Open Collections

UBC Undergraduate Research

Investigation of damaged tibiofemoral articular cartilage through susceptibility-weighted MR phase imaging… Yuen, Joanna Apr 30, 2014

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

Item Metadata


52966-Yuen_Joanna_PHYS_449_2014.pdf [ 1.46MB ]
JSON: 52966-1.0085972.json
JSON-LD: 52966-1.0085972-ld.json
RDF/XML (Pretty): 52966-1.0085972-rdf.xml
RDF/JSON: 52966-1.0085972-rdf.json
Turtle: 52966-1.0085972-turtle.txt
N-Triples: 52966-1.0085972-rdf-ntriples.txt
Original Record: 52966-1.0085972-source.json
Full Text

Full Text

THE UNIVERSITY OF BRITISH COLUMBIAInvestigation of Damaged TibiofemoralArticular Cartilage throughSusceptibility-Weighted MR PhaseImaging and Frequency MappingbyJoanna YuenA thesis submitted in partial fulfillment for thedegree of Bachelor of Science, Honours Biophysicsin theFaculty of ScienceDepartment of Physics and AstronomyApril 2014c© Joanna Yuen 2014“Out of clutter, find simplicity.From discord, find harmony.In the middle of difficulty lies opportunity”Albert EinsteinTHE UNIVERSITY OF BRITISH COLUMBIAAbstractFaculty of ScienceDepartment of Physics and AstronomyBachelor of Scienceby Joanna YuenDegenerative articular cartilage damage, such as osteoarthritis, is one of the most com-mon causes of chronic disability, so early detection and accurate assessment is essentialfor effective treatment. Magnetic resonance imaging, a non-invasive imaging modality,provides excellent soft tissue contrast in the knee and enables visualization of all zonesof cartilage. In this project, a 3D multiple echo gradient echo sequence was used toassess articular cartilage damage in patients with osteoarthritis or other cartilage dis-orders. Frequency maps and susceptibility weighted images were generated to examinethe anisotropic and isotropic orientation of collagen fibres and the presence of cartilagedamage. It is anticipated that these techniques could be a more sensitive method ofdetecting changes in cartilage structure than current clinical procedures.PrefaceMy administrative contributions to this project include writing ethics amendments forsubmission and approval by the University of British Columbia (UBC) Clinical ResearchEthics Board and modifying and editing the MRI protocol proposal for review by theUBC MRI Research Centre Protocol Proposal Committee.I was also responsible for the clinical coordination for this project, which included di-rectly contacting participants, booking MR time, reviewing contraindictions, and ob-taining research consent. Two control participants and three patients were recruitedthrough me. The remaining patients were recruited through Dr. Robert McCormack’sclinical practice in New Westminister.All MRI scans, except the first four control data sets, were conducted by the MRI tech-nologists with me being present to provide direction and feedback under Dr. AlexanderRauscher’s guidance.All data analysis and figures including MRI images and segmentation (unless otherwisespecified) were conducted and produced by me under the supervision of Dr. AlexanderRauscher. Certain parts of the post-processing algorithms were provided by current andprevious lab members and were modified by me for this project.iiiContentsAbstract iiPreface iiiList of Figures viList of Tables viiAbbreviations viiiAcknowledgements ix1 Introduction 11.1 Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . . . . . . . . . 11.1.1 Magnetization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1.2 Relaxation Processes . . . . . . . . . . . . . . . . . . . . . . . . . . 41.1.3 Spatial Localization . . . . . . . . . . . . . . . . . . . . . . . . . . 61.1.4 Pulse Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.1.5 Image Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . 101.2 Phase and Susceptibility-Weighted Imaging . . . . . . . . . . . . . . . . . 101.3 Knee Articular Cartilage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121.3.1 Healthy Cartilage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131.3.2 Cartilage Damage and Degeneration . . . . . . . . . . . . . . . . . 141.3.3 Methods of Diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . 161.4 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 Methods 212.1 Recruitment of Study Participants . . . . . . . . . . . . . . . . . . . . . . 212.2 MRI Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.3 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.3.1 T2 and T2∗maps, T1-weighted Images, STIR Sequences . . . . . . 232.3.2 Post-Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243 Results and Discussion 273.1 Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27ivContents v3.1.1 Protocol and Post-Processing Development . . . . . . . . . . . . . 273.1.2 Comparison of T2, T2∗, and Frequency Maps, Magnitude Images,and SWI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.2 Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.3 Other Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344 Conclusion 375 Future Work 38A Rescaling Factor Calculation 40Bibliography 41List of Figures1.1 Zeeman diagram showing splitting of the two spin states . . . . . . . . . . 21.2 Net magnetization of spins in the presence of an external magnetic field . 31.3 Radiofrequency pulse excitation . . . . . . . . . . . . . . . . . . . . . . . . 31.4 Longitudinal relaxation processes . . . . . . . . . . . . . . . . . . . . . . . 51.5 Transverse relaxation processes . . . . . . . . . . . . . . . . . . . . . . . . 61.6 Application of gradients to achieve spatial localization of nuclei . . . . . . 71.7 Spin echo sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.8 Simple gradient echo sequence . . . . . . . . . . . . . . . . . . . . . . . . . 91.9 3D gradient echo sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . 91.10 Anatomy of the knee joint . . . . . . . . . . . . . . . . . . . . . . . . . . . 121.11 Arcade model of collagen in healthy tibiofemoral articular cartilage . . . . 141.12 Current methods of diagnosing damaged knee cartilage . . . . . . . . . . . 172.1 Segmentation of cartilage . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.2 Post-processing methods to create susceptibility weighted images . . . . . 263.1 Magnitude images at 4 different resolutions . . . . . . . . . . . . . . . . . 283.2 Post-processed phase images at 4 different resolutions . . . . . . . . . . . 283.3 Magnitude images of 5 echoes . . . . . . . . . . . . . . . . . . . . . . . . . 293.4 Raw and post-processed phase images of 3 echoes . . . . . . . . . . . . . . 303.5 Comparison of T2, T2∗, and frequency maps, magnitude images, and SWI 313.6 Comparison of T2, T2∗, Frequency Maps, SWI, T1W, and STIR in patientwith characteristic signs of OA . . . . . . . . . . . . . . . . . . . . . . . . 333.7 Comparison of T2, T2∗, frequency maps, SWI, T1W, and STIR in patientwith suspected mild cartilage damage . . . . . . . . . . . . . . . . . . . . 333.8 Enhanced features of SWI in 2 patients . . . . . . . . . . . . . . . . . . . 353.9 Imaging artifacts caused by magnetic susceptibilities . . . . . . . . . . . . 36viList of Tables1.1 Tissue relaxation parameters . . . . . . . . . . . . . . . . . . . . . . . . . 51.2 Outerbridge arthroscopic scoring system . . . . . . . . . . . . . . . . . . . 17viiAbbreviationsBMI Body Mass IndexECM Extracellular MatrixFID Free Induction DecayFW Filter WidthFOV Field Of ViewFT Fourier TransformGAG GlycosaminoglycansGRE Gradient EchoIL-1 Interleukin-1LF Lorentz FactorMRI Magnetic Resonance ImagingNMR Nuclear Magnetic ResonanceOA OsteoarthritisRA Rheumatoid ArthritisRF RadiofrequencySNR Signal to Noise RatioSTIR Short Tau Inversion RecoverySWI Susceptibility Weighted ImagingSE Spin EchoTE Echo TimeTNF Tissue Necrosis FactorTR Repetition TimeUBC The University of British ColumbiaviiiAcknowledgementsThis project is a culmination of efforts made by several members of the UBC MRI Re-search Centre and UBC Department of Orthopaedics. Firstly, I would like to thank Dr.Alexander Rauscher, my supervisor, for giving me the opportunity to contribute to thisresearch. His steadfast patience, excellent mentorship, and encouragement throughoutthe project was greatly appreciated. I would like to acknowledge Vanessa Wiggermannfor the design of analytical methods. I would also like to thank Evan Chen and AndreaSpicthtinger for their patience and time in explaining concepts. Other members of theteam, Enedino Hernandez and Mike Jarrett, lent their support throughout the year.Furthermore, I am grateful for technical support from the MRI technologists, TraceyMaile and Alex Mazur. Lastly, I would like to thank Linda Chandler and Marie Punza-lan for their administrative support.This research would not be possible without the support and communication betweenour collaborators at the UBC Department of Orthopaedics, in particular Dr. Agnesd’Entremont and Dr. Robert McCormack for their input and guidance during the par-ticipant recruitment process.I am greatly appreciative of all study participants and their families for their contri-butions to this research as we hope to gain further understanding and to improve earlydiagnosis of this debilitating condition.I would also like to thank my friends and colleagues in Biophysics and Physics, An-nie Park, Melody Wong, Alice Liang, John Luu, Wonjin Kim, Nathan Lee, Lilly Li,Alice Chik, Euweng Chan, Jessie Fu, Lisa Wei, Samantha Tan, Sunny Li, and MatthewBundala for enriching my undergraduate experience and driving my academic inter-ests. Importantly, I would like to thank the kind-hearted and loyal Melissa Ng and thedown-to-earth and considerate Janine Reeves for providing moral support, transporta-tion assistance during holidays and weekends, and delicious baked products.Funding for this work was made possible by the Natural Sciences and Engineering Re-search Council of Canada (NSERC).ixDedicated to my parents for their care and support, my paw-pawfor encouraging me to strive for excellence, my sister for inspiringme to persevere and work hard for my dreams and goals, and mymentor Linda for demonstrating courage and resilience amidstinsurmountable odds.xChapter 1Introduction1.1 Magnetic Resonance ImagingMagnetic resonance imaging (MRI) is an imaging modality based on the phenomenon ofnuclear magnetic resonance (NMR), but with additional spatial localization [1]. Histor-ically, Bloch and Purcell independently described and demonstrated NMR in condensedmatter, both receiving the 1952 Nobel Prize in Physics. Bloch’s induction method, whichdominates modern NMR, involved arranging two coils orthogonal to a water sample, andapplying a strong static magnetic field perpendicular to both coils [2]. After applying analternating current in one coil (transmitter), the induction signal at a specific frequencycould be detected by the other (receiver). Lauterbur and Mansfield demonstrated NMR’spotential as an imaging technique and were awarded the 2003 Nobel Prize in Physiologyor Medicine. The use of Fourier analysis for NMR was explored by Ernst and Ander-son, and their contributions facilitated data collection, resulting in Ernst receiving the1991 Nobel Prize in Chemistry [3]. In modern MRI, three types of magnets are used:a static magnetic field within the bore of the scanner, oscillating radiofrequency pulses(RF), and gradient systems, changing magnetic fields. These cause spin magnetization,region-of-interest selection, and image generation.1.1.1 MagnetizationNMR is based on the quantum magnetic property of nuclear spin, which is exhibitedby atoms with an odd number of protons, neutrons, or both. Commonly used nucleiinclude 1H, 13C, 19F, 23Na, and 31P. The majority of in vivo imaging is based on theproton or hydrogen nucleus, 1H, as it is present in water and is abundant in biologicalsoft tissue. The basic volume unit of MRI is the voxel, which is approximately on the1Chapter 1. Introduction 2Figure 1.1: Zeeman diagram showing splitting of the two spin states. Due to theinteraction of the proton’s magnetic moment and external magnetic field, there are twopossible states for