.46 8.08±0.65 12.36±1.43  Table 4.4: Mean MWFs with standard errors for each region and averaged over all white matter, calculated with segmented myelin maps and 10-40ms T window. 2  4.4 Correlations between MWF and Metabolite Concentrations  Figures 4.4. to 4.13 show correlations between each metabolite concentration and myelin water fraction. Absolute concentrations of N A A , choline and creatine are plotted against the myelin water fraction for all 44 white matter voxels. Each plot shows the correlation coefficient R and the significance (if there is one) of the slope between the two variables. 2  A n R =1 indicates a perfect correlation between two variables where all the variation in 2  one variable is completely accounted for by the other. Significances were determined using an F distribution. A significant correlation was found between [NAA] and myelin water fraction and [Cre] and myelin water fraction. The correlation was stronger between [NAA] and M W F (pO.Ol, R = 0.17) than between [Cre] and M W F (p<0.05, R =0.11). 2  2  63  The plot of [NAA] vs M W F , Figure 4.4, illustrated that M W F had a much wider range than [NAA] and error bars indicated that large MWFs also have large standard deviations. For all white matter the plot had an outlying point which could have been influencing the results but when the data was plotted without the outlier, the slope was still significant, although more weakly (p<0.05, R = 0.12). A plot of [Cho] vs M W F did 2  not have a significant slope.  The correlations of metabolite concentrations and M W F  were also examined within each white matter structure.  Between structures different  results were observed. For [NAA] vs M W F , the correlation within the splenium was highly significant (p<0.01, R = 0.57). The slope of [NAA] vs M W F was not significant 2  in any other white matter structure.  [Cho] vs M W F slope only reached significance  within frontal white matter (p<0.05, R = 0.40) and [Cre] vs M W F slope reached 2  significance within occipital white matter (p<0.05, R = 0.38). Separating white matter 2  voxels into individual structures also revealed that [NAA] in F W matter had a much wider range (7 - 10.5 mmol) than occipital white where [NAA] ranged from 9.5 - 11 mmol (See Figure 4.5). [Cho] had a narrower range throughout white matter than [NAA] or [Cre] and thus most structures had similar ranges of choline concentration (Figure 4.7). The plot of [Cre] vs M W F for individual structures (Figure 4.9) illustrated clear groupings of data for structures as each had a narrow range which varied from structure to structure. For example in the splenium [Cre] ranged from 4-6 mmol and from 8-10 mmol in frontal white matter.  Thus there are clear differences in metabolite  concentrations between structures. Relative concentrations NAA/Cre and Cho/Cre were also plotted against myelin water fraction. The plot of NAA/Cre vs M W F (Figure 4.10) over all white matter voxels exhibited a highly significant slope (p<0.0005) but when divided into individual structures no significant slopes existed. This differed from the plot of [NAA] vs M W F where the splenium had a significant correlation. No significant correlations occurred between Cho/Cre and M W F in all white matter voxels or within individual structures (Figures 4.12 and 4.13).  64  Figure 4.4: NAA concentration vs myelin water fraction plotted for all white matter voxels exhibited a highly significant correlation. Error bars represent standard deviation of each measure.  40  -r  35 -  x SP: R = 0.57, p < 0.01 2  A P I C : R = 0.12 2  o  N  S O  a U  «  30 -  • O W : R = 0.14 2  • F W : R = 0.077 2  25 20 15 -  yel  e  10 504.0  5.0  6.0  7.0  8.0  9.0  10.0  11.0  12.0  13.0  [NAA] (mmol) Figure 4.5: NAA concentration vs myelin water fraction plotted for each individual structure. Splenium (SP) = x, posterior internal capsules (PIC) = A, occipital white matter (OW) = • and frontal white matter (FW) = •.  65  Figure 4.6: Choline concentration vs myelin water fraction plotted for all white matter voxels did not exhibit a significant correlation. Error bars represent standard deviations of each measure.  40 -j  x S P : R = 0.1 A P I C : R = 0.033 • OW: R = 0.0095 • FW: R 0.40, p < 0.05 2  35 -  2  30 o o  2  w 25 -  u 20 z> a  15 -  GJ  ^ 10 500.0  0.5  1.0  1.5  2.0  2.5  3.0  3.5  [Choline] (mmol) Figure 4.7: Choline concentration vs myelin water fraction plotted for each individual structure. Splenium (SP) = x, posterior internal capsules (PIC) = A, occipital white matter (OW) = • and frontal white matter (FW) = • .  66  40  R = 0.11 p < 0.05 2  4.0  6.0  12.0  8.0  [Creatine] (mmol) Figure 4.8: Creatine concentration vs myelin water fraction plotted for all white matter voxels exhibited a significant correlation. Error bars represent standard deviations for each measure.  40 j 35 30 -  a  o w  «  x • • •  SP: R = 0.074 PIC: R = 0.021 OW: R = 0.38, p < 0.05 F W : R = 0.011 2  2  2  25 20 15 -  yel  a 10 5 0 -  o.o  2.0  4.0  6.0  8.0  10.0  12.0  [Creatine] (mmol) Figure 4.9: Creatine concentration vs myelin water fraction plotted for each individual structure. Splenium (SP) = x, posterior internal capsules ( P I C ) = A, occipital white matter ( O W ) = • and frontal white matter (FW) = • .  67  40 -j 35 -  R = 0.29 p < 0.0005  a o  30 -  w 25 -  «  i- 20 15 -  s 10 5 0 0.0  2.0  1.0  3.0  4.0  5.0  6.0  NAA/Creatine Figure 4.10: NAA relative to creatine vs myelin waterfractionplotted for all white matter voxels exhibited a highly significant correlation. Error bars represent standard deviations of each measure.  40  x S P : R = 0.017 A P I C : R = 0.15 • O W : R = 0.38 • F W : R = 0.044 2  35  2  IS  30  2  e o  2  tj 25  « u u 20 rt £ 15 <B CJ  >> 10  0.0  1.0  2.0  3.0  4.0  5.0  6.0  NAA/Creatine Figure 4.11: NAA relative to creatine vs myelin water fraction plotted for each individual structure. Splenium (SP) = x, posterior internal capsules (PIC) = • , occipital white matter (OW) = • and frontal white matter (FW) = • .  68  40  0.0  0.5  1.0  1.5  2.0  2.5  3.0  C hoIine/C reatine Figure 4.12: C h o l i n e relative to creatine vs myelin water fraction plotted for a l l white matter voxels was not significantly correlated. Error bars represent standard deviation o f each measure.  40 -,  x S P : R = 0.018 A PIC: R = 0.056 • OW: R = 0.20 • FW: R = 0.26 2  35  , ^  2  30  2  e  2  u 25 M s. s. 20  15  e % 10 -  S  5 0 -  o.o  0.5  1.0  1.5  2.0  2.5  3.0  Choline/Creatine Figure 4.13: C h o l i n e relative to creatine vs myelin water fraction plotted for each individual structure. Splenium (SP) = x, posterior internal capsules (PIC) = A, occipital white matter ( O W ) = • and frontal white matter ( F W ) = • .  69  4.5 Correlations between Grey and White Matter Fractions and Metabolite Concentrations Voxels in different white matter structures varied widely in their white matter fraction (WMF) and although the myelin water fraction was calculated using only white matter voxels within the spectroscopy voxel, metabolite concentrations could not be similarly corrected for partial voluming. The variation in white matter fraction between structures is illustrated in Table 4.5, which lists the mean white matter fraction for each region and averaged over all white matter. Using the two tailed, unequal variance Student's t-test significant differences were found between structures.  Splenium voxels had a large  amount of partial voluming and had the lowest white matter fraction, 0.58±0.03. The white matter fraction in SP was significantly lower than all other white matter structures. Frontal white voxels had the highest white matter fraction, 0.89±0.02, significantly higher than all other white matter structures.  Mean White Matter Fraction ±SE  Region A l l White Matter Frontal W M Occipital W M Posterior IC Splenium  0.79±0.01 0.89±0.02 0.78±0.03 0.70±0.03 0.58±0.03  Table 4.5: Mean white matter fractions with standard errors for each white matter (WM) region and averaged over all W M .  To examine whether differences in metabolite concentrations between structures could be due to the amount of white matter in the voxel, the white matter fraction and the grey matter fraction was plotted against metabolite concentration for all white matter voxels. The plots are shown in Figures 4.14-4.16. These plots should not be symmetric when comparing white matter and grey matter trends because many voxels contained a significant CSF fraction thus W M F + G M F + 0.  For all three metabolites, N A A , Cho  and Cre, there was a significant correlation between absolute metabolite concentration and white matter fraction. For N-acetyl-aspartate, the correlation was negative between [NAA] and white matter fraction (R = 0.12, p<0.05) and a significant positive 2  correlation [NAA] and grey matter fraction (R = 0.19, p<0.005) was also found (Figure 2  70  4.14).  For [Cho] and [Cre] a positive significant correlation between absolute  concentration and white matter fraction was found, but without a corresponding significant correlation for grey matter. 14.0 12.0 10.0 8.0 <  6.0 4.0  • Fraction of White Matter: R =0.12, p<0.05 • Fraction of Grey Matter: R = 0.19, p<0.005  2.0  2  0.0 0.2  0.4  0.6  0.8  Grey/White Matter Fraction Figure 4.14: Plot of absolute NAA concentration vs grey and white matter fraction. Both tissue types exhibited a significant trend. 3.5 3.0  1.0  • Fraction of White Matter: R =0.18, p<0.005 Fraction of Grey Matter: R = 0.018  0.5  2  0.0 0  0.2  0.4  0.6  0.8  1  Grey/White Matter Fraction Figure 4.15: Plot of absolute choline concentration vs grey and white matter fraction. Trend in white matter was significant and positive, there was no significant trend in grey matter.  71  12.0  0.0 -I 0  1  1  1  1  1  0.2  0.4  0.6  0.8  1  Grey/White Matter Fraction Figure 4.16: Plot o f absolute creatine concentration vs grey and white matter fraction. Trend in white matter was significant and positive, there was no significant trend in grey matter.  4.6 Discussion  The myelin water fraction was calculated for all white matter and for four white matter structures, frontal white matter, occipital white matter, posterior internal capsules and splenium. M W F was calculated with 6 different methods and all found the same trend. The splenium of the corpus callosum had a significantly higher myelin water fraction than any other white matter structure.  