Transverse Relaxation based Magnetic Resonance Techniques ... · Magnetic resonance imaging (MRI) is a medical imaging technique that allows non-invasive assessment of the microstructure

Transverse Relaxation based Magnetic Resonance

Techniques for Quantitative Assessment of Biological

Tissues

Tonima S. Ali

BSc (Electrical Engineering), MSc (Biomedical Engineering)

Principal Supervisor: Dr Konstantin Momot

Associate Supervisor: Prof. Yin Xiao

School of Chemistry, Physics and Mechanical Engineering

Science and Engineering Faculty

Queensland University of Technology

2019

Submitted by Tonima Sumya Ali to the Science and Engineering Faculty, Queensland

University of Technology, in fulfilment of the requirements for the degree of Doctor

of Philosophy

ii

iii

Keywords

_____________________________________________________________________

Nuclear magnetic resonance, magnetic resonance imaging, transverse spin relaxation,

quantitative T2, magic angle effect, articular cartilage, collagen anisotropy, breast

cancer, mammographic density, portable NMR, post-traumatic osteoarthritis

iv

v

Abstract

_______________________________________________________________________________________________________

Magnetic resonance imaging (MRI) is a medical imaging technique that allows

non-invasive assessment of the microstructure and composition of biological tissues.

MRI is based on the principles of Nuclear Magnetic Resonance (NMR) that primarily

rely on the signal from 1H (the proton nucleus). The NMR relaxation decays originated

from 1H can be characterised by certain parameters like longitudinal relaxation time

(T1) and transverse spin relaxation time (T2). T2 is sensitive to the structural anisotropy

in the local imaging environment and to the distribution of tissue water, and in some

cases, to the distribution of 1H in tissues. Therefore, transverse relaxation based MRI

has the potential to indirectly probe both the structural components and the chemical

composition of biological tissues. Quantitative MRI allows the measurement of voxel-

based T2 and parametric maps of T2 by employing specialised imaging protocol and

subsequent computational analysis. For example, T2 weighted relaxation decays can be

achieved by using Curr-Purcell-Meiboom-Gill (CPMG) sequence. By incorporating

additional gradients in the CPMG sequence, Multi-Slice-Multi-Echo (MSME)

sequence can obtain T2 weighted echoes for multiple slices. Then, the voxel-specific T2

can be measured by iterative fitting of the transverse relaxation data to the mathematical

model of the T2 relaxation decay.

The aim of this thesis was to evaluate and to demonstrate the analytical capacity

of transverse relaxation based MR imaging techniques for non-invasive quantitative

evaluation of the structure and composition of biological tissues. Accordingly, three

semi-independent case-studies were conducted that evaluated the application of

quantitative T2 measurements and transverse relaxation based MRI in three different

tissue type scenarios. The first case study identified the collagen fibre alignment in

articular cartilage (AC) of kangaroo by applied a specialised T2 MRI technique called

magic angle effect. It was the first MRI study to investigate the collagen architecture in

kangaroo knee cartilages. Using MSME sequence and quantitative analysis in a high-

resolution micro-MRI (µMRI) system, voxel-based R2 (R2 = 1/T2) maps were measured

vi

from cartilage samples obtained from femoral hyaline cartilage, tibial hyaline cartilage

and tibial fibrocartilage of red kangaroo (Macropus rufus). This study introduced the

technique for measuring relative depth profile of anisotropic R2 or 𝑅2𝐴, which allowed

the identification of histological zones within cartilage samples. The histological zones

were defined based on the nature of collagen organization. Most notably, wide

superficial zones were identified in femoral hyaline cartilage samples, which also had

relatively narrow radial zones with low anisotropy in collagen distribution; tibial

hyaline cartilage samples had exceptionally wide radial zones with highly aligned

collagen fibres while those had very narrow superficial zones. Calcification was

identified on the radial zones of the tibial fibrocartilage samples, which gradually

increased closer to the subchondral bones. This was the first MRI study to investigate

the collagen architecture in kangaroo knee cartilages. The application of this work is

toward achieving a better understanding of the collagen scaffold in AC. This work also

identified the zone and cartilage specific collagen organization in kangaroo AC that

supports the extra-ordinary biomechanics of kangaroo knee. This in turn, may inspire

new designs for cartilage tissue engineering.

The second case study of this thesis developed a novel transverse relaxation

based technique for assessing the chemical composition of breast tissue by portable

NMR. Breast tissue mainly consists of two components: adipose tissue (fat) and

fibroglandular tissue (FGT), the prevalence of FGT is directly related to the tissue water

content. In conventional X-ray mammography, mammographic density (MD) of breast

tissue is determined by the FGT/adipose ratio. However, in this study, T2 weighted

transverse relaxation decays were measured from breast slices using CPMG sequence.

The relaxation curves were then converted into T2 distributions by inverse Laplace

transform. The presence of fat and water were unambiguously identified within the

samples by H2O-D2O replacement. T2 peaks centred approximately at 10 ms

corresponded to water and the T2 peaks centred close to 80 ms corresponded to fat. T2

distributions measured from high MD (HMD) regions featured two major peaks

corresponding to water and fat whereas only the fat-peaks were prominent in the T2

distributions measured from Low MD (LMD) regions. This study demonstrated that

transverse relaxation based quantitative analysis can detect the presence of adipose

tissue and FGT in breast, provide quantitative information on the relative prevalence of

these components, and can identify breast regions with HMD and LMD. The

vii

identification of HMD/LMD region by this novel method was in agreement with the

results of X-ray mammograms. In comparison to the conventional X-ray

mammography, the use of portable NMR may provide means for an informative and

safer alternative for MD screening while it promises an affordable price for MR based

breast examination.

The third and the final case study of this thesis aimed to detect the pathological

alterations in multiple tissues of an organ, both structural and compositional, by using

transverse relaxation based quantitative MRI techniques. Post-traumatic Osteoarthritis

(PTOA) was induced in rat knee joints by complete medial meniscectomy and the knee

joints were scanned every week for eight weeks using MSME sequence in a µMRI

scanner. During the progression of PTOA, three physical quantities: the thickness of

AC, T2 of AC, and T2 of epiphysis were identified to consistently evolve with strong

monotonic trends. The thickness of AC and its T2 were strongly correlated (p < 0.05)

throughout the study period. However, by making comparisons between these

quantities, in relation to the pathobiology of AC defined by histological analysis,

quantitative T2 of AC was identified as both an earlier and a more reliable indicator for

understanding the course of PTOA than AC thickness. Overall, the following

developmental pathway was identified to precede advanced PTOA: meniscal injury →

AC swelling → bone remodelling in subchondral and trabecular region → gradual

depletion of proteoglycan and loss of cellular density → severe proteoglycan loss and

free-water influx → erosion of the cartilage. This study has demonstrated that the use

of transverse relaxation based µMRI is sufficient to obtain adequate information about

the development of knee PTOA in rat models.

The findings presented in this thesis have evaluated the use of transverse

relaxation based analysis by MRI and NMR for assessing the structural and

compositional detail of biological tissues. The efficacy of transverse relaxation based

analysis was demonstrated by the results of three experimental case studies that have

identified previously unknown collagen architecture in kangaroo AC, introduced

transverse relaxation based technique for MD assessment by portable NMR and have

established a MRI-only measurement protocol for the evaluation of whole knee joint

that delineated the developmental pathway of PTOA. The imaging and analysis

protocols developed in these works are completely non-invasive and are transferrable

viii

to clinical scanners in principle. Further research investigations are required to assess

the suitability of these techniques for clinical application.

ix

x

Acknowledgement

_______________________________________________________________________________________________________

I have been fortunate to be surrounded by a remarkable group of people who

have made this challenging journey an enjoyable one. I am thankful to many for their

unwavering support during the last five years.

First and foremost, I thank my supervisor Dr. Konstantin Momot for guiding

me with patience and kindness throughout my research endeavour. I thank him for

giving me the freedom to pursue my own research interest, for motivating me and being

a mentor, and for being generous with time and knowledge. Thanks goes to QUT and

IHBI for allowing me to undertake the research investigations by providing my

scholarship (QUTPRA), travel and research funding.

I thank my associate supervisor Prof. Yin Xiao, Dr. Mark Wellard and Prof Rik

Thompson for their generous support, advice and encouragement, my collaborators,

Prof YuanTong Gu, Dr. Namal Thibbotuwawa, Dr. Monique Tourell and Dr. Indira

Prasadam for their contributions to my thesis. In addition, I thank the panel members

and the reviewers for taking interest in my work and for their constructive feedback.

I thank my colleagues in the MRI research group, Dr. Sirisha Tadimalla, Dr.

Monique Tourell, Monika Madhavi and Dr. Sean Powell for their friendship, support

and time.

I thank my parents, Prof Hafiza Khatun and Prof Md. Hazrat Ali for gifting me

the love of science and for teaching me the essence of perseverance. My sisters, Tania

Ali and Behnaz Ahmed, brothers, Tahseen Ali and Atiqur Rahman for their steady

support in this arduous journey. My sons, Reeyan and Raviv, my niece, Tazara and my

nephews Arziyan, Tahiyat and Ayaat for the abundant laughter and love. Most of all, I

thank my husband, Suffat Younus Ovee for his patience and his confidence in my

abilities, for sharing the overwhelming experience of early parenthood combined with

PhD research and for making this journey together. I dedicate this work to my family.

xi

xii

xiii

Statement of Original Authorship

_______________________________________________________________________________________________________

The work contained in this thesis has not been previously submitted to meet

requirements for an award at this or any other higher education institution. To the best

of my knowledge and belief, this thesis does not contain material previously published

or written by another person except where due reference is made.

Signature:

Date: 25-06-2019

QUT Verified Signature

xiv

xv

Table of Contents

_______________________________________________________________________________________________________

Abstract ......................................................................................................................... v

Acknowledgement ........................................................................................................ x

Statement of Original Authorship .......................................................................... xiii

Table of Contents ....................................................................................................... xv

List of Publications ................................................................................................ xviii

List of Abbreviation and Symbols ........................................................................... xxi

List of Figures .......................................................................................................... xxvi

Chapter 1: Introduction .............................................................................................. 1

1.1 Research Motivation ...................................................................................... 1

1.2 Thesis Aims and Objectives......................................................................... 11

1.3 Thesis Structure and Overview .................................................................... 12

Chapter 2: Background and Theory ........................................................................ 15

2.1 Imaging by Magnetic Resonance ................................................................. 16

2.1.1 Basics of Nuclear Magnetic Resonance ................................................. 16

2.1.2 RF Excitation ........................................................................................ 17

2.1.3 Spin Relaxation ...................................................................................... 19

2.1.3.1 Dipolar Interactions ........................................................................ 19

2.1.3.2 Chemical Exchange ........................................................................ 20

2.1.3.3 Free Induction Decay ...................................................................... 21

2.1.4 Signal Localization ................................................................................ 23

2.1.4.1 Slice-Selective Gradient ................................................................ 23

2.1.4.2 Frequency Encoding ...................................................................... 24

2.1.4.3 Phase Encoding.............................................................................. 25

2.1.5 K-space Acquisition and Image Reconstruction .................................... 25

2.1.6 Transverse Relaxation Analysis ............................................................. 26

2.1.6.1 Imaging Sequence for Transverse Relaxation based MRI ............ 26

2.1.6.2 T2 Mapping and Analysis ............................................................... 28

xvi

2.1.7 Specialised Scanner for MRI and NMR ................................................ 31

2.1.7.1 Micro-MRI for High Resolution MRI ........................................... 31

2.1.7.2 Single-Sided Portable NMR Scanner ............................................ 32

2.2 Knee Joint .................................................................................................... 33

2.2.1 Articular Cartilage ................................................................................. 33

2.2.2 Assessment of Collagen Fibre Architecture in AC by MRI .................. 36

2.2.3 Osteoarthritis in Knee Joint ................................................................... 39

2.2.3.1 Articular Cartilage – Effects of OA and Diagnosis by MRI .......... 40

2.2.3.2 Subchondral Bone – Anatomy, Effects of OA, and Diagnosis by MRI

.........................................................................................................42

2.2.3.3 Ligament – Anatomy, Effects of OA, and Diagnosis by MRI ....... 43

2.2.3.4 Menisci – Anatomy, Effects of OA, and Diagnosis by MRI .......... 43

2.2.3.5 Synovial Tissue – Anatomy, Effects of OA and Diagnosis by MRI

.........................................................................................................44

2.3 Mammographic Density............................................................................... 45

2.3.1 Mammographic Density – Clinical Significance and Methods of

Assessment ........................................................................................................... 45

2.3.2 Assessment of Mammographic Density using Portable NMR ............. 47

2.4 Transverse Relaxation in Biological Tissues ............................................... 48

Chapter 3: Transverse Relaxation based Assessment of Collagen Architecture

in Cartilage ............................................................................................................ 52

3.1 Prelude ......................................................................................................... 52

3.2 Statement of Co-author Contribution........................................................... 55

3.3 MRI magic-angle effect in femorotibial cartilages of the red kangaroo ..... 56

Chapter 4: Mammographic Density Assessment by Transverse Relaxation

based NMR ............................................................................................................74

4.1 Prelude ......................................................................................................... 74

4.2 Statement of Co-author Contribution........................................................... 76

4.3 Transverse relaxation-based assessment of mammographic density and breast

tissue composition by single-sided portable NMR .................................................. 77

Chapter 5: Detection of the Developmental Pathway of Osteoarthritis by

Transverse Relaxation based MRI ......................................................................... 101

5.1 Prelude ....................................................................................................... 101

5.2 Statement of Co-author Contribution......................................................... 104

5.3 Progression of Post-Traumatic Osteoarthritis in rat meniscectomy models:

Comprehensive monitoring using MRI ................................................................. 105

Chapter 6: Summary and Future Scope ................................................................ 132

xvii

References ................................................................................................................. 145

Appendix 1: Supporting Information for Chapter 4 ............................................ 170

Appendix 2: Preliminary Investigation for Chapter 5 ......................................... 175

A2.1 Methods...................................................................................................... 175

A2.1.1 Development of Rat OA Model ....................................................... 175

A2.1.2 MRI Protocol .................................................................................... 176

A2.1.3 Scanning by MRI and Image Processing ......................................... 180

A2.1.3.1 T1 weighted Imaging .................................................................... 181

A2.1.3.2 T2 weighted Imaging .................................................................... 183

A2.1.3.3 T2* weighted Imaging ................................................................... 185

A2.2 Results ........................................................................................................ 186

A2.2.1 Quantitative T1 Analysis................................................................... 186

A2.2.2 Quantitative T2 Analysis................................................................... 187

A2.2.3 Quantitative T2* Analysis................................................................. 191

A2.3 Conclusions ................................................................................................ 192

A2.4 References .................................................................................................. 194

xviii

List of Publications

_______________________________________________________________________________________________________

Refereed Publications in this Thesis

Ali, T.S., Thibbotuwawa, N., Gu, Y. and Momot, K.I., MRI magic-angle

effect in femorotibial cartilages of the red kangaroo. Magnetic Resonance

Imaging, 43 (2017) 66-73 (doi: 10.1016/j.mri.2017.07.010)

Ali, T.S., Tourell, M.C., Hugo, H.J., Pyke, C., Yang, S., Lloyd, T., Thompson,

E.W. and Momot, K.I., 2018. T2-based assessment of mammographic density and breast

tissue composition by single-sided portable NMR. Magnetic Resonance in

Medicine, 82(3) (2019) 1199-1213 (doi: 10.1002/mrm.27781)

Ali, T.S., Prasadam, I., Xiao, Y. and Momot, K.I., Progression of post-traumatic

osteoarthritis in rat meniscectomy models: Comprehensive monitoring using

MRI. Scientific Reports, 8(1) (2018) 6861 (doi: 10.1038/s41598-018-25186-1)

Related Refereed Publication throughout Candidature

Tourell, M.C., Ali, T.S., Hugo, H.J., Pyke, C., Yang, S., Lloyd, T., Thompson,

E.W. and Momot, K.I., T1‐based sensing of mammographic density using single‐sided

portable NMR. Magnetic resonance in medicine, 80(3) (2018) 1243-1251 (doi:

10.1002/mrm.27098)

Huang, X., Ali, T.S., Blick, T., Haupt, L., Lloyd, T., Thompson, E.W., Momot,

K.I. and Hugo, H.J., Correlation of Micro-CT with single-sided NMR T1 values as a

measure of mammographic density (2019) (to be submitted to Magnetic Resonance

Imaging)

https://doi.org/10.1016/j.mri.2017.07.010

https://doi.org/10.1038/s41598-018-25186-1

https://doi.org/10.1002/mrm.27098

xix

Conference Presentations

Oral Presentations

Tonima S Ali, Namal Thibbotuwawa, Yuan T Gu, Konstantin I Momot,

Collagen anisotropy in tibiofemoral cartilages of kangaroo using magic-angle-effect,

Oral Presentation, Australian Institute of Physics Congress, Brisbane, Queensland,

Australia (2016)

Tonima S Ali, Indira Prasadam, Yin Xiao, Konstantin I Momot, Quantitative

micro-MRI of murine models of PTOA, Oral Presentation, ISMRM Workshop on:

Osteoarthritis Imaging, Sydney, New South Wales, Australia (2017)

Poster Presentations

Tonima S Ali, Indira Prasadam, Yin Xiao and Konstantin I Momot,

Pathogenesis cascade of post-traumatic osteoarthritis in rat models by MRI, Australian

and New Zealand Bone and Mineral Society Congress, Brisbane, Queensland, Australia

(2017)

Tonima S Ali, Indira Prasadam, Yin Xiao and Konstantin I Momot,

Progression of Post-Traumatic Osteoarthritis in rat meniscectomy models:

Comprehensive monitoring using MRI, The Australian and New Zealand Society of

Magnetic Resonance Conference, Kingscliff, New South Wales, Australia (2017)

Monique C Tourell, Tonima S Ali, Patricia O’Gorman, Honor J Hugo, Thomas

Lloyd, Erik W Thompson, Konstantin I Momot, Oral Presentation, Development of

single-sided portable NMR methods for the sensing of mammographic density, Joint

Annual Meeting ISMRM – ESMRMB, Paris, France (2018)

Tonima S Ali, Monique C Tourell, Honor J Hugo, Chris Pyke, Yang, S.,

Thomas Lloyd, Erik W Thompson, Konstantin I Momot, T2-based assessment of

mammographic density and breast tissue composition by NMR MOUSE: a safe and

economical alternative, Poster Presentation, IHBI Inspires, Brisbane, Queensland,

Australia (2018)

xx

xxi

List of Abbreviation and Symbols

_______________________________________________________________________________________________________

α Flip angle

ω0 Larmor frequency

γ Gyromagnetic ratio

µ Magnetic moment

µ0 Magnetic permeability constant

µMRI Micro magnetic resonance imaging

𝛾𝑘, 𝛾𝑙 Gyromagnetic ratios of spins k and l

ρ(x), ρ(y) Spin distribution

Φ Phase angle

Θ Tilt angle of net magnetisation

θF Predominant angle of collagen fibre alignment relative to B0

ΔE Energy difference between two spin states

AC Articular cartilage

ACL Anterior cruciate ligament

ADC Apparent diffusion coefficient

AF Area fraction

B0 Static magnetic field

xxii

B1 RF pulse

BC Breast cancer

BIRADS Breast Imaging Reporting and Data System

BML Bone marrow lesion

COOH Carboxyl group

CPMG Curr-Purcell-Meiboom-Gill

CT Computer tomography

𝐷𝑘𝑙 Dipolar coupling constant between spins k and l

dGEMRIC Gadolinium-Enhanced MRI

DTI Diffusion tensor imaging

DWI Diffusion weighted imaging

ECM Extracellular matrix

FGT Fibroglandular tissue

FID Free Induction Decay

FOV Field of view

FS Fat suppressed

FSE Fast spin echo

Gz Gradient magnetic field along z axis

Gx Gradient magnetic field along x axis

Gy Gradient magnetic field along y axis

GAG Glycosaminoglycan

xxiii

gmT2 Geometric mean T2

GRE Gradient echo

1H Hydrogen nuclei; proton

H Plank’s constant

Ħ Modified Plank’s constant

HD Dipolar Hamiltonian

HMD High mammographic density

𝑰𝒌, 𝑰𝒍 Operators of spins k and l

𝑘𝑒𝑥 Rate of chemical exchange

LMD Low mammographic density

M Bulk magnetisation

Mz Longitudinal magnetisation

𝑀𝑧0 Thermal equilibrium value for M

Mxy Transverse magnetisation

MD Mammographic density

MRI Magnetic Resonance Imaging

MSME Multi slice multi echo

NMR Nuclear Magnetic Resonance

NMR-MOUSE A single sided portable NMR instrument

OA Osteoarthritis

PCL Posterior cruciate ligament

xxiv

PD Proton density

PG Proteoglycan

PLM Polarised light microscopy

PTOA Post traumatic osteoarthritis

𝒓𝒌𝒍 Dipole-dipole vector between spins k and l

R2 Transverse spin relaxation rate

𝑅2𝐼 Isotropic R2

𝑅2𝐴 Anisotropic R2

RDC Residual dipolar coupling

S MR signal readout

SEM Scanning electron microscopy

SO4 Sulphate group

SNR Signal to noise ratio

STIR Short tau inversion recovery

T1 Longitudinal spin relaxation time

T2 Transverse spin relaxation time

T2* Apparent transverse spin relaxation time

TE Echo time

TR Repetition time

Tpe Time for frequency encoding

xxv

xxvi

List of Figures

_____________________________________________________________________

Chapter 2

Figure 1. Nuclear magnetic moment vectors oriented randomly at thermal equilibrium

(A) and aligned in the direction of external magnetic field (B) ................................... 16

Figure 2. Zeeman splitting for a spin 12 system in presence of magnetic field B0 .... 17

Figure 3. Time evolution of bulk magnetization M. A constant external magnetic field

is applied along the z axis. The three dimensional behaviour of M over time is depicted

in black line in (A). M rotates around the z/z’ axis at the Larmor frequency and returns

to equilibrium, its vector behaviour along x, y and z axes are shown in (B), (C) and (D).

Longitudinal magnetization Mz grows along the z axis and transverse magnetization Mxy

decays as time progresses. ........................................................................................... 22

Figure 4. A pulse sequence for achieving voxel-specific frequency specificity in a 3D

imaging object. A linear gradient magnetic field Gz is applied along the z-axis for slice

selection; a linear gradient magnetic field Gx is applied along the x-axis for frequency

encoding, a linear gradient magnetic field Gy is applied for time Tpe along the y-axis.

...................................................................................................................................... 24

Figure 5. The de-phasing of isochromats that have different precessing speeds, after

the initial 90° RF pulse. ............................................................................................... 27

Figure 6. Basic CPMG pulse sequence for pure T2 imaging. One k-space line is

acquired for each phase encoding gradient, Gy. Due to the multiple 180° pulses and read

gradients, k-space lines are acquired in alternating directions. The dotted lines on the

right side of the figure indicates that the sequence continues for a predetermined number

of echoes within a single TR. ....................................................................................... 28

Figure 7. Formation of spin echoes by a CPMG sequence. Initial 90° RF pulse produces

a FID, which quickly disappears as the spins de-phase. The first 180° RF pulse at time

τ flips the equatorial plane and refocuses the spins that produce an echo at time 2τ.

Another 180° RF pulse is applied at time 3τ that generates an echo at time 4τ. .......... 29

Figure 8. A cartoon sketch of the human knee joint. Articular cartilage is represented

by the shaded regions lining the femoral condyle and tibial plateau. Courtesy of Dr.

Sirisha Tadimalla, Queensland University of Technology. ......................................... 34

Figure 9. Schematic diagram of the collagen fibre arrangement in articular cartilage.

From articular surface to bone: the superficial zone containing collagen fibres aligned

parallel to the surface, the transitional zone containing fibres with no particular

alignment, the radial zone containing collagen fibres aligned perpendicular to the

xxvii

articular surface, followed by a layer of calcified cartilage. Courtesy of Dr Monique

Tourell, Queensland University of Technology. .......................................................... 35

Figure 10. R2 anisotropy in AC: (A) R2 map of a bovine bone-cartilage plug oriented

perpendicular to the applied magnetic field B0; (B) R2 map of the same sample oriented

nearly at the magic angle (55°) relative to B0. In both maps, white corresponds to R2 =

0.13 ms−1. (C) R2 depth profiles constructed from maps A and B. Reprinted from [8],

with permission from IOS Press. ................................................................................. 38

Chapter 3

Figure 1. The typical anatomical locations (A–C) and representative T2-weighted (TE

= 8.46ms) MR images (D–F) of the samples used in the study: (A, D) femoral hyaline

cartilage; (B,E) tibial hyaline cartilage; and (C,F) tibial fibrocartilage. The cylindrical

samples were excised using a holesaw drill and the square sample was excised using a

hand-held saw, as described in Section 2.1. In (A), the sample used for the

measurements was taken from the upper-right of the two holes seen in the photograph;

the bottom-left hole is an auxiliary channel used to release the sample from the main

bone. ............................................................................................................................. 64

Figure 2. Representative maps and the corresponding relative-depth profiles of the

transverse relaxation rate constant (R2): (A–C) Maps and relative-depth profiles of R20

(sample orientation S = 0o); (D–F) same data for R255 (S = 55o); and (G – I) the relative-

depth profiles of the anisotropic component of R2 (R2A), computed as described in

section 2.5 (see Eqs. (3) and (4)). Each column represents the data from a single

imaging slice of a single cartilage sample: column 1, femoral hyaline cartilage sample

3 slice 1; column 2, tibial hyaline cartilage sample 1 slice 2; column 3, tibial

fibrocartilage sample 2 slice 2. The three-zone structure is readily apparent in each R2A

profile. .......................................................................................................................... 65

Figure 3. Average relative-depth R2 profiles obtained by averaging of the nine

respective individual profiles (three samples of each cartilage type, three imaging slices

per sample): (A, C, E) average profiles of R20 and R2

55; (B, D, F) average profiles of the

anisotropic component, R2AS, determined from R2

0 and R255 as described in section 2.5

(see Eqs. (3) and (4)). The three-zone structure is apparent in all R2A profiles. Note the

rapid increase of R255 between x=0.88 and x=1 in tibial fibrocartilage (“the attachment

sub-zone”, see Discussion). ......................................................................................... 67

Chapter 4

Figure 1. A, A photograph and B, a mammogram of a representative breast slice

(Patient 1-Slice 2) used in this study. B, The HMD and LMD regions specified by the

radiologist are shown as white circles. A, The black dashed squares show the HMD and

LMD regions excised from the full slice. HMD, high mammographic density; LMD,

low mammographic density ......................................................................................... 86

Figure 2. Histograms of the intensities of HMD and LMD regions in slice

mammograms of A, Patient 1-Slice 1; B, Patient 1-Sice 2; and C, Patient 1-Slice 3. The

horizontal axis represents the pixel greyscale values. The vertical axis shows the bin

xxviii

counts, or the abundance, of the respective greyscale values. HMD, high

mammographic density; LMD, low mammographic density ...................................... 87

Figure 3. Representative T2 distributions obtained from A, excised HMD and B,

excised LMD breast tissue samples. The samples shown were excised from Patient 1-

Slice 2. Each panel shows the T2 distribution in the native tissue (labelled “b”) and after

H2O-D2O replacement (labelled “a”). The peak near T2 = 10 ms, which disappears upon

H2O-D2O replacement, was identified as water. The measurements shown were taken

at the 4-mm tissue depth of the respective samples (Depth 2, P1-S2-D2). In these and

all subsequent ILT spectra, the T2 range from 0.1 ms to 1000 ms with logarithmic

spacing of bins was used. However, as no T2 contributions were observed for T2 < 3

ms, all ILT T2 distributions were plotted in the range from 1 ms to 1000 ms. The

boundaries of the T2 peaks were selected individually for each T2 spectrum, either as

the first bin whose value was above the baseline or as the bin closest to the minimum

between the two peaks. As an example, for spectrum “b” in panel A, the peak

boundaries were defined as 4.98 ms to 22.1 ms for water and 24.2 ms to 359 ms for fat.

In panel B, the respective boundaries were defined as 4.13 ms to 13.8 ms and 15.2 ms

to 394 ms for spectrum “b” and 20.1 ms to 327 ms for spectrum “a”. HMD, high

mammographic density; LMD, low mammographic density ...................................... 89

Figure 4. The T2 distributions obtained from the breast tissue regions excised from

the 5 slices used in the study. A, Excised HMD samples before H2O-D2O replacement;

B. same samples after H2O-D2O replacement; C, excised LMD samples before H2O-

D2O replacement; and D, same samples after H2O-D2O replacement. The individual

distributions represent measurements at a specific depth within a given slice: Patient 1-

Slice 1-Depth 1 (P1-S1-D1), Patient 1-Slice 1-Depth 2 (P1-S1-D2), Patient 1-Slice 2-

Depth 1 (P1-S2-D1), Patient 1-Slice 2-Depth 2 (P1-S2-D2), Patient 1-Slice 3-Depth 1

(P1-S3-D1), Patient 1-Slice 3-Depth 2 (P1-S3-D2), Patient 2-Slice 1-Depth 1 (P2-S1-

D1), Patient 3-Slice 1-Depth 1 (P3-S1-D1) and Patient 3-Slice 1-Depth 2 (P3-S1-D2).

HMD, high mammographic density; LMD, low mammographic density ................... 90

Figure 5. The T2 distributions obtained from the full breast slice and from the

excised regions of Patient 1-Slice 2: A, HMD region and B, LMD region. The full-slice

measurements were taken with the respective region positioned above the centre of the

NMR-MOUSE sensing coil. All the measurements shown are from the 2-mm tissue

depth (Depth 1, P1-S2-D1). HMD, high mammographic density; LMD, low

mammographic density ................................................................................................ 91

Figure 6. The T2 distributions obtained from the full breast slices used in this study.

A, HMD regions within the full breast slices; and (B): LMD regions within the full

slices. The measurements were taken with the respective region positioned above the

centre of the NMR-MOUSE sensing coil. The individual distributions represent the

measurements made at a specific depth within a given slice (see the legend of Figure 4

for the nomenclature). HMD, high mammographic density; LMD, low mammographic

density 92

Figure 7. The geometric mean T2 (gmT2) values and the area fractions (AF) of the

water and fat peaks A, measured from excised breast tissue samples and B, the

respective regions within the full slices. The gmT2 values represent the geometric-

average T2 of the water and fat, while the AF values reflect the relative prevalence of

xxix

the respective chemical species within the sample. This Figure includes the HMD and

LMD regions from all five breast tissue slices studied. HMD, high mammographic

density; LMD, low mammographic density ................................................................ 94

Chapter 5

Figure 1. The MRI scan locations are shown in an axial slice of control knee joint. The

position of coronal slice inside MRI gantry is shown in inset. Here 1, 2 and 3 refer to

the anterior, central and posterior slices, respectively. These slice orientations were

maintained for all scans of CTRL, MSX and CLAT joints. The femoral and tibial AC,

the menisci, cortical and trabecular bone of the epiphysis, ligaments and fat tissues were

clearly visible in the T2-weighted coronal slices acquired maintaining this protocol. The

schematic outline of the knee in the inset is reproduced from

https://en.wikipedia.org/wiki/Knee#/media/File:Knee_skeleton_lateral_ante--

rior_views.svg in accordance with the terms of the CC BY 2.5 license. ................... 109

Figure 2. Cartilage sections of medial condyles of CTRL and MSX joints (A) stained

with safranin-O fast green, which provided colour discrimination between bone and

cartilage. Here, the cartilage matrix proteoglycan is stained red, cell nuclei black,

cytoplasm grey green, and the underlying bone green [362]. Week 1 (CTRL) showed

abundant proteoglycan, week 4 (MSX) showed proteoglycan depletion while week 8

(MSX) showed major proteoglycan loss. Gradual thinning of cartilage was observed at

week 4 and week 8 as shown in (B).The Mankin scores of these slices are plotted in

(C). ............................................................................................................................. 110

Figure 3. The cartilage thickness measurement procedure shown in a T2 weighted MR

image of a MSX joint at TE = 12 ms (A). The straight line bordering AC is shown in

yellow and denoted by a. The perpendicular line drawn from femur to tibia, b, is shown

in blue in the inset, the nearest voxels of line b are shown in red. The corresponding T2

profile in (B) represents femoral cortical bone in pixel 1-4, cartilage in pixel 5, partial

volume of cartilage in pixel 6 – 8 and tibial cortical bone in pixel 8 – 9. The partial

volume effect observed in pixels 6 – 8 was corrected by using Eqs (2) and (3). Cartilage

thickness was computed by multiplying the total number of voxels representing

cartilage with voxel resolution (78 µm). All of the perpendicular lines b and

corresponding T2 profiles are shown in (C). The partial volumes of each profile was

corrected as above and a thickness was computed. The mean cartilage thickness was

computed by averaging the thicknesses of these intensity profiles. .......................... 111

Figure 4. Cartilage thickness and cartilage T2 evolution of MSX joints over the eight

week observation period post meniscectomy. The CTRL data of week 1 and week 8 are

also presented here. Cartilage T2 exhibited little change between week 1 and week 3, as

well as between week 4 and week 6. The data represent cartilage from the medial

condyle of central coronal slice. Data plotted as mean ± SE. .................................... 112

Figure 5. Mean T2 of medial epiphysis of CTRL, MSX and CLAT joints over the eight

week observation period post meniscectomy (A). The data represent epiphyseal T2

measured from central coronal slice location. The epiphyseal T2 of medial condyles of

anterior (slice 1), central (slice 2) and posterior slice (slice 3) locations for the MSX

joints are shown in (B). Data plotted as mean ± SE. ................................................. 113

xxx

Figure 6. Changes in the tissues of the medial condyles of CLAT joints, in comparison

to controls, during the eight week observation period post meniscectomy. Cartilage

thickness (A), cartilage T2 (B) and T2 of epiphysis (C) of CTRL and CLAT joints are

plotted for week 1-week 8 for central coronal slice location. Data plotted as mean ± SE.

.................................................................................................................................... 115

Figure 7. Changes in the tissues of the lateral condyles of MSX joints, in comparison

to lateral condyle controls, during the eight week observation period post

meniscectomy. Cartilage thickness (A), cartilage T2 (B) and T2 of epiphysis (C) of

CTRL and MSX joints are plotted for week 1-week 8 for central coronal slice location.

Data plotted as mean ± SE. ........................................................................................ 116

Appendix 1

FIGURE S1 Comparison of slice mammograms of a A, fresh and B, frozen breast

tissue slice. The two images are of the same physical slice; image A was obtained from

the fresh slice immediately after excision; image B was obtained from the frozen slice

following a 1‐ year 9‐month storage at –80⁰C. The slice shown was not used in the main

part of this study but is representative of the breast tissue slices used. Freezing-and-

thawing cycle causes slight changes in the topography of the sample and local

nonuniformity of the sample thickness; any areas thus affected were avoided when

selecting the measurement regions. The red circles show the HMD and LMD regions

of interest (ROIs) selected by the radiologist to match the same topographical features

in the fresh and frozen sample. The areas of the ROIs were A, 20.4 mm2 (LMD) and

3.8 mm2 (HMD); B, 13.2 mm2 (LMD) and 7.5 mm2 (HMD). The absorbed dose per

unit mass was A, 2452 ± 41 Gy (LMD) and 3052 ± 79 Gy (HMD); B, 2477 ± 76 Gy

(LMD) and 3089 ± 137 Gy (HMD). The absorbed doses are similar between the fresh

and the frozen sample, indicating that freezing and prolonged storage at –80⁰C do not

have a significant effect on the distribution of the mammographic density of the sample.

HMD, high mammographic density; LMD, low mammographic density ................. 170

FIGURE S2 Effect of the ILT regularization parameter α on the computed ILT spectra:

A, The main plot is a representative CPMG dataset with n = 4000 echoes. Each sample

point corresponds to one echo integrated from −8 s to +8 s from the echo centre. The

SNR value is 18, which is representative of the remaining data sets. The inset shows

the plot of χ2 versus the regularization parameter for a wide range of α values (see

section 2.4 in the main text). This plot is approximately L‐shaped. The corner of the

“L”, which was selected after visual inspection as the point of the apparent maximum

of the second derivative of the plot, corresponds to the optimal range of α in the ILT.

The circled points labelled b, c, and d in the inset correspond to the values of α used to

compute the ILT spectra in panels B, C, and D, respectively. B, An underregularized

ILT spectrum computed with α set too low. This makes the ILT smooth the physical

features of the T2 spectrum as well as the noise; the resulting oversmoothed spectrum

does not reliably distinguish between the fat and water T2 peaks). A properly

regularized ILT spectrum with the in the optimal range. This spectrum reliably

distinguishes between the fat and water T2 peaks without introducing spurious peaks).

An overregularized ILT spectrum with the α set too high, making the ILT overly

sensitive to noise and resulting in the introduction of spurious T2 peaks. HMD, high

mammographic density; ILT, inverse Laplace transform; LMD, low mammographic

density ........................................................................................................................ 171

xxxi

Appendix 2

Figure 1. Cartoon sketch of a mouse knee joint fixed by two Teflon plugs in a 25 mm

NMR tube. The sample is immersed in PBS. Imaging planes are shown by dotted lines

along axial (a), coronal (b), and sagittal (c) orientations. The schematic outline of the

knee in the inset is reproduced from

https://en.wikipedia.org/wiki/Knee#/media/File:Knee_skeleton_lateral_ante--

rior_views.svg in accordance with the terms of the CC BY 2.5 license. ................... 177

Figure 2. First echoes obtained by MSME sequence (TE = 6 ms) in axial (a), coronal

(b), and sagittal (c) orientation. .................................................................................. 178

Figure 3. Second echoes obtained by a MSME sequence (TE = 12 ms) of a coronal

slice with thickness of 0.25 mm (a), 0.5 mm (b), 0.75 mm (c), and 1 mm (d). ......... 179

Figure 4. Second echo of MSME sequence (TE = 12 ms) of a sagittal slice with FOV

of 20x20 mm (a), 25x25 mm (b), and 30x30 mm (c). ............................................... 179

Figure 5. Second echo of MSME sequence of a sagittal slice with image matrix of

128x128 (a), 256x256 (b), and 512x512 (c). ............................................................. 180

Figure 6. First echo of MSMEVTR sequence of a sagittal slice, the voxel selected for

analysis is highlighted in yellow (a), signal magnitude measured at 22 echo peaks (for

22 different values of TR) shown in blue and mathematically calculated fit shown in

red (b), and the fitting residuals of the fit in b (c). ..................................................... 182

Figure 7. The data fitting method for T1 weighted decay. This method was repeated for

the data acquired from every voxel of an imaging plane. The mathematical model was

defined for unconstrained fitting by non‐linear least squares method with optimization

based on trust-region algorithm. ................................................................................ 183

Figure 8. The first echo of MSME sequence of a coronal slice (TE = 6 ms), the voxel

selected for analysis is highlighted in yellow (a), signal magnitude at 25 echo peaks

shown in blue and mathematically calculated fit shown in red (b), and the residuals of

the calculated fit (c). .................................................................................................. 184

Figure 9. The first echo obtained by a MGE sequence of a coronal slice (a), the voxel

selected for analysis is highlighted in yellow, signal magnitude at 12 echo peaks shown

in blue and the mathematically calculated fit shown in red (b), and the residuals of the

calculated fit (c). ........................................................................................................ 186

Figure 10. The First echo obtained by MSMEVTR sequence, the region selected for

analysis is outlined by a blue rectangle (a), the R1 relaxation rate map computed from

22 echoes (b), and the results of the Run test results where 0 (black) = pass, 1 (white)

= fail (c). .................................................................................................................... 187

Figure 11. The first echo obtained by MSME sequence (TE = 6ms) along the coronal

plane, the region selected for analysis is outlined by a blue rectangle (a), the R2

relaxation map computed from 25 echoes (b), and the results of Run test where 0 (black)

= pass, 1 (white) = fail (c). ......................................................................................... 188

xxxii

Figure 12. R2 relaxation map of a coronal slice (a) and the result from edge detection

(b). .............................................................................................................................. 188

Figure 13. Voxels of AC are highlighted in purple (a) and the T2 distribution computed

from the data of corresponding voxels (b). Here the longitudinal axis presents T2 times

in milliseconds while the horizontal axis has no scale. ‘Jitter’ method has been used

for distributing the points to minimise overlaps. ....................................................... 189

Figure 14. In a T2 weighted echo (TE = 6ms), a region is outlined in the tibial epiphysis

by a closed ROI (a). The T2 distribution computed from the voxel T2 measurements

obtained from voxels within the outlined region (b). ................................................ 190

Figure 15. In a T2 weighted echo (TE = 6ms), trabecular region is outlined in the tibial

metaphysis and diaphysis by a closed ROI (a). The T2 distribution computed from the

voxel T2 measurements obtained from voxels within the outlined region (b). .......... 190

Figure 16. T2 distributions computed from tissues of two CLAT joints. .................. 191

Figure 17. The first echo obtained by a MGE sequence, the region selected for analysis

is outlined by a blue rectangle (a), the relaxation map of R2* computed from 12 echoes

(b), and the results of Run test where 0 = pass, 1 = fail (c). ...................................... 192

xxxiii

Chapter 1: Introduction

1


_____________________________________________________________________

1.1 Research Motivation

Medical imaging provides insight into the anatomical and physiological details

of human body for the purposes of diagnosis, disease monitoring and treatment.

Magnetic resonance imaging (MRI) is a medical imaging technique based on Nuclear

Magnetic Resonance (NMR) phenomenon that provides a non-invasive, quantitative

and often spatially resolved means of evaluating the structural organization and

chemical composition of organs and biological tissues. Many NMR and MRI

measurements rely on signal from 1H (the proton nucleus), which is present in

abundance in the water of cells and extracellular matrix (ECM) as well as in other tissue

components like fat. This allows excellent soft tissue contrast in MRI, which can be

further enhanced by manipulating the imaging parameters and pulse sequences. MRI

instrumentation and specialised sequences have evolved with features appropriate for

usage in research (~ 15 – 100 µm resolution) [1], pre-clinical (~ 20 – 200 µm resolution)

[2] and clinical setting (~ 500 µm – 3 mm resolution) [3-5]. Clinical MRI scans are

commonly acquired for disease diagnosis and treatment planning. For clinical scans,

the imaging sequences are designed for rapid image acquisition while maintaining a

resolution adequate for diagnostic purposes. Pre-clinical studies, requiring imaging and

visualization of living animal models of human diseases, are commonly used to

investigate the efficacy of disease diagnosis by MRI or to study the effects of treatment

procedures. For these purposes, MRI sequences are designed to maintain good

resolution that is adequate for detailed understanding of the subject under investigation.

At the same time, the sequences are also designed such that the scans complete within

a time frame that a living animal can remain steady or anesthetised. In research, MRI

is also used for ex vivo imaging where the sequences are designed for attaining very

high resolution with a high signal to noise ratio (SNR). High resolution NMR and MRI

have demonstrated the potential for non-invasive probing of structural features that


2

underpin the biomechanical functionality of tissues [6-9] and for evaluation of tissue

composition in normal and pathological conditions [10-17].

The imaging superiority of MRI, in comparison to other medical imaging

modalities, lies in its inherent ability to non-invasively analyse multiple tissue

structures in great detail and in a three-dimensional perspective, combined with the

ability to exploit a range of tissue properties for image contrast. The NMR signal

exhibited by 1H can be characterised by certain parameters, for example, longitudinal

relaxation time (T1) and transverse spin relaxation time (T2). The conventional clinical

interpretation of MRI focuses on qualitative assessment of anatomical features that are

visually apparent in MR images. Using this method of interpretation, diagnostic MRI

is commonly used to identify gross anatomical changes caused by diseases, which can

be visibly defined by differences in the pixel intensities in MR images. The pixel-values

of qualitative MR images are contrast weighted and these values are dependent on a

complex combination of proton density (PD), T1 and T2. In spite of the fact that the

relative contribution of these weighting factors can be varied by adjusting imaging

parameters, which can make MR images primarily PD weighted, T1-weighted or T2

weighted, the pixel-values are always influenced by the non-primary weighting factors

as well. Consequently, the pixel-values of conventional qualitative MR images are

informative only in relation to the other pixels of the same image. On the contrary,

quantitative MRI allows the measurement of biophysical parameters on a pixel-by-pixel

basis by utilizing the large degrees of freedom in designing the imaging sequences.

