You may notice some images loading slow across the Open Collections website. Thank you for your patience as we rebuild the cache to make images load faster.

Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Frequency shift mapping in spinal cord models of white matter demyelination Chen, Evan I-Wen 2016

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

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

Item Metadata


24-ubc_2016_may_chen_evaniwen.pdf [ 28.59MB ]
JSON: 24-1.0300307.json
JSON-LD: 24-1.0300307-ld.json
RDF/XML (Pretty): 24-1.0300307-rdf.xml
RDF/JSON: 24-1.0300307-rdf.json
Turtle: 24-1.0300307-turtle.txt
N-Triples: 24-1.0300307-rdf-ntriples.txt
Original Record: 24-1.0300307-source.json
Full Text

Full Text

Frequency shift mapping in spinal cordmodels of white matter demyelinationbyEvan I-Wen ChenB.Sc., The University of British Columbia, 2012A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinThe Faculty of Graduate and Postdoctoral Studies(Physics)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)April 2016c© Evan I-Wen Chen 2016AbstractThe behavior of MR phase and frequency in demyelination and damage in central nervous tissuewhite matter arises not only from traditionally associated bulk susceptibility changes, but alsofrom changes to its tissue microstructure. A recently proposed generalized Lorentzian model ofmicrostructure-related magnetic susceptibility effects predicts an increase in MR frequency dueto damage in myelin in MS lesions. The same model also predicts reduction in MR frequencydue to axonal degeneration. Here, we investigate the effect of both myelin and axonal damagethrough transection of white matter fibers in the dorsal column of rat cervical spinal cord. Thisinjury generates secondary damage consisting of neurodegeneration along nerve tracts bilateralto the transection site, producing cases of Wallerian and retrograde degeneration free of excessivehemorrhage and inflammation. High-resolution frequency maps of degenerating tracts were cor-related with histopathology for axons, myelin, degenerated myelin, and macrophages. Damageto myelin sheaths is prominent in Wallerian degeneration, where we observe strong correlationswith increasing frequency up to 8 weeks post-injury. Retrograde degeneration, which consists pre-dominantly of axonal damage, produces decreased frequency shift over time. The MR frequencyshifts are sensitive to the effects of macrophage infiltration and debris clearance, which vary withwhite matter fiber density and affect rates of degeneration. We demonstrate how MR frequencycan successfully characterize injury in rat spinal cord white matter in a manner consistent withpredictions outlined by the Generalized Lorentzian Approximation Model, and conclude thatthese results suggest potential applications of MR frequency to supplement or replace currentclinical techniques, such as myelin water and diffusion weighted imaging, as a non-invasive andquantitative method of assessing white matter damage in CNS.iiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi1 Introduction and Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Introduction and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 The Spinal Cord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2.1 Spinal Cord and Injury . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2.2 Myelin and Axons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.3 Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.3.1 NMR Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.3.2 Non-Uniform Magnetic Fields: Field Gradients . . . . . . . . . . . . . . . . 71.3.3 Magnetic Susceptibility and Tissue Orientation . . . . . . . . . . . . . . . 71.3.4 Basic Components of MRI and Pulse Sequences . . . . . . . . . . . . . . . 91.4 MR Imaging Techniques Used to Characterize WM . . . . . . . . . . . . . . . . . 101.4.1 Myelin Water Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101.4.2 Diffusion Tensor Imaging (DTI) . . . . . . . . . . . . . . . . . . . . . . . . 131.4.3 Phase Imaging, Magnetic Susceptibility, and Tissue Orientation . . . . . . 152 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172.1 Experiment: Rat DC-Tx Model and Comparison to Histology . . . . . . . . . . . 172.2 Rat Spinal Cord Injury Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.2.1 Injury Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.2.2 Perfusion and Spinal Cord Extraction . . . . . . . . . . . . . . . . . . . . . 182.3 MR Apparatus, Scan Parameters, and Data Analysis Methods . . . . . . . . . . . 192.3.1 Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.3.2 Scan Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202.3.3 Registration and ROIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21iiiTable of Contents3 Results and Discussion: Frequency Shifts of Myelin Changes in DC-Tx Pathol-ogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.1 Dorsal Column: Anatomy and Pathology . . . . . . . . . . . . . . . . . . . . . . . 223.2 Evaluating Dorsal Column Transection Model in Rat Spinal Cord . . . . . . . . . 233.2.1 Healthy Cord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2.2 Cranial to Injury Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2.3 Caudal to Injury Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.3.1 Frequency Shifts in Retrograde Degeneration . . . . . . . . . . . . . . . . . 283.3.2 Frequency Shifts in Wallerian Degeneration . . . . . . . . . . . . . . . . . . 283.3.3 Characterization of White Matter Damage in Spinal Cord . . . . . . . . . 293.3.4 Histological Correlates of White Matter Frequency Shifts . . . . . . . . . . 293.4 Frequency Changes in DC-Tx Evaluated Along the Spinal Cord . . . . . . . . . . 313.5 Applications to Assessment of Spinal Cord Injury and Repair and Future Work . 33Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36ivList of Tables3.1 WM tract orientations in the dorsal column. Tract degeneration patterns relativeto cranial-caudal direction from injury. As an ascending tract, the FG experiencesWallerian in the cranial direction, while the descending CST experiences Walleriandegeneration caudally. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.2 Multiple regression analysis of histology parameters, generating models for fre-quency shifts in Wallerian (R2 = 0.839) and Retrograde (R2 = 0.536) mechanisms,with predictor importance listed for each histology parameter. . . . . . . . . . . . . 29vList of Figures1.1 Lateral view of a human spinal cord with labelled vertebral regions. Damage inthe spinal cord at level C7 and above most likely results in Quadriplegia. BelowC7, paraplegia is likely, with varying degrees based on location of injury. Injuriesin the thoracic region include paralysis above the waist, whereas injuries below L2in the lumbar region will most likely not. Adapted from Gray’s Anatomy, 1918[32] 31.2 Wallerian degeneration following axon transection. Degeneration of the axon oc-curs at the injury site, travelling distally away from the neuron cell body. Macrophagespresent after acute inflammation help clear the myelin debris over time. In thespinal cord, Wallerian degeneration will occur in the direction of axonal travel,either ascending or descending. Adapted from Wang et al., 2012 [86] . . . . . . . . 41.3 Illustration of spin orientation in an external magnetic field. Spins oriented inan external magnetic field B will align in spin-up or spin-down states, with anet magnetization vector M0. Excitation RF at the larmor frequency ω0 tips thisvector in the transverse direction, resulting in the precession pattern shown aroundB0. Adapted from Xiang, Q. 2004 . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.4 Zeeman energy splitting. For protons with spin-up and -down states, E+ andE− energy states result from interactions with external magnetic field B, with anenergy difference γh¯B. Adapted from Xiang, Q. 2004 . . . . . . . . . . . . . . . . . 61.5 Slice selection using gradients. Linear field gradients applied in the Z directioncombined with nonchanging fields in the X and Y direction affects resonance fre-quencies only along Z. By applying pulses specific for a certain frequency, ’slices’in the XY plane can be selected. In patients, field gradients in X, Y, or Z allowvisualization of body tissues at almost any orientation. Adapted from Xiang, Q.2004 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.6 Transverse (MT ) decay and longitudinal (ML) recovery of a tipped magnetizationvector M(t) as it relaxes longitudinally in parallel with the main magnetic field(B0). Adapted from Xiang, Q. 2004 . . . . . . . . . . . . . . . . . . . . . . . . . . 81.7 Components of myelin bilayer and intra/extracellular water. Signal amplitudesfor corresponding T2 relaxation times, with peaks at: 1) T2 < 30ms related towater trapped between myelin sheath bilayers, 2) between 40ms < T2 < 100msrepresenting water inside and outside the axon, and 3) from water in cerebrospinalfluid (CSF), at T2 > 1000ms. CSF peak not shown. Adapted from Laule et al. [41] 11viList of Figures1.8 Vector diagram of CPMG. At t=0, 90 ◦x’ tips the magnetization vector into thetransverse plane. Following a dephasing period τ , a 180 ◦y’ pulse is applied to flipthe spins, reversing their direction. The spins rephase after τ , and the sequencecan be applied again for multiples echoes. Note that the the signal shown on thebottom is after 2τ , where subsequent echoes are observed at 2nτ intervals. Theoverall signal decreases over increasing TE, producing a T2 decay curve. Adaptedfrom Xiang, Q. 2004 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121.9 Illustration of the diffusion tensor. A) Diffusion in isotropic media is expected to beequal in all dimensions, with the expected root mean squared displacement forminga diffusion sphere. B) In anisotropic media, bias in longitudinal (λ1) diffusion isreflected in a more ellipsoid shape. . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.1 Preparation of rat spinal cord sections for MRI scanning. Extracted cords (left)are loaded in plastic cylindrical tubes with PF into the solenoid transmit/receivecoil (center) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.1 Cross-sectional slice of cervical rat spinal cord showing WM/GM contrast andcorresponding anatomical features. (A) Transverse magnitude slice, with ROIs ofthe Fasciculus Gracilis (orange) and Corticospinal Tract (green). Areas of interestdenoted: Dorsal/Ventral Horns (DH, VH), Central Canal (CC), Ventral MedianFissure (VMF), Ventral White Matter (VWM), and Lateral White Matter (LWM).Gray matter lamiae transition (Tr) into WM, from the DH to LWM (B) Sameregions in a MR frequency map of the corresponding PF fixed section, with graymatter appearing brighter than white matter. Scale bar = 1mm. Frequency scaledto ± 1.446 ppb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.2 Electron microscope micrographs of degenerating WM. Retrograde degenerationin the FG (A, B) show myelin sheaths loosening over time, maintaining concentricorganization around the axons. Internal axon structure is lost in many axonsby 8-weeks post injury. Wallerian degeneration in the CST (C) produces acuteaxonal loss and moderate demyelination. Similarly in the FG at 8-wpi (D), axondegeneration forms cavitations, with myelin forming debris fragments. Evidenceof remyelination appears at 8-wpi with sparsely myelinated axons (see red star).6300x magnification, scale bar = 5um. . . . . . . . . . . . . . . . . . . . . . . . . . 24viiList of Figures3.3 Enlarged images of dorsal column, both cranial and caudal to injury. The cranialimages show (1) healthy, (2) 3-week post injury, and (3) 8-week post injury cords+5mm from injury. (A) Frequency contrast in FG and CST produced by degen-eration, with brighter gray matter remaining unchanged. (B) Eriochrome stainingof myelin phospholipid content with brighter areas denoting increased staining dueto increased surface area. (C) Immunohistochemistry for major neuronal compo-nents: neurofilament/tubulin III (NF/TubIII, blue, marker of axonal content), anddegen-MBP (dMBP, red, marker of degenerated myelin content). (D) Immunohis-tochemistry for ED-1 macrophage marker. Corresponding caudal sections in (4),(5), and (6). Both the FG and CST shrink physically as damage and tissue re-moval progresses over time: caudally for the CST, and cranially for the FG. Thecharacteristic V-shape Post-synaptic Dorsal Column Pathway that borders the FGis most prominent in D5 and D6 (red arrows), and observable in frequency mapsA2, A3, A5, A6. Minimal tearing (tg) due to tissue preparation is mostly limitedto GM. Histology at 100x magnification, scale bar = 250µm. Frequency scaled to± 1.446 ppb. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.4 Frequency shifts in paraformaledyde-fixed control WM (ventrolateral) and GM,averaged over all subjects in each treatment, showing no significant variabilityacross subjects over time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.5 Comprehensive schematic of DC-Tx in rat spinal cord. Central figure depicts WMdegeneration over time divided by tract (FG, CST) and mechanism (Wallerian,retrograde), with general visual representation of myelin and axon damage basedon histology and EM micrographs. Frequency for each treatment plotted with er-ror bars, with significances between time-points for all histology measures markedabove/below. In Wallerian degeneration, neuronal damage is extensive, showingsignificant loss in myelin and axonal content over time, matched with increasedpresence of macrophages and myelin fragments. Remaining tissue is mostly debrisby 8-wpi, with relatively more damage in the FG. At 3-wpi and later, frequencyshifts show a significantly increasing trend compared to healthy baseline. Ret-rograde degeneration is a relatively slower degeneration pattern characterized bychanges in eriochrome and axon staining. ED-1 and degen-MBP differences arepresent in CST only. Frequency shifts trend negatively in retrograde degeneration,reaching a significantly negative frequency shift (-5 to -6 * 10−3 ppm). Differ-ences in tract anatomy are a contributing factor affecting injury pathology andsubsequent immune response in the FG and CST, as reflected in the central diagram. 30viiiList of Figures3.6 Frequency to histology correlations of degeneration. Wallerian degeneration (Cra-nial FG, Caudal CST) and retrograde degeneration (Caudal CST, Cranial FG)were stained with eriochrome stain of myelin content (A, D), immunohistochem-istry stain of Neurofilament/TubulinIII axonal content (B, E), and degen-MBP fordegenerated myelin content (C, F). Tracts denoted by color (Orange FG, GreenCST) and divided into healthy (circle), 3-wpi (triangle), and 8-wpi (square) time-points. Data sets encircled in groups to improve visualization. Correlations ofeach histology measure to frequency shifts shown in dotted lines with correspond-ing R2 values. Frequency shifts in Wallerian degeneration correlate strongly witheriochrome in FG and CST, and with degen-MBP in FG. Retrograde correlationssignificant with NF/TubIII axon stain in FG and CST, with eriochrome in FG,and with degen-MBP in CST. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.7 Frequency maps of rat spinal cord dorsal column cranial (A, B) and caudal (D, E)to injury site (C). Distance from injury labelled above. Fasciculus Gracilis markedin orange, Corticospinal Tract marked in green. . . . . . . . . . . . . . . . . . . . . 323.8 Frequency shifts values bilateral to injury. Frequency shift values shown for cra-nial (+8mm) to caudal (-8mm) along healthy (black), 3-weeks (dark), and 8-weeks(light) post-injury spinal cords for the Fasciculus Gracilis (A, orange) and Corti-cospinal Tract (B, green). Injury site located at 0mm (red highlight), with severelydamaged adjacent slices (yellow highlight). Areas of Wallerian and retrograde de-generation for each tract are labelled. . . . . . . . . . . . . . . . . . . . . . . . . . 32ixAcknowledgementsMuch thanks and appreciation goes to Dr. Alexander Rauscher and Dr. Piotr Kozlowski forsupervising and directing the development of this project. In addition, support from the staffand researchers at the UBC MRI Research Center has been instrumental to this thesis: HenryChen and Andrew Yung at the 7T, for MATLAB tips, coil construction, and MRI scanningprotocols; Vanessa Wiggermann, Edenino Hernandez, Jojo Yuen, and Jachin Hung for MRIand myelin background and theory; Jie Liu, Yuan Jiang, and Dr. Wolfram Tetzlaff at ICORD,Vancouver Coastal Health, for providing animal work, histology, and direction. This project andits continued work is a joint effort of highly skilled researchers and technicians in their respectivefields, and their contributions have significantly helped my understanding of our project, and thefuture directions it leads to. Thanks also go out to colleagues and friends who have supportedand helped in this endeavour.xDedicationThis is dedicated to my family, who are supportive of my interests in research, and tolerant ofsome of my late nights in the MRI lab. It has allowed me to expand my interests and learn severalcritical skills, much of which would not have been possible without their support.xiChapter 1Introduction and Theory1.1 Introduction and MotivationThe spinal cord communicates between the brain and peripheral nervous system (PNS), facili-tating almost all functions in the body. The senses, motor control, pain, and reflexes all rely onproper signal conduction in the cord. Not only does injury in the spinal cord affect these func-tions, the level (along the spinal cord; cervical, thoracic, lumbar, sacral) of neurological injurycorrelates with potential significant loss in sensory and motor function of the peripheral body,often in the form of paraplegia or quadriplegia (Fig. 1.1). Damage as a result of spinal cordinjury (SCI) is accompanied by secondary mechanisms that further the functional and physio-logical impact of injury. Recovery and treatment of SCI involves addressing acute and long-termsecondary effects following the primary mechanical injury, which include inflammation, injury-induced programmed neuron death (apoptosis), increased toxicity in the tissue environment, anddemyelination [5][60]. The biochemical environment accompanying the mechanism of interactionbetween neural death and regeneration post-injury is not currently fully understood, and whilemost patients with SCI do experience partial recovery of function, a complete cure does not cur-rently exist. While the acute effects are drastic, many SCI patients must also adapt to long-termchanges in lifestyle, rehabilitation, and overall health. With over 2.5 million current cases ofSCI around the world, significant research has investigated methods of treating and eventuallyrepairing spinal cord damage [73].A large part of recovery and treatment of SCI involves the acute and long-term secondaryeffects of the injury following the primary mechanical injury, which include inflammation, injury-induced programmed neuron death (apoptosis), increased toxicity in the tissue environment,and demyelination [60]. While the complex mechanism of interaction between neural death andregeneration post-injury is not fully understood, most patients with SCI do experience partialrecovery of function, though no complete cure currently exists.One fundamental component of SCI characterization that has been difficult to assess is theneuronal changes in recovering cords over time, due to lack of effective non-invasive methods avail-able for longitudinal assessment in-vivo. As a major component of the central nervous system(CNS), myelin plays an integral role in maintaining regular synaptic transmission. Demyelina-tion, or myelin degeneration, is caused by the damage and loss of myelin sheaths in neurons ofthe nervous system, leading to impairment of neural activity regulating movement, propriocep-tion (relative positioning), cognition, and memory, some of which are primary characteristics ofdiseases such as multiple sclerosis (MS), Alzheimer’s disease, and Huntington’s disease [20] in thebrain.The application of magnetic resonance (MR) techniques to the study these demyelinationdisease in white matter (WM) of the CNS [17][66][67] have been conducted in conjunction withthe monitoring of changes in cord tissue over time. Using SCI in rats as a model, different imagingtechniques such as diffusion tensor imaging (DTI), myelin water imaging (MWI), and Gradient11.2. The Spinal CordEcho imaging (GRE) can be used to measure different properties of water in and around thecord. Investigations in WM phase from GRE suggest the method is capable of distinguishingbetween tissue of similar magnetic susceptibilities based on micro-structural differences, a featurenot reflected in traditionally used magnitude images [17][90][92].One goal of the proposed experiments will apply the modifications of the generalized Lorentzianmodel used to predict frequency shifts in phase based on microstructure effects in tissues of thespinal cord. Organized bundles of WM tracts such as the fasciculus gracilis, fasciculus cuneatus,corticospinal tract are the main focus. Evaluation of dorsal column (DC) transection injury in arat model ex vivo with phase imaging will be compared to previously observed results which useddiffusion tensor and myelin water imaging [37], and will lay the groundwork for more quantitativeanalysis methods involving in vivo CNS studies in rats, and eventually human applications.We show that significant contribution from tissue microstructure and orientation is a majorcause of changes in magnetic frequency observed in tissues of the spinal cord by applying He andYablonskiy’s [35] modifications of the Lorentzian model. This will be a major step in establishingthe applications of phase imaging in clinical MRI protocols, and will further our understanding ofphase shifts and changes in tissue magnetic susceptibility caused by degenerative myelin diseases,axon degeneration, and tissue microstructure.1.2 The Spinal Cord1.2.1 Spinal Cord and InjuryIn SCI, the complexity of cellular and subcellular interactions in healthy and damaged cordhave been extensively studied (biochemically), exploring the effects of post-injury hemorrhaging,release of neurotoxic factors, and inflammation [5]. Most injuries are primarily trauma-related,caused by blunt force or axonal transection. Because imaging a moderate or large population ofhuman SCI patients is both difficult and costly, research is initially conducted in animal models,such as rats, before moving to clinical studies in human patients.Inflammation [5] involves complex short- and long- term interactions between immune cellsrecruited to the injury site, including monocytes, macrophages, and neutrophils. However, thepresence of inflammatory cytotoxic factors associated with these cells contribute to axonal degen-eration. As a result, axons severed at the point of injury atrophy, leaving neural and myelin debristo be cleared over time by macrophages. Activated macrophages persist at the injury site over aperiod of several weeks, secreting neuroregenerative factors that facilitate axonal regrowth[30][63].To study the effects of myelin and axon damage over time, we focused on transection injuriesusing a rat spinal cord model, specifically in the dorsal column white matter. While gray matteris isotropically distributed in the inner cord, the white matter neurons surrounding it in the outercord have long axons linearly oriented along the length of the overall cord. The dorsal columntransection (DC Tx) creates an injury site by severing axons only in the dorsal column WM.Severed axons in the spinal cord WM are unable to propagate signals, terminating the inter-action between the CNS and PNS. Depending on the tracts that are transected, the interactionlost could be cranial/ascending (from the PNS to CNS), caudal/descending (from the CNS toPNS), or both, and is often accompanied by significant loss in sensory and/or motor function.Aggregate bundles of axons sharing similar directionality are anatomically referred to as ’tracts’.Post-transection, axonal degeneration is not fully observed for 24 to 48 hours in rats [48], after21.2. The Spinal CordFigure 1.1: Lateral view of a human spinal cord with labelled vertebral regions. Damage in thespinal cord at level C7 and above most likely results in Quadriplegia. Below C7, paraplegia islikely, with varying degrees based on location of injury. Injuries in the thoracic region includeparalysis above the waist, whereas injuries below L2 in the lumbar region will most likely not.Adapted from Gray’s Anatomy, 1918[32]31.2. The Spinal CordFigure 1.2: Wallerian degeneration following axon transection. Degeneration of the axon occursat the injury site, travelling distally away from the neuron cell body. Macrophages presentafter acute inflammation help clear the myelin debris over time. In the spinal cord, Walleriandegeneration will occur in the direction of axonal travel, either ascending or descending. Adaptedfrom Wang et al., 2012 [86].which the cytoskeleton begins to dissociate and form debris [4]. This phenomenon, Walleriandegeneration 1.2, is followed by disintegration of the myelin sheath, where loss of axon-secretedsurvival signals cause oligodendrocytes to undergo apoptosis [2]. As mentioned, clearance of thisdebris occurs over several weeks post-injury.1.2.2 Myelin and AxonsThe CNS consists heavily of neurons making up white and gray matter (WM, GM) in the brainand spinal cord, with WM appearing lighter due to prominent myelination. WM tracts con-sists of fiber pathways in bundles traveling between similar origins and destinations, resulting inthe anisotropic orientation of fibers compared to GM in the cord and brain, which is relativelyisotropic. GM nerves, on the other hand, relatively sparsely myelinated and are mostly isotrop-ically oriented in the spinal cord. By understanding this anatomy, MRI can focus on producingWM and GM contrast in several different ways.Myelin, produced by oligodendrocyte cells in the CNS, is found in the brain and spinal cordin the form of multi-layered sheaths wrapped around the axons of neurons. The main mecha-nism of myelination involves oligodendrocytes tightly wrapping continuously around an axon toincrease overall axonal membrane resistance and decrease the capacitance, both critical elementsin increasing propagation speed of neural signals. Myelin sheath layers (12 nm) are much thinnercompared to their corresponding axon diameters (approximately 15µm on average). Around halfof total WM is myelin, which consists of high lipid content and contributes structurally to overallWM anisotropy [1][18][35].Axons are the channels with which neural impulses are conducted. They support myelinsheaths and traverse the entire length of the spinal cord, connecting the peripheral nervous system41.3. Magnetic Resonance Imagingwith the central. Like a bridge or highway, their function and health come from maintaining theirstructural integrity. For example, if a spinal cord injury severs some axons, the axons atrophy andbegin degrading after 24 to 48 hours [4][48]. This phenomenon, called Wallerian degeneration (Fig.1.2), is followed by disintegration of the myelin sheath[2] into isotropic bundles. Clearance of thisdebris occurs over several weeks post-injury. A similar but distinct pathology also occurs in theother direction along the axon (Retrograde degeneration)[72]. By understanding this anatomy,certain disease and injury pathology can be studied by focusing on their effects on local tissuemicrostructure.1.3 Magnetic Resonance ImagingMagnetic resonance imaging has become an important tool due to its ability to image internalbiological tissue non-invasively, without the risks associated with radioisotopes or x-rays. In recentdecades, it has been used to look at a wide variety of diseases and conditions throughout the body,with significant discoveries being made in groups concentrating on characterizing neurologicaldiseases in the brain and spinal cord.In essence, MRI relies on radiofrequency (RF) pulses and magnetic fields to measure the con-trast between signals received from different tissues in the body. Manipulating these factors hasresulted in a wide field of available MRI techniques including (but not limited to) diffusion MRI,T2-weighted MRI, and myelin water imaging. Overall, 3 types of magnetic fields are necessary:Weak magnetica field oscillating at radiofrequency, magnetic field gradienta switched rapidly, anda strong static magnetic field.1.3.1 NMR TheoryNuclear Magnetic Resonance (NMR) describes the phenomenon of energy absorption and emissionin magnetic nuclei. The widely measured nuclei in MRI are H1 protons, found in water andlipids of the body’s biological tissues. Protons possess an intrinsic magnetic moment (m), whichinteracts with external magnetic fields B0 (in the z direction), with energy defined as:E = −m ·B = −mzB. (1.1)The spin quantum number (s = 12) of the proton in the z-component is limited to two states,with the resulting energies E+ and E− of protons oriented in magnetic field B, referred to asspin-up and spin-down states. The phenomenon, Zeeman energy splitting 1.4, is a quantummechanical effect fundamental to the function of NMR and MRI.P (E−) = CeE−kT (1.2)P (E+) = Ce−E+kT (1.3)P (E−)P (E+)= e∆EkT (1.4)Division of protons into the two spin states depends on temperature and magnetic field asdescribed in the Boltzmann distribution, which can be defined as P (E+) and P (E−) the prob-ability of spin-up or spin-down states (1.2) (1.3), with constant C, Boltzmann constant k, andtemperature T. The ratio of both probabilities (1.4) can be approximated as 1 + ∆EkT .51.3. Magnetic Resonance ImagingFigure 1.3: Illustration of spin orientation in an external magnetic field. Spins oriented in anexternal magnetic field B will align in spin-up or spin-down states, with a net magnetizationvector M0. Excitation RF at the larmor frequency ω0 tips this vector in the transverse direction,resulting in the precession pattern shown around B0. Adapted from Xiang, Q. 2004Figure 1.4: Zeeman energy splitting. For protons with spin-up and -down states, E+ and E−energy states result from interactions with external magnetic field B, with an energy differenceγh¯B. Adapted from Xiang, Q. 200461.3. Magnetic Resonance ImagingUsing a 3T scanner at room temperature (298K), this ratio is 1.000005, with small excess ofthe energetically favourable spin-up state, producing a net magnetic moment M0 in the directionof B0, oriented along the longitudinal axis. Signal from longitudinally oriented spins is almostimpossible to measure, as the magnitude of M0 is usually on the scale of µTeslas, and oftenweaker than earth’s magnetic field (up to 60µT). Initially, transverse magnetization is randomlyoriented over every proton, producing a zero net transverse magnetization.The phenomenon of nuclear magnetic resonance (NMR) with induction was first demonstratedby F. Bloch (1946) using orthogonally placed coils exposed to a strong static magnetic fieldperpendicular to both coils. His experiment revealed that an induction current could be createdin one coil if an alternating current was applied in the other with a resonance frequency ω0 [7].Resonance in nuclei can be achieved by application of radio frequency (RF) pulses at ω0causing M0 to tip transversely away from B0 towards the transverse (xy) plane, which generatesa detectable signal. This excitation can be manipulated to control the degree of transverserotation, often by changing excitation times.The Larmor equation (1.5) describes the basic relationship between the resonance frequencyω0, the external magnetic field B0, and the gyro-magnetic ratio γ (a proportionality constantdependent on the type of substance).ω0 = γB0 (1.5)Precession of the magnetization vector occurs when it is perturbed away from the directionof field B0. However, in practicality the protons are in non-isolated systems, and experienceboth spin-lattice (proton-environment) and spin-spin (proton-proton) interactions which returnthem to thermal equilibrium. Over time, relaxation of the tipped magnetization vector M′towards B0 occurs in both longitudinal and transverse directions, and are referred to as T1 andT2 relaxation times (TR), respectively. Figures 1.3 and 1.6 illustrate the general principles ofmagnetic resonance and transverse decay.1.3.2 Non-Uniform Magnetic Fields: Field GradientsFor a patient aligned in an external magnetic field Bz, the temporary application of linear fieldgradients in a chosen direction creates a range of Larmor frequencies, each spatially defined bythis field gradient. As a result, ’slices’ the patient in the XY plane can be chosen by applyingexcitation pulses matching slice to its corresponding Larmor frequency 1.5. Similarly, gradientsapplied in the X or Y direction allow for slice selection in the YZ and XZ planes, respectively.Changing the field gradient strength (slope) or excitation pulse bandwidth enables control overthe signal range in the Z direction, translating to slice thickness.The ability to create magnetic field gradients to vary resonance frequencies with spatial lo-cation, combined with the method of Fourier analysis introduced by Ernst et al. [26], allows forsignificant biomedical applications of magnetic resonance imaging.1.3.3 Magnetic Susceptibility and Tissue OrientationMagnetic susceptibility ”χ” is a property of a substance to become magnetized in an externalmagnetic field. When placed in a magnetic field of strength H, the overall magnetization of anobject M is defined by 1.6, with paramagnetic substances having positive χ (from the effects ofexternal magnetic field on unpaired electrons), and diamagnetic substances having negative χ.71.3. Magnetic Resonance ImagingFigure 1.5: Slice selection using gradients. Linear field gradients applied in the Z directioncombined with nonchanging fields in the X and Y direction affects resonance frequencies onlyalong Z. By applying pulses specific for a certain frequency, ’slices’ in the XY plane can beselected. In patients, field gradients in X, Y, or Z allow visualization of body tissues at almostany orientation. Adapted from Xiang, Q. 2004Figure 1.6: Transverse (MT ) decay and longitudinal (ML) recovery of a tipped magnetizationvector M(t) as it relaxes longitudinally in parallel with the main magnetic field (B0). Adaptedfrom Xiang, Q. 200481.3. Magnetic Resonance ImagingIn MRI images, paramagnetic materials strengthen magnetic field, while diamagnetic materials(such as most of the body’s tissues) weaken the field.M = χH (1.6)Magnetic susceptibility is known to be a source of shifts in Larmor frequency. Frequencyshifts directly affect phase contrast, resulting in a source of ambiguity in phase data. To addressthe effects of local magnetic fields created by magnetically polarized particles in contributing tothese frequency shifts, the concept of a Lorentzian sphere is widely used as a good approximationfor isotropic structures [46]. The equation (1.7) is a proportion of the ∆f frequency shift fromthe base Larmor frequency f0, with χ being the volumetric magnetic susceptibility of the object.∆ff0=43pi · χ (1.7)While the exact origin of magnetic susceptibility variations in tissues has not been fullyestablished, suspected sources include presence of iron, ferritin, and lipids [24][35][61]. Comparedto the existing studies that investigate iron’s contribution to magnetic susceptibility [50][95], ithas been recently proposed that tissue orientation also affects to magnetic susceptibility, and hasmost recently been applied to studies in the brain [17]The anisotropic and elongated structures in the CNS such as spinal column WM tracts possessmyelin arranged in longitudinal bundles along the axon that are difficult to approximate with aLorentzian sphere. Because magnetic susceptibility contributions from longitudinal structures arenot factored into the sphere model, a generalized Lorentzian model (equation 1.8) was introducedby He and Yablonskiy [35] using the shape of a Lorentzian cylinder (ellipsoid) for circular axonalbundles in external isotropic media.∆fWM and ∆fe denote frequency shifts in WM and and its surrounding media, respectively.αm is the angle between B0 and the long axis of the ellipsoid aligned with the length of WMtracts. (1.9) is the expected total contribution χtotal from the magnetic susceptibilities of: whitematter (χWM ), macroscopic contribution of the object (χa), and the external media (χ).∆fWM −∆fef0= −2pi · (χtotal) · (cos2 αm − 13) (1.8)χtotal = χa − χWM + χ (1.9)Equation 1.8 was shown to quantitatively explain shifts in frequency and phase contrastbetween GM and CSF, or WM and GM in the motor cortex, that were previously suspectedbut unseen in spin-echo magnitude images. The introduction of this model also suggests thatphase shifts vary due to the orientation of the tissue in relation to the main magnetic field [35].A recent study of cerebral WM with multi-echo gradient-echo imaging demonstrated that WMphase not only depends on magnetic susceptibility, but also on the angle αm between axon bundleorientation and main magnetic field B0 [18].1.3.4 Basic Components of MRI and Pulse SequencesIn modern clinical MR imaging, patterns of RF pulses and gradients, called MR sequences, areused to create MR images. These sequences involve of several important components:91.4. MR Imaging Techniques Used to Characterize WM1. The RF system, with a coil interacting with the subject being scanned. The coil can act asboth the transmitter for excitation, and the receiver for signal detection. The RF excitationpulse applied at 90 ◦ rotates the spin from a longitudinal orientation to transverse.2. The Gradient system, with paired coils oriented to generate field gradients for spatial en-coding in 2D or 3D.3. A computer, which synchronizes pulse sequences and receives raw data to be analyzed.Various available programs and techniques can be used to reconstruct the data into images.1.4 MR Imaging Techniques Used to Characterize WMThe diversity found in available MRI pulse sequences and methods enables generation of imagesthat reflect different properties in tissue.1.4.1 Myelin Water ImagingAlthough there are no current non-invasive methods of tracking myelin in spinal cord in-vivo,several indirect methods of myelin detection are available, including myelin water imaging. Thefocus on myelin water, as opposed to myelin itself, arises from the inability to directly detectthe short T2 relaxation times (<1ms) of myelin protons [36]. Instead, the water trapped be-tween myelin bi-layers has a T2 time <50ms, and thus can be imaged. Using multi-echo spinecho techniques (CPMG sequence), a sequence of images are acquired to sample the T2 decayat increasing echo times, covering the range of water compartments: within myelin bi-layers,intra/extracellular water, and cerebrospinal fluid (CSF). Fitting each voxel of the multi-echo T2data (from <7ms to >1000ms) is done with a regularized non-negative least squares (NNLS)algorithm. This produces per-voxel distribution of relaxation rates, which is used to determinethe ”fraction” of water molecules with faster relative decay (within myelin bi-layers) in that par-ticular voxel. The ”fraction” of water bound by the myelin bi-layers is calculated by dividing itsrespective component (1) over the sum of the total weights of all the water components (1) +(2) + (3), where (2) and (3) are the intra/extracellular and CSF components, respectively (Fig.1.7). Quantitative myelin water measurements have been shown to correlate well anatomicallywith myelin distribution [56][41][89].Figure 1.8 illustrates the multi-spin-echo Carr-Purcell-Meiboom-Gill (CPMG) sequence, whichgenerates a T2 relaxation curve showing distinct water compartments (WC) at different T2 re-laxation times. A basic spin echo sequence is modified with additional 180 ◦ at 2τ intervalsto produce multiple echoes. Over multiple echoes, nonrecoverable signal loss from T2* effectsmanifest as gradual signal decay, which can be fitted to a curve to obtain per-voxel T2.This method has been applied to show myelin loss in normal appearing white matter (NAWM)in healthy patients compared to multiple sclerosis patients [51][55]. More recently, the applicationof high-resolution MWF maps to study rat spinal cords in-vivo and ex-vivo showed average MWFvalues for WM and GM that agreed with previously published results, demonstrating the potentialof using rat models to study WM damage in spinal cord injury [38][37][39].While MWI is able to detect presence of overall myelin content, it cannot differentiate betweenintact myelin on healthy axons, or damaged myelin. In the rat spinal cord injury model, weexpect to differentiate between healthy and damaged myelin based on differences in magneticsusceptibility caused mostly by changing tissue orientation, with healthy myelin being relatively101.4. MR Imaging Techniques Used to Characterize WMFigure 1.7: Components of myelin bilayer and intra/extracellular water. Signal amplitudes forcorresponding T2 relaxation times, with peaks at: 1) T2 < 30ms related to water trapped betweenmyelin sheath bilayers, 2) between 40ms < T2 < 100ms representing water inside and outsidethe axon, and 3) from water in cerebrospinal fluid (CSF), at T2 > 1000ms. CSF peak not shown.Adapted from Laule et al. [41]111.4. MR Imaging Techniques Used to Characterize WMFigure 1.8: Vector diagram of CPMG. At t=0, 90 ◦x’ tips the magnetization vector into thetransverse plane. Following a dephasing period τ , a 180 ◦y’ pulse is applied to flip the spins,reversing their direction. The spins rephase after τ , and the sequence can be applied again formultiples echoes. Note that the the signal shown on the bottom is after 2τ , where subsequentechoes are observed at 2nτ intervals. The overall signal decreases over increasing TE, producinga T2 decay curve. Adapted from Xiang, Q. 2004121.4. MR Imaging Techniques Used to Characterize WManisotropic. Damaged myelin would be more isotropic due to dissociation from the axon, andappear as myelin debris distributed evenly throughout the cellular environment.1.4.2 Diffusion Tensor Imaging (DTI)Applications of DTI in studying Alzheimers disease, multiple sclerosis, brain tumors, strokes, andspinal cord damage are well reviewed [8][14][62]. The ability of DTI to reveal details on tissuemicrostructure fit well with the WM/GM organization of the brain and spinal cord, making it wellsuited in detecting changes in white matter pathways of the central nervous system [15][57][65].Myelinated WM in the CNS, which produces anisotropic diffusion due to its linear near-cylindrical structure, is structurally different than isotropically distributed GM, and biases waterdiffusion in the spinal cord. Though this behavior in diffusion is not specifically due to presenceor absence of myelin, previous studies have shown that myelin and myelin damage does contributeto diffusion parameters [22][28][76][? ].Diffusion can be measured through careful application of signal-dephasing and rephasing gra-dients. First, a gradient is applied to dephase the nuclei. Applying second gradient reverses theeffects of the first gradient to rephase the spins. Although full rephasing is expected, random dis-placement of water molecules after the first gradient places them at locations different from theirinitial position. When the second gradient is applied, the displaced molecules are not perfectlyrephased, resulting in reduced signal.The gradient can only be applied in one direction at a time, and optimal diffusion imagingcollects images from multiple directions to generate a diffusion tensor, typically given in the formD =∣∣∣∣∣∣Dxx Dxy DxzDyx Dyy DyzDzx Dzy Dzz∣∣∣∣∣∣which possess 6 unique terms Dxx, Dxy, Dxz, Dyy, Dyz, andDzz. 6 sets of data with gradientsapplied in 6 different directions (in all 3 x, y, z, axes) for each set, combined with a baseline B0image with no gradient applied, produce enough data to mathematically produce this diffusiontensor.The diffusion tensor is diagonalized to produce a set of eigenvectors and eigenvaluesΛ =∣∣∣∣∣∣λ1 0 00 λ2 00 0 λ3∣∣∣∣∣∣which characterize a ’diffusion ellipsoid’ (Fig. 1.9). Conventionally, the λ are labelled in orderof magnitude, with λ1 ≥ λ2 ≥ λ3.Structurally, WM has long axons, which are all longitudinally oriented in one axis, whereasGM neurons possess short axons that are generally oriented in all directions. As a result, weexpect biased water diffusion, where water diffusion longitudinally (axial, λ1) to white matteraxons is different than water diffusion transverse (radial, λ2 and λ3) to the axon 1.9.Longitudinal (axial) Diffusivity DLong, Transverse (radial) Diffusivity (TD), Fractional Anisotropy(FA) and Apparent Diffusion Coefficient (ADC) parameters generated from the diffusion tensoreigenvectors describe different aspects of water diffusion. In the rat spinal cord, certain parame-ters correlate with WM axon degeneration, and others with myelin degeneration.131.4. MR Imaging Techniques Used to Characterize WMFigure 1.9: Illustration of the diffusion tensor. A) Diffusion in isotropic media is expected to beequal in all dimensions, with the expected root mean squared displacement forming a diffusionsphere. B) In anisotropic media, bias in longitudinal (λ1) diffusion is reflected in a more ellipsoidshape.Longitudinal Diffusion Of the three diffusion eigenvalues, λ1 is biased when diffusion isanisotropic to reflect longitudinal diffusion, and helps differentiate between ordered WM andGM based on such diffusion patterns along the long axon structure of WM. Decreases in lon-gitudinal (axial) diffusivity correlate with decreases in anisotropy from axonal damage or axondegeneration [19][76], and can be used as a measure of axon integrity [47].λ‖ = λ1 (1.10)Transverse Diffusion Transverse (radial) diffusivity is sensitive towards myelin amount, andcan differentiate between axonal damage and myelin damage. Demyelination is found to increasetransverse diffusion without changing longitudinal diffusion [77][78], and increases in transverse(radial) diffusivity correlate well with reduced myelination and demyelination [12][76][85].λ⊥ = (λ2 + λ3)/2 (1.11)Apparent Diffusion Coefficient ADC represents the average diffusion of water per voxel. Asan average, higher ADC denotes fast diffusion, whereas slow ADC indicates low diffusion. Theresultant intensity in an ADC map thus reflects the mean or avreage measured diffusion ratesin each voxel. In the spinal cord, ADC values correlate with axon counts negatively, with ADC[37][70].ADC = (λ1 + λ2 + λ3)/3 (1.12)Fractional Anisotropy Anisotropy in diffusion is defined by unequal diffusion in any or alldirections, and is a good measure of fiber and tract orientation in specific tissues such as thespinal cord, where such tracts rarely cross. FA is defined according to the equation 1.13. In thespinal cord, WM tracts limit the direction of water diffusion, resulting in higher FA [3]. Axondegeneration is shown to decrease FA in WM tracts [37]. Based on bias from either λ1, λ2, orλ3,141.4. MR Imaging Techniques Used to Characterize WMthe FA equation 1.13 can be simplified to 3 general cases of a spherical 1.14, linear 1.15, or planar1.16 shape of the diffusion ellipsoid. Axon degeneration is shown to decrease FA in WM tracts[37].FA =√12√(λ1 − λ2)2 + (λ1 − λ3)2 + (λ2 − λ3)2√λ21 + λ22 + λ23(1.13)Linear =λ1 − λ2λ1 + λ2 + λ3, forλ1  λ2 ' λ3 (1.14)Planar =2(λ2 − λ3)λ1 + λ2 + λ3, forλ1 ' λ2  λ3 (1.15)Spherical =3λ3λ1 + λ2 + λ3, forλ1 ' λ2 ' λ3 (1.16)1.4.3 Phase Imaging, Magnetic Susceptibility, and Tissue OrientationImaging of the brain and spinal cord using gradient echo (GRE) MRI has yielded excellent phasecontrast in white and gray matter (WM, GM) [67] [24], especially at high field. Recent focususing this technique has shown GRE phase imaging to be a potentially powerful tool in revealingWM abnormalities related to neuro-degeneration and injury, distinct from other standard clinicalmethods [24], [90], [35], [92]. The underlying nature of this phase contrast is not yet clear, thoughseveral possible sources being studied include bulk susceptibility changes [42], tissue susceptibilityanisotropy [43], chemical exchange [75], and presence of iron [25].Phase is limited to a domain spanning over 2pi (−pi < n < pi) and causes aliasing (’phasewraps’) in the mapped data. Though fully unwrapping the phase is often impossible or imprac-tical, alternative methods to correct for phase wraps (homodyne filter, Laplace filter) have beenused to produce rapid and relatively accurate phase images[54] [59] [71] [90].Magnetic susceptibility ”χ” is a property of a substance to become magnetized in an externalmagnetic field. In MRI images, paramagnetic materials strengthen the magnetic field, whilediamagnetic materials (such as most of the body’s tissues) weaken the field. Because water is thepredominant source of (paramagnetic) susceptibility in the bulk tissue, local field inhomogeneitiesarise from sources that have susceptibilities either more negative (diamagnetic) or more positive(paramagnetic), relative to water’s susceptibility.Protons will experience a shift in frequency ∆f from the Larmor frequency f0 due to magneticsusceptibility effects from their local environment, on top of bulk magnetic susceptibility effects.Sources with more paramagnetic susceptibility compared to water will increase the field, causingincreased precession, which increases the phase (and vice versa for diagmagnetic sources). Becausethe phase information is what the MRI provides, figuring out its associated frequency informationis deemed a ’reverse’ problem. Ideally, once the frequency map is obtained from the phase map, itcan reveal information as to which factors in the local environment contributed to the frequencyshifts.In tissue, the measured MR frequency is shifted due to local magnetic susceptibility varia-tions in tissue components, including proteins, lipids, and iron (deoxyHb or ferritin), which alsocontribute to overall tissue microstructure. Traditionally, frequency shifts are predicted using astandard Lorentzian Sphere approximation, which considers magnetic susceptibility inclusions ina local area of interest as point dipoles. However, this method was shown to be unable to predict151.4. MR Imaging Techniques Used to Characterize WMloss of frequency shift contrast between demyelinated WM and GM (reflected in experimentalresults), [29] [44]. Due to the relative anisotropic organization of WM fibers in the CNS, recentstudies have also looked into GRE phase contributions from structural ordering and disorderingat the cellular and molecular level [88] [81] [94].The traditional method of using a standard Lorentzian Sphere (Eqn.(1.17)) to approximatefrequency shifts from phase considered magnetic susceptibility inclusions in a local area of in-terest as point dipoles. Because WM is not isotropic, nor spherically oriented, this method wasshown to be unable to accurately model and predict the frequency shift between demyelinatedWM and GM [29][44]. He et al., [35] proposed using a General Lorentzian Approximation (GLA)method, which improves on the conventional Sphere approach by incorporating Lorentzian Cylin-der approximations to account for the highly anisotropic composition of WM axons and myelinsheaths.∆ff0=43pi · χ (1.17)∆fWM −∆fef0= −2pi · (χtotal) · (cos2 αm − 13) (1.18)Equation (1.18) was shown to quantitatively explain shifts in frequency and phase contrastbetween WM and the other spinal tissues. Previously, the tissue contrast predicted with theSphere model only did not agree with contrast in the actual frequency images. The model alsosuggested that phase shifts vary due to the orientation of the tissue relative to the main magneticfield [35]. Recent studies of cerebral WM with multi-echo gradient-echo imaging demonstratedthat WM phase not only depends on magnetic susceptibility, but also on the angle between axonbundle orientation and main magnetic field B0 [18][49].Yablonskiy et al., [92] constructed a model of frequency shifts in MS lesions incorporatingthe GLA, focusing on scenarios of demyelination facilitated by damage to myelin and axons.Due to its makeup consisting mostly of lipids, myelin (relatively diamagnetic) damage has theopposite effect on the frequency shift than damage to proteinous axonal neurofilaments (relativelyparamagnetic), and thus a combination of myelin and axonal damage could result in overall losscontrast. This theory has been coupled with experimental results demonstrating the ability offrequency shifts to predict lesion formation at very early stages in MS patients [90].However, the pathology of MS lesions in the human CNS is complex, subject to hemorrhagesand glial cell infiltration resulting in hemosiderin leaks, formation of reactive oxygen species [34],and iron accumulation by macrophages [83]. Both the healthy and damaged environment in thespinal cord injury are less complex. Rat SCI models produce a relatively cleaner environmentof degeneration in several nerve tracts. Axonal death is expected to occur along the tracts bi-directionally to injury, resulting in progressive demyelination and axonal degeneration over severalweeks [37]. The simple geometry of the cord allows placement of the cord parallel to B0, removingcontributions to MR signal phase from the cord’s external shape [49]. Hence, the frequency shiftsobserved can be attributed primarily due to axonal and myelin damage across several timepoints.With frequency shift mapping, it is possible to assess animals both in-vivo and ex-vivo at higherresolution and faster than DTI or MWI.16Chapter 2Materials and Methods2.1 Experiment: Rat DC-Tx Model and Comparison toHistologyIn this study, we use the Dorsal Column Transection (DC-Tx) injury in rat cervical spinal cord toproduce a secondary degenerative injury along several nerve tracts of the dorsal column. Not onlyis the DC-Tx model widely used and histologically validated, it also produces well defined patternsof axonal degeneration and demyelination over several weeks [11]. The C6 DC-Tx injury modelcauses damage bidirectionally from the transection site, producing secondary neural degenerationover several weeks. Optimal time-points to harvest spinal cords are at 3- and 8-weeks post injury(wpi), corresponding to periods of extensive myelin debris formation and clearance respectively[38]. Our study focuses on two closely bundled tracts in the dorsal column that run in oppositedirections: the ascending Fasciculus Gracilis, and the descending Corticospinal tract. Axonsdistal to the site of injury undergo Wallerian degeneration (cranial FG, caudal CST), whereasthe proximal axon segments undergo retrograde degeneration (caudal FG, cranial CST) [38].While Wallerian degeneration is characterized by acute axonal loss distal to injury within thefirst 72 hours followed closely by myelin sheath degeneration [31] [80], retrograde degeneration isa relatively slower process with axonal and subsequent myelin loss occurring over several weeks[91] [72]. As part of the inflammatory response, macrophages appear 2-3 weeks post injury toclear debris, and can persist for up to 90 days [31].By tracking both Wallerian and retrograde degeneration post-injury over time and usingimmunohistochemistry, we matched ex-vivo spinal cord MR images to histological tissue sectionsstained for myelin and axon neurofilaments (as a gold standard), then correlated the measuredfrequency shifts to distinct changes in WM axons and myelin. To minimize any effects fromhemorrhage, blood clots, and inflammation, neurodegeneration was monitored 5mm from the siteof injury. The simple cylindrical geometry of the cord allows for orientation of the cord parallelto B0, an important feature which removes microstructural contributions to MR signal phase inhealthy WM.Degen-MBP has been used as indicator of substantial change in myelin microstructure fromWM damage, including cases following ischemia and hemorrhagic events [53] [58]. Very few studieshave investigated the relationship between degen-MBP and MRI white matter injury biomarkers[84] [39]. In one case, degen-MBP has previously shown no significant correlation with myelinwater fraction [39]. Using degen-MBP to differentiate between healthy and damaged myelin atprogressive stages of neurodegeneration, we present frequency shifts as the first MR parameterto correlate significantly with degen-MBP in damaged spinal cord white matter, with a high ofdegree sensitivity for myelin degeneration previously unseen with other MR biomarkers.In this study, we explore MR frequency shifts in a spinal cord injury model and compare MRIresults with electron microscopy and histopathology.172.2. Rat Spinal Cord Injury Model2.2 Rat Spinal Cord Injury ModelSprague-Dawley rats (n = 21) were obtained from UBC and Charles River, and kept healthy inaccordance with approved Animal Card protocols at the ICORD Research Center in Vancouver,B.C. The rats were divided in per treatment: Healthy (n=6), 3-Week Post Injury (n=7), and8-Week Post Injury (n=8). After perfusion, spinal cords were extracted and fixed in paraformalde-hyde (PF) for a minimum of 3 hours before scanning. All scans were completed at room tem-perature within 24 hours of extraction, after which the cords were transferred to 24% sucrosesolution for storage and histology. All images were reconstructed from raw FID data, with dataprocessing performed using MATLAB (The MathWorks, Inc., Natick, MA, USA).2.2.1 Injury ProtocolEstablished methods for dorsal column transection (DC Tx) were used to injure anesthesizedhealthy rats, between C4 and C5 [37]. Anaesthetized rats underwent surgery in which a dorsalincision into the tissue an muscle in the cervical region was made to access the cervical spine. Alaminectomy performed in the C4-C5 region allowed access to the dorsal spinal cord. A scalpelcutting perpendicular to the spine at a depth of 2.5mm into the dorsal column severed the axonsin the region. After performing the injury, the dorsal incision was stitched closed before allowingthe animal to recover.2.2.2 Perfusion and Spinal Cord ExtractionExtraction of all cords were performed at the iCORD Research Center, with the following proce-dure:1. Deeply anesthesize rats2. Surgically open the thoracic cavity and expose the heart3. Using a pump with the output attached through the left ventricle and into the aortic archto flush intracardially with 100mL of phosphate buffered saline (PBS). An incision made inthe right atria allows blood to be pumped out of the cardiovascular system. This perfusionstep usually takes up to 5 minutes.4. A secondary perfusion is performed with fresh 4% PF (prepared in 0.1M sodium phosphatebuffer at physiological pH). This fixes the tissue and prepares the spinal cord for extractionand further fixation.5. Similar to the injury protocol above, incisions made dorsally allow access to the spinal cord.An extended laminectomy is performed to expose up to 20mm in length of spinal cord.6. Using a scalpel, the spinal cord is cut, making sure to have the injury centered in thesegment. Ideally, there is 10mm of tissue extracted both cranially and caudally to theinjury site.7. Ex-vivo cords were post-fixed in 4% PF, stored overnight at 4◦C, and scanned the followingday.182.3. MR Apparatus, Scan Parameters, and Data Analysis MethodsFigure 2.1: Preparation of rat spinal cord sections for MRI scanning. Extracted cords (left) areloaded in plastic cylindrical tubes with PF into the solenoid transmit/receive coil (center). Thecord and coil are placed in the 7T magnet for scanning (right). Note: Using this method, thecord is oriented perpendicular to the main magnetic field.8. After scanning, cords were post-fixed in daily increasing concentrations of sucrose (8%, 12%,and 24%) in PBS.9. After 24 hours in 24% sucrose, cords were frozen for histopathlogy (see Histopathologybelow).2.3 MR Apparatus, Scan Parameters, and Data AnalysisMethodsMRI experiments were conducted on a 7T Biospec 70/30 scanner (Bruker BioSpin Gmbh, Et-tlingen, Germany) at the UBC MRI Research Center 7T High-Field Lab in the Life SciencesInstitute. Custom solenoid and surface coils were used to obtain all the scans, which were per-formed with guidance from the staff at the 7T lab. Additional in-vivo scanning was conducedwith a phased-array coil. Figure 2.1 shows an overview of sample handling and setup.2.3.1 ApparatusA custom solenoid coil (13mm inner diameter, 20mm long, four-turn) was used for pulse trans-mission and signal reception, primarily due to its high-quality data output and low signal-to-noise(SNR) ratio. Fixed cords were immersed in PF and placed in a plastic tube (4.5mm diameter),with additional plastic rods to keep the cord aligned straight. Placement of the cord is perpen-dicular to the main magnetic field, as shown in Fig. 2.1, center. Phase images were acquired withcord placed parallel to the main magnetic field, on a custom surface coil.Spinal Cord Scanning protocolPrepared cord were scanned with the following protocol:1. Tripilot (12 seconds)2. 3D Orientation (FLASH Bas) (30 seconds)3. Longitudinal Slices (RARE 8) (2 minutes)192.3. MR Apparatus, Scan Parameters, and Data Analysis Methods4. 3D Gradient Echo (FLASH 3DD) (38 minutes)5. Myelin Water Imaging (cpmg orig) (48 minutes, repeated for each slice of interest)6. Diffusion Tensor Imaging (dti check) (1hr 40 minutes)2.3.2 Scan ParametersMyelin Water Imaging1mm single-slice CPMG [37][64] scans were conducted for myelin water measurements, withEcho-Time/Repetition-Time (TE/TR) at 6.673ms/1500ms, 32 echoes, and total square FOV of25.6 mm with 100µm x 100µm x 1 mm voxels in a 256×256 matrix, over 6 averages. One slicewas obtained for each healthy cord (38 minutes), and 3 slices per injured cord (total 1hr 52 mins).Diffusion Tensor ImagingDTI scans used a multi-slice (3) multi-echo sequence with TE/TR of 21.337ms/1500ms, in a totalsquare FOV of 12.8mm with 100µm x 100µm x 1mm voxels in a 128×128 matrix, over 4 averages.Magnetic field b-factors were set at 50s/mm2 and 750s/mm2 in six non-collinear directions usingvarious diffusion sensitizing gradients according to established methods [37][52], with a total scantime of 1 hour and 44 minutes.PhaseMulti echo gradient echo (MGE) sequence was used to generate 3D data along the entire cord,divided into 1mm slices during reconstruction. Parameters: TE = 4.2ms, subsequent echoesspaced at 3.7ms intervals, TR = 35ms, 20 flip angle, Echoes = 6, and 16 averages, and totalFOV of 6.4mm with 33µm x 33µm x 1mm voxels in a 192x192 matrix. This method allowed dataacquisition in relatively short scan times (38 minutes) with a high in-plane resolution (33µm x33µm), while maintaining a 3:1 pixel ratio to MWI and DTI resolutions. Homodyne filters (high-pass filters) were then used to remove background field inhomogenieties, and generate relativelyunwrapped phase images.HistopathologyCords were fixed in 24 percent sucrose and cryoprotected, and 20um-thick cross sections forstaining. To generate reliable quantitative results, histology of the spinal cord in sections corre-sponding to the obtained scans were used (at 5mm cranial to, 5mm caudal to, and at the injurysite). Histochemical and Histoimmunochemical stains:1. Myelin Basic Protein (MBP, a validated stain for myelin proteins2. Degen Myelin Basic Protein (dMBP), for degenerating myelin3. Neurofilament/Tubulin antibodies (NF/Tub), for axon staining4. Eriochrome, a more effective myelin stain202.3. MR Apparatus, Scan Parameters, and Data Analysis MethodsStains 1 through 3 were used previously in [37]. NF/Tub staining is expected to correlatewell with DLong values involving axonal damage. LFB and Eriochrome staining for myelin hasshown to correlate well with MWF, and also correlate somewhat with FA. Conversely, MBPwas not found to be reliable in measuring myelin post-injury primarily due to the processesof Wallerian degeneration, and thus was not used in this study. Instead, dgen-MBP stain wasused to identify degenerated myelin [37]. Stained sections were imaged and digitized underhalogen light (fluorescent for immunostained sections) with a Zeiss Axioplan 2 microsope withNorthern Exposure software. Average optical densities (OD) in each ROI, normalized to healthyventrolateral WM, were measured with Sigma Pro 5.0 software (SPSS Inc.).2.3.3 Registration and ROIsRegion-of-interest (ROI) analysis was used to individually analyze each WM tract in the dorsalcolumn for each image of interest. Using MATLAB, ROIs were drawn based on the image for eachdataset producing the highest contrast to improve accuracy: the first echo from MWI CPMGimages, the B0 image from DTI, and the magnitude images from GRE.While phase images acquired in these experiments had nine times the spatial resolution ofDTI and MWI images, the highest resolution scans of cord slices were produced from opticalmicroscope images of stained cord sections. ROI analysis in histology images was done in thesame method as above: drawing ROIs around WM tracts of interest, and quantifying stainedmyelin and axons. To normalize these measurements, average ROI values from sections of lateralwhite matter not involved in injury were used. In the case of phase, normalizing to lateral WMalso accounted for any transverse phase wraps present across slices.Overlaying the high-resolution histology images to their corresponding images was required.This process, called registration, requires feature-based transformation of the histology imagesetto match the orientation, size, and fit of the MRI image. This method has been used [37], andsimilar microscopy of histology stains has been done to quantify and observe axon damage [85].Electron MicroscopyTissue blocks were fixed by cross-linking proteins with 1% osmium tetraoxide in 0.1M CaCl, thenstained with 2% uranyl acetate in 50% ethanol, dehydrated, and set in resin. 90nm-thick sectionswere cut and placed on 200 square/inch SPI TEM grids for imaging and digitization on a Zeisselectron microscope.StatisticsFor statistical analysis of results, data from each treatment and timepoint were grouped andanalyzed for normality (Kolmogorov-Smirnov). For comparison analysis between groups, corre-lations (Pearson) and pair-wise multiple comparison (Tukey-Kramer for normal, Kruskal-Wallisfor non-normal groups) was performed with Medcalc (Ostend, Belgium), with significances de-termined at p < 0.05, p < 0.01, and p < 0.001. Multiple regression models of Wallerian andretrograde degeneration were conducted with all compiled data, using bootstrap aggregation over1000 bootstrap samples in SPSS (IBM Corp. Armonk, NY).21Chapter 3Results and Discussion: FrequencyShifts of Myelin Changes in DC-TxPathology3.1 Dorsal Column: Anatomy and PathologyFigure 3.1 shows DC WM segmented into sensory (Fasciculus Gracilis, FG, ascending), motor(Corticospinal, CST, descending), and mixed (Fasciculus Cuneatus, FC) tracts. The resolutionachieved in the spinal cord images (33µm in-plane at 1mm slice thickness) are relatively high[38] [79], showing excellent GM/WM contrast at 7T. Anatomical features such as the dorsal andventral nerve roots, the central canal, and ventral median fissure are well defined in both MRmagnitude and frequency maps. In addition, reduced partial volume effects reveal previouslyunseen features, including the corticospinal tract, and spinal gray matter laminae that transitioninto lateral white matter (Fig. 3.1, A) [38]. By diameter, axons in the dorsal column are largestin the FC, smaller in the FG, and smallest in the CST [82]. Similarly, myelin sheaths are thickestin the FC, thinner in the FG, and thinnest in the CST [13] [23]. GM in the spinal cord is notmyelinated, though certain myelinated white matter fibers permeate areas within GM [33].Results are discussed with respect to the location of neuronal cell bodies relative to the injurysite (Table 3.1). As an ascending tract, the FG proximal segment is caudal to injury, while itsdistal segment is cranial to injury. Conversely, the CST travels in the opposite direction, and hasits proximal segment cranial to injury.Table 3.1: WM tract orientations in the dorsal column. Tract degeneration patterns relative tocranial-caudal direction from injury. As an ascending tract, the FG experiences Wallerian in thecranial direction, while the descending CST experiences Wallerian degeneration caudally.223.2. Evaluating Dorsal Column Transection Model in Rat Spinal CordFigure 3.1: Cross-sectional slice of cervical rat spinal cord showing WM/GM contrast andcorresponding anatomical features. (A) Transverse magnitude slice, with ROIs of the FasciculusGracilis (orange) and Corticospinal Tract (green). Areas of interest denoted: Dorsal/VentralHorns (DH, VH), Central Canal (CC), Ventral Median Fissure (VMF), Ventral White Matter(VWM), and Lateral White Matter (LWM). Gray matter lamiae transition (Tr) into WM, fromthe DH to LWM (B) Same regions in a MR frequency map of the corresponding PF fixed section,with gray matter appearing brighter than white matter. Scale bar = 1mm. Frequency scaled to± 1.446 ppb3.2 Evaluating Dorsal Column Transection Model in Rat SpinalCordFor WM neurons, the MR frequency shift (∆ff0 ) can be described as∆ff0= LF · χ, a product ofthe Lorentz Factor coefficient and magnetic susceptibility [92]. The secondary microstructuralcontributor to frequency is an object’s general external shape ∆ff0 = SF (α) · χ, also called theShape Factor (SF), with angle α. In the case of WM axons oriented parallel to B0 as we have inspinal cord, the SF contribution is zero, making LF the primary contributor to microstructure-related frequency shifts. As spinal cord microstructure lose structural integrity, the LF shifts fromcylindrical (LF = 0) towards spherical (LF = 4pi3 ), revealing magnetic susceptibility contributionsfrom underlying tissue components with paramagnetic (lipids) or diamagnetic (proteins) χ relativeto water.Tight concentric myelin bilayers are anisotropically oriented along the axon, producing an MRphase-dependent WM tissue magnetic architecture, which affects observed phase contrast in areasof myelination [10] [88] [69] [93]. As myelin sheaths degenerate, their bilayers pull apart. The over-all anisotropic microstructure transforms into isotropic lipid-rich myelin bundles or myelin ovoids,causing frequency shifts to increase proportionately to the degree of myelin damage [31][92]. Theorganized cytoskeleton of axons are oriented cylindrically, and placed parallel to B0 (main mag-netic field). When damaged, neurofilaments and tubulin break down into isotropic fragmentsthat cause similarly proportionate frequency shifts.233.2. Evaluating Dorsal Column Transection Model in Rat Spinal CordFigure 3.2: Electron microscope micrographs of degenerating WM. Retrograde degeneration inthe FG (A, B) show myelin sheaths loosening over time, maintaining concentric organizationaround the axons. Internal axon structure is lost in many axons by 8-weeks post injury. Walleriandegeneration in the CST (C) produces acute axonal loss and moderate demyelination. Similarly inthe FG at 8-wpi (D), axon degeneration forms cavitations, with myelin forming debris fragments.Evidence of remyelination appears at 8-wpi with sparsely myelinated axons (see red star). 6300xmagnification, scale bar = 5um.243.2. Evaluating Dorsal Column Transection Model in Rat Spinal Cord3.2.1 Healthy CordBoth histology and MRI of uninjured spinal cords (Fig. 3.2.1, Rows 1 and 4) show clear contrastbetween WM and GM, as well as some contrast within WM tracts of the DC. Eriochrome images(B1, B4) show high myelin content in CST and FG and low myelin content in the GM. There issome contrast between FG/CST and the fasciculus cuneatus on eriochrome. Axonal stains of thehealthy CST appear less granular with a deeper blue than the FG, consistent with having moredensely packed axons than surrounding WM (Fig. 3.2.1, C1). ED-1 staining (D1) shows almostcomplete absence of macrophages. There was very little variability in frequency in non-injuredWM tracts and GM across all time points, both cranially and caudally (Fig. 3.4).3.2.2 Cranial to Injury SiteWallerian Degeneration (Fasciculus Gracilis) At 3-wpi there is loss of axons and dis-integration of the myelin sheaths, as indicated by the reduction in NF/TUB and an increasein eriochrome and degen-MBP stains. Concurrently, macrophages have infiltrated the tract.Frequency in the FG remains unchanged but there is a slight frequency increase in in the ad-jacent cuneatus that corresponds to a reduction in eriochrome staining. Transmission electronmicroscopy (TEM) of tissue shows near-complete axonal loss that produce axonal cavities, andmyelin sheaths pulling apart (Fig. 3.2).By 8-wpi, there is significant increase in frequency (A3). Eriochrome reveals reduction inmyelin (B3) and there is further increase in degenerated myelin (C3) compared to 3 weeks afterinjury. Macrophage activity is still present at eight weeks and limited to the FG, but in asmaller area due to shrinkage of the tract as damaged neurons are cleared. The region of highmacrophage activity overlaps with the area of reduced eriochrome stain, suggesting that myelinhas been removed or has been taken up by the macrophages. Both TEM and NF/TubIII stainsshow axon content remaining low. Additionally, myelin is severely disintegrated, forming sheathfragments of varying shape and size. Of the very few intact axons that remain, (sparse) evidenceof initial remyelination can be observed in the FG (Fig. 3.2).Retrograde degeneration (Corticospinal Tract) At 3-wpi, frequency decreases. Eriochromeincreases significantly and NF/TUB experiences a significant decrease. There is no significantinfiltration with macrophages and no significant change in degen-MBP. Though eriochrome stainsdo not increase significantly at 8-wpi, a significant increase in degen-MBP indicates more myelindamage. It is possible that the myelin sheaths have begun dissociating, keeping their cylindricalorientation while experiencing more swelling between concentric sheath layers. However, suchmorphological changes can only currently be assessed with quantitative EM. Axonal stains dontchange from 3 weeks, but closer inspection with TEM shows more cavitations and shrinkage inaxons at 8 weeks (Fig. 3.2). These changes contribute to a net frequency decrease.3.2.3 Caudal to Injury SiteWallerian degeneration (Corticospinal Tract) Eriochrome, degen-MBP, and ED-1 in-crease significantly by three weeks after injury, while stains and TEM show axon integrity de-creases significantly. These differences correspond with a strong decrease in frequency (Fig. 3.2.1,A5). Compared to cranial FG three weeks after injury, the degree of myelin debris formation andmacrophage presence in the CST is less abundant (Fig. 3.2.1, D2 and D5).253.2. Evaluating Dorsal Column Transection Model in Rat Spinal CordFigure 3.3: Enlarged images of dorsal column, both cranial and caudal to injury. The cranial im-ages show (1) healthy, (2) 3-week post injury, and (3) 8-week post injury cords +5mm from injury.(A) Frequency contrast in FG and CST produced by degeneration, with brighter gray matter re-maining unchanged. (B) Eriochrome staining of myelin phospholipid content with brighter areasdenoting increased staining due to increased surface area. (C) Immunohistochemistry for majorneuronal components: neurofilament/tubulin III (NF/TubIII, blue, marker of axonal content),and degen-MBP (dMBP, red, marker of degenerated myelin content). (D) Immunohistochemistryfor ED-1 macrophage marker. Corresponding caudal sections in (4), (5), and (6). Both the FGand CST shrink physically as damage and tissue removal progresses over time: caudally for theCST, and cranially for the FG. The characteristic V-shape Post-synaptic Dorsal Column Pathwaythat borders the FG is most prominent in D5 and D6 (red arrows), and observable in frequencymaps A2, A3, A5, A6. Minimal tearing (tg) due to tissue preparation is mostly limited to GM.Histology at 100x magnification, scale bar = 250µm. Frequency scaled to ± 1.446 ppb.263.2. Evaluating Dorsal Column Transection Model in Rat Spinal CordFigure 3.4: Frequency shifts in paraformaledyde-fixed control WM (ventrolateral) and GM,averaged over all subjects in each treatment, showing no significant variability across subjectsover time.Macrophages become very active at 8-wpi, with elevated degen-MBP stains and decreasedmyelin content indicating that remaining myelin sheaths are mostly damaged (Fig. 3.2.1, Row6), which agrees with TEM. Axon stains remain low, with frequency shifts showing a significantincrease from three weeks.Retrograde degeneration (Fasciculus Gracilis) At 3-wpi, frequency shifts decrease, cou-pled with a decrease in axons (NF/TubIII), minor increase in eriochrome myelin, and no changein ED-1 macrophage activity (Fig. 3.2.1, Row 5). Similar to retrograde CST, we believe myelindegeneration could be characterized by intra-sheath swelling as a result of axonal shrinkage andcavitation (TEM, Fig. 3.2), though further quantitative EM results would provide a clearermorphological assessment of sheath integrity. More structural changes occur in the axons ratherthan the myelin at 3-wpi, correlating with decreased frequency shifts. Through 8-wpi, there is nosignificant change in all histology stains compared to 3 weeks, with overall condition in the cordchanging minimally from 3-wpi (Fig. 3.2.1, Row 6). Macrophages continue to remove both axonand myelin debris, and overall frequency shifts remain decreased.Overall, frequency shifts in Wallerian degeneration correlated strongly with Eriochrome in bothFG (R2=0.0.7975) and CST (R2=0.8187), and degen-MBP in FG (R2=0.7287). Frequency shiftsin retrograde degeneration correlated strongly with Eriochrome in FG (R2=0.0.4655), NF/TubIIIin both FG (R2=0.5479) and CST (R2=0.5365), and degen-MBP in CST (R2=0.0.5305)273.3. Discussion3.3 Discussion3.3.1 Frequency Shifts in Retrograde DegenerationWith continued impulse activity and survival factors [2], axons still linked to the cell body re-main functional post-injury, during which retrograde degeneration progresses proximally from thesite of transection. Both the CST (cranial) and FG (caudal) undergo such degeneration, withTEM showing moderate levels of demyelination and axon loss (Fig. 3.2, A and B). While lightmicroscopy is limited in resolution up to 0.2 /microm2, greater TEM resolving power enablesmore accurate assessment of tissue pathology, notably in cases when distribution of damagedand healthy neurons is non-homogenous. At 3-weeks post injury, TEM shows myelin appearingrelatively intact with some swelling between sheaths. Some axonal loss is also observed. Theeriochrome myelin stain appears brighter for this damaged myelin due to the increased surfacearea available to stain, and lack of degen-MBP and macrophages indicate no debris removal.In FG, none of the histopathology parameters show significant change at 8-wpi. EM revealsa significant population of axons lose internal structure by 8-wpi, while myelin sheaths continueswelling without disintegrating (Fig. 3.2, B). These changes drive frequency in opposite directions,resulting in no net frequency shift change at 8-wpi in FG.CST pathology differs from FG at 8-wpi. In addition to the same structural changes observedin FG at 8-wpi, the increase in CST degen-MBP and macrophages suggests debris removal.Densely packed small axons in CST form a physically restrictive environment in which sheathsare less able to freely dissociate into isotropic debris bundles. The bulk of remaining myelinsheaths may hinder the axons from dissociating as quickly as observed in the FG.With FG axons bundled relatively less densely, myelin debris form and diffuse more freely,leading to a weaker correlation between frequency and degen-MBP. The lack of significant histol-ogy and frequency changes in FG between 3- and 8-wpi could be explained by the existence of anequilibrium state between axonal loss, myelin debris formation, and debris clearance. Noting thephysical shrinkage of FG (Fig. 3.2.1, Caudal) is key to showing that this equilibrium involves con-tinued tract degeneration, rather than a static injury state. EM analysis at the microstructurallevel reveals more fragmented myelin in FG compared to CST at 8-wpi (Fig. 3.2, Retrograde),and also reveals presence of some remyelinating axons in FG (Fig. 3.2, D).3.3.2 Frequency Shifts in Wallerian DegenerationAt 3- and 8-wpi, myelin sheaths become damaged throughout the entire distal segment, forminghighly dissociated sheath fragments that fill both intracellular cavitations (voided by axonal loss),and extracellular space. Acute axonal loss within the first 72 hours post-injury form axonal de-bris, which is then phagocytized by 3-wpi (Fig. 3.2.1: Row 2 for FG, Row 5 for CST). Increasedfrequency shifts expected as a result of microstructural changes in myelin could be masked bybulk susceptibility effects of removing axonal protein content, for no overall change in frequencyshifts at 3-wpi were observed (Fig. 3 A-C). However, CST shows a significant frequency decreaseunobserved in the FG. Additionally, differences in axon size and density are apparent in histol-ogy, with macrophage penetration stronger in the FG. Densely packed small axons in the CSTphysically compress the intercellular environment, inhibiting widespread macrophage infiltrationand limiting myelin sheath separation. The CST experiences similar increases in degen-MBP toFG indicating similar myelin damage, but relatively less severe NF/TubIII decrease at 3-wpi sug-gests microstructural axon changes present in CST and not in FG are the source of the observed283.3. DiscussionTable 3.2: Multiple regression analysis of histology parameters, generating models for frequencyshifts in Wallerian (R2 = 0.839) and Retrograde (R2 = 0.536) mechanisms, with predictor im-portance listed for each histology parameter.frequency shift decrease in 3-wpi CST. Based on factors related to tract anatomy and the localmicro-environment, the resulting Wallerian degeneration can occur at slightly different rates. By8-wpi, pathology in both tracts show an environment dominated by myelin debris fragments,macrophages, and very few viable axons (Fig. 3.2.1, Rows 3, 6).3.3.3 Characterization of White Matter Damage in Spinal CordWhile the effect of WM tissue microstructure on frequency has only been recently suggested [35],it has been shown to play a significant role in detecting WM abnormalities in the brain, notablyin demyelinating MS lesions[90] [92]. With the DC-Tx model, we investigate frequency shifts inspecific patterns of myelin and axonal degeneration, comparing experimental evidence with theseproposed theories.Figure 3.5 is a combined diagram of degeneration observed in this DC-Tx study, outlining therelative levels of demyelination and axonal loss in the FG and CST, as measured through histologyand EM. We show that not only are frequency shifts sensitive to the underlying integrity of theneuronal microstructure, but they can also characterize and differentiate between Wallerian andretrograde degeneration over time.3.3.4 Histological Correlates of White Matter Frequency ShiftsResults from univariate regression analysis (Fig. 3.6) generates trends of decreasing and increas-ing frequency profiles over time in retrograde and Wallerian degeneration, respectively. Multipleregression analysis of all histological measures (Table 3.2) suggest myelin/axon integrity as pri-mary drivers of frequency shift change. Wallerian degeneration can be well observed with histology(R2=0.839). As expected, acute axonal loss makes NF/TubIII a relatively insignificant predictor.As retrograde degeneration is a relatively slower process, the cellular environment is a con-stantly changing balance of healthy and damaged axons and myelin. While macrophages dofunction to remove debris, their presence is minimal compared to the macrophage response inWallerian degeneration. Remaining myelin disintegrates slowly, inhibiting macrophage access tothe internal axonal debris. The overall frequency shifts in retrograde degeneration are nearlyevenly balanced between eriochrome (0.58) and NF/TubIII (0.40), with ED-1 and degen-MBPas relatively insignificant predictors. An R2 of 0.536 suggests alternative contributors to theobserved frequency shifts, one of which may be related to possible changes in tissue bulk sus-ceptibility from microglial activity. However, it is evident that overall changes in microstructure293.3. DiscussionFigure 3.5: Comprehensive schematic of DC-Tx in rat spinal cord. Central figure depicts WMdegeneration over time divided by tract (FG, CST) and mechanism (Wallerian, retrograde), withgeneral visual representation of myelin and axon damage based on histology and EM micrographs.Frequency for each treatment plotted with error bars, with significances between time-points forall histology measures marked above/below. In Wallerian degeneration, neuronal damage is ex-tensive, showing significant loss in myelin and axonal content over time, matched with increasedpresence of macrophages and myelin fragments. Remaining tissue is mostly debris by 8-wpi,with relatively more damage in the FG. At 3-wpi and later, frequency shifts show a signifi-cantly increasing trend compared to healthy baseline. Retrograde degeneration is a relativelyslower degeneration pattern characterized by changes in eriochrome and axon staining. ED-1 anddegen-MBP differences are present in CST only. Frequency shifts trend negatively in retrogradedegeneration, reaching a significantly negative frequency shift (-5 to -6 * 10−3 ppm). Differencesin tract anatomy are a contributing factor affecting injury pathology and subsequent immuneresponse in the FG and CST, as reflected in the central diagram.303.4. Frequency Changes in DC-Tx Evaluated Along the Spinal CordFigure 3.6: Frequency to histology correlations of degeneration. Wallerian degeneration (CranialFG, Caudal CST) and retrograde degeneration (Caudal CST, Cranial FG) were stained with eri-ochrome stain of myelin content (A, D), immunohistochemistry stain of Neurofilament/TubulinIIIaxonal content (B, E), and degen-MBP for degenerated myelin content (C, F). Tracts denotedby color (Orange FG, Green CST) and divided into healthy (circle), 3-wpi (triangle), and 8-wpi(square) time-points. Data sets encircled in groups to improve visualization. Correlations ofeach histology measure to frequency shifts shown in dotted lines with corresponding R2 values.Frequency shifts in Wallerian degeneration correlate strongly with eriochrome in FG and CST,and with degen-MBP in FG. Retrograde correlations significant with NF/TubIII axon stain inFG and CST, with eriochrome in FG, and with degen-MBP in CST.are related to changes in frequency shifts. Figures 3.6 and 3.5 combined offer a comprehensiveassessment of WM frequency shifts in rat spinal cord.3.4 Frequency Changes in DC-Tx Evaluated Along the SpinalCord3D frequency shift maps of spinal cords bi-directional to the injury were generated from 3D multi-gradient echo (MGE) phase, and progressive degeneration evaluated both cranially and caudallyfrom injury, over time, in the FG and CST.Significant damage to the DC at the injury site (0 mm) and immediately adjacent slices (±1mm) due to mechanical injury as well as local effects (inflammation and repair) are not Wallerianor retrograde, and thus not evaluated (Fig. 3.8, red and yellow highlights). Slices further away,up to ± 8 mm, are evaluated as well. Frequency shift values of healthy FG and CST WM in Fig.3.8 (black lines) are consistent in each tract across several animals and along the spinal cord,providing a pre-injury (control) baseline.Wallerian degeneration occurs relatively quickly after axonal injury, starting with disinte-gration of most axons (3-4 days post injury) followed by myelin debris formation and removal313.4. Frequency Changes in DC-Tx Evaluated Along the Spinal CordFigure 3.7: Frequency maps of rat spinal cord dorsal column cranial (A, B) and caudal (D, E)to injury site (C). Distance from injury labelled above. Fasciculus Gracilis marked in orange,Corticospinal Tract marked in green.Figure 3.8: Frequency shifts values bilateral to injury. Frequency shift values shown for cranial(+8mm) to caudal (-8mm) along healthy (black), 3-weeks (dark), and 8-weeks (light) post-injuryspinal cords for the Fasciculus Gracilis (A, orange) and Corticospinal Tract (B, green). Injury sitelocated at 0mm (red highlight), with severely damaged adjacent slices (yellow highlight). Areasof Wallerian and retrograde degeneration for each tract are labelled.323.5. Applications to Assessment of Spinal Cord Injury and Repair and Future Workover several weeks. This is reflected in the consistent frequency shifts measured along the path ofWallerian degeneration in the FG (1×10−6 ppm decrease, + 2mm to + 8mm) and CST (1×10−6ppm increase, -2mm to -8mm) at 3-weeks post injury (Fig. 3.8 A, B, darker line, Wallerian). At8-weeks, frequency shifts in both the FG and CST are significantly decreased compared to controland 3-week post injury values (Fig. 3.8 A, B, lighter line, Wallerian).Retrograde degeneration is a relatively slower process compared to Wallerian degeneration,characterized by shrinking of axons (axonal die-back), leaving some myelin debris in its wake.Frequency shifts measured along retrograde degeneration paths in the FG (+ 2mm to + 8mm)and CST (- 2mm to - 8mm) are similar to baseline values at up to 3 mm from the injury site,then show increased frequency (around 2*10−6 ppm) consistently for the remaining length, whichpersists at 8-weeks post injury (Fig. 3.8 A, B, Retrograde).Both types degeneration are mediated by inflammatory agents (microglia, astrocytes, andmacrophages) that have been known to proliferate at different rates in each tract [87]. TheCST microstructure has a smaller inter-axonal distance compared with the FG, and may restrictdiffusion of these agents. In turn, degeneration of myelin and its subsequent removal may occurat different rates in the FG compared to CST in the Wallerian degeneration case, leading to theobserved difference in frequency shifts at 3-weeks post injury (decreases for FG, increases for CST).Due to the relatively slow progression of retrograde degeneration, proliferation of inflammatoryagents is (most likely) consistent in both tracts, producing similar frequency shift patterns (Fig.3.8 A, B, Retrograde).Our results, (especially at ±5mm from the injury site), agree with previous studies thatcorrelated MR of spinal cord injury with histology for myelin and axons [38]. While a clearpattern is seen for frequency shifts in the retrograde degeneration of the FG and CST, thepattern is less distinct in the two tracts for Wallerian degeneration. However, these resultsprovide characterization of both types of degeneration that are distinct and differentiable usingfrequency shift mapping. Furthermore, this study suggests that 3D GRE imaging of longer spinalcord segments can be adapted for in-vivo imaging in future studies to provide more significantinsights into spinal cord injury.3.5 Applications to Assessment of Spinal Cord Injury andRepair and Future WorkOur results demonstrate frequency shift mapping as a potential technique capable of assess WMdamage and repair non-invasively with high sensitivity. We show that in agreement with theoreti-cal predictions [92] and studies in multiple sclerosis [90], changes in neuronal microstructure fromthe axon cytoskeleton and myelin sheaths are contributors to frequency shifts. The relativelyfast and robust 3D-MGE method used in this study produces high-spatial resolution images ofex-vivo spinal cord with excellent WM/GM contrast, and overcomes some previous difficultieswith observing partial volume effects in small DC tracts with diffusion tensor imaging [38] [37].In comparison to other studies investigating WM frequency shifts, our approach uses a cleanenvironment of neuronal degeneration with minimal confounding factors. The area of degenera-tion studied is far enough from the transection site to avoid effects of hemorrhage and mechanicalinjury. Prussian blue iron staining of spinal cord sections showed negligible amounts of iron inWM (data not shown). The cylindrical structure of spinal cord is ideal, as placement parallel333.5. Applications to Assessment of Spinal Cord Injury and Repair and Future Workto B0 is critical for zero frequency shift contribution from the WM fibers. Ex-vivo samples wereimmersed in paraformaldehyde (PF) to minimize frequency artifacts from air-tissue boundaries[67] [74]. Although offsets in frequency shifts due to fixation [49] were present, such offsets aremeasured and remain consistent across all healthy and damaged cords (Fig. 3.4).Histology remains the gold standard against which MRI parameters are validated. Our re-sults use eriochrome and myelin basic protein biomarkers, both widely used as a measure ofoverall myelin content and known to correlate well with certain MRI parameters for measuringmyelin [39]. In addition to showing frequency shift correlations with myelin, axonal, and mi-croglial biomarkers, we also produce evidence that frequency shifts correlated with degen-MBPin Wallerian FG (R2=0.7287) and retrograde CST (R2=0.7287), which is the first such finding ofits kind. Compared to other measures, degen-MBP correlates strongest with ED1 (r = 0.7315,p < 0.0001) across all data points regardless of degeneration mechanism or time-point, indicat-ing an increased microglial activity is tied to closely to formation of fragmented myelin sheaths.Concurrently, such fragmentation is a key microstructural change that increases frequency shiftsin FG undergoing Wallerian degeneration, though we believe the presence of axons, as well asanatomical differences between the FG and CST, are contributing factors to the frequency shiftdecrease observed in retrograde degeneration and caudal CST.Current understanding of white matter disease pathologies and their relation to clinical MRparameters is limited by difficulty in obtaining histology for comparative assessment. With theDC-Tx model, we correlate quantitative measures of neuronal integrity with MR frequency shiftsensitivity to several patterns of microstructural damage in spinal cord. The DC-Tx is a widelyused and well characterized model of spinal cord injury, one of several other well-known models,such as contusion and compression [11]. In addition, recent work has shown that degenerationfrom secondary injury is structurally similar in-vivo and ex-vivo, suggesting viability of assessingmultiple injury models in-vivo[39].Concurrent myelin and axonal degeneration in MS lesions can result in an overall loss offrequency contrast [92]. Unlike in MS lesions, macrophages act more rapidly in the DC-Tx injurymodel, making assessment of non-bulk and microstructure-only related changes in WM difficultto assess in rat spinal cord. Rather, we note two interesting cases where frequency shifts remainunchanged in the cranial FG, from healthy to 3-wpi, and in the caudal FG, from 3-wpi to 8-wpi.Further work developing frequency shift mapping as a viable clinical method requires thisambiguities to be resolved without comparative histology. A companion method such as myelinwater imaging or diffusion tensor imaging (DTI) can provide additional information. In a previousstudy, MWI and DTI of Wallerian degeneration at healthy to 3-wpi in the FG did show measurabledifferences, however these differences were not present in retrograde degeneration at 3 to 8-wpi[37]. Combined with the lack of strong correlation between MWI with degen-MBP, our resultssuggest frequency shift mapping brings a higher degree of sensitivity to tissue microstructure inquantitative MR assessment of white matter injury.Though this study does not assess recovery post-injury, we aim to expand the scope of futureexperiments to include frequency shift mapping of remyelination, which could reveal potentialsensitivity to WM repair activity. There is evidence of stronger dieback at 5mm after 16 weeks(Seif et al., 2007) which may reveal later stages of Wallerian degeneration. EM of demyelinationis similar to damage observed in EM of mild traumatic brain injury. Adapting frequency shiftmapping to monitoring TBI white matter disruption can be a possibility, though it is madeharder with presence of iron from hemorrhahges related to the TBI [21]. Our observation of someremyelination beginning to appear (Fig. 3.2) marks a significant point of interest for assessing343.5. Applications to Assessment of Spinal Cord Injury and Repair and Future Workrepair and SCI recovery, and to investigate frequency shifts sensitivity for re-myelination.Our data agree with established theories regarding WM microstructure and its role in MRphase contrast of CNS tissue in brain, and suggests that frequency mapping can complementcurrent clinical methods as a non-invasive and robust way of assessing both acute and long-termWM damage in spinal cord injury patients [9]. We demonstrate that these maps can be acquiredex-vivo at resolutions high enough to separate individual tract bundles. More significantly, we findstronger correlations between frequency shifts and myelin integrity than observed in similar stud-ies using MWI [37][38]. With modifications, frequency shift mapping with 3D-MGE can becomea viable alternative to MWI for SCI assessment, offering advantages in spatial resolution, shorterscan times, motion correction, in-vivo imaging, and for potential applications using quantitativesusceptibility mapping [67][68][24][45]. The potential uses of frequency shift mapping in trackingpathology and for revealing underlying damage-related microstructural changes can bring aboutnew insight into numerous WM diseases, including demyelination in Parkinson’s and Alzheimers,and evolution of MS lesions. The implication of having a non-invasive way of quantifying andmonitoring recovery will contribute greatly to evaluating effectiveness of therapeutic interventions[40], and to further understanding of repair mechanisms.35Bibliography[1] Shweta Aggarwal, Larisa Yurlova, and Mikael Simons. Central nervous system myelin: struc-ture, synthesis and assembly. Trends Cell Biol, 21(10):585–593, Oct 2011.[2] B. A. Barres, M. D. Jacobson, R. Schmid, M. Sendtner, and M. C. Raff. Does oligodendrocytesurvival depend on axons? Curr Biol, 3(8):489–497, Aug 1993.[3] P. J. Basser, J. Mattiello, and D. LeBihan. Mr diffusion tensor spectroscopy and imaging.Biophys J, 66(1):259–267, Jan 1994.[4] Bogdan Beirowski, Robert Adalbert, Diana Wagner, Daniela S. Grumme, Klaus Addicks,Richard R. Ribchester, and Michael P. Coleman. The progressive nature of wallerian de-generation in wild-type and slow wallerian degeneration (wlds) nerves. BMC Neurosci, 6:6,2005.[5] Larry I. Benowitz and Phillip G. Popovich. Inflammation and axon regeneration. Curr OpinNeurol, 24(6):577–583, Dec 2011.[6] W. F. BLAKEMORE. Invasion of schwann cells into the spinal cord of the rat followinglocal injections of lysolecithin. Neuropathology and Applied Neurobiology, 2(1):21–39, 1976.[7] F. Bloch. Nuclear induction. Phys Rev, 7:460–474, 1946.[8] R. Bosma and Patrick W. Stroman. Diffusion tensor imaging in the human spinal cord:development, limitations, and clinical applications. Crit Rev Biomed Eng, 40(1):1–20, 2012.[9] Anthony Bozzo, Judith Marcoux, Mohan Radhakrishna, Julie Pelletier, and Benoit Goulet.The role of magnetic resonance imaging in the management of acute spinal cord injury.Journal of neurotrauma, 28(8):1401–1411, 2011.[10] Way Cherng Chen, Sean Foxley, and Karla L Miller. Detecting microstructural properties ofwhite matter based on compartmentalization of magnetic susceptibility. Neuroimage, 70:1–9,2013.[11] T Cheriyan, DJ Ryan, JH Weinreb, J Cheriyan, JC Paul, V Lafage, T Kirsch, and TJ Errico.Spinal cord injury models: a review. Spinal cord, 52(8):588–595, 2014.[12] Matthew M. Cheung, Darrell T H. Li, Edward S. Hui, Shujuan Fan, Abby Y. Ding, YongHu, and Ed X. Wu. In vivo diffusion tensor imaging of chronic spinal cord compression inrat model. Conf Proc IEEE Eng Med Biol Soc, 2009:2715–2718, 2009.[13] Chih-Liang Chin, Felix W Wehrli, Yingli Fan, Scott N Hwang, Eric D Schwartz, JonathanNissanov, and David B Hackney. Assessment of axonal fiber tract architecture in excisedrat spinal cord by localized nmr q-space imaging: Simulations and experimental studies.Magnetic resonance in medicine, 52(4):733–740, 2004.36Bibliography[14] Chris A. Clark and David J. Werring. Diffusion tensor imaging in spinal cord: methods andapplications - a review. NMR Biomed, 15(7-8):578–586, 2002.[15] T. E. Conturo, N. F. Lori, T. S. Cull, E. Akbudak, A. Z. Snyder, J. S. Shimony, R. C.McKinstry, H. Burton, and M. E. Raichle. Tracking neuronal fiber pathways in the livinghuman brain. Proc Natl Acad Sci U S A, 96(18):10422–10427, Aug 1999.[16] Cynthia A. DeBoy, Jiangyang Zhang, Sonny Dike, Irina Shats, Melina Jones, Daniel S. Reich,Susumu Mori, Thien Nguyen, Brian Rothstein, Robert H. Miller, John T. Griffin, Douglas A.Kerr, and Peter A. Calabresi. High resolution diffusion tensor imaging of axonal damage infocal inflammatory and demyelinating lesions in rat spinal cord. Brain, 130(Pt 8):2199–2210,Aug 2007.[17] Christian Denk, Enedino Hernandez Torres, Alex MacKay, and Alexander Rauscher. Theinfluence of white matter fibre orientation on mr signal phase and decay. NMR Biomed,24(3):246–252, Apr 2011.[18] Christian Denk and Alexander Rauscher. Susceptibility weighted imaging with multipleechoes. J Magn Reson Imaging, 31(1):185–191, Jan 2010.[19] Aparna A. Deo, Raymond J. Grill, Khader M. Hasan, and Ponnada A. Narayana. In vivoserial diffusion tensor imaging of experimental spinal cord injury. J Neurosci Res, 83(5):801–810, Apr 2006.[20] Suhayl Dhib-Jalbut, Douglas L. Arnold, Don W. Cleveland, Mark Fisher, Robert M. Fried-lander, M Maral Mouradian, Serge Przedborski, Bruce D. Trapp, Tony Wyss-Coray, andV Wee Yong. Neurodegeneration and neuroprotection in multiple sclerosis and other neu-rodegenerative diseases. J Neuroimmunol, 176(1-2):198–215, Jul 2006.[21] Virginia Donovan, Claudia Kim, Ariana K Anugerah, Jacqueline S Coats, Udochuwku Oyoyo,Andrea C Pardo, and Andre Obenaus. Repeated mild traumatic brain injury results in long-term white-matter disruption. Journal of Cerebral Blood Flow & Metabolism, 34(4):715–723,2014.[22] P. Douek, R. Turner, J. Pekar, N. Patronas, and D. Le Bihan. Mr color mapping of myelinfiber orientation. J Comput Assist Tomogr, 15(6):923–929, 1991.[23] Adrienne N Dula, Daniel F Gochberg, Holly L Valentine, William M Valentine, and Mark DDoes. Multiexponential t2, magnetization transfer, and quantitative histology in white mat-ter tracts of rat spinal cord. Magnetic Resonance in Medicine, 63(4):902–909, 2010.[24] Jeff H. Duyn, Peter van Gelderen, Tie-Qiang Li, Jacco A. de Zwart, Alan P. Koretsky, andMasaki Fukunaga. High-field mri of brain cortical substructure based on signal phase. ProcNatl Acad Sci U S A, 104(28):11796–11801, Jul 2007.[25] Jozef H. Duyn. High-field mri of brain iron. Methods Mol Biol, 711:239–249, 2011.[26] R.R. Ernst and W.A. Anderson. Application of fourier transform spectroscopy to magneticresonance. Rev sci instr, 37:93–102, 1965.37Bibliography[27] C. C. Ford, T. L. Ceckler, J. Karp, and R. M. Herndon. Magnetic resonance imaging ofexperimental demyelinating lesions. Magn Reson Med, 14(3):461–481, Jun 1990.[28] J. C. Ford, D. B. Hackney, D. C. Alsop, H. Jara, P. M. Joseph, C. M. Hand, and P. Black.Mri characterization of diffusion coefficients in a rat spinal cord injury model. Magn ResonMed, 31(5):488–494, May 1994.[29] M. Fukunaga, T. Q. Li, P. van Gelderen, J. A. de Zwart, K. Shmueli, B. Yao, J. Lee,D. Maric, M. A. Aronova, G. Zhang, R. D. Leapman, J. F. Schenck, H. Merkle, and J. H.Duyn. Layer-specific variation of iron content in cerebral cortex as a source of MRI contrast.Proc. Natl. Acad. Sci. U.S.A., 107(8):3834–3839, Feb 2010.[30] John C. Gensel, Satoshi Nakamura, Zhen Guan, Nico van Rooijen, Daniel P. Ankeny, andPhillip G. Popovich. Macrophages promote axon regeneration with concurrent neurotoxicity.J Neurosci, 29(12):3956–3968, Mar 2009.[31] Rachel George and John W Griffin. Delayed macrophage responses and myelin clearanceduring wallerian degeneration in the central nervous system: the dorsal radiculotomy model.Experimental neurology, 129(2):225–236, 1994.[32] Henry Gray. Anatomy of the human body. Lea & Febiger, 20th edition, 1918.[33] V Gulani, AG Webb, ID Duncan, and PC Lauterbur. Apparent diffusion tensor measure-ments in myelin-deficient rat spinal cords. Magnetic Resonance in Medicine, 45(2):191–195,2001.[34] Lukas Haider, Marie T Fischer, Josa M Frischer, Jan Bauer, Romana Ho¨ftberger, Gergo¨Botond, Harald Esterbauer, Christoph J Binder, Joseph L Witztum, and Hans Lassmann.Oxidative damage in multiple sclerosis lesions. Brain, 134(7):1914–1924, 2011.[35] Xiang He, Dmitriy A. Yablonskiy, and Marcus E. Raichle. Biophysical mechanisms of phasecontrast in gradient echo mri. Proceedings of the National Academy of Sciences of the UnitedStates of America, 106(32):pp. 13558–13563, 2009.[36] R. M. Henkelman, G. J. Stanisz, and S. J. Graham. Magnetization transfer in mri: a review.NMR Biomed, 14(2):57–64, Apr 2001.[37] Piotr Kozlowski, Jie Liu, Andrew C. Yung, and Wolfram Tetzlaff. High-resolution myelinwater measurements in rat spinal cord. Magn Reson Med, 59(4):796–802, Apr 2008.[38] Piotr Kozlowski, Disha Raj, Jie Liu, Clarrie Lam, Andrew C. Yung, and Wolfram Tetzlaff.Characterizing white matter damage in rat spinal cord with quantitative mri and histology.J Neurotrauma, 25(6):653–676, Jun 2008.[39] Piotr Kozlowski, Paulina Rosicka, Jie Liu, Andrew C Yung, and Wolfram Tetzlaff. In vivolongitudinal myelin water imaging in rat spinal cord following dorsal column transectioninjury. Magnetic resonance imaging, 32(3):250–258, 2014.[40] Brian K Kwon, Lali H Sekhon, and Michael G Fehlings. Emerging repair, regeneration, andtranslational research advances for spinal cord injury. Spine, 35(21S):S263–S270, 2010.38Bibliography[41] C. Laule, I. M. Vavasour, G R W. Moore, J. Oger, D K B. Li, D. W. Paty, and A. L. MacKay.Water content and myelin water fraction in multiple sclerosis. a t2 relaxation study. J Neurol,251(3):284–293, Mar 2004.[42] Jongho Lee, Karin Shmueli, Byeong-Teck Kang, Bing Yao, Masaki Fukunaga, Peter vanGelderen, Sara Palumbo, Francesca Bosetti, Afonso C. Silva, and Jeff H. Duyn. The contri-bution of myelin to magnetic susceptibility-weighted contrasts in high-field mri of the brain.Neuroimage, 59(4):3967–3975, Feb 2012.[43] Wei Li, Bing Wu, Alexandru V. Avram, and Chunlei Liu. Magnetic susceptibility anisotropyof human brain in vivo and its molecular underpinnings. Neuroimage, 59(3):2088–2097, Feb2012.[44] C. Liu, W. Li, G. A. Johnson, and B. Wu. High-field (9.4 T) MRI of brain dysmyelinationby quantitative mapping of magnetic susceptibility. Neuroimage, 56(3):930–938, Jun 2011.[45] Chunlei Liu, Wei Li, Karen A Tong, Kristen W Yeom, and Samuel Kuzminski. Susceptibility-weighted imaging and quantitative susceptibility mapping in the brain. Journal of MagneticResonance Imaging, 2014.[46] Lorentz. The Theory of Electrons. Leipzig, 1909.[47] David N. Loy, Joong Hee Kim, Mingqiang Xie, Robert E. Schmidt, Kathryn Trinkaus, andSheng-Kwei Song. Diffusion tensor imaging predicts hyperacute spinal cord injury severity.J Neurotrauma, 24(6):979–990, Jun 2007.[48] L. Lubinska. Early course of wallerian degeneration in myelinated fibres of the rat phrenicnerve. Brain Res, 130(1):47–63, Jul 1977.[49] J. Luo, X. He, and D. A. Yablonskiy. Magnetic susceptibility induced white matter MRsignal frequency shifts-experimental comparison between Lorentzian sphere and generalizedLorentzian approaches. Magn Reson Med, May 2013.[50] Jie Luo, Xiang He, D Andre’ d’Avignon, Joseph J H. Ackerman, and Dmitriy A. Yablonskiy.Protein-induced water 1h mr frequency shifts: contributions from magnetic susceptibilityand exchange effects. J Magn Reson, 202(1):102–108, Jan 2010.[51] A. L. MacKay, I. M. Vavasour, A. Rauscher, S. H. Kolind, B. Mdler, G R W. Moore, A. L.Traboulsee, D K B. Li, and C. Laule. Mr relaxation in multiple sclerosis. Neuroimaging ClinN Am, 19(1):1–26, Feb 2009.[52] Saaussan Madi, Khader M. Hasan, and Ponnada A. Narayana. Diffusion tensor imaging ofin vivo and excised rat spinal cord at 7 t with an icosahedral encoding scheme. Magn ResonMed, 53(1):118–125, Jan 2005.[53] A Matsuo, GC Lee, K Terai, K Takami, WF Hickey, EG McGeer, and PL McGeer. Unmask-ing of an unusual myelin basic protein epitope during the process of myelin degeneration inhumans: a potential mechanism for the generation of autoantigens. The American journalof pathology, 150(4):1253, 1997.39Bibliography[54] Kelly C. McPhee, Christian Denk, Zeinab Al-Rekabi, and Alexander Rauscher. Bilateralfiltering of magnetic resonance phase images. Magn Reson Imaging, 29(7):1023–1029, Sep2011.[55] Evan P. Minty, Thorarin A. Bjarnason, Cornelia Laule, and Alex L. MacKay. Myelin watermeasurement in the spinal cord. Magn Reson Med, 61(4):883–892, Apr 2009.[56] G. R. Moore, E. Leung, A. L. MacKay, I. M. Vavasour, K. P. Whittall, K. S. Cover, D. K.Li, S. A. Hashimoto, J. Oger, T. J. Sprinkle, and D. W. Paty. A pathology-mri study of theshort-t2 component in formalin-fixed multiple sclerosis brain. Neurology, 55(10):1506–1510,Nov 2000.[57] S. Mori, B. J. Crain, V. P. Chacko, and P. C. van Zijl. Three-dimensional tracking of axonalprojections in the brain by magnetic resonance imaging. Ann Neurol, 45(2):265–269, Feb1999.[58] Iska Moxon-Emre. White matter damage and inflammation in rat models of ischemic andhemorrhagic stroke. PhD thesis, University of Toronto, 2010.[59] D. C. Noll, D. G. Nishimura, and A. Macovski. Homodyne detection in magnetic resonanceimaging. IEEE Trans Med Imaging, 10(2):154–163, 1991.[60] Michael D. Norenberg, Jon Smith, and Alex Marcillo. The pathology of human spinal cordinjury: defining the problems. J Neurotrauma, 21(4):429–440, Apr 2004.[61] R. J. Ogg, J. W. Langston, E. M. Haacke, R. G. Steen, and J. S. Taylor. The correlationbetween phase shifts in gradient-echo mr images and regional brain iron concentration. MagnReson Imaging, 17(8):1141–1148, Oct 1999.[62] C. Oppenheim, D. Ducreux, S. Rodrigo, J. Hodel, T. Tourdias, F. Charbonneau, S. Pier-refitte, and Jf Meder. [diffusion tensor imaging and tractography of the brain and spinalcord]. J Radiol, 88(3 Pt 2):510–520, Mar 2007.[63] Charles Aidemise Oyinbo. Secondary injury mechanisms in traumatic spinal cord injury: anugget of this multiply cascade. Acta Neurobiol Exp (Wars), 71(2):281–299, 2011.[64] C. S. Poon and R. M. Henkelman. Practical t2 quantitation for clinical applications. J MagnReson Imaging, 2(5):541–553, 1992.[65] C. Poupon, C. A. Clark, V. Frouin, J. Rgis, I. Bloch, D. Le Bihan, and J. Mangin. Regu-larization of diffusion-based direction maps for the tracking of brain white matter fascicles.Neuroimage, 12(2):184–195, Aug 2000.[66] Alexander Rauscher, Markus Barth, Jrgen R. Reichenbach, Rudolf Stollberger, and EwaldMoser. Automated unwrapping of mr phase images applied to bold mr-venography at 3 tesla.J Magn Reson Imaging, 18(2):175–180, Aug 2003.[67] Alexander Rauscher, Jan Sedlacik, Markus Barth, Hans-Joachim Mentzel, and Jrgen R.Reichenbach. Magnetic susceptibility-weighted mr phase imaging of the human brain. AJNRAm J Neuroradiol, 26(4):736–742, Apr 2005.40Bibliography[68] Alexander Rauscher, Jan Sedlacik, Andreas Deistung, Hans-Joachim Mentzel, and Jrgen R.Reichenbach. Susceptibility weighted imaging: data acquisition, image reconstruction andclinical applications. Z Med Phys, 16(4):240–250, 2006.[69] David A Rudko, L Martyn Klassen, Sonali N de Chickera, Joseph S Gati, Gregory A Dekaban,and Ravi S Menon. Origins of orientation dependence in gray and white matter. Proceedingsof the National Academy of Sciences, 111(1):E159–E167, 2014.[70] Eric D. Schwartz, Emily T. Cooper, Yingli Fan, Abbas F. Jawad, Chih-Liang Chin, JonathanNissanov, and David B. Hackney. Mri diffusion coefficients in spinal cord correlate with axonmorphometry. Neuroreport, 16(1):73–76, Jan 2005.[71] Ferdinand Schweser, Andreas Deistung, Berengar Wendel Lehr, and Ju¨rgen Rainer Reichen-bach. Quantitative imaging of intrinsic magnetic tissue properties using mri signal phase:an approach to in vivo brain iron metabolism? Neuroimage, 54(4):2789–2807, 2011.[72] Gamal I Seif, Hiroshi Nomura, and Charles H Tator. Retrograde axonal degeneration(dieback) in the corticospinal tract after transection injury of the rat spinal cord: a con-focal microscopy study. Journal of neurotrauma, 24(9):1513–1528, 2007.[73] L. H. Sekhon and M. G. Fehlings. Epidemiology, demographics, and pathophysiology ofacute spinal cord injury. Spine (Phila Pa 1976), 26(24 Suppl):S2–12, Dec 2001.[74] K. Shmueli, J. A. de Zwart, P. van Gelderen, T. Q. Li, S. J. Dodd, and J. H. Duyn. Magneticsusceptibility mapping of brain tissue in vivo using MRI phase data. Magn Reson Med,62(6):1510–1522, Dec 2009.[75] Karin Shmueli, Stephen J. Dodd, Tie-Qiang Li, and Jeff H. Duyn. The contribution ofchemical exchange to mri frequency shifts in brain tissue. Magn Reson Med, 65(1):35–43,Jan 2011.[76] Sheng-Kwei Song, Shu-Wei Sun, Won-Kyu Ju, Shiow-Jiuan Lin, Anne H. Cross, andArthur H. Neufeld. Diffusion tensor imaging detects and differentiates axon and myelindegeneration in mouse optic nerve after retinal ischemia. Neuroimage, 20(3):1714–1722, Nov2003.[77] Sheng-Kwei Song, Shu-Wei Sun, Michael J. Ramsbottom, Chen Chang, John Russell, andAnne H. Cross. Dysmyelination revealed through mri as increased radial (but unchangedaxial) diffusion of water. Neuroimage, 17(3):1429–1436, Nov 2002.[78] Sheng-Kwei Song, Jun Yoshino, Tuan Q. Le, Shiow-Jiuan Lin, Shu-Wei Sun, Anne H. Cross,and Regina C. Armstrong. Demyelination increases radial diffusivity in corpus callosum ofmouse brain. Neuroimage, 26(1):132–140, May 2005.[79] David P Stirling, Kourosh Khodarahmi, Jie Liu, Lowell T McPhail, Christopher B McBride,John D Steeves, Matt S Ramer, and Wolfram Tetzlaff. Minocycline treatment reduces delayedoligodendrocyte death, attenuates axonal dieback, and improves functional outcome afterspinal cord injury. The Journal of neuroscience, 24(9):2182–2190, 2004.[80] Guido Stollg and Sebastian Jander. The role of microglia and macrophages in the patho-physiology of the cns. Progress in neurobiology, 58(3):233–247, 1999.41Bibliography[81] Alexander L Sukstanskii and Dmitriy A Yablonskiy. On the role of neuronal magnetic suscep-tibility and structure symmetry on gradient echo mr signal formation. Magnetic Resonancein Medicine, 71(1):345–353, 2014.[82] Ben G Szaro, Mark H Whitnall, and Harold Gainer. Phosphorylation-dependent epitopeson neurofilament proteins and neurofilament densities differ in axons in the corticospinaland primary sensory dorsal column tracts in the rat spinal cord. Journal of comparativeneurology, 302(2):220–235, 1990.[83] CH Toh, K-C Wei, C-N Chang, P-W Hsu, H-F Wong, S-H Ng, M Castillo, and C-P Lin. Dif-ferentiation of pyogenic brain abscesses from necrotic glioblastomas with use of susceptibility-weighted imaging. American Journal of Neuroradiology, 33(8):1534–1538, 2012.[84] Yanina Tsenkina. Behavioural, genetic and epigenetic determinants of white matter pathol-ogy in a new mouse model of chronic cerebral hypoperfusion. 2013.[85] Clare K. Underwood, Nyoman D. Kurniawan, Tim J. Butler, Gary J. Cowin, and Robyn H.Wallace. Non-invasive diffusion tensor imaging detects white matter degeneration in thespinal cord of a mouse model of amyotrophic lateral sclerosis. Neuroimage, 55(2):455–461,Mar 2011.[86] Jack T. Wang, Zachary A. Medress, and Ben A. Barres. Axon degeneration: molecularmechanisms of a self-destruction pathway. J Cell Biol, 196(1):7–18, Jan 2012.[87] Li Wang, Bing Hu, Wai M. Wong, Peihua Lu, Wutian Wu, and Xiao-Ming Xu. Glial andaxonal responses in areas of wallerian degeneration of the corticospinal and dorsal ascendingtracts after spinal cord dorsal funiculotomy. Neuropathology, 29(3):230–241, Jun 2009.[88] Samuel Wharton and Richard Bowtell. Effects of white matter microstructure on phase andsusceptibility maps. Magnetic Resonance in Medicine, 2013.[89] K. P. Whittall, A. L. MacKay, D. A. Graeb, R. A. Nugent, D. K. Li, and D. W. Paty. In vivomeasurement of t2 distributions and water contents in normal human brain. Magn ResonMed, 37(1):34–43, Jan 1997.[90] V. Wiggermann, E. Hernandez Torres, I. M. Vavasour, G. R. Moore, C. Laule, A. L. Mackay,D. K. Li, A. Traboulsee, and A. Rauscher. Magnetic resonance frequency shifts during acuteMS lesion formation. Neurology, 81(3):211–218, Jul 2013.[91] Santiago Ramo´n y Cajal. Cajal’s degeneration and regeneration of the nervous system.Oxford University Press, 1991.[92] D. A. Yablonskiy, J. Luo, A. L. Sukstanskii, A. Iyer, and A. H. Cross. Biophysical mecha-nisms of MRI signal frequency contrast in multiple sclerosis. Proc. Natl. Acad. Sci. U.S.A.,109(35):14212–14217, Aug 2012.[93] Dmitriy A Yablonskiy and Alexander L Sukstanskii. Biophysical mechanisms of myelin-induced water frequency shifts. Magnetic Resonance in Medicine, 71(6):1956–1958, 2014.[94] Dmitriy A Yablonskiy and Alexander L Sukstanskii. Generalized lorentzian tensor approach(glta) as a biophysical background for quantitative susceptibility mapping. Magnetic Reso-nance in Medicine, 73(2):757–764, 2015.42Bibliography[95] I. R. Young, I. J. Cox, D. J. Bryant, and G. M. Bydder. The benefits of increasing spatialresolution as a means of reducing artifacts due to field inhomogeneities. Magn Reson Imaging,6(5):585–590, 1988.43


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items