The myelin water fraction decreased from  structure to structure in the following order SP>PIC>OW>FW and most of the differences between structures were found to be significantly different with an unequal variance, paired Student's t-test. The finding that the frontal white matter had the lowest myelin water fraction of white matter structures is supported by previous work in 1997 which found that the minor forceps (located in frontal white matter) had a significantly lower M W F than the major forceps (located in occipital white matter) [41]. A comparison is shown in Table 4.6. In that study the highest myelin water fraction was measured in the internal capsules, not the splenium. Also the mean myelin water fractions in the 1997 study are higher than those measured in the current study. This may be because in the  72  1997 study the myelin water fraction was calculated by drawing a precise region of interest around the structure and by examining only the voxels inside the ROI whereas the current study used large single voxel spectroscopy voxels to approximate the shape of the structure. Also, in the 1997 study the myelin water fraction was defined as tissue with a T time between 1 0 - 5 0 ms, a larger window than used in the current study. Therefore, 2  although the SVS results were corrected for partial voluming, the results still reflect that the spectroscopy voxel contains some voxels of white matter surrounding the structure of interest or even voxels of grey matter and CSF. Secondly the pulse sequences used to measure T relaxation in each study were very different where the current study included 2  composite pulses and 48 echoes and was also carried out on a different M R scanner than the 1997 study.  Mean MWF (%) 1997 Results[41] 11.28  Mean MWF (%) Current Study 7.21  Frontal White Matter  8.40  3.65  Occipital White Matter  10.11  4.74  Posterior Internal Capsules  15.00  8.08  Splenium of Corpus Callosum  13.05  12.36  Region Average over White Matter  Table 4.6: Comparison of mean myelin water fraction in various white matter regions from 1997 study and the current study.  NAA  Absolute Concentration in FW (mmol) 2000 study[321 10.0  Absolute Concentration in FW (mmol) Current Study 9.2  Choline  2.0  2.1  Creatine  8.0  7.5  Metabolite  Table 4.7: Comparison of mean absolute metabolite concentrations measured in frontal white matter (FW) in two similar studies.  Mean absolute and relative metabolite concentrations were calculated for each white matter region and throughout all white matter. The absolute metabolite  73  concentration results in frontal white matter were compared to a previous study (2000) designed to verify the quantification technique used in the current study (Table 4.7) [32]. The values are comparable which is a strong indicator that the technique is highly reproducible. Mean metabolite concentrations were found to vary significantly between white matter structures. This was true for both absolute concentrations and relative concentrations although the significant differences varied between the two types of measurements.  The standard errors of mean metabolite concentrations also varied  between structures especially for [NAA] which had the largest standard error, yet creatine measurements had the same standard errors for all structures. For all the measurements (except NAA/Cre) the frontal white matter had the highest variability.  This is an  interesting result considering that frontal white matter had the least amount of partial voluming with CSF and grey matter. Perhaps the white matter in this region is heterogeneous compared to the other regions. [NAA] was significantly low in frontal white matter. NAA/Cre was also significantly lower in frontal white matter than the average over all white matter, splenium and occipital white matter. These results could, in part, be accounted for by differences in white matter fraction between structures. Frontal white matter voxels had the highest white matter fraction of all the white matter structures.  The current study also found that [NAA] decreases significantly as white  matter fraction increases and although the R of this plot is low (R = 0.12, Figure 4.14) it 2  2  could account for the significantly low concentration of N A A in frontal white matter. This trend, combined with the opposite trend in creatine which leads to low creatine in the splenium, account for significantly high NAA/Cre in the splenium which had the lowest white matter fraction of the four structures. The same trend was seen in absolute creatine concentration. [Cre] was significantly high in the posterior internal capsules and significantly low in the splenium. The significant positive trend between white matter fraction and [Cre] could explain why the PIC with a relatively high white matter fraction corresponds to a high creatine concentration and similarly a low white matter fraction for splenium leads to low [Cre]. Significantly low choline concentration in the splenium can be explained by a significant positive slope between [Cho] and white matter fraction. Low Cho/Cre in the PIC is explained by high [Cre] in the PIC and also high Cho/Cre in SP is accounted for by low [Cre] in the SP. White matter fraction cannot account for the  74  majority of the variation of metabolite concentrations between structures because frontal white matter did not measure the highest choline or creatine concentrations. If significant trends between white matter fraction and metabolite concentration can explain regional distribution of metabolite concentrations then the meaning of the trend between white matter fraction and metabolite concentration must be examined. Dealt with simply, i f the amount of metabolite concentration varies with white matter fraction then there must be a significant difference in metabolite concentration between white matter, grey matter and CSF.  As detailed in the literature review many studies  have attempted to determine metabolite concentration differences between white and grey matter.  A popular method to do so is to carry out a linear regression by segmenting  several voxels and plotting metabolite concentration against white or grey matter fraction. Pure white and grey matter metabolite concentrations are extrapolated from the plot of metabolite concentration and white/grey matter fraction [67,95,56].  This technique is  usually carried out with MRSI because many measurements can be taken in a single subject.  The absolute quantification technique for SVS used in this study normalized  metabolite concentrations across subjects so that the data from many subjects could be combined into a single data set. This created enough data to perform a linear regression between single voxel metabolite measurements and white/grey matter fractions. Single voxel metabolite concentration measurements have much higher signal to noise than M R S I measurements and are more reliable for a linear regression analysis.  The results  of the current study were compared to the results from the literature. Most studies have found that choline in white matter is higher than grey matter [39,31,57,58,29,59,32,60]. The results from the current study support these results, i f the positive trendline in Figure 4.15 is extrapolated to W M F = 0 (pure grey matter) and W M F = 1 (pure white matter) [Cho] is higher in white matter than in grey matter. The plot of creatine concentration vs white/grey matter does not fall in line with the current literature in which most studies found that [Cre] is higher in grey matter than white matter [62,39,31,64,59,36,60]. The positive correlation in Figure 4.16 predicts the opposite which supports the findings of one study [58]. The literature on the regional variation of N A A does not contain a clear message. The same number of studies have found that N A A is higher in white matter as have found the opposite.  The current study found a significant negative correlation  75  between [NAA] and white matter fraction which supports the body of literature that finds that [NAA] is greater in grey matter than in white matter [56,66,30,67]. This result is consistent with the traditional view that N-acetyl-aspartate is a neuronal marker. The density of neuronal bodies is higher in grey matter and thus the concentration of N A A should be higher in grey matter. The correlation between N A A and white matter fraction in this study is weak and likely the difference between N A A concentration in grey and white matter is small given the high number of contradictory studies. Although trends between metabolite concentrations and grey/white matter fractions can be used to explain in part the regional variation of metabolites observed in this study, it is still unclear exactly what the grey/white differences are for creatine and N A A . Regional variation of metabolite concentration throughout white matter was observed in the current study.  Comparable previous studies using single voxel  spectroscopy found that of the regions studied (frontal, parietal and occipital tissue) the only significant difference observed was elevated [NAA] in the occipital lobe [36,68,69]. The current study which found that [NAA] was decreased in frontal white matter and therefore [NAA] in occipital white matter was significantly higher than frontal white matter, but not significantly different from other structures. Studies have also seen low [Cho] in the occipital lobe [56,36].  Again, these results are not consistent with the  current study but these studies of regional variation are not white matter specific. Other studies examining different white matter structures than the current study found significant differences in N A A between structures and also in N A A throughout grey matter [62,70,58]. Few results are presented for choline and creatine because they are hard to quantify.  Therefore it is unsurprising that regional differences in N A A , choline  and creatine have been measured throughout white matter. Metabolite concentrations were plotted against myelin water fractions to observe whether trends between the two correspond. [NAA] vs M W F in all white matter had a significant positive correlation with a low R = 0.17 implying that only about 17% of the 2  variation in one variable is caused by the other. NAA/Cre vs W M F also exhibited a significant positive correlation. The [NAA] data were plotted for individual structures and a strong correlation (R = 0.57) was found for splenium which could be driving the 2  correlation for all white matter. No correlations were found within individual structures  76  for NAA/Cre. The strong correlation for splenium arises because splenium had a high average myelin water fraction and a high [NAA] concentration. Splenium also had the lowest white matter fraction of all white matter voxels which results in a large CSF correction for metabolite quantification and white matter correction for M W F .  This  could indicate that the corrections in the quantification technique significantly affect the results. Metabolite concentrations cannot be corrected for partial voluming because the spectrum represents the whole voxel, even i f it isn't all white matter. Concentrations in the splenium would be most affected by this since they have the lowest white matter fraction.  