Although quantitative and qualitative MRI use the same technological platform and

offer complementary medical information, because of the above mentioned reasons,

conventional practice of qualitative MRI is relatively inefficient in comparison to

quantitative MRI for extracting MR information from tissues and organs.

The relaxation based quantitative MRI involves frequent sampling of MR

relaxation decays in order to accurately capture the nature of the relaxation decays. The

mathematical model of the MR relaxation decay is fitted individually to the relaxation

data acquired from each voxel (3D pixel). For example, in order to measure the T2 of a

specific voxel, the T2 weighted decay generated from that particular voxel is sampled

at regular intervals and the voxel-specific T2 is measured by iterative fitting of the

mathematical model of the T2 relaxation decay to the relaxation data. The measured

voxel T2 therefore exclusively represents the transverse relaxation characteristics of the


3

tissue or the combination of tissues that correspond to that particular voxel. A complete

T2 map of a MR imaging slice can be obtained by repeating the above procedure for

every voxel of the imaging plane. However, in order to accommodate for large number

of sampling points and high SNR, quantitative MRI requires longer acquisition time in

comparison to qualitative MRI. Quantitative MRI is capable of measuring various

biophysical parameters including PD, T1, T1ρ, T2 and T2* in a voxel-by-voxel basis.

Theoretically, these quantitative measures are absolute and are independent of the

experimental setting [18]. Consequently, the quantitative MR parameters are less

dependent on visual assessment and are comparable between different scanners and

between images acquired at different time points. The parametric maps obtained by

quantitative MRI can also be post-processed to explore MR information further, which

can then be used for various purposes, such as, image segmentation and analysis based

on biophysical properties, disease diagnosis based on altered biophysical parameters

and computation of distribution histograms of specific biophysical parameters from

voxels corresponding to one or more tissues.

The transverse relaxation decay and T2 are sensitive to the distribution of tissue

water, both intra-cellular and extra-cellular, due to the interactions of 1H population

with the local micro-environment. T2 measurements are also sensitive to the structural

anisotropy in the local imaging environment that causes restricted motion of water

molecules. Transverse relaxation, therefore, has the potential to indirectly probe the

composition and the structural organization of the tissue that hosts the water-molecule

or 1H population. T2 weighted MRI has been observed to produce excellent contrast

between fat (high signal intensity) and muscle (intermediate signal intensity) and

therefore is a popular choice for studying skeletal muscles [19]. When qualitative T2

MRI is used for diagnosis, the pathological conditions are identified based on the T2

weightings of the voxels. Pathological conditions induce structural and compositional

alterations of tissues at the cellular level. These alterations result in different T2

weightings for pathological tissues in comparison to the T2 weightings of the native

tissues. In clinical practice, qualitative T2 MRI is commonly used to investigate

pathological conditions in non-calcified tissues, such as, muscle [20], cardiovascular

tissues [21], breast [22, 23], tissues of the nervous system [24-26] and liver [27]. On

the other hand, quantitative T2 is more commonly used in µMRI studies to study the

structure and integrity of cartilage [5, 7, 28-31] and to investigate water micro-


4

compartments in normal neural tissues and in neurological conditions [12, 13, 17, 25,

32-34]. Quantitative T2 have also had limited use in analysing the mono- and multi-

exponential transverse relaxation decays measured from other types of tissues, such as,

the tissue of liver [35, 36], prostate [37], heart [38] and skeletal muscle [39, 40].

The primary goal of quantitative T2 MRI is to measure the voxel specific T2. In

theory, under the influence of the static magnetic field of a MR system, the transverse

component of the free induction decay (FID) of 1H population decays exponentially

with a T2 weighting and the voxel-specific T2 is attainable from the FID measured from

the same voxel. However, in reality, the FID of a spin system holds the T2

characteristics of a spin system only if the magnetic field experienced by the spins in

the imaging sample is perfectly homogeneous. The magnetic fields created by the MR

systems are often inhomogeneous due to the practical limitations of the MR hardware

and the distortion of the main magnetic field upon the placement of an imaging object.

The magnetic field inhomogeneity results in a distribution of precessional frequencies

for magnetised protons and therefore the spins quickly go out of phase with time. This

process leads to a faster decay of the bulk magnetization and the FID carries a T2*

weighting instead of the T2 weighting (T2* < T2). Nevertheless, pure T2 weighted

relaxation decay is achievable by using the specialised Curr-Purcell-Meiboom-Gill

sequence [41], commonly known by CPMG, which applies multiple refocusing pulses

to generate echoes whose amplitudes bear the T2 weighting. Although the use of CPMG

sequence permits the acquisition of MR relaxation data with uncontaminated T2

weighting, the acquisition time required for CPMG is substantially long and that limits

its use in multi-slice imaging. Multi-Slice-Multi-Echo (MSME) sequence, which is

built upon the original CPMG sequence for multi slice imaging, is commonly used in

research for measuring T2 in quantitative MRI. MSME sequence uses imaging gradients

for fast acquisition of the k-space data. However, because of using the imaging

gradients in MSME sequence, the R2 (1 𝑇2⁄ ) measured from MSME always contains a

diffusion contribution and the T2 obtained by MSME is always shorter than true T2 or

the T2 obtained by CPMG. For short T2 values, the contribution of diffusion is not

significant. For example, cartilage has short intrinsic T2 and the effect of diffusion can

be ignored when measuring cartilage T2 using MSME sequence. However, the diffusion

contribution can dominate in water-rich soft tissues (e.g. muscle) and the use of MSME

may be unsuitable for measuring voxel based T2 or for T2 mapping in those tissues.


5

This thesis discusses three case studies that were undertaken in order to evaluate

and to illustrate the application of quantitative T2 measurements and transverse

relaxation based MRI in different tissue type scenarios. The first case study

demonstrated the application of quantitative T2 measurements for identifying the site-

specific structural composition of normal femorotibial cartilages and to provide insights

into the biomechanical functions of cartilage in relation to its structural heterogeneity.

In mammals, the articular cartilage (AC) contains chondrocytes and a large proportion

of ECM. The ECM is principally composed of collagen (~15%-20%) [42],

proteoglycan (PG) (~3%-10%), lipid (~1%-5%) and water [7, 31, 43]. Collagen (type I

and type II) is the most abundant protein in body and a major constituent of the tissue

ECM that offers structural support for tissue cells [44, 45]. The cross linked collagen

network makes the structural scaffold of AC. The nature of collagen alignment and

distribution varies across the depth of AC and that typically creates three histological

zones in cartilage ECM: superficial zone, transitional zone and radial zone [46, 47]. AC

plays key roles in joint movement by creating a low friction protective barrier for

gliding and by distributing stress and transmitting loads to the underlying bones [48,

49]. It is postulated that the three-zone structure governs the response of cartilage to

dynamic loading during movement [50, 51]. In addition, the shear and tensile properties

of AC are also dependent on the underlying collagen scaffold in cartilage ECM [46,

52].

Collagen fibre organisation in AC can be interrogated by several experimental

techniques, most notably, scanning electron microscopy [53-56] and polarised light

microscopy [53, 54, 57]. Although these techniques provide high resolution (< 1 µm)

insight into the collagen alignment, and can be used to assess changes in the collagen

organisation, both of these techniques are destructive and therefore are unsuitable for

longitudinal studies or for in vivo evaluation. On the contrary, because of the fact that

collagen macromolecules restrict the movement of water molecules in cartilage ECM,

the quantitative T2 measurements obtained from AC are sensitive to the anisotropy in

collagen organization. In transverse relaxation based MRI, the anisotropic collagen

distribution in AC often results in an artefact that results in visually observable laminar

patterns [58-61]. The nature of this laminar appearance varies with the change in the

orientation of the imaged cartilage with respect to the static magnetic field used for


6

MRI [43]. This orientational dependence of the measured quantitative T2, on the

collagen anisotropy, is commonly known as the magic angle effect [1, 30, 62].

Previously, the magic angle effect has been useful in investigating collagen

architecture in AC by measuring cartilage T2 at specific orientations, in both clinical

and research setting [1, 28-30, 63, 64]. The method of obtaining quantitative T2 MRI is

both non-destructive and non-invasive. It has been shown that, by using an empirically

derived formula, the magic angle effect of T2 MRI can assess collagen fibre alignment

in ligaments [65] and in regions of AC [66, 67]. To date, the collagen architecture has

been studied in considerable detail in the AC obtained from human [28, 68], bovine [8,

29] and canine [62] joints. Consequently, attempts have been made to establish links

between the collagen organisations observed in AC samples with the inherent

biomechanical functionalities of the same tissue.

According to the results of the previous research investigations, the thickness of

the histological zones of AC, as well as the composition and organization of the major

molecular components, may vary across species and even across different sites in the

same joint [69-72]. The gait pattern of an animal sets the requirements for the functions

of its knee joint, which in turn impacts the structural make-up of its AC. Kangaroos

possess an exceptional locomotory behaviour that allow them to cross long distances

within a short time by repetitive hopping. The knee joints of kangaroo experience very

high ground reaction forces at every hop. The femorotibial cartilages of kangaroos are

examples of extremely robust, adaptive and durable cartilages. However, the nature of

the collagen architecture in kangaroo AC is still unknown. A detailed and quantitative

understanding of the collagen distribution in femorotibial cartilages of kangaroo may

benefit the assessment of the biomechanical capacities of the femoral and tibial

cartilages in kangaroo knee. The literary findings discussed above suggest that, with an

appropriate analytical approach, quantitative T2 MRI may be suitable for non-invasive

and site-specific assessment of the collagen scaffold – the structural framework of the

femorotibial cartilages of kangaroo.

The second case study discussed in this thesis has evaluated the suitability of

the use of quantitative T2 measurements for compositional assessment of breast tissue.

The breast tissue mainly consists of two components: fibroglandular tissue (FGT) - a

mixture of fibrous connective tissue (the stroma) and the glandular epithelial cells that


7

line the ducts of the breast (the parenchyma) and adipose tissue (fat). The density of

breast, commonly known as mammographic density (MD), is the measure of the

relative amount of FGT as opposed to the amount of adipose tissue in breast. The

measurement of MD is of particular clinical importance because high MD (HMD) has

been identified as a precursor of breast cancer (BC) that may estimate the risk for a

patient to develop BC in future [73-75]. Previous applications of transverse relaxation

based MRI have demonstrated that T2 is sensitive to the compositional heterogeneity of

biological tissues including breast tissues and that tissue T2 may be influenced by the

compositional alterations in tissues caused by pathological conditions [22, 25, 35, 76-

79]. Yet, the specific effect of tissue composition on T2 variation has not been evaluated

in these studies. In the presence of a pathological condition, it is also not possible to

isolate the exclusive effect of tissue composition on T2 variation due to the various

anatomical changes that occur during the development of the disease. Promisingly, the

evaluation of varying MD by T2 measurements provides a relatively simple analytical

problem where it is possible to understand the T2 measurements in relation to the

varying distribution of FGT/fat in the scanned tissue. Contrary to the composition of

cartilage, breast tissue has no known structural heterogeneity that may influence the T2

measured from breast. Therefore, transverse relaxation based MD assessment may

demonstrate the direct interrelation between FGT/fat composition and the

corresponding T2 measurements.

X-ray mammography is the current clinical standard for screening MD and BC

[80]. Although X-ray mammogram has been proven to be beneficial for BC detection,

it also exposes patients to ionizing radiation, which is harmful to patients.

Mammograms also suffer from other limitations such as projectional imaging artefact

and reduced sensitivity in dense breast, which sometimes result in erroneous diagnosis.

Conversely, in comparison to X-ray mammography, MRI is more sensitive in detecting

breast tumours or BC [81, 82]. It is also capable of producing more detailed information

concerning the MD and the extent, character and position of breast lesions [83-87]. MRI

results have shown good correlation with MD measurements acquired from

corresponding mammograms [79]. On the downside, MRI is significantly more

expensive than mammography. At present, a breast MRI scan is expected to cost

approximately $700 in Australia. BC is the most commonly diagnosed cancer among

females all over the world [88] and it is recommended that every woman between 50


8

and 74 years of age, should undergo a screening test for BC once every 2 years [80].

Therefore, in spite of the advantages of MRI, it is still not a feasible choice for routine

screening of such large population because of the associated cost involvement. The set-

up and maintenance cost of a clinical MRI unit is very high and is unlikely to reduce in

future. Consequently, MRI is only recommended for patients at high risk for BC,

patients with confirmed cases of BC and women who are extra-susceptible to ionizing

radiation. MRI is also performed in cases where mammogram results in poor diagnosis.

An alternative to clinical MRI is a single-sided desktop NMR system that employs the

same fundamental principles as MRI to probe the 1H within a sample. Portable NMR

instruments are designed as low-cost and low-maintenance units based on permanent

magnets [89-91]. The mobility and the low-cost of portable NMR has encouraged its

use in investigating silicone breast implants [92] and various biological tissues,

including skin [93], tendon [94], cartilage [95], and trabecular bone [96]. Due to the

location and the anatomy of the breast tissues, it is possible to employ such an

instrument for obtaining MD measurements in vivo.

Portable NMR instruments can measure transverse relaxation decays by using

CPMG sequence for scanning. However, only one transverse relaxation decay is

measured from the specimen right above the sensing area (~ 15mm x 15mm) [89, 97].

Multiple tissue components are expected to co-exist within that region and the

measured relaxation decay is likely to be a combined decay with multiple T2 relaxation

components. Such decays can be analysed using one dimensional Inverse Laplace

Transform, which decomposes the multi-exponential decay into a sum of mono-

exponential decays. The resulting T2 distribution contains distinctive T2 peaks where

each peak correspond to a unique tissue component. The distribution of T2 peaks show

the relative contribution of each T2 to the total NMR signal that can be interpreted as

the relative prevalence of each tissue component (with distinct T2) within the imaging

sample. In order to avoid ambiguity, the possible effect of diffusion contribution on the

measured T2 should be considered while interpreting the T2 distribution. Previously,

clinical MRI has been used to quantify the proportion of FGT in breast tissue [79].

Therefore, in principle, the T2 distribution obtained by portable NMR may also

demonstrate the FGT and fat composition in the tissue under examination. Recently,

using T1-based analysis, portable NMR has been successful in discerning between

breast tissue with HMD from low MD (LMD) [97]. Transverse relaxation based


9

analysis by portable NMR has the potential to determine the composition of breast

tissue and measure MD. The study of transverse relaxation based MD assessment by

portable NMR may aid in establishing a sensitive and inexpensive platform for MD

screening while demonstrating the effectiveness of using T2 measurements for

identifying specific chemical species (FGT/fat in this case) in breast tissue.

After the studies on the transverse relaxation based quantitative assessment of

the structural component and the chemical composition of biological tissues, the third

case study presented in this thesis aimed to evaluate the applicability of transverse

relaxation based quantitative MRI for comprehensive assessment of an organ - knee

joint, which consist multiple types of connective tissues, muscle and calcified tissues.

A well-established knee post-traumatic Osteoarthritis (PTOA) model was chosen for

this study. The goal of this component was to identify the alterations in the tissues of

the knee joint, caused by PTOA, from the measurements obtained by transverse

relaxation based MRI, and consequently identify the developmental pathway of PTOA.

Research of the last 20-25 years has demonstrated that OA is a whole joint disease and

that it is characterised by degenerative changes in joint structures including AC,

subchondral bone, menisci, synovial tissues, and ligaments. Currently, OA is the most

common joint disease worldwide and a leading cause of chronic pain and disability [98-

103]. OA is non-curable, and the optimal clinical outcomes in OA cases rely on

appropriate preventative measures or clinical intervention within the ‘treatment

window’ in early stages of the disease [104]. However, OA cases are commonly

reported after the patients present with joint pain and discomfort at the advanced stage

of the disease. Then, OA is diagnosed based on the physical examination and X-ray

radiography that primarily focus on the misalignment of bones in the affected joint,

which takes place after the AC had either partly or completely degraded. Consequently,

the manifestation of early OA as well as the pathogenesis cascade that define the

developmental pathway of OA remain elusive to clinical diagnosis. Improvement of

OA management requires detailed information on its initiation and tissue alterations at

different developmental stages of the disease.

The degradation of AC is often regarded as the structural hallmark of OA

progression. Osteoarthritis causes loss of PG in AC, which disrupts the pre-existing

collagen network and results in ECM degradation [46]. The collagen content is also

reduced in advanced OA [105]. Quantitative MRI exploits these macromolecular


10

changes to provide a quantitative understanding of the AC breakdown process.

Quantitative T2 measurements of cartilage are sensitive to the water content in AC and

to the integrity of the PG–collagen matrix. In a previous study, areas of damaged

cartilage were identified using quantitative T2 MRI, which showed that damaged

regions of cartilage had higher T2 values than usual along with lower cartilage volumes

and lower cartilage thicknesses [106]. Ex vivo studies on AC T2 have revealed

sensitivity of T2 imaging to changes in collagen content and distribution [107].

Measurements based on qualitative T2 MRI techniques have also been effective in

measuring changes in cartilage thickness resulting from OA [108-111]. In OA-affected

knee joints, bone marrow lesion (BML) and bone marrow edema (BME) have been

detected in the subchondral bone of both the tibia and the femur [112, 113]. The

presence of BML and BME is often correlated with the damage to neighbouring

cartilage [114, 115]. Previously, T2 based MRI has been used to identify and assess

BML in OA [98]. In the presence of OA, MRI has also been effective in identifying

abnormalities in ligaments [116-118], in detecting damage to menisci [119] and also in

assessing synovial inflammation [120].

Although OA is a whole-joint disease, previous OA-related research

investigations have mostly focused on individual features of OA, such as AC

degradation, BML, or meniscal or ligament injury. The interrelations of such changes

have not yet been established. Promisingly, a 3D T2 map is attainable for a whole knee

joint using MSME sequence while quantitative T2 MRI has the potential for diagnosing

OA-induced changes in multiple tissues of the knee joint. At the experimental level, an

animal model is appropriate for investigating the effects of OA on the tissues of the

knee joint. The use of a small animal will permit the use of micro-MRI (µMRI) system

for obtaining MR images with high resolution and good SNR. A transverse-relaxation

based longitudinal study of whole knee PTOA in an animal model, along with the study

of control joints, starting from the initiation of the disease, and continuing until the

disease reaches its advanced stage, may detect the tissue alterations that take place in

knee joint during this process. This information would be useful for attaining a

comprehensive understanding of OA development. When early OA-induced changes

are identified in a patient, this information would be particularly beneficial for initiating

site-specific and timely treatment to inhibit further progression of OA. In addition, this

study will demonstrate the effectiveness of transverse relaxation based MRI for


11

quantitative assessment of whole knee joints in normal condition as well as and in

pathological condition.

1.2 Thesis Aims and Objectives

The aim of this thesis was to evaluate and to demonstrate the analytical capacity

of transverse relaxation based MR imaging techniques for non-invasive quantitative

evaluation of the structure and composition of biological tissues in normal and

pathological conditions.

The specific objectives of this thesis were:

Investigate the collagen architecture in the femorotibial cartilages of red

kangaroo (Macropus rufus) by using magic angle effect of T2 MRI. In the

AC of bipedal and quadrupedal mammals, the collagen orientation and

distribution have been studied using magic angle effect [8, 28, 29, 62, 68] of T2

MRI. The shear and tensile properties of these AC were observed to be

dependent on the underlying collagen network in cartilage ECM [46, 52]. Here,

the aim was to measure R2 (= 1/T2) anisotropy maps of femoral hyaline cartilage,

tibial hyaline cartilage and tibial fibrocartilage of kangaroo, interpret the

collagen distribution in the respective cartilages and consequently identify the

biomechanical functionalities of these cartilages in relation to the collagen

architecture.

Assess the compositional make-up of breast tissue, identify the effect of

tissue composition on T2 variation and measure MD analogues quantities

from transverse relaxation decays obtained using portable NMR

instrument. MRI results have shown good correlation with MD measurements

acquired from matching mammograms [79]. In breast tissue, MD is determined

by the ratio of FGT to adipose tissue [75] while the prevalence of FGT is highly

correlated with the water content [75]. T2 is highly sensitive to the water content

and distribution in biological tissues [1, 10, 25, 29, 68, 121]. Quantitative T2

relaxation analysis in MRI has been used to characterise multiple water micro-

compartments within the same imaging voxels [12, 13, 25, 33, 122, 123].

Therefore, quantitative analysis of T2 NMR relaxation decays measured from


12

excised breast slices may have the capacity to determine the T2 values particular

to FGT and adipose tissue. This analysis may identify and quantify the FGT and

adipose tissue in the specimen. The relative prevalence of FGT/adipose tissue

can then be used to estimate the MD of the breast tissue.

Identify structural and compositional changes in knee joint tissues from

measurements obtained by transverse relaxation based MRI that define the

developmental pathway of post-traumatic OA (PTOA). Transverse

relaxation based MRI has been used to detect cartilage degradation in OA [106,

108-111] and to assess BML in the presence of OA [98]. MRI has also been

effective in identifying abnormalities in ligaments [116-118] and menisci [119].

Here, a rat PTOA model was chosen for studying the development of PTOA

due to its size (joint size small enough to fit inside the gantry of a µMRI scanner)

and its similarity to human PTOA. Examination by µMRI system allows high

resolution and good SNR required for quantitative assessment. Using transverse

relaxation based imaging techniques in a µMRI system, the damage and

degradation of all tissues in a rat knee joint can be monitored at regular intervals

from the initiation of PTOA to the advanced PTOA. The measurements thus

obtained can then be combined to gain a thorough understanding of the

pathogenesis cascade that precedes advanced PTOA.

1.3 Thesis Structure and Overview

The following chapters in this thesis are organised as:

Chapter 2 provides an overview of the theory and literature relevant to this

thesis. It begins by introducing the basics of NMR principle and the methods of

image formation in MRI. This is followed by a brief overview of the imaging

sequence and analysis procedures of transverse relaxation decays. µMRI

scanner and portable NMR instruments are then briefly discussed, which are

used for specialised MR imaging. This is followed by an overview of the

structure of the knee joint. Emphasis is given to the macromolecular

composition and collagen architecture of cartilage ECM. The magic angle

effect, a particular transverse relaxation based MRI technique used for

determining collagen alignment in cartilage is presented. Then, a literature


13

review is provided on the most common knee joint disease, OA. Knee joint

anatomy is discussed in conjunction with the effects of OA on the respective

tissues and associated MRI-based diagnosis techniques. The measurement

procedures of MD are then discussed followed by the analytical capacities of

portable NMR for investigating biological tissues. Finally, some applications of

transverse relaxation based MR are discussed that have been employed to assess

biological tissues and pathological conditions.

Chapter 3 presents a journal article [124] published in Magnetic Resonance

Imaging that demonstrates the use of magic angle effect for identifying the

collagen alignment in the ECM of the femorotibial cartilages of red kangaroo

(Macropus rufus). Spatially resolved R2 maps were measured from femoral

hyaline cartilage, tibial hyaline cartilage and tibial fibrocartilage at 0° and 55°

(magic angle) orientation with respect to the static magnetic field of µMRI

scanner. R2 anisotropy profiles were computed for each cartilage type and the

associated collagen organisation was identified. Based on the variations in

collagen arrangement in these cartilage samples, the characteristic collagen

arrangement suitable for particular biomechanical functions were classified.

Chapter 4 is in the form of a journal article that has been published in Magnetic

Resonance in Medicine [125]. This chapter evaluates the potential of the use of

transverse relaxation measurements by portable NMR for measuring tissue

composition and MD analogous quantities. Using CPMG sequence, T2

relaxation decays were measured from excised breast slices. Each relaxation

decay was converted into a T2 distribution using one dimensional inverse

Laplace transform. The T2 peaks corresponding to FGT (water) and adipose

tissue (fat) were unambiguously identified using H2O-D2O replacement and the

relative prevalence of water/fat were estimated from the distribution of T2 peaks.

The densities of breast estimated from the relative prevalence of FGT and fat

were compared against the MD measurements previously specified from X-ray

mammograms by a clinical radiologist.

Chapter 5 is in the form of a journal article [97] published in Scientific Reports

that describes an experimental study of a knee PTOA model in rats that was

examined by transverse relaxation based µMRI to identify the


14

pathophysiological pathway of OA development. PTOA was initiated in rat

knee joints by complete removal of medial meniscus (meniscectomy). The

whole rat knee joints were examined at weekly time points for 8 weeks where

week 0 marked the time of meniscectomy and week 8 marked advanced PTOA.

All tissues of the knee joints were examined by transverse relaxation based MRI

to identify alterations in tissues that evolved with the development of OA over

time. The methods developed in this study illustrated the analytical capabilities

of transverse relaxation based quantitative µMRI, which showed efficacy in

early diagnosis of PTOA and provided information on the anatomical and

compositional changes in knee joint tissues during OA development. It also

demonstrated the use of quantitative T2 as a biomarker for PTOA progression.

Chapter 6 summarises the works presented in this thesis and provides

directions for future research investigations in this area.

Chapter 2: Background and Theory

15


_____________________________________________________________________

Magnetic Resonance Imaging (MRI) and the scanning by Nuclear Magnetic

Resonance (NMR) are based on the same phenomenon of NMR. The signal that results

from the NMR exhibited by the hydrogen nucleus (1H) in tissue can be characterised

by certain parameters, for example, longitudinal relaxation time constant, T1 and

transverse relaxation time constant, T2. These relaxations/quantities are sensitive to the

local chemical and structural micro-environment experienced by the 1H population.

Therefore, careful evaluation of longitudinal and transverse relaxations, measured from

a biological tissue, allow the investigation of the structural organization and chemical

composition of the native tissue. This thesis focuses on the applications of transverse

relaxation based techniques that have been developed for the assessment of particular

biological tissues and pathological conditions. Therefore, the literature and theory

relevant to this thesis begins by describing the basics of NMR and the methods of image

formation using magnetic resonance. This is followed by a brief discussion on the

imaging sequence used to obtain transverse relaxation decays and the associated

analysis procedures. This chapter also describes two instruments used for specialised

MR imaging: micro-MRI (µMRI) scanner and portable NMR scanner that have been

used in the studies presented in this document. The second part of this chapter provides

a basic understanding of the tissues of knee joint and of Osteoarthritis (OA), which is

the most common disease of knee joint. It reviews literature on magic angle effect, a

particular transverse relaxation based phenomenon that can be used for determining

collagen alignment. It further explores the MRI based assessment methods used in

literature to evaluate OA. Finally, the third part of this chapter discusses NMR and

MRI-based techniques used to assess mammographic density (MD) and the plausible

application of portable NMR for evaluating MD in vivo.


16

2.1 Imaging by Magnetic Resonance

2.1.1 Basics of Nuclear Magnetic Resonance

NMR and MRI interpret the collective behaviour of an ensemble of a large

number of nuclei. Only 1H or proton in water molecules and adipose tissue are

considered in the context of this thesis. The magnetic behaviour of 1H can be modelled

using classical physics [126]. A proton has odd atomic number and odd atomic weight

and it possesses an angular momentum known as spin. Having a non-zero spin, a 1H

creates a magnetic field, known as magnetic moment μ, analogous to a bar magnet

[127]. At thermal equilibrium, these magnetic moment vectors are randomly oriented

due to thermal random motion and their vector sum approaches zero. When magnetic

moments are exposed to a strong static magnetic field, spin orientation is quantised

along the external magnetic field while spin transverse components remain random.

Some of the spins may then take one of the two possible orientations: parallel and anti-

parallel as shown in Fig. 1.

Figure 1. Nuclear magnetic moment vectors oriented randomly at thermal equilibrium (A) and

aligned in the direction of external magnetic field (B)

The energy difference between the two spin states of alignment is expressed by

𝛥𝐸 = 𝛾ħ𝐵0, (1)

where B0 is the strength of the applied magnetic field, γ is the gyromagnetic ratio with

a nucleus dependent value and ħ is the Plank’s constant divided by 2π. The nonzero

difference in energy level between two spin states is known as the Zeeman splitting

phenomenon as illustrated in Fig. 2.


17

The parallel spins are at a lower energy state with higher stability. At equilibrium, it

results in an uneven spin distribution between the two spin states with higher number

of spins in parallel orientation as defined by the Boltzmann distribution. The population

difference between the two spin states generates an observable macroscopic

magnetization M along the direction of the external magnetic field. By common MRI

convention, both B0 and M are aligned along the z-axis of the imaging system.

Figure 2. Zeeman splitting for a spin 1

2 system in presence of magnetic field B0

Under the influence of an external static magnetic field, the angular frequency

ω0 of a spin system is defined by the Larmor equation:

𝜔0 = 𝛾𝐵0. (2)

Here, B0 is the strength of the external static magnetic field and γ is the

gyromagnetic ratio. The angular frequency ω0 is the Larmor frequency, which is

linearly dependent on both B0 and γ. When multiple spin systems co-exist in a system,

as observed in a biological environment, the Larmor frequency is the physical basis for

achieving nucleus specificity. With known γ for 1H and by using a specific strength of

B0, the ω0 sensitive to 1H spin system can be computed and thus targeted for imaging.

2.1.2 RF Excitation

While the system is under the influence of B0, the magnetic moments precess

with a specific longitudinal orientation but with random phases. As a result, the

combined transverse magnetization component (Mxy) remains null. In order to establish

phase coherence among the precessing spins, an additional temporary magnetic field

B1 is applied. B1 is generated by Radio Frequency (RF) pulses; it is short-lived and


18

oscillates in the RF range. For conceptual simplicity, a rotating frame of reference is

commonly used in describing RF pulses and signals resulting from magnetization. This

rotating frame is a three-dimensional co-ordinate system where the orthogonal axes of

the frame is denoted by x’, y’ and z’ and the associated unit vectors are denoted by i’,

j’ and k’. The transverse plane of the rotating frame is assumed to rotate clockwise at

the Larmor frequency, ω0.

The time-dependent behaviour of bulk magnetization M in response to B1 is

described by the Bloch equation [127, 128]:

𝑑�⃑⃑�

𝑑𝑡= 𝛾�⃑⃑� × �⃑� 1 −

𝑀𝑥

𝑇2 𝑖→ −

𝑀𝑦

𝑇2 𝑗→ −

𝑀𝑧−𝑀𝑧0

𝑇1 𝑘→ . (3)

Here, 𝑀𝑧0 is the thermal equilibrium value for Mz in the presence of B0 only. Mx,

My and Mz are the magnetisation component of M along x, y and z axis, respectively. T1

and T2 are time constants characterising the relaxation process of a spin system after

the system has been perturbed by the magnetic field B1. For simplification, the

behaviour of M can be analysed in two steps: excitation during the RF pulse and

relaxation after the RF pulse is over.

As the duration of an RF pulse is very short compared to both T1 and T2, the

Bloch equation takes the following form during an RF excitation period [127]:

𝑑�⃑⃑�

𝑑𝑡= 𝛾�⃑⃑� × �⃑� 1 (4)

Ideally, B1 is applied with an angular frequency ωrf, which is equal to the

resonance frequency of the spin system (ωrf = ω0 = γB0). When B1 is applied along the

x’, the magnetization of a spin system with a single isochromat can be described by the

following list of equations [127]. Isochromat is the group of 1H that shares the same

resonance frequency in a 1H spin system:

𝑀𝑥′(𝑡) = 0 (5)

𝑀𝑦′(𝑡) = 𝑀𝑧0 sin (∫ 𝛾 𝐵1 (�̂�)𝑑�̂�

𝑡

0) 0 ≤ t ≤ τp (6)

𝑀𝑧′(𝑡) = 𝑀𝑧0 cos (∫ 𝛾 𝐵1 (�̂�)𝑑�̂�

𝑡

0) 0 ≤ t ≤ τp (7)

As shown above, a RF pulse along x’ tips M away from its original position

along z’ axis and makes M precesses about the x’ axis. The angle between M and the


19

positive z/z’ axis is known as the flip angle α. The value of α depends on both the

strength of B1 and the exposure time [127].

𝛼 = ∫ 𝛾𝐵1(𝑡)𝑑𝑡𝜏𝑝

0 (8)

For the MRI sequences that were used in the studies discussed in this thesis, the

common choices for flip angles were 90˚ and 180˚.

2.1.3 Spin Relaxation

The decay of NMR signal to the equilibrium is described by spin relaxation,

which is governed by interactions of the spins with one another and with the

surrounding environment. There are several mechanisms for spin relaxation such as

dipolar interactions, chemical exchange, J-coupling and quadrupolar coupling [129,

130]. The contributions of these mechanisms to spin relaxation depends on the state

and anisotropy of the sample. In a water molecule, the 1H are magnetically equivalent

and the precision frequency of the 1H spins in the ensemble is the same. Additionally,

the spin-rotation interactions can be neglected in water for temperatures below 373 K

[131]. Therefore, dipolar interaction is the dominant relaxation mechanism for spin

relaxation of 1H spins in bulk water [131, 132]. Additionally, different chemical

environments coexist in biological tissues where spin relaxation may occur as a result

of exchange of spins between chemically different environments.

2.1.3.1 Dipolar Interactions

Dipolar interaction, also known by dipolar coupling, refers to the direct

interactions between two magnetic dipoles. For example, if a spin pair is considered,

each spin creates its own magnetic field while it also experiences the magnetic field of

the other spin. Their roles are reversible, which results in pair-wise interactions between

all spins in a spin ensemble. The interaction energy between two spins k and l can be

expressed by the dipolar Hamiltonian [130, 133] as below:

𝐻𝐷 = µ0

4𝜋∑

𝛾𝑘𝛾𝑙ħ

𝑟𝑘𝑙3 {𝑰𝑘. 𝑰𝑙 − 3

(𝑰𝑘.𝒓𝑘𝑙)(𝑰𝑙.𝒓𝑘𝑙)

𝑟𝑘𝑙2 }. (9)

Here, µ0 is the magnetic constant or permeability of free space, γk and γl are the

gyromagnetic ratios of the coupled pair of nuclei, rkl is the distance between the spins,

and Ik and Il are the corresponding spin operators. The interaction energy HD is


20

inversely related to the cube of the distance between the spins. Accordingly, the

strongest dipolar interaction exists between 1H of the same water molecule, which is

known as intramolecular dipolar coupling. Conversely, dipolar interactions between

spins of different molecules, which are lower in energy, are known as intermolecular

dipolar couplings. The strength of dipolar coupling, expressed by Dkl, also depends on

the direction of the dipolar vector, the vector that joins the two spins involved in dipolar

coupling. When θkl is the angle between the dipolar vector and the static magnetic field

B0, Dkl can be expressed by:

𝐷𝑘𝑙 = µ0

4𝜋 𝛾𝑘𝛾𝑙ħ

4𝑟𝑘𝑙3 (1 − 3𝑐𝑜𝑠2𝜃𝑘𝑙) (10)

As shown in the above equation, the dipolar interaction energy varies with the

change in the direction of the dipolar vector. The direction of the dipolar vector changes

due to molecular tumbling that eventually results in spin relaxation. When the

molecular tumbling is rapid and unrestricted, as observed in a “free” water pool, dipolar

couplings average to zero. Conversely, when molecular motion is restricted, as

sometimes observed in biological tissues, the dipolar couplings do not average out. The

remaining dipolar coupling is called residual dipolar coupling (RDC). However, at an

angle 𝜃𝑘𝑙 = 𝑐𝑜𝑠−1 (1√3

⁄ ) ≈ 54.7°, the strength of the dipolar interaction becomes zero

and therefore do not contribute to spin relaxation. This particular angle is known as the

magic angle (54.7°).

2.1.3.2 Chemical Exchange

B0 induces current in the electron clouds surrounding each 1H spin, which

induces a magnetic field. This, in turn, changes the total magnetic field felt by the spin.

Consequently, the precession frequency of a spin depends on its location in a molecule

or on its chemical environment. The exchange of spins between different populations

that represent different chemical environments cause spin relaxation [134]. When two

chemically distinct pools, A and B, are involved in chemical exchange, the dynamic

equilibrium equation is expressed by:

𝐴 ⇄ 𝐵. (11)

Because the chemical environment (atoms surrounding the spins) is different

for each pool, the resonance frequency of the two pools differ by Δω. The chemical


21

shift between the pools is then defined by Δω/ω0. In the absence of any chemical

exchange, these pools are represented by individual spectral lines in the NMR spectrum

that are separated by Δω. If the exchange rate is slow so that the rate of exchange of

spins (kex) between the pools is significantly lower than the chemical shift (kex << Δω),

then the spins spend enough time in each pool for the resonance frequencies to be

observed distinctly in the NMR spectrum. Conversely, if exchange rate is fast (kex >>

Δω), only a single resonance in observed at the population-weighted resonance

frequency of the two pools. In both situations, the spins lose coherence as they are

exchanged between chemically different pools, which results in spin relaxation. In this

thesis, the pools A and B will predominantly represent the free and bound water pools

of the extra-cellular matrix (ECM) water [7].

2.1.3.3 Free Induction Decay

After the application of a RF pulse, if no other external forces are applied, the

spin system returns back to its thermal equilibrium state by spin relaxation. It involves

simultaneous return of the longitudinal and transverse components of magnetization to

equilibrium. Given that B0 is always present, the relaxation process is characterised by

the precession of M about the B0 field, which is known as the free induction decay

(FID). The transverse and longitudinal magnetization components are expressed by the

following equations [127].

𝑀𝑥′𝑦′(𝑡) = 𝑀𝑥′𝑦′(0+)𝑒−

𝑡

𝑇2 𝑒−𝑖𝜔0𝑡 (12)

𝑀𝑧′(𝑡) = 𝑀𝑧′0 (1 − 𝑒

−𝑡

𝑇1) + 𝑀𝑧′(0+)𝑒−

𝑡

𝑇1 (13)

Here 𝑀𝑥′𝑦′(0+) and 𝑀𝑧′(0+) are the magnetization components on the

transverse plane and along the z’ axis immediately after the RF pulse and 𝑀𝑧′0 is the

thermal equilibrium value for Mz in the presence of B0 only. T1 and T2 are time

constants. With time, Mz’ recovers to its original magnitude and Mx’y’ is diminished by

the relaxation process. An example of this process is depicted in Fig. 3. Here, the initial

90° RF pulse (α = 90°) was applied along the x’ axis, which repositioned the bulk

magnetization M along y’ axis. The evolution of Eqn. 12 and 13 are illustrated in a

three-dimensional space in Fig. 3A. The orthogonal magnetization components are


22

shown in Fig. 3B-D. Simpler decay curves, as shown by the black dotted lines, can be

obtained by using the rotating frame of reference as mentioned previously.

Figure 3. Time evolution of bulk magnetization M. A constant external magnetic field is

applied along the z axis. The three dimensional behaviour of M over time is depicted in black

line in (A). M rotates around the z/z’ axis at the Larmor frequency and returns to equilibrium,

its vector behaviour along x, y and z axes are shown in (B), (C) and (D). Longitudinal

magnetization Mz grows along the z axis and transverse magnetization Mxy decays as time

progresses.

Longitudinal spin relaxation or T1-relaxation describes the return of the

longitudinal component of the magnetisation, Mz’, back to its equilibrium state, 𝑀𝑧′0 .

Dipolar interactions contribute to T1 relaxation while it is dependent on the population

difference between the spin states. The rapid transitions between the spin states, which

result from the energy exchange between the spin ensemble and the degrees of freedom

in the surrounding environment (e.g. rotational and translational motion of the spins)

return the longitudinal magnetisation to equilibrium [135]. This relaxation is

characterised by T1, commonly known as the longitudinal relaxation time or the spin-

lattice relaxation time. Chemical exchange is a relatively low frequency process and

therefore does not affect T1 relaxation.

Transverse spin relaxation or T2-relaxation describes the return of the transverse

component of the magnetisation, Mx’y’, back to zero (the transverse component of M is

Time (s)

Time (s)

Time (s)

Rel

ativ

e am

plit

ud

e

Rel

ativ

e am

plit

ud

e

Rel

ativ

e am

plit

ud

e

Rel

ativ

e am

plit

ud

e


23

zero at equilibrium). Following the RF pulse, the spins in an ensemble loose coherence

as a result of dephasing of the spin magnetisation vectors, which causes eventual loss

of transverse magnetisation. This relaxation is characterised by T2, commonly known

as the transverse relaxation time or the spin-spin relaxation time. Both dipolar

interactions and chemical exchange contribute to T2 relaxation. The T2-relaxation

analysis and measurement of T2 is of particular interest in this thesis and these will be

discussed in further detail later in this chapter.

Although T1-relxation and T2-relaxation are the most common types of spin

relaxations, there are also other relaxation mechanisms that can be described by other

MRI parameters [136, 137]. However, those are outside the scopes of the studies

presented in this thesis.

2.1.4 Signal Localization

The NMR signal discussed above is generated from an ensemble of 1H spins

irrespective of their spatial positions. However, for MRI, spatial localization of NMR

signal is necessary to differentiate measured signals from different parts of an imaging

object and to generate a 2D/3D representation of the sample under examination. In

order to localise a voxel of the imaging object, which appears as a pixel in a 2D MR

image, three independent orthogonal localization systems are used, which are

commonly known as the gradient coils. The gradient coil system produces time varying

magnetic fields of controlled spatial non-uniformity for signal localization.

2.1.4.1 Slice-Selective Gradient

The imaging plane of a 2D MRI image (obtained from a 3D object) can be

defined by a slice selective gradient. Figure 4 shows a slice selective pulse sequence

that selects an imaging slice orthogonal to z axis by selective excitation method. The

effect of a slice selective gradient on the resonant frequency can be explained by the

following equations.

𝐵 = 𝐵0 + 𝑧𝐺𝑧 = 𝐵 + ∆𝐵 (14)

𝜔 = 𝛾(𝐵0 + ∆𝐵) = 𝜔0 + 𝛾𝑧𝐺𝑧 (15)

If the static magnetic field B0 is homogeneous for the entire imaging object,

then the entire volume of the imaging object should oscillate at the Larmor frequency


24

𝜔0. Here, the application of the gradient magnetic field Gz has made the resonant

frequency 𝜔 position dependent. The 90° RF pulse applied with the Gz pulse can select

a slice of thickness Δz. Slices at different depths can be chosen by the manipulation of

the Gz pulse. The volumetric MR image of a 3D object can be created by assembling

such individual planar images. The bandwidths of the transmitting coils determine the

minimum slice width that can be selected around a focused plane.

Figure 4. A pulse sequence for achieving voxel-specific frequency specificity in a 3D imaging

object. A linear gradient magnetic field Gz is applied along the z-axis for slice selection; a linear

gradient magnetic field Gx is applied along the x-axis for frequency encoding, a linear gradient

magnetic field Gy is applied for time Tpe along the y-axis.

2.1.4.2 Frequency Encoding

A frequency encoding pulse is shown in Fig. 4 where the MR signal is frequency

encoded along the x axis by a gradient Gx. Frequency encoding makes the resonance

frequency linearly dependent on its spatial location. When the linear gradient vector Gx

is applied along the x axis, the imaging object experiences the homogeneous B0 field

added with the linear gradient field Gxx. Therefore, the Larmor frequency at position x

will be

𝜔 = 𝜔0 + 𝛾𝑥𝐺𝑥 (16)

Ignoring the influence of transverse relaxation, the FID decay can then be expressed by

𝑑𝑆(𝑥, 𝑡) = 𝜌(𝑥)𝑑𝑥𝑒𝑖𝛾(𝐵0+𝑥𝐺𝑥)𝑡 , (17)

where ρ(x) is the spin distribution in the imaging object [127]. Frequency encoding in

a 2D imaging plane defines a family of isofrequency lines, which are all perpendicular


25

to the frequency gradient vector. One dimensional spatial localization of signal is

achieved by this.

2.1.4.3 Phase Encoding

Phase encoding makes the phase of NMR signal linearly dependent on its spatial

origin. A phase encoding pulse sequence is shown in Fig. 4, which is built on the

previous slice selection and frequency encoding pulse. Here, NMR signal is phase

encoded along the y-axis by the gradient vector Gy during the free precession period.

Immediately after the RF pulse, a linear gradient field is applied for a short interval Tpe

when the local signal is frequency encoded (0 ≤ t ≤ Tpe). This results in different phase

angle accumulation for signals at different y positions after the time Tpe. The signal

measured after this time bears an initial phase angle

𝜑(𝑦) = −𝛾𝑦𝐺𝑦𝑇𝑝𝑒 (18)

Ignoring the influence of transverse relaxation, the signal under the influence of

this gradient can be expressed by

𝑑𝑆(𝑦, 𝑡) = {𝜌(𝑦)𝑒−𝑖𝛾𝑦𝐺𝑦𝑇𝑝𝑒𝑒−𝑖𝛾𝐵0𝑡, 𝑇𝑝𝑒 ≤ 𝑡

𝜌(𝑦)𝑒−𝑖𝛾𝑦𝐺𝑦(𝐵0+𝑦𝐺𝑦)𝑡, 0 ≤ 𝑡 ≤ 𝑇𝑝𝑒.