the proton, spin-up and spin-down. The population distribution isdescribed by Boltzmann’s distribution. Taken from [2].mm3 scale. Spins in a voxel are summed to yield a net macroscopic magnetizationvector. In the absence of an external magnetic field, B0, the nuclear magnetic momentsare randomly arranged. Macroscopically, nuclei with opposite orientations are cancelledout, and there is no net magnetization. As the magnetic field B0 is applied, the proton’smagnetic moment mz interacts with the field as follows:E = −mzB0 = −γjzB0 = −γ~sB0, (1.1)where E is the energy, γ is the gyromagnetic ratio characteristic to the atom, ~ isPlanck’s constant divided by 2pi, and s is the spin which exists as ±12 and correspondsto the quantum number for the z component of angular momentum jz.The spin-up state has parallel alignment of magnetic moments to the magnetic field andhas a lower energy than the spin-down state, in which magnetic moments are orientedanti-parallel. This splitting of states is known as Zeeman energy splitting and theprobability that the proton exists in either state is described by Boltzmann’s distribution(Figure 1.1):P (E) = Ce−E/kT , (1.2)where C is a constant, T is the absolute temperature, and k is the Boltzmann constant.Since orientation associated with the lower energy state is thermodynamically favoured,the spin-up and spin-down states are not equally populated. The ratio of the number ofspins in the lower states (N−) to higher states (N+) is:N−N+= e−∆E/kT (1.3)At equilibrium, there is a small excess of energetically-favourable spin-up states, resultingin a net magnetization aligned parallel to the field (Figure 1.2). The state of thermalequilibrium depends on the temperature and magnetic field strength, as illustrated inequations 1.2 and 1.1. At a lower temperature and higher magnetic field, there is aChapter 1. Introduction 3Figure 1.2: When there is no external magnetic field, spins are randomly orientedand cancelled out. Upon application of an external magnetic field, there is a net mag-netization of spins parallel to the magnetic field as this orientation is associated with alower energy state. Taken from [5].greater difference in energy between the two spin states, so fewer spins reach the higherspin state as more energy is required [4]. This results in the increasing excess of thespin-up state and higher MR signal.The spins rotate along the axis at the Larmor frequency, ωo, which is affected by themagnetic field strength (B0) and gyromagnetic ratio (γ) [6]. This movement is describedas precession or resonance, as illustrated in the following equation:ω0 = γB0, (1.4)The gyromagnetic ratio of hydrogen is 42.28MHzT−1, so for a 3.0 T MRI system, theLarmor frequency is 128 MHz.When an oscillating electromagnetic radiofrequency pulse (RF) from a transmit coil isapplied, RF excitation can occur if the pulse frequency is equal to the nuclei’s Larmorfrequency, and the magnetization vector will the tip away from its equilibrium position(Figure 1.3). The angle at which the magnetization vector tips is defined as the flipangle, and it depends on the amplitude and duration of the RF field.Figure 1.3: At equilibrium, the net magnetization, M0, is aligned along the z-axisin the direction of the external magnetic field. When an RF pulse is applied, themagnetization vector makes an angle α with the z-axis. The magnetization vector thenrotates around the axis. The magnetization can be split into components Mz and Mxy.Taken from [7].Chapter 1. Introduction 41.1.2 Relaxation ProcessesRelaxation of the magnetic moments back to equilibrium occurs longitudinally or trans-versely to the external magnetic field B0 and is described by the respective relaxationtimes T1 and T2. The relaxation of the magnetization vector back to equilibrium posi-tion can be described by the Bloch equation:dMdT= γM×B−Mxxˆ+MyyˆT2−(Mz −M0)zˆT1, (1.5)where B is the magnetic field, Mx, My, Mz are the components of the magnetizationvector and xˆ, yˆ, zˆ are the unit vectors. The solutions of the Bloch equation describe thechange of the magnetization vectors over time.T1, known as spin-lattice, is the time required for the spin to regain 63% of its longi-tudinal magnetization after a 90◦ RF pulse has been applied (Figure 1.4). Eventually,signal in the longitudinal direction is recovered and there is an exchange of energy withthe surroundings. The solution of the longitudinal equation of motion after applying a90◦ RF pulse is:Mz(t) = Mz(0) · e− tT1 +Mz0(1− e− tT1 ) (1.6)T1 is affected by thermal interactions between nuclei in the lattice environment. Theseinteractions allow the energy absorbed by the protons during resonance to be dispersedto other nuclei in the lattice. T1 describes the relationship between the frequency ofmolecular motions (vibrational, rotational, translational) and Larmor frequency. Whenthe two frequencies are similar, T1 is short, and vice versa. Water, a small molecule,moves rapidly and has higher natural frequencies than the clinical range of Larmorfrequencies, so it has a longer T1. Long T1 times are exhibited by large molecules likeprotein which move slowly and have lower natural frequencies. When water is trappedin hydration layers, its motion slows and its T1 relaxation becomes much shorter.Upon excitation, the transverse magnetization has its maximum amplitude, and all thenuclei are in phase. Immediately after, the random and time-dependent variation ofadjacent spins causes local changes in the individual magnetic field experienced by thespin. This causes fluctuations in Larmor frequency. There is irreverisible dephasingof nearby spins as faster spins move faster and vice versa, resulting in a fan-shapeddistribution and decay in the transverse magnetization [5]. T2, known as spin-spinrelaxation, is defined as the time for the x-y component to decay to 37% of the originalmagnetization. The solution of the time derivative for the transverse component is:Mxy(t) = Mxy(t = 0) · e− tT2 (1.7)Chapter 1. Introduction 5Figure 1.4: Longitudinal relaxation processes of the magnetization vector. Afterexcitation from a 90◦ RF pulse, the z-component of the magnetization, Mz, decreasesto 0, but slowly recovers to equilibrium. The time at which the spin regains 63% of itsequilibrium value is defined as T1. Modified from [7].Local magnetic field inhomogeneities also contribute to the decay process, and thisrelaxation time is referred as T2′. These variations in magnetic field (∆B) includescanner imperfections, magnetic suceptibilities (tissue’s local response to the externalmagnetic field), and chemical shift. The combination of both static and time variantprocesses is known as the apparent transverse relaxation time, T2∗:1T2∗=1T2+12γ∆B=1T2+1T2′(1.8)Because T2∗ includes more decay processes, there is a greater loss of signal than T2 andT2∗ times are always shorter than T2. Due to the inverse linear relationship of the T2∗components, inverse relaxation rates are sometimes used:R2∗ =1T2∗(1.9)The relaxation of T1 and T2 are tissue specific. These relaxation times affect the contrastobserved in MRI images, and tissue can be visualized and characterized (Table 1.1).Table 1.1: Relaxation times of different musculoskeletal human tissue at 3T, 37◦C [8]Tissue T1 (ms) T2 (ms)Muscle 1420±38.1 31.7 ± 1.9Cartilage 1240 ±107 36.9 ± 3.81Synovial Fluid 3620 ± 320 767 ± 48.8Subcutaneous fat 371 ± 7.94 133 ± 4.43Marrow fat 365 ± 9.0 133 ± 6.14Chapter 1. Introduction 6Figure 1.5: Transverse relaxation processes. After a 90◦ RF pulse, the transversemagnetization, Mxy has its maximum amplitude as the population of spins rotate inphase. The amplitude of Mxy decays over time as the spins move out of phase withone another. T2 and T2∗ are the times required for the magnetization to be reduced to37%. T2 is caused by time-varying spin-spin interactions, and T2∗ is the combinationof T2 and dephasing caused by static homogeneities in the magnetic field. Modifiedfrom [7].1.1.3 Spatial LocalizationSince the strength of the magnetic field, B0 affects the precessional frequency of thespins (as in equation 1.4), the application and superposition of non-uniform magneticfield gradients along the x, y, and z-axes on top of the external magnetic field causesspatial localization of the signal. These three spatial localization techniques include sliceselection, frequency encoding, and phase encoding.Firstly, slice selection involves the application of a magnetic field gradient in the zdirection:B(r) = (B0 + r ·G)zˆ, (1.10)where G is the gradient strength, r is the distance from the origin, the isocenter of thegradient coils. The direction of z is arbitrary and is automatically set to the direction ofthe slice selection gradient; this versatility enables slices to be imaged in any orientation.Spins in the x-y plane precess at the same frequency while those in the z-axis will precessat different frequencies, so only the z slice with frequency ω corresponding to the RFpulse will exhibit resonance and have its spins tipped out of equilibrium (Figure 1.6A).Spins outside the frequency bandwidth remain unaffected, so different slices along thez-axis could be excited by applying an RF pulse at a different Larmor frequency.Chapter 1. Introduction 7(a) Slice Selection (b) Frequency encoding (c) Phase encodingFigure 1.6: Application of gradients in the x, y, and z axes to achieve spatial local-ization of nuclei in the volume. A slice selection gradient along the z-axis only excitesspins at a specific z-distance as they precess at the Larmor frequency corresponding tothe frequency of the RF pulse (A). A magnetic gradient is applied along in the x-axis,causing spins along this axis to precess at different frequencies (B). The y-axis mag-netic gradient causes the spins to be phase-shifted (C). Each nucleus in a given planehas a particular frequency and phase shift, enabling the determination of its spatialdistribution through the Fourier transform. Modified from [2].After exciting a slice, a magnetic gradient in the orthogonal x-direction is applied, re-sulting in spins in different x-positions to precess at different Larmor frequencies as thestrength of the magnetic field gradient varies linearly in that direction (Figure 1.6B). Asa result, the Larmor frequency is made position-dependent along the x-axis by a fieldgradient. This spatial localization technique is known as frequency encoding.For phase encoding, a gradient is applied in the y-axis for a brief time period t (Fig-ure 1.6C). This causes the spins along this axis to momentarily precess at differentfrequencies before returning back to their respective frequencies. The direction of themagnetization vector gets phase-shifted byΦ = γ · y ·Gy · t, (1.11)where Φ is the phase and Gy is the gradient along the y-axis. Sometimes, phase encodingis applied along two axes in order to achieve 3-dimensional distribution of the signal.The overall signal, which contains a mixture of different frequencies and phases, can bebroken back into its individual spatial frequency and phase components by the Fouriertransform.1.1.4 Pulse SequencesThe timing of RF pulses, magnetic field gradients, and signal acquisition is known as thepulse sequence and is programmed into a computer. The time between the RF pulse andsignal acquisition is known as echo time (TE) and the amount of time between successivepulse sequences is the repetition time (TR). Pulse sequences not only enable calculationsof relaxation times for contrast based on T1 and T2 tissue differences, but also enableChapter 1. Introduction 8measurements of other physical properties (e.g. laminar flow, self-diffusion coefficients)for anatomical and functional information. Sequences for T1 measurement include sat-uration recovery and inversion recovery. The most conventional MRI sequences are thespin echo sequence and the gradient echo sequence (GRE).The spin echo sequence involves the application of a 90◦ RF pulse that flips the magne-tization vector (Figure 1.7). The spins precess and dephase due to field inhomogeneitiesand tissue-specific fields. After time τ , a 180◦ pulse flips the spins