If [NAA] decreases with white matter fraction then the measured [NAA] of  splenium is high and this could account for the significant correlation between [NAA] and M W F in the splenium. Given the negative slope between [NAA] and white matter fraction and the belief that N A A is a neuronal marker it seems plausible that N A A would decrease with myelin water fraction because areas with high myelination would leave little space for neuronal bodies and axons which contain N A A (according to immunocytochemical studies [3]). The function of N A A is not well understood and these results could be used to support or contradict a number of postulates. Other sources in the literature use N A A as a marker of neuronal integrity [26], in which case the current study's results could imply that neurons with high myelin also measure high N A A . Links between N A A and myelin lipid synthesis are supported by high concentrations of N A A in white matter and aging studies show links between N A A and myelin[63,15]. Figure 4.5 indicates that the trend between N A A and M W F varies between structures, whether significant or not. This could be indicative of the pathology of the structure, the size of axons, the amount of myelination, the orientation of the neurons or the amount of intra and extra cellular water. Although well characterized, it is not well understood why the myelin water fraction varies widely between white matter structures. The variation in N A A is even less well understood as its function and localization has not been completely explained.  The current study fails to provide conclusive links between  [NAA] and myelin because trends between [NAA] and white matter fraction and between [NAA] and myelin water fraction are both weak and contradict each other. Choline concentration plotted against myelin water fraction did not exhibit a significant correlation although it had a negative slope.  When plotted for individual  77  structures a significant correlation was found within frontal white matter. Frontal white matter had the highest white matter fractions but lowest myelin water fractions. Choline also had the highest concentration in frontal white matter although it was not significantly higher. Choline is involved in lipid synthesis and thus some papers associate it with myelin. In this case it would be expected that a positive correlation would exist between the two variables. This was not observed in the current study. In the literature choline has been shown to be higher in white matter than in grey matter therefore high partial voluming in voxels with high myelin water fractions could have produced decreased choline concentrations and thus no correlation with myelin water fraction. Cho/Cre had no significant correlation with myelin water fraction in either white matter or within individual structures. The lack of significant correlation for choline is probably due to the small range of choline concentrations (1.5 - 2.5 mmol) since the significant correlation for frontal white matter was driven by one data point which had a choline concentration of 3.0 mmol. Creatine concentration and myelin water fraction were significantly negatively correlated but the correlation is weak and i f one point with high myelin water fraction is removed the correlation weakens and does not reach significance. When creatine is plotted against M W F for individual structures some significant differences between structures are apparent. Splenium measurements have a narrow range of low creatine concentrations and PIC measurements are all within a narrow range of high creatine concentrations. This difference is also observed in NAA/Cre and Cho/Cre plots which see the opposite effect as the measurements are relative to creatine. Therefore NAA/Cre and Cho/Cre have low values in the PIC and high values in the SP. A significant negative trend was observed between [Cre] and M W F in occipital white matter. This correlation is fairly strong (R = 0.38) and can not be fully explained with the current 2  results but previous studies have found interesting results in the occipital lobe such as increased N A A and decreased choline.  78  4.7 Summary Myelin water fractions measured within four white matter regions reproduced previous work and confirmed that myelin water fraction varies throughout white matter and in this study, M W F increased from frontal white matter to occipital white matter, posterior internal capsules and splenium of the corpus callosum. Absolute concentrations of N A A , choline and creatine also reproduced previous work on a novel method of quantifying metabolite concentrations using external water standards and T relaxation methods. 2  Four white matter structures were studied and significant differences between them were observed. Frontal white matter voxels had the lowest myelin water fraction and also low [NAA] and NAA/Cre which was, in part, due to a significant negative slope between [NAA] and white matter fraction. The trend between [NAA] and M W F in all white matter was significant with a positive correlation. This trend was particularly strong within the splenium where N A A concentrations may be artificially high due to low white matter content within the spectroscopy voxel. The splenium measured the highest myelin water fraction and also significantly low choline and creatine. Low choline and creatine in the splenium and high creatine in the posterior internal capsules were due, in part, to positive correlations between choline and creatine and white matter fractions. Significant differences in creatine concentration call into question the common practice of using creatine as a constant for calculating relative concentrations. Choline results in this study were consistent with the literature but regional differences between grey and white matter of creatine and N A A remain unclear, both from the current results and from conflicting reports in the literature.  Studies into the distribution of metabolite concentrations  throughout similar white matter regions are few and report only increased N A A and decreased choline in occipital white matter, results which were not reproduced with MRSI.  Standard deviations of mean metabolite concentrations indicate that for some  regions the metabolite concentration varied widely between volunteers suggesting that the technique is sensitive to biological variability between subjects for certain structures. The current study attempted to link myelin water fraction and metabolite concentrations in white matter. Results show that the trends vary between structures for reasons which could include the pathology of the structure, the size of axons, amount of myelination, the  79  orientation of the neurons and the amount of intra and extra cellular water. [NAA] was positively and significantly correlated with myelin water,fraction both in all white matter and within the splenium. A positive significant trend between [NAA] and M W F could illustrate that areas with high myelin water fractions also have high amounts of N A A . The body of literature on the function of N A A is contradictory and the results of this study could be used to support or contradict several postulates. Choline had a significant positive trend within frontal white matter which is consistent with literature linking choline to myelin lipid synthesis.  Correlations between creatine concentration and  myelin water fraction were weak for all white matter and significant within the occipital lobe where other studies have also observed regional differences. Regional distributions of myelin water fraction and metabolite concentrations and trends between the two add to the conflicting literature on metabolite concentrations and their functions.  80  Chapter 5  Spectroscopic Imaging Study Results  5.1 Introduction  Magnetic resonance spectroscopic imaging (MRSI) was used to measure metabolite concentrations in multiple voxels throughout a 10 mm slice of brain. Myelin water fractions were also calculated and trends between the two measurements will be examined in three subjects.  Significant differences in metabolite concentrations and  myelin water fractions between white matter structures will be presented. Metabolite concentrations will also be plotted against white and grey matter fractions. These results will be compared to the results of the single voxel spectroscopy study.  5.2 Metabolite Concentrations and Myelin Water Fractions  Magnetic spectroscopic imaging measures spectra from multiple voxels throughout the brain. For the current study, 32x32 voxels were acquired from the entire field of view. Figure 5.1 shows a grid of spectra superimposed onto a C P M G M R I image. PRESS was used to localize and shim a region within the brain to improve the spectra in the study's regions of interest. Although this technique produces a large quantity of data, the low signal to noise of MRSI means that many of the spectra will not produce reliable measurements of metabolite concentration. Several filters were used to eliminate these measurements.  LCModel was used to calculate concentrations of N A A , choline and  creatine. These concentrations were not normalized or measured in mmol. They cannot be compared between subjects but can be used to examine trends within each subject's brain. These trends will be compared between subjects and to single voxel spectroscopy results. LCModel returns a %SD for each concentration measurement and a signal to  81  noise ratio for each spectrum. These values were used to separate good spectra from bad spectra.  Figure 5.1: Image from the first echo in a 48 echo sequence CPMG MRI sequence, superimposed with 25x25 grid of MRSI spectra. Spectra werefilteredwith metabolite %SD and SNR cutoffs and only spectra from white matter voxels were used.  N A A is the largest peak in the human in vivo spectrum and is the easiest to measure. The LCModel %SDs for N A A in the first study's single voxel spectra were, on average, 6%. Since the signal to noise ratio was much lower in the smaller MRSI voxels, spectra with %SD < 20% were accepted as good data [87]. Choline and creatine peaks were much smaller and had larger %SD than N A A . For single voxel spectra the %SDs were on average 11%, thus a higher %SD threshold for choline and creatine of 40% was used. The average %SD of all MRSI spectra could not be used to determine the cutoff because it was very high due to the numerous 999% SDs which indicates a spectrum which could not be fit by LCModel. After voxels were sorted by %SD they were sorted by white matter fraction. Voxels were segmented using the same algorithm as in the single voxel study.  