(19)

Here, ρ(y) is the spin distribution in the imaging object. Phase encoding along

any arbitrary direction can be achieved by simultaneous use of Gx, Gy and Gz during the

phase encoding period. The phase angle can be adjusted by controlling the strength of

phase encoding gradient or the time interval for phase encoding.

2.1.5 K-space Acquisition and Image Reconstruction

The k-space is an extension of the concept of Fourier space that temporarily

holds the raw MR signal. As the bulk magnetisation relaxes back to equilibrium, the

MR signal induces current in receive coils. This MR signal is transformed by Fourier

Transform to the frequency-domain MR signal, which is then saved at the 2D or 3D k-

space [128, 138]. The mapping of the k-space is directly related to the spatial encoding

gradients applied to the imaging object. A uniform coverage of the k-space is essential

to obtain an image of good quality that preserves the necessary information. By

adjusting the spatial encoding gradients, k-space data are collected in trajectories until

the k-space is full or complete. For visual interpretation and analysis, the k-space data


26

is converted into a time domain image; this process is commonly known as image

reconstruction. The image reconstruction techniques vary considerably for different

spatial encoding methods used for imaging. The mathematical algorithm Fourier

transform is generally used that defines the relationship between a continuous signal in

the time domain and its representation in the frequency domain. In practice, only a finite

number of k-space points are sampled at regular intervals. Inverse fast Fourier transform

(iFFT) [128] is applied for k-spaces to reconstruct the digital MR images.

2.1.6 Transverse Relaxation Analysis

In MRI, an image is a multidimensional function of spin density, diffusion,

relaxation times, and other factors. Eqn. 12 and 13 describe the simplified decay

behaviour of transverse and longitudinal magnetization of free induction decay that

relies on the values of T2 and T1. Magnetic field inhomogeneity experienced by spin

system contributes to another weighting factor known by T2*. As the magnetization

decays are generated from an ensemble of protons, proton density or PD works as a

weighting factor of the decays. The image sampling time (t) is variable and the preferred

contrast can be achieved by choosing an appropriate value for t.

2.1.6.1 Imaging Sequence for Transverse Relaxation based MRI

The FID of a spin system holds the true T2 characteristics of a spin system only

if the magnetic field experienced by the spins in the imaging sample is perfectly

homogeneous. However, when an object is placed is placed in the B0 field, the main

magnetic field gets distorted. The variations in magnetic field strength results in a

distribution of precessional frequencies for magnetised protons Due to the

inhomogeneity in B0 and the presence of chemical shift, a spin system is likely to have

a distribution of isochromats. There, the spins quickly go out of phase owing to the

differences in precessing speeds as shown in Fig. 5. This process leads to faster decay

of bulk magnetization and the FID no longer carries the T2 weighting. In a situation like

this, the overall decay of the NMR signal can be characterised by the apparent

transverse relaxation time T2*,

1

𝑇2∗=

1

𝑇2+

1

𝑇2′ (20)


27

where 1/T2’ represents the shimming linewidth. It is obvious that T2* < T2 and therefore

the signal decays faster than predicted by T2 in an ideal situation. However, the effects

of T2’ on the decay of the NMR signal is reversible if the position of the spins remain

constant during the imaging experiment.

Figure 5. The de-phasing of isochromats that have different precessing speeds, after the initial

90° RF pulse.

The Curr-Purcell-Meiboom-Gill sequence [41], commonly known by CPMG,

has been established as a reliable technique to measure T2 weighted decays. CPMG is

a modified and developed version of Carr-Purcell sequence [139] originally based on

the spin-echo phenomenon [140, 141] discovered by Hahn. It uses the phase

incoherence of the magnetised isochromats to its advantage in order to generate echoes

of the original FID by applying multiple RF pulses. In a CPMG sequence, an initial 90°

RF pulse is followed by multiple 180° RF pulses at regular intervals:

90°𝑦′ − 𝜏 − (180°𝑥′ − 2𝜏)𝑁. (21)

Parts of CPMG imaging sequence is shown in Fig. 6. The initial 90°y’ pulse

rotates the bulk magnetic moment vector along the x’ axis. Due to the spin-spin

interactions, the FID signal dephase and decays exponentially with T2* envelope before

the 180° pulse is applied (0 < t < τ). The 180°x’ pulse applied after the first τ delay flips

the equatorial plane completely around the x’ axis. The magnetic moments continue to

dephase during the second τ delay. However, because of the 180° added phase, the

dephasing results in the re-phasing of the magnetic moments and the signal reaches a

maximum value at 2τ generating the first echo. The time period from the initial 90° RF

pulse to the formation of the first spin echo is called echo time (TE = 2τ). For the

following echoes, 180°x’ pulses are applied at the odd multiples of τ ((2n + 1) τ) and the

decay is measured at even multiples of τ (2nτ, n = 1, 2, 3, ...), to sample echo times and

echo magnitudes. These echo magnitudes decay exponentially following the T2


28

envelope as shown in Fig. 7. Such pulse sequences are often repeated after repetition

time (TR) to acquire images from different slices and also to average the signal in order

to obtain images with improved signal to noise ratio (SNR).

Figure 6. Basic CPMG pulse sequence for pure T2 imaging. One k-space line is acquired for

each phase encoding gradient, Gy. Due to the multiple 180° pulses and read gradients, k-space

lines are acquired in alternating directions. The dotted lines on the right side of the figure

indicates that the sequence continues for a predetermined number of echoes within a single TR.

The Multi-Slice-Multi-Echo (MSME) sequence is built upon the original

CPMG sequence for multi slice imaging by incorporating multiple slice selection

gradients that are applied at the same time as the refocusing pulses. Using MSME, the

transverse relaxation data is acquired for multiple slices within 1 TR and thereby

requires less time for completing multi-slice scanning in comparison to scanning by

CPMG.

2.1.6.2 T2 Mapping and Analysis

Using the CPMG sequence, T2 weighted signal decays are achieved by using a

single long TR (TR ~ 3xT1 to 5xT1), and multiple TE values within the TR. The TEs are

evenly spaced until the signal decay is complete or it reaches the noise floor. The use

of multiple TE allows the measurement of different degrees of T2 weightings. This train

of pulses can be repeated after every TR, which can be combined with appropriate

spatial encoding pulses to obtain a T2-weighted MR image. For a mono-exponential

decay, the measured T2 weighted signal can be expressed as follows.

𝑆 = 𝑆0 𝑒

−𝑡

𝑇2 + 𝑆𝑜𝑓𝑓𝑠𝑒𝑡 (22)

Here, S is the signal amplitude measured at time t (t = n x TE, n = 1, 2, 3 ...), S0

is the full signal intensity at time 0 and Soffset is the magnitude of noise. In magnitude


29

echoes, the noise is primarily governed by Rician distribution [142]. With measured S

and known t, the value for S0, T2 and Soffset are obtained by least-square iterative fitting

of the above equation to the measured decay. Spatial T2 map or transverse relaxation

rate map (R2 = 1/T2) can be obtained by repeating this procedure voxel-by-voxel for the

entire MR image.

Figure 7. Formation of spin echoes by a CPMG sequence. Initial 90° RF pulse produces a FID,

which quickly disappears as the spins de-phase. The first 180° RF pulse at time τ flips the

equatorial plane and refocuses the spins that produce an echo at time 2τ. Another 180° RF pulse

is applied at time 3τ that generates an echo at time 4τ.

However, in case of a bi-exponential decay, the following equation is fitted to

the measured data.

𝑆 = 𝐴1 𝑒

− 𝑡

𝑇21 + 𝐴2 𝑒

− 𝑡



30

A bi-exponential T2 decay with two different T2 components may result from

one imaging volume that contains two different 1H spin systems with different

transverse relaxation rates and relative weightings (A1, A2). In a more complicated

situation, when multiple T2 components are expected in a T2 weighted decay, the decay

can be analysed using one dimensional inverse Laplace transform, which can

decompose the multi-exponential decay into a sum of mono-exponential decays. The

T2 relaxation decay can then be expressed as

𝑆(𝑡𝑖) = 𝑔𝑖 = ∑ 𝐴(𝑇𝑗)𝑒−

𝑡𝑖𝑇𝑗

𝑗 + 𝜀𝑖, (24)

where, Tj are the relaxation times, A(Tj) are the relaxation time distribution of the

acquired signal and ɛi is the noise magnitude of the signal. A(Tj) can be determined by

inversing the T2 relaxation data using inverse Laplace transform or by non-negative

least square (NNLS) matrix-fitting algorithm by minimizing the error χ2 [143] as below.

𝑚𝑖𝑛 {𝜒2 = ∑ (𝑔𝑖𝑛𝑖=1 − ∑𝐴(𝑇𝑗)exp (−

𝑡𝑖𝑇𝑗

⁄ ))2} (25)

A regularization function weighted by a smoothing parameter δ is added in order

to achieve a robust fit in presence of noise [144-148]. The new function for minimizing

χ2 then takes the following form.

𝑚𝑖𝑛 {𝜒2 = ∑ (𝑔𝑖𝑛𝑖=1 − ∑𝐴(𝑇𝑗)exp (−

𝑡𝑖

𝑇𝑗))2 + δ−1 ∑ (2𝐴(𝑇𝑗) − 𝐴(𝑇𝑗−1) −𝑚

𝑗=1

𝐴(𝑇𝑗+1))2} (26)

The NNLS fitting is performed with a pre-defined list of Tj that contains

logarithmically spaced discrete T2 values. The resulting T2 distribution contains log-

normal-like curves with distinctive peaks where each peak correspond to a distinct

micro-environment for 1H that has a unique T2. Two reproducible measures, area

fraction (AF) and geometric mean T2 (gmT2), are commonly used to assess such

distributions [12, 17, 33, 34, 149-152]. AF and gmT2 are measured for T2 distribution

peaks within a specified T2 range: T2min to T2max.

𝐴𝐹 =∑ 𝐴(𝑇2j)

𝑇2𝑚𝑎𝑥𝑇2𝑚𝑖𝑛

∑𝐴(𝑇2j)⁄ (27)

𝑔𝑚𝑇2 = 𝑒𝑥𝑝 (∑ 𝐴(𝑇2j) 𝑙𝑜𝑔


𝑇2

∑ 𝐴(𝑇2j)𝑇2𝑚𝑎𝑥𝑇2𝑚𝑖𝑛

⁄ ) (28)


31

AF measures the relative contribution of the measured T2 peak with respect to

the T2 distribution and gmT2 measures the average T2 of the particular T2 peak on a log

scale. The choice of T2min and T2max are subjective and the choice depends on the nature

of the particular T2 distribution under investigation. The AF and gmT2 of T2 distribution

peaks are used to study the underlying water micro-environment and tissue

characteristics responsible for different T2 [12, 33, 122].

2.1.7 Specialised Scanner for MRI and NMR

Clinical MRI scanners use the principles of magnetic resonance discussed above

for diagnostic purposes. The structural components and associated parameters of

clinical MRI scanners are particularly designed to suit the size of human body.

However, research studies often have different requirements than clinical standards,

such as, high resolution, improved SNR, and low-cost involvement. Accordingly, with

recent advancements in MRI technology and software capabilities for image analysis,

some specialised MRI tools have been developed to suit the requirements of certain

purposes. µMRI system and portable NMR are two such specialised MR instruments

that have been used in the studies presented in this thesis.

2.1.7.1 Micro-MRI for High Resolution MRI

In MRI, the signal localization gradients define small regions of the sample

containing nuclei precessing at a frequency and a phase exclusive to that region. Such

a region is called a voxel. The resolution of an image thus formed is expressed in terms

of the size of each voxel [153]. The static magnetic field strength for clinical scanners

are commonly in the 1.5 T – 3 T range. µMRI systems have higher static magnetic

fields (usually greater than 4T) compared to conventional MR systems. Higher static

magnetic field B0 generates greater energy difference between parallel and anti‐parallel

orientation of nuclei. For a given spin density, this increased energy difference between

the eigenstates increases the strength of the net magnetization. This results in a larger

output signal from the sample and hence greater SNR. Additionally, the imaging

gradients of the µMRI scanners are approximately an order of magnitude stronger than

that in the clinical scanners. Consequently, magnetic resonance micro‐imaging or

µMRI allows the acquisition of MR images at higher resolution than the clinical

scanners. It is worth noting that, the relaxation parameters, T1 and T2, are also dependent


32

on the field strength used for MRI: T2 decreases and T1 increases with the increase of

the field strength. Therefore, the T1 and T2 measurements are not absolute and they can

only be defined for a specific imaging object/tissue at a particular field strength.

The hardware of µMRI is designed to trade high spatial resolution for large

field-of-view. RF coils of µMRI systems are significantly smaller than the RF coils in

clinical MRI, which often restricts the imaging of large samples. Usually, µMRI is done

in the context of research for morphological as well as physiological imaging of small

animals or ex-vivo samples. For example, researches have reported µMRI studies that

used 4 – 16.4 T magnet and achieved 17 – 156 µm voxel sizes, 0.5 – 2 mm slice

thicknesses [2, 110, 154-158] and maximum FOV of 2.6 x 3 cm [155]. MRI with such

high resolution can provide insight into the detailed anatomical structures and

functional properties of the imaging sample. The high SNR achieved in such system

also facilitates quantitative imaging of small samples that may reveal vital information

about the local water micro-environment (of each imaging voxel). In research,

quantitative µMRI is often used to investigate the structural organization of

macromolecules [8, 64, 159] and to study the changes in tissues induced by particular

diseases [12, 13, 110, 111].

2.1.7.2 Single-Sided Portable NMR Scanner

Single-sided and mobile NMR is a powerful tool commonly used for well-

logging [160-162] and process and quality control [163-166]. It is also used in various

other research fields for characterizing arbitrary large samples [89, 91]. The NMR

instrument uses unilateral magnet arrays and surface RF coils to excite and detect NMR

signals from a sample external to the sensor. This approach allows the interrogation of

the near surface structures of large samples [93]. Because MR signals are generated in

the stray field of open magnets in the mobile NMR units, inhomogeneity is common in

B0 and B1 fields. However, researchers have developed methods that has now enabled

single-sided NMR to measure relaxation times [140, 167], material density [168, 169],

self-diffusion coefficients [170], relaxation-diffusion [171, 172], and diffusion-

diffusion correlation functions [160]. In addition, in comparison to a clinical MRI

scanner, the installation and maintenance cost for a portable single sided NMR

instrument is significantly lower. This has encouraged the use of mobile NMR units in

medical applications [93]. For example, a portable NMR system (commercially known


33

as NMR-MOUSE [89, 90] (Magritek, Wellington, New Zealand)) has been used to

investigate silicone breast implants [92] and various biological tissues, including

tendon [94], articular cartilage [95], skin [93] and trabecular bone [96]. Additionally, a

recent investigation has shown that T1 relaxation time constants measured using

portable NMR can distinguish between regions with high mammographic density from

regions with low mammographic density in human breast tissue [97]. NMR-MOUSE

is designed as a low-cost and low-maintenance mobile unit [91] based on permanent

magnets. The second study presented in this thesis uses a commercially available NMR-

MOUSE unit for investigating the tissue composition of human breast tissue.

2.2 Knee Joint

The knee joint is a hinge type synovial joint that permits flexion and extension

as well as a slight medial and lateral rotation. It consists of multiple bones, cartilage

layers and an extensive network of ligaments, tendons, and muscles. Figure 8 presents

the cartoon sketch of a human knee joint that shows the major structural components.

The femur is commonly known as the thigh bone that connects the hip joint with the

knee joint. The two femurs of the legs converge medially towards the knees, where they

articulate with the proximal ends of the tibia. The tibia is the long bone that connects

the knee with the ankle bones. The knee joints support the entire body weight during

movement. This section describes the basic anatomy and functions of the tissues of

knee joint. This section also provides a brief overview of knee joint Osteoarthritis (OA)

and its effects on the tissues of the knee joint. The MRI-based diagnosis methods

available for knee joint OA are also discussed.

2.2.1 Articular Cartilage

Articular cartilage (AC) is a thin layer of connective tissue that covers the

articulating surfaces of the synovial joints. In a human knee joint, healthy adult cartilage

is 2-4 mm thick [173], which is found on the femoral condyles and on the tibial plateau.

The primary functional roles of AC include: creating a low friction protective barrier

for the underlying bones, transmitting loads to the underlying bones [49] and

distributing pressure exerted on the joint over a wider area to reduce stresses sustained

by the contacting bone surfaces [48]. In mammals, the AC is made of a specialised

tissue known as hyaline cartilage. Hyaline cartilage is an avascular tissue with a


34

complex three-dimensional architecture where chondrocytes are embedded in an

extracellular matrix (ECM) principally composed of collagen (~15%-20%) [42],

proteoglycan (~3%-10%), lipid (~1%-5%) and water [7, 31, 43].

Figure 8. A cartoon sketch of the human knee joint. Articular cartilage is represented by the

shaded regions lining the femoral condyle and tibial plateau. Courtesy of Dr. Sirisha Tadimalla,

Queensland University of Technology.

Proteoglycan (PG) are complex macromolecules with a protein core that are

covalently bound to glycosaminoglycan (GAG) chains. The GAG chains have

repeating carboxyl (COOH) and sulphate (SO4) groups that remain highly anionic in

ECM. Due to these negative charges, PG macromolecules are highly hydrophilic, which

contributes to the high osmotic pressure within the AC. This osmotic pressure is

counteracted by the collagen macromolecule network within the AC. Type II is the

primary (90 – 95%) collagen [49, 174, 175] in hyaline cartilage, which is assembled

into fibrils that are arranged into fibres with diameters ranging from 20 nm to 150 nm

in diameter depending on age and species [69, 176]. The collagen fibres form a cross-

linked network that works as the structural scaffold of ECM and contributes to the shear

and tensile properties of AC [46, 52]. The overall biomechanical properties of AC

depends on both the collagen network structure and hydrostatic interplay between the

negatively charged PGs and water.


35

The ECM of AC is both structurally heterogeneous and mechanically

anisotropic across the depth of the tissue [64]. The anisotropy is associated with the

alignment and organization of collagen fibres, which vary with depth, as does the

collagen volume fraction. These variations typically create three histological zones in

cartilage ECM: the superficial, transitional and the radial zones [46, 47]. As shown in

Fig. 9, superficial zone is the zone closest to the articular surface (AS). This zone makes

up ~ 3% – 12% of the AC thickness [29] while it contains both the highest water (~

75% wet weight or w/w) and collagen (~ 20% w/w) content, and the lowest PG fraction

of all three zones (~ 4% w/w) [29, 48]. Smaller fibres are predominantly aligned

parallel to the AS. This organization facilitates stress distribution [177, 178] and

enables fast tissue response at high loading rates [179].

Figure 9. Schematic diagram of the collagen fibre arrangement in articular cartilage. From

articular surface to bone: the superficial zone containing collagen fibres aligned parallel to the

surface, the transitional zone containing fibres with no particular alignment, the radial zone

containing collagen fibres aligned perpendicular to the articular surface, followed by a layer of

calcified cartilage. Courtesy of Dr Monique Tourell, Queensland University of Technology.

The radial zone is closest to the bone and is the thickest zone of the three

cartilage zones (> 50% of the tissue thickness [64]). Here, the collagen content is ~ 65%

w/w [64] with large fibre bundles formed by woven collagen fibres that are oriented

primarily perpendicular to the articular surface [48]. The concentration of PG increases

with the depth of cartilage, the highest PG concentration is observed in the radial zone

[29, 30, 46]. The PG-rich radial zone facilitates the response of cartilage to compressive

loading by reducing excessive deformation [50, 179, 180]. The transitional zone lies


36

between the superficial and the radial zone and has ~ 15% w/w collagen [64]. In this

zone, the alignment of collagen fibres undergo a transition from their alignment in the

radial zone, to that in the superficial zone. Therefore, the collagen fibres in the

transitional zone appear to have no preferred direction of alignment [56, 181]. The

cartilage is separated from the subchondral bone by a calcified zone where the

calcification progressively increases closer to the bone.

2.2.2 Assessment of Collagen Fibre Architecture in AC by MRI

The three-zone structure governs the response of cartilage to dynamic loading

during movement [50, 51]. However, the thickness of each zone, as well as the

composition and organization of the major molecular components, have been observed

to vary across species and even across different sites in the same joint [69-72].

Consequently, a comprehensive understanding of the collagen architecture in AC is

essential in order to study the joint functionalities. This understanding is also necessary

to diagnose and assess the effects of disease, such as Osteoarthritis (OA), on AC.

Previously, several experimental techniques have been used to investigate the

organization of collagen fibres in AC, for example, scanning electron microscopy

(SEM) [53-56], X-ray scattering [182-185], and polarised light microscopy (PLM) [53,

54, 57]. Although these methods provide high resolution (< 1 µm) information on the

collagen fibre architecture, all of these techniques are destructive and hence cannot be

used for evaluation of sample-specific responses to external stimulants or for in vivo

applications. Confocal microscopy allows non-invasive evaluation of tissue

microstructure [186, 187]. However, the determination of collagen fibre orientation in

AC using fibre optic confocal microscopy is an invasive technique since it requires

administration of a specific dye for collagen [187]. On the other hand, T2-based MRI

probes collagen fibre alignment in AC using a specialised technique called the “magic-

angle effect” [8, 29, 43, 64], which is both non-destructive and non-invasive.

The transverse relaxation time or the T2 of nuclear spins is sensitive to the local

molecular environment of the 1H population inside the imaging sample [188, 189]. The

value of T2 is sensitive to the molecular hydrodynamics of water, which in turn depends

on the viscosity, water content, biomolecular composition and microscopic

organisation of the tissue [8]. Consequently in AC, the anisotropic distribution of

collagen results in T2 anisotropy – this was unambiguously demonstrated and explained


37

through the works of Xia et al.[1, 29]. In research, the transverse relaxation rate R2 (R2

= 1/T2) is commonly used for transverse relaxation mapping. The R2 in cartilage is the

summation of its isotropic and anisotropic contributions, 𝑅2𝐼 and 𝑅2

𝐴 [1, 94]:

𝑅2 = 𝑅2𝐼 + 𝑅2

𝐴 = 𝑅2𝐼 + 𝑅2

𝐴0 ( 3 𝑐𝑜𝑠2𝜃𝐹 − 1

2)2

. (29)

Here, the angle θF is the predominant angle of collagen fibre alignment relative

to static magnetic field B0. This orientational dependence of R2 is known as the magic

angle effect because the anisotropic term (𝑅2𝐴) vanishes when θF equals θMA =

𝑎𝑟𝑐𝑐𝑜𝑠(1 √3⁄ ) ≈ 54.7°, the magic angle. This behaviour of 𝑅2 can be explained by

the residual intramolecular dipolar couplings (described in section 2.1.3.1) in water

molecules [64, 190, 191]. The extracellular water in AC is present in two chemically

exchanging pools: free water (FW) and water bound to ECM macromolecule (BW).

The relaxation rate of FW is represented by 𝑅2𝐼 . The water molecules near collagen

fibres are loosely bound to the fibres and therefore are no longer as mobile as they are

in viscous solution. The restricted motion of these water molecules result in 𝑅2𝐴 that

qualitatively follows the (3cos2θF - 1)2 curve at all depths from AS [1]. The measured

relaxation rate R2 is the population-weighted sum of the relaxation rates in the FW and

BW pools. R2 reaches near minimum value when 𝜃F = 55° and attains its maximum

value when 𝜃F = 0°. However, the two-pool approximation of the distribution of water

in AC is a simplified approximation of the real situation. In reality, there exists several

bound pools within the exchange equilibrium that consist water molecules bound to

collagen macromolecules present in several layers of hydration as well as water

molecules bound to proteoglycan aggregates.

The magic angle effect in AC is illustrated in Fig. 10 where R2 maps of a bone-

cartilage plug is shown at two different orientations. At the 0° orientation (B0

perpendicular to AS), the R2 relaxation rate map of AC show a distinct laminar

appearance where two dark bands mark the cartilage regions with low collagen

anisotropy and a bright band marks the region with high anisotropy that likely represent

the radial zone with highly aligned collagen molecules [8]. This laminar appearance

disappears when the AS makes an angle of 55°with B0. The R2 profiles computed from

these R2 maps show that, based on the depth-wise variations in the orientational

dependence of R2, the AC can be divided into three zones: depths 0 mm – 0.2 mm from


38

AS represent the superficial zone, depths 0.2 mm – 0.4 mm from AS represent the

transitional zone, and depths 0.5 mm – 1.3 mm from AS represent the radial zone. The

value of R2 is independent of the orientation of the cartilage plug in the transitional zone

of AC where the collagen network lacks a preferred direction of fibre alignment.

Conversely, the value of R2 differs significantly between the 0° and 55° orientation in

both superficial and radial zones, which indicates the presence of a predominant

alignment in the collagen network.

Figure 10. R2 anisotropy in AC: (A) R2 map of a bovine bone-cartilage plug oriented

perpendicular to the applied magnetic field B0; (B) R2 map of the same sample oriented nearly

at the magic angle (55°) relative to B0. In both maps, white corresponds to R2 = 0.13 ms−1. (C)

R2 depth profiles constructed from maps A and B. Reprinted from [8], with permission from

IOS Press.

The degree of R2 anisotropy can be used as a qualitative indicator of the degree

of collagen alignment in AC and thus indirectly identify the cartilage zones with

preferred collagen alignments. Theoretically, the 3D alignment of the collagen fibres in

different zones of the cartilage plug can be examined by rotating/tilting a cartilage plug

about different axes and repeating the same R2 measurement protocol. In the same way,

the collage alignment in the AC of whole knee can be interpreted by R2 mapping of the


39

whole knee and subsequent analysis. Diffusion Tensor Imaging by MRI is also sensitive

to the collagen alignment in AC. However, the focus of this thesis is the transverse

relaxation or T2/R2 based MRI techniques. An experimental study on the magic angle

effect of T2 anisotropy in AC is presented later in this thesis.

2.2.3 Osteoarthritis in Knee Joint

Osteoarthritis refers to a degenerative condition that commonly affects the large

weight bearing joints, such as the knees and the hips. OA also affects ankles, hands and

spines. In 2014-15, approximately 2.1 million Australians have been diagnosed with

OA with a prevalence rate of 9% [192]. Although OA is not yet curable, progression of

OA can be prevented or decelerated by early diagnosis and effective treatment planning

within the treatment window. According to literature, knee joint OA may induce

changes in AC, subchondral bones, bone marrow, menisci, ligaments, and synovial

tissues. The conventional clinical practice for OA diagnosis combines physical

examination and X-ray radiography. However, the considerable discordance between

radiological and clinical OA findings [193] highlights major limitations in this

approach. In knee joint OA, AC is usually the first tissue that gets affected. AC is

translucent to X-ray and therefore the early changes of OA are undetectable by

radiological investigation. The radiological diagnosis of OA is primarily based on the

misalignment of bones in the affected joint, which is observed at the advanced stages

of OA after the AC is partly or completely degraded.

Recently, MRI has become an invaluable tool to diagnose as well as to study

the pathophysiology of OA and has the potential to be the imaging modality of choice

in therapeutic trials. The imaging superiority of MRI lies in its inherent ability to

analyse multiple tissue structures concurrently in great detail and in a three-dimensional

(3D) perspective. MRI is completely non-invasive and it can focus on different tissues

of the knee joint by manipulating image contrast. The fundamentals of OA diagnosis

by MRI depend on the selection of sensitive sequences, using appropriate evaluation

methods, and thorough knowledge of the characteristic imaging manifestations. The

commonly used MRI sequences in the OA research and diagnostics include fluid

sensitive fat suppressed (FS) sequences, i.e. T2 weighted (long repetition time (TR) and

long echo time (TE), e.g. TR/TE = 3500/120 ms), PD weighted (long TR and short TE,

e.g. TR/TE = 3500/20 ms) or intermediate-weighted (Iw) (PD weighted with a TE of


40

about 40 ms, e.g. TR/TE = 3500/40) fast spin echo (FSE) or gradient echo (GRE)

sequences, or a short tau inversion recovery (STIR) sequence [98-103]. Specifically,

these MRI contrasts/sequences have been shown to be effective in identifying AC

degradation, bone marrow lesion (BML), synovitis, and injury to ligaments and

menisci. The following sections present a brief overview of the effects of OA on knee

joint tissues and the MRI techniques used to diagnose OA. Later in this thesis, an

experimental MRI study particularly designed for early diagnosis and comprehensive

monitoring of the developmental pathway of knee joint OA will be presented.

2.2.3.1 Articular Cartilage – Effects of OA and Diagnosis by MRI

The degeneration of AC is often regarded as the structural hallmark of OA

progression. The earliest events associated with OA are stress‐induced production of

cytokines, chemokines, other inflammatory mediators, and cartilage‐degrading

proteinases by chondrocytes. These change the composition of the cartilage ECM,

specifically the loss of proteoglycans and disturbance of collagen orientation, without

major change in collagen content. This loss in homeostasis results in a cascade of

events, which eventually lead to cartilage damage [46]. The collagen content are also

reduced in advanced OA [105], which disrupts the pre-existing collagen network and

results in matrix degradation. MRI provides superior soft tissue contrast in comparison

to other diagnostic imaging modalities and therefore it is more effective in diagnosing

the early degradation in AC. T2* weighted SE/FSE sequence and T1 weighted GRE

sequences are commonly used to measure cartilage thickness in presence of OA [108-

111].

In recent years, MR technology and image processing have evolved to provide

insights into the OA physiology. Quantitative MRI exploits the macromolecular

changes of AC that take place in the presence of OA to provide a quantitative

understanding of the breakdown process of AC. T2 measurements of cartilage are

sensitive to the water content in AC and the integrity of the proteoglycan–collagen

matrix. It thereby provides a useful non‐invasive marker of the hydration, composition

and overall structure of AC. As described earlier in this chapter, the MRI signals

obtained from voxel-specific pure T2 weighted decays can be mathematically processed

to compute the spatial distribution of T2 relaxation times throughout articular cartilage.

It has been observed that the cartilage areas of increased or decreased water content


41

result in higher or lower T2 than usual, which generally correlate with areas with

damaged cartilage [106]. In whole knee joint imaging, a multi-echo spin-echo (MSME)

sequence is commonly used to shorten scan time and mono-exponential or multi-

exponential relaxation models are fitted to the measured decays for T2 mapping [5]. Ex

vivo studies on T2 relaxation in articular cartilage have revealed sensitivity of T2

imaging to changes in collagen content and distribution [107]. Although it is clear that

T2 relaxation in cartilage is strongly dependent on collagen content and distribution, its

relationship with PG content is yet to be determined [194].

Diffusion weighted imaging (DWI) measures the diffusion pattern of water

protons providing insights into structure and organization of AC at the micro-level. In

a healthy cartilage, the motion of water is restricted due to its inherent architecture,

times required for diffusion are long and the apparent diffusion coefficient (ADC) is

low, whereas in a disordered matrix, water protons move more freely reducing diffusion

time and thereby increasing the ADC [195, 196]. Diffusion anisotropy measured by

DTI, which is a subclass of DWI, has been particularly useful in determining cartilage

microstructure and degradation [197]. On the contrary, T1ρ mapping is sensitive to the

macromolecular content of tissues and therefore is very effective in detecting early

changes in OA [4, 198]. In this method, the movement of water protons are restricted

by “spin-locking” and the interaction between motion-restricted water molecules and

their extracellular environment is measured by T1ρ [199]. When PG depletion occurs in

the earliest phases of OA, the physio-chemical interactions in the macromolecular

environment are disrupted that cause an uneven distribution of T1ρ. When compared

with normal cartilage, osteoarthritic knee cartilage show elevated T1ρ values [200-202].

Degenerated cartilage can also be detected using sodium (23Na) MRI that relies on the

negative fixed charge density within the ECM. In normal AC, high concentrations of

positively charged 23Na are associated with the negatively charged GAG side chains,

which hold abundant negatively charged carboxyl and sulphate groups. With PG

depletion, GAGs are damaged and sodium signals are declined [203-205]. Therefore,

the strength of sodium signal differentiates degenerated cartilage from normal AC.

Based on the same principles, using the contrast agent Gadolinium, delayed

Gadolinium-Enhanced MRI (dGEMRIC) uses the fixed charge density ions in the

extracellular fluid to reflect the quantity of PG content in cartilage [206].


42

2.2.3.2 Subchondral Bone – Anatomy, Effects of OA, and Diagnosis by MRI

Subchondral bone is the layer of bone immediately below the AC that acts as

an anchor for the AC. Trabecular bone is light weighted and porous with a spongy

appearance that encloses numerous large spaces often filled with marrow. Bone marrow

is the soft tissue found in the spaces between trabeculae of the trabecular bone.

Subchondral bone, trabecular bone and bone marrow are found in the two large bones

of knee joint: tibia and femur. In adults, marrow in these bones produces red blood cells

in a process known as hematopoesis, and uses the bone marrow vasculature as a conduit

to the body's systemic circulation.

In the presence of OA in the knee joint, subchondral bones of both tibia and

femur experience substantial modifications [207-209]. In particular, the presence of

bone marrow lesion (BML) and bone marrow edema (BME) has been confirmed in

tibia and femur [112, 113]. BME occurs when excess fluid in the bone marrow builds

up and causes swelling. This condition is often caused by a protective reaction of the

body in response to an injury or inflammation, such as the injuries in OA. BMLs are

degenerative lesions consisting of edema, bone marrow necrosis, fibrosis, and

trabecular abnormalities [210]. They are often detected in conjunction with

neighbouring cartilage damage. A few studies have also demonstrated a correlation

between BMLs and progressive cartilage damage [114, 115].

The following MRI sequences are commonly used for the identification and

assessment of BML: T2 weighted (long repetition time (TR) and long echo time (TE),

e.g. TR/TE = 3500/120 ms), PD weighted (long TR and short TE, e.g. TR/TE = 3500/20)

or intermediate-weighted (e.g. TR/TE = 3500/40) fast spin echo (FSE) sequences and

short tau inversion recovery (STIR) sequence [98, 99, 112]. Several semi-quantitative

scoring systems, such as, Whole Organ Magnetic Resonance Imaging Score (WORMS)

[211], Knee Osteoarthritis Scoring System (KOSS), Boston Leeds Osteoarthritis Knee

Score (BLOKS) and Magnetic Resonance Imaging Osteoarthritis Knee Score

(MOAKS) [212] are available that allow cross-sectional and longitudinal evaluation of

BMLs. Quantitative measurements of BMLs using MRI provide a tool for evaluating

and monitoring lesions that are less observer-dependent than subjective evaluation. In

quantitative MRI, BMLs are manually or semi automatically segmented using a


43

greyscale threshold [213]. Compared to manual quantification, threshold-based

segmentation of BML has shown higher intra- and inter-observer reliability [212].

2.2.3.3 Ligament – Anatomy, Effects of OA, and Diagnosis by MRI

Ligament is the fibrous connective tissue that connects bones to other bones in

and around joints. It is composed mainly of long, stringy collagen molecules. Four

ligaments are present in the knee joint: anterior cruciate ligament (ACL), posterior

cruciate ligament (PCL), medial collateral ligament (MCL) and lateral collateral

ligament (LCL). These ligaments provide stability and strength to the knee joint by

limiting its movement. ACL stretches from the lateral condyle of the femur to the

anterior intercondylar area. It prevents the tibia from being pushed too far anterior

relative to the femur [214, 215].

Damage to ACL can often predispose a joint to early OA [216]. The risk for OA

seems to increase for ACL ruptures with combined ligamentous injuries [217]. Knee

joint alignment and laxity, which are primarily dependent on the structural and

functional roles of ligaments, have also been found to be related to the risk of OA

progression [218, 219]. PD and T2 weighted SE sequences are commonly used for

evaluating knee ligaments. These methods have shown sensitivity and specificity of

96% and 98% for detecting ACL damage [116]. In knee joints with OA, MR images

are also used to get insight into the pathophysiology of ligaments, particularly for ACL

and MCL [117, 118].

2.2.3.4 Menisci – Anatomy, Effects of OA, and Diagnosis by MRI

Menisci are the two articular disks of the knee-joint, medial meniscus and lateral

meniscus. They are made of connective tissue with extensive collagen fibers containing

cartilage-like cells. The menisci serve to protect the ends of the bones from rubbing on

each other and provide shock absorption and load transmission in both active and static

loading [220]. The common causes for meniscal tear or loss include injuries to the ACL

and the MCL and traumatic injures with brittle cartilage, which is common in athletes

and in older patients.

Injury to menisci often results in the development of OA [221]. Partial or total

meniscectomy disturbs the natural loading mechanism of a knee joint by increasing the


44

strain on articular cartilage. It causes dynamic deformation in knee joint areas which

may contribute to OA development [222]. Studies have demonstrated a positive

correlation of meniscal abnormalities with other structural joint abnormalities in OA,

such as subchondral BML and cartilage loss [223, 224]. Commonly used MRI

sequences for menisci imaging include PD-weighted SE with or without fat saturation

and T1 weighted and GRE sequences [225]. Emphasis is given to maintain short TE in

order to reduce TR and scan time, improve SNR and to decrease susceptibility artefacts

[225]. Addition of fat saturation to PD sequences is common [225], 93% sensitivity and

97% specificity has been reported for diagnosing meniscal tears using a fat-saturated

conventional PD-weighted SE sequence [119]. Although FSE sequence has also been

used to evaluate meniscal tears, it trades time for sensitivity and therefore is not

recommended [119].

2.2.3.5 Synovial Tissue – Anatomy, Effects of OA and Diagnosis by MRI

The thin, loose and vascular connective tissue that makes up the membranes

surrounding knee joint is the synovial tissue. It consists of synovial cells, which secrete

a viscous liquid called synovial fluid or synovium; this liquid contains protein and

hyaluronic acid that serves as a lubricant and nutrient for the joint cartilage surfaces.

The inflammation of the synovial membrane, known as synovitis, plays a major role in

the advanced OA [226]. The degree of synovitis closely correlates with joint swelling,

inflammatory pain, functional impairment and disability in patients with end stage knee

OA [227]. Although the precise inflammatory mechanisms of synovitis remain to be

elucidated, the synovium phagocytes degrade cartilage and bone [228] and it appears

that synovitis may be a secondary phenomenon in OA.

MRI has been shown to be an effective method for the assessment of synovial

tissues due to its superior contrast in comparison to radiography [229]. However,

relatively few publications are available on MRI measurements of synovium in knee

OA. Extensive studies on the quantification of synovitis in rheumatoid arthritis [230]

suggests that MRI can be useful in the diagnosis of knee synovitis because of their

anatomical similarities. Literature suggests that synovitis in OA can be identified by

T1 weighted sequence, fat suppressed PD or T2 weighted SE sequence [112, 120] and

also with the addition of intravenous administration of gadolinium contrast agent [226].


45

2.3 Mammographic Density

Mammographic density (MD), also known as breast density, is a well-

established predictor for breast cancer (BC) risk [231, 232] along with increasing age,

genetic mutation, family history or personal history of BC. Research has shown that the

risk for BC predicted by MD is independent of other risk factors [232]. Breast cancer

(BC) is the most commonly diagnosed cancer among females all over the world. It is

estimated that, in 2018, 18,235 females and males will be diagnosed with BC in

Australia with that may result in 3,157 deaths in the year [88]. Although the prevalence

of BC has been increasing every year for decades, the survival rate has also been

improving in the population living with BC. Between 1984-1988 and 2009-2013, 5-

year survival from BC improved from 72% to 90% [88]. This remarkable improvement

in BC survival owes to two main factors: early diagnosis and effective treatment

planning. In current clinical practise, a specialised medical imaging technique called

mammography is used for screening BC. Mammography uses low-dose X-ray to

acquire a mammogram that aids in the detection and diagnosis of BC. Radiographically,

the breast mainly consists of two components: fibroglandular tissue (FGT) and fat.

FGT is a mixture of fibrous connective tissue (the stroma) and the glandular epithelial

cells that line the ducts of the breast (the parenchyma). FGT has a high X-ray

attenuation coefficient and therefore appears bright on a mammogram. Conversely, fat

has a lower X-ray attenuation coefficient than FGT, and therefore appears dark on a

mammogram. Therefore, MD is a measure of the relative amount of radio-dense FGT

as opposed to the amount of the radiolucent adipose tissue in the breast.

2.3.1 Mammographic Density – Clinical Significance and Methods of

Assessment

Research investigations have shown that, after correcting for body mass index

and age, women in the highest MD quartile are 4 to 6 times more likely to develop BC

over time than those in the lowest 10% [232, 233]. MD appears to be the second highest

risk factor for BC (after BRCA mutation) [234] and this risk of BC has been shown to

persist over 10 years follow-up [235]. MD refers to the degree of radio-opacity of the

breast as observed on a mammogram. Based on the distribution of MD observed in a

mammogram, the relative prevalence of FGT and fat in the breast can be inferred [75].


46

A density classification scheme, known as Breast Imaging Reporting and Data System

(BI-RADS) [74, 236] is widely used for reporting the findings on mammography.

Based on the MD, it classifies the examined breast into four categories: BI-RADS 1

indicates a predominantly fatty breast; BI-RADS 2 scattered fibroglandular densities;

BIRADS 3 a breast that is heterogeneously dense; and BIRADS 4 an extremely dense

breast that can potentially obscure a lesion [75, 237]. A recent study has reported on

the presence of increased dense connective tissue stroma and small and low-complexity

glands in HMD regions [238]. Although the altered PG content has been considered

responsible for the physical properties of MD and the associated increase in BC risk

[234], the molecular basis and the pathobiology of MD remains elusive to date.

Nevertheless, the identification of HMD and assessment of the distribution of MD in

breast is of fundamental importance for predicting BC.

Currently, mammography is the only available method for measuring MD.

However, it suffers from several limitations. Due to the use of ionizing radiation,

mammography is not recommended for frequent screening, usually no more than once

every 2 years [239]. For the same reason, mammography is not recommended for young

women or women who have inherited syndromes that are associated with radio-

sensitivity and/or cancer risk [240, 241]. In addition, the accuracy of mammography is

affected by breast compression that leads to tissue overlap and consequently

projectional imaging artefact [242]. The accuracy of mammography is also inversely

correlated with MD [243, 244]. This means that the high BI-RADS groups, which are

more likely to have BC are also more prone to erroneous diagnosis by mammography.

In fact, in order to be certain about BC, a high BI-RADS score warrants additional tests

that are less affected by density, such as ultrasound or MRI [75]. Currently, researches

are in the lookout for an alternative imaging technique that can measure MD-analogous

quantities without using ionizing radiation [245, 246]. Although several imaging

techniques have been suggested for this, MRI seems particularly suited for the purpose

for the following reasons: MRI allows spatially resolved assessment of MD, MRI

results have shown good correlation with MD measurements acquired from

corresponding mammograms [79], the proportion of FGT in breast has been quantified

using clinical MRI [79], and MRI-derived volumetric measurements of MD, using

semi-automated or fully-automated clustering or segmentation algorithm, have shown

good agreement with MD measurements in several studies [84-87, 232, 247-251].


47

However, the cost of a MRI scan is substantially higher than that of a mammogram and

therefore the adaptation of conventional MRI for routine breast screening is unlikely.

On the contrary, portable NMR instruments [89, 90] provide a low-cost alternative to

MRI for breast screening while it shows the potentials to follow the same fundamental

principles for MD assessment like MRI.

2.3.2 Assessment of Mammographic Density using Portable NMR

Portable NMR uses an assembly of permanent magnets to induce magnetization

in order to generate measurable NMR signal in the specimen under examination. It

employs the same fundamental NMR phenomenon as MRI to study the 1H within a

sample. An overview and the common use of such NMR instruments have been

discussed in a previous section of this chapter (see section 2.1.7.2). The architecture

and functionalities of the surface RF coils vary among different models of portable

NMR and consequently varies the size of the sensing area, achievable penetration or

sampling depth and resolution [252]. Although it offers depth-wise resolution, unlike

the in plane 2D resolution of MRI, it only acquires an average signal from the FOV and

thereby offers 1D in plane resolution.