and afterwards thespins continue precessing. This pulse enables the spins to refocus and be in phase atthe same point at time 2τ , producing an induction signal called spin echo. Dephasingfrom local field inhomogeneities is eliminated by this refocusing pulse, and only T2 re-laxation is observed as the spins are still affected by the random thermal agitation ofthe spin-spin interactions.Figure 1.7: In a Spin echo sequence, a 90◦ pulse is applied, followed by a 180◦ pulsewhich refocuses the spins and eliminates dephasing from local field inhomogeneities.This results in only T2 relaxation being measured. Taken from [5].The gradient echo sequence has no refocusing pulse. Instead, a gradient is appliedin the negative direction for time T , followed by a gradient in the positive directionfor a duration of 2T (Figure 1.8). The spin phase effects of the negative and positivegradient applications cancel and a strong signal is formed at the time when the totalarea under the gradient waveform is zero. This image contrast is not specific to T1 or T2relaxation; however, the signal is sensitive to local field inhomogeneities such as thosecaused by magnetic susceptibilities (the tissue’s local response to a magnetic field), χ,as demonstrated in the following equation:B = µ0(1 + χ)H, (1.12)where µ0 is the permeability (4pi · 10−7 A−2) and H is the magnetizing field. GREsequences can be made more sensitive to T2∗ decay and predominately T2∗ weighted byusing a low flip angle, long TE, and long repetition time [9]. With a longer TE, moresignal loss occurs, resulting in more dephasing and T2∗ sensitivity while a low flip anglereduces the T1 influence, as the longitudinal magnetization remains close to the fullyrelaxed state.Chapter 1. Introduction 9Figure 1.8: Gradient echo sequence. A negative x-gradient is applied for time periodT. Afterwards, the gradient switches polarity and the signal is collected at 2T. Takenfrom [2].For data acquisition in this thesis, a 3D multiple echo gradient echo sequence was used toobtain phase and magnitude information. The principles of phase encoding are extendedinto two dimensions, enabling the true generation of 3 dimensional images for increasedsignal-to-noise ratio. It is a fast acquisition method, and there is possible reduction inTR as low flip angles are used and no refocusing pulses have to be applied to preparethe signal for sampling [10]. Only a single excitation is required to acquire multipleechoes. All echoes following the first detect the magnetization which is left from theinitial pulse and has not yet decayed. The acquisition of multiple echoes at differentecho times enables the calculation of T2∗ or R2∗ contrast images.Figure 1.9: 3D gradient echo sequence. The slice selection is replaced by an additioanlphase encoding gradient in the z-axis, enabling true 3-dimensional images to be acquiredfrom the volume of interest. Taken from [2].Chapter 1. Introduction 101.1.5 Image GenerationThe relaxation of the magnetization vector induces a change in magnetic field. Thischange in flux, governed by Faraday’s law, is detected in the receiver coil and is digitallysampled:Emf =dΦdt, (1.13)where Emf is the electromotive force, and Φ is the magnetic field flux.The signal intensity is proportional to the transverse component of the magnetizationvector, which decays after excitation and results in a free induction decay signal (FID).In MRI, the signals measured are a combination of signals acquired from the objectbeing imaged. So, at a particular time point, the received signal, s(t), is the sum ofall precessing transverse spin magnetization in the volume of interest. All signals s(t)correspond to the Fourier coefficients which are the pixel values of the image in theFourier space. After acquiring all Fourier coefficients, a representation of the originalimage called k-space is acquired. Since the signals in k-space have been subjected to aFourier Transform (FT), the original image in real space can be obtained by applying theinverse FT. Every digital image can be represented as a spatial distribution of grey valueswithin a plane. These pixel values are complex-valued, so the MR signal is composed ofreal and imaginary components.1.2 Phase and Susceptibility-Weighted ImagingIn MR physics, phase is defined as the changing orientation of the magnetization vectorin the transverse plane [11]. The magnetization vector can be described in terms of itslength and direction. So, the transverse magnetization vector, Mxy(r, t) at position rand time t can be written and expressed in complex notation as:Mxy(r, t) = Mxy(r, t) · eiφ(r,t) = |ρ| · (cosφ+ i sinφ), (1.14)where |ρ| =√x2 + y2 is the magnitude or modulus of the vector, φ is the phase, the anglebetween the reference axis and the magnetization vector measured counter clockwise.More specifically, phase can be described as:φ = tan−1(y/x) (1.15)Chapter 1. Introduction 11From 1.15, it is apparent that phase is defined from (-pi, pi). Once the phase values lieoutside this range, drastic contrast changes called phase wraps can occur even in regionswith no changes in magnetic susceptibilities. This ambiguity between measured andactual phase is described as aliasing, and the actual phase can be described as:φactual = φmeasured + n · 2pi (1.16)To remove these phase wraps, unwrapping algorithms have to be implemented to deter-mine n in order to recover a continuous function and determine the correct, wrap-freephase information. These algorithms include Laplacian-based algorithms and filteringmethods [10, 12, 13]. Afterwards, algorithms such as high-pass filtering and frequencydifference mapping need to be applied to remove low-frequency background field in-formation from non-localized sources (e.g. sources outside the field of view, magneticsusceptibilities outside region of interest) so that phase images of localized contrast canbe acquired. The difficulty and time-consumption of these post-processing algorithmsare contributing factors for the underuse of phase imaging in radiology (as compared tomagnitude).Nevertheless, phase images have several advantages over conventional magnitude images.Phase offers high spatial resolution of anatomic structures and better signal to noise(SNR) and contrast to noise ratio (by ten-fold) than conventional magnitude images[14, 15]. As mentioned in previous sections, phase can be used to encode spatial infor-mation; however, there are other forms of phase present. Biophysical mechanisms suchas magnetic susceptibilities of different inclusions (e.g.deoxyhemoglobin, iron deposits,air-tissue interfaces, paramagnetic contrast agents) of the tissue, chemical exchange,and changes in fiber orientation and tissue microstructure at the subcellular and cel-lular levels, can create macroscopic and microscopic fields that alter the magnetic fieldoutside and within the tissue [11, 16–19]. These alterations then induce changes inthe precessional/resonant frequencies of the spins (relative to Larmor frequency) in thesurrounding tissue. Phase is related to resonant frequency (ω) by:φ = ∆ω · t. (1.17)These changes in frequency result in changes in phase, as phase measures the orientationof the magnetization vector relative to the location the vector would be pointing if itprecessed at the Larmor frequency. The above mechanisms have been proposed as phasecontrast mechanisms, and current investigation is still underway to study the degreeof their influence. Due to its sensitivity to magnetic susceptibilities and orientation ofanisotropic tissue (relative to B0), phase can pick up microstructure details in tissue thatChapter 1. Introduction 12are not normally detected in conventional magnitude information. Phase imaging hasbeen an established technique to visualize and map blood vessels and blood flow. Moreimportantly for this thesis, calculated phase masks called venograms can be incorporatedwith magnitude images to form susceptibility weighted images (SWI) [15].1.3 Knee Articular CartilageThe knee joint is formed by three bones: the femur, tibia, and patella. The distal endof the femur and proximal end of the tibia each contain a lateral and medial condyle.So, the three joint articulations of the knee are located between each correspondingfemoral and tibial condyle and between the patella and femur (Figure 1.10A). Thesejoints are covered in articular cartilage and supported by ligaments. Additional supportand cushioning is provided by muscles, crescent-shaped menisci, bursae, and fat pads(Figure 1.10B). The health and maintenance of these components are important forprotecting the articular cartilage from damage and degeneration.(a) Coronal View of the Knee (b) Sagittal View of the KneeFigure 1.10: Anatomy of the knee joint. The femur and tibia are divided into medialand lateral regions, so the two articulations and the patellofemoral articulation make theknee joint (A). Region S represents the area below the tibial spines, where menisci andligaments are attached. Cartilage, muscle, menisci, synovial fluid, bursae, ligaments,and fat pads are components that collectively support the knee joint (B). Modified from[20] and [21].Chapter 1. Introduction 131.3.1 Healthy CartilageCartilage is a flexible connective tissue that can be classified as three forms: hyaline,elastic, and fibrocartilage [22]. Elastic cartilage, the most flexible form, is found in theear, epiglottis, and larynx. Fibrocartilage, a more dense and fibrous form, is a componentin the spinal column, knee menisci, tendon, and rib and jaw joints. Hyaline cartilage,the most common, is found in joint (articular) surfaces. Healthy articular cartilage hasstiffness to compression, firm consistency, and some elasticity. It provides a smoothsurface for joint motion by reducing friction between gliding surfaces of the bone, andacts as a resilient, load-bearing medium which absorbs shock and minimalizes pressurefrom bone articulations [22].Articular cartilage is a complex three dimensional structure composed of chondrocytesembedded in the extracellular matrix (ECM). Chondrocytes, which represent 1% of thecartilage volume, are responsible for secreting and maintaining the organic componentsof the ECM, which consist of hyaluronic acid, type II collagen (approximately 15% wetweight), proteoglycans (5% wet weight), and extracellular water (65-80% wet weight)[23]. On a molecular level, the interaction between these components affect the car-tilage’s biomechanical properties. Proteoglycans are proteins with the carbohydrateglycosaminoglycan (GAG), and the most common proteoglycan in articular cartilage isaggrecan. Aggrecan’s structure resembles a bottlebrush, with the protein as the core,and the two glycosaminoglycan chains of keratin sulfate and chondroiton sulfate as thebrush. Aggrecan can bind with hyaluronate molecules, resulting in their aggregation.Fixed aggrecan with the negatively charged GAG chains then attracts positive ions suchas sodium and traps water due to the increase in osmolarity [24].Macroscopically, the negatively-charged glycosaminoglycans are enmeshed in the colla-gen matrix, and their electrostatic repulsion helps maintain the stability of the collagenfibres and provide the cartilage with the resiliency to withstand pressure [25]. Thetrapped water causes swelling of the ECM, and this swelling force, driven by the os-motic pressure, is balanced by tension from the collagen fibres [26]. Thus, even at anunloaded state, the collagen network is still subjected to tensile pre-stress. When car-tilage is further compressed by an increased external load, some water is squeezed outof the cartilage until the increased repulsion from the GAGs reaches a new equilibriumand gives cartilage the resiliency to withstand compression.The arcade model describes the orientation of the collagen fibres in a healthy adult’scartilage as they are organized into the calcified, radial, transitional, and tangential zonesto form a network to provide stability (Figure 1.11). In the calcified zone, calcifiedcartilage attaches cartilage to the underlying subchondral bone. Fibres in the radialChapter 1. Introduction 14Figure 1.11: Arcade model of collagen in healthy tibiofemoral articular cartilage.From the bone surface, fibres are organized perpendicular to the surface (radial zone),then transition into having no orientation preference (transitional zone), and finally areoriented tangential to the bone surface (tangential zone). The angle αf characterizesthe angle the fibre deviates from its initial radial orientation. Taken from [27].(deep) zone are highly anisotropic and are arranged radially from the bone surface. Thecollagen fibres are the thickest in this layer. In the transitional (intermediate) zone,fibre orientation is isotropic with no preferred orientation as it connects the radial andtangential fibres [27]. In the tangential (superficial) zone, fibres are arranged parallelto the joint surface. This layer provides tensile strength and resists swelling from thedeeper layers; furthermore, this pattern of fibres is resistive to shear stress and formsa gliding surface for the cartilage. The outer layer of the tangential zone is covered bythe lamina splendens, a thin cell-free layer of tightly packed and tangentially orientedcollagen fibres. This layer of collagen fibres acts as a filter to prevent the passage of largemacromolecules. Histological analysis of these zones has determined that the radial zoneis the thickest (over 50% of total thickness) while the superficial zone is the thinnest(3-12%). As tissue depth increases, chondrocytes become bigger. The concentrationof water is greatest in the tangential zone and the concentration of proteoglycans isgreater in the transitional and radial zones. Variations in zone thickness and componentcontent exist in cartilage from different sites along the joint, and these variations maybe attributed to the differences in mechanical demands.1.3.2 Cartilage Damage and DegenerationOsteoarthritis (OA) is a progressive joint disease characterized pathologically by articu-lar cartilage break-down and loss, subchondral sclerosis, and cyst and osteophyte (bonespur) formation. Osteoarthritic cartilage deforms more readily and loses more fluid asChapter 1. Introduction 15compared to healthy cartilage with the same force application [22]. Articular cartilagemay be subjected to traumatic injury or degenerative change. In a healthy knee joint,daily activities of repetitive loading (e.g. sports, work) can cause damage which is bal-anced by consistent repair. However, when the damage persistently exceeds repair, thecartilage’s subsequent degeneration often leads to OA. The exact pathophysiology ofOA onset is still under investigation and may vary among individuals; however, it is ac-knowledged that this complex condition results from various influences. Factors linkedwith this disease include age, obesity, gender (3:2 female/male ratio), repetitive stressinjuries, genetics, illness such as rheumatoid arthritis (RA), iron overload, and excessgrowth hormone.The first signs of OA include a decrease in freedom of active joint movement, and they aretypically noticed after a minor joint injury or strenous physical activity [28]. In the earlystages of OA, there is increased synthesis of ECM components, DNA, and proteolyticenzymes. Although the orientation of the collagen fibres is disrupted, collagen contentis maintained, and the appearance of cartilage remains unchanged macroscopically [23].Due to disruption in the collagen network, the proteoglycans are no longer constrained bytension of the collagen network and can bind more water molecules, resulting in increasedwater content. Eventually, there is a decrease in proteoglycan content, resulting theweakened ability of cartilage to resist deformation because of the decreased repulsionof negatively charged GAGs. As a result, the load is redistributed, with an increasingportion being carried by the solid components of the ECM, and this leads to cartilageflaking of the articular surface [29].As OA progresses, there is an increase in signalling molecules such as Interleukin-1 (IL-1),tumor necrosis factor (TNF), and nitrous oxide. IL-1 and TNF suppress the synthesis ofcollagen II and induce catabolic enzymes while nitric oxide acts a pathogenic mediator.Furthermore, apoptosis, programmed cell death, increases and is likely responsible forthe decrease in the number of functioning chondrocytes. Consequently, chondrocytescan no longer keep up with the rate at which their synthesized components degrade,and there is a decreased synthesis of ECM components and decrease in water content[29]. The decreased synthesis of type II collagen and increased breakdown of pre-existingcollagen lead to further weakening and disorientation of the collagen network. Surfaceirregularities deepen and become fissures that eventually reach the subchondral bone.In the advanced stages, cartilage thickness decreases, subchondral bone eventually isexposed, and the pain becomes severe and constant.Although softening and thinning of articular cartilage is one of the most defining charac-teristics of OA, it is important to note that abnormalities in other structures of the kneecan also accelerate cartilage damage and joint degeneration. Since cartilage is aneural,Chapter 1. Introduction 16pain in OA likely arises from these structures. Meniscal tears and displacement areoften observed in OA knee joints [30]. Furthermore, since articular cartilage is fairlyelastic in nature, it is dependent on its subchondral base for its continued integrity. So,it has been proposed that bone marrow lesions and edema are risk factors for structuralprogression of knee OA and may play a role in cartilage degeneration and pain [31, 32].1.3.3 Methods of DiagnosisOA is often diagnosed based on patient history and physical findings. Current clinicalpractices for diagnosing cartilage damage include X-ray radiography, arthroscopy, andMRI.The imaging principles of X-ray radiography are based on tissue absorption of ionizing,electromagnetic high energy radiation (20-150 keV). Gross changes such as the appear-ance of osteophytes, major loss of cartilage, and joint malalignment can be visualized[22]. Cartilage loss is indirectly inferred based on the narrowing of joint space (Figure1.12A). Despite its high spatial resolution, relatively low cost, and fast acquisition time,radiography is limited by its inability to detect the early signs of cartilage damage andits insensitivity to focal cartilage loss [33]. Furthermore, it has been shown that thedegree of joint space narrowing correlates poorly with the incidence and magnitude ofjoint pain [34].Arthroscopy is a surgical procedure in which an arthroscope, a small fibreoptic camera,is inserted into the joint through a small incision for direct visualization (Figure 1.12B).More incisions may be made for visualization of other knee compartments and sometimestreatment (e.g. cartilage microfracture) may accompany the procedure. Because of itshigh specificity (true negative rate) and sensitivity (true positive rate), it is consideredthe “gold standard” by clinicians as it enables accurate appraisal of the joint [35, 36].However, arthroscopy is ultimately an assessment of the surface only, and is inaccuratein estimating cartilage thickness and lesion depth.MRI relies on the stimulation and emission of non-ionizing electromagnetic radiationof low energy (10−8 - 10−6eV). In comparison to the previous methods, MRI is a 3Dtechnique which visualizes the knee as a whole organ, offers superior soft tissue contrast,and enables non-invasive evaluation of cartilage morphology and function (Figure 1.12C).Different zones of cartilage can be visualized at different imaging planes, making thisversatile imaging modality a promising procedure in evaluating cartilage degeneration[22].Chapter 1. Introduction 17(a) X-ray radiography (b) Knee arthroscopy (c) MRIFigure 1.12: Current methods of knee cartilage diagnosis include X-ray radiographyof the joint space width (A), arthroscopy for direct visualization (B), and MRI for wholeorgan visualization and quantative assessment (C). Taken from [37] and [38].Cartilage scoring systems have been created for more uniform and consistent diagnosisamongst physicians. The Kellgren-Lawrence scale measures cartilage loss and subchon-dral bone reaction of radiographic images by the presence of osteophytes, joint spaceloss, subchondral sclerosis, and cyst formation. For arthroscopy, the most common in-ternational scoring system is the Outerbridge scoring system, which classifies cartilageby the severity of damage and size (Table 1.2). The Noyes scoring system distinguisheslesions into 3 surface appearance grades which are subdivided based on depth of involve-ment, diameter, and location for an overall joint score calculation. For MRI, there areseveral scoring systems including Recht, which is a modified Noyes classification, andWhole Organ Resonance Imaging Score (WORMS), a semi-quantitative scoring methodfor multi-feature, whole-organ evaluation of the knee. WORMS combines assessments ofcartilage lesion depth and size, meniscal morphology, cysts, bone attrition, subarticularmarrow edema, osteophytes and ligaments. [20, 22]Table 1.2: The Outerbridge arthroscopic scoring system of cartilage lesions [22]Grade Surface description of articular cartilage DiameterI softening and swelling not evaluatedII fragmentation and fissuring <1/2 inchIII fragmentation and fissuring >1/2 inchIV erosion of cartilage down to bone not evaluated1.4 MotivationOsteoarthritis, the most common form of arthritis, affects one in eight Canadians [39].Symptoms include joint pain, stiffness, and swelling. Once OA has developed, patientsChapter 1. Introduction 18suffer from the disease for the remainder of their lives, and long-term disability becomessignificant as the pain severity increases and joint function becomes limited. Conse-quently, OA is the third most common diagnosis in the elderly and is the leading causeof functional decline and morbidity for seniors [40]. According to the Arthritis Allianceof Canada, it is estimated that there were 4.4 million Canadians living with OA in 2010and that the financial burden of this condition was $27.5 billion. Data on prevalenceand incidence, population demographics, risk factor and demographic modules, healthcare utilization, and wage-based productivity loss have been implemented in simula-tions. These simulations strongly suggest that by 2040, the prevalence will increase to10.4 million people with a total economic burden of $1455.5 billion [39].At present, there is no medically-established cure for OA, and symptoms can worsen withtime; however, there is treatment for slowing cartilage degeneration, decreasing pain, andimproving joint function. This includes lifestyle changes, pharmacologic treatment, andsurgical procedures. Exercise and weight loss is considered the most effective methodof preventing further damage as the weight-bearing load of the knee is directly lessenedand muscle may be strengthened [22]. Meanwhile, pharmacologic agents such as dietarysupplements and hyaluronic acid have been proposed to preserve cartilage or treat dam-aged cartilage, but their effectiveness has been debated as they have not been provento provide significant benefit, and their biochemical mechanism is not fully understood[41–43]. Surgical options for knee cartilage aim to ease symptoms and postpone OAonset. These include reparative (e.g. microfracturing, drilling) and reconstructive (e.g.autologous chondrocytes implantation, osteoarticular transfer) techniques. At end-stageOA, total joint replacement with artificial components is the definitive treatment; how-ever, the limited lifespan of these prostheses may be unable to meet the growing demandfrom younger and more active patients [44]. As more therapeutic options (3D cartilagetissue engineering, stem cell therapy, drugs) are being explored, assessment of earlychanges is important for evaluating the feasibility of these therapies [45–47]. Follow-upimaging and monitoring of these therapies may require advanced and sensitive imagingtechniques. As a result, there is a growing need for techniques that are sensitive andaccurate at detecting earlier stages of OA so that patients have more options to makeproper lifestyle and health choices to improve their quality of life.Since OA involves many structures within a joint, MRI permits non-invasive and non-ionizing visualisation of all these structures in three dimensions, and thereby allowingother sources of joint pain to be considered. Imaging requirements imposed by physicaldimension (as cartilage is only a few mm thick), gemometry, anatomic location, and thebiochemical composition of cartilage affect MRI scanning parameters. Early signs of car-tilage damage can be visualized with changes in contour, thickness, and signal intensity[22, 48]. T2 mapping describes the relationship between collagen network integrity andChapter 1. Introduction 19water interaction and there is excellent contrast between the cartilage and surroundingsynovial fluid [49]. In the early stages, there is an increase in water content due todisorganization of the collagen matrix. This results in a higher T2 relaxation time fromthe proton spins [50–53]. This change is a sensitive method of assessing cartilage health,and T2 mapping is becoming more established as a compositional imaging technique.However, T2-weighted imaging has a long imaging duration, ranging from 10 to 20 min-utes, depending on the size of imaged volume. This can lead to motion artifacts as it isdifficult for OA patients to remain completely still due to pain. T1-weighted and ShortTau Inversion Recovery (STIR) MRI can also be used to confirm cartilage damage sincebone edema and bone bruising, which are risk factors for cartilage damage, can be visu-alized. A relatively new imaging contrast of cartilage damage being investigated is T2∗,as it can be acquired at much shorter acquisition times (no refocusing pulse). However,as this is an emerging field, there are only a few clinical studies presently available.Fairly recently, a model has been developed by He and Yablonskiy to explain how struc-tural disordering can be a mechanism of phase contrast (and frequency shifts). It ex-plores the effect of fibre orientation and magnetic susceptibility differences in the brain.MR frequency shifts are described as the sum of the following two equations:∆ff0= LF · χ, (1.18)∆ff0= A · χ, (1.19)where ∆f is the tissue frequency shift, f0 is the nuclei’s Larmor frequency, LF is theLorentz factor, χ is the magnetic susceptibility (product of volume magnetic suscepti-bility and volume fraction), and A is the parameter depending on the object’s specificshape. Equation 1.18 describes effects of the Lorentzian sphere (in which LF is 4pi3 forrandomly scattered inclusions). In this approach, a Lorentzian sphere is drawn aroundmagnetic inclusions, which are considered as point dipoles altering the magnetic fieldthrough their magnetic susceptibilities. However, for intact longitudinal structures likeaxons, LF approaches zero. Equation 1.19 explains the effects of the object’s generalexternal shape on frequency shift. For circular longitudinal structures, this equation isfurther described by:∆ff0= −2pi ·∆χ ·(cos2 α−13), (1.20)where ∆χ is the difference in magnetic susceptibility between isotropic componentsand external medium, and α is the angle between the fibre and magnetic field [16,17]. This model also predicts that longitudinal structures such as myelin sheaths andneurofilaments do not contribute to MR signal frequency shift if they are oriented parallelChapter 1. Introduction 20to the magnetic field (due to Maxwell’s equations), thus making the frequency shiftcompletely due to equation 1.18.This model has been recently explored in multiple sclerosis, a disease in which struc-turally anisotropic healthy white matter transforms into structurally isotropic debris.This transformation due to tissue damage leads to an increase in LF from 0 to 4pi3 , re-sulting in a frequency shift. Literature has reported that MR frequency increased withearly MS lesion formation [12, 54]. Similarly, in healthy tibiofemoral cartilage, collagenfibres are organized into isotropic and anisotropic regions [27]. As cartilage gets dam-aged, the collagen fibre matrix becomes disorganized, and anisotropic layers can becomeisotropic lesions. One of the goals of this project was to investigate these changes byusing a frequency-based technique, and to observe whether structural changes in theanisotropic zones of cartilage will result in an MR frequency shift when cartilage is dam-aged. Frequency maps were compared with T2∗ maps, T2 maps, T1-weighted images,and STIR scans. This project also explored whether frequency maps could visualize theradial and transitional zones as distinct regions.SWI has been studied for clinical applications in the brain, in particular for pathologiessuch as tumors, stroke, and multiple sclerosis lesions [10, 15, 19]. However, it hasnot been applied yet in knees with cartilage damage, so this project also qualitativelyexplored the potential of this technique.Phase and frequency information from 3D multiple echo GRE scans has the potential toenhance magnitude images and be a more sensitive measure of early cartilage damage.3D multiple echo gradient echo sequences also enable high resolution knee images, notpreviously implemented in literature, to be acquired at relatively short acquisition times.So, results from this project may aid in earlier identification of damage to cartilagefor prompt treatment and lifestyle choices. Consequently, the concepts and methodsproposed by this study could be applied for monitoring, evaluating, and comparing theeffect of different treatments to prevent potential long-term disability.Chapter 2Methods2.1 Recruitment of Study ParticipantsThis study received approval from the Clinical Research Ethics Boards. Patients withcartilage damage were recruited either personally or through referral from the New WestOrthopaedic and Sports Medicine Centre. Their cartilage damage had been confirmedpreviously by a physician through arthroscopy or X-ray. Pre-screening telephone in-terviews and MRI bookings were conducted by the author who reviewed inclusion andexclusion criteria for participation. Approved informed consent was then forwarded topatients using a standardized letter of invitation with the following inclusion criteria:clinically definite osteoarthritis or other cartilage damage, ages 18 to 65 years, and theability to provide informed consent. Exclusion criteria included inability to provideinformed consent and possessing contraindictions to MRI. 6 patients diagnosed withosteoarthritis or cartilage damage, ages 48 - 70 years (mean age 58.3±3.6 years, aver-age BMI 29.9±2.3, 5 males, 1 female), were recruited as well as 4 controls (mean age24.3±1.9 years, average BMI 24.5±1.9, 2 males, 2 females).2.2 MRI ProtocolData were acquired on a Phillips Achieva 3.0T research scanner located in the UBCMRI Research Centre (Purdy Pavillion, UBC Hospital). All sequences were performedusing a SENSE 8-channel transmit-receive knee coil with a survey and reference scanbeing conducted prior to the sequences listed below. Research subjects were positionedsupine with the axis of the knee aligned parallel to the static magnetic field.21Chapter 2. Methods 22Since this was a proof-of-concept project, MRI protocol parameters for the controls werebeing modified and optimized for protocol development throughout the project, and afew of the controls were scanned multiple times at separate sittings (4 controls for 6 datasets taken at different dates). Susceptibility weighted data (for SWI images, T2∗, andfrequency maps) were acquired using a 3D multiple echo GRE (which was set to acquiremagnitude and phase information), and parameters investigated included resolution,field of view (FOV), number of echoes, repetition time, and echo time. Listed below arethe sequence parameters discussed in the results and discussion:• Acquisition/reconstructed matrix = 456 x 457/639 x 640, acquisition/reconstructedvoxel size = 0.35 x 0.35 x 2.00 mm3/0.25 x 0.25 x 1.00 mm3, FOV = 160 x 160 x48 mm3, TR/TE = 29/5 ms, 5 echoes, 17◦ flip angle, 48 slices, acquisition time =5.46 min• Acquisition/reconstructed matrix = 560 x 561/700 x 701, acquisition/reconstructedvoxel size = 0.25 x 0.25 x 3.00 mm3/0.20 x 0.20 x 1.50 mm3, FOV = 140 x 140 x48 mm3, TR/TE = 40/5 ms, 5 echoes, 17◦ flip angle, 32 slices, acquisition time =6.21 min• Acquisition/reconstructed matrix = 700 x 700/ 960 x 960, acquisition/recon-structed voxel size = 0.20 x 0.20 x 3.00 mm3/0.15 x 0.15 x 1.50 mm3, FOV =140 x 140 x 48 mm3, TR/TE = 64/5 ms, 5 echoes, 17◦ flip angle, 32 slices, acqui-sition time = 11.31 min• Acquisition/reconstructed matrix = 932 x 932/ 1200 x 1200, acquisition/recon-structed voxel size = 0.15 x 0.15 x 3.00 mm3/0.12 x 0.12 x 1.50 mm3, FOV = 140x 140 x 48 mm3, TR/TE = 111/5 ms, 5 echoes, 17◦ flip angle, 32 slices, acquisitiontime = 13.29 minA spin-echo sequence (Acquisition/Reconstructed Matrix = 440 x 437/511 x 521, acqui-sition/reconstructed voxel size = 0.36 x 0.37 x 3.00 mm3/0.31 x 0.31 x 3.00 mm3, FOV= 160 x 160 x 39 mm3, TR/TE = 1050/13 ms, 7 echoes, 11 slices, acquisition time =6.49 min) was also applied so that T2 maps could be calculated.The sequence with an acquisition matrix of 700 x 700 offered the highest resolutionwithout compromising image quality, and it was used for patients. Furthermore, thefield of view along the lateral-medial axis was increased so that both condyles of theknee were imaged.For patients, the sequences implemented were:Chapter 2. Methods 23• Multiple-Echo 3D GRE sequence: acquisition/reconstructed matrix = 700 x 700/960 x 960, acquisition/reconstructed voxel size = 0.20 x 0.20 x 2.60 mm3/0.15 x0.15 x 1.30 mm3, FOV = 140 x 140 x 71.5 mm3, TR/TE = 37/5 ms, 3 echoes, 17◦flip angle, 55 slices, acquisition time = 12.09 min• Spin-echo sequence: acquisition/reconstructed matrix = 440 x 437/511 x 521,acquisition/reconstructed voxel size = 0.40 x 0.40 x 3.00 mm3/0.31 x 0.31 x 3.00mm3, FOV = 180 x 160 x 93 mm3, TR/TE = 1050/13 ms, 7 echoes, 26 slices,acquisition time = 15.40 min• T1 weighted images: acquisition/reconstructed matrix = 484 x 481/560 x 560,acquisition/reconstructed voxel size = 0.33 x 0.33 x 3.00 mm3/0.29 x 0.29 x 3.00mm3, FOV = 160 x 160 x 89.25 mm3, TR/TE = 544/20 ms, 1 echo, 24 slices,acquisition time = 5.52 min• STIR sequence: acquisition/reconstructed matrix = 308 x 300/576 x 570, acqui-sition/reconstructed voxel size = 0.55 x 0.57 x 3.50 mm3/0.3 x 0.3 x 3.50 mm3,FOV = 170 x 170 x 89 mm3, TR/TE = 3485/190 ms, 1 echo, 20 slices, acquisitiontime = 4.46 minAlthough the multiple-echo 3D GRE sequence was approximately 3 minutes shorterthan the spin echo sequence, it had a much higher spatial resolution. Having a spin-echo sequence at same resolution of the 3D GRE would result in greater acquisitiontimes (initial tests indicated acquisition times ranging from 25 - 30 minutes).2.3 Data AnalysisData were analyzed using MATLAB (The MathWorks, Inc., Natick, MA) and FSLVIEW(FMRIB, Oxford, UK) software. The MRI scanner made output files in .REC and .PARformat, which was in binary format. So, data had to be converted into standard .niifiles for direct viewing on FSLVIEW or loaded and converted in MATLAB for furtherpost-processing and analysis. Masks were manually drawn for regions of interest usingFSLVIEW (Figure 2.1).2.3.1 T2 and T2∗maps, T1-weighted Images, STIR SequencesT2 maps from the spin echo sequences were constructed by the MRI scanner, extractedusing the dcm2nii routine (MRIcon), and overlayed with the 2nd echo of the imagesChapter 2. Methods 24Figure 2.1: Articular cartilage was manually segmented from the other tissues inFSLVIEW.acquired from the spin-echo sequence. From the magnitude images of the 3D multi-echo GRE sequence, T2∗ maps were computed automatically by the scanner and wereoverlayed with corresponding average magnitude images. These T2∗ values were off by ascaling factor and had to be rescaled to their correct values by using parameters obtainedfrom .PAR files.T1-weighted images and images from STIR sequences were extracted using dcm2nii, andno further processing was implemented.2.3.2 Post-ProcessingPhase images were processed by a homodyne filter, a high pass filter which removesboth phase wraps and non-local background field contributions simultaneously. A 2DHanning window was applied slice-by-slice to 2D k-space complex data (Fourier space)to produce low-pass filtered images. The width of the Hanning window was determinedby visual inspection of the resulting phase images. For increasing TE, the number ofphase wraps increased, so the filter width at different TE was adjusted. The filterwidths implemented were 106 at 5 ms, 186 at 10 ms, and 286 at 15 ms. They were madeindependent of the dimension of the images as spatial frequencies of field inhomogeneitiesshould not change with matrix size. Complex division of the original complex imageswith low-pass filtered complex images generated high-pass filtered images.Frequency maps (ω) were generated by dividing the filtered phase images (φ) by theecho time and normalizing with the Larmor frequency (ω0):∆ω[ppm] = φ/(TE · ω0). (2.1)MR frequency maps provide information about the resonance frequency in each imagingvoxel and with Larmor frequency normalization, they could be compared across differentscanners and field strengths [12]. These frequency maps were averaged over 3 echoesChapter 2. Methods 25(5 ms, 10 ms, and 15 ms) to reduce noise and were overlayed with the correspondingaverage magnitude images.For susceptibility weighted images, also known as venograms, a phase mask based on thefiltered phase data was calculated. It was assumed that voxels with veins had increasedMR frequency values (due to increased deoxyhemoglobin) and the phase in these voxelswas set to 1, while other voxels were scaled to a range of 0 to 1. The phase mask (fn(x))was then multiplied by the original magnitude image (ρ(x)) to create susceptibilityweighted images (ρswi(x)) for each echo [11]:ρswi(x) = fn(x) · ρ(x) (2.2)The fourth power (n=4) was chosen to optimize the SNR of the SWI. SWI of the threeechoes (at 5 ms, 10 ms, and 15 ms) were then averaged. This technique had been appliedpreviously in the brain for imaging tumors, stroke, hemorrhages, and MS, but it has notbeen studied in knees. A summary of the post-processing steps for knee SWI is presentedbelow (Figure 2.2).Chapter 2. Methods 26Figure 2.2: Summary of post-processing methods to create susceptibility weightedimages. For each echo, the homodyne filter was applied to unwrap the phase, removebackground field contributions, and create a high-pass filtered image. The filtered phaseimage was then converted into a phase mask. The fourth power of the phase mask wasmultiplied with the respective magnitude image to obtain the final SWI for each echo.Chapter 3Results and Discussion3.1 Controls3.1.1 Protocol and Post-Processing DevelopmentInitial scans were taken at 4 different resolutions with reconstructed matrix sizes of 639x 640, 700 x 701, 960 x 960, and 1200 x 1200 and respective voxel sizes of 0.25 x 0.25x 1.00 mm3, 0.20 x 0.20 x 1.50 mm3, 0.15 x 0.15 x 1.50 mm3, and 0.12 x 0.12 x 1.50mm3. There was a trade off between resolution and SNR. To achieve higher resolutions,smaller voxel sizes are required; consequently, there is less signal in each voxel [2].For the magnitude images, excellent SNR was maintained at lower resolutions. At thehighest resolution, there was slight observable noise, and the images began looking moregrainy. For the corresponding post-processed phase images, at the highest resolution,the effect of decreased SNR was more pronounced, and the image was much noisier thanthe magnitude. These observed differences in phase and magnitude correspond withthe theory that phase variation is much smaller if the SNR is sufficiently large (evenwhen there are equal variances in the real and imaginary components) [11]. So, noisein the magnitude and phase images was not correlated [11]. Scans for the patients wereacquired at a reconstructed voxel size of 0.15 x 0.15 x 1.50 mm3, as these parametersoffered a high resolution without significant decrease in SNR. Unless otherwise specified,all images presented in this thesis had this reconstructed voxel size.Furthermore, initial scans were conducted with 5 echoes with a TE of 5 ms. As theecho time increased, the signal decayed, and this decrease in SNR became more evidentafter the third echo (Figure 3.3). In particular, it became difficult to visualize thearticular cartilage around the tibia. Consequent scans were then acquired with 3 echoes27Chapter 3. Results and Discussion 28Figure 3.1: Magnitude images (single echo) of right knee, medial condyle at 4 differentresolutions with voxel size of 0.25 x 0.25 x 1.00 mm3 (A), 0.20 x 0.20 x 1.50 mm3 (B),0.15 x 0.15 x 1.50 mm3 (C), and 0.12 x 0.12 x 1.50 mm3 (D). At the highest resolution,the image became grainier.Figure 3.2: Post-processed images (single echo) of right knee, medial condyle at 4different resolutions with voxel size of 0.25 x 0.25 x 1.00 mm3 (A), 0.20 x 0.20 x 1.50mm3 (B), 0.15 x 0.15 x 1.50 mm3 (C), and 0.12 x 0.12 x 1.50 mm3 (D). At the highestresolution, the image was noisier than the lower resolutions.Chapter 3. Results and Discussion 29to preserve SNR and decrease scan time. Magnitude images were then averaged over 3echoes to reduce noise.Figure 3.3: Magnitude images of right knee, medial condyle at echo times of 5 ms(A), 10 ms (B), 15 ms (C), 20 ms (D), and 25 ms (E). By the fourth echo, the decreasein SNR makes it difficult to distinguish certain regions of the cartilage (red arrow).As the echo time increased, the number of phase wraps increased and the effects ofconstant vs. adaptive homodyne filtering were apparent (Figure 3.4). With a constantfilter width of 106, phase wraps began appearing and degrading the phase image at aTE of 10 ms (Figure 3.4 E, F). Modified filter widths of 186 at 10 ms and 286 at 15ms enabled good suppression of the phase wraps while maintaining some phase contrast(Figure 3.4 H, I). In this particular slice, it can be noted that the boundary betweenthe articular cartilage of the tibia and femur became slightly less distinct. This may bedue to the inherent feature of the homodyne filter, in which more phase wraps can beremoved at the expense of losing low spatial frequency information. There were someresidual phase wraps at the tissue-air interface; however, these did not affect the regionof interest.3.1.2 Comparison of T2, T2∗, and Frequency Maps, Magnitude Im-ages, and SWIIn the T2 map and T2∗ maps, cartilage had a laminar appearance. T2 values were lowestin the highly anisotropic radial zone and increased towards the superficial zone due toincreased water content in the superficial zone and increased dipole-dipole interactionChapter 3. Results and Discussion 30Figure 3.4: Unprocessed phase images of right knee, medial condyle at echo times of5 ms (A), 10 ms (B), and 15 ms (C) and corresponding post-processed phase imagesat a constant filter width of 106 (D, E, F) and adjusted filter widths (FW = 186 at 10ms, FW = 286 at 15 ms) for echo time (G, H, I) to remove residual phase wraps (redarrows).between water molecules (Figure 3.4A). These T2 values were within the healthy rangeof 15 ms to 60 ms. In addition, this pattern of increase was in agreement with past MRIand histology studies [49, 50, 55]. There was an increase in T2 values in the anterior andposterior regions, and these differences may be either caused by the magic angle effector regional tissue component differences (as these regions are not load-bearing). Magicangle effect is a common effect which occurs in cartilage (knee, shoulder, ankle, wrist) ofhumans, bovine, canine, and porcine. It is linked to the structure of collagen fibres, theirorientation in the magnetic field, and the water-proteoglycan interactions that amplifythe prevailing orientation of the collagen fibre network. This results in higher T2 valueswhen the collagen fibres are oriented 54.74◦ relative to the main magnetic field [56].A similar pattern of contrast was observed for T2∗ maps; these times had a smallerrange because in T2∗, more decay contributions (e.g. magnetic susceptibilities) areChapter 3. Results and Discussion 31measured and picked up during the 3D GRE sequence (Figure 3.5B). Similar findingswere observed in an earlier study which reported a significant correlation between T2and T2∗ in healthy controls [55].Articular cartilage in the frequency map had a heterogeneous contrast pattern (Figure3.5C). In the tibial cartilage, there was a decrease in frequency from the radial zoneto the superficial zone. In the femoral cartilage, anisotropic collagen fibres were ori-ented at different angles relative to the external magnetic field (due to the contour ofthe condyle). The frequency values of the radial zone appeared to oscillate betweennegative (anterior and posterior regions) and positive (central zone) frequency values.Frequency also seemed to alter in the superficial zone (positive frequency values in theanterior and posterior regions, negative frequency values in the central zone). Thesequalitative patterns of the observed frequency and fibre orientation (relative to the ex-ternal magnetic field) are a clinical confirmation of He and Yabloskiy’s model predictinga relationship between fiber orientation and frequency (Equation 1.20). Previous workon healthy knee controls has quantified this relationship in the radial and superficialzones of the femoral cartilage [57].The SWI image (Figure 3.5E) was able to provide more tissue details (e.g. patellar ten-don, fat) than the conventional magnitude images, averaged over 3 echoes (Figure 3.5D).The tibial articular cartilage had a distinct, striated pattern in the SWI. The details ofthis striation was not observed in the average magnitude images. This striated patternmay be due to the magnetic properties of the collagen fibres detected by the phase. Itcan be noted that there was a decreasing intensity gradient in the striation pattern fromanterior to posterior, and there was a similar contrast pattern in the corresponding T2∗map. This contrast pattern pattern was not observed in the T2 map. This adds supportthat magnetic susceptibilities from the underlying tissue microstucture at the cellularand subcellular levels may be enhancing the details observed in SWI. Furthermore, theT2∗ map had greater times in the superficial zone of the tibial cartilage; this suggeststhat the brightness in the corresponding region of the SWI was not due to an imageprocessing artifact from high-pass filtering.3.2 PatientsPresented in this thesis are three representative cases from 3 patients, two with char-acteristic signs of OA (Figure 3.6, 3.8) and one with suspected mild cartilage damage(Figure 3.7, 3.8).Chapter 3. Results and Discussion 32Figure 3.5: Comparison of T2 (A), T2∗ (B), Frequency Maps (C), Average Magni-tude Image (D), and Average SWI (E) of the right knee, medial condyle in controlparticipant. Averaging was done over 3 echo times of 5 ms, 10 ms, and 15 ms. (A)had a reconstructed voxel size of 0.31 x 0.31 x 3.00 mm3. (B), (C), (D), and (E) had areconstructed voxel size of 0.15 x 0.15 x 1.50 mm3.Chapter 3. Results and Discussion 33In one of the patients with characteristic OA signs, there were observable changes in thecartilage contour and thickness. The boundary between the cartilage and surroundingtissue appeared more jagged than the control patient, and there was cartilage thinning(Figure 3.6A, red arrow). In the T2 map, there was an increase in T2 relaxation timesin the affected region, indicating increased water content and mobility due to collagendisorganization and damage. Furthermore, the spatial heterogeneity of T2 values alsoincreased. Literature has reported similar findings in individuals with OA or risk factorsof OA [23, 49, 52, 58]. In the T2∗ map, there was a decrease in T2∗ time (Figure 3.6B).It is possible that this decrease in time may be due to the fact that T2∗ is sensitiveto other decay processes not detected in T2. Decreasing T2∗ times at 3T have beenreported in increasing OA severity of hip joints [59]. Furthermore, Mamisch et al. hadreported decreased T2∗ and T2 times in femoral cartilage at 3T [55]. Although thisdecrease in T2 time was contrary to what we found, it is important to note that hisstudy was conducted on participants after microfracture therapy (surgical procedure inwhich collagen production is stimulated) while this participant did not report havingthis procedure. However, it is also possible that the observed increase in T2 may beattributed to the magic angle effect. Since T2∗ is a relatively new image contrast forarticular cartilage, there are very few clinical studies comparing the correlation of T2and T2∗ values in damaged cartilage, so findings from this project may contribute tothis growing field of research. For the frequency map, in the damaged area, there was anincreased variation in frequency, and there was a decrease in frequency (Figure 3.6C). Inthe SWI image, although the cartilage was thin, SWI offered good visualization of thethinning cartilage (Figure 3.6D). Hypointensity in the corresponding region of the T1-Weighted image and hyperintensity in the STIR image indicated the presence of boneedema and further confirmed the presence of cartilage damage (Figure 3.6 D, E).In the patient with suspected mild cartilage damage, there was a meniscal tear identifiedin the spin-echo sequence, and there may have been cartilage damage in the articularcartilage above this tear (Figure 3.7A). There was a minor decrease in T2 time andincrease in signal intensity in SWI; however, statistical analysis will need to be conductedin the future to check if these changes were significant (Figure 3.7 A, D). Future studieswould involve a larger cohort of patients with mild cartilage damage. No abnormalitieswere qualitatively observed in the T2∗ maps, T1-Weighted, and STIR images (Figure3.7 B, E, F). In the frequency map, a distinct decrease in frequency was observed in thesuperficial zone of the suspected area of damage. This suggests that frequency mapsmay have the potential to be a sensitive biomarker for cartilage damage.Early collagen disorganization may have a negative frequency shift, as demonstrated inthe above two patients. He and Yablonskiy’s model for frequency shifts had predictedand reported a positive frequency shift for myelin as lipids have a positive magneticChapter 3. Results and Discussion 34Figure 3.6: Comparison of T2 (A), T2∗ (B), Frequency Maps (C), Average SWI (D),T1-Weighted Image (E), and STIR (F) of the right knee, lateral condyle. Averagingwas done over 3 echo times of 5 ms, 10 ms, and 15 ms. (A) had a reconstructed voxelsize of 0.31 x 0.31 x 3.00 mm3. (B), (C), and (D) had a reconstructed voxel size of0.15 x 0.15 x 1.50 mm3. (E) and (F) had a reconstructed voxel size of 0.29 x 0.29 x3.00 mm3 and 0.3 x 0.3 x 3.50 mm3 respectively. Patient had characteristic signs ofOA such as cartilage thinning (red arrow).Chapter 3. Results and Discussion 35susceptibility [16, 17]. A negative frequency shift was reported for neurofilaments, whichare composed of proteins with a negative magnetic susceptibility. Similarly, literaturehas reported that a collagen fibril contains 104 to 106 collagen molecules and that eachcollagen molecule has a magnetic susceptibility of -1 x 10−25 JT−2 [60]. However, it isimportant to note that in the control participant, the frequency shifts in the femoralcartilage were most likely due to changes in fibre orientation relative to the externalmagnetic field. So, the negative frequency shift observed in the patients cannot be fullyattributed to the increase in isotropic components in tissue microstructure. At thisstage of the project, it is difficult to determine the degree in which both contribute tothe observed frequency shift. Future work in ex vivo cartilage tissue should be conductedto explore the details of this relationship on frequency shifts.Figure 3.8 demonstrated the potential use of SWI in imaging and evaluation of the kneeas a whole organ system. In one patient with characteristic signs of OA, there wereosteophytes at the peripheral regions of the femur, and cyst formation was observed inthe posterior region of the joint (Figure 3.8B). These clinical manifestations appearedmore distinct in the SWI image, in comparison to the average magnitude image (Figure3.8A). Although the cartilage was very thin in the regions of osteophyte growth, itwas still possible to see its boundaries because of the high resolution of the multi-echo3D GRE. In the SWI, there was also an enhancement of tissue details in the anteriorregion of the knee. In the patient with suspected mild cartilage damage, SWI of themedial and lateral condyles also revealed a distinct, striated pattern in the radial zoneof tibial cartilage that was previously observed in control participants (Figure 3.8 C, D).The medial condyle did not have a distinct contrast like the lateral condyle; however,striation details were still observable. Ex vivo MRI and pathology should be conductedto study the nature of these striations. The striation pattern along with the contrast andhigh resolution of the multi-echo 3D GRE could help in making maps of zonal thickness.All results presented in this thesis were qualitative as this was a proof-of-concept project,and patient records from the clinician have not been received yet. After receiving andreviewing these records, quantitative analysis will be conducted to ensure that the ob-servations made had statistical signifance.3.3 Other LimitationsThis was a proof-of-concept project, and controls were chosen for protocol developmentand healthy cartilage evaluation. So, they were not age- and BMI-matched with thepatients, and differences due to age or weight could not be taken into account. Oldercollagen has higher viscosity and faster gelation rates. It has been reported that duringChapter 3. Results and Discussion 36Figure 3.7: Comparison of T2 (A), T2∗ (B), Frequency Maps (C), Average SWI (D),T1-Weighted Image (E), and STIR (F) of the right knee, lateral condyle. Averagingwas done over 3 echo times of 5 ms, 10 ms, and 15 ms. (A) had a reconstructed voxelsize of 0.31 x 0.31 x 3.00 mm3. (B), (C), and (D) had a reconstructed voxel size of 0.15x 0.15 x 1.50 mm3. (E) and (F) had a reconstructed voxel size of 0.29 x 0.29 x 3.00mm3 and 0.3 x 0.3 x 3.50 mm3 respectively. Patient had a meniscal tear (red arrow).Mild cartilage damage above the tear was suspected.Chapter 3. Results and Discussion 37Figure 3.8: Enhanced features of SWI in 2 patients. Average magnitude (A) andcorresponding average SWI (B) of left knee, medial condyle in one patient. Red arrowsemphasized features of OA. Average SWI of right knee, lateral condyle (C) and medialcondyle (D) of patient suspected mild cartilage damage. Averaging was done over 3echo times of 5 ms, 10 ms, and 15 ms. (A), (B), (C), and (D) had a reconstructed voxelsize of 0.15 x 0.15 x 1.50 mm3.Chapter 3. Results and Discussion 38aging, the collagen monomer structures are altered, thus affecting the architecture of theresulting collagen fibers and the macroscopic properties of the network [61]. So, thesechanges along with cartilage may have affected the magnetic environment which wasdetected in the resulting images.Only sagittal scans were acquired in this study as they offer visualization of the entireknee joint, including the femor trochlea (knee hinge joint with patella), tibiofemoraljoint, and posterior convexities of femoral condyles. A predominantly medial-lateralpattern of cartilage loss for tibiofemoral OA has been reported in literature; however,this pattern would be more difficult to detect on sagittal images than coronal [62]. So,future studies may need to examine the possibility of combining coronal and sagittalscans into the protocol.It should be noted that the 3D multiple echo GRE sequence is sensitive to local field inho-mogeneities. Scanner imperfections and embedded objects may cause changes unrelatedto disease etiology in the acquired and reconstructed image (Figure 3.9). So, patientswho undergo treatments that require metallic fastenings (e.g. screws, bolts, plates) maybe unable to have their cartilage damage and treatment monitored accurately.Figure 3.9: Imaging artifacts due to field inhomogeneities. In this image, the screwembedded in the tibia is the likely source of the imaging artifacts (red arrows).Chapter 4ConclusionIn conclusion, knee MRI offers excellent visualization of cartilage and surrounding tis-sue structures. In particular, the 3D multiple echo GRE sequence offers high resolutionat relatively shorter acquisition times than the more conventional sequences such asspin-echo. SWI reveals certain details of the knee joint structures and may enhance thevisualization of any macroscopic damage (e.g. cyst formation). Meanwhile, frequencyshift information has the potential to be a sensitive biomarker of cartilage health. Theincorporation of these two techniques may aid in earlier detection of degenerative carti-lage damage, allowing timely treatment and monitoring and thus potentially improvingthe quality of life of at-risk patients.39Chapter 5Future WorkThis study was a proof-of-concept project; consequently, there are many fields thisproject’s direction can undertake for a more rigorous investigation.Firstly, all the MRI images presented in this thesis, especially those with suspectedtissue damage, need confirmation with a physician to provide qualitative assessment ofour findings. We have yet to receive patient clinical information from our collaborators.Comparison with histopathology will also be an accurate confirmation. Afterwards,quantitative and statistical analysis can be conducted for a more objective comparisonbetween patients and controls. Scans of the controls will need to be retaken with theexact scan parameters of the patients, and they should be age and BMI-matched sothese differences can be taken into account.Because of the high resolution and excellent SNR of 3D multi-echo GRE, segmentationof cartilage across several slices can be conducted, and 3D zonal thickness and cartilagesurface maps and can be generated. Future work can compare the specificity and sensi-tivity of these surface maps with arthroscopy which reportedly has high specificity andsensitivity of surface defects [35, 36]. Maps of cartilage thickness can also be comparedwith histology.Furthermore, since SWI has been reported to be effective at imaging strokes and brainhemorrhages, future studies can also investigate the use of SWI in imaging knee jointsof patients with blood disorders. Hemophiliac patients are more prone to hemarthrosis,bleeding in joint spaces such as the knee [63]. Meanwhile, patients with thalassemia, adisorder in which hemoglobin production is impaired, can experience complications fromiron overload which can accumulate in knee joints. So, SWI, frequency, and T2∗ mappingmay be able to detect these abnormalities through magnetic susceptibility changes.40Chapter 5. Future Work 41A thorough examination of the mechanisms of phase contrast in cartilage is also neededto determine the exact nature of the striated pattern observed in the radial zone ofSWI images and frequency shifts in patients with mild to severe cartilage damage. Exvivo cartilage tissue should be imaged using the 3D multi-echo GRE sequence and thencompared to electron micrographs and histology so that a more accurate and detailedrelationship between iron content, orientation of collagen, tissue microstructure, andMR phase and frequency can be established.Appendix ARescaling Factor CalculationT2∗ maps were computed automatically by the scanner. However, these T2∗ values wereoff by a scaling factor and had to be rescaled to their correct values by using parametersobtained from .PAR files. The rescaling factor was determined by the following formulas:FP =DVRS · SS, (A.1)DV = PV ·RS +RI, (A.2)where FP is the floating point value (correct value), DV is the displayed value onconsole, RS is the rescale slope, SS is the scale slope, RI is the rescale intercept, andPV is the pixel value in the file.For T2∗ maps, these values from the .PAR file were determined to be .03712 (RS), 2(RI), and 26.9408 (SS).42Bibliography[1] Donald B Plewes and Walter Kucharczyk. Physics of MRI: A primer. Journal ofMagnetic Resonance Imaging, 35(5):1038–1054, 2012.[2] Qiang S Xiang. Physics 404: Introduction to Medical Physics. 2013.[3] Peter A. Rinck. Magnetic Resonance in Medicine, The Basic Textbook of the Euro-pean Magnetic Resonance Forum, 2013. URL[4] Philip Nelson. Biological physics, volume 2. Freeman New York, 2004.[5] D C Noll. A Primer on MRI and Functional MRI, 2001. URL[6] David Griffiths. Introduction to Quantum Mechanics, volume 2. Pearson Education,2005.[7] John P Ridgway. Cardiovascular magnetic resonance physics for clinicians: part I.J Cardiovasc Magn Reson, 12(1):71, 2010.[8] Garry E Gold, Eric Han, Jeff Stainsby, Graham Wright, Jean Brittain, and Christo-pher Beaulieu. Musculoskeletal MRI at 3.0 T: relaxation times and image contrast.American Journal of Roentgenology, 183(2):343–351, 2004.[9] Govind B Chavhan, Paul S Babyn, Bejoy Thomas, Manohar M Shroff, and E MarkHaacke. Principles, techniques, and applications of T2*-based MR imaging and itsspecial applications. Radiographics, 29(5):1433–1449, 2009.[10] Christian Denk and Alexander Rauscher. Susceptibility weighted imaging withmultiple echoes. Journal of Magnetic Resonance Imaging, 31(1):185–191, 2010.[11] E Mark Haacke and Ju¨rgen R Reichenbach. Susceptibility weighted imaging in MRI:basic concepts and clinical applications. John Wiley & Sons, 2011.43Bibliography 44[12] Vanessa Wiggermann. MR Phase Post-Processing and MR Susceptibility Mapping- A comprehensive comparison. PhD thesis, The University of British Columbia,2013.[13] Wei Li, Bing Wu, and Chunlei Liu. Quantitative susceptibility mapping of humanbrain reflects spatial variation in tissue composition. Neuroimage, 55(4):1645–1656,2011.[14] Jeff H Duyn, Peter van Gelderen, Tie-Qiang Li, Jacco A de Zwart, Alan P Koretsky,and Masaki Fukunaga. High-field mri of brain cortical substructure based on sig-nal phase. Proceedings of the National Academy of Sciences, 104(28):11796–11801,2007.[15] Alexander Rauscher, Jan Sedlacik, Markus Barth, Hans-Joachim Mentzel, andJu¨rgen R Reichenbach. Magnetic susceptibility-weighted MR phase imaging ofthe human brain. American journal of neuroradiology, 26(4):736–742, 2005.[16] Xiang He and Dmitriy A Yablonskiy. Biophysical mechanisms of phase contrastin gradient echo MRI. Proceedings of the National Academy of Sciences, 106(32):13558–13563, 2009.[17] Dmitriy A Yablonskiy, Jie Luo, Alexander L Sukstanskii, Aditi Iyer, and Anne HCross. Biophysical mechanisms of mri signal frequency contrast in multiple sclerosis.Proceedings of the National Academy of Sciences, 109(35):14212–14217, 2012.[18] Samuel Wharton and Richard Bowtell. Fiber orientation-dependent white mattercontrast in gradient echo MRI. Proceedings of the National Academy of Sciences,109(45):18559–18564, 2012.[19] Christian Denk, Enedino Hernandez Torres, Alex MacKay, and AlexanderRauscher. The influence of white matter fibre orientation on MR signal phaseand decay. NMR in Biomedicine, 24(3):246–252, 2011.[20] CG Peterfy, A Guermazi, S Zaim, PFJ Tirman, Y Miaux, D White, M Kothari,Y Lu, K Fye, S Zhao, et al. Whole-organ magnetic resonance imaging score (worms)of the knee in osteoarthritis. Osteoarthritis and Cartilage, 12(3):177–190, 2004.[21] Johnson David C and Timothy S Johnson. What are some common knee problems?,2007. URL[22] Sharmila Majumdar. Advances in MRI of the Knee for Osteoarthritis. WorldScientific, 2010.Bibliography 45[23] AJ Palmer, CP Brown, EG McNally, AJ Price, I Tracey, P Jezzard, AJ Carr, andS Glyn-Jones. Non-invasive imaging of cartilage in early osteoarthritis. The bone& joint journal, 95(6):738–746, 2013.[24] NS Landnez-Parra, DA Garzn-Alvarado, and Vanegas-Acosta. Mechanical Be-havior of Articular Cartilage, Injury and Skeletal Biomechanics, 2012. URL[25] Joseph A Buckwalter, Henry J Mankin, and Alan J Grodzinsky. Articular Carti-lage and Osteoarthritis. INSTRUCTIONAL COURSE LECTURES-AMERICANACADEMY OF ORTHOPAEDIC SURGEONS, 54:465, 2005.[26] Van C Mow, Mark H Holmes, and W Michael Lai. Fluid transport and mechanicalproperties of articular cartilage: a review. Journal of biomechanics, 17(5):377–394,1984.[27] Wilfried Gru¨nder. MRI assessment of cartilage ultrastructure. NMR inBiomedicine, 19(7):855–876, 2006.[28] Joseph A Buckwalter and James A. Martin. Osteoarthritis. Advanced Drug DeliveryReviews, 58(2):150–167, 2006.[29] Leonore C Dijkgraaf, Lambert GM de Bont, Geert Boering, and Robert SB Liem.The structure, biochemistry, and metabolism of osteoarthritic cartilage: a reviewof the literature. Journal of oral and maxillofacial surgery, 53(10):1182–1192, 1995.[30] Connie Y Chang and Ambrose J Huang. Mr of articular cartilage lesions of theknee. Applied Radiology, 40(9):5, 2011.[31] David J Hunter, Yuqing Zhang, Jingbo Niu, Joyce Goggins, Shreyasee Amin,Michael P LaValley, Ali Guermazi, Harry Genant, Daniel Gale, and David T Fel-son. Increase in bone marrow lesions associated with cartilage loss: a longitudinalmagnetic resonance imaging study of knee osteoarthritis. Arthritis & Rheumatism,54(5):1529–1535, 2006.[32] Peter F Sharkey, Steven B Cohen, Charles F Leinberry, Javad Parvizi, et al. Sub-chondral bone marrow lesions associated with knee osteoarthritis. Am J Orthop,41(9):413–417, 2012.[33] Ali Guermazi, Jingbo Niu, Daichi Hayashi, Frank W Roemer, Martin Englund,Tuhina Neogi, Piran Aliabadi, Christine E McLennan, and David T Felson. Preva-lence of abnormalities in knees detected by MRI in adults without knee osteoarthri-tis: population based observational study (Framingham Osteoarthritis Study).BMJ: British Medical Journal, 345, 2012.Bibliography 46[34] John Bedson and Peter R Croft. The discordance between clinical and radiographicknee osteoarthritis: a systematic search and summary of the literature. BMC mus-culoskeletal disorders, 9(1):116, 2008.[35] Stuart Brooks and Mamdouh Morgan. Accuracy of clinical diagnosis in kneearthroscopy. Annals of the Royal College of Surgeons of England, 84(4):265, 2002.[36] Richard Nickinson, Clare Darrah, and Simon Donell. Accuracy of clinical diagnosisin patients undergoing knee arthroscopy. International orthopaedics, 34(1):39–44,2010.[37] Arthroscopy, 2014. URL[38] X-ray of knee arthritis, 2014. URL[39] C Bombardier, G Hawker, and D Mosher. The Impact of Arthritis in Canada:Today and Over the Next 30 Years. Arthritis Alliance of Canada, 2011.[40] David T Felson, Reva C Lawrence, Paul A Dieppe, Rosemarie Hirsch, Charles GHelmick, Joanne M Jordan, Raynard S Kington, Nancy E Lane, Michael C Nevitt,Yuqing Zhang, et al. Osteoarthritis: new insights. Part 1: the disease and its riskfactors. Annals of internal medicine, 133(8):635–646, 2000.[41] Eric J Strauss, Jennifer A Hart, Mark D Miller, Roy D Altman, and Jeffrey E Rosen.Hyaluronic Acid Viscosupplementation and Osteoarthritis Current Uses and FutureDirections. The American journal of sports medicine, 37(8):1636–1644, 2009.[42] Sascha Colen, Michel PJ van den Bekerom, Michiel Mulier, and Danie¨l Haverkamp.Hyaluronic Acid in the Treatment of Knee Osteoarthritis. BioDrugs, 26(4):257–268,2012.[43] Grace H Lo, Michael LaValley, Timothy McAlindon, and David T Felson. Intra-articular hyaluronic acid in treatment of knee osteoarthritis: a meta-analysis. Jama,290(23):3115–3121, 2003.[44] Steven Kurtz, Kevin Ong, Edmund Lau, Fionna Mowat, and Michael Halpern.Projections of primary and revision hip and knee arthroplasty in the United Statesfrom 2005 to 2030. The Journal of Bone & Joint Surgery, 89(4):780–785, 2007.[45] Wan-Ju Li, Richard Tuli, Chukwuka Okafor, Assia Derfoul, Keith G Danielson,David J Hall, and Rocky S Tuan. A three-dimensional nanofibrous scaffold forBibliography 47cartilage tissue engineering using human mesenchymal stem cells. Biomaterials, 26(6):599–609, 2005.[46] Wael Kafienah, Sanjay Mistry, Sally C Dickinson, Trevor J Sims, Ian Learmonth,and Anthony P Hollander. Three-dimensional cartilage tissue engineering usingadult stem cells from osteoarthritis patients. Arthritis & Rheumatism, 56(1):177–187, 2007.[47] Brian O Diekman and Farshid Guilak. Stem cell-based therapies for osteoarthri-tis: challenges and opportunities. Current opinion in rheumatology, 25(1):119–126,2013.[48] Hollis G Potter and Li F Foo. Magnetic Resonance Imaging of Articular CartilageTrauma, Degeneration, and Repair. The American journal of sports medicine, 34(4):661–677, 2006.[49] Timothy J Mosher and Bernard J Dardzinski. Cartilage mri t2 relaxation timemapping: overview and applications. In Seminars in musculoskeletal radiology,volume 8, pages 355–368. Copyright c© 2004 by Thieme Medical Publishers, Inc.,333 Seventh Avenue, New York, NY 10001 USA., 2004.[50] Xiaojuan Li, C Benjamin Ma, Thomas M Link, D-D Castillo, Gabrielle Blu-menkrantz, Jesus Lozano, Julio Carballido-Gamio, Michael Ries, and SharmilaMajumdar. In vivo T1ρ and T2 mapping of articular cartilage in osteoarthritisof the knee using 3T MRI. Osteoarthritis and cartilage, 15(7):789–797, 2007.[51] Gabrielle Blumenkrantz, Colleen T Lindsey, Timothy C Dunn, Hua Jin, Michael DRies, Thomas M Link, Lynne S Steinbach, and Sharmila Majumdar. A pilot,two-year longitudinal study of the interrelationship between trabecular bone andarticular cartilage in the osteoarthritic knee. Osteoarthritis and cartilage, 12(12):997–1005, 2004.[52] Gabby B Joseph, Thomas Baum, Julio Carballido-Gamio, Lorenzo Nardo, Wara-pat Virayavanich, Hamza Alizai, John A Lynch, Charles E McCulloch, SharmilaMajumdar, and Thomas M Link. Texture analysis of cartilage t2 maps: individualswith risk factors for oa have higher and more heterogeneous knee cartilage mr t2compared to normal controls-data from the osteoarthritis initiative. Arthritis ResTher, 13(5):R153, 2011.[53] G Chang, X Ding, O Sherman, E Strauss, L Jazrawi, MP Recht, and RegatteRR. High resolution morphologic imaging and T2 mapping of cartilage at 7 Tesla:comparison of cartilage repair patients and healthy controls. Magnetic ResonanceMaterials in Physics, Biology & Medicine, 26(6):539–548, 2013.Bibliography 48[54] Vanessa Wiggermann, Enedino Herna´ndez Torres, Irene M Vavasour, GR WayneMoore, Cornelia Laule, Alex L MacKay, David KB Li, Anthony Traboulsee, andAlexander Rauscher. Magnetic resonance frequency shifts during acute MS lesionformation. Neurology, 2013.[55] Tallal Charles Mamisch, Timothy Hughes, Timothy J Mosher, Christoph Mueller,Siegfried Trattnig, Chris Boesch, and Goetz Hannes Welsch. T2 star relaxationtimes for assessment of articular cartilage at 3 T: a feasibility study. Skeletal radi-ology, 41(3):287–292, 2012.[56] Yang Xia. Magic-angle effect in magnetic resonance imaging of articular cartilage:a review. Investigative radiology, 35(10):602–621, 2000.[57] Jachin Hung. The Effects of Cartilage Orientation on MR Phase and the Assessmentof Tibiofemoral Articular Cartilage through R2∗ Mapping, 2012.[58] Cynthia F Maier, Steve G Tan, Hari Hariharan, and Hollis G Potter. T2 quantita-tion of articular cartilage at 1.5 t. Journal of Magnetic Resonance Imaging, 17(3):358–364, 2003.[59] B Bittersohl, FR Miese, HS Hosalkar, M Herten, G Antoch, R Krauspe, andC Zilkens. T2 star mapping of hip joint cartilage in various histological gradesof degeneration. Osteoarthritis and Cartilage, 20(7):653–660, 2012.[60] Jim Torbet and Marie-Claire Ronziere. Magnetic alignment of collagen during self-assembly. Biochem. j, 219:1057–1059, 1984.[61] Sul-Suso J Torbet J Jeannesson P Sockalingum GD Yang Y. Wilson SL, Guilbert M.A microscopic and macroscopic study of aging collagen on its molecular structure,mechanical properties, and cellular response. FASEB journal, 28(1):1–12, 2014.[62] F Eckstein, Flavia Cicuttini, J-P Raynauld, JC Waterton, and C Peterfy. Magneticresonance imaging (mri) of articular cartilage in knee osteoarthritis (oa): morpho-logical assessment. Osteoarthritis and cartilage, 14:46–75, 2006.[63] E Carlos Rodriguez-Merchan. Musculoskeletal complications of hemophilia. HSSjournal, 6(1):37–42, 2010.


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