A white matter fraction for the entire spectroscopy voxel was  determined with the segmentation routine so that spectra could be classified into white or grey matter spectra. Voxels with white matter fractions greater than 70% were accepted as pure white matter voxels. This was decreased from the 80% threshold used in the SV study because the MRSI spectroscopy voxels were larger than the individual imaging voxels (segmented in the SVS) and one would expect them to have lower white matter fractions. White matter maps produced by the segmentation algorithm with a 70% cutoff  82  only included voxels in reasonable white matter regions of the brain. The data set of spectra included after the two filters was examined closely. Some spectra were from voxels in areas of the brain which were very badly shimmed, too close to the skull or close to the ventricles. A final filter was used to eliminate these spectra. Only spectra with SNR (as measured by LCModel) greater than one were accepted as good data. Single voxel spectra, with good signal to noise, had SNR ~ 8. After the filters the mean %SD of good N A A spectra was ~ 16%, 31% for choline spectra and 27% for creatine spectra. After all the filters were applied the voxels were separated in the four white matter regions depending on their location within the brain. For each subject the mean myelin water fraction and mean metabolite concentration for all accepted voxels were calculated for each white matter structure and averaged over all white matter. If there were fewer than 5 good voxels for a given white matter structure the voxels were included in the data set for the average over all white matter, but the structure was not included in the analysis for significant differences between structures. For each subject the number of good voxels for each structure varied (between structures and between subjects).  The mean myelin water fraction and the  number of good voxels used to calculate the means are presented for each subject in Table 5.1. The number of good voxels for the myelin water measurement was determined by combining the good voxels from each metabolite measurement. From Table 5.1 some trends are clear, the splenium of all three subjects had the highest myelin water fraction and the frontal white matter had the lowest M W F .  Mean MWFs were examined for  significant differences with the unequal variance paired Student's t-test. For subject A , M W F in occipital white matter was significantly lower than SP (p<0.01) and the average over all white matter (p<0.05) and the SP was significantly higher than the average over all white matter (p<0.05). For subject B , the M W F in frontal white was significantly lower than all other white matter structures (p<0.05). For subject C, the only significant difference existed between F W and PIC (p<0.05). Subject C had much lower myelin water fractions than subject A and B.  83  Subject  A  i  B  C  Region  Mean MWF %±%SE  Number of good voxels  WM  8.6±0.3  27  FW  N/A  0  OW  6.9±0.1  9  PIC  8.5±0.2  9  SP  10.4±0.2  9  WM  6.89±0.04  59  FW  4.55±0.07  19  OW  6.7±0.1  13  PIC  7.2±0.1  16  SP  10.7±0.3  11  WM  4.56±0.03  57  FW  3.36±0.09  15  OW  4.05±0.12  16  PIC  5.19±0.08  20  SP  6.49±0.4  6  Table 5.1: For each subject (A,B and C) the mean MWF with standard errors and number of good voxels is presented for each white matter region and averaged over all white matter.  Metabolite concentrations were also calculated for each subject and in each region. The number of good voxels in each region and for each subject varied widely. The mean concentration of metabolites in each region is shown in Figure 5.2.  The  magnitude of the concentrations can be compared within each subject but not between subjects because they have not been absolutely quantified. Where myelin water fraction had a consistent trend between structures, metabolite concentrations do not. example,  [NAA]  increased  from  PIC<OW<WM  in  subject  A,  For from  FW<OW<WM<PIC<SP for subject B and OW<SP<WM<FW<PIC for subject C. Plots in Figure 5.2 suggest that there are few differences between structures when metabolite concentration is measured with MRSI and calculated with LCModel.  The unequal  variance, paired Student's t-test was used to test quantitatively for significant differences  84  between structures.  No significant differences were found for subject A . [Cre] in  occipital white matter for subject B was found to be significantly lower than [Cre] in all white matter, frontal white matter and posterior internal capsules (p<0.05).  Subject C  had significantly lower [NAA] in occipital white matter when compared to F W and PIC (p<0.05). Subject C also had significant differences in creatine concentration. [Cre] was significantly lower in the occipital white matter when compared to F W (p<0.05), PIC and the average over all white matter (p<0.01).  NAA  Cho  Cre  Figure 5.2: Plots of mean metabolite concentration for three subjects (A,B and C). For each metabolite the concentration was calculated in five brain regions: (from left to right) all white matter, frontal white matter, occipital white matter, posterior internal capsules and splenium. For some subjects and metabolites there were too few measurements in a region to calculate a mean. Standard errors were too small to display.  5.3 Correlations between M W F and metabolite concentrations  Metabolite concentrations were plotted against myelin water fraction for each subject. For each metabolite all good white matter voxels were plotted to determine whether there was a trend in all white matter. Good white matter voxels were defined as voxels that  85  had a white matter fraction of at least 70% and which were accepted as good data on the basis of LCModel outputs such as %SD and SNR.  Figures 5.3, 5.4 and 5.5 show [NAA],  [Cho] and [Cre] concentration plotted against M W F respectively. On each plot the data, trend lines and R values are displayed for each of three subjects. No significant 2  correlations were observed. For each set of data the good white matter voxels were divided into white matter structures and plotted to examine whether trends between myelin water fraction and metabolite concentrations vary between structures.  If a structure had fewer than five  measurements the structure was not plotted and examined for correlations. However these measurements were included in the plot for all white matter voxels.  Plots of  [NAA], [Cho] and [Cre] against myelin water fraction in individual structures for each subject are displayed in Figures 5.6-5.14. The only plot with significant correlations was Figure 5.8, [NAA] vs myelin water fraction for individual white matter structures for subject C. [NAA] in occipital white matter had a significant positive correlation with myelin water fraction (p<0.05, R = 0.47) and [NAA] in the posterior internal capsules 2  had a significant negative correlation with myelin water fraction (p<0.05, R = 0.34). There were more measurements in the PIC (n=15) than in O W (n=8). Figures 5.3-5.5 show that although concentrations were not corrected or normalized, the range of [NAA], [Cho] and [Cre] within each subject was similar.  86  Figure 5.3: [NAA] plotted against myelin water fraction in all good white matter voxels for three subjects. None of the data sets show a significant correlation.  Figure 5.4: [Choline] plotted against myelin water fraction in all good white matter voxels for three subjects. None of the data sets show a significant correlation.  87  40 35 =\30  e  •A  = 0.051  xB  R = 0.024 2  • C R = 0.0014 2  o  tJ25 «  u20  «  ^15  a  GJ  .^10  • 100  tax 200  300  400  500  600  [Creatine] (LCModel units) Figure 5.5: [Creatine] plotted against myelin water fraction in all good white matter voxels for three subjects. None of the data sets show a significant correlation.  Figure 5.6: [NAA] plotted against myelin water fraction in two white matter structures, occipital white matter (OW) and splenium (SP), for subject A.  88  Figure 5.7: [NAA] plotted against myelin water fraction in four white matter structures, frontal white matter (FW), occipital white matter (OW), posterior internal capsules (PIC) and splenium (SP) for subject B.  25  Subject C • F W R = 0.32 • OW R = 0.47, p<0.05 A P I C R = 0.34, p<0.05 x S P R = 0.43 2  r20  X  2  2  s  2  o os j.  u |10  \  A A A  a  ° 200  300  400  a 500  A  x  A  X  600  +  700  [NAA] (LCModel units)  *  800  X  A  A  900  1000  Figure 5.8: [NAA] plotted against myelin water fraction in four white matter structures, frontal white matter (FW), occipital white matter (OW), posterior internal capsules (PIC) and splenium (SP) for subject C.  89  Figure 5.9: [Choline] plotted against myelin water fraction in one white matter structure, posterior internal capsules, for subject A. All other white matter structures had too few measurements to be plotted.  100  200  300  400  500  600  [Choline] (LCModel units) Figure 5.10: [Choline] plotted against myelin water fraction in two white matter structures, frontal white matter (FW) and posterior internal capsules (PIC) for subject B.  90  10  Subject C • F W R = 0.034 • O W R = 0.14  9  2  8  a o  7  « a  6  2  A PIC R  2  = 0.00001  5 4 3 2  A A  1  100  200  300  400  500  600  700  [Choline] (LCModel units) Figure 5.11: [Choline] plotted against myelin water fraction in three white matter structures, frontal white matter (FW), occipital white matter (OW) and posterior internal capsules (PIC) for subject C.  16 -|  Subject A 14^  D O W R A PIC R  2  2  = 0.63 = 0.009  ^ 1 2 - x S P R = 0.17 2  0 -I  1  1  0  50  100  -,  150  1  200  !  r  250  300  1  1  350  400  [Creatine] (LCModel units) Figure 5.12: [Creatine] plotted against myelin water fraction in three white matter structures, occipital white matter (OW), posterior internal capsules (PIC) and splenium (SP), for subject A .  91  40 35  Subject B • F W R • O W R  C30  = 0.011  2  = 0.0001  2  c  APICR  o  x S P R = 0.025  = 0.11  2  2  '•§25 cs  u i-20 rt ^15  c  >>10  100  200  300  400  500  600  [Creatine] (LCModel units) Figure 5.13: [Creatine] plotted against myelin water fraction in four white matter structures, frontal white matter (FW), occipital white matter (OW), posterior internal capsules (PIC) and splenium (SP) for subject B.  25  Subject C • F W R = 0.020 2  ;^20  • O W R = 0.050  X  2  A P I C R = 0.067 2  c  x S P R = 0.023 2  o  £10  A  s  S  A  • 5  " •  X  %  •  100  ^ X 4>  • *  •  •  200  '  300  400  500  600  [Creatine] (LCModel units)  Figure 5.14: [Creatine] plotted against myelin water fraction in four white matter structures,frontalwhite matter (FW), occipital white matter (OW), posterior internal capsules (PIC) and splenium (SP) for subject C.  92  5.4 Correlations between grey and white matter fractions and metabolite concentrations. To determine whether metabolite concentration varies with the amount of white or grey matter these two measurements were plotted against each other for each subject. For N A A , all voxels with %SD < 20% were plotted against their white matter and grey matter fraction.  For choline and creatine all voxels with %SD < 40% were  plotted.  A  significant correlation between [NAA] and white/grey matter fraction was only found for subject C where R = 0.063 and p<0.05 for white matter and R = 0.055 and p<0.05 for 2  2  grey matter. These results are shown in Figure 5.15. 2000 j 1800 -  Subject C • Fraction of White Matter, R = 0.063, p<0.05 • Fraction of Grey Matter, R = 0.055, p<0.05 "S 1400 " z  1600 -  2  del  a o  U  < <  z  1200 " 1000 800 i\ 600 |_ J  400 200 0 0  0.2  0.4  0.6  0.8  1  White and Grey Matter Fraction Figure 5.15: [NAA] plotted against white and grey matter fraction only shows a significant correlation for subject C.  For choline, significant correlations were observed between [Cho] and white matter (R = 0.028, p<0.05), but not between choline and grey matter in subject A (R = 0.018). 2  Correlations for subjects B and C were not significant. Figure 5.16 illustrates [Cho] vs white and grey matter for subject A . Subject B had a significant correlation between [Cre] and grey matter fraction (R = 0.036, p<0.05). Figure 5.17 contains this plot. None 2  of the plots had a significant correlation between white matter and [Cre]. A l l white/grey matter fraction plots have many measurements at W M F = 1. These data points do not represent pure white matter voxels, most of these voxels are close to the skull or even  93  outside the head because the segmentation algorithm classifies these dark voxels (due to lack of signal) as white matter voxels and unfortunately the %SD filter did not eliminate all such voxels. 2000  Subject A ~ ~ • Fraction of White Matter, R = 0.028, p<0.05 = 0.018  1800  2  1600 1400  •  •  CLP  TS  O 1200  s  •  I  1000 • • •  800  a "3 600 i 400  1  • •  •  "' 1  '  •  • • •  ° q n  N 'b  ,  "  ,  f  -  >  »  M  200  •  JB. - - - - *  .  =  •  •  i •  m  • n •  D  j  1  0 0  0.2  0.4  0.6  0.8  1  White and Grey Matter Fraction Figure 5.16: [Cho] plotted against white and grey matter fraction only shows a significant correlation for subject A.  800 700 600  Subject B • Fraction of White Matter, R = .0012 Fraction of Grey Matter, R = 0.036, p<0.05 z  2  •o 500 O u 400 J  300  a o»  u  •  •, • •  ***  T  ° o. -  •  -».--- s  :•B " " • n  200 100 0 0.2  0.4  0.6  0.8  White and Grey Matter Fraction Figure 5.17: [Cre] plotted against white and grey matter fraction only shows a significant correlation for subject B.  94  5.5 Discussion  Several measurements of myelin water fraction were made within four white matter structures of three subjects using MRSI. A l l subjects had the same trend where the M W F was highest in the splenium and lowest in frontal white matter.  The myelin water  fraction of the four structures increased in the following order FW<OW<PIC<SP, and most of these differences were found to be significant using an unequal variance paired Student's t-test. Some differences were not significant because too few measurements were made in the structure. Single voxel spectroscopy also found that the M W F in SP was significantly higher than other white matter structures and that F W was significantly lower. The same trend across all four structures was observed in both studies. This trend also confirms previous work which found that the minor forceps in frontal white matter had a lower myelin water fraction than the major forceps in the occipital white matter region [41]. In the current work the splenium of the corpus callosum was found to have the highest myelin water fraction. This contradicts previous work which found that the internal capsules had a higher myelin water fraction than splenium [41]. Previous studies calculated the myelin water fraction by drawing a very precise region of interest around the internal capsules. In the current study the region encompassed by the voxels was much larger and would contain some deep grey matter structures locate near the PIC and white matter which was not attributed to the posterior internal capsules in the previous study. The myelin water fractions calculated in this study are well replicated among subjects and between techniques, reproduces previous work and confirms that myelin water fraction varies with a consistent pattern throughout white matter. Trends in metabolite concentrations varied for the three subjects studied with MRSI.  Few significant differences were found when metabolite concentration means  were examined with the unequal variance, paired Student's t-test.  [NAA] was  significantly low in occipital white matter when compared to F W and PIC for subject C and [Cre] was significantly low in OW when compared to FW, PIC and all W M for both subject B and subject C, although the significance was stronger for subject C. Differences in the results between subjects could be explained by several reasons. The  95  magnetic field shimmed differently for each person and from the grids of spectra it is clear that some regions shimmed better than others. For example subject A had no good spectra in the frontal white region, or on the left occipital lobe but several good spectra were obtained from the right occipital lobe. These effects lead to wide differences in the number of good measurements in each region for each subject which makes it difficult to find significant differences and to compare one set of significant differences to another. MRSI found much fewer significant differences between the four regions than single voxel spectroscopy did.  Differences in MRSI and SVS results will point to the effects of  corrections in the absolute quantification of metabolite concentrations and are largely due to large differences in SNR between the two techniques. SVS found that [NAA] was significantly lower in frontal white matter and this contradicts the findings of MRSI in subject C which found that [NAA] was lower in occipital white matter than FW. Previous studies found elevated [NAA] in the occipital lobe (both white and grey matter) which is confirmed with SVS but not with MRSI [36,68,69].  S V spectroscopy found  significant differences in the splenium which were not found with MRSI, probably because it was difficult to get many measurements within the splenium.  These  differences could also be due to the high CSF fraction in SVS splenium voxels. The CSF correction in the quantification procedure had a large effect on metabolite measurements in the splenium. In SVS the splenium had significantly low choline and creatine. The posterior internal capsules had significantly high creatine when measured by SVS. Creatine in the SP was higher than OW but this difference was not significant in either subject B or C. MRSI measured low creatine in occipital white matter for two subjects. Previous M R S I studies have measured low choline and high N A A in occipital white matter but no significant differences were found for creatine concentrations [36,56]. One of the goals of this study was to find and account for variations in metabolite concentrations throughout white matter. matter structure, occipital white matter.  MRSI found differences in only one white [NAA] and [Cre] were measured to be  significantly low in occipital white matter. Trends between [NAA], [Cho] and [Cre] and white matter fraction were examined to determine whether metabolite concentrations could depend on the amount of white matter in the voxel. white/grey matter were significantly correlated.  In subject C, [NAA] vs  The trend between [NAA] vs white  96  matter was significantly positive indicating that voxels with high white matter fractions should also have high N A A concentrations. This result was opposite to the single voxel study. [NAA] was also significantly lower in occipital white matter than in other white matter structures. This was not due to a W M effect because no significant differences in white matter fraction were found between subject C's structures.  Low creatine  concentration in subject C's occipital white matter cannot be explained by trends between [Cre] and white matter fraction because no significant trend was observed for subject C. A significant trend between [Cre] and grey matter fraction was observed for subject B , who also measured significantly low creatine levels in occipital white matter.  The  positive trend between grey matter and [Cre] means that [Cre] is higher in white matter than grey matter, which would lead to high [Cre] in voxels with high white matter fractions. Unfortunately no significant differences were found between voxels of the four white matter structures, so significantly low creatine concentration in occipital white matter did not correspond to low white matter fraction. A significant trend between choline and white matter fraction was observed for subject A , but no other significant differences in metabolite concentrations were observed for subject A .  Regional  distributions of metabolite concentration through white matter as measured by M R S I were not related to the white matter fraction of white matter structures' voxels. Metabolite concentrations were plotted against myelin water fractions and few significant differences were found compared to the results from the single voxel study. Most of the correlations in the single voxel study could be explained by white matter fractions and outlying points which did not have the same effect for the MRSI data. Plots of [NAA], [Cho] and [Cre] vs M W F for all white matter did not have any significant correlations for any of the three subjects in the MRSI study.  When metabolite  concentration was plotted against M W F for individual structures only two significant slopes were observed in all the metabolites, structures and subjects studied. For [NAA] vs M W F a significant negative slope was measured in the PIC of subject C and a significant positive slope was observed in OW of subject C. Interestingly subject C was the only subject who had a significant trend between [NAA] and white matter fraction. The PIC had the lowest white matter fraction of the four voxels, but it was not a significant difference, also PIC voxels had high [NAA] when compared to O W and SP  97  voxels. In the single voxel study PIC had significantly higher [NAA] when compared to F W only. This was not consistent with the positive trend between [NAA] and white matter fraction which predicts that voxels with low white matter also have low N A A concentrations. R  2  The significant trends in PIC and OW were convincing with fairly high  and no outlying points.  In subject B , where trends in PIC and O W were also  measured, the slope for PIC was positive instead of negative and the slope for O W was positive but not significant as in subject C.  Therefore different subjects display different  results within the same study. No trends were observed within structures for choline or creatine concentrations. Choline had the fewest good spectra of the metabolites and thus it was hard for differences to be found significant with few measurements within a structure and many structures which didn't have enough measurements to be studied.  5.6 Summary  Mean myelin water fractions measured in MRSI voxels within four white matter structures exhibited the same trend in all three subjects in the study. The trend also reproduced the single voxel spectroscopy study results and previously published work. Uncorrected mean metabolite concentrations measured with MRSI did not find the same regional distribution of metabolites throughout white matter as was demonstrated with single voxel spectroscopy.  Effects introduced with normalization with an external  standard, T i , T and CSF corrections can partially account for differences between SVS 2  and M R S I results. Significant trends between white matter fractions and metabolite concentrations could not explain regional distribution of metabolite concentrations since no significant differences were found in mean white matter fractions between regions. Regional distribution of metabolites varied between subjects. This is largely due to the fact that different numbers of good spectra were obtained in each subject and within each region due to shimming. [NAA] was lower in occipital white matter for subject A and creatine was significantly low in occipital white matter for subjects B and C. Interestingly, literature on regional distribution of metabolites in white matter also only cite differences in occipital white matter, although the differences were not confirmed by the current study. Metabolite concentrations plotted against myelin water fraction had  98  few significant correlations. Among all metabolites, structures and subjects only two significant trends emerged.  In subject C, [NAA] vs M W F in the posterior internal  capsules and occipital white matter were significantly correlated. These trends were not reproduced in other subjects or in the single voxel study. These correlations were strong and convincing but the specific implications are unclear. MRSI voxels have much lower signal to noise for measuring metabolite concentrations than single voxel spectroscopy and thus less significant differences were found with MRSI and they were not consistent between subjects. MRSI may not be sensitive enough to detect regional distribution of metabolite concentrations which, from the conflicting literature, are difficult to detect and to correlate with the trends in myelin water fraction which show more consistent results for regional distribution throughout white matter.  99  Chapter 6  Conclusions  6.1 Conclusions The goals of this thesis were to measure metabolite concentrations and myelin water fraction with single voxel spectroscopy and magnetic resonance spectroscopic imaging and to examine the regional distribution of these two measurements both individually and together throughout white matter by studying four white matter structures.  Regional  variation of myelin water fraction has been well characterized by previous studies [41,51]. High myelin water fractions in certain white matter structures were confirmed with measurements within single voxel spectroscopy voxels and MRSI voxels placed in the region of the splenium and the posterior internal capsules.  Low myelin water  fractions were calculated for frontal and occipital white matter regions with both studies. High myelin water fractions correspond to regions in the brain where large neuronal fiber tracts are located such as the posterior internal capsules where white matter fibers connect the frontal region to the occipital region and the splenium of the corpus callosum where fiber tracts run between the left and right hemispheres of the brain. Mean metabolite concentrations were measured with single voxel spectroscopy and were absolutely quantified after applying T i , T and CSF corrections and normalizing 2  with an external standard. Metabolite concentrations were also measured using MRSI but they were not normalized or corrected since the Ti and T varies throughout the brain 2  and these values could not be known for the multitude of spectra measured with M R S I [88,89]. M R S I voxels were much smaller than single voxel spectroscopy voxels and thus had better spatial resolution and lower partial voluming at the expense of signal to noise. Each M R S I scan produced approximately 25x25 voxels with spectra within the brain but after applying several filters the number of remaining voxels for each subject and within  100  each structure were only just sufficient to carry out this study. SVS voxels were large and contained much more tissue besides the white matter structure being studied. Although the myelin water fractions were corrected for partial voluming by only using voxels with a minimum 80% white matter fraction the spectra could not be similarly corrected.  This was a large effect for splenium voxels which had the highest CSF  fractions and was minimal for frontal white voxels which were mostly white matter. This could explain the lower myelin water fractions obtained in this study when compared to previous work. The quantification technique corrected voxels for T i and T variations 2  throughout the brain but accurate Ti and T values were not available for all the regions 2  studied and were approximated by using values for parietal white matter.  The  quantification technique would be improved i f these values were obtained for several regions within white matter. Each technique had its advantages and disadvantages, SVS voxels had poor spatial resolution and significant partial voluming, MRSI voxels had low signal to noise and few reliable spectroscopy measurements relative to the large quantities of data which can be obtained in a single scan. Differences in the results between the two techniques can largely be accounted for in terms of the issues discussed above. Single voxel spectroscopy measurements detected many significant differences in the regional distribution of metabolite concentrations but many were related to the white matter fraction of the voxels. MRSI was not sensitive enough to detect any reliable trends in metabolite concentrations which could be reproduced between subjects. Absolute metabolite concentrations measured in frontal white matter were consistent with those calculated in a study designed to verify the quantification technique [32].  Regional distributions of metabolite concentrations measured in each study were  tested with an unequal variance, paired Student's t-test. SVS found that [NAA] and NAA/Cre were increased in frontal white matter but MRSI found [NAA] was significantly low in occipital white matter in one subject. SVS frontal white results were partially related to high white matter fractions and a negative correlation between [NAA] and white matter fraction.  Previous work found significantly high [NAA] within  occipital white matter which was confirmed by the SVS study where N A A concentration was higher in occipital white matter than frontal white matter [36,68,69]. Significantly low choline and creatine was measured in the splenium of the corpus callosum with SVS.  101  These results were partially due to low white matter fractions in splenium voxels and a positive correlation between concentrations and white matter fraction. No significant results were found within the splenium with MRSI because in most subjects and with most metabolites too few good spectra were obtained from the splenium due to its proximity to the  ventricles and poor shimming.  Significantly  high creatine  concentrations were obtained with SVS within the posterior internal capsules and this too was, in part, the result of a high white matter fraction. MRSI measured significantly low creatine concentrations within the occipital white matter of two subjects.  Previous  studies also detected many significant differences within the occipital lobe, namely elevated N A A and decreased choline [36,69,68,56]. Regional differences in metabolite concentrations were detected throughout white matter with both single voxel spectroscopy and MRSI. Some of these differences were related to the amount of white matter in the voxel but these correlations were weak and could not account for all the variation of metabolite concentrations. The dependence of metabolite regional distribution on white matter fraction implies significant differences in concentration between grey and white matter. For choline the grey/white ratio is well established in the literature with higher choline in white matter than grey matter [56] and this trend was confirmed in the current study.  High choline in white matter is consistent  with choline's role in lipid synthesis which is also supported by pathology studies showing elevated choline during periods of demyelination when myelin lipids are being broken down [72]. Creatine and N A A grey/white differences have been studied but the results are contradictory both in the literature and within the current study.  Creatine is  essential for energy production and possibly regions with increased creatine such as the posterior internal capsules are involved in energy production for the brain. The regional distribution of N A A is under debate and the resolution will have strong influences on the debate surrounding the function of N A A . Different aspects of the current study support different sides of the debate. A negative correlation between N A A and white matter fraction predicts that N A A is higher in grey matter than white matter which supports the theory that N A A is a neuronal marker and is related to density of neurons which is higher in grey matter than white matter. A significant correlation between N A A and myelin water fraction implies that N A A could be a marker of neuronal integrity.  102  Correlations between metabolite concentrations and myelin water fraction give additional information about metabolite distributions. A weak correlation was observed between N A A and M W F which was observed with both absolute and relative concentrations in SVS and within occipital white matter and posterior internal capsules with MRSI.  This result is not consistent with the postulate that N A A is a neuronal  marker since areas with high myelination do not necessarily have more neurons as myelin takes up a lot of space, it may be more congruous with the theory that N A A is a marker of neuronal integrity with is also supported by a strong body of literature. A negative trend between creatine and myelin water fraction found with SVS is consistent with findings that creatine is associated with astrocytes rather than neurons [33].  No  significant trends were established between choline and myelin water fraction which is surprising since choline and myelin have been linked in several previous studies [36,59,63]. Metabolite concentrations are hard to measure absolutely and to quantify and thus many contradicting studies have attempted to determine their regional distribution and link their results to the function of several metabolites. Establishing relationships between metabolite concentrations and other quantitative M R I measures will give additional information about metabolites and their functions.  The current study has  established weak relationships between N A A and creatine and myelin water fraction which are consistent with previous studies in the literature.  6.2 Future W o r k  The results of the MRSI study were inconclusive and the S V study established weak relationships between several variables. The magnetic resonance spectroscopic imaging study did not uncover any trends in metabolite concentrations between structures or subjects. In future MRSI studies, the SNR must be improved by using a thicker slice or by increasing acquisition times. Also more data points must be collected to observe a trend thus more volunteers should be examined and a quantification technique could be developed for MRSI so that data from several volunteers could be combined into one data set.  Some groups are attempting quantitative MRSI [96,95] but not with the rigorous  techniques presented here.  The reproducibility of non-quantitative and quantitative  103  M R S I should also be confirmed. Rigorous reproducible filter techniques need to be developed for MRSI so that all the spectra studied are providing reliable and accurate data. Although most current studies of regional distributions of metabolites use MRSI because of the large amount of information which can be collected in a relatively short time, these studies do not address Ti or T regional variability throughout the brain or use 2  a correction between volunteers with an external standard. The current study shows that M R S I studies must be conducted with caution and that the single voxel spectroscopy techniques presented here, although time consuming, are more effective for uncovering trends of metabolite concentrations throughout the brain. In single voxel spectroscopy, absolute quantification of metabolite concentrations could be improved by acquiring the correct Ti and T times for all the regions examined 2  in this study and in future studies. Also this work could be extended to the study of regional distribution of metabolites throughout grey matter. The weak correlation between metabolite concentrations and myelin water fractions could be studied further in clinical cases. It would be interesting to observe the change in myelin water fraction in cases such as reversible N A A changes in M S lesions or choline increases during demyelination.  One of the goals of this study was to learn more about metabolite  concentrations by correlating them with other quantitative M R I measures. Results with myelin water fraction were inconclusive but perhaps future studies linking metabolite concentrations to quantities such as water content, M T R or diffusion will provide more information.  104  Bibliography  1. K.M.Van De Graaf. Human Anatomy. McGraw-Hill, New York, 6th Edition (2002). 2. D.L.Birken and W.H.Oldendorf. N-acetyl-L-aspartic acid: A literature review of a compound prominent in 1H-NMR spectroscopic studies of brain. Neuroscience and Behavioural Review. 13:23-31 (1989). 3. M.L.Simmons, C.G.Frondoza, and J.T.Coyle. Immunocytochemical localization of N-acetyl-aspartate with monoclonal antibodies. Neuroscience. 45:37-45 (1991). 4. J.V.Nadler and J.R.Cooper. N-acetyl-L-aspartic acid content of human neural tumours and bovine peripheral nervous tissues. Journal of Neurochemistry. 19:313-319(1972). 5. K.J.Koller, R.Zaczek, and J.T.Coyle. N-acetyl-aspartyl-glutamate: Regional levels in rat brain and the effects of brain lesions as determined by a new H P L C method. Journal of Neurochemistry. 43:1136-1142 (1984). 6. G.Tsai and J.T.Coyle. N-acetylaspartate in neuropsychiatric disorders. Progress in Neurobiology. 46:531-540 (1995). 7. R.E.Jung, W.M.Brooks, R.A.Yeo, S.J.Chiulli, D.C.Weers, and W.L.Jr Sibbitt. Biochemical markers of intelligence: A proton M R spectroscopy study of normal human brain. Proceedings of Royal Society of London. Series B: Biological Sciences. 266:1375-1379 (1999). 8. J.Urenjak, S.R.Williams, D.G.Gadian, and M.Noble. Specific expression of N acetyl aspartate in neurons, oligodendrocyte-type-2 astrocyte progenitors and immature oligodendrocytes in vitro. Journal of Neurochemistry. 59:55-61 (1992). 9. A.P.Burlina, F.Ferrari, L.Facci, S.D.Skaper, and A.B.Burlina. Mast cells contain large quantities of secretagogue-sensitive N-acetylaspartate. Journal of Neurochemistry. 69:1314-1317 (1997). 10. K.K.Bhakoo and D.Pearce. In Vitro expression of N-acetyl aspartate by oligodendrocytes: implication for proton magnetic resonance spectroscopy signal in vivo. Journal of Neurochemistry. 74:254-262 (2000). 11. R.Matalon, R.Kaul, J.Casanova, K.Michals, A.Johnson, E.Rapin, P.Gashkoff, and M.Deanching. SSIEM Award: Aspartoacylase deficiency: The enzyme defect in  105  Canavan's disease. Journal of Inherited Metabolic Disease. 12 (SuppL 2):329331 (1989). 12. P.Barker. N-acetyl aspartate: A neuronal marker? Annals of Neurology. 49:423424 (2001). 13. E.Martin, A.Capone, and J.Schneider. Absence of N-acetyl aspartate in the human brain: Impact on neurospectroscopy? Annals of Neurology. 49:518-521 (2001). 14. C.Bjartmar, J.Battistuta, N.Terada, E.Dupree, and B.D.Trapp. N-acetyl aspartate is an axon specific marker of mature white matter in vivo: A biochemical and immunohistochemical study on the rat optic nerve. Annals of Neurology. 51:51-58 (2002). 15. A.F.D'Adamo, L.I.Gidez, and F.M.Yatsu. Acetyl transport mechanisms. Involvement of N-acetyl aspartic acid in de novo fatty acid biosythesis in the developing rat brain. Experimental Brain Research. 5:267-273 (1968). 16. D.D.Clarke, S.Greenfield, E.Dicker, L.J.Tirri, and E.J.Ronan. A relationship of N acetyl aspartate biosynthesis to neuronal protein synthesis. Journal of Neurochemistry. 24:479-485 (1975). 17. D.R Curtis and J.C.Watkins. The excitation and depression of spinal neurons by structurally related amino acids. Journal of Neurochemistry. 6:117-141 (1960). 18. M.H.Baslow. Functions of N-acetyl-L-aspartate and N-acetyl-L-aspartyl glutamate in the vertebrae brain: role in glial cell signaling. Journal of Neurochemistry. 75:453-459 (2000). 19. G.Chakraborty, P.Mekala, D.Yahya, G.Wu, and R.W.Ledeen. Intraneuronal N acetylaspartate supplies acetyl groups for myelin lipid synthesis: evidence for myelin-associated aspartoacylase. Journal of Neurochemistry. 78:736-745 (2001). 20. M.H.Baslow. Evidence supporting a role for N-acetyl-L-aspartate as a molecular water pump in myelinated neurons in the central nervous system: A n analytical review. Neurochemistry International. 40:295-300 (2002). 21. D.Sappey-Marinier, G.Calabrese, H.P.Hetherington, S.N.G.Fisher, R.Deicken, C.V.Dyke, G.Fein, and M.W.Weiner. Proton magnetic resonance spectroscopy of human brain: applications to normal white matter, chronic infarction and M R I white matter signal hyperintensities. Magnetic Resonance in Medicine. 26:313327 (1992). 22. S.H.Graham, D.J.Meyerhoff, L.Bayne, F.R.Sharp, and M.W.Weiner. Magnetic resonance spectroscopy of N-acetylaspartate in hypoxic-ischemic encephalopathy. Annals of Neurology. 35: 490-494 (1994).  106  23. B.L.Miller, R.A.Moats, T.Shonk, T.Ernst, S.Woolley, and B.D.Ross. Alzeimer's disease: Depiction of increased cerebral myo-inositol with proton M R spectroscopy. Radiology. 37:1064-1067 (1993). 24. A.Shiino, M.Matsuda, S.Morikawa, T.Inubushe, I.Akiguchi, and J.Handa. Proton magnetic resonance spectroscopy with dementia. Surgical Neurology. 39:143-147 (1993). 25. D.L.Arnold, P.M.Matthews, F.Francis, and J.Antel. Proton magnetic resonance spectroscopy of human brain in vivo in the evaluation of multiple sclerosis: Assessment of the load of disease. Magnetic Resonance in Medicine. 14:154-159 (1990). 26. R.I.Grossman, R.E.Lenkinski, K.N.Ramer, F.Gozalez-Scarano, and J.A.Cohen. M R proton spectroscopy in multiple sclerosis. American Journal of Neuroradiology. 13:1535-1543 (1992). 27. A.L.Lehniger, D.L.Nelson, and M.M.Cox. Principles of Biochemistry. Worth Publishers, New York, 3rd Edition (2000). 28. O.Henriksen. M R S of Brain in Magnetic Resonance Spectroscopy in Biology and Medicine. J.D.De Certaines, W.M.M.J.Bovee, and F.Podo. Pergamon Press: New York, 411-435 (1992). 29. J.Frahm, H.Bruhm, M.L.Gyngell, K.D.Merboldt, W.Hanicke, and R.Sauter. Localized proton N M R spectroscopy in different regions of the human brain in vivo: relaxation times and concentrations of cerebral metabolites. Magnetic Resonance in Medicine. 11:47-63 (1989). 30. C.C.Hanstock, D.L.Rothman, J.W.Pritchard, T.Jue, and R.G.Shulman. Spatially localized 1H N M R spectra of metabolites in the human brain. Proceedings of National Academy of Science USA. 85:1821-1825 (1988). 31. T.Michaelis, J.K.Merboldt, H.Bruhm, W.Hanicke, and J.Frahm. Absolute concentrations of metabolites in the adult human brain in vivo: Quantification of localized proton M R spectra. Radiology. 187:219-227 (1993). 32. E.Brief. 1H Magnetic Resonance Spectroscopy of Human Brain: Relaxation and Concentrations. U B C , Department of Physics and Astronomy:Ph.D. (2000). 33. U.Flogel, T.Niendorf, N.Serkowa, A.Brand, J.Henke, and D.Liebfritz. Changes in organic solutes, volume, energy state, and metabolism associated with osmotic stress in a glial cell line: a multinuclear N M R study. Neurochemical Research. 20:793-802 (1995). 34. B.Ross and S.Bluml. Magnetic Resonance Spectroscopy of the Human Brain. The Anatomical Record (New. Anat.). 265:54-84 (2001).  107  35.  H.P.Hetherington, G.F.Mason, J.W.Pan, S.L.Ponder, J.T.Vaughan, D.B.Twieg, and G.M.Pohost. Evaluation of cerebral gray and white matter metabolite differences by spectroscopic imaging at 4. IT. Magnetic Resonance in Medicine. 32:565 -571 (1994).  36.  P.J.W.Pouwels and J.Frahm. Regional metabolite concentrations in human brain as determined by quantitative localized proton MRS. Magnetic Resonance in Medicine. 39:53-60 (1998).  37.  S.S.Gill, R.K.Small, D.G.Thomas, P.Patel, R.Porteous, N.van Bruggen, G.Gadian, R.A.Kauppinen, and S.R.Williams. Brain metabolites as 1H N M R markers of neuronal and glial disorders. NMR in Biomedicine. 2:196-200 (1993).  38.  C.A.Davie, C.P.Hawkins, G.J.Barker, A.Breenan, P.S.Tofts, D.H.Miller, and et al. Detection of myelin breakdown products by proton magnetic resonance spectroscopy. Lancet. 341:630-631 (1993).  39.  R.Kreis, B.D.Ross, N.A.Farrow, and Z.Ackerman. Metabolic disorders of the brain in chronic hepatic encephalopathy detected with H - l M R spectroscopy. Radiology. 182:19-27 (1992).  40.  A.L.MacKay, K.P.Whittall, J.Adler, D . K . B . L i , D.W.Paty, and D.Graeb. In vivo visualization of myelin water in brain by magnetic resonance. Magnetic Resonance in Medicine. 31:673-677 (1994).  41.  K.P.Whittall, A.L.MacKay, D.A.Graeb, R.A.Nugent, D . K . B . L i , and D.Paty. In vivo measurement of T2 distributions and water contents in normal human brain. Magnetic Resonance in Medicine. 37:34-43 (1997).  42.  R.M.Henkelman, G.J.Stanisz, J.K.Kim, and M.J.Bronskill. Anisotropy of N M R properties of tissues. Magnetic Resonance in Medicine. 32:592-601 (1994).  43.  R.S.Menon and P.S.Allen. Application of continuous relaxation time distributions to the fitting of data from model systems and excised tissue. Magnetic Resonance in Medicine. 20:214-227 (1991).  44.  R.Harrison, M.J.Bronskill, and R.M.Henkelman. Magnetization transfer and T2 relaxation components. Magnetic Resonance in Medicine. 33:490-496 (1995).  45.  B.K.Rutt and J.D.Mitchell. Myelin selective imaging with 3D fast spin echo MRI. Proceedings of Society of Magnetic Resonance in Medicine. 12: (1993).  46.  M . D .Does and R.E.Snyder. T2 relaxation of peripheral nerve measured in vivo. Magnetic Resonance Imaging. 16:67-79(1990).  47.  R.V.Mulkern, S.T.S.Wong, P.Jakab, A.R.Bleier, T.Sandor, and F.A.Jolesz. C P M G imaging sequences for high field in vivo transverse relaxation studies. Magnetic Resonance in Medicine. 16:67-79 (1990).  108  48.  P.Christiansen, R.B.Toft, P.Gideon, E.R.Danielson, P.Ring, and O.Henriksen. MR-visible water content in human brain: a proton MRS study. Magnetic Resonance Imaging. 12:1237-1244 (1994).  49.  J.-P.Armspach, D.Gounot, L.Rumbach, and J.Chambron. In vivo determination of multiexponential T2 relaxation in the brain of patients with multiple sclerosis. Magnetic Resonance Imaging. 9:107-113 (1991).  50.  W.A.Stewart, A.L.MacKay, K.P.Whittall, G.R.W.Moore, and D.W.Paty. Spinspin relaxation in experimental allergic encephalomyelitis. Analysis of C P M G data using a non-linear least squares methods and linear inverse theory. Magnetic Resonance in Medicine. 29:767-775 (1993).  51.  A.L.MacKay, C.Laule, B.Madler, I.M.Vavasour, and K.P.Whittall. Magnetic resonance distinguishes different water reservoirs in brain. Physics in Canada. March/April: 135-143 (2002).  52.  G.R.W.Moore and others. A pathology-MRI study of the short-T2 component in formalin-fixed multiple sclerosis brain. Neurology. 55:1506-1510 (2000).  53.  B.Madler and et al. Correlation of diffusion tensor anisotropy measures, T2 relaxation distributions and myelin water content in human white matter brain. Proceedings of International Society for Magnetic Resonance in Medicine. 9: (2001).  54.  V.Dousset, B.Brochet, A.Vital, C.Gross, A.Benazzouz, A.Boulleme, A.M.Bidabe, A . M . G i m , and J.M.Caille. Lysolecithin-induced demyelination in primates: Preliminary in vivo study with M R and magnetization transfer. American Journal of Neuroradiology. 16:225-231 (1995).  55.  C.Laule, I.M.Vavasour, D.W.Paty, D . K . B . L i , D.L.Arnold, and A.L.MacKay. Correlation between magnetizaation transfer ratio and myelin water content in normal white matter and MS lesions. Proceedings of International Society for Magnetic Resonance in Medicine. 10: (2002).  56.  S.M.Noworolski, S.J.Nelson, R.G.Henry, M.R.Henry, M.R.Day, L.L.Wald, J.Star-Lack, and D.B.Vigneron. High spatial resolution 'H-MRSI and segmented M R I of cortical grey matter and subcortical white matter in three regions of the human brain. Magnetic Resonance in Medicine. 41:21-29 (1999).  57.  B.J.Soher, P.C.M.van Zijl, J.H.Duyn, and P.B.Barker. Quantitative proton M R spectroscopic imaging of the human brain. Magnetic Resonance in Medicine. 35:356-363 (1996).  58.  G.Tedeschi, A.Bertolino, A.Righini, G.Campbell, R.Raman, J.H.Duyn, C.T.W.Moonen, J.R.Alger, and G.Di Chiro. Brain regional distribution pattern of metabolite signal intensities in young adults by proton magnetic resonance spectroscopic imaging. Neurology. 45:1384-1391 (1995). 109  59. N.Schuff, F.Ezekiel, A.C.Garnst, D.L.Amend, A.A.Capizzano, A.A.Maudsley, and M.W.Weiner. Region and tissue differences of metabolites in normally aged brain using multislice 1H magnetic resonance spectroscopic imaging. Magnetic Resonance in Medicine. 45:899-907 (2001). 60. D.Lopez-Villegas, H.Kimura, S.Tunlayadechanont, and R.E.Lenkinski. High spatial resolution M R I and proton M R S of human frontal cortex. NMR in Biomedicine. 9:297-304 (1996). 61. C.L.Florian, S.R.Williams, K.K.Bhakoo, and M.D.Nobel. Regional and developmental variations in metabolite concentration in the rat brain and eye: A study using 1H N M R spectroscopy and high performance liquid chromatography. Neurochemical Research. 21:1065-1074(1996). 62. D.Weidermann, N.Schuff, G.B.Matson, B.J.Soher, A.T.Du, A.A.Maudsley, and M.W.Weiner. Short echo time multislice proton magnetic resonance spectroscopic imaging in human brain: metabolite distributions and reliability. Magnetic Resonance Imaging. 19:1073-1080(2001). 63. PJ.W.Pouwels, J.Brockmann, B.Kruse, B.Wilken, M.Wick, F.Hanefeld, and J.Frahm. Regional age dependence of human brain metabolites from infancy to adulthood as detected by quantitative localized proton M R S . Pediatric Research. 46:474-485 (1999). 64. M.A.McLean, F.G.Woermann, GJ.Barker, and J.S.Duncan. Quantitative analysis of short echo time 1H-MRSI of cerebral grey and white matter. Magnetic Resonance in Medicine. 44:401-411 (2000). 65. O.A.C.Petroff, D.D.Spencer, J.R.Alger, and J.W.Pritchard. High field proton magnetic resonance spectroscopy of human cerebrum obtained during surgery for epilepsy. Neurology. 39:1197-1202 (1989). 66. T.J.Doyle, B.J.Bedell, and P.A.Narayana. Relative concentrations of proton M R visible neurochemicals in grey and white matter in human brain. Magnetic Resonance in Medicine. 33:755-759 (1995). 67. K.O.Lim and D.M.Spielman. Estimating N A A in cortical grey matter with applications for measuring changes due to aging. Magnetic Resonance in Medicine. 37:372-377 (1997). 68. P.Christiansen, P.Toft, H.B.W.Larsson, M.Stubgaard, and O.Henriksen. The concentration of N-acetyl aspartate, creatine + phospocreatine and choline in different parts of the brain in adulthood and senium. Magnetic Resonance Imaging. 11:799-806(1993). 69. J.Hennig, H.Pfister, T.Ernst, and D.Ott. Direct absolute quantification of metabolites in the human brain with in vivo localized proton spectroscopy. NMR Biomed. 5:193-199(1992).  110  70.  G.Tedeschi, A.Righini, A.Bizzi, A.S.Barnett, and J.R Alger. Cerebral white matter in the centrum semiovale exhibits a larger N-acetyl signal than does gray matter in long echo time H magnetic resonance spectroscopic imaging. Magnetic Resonance in Medicine. 33:127-133 (1995). !  71.  D.T.Chard, M.A.McLean, C.M.B.Griffm, A.J.Thompson, and D.H.Miller. Proton magnetic resonance spectroscopic findings in normal appearing grey and white matter in early relapsing remitting multiple sclerosis. Proceedings of International Society for Magnetic Resonance in Medicine. 10: (2002).  72.  L.Rogers, N.Schuff, and M.W.Weiner. Gender effects on N A A in brain grey and white matter revealed by mixed effects statistical analysis. Proceedings of International Society for Magnetic Resonance in Medicine. 10: (2002).  73.  Z.Caramanos and others. Axonal integrity and fractional anisotropy in the normal appearing white matter of patients with multiple sclerosis: relations to cerebrofunctional reorganization and clinical disability. Proceedings of International Society for Magnetic Resonance in Medicine. 10: (2002).  74.  W.Huang and others. Correlation studies of multiple sclerosis using 1H M R S , volumetric M R I and cognitive tests. Proceedings of International Society for Magnetic Resonance in Medicine. 10: (2002).  75.  M.Wylezinska, A.Cifelli, P.Matthews, J.Palace, and P.Jezzard. Neuronal damage in thalamic grey matter in relapsing-remitting multiple sclerosis. Proceedings of International Society for Magnetic Resonance in Medicine. 10: (2002).  76.  I.I.Rabi, J.R.Zacharia, S.Millman, and P.Kusch. A new method of measuring nuclear magnetic moment (letter). Physics Review. 53: (1938).  77.  C.P.Slichter. Principles of Magnetic Resonance. Springer-Verlag, New York, 3rd Edition (1996).  78.  E.L.Hahn. Spin Echos. Physics Review. 80:580-584 (1950).  79.  H . Y.Carr and E.M.Purcell. Effects of diffusion on free precession in nuclear magnetic resonance experiments. Physics Review. 94:630-638 (1954).  80.  S.Meiboom and D.Gill. Modified spin echo method for measuring relaxaion times. Review of Scientific Instruments. 29:688-691 (1958).  81.  P.C.M.van Zijl. Overview of water suppression. MR Spectroscopy Course, Proceedings of Internations Society Magnetic Resonance in Medicine. 10:119125 (2002).  82.  A.Haase, J.Frahm, W.Hanicke, and D.Matthaei. 1H N M R chemical shift selective (CHESS) imaging. Physics in Medicine and Biology. 30:341-344 (1985).  1.11  83.  A.A.Maudsley. Spectroscopic imaging. MR Spectroscopy Course, Proceedings of International Society Magnetic Resonance in Medicine. 10:17-23 (2002).  84.  E.Schneider and G.Glover. Rapid in vivo shimming. Magnetic Resonance in Medicine. 20:113-122(1991).  85.  D.Spielman. Magnetic field homogeneity and shimming. MR Spectroscopy Course, Proceedings of International Society Magnetic Resonance in Medicine. 10:24-29 (2002).  86.  S.Provencher. Estimation of metabolite concentrations from localized in vivo proton N M R spectra. Magnetic Resonance in Medicine. 30:672-679 (1993).  87.  S.W.Provencher. LCModel and LCMgui user's manual (2000).  88.  E.Brief, K.P.Whittall, D . K . B . L i , and A.L.MacKay. Metabolite TI relaxation differs between and within regions in normal human brain. Proceedings of International Society for Magnetic Resonance in Medicine. 8: (2000).  89.  E.Brief, K.P.Whittall, D . K . B . L i , and A.L.MacKay. In vivo normal brain metabolite T2 with large TE range. ESMRMB. (2000).  90.  E.Brief, K.P.Whittall, A.L.MacKay, and D.K.B.Li. Precision of M R S metabolite peak area ratios in human brain at short TE. Proceedings of International Society for Magnetic Resonance in Medicine. 7: (1999).  91.  K.P.Whittall and A.L.MacKay. Quantitative interpretation of N M R relaxation data. Journal of Magnetic Resonance Imaging. 84:134-152 (1989).  92.  M . Skinner. Effect of variable echo spacing in multi-echo sequences for resolving long T2 components in multiple sclerosis. U B C , Department of Physics and Astronomy:M.Sc. (2001).  93.  C.Laule. Optimization of a 48 echo MRI sequence using variable TR data acquisition. U B C , Department of Physics and Astronomy:M.Sc. (2001).  94.  R.E.Walpole. Introduction to Statistics. MacMillan Publishing Co. Inc., New York, 2nd Edition (1974).  95.  J.W.Pan, D.B.Twieg, and H.P.Hetherington. Quantitative spectroscopic imaging of the human brain. Magnetic Resonance in Medicine. 40:363-369 (1998).  96.  D.T.Chard, M.A.McLean, G.J.M.Barker, D.G.MacManus, and D.H.Miller. Reproducibility of in vivo metabolite quantification with proton magnetic resonance spectroscopic imaging. Journal of Magnetic Resonance Imaging. 15:219-225 (2002).  112  

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}]}"
                            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:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0090620/manifest

Comment

Related Items