In the context of MD assessment, portable NMR provides a range of

measurement options that can be used to infer the density of the tissue under

examination. The thickness of the sample and the generic distribution of 1H within the

sample can be obtained by PD weighted scan as a function of depth. Additionally, the

measurement of self-diffusion coefficient or diffusion tensor may provide a means to

identify the distribution as well as the extent of 1H mobility in tissue that may provide

important information on the associated MD. Breast tissue with HMD is known to

contain a larger proportion of FGT tissue in comparison to tissue with LMD. At the

same time, the water content of breast tissue is highly correlated with the prevalence of

FGT [75]. Owing to this difference in the water content of tissue with HMD and LMD,

T1 – weighted MR images can provide contrast between water-rich FGT and adipose

tissue. A recent research investigation has demonstrated that T1 relaxation time

measured by portable NMR can distinguish between breast tissue regions with HMD

from regions with LMD. The T1 values measured from regions with HMD (T1 = 170 ±

30 ms in full breast slice and T1 = 160 ± 30 ms in excised regions) were significantly


48

different (P < 0.001) from those measured in regions with LMD (T1 = 120 ± 10 m in

both full slices and excised regions) [97].

Portable NMR also allows the measurement of T2 NMR relaxation times by

CPMG sequence. However, in a large sensing area (~15 x 15 mm in the model used in

the study presented in this thesis), multiple tissue components, each with a distinctive

T2 value, are expected to co-exist within this FOV. Therefore, the T2 relaxation decay

measured from a breast tissue region is likely to be a compound relaxation decay.

Nevertheless, when multiple T2 components are expected in a T2 relaxation decay, the

signal can be analysed using one dimensional inverse Laplace transform in the form of

a T2 distribution (see section 2.1.6.2). The resulting T2 distribution holds distinctive T2

peaks where each peak correspond to a tissue component that has a unique T2 while it

also shows the relative contribution of each T2 to the total NMR signal that may relate

to the relative prevalence of each tissue component (with distinct T2). A research

investigation conducted by portable NMR is presented in chapter 5 of this thesis where

T2-based NMR was employed to identify the presence of adipose tissue and water in

breast tissue and also to quantify their distribution.

2.4 Transverse Relaxation in Biological Tissues

The transverse relaxation of 1H NMR is a fundamental physical phenomenon

sensitive to the chemical and anatomical environment of native 1H population. In

biological tissues, the transverse relaxation may originate from a number of distinct

scales – molecular (in the ranges of nanometres (nm)), cellular (in the ranges of

micrometres (µm)), and macroscopic (in the ranges of millimetres (mm)) [253]. The

nature of the transverse relaxation decay is modulated by the transverse relaxation time

constant T2 (Eqn. (12) in Section 2.1.3.3). T2 is the time required for the transverse

magnetization to fall to approximately 37% of its initial value. Clinical MRI scanners

are capable of identifying the changes in transverse relaxation or T2 at the macroscopic

level.

In clinical practise, T2 weighted images are commonly used to investigate

pathological conditions in non-calcified tissues. In order to achieve adequate T2

weighting, the values for TR and TE are chosen so that the sequence weighting

highlights differences in T2 of tissues under examination. Tissues/pathological


49

conditions with particular T2 weightings are recognised based on the voxel intensity

level observed in clinical MRI. In general, the acquisition of T2 weighted images are

multiple-times faster in comparison to the T1 weighted imaging. The time efficiency of

T2 imaging is often beneficial for in vivo imaging. T2 weighted sequences are used in

cardiovascular magnetic resonance (CMR) to identify myocardial edema associated

with ischemia, inflammation, vasculitis, or intervention in the myocardium [21]. T2

weighted CMR has also been shown to identify acute or recent myocardial ischemic

injury and to distinguish acute coronary syndrome (ACS) from non-ACS as well as

acute from chronic myocardial infraction [254]. T2 mapping using clinical scanners has

shown the potentials to detect and quantify myocardial edema [255].

T2 weighted MRI is capable of producing excellent contrast between fat (high

signal intensity) and muscle (intermediate signal intensity) and therefore is often

employed to study skeletal muscles [19]. Abnormal signal intensity within skeletal

muscle, as observed in T2 weighted MRI, is often related to various pathological

conditions including traumatic, infectious, autoimmune, inflammatory, neoplastic,

neurologic, and iatrogenic conditions [20]. T2 relaxation time has been identified as a

reliable quantifier of muscle inflammation in children with juvenile dermatomyositis

[40]. T2 mapping has been useful in measuring muscular fat that differentiated subjects

with Duchenne muscular dystrophy (DMD) from healthy subjects [39, 256]. T2

measured from skeletal muscle water has been proven to be a sensitive biomarker of

the disease status in DMD [257]. Muscle specific T2 has been used to access the effects

of exercise [258] and to identify changes in muscle volume and body composition after

space flight [259].

Parametric MRI is used extensively to assess breast tissues and diagnose breast

lesions. Breast MRI is increasingly used as an adjunct to conventional mammography,

particularly in diagnostic problem cases and for pre-operative staging [78]. MRI allows

spatially resolved localisation of breast lesions and detailed mapping of breast tumours.

Breast MRI protocols commonly include a T2 weighted unenhanced sequence, with or

without fat suppression. T2 weighted images with fat suppression allows easier visibility

of fluid intensity at the expense of spatial resolution while T2 weighted images without

fat suppression provide better depiction of the normal tissue architecture and lesion

morphology. It has been observed that, careful analysis of T2 weighted images can

reduce the false-positive rates and can distinguish rare well-circumscribed breast


50

carcinomas from common benign breast masses [22]. Visual assessment of lesion

appearance in T2 weighted images allowed the distinction between fibroadenomas and

breast cancers with high sensitivity and specificity [23]. Transverse relaxation based

MRI is also used to measure mammographic density, which is an independent risk

factor for breast cancer. The volumetric breast density measurements obtained from

MRI-based semi-automated or fully automated clustering or segmentation algorithms

have shown good agreement with conventional measurements of mammographic

density in multiple studies [84-87, 249-251].

Transverse relaxation based MRI is widely used to study the anatomy and

functionalities of nervous system as well as to investigate brain pathologies. T2

weighted MRI is particularly sensitive to the extent of water compartmentalization at

the micro-level that enables precise discrimination between white matter (WM), grey

matter (GM) and cerebrospinal fluid (CSF) [17]. T2 distributions measured from T2

maps obtained from normal brain show peaks from myelin water, intra/extracellular

water and CSF. These measurements can be used to provide estimates of total water

content (total area under the T2 distribution) and myelin water fraction (MWF,

fractional area under the myelin water peak) and identify different white matter

structures that have characteristic MWFs [33]. Multiple sclerosis (MS) is a chronic

disease of central nervous system that involves demyelination of WM. Multi-echo T2

relaxation analysis has been shown to be effective in myelin water imaging for the

detection of MS [24]. Quantitative analysis of T2 distributions have shown that normal-

appearing white matter (NAWM) in MS brain possesses a higher water content and

lower MWF than controls, which is consistent with histopathological findings [33].

Long T2 related abnormalities have been identified in the WM of subjects with either

phenylketonuria or MS [25]. However, shortening of T2 has been detected by abnormal

iron deposition in the brains of MS patients that may relate to neuronal degeneration in

MS [260]. Subjects with schizophrenia were found to have significantly reduced MWF

in the minor forceps and genu of the corpus callosum when compared to controls, which

suggested that that reduced frontal lobe myelination plays a role in schizophrenia [33].

A research investigation has shown that combination of T2 and T2* weighted MRI can

differentiate Parkinson's disease patients from normal age‐matched controls based on

the differences in iron content within the substantia nigra [261]. T2 weighted MRI has

been used to detect structural abnormalities of GM in migraine patients with brain T2


51

visible lesions using voxel-based morphometry. It was postulated that such changes in

GM comprise areas with increased or reduced density that are likely related to the

pathological substrates associated with this disease [262].

MRI assessment is routinely employed in the care of the brain tumour patients.

Transverse relaxation based sequences are employed to diagnose brain tumour and also

to identify the structure and distribution of the tumour tissues with detailed spatial

information. Metastatic brain tumours from gastric and colon cancers have been

revealed by hypo intensity on T2 weighted MRIs [263]. Progression of high grade

glioma has been linked to significant non-enhancing signal increase in T2 weighted

images [26]. Additionally, attempts have been made to establish automated tumour

segmentation framework based on T2 weighted and T1 weighted brain MRI [264]. It

should be noted that the transverse relaxation decays measured in brain are often multi-

exponential and are analysed by specialised quantitative analysis technique to identify

individual T2 components in the relaxation decay. Quantitative or quantified T2

relaxation times have been effective in diagnosing glioma and in monitoring tumour

progression [15, 26]. In addition, quantitative analysis of transverse relaxation using

µMRI has been successful in detecting pathological water compartments with particular

T2 components in murine models of glioblastoma [12, 13]. These µMRI studies have

examined transverse relaxation mechanism at the cellular level scale or in the ranges of

µm.

Transverse relaxation analysis approaches, including both T2 weighted MRI and

T2 mapping, are extensively used to measure cartilage thickness, identify cartilage

abnormalities and to identify active healthy subjects at higher risk for developing

cartilage pathology [3, 68, 108, 265]. Quantitative T2 has been established as a reliable

biomarker by µMRI studies to interpret the collagen scaffold of articular cartilage and

thereby to interrogate the structural integrity of cartilage [5, 7, 28-31]. The

applications of transverse relaxation based analysis for cartilage assessment have

been discussed in detail in previous sections (section 2.2.2 and section 2.2.3.1).

Transverse relaxation based MRI has also been employed to investigate various

biological tissues and organs, including, but not limited to, liver [27], tendon [266],

ligament [267], and bone [112, 268]. T2 weighted MRI has also been used to detect and

examine prostate cancer in conjunction with diffusion weighted or T1 weighted MRI

[269, 270].

Chapter 3: Assessment of Collagen Architecture

52

Chapter 3: Transverse Relaxation based Assessment of Collagen

Architecture in Cartilage

____________________________________________________________________________________________

3.1 Prelude

Articular cartilage (AC) plays key roles in joint movement by creating a low

friction protective barrier for gliding and by distributing stress and transmitting loads

to the underlying bones [48, 49]. The structural scaffold of AC is made up of cross-

linked collagen networks, which also comprise the majority of the dry weight in AC

[271]. Collagen (type I and type II) is the most abundant protein in the body and a major

constituent of the tissue extra cellular matrix (ECM) that offers structural support for

tissue cells [44, 45]. Collagen macromolecules restrict the movement of water

molecules in cartilage ECM. In transverse relaxation based MRI, the anisotropic

collagen distribution in AC often results in an artefact that manifests as laminar patterns

in AC [58-61]. This laminar appearance varies with the change in orientation of the

imaged cartilage with respect to the static magnetic field used for MRI [43]. This

orientational dependence of the measured T2 on the collagen anisotropy is commonly

known as the magic angle effect [1, 30, 62].

The nature of collagen alignment and distribution varies across the depth of AC

and that typically creates three histological zones in cartilage ECM: superficial zone,

transitional zone and radial zone [46, 47]. It is postulated that the three-zone structure

governs the response of cartilage to dynamic loading during movement [50, 51]. In

addition, the shear and tensile properties of AC are also dependent on the underlying

collagen scaffold in cartilage ECM [46, 52]. Collagen fibre organisation in AC can be

interrogated by several experimental techniques, most notably, scanning electron

microscopy (SEM) [53-56] and polarised light microscopy (PLM) [53, 54, 57]. These

techniques provide high resolution (< 1 µm) insight into the collagen alignment and can

be used to assess changes in the collagen organisation. However, both of these

techniques are destructive and therefore are unsuitable for longitudinal studies or for in


53

vivo evaluation. On the contrary, T2 is sensitive to the restricted rotational and

translational motion of water molecules bound to collagen fibres, which makes the

measured T2 dependent on the orientation of the collagen fibres relative to the static

magnetic field used for MRI [43]. The method of achieving T2 weighted MRI is both

non-destructive and non-invasive. Using an empirically derived formula, the magic

angle effect of T2 MRI has been used to assess collagen fibre alignment in ligaments

[65] and in AC [1, 28-30, 63, 64, 66, 67]. To date, the collagen architecture has been

investigated in considerable detail in the AC obtained from human [28, 68], bovine [8,

29] and canine joints [62]. Consequently, attempts have been made to establish links

between the collagen organisations observed in AC samples with the inherent

biomechanical functionalities of the same sample.

The gait pattern of an animal sets the requirements for the functions of its knee

joint, which in turn impacts the structural make-up of its AC. The hopping locomotion

of kangaroo has gait parameters that are significantly different from that of bipedal and

quadrupedal running mammals. A large kangaroo hops at an average speed of 40 km/h

and may reach a speed of 50 – 65 km h-1 in short bursts [272]. During this movement,

the ground reaction force and load experienced by the kangaroo knee joints are several

times higher than that of human walking and running [273, 274]. The stride frequency

of a hopping kangaroo is also higher than that of a human when progressing at the same

velocity [273]. The unusual makeup (in comparison to mammals) of the knee cartilage

in kangaroo is believed to support the very high ground reaction force experienced

frequently during hopping locomotion. The articulating surfaces of kangaroo knee joint

are lined with a combination of hyaline cartilage and fibrocartilage. Contrary to the type

II collagen in hyaline cartilage, the fibrocartilage contains type I collagen as the major

constituent of ECM [19, 20]. However, the research on kangaroo cartilage remains

limited and the overall organization of collagen fibres in AC is not well understood.

The collagen distribution identified by transverse relaxation based MRI studies

have mostly used samples from human, bovine, equine and canine cartilages due to the

size, ease of use and availability. Consequently, the three zone model of collagen

organization also represent the collagen architecture in AC only specific to these

species. However, the thickness of the histological zones of AC, as well as the

composition and organization of the major molecular components, have been observed

to vary across species and even across different sites in the same joint [69-72]. The


54

cartilages of a kangaroo knee joint - the femoral hyaline cartilage, the tibial hyaline

cartilage and the tibial fibrocartilage - each plays a unique role in the hopping

locomotion. Therefore, each of these cartilage types is also expected to have a unique

collagen organization allowing the accommodation of its respective biomechanical

demands. Although the kangaroo cartilage has been examined by histological analysis

and using polarised light microscopy, little is known about the collagen distribution at

different cartilages and their corresponding functionalities. A thorough understanding

of the collagen organization is therefore necessary in order to comprehend the

mechanical properties of AC.

This chapter presents an experimental study of transverse relaxation based

assessment of the collagen architecture in kangaroo femorotibial cartilages that

addresses the first objective of this thesis. In this study, magic angle effect was used to

probe the collagen distribution in femoral hyaline cartilage, tibial hyaline cartilage and

tibial fibrocartilage of the red kangaroo (Macropus rufus). The aim of this work was to

identify the arrangement of the collagen fibres specific to these cartilage types. A good

understanding of the structural make-up of kangaroo cartilage may inspires new designs

for tissue engineering while the analysis method established by this study can be used

as a valuable technique for the in vivo assessment of cartilage and for regenerative

therapies.

This study is presented in the form of a journal article (doi:

10.1016/j.mri.2017.07.010) published in the Magnetic Resonance Imaging [124]. This

chapter is self-contained with headings, the main body of the article, figures and tables

as they appear in the accepted manuscript. The references are merged with the

bibliography at the end of this thesis.


55

3.2 Statement of Co-author Contribution

The authors listed below have certified that:

they meet the criteria for authorship in that they have participated in the conception,

execution, or interpretation, of at least that part of the publication in their field of

expertise;

they take public responsibility for their part of the publication, except for the

responsible author who accepts overall responsibility for the publication;

there are no other authors of the publication according to these criteria;

potential conflicts of interest have been disclosed to (a) granting bodies, (b) the editor

or publisher of journals or other publications, and (c) the head of the responsible

academic unit, and

they agree to the use of the publication in the student’s thesis and its publication on the

QUT’s ePrints site consistent with any limitations set by publisher requirements.

In the case of this chapter:

MRI magic-angle effect in femorotibial cartilages of the red kangaroo

Tonima S. Ali, Namal Thibbotuwawa, YuanTong Gu, Konstantin I. Momot

Published: Magnetic Resonance Imaging 43 (2017) 43: 66-73

Contributor Statement of contribution

Tonima S. Ali

Conceived and designed the study, assisted in sample preparation,

performed MRI experiments and data analysis, co-wrote

manuscript.

Signature

Date

Namal Thibbotuwawa Conceived and designed the study, acquired samples, assisted in

data analysis and interpretation, revised manuscript.

YuanTong Gu Conceived and designed the study, revised manuscript.

Konstantin I. Momot Conceived and designed the study, prepared samples, co-wrote the

manuscript, supervised the study.

Principal Supervisor Confirmation

I have sighted email or other correspondence from all co-authors conforming their

certifying authorship.

Konstantin I. Momot

Name Signature Date




56

3.3 MRI magic-angle effect in femorotibial cartilages of the red

kangaroo

Tonima S. Alia,b, Namal Thibbotuwawa a, YuanTong Gu a and Konstantin I.

Momota,b*

a Queensland University of Technology (QUT), Brisbane, Queensland, Australia

b Institute of Health and Biomedical Innovation, Kelvin Grove, QLD 4059, Australia

* Corresponding author:

Dr. Konstantin I. Momot


Queensland University of Technology (QUT)

GPO Box 2434, QLD 4001, Brisbane, Australia

Phone: +61-7-3138-1173

Fax: +61-7-3138-9079

Email: [email protected]


57

Keywords:

Hyaline cartilage, Fibrocartilage. Red kangaroo (Macropus rufus), Collagen

architecture, T2 anisotropy, Magic-angle effect

Abbreviations:

AC, articular cartilage

AS, articular surface

B0, static magnetic field

BW, bound water

ECM, extracellular matrix

FW, free water

MRI, magnetic resonance imaging

PBS, phosphate-buffered saline

PG, proteoglycans

R2, transverse relaxation rate constant

R20, R2 when the sample is perpendicular to the static magnetic field

R255, R2 when the sample is at 55o to the static magnetic field

R2A, anisotropic component of the transverse relaxation rate constant

R2I, isotropic component of the transverse relaxation rate constant

T2, transverse relaxation time constant

x, relative depth from the articular surface

µMRI, magnetic resonance microimaging

, angle between the static magnetic field and the predominant collagen direction

S, angle between the normal to the articular surface and the static magnetic field


58

ABSTRACT

Objective: Kangaroo knee cartilages are robust tissues that can support knee flexion

and endure high levels of compressive stress. This study aimed to develop a detailed

understanding of the collagen architecture in kangaroo knee cartilages and thus obtain

insights into the biophysical basis of their function.

Design: Cylindrical/square plugs from femoral and tibial hyaline cartilage and tibial

fibrocartilage were excised from the knees of three adult red kangaroos. Multi-slice,

multi-echo MR images were acquired at the sample orientations 0° and 55° (“magic

angle”) with respect to the static magnetic field. Maps of the transverse relaxation rate

constant (R2) and depth profiles of R2 and its anisotropic component (R2A) were

constructed from the data.

Results: The R2A profiles confirmed the classic three-zone organisation of all cartilage

samples. Femoral hyaline cartilage possessed a well-developed, thick superficial zone.

Tibial hyaline cartilage possessed a very thick radial zone (80% relative thickness) that

exhibited large R2A values consistent with highly ordered collagen. The R2

A profile of

tibial fibrocartilage exhibited a unique region near the bone (bottom 5-10%) consistent

with elevated proteoglycan content (“attachment sub-zone”).

Conclusions: Our observations suggest that the well-developed superficial zone of

femoral hyaline cartilage is suitable for supporting knee flexion; the thick and well-

aligned radial zone of tibial hyaline cartilage is adapted to endure high compressive

stress; while the innermost part of the radial zone of tibial fibrocartilage may facilitate

anchoring of the collagen fibres to withstand high shear deformation. These findings

may inspire new designs for cartilage tissue engineering.


59

1. Introduction

Articular cartilage (AC) covers the articulating surfaces of the femorotibial joint

and provides a low-friction surface that facilitates joint movement [29, 275]. AC also

absorbs shocks and distributes mechanical stress; this enables a lifetime function of the

joint. Mammalian AC, composed of hyaline cartilage, is an avascular tissue with a

complex three-dimensional architecture, where chondrocytes are embedded in an

extracellular matrix (ECM) principally made of type II collagen fibres, proteoglycans

(PG), and water [29, 275].

Collagen is the most abundant structural macromolecule of cartilage ECM

(1020% of the total wet weight) that contributes to the shear and tensile properties of

AC [46, 276]. Collagen forms a cross-linked fibre network, whose degree of alignment

and the predominant direction of alignment vary with depth, as does the collagen

volume fraction. This variation typically creates three histological zones in cartilage

ECM: the superficial, transitional and radial zones [46, 47]. The superficial zone

contains densely packed collagen fibrils aligned parallel to the articular surface (AS)

[46]. This facilitates stress distribution [277, 278] and enables fast tissue response at

high loading rates [179]. In the transitional zone, the collagen fibrils are orientationally

disordered [46]. The radial zone has highly aligned collagen fibrils oriented primarily

perpendicular to AS. The concentration of PG tends to increase with the depth, with a

maximum in the radial zone [29, 30, 46]. PG-rich radial zone facilitates the response of

cartilage to compressive loading by reducing excessive deformation [179, 180, 279]. A

calcified zone marks the separation between cartilage and subchondral bone, with the

calcification progressively increasing closer to the bone. This three-zone structure

governs the response of the cartilage to dynamic loading during movement [279, 280];

its development has been linked to the functional adaptation to weight-bearing in early

life [281-283].

This zonal model of cartilage has been derived primarily from data obtained

from human, bovine, equine and canine joints. At the same time, research of kangaroo

knee cartilage remains limited, despite the extraordinary biomechanics of kangaroo

knee joints and consequent high biomechanical demands placed on the cartilage tissue,

as well as the unusual makeup of knee cartilage in these species. The ground reaction

force exerted on a kangaroo knee joint is several times higher and more frequent than


60

the force experienced by human joints because kangaroos hop at an average speed of

40 km/h and with a higher stride frequency than a human running at the same speed

[272-274]. Kangaroo knee joint is articulated with a combination of hyaline cartilage

and fibrocartilage: hyaline cartilage lines the femur and the edges of the tibial plateau,

while the central part of the tibial plateau is covered with fibrocartilage (which contains

type I collagen as the major ECM constituent) [284, 285]. This unique combination of

hyaline and fibrocartilage is believed to support the high level of dynamic stresses

experienced by kangaroo knee joints during hopping locomotion.

Magnetic Resonance Imaging (MRI) magic-angle effect can probe collagen

fibre alignment in AC [1, 6, 9, 29, 173, 286, 287]. Extracellular water in cartilage is

present in two chemically exchanging “pools”: free water (FW) and water bound to the

ECM macromolecules (BW) [275, 288-290]. The transverse relaxation rate constant

(R2 = 1/T2) of water protons is the sum of the isotropic and the anisotropic contributions,

R2I and R2

A. The anisotropic contribution (R2A) is dependent upon the angle () between

MRI static magnetic field (B0) and the predominant collagen direction [1, 64, 94]:

𝑅2 = 𝑅2𝐼 + 𝑅2

𝐴 = 𝑅2𝐼 + 𝑅2

𝐴0 (3𝑐𝑜𝑠2Ɵ−1

2)2

(1)

In an aligned collagen network, R2A is usually positive; it approaches zero when

= 54.74° (the magic angle) [291], or when the collagen network lacks an alignment

order. Spatially resolved R2 maps and depth-resolved R2 profiles therefore reflect the

spatial organisation and zonal structure of the cartilage ECM [1, 6, 8, 30, 43, 173].

We used micro-MRI (µMRI) profiling of the R2 magic-angle effect to probe the

collagen architecture in the three types of kangaroo femorotibial cartilage: femoral

hyaline cartilage, tibial hyaline cartilage and tibial fibrocartilage. We hypothesise that

each type has a unique collagen organisation adapted to accommodate its respective

biomechanical demands in the hopping locomotion. We also hypothesise that the

differences in the MRI properties between different types of cartilage provide insights

into the biomechanical differences between them. We discuss the likely relationship

between collagen architecture in different types of cartilage and their respective

biomechanical roles in the knee joint.


61

2. Materials and methods

2.1. Sample preparation

Visually normal whole knee joints of three red kangaroos (Macropus rufus, ~5

years of age) were purchased from a local abattoir within 24 hours of slaughter and

stored at 18°C. Bone-cartilage plugs were excised for the MRI measurements. The

samples of femoral hyaline cartilage were cylindrical (8 mm diameter) and excised with

a hole saw. Tibial hyaline cartilage samples were rectangular and excised using a

regular saw due to the difficulty of holes aw positioning. Tibial fibrocartilage samples

included one cylindrical plug (hole saw) and two square plugs (regular saw). Each

sample had a nearly flat (within 10°) AS and a layer of subchondral bone. The samples

were placed in phosphate-buffered saline (PBS; pH 7.4, NaCl 0.138 M, KCl 0.0027 M;

prepared from PBS concentrate, Sigma-Aldrich, Australia) for two hours before MRI.

The Animal ethics approval (1200000376) was granted by Queensland University of

Technology.

2.2. MRI protocol

MRI measurements were conducted at room temperature on a Bruker Avance

NMR spectrometer (Bruker, Germany) operating at 7 T and equipped with a 1.5 Tm1

(120 Gcm1) triple-axis gradient set and a Micro2.5 microimaging probe with a 15mm

1H birdcage radiofrequency coil. The sample was placed in PBS inside a 15mm NMR

tube (Wilmad, USA) between two purpose-built Teflon inserts [57, 292, 293] that kept

the sample in the required orientation and prevented its movement. For each sample,

two sets of MR images were obtained: with the AS perpendicular and at 55° to B0 (S

= 0o and S = 55o, respectively).

For every sample, three vertical slices were imaged using multi-slice multi-echo

(MSME) sequence (TR = 4555.52 ms, TE = 4.23 ms, 25 echoes, 1 mm slice thickness,

zero slice gap, FOV 30 mm 15 mm, 256 128 voxels, 16 averages). The FOV

contained a complete cross-section of the cartilage-subchondral bone plug. In order to

minimise the MRI signal from PBS, a slice-selective inversion pulse was applied 1400

ms prior to the imaging pulse sequence. Nine datasets (3 samples 3 imaging slices)

were thus obtained for each type of cartilage studied. The minimum SNR (defined

relative to the background noise) was 17:1.


62

2.3. Relaxation mapping

The transverse relaxation rate constant, R2 = 1/T2, was determined individually

for every voxel by fitting a three-parameter mono-exponential decay function to the

MSME data:

𝑆(𝑡) = 𝑆0𝑒(−𝑡 × 𝑅2) + 𝑆𝑜𝑓𝑓𝑠𝑒𝑡, (2)

where S is the voxel signal intensity and t is echo time (ranging in 25 equidistant

increments from TE to 25TE). A maximum of 100 fitting iterations were allowed for

the pixels with S0 > 5 × Soffset. Fitting residuals were checked for randomness by Runs

Test [294] (α = 0.05) to verify mono-exponentiality of the decay. In order to ensure

physically meaningful fits, only the voxels within the range 0.0005 ms1 < R2 < 0.5

ms1 were included in further analysis. An R2 map was constructed for every imaging

slice. All data analysis was performed using in-house code written in MATLAB

R2014a (MathWorks, USA).

2.4. R2 depth profiles

For every imaging slice, a series of approximately equidistant points were

specified on the cartilage AS in the second T2-weighted MSME image (TE = 8.46 ms).

A third-degree polynomial curve was fitted to the points. Another series of points was

specified on the cartilage-subchondral bone interface; these were also fitted with a

third-degree polynomial. The cartilage tissue enclosed between the two fitted curves

was taken as the region of interest (ROI), with the lateral edges excluded in order to

avoid susceptibility artefacts. Minimum geometric distances from AS (DAS) and from

cartilage-subchondral bone interface (DSB) and the relative depth from AS (x = DAS /

(DAS + DSB)) were measured for each ROI voxel. A relative-depth profile of R2 was then

constructed by making a histogram of the x values of the ROI voxels and calculating

the average R2 of the voxels contained in each bin of the histogram. Two relative-depth

R2 profiles were constructed for each imaging slice of each sample, one for S = 0o and

the other for S = 55o.

2.5. R2A depth profiles and zone boundary identification

For every imaging slice, a relative-depth profile of the anisotropic component

of R2 (R2A) was computed. Initially, the provisional R2

A profile was defined as the

difference between the R2 profiles at the two sample orientations (S = 0° and 55°):


63

𝑅2𝐴0 = 𝑅2

0 − 𝑅255 (3)

From the provisional R2A profile, the isotropic transitional zone was identified

by the absence of discernible anisotropy (R2A ~ 0). Starting from the apparent centre of

the transitional zone, two moving averages of R2A and their associated standard

deviations (SD) were computed in both directions, towards the superficial and the radial

zones. The borders of the transitional zone were determined as the depths where two

consecutive moving averages of R2A differed by more than one SD. The superficial and

the radial zones were identified as the zones above and below the transitional zone,

respectively.

The provisional R2A profile was then modified as follows. At S = 0°, R2

A0 can

be assumed to reach its maximum in the radial zone, where the predominant collagen

alignment is approximately parallel to B0 and ((3cos2 – 1)/2)2 = 1. However, in the

superficial zone at the same sample orientation, R2A0 reaches only a quarter of its

maximum value due to the predominant collagen alignment being perpendicular to B0

and ((3cos2 – 1) / 2)2 = 0.25. Therefore, the R2A0 values in the superficial zone were

corrected by a factor of 4:

𝑅2𝐴𝑆 = 4 × (𝑅2

0 − 𝑅255) (4)

For each cartilage type and each sample orientation, an average R2 profile was

computed by pooling together the pairs [86] for all ROI voxels from nine R2 maps (3

slices 3 samples). For each cartilage type, the average R2A profile was then computed

from the average R20 and R2

55 profiles using Eqs. (3) and (4).

3. Results

Figure 1(AC) shows the typical anatomical locations of the bone-cartilage

plugs analysed in this study. Representative T2-weighted images are shown in Fig. 1(D–

F).

3.1. Femoral hyaline cartilage

The first column of Fig. 2 shows the R2 maps and the relative-depth R2 profiles

of a representative femoral hyaline cartilage (FHC) sample. The cartilage thickness in

this sample varied between 1.05 mm and 1.29 mm. At S = 0° (Fig. 2A), a gradual

increase of R2 with depth was observed throughout most of the radial zone, with the


64

greatest R2 (0.07 ms1) in the region x = 0.851. The corresponding depth profile at S

= 55° (Fig. 2D) was nearly flat and exhibited a relatively minor R2 increase in the radial

zone. The relative-depth profile of R2A (Fig. 2G) showed a region with R2

A ~ 0 that was

identified as the transitional zone, x ~ 0.25–0.5. Non-zero R2A was observed in the

superficial (x < 0.25) and the radial zone (x > 0.5).

Figure 1. The typical anatomical locations (A–C) and representative T2-weighted (TE =

8.46ms) MR images (D–F) of the samples used in the study: (A, D) femoral hyaline cartilage;

(B,E) tibial hyaline cartilage; and (C,F) tibial fibrocartilage. The cylindrical samples were

excised using a holesaw drill and the square sample was excised using a hand-held saw, as

described in Section 2.1. In (A), the sample used for the measurements was taken from the

upper-right of the two holes seen in the photograph; the bottom-left hole is an auxiliary channel

used to release the sample from the main bone.

Fig. 3(A, B) presents the R2 depth profiles averaged over the three FHC samples.

The corresponding average depth profile of R2A is shown in Fig. 3B; it shows

anisotropic superficial and radial zones separated by isotropic transitional zone. The

superficial zone comprised on average nearly 30% of FHC thickness and exhibited the

maximum R2A of 0.02 ms1 (see Table 1); the characteristics of the radial zone were

similar to those of the sample in Fig. 2.

3.2. Tibial hyaline cartilage

The second column of Fig. 2 shows the results for a representative tibial hyaline

cartilage (THC) sample. The cartilage had a non-uniform thickness ranging from 1.17

mm to 2.34 mm. The R2 map at S = 0° (Fig. 2B) exhibited a band of high R2 values

next to the subchondral bone; the corresponding depth profile exhibited a very wide


65

region of high R2 values. The high-R2 band was absent at S = 55° (Fig. 2E), and the

corresponding depth profile was relatively flat with a slight increase in R2 towards the

subchondral bone. The relative-depth profile of R2A (Fig. 2H) revealed a thin superficial

zone with low R2 anisotropy and a very wide radial zone (over 80% of the total

thickness) with high anisotropy (maximum R2A 0.17 ms1).

Figure 2. Representative maps and the corresponding relative-depth profiles of the transverse

relaxation rate constant (R2): (A–C) Maps and relative-depth profiles of R20 (sample orientation

S = 0o); (D–F) same data for R255 (S = 55o); and (G – I) the relative-depth profiles of the

anisotropic component of R2 (R2A), computed as described in section 2.5 (see Eqs. (3) and (4)).

Each column represents the data from a single imaging slice of a single cartilage sample:

column 1, femoral hyaline cartilage sample 3 slice 1; column 2, tibial hyaline cartilage sample

1 slice 2; column 3, tibial fibrocartilage sample 2 slice 2. The three-zone structure is readily

apparent in each R2A profile.

The average R2 profiles of THC are shown in Fig. 3(C, D). The average

superficial zone had a maximum R2A of 0.04 ms1

and comprised ~ 10% of the total

THC thickness, while the average radial zone had a maximum R2A of 0.15 ms1

and

comprised 80% of the total thickness.


66

3.3. Tibial fibrocartilage

The results for a representative TFC sample are presented in the third column

of Fig. 2. Cartilage thickness varied from 2.11 mm to 3.86 mm, thicker than the hyaline

cartilage samples. At S = 0°, a high-R2 region was observed at x > 0.5, with the R2

increasing all the way to the subchondral bone (Fig. 2C). The relative-depth profile of

R255 (Fig. 2F) showed an accelerating increase close to the subchondral bone. The

lowest anisotropy (R2A ~ 0) was observed at x = 0.3; the maximum anisotropy (R2

A ~

0.09 ms1) was observed next to the subchondral bone. Unlike hyaline cartilage, the R2A

continued to increase throughout the radial zone of TFC.

Figure 3(E, F) shows the average R2 profiles computed from the nine

fibrocartilage R2 maps (3 slices 3 samples). The superficial zone had a maximum R2A

of 0.07 ms1 with a relative width of 33%. The greatest R2

A (0.1 ms1) was observed in

the radial zone, adjacent to the subchondral bone. The depth profile of the isotropic

component of R2 (Fig. 3E, S = 55°) exhibited a relatively rapid increase from x = 0.88

to x = 1.

4. Discussion

To our knowledge, this is the first MRI study of the femorotibial cartilages of

the red kangaroo (Macropus rufus) to date. Past studies have investigated the

ultrastructure of kangaroo cartilage using histology or optical microscopy [186, 273,

295], where the collagen alignment information is derived from small ROIs that do not

necessarily cover the entire cartilage. MRI, while providing a lower spatial resolution,

has the advantage of affording a ROI sufficiently large for a whole-sample view of

collagen organisation. Comparison of Figs. 2 and 3 reveals a subtle variability of

collagen alignment patterns within each cartilage type. This variability has generally

been overlooked in the past studies of kangaroo cartilage [282].


67

Figure 3. Average relative-depth R2 profiles obtained by averaging of the nine respective

individual profiles (three samples of each cartilage type, three imaging slices per sample): (A,

C, E) average profiles of R20 and R2

55; (B, D, F) average profiles of the anisotropic component,

R2AS, determined from R2

0 and R255 as described in section 2.5 (see Eqs. (3) and (4)). The three-

zone structure is apparent in all R2A profiles. Note the rapid increase of R2

55 between x=0.88 and

x=1 in tibial fibrocartilage (“the attachment sub-zone”, see Discussion).

4.1. Femoral hyaline cartilage

The R2A depth profile of a representative FHC sample is shown in Fig. 2G. All

FHC samples exhibited the typical three-zone appearance (superficial, transitional and

radial zones), similar to the known patterns seen in other mammals [6, 29, 60, 64, 282,

283, 296]. We have observed some variation in the thickness of the superficial zone

both between and within individual samples: the relative thickness of the superficial


68

zone in the three samples was 25 2, 28 14 and 33 3%, where the intra-sample

standard deviation was calculated on the basis of the three imaging slices. The R2

anisotropy data are summarised in Fig. 3B and Table 1.

The relative thickness of the superficial zone averaged over the three samples

was 28 8%. This value is large compared to, for example, bovine knee cartilage,

where the superficial zone often occupies less than 5% of the cartilage thickness. We

suggest that this is likely related to the biomechanical requirements of kangaroo

locomotion. The superficial zone, where collagen fibres are parallel to the AS, is crucial

to the tensile and shearing resistance of AC [46] and ensures an extremely low

coefficient of AS friction. During the hopping locomotion, the surface of the femoral

cartilage experiences a significant excursion due to large knee flexion, while also

transmitting the weight of the upper body to the tibia. We suggest that the large knee

flexion of the kangaroo (compared to most other mammals) necessitates the presence

of a femoral AC with a thick superficial zone, as seen here.

The radial zone of FHC exhibited a relatively low R2A (0.05 ms1). This is significantly

lower than the maximum R2A in the other two types of kangaroo knee cartilage.

Although the radial zone occupied ~60% of the total thickness of FHC (Fig. 3B and

Table 1), only ~20% corresponded to regions with R2A greater than two typical standard

deviations, or 0.03 ms1. Since R2A is normally well-correlated with the degree of

collagen alignment, the low R2A values observed in the radial zone of FHC suggest a

limited degree of collagen alignment there. Highly aligned collagen in the radial zone

is known to provide resistance to compressive forces [297]. Therefore, our observation

suggests that kangaroo FHC may experience limited amount of compressive stress,

possibly due to the large area of cartilage-covered femoral condyles combined with the

prominence of knee flexion in the hopping locomotion.

4.2. Tibial hyaline cartilage

The R2A depth profiles of THC also showed the typical three-zone structure. In

kangaroo tibia, hyaline AC is present only at the periphery of the tibial plateau, covering

~ 50% of its area [295]. Its AS is covered by two meniscal discs that protect the

superficial zone from friction and contribute to shock absorption and load transmission

in both active and static loading [298]. Therefore, the superficial zone of THC has only

a limited role in protecting the cartilage against shear during movement. This is


69

consistent with the narrow superficial zone observed in the THC: 7 2%, 13 1% and

13 1% in the three samples, or 11 3% on average.

The average R2A depth profile of THC (Fig. 3D) reveals the radial zone that is

significantly thicker (80 6%) than that of FHC (60 10%). Both the average R2A (0.09

0.04 ms1) and the maximum R2A (0.15 0.01 ms1) were significantly higher than

R2A in the radial zone of the other two types of cartilage (see Table 1). This suggests

that collagen in the radial zone of THC was more strongly aligned than in the radial

zone of the other types of kangaroo cartilage. Furthermore, the relatively high values

of R255 in this zone (Fig. 3) suggest a greater PG density than in all zones of the other

types of cartilage. The ground reaction force generated during hopping is several times

higher and cycled at more frequent intervals than in a walking or running human [273,

274]. In the human knee, the resistance to compressive forces provided by the radial

zone of AC is associated with its highly aligned and dense collagen network [297]. We

therefore posit that high PG density and high collagen alignment in the radial zone of

THC are a consequence of adaptation to the high-amplitude, high-frequency

compressive stresses experienced by the kangaroo tibia.

Table 1

The mean and maximum values of the anisotropic component of the transverse relaxation rate

constant (R2A) in each histological zone and the relative depths of the zones in each type of

kangaroo cartilage. For each cartilage type, the values are based on the combined data of the

nine imaging slices (3 samples 3 slices per sample). The data are presented as mean ± standard

deviation.

Superficial zone Transitional zone Radial zone

Mean anisotropic relaxation rate constant, 𝑅2

𝐴 (ms1)

Femoral hyaline cartilage 0.003 ± 0.002 0.001 ± 0.000 0.03 ± 0.02

Tibial hyaline cartilage 0.007 ± 0.003 0.006 ± 0.001 0.09 ± 0.04

Tibial fibrocartilage 0.009 ± 0.006 0.004 ± 0.001 0.04 ± 0.03

Maximum anisotropic relaxation rate constant, (R2A)max (ms1)

Femoral hyaline cartilage 0.020 ± 0.007 0.003 ± 0.005 0.05 ± 0.01

Tibial hyaline cartilage 0.040 ± 0.009 0.007 ± 0.005 0.15 ± 0.01

Tibial fibrocartilage 0.07 ± 0.02 0.004 ± 0.005 0.10 ± 0.05

Relative depth of histological zones (%)

Femoral hyaline cartilage 28 ± 8 13 ± 7 60 ± 10

Tibial hyaline cartilage 11 ± 3 9 ± 4 80 ± 6

Tibial fibrocartilage 33 ± 9 7 ± 3 60 ± 10


70

4.3. Tibial fibrocartilage

The R2A depth profiles of TFC also showed anisotropic superficial and radial

zones separated by an isotropic transitional zone. A previous study [295] reported that

the radial zone of kangaroo TFC is organised in domains of the typical size 10 – 100

m, in which the collagen alignment alternates between near-parallel and near-

perpendicular to the AS. It was hypothesised in that study that this architecture may

impart to the fibrocartilage “improved compression absorption, extensibility, elasticity

and resilience” [295]. The heterogeneous domain-like architecture is consistent with

our observation that both the mean and the maximum values of R2A in the radial zone

of TFC are relatively low compared to the radial zone of the adjacent THC (see Table

1, Fig. 3). At the same time, the fact that the mean R2A in the TFC exhibits a continuous

increase with the increasing depth (see Fig. 3F) suggests that the relative contributions

of the “parallel” and “perpendicular” domains vary with the depth, and the contribution

of the “parallel” domains likely increases closer to the bone.

It is noteworthy that the shape of the R2A depth profile in the radial zone of TFC

differed from the radial zones of both femoral and tibial hyaline cartilage. In the radial

zone of both FHC and THC, the mean R2A exhibited an increase with the depth at

relatively low depths, followed by a decrease closer to the bone; the reversal of the

trend was observed at x ~ 85% in FHC and ~ 75% in THC. The radial zone of TFC

showed no such reversal, and the mean R2A continued to increase with the depth all the

way to the subchondral bone. TFC depth profiles of R20 and R2

55 exhibited similar

behaviour. Furthermore, the R255 profile exhibited an inflection point at x ~ 88%, where

the rate of the increase of R255 with depth accelerated (see Fig. 3E). Since R2

55 is

interpreted as the “isotropic” R2 contribution that is dependent on the chemical

composition of the tissue but not on collagen alignment, the relatively rapid increase of

R255 past x = 0.88 could be interpreted as being due either to the increase in PG

concentration or to a highly undulated interface between the radial zone and the

calcified zone. The possible PG-related origin of this increase is supported by the results

of He et al [295], who have observed a distinctly different histomorphological texture,

high density of cell nuclei, a relatively high PG content and a non-uniform PG

distribution in the bottom 5-10% of the radial zone of kangaroo TFC in safranin O –

fast green stains (see especially Fig. 4A, ref. [295]). The possible role of calcification

in the sharp increase in R255 is supported by the results of Xia, who has observed a


71

similar increase in the bottom 10% of the radial zone of canine AC and attributed it to

the partial-voluming between the radial and the calcified zones amplified by their

undulated interface [29, 30]. The high PG content at the bottom of the radial zone, seen

by He, was unique to the tibial fibrocartilage and was not seen in the adjacent THC

[295], while the undulation noted by Xia was observed in hyaline AC rather than

fibrocartilage. We therefore hypothesise the origin of the rapid increase in R255 lies in

the elevated PG content just above the cartilage tidemark. It is, however, impossible to

exclude the role of undulated radial-calcified interface in this behaviour, and further

detailed studies are needed to unambiguously ascertain its origin. We empirically term

the region of rapid R255 increase between x = 0.88 and x = 1 the “attachment sub-zone”

to reflect its likely role as a transition from the radial zone to the tidemark, where

collagen fibres are anchored to the subchondral bone.

The fibrocartilage pad of the kangaroo tibia is absent in most mammalian knee

joints, including human, bovine, equine, canine, porcine and murine, i.e. the animals

whose predominant modes of movement are walking or running but not jumping.

Kangaroo TFC differs from tibial hyaline cartilage in chondrocyte density, PG content,

and collagen and elastin architecture [295]. These unique features render TFC more

easily compressible under mechanical load than hyaline cartilage. In studies involving

similar macropods (e.g. the agile wallaby, Macropus agilis [273]) it has been postulated

that the relatively high deformability of TFC is a biomechanical load-processing

mechanism that has evolved in response to high articular stresses involved in hopping;

it maximises articular contact surface and thus minimises peak loads in the regions of

contact between the tibial plateau and femoral condyles. The combination of high

deformability and high mechanical loads means that TFC can be expected to experience

greater relative deformations (both compressive and shear) during the hopping cycle

than the adjacent hyaline cartilages. We hypothesise that the prominent “attachment

sub-zone” at the bottom of TFC may be an evolutionary adaptation that enables TFC to

withstand large deformations inherent in hopping locomotion.

4.4. Summary

Our investigation of the MRI magic-angle effect has identified the characteristic

zonal structure in femoral and tibial cartilages of the red kangaroo (Macropus rufus).

The R2A depth profiles have confirmed the presence of superficial, transitional and


72

radial histological zones in femoral hyaline cartilage, tibial hyaline cartilage and tibial

fibrocartilage. The apparent degree of collagen fibre alignment in each zone and the

relative thickness of different zones were distinctly different in different types of

cartilage. Femoral hyaline cartilage had a well-developed superficial zone consistent

with the need for withstanding shear deformations. In tibial hyaline cartilage, the radial

zone occupied nearly 80% of the total thickness and exhibited the highest apparent

degree of collagen alignment (the largest R2A values) of any zone of any type of

kangaroo knee cartilage. The “attachment sub-zone” at the bottom of the radial zone of

tibial fibrocartilage suggests a collagen-fibre anchoring mechanism that may enhance

the ability to withstand high compressive and shear deformations. MRI studies

combined with biomechanical testing, similar to the existing methodologies [299-304],

would provide a more detailed understanding of the role of these features of cartilage

ECM in the biomechanical function of kangaroo knee.

The zonal structure of kangaroo knee cartilages is broadly similar to that of other

mammals, but at the same time is subtly different and appears to be adapted to large-

amplitude, high-frequency compressive and shearing stresses involved in hopping

locomotion. The unique features of the collagen architecture of kangaroo knee

cartilages may inspire new designs for cartilage tissue engineering. The observed

differences between the R2 depth profiles of fibrocartilage and hyaline cartilage could

be useful for non-invasive identification of the two types of cartilage in vivo, which in

turn can play a crucial role in informing the development of cartilage tissue regenerative

therapies [305, 306].


73

Acknowledgements

The authors would like to thank the staff at the Medical Engineering research

laboratory for facilities for sample preparation and Dr R. Mark Wellard for assistance

with the MRI measurements.

Contributions

All authors were involved in drafting the article or revising it critically for

important intellectual content and all authors approved the final version to be submitted.

Study design and conception: Ali, Thibbotuwawa, Gu and Momot.

Acquisition of samples and data: Ali, Thibbotuwawa and Momot.

Analysis and interpretation of the data: Ali, Momot and Thibbotuwawa.

Dr. Konstantin I. Momot ([email protected]) takes the responsibility for the

integrity of the work as a whole.

Role of the funding source

YTG acknowledges support from the Australian Research Council (grants

DP150100828 and LP150100737). Australian Research Council had no involvement in

the design of the study; collection, analysis and interpretation of the data; writing of the

manuscript; or the decision to submit the manuscript for publication.

Competing Interests

There is no financial or personal relationship of the authors with other people or

organisations that could potentially and inappropriately influence this work and its

conclusions.

Chapter 4: Mammographic Density by T2 NMR

74

Chapter 4: Mammographic Density Assessment by Transverse

Relaxation based NMR

_____________________________________________________________________

4.1 Prelude

This chapter presents an experimental study of transverse relaxation based

assessment of mammographic density (MD) using a single sided and portable nuclear

magnetic resonance (NMR) instrument. Increased MD has been established as an

independent risk factor for breast cancer (BC) after adjustment for age and body mass

index (BMI) [75, 231, 307, 308]. BC is the most commonly diagnosed cancer among

females all over the world. Approximately, 8% of all women aged 40-74 have

extremely dense breasts [309], who are at high risk for developing cancer in their

lifetime. X-ray mammography is the current standard for measuring MD in clinical

practice. However, the presence of ionizing radiation in X-rays imposes limits on

patient age and the frequency of mammogram examination. X-rays also pose health

risk to women with inherited syndromes associated with radio-sensitivity and/or cancer

risk [240, 241]. It addition, mammography suffers from projectional imaging artefact

because of breast compression in mammograms and from mammographic masking in

dense breasts [75, 243, 244]. At the same time, mammography is inefficient for

diagnosing BC in this particular group because of the reduced sensitivity of X-ray

mammograms in dense breast.

The breast tissue with high MD (HMD) contains a significantly greater

proportion of fibroglandular tissue (FGT) and less adipose (fat) tissue than the breast

tissue with low MD (LMD) [73, 238, 310-312]. The water content of breast tissue is

highly correlated with the prevalence of FGT [75]. Consequently, the tissue T2

measured using NMR/MRI is likely to be sensitive to the FGT and fat distribution in

the breast tissue. Although MRI has shown the potentials for spatially resolved

assessment of MD [84-87], the acquisition of a MRI scan is substantially expensive and

therefore inappropriate for breast screening on a regular basis. Portable NMR employs

the same fundamental principles as MRI for measuring T2 relaxation decays while it is


75

designed as a low-cost and low-maintenance unit based on permanent magnets [89-91].

The goal of this study was to establish a transverse relaxation based MD measurement

protocol using portable NMR that may provide a safe and cost-effective alternative to

mammography for screening MD in patients. This experimental work addresses the

second objective of this thesis.

This study is presented in the form of a journal article (doi:

10.1002/mrm.27781) Magnetic Resonance in Medicine [125]. This chapter comprises

the main body of the article, figures and tables as contained in the manuscript. The

supporting information of the journal article is presented in Appendix 1. The references

are, however, merged with the bibliography at the end of this thesis.


76





expertise;






academic unit, and




Transverse Relaxation based assessment of mammographic density and breast tissue

composition by single-sided portable NMR

Tonima S. Ali, Monique C. Tourell, Honor J. Hugo, Chris Pyke, Samuel Yang, Thomas Lloyd,

Erik W. Thompson, Konstantin I. Momot

Published: Magnetic Resonance in Medicine 82 (3) (2019) 1199-1213


Tonima S. Ali

Conducted NMR experiments, performed data analysis and wrote

the manuscript. Date

Monique C. Tourell Conducted NMR experiments.

Honor J. Hugo Designed the study, revised the manuscript

Chris Pyke Performed surgery and prepared samples from excised tissues.

Samuel Yang Performed surgery and prepared samples from excised tissues.

Thomas Lloyd Conducted radiological investigation.

Erik W. Thompson Conceived and designed the study, revised the manuscript.

Konstantin I. Momot Conceived and designed the study, wrote the manuscript and

supervised the study.




Konstantin I. Momot

Name Signature Date

QUT SignatureVerified


77

4.3 Transverse relaxation-based assessment of mammographic density

and breast tissue composition by single-sided portable NMR

Tonima S Ali1, 2

Monique C. Tourell1, 2

Honor J Hugo2, 3, 4

Chris Pyke5

Samuel Yang6

Thomas Lloyd7

Erik W. Thompson2, 3, 4, 8

*Konstantin I. Momot1, 2

1School of Chemistry, Physics and Mechanical Engineering, Queensland University of

Technology (QUT), Brisbane, Australia

2Institute of Health and Biomedical Innovation, Queensland University of Technology

(QUT), Brisbane, Australia

3School of Biomedical Sciences, Faculty of Health, Queensland University of

Technology (QUT), Brisbane, Australia

4 Translational Research Institute, Woolloongabba, Australia

5Department of Surgery, Mater Hospital, University of Queensland, St Lucia, Australia

6Department of Plastic and Reconstructive Surgery, Greenslopes Private Hospital,

Brisbane, Australia

7Division of Radiology, Princess Alexandra Hospital, Woolloongabba, Australia

8University of Melbourne Department of Surgery, St Vincent’s Hospital, Melbourne,

Australia

* Corresponding Author: Dr Konstantin I. Momot

School of Chemistry, Physics and Mechanical

Engineering


GPO Box 2434

Brisbane, QLD 4001

Australia


mailto:[email protected]


78

ABSTRACT

Purpose: Elevated mammographic density (MD) is an independent risk factor for

breast cancer (BC) as well as a source of masking in X-ray mammography. High‐

frequency longitudinal monitoring of MD could also be beneficial in hormonal BC

prevention, where early MD changes herald the treatment’s success. We present a novel

approach to quantification of MD in breast tissue using single‐sided portable NMR. Its

development was motivated by the low cost of portable‐NMR instrumentation, the

suitability for measurements in vivo, and the absence of ionizing radiation.

Methods: Five breast slices were obtained from three patients undergoing prophylactic

mastectomy or breast reduction surgery. Carr-Purcell-Meiboom-Gill (CPMG)

relaxation curves were measured from: (1) regions of high and low MD (HMD and

LMD, respectively) in the full breast slices; (2) the same regions excised from the full

slices; and (3) the excised samples after H2O-D2O replacement. T2 distributions were

reconstructed from the CPMG decays using inverse Laplace Transform.

Results: Two major peaks, identified as fat and water, were consistently observed in

the T2 distributions of HMD regions. LMD T2 distributions were dominated by the fat

peak. The relative areas of the two peaks exhibited statistically significant (P < .005)

differences between HMD and LMD regions, enabling their classification as HMD or

LMD. The relative‐area distributions exhibited no statistically significant differences

between full slices and the excised samples.

Conclusion: T2-based portable-NMR analysis is a novel approach to MD

quantification. The ability to quantify tissue composition, combined with the low cost

of instrumentation, make this approach promising for clinical applications.

Keywords: breast cancer, mammographic density, NMR-MOUSE, nuclear magnetic

resonance, single-sided portable NMR, transverse spin relaxation time constant (T2)


79

1 INTRODUCTION

Mammographic density (MD), also known as breast density, is estimated in clinical

practice from X-ray mammograms and serves as an indicator of breast tissue

composition. High MD (HMD) is associated with a relatively large proportion of

stroma, collagen and epithelial tissue and relatively low adipose tissue content in the

breast. Conversely, low MD (LMD) is associated with a relatively large adipose tissue

content [73, 238, 310-312]. Elevated MD has been established, along with family or

personal history of breast cancer (BC), age and genetic mutations, as a significant

independent risk factor for BC [75, 231, 307, 308]. Women in the highest MD quartile,

after adjustment for age and body mass index, are four to six times more likely to

develop BC over time than the women in the low-MD group [232, 233]. Besides being

a significant risk factor for breast cancer, HMD acts as a masking factor in

mammography, often making mammographic detection of BC in dense breasts difficult

[75, 243, 244].

While mammography remains a universally accepted standard for MD assessment, it

has a number of important limitations. First, it is a 2D technique used to visualise a 3D

anatomical structure; it therefore suffers from projectional imaging artefacts. The

second is its use of ionizing radiation, which limits its suitability for young women and

women with inherited syndromes that are associated with radio-sensitivity and/or

cancer risk [240, 241]. Importantly, it also limits the clinically acceptable screening

frequency (the normal guideline is no more than once every 2 years). There are

scenarios where frequent longitudinal monitoring of MD would be of clinical benefit,

e.g. tamoxifen treatment for BC prevention, where early MD changes are currently the

only known biomarker of the eventual success or failure of the treatment [234, 313].

All these factors have both encouraged and necessitated the development of

nonionizing alternatives for breast screening that may measure MD-analogous

quantities.

We have recently shown that quantitative T1 measurements using single-sided portable-

NMR instrumentation are capable of distinguishing between HMD and LMD regions

in excised breast tissue slices [97]. This suggests that portable NMR could potentially

complement the other nonionizing techniques developed or adapted for the

measurement of breast density-equivalent quantities [245, 246]: ultrasound [314, 315],


80

bioimpedance [316], transillumination [317, 318] and MRI [84, 86, 232, 247, 248].

Magnetic resonance in general, and portable NMR in particular, appear promising for

quantification of MD because of the great signal editing flexibility offered by MR.

Multisequence clinical MRI followed by automatic segmentation has been used for

quantification of fibroglandular breast tissue (FGT) content, with the conclusion that

MRI “provides a reproducible assessment of the proportion of FGT, which correlates

well with mammographic assessment of breast density [based on Breast Imaging

Reporting and Data System (BI-RADS)]” [79]. The MRI-based volumetric breast

density measurements, using semi-automated or fully automated clustering or

segmentation algorithms, have shown good agreement with conventional MD

measurements in multiple studies [84-87, 249-251]. Portable NMR offers the ability to

quantify tissue spin-relaxation and diffusion properties, which have been shown to

provide reliable quantification of MD in conventional breast MRI. Portable NMR also

has the added advantages of low purchasing and running cost and low maintenance,

largely due to the absence of superconducting magnets (which obviates the need for

cryogenic maintenance). Portable-NMR systems, most notably the NMR-MOUSE [89-

91], are commercially available and have been used in a number of biomedical

applications including the testing of silicone breast implants [92] and studies of various

biological tissues, including tendon [94], articular cartilage [95, 319], skin [93, 320]

and trabecular bone [96].

In the present study, we follow up on the T1-based portable-NMR quantification of MD

reported earlier [97] and explore the capabilities of T2-based portable-NMR analysis

for the assessment of MD in human breast tissue. Transverse spin relaxation in

biological tissues is sensitive to the chemical composition and microscopic organisation

of the tissue [38, 64, 76, 97, 124, 154, 299, 301, 321-324]. In 1H NMR of breast tissue,

two major sources of the NMR signal are present: water (the principal signal source in

FGT) and fat (which dominates in adipose tissue). These two chemical components

exhibit significantly different T2 values and can be resolved in the T2 relaxation spectra

obtained from inverse Laplace transforms (ILTs) of CPMG relaxation decays [325]. In

portable NMR, ILT-based T2 analysis has been used in a wide variety of applications,

including skin [93]. Outside portable NMR, MRI-based multi-exponential T2 relaxation

analysis has been applied to study normal breast [326], liver [27] and prostate tissues

[37] as well as pathological conditions in brain [12, 25, 32, 327, 328]. The ILT-based


81

relaxometry has also been applied to assess water and fat distribution in processed food

products [149].

We demonstrate that, in excised breast tissue samples, the relative area fractions (AF)

of fat and water peaks in ILT T2 spectra enable the discrimination between HMD and

LMD regions. To our knowledge, this is the first time ILT T2 spectra have been used

for compositional quantification of soft biological tissues. The key advantage of this

approach is that, unlike T1-based measurements [97], it enables explicit quantification

of the water: fat ratio in breast tissue. We discuss how this approach could be used for

quantification of MD and evaluation of the relative amount of FGT within breast tissue.

2 METHODS

2.1 Tissue selection and preparation

Patients presenting with ductal carcinoma in-situ and/or micro-calcifications on

radiological investigation were excluded from this study. Five breast slices (the same

as those used in our previously reported study [97]) were obtained from three women

who underwent breast reduction surgery (Patient 1) or prophylactic mastectomy

(Patients 2 and 3). Immediately after the surgery, excised tissues were transported on

ice to the pathology suits, and cranio-caudal slices of breast tissue were resected in a

sterile environment [238, 307, 310, 329]. The breast slices were assessed for

abnormalities by a pathologist. Breast slices that were surplus to pathologists’ needs

were used for the present study. For Patient-1 and Patient-3, the slices were transported

for mammography fresh (on ice) immediately after accrual. For Patient-2, the slice was

stored at -80°C long term after accrual and was transported on dry ice for

mammography. Further details can be found in Table 1, ref. [97].

The study was approved by the Peter MacCallum Human Research Ethics Committee

(#08/21), Metro South Hospital and Health Services, Queensland

(HREC/16/QPAH/107), Mater Research (RG-16-028-AM02, MR-2016-32), and

administratively approved by Queensland University of Technology (QUT)

(#1600000261). The study was conducted in accordance with the Australian National

Statement on Ethical Conduct in Human Research (2007).


82

2.2 Slice Mammogram Acquisition and Analysis

Mammography of the breast slices (Mo target / Mo filter; tube voltage 28 kV; exposure

40 mAs) was performed at the Radiology suite at Princess Alexandra Hospital (PAH).

Mammograms for Patient-1, Patient-2, and Patient-3 were acquired from fresh, frozen,

and fresh slices, respectively. In the mammogram of each slice, one HMD region and

one LMD region were identified and outlined by a clinical radiologist (TL). Following

mammography, all slices were kept frozen (–80 oC) and were later transferred to a

freezer (-20oC) at QUT’s Gardens Point campus, where they were kept until portable-

NMR measurements. The use of frozen sample is consistent with the previously

established experimental protocol [97]; Supporting Information Figure S1 illustrates

the absence of significant effects of freezing on the spatial distribution of

mammographic density of the samples.

The JPEG images of the mammograms of the three slices of Patient-1 were read

in MATLAB R2014a (MathWorks, Natick, MA). Rectangular regions of interest

(ROIs), approximately the same size as the portable NMR sensing coil, were identified

in the HMD and the LMD regions of the slices. Greyscale pixel values of the ROIs were

used to construct histograms for further analysis, which was performed using an in-

house MATLAB code.

2.3 Portable NMR Measurements

The breast slices were defrosted prior to the NMR measurements and kept at room

temperature during the measurements. Portable-NMR measurements were performed

using a PM5 NMR-MOUSE® instrument (Magritek, New Zealand). This instrument is

a single-sided NMR scanner that uses an assembly of permanent magnets to create a

horizontal magnetic field of the strength B0 = 0.47 T and a vertical permanent field

gradient G0 = 22.5 T/m. It uses a surface coil for excitation and signal detection. The

instrument allowed the selection of a horizontal sensing slice with an approximate

sensing area of 15 x 15 mm (determined by the dimensions of the surface coil) and 50

µm thickness (determined by the amplitude of the magnetic field gradient, RF field

strength and the acquisition dwell time). The NMR-MOUSE setup and sample

placement have been described in detail in our previous work [97]. T2 relaxation curves


83

were obtained using the CPMG pulse sequence (TE = 120 µs, TR = 10 s, 4000 integrated

echoes and 64 averages, scan time was 11 minutes per scan).

One HMD and one LMD region were identified in each breast slice by visual

comparison of the topography of the physical slice with the slice mammogram, where

HMD and LMD regions had previously been marked (section 2.2). Three sets of T2

relaxation data were acquired from each slice. First, the HMD and LMD regions were

measured within the full slice: the slice was placed such that the required region was

located above the centre of the NMR-MOUSE sensing coil [97]. A depth profile of the

region was acquired in order to check the uniformity of the sample. The CPMG decays

were acquired at the depths of 2 mm and 4 mm for the four samples that were ~ 10 mm

thick (Patient1-Slice1, Patient1-Slice2, Patient1-Slice3 and Patient3-Slice1). A single

CPMG decay was acquired at the 2-mm depth for the thinner sample (Patient2-Slice1,

~4 mm thick).

The second set of CPMG data was obtained from the HMD and LMD regions excised

from the respective full slices. The regions (smaller than the sensing area of the NMR-

MOUSE) were excised using sterile blades in a Physical Containment level 2 (PC2)

laboratory. The CPMG decays were acquired for all excised HMD and LMD regions

using the same protocol as used for the full-slice measurements.

The third set of CPMG data was obtained from the excised regions subjected to H2O –

D2O replacement. The excised tissue samples were soaked in 0.01 M phosphate

buffered saline (PBS) solution, made with 99% D2O for 16 to 18 hours at +4°C, after

which portable-NMR measurements were repeated again using the same protocol. The

full dataset, therefore, comprised 54 CPMG decay curves (full slice, excised, and

excised after H2O – D2O replacement) from each of the 9 different HMD and 9 different

LMD locations (two depths each in four of the slices and a single depth in the fifth slice;

see previous discussion). Data acquisition was completed over three days, with the

samples being alternated between room temperature (when being measured) and +4°C

(between measurements). Two control samples (excised from Patient1-Slice1) were

subjected to the same experimental protocol for three days to check for signs of tissue

degradation (as seen in T2 relaxation measurements).


84

2.4 T2 Relaxation Analysis

The T2 relaxation curves were analysed using one-dimensional ILT [148]. The time-

dependent signal describing a multicomponent T2 relaxation decay can be written as

𝑆(𝑡𝑗) = 𝑔𝑗 = ∑ 𝐴(𝑇𝑖)𝑖 exp (−𝑡𝑗

𝑇𝑖) + ɛ𝑗, (1)

where i = 1 ... m (the number of relaxation-time components); Ti are the respective

relaxation time constants; A(Ti) are the (nonnegative) relative amplitudes of the

relaxation-time component; ɛ is the noise; and j = 1 ... n (the number of sampled

echoes). The amplitudes A(Ti) can in principle be determined by inverting the T2

relaxation curve using a nonnegative least-squares algorithm that minimises the χ2 value

[143].

𝑚𝑖𝑛 {𝜒2 = ∑ (𝑔𝑗𝑛𝑗=1 − ∑ 𝐴(𝑇𝑖)

𝑚𝑖=1 exp (−

𝑡𝑗𝑇𝑖

⁄ ))2} (2)

A robust fit in presence of noise requires a regularization function weighted by a

regularization parameter α [144, 145, 147, 330]. The new minimization function takes

the following form [148]:

𝑚𝑖𝑛 {𝜒2 = ∑ (𝑔𝑗𝑛𝑗=1 − ∑ 𝐴(𝑇𝑖)

𝑚𝑖=1 exp (−

𝑡𝑗𝑇𝑖

⁄ ))2 + 𝛼−1 ∑ (2𝐴(𝑇𝑖) − 𝐴(𝑇𝑖−1) −𝑚𝑖=1

𝐴(𝑇𝑖+1))2} (3)

The values n = 4000 (the number of echoes) and m = 100 (the number of T2 bins) were

used. A code originally designed by Venkataramanan et al. and subsequently modified

was used for solving Eq. [3] [148, 331] on MATLAB platform. The code can be

obtained from Magritek ([email protected]). In order to determine the appropriate

value of the regularization parameter α for each T2 relaxation curve, a wide range of α

(~106 – ~1012) was specified, and χ2 was calculated at 20 values of α covering this range.

Supporting Information Figure S2 illustrates the effect of the regularization parameter

α on ILT spectra. The curve of χ2 vs (see Figure S2A) was then plotted. The “best”

value of α (corresponding to the best trade-off between over-smoothing and an ill-posed

inversion) was selected after visual inspection as the point of the apparent maximum of

the second derivative (the “L-bend”) of this curve. This value was used for the



85

subsequent inversion of the respective CPMG curve. The optimal choice of α ensured

that the resulting T2 distribution was insensitive to noise while reproducing the different

relaxation components contributing to the CPMG decay curve. No obvious L-bend

point was observed for the HMD full-slice measurement Patient2-Slice1-Depth1. This

T2 distribution was not included in the analysis of the results. However, for a complete

understanding of the data used in this study, it was included in Supporting Information.

The resulting T2 distributions were plotted in semilog coordinates (signal amplitude

versus logT2). They typically exhibited distinct peaks, which were interpreted as arising

from either water or fat, as described in Results. Area fraction (AF) and geometric mean

T2 (gmT2), two measures commonly used for assessing such distributions [12, 17, 33,

34, 149-152], were used to characterise the peaks:

𝐴𝐹 =∑ 𝐴(𝑇2)


∑𝐴(𝑇2)(4)

𝑔𝑚𝑇2 = exp(∑ 𝐴(𝑇2) 𝑙𝑜𝑔


𝑇2

∑ 𝐴(𝑇2)𝑇2𝑚𝑎𝑥𝑇2𝑚𝑖𝑛

) (5)

where T2min and T2max are the left and right boundaries of the respective peak. Welch’s

unequal variance t test was used to evaluate the statistical significance of the difference

between the groups of gmT2 and AF values corresponding to HMD and LMD regions;

water and fat peaks; and measurements made from full slices versus excised samples.

3 RESULTS

The photograph and the mammogram of a representative breast tissue slice (Patient1-

Slice2) are shown in Figure 1. The ROIs (dashed rectangles in Figure 1A), were selected

to correspond in size and shape to the sensing area of the RF surface coil of the NMR-

MOUSE. Rectangular HMD and LMD ROIs were chosen in this way for each breast

slice.

The three slice mammograms obtained from Patient-1 slices were analysed to identify

the 8-bit greyscale values associated with the HMD and LMD regions. Figure 2 presents

the histograms obtained from the HMD and LMD regions of the three Patient-1 slice

mammograms, with the mammogram pixel values quantified using the 0-255 range (0

is “black” and 255 “white”). The histograms derived from the HMD regions had higher


86

greyscale values than those from LMD regions. However, the HMD and LMD

histograms exhibited an overlap, which in some cases was significant.

Figure 3A shows representative T2 distributions measured from an excised HMD region

(as shown in Figure 1) before and after H2O-D2O replacement. D2O is “silent” in 1H

NMR, and therefore the T2 values in distribution a (shown in orange) can be interpreted

as those of fat. The T2 distribution b (shown in blue) was obtained from the native (H2O-

containing) tissue; the T2 modes in this distribution can be interpreted as corresponding

to tissue fat (T2 ~ 90 ms) and water (T2 ~ 10 ms). Figure 3B shows the equivalent T2

distributions measured from an excised LMD region (see Figure 1). There, the T2 values

of distribution a (in purple, measured from the D2O-replaced sample) can be interpreted

as the fat component of the LMD sample. The T2 distribution of the native H2O-

containing tissue (distribution b, in green) had two well-separated modes. The dominant

mode (84.99% of the signal, T2 ~ 90 ms) can be interpreted as tissue fat, while the

smaller mode at T2 ~ 10 ms can be interpreted as tissue water. Figure 4 shows the T2

distributions measured from all excised HMD and LMD samples used in this study both

before and after the H2O-D2O replacement. The AF and gmT2 values measured from

these distributions are summarised in Supporting Information Tables S1 and S2.

Figure 1. A, A photograph and B, a mammogram of a representative breast slice (Patient

1-Slice 2) used in this study. B, The HMD and LMD regions specified by the radiologist are

shown as white circles. A, The black dashed squares show the HMD and LMD regions excised

from the full slice. HMD, high mammographic density; LMD, low mammographic density


87

Figure 2. Histograms of the intensities of HMD and LMD regions in slice mammograms

of A, Patient 1-Slice 1; B, Patient 1-Sice 2; and C, Patient 1-Slice 3. The horizontal axis

represents the pixel greyscale values. The vertical axis shows the bin counts, or the abundance,

of the respective greyscale values. HMD, high mammographic density; LMD, low

mammographic density

Figure 5A shows a representative T2 distribution measured from an HMD region of a

full slice (distribution f, shown in light-blue) as well as the T2 distribution measured

from the same HMD region after its excision (distribution e, in brown). The “fat” and

“water” peaks are evident in both T2 distributions. The corresponding peaks in each


88

distribution exhibit approximately equal most-probable T2 values, while the relative

amplitudes and the AF values (see Supporting Information Tables S1 and S3) of these

peaks differed between the two distributions. Figure 5B shows the equivalent two T2

distributions measured from an LMD region of the same slice (distribution f, light-

green, within the full slice; distribution e, magenta, from the excised LMD region).

Both peaks are again evident and exhibit approximately equal most-probable T2 values

in the two distributions (see Supporting Information Tables S2 and S3). Figure 6 shows

the T2 distributions obtained from all HMD and LMD regions within the full slices. The

AF and gmT2 values measured from these distributions are presented in Supporting

Information Table S3.

In order to check the compositional stability of the tissue over the course of the

measurements, T2 distributions were measured from two control samples, CTRL 1 and

CTRL 2, on three consecutive days. In the CTRL 1 sample, the “fat” peak had the most-

probable T2 at 81.10 ms, 88.92 ms, and 81.10 ms; gmT2 at 80.11 ms, 76.70 ms, and

79.27 ms; and AF of 81.80%, 85.62%, and 80.17% in days 1, 2, and 3, respectively.

The “water” peak had the most-probable T2 9.54 ms, 7.92 ms, and 9.54 ms; gmT2 10.18

ms, 7.55 ms and 10.26 ms; and AF 17.87%, 14.37% and 19.35% at the same time points.

(The sum of water and fat AF values was < 100% because the area of the entire T2

distribution was taken as 100%). The corresponding values in the CTRL 2 sample were

“fat” peak, most probable T2 81.10, 88.92, and 81.10 ms; gmT2 = 76.41, 77.61, and

79.71 ms; AF = 93.15%, 91.42%, and 87.28%; “water” peak, most probable T2 = 9.54,

7.92, and 11.50 ms; gmT2 = 10.98, 8.51, and 11.39 ms; AF = 6.83%, 8.54%, and

11.71%. These results indicate that there was no statistically significant change in the

most probable T2, gmT2 or AF values during the measurement cycle that could be

attributed to sample degradation. This result is in agreement with T1 study, where T1

values were found to be consistent throughout the three days of the measurement [97]


89

Figure 3. Representative T2 distributions obtained from A, excised HMD and B, excised

LMD breast tissue samples. The samples shown were excised from Patient 1-Slice 2. Each

panel shows the T2 distribution in the native tissue (labelled “b”) and after H2O-D2O

replacement (labelled “a”). The peak near T2 = 10 ms, which disappears upon H2O-D2O

replacement, was identified as water. The measurements shown were taken at the 4-mm tissue

depth of the respective samples (Depth 2, P1-S2-D2). In these and all subsequent ILT spectra,

the T2 range from 0.1 ms to 1000 ms with logarithmic spacing of bins was used. However, as

no T2 contributions were observed for T2 < 3 ms, all ILT T2 distributions were plotted in the

range from 1 ms to 1000 ms. The boundaries of the T2 peaks were selected individually for each

T2 spectrum, either as the first bin whose value was above the baseline or as the bin closest to

the minimum between the two peaks. As an example, for spectrum “b” in panel A, the peak

boundaries were defined as 4.98 ms to 22.1 ms for water and 24.2 ms to 359 ms for fat. In panel

B, the respective boundaries were defined as 4.13 ms to 13.8 ms and 15.2 ms to 394 ms for

spectrum “b” and 20.1 ms to 327 ms for spectrum “a”. HMD, high mammographic density;

LMD, low mammographic density


90

Figure 4. The T2 distributions obtained from the breast tissue regions excised from the 5

slices used in the study. A, Excised HMD samples before H2O-D2O replacement; B. same

samples after H2O-D2O replacement; C, excised LMD samples before H2O-D2O replacement;

and D, same samples after H2O-D2O replacement. The individual distributions represent

measurements at a specific depth within a given slice: Patient 1-Slice 1-Depth 1 (P1-S1-D1),

Patient 1-Slice 1-Depth 2 (P1-S1-D2), Patient 1-Slice 2-Depth 1 (P1-S2-D1), Patient 1-Slice

2-Depth 2 (P1-S2-D2), Patient 1-Slice 3-Depth 1 (P1-S3-D1), Patient 1-Slice 3-Depth 2 (P1-

S3-D2), Patient 2-Slice 1-Depth 1 (P2-S1-D1), Patient 3-Slice 1-Depth 1 (P3-S1-D1) and

Patient 3-Slice 1-Depth 2 (P3-S1-D2). HMD, high mammographic density; LMD, low

mammographic density


91

Figure 5. The T2 distributions obtained from the full breast slice and from the excised

regions of Patient 1-Slice 2: A, HMD region and B, LMD region. The full-slice measurements

were taken with the respective region positioned above the centre of the NMR-MOUSE sensing

coil. All the measurements shown are from the 2-mm tissue depth (Depth 1, P1-S2-D1). HMD,

high mammographic density; LMD, low mammographic density

As a visual summary of the T2-based analysis, the AF of the T2 peaks were plotted

against their respective gmT2 values for the following groups of measurements: excised

HMD regions, excised LMD regions, HMD regions within the full slice, and LMD

regions within the full slice. The results are shown in Figure 7A and 7B for the excised

and full slice samples, respectively. Figure 7 shows the large difference between the

T2s of the “fat” and “water” peaks, which can be clearly distinguished based on their

gmT2 values. Table 1 shows the results of Welch’s t test for the gmT2 values; these

demonstrate that the gmT2 distributions of water peaks were significantly different from

those of the fat peaks in all samples. There were no statistically significant differences

between the gmT2 distributions of either water or fat peaks between the HMD and LMD

regions in either group of samples. Table 2 shows that the distributions of the AF values

measured from HMD regions were significantly different from those of the LMD


92

regions, both for excised samples and full-slice samples. There were no statistically

significant differences between the AF (Table 2) or gmT2 (Table 1) measurements

between full-slice and excised samples.

Figure 6. The T2 distributions obtained from the full breast slices used in this study. A,

HMD regions within the full breast slices; and (B): LMD regions within the full slices. The

measurements were taken with the respective region positioned above the centre of the NMR-

MOUSE sensing coil. The individual distributions represent the measurements made at a

specific depth within a given slice (see the legend of Figure 4 for the nomenclature). HMD,

high mammographic density; LMD, low mammographic density

4 DISCUSSION

The radiographic appearance of the breast is determined by the ratio of FGT and

adipose tissue: HMD regions are known to contain a larger proportion of FGT than

LMD regions [75, 332-334]. The effective transverse spin-relaxation time constants

(T2eff) of water and fat are determined by the relative amounts of the intracellular and

extracellular water, as well as the association of extracellular water with the

biopolymers of the extracellular matrix [332-334]. The T2eff values measured under

portable-NMR conditions are further dependent upon the diffusion properties of water

and fat [93, 335]:

1

𝑇2𝑒𝑓𝑓=

1

𝑇2+

𝐷 𝛾2𝐺02

12𝑇𝐸2, (6)

where D is the diffusion coefficient of the relevant chemical species and G0 is the

magnetic field gradient strength. The time constant T2eff is therefore a composite

function of the true intrinsic T2 and diffusion properties of the relevant chemical


93

species. In this study we analysed how the ILT-derived distributions of T2eff values

could be used as “signatures” of HMD and LMD regions of breast tissue samples.

Slice mammograms provide a “gold-standard” reference for identification of HMD and

LMD regions within the breast slices. Figure 2 illustrates that, in mammograms, FGT-

rich HMD regions tends to exhibit higher X-ray attenuation coefficient and

consequently higher image intensity than (adipose-rich) LMD regions [75]. Overlaps

between the HMD and LMD distributions suggest that a given ROI may contain both

FGT and adipose tissue. This is consistent with the observed distributions of T2eff, which

demonstrate the coexistence of water and fat in all HMD and LMD regions measured.

The nomenclature “HMD region” or “LMD region” is therefore used here to indicate

the preponderance of a given tissue type in a given ROI, rather than an exclusive

presence of FGT or adipose tissue in that region.

Another feature evident in Figure 2 is that the histograms obtained from LMD regions

were more homogeneous than those of HMD regions. This observation is also in

agreement with the T2eff distributions, which show that the contribution of water to the

NMR signal was very low in all LMD regions (significantly lower than the signal

contribution from fat in the HMD regions).


94

Figure 7. The geometric mean T2 (gmT2) values and the area fractions (AF) of the water

and fat peaks A, measured from excised breast tissue samples and B, the respective regions

within the full slices. The gmT2 values represent the geometric-average T2 of the water and fat,

while the AF values reflect the relative prevalence of the respective chemical species within the

sample. This Figure includes the HMD and LMD regions from all five breast tissue slices

studied. HMD, high mammographic density; LMD, low mammographic density

The H2O-D2O replacement measurements enabled identification of the two principal

peaks in T2 distributions as water (T2 ~ 10 ms) and fat (T2 ~ 80 ms). Figure 4 and 6 and

Table 1 show that there was no significant difference in the T2 values of either water or

fat peaks between HMD and LMD regions. This suggests that the microenvironments

experienced by both water and fat molecules are similar in HMD and LMD regions,

which in turn suggests that the mixing of FGT and adipose tissue occurs on a

macroscopic length scale.


95

TABLE 1: Results of the Welch’s unequal t test between the geometric mean T2 (gmT2) measurements of water-peaks and fat-peaks that were obtained

from the T2 distributions of excised HMD and LMD regions of breast tissue

Full Slice Excised Regions

HMD LMD HMD LMD

Water Fat Water Fat Water Fat Water Fat

Full Slice HMD Water 5.10 x 10-10 0.25 1.89 x 10-8 0.39 1.07 x 10-9 0.03 1.54 x 10-15

Fat 5.10 x 10-10 4.92 x 10-11 0.29 7.95 x 10-11 0.11 5.06 x 10-11 0.77

LMD Water 0.25 4.92 x 10-11 2.66 x 10-9 0.11 1.30 x 10-10 0.02 1.49 x 10-11

Fat 1.89 x 10-8 0.29 2.66 x 10-9 8.05 x 10-9 0.03 7.41 x 10-9 0.17

Excised

Regions

HMD Water 0.39 7.95 x 10-11 0.11 8.05 x 10-9 3.96 x 10-10 0.18 1.93 x 10-17

Fat 1.07 x 10-9 0.11 1.30 x 10-10 0.03 3.96 x 10-10 3.47 x 10-10 0.18

LMD Water 0.03 5.06 x 10-11 0.02 7.41 x 10-9 0.18 3.47 x 10-10 1.50 x 10-17

Fat 1.54 x 10-15 0.77 1.49 x 10-11 0.17 1.93 x 10-17 0.18 1.50 x 10-17

Note: Shaded cells indicate statistically significant difference between the respective sample groups (P < 0.005).

Abbreviations: HMD, high mammographic density; LMD, low mammographic density.

TABLE 2: Results of the Welch’s unequal t test between the area fraction (AF) measurements of water peaks and fat peaks, which were obtained from

the T2 distributions of excised HMD and LMD regions of breast tissue

Full Slice Excised Regions

HMD LMD HMD LMD

Water Fat Water Fat Water Fat Water Fat

Full Slice HMD Water 9.38 x 10-5 3.33 x 10-5 1.02 x 10-8 0.59 2.4 x 10-3 6.25 x 10-5 1.23 x 10-7

Fat 9.38 x 10-5 1.33 x 10-7 8.45 x 10-6 2.8 x 10-3 0.61 2.02 x 10-7 1.12 x 10-4

LMD Water 3.33 x 10-5 1.33 x 10-7 4.22 x 10-10 2.61 x 10-4 3.69 x 10-6 0.25 1.24 x 10-8

Fat 1.02 x 10-8 8.45 x 10-6 4.22 x 10-10 1.82 x 10-6 7.54 x 10-5 1.03 x 19-14 0.01

Excised

Regions

HMD Water 0.59 2.8 x 10-3 2.61 x 10-4 1.82 x 10-6 0.02 5.58 x 10-4 9.72 x 10-6

Fat 2.4 x 10-3 0.61 3.69 x 10-6 7.54 x 10-5 0.02 7.36 x 10-6 5.40 x 10-4

LMD Water 6.25 x 10-5 2.02 x 10-7 0.25 1.03 x 19-14 5.58 x 10-4 7.36 x 10-6 8.58 x 10-17

Fat 1.23 x 10-7 1.12 x 10-4 1.24 x 10-8 0.01 9.72 x 10-6 5.40 x 10-4 8.58 x 10-17

Note: Shaded cells indicate statistically significant difference between the respective sample groups (P < 0.005).

Abbreviations: HMD, high mammographic density; LMD, low mammographic density.


96

Fat was identified as the dominant tissue constituent in the LMD regions, with the fat

peak consistently having the area fraction >75% in the LMD T2 distributions (Figures

3B, 4C, 5B, and 6B; Supporting Information Tables S2 and S3). Fat was also a major

tissue constituent in the HMD regions. The relative amplitudes of water peaks were

higher in HMD than in LMD regions but exhibited significant variability (HMD water

AF between 22.42% and 65.71%; see Figures 3A, 4A, 5A, and 6A; Supporting

Information Tables S1 and S2). Nevertheless, the distributions of the AF values of water

and fat peaks were significantly different between the HMD and LMD regions both in

excised and full-slice samples (Table 2). We therefore conclude that the relative

amounts of tissue fat and water measured from T2 distributions can be used to

distinguish between HMD and LMD regions.

The ultimate aim of this research is to adapt the portable-NMR methodology for

characterisation of MD in the full breast in vivo. There, the presence of intertwined

FGT and adipose tissue domains can potentially lead to the partial-volume effect and

affect the T2 distributions measured. In order to assess the significance of the partial-

volume effect, we have measured T2 distributions from the same ROIs within the full

slice and after the ROIs were excised (Figures 4, 6, and 7). The T2 distributions

measured from full-slice HMD regions exhibited water and fat peaks of comparable

amplitudes (Figure 6A). The T2 distributions measured from full-slice LMD regions

were dominated by fat peaks, with minor water peaks (Figure 6B). A comparison of T2

distributions acquired from the same tissue regions before and after excision can be

seen in Figure 5; this figure demonstrates that excision can affect the apparent fat: water

ratio measured from T2 distributions. However, application of the Welch’s t test to the

respective distributions shows that, both for HMD and LMD regions, there was no

statistically significant difference between the AF values measured from the full slices

and from the excised samples (Table 2).

Figure 7 presents a visual summary of the ability of T2-based portable-NMR analysis

to discriminate between High-MD and Low-MD regions. This figure demonstrates that

water and fat peaks were readily distinguishable on the basis of their gmT2 values. The

gmT2 values of both water and fat peaks were similar between all groups of samples

(excised and full-slice, HMD and LMD). The HMD regions could be reliably

discriminated from LMD on the basis of their relative fat and water content, which is


97

presented in Figure 7 as the respective AF values. The HMD water peaks had

significantly higher AF values than LMD water peaks, in both excised and full-slice

samples. It can also be seen that the water/fat ratio can vary substantially from one

HMD region to another, which is consistent with the wide variation of MD patterns

observed in patients.[237, 243, 244] To re-cap the key findings of this study: 1) the

spin-relaxation properties of both fat and water are equivalent between HMD and LMD

regions, suggesting that the physical microenvironments of both FGT and adipose

tissue are identical between HMD and LMD; 2) the AF values of fat and water peaks

are reliable markers for distinguishing between HMD and LMD regions of breast tissue;

and 3) the distributions of fat and water AF values were statistically equivalent between

full-slice and excised regions, which augurs well for application of the present MD

analysis in vivo.

We hypothesise that the approach illustrated by Figure 7 can enable classification of

breast tissue samples with unknown MD into HMD and LMD groups. To provide

comprehensive coverage, we propose that Figure 7 should be extended to include a

large number of HMD and LMD regions acquired from patients across the full range

of BI-RADS scores (from 1 to 4). A “library” of T2 characteristics for each BI-RADS

category may enable a more refined and targeted HMD/LMD classification.

Our previous work has shown that T1-based portable-NMR analysis enables

discrimination between HMD and LMD breast tissue [97]. The results of the present

study demonstrate T1-based and T2-based portable-NMR analyses are potentially

complementary MD assessment tools. In particular, the T2-based analysis presented

here provides explicit information about tissue water and fat content, which may be

beneficial for understanding the physiological basis of MD.

4.1 Limitations and future work

The ILT, which was used to reconstruct the T2 distributions of breast tissue samples, is

well-known to be an ill-posed numerical problem that requires regularisation of noisy

data [145, 147, 148, 330]. The regularisation parameter (α) in this work was selected

visually, based on the identification of the L-bend in the χ2 vs curve (see Methods).

The L-bend was not always unambiguously identifiable, which leaves the possibility of

the reconstructed T2 distributions being either over-smoothed or undersmoothed.


98

Examination of the most-probable T2 and gmT2 values given in Supporting Information

Tables S1 to S3 suggests that there could have been some measurements where the

performance of ILT was sub-optimal. For example, the T2 values of the fat peak in the

excised HMD sample P3-S1 differ significantly between native and D2O-replaced

tissue. This is contrary to the expectation that the fat chemical environment should be

unaffected by H2O-D2O replacement, and such large differences not be observed for

the majority of the samples. We hypothesise that the origin of these differences lies in

the performance of the ILT procedure. Further investigation of the performance of ILT

in breast tissue is warranted.

The sample size used (five breast slices from three different patients) is another

potential limitation of the present study. This limitation is alleviated by the fact that the

ROIs were measured in both the full slices and excised samples, and in most samples

the measurements were performed at two different depths. This provided a total of 18

HMD and 18 LMD measurements, which increased the robustness of the statistics. The

p values reported in Tables 1 and 2, as well as the very good separation of HMD and

LMD points in Figure 7, suggest a high degree of confidence that the present analysis

reliably distinguishes between HMD and LMD tissue. This is consistent with our earlier

T1‐based study [97], which used exactly the same set of physical samples. Nevertheless,

further studies with a larger sample size would be beneficial. Furthermore, studies of

the temperature dependence of breast tissue T2s under portable‐NMR conditions would

benefit the understanding of how the present results might transfer to measurements

performed in vivo at physiological temperature.

Unlike MRI, portable NMR is a volume‐selective spectroscopic rather than a true

imaging technique. This imposes restrictions on how much of the breast volume could

be covered in a single measurement in vivo. Approaches to addressing this issue were

discussed in our previous work [97]. Penetration depth of portable‐NMR sensors is

another potential limitation in vivo. This issue can be mitigated by the selection of

instrumentation models: e.g., the commercially available NMR‐MOUSE model PM25

offers the penetration depth of 25 mm (as opposed to 5 mm by the PM5 model used

here); we hypothesize that the former should be sufficient for MD sensing in the

majority of clinical scenarios. Furthermore, the instrumentation used in the present

study offered a limited thickness of the sensing slice (50 μm). Although all the samples


99

displayed good agreement between measurements taken at different depths, it is

conceivable that in some situations measurements could be sensitive to the precise

positioning of the sensor. Alternative designs of portable‐NMR instrumentation may be

able to alleviate both these issues by enabling a larger sensing volume [95, 336-340].

Finally, transferability of these results to measurements in vivo could be affected by

motional artefacts resulting from patient movement and blood flow. Development of

portable NMR‐specific motion‐compensated acquisition schemes [341] will be able to

address this issue and provide robust acquisition approaches suitable for clinical

measurements.

5 CONCLUSIONS

Portable-NMR T2-based analysis can unambiguously identify HMD and LMD regions

in slices of human breast tissue. Importantly, it also provides information about the

relative quantities of fat and water within the respective regions, which represents a

unique and novel way of assessing breast tissue composition. In both excised and full-

slice samples, HMD regions were found to contain higher proportions of water than

LMD regions. This is consistent with the relatively high FGT and low adipose tissue

content in HMD tissue. Our analysis is in agreement with the identification of HMD

and LMD breast tissue regions based on conventional slice X-ray mammograms, as

well as the T1-based portable-NMR analysis reported earlier. T2-based portable-NMR

analysis has the potential as an informative, cost-effective and safe alternative suitable

for high-frequency monitoring of MD. We envisage that it will have clinical utility in

breast density screening, as well as predicting the efficacy of hormonal treatments for

breast cancer prevention.


100

ACKNOWLEDGEMENTS

We thank Dr Andrew Coy and Dr Robin Dykstra (Magritek Ltd) and A/Prof Petrik

Galvosas (Victoria University Wellington, New Zealand) for the loan of PM5NMR-

MOUSE and invaluable discussions, and Dr R. Mark Wellard (QUT) for useful

discussions concerning experimental design. The authors thank the women who gave

permission for their breast tissue to be used for this study, and Ms Gillian Jagger (PAH)

and Ms Claire Davies (Mater Hospital) for tissue accrual coordination.

FUNDING INFORMATION

Funding from Princess Alexandra Research Foundation (ALH Breast Cancer Project

Grant and Translational Research Innovation Award) and Translational Research

Institute (SPORE Grant) is gratefully acknowledged. The Translational Research

Institute is supported by a grant from the Australian Government.

Chapter 5: Developmental pathway of PTOA

101

Chapter 5: Detection of the Developmental Pathway of

Osteoarthritis by Transverse Relaxation based MRI

____________________________________________________________________________________________

5.1 Prelude

Osteoarthritis (OA) is the most common joint disease world-wide and a leading

cause of pain and disability. It involves degradation, degeneration and inflammation in

multiple tissues of the affected joint. It is still non-curable and its prevalence rate has

been consistently rising with the increase in aged population. There has been a 38% rise

in the rate of total knee replacements for OA in Australia from 2005-06 to 2015-16

[192]. OA cases are commonly diagnosed after the patients present cases with joint pain

and discomfort, which take place at the advanced stage of OA followed by structural

damage and functional impairment. Therefore, the manifestation of early OA as well as

the histopathological alterations that define the developmental pathway of OA remain

elusive. Literature shows that MRI has been effective in detecting OA and in identifying

the OA induced changes in the knee joint. However, the previous MRI studies were

mostly concentrated on individual features of OA, such as, AC degradation, BML,

menisci or ligament injury. The interrelations of such changes have not yet been

established. It is therefore important to develop a holistic understanding of OA

progression, which is necessary for optimal outcome in OA management by appropriate

treatment planning.

In OA, the complex morphological and physiological changes may take several

decades to develop and are usually influenced by multiple genetic and environmental

factors. Various types of OA are observed in human, which are initiated by several

types of injuries and/or incidents. It is therefore not possible to establish a standard OA

model in human. In addition, human OA model is not appropriate to study OA

development due to the late diagnosis in patients and the large variability in human

volunteer data due to age, genetics, physical structure, life style, and other factors.

However, it is possible to develop an animal model of OA, which can represent a

specific type of human OA in a controlled environment while minimizing the internal


102

variability within the experimental animals. Recent research work at QUT has shown

substantial progression in establishing animal model of post-traumatic OA (PTOA) in

rats that resembles human PTOA with significant similarity [342-344]. Longitudinal

examination of the rat joints with PTOA, from the onset of the disease to advanced

PTOA, may give insights on how various tissues are affected during the development

of PTOA.

MRI allows non-invasive and detailed analysis of multiple tissue structures of a

joint in a three-dimensional perspective. MRI can focus on different tissues of the knee

joint by manipulating image contrast. The use of small animals like rats permits the use

of µMRI for examination and analysis of OA. In comparison to clinical MRI, imaging

by µMRI provides higher resolution and improved SNR, which is suitable for

monitoring the morphological as well as the physiological changes [2, 345] in tissue

like AC [11, 155-157]. This research component aimed at establishing a MRI only

protocol for the assessment of knee joint PTOA that will be capable of identifying and

delineating the alterations in knee joint tissues during the development of PTOA. The

ultimate goal of this research was to gain a comprehensive understanding of the

pathogenesis cascade leading from the initial knee injury towards advanced PTOA.

A rat PTOA model was chosen for investigating the progression of PTOA for

this purpose. PTOA was induced in rat knee joints by complete removal of medial

meniscus in the right hind knee joints. Excised whole knee joints were examined by

µMRI at multiple time points that ranged from disease onset to advanced PTOA. In

order to evaluate the efficacy of different quantitative MRI methods for assessing knee

joint tissues, three contralateral knee joints were imaged using MSMEVTR (Multi-

Slice-Multi-Echo-Variable TR), MSME (Multi-Slice-Multi-Echo) and MGE (Multiple-

Gradient-Echo) sequences in the sagittal, coronal and axial orientations. Using custom

designed MATLAB codes, 2D parametric maps (quantitative T1 map, quantitative T2

map and quantitative T2* map) were computed for all MRI slices. Appendix 2 presents

the details of the image acquisition, image processing and the preliminary results

obtained from T1, T2 and T2* MRI.

It was observed that the voxel intensities of T2* weighted images were adequate

for identifying cartilage and measuring the thickness of cartilage. However, the voxel

measurements of T2* were not sensitive to the integrity of AC. On the other hand, the


103

acquisition of sufficient T1 weighted echoes for quantitative T1 analysis, with good

resolution (97 x 97 µm pixel), required several hours (> 5 hours), which resulted in

degradation of the tissues of the excised rat joints. Conversely, the acquisition of

adequate number of T2 weighted echoes for quantitative analysis, with better resolution

(78 x 78 µm pixel), could be completed approximately in 1 hour. However, at this

resolution, a minimum slice thickness of 500 µm was necessary in order to maintain

the SNR above 9:1. This choice of slice thickness may have introduced partial voluming

effect in the T2 weighted echoes. The T2 weighted echoes allowed measurement of

cartilage thickness while the quantitative T2 map was informative. The voxel

measurements of T2 were sensitive to the integrity of cartilage and to the water content

of the tissues of knee joint. Therefore, considering the issues mentioned above, only T2

weighted images were acquired for all control joints, all joints that were subjected to

meniscectomy and the remaining contralateral joints. The transverse relaxation based

MR images were then analysed to identify the developmental pathway of PTOA.

The next sections of this chapter present an experimental study of the PTOA

model in rats that was examined by transverse relaxation based µMRI to identify the

pathophysiological pathway of OA progression. This study is significant because it has

developed a MRI only protocol for whole knee joint evaluation that can be used to

assess whole joint OA and other diseases of the knee joint. In addition, the information

obtained from this study will enhance the knowledge of OA induced changes in joint

tissues and thereby may benefit the development of preventative measures for OA

management. This study is presented in the form of a journal article

(doi:10.1038/s41598-018-25186-1) published in the Scientific Reports [346]. The

presentation format is that of the published paper, including all the text, figures and

tables contained in the original article. The references are merged with the bibliography

at the end of this thesis.


104





expertise;






academic unit, and




Progression of Post-Traumatic Osteoarthritis in rat meniscectomy models:

Comprehensive monitoring using MRI

Tonima S. Ali, Indira Prasadam, Yin Xiao, Konstantin I. Momot

Published: Scientific Reports 8 (2018) 6861


Tonima S. Ali

Conducted animal study, designed and conducted MRI

experiments, performed data analysis and wrote the manuscript. Date

Indira Prasadam Conducted animal study, prepared samples, performed histology

measurements and co-wrote manuscript.

Yin Xiao Conceived and designed the study, co-wrote manuscript.

Konstantin I. Momot Conceived and designed the study, wrote the manuscript and

supervised the study.




Konstantin I. Momot

Name Signature Date



105

5.3 Progression of Post-Traumatic Osteoarthritis in rat meniscectomy

models: Comprehensive monitoring using MRI

Tonima S. Ali1,2, Indira Prasadam1,2, Yin Xiao1,2 & Konstantin I. Momot1,2

1 Queensland University of Technology (QUT), Brisbane, Queensland (QLD),

Australia

2 Institute of Health and Biomedical Innovation, Kelvin Grove, QLD 4059, Australia

Correspondence and requests for materials should be addressed to:

Dr. Konstantin I. Momot



GPO Box 2434, QLD 4001, Brisbane, Australia

Phone: +61-7-3138-1173

Fax: +61-7-3138-9079




106

Abstract

Knee injury often triggers post-traumatic osteoarthritis (PTOA) that affects articular

cartilage (AC), subchondral bone, meniscus and the synovial membrane. The available

treatments for PTOA are largely ineffective due to late diagnosis past the “treatment

window”. This study aimed to develop a detailed understanding of the time line of the

progression of PTOA in murine models through longitudinal observation of the

femorotibial joint from the onset of the disease to the advanced stage. Quantitative

magnetic resonance microimaging (µMRI) and histology were used to evaluate PTOA-

associated changes in the knee joints of rats subjected to knee meniscectomy.

Systematic longitudinal changes in the articular cartilage thickness, cartilage T2 and the

T2 of epiphysis within medial condyles of the tibia were all found to be associated with

the development of PTOA in the animals. The following pathogenesis cascade was

found to precede advanced PTOA: meniscal injury → AC swelling → subchondral

bone remodelling → proteoglycan depletion → free water influx → cartilage erosion.

Importantly, the imaging protocol used was entirely MRI-based. This protocol is

potentially suitable for whole-knee longitudinal, non-invasive assessment of the

development of OA. The results of this work will inform the improvement of the

imaging methods for early diagnosis of PTOA.


107

Introduction

A common consequence of joint injury is post-traumatic osteoarthritis (PTOA), which

accounts for 12% of all cases of osteoarthritis (OA) [347]. Trauma sustained by joint

tissues, particularly tears of the meniscus or anterior cruciate ligament (ACL) can result

in injuries to articular cartilage (AC) and lead to the development of PTOA within a

10-to-15 years’ time window. Characterised primarily by gradual degradation of

articular cartilage (AC) [46], the pathogenesis of PTOA also includes bone remodelling

[98, 207, 208, 221, 348], meniscal modification [224] and synovial inflammation [226,

349]. However, PTOA is often detected in advanced stage after it becomes

radiographically apparent by the altered alignment of the major bones caused by severe

cartilage damage. The cartilage is translucent to the X-ray radiography used in clinical

practice and therefore is not capable of detecting the cartilage degradation at the onset

of PTOA. The discordance between radiological and clinical OA findings [193] also

highlights major limitations of the conventional radiography. With no cure available

until now, preventative measures or clinical intervention within the ‘treatment window’

of early PTOA may provide the optimal clinical outcome for disease management

[104].

PTOA cases are commonly reported by patients as a result of pain in the joints

followed by structural damage and functional impairment at the advanced stage [350-

352], which makes it difficult to investigate the early PTOA in human patients. With

an incomplete understanding of early PTOA, the sequence of events leading towards

symptomatic PTOA remains unclear as well. In the present study, we investigated the

progression of PTOA in rat knee joints that underwent meniscectomy (MSX), a

standard protocol for PTOA initiation known to replicate human PTOA with a

significant degree of similarity [342, 344]. The use of this model has also eliminated

the variabilities due to age, weight, genetics, and environmental conditions that may

result in a significant variability of the clinical manifestation of the disease. The whole

knee joints were monitored weekly over an eight-week post-injury time window in

order to capture the gradual developmental changes from very early PTOA preceding

to the severe stage.

Magnetic Resonance Imaging (MRI), with its recent advancements, is now

extremely sensitive to the changes in cartilage [7, 110, 111, 353], subchondral bone


108

marrow [98] and synovial membrane [354, 355]. The transverse spin relaxation time

constant (T2) of MRI is sensitive to the state of the water in biological tissues and also

to the 3D architecture of the collagen scaffold within the cartilage extracellular matrix

(ECM) [1, 64, 299, 301]. Consequently, T2 mapping allows an indirect assessment of

the integrity and the microscopic organisation of the ECM of cartilage [3, 9, 77, 107,

286, 356], which generally correlate with cartilage damage observed in model PTOA

[106, 357]. The resolution and signal to noise ratio (SNR) of MRI can be enhanced

further by using stronger magnetic fields in the micro-MRI (µMRI) system. With the

advantage of 3D imaging capability of µMRI, we have examined all tissues of the whole

knee joints of our rat PTOA model using µMRI and quantitative T2 relaxation mapping.

In our analysis, we have further emphasised on the tibial hyaline cartilage and tibial

tissues, which remain less studied by MRI mostly due to their irregular shapes and the

difficulty in isolating them from adjacent structures. The µMRI results were compared

against histology assays in order to confirm that the joints that underwent MSX reached

severe PTOA within the 8-week observational time period.

To date, numerous studies have investigated the tissues of the knee joint both in

normal and in PTOA-affected states [31, 46, 98, 207, 208, 221, 224, 226, 342, 343, 348,

358, 359]. Nevertheless, the tissue alteration pathway leading towards symptomatic

PTOA has not yet been identified due to late diagnosis and analytical limitations. The

objectives of the present study were (1) to enhance the analytical capabilities of

quantitative MRI for early detection of PTOA and (2) to identify the sequential changes

in tibial tissues leading from the initial knee injury towards advanced PTOA.

Results

Figure 1 shows the location and the orientation of the three coronal MRI slices acquired

from a control (CTRL) joint at week 1. All of the knee joints that were investigated in

this study were imaged with the same sample position and slice orientation. The femoral

and tibial AC were in direct contact in the central (second) slice, which also contained

the largest cross-section of ACL and posterior cruciate ligament (PCL). In this slice

location, osteophyte-like growths were observed at the medial tibial condyles of the

joints that underwent meniscectomy (MSX joints) at every weekly time point between

week 4 and week 8. In contrast, the lateral condyles did not exhibit osteophytes-like


109

growths or any significant changes in the AC or subchondral bone over the study period

(according to Mann-Kendall trend test).

Figure 1. The MRI scan locations are shown in an axial slice of control knee joint. The position

of coronal slice inside MRI gantry is shown in inset. Here 1, 2 and 3 refer to the anterior, central

and posterior slices, respectively. These slice orientations were maintained for all scans of

CTRL, MSX and CLAT joints. The femoral and tibial AC, the menisci, cortical and trabecular

bone of the epiphysis, ligaments and fat tissues were clearly visible in the T2-weighted coronal

slices acquired maintaining this protocol. The schematic outline of the knee in the inset is

reproduced from https://en.wikipedia.org/wiki/Knee#/media/File:Knee_skeleton_lateral_ante-

-rior_views.svg in accordance with the terms of the CC BY 2.5 license.

..........................................................................................................................................

Histological Analysis. Three histology slices of the AC are shown in Fig. 2.

These were sectioned from the medial tibial compartment of a CTRL joint at week 1, a

MSX joint at week 4, and a MSX joint at week 8. Cartilage surface roughness,

fibrillation, small osteophytes and areas with peripheral fibrous tissue proliferation

were observed in the week 4 and week 8 MSX samples. The proteoglycan content was

lower in the week 8 MSX sample than in the week 4 MSX sample, which in turn was

lower than in the CTRL sample. The same pattern was observed for the AC thickness

of the three samples (Fig. 2B). The Mankin Scores [360] of these samples (Fig. 2C)

validate the presence of PTOA in the MSX joints and confirm that the disease had

advanced in severity from week 4 to week 8. Mankin score takes account of PG

depletion and cell count in cartilage and is considered to be the standard procedure for

evaluating Osteoarthritis [361].

Thickness and T2 of Articular Cartilage: Temporal Evolution and Mutual

Correlation. With partial volume correction, as shown in Fig. 3 for a MSX

joint, the thickness of AC in medial tibial condyle varied between 2 and 4 pixels. Fig.

4 shows the temporal evolution of the AC thickness and the AC T2 for the medial tibial


110

condyle of the MSX joints, both measured on the central coronal slice. The cartilage

thickness showed a tendency to increase for the first 7 weeks with a prominent

increment at week 5 and a substantial drop after week 7. A significant monotonic trend

(Mann-Kendall trend test, p < 0.02) was observed for the week 1 – week 7 time period.

No net trend was observed when week-8 data was included. The AC T2 also showed a

tendency to increase for the first 7 weeks following the surgery with a strong monotonic

Figure 2. Cartilage sections of medial condyles of CTRL and MSX joints (A) stained with

safranin-O fast green, which provided colour discrimination between bone and cartilage. Here,

the cartilage matrix proteoglycan is stained red, cell nuclei black, cytoplasm grey green, and

the underlying bone green [362]. Week 1 (CTRL) showed abundant proteoglycan, week 4

(MSX) showed proteoglycan depletion while week 8 (MSX) showed major proteoglycan loss.

Gradual thinning of cartilage was observed at week 4 and week 8 as shown in (B).The Mankin

scores of these slices are plotted in (C).

..........................................................................................................................................


111

trend (Mann-Kendall trend test, p < 0.02). The T2 increased rapidly between week 6

and week 7 followed by a quick drop between week 7 and week 8. During the 8–week-

long observation period after meniscectomy, both the highest AC thickness and the

longest AC T2 were observed at week 7. Additionally, the thickness of AC and its T2

were found to have strong correlation (Spearman’s rank order correlation, p < 0.05)

throughout these eight weeks.

Figure 3. The cartilage thickness measurement procedure shown in a T2 weighted MR image

of a MSX joint at TE = 12 ms (A). The straight line bordering AC is shown in yellow and

denoted by a. The perpendicular line drawn from femur to tibia, b, is shown in blue in the inset,

the nearest voxels of line b are shown in red. The corresponding T2 profile in (B) represents

femoral cortical bone in pixel 1-4, cartilage in pixel 5, partial volume of cartilage in pixel 6 – 8

and tibial cortical bone in pixel 8 – 9. The partial volume effect observed in pixels 6 – 8 was

corrected by using Eqs (2) and (3). Cartilage thickness was computed by multiplying the total

number of voxels representing cartilage with voxel resolution (78 µm). All of the perpendicular

lines b and corresponding T2 profiles are shown in (C). The partial volumes of each profile was

corrected as above and a thickness was computed. The mean cartilage thickness was computed

by averaging the thicknesses of these intensity profiles.

..........................................................................................................................................

Temporal Evolution of the Epiphyseal T2. The average T2 of medial

epiphysis, which included both the cortical/subchondral bone and the trabecular bone

within the epiphysis, exhibited gradual changes over the 8-week observation period in

both MSX and contralateral (CLAT) joints. Figure 5A shows the evolution of

epiphyseal T2 in the medial tibial condyle (the central coronal slice) for the CTRL, MSX


112

and CLAT joints. The CTRL joints showed no significant changes between week 1 and

week 8, the epiphyseal T2 remained over 25 ms at both time points. For the MSX joints,

the epiphyseal T2 continually decreased for seven weeks after the surgery and reached

9.2 ms at week 7 (See Table 1). This was followed by a slight increase at the week 8

time point. This temporal evolution of epiphyseal T2 was associated with a very strong

monotonic trend (Mann-Kendall trend test, p < 0.01) for the week 1 – week 8 time

period. By visual observation, thinning of the subchondral bone was identified in the

first two weeks following surgery, which was followed by gradual thickening of the

subchondral bone from week 3 to week 8.

Figure 4. Cartilage thickness and cartilage T2 evolution of MSX joints over the eight week

observation period post meniscectomy. The CTRL data of week 1 and week 8 are also presented

here. Cartilage T2 exhibited little change between week 1 and week 3, as well as between week

4 and week 6. The data represent cartilage from the medial condyle of central coronal slice.

Data plotted as mean ± SE.

..........................................................................................................................................

In the CLAT joints, the epiphyseal T2 was observed to continually decrease for

the eight weeks after surgery. The epiphyseal T2 was 11.2 ms by week 8. A very strong

monotonic trend was identified for the epiphyseal T2 of CLAT joints (Mann-Kendall

trend test, p < 0.01) in the week 1–week 8 time period. Visual observation confirmed

initial thinning of subchondral bone (week 1 to week 3) that was followed by gradual

thickening (week 4 to week 8) within the study period. The epiphyseal T2 in the MSX

and CLAT joints were found to be statistically correlated with each other (Spearman’s

rank order correlation analysis, p < 0.01).


113

Figure 5B illustrates the spatial variations in the epiphyseal T2 of medial tibia

for the three coronal slice locations of MSX joints. The epiphyseal T2 exhibited similar

patterns in progression in all three slices with a gradual decrease for eight weeks. A

slight increase at week 6 time point was also observed in all three slices. The epiphyseal

T2 of the anterior slice and the central slice showed a slight increase at week 8, reaching

17.9 ms and 11.3 ms, respectively, by week 8. However, in the posterior slice, the

epiphyseal T2 continually decreased for eight weeks and reached 8.4 ms at week 8. The

epiphyseal T2 in the anterior, central and posterior slices were found to be statistically

correlated according to Spearman’s rank order correlation analysis (anterior and central

slice: p < 0.01, posterior and central slice: p < 0.01, anterior and posterior slice: p <

0.05). Initial thinning and gradual thickening of subchondral bone was identified in all

Figure 5. Mean T2 of medial epiphysis of CTRL, MSX and CLAT joints over the eight week

observation period post meniscectomy (A). The data represent epiphyseal T2 measured from

central coronal slice location. The epiphyseal T2 of medial condyles of anterior (slice 1), central

(slice 2) and posterior slice (slice 3) locations for the MSX joints are shown in (B). Data plotted

as mean ± SE.

..........................................................................................................................................


114

three slice locations by visual observation.

Whole-Joint Evaluation of the Longitudinal Changes. AC thickness, AC T2 and

epiphyseal T2 were measured throughout the observation period, from week 1 to week

8 for CTRL, MSX and CLAT joints. These quantities are plotted in chronological

orders in Fig. 6 for the medial condyles of CLAT joints. Figure 7 exhibits the same

quantities measured from the lateral condyles of the MSX joints. The CTRL data of

each quantity was also plotted in both Figs 6 and 7 in order to demonstrate the changes

only due to the joint maturation process from week 1 to week 8.

Table 1 reports the thickness of AC, its T2 and the T2 of epiphysis for the CTRL, MSX

and CLAT joints for each observation week. Table 2 presents the intragroup changes

in AC thickness, its T2 and the T2 of epiphysis of the medial tibial condyle from week

1 to week 7 and from week 1 to week 8 for CTRL, MSX and CLAT joints.

Cartilage Thickness (µm)

Week 1 Week 2 Week 3 Week 4 Week 5 Week 6 Week 7 Week 8

CTRL 141 ± 10 - - - - - - 144 ± 1

MSX 182 ± 0 184 ± 10 194 ± 7 185 ± 6 214 ± 9 208 ± 23 219 ± 8 161 ± 31

CLAT 179 ± 3 161 ± 12 129 ± 7 - 128 ± 6 130 ± 15 139 ± 9 162 ± 16

Cartilage T2 (ms)

CTRL 21.7 ± 2.5 - - - - - - 21.7 ± 1.4

MSX 24.4 ± 3.2 24.1 ± 4.1 24.9 ± 2.7 31.8 ± 1.7 31.4 ± 2.6 33.2 ± 2.2 59.3 ± 4.6 28.4 ± 5.7

CLAT 22.5 ± 1.5 21.2 ± 2.2 18.2 ± 1.5 - 22.3 ± 2.8 19.0 ± 1.1 21.8 ± 1.9 17.6 ± 4.4

T2 of Epiphysis (ms)

CTRL 27.2 ± 1.9 - - - - - - 25.9 ± 3.3

MSX 29.2 ± 7.6 26.7± 10.8 17.4 ± 2.1 16.5 ± 1.2 12.5 ± 3.2 13.4 ± 2.1 9.2 ± 1.0 11.3 ± 0.7

CLAT 27.5 ± 3.4 25.2 ± 4.1 22.5 ± 0.1 - 15.6 ± 2.7 16.9 ± 0.9 13.9 ± 1.5 11.2 ± 1.8

Table 1. The thickness of AC, its T2 and the T2 of epiphysis of the medial tibial condyle for the

CTRL, MSX and CLAT groups for each observation week. All of these physical quantities are

measured from the T2-weighted MR images and T2 maps of the central coronal slice location

shown in Fig. 1. Data is presented as mean ± standard error.

..........................................................................................................................................


115

Figure 6. Changes in the tissues of the medial condyles of CLAT joints, in comparison to

controls, during the eight week observation period post meniscectomy. Cartilage thickness (A),

cartilage T2 (B) and T2 of epiphysis (C) of CTRL and CLAT joints are plotted for week 1-week

8 for central coronal slice location. Data plotted as mean ± SE.

..........................................................................................................................................


116

Figure 7. Changes in the tissues of the lateral condyles of MSX joints, in comparison to lateral

condyle controls, during the eight week observation period post meniscectomy. Cartilage

thickness (A), cartilage T2 (B) and T2 of epiphysis (C) of CTRL and MSX joints are plotted for

week 1-week 8 for central coronal slice location. Data plotted as mean ± SE.

..........................................................................................................................................

Discussion

This study investigated the development of PTOA in a rat model where the disease was

induced by complete removal of medial meniscus or meniscectomy [344]. The menisci

of the knee joint protect the ends of the bones from rubbing against each other and

provide shock absorption and load transmission [298, 363]. Meniscectomy, either


117

partial or total, disturbs the natural loading mechanism of a knee joint, which in turn

increases the amount of strain on the AC. The absence of a meniscus in the knee joint

has been linked to “over-compression” of the cartilage as well as a slower post-load

recovery. These effects, in turn, have been postulated to set off a cascade of cellular

and structural events leading to the development of OA [221, 222, 364]. In our rat

meniscectomy model, the AC layers covering the medial condyles, both in tibia and in

femur, were exposed to an increased risk of ECM degradation or OA initiation. We

examined all tissues of the MSX and CLAT joints and compared the respective tissues

against that of the age matched controls using µMRI. By this, three physical quantities

were identified that consistently evolved with the progression of PTOA: the thickness

of AC, T2 of AC, and T2 of epiphysis. The use of age-matched controls ensured that

there was no bias inherent to the maturation process of the rat joints.

The imaging pulse sequence used in this study, Multi-Slice Multi-Echo

(MSME) imaging, is not the most common imaging sequence for anatomical

visualisation of articular cartilage; fast gradient echo-based pulse sequences such as

FLASH are more commonly used for this purpose30. Nevertheless, MSME was used in

our study because one of its key objectives was to obtain quantitative T2 maps of the

joints, and MSME allowed obtaining these in a time-efficient manner.

Swelling and Degradation of Articular Cartilage. The partial volume

correction by Eqs (2) and (3) allowed the determination of AC thicknesses that were

not integer multiples of the voxel size (78 µm), which effectively improved the

precision of thickness measurement in MRI. The thickening of AC was observed as

early as week 1 in the MSX joints, which continued to thicken for seven consecutive

weeks with a strong monotonic trend (Mann-Kendall trend test, p < 0.02). The rate of

change in the MSX joints significantly exceeded that in the CTRL joints (see Table 1).

By comparing these results with the results of histological analysis, it was concluded

that weeks 1–7 corresponded to the gradual depletion of proteoglycan and cellular loss,

which in turn allowed the AC to swell. At week 8, a severe loss of AC thickness was

observed in the T2-weighted image, with the average week-8 thickness being lower than

that at week 1. Histological results indicated that week 8 corresponded to the erosion

of AC (Fig. 2A). It should be noted that the histology-based thickness measurements

were performed on dehydrated sections, which did not reflect the swelling of AC


118

present in the native samples. Therefore, the AC thickness in the histological samples

cannot be taken as an indicator of the AC thickness in the actual intact knee.

Slice 1

(anterior)

Slice 2

(middle)

Slice 3

(posterior)

Change in Thickness of

Articular Cartilage (µm)

CTRL Δ8 – 1 -15 ± 5 2 ± 10 -3 ± 10

MSX Δ7 – 1 54 ± 19 36 ± 8 3 ± 46

Δ8 – 1 -30 ± 26 -21 ± 31 -33 ± 22

CLAT Δ7 – 1 12 ± 13 -40 ± 9 47 ± 11

Δ8 – 1 3 ± 12 -16 ± 16 27 ± 6

Change in T2 of Articular

Cartilage (ms)

CTRL Δ8 – 1 0.8 ± 3.1 -0.04 ± 2.6 2.4 ± 3.3

MSX Δ7 – 1 19.4 ± 3.4 34.9 ± 5.6 20.6 ± 2.6

Δ8 – 1 24.1 ± 5.3 4.0 ± 6.4 40.5 ± 2.8

CLAT Δ7 – 1 -2.3 ± 0.7 -0.7 ± 2.4 -7.8 ± 1.6

Δ8 – 1 -3.3 ± 2.2 -4.9 ± 4.6 -5.8 ± 2.2

Change in T2 of Epiphysis

(ms) CTRL Δ8 – 1 -4.3 ± 2.0 -1.3 ± 2.7 -3.7 ± 1.7

MSX Δ7 – 1 -11.4 ± 2.7 -20.0 ± 7.7 -11.2 ± 7.4

Δ8 – 1 -5.9 ± 2.7 -17.9 ± 7.6 -14.4 ± 7.6

CLAT Δ7 – 1 -7.1 ± 4.1 -13.6 ± 3.7 -7.2 ± 2.1

Δ8 – 1 -7.8 ± 4.1 -16.4 ± 3.8 -10.2 ± 1.3

Table 2. The intragroup changes in AC thickness, AC T2 and T2 of epiphysis of the medial tibial

condyle, for CTRL, MSX and CLAT joints. 7 1 is defined as the difference between week 7

and week 1 for the given physical quantity (A), the given slice location, and the given group:

71 = A7 (group, slice) – A1 (group, slice). Equivalent definition was used for 8 1: 8 1 = A8

(group, slice) – A1 (group, slice). The respective quantities were measured from T2-weighted

MR images and T2 maps at the three different coronal slice locations shown in Fig. 1. Data is

presented as mean ± standard error.

..........................................................................................................................................

Cartilage thickening or swelling in early OA have been observed by MRI in

previous studies [110, 111, 353]. A study of a spontaneous model of OA in guinea-pigs

reported an initial increase of AC thickness for 24 weeks followed by a decrease for the

next 28 weeks [353]. In a rabbit partial meniscectomy model of PTOA, the cartilage

thickness at the weight-bearing area of the femoral medial condyle increased for eight

weeks following surgery and then decreased at week 10 [110, 111]. However, the

thickness of tibial cartilage showed a significant increase only at week 6 [110]. Our

study, along with those cited above, supports the model whereby the AC thickens

following the removal of meniscus, and continues to thicken until it reaches a maximum

thickness supported by its ECM. After this time point, AC loses its thickness due to

erosion of the ECM. The complete medial meniscectomy employed in our model can

be assumed to have more severe effects on weight bearing than partial meniscectomy.


119

Therefore, it can be expected to result in a more aggressive progression of PTOA

comparing to the above mentioned PTOA models, which is in agreement with our

observations.

Evolution of T2 in Articular Cartilage. The transverse spin relaxation time

constant (T2) is a valuable indicator of the water content in AC. The water content, in

turn, is strongly related to the integrity of the proteoglycan-collagen matrix [3, 107].

The reduced proteoglycan content in PTOA disrupts the collagen network in AC [105],

which increases its permeability to water; this results in an elevated water content and

longer T2 values [68]. Literature suggests that the modified T2 of cartilage could be used

as a harbinger of the onset of PTOA before the disease reaches its radiographic stage

[365]. Our weekly observations of quantitative T2 maps confirmed the presence of T2

changes in the AC of the MSX joints from week 1 to week 8.

The spin relaxation rate constants, 1/T2, of AC can be viewed as the weighted

average of the contributions from two components: bound water (BW), which is

transiently associated with the ECM biomacromolecules (collagen and proteoglycans)

and has a short intrinsic T2, and free water (FW), which experiences a molecular

environment similar to that of bulk water and has a long intrinsic T2 [7, 64, 366, 367].

The T2 values therefore exhibit an inverse relationship with the local concentration of

proteoglycans [68]. However, cartilage T2 is also influenced by the T2 magic-angle

effect [29, 64, 190], which is due to the aligned collagen fibres anisotropically

restricting the rotational dynamics of bound water [57, 159]. The collagen orientation

varies throughout the AC, which typically divides the AC into three histological zones:

superficial, transitional and radial that differ in T2 values [31, 43]. The AC observed in

this study was rather thin and it was not possible to differentiate one zone from another.

For every sample, therefore, an average T2 value was computed from AC that combined

these three zones. By maintaining the same position and orientation of the samples

during µMRI, equal influence of collagen alignment on the T2 values was ensured for

all samples. This influence, if present, can be expected to cancel out in comparisons

between the samples.

Based on the relative contribution of BW and FW, to the observed 1/T2, the T2

evolution seen in the MSX joints (central coronal slice, Fig. 4 and Table 1) can be

related to the physiological changes in AC during the study period of eight weeks. The


120

minor positive T2 changes observed from week 1 to week 3 is interpreted as the excess

FW due to the swelling initiated in week 1. Further T2 elevation was observed from

week 4 to week 6, during which continued loss of PGs led to a further increase in FW

and a simultaneous decrease in BW. This hypothesis is supported by the histology

results that identified proteoglycan depletion and cellular loss at the week 4 time point

(Fig. 2A). A rapid T2 increase of 37.6 ms was observed at week 7 as the result of severe

proteoglycan loss, consequent decrease of the BW/FW ratio and increase in the free-

water content. A pronounced inversion of the temporal trend of AC T2 was observed at

week 8, when the T2 decreased by ~30 ms between weeks 7 and 8; this is attributed to

the erosion of the cartilage and the accompanying loss of FW from AC. Nevertheless,

a residual AC swelling appeared to remain because the T2 of the remaining AC was still

longer than that of CTRL (Fig. 4). However, the AC T2 was observed to continually

increase from week 1 to week 8 in the anterior and the posterior coronal slices (slice

locations as shown in Fig. 1) and no decrease in T2 or AC erosion were observed (Table

2). It suggests that the progression of PTOA was more aggressive in the central

(maximum load-bearing) coronal region than in the anterior or posterior regions of the

joint.

Cartilage Thickness and T2 as Biomarkers of PTOA. In this study, both the

thickness and T2 of AC were observed to experience continuous changes during the

progression of PTOA. T2 mapping has been established to be a sensitive marker of

collagen content, distribution and orientation [5, 14, 68, 368, 369] as well as of the

proteoglycan content [194, 370, 371], which influences the content of the free and

bound water. Because the development of PTOA involves both proteoglycan depletion

and collagen disruption prior to the erosion of AC, a marker sensitive to these changes

would have definite advantages in identifying the physiological changes occurring in

early PTOA. In a previous study, the correlation between the T2 values and the relative

water content of AC was reported in a rat OA model at two time points [10]. In a partial-

meniscectomy rat PTOA model, the AC of the weight bearing areas of the medial

condyles exhibited a significant increase in T2 three weeks after surgery [155]. In

another rat model, where PTOA was induced by the transection of ACL (ACLT),

significantly higher T2 was observed in the AC of the operated knees at week 4 and

week 13 post-surgery, the T2 value at week 13 was also significantly higher than that at

week 4 [10]. However, the results of these studies were insufficient to explain the


121

relationship between PTOA and its effect on the T2 of AC due to the limited number of

observational time points.

Our study has shown that the thickness of AC and its T2 were strongly correlated

(Spearman’s rank order correlation, p < 0.05) throughout the eight-week post-

meniscectomy observation period, where week 8 marks advanced PTOA according to

the Mankin score. However, there are also noticeable dissimilarities between the

sensitivity of AC T2 and its thickness (Fig. 4) to the development of PTOA. The AC

thickness in the MSX joints go through both positive (week 2, 3, 5 and 7) and negative

(week 4 and 6) shifts, as seen in Fig. 4. These shifts were observed before the erosion

of AC and therefore the reason for these apparent transient reductions in thickness is

unclear. Additionally, both positive and negative shifts of the AC thickness were

observed in the medial condyles of CLAT joints (Fig. 6) and in the lateral condyles of

MSX joints (Fig. 7). The underlying reason for these changes in AC thickness cannot

be explained within the scopes of this study. In contrast, the AC T2 continually

increased from week 1 to week 7 (Fig. 4) and decreased with the commencement of

cartilage erosion at week 8 in the medial condyle of MSX joints. Cartilage T2 remained

steady in the medial condyles of CLAT joints (Fig. 6) and in the lateral condyles of

MSX joints (Fig. 7). This dissimilarity can be attributed to the fact that AC thickness is

a gross measure of the after-effect of the changes occurring in AC, while T2 provides

insight into subtle compositional changes before AC erosion occurs. For example, the

first significant change in T2 seen in the MSX joints (week 4) precede that in AC

thickness (week 5) by one week. Similarly, the spike in T2 seen in week 7 precedes by

one week the erosion of AC (week 8). The AC thickness measurement is also more

severely affected by the resolution of the MR image in comparison to the cartilage T2

measurements. Therefore, T2 values of AC appear to be both an earlier and a more

reliable indicator for understanding the course of PTOA than AC thickness.

Changes in the Cortical and Trabecular Bone Volume of Epiphysis. Our results

have identified significant decrease in epiphyseal T2 in the anterior, central and

posterior coronal slices (slice location as shown in Fig. 1) of MSX and CLAT joints

(Fig. 5), which were mutually correlated. The epiphyseal T2 represents the T2 of both

subchondral bone and trabecular bone and therefore demonstrates the gradual reduction

of water content within the tibial epiphysis from wee 1 to week 8. Our visual

observation of MR weighted images (TE = 24 ms) have identified two distinctive trends


122

of subchondral bone remodelling: initial thinning followed by gradual thickening,

between disease onset (week 1) and advanced PTOA (week 8).

The rats that underwent surgery had limited movement following surgery due

to the surgical trauma and pain. Reduced movement of rats resulted in reduced amount

of load on their tibial condyle. After the removal of medial meniscus, the joints had

adopted to an alternative load distribution technique while the rats continued to move

with functioning joints. According to our observations, this altered load distribution

was the primary cause for the changed subchondral/trabecular bone volume ratio due

to the bone remodelling within tibial epiphysis that resulted in gradual decrease in

epiphyseal T2 from week 1 to week 8. According to the results presented in Fig. 5B, it

can be stated with certainty that the epiphysis of the medial condyle experienced

substantial alteration in all regions of the joint at each time point of observation.

In conventional practice, the quality of the subchondral bone is assessed by bone

mineral density (BMD) and bone versus tissue volume ratio (BV/TV) measured over

small cylindrical ROIs (few mm in diameter) [343, 372]. BMD and BV/TV are

computed from microcomputer-tomography (µCT) or X-ray scans of excised samples.

In two PTOA rat models, where knee joints were subjected to ACLT alone or the

combination of ACLT and MSX, damage to AC and subchondral bone loss was

observed within 2 weeks of surgery [208]. This was followed by a significant increase

in subchondral bone volume up to 10 weeks [208]. In a PTOA model of rabbit knee,

bone loss or decreasing volume BMD were observed 4 and 8 weeks post-ACLT, and

recovery to control values was observed at 12 weeks [157]. In a rabbit MSX model,

initial changes of cartilage were associated with a decrease in BMD of the proximal

tibia [372]. In a canine ACLT-MSX model, thinning and porosity of subchondral bone

were observed in the medial condyles 12 weeks after the operation [207]. With the

support of histological analysis and µCT data, thinning of subchondral bone was

identified as a localised phenomenon related to cartilage degeneration, while trabecular

bone changes were found to be related to mechanical loading [207]. In human post-

mortem samples with early OA in proximal tibia, significant deterioration in the three

dimensional architecture of cancellous bone and increased trabecular thickness and

density with relatively decreased connectivity were observed, which suggested a

mechanism of bone remodelling [358]. Due to the wide variability among the

OA/PTOA models and the variable time points of measurements (that represent


123

different developmental stages of PTOA), it is not possible to identify the exact time

line of subchondral/trabecular bone remodelling in PTOA from the above mentioned

studies. Additionally, due to the small field of view of µCT scanners, the sample must

be excised before scanning, which is unsuitable for clinical practice or for monitoring

the progression of PTOA. In contrast to the established CT-based protocol, we

employed MRI to measure the epiphyseal T2 for the assessment of subchondral bone

and trabecular bone within epiphysis, which allows non-invasive evaluation.

Effects of PTOA on Contralateral Joints. A very interesting finding of this

study was the temporal variations of the epiphyseal T2 in medial tibial condyles of the

contralateral joints. Without being subjected to any surgical procedure, the epiphyseal

T2 of contralateral joints had significant deviation from controls and demonstrated

striking resemblance to the joints subjected to meniscectomy with significant

correlation (Spearman’s Rank Order Correlation analysis, p < 0.01). The epiphyseal T2

of the CLAT joints, in the central coronal slice, showed consistent reductions that

continued up to week 8 (Fig. 5A). At week 8, the epiphyseal T2 of the CLAT joint

reached 11.2 ms (see Table 1). This epiphyseal T2 is comparable to the epiphyseal T2 of

the MSX joints between week 6 and week 7. This gradual reduction of epiphyseal T2

was identified in anterior and posterior coronal slices of the CLAT joints as well (see

Table 2), yet, from week 1 to week 7, the epiphyseal T2 was higher in magnitude in the

CLAT joints in comparison to the MSX joints.

In studies concerning the development of experimental PTOA, contralateral

joints are commonly ignored and considered to be unaffected. In fact, contralateral

joints are also used as control data [348]. Figure 6 shows the comparisons between the

following quantities of CLAT and CTRL joints: thickness of AC, T2 of AC and T2 of

epiphysis. It is obvious that these properties of the normal or control (CTRL) are not

similar to that of the contralateral (CLAT), particularly for the T2 of epiphysis. If the

contralateral was taken as the control and if the effects of PTOA were determined by

making comparisons, this study would likely result in a wrong understanding of PTOA

development.

Limitations of the Study. The resolution of the MR images acquired in this study

was limited to the voxel size of 78 x 78 µm2. Although this choice of resolution allowed

the T2 mapping of the intact limb containing the knee joint within a two-hour time slot,


124

it has the potential to introduce a partial volume effect that could affect the quantitative

accuracy of the MRI measurements. The AC of the rat joints was thin, and therefore it

was not possible to differentiate one cartilage zone from another one at this resolution.

The partial volume effect in the AC was mitigated by mathematical processing based

on Eqs (2) and (3) (see Materials and Methods). Nevertheless, a higher MRI spatial

resolution would undoubtedly be beneficial to the quantitative accuracy of the

measurements of the thickness and T2 of AC.

This study identified a gradual decrease of epiphyseal T2 in the medial epiphysis

of MSX and CLAT joints that indicated a continuous remodelling of the subchondral

and the trabecular bone within the medial epiphysis with the progression of the PTOA.

However, the epiphyseal T2 could not unambiguously differentiate the T2 representing

the subchondral bone from the T2 associated with the trabecular bone. Overall, the data

acquired in this study did not reveal the exact nature of bone remodelling, or the factors

underpinning the changes in epiphyseal T2. Further investigation is required, preferably

involving both µMRI and µCT measures for a complete understanding of the

epiphyseal bone remodelling in both MSX and CLAT joints.

Feasibility of Monitoring PTOA Progression Using MRI. This study has

shown that the three physical quantities: thickness of AC, its T2, and the epiphyseal T2,

that are sensitive to the development of PTOA, can be measured from the T2-weighted

images and the quantitative T2 maps. The use of the MSME imaging sequence allowed

fast measurement of all three characteristics and minimised the time of sample exposure

to room temperature (and therefore tissue degradation). The T2 maps were computed

from a series of T2-weighted images acquired using the MSME sequence in the µMRI

system. Partial volume effect, which often affects MRI measurements, was mitigated

by mathematical processing. Here, a single imaging modality (µMRI) was able to

provide adequate information about the development of PTOA in the rat models. This

marks a major methodological advance in the analysis of PTOA, in comparison with

the standard practice, where a minimum of two diagnostic imaging modalities are used

to assess the knee joint tissues: MRI for cartilage and CT for subchondral bone [157].

Due to the limitation of the bore size of the µMRI spectrometer used in this study, the

limb containing the knee joint had to be removed and imaged on its own. Nevertheless,

the knee was kept intact, including muscle and skin, and the limbs were maintained in

osmotic conditions mimicking the physiological environment during the imaging. This


125

augurs well for MRI-based comprehensive evaluation of PTOA in vivo, as our approach

is in principle transferrable to clinical MRI scanners. At the same time, it must be kept

in mind that imaging in vivo entails additional factors not present in sacrificed animals

(most notably, motional and susceptibility artifacts resulting from the presence of active

blood vessels), and the suitability of the present protocol for PTOA imaging in vivo

ought to be demonstrated by further research.

Our imaging protocol and subsequent analysis identified the sequential changes

in tibial cartilage and tibial epiphysis of rat knee joints by weekly observation for eight

weeks following complete medial meniscectomy. Gradual swelling of AC was

observed during the first week and continued for the next six weeks, while elevated

water content resulted in an increase of the T2 values. Depletion of proteoglycan was

identified in the fourth week that led to proteoglycan loss by the seventh week. Erosion

of AC was observed in the eighth week, accompanied by a drop in T2 values. Although

the thickness and T2 of AC were strongly correlated, T2 was clearly a more sensitive

marker of the integrity of AC. The average T2 of epiphysis continued to decrease with

the progression of PTOA. Integrating these observations, we identified the following

disease development pathway that lead to advanced PTOA: meniscal injury → AC

swelling (week 1 – week 7) → bone remodelling in subchondral and trabecular region

(week 2 – week 8) → gradual depletion of proteoglycan and loss of cellular density

(week 4 – week 6) → severe proteoglycan loss and free-water influx (week 7) →

erosion of the cartilage (week 8). Surprisingly, the contralateral joints also

demonstrated altered epiphyseal T2, which evolved with time.

Materials and Methods

Rat OA Model. A total of 30 Male Wistar Kyoto rats (11-12 weeks old, 300-350

grams weight) were purchased from the Medical Engineering Research Facility

(MERF, Brisbane, Australia) and housed in controlled day-night cycle (light/dark,

12/12 h) and controlled temperature (23 ± 1 ˚C). The 6 rats of the control group did not

undergo surgery. PTOA was induced in the remaining 24 rats by complete medial

meniscectomy (MSX) on the right hind knee joint [343, 344]. The rats were

anesthetised via intra-peritoneal injection with Zoletil (tiletamine 15 mg/kg, zolazepam

15 mg/kg) and Xylazil (xylazine 10 mg/kg), the medial collateral ligament was

transected just below its attachment to the meniscus, the meniscus reflected towards the


126

femur when the joint space opened and then the meniscus was cut at its narrowest point.

This resulted in complete transection of the medial meniscus. Care was taken to avoid

damaging the tibial surface. The surgical wound was closed by suturing the

subcutaneous tissue and skin in two different layers. No surgery was carried out on the

left hind knee.

The rats were allowed to walk freely in the cage after surgery. Pain killers

(Buprenorphine 0.05 mg/kg) and antibiotics (Gentamycin 5 mg/kg) were given to the

rats that underwent surgery. Following surgery, 3 rats were sacrificed every week (week

1 – week 8) and 6 knee samples were harvested: 3 samples of the MSX joints and 3

samples of the CLAT joints. Three CTRL joints were harvested at week 1 and 3 were

harvested at week 8. Each whole-joint knee sample extended from the middle of

femoral diaphysis to the middle of tibial diaphysis (Fig. 1). The muscle and skin

surrounding the knee joint were left intact in order to maintain an anatomically realistic

environment. The samples were then subjected to MRI measurements. Animal ethics

approval for this project was granted by the Queensland University of Technology

(QUT) and the Prince Charles Hospital Ethics Committees (QUT Ethics approval

number: 0900001134). All methods were carried out in accordance with the relevant

guidelines and regulations of QUT.

MRI Protocol. MR images were acquired at room temperature using a Bruker

Avance NMR spectrometer (Bruker, Germany) at 7 T using 1.5 T m-1 (150 G cm-1)

triple-axis gradient set, a Micro2.5 microimaging probe and a 25 mm radiofrequency

(RF) birdcage 1H resonator coil. In order to maintain physiological osmotic conditions

in the tissues imaged, the sample was hydrated for 2 hours in 0.01 M phosphate buffered

saline (Sigma-Aldrich, USA) and then immersed in Fomblin (Sigma-Aldrich, USA)

inside a 25 mm diameter NMR tube. The sample was positioned using purpose-built

Teflon plugs [57, 292, 293], with the axis of limb approximately parallel to the NMR

tube axis and the static magnetic field (B0), which was maintained for all MRI scans.

The field of view (FOV) was determined by a 3D gradient-echo localiser scan

using Fast Low-Angle SHot (FLASH) MRI sequence with repetition time (TR) / echo

time (TE) of 100/5 ms, 2 mm slice thickness, 70 mm x 70 mm FOV, and 128 x 128

pixel matrix. Ten axial slices were acquired by multi-slice multi-echo (MSME)

sequence with TR/TE of 1000/6 ms, 0.5 mm slice thickness, 30 mm x 30 mm FOV,


127

128 x 128 pixel matrix and 4 averages. Using these axial slices as references, 3 coronal

slices were obtained by MSME with TR/TE of 5000/6 ms, 32 echoes, 0.5 mm slice

thickness, 0.5 mm slice spacing, 20 mm x 20 mm FOV, 256 x 256 pixel matrix and 8

averages. The second coronal slice (as shown in Fig. 1) was positioned to contain the

largest cross-section of ACL and PCL. The SNR was computed by taking the ratio of

the mean pixel intensity in a region of interest (ROI) within the sample to the noise

amplitude in a ROI of the background air (noise-only region). The noise amplitude was

computed as the square root of the sum of the squared mean and the squared standard

deviation of the signal in a noise-only region in a magnitude image (noise and noise,

respectively):

2 2

noise noiseNoise (1)

This measurement was repeated for ROIs in cartilage, muscle and tibial epiphysis. The

SNR was maintained at a minimum of 9:1 for all tissues.

The week 4 CLAT joint samples were used to standardise the MR imaging

protocol. The remaining 57 knee joints underwent the identical MRI data acquisition

procedure discussed above. The data was divided into three groups: MSX group (24

right knee joints subjected to MSX: week 1–week 8), CLAT group (21 contralateral left

knee joints: week 1–week 3 and week 5–week 8) and CTRL group (6 control right knee

joints: week 1 and week 8).

Histology. Soft tissues were removed from the joints after MRI. The joints were

fixed in 4% paraformaldehyde, decalcified in 10% Ethylenediaminetetraacetic acid

(EDTA), dehydrated and embedded in paraffin. A series of 5 µm coronal sections that

matched the orientation and location of Slice 2 in Fig. 1 were then prepared from the

medial tibial condyle of the joint. These sections were stained with Safranin-O / Fast

Green [344, 373], which provided colour discrimination between bone and cartilage.

The depth or thickness of AC was measured from histology stains using ImageJ

(National Institutes of Health, USA) from the average distance (of three distance

measurements) between the superficial borders of cartilage to the boundary with the

calcified cartilage zone. The severity of PTOA was evaluated according to modified

Mankin’s histologic grading system, ranked between 0-14, where 0 is the rank for

normal and 14 is the rank for most severe OA [344, 360].


128

MRI Measurement of Articular Cartilage Thickness. The thickness of AC was

measured from the coronal MSME data sets at the relatively flat surface of medial tibial

condyles, between the medial intercondylar tubercle of intercondylar eminence and the

edge of the condyle. A straight line was drawn bordering the AC on the T2-weighted

image (TE = 12 ms) as shown in Fig. 2 (line “a”, the yellow line). Ten lines

perpendicular to line “a” were computed from femur to tibia (line “b”, the blue lines),

the voxels nearest each line “b” were identified by rounding and a signal intensity

profile was plotted along each line “b”. Three voxel intensities were specified: IC for

cartilage with the highest signal intensity, IF for femoral cortical bone with no/minimal

signal at the femoral end of cartilage and IT for tibial cortical bone with no/minimal

signal at the tibial end of cartilage. To correct for the partial volume effect, the volume

fraction of articular cartilage, PAC was computed using Eq. (2) for voxels located at the

interface between the femoral cortical bone and cartilage, and using Eq. (3) for voxels

at the interface between cartilage and the tibial cortical bone. PAC was 1 for the voxels

located entirely within cartilage. Partial volume correction was based on the assumption

that the pixel with partial volume can only have two tissue types: cartilage and cortical

bone of femur/tibia. Considering the signal intensity variation by slice position, I, IC, IT

and IF were specified individually from the intensity profiles obtained from the T2-

weighted image of each MRI slice.

𝑃𝐴𝐶 =𝐼−𝐼𝐹

𝐼𝐶− 𝐼𝐹 (2)

𝑃𝐴𝐶 =𝐼−𝐼𝑇

𝐼𝐶− 𝐼𝑇 (3)

Here, I is the signal intensity of a voxel at an interface between two tissue types.

The cartilage thickness was computed by multiplying the voxel dimension, 78

µm, by the sum of the PAC measurements of each intensity profile. The mean of the 10

thickness measurements was taken as the AC thickness of the medial tibial condyle.

The same procedure was followed to measure the AC thickness in the lateral condyle.

The data analysis procedures, described in this section and in the following sections,

were performed by in-house codes written in MATLAB R2014a (MathWorks, Natick,

MA, USA).


129

Measurement of Articular Cartilage T2. The quantitative T2 maps were computed

from the coronal MSME data. The multi-echo data of every voxel was fitted with a

three-parameter mono-exponential relaxation decay according to:

𝑆 = 𝑆0𝑒−𝑡


Here, S is the voxel signal intensity measured at the sequential MSME echoes, and t is

the cumulative echo time (ranging in each data set through 32 equidistant values from

TE to 32TE). The fit parameters were S0 (full signal intensity), T2 (apparent spin

relaxation time), and Soffset (the mean of the magnitude noise). With measured S and

known t, the values of T2, S0 and Soffset were obtained by iterative least-squares fitting

(LSF). A maximum of 100 LSF iterations were allowed for the voxels with S0 > 5 ×

Soffset. All three LSF parameters were determined individually for every voxel. The

number of voxels with S0 > 5 × Soffset varied between the imaging slices. However, in

any given slice fewer than 3% of the voxels were identified as having S0 < 5 × Soffset.

The SNR was maintained at a minimum of 9:1 for all tissues in all slices. Fitting

residuals were checked for randomness by Runs Test at α = 0.05 to verify the suitability

of the mono-exponential fit given by Eq. (4).

For AC T2 of medial tibial condyle, voxels with PAC > 0.5 were isolated from

the voxel intensity profiles. The corresponding T2 values were then extracted from

quantitative T2 maps using the voxel coordinates and the mean T2 was recorded. The

voxels near intercondylar eminence and curved edges were excluded in order to avoid

susceptibility artefacts. The same procedure was followed in the measurement of the

cartilage T2 in the lateral condyle of the tibia.

Measurement of T2 of Epiphysis. The mean T2 of tibial epiphysis was measured

individually for the medial and lateral condyles. The coronal cross section of the tibial

epiphysis is bordered by AC superiorly and by the growth plate or epiphyseal cartilage

(EC) inferiorly (Fig. 1) where both AC and EC have much higher signal intensities

compared to the cortical bone. Using the Sobel edge-detection filter on a T2-weighted

image (TE = 24 ms), the medial compartment of the tibial epiphysis was outlined. A

rectangular ROI was drawn at the centre enclosing 50% of epiphyseal area, two sides

of epiphysis were excluded in order to avoid the chemical shift artefact. The

corresponding T2 values were extracted from the quantitative T2 map using the voxel


130

coordinates and the mean T2 was recorded. The same procedure was followed to

measure the mean T2 in the lateral tibial epiphysis.

Statistical Analysis. The measurements of AC thickness, AC T2 and the T2 of

epiphysis for the MSX, CTRL and CLAT groups over the 8-week observation period

were entered in Matlab. For each physical quantity measured from each slice of each

sample, a mean and a standard error was calculated every week from week 1 to week

8. This was done separately for the medial and the lateral condyles. The following nine

data series (3 physical quantities x 3 groups of animals) were analysed for every slice

location of medial condyle: AC thickness of MSX, AC thickness of CTRL, AC

thickness of CLAT, AC T2 of MSX, AC T2 of CTRL, AC T2 of CLAT, epiphyseal T2

of MSX, epiphyseal T2 of CTRL and epiphyseal T2 of CLAT. The identical analysis

was performed for the lateral condyle.

The Mann-Kendall trend test [374, 375] was performed individually on each of

the thirty six data series (2 condyles x 3 slices x 3 physical quantities x MSX and CLAT

group) to ascertain the presence of a temporal trend. For the data series of each slice (2

condyles x 3 slices), Spearman’s Rank Order Correlation analysis [376] was performed

to evaluate the correlation between the thickness of AC and its T2 values. In order to

check for inter-slice correlations of epiphyseal T2 between MSX and CLAT joints and

between the slices of the MSX joints, the Spearman’s Rank Order Correlation analysis

[376] was performed between the epiphyseal T2 data series of these groups at each slice

location (2 condyles x 3 slices).


131

Acknowledgements

This work was supported by the program grant from the Prince Charles Hospital

Research Foundation (MS2014-12). The authors would like to thank the staff at the

Medical Engineering research facility for assisting with the animal care and Dr R. Mark

Wellard for assistance with the MRI measurements.

Author contributions

Y.X. and K.I.M. designed the study. I.P. and T. A. conducted the animal study. I. P.

conducted histology measurements. T.A. conducted the MRI measurements and

analysed the MRI data. K.I.M. and Y.X. supervised the work. T.A., I.P. and K.I.M.

wrote the manuscript. All authors have read and approved the final manuscript.

Additional Information

Competing Interests: The authors declare no competing interests.

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims

in published maps and institutional affiliations.

Chapter 6: Summary and Future Scope

132


_____________________________________________________________________

The aim of this thesis was to explore the applications of transverse relaxation

based MR imaging techniques for non-invasive quantitative evaluation of the structure

and composition of biological tissues. The three case studies presented in this thesis

have experimentally investigated and evaluated the analytical efficacy of the transverse

relaxation based techniques and quantitative T2 measurements for the assessment of

structural scaffold of cartilage, for identifying chemical composition of breast tissue

and for detecting pathological alterations in multiple tissues of the knee joint. For this,

three different tissue type scenarios were investigated using established transverse

relaxation based sequences and analysis techniques while some techniques were

modified and new techniques were introduced when required. The case studies have

also identified previously unknown information on the composition of native and

pathological tissues and thereby demonstrated the suitability of the application of

transverse relaxation based techniques for comprehensive assessment of biological

tissues and organs.

This thesis focuses on the signal generated from the MR of the 1H population

present in biological tissues.1H is abundantly present in the water of extracellular matrix

(ECM) and cells as well as in other tissue components, such as, fat. The nature of the

transverse relaxation decay is primarily influenced by two factors: dipolar interaction,

which is the direct interaction between two magnetic dipoles, and chemical exchange,

which is the exchange of spins between different 1H populations that represent different

chemical species (see Spin Relaxation in section 2.1.3). The dipolar interaction is

influenced by restricted movement of water molecules due to the structural anisotropy

in biological tissues. On the other hand, chemical exchange is sensitive to the water and

1H content and their distribution in tissues. Consequently, transverse relaxation based

MR techniques can indirectly probe both the structural heterogeneity and the chemical

composition of the tissue that hosts the 1H population. Transverse relaxation based or

T2 imaging has recently been incorporated into clinical imaging protocols for measuring


133

cartilage thickness [265] as well as for evaluating collagen network and proteoglycan

content in cartilage ECM [3, 108, 377].

The structural scaffold of cartilage is made up of networks of collagen, which

restricts the movement of water molecules in cartilage ECM. The orientational

dependence of transverse relaxation or T2 can be used to infer the 3D fibre architecture

of the collagen scaffold by rotating a cartilage sample around the static magnetic field

of a MRI system, B0, and by measuring quantitative T2 maps at different orientations

[66, 67]. The orientational dependence is emphasised at the 0° orientation of the sample

with respect to B0 while it approaches zero at the 55° orientation of the sample (magic

angle effect). The magic angle effect has been used for delineating the extent of

anisotropy observed in the collagen network [1, 8, 29, 30, 43]. Previously, the collagen

architecture in articular cartilage (AC) has been examined using histology and various

imaging techniques, including scanning electron microscopy, polarised light

microscopy, NMR and MRI. The general consensus on cartilage research proposes a

three zone histological model on the basis of the orientation of the collagen fibres and

the collagen volume fraction. The superficial zone borders the articular surface (AS)

and contains densely packed collagen fibres aligned parallel to the AS. The tensile

strength of the collagen fibre is the greatest at this particular orientation [378]. It

therefore facilitates stress distribution and enables fast tissue response at high loading

rates [177-179]. The transitional zone lies next to the superficial zone, the collagen

fibres are orientationally disordered at this zone. The radial zone lies in between the

transitional zone and the calcified zone. It contains highly aligned collagen fibres that

are primarily perpendicular to the AS. This particular organizational pattern of the

collagen fibres, in conjunction with the high proteoglycan (PG) content in the radial

zone, restricts extreme deformation of cartilage ECM in response to compressive

loading [50, 179, 180].

In human, the superficial zone was measured as the thinnest (3%-12%) and the

radial zone was measured as the thickest (>50% of total thickness) histological zone

[379]. However, research has also shown that the thickness of the histological zones of

AC, as well as the composition and organization of the major molecular components,

may vary across species and even across different sites in the same joint [69-72].

Although many research investigations have explored the collagen organization in

cartilage samples using MRI, mostly acquired from bovine and canine cartilages, the


134

cartilage specific zonal distribution remains unknown to date. For example, cartilage

samples obtained from knee joints have been examined but the zonal distribution

particular to femoral cartilage and tibial cartilage is not defined. The first case study of

this thesis aimed to gain an understanding of the cartilage specific zonal distribution in

the femorotibial cartilages of kangaroo and thereby demonstrate the use of transverse

relaxation based techniques for site and cartilage specific collagen assessment.

Chapter 3 presented the first case study of this thesis that used magic angle

effect to identify the collagen fibre architecture in the femoral hyaline cartilage (FHC),

tibial hyaline cartilage (THC) and tibial fibrocartilage (TFC) of adult red kangaroo

(Macropus rufus) [124]. Using µMRI, sample-specific relative depth profiles of R2 (R2

= 1/T2) were obtained at 0° (𝑅20) and 55° (𝑅2

55) orientations of cartilage samples with

respect to B0. From these relative depth profiles, the relative depth profiles of the

anisotropic R2 component, 𝑅2𝐴 was computed. The 𝑅2

𝐴 depth profiles showed the

variations in collagen anisotropy across the depth of cartilage samples. By observing

and analysing the relative depth profiles of 𝑅2𝐴, the three histological zones were

identified in each cartilage sample.

The FHC samples exhibited the typical three-zone structure with collagen

distribution similar to the known patterns seen in other mammals. However, the average

relative thickness of the superficial zone was 28 ± 3% of the total thickness of the

cartilage. This value is substantially large in comparison to the superficial zone in

bovine knee cartilage, which often occupies <10% of the cartilage thickness [29, 30].

Therefore, the presence of the thick superficial zone indicated that this zone plays a

crucial role by maintaining the tensile and shearing resistance of AC and ensures an

extremely low coefficient of friction in AS during the large flexion of the kangaroo

knee. The radial zone occupied nearly 60% of the total thickness of FHC in kangaroo,

yet only ~20% of the total thickness had 𝑅2𝐴 greater than two typical standard deviations

(𝑅2𝐴 > 0.03ms-1). The less ordered collagen fibres in the radial zone indicated that

kangaroo FHC experiences limited amount of compressive force in comparison to that

of other mammals that have highly ordered collagen fibres in the radial zone.

The THC is present at the periphery of the tibial plateau of the kangaroo knee

and covers ~50% of its area. The 𝑅2𝐴 measured from the THC samples showed that

although all three histological zones were present in the THC samples, the radial zones


135

were considerably thicker and contained very highly ordered collagen fibres in all

samples. The average relative thickness of the radial zone was 80 ± 6% of the total

cartilage thickness. Both the average 𝑅2𝐴 (0.09 ± 0.04 ms−1) and the maximum 𝑅2

𝐴 (0.15

± 0.01 ms−1) were significantly higher than the respective 𝑅2𝐴 measurements obtained

in the radial zones of the FHC and TFC. In addition, a greater PG density was identified

in the radial zone of THC than in all zones of FHC and TFC. The highly ordered

collagen fibres and the high PG density indicated that the THC is particularly adapted

for the frequent high-amplitude compressive stress that the kangaroo THC experience

during the hopping locomotion.

The fibrocartilage pad of the kangaroo tibia is absent in most mammalian knee

joints. However, similar structures have been identified in knee joints of animals whose

predominant mode of movement is jumping. Histologically, TFC differ from THC in

chondrocyte density, PG content, and collagen and elastin architecture [295]. Our

𝑅2𝐴 depth profiles showed that TFC samples had anisotropic superficial and radial zones

that were separated by isotropic transitional zones. In the radial zone, the collagen fibres

were less aligned in TFC in comparison to that in THC. A rapid increase in the PG

content was identified near ~88% depth of cartilage that continued to increase for the

full thickness of TFC. This zone was named “attachment sub-zone” for its probable role

as a transition from radial to the tidemark, where collagen fibres are anchored to the

subchondral bone. It was postulated that the TFC adapts to the high compressive stress

by controlled deformation that maximises articular contact surface and minimises peak

loads in the regions of contact between the tibial plateau and the femoral condyle. The

strong anchoring base provided by the “attachment sub-zone” enables TFC to withstand

the large deformation required for the hopping locomotion.

This is the first MRI study of the knee cartilages of kangaroo where the use of

magic angle effect had identified the characteristic zonal structures of three cartilage

types. The measurement technique of 𝑅2𝐴 depth profile was introduced by this study.

The cartilage specific 𝑅2𝐴 depth profiles were particularly useful for the identification

of the functional roles of FHC, THC and TFC. According to our results, the cartilage

specific collagen networks were optimally structured to meet the high biomechanical

demands placed on the kangaroo knee cartilages during hopping: FHC was specially

structured with thick superficial zone for resisting shear deformations; THC featured

robust radial zone with highly aligned collaged fibres and high PG content for enduring


136

extraordinary compressive stress and TFC contained “attachment sub-zone” for

facilitating anchoring of collagen fibres for persisting cartilage deformation. In

addition, it is worth noting that, in all of the FHC, THC and TFC cartilage samples, the

𝑅255 measurements were observed to increase closer to the subchondral bone. Since

the𝑅255 is the isotropic contribution of R2, any change in 𝑅2

55 is likely to be related to

the chemical composition of the tissue and not to the collagen anisotropy. Therefore,

the increase in 𝑅255 can be attributed to the increase in PG content in the radial zones,

which is supported by the findings in literature that identified increased PG content in

the radial zone in comparison to other histological zones of AC [50, 179, 180]. Our

transverse relaxation based analysis allowed investigation of ROI sufficiently large for

a whole sample view of collagen organisation. Nevertheless, there were subtle

variabilities in the collagen alignment patterns within each cartilage type. This

variability has generally been overlooked in the past studies of kangaroo cartilage that

examined collagen organisation in small ROIs using histology or optical microscopy

[186, 273, 295]. These observations also bring attention to the importance of follow-up

studies for investigating whole cartilages in kangaroo joint for obtaining a more

comprehensive understanding of the collagen architecture and PG distribution.

The experimental protocol used in this study can be used to interrogate the

collagen architecture of the cartilage tissue in any species and consequently infer the

associated biomechanical capacities. Future research studies may incorporate non-rigid

image registration to match the cartilage voxels between the R2 maps obtained at 0⁰ and

55⁰ orientation of the same sample. Then, a 2D map of the 𝑅2𝐴distribution can be

computed that may allow direct visual interpretation of cartilage anisotropy. In the same

way, by acquiring R2 measurements at the 0⁰ and 55⁰ orientations from the whole

cartilage samples and by computing 𝑅2𝐴 maps, 3D models of collagen scaffold can be

developed for the complete cartilage tissues. However, the natural curvature of cartilage

tissues should be considered while computing the 𝑅2𝐴 maps. A 3D model of collagen

scaffold in cartilage will be helpful in establishing the relation between collagen

architecture and the biomechanical capacity of cartilage associated with its structural

heterogeneity. A model of collagen scaffold can also be used as a standard for artificial

cartilage development by tissue engineering and for evaluating the timely outcome of

cartilage tissue regenerative therapies.


137

Kangaroos hop at an average speed of 40 km/h and may reach a speed of 50 –

65 km h-1 in short bursts [272]. The ground reaction force and load experienced by the

kangaroo knee joints are several times higher than that of human walking and running

at the same speed [273, 274]. The kangaroo knee cartilages are extremely robust and

durable to sustain such high level of dynamic stresses at regular intervals for a lifetime.

Our results have identified the collagen distribution particular to kangaroo knee

cartilages and have proposed methods that may allow the computation of the 3D

collagen scaffold of whole knee joints. These information and techniques may inspire

new designs for cartilage tissue engineering in future. Research investigation by Brama

et al. showed that the biochemical composition of cartilage is uniform at birth [282].

The biochemical and therefore the biomechanical heterogeneities in cartilage are

influenced by the functional adaptation to weight bearing in early life [283]. Exercise

plays a crucial role in the adaptation of AC to different stress environments.

Consequently, the knowledge of the collagen structure and the expected functionalities,

as identified by our study, may benefit cartilage specific physiotherapy and trainings.

The second case study of this thesis (chapter 4) particularly focused on the

capacity of transverse relaxation based analysis for probing the chemical composition

of tissues. Using a single sided portable NMR instrument, it experimentally evaluated

the applicability of T2 measurements for identifying the composition of breast tissue

and for providing quantitative information on the relative prevalence of the chemical

species in tissue. The breast tissue mainly consists of two components: fibroglandular

tissue (FGT) and adipose tissue (fat). The mammographic density (MD) is the measure

of the relative amount of FGT as opposed to the amount of adipose tissue in breast. The

breast tissue with HMD contains a significantly greater proportion of FGT and also less

fat than the breast tissue with LMD. The prevalence of FGT is highly correlated with

the water content in breast ECM while T2 is highly sensitive to water content and its

distribution in biological tissues. Additionally, NMR measurements can focus on signal

from 1H, which is present in abundance in both adipose tissue (fat) and in water

associated with FGT. Theoretically, according to the inherent nature of transverse

relaxation decays and the different biochemical composition of adipose tissue and

water, the adipose tissue and water should result in distinctive T2 values in quantitative

assessment. In this experiment, the NMR measurements were obtained from breast

slices acquired from patients. Using Carr-Purcell-Meiboom-Gill sequence, pure T2


138

relaxation decays were measured from 1) the regions of HMD and LMD in the full

breast slices, 2) the excised HMD and LMD regions and 3) the same regions after H2O-

D2O replacement. The measured T2 relaxation decays were converted into T2

distributions using one-dimensional inverse Laplace transform.

The presence of two major peaks were identified in the T2 distributions

measured from the HMD regions whereas only one peak was prominent in the T2

distributions measured from the LMD regions. Peak specific geometric mean T2 (gmT2)

and area fraction (AF) values were measured from the T2 distributions. The native breast

tissue contains H2O. The T2 distribution measured from native tissue sample with HMD

demonstrated the presence of two T2 peaks, one for fat and one for water. After the

H2O-D2O replacement, the T2 distribution measured from the same sample featured

only one T2 peak. Because deuterium (2D) is not responsive to MR, the remaining T2

peak was confirmed to represent the 1H population in fat. This procedure was repeated

for all samples in the study and it was unambiguously confirmed that the T2 peaks

centred approximately at 10 ms corresponded to water and the T2 peaks centred close

to 80 ms corresponded to fat within the sample. Additionally, the gmT2 values of water

or fat, whether measured from HMD region or LMD region, showed no significant

difference (P < 0.005). This observation indicated that the fat present in LMD regions

of breast tissue is the same chemical species as the fat observed in the HMD regions.

For the same reason, the water present in HMD regions represent the same chemical

environment as the water present in the LMD regions of breast tissue.

On the contrary, the AF of the water-peaks measured from HMD regions were

significantly different (P < 0.005) than that of the LMD regions. Similarly, the AF of

the fat-peaks measured from HMD regions were also significantly different (P < 0.005)

than that of the LMD regions. Breast tissues with HMD contained significantly higher

proportion of water and consequently higher AF measurements for the water peaks in

comparison to the tissue with LMD. No statistically significant difference was found

between the measurements obtained from full breast slices versus the measurements

obtained from the excised regions. The densities of breast estimated from this study

were compared against the MD measurements previously specified from X-ray

mammograms by a clinical radiologist. It was shown that the combination of gmT2 and

AF measurements can unmistakable differentiate between breast tissue regions with

HMD from breast tissue regions with LMD. These measurements can also provide


139

quantitative information on the relative proportion of fat (corresponding to adipose

tissue) and water (corresponding to FGT) within the sample. Although the primary

focus of many MRI interpretations is the water 1H, the results obtained in this study

have shown that the fat in breast tissue is a good source of 1H and of the associated T2.

However, because this was a pilot study for MD assessment, there were several

limitations that future studies should address. The number of breast samples should be

increased to include samples from patients with varying Breast Imaging Reporting and

Data System (BI-RADS) scores (BI-RADS 1 – BI-RADS 4). Accordingly, the gmT2

and the AF measurements should be obtained from large number of HMD and LMD

regions. The numeric standards (gmT2 and AF measurements) particular to HMD and

LMD regions can then be computed by appropriate statistical analysis. In future, in

order to measure the MD of a breast, we recommend the measurement of T2 relaxation

decays at several key locations in the breast at multiple depths. Then, by comparing the

gmT2 and AF measured from such relaxation decays, with the established gmT2 and AF

standards specific to HMD and LMD, the MD of the breast under examination may be

identified while the presence and the relative proportion of water and fat in breast may

be inferred from the gmT2 and AF measurements.

The work presented in chapter 4 is the first study on transverse relaxation based

assessment of breast tissue using portable NMR instrument. It has introduced a novel

method for assessing MD and breast tissue composition using NMR. X-ray

mammogram is the current clinical standard for assessing MD, however,

mammography exposes patients to ionizing radiation and suffers from certain

limitations, such as, projectional imaging artefact and mammographic masking.

Although transverse relaxation based MRI is capable of measuring spatially resolved

MD, it also requires substantially higher cost involvement in comparison to

mammography, and hence is less likely to be adopted for routine breast screening.

Conversely, the use of portable NMR is cost-effective and it has the competency of

performing quantitative T2 analysis on individual relaxation decays. The use of portable

NMR for therapeutic purposes or for MD screening is expected to bring definite

advantages to women susceptible to radiation and mammographic masking. The results

of this study have also demonstrated that the transverse relaxation decay and

quantitative T2 measurements are sensitive to the water content and chemical

composition of biological tissues. Breast tissue has no known structural heterogeneity


140

like cartilage that may influence the T2 measured from breast. Hence, the results

obtained in this study illustrated the direct interrelation between FGT/fat composition

and the corresponding T2 measurements.

The sensitivity of transverse relaxation decays, towards the chemical

composition of tissues, was also identified in our first case study on kangaroo knee

cartilages. It was demonstrated by the increases in the anisotropic R2 components in

relation to the increases in PG contents in the radial zones of cartilages. Although

collagen anisotropy was the dominant factor that influenced the R2 measurements in

that study, variation in the chemical composition of ECM was also identifiable by the

transverse relaxation. Previously, in neuroimaging, T2 distributions measured from

normal brain has shown distinct peaks for myelin water, intra/extracellular water and

cerebrospinal fluid. The measurements obtained from T2 peaks were used to provide

estimates of total water content (total area under the T2 distribution) and myelin water

fraction (MWF, fractional area under the myelin water peak) and identify different

white matter structures that had characteristic MWFs [33]. The transverse relaxation

decays measured in brain MRI are often multi-exponential and are analysed by

specialised quantitative analysis techniques to identify the individual T2 components in

the relaxation decays. Quantitative T2 measurements obtained using µMRI have been

successful in detecting pathological water compartments with particular T2 values in

murine models of glioblastoma [12, 13]. However, in the presence of a pathological

condition, it is not possible to focus on the exclusive effect of tissue composition on T2

variation due to the various anatomical changes that occur during the development of

the disease. The relatively simple composition of the tissues analysed in this study

allowed the definite identification of the T2 values that correspond to two specific

chemical species or source of 1H population: 1H in fat and 1H in water associated with

FGT. For future studies, this approach of investigation can be used to identify

quantitative T2 values specific to the other chemical species present in biological

tissues, which also host the 1H population. The quantitative NMR analysis can also be

incorporated to quantitative MRI in order to identify the T2 and corresponding chemical

species on a voxel-by-voxel basis. These information on tissue specific T2 will be

valuable for characterizing the chemical composition of native tissues as well as for

diagnosing pathological conditions based on the alteration in native T2.


141

The final case study of this thesis is presented in chapter 5 that used

transverse relaxation based µMRI in order to obtain a comprehensive and quantitative

understanding of the developmental pathway of post traumatic osteoarthritis (PTOA)

in rat knee joints [97]. This study investigated whole knee joints using T2 weighted

echoes and voxel T2 measurements obtained by using MSME sequences and subsequent

quantitative analyses. PTOA was induced at week 0 by complete medial meniscectomy.

The rat knee joints were examined at weekly time points for the following 8 weeks in

order to capture the gradual morphological and physiological changes from very early

PTOA preceding to the severe stage. All tissues of the whole knee were studied in the

joints that were subjected to meniscectomy (MSX), in the contralateral (CLAT) joints

and in the control (CTRL) joints. Three physical quantities were identified in the medial

tibia that consistently evolved with the progression of PTOA: the thickness of AC, T2

of AC, and T2 of epiphysis. This study introduced the method of partial volume

correction (Eqs (2) and (3) in [97]), which effectively improved the precision of

thickness measurement in T2 weighted MRI. This allowed the measurement of AC

thicknesses that were not integer multiples of the voxel size (78 μm in this case). The

improved precision in measurements allowed the identification of changes in AC

thickness at week 1, which is the earliest time point to have reported AC swelling in

PTOA. The thickening of medial AC was observed at week 1 in the MSX joints and it

continued to thicken for seven consecutive weeks with a strong monotonic trend (p <

0.02). By making comparison with the results of histological analysis, it was identified

that, weeks 1–7 corresponded to the gradual depletion of PG and cellular loss, which in

turn allowed the AC to swell. A severe loss of AC thickness was observed at week 8,

which marked severe PTOA according to the Mankin score.

The quantitative T2 measurements confirmed the presence of T2 changes in the

tibial AC of the medial compartment of the MSX joints from week 1 to week 8. At the

imaging resolution used in this study (78 x 78 µm pixel), two or three histological zones

of the rat cartilage (thickness = 141 ± 10 µm in control joint at week 1) were likely

presented by the same voxel and therefore an average T2 value was computed from the

AC that combined the three histological zones. The magic angle effect of T2 was largely

ignored in these measurements. Minor positive shifts were observed in T2 from week 1

to week 3, which was interpreted as the excess free water (FW) due to the swelling

initiated in week 1. T2 continued to increase further from week 4 to week 6, which was


142

attributed to the continued loss of PGs that led to an additional increase in FW and a

simultaneous decrease in bound water (BW). Severe PG loss, consequent decrease of

the BW/FW ratio and increase in the FW content was detected at week 7, which was

represented by a rapid T2 increase of 37.6 ms. The T2 decreased by ~30 ms between

weeks 7 and 8, which indicated the erosion of the cartilage and the accompanying loss

of FW from AC. These results demonstrated that quantitative T2 of cartilage was a

valuable indicator of cartilage physiology and that T2 can be used as a harbinger for

assessing the degradation of AC. The increase in T2 with PG depletion is also in

agreement with our observations from normal kangaroo cartilages (chapter 3) where

increased R2 (1/T2) was identified in cartilage areas with high PG content.

Additionally, in this study, the thickness of AC and its T2 were strongly

correlated (p < 0.05) throughout the eight-week post-meniscectomy observation period.

However, careful evaluation of these parameters revealed dissimilarities between the

sensitivity of AC T2 and its thickness. Previously, few studies have reported AC

swelling or thickening in early OA [110, 111, 353]. However, because of the fact that

AC thickness is a gross measure of the after-effect of the changes occurring in AC, the

thickness measurements were insufficient to diagnose the sequential alterations in AC

during OA development. On the other hand, our results demonstrated that, T2 provided

insight into the subtle compositional changes that occurred before the erosion of AC.

Accordingly, quantitative T2 of AC was both an earlier and a more reliable indicator for

understanding the course of PTOA than AC thickness.

The quantitative T2 maps measured from tibial epiphysis showed significant

decrease in epiphyseal T2 in both MSX and CLAT joints, which were mutually

correlated (p < 0.01). By convention, the subchondral bone is assessed by measuring

bone mineral density and bone versus tissue volume ratio using microcomputer-

tomography (μCT) or X-ray scans of excised samples. Instead, this study employed

μMRI to measure the epiphyseal T2 for the assessment of subchondral bone and

trabecular bone within epiphysis, which allowed non-invasive evaluation. The

measured epiphyseal T2 represented the average T2 of both subchondral bone and

trabecular bone. Because the cortical bone is non-responsive to MR, the epiphyseal T2

corresponded to the trabecular bone or more specifically to the bone marrow that filled

up the space between the trabeculae. Bone marrow is composed of hematopoietic cells,

marrow adipose tissue and supportive stromal cells. Our case study on breast NMR has


143

shown that fat is a good source of 1H in breast tissue and of the associated T2. Likewise,

it is reasonable to postulate that, the average T2 measured from the tibial epiphysis was

a measure of the joint contribution of the 1H population that resided in the intra-cellular

water and in the adipose tissue in trabecular bone. Therefore, the gradual reduction in

the epiphyseal T2 over the eight-week long study period can be attributed to the ongoing

bone remodelling process that resulted in progressive calcification, which in turn

caused loss of water and fat content within the tibial epiphysis from week 1 to week 8.

By integrating the above results, the following developmental pathway was determined

that lead to advanced PTOA: meniscal injury (week 0) → AC swelling (week 1 – week

7) → bone remodelling in subchondral and trabecular region (week 2 – week 8) →

gradual depletion of proteoglycan and loss of cellular density (week 4 – week 6) →

severe proteoglycan loss and free-water influx (week 7) → erosion of the cartilage

(week 8).

This study demonstrated that the use of transverse relaxation based µMRI alone

was able to provide adequate information about the development of knee PTOA in the

rat models. During the progression of PTOA, it detected changes in cartilage thickness,

which is a structural alteration. It also detected changes in cartilage T2 and epiphyseal

T2, which are related to the compositional alterations in biological tissues. In essence,

it combined the analytical approaches of the first two case studies and aimed to identify

both structural and compositional alterations in knee joint tissues that take place during

the development of PTOA. MSME imaging sequence was used in this study for time-

efficient acquisition of sufficient T2 echoes required for computing quantitative T2

maps. However, the use of MSME sequence limited the resolution of the MR images

acquired in this study to the voxel size of 78 x 78 µm2. Although the partial volumes

were corrected while measuring the thickness of cartilage, the plausible effects of

partial voluming in quantitative measurements cannot be completely disregarded at this

resolution. In future, MRI with higher spatial resolution is recommended for improved

quantitative accuracy of the measurements. In addition, by using MRI with higher

spatial resolution or by using a bigger joint of another species instead of rats, and by

following the 𝑅2𝐴 measurement protocol explained in chapter 3, the changes in collagen

scaffold can be also be estimated during the development of PTOA.

In standard practice, a minimum of two diagnostic imaging modalities are used

to assess the knee joint tissues: MRI for soft tissues and CT for subchondral bone. This


144

study, therefore, marks a major methodological advance in the analysis of PTOA using

MRI. Yet, the data acquired from this study were not sufficient to elucidate the exact

nature of bone remodelling in the subchondral bone and the trabecular bone or the

factors responsible for the changes observed in the epiphyseal T2. It is recommended

that future investigations should be carried out involving both μMRI and μCT measures

for a complete understanding of the epiphyseal bone remodelling in both MSX and

CLAT joints. Previously, quantitative T2 measurements have been used to differentiate

between native and abnormal regions of femoral AC in human [76]. The imaging and

analysis protocol developed in this study is in principle transferrable to clinical MRI

scanners. It is non-invasive and is potentially suitable for longitudinal assessment of

whole-knee OA in patients. However, due to the presence of motional and susceptibility

artefacts resulting from the presence of active blood vessels in live animals, further in

vivo research investigations are necessary to assess the suitability of this protocol for

clinical application.

Transverse relaxation based imaging techniques have been used to assess the

microstructure of biological tissues like cartilage since late 90s. In addition to the well-

known usage of transverse relaxation based analyses, we have developed new

transverse relaxation based quantitative approaches to characterise the progression of

OA in animal models and we were the first to apply this type of characterisation for the

assessment of MD. We have also identified the collagen architecture in the femorotibial

cartilages of kangaroo using MRI for the first time. Our results reinforce the well-

known fact that quantitative assessment of biological tissues using transverse relaxation

is a non-trivial process and the analytical approach often needs to be customised

according to the nature of the tissue under examination. Therefore, three semi-

independent case studies investigated three distinctive tissue type scenarios for focusing

particularly at the structural and/or chemical composition of tissues. This thesis has

suggested new applications for transverse relaxation based assessment while it also

highlighted the limitations of this technique. Finally, it provided directions for future

works to advance these investigations further and for broadening the applicability of

transverse relaxation based quantitative evaluation of biological tissues and organs.

References

145

References

_____________________________________________________________________

1. Xia Y. Relaxation Anisotropy in Cartilage by NMR Microscopy (µMRI) at 14-Μm Resolution. Magnetic Resonance in Medicine. 1998;39(6):941-9.

2. Goebel JC, Bolbos R, Pham M, Galois L, Rengle A, Loeuille D, Netter P, Gillet P, Beuf O, Watrin-Pinzano A. In Vivo High-Resolution MRI (7T) of Femoro-Tibial Cartilage Changes in the Rat Anterior Cruciate Ligament Transection Model of Osteoarthritis: A Cross-Sectional Study. Rheumatology. 2010;49(9):1654-64.

3. Mosher TJ, Dardzinski BJ. Cartilage MRI T2 Relaxation Time Mapping: Overview and Applications. Seminars in Musculoskeletal Radiology. 2004;8(4):355-68.

4. Mosher TJ, Zhang Z, Reddy R, Boudhar S, Milestone BN, Morrison WB, Kwoh CK, Eckstein F, Witschey WRT, Borthakur A. Knee Articular Cartilage Damage in Osteoarthritis: Analysis of MR Image Biomarker Reproducibility in Acrin-Pa 4001 Multicenter Trial. Radiology. 2011;258(3):832-42.

5. Smith H, Mosher T, Dardzinski B, Collins B, Collins C, Yang Q, Schmithorst V, Smith M. Spatial Variation in Cartilage T2 of the Knee. Journal of Magnetic Resonance Imaging. 2001;14(1):50-5.

6. Xia Y, Moody JB, Burton-Wurster N, Lust G. Quantitative in Situ Correlation between Microscopic MRI and Polarized Light Microscopy Studies of Articular Cartilage. Osteoarthritis and Cartilage. 2001;9(5):393-406.

7. Xia Y, Momot KI, editors. Biophysics and Biochemistry of Cartilage by NMR and MRI. Cambridge, UK: The Royal Society of Chemistry; 2016.

8. Momot KI. Microstructural Magnetic Resonance Imaging of Articular Cartilage. Biomedical Spectroscopy and Imaging. 2012;1(1):27-37.

9. Carballido-Gamio J, Blumenkrantz G, Lynch JA, Link TM, Majumdar S. Longitudinal Analysis of MRI T2 Knee Cartilage Laminar Organization in a Subset of Patients from the Osteoarthritis Initiative. Magnetic Resonance in Medicine. 2010;63(2):465-72.

10. Chou MC, Tsai PH, Huang GS, Lee HS, Lee CH, Lin MH, Lin CY, Chung HW. Correlation between the Mr T2 Value at 4.7 T and Relative Water Content in Articular Cartilage in Experimental Osteoarthritis Induced by Acl Transection. Osteoarthritis and Cartilage. 2009;17(4):441-7.

11. Huang GS, Lee HS, Chou MC, Shih YYI, Tsai PH, Lin MH, Lin CY, Lee CH, Chung HW. Quantitative Mr T2 Measurement of Articular Cartilage to Assess the Treatment Effect of Intra-Articular Hyaluronic Acid Injection on Experimental Osteoarthritis Induced by Aclx. Osteoarthritis and Cartilage. 2010;18(1):54-60.

12. Ali TS, Bjarnason TA, Senger DL, Dunn JF, Joseph JT, Mitchell JR. Quantitativet2: Interactive Quantitative T2 MRI Witnessed in Mouse Glioblastoma. Journal of Medical Imaging. 2015;2(3):036002.

13. Dortch RD, Yankeelov TE, Yue Z, Quarles CC, Gore JC, Does MD. Evidence of Multiexponential T2 in Rat Glioblastoma. NMR in Biomedicine. 2009;22(6):609-18.

References

146

14. Dunn TC, Lu Y, Jin H, Ries MD, Majumdar S. T2 Relaxation Time of Cartilage at Mr Imaging: Comparison with Severity of Knee Osteoarthritis. Radiology. 2004;232(2):592-8.

15. Hattingen E, Jurcoane A, Daneshvar K, Pilatus U, Mittelbronn M, Steinbach JP, Bähr O. Quantitative T2 Mapping of Recurrent Glioblastoma under Bevacizumab Improves Monitoring for Non-Enhancing Tumor Progression and Predicts Overall Survival. Neuro-Oncology. 2013;15(10):1395-404.

16. Poon CS, Henkelman RM. Practical T2 Quantitation for Clinical Applications. Journal of Magnetic Resonance Imaging. 1992;2(5):541-53.

17. Whittall KP, Mackay AL, Graeb DA, Nugent RA, Li DKB, Paty DW. In Vivo Measurement of T2 Distributions and Water Contents in Normal Human Brain. Magnetic Resonance in Medicine. 1997;37(1):34-43.

18. Chang KJ, Jara H. Applications of Quantitative T1, T2, and Proton Density to Diagnosis. Applied radiology. 2005;34.

19. Murphy WA, Totty WG, Carroll JE. MRI of Normal and Pathologic Skeletal Muscle. American Journal of Roentgenology. 1986;146(3):565-74.

20. May DA, Disler DG, Jones EA, Balkissoon AA, Manaster BJ. Abnormal Signal Intensity in Skeletal Muscle at Mr Imaging: Patterns, Pearls, and Pitfalls. RadioGraphics. 2000;20(suppl_1):S295-S315.

21. Amano Y, Tachi M, Tani H, Mizuno K, Kobayashi Y, Kumita S. T2-Weighted Cardiac Magnetic Resonance Imaging of Edema in Myocardial Diseases. The Scientific World Journal. 2012;2012.

22. Westra C, Dialani V, Mehta TS, Eisenberg RL. Using T2-Weighted Sequences to More Accurately Characterize Breast Masses Seen on MRI. American Journal of Roentgenology. 2014;202(3):W183-W90.

23. Kuhl CK, Klaschik S, Mielcarek P, Gieseke J, Wardelmann E, Schild HH. Do T2-Weighted Pulse Sequences Help with the Differential Diagnosis of Enhancing Lesions in Dynamic Breast MRI? Journal of Magnetic Resonance Imaging. 1999;9(2):187-96.

24. MacKay AL, Laule C. Magnetic Resonance of Myelin Water: An in Vivo Marker for Myelin. Brain Plasticity. 2016;2(1):71-91.

25. Laule C, Vavasour IM, Mädler B, Kolind SH, Sirrs SM, Brief EE, Traboulsee AL, Moore GRW, Li DKB, MacKay AL. Mr Evidence of Long T2 Water in Pathological White Matter. Journal of Magnetic Resonance Imaging. 2007;26(4):1117-21.

26. Radbruch A, Lutz K, Wiestler B, Bäumer P, Heiland S, Wick W, Bendszus M. Relevance of T2 Signal Changes in the Assessment of Progression of Glioblastoma According to the Response Assessment in Neurooncology Criteria. Neuro-Oncology. 2012;14(2):222-9.

27. Lupu M, Thomas CD, Mispelter J. Retrieving Accurate Relaxometric Information from Low Signal-to-Noise Ratio 23na MRI Performed in Vivo. Comptes Rendus Chimie. 2008;11(4–5):515-23.

28. Mosher TJ, Smith H, Dardzinski BJ, Schmithorst VJ, Smith MB. Mr Imaging and T2 Mapping of Femoral Cartilage: In Vivo Determination of the Magic Angle Effect. American Journal of Roentgenology. 2001;177(3):665-9.

29. Xia Y. Magic-Angle Effect in Magnetic Resonance Imaging of Articular Cartilage: A Review. Investigative Radiology. 2000;35(10):602-21.

References

147

30. Xia Y. Relaxation Anisotropy as a Possible Marker for Macromolecular Orientations in Articular Cartilage. Spatially Resolved Magnetic Resonance. Weinheim: Wiley-VCH; 1998. p. 351-62.

31. Xia Y. MRI of Articular Cartilage at Microscopic Resolution. Bone & Joint Research. 2013;2(1):9-17.

32. Laule C, Kozlowski P, Leung E, Li DKB, MacKay AL, Moore GRW. Myelin Water Imaging of Multiple Sclerosis at 7 t: Correlations with Histopathology. NeuroImage. 2008;40(4):1575-80.

33. MacKay A, Laule C, Vavasour I, Bjarnason T, Kolind S, Mädler B. Insights into Brain Microstructure from the T2 Distribution. Magnetic Resonance Imaging. 2006;24(4):515-25.

34. Whittall KP, MacKay AL, Li DKB, Vavasour IM, Jones CK, Paty DW. Normal-Appearing White Matter in Multiple Sclerosis Has Heterogeneous, Diffusely Prolonged T2. Magnetic Resonance in Medicine. 2002;47(2):403-8.

35. Guimaraes AR, Siqueira L, Uppal R, Alford J, Fuchs BC, Yamada S, Tanabe K, Chung RT, Lauwers G, Chew ML, Boland GW, Sahani DV, Vangel M, Hahn PF, Caravan P. T2 Relaxation Time Is Related to Liver Fibrosis Severity. Quantitative imaging in medicine and surgery. 2016;6(2):103-14.

36. Papakonstantinou O, Kostaridou S, Maris T, Gouliamos A, Premetis E, Kouloulias V, Nakopoulou L, Kattamis C. Quantification of Liver Iron Overload by T2 Quantitative Magnetic Resonance Imaging in Thalassemia: Impact of Chronic Hepatitis C on Measurements. Journal of pediatric hematology/oncology. 1999;21(2):142-8.

37. Storås TH, Gjesdal K-I, Gadmar ØB, Geitung JT, Kløw N-E. Prostate Magnetic Resonance Imaging: Multiexponential T2 Decay in Prostate Tissue. Journal of Magnetic Resonance Imaging. 2008;28(5):1166-72.

38. Zia MI, Ghugre NR, Connelly KA, Strauss BH, Sparkes JD, Dick AJ, Wright GA. Characterizing Myocardial Edema and Hemorrhage Using Quantitative T2 and T2* Mapping at Multiple Time Intervals Post St-Segment Elevation Myocardial Infarction. Circulation: Cardiovascular Imaging. 2012;5(5):566-72.

39. Kim HK, Serai S, Lindquist D, Merrow AC, Horn PS, Kim DH, Wong BL. Quantitative Skeletal Muscle MRI: Part 2, Mr Spectroscopy and T2 Relaxation Time Mapping—Comparison between Boys with Duchenne Muscular Dystrophy and Healthy Boys. American Journal of Roentgenology. 2015;205(2):W216-W23.

40. Maillard S, Jones R, Owens C, Pilkington C, Woo P, Wedderburn L, Murray K. Quantitative Assessment of MRI T 2 Relaxation Time of Thigh Muscles in Juvenile Dermatomyositis. Rheumatology. 2004;43(5):603-8.

41. Meiboom S, Gill D. Modified Spin‐Echo Method for Measuring Nuclear Relaxation Times. Review of Scientific Instruments. 1958;29(8):688-91.

42. Eyre DR, Weis MA, Wu J-J. Advances in Collagen Cross-Link Analysis. Methods (San Diego, Calif). 2008;45(1):65-74.

43. Xia Y, Moody JB, Alhadlaq H. Orientational Dependence of T2 Relaxation in Articular Cartilage: A Microscopic MRI (µMRI) Study. Magnetic Resonance in Medicine. 2002;48(3):460-9.

44. Vuorio E, De Crombrugghe B. The Family of Collagen Genes. Annual review of biochemistry. 1990;59(1):837-72.

References

148

45. Gordon MK, Hahn RA. Collagens. Cell and tissue research. 2010;339(1):247-57.

46. Pearle AD, Warren RF, Rodeo SA. Basic Science of Articular Cartilage and Osteoarthritis. Clinics in Sports Medicine. 2005;24(1):1-12.

47. Huber M, Trattnig S, Lintner F. Anatomy, Biochemistry, and Physiology of Articular Cartilage. Investigative Radiology. 2000;35(10):573-80.

48. Mow V, Hung C. Biomechanics of Articular Cartilage and Determination of Material Properties. Medicine & Science in Sports & Excersize. 2008:40(2):193–99.

49. Mow VC. Structure and Function of Articular Cartilage and Meniscus. Basic orthopaedic biomechanics. 1997:31-58.

50. Korhonen RK, Laasanen MS, Töyräs J, Lappalainen R, Helminen HJ, Jurvelin JS. Fibril Reinforced Poroelastic Model Predicts Specifically Mechanical Behavior of Normal, Proteoglycan Depleted and Collagen Degraded Articular Cartilage. Journal of Biomechanics. 2003;36(9):1373-9.

51. Laasanen M, Töyräs J, Korhonen R, Rieppo J, Saarakkala S, Nieminen M, Hirvonen J, Jurvelin J. Biomechanical Properties of Knee Articular Cartilage. Biorheology. 2003;40(1):133-40.

52. He B, Wu JP, Kirk TB, Carrino JA, Xiang C, Xu J. High-Resolution Measurements of the Multilayer Ultra-Structure of Articular Cartilage and Their Translational Potential. Arthritis Research & Therapy. 2014;16(2):205.

53. Changoor A, Nelea M, Methot S, Tran-Khanh N, Chevrier A, Restrepo A, Shive M, Hoemann C, Buschmann M. Structural Characteristics of the Collagen Network in Human Normal, Degraded and Repair Articular Cartilages Observed in Polarized Light and Scanning Electron Microscopies. Osteoarthritis and Cartilage. 2011;19(12):1458-68.

54. Hughes L, Archer C, Ap Gwynn I. The Ultrastructure of Mouse Articular Cartilage: Collagen Orientation and Implications for Tissue Functionality. A Polarised Light and Scanning Electron Microscope Study and Review. Eur Cell Mater. 2005;9(68):e84.

55. Gottardi R, Hansen U, Raiteri R, Loparic M, Düggelin M, Mathys D, Friederich NF, Bruckner P, Stolz M. Supramolecular Organization of Collagen Fibrils in Healthy and Osteoarthritic Human Knee and Hip Joint Cartilage. PLOS ONE. 2016;11(10):e0163552.

56. Jeffery A, Blunn G, Archer C, Bentley G. Three-Dimensional Collagen Architecture in Bovine Articular Cartilage. The Journal of bone and joint surgery British volume. 1991;73(5):795-801.

57. de Visser SK, Bowden JC, Wentrup-Byrne E, Rintoul L, Bostrom T, Pope JM, Momot KI. Anisotropy of Collagen Fibre Alignment in Bovine Cartilage: Comparison of Polarised Light Microscopy and Spatially Resolved Diffusion-Tensor Measurements. Osteoarthritis and Cartilage. 2008;16(6):689-97.

58. Lehner KB, Rechl HP, Gmeinwieser JK, Heuck AF, Lukas HP, Kohl HP. Structure, Function, and Degeneration of Bovine Hyaline Cartilage: Assessment with Mr Imaging in Vitro. Radiology. 1989;170(2):495-9.

59. Hayes C, Sawyer R, Conway W. Patellar Cartilage Lesions: In Vitro Detection and Staging with Mr Imaging and Pathologic Correlation. Radiology. 1990;176(2):479-83.

60. Modl JM, Sether LA, Haughton VM, Kneeland JB. Articular Cartilage: Correlation of Histologic Zones with Signal Intensity at Mr Imaging. Radiology. 1991;181(3):853-5.

References

149

61. Hayes CW, Parellada JA. The Magic Angle Effect in Musculoskeletal Mr Imaging. Topics in magnetic resonance imaging: TMRI. 1996;8(1):51-6.

62. Xia Y, Farquhar T, Burton‐Wurster N, Ray E, Jelinski LW. Diffusion and Relaxation Mapping of Cartilage‐Bone Plugs and Excised Disks Using Microscopic Magnetic Resonance Imaging. Magnetic Resonance in Medicine. 1994;31(3):273-82.

63. Ludescher B, Effelsberg J, Martirosian P, Steidle G, Markert B, Claussen C, Schick F. T2‐and Diffusion‐Maps Reveal Diurnal Changes of Intervertebral Disc Composition: An in Vivo MRI Study at 1.5 Tesla. Journal of Magnetic Resonance Imaging. 2008;28(1):252-7.

64. Momot KI, Pope JM, Wellard RM. Anisotropy of Spin Relaxation of Water Protons in Cartilage and Tendon. NMR in Biomedicine. 2010;23(3):313-24.

65. Seidel T, Hammer N, Garnov N, Schneider G, Steinke H. An Algorithm for the Calculation of Three‐Dimensional Collagen Fiber Orientation in Ligaments Using Angle‐Sensitive MRI. Magnetic Resonance in Medicine. 2013;69(6):1594-602.

66. Zheng S, Xia Y. The Collagen Fibril Structure in the Superficial Zone of Articular Cartilage by µMRI. Osteoarthritis and Cartilage. 2009;17(11):1519-28.

67. Zheng S, Xia Y, Badar F. Further Studies on the Anisotropic Distribution of Collagen in Articular Cartilage by µMRI. Magnetic Resonance in Medicine. 2011;65(3):656-63.

68. Dardzinski BJ, Mosher TJ, Li S, Slyke MAV, Smith MB. Spatial Variation of T2 in Human Articular Cartilage. Radiology. 1997;205(2):546-50.

69. Clarke IC. Articular Cartilage: A Review and Scanning Electron Microscope Study. Ii. The Territorial Fibrillar Architecture. Journal of Anatomy. 1974;118(Pt 2):261-80.

70. Lee JH, Badar F, Kahn D, Matyas J, Qu X, Xia Y. Loading-Induced Changes on Topographical Distributions of the Zonal Properties of Osteoarthritic Tibial Cartilage–a Study by Magnetic Resonance Imaging at Microscopic Resolution. Journal of Biomechanics. 2015;48(13):3634-42.

71. Lee JH, Badar F, Kahn D, Matyas J, Qu X, Chen CT, Xia Y. Topographical Variations of the Strain-Dependent Zonal Properties of Tibial Articular Cartilage by Microscopic MRI. Connective tissue research. 2014;55(3):205-16.

72. Darling EM, Hu JC, Athanasiou KA. Zonal and Topographical Differences in Articular Cartilage Gene Expression. Journal of Orthopaedic Research. 2004;22(6):1182-7.

73. Huo CW, Chew GL, Britt KL, Ingman WV, Henderson MA, Hopper JL, Thompson EW. Mammographic Density—a Review on the Current Understanding of Its Association with Breast Cancer. Breast Cancer Research and Treatment. 2014;144(3):479-502.

74. Liberman L, Menell JH. Breast Imaging Reporting and Data System (Bi-Rads). Radiologic Clinics. 2002;40(3):409-30.

75. Yaffe MJ. Mammographic Density. Measurement of Mammographic Density. Breast Cancer Research. 2008;10(3):209.

76. Mamisch TC, Trattnig S, Quirbach S, Marlovits S, White LM, Welsch GH. Quantitative T2 Mapping of Knee Cartilage: Differentiation of Healthy Control Cartilage and Cartilage Repair Tissue in the Knee with Unloading—Initial Results. Radiology. 2010;254(3):818-26.

77. Raya JG, Ferizi U. Quantitative MRI for Detection of Cartilage Damage. In: Xia Y, Momot KI, editors. Biophysics and Biochemistry of Cartilage by NMR and MRI: The Royal Society of Chemistry; 2016. p. 575-627.

References

150

78. Kuhl CK. MRI of Breast Tumors. European Radiology. 2000;10(1):46-58.

79. Petridou E, Kibiro M, Gladwell C, Malcolm P, Toms A, Juette A, Borga M, Dahlqvist Leinhard O, Romu T, Kasmai B, Denton E. Breast Fat Volume Measurement Using Wide-Bore 3 t MRI: Comparison of Traditional Mammographic Density Evaluation with MRI Density Measurements Using Automatic Segmentation. Clinical Radiology. 2017;72(7):565-72.

80. Australia CC. Breast Cancer Screening 2016 [Available from: https://www.cancer.org.au/about-cancer/early-detection/screening-programs/breast-cancer-screening.html].

81. Lehman CD, Gatsonis C, Kuhl CK, Hendrick RE, Pisano ED, Hanna L, Peacock S, Smazal SF, Maki DD, Julian TB, DePeri ER, Bluemke DA, Schnall MD. MRI Evaluation of the Contralateral Breast in Women with Recently Diagnosed Breast Cancer. New England Journal of Medicine. 2007;356(13):1295-303.

82. Kriege M, Brekelmans CTM, Boetes C, Besnard PE, Zonderland HM, Obdeijn IM, Manoliu RA, Kok T, Peterse H, Tilanus-Linthorst MMA, Muller SH, Meijer S, Oosterwijk JC, Beex LVAM, Tollenaar RAEM, de Koning HJ, Rutgers EJT, Klijn JGM. Efficacy of MRI and Mammography for Breast-Cancer Screening in Women with a Familial or Genetic Predisposition. New England Journal of Medicine. 2004;351(5):427-37.

83. Harms SE. MRI in Breast Cancer Diagnosis and Treatment. Current Problems in Diagnostic Radiology. 1996;25(6):192-215.

84. Klifa C, Carballido-Gamio J, Wilmes L, Laprie A, Lobo C, DeMicco E, Watkins M, Shepherd J, Gibbs J, Hylton N, editors. Quantification of Breast Tissue Index from Mr Data Using Fuzzy Clustering. The 26th Annual International Conference of the IEEE Engineering in Medicine and Biology Society; 2004 1-5 Sept. 2004.

85. Klifa C, Carballido-Gamio J, Wilmes L, Laprie A, Shepherd J, Gibbs J, Fan B, Noworolski S, Hylton N. Magnetic Resonance Imaging for Secondary Assessment of Breast Density in a High-Risk Cohort. Magnetic Resonance Imaging. 2010;28(1):8-15.

86. Khazen M, Warren RML, Boggis CRM, Bryant EC, Reed S, Warsi I, Pointon LJ, Kwan-Lim GE, Thompson D, Eeles R, Easton D, Evans DG, Leach MO. A Pilot Study of Compositional Analysis of the Breast and Estimation of Breast Mammographic Density Using 3d T1-Weighted Mr Imaging. Cancer Epidemiology, Biomarkers & Prevention. 2008;17(9):2268-74.

87. Wei J, Chan H-P, Helvie MA, Roubidoux MA, Sahiner B, Hadjiiski LM, Zhou C, Paquerault S, Chenevert T, Goodsitt MM. Correlation between Mammographic Density and Volumetric Fibroglandular Tissue Estimated on Breast Mr Images. Medical Physics. 2004;31(4):933-42.

88. Australian Institute of Health and Welfare: Cancer in Australia 2017. Canberra: AIHW, 2017.

89. Blümich B, Blümler P, Eidmann G, Guthausen A, Haken R, Schmitz U, Saito K, Zimmer G. The NMR-MOUSE: Construction, Excitation, and Applications. Magnetic Resonance Imaging. 1998;16(5):479-84.

90. Eidmann GS, R; Blumler, P; Blumich, B. The NMR MOUSE, a Mobile Universal Surface Explorer Journal of Magnetic Resonance. 1996;122(Series A):104-9.

91. Casanova F, Perlo J, Blümich B. Single-Sided NMR. In: Casanova F, Perlo J, Blümich B, editors. Single-Sided NMR. Berlin, Heidelberg: Springer Berlin Heidelberg; 2011. p. 1-10.

92. Krüger M, Schwarz A, Blümich B. Investigations of Silicone Breast Implants with the NMR-MOUSE. Magnetic Resonance Imaging. 2007;25(2):215-8.

https://www.cancer.org.au/about-cancer/early-detection/screening-programs/breast-cancer-screening.html

https://www.cancer.org.au/about-cancer/early-detection/screening-programs/breast-cancer-screening.html

References

151

93. Landeghem MV, Danieli E, Perlo J, Blümich B, Casanova F. Low-Gradient Single-Sided NMR Sensor for One-Shot Profiling of Human Skin. Journal of Magnetic Resonance. 2012;215:74-84.

94. Navon G, Eliav U, Demco DE, Blümich B. Study of Order and Dynamic Processes in Tendon by NMR and MRI. Journal of Magnetic Resonance Imaging. 2007;25(2):362-80.

95. Rössler E, Mattea C, Stapf S. Feasibility of High-Resolution One-Dimensional Relaxation Imaging at Low Magnetic Field Using a Single-Sided NMR Scanner Applied to Articular Cartilage. Journal of Magnetic Resonance. 2015;251:43-51.

96. Brizi L, Barbieri M, Baruffaldi F, Bortolotti V, Fersini C, Liu H, Nogueira d'Eurydice M, Obruchkov S, Zong F, Galvosas P, Fantazzini P. Bone Volume–to–Total Volume Ratio Measured in Trabecular Bone by Single-Sided NMR Devices. Magnetic Resonance in Medicine. 2018;79(1):501-10.

97. Tourell MC, Ali TS, Hugo HJ, Pyke C, Yang S, Lloyd T, Thompson EW, Momot KI. T1-Based Sensing of Mammographic Density Using Single-Sided Portable NMR. Magnetic Resonance in Medicine. 2018;80(3):1243-51.

98. Roemer F, Frobell R, Hunter D, Crema M, Fischer W, Bohndorf K, Guermazi A. MRI-Detected Subchondral Bone Marrow Signal Alterations of the Knee Joint: Terminology, Imaging Appearance, Relevance and Radiological Differential Diagnosis. Rheumatology. 2009;17(9):1115-31.

99. Roemer F, Hunter D, Guermazi A. MRI-Based Semiquantitative Assessment of Subchondral Bone Marrow Lesion in Osteoarthritis Research. Osteoarthritis Cartilage. 2009;17(3):414-15.

100. Shapiro LM, McWalter EJ, Son M-S, Levenston M, Hargreaves BA, Gold GE. Mechanisms of Osteoarthritis in the Knee: MR Imaging Appearance. Journal of Magnetic Resonance Imaging. 2014;39(6):1346-56.

101. Conaghan P. Is MRI Useful in Osteoarthritis? Best Practice & Research Clinical Rheumatology. 2006;20(1):57-68.

102. Conaghan PG, Felson D, Gold G, Lohmander S, Totterman S, Altman R. MRI and Non-Cartilaginous Structures in Knee Osteoarthritis. Osteoarthritis and Cartilage. 2006;14(Supplement 1):87-94.

103. Roemer FW, Crema MD, Trattnig S, Guermazi A. Advances in Imaging of Osteoarthritis and Cartilage. Radiology. 2011;260(2):332-54.

104. Berenbaum F. The Quest for the Holy Grail: A Disease-Modifying Osteoarthritis Drug. Arthritis Research & Therapy. 2007;9(6):1-2.

105. Dijkgraaf LC, de Bont LGM, Boering G, Liem RSB. The Structure, Biochemistry, and Metabolism of Osteoarthritic Cartilage: A Review of the Literature. Journal of Oral and Maxillofacial Surgery. 1995;53(10):1182-92.

106. Stahl R, Blumenkrantz G, Carballido-Gamio J, Zhao S, Munoz T, Hellio Le Graverand-Gastineau M, Li X, Majumdar S, Link T. MRI-Derived T2 Relaxation Times and Cartilage Morphometry of the Tibio-Femoral Joint in Subjects with and without Osteoarthritis During a 1-Year Follow-Up. Osteoarthritis and Cartilage. 2007;15(11):1225-34.

107. Nissi M, Rieppo J, Toyras J, Laasanen M, Kiviranta I, Jurvelin J, Nieminen M. T2 Relaxation Time Mapping Reveals Age- and Species-Related Diversity of Collagen Network Architecture in Articular Cartilage. Osteoarthritis and Cartilage. 2006;14(12):1265-71.

References

152

108. Crema MD, Roemer FW, Marra MD, Burstein D, Gold GE, Eckstein F, Baum T, Mosher TJ, Carrino JA, Guermazi A. Articular Cartilage in the Knee: Current Mr Imaging Techniques and Applications in Clinical Practice and Research. RadioGraphics. 2011;31(1):37-61.

109. Ronga M, Angeretti G, Ferraro S, De Falco G, Genovese EA, Cherubino P. Imaging of Articular Cartilage: Current Concepts. Joints. 2014;2(3):137-40.

110. Calvo E, Palacios I, Delgado E, Ruiz-Cabello J, Hernández P, Sánchez-Pernaute O, Egido J, Herrero-Beaumont G. High-Resolution MRI Detects Cartilage Swelling at the Early Stages of Experimental Osteoarthritis. Osteoarthritis and Cartilage. 2001;9(5):463-72.

111. Calvo E, Palacios I, Delgado E, Sánchez-Pernaute O, Largo R, Egido J, Herrero-Beaumont G. Histopathological Correlation of Cartilage Swelling Detected by Magnetic Resonance Imaging in Early Experimental Osteoarthritis. Osteoarthritis and Cartilage. 2004;12(11):878-86.

112. Hayashi D, Roemer W, Guermazi A. Osteoarthritis Year 2011 in Review: Imaging in Oa - a Radiologist's Perspective. Osteoarthritis and Cartilage. 2012;20(3):207-14.

113. Tan Y, Conaghan G. Insights into Osteoarthritis from MRI. International Journal of Rheumatic Diseases. 2012 Feb;15(1):1-7.

114. Felson D, McLaughlin S, Goggins J, Lavalley M, Gale M, Totterman S, Li W, Hill C, Gale D. Bone Marrow Edema and Its Relationship to Pregression of Knee Osteoarthritis. Annals of Internal Medicine. 2003;139(5):330-6.

115. Hunter D, Zhang Y, Niu J, Goggins J, Amin S, LaValley M, Guermazi A, Genant H, Gale D, Felson D. Increase in Bone Marrow Lesions Associated with Cartilage Loss. Arthritis and Rheumatism. 2006;54(5):1529-35.

116. Ha T, Li K, Beaulieu C, Bergman G, Ch'en I, Eller D, Cheung L, Herfkens R. Anterior Cruciate Ligament Injury: Fast Spin-Echo Mr Imaging with Arthroscopic Correlation in 217 Examinations. Americal Journal of Roentgenology. 1998;170(5):1215-19.

117. Pham X, Monteiro I, Judet O, Sissakian J, Plantin P, Aegerter P, Le Parc J. Magnetic Resonance Imaging Changes in Periarticular Soft Tissues During Flares of Medial Compartment Knee Osteoarthritis. Preliminary Study in 10 Patients. Revue de Rhumatisme. 1999;66(7-9):398-403.

118. Hill C, Seo G, Gale D, Totterman S, Gale M, Felson D. Cruciate Ligament Integrity in Osteoarthritis of the Knee. Arthritis Rheumatology. 2005;52(3):794-99.

119. Blackmon G, Major N, Helms C. Comparison of Fast Spin-Echo Versus Conventional Spin-Echo MRI for Evaluating Meniscal Tears. American Journal of Roentgenology. 2005:184:1740-43.

120. Østergaard M, Stoltenberg M, Lovgreen-Nielsen P, Volck B, Jensen C, Lorenzen I. Magnetic Resonance Imaging-Determined Synovial Membrane and Joint Effusion Volumes in Rheumatoid Arthritis and Osteoarthritis. Comparison with the Macroscopic and Microscopic Appearance of Sinovium. Arthritis Rheumatology. 1997;40(10):1856-67.

121. Ali T. Real-Time Voxel-Based Quantitative T2 MRI Analysis and Applications [M. Sc. Thesis]. Calgary: University of Calgary; 2010.

122. Bjarnason TA, Mitchell JR. Analyzennls: Magnetic Resonance Multiexponential Decay Image Analysis. Journal of Magnetic Resonance. 2010;206(2):200-4.

References

153

123. Blasiak B, Tomanek B, Abulrob A, Iqbal U, Stanimirovic D, Albaghdadi H, Foniok T, Lun X, Forsyth P, Sutherland GR. Detection of T2 Changes in an Early Mouse Brain Tumor. Magnetic Resonance Imaging. 2010;28(6):784-9.

124. Ali TS, Thibbotuwawa N, Gu Y, Momot KI. MRI Magic-Angle Effect in Femorotibial Cartilages of the Red Kangaroo. Magnetic Resonance Imaging. 2017;43:66-73.

125. Ali TS, Tourell MC, Hugo HJ, Pyke C, Yang S, Lloyd T, Thompson EW, Momot KI. Transverse Relaxation-Based Assessment of Mammographic Density and Breast Tissue Composition by Single-Sided Portable NMR. Magnetic Resonance in Medicine. 2019;82(3):1199-213.

126. Hanson L. Is Quantum Mechanics Necessary for Understanding Magnetic Resonance? Concepts in Magnetic Resonance. 2008;32A(5):329-40.

127. Liang Z, Lauterbur P. Principles of Magnetic Resonance Imaging: A Signal Processing Perspective. New York: Wiley-IEEE Press 2000.

128. Bracewell RN. The Fourier Transform and Its Applications. Second edition ed. New York: McGraw-Hill; 1978.

129. Farrar TC, Becker ED. Pulse and Fourier Transform NMR: Introduction to Theory and Methods: Elsevier; 2012.

130. Keeler J. Understanding NMR Spectroscopy: John Wiley & Sons; 2011.

131. Calero C, Martí J, Guàrdia E. 1h Nuclear Spin Relaxation of Liquid Water from Molecular Dynamics Simulations. The Journal of Physical Chemistry B. 2015;119(5):1966-73.

132. Derome AE. Modern NMR Techniques for Chemistry Research: Elsevier; 2013.

133. Richter W, Warren WS. Intermolecular Multiple Quantum Coherences in Liquids. Concepts in Magnetic Resonance. 2000;12(6):396-409.

134. Cavanagh J, Fairbrother WJ, Palmer III AG, Skelton NJ. Protein NMR Spectroscopy: Principles and Practice: Elsevier; 1995.

135. Abragam A. Principles of Nuclear Magnetism (International Series of Monographs on Physics). Clarendon Press, Oxford; 1961.

136. Umamaheswar D, Ravinder R, D. PS, H. KJ, Bruce KJ, S. LJ. T1ρ‐Relaxation in Articular Cartilage: Effects of Enzymatic Degradation. Magnetic Resonance in Medicine. 1997;38(6):863-7.

137. Henkelman R, Huang X, Xiang QS, Stanisz GJ, Swanson SD, Bronskill MJ. Quantitative Interpretation of Magnetization Transfer. Magnetic Resonance in Medicine. 1993;29(6):759-66.

138. Papoulis A. Signal Analysis. New York, United States: McGraw-Hill Education - Europe; 1977.

139. Carr H, Purcell E. Effects of Diffussion on Free Precession in Nuclear Magnetic Resonance Experiments. Physical Review. 1954;94(3):630-38.

140. Hahn E. Spin Echoes. Physical Review. 1950;80(4):580-94.

141. Hahn EL. Nuclear Induction Due to Free Larmor Precession. Physical Review. 1950;77(2):297-8.

142. Gudbjartsson H, Patz S. The Rician Distribution of Noisy MRI Data. Magnetic Resonance in Medicine. 1995;34(6):910-4.

References

154

143. Lawson CA, Hanson R. Solving Least Squares Problems. Englewood Cliffs, NJ: Prentice-Hall; 1974.

144. Provencher SW. Inverse Problems in Polymer Characterization: Direct Analysis of Polydispersity with Photon Correlation Spectroscopy. Die Makromolekulare Chemie. 1979;180(1):201-9.

145. Provencher SW. A Constrained Regularization Method for Inverting Data Represented by Linear Algebraic or Integral Equations. Computer Physics Communications. 1982;27(3):213-27.

146. Provencher SW, editor A General-Purpose Constrained Regularization Method for Inverting Photon Correlation Data1983; Berlin, Heidelberg: Springer Berlin Heidelberg.

147. Borgia GC, Brown RJS, Fantazzini P. Uniform-Penalty Inversion of Multiexponential Decay Data. Journal of Magnetic Resonance. 1998;132(1):65-77.

148. Venkataramanan L, Yi-Qiao S, Hurlimann MD. Solving Fredholm Integrals of the First Kind with Tensor Product Structure in 2 and 2.5 Dimensions. IEEE Transactions on Signal Processing. 2002;50(5):1017-26.

149. Oztop MH, Bansal H, Takhar P, McCarthy KL, McCarthy MJ. Using Multi-Slice-Multi-Echo Images with NMR Relaxometry to Assess Water and Fat Distribution in Coated Chicken Nuggets. LWT - Food Science and Technology. 2014;55(2):690-4.

150. Bjarnason TA, McCreary CR, Dunn JF, Mitchell JR. Quantitative T2 Analysis: The Effects of Noise, Regularization, and Multivoxel Approaches. Magnetic Resonance in Medicine. 2010;63(1):212-7.

151. Laule C, Kolind SH, Bjarnason TA, Li DKB, MacKay AL. In Vivo Multiecho T2 Relaxation Measurements Using Variable Tr to Decrease Scan Time. Magnetic Resonance Imaging. 2007;25(6):834-9.

152. Meyers SM, Vavasour IM, Mädler B, Harris T, Fu E, Li DKB, Traboulsee AL, MacKay AL, Laule C. Multicenter Measurements of Myelin Water Fraction and Geometric Mean T2: Intra- and Intersite Reproducibility. Journal of Magnetic Resonance Imaging. 2013;38(6):1445-53.

153. Brown M, Semelka R. MRI: Basic Principles and Applications. New Jersey: John Wiley & Sons, Inc.; 2003.

154. Watrin-Pinzano A, Ruaud JP, Cheli Y, Gonord P, Grossin L, Gillet P, Blum A, Payan E, Olivier P, Guillot G, Netter P, Loeuille D. T2 Mapping: An Efficient Mr Quantitative Technique to Evaluate Spontaneous Cartilage Repair in Rat Patella. Osteoarthritis and Cartilage. 2004;12(3):191-200.

155. Spandonis Y, Heese FP, Hall LD. High Resolution MRI Relaxation Measurements of Water in the Articular Cartilage of the Meniscectomized Rat Knee at 4.7 T. Magnetic Resonance Imaging. 2004;22(7):943-51.

156. Chou MC, Tsai PH, Huang GS, Lee HS, Lee CH, Lin MH, Lin CY, Chung HW. Correlation between the MR T2 Value at 4.7 T and Relative Water Content in Articular Cartilage in Experimental Osteoarthritis Induced by Acl Transection. Osteoarthritis and cartilage / OARS, Osteoarthritis Research Society. 2009;17(4):441-7.

157. Batiste DL, Kirkley A, Laverty S, Thain LMF, Spouge AR, Holdsworth DW. Ex Vivo Characterization of Articular Cartilage and Bone Lesions in a Rabbit Acl Transection Model of Osteoarthritis Using MRI and Micro-Ct. Osteoarthritis and Cartilage. 2004;12(12):986-96.

References

155

158. Shajan G, Hoffmann J, Balla DZ, Deelchand DK, Scheffler K, Pohmann R. Rat Brain MRI at 16.4 T Using a Capacitively Tunable Patch Antenna in Combination with a Receive Array. NMR in Biomedicine. 2012;25(10):1170-6.

159. Tadimalla S, Momot KI. Effect of Partial H2O-D2O Replacement on the Anisotropy of Transverse Proton Spin Relaxation in Bovine Articular Cartilage. PLOS ONE. 2015;9(12):e115288.

160. Brown RJ. Nuclear Magnetism Logging at the Coyote Institute. Magnetic Resonance Imaging. 1996;14(7-8):811-7.

161. Woessner DE. The Early Days of NMR in the Southwest. Concepts in Magnetic Resonance: An Educational Journal. 2001;13(2):77-102.

162. Ellis DV, Singer JM. Well Logging for Earth Scientists: Springer; 2007.

163. Maciel GE. NMR in Industrial Process Control and Quality Control. In: Maciel GE, editor. Nuclear Magnetic Resonance in Modern Technology. Dordrecht: Springer Netherlands; 1994. p. 225-75.

164. Pearson RM, Ream LR, Job C, Adams J. Use of NMR Spectrometers for Process Control. Cereal Foods World. 1987;32:822-6.

165. Pearson RM. Using Industrial Magnetic Resonance for Moisture Monitoring. Cereal Foods World. 1987;32(9):653.

166. Persyn G, Rollwitz W. Transient NMR Quantitative Measurements. Journal of the American Oil Chemists Society. 1971;48(2):67-9.

167. Coates GR, Xiao L, Prammer MG. NMR Logging: Principles and Applications: Haliburton Energy Services Houston; 1999.

168. Ernst RR, Bodenhausen G, Wokaun A. Principles of Nuclear Magnetic Resonance in One and Two Dimensions: Clarendon Press; 1987.

169. Blumich B. NMR Imaging of Materials: OUP Oxford; 2000.

170. Hahn E. Lille Conference Talk. Journal of Magnetic Resonance. 2006;179(1):9-10.

171. Carr HY, Purcell EM. Effects of Diffusion on Free Precession in Nuclear Magnetic Resonance Experiments. Physical Review. 1954;94(3):630-8.

172. Meiboom S, Gill D. Modified Spin-Echo Method for Measuring Nuclear Relaxation Times. Review of Scientific Instruments. 1958;29(8):688-91.

173. Wellard R, Ravasio J-P, Guesne S, Bell C, Oloyede A, Tevelen G, Pope J, Momot K. Simultaneous Magnetic Resonance Imaging and Consolidation Measurement of Articular Cartilage. Sensors. 2014;14(5):7940.

174. Fox A, Bedi A, Rodeo S. The Basic Science of Articular Cartilage: Structure, Composition and Function. Sports Health: A Multidisciplinary Approach. 2009:461-68.

175. Kheir E, Shaw D. Hyaline Articular Cartilage. Orthopaedics and Trauma. 2009;23(6):450-5.

176. Minns RJ, Steven FS. The Collagen Fibril Organization in Human Articular Cartilage. Journal of Anatomy. 1977;123(Pt 2):437-57.

177. Mononen ME, Mikkola MT, Julkunen P, Ojala R, Nieminen MT, Jurvelin JS, Korhonen RK. Effect of Superficial Collagen Patterns and Fibrillation of Femoral Articular Cartilage on Knee Joint Mechanics - a 3d Finite Element Analysis. Journal of Biomechanics. 2012;45(3):579-87.

References

156

178. Nickien M, Thambyah A, Broom N. How Changes in Fibril-Level Organization Correlate with the Macrolevel Behavior of Articular Cartilage. Wiley Interdisciplinary Reviews: Systems Biology and Medicine. 2013;5(4):495-509.

179. Julkunen P, Jurvelin JS, Isaksson H. Contribution of Tissue Composition and Structure to Mechanical Response of Articular Cartilage under Different Loading Geometries and Strain Rates. Biomechanics And Modeling In Mechanobiology. 2010;9(2):237-45.

180. Oloyede A, Broom N. The Biomechanics of Cartilage Load-Carriage. Connective tissue research. 1996;34(2):119-43.

181. Moskowitz R. Adult Articular Cartilage. Arthritis & Rheumatism. 1980;23(1):135-.

182. Mollenhauer J, Aurich M, Muehleman C, Khelashvilli G, Irving T. X-Ray Diffraction of the Molecular Substructure of Human Articular Cartilage. Connective tissue research. 2003;44(5):201-7.

183. Yarker Y, Hukins D, Naves C. X-Ray Diffraction Studies of Tibial Plateau Cartilage Using Synchrotron Radiation. Connective tissue research. 1984;12(3-4):337-43.

184. Tadimalla S, Tourell MC, Knott R, Momot KI. Assessment of Collagen Fiber Orientation Dispersion in Articular Cartilage by Small-Angle X-Ray Scattering and Diffusion Tensor Imaging: Preliminary Results. Magnetic Resonance Imaging. 2018;48:115-21.

185. Tadimalla S, Tourell M, Knott R, Momot K. Quantifying Collagen Fibre Architecture in Articular Cartilage Using Small-Angle X-Ray Scattering. Biomedical Spectroscopy and Imaging. 2017;6(1-2):37-57.

186. He B, Wu JP, Chim SM, Xu J, Kirk TB. Microstructural Analysis of Collagen and Elastin Fibres in the Kangaroo Articular Cartilage Reveals a Structural Divergence Depending on Its Local Mechanical Environment. Osteoarthritis and Cartilage. 2013;21(1):237-45.

187. Wu JP, Kirk TB, Zheng MH. Study of the Collagen Structure in the Superficial Zone and Physiological State of Articular Cartilage Using a 3d Confocal Imaging Technique. Journal of Orthopaedic Surgery and Research. 2008;3:29.

188. Fullerton GD, Potter JL, Dornbluth NC. NMR Relaxation of Protons in Tissues and Other Macromolecular Water Solutions. Magnetic Resonance Imaging. 1982;1(4):209-26.

189. Xia Y. Introduction to Magnetic Resonance. In: Blu¨mler P, Blu¨mich B, Botto R, Fukushima E, editors. Spatially Resolved Magnetic Resonance. Weinheim, Germany: Wiley-VCH; 1998.

190. Fullerton GD, Rahal A. Collagen Structure: The Molecular Source of the Tendon Magic Angle Effect. Journal of Magnetic Resonance Imaging. 2007;25(2):345-61.

191. Mlynarik V, Szomolanyi P, Toffanin R, Vittur F, Trattnig S. Transverse Relaxation Mechanisms in Articular Cartilage. Journal of Magnetic Resonance. 2004;169(2):300-7.

192. Australian Institute of Health and Welfare. Osteoarthritis. Canberra, 2017.

193. Bedson J, Croft P. The Discordance between Clinical and Radiographic Knee Osteoarthritis: A Systematic Search and Summary of the Literature. BMC Musculoskeletal Disorders. 2008;9(1):116.

194. Nishioka H, Hirose J, Nakamura E, Oniki Y, Takada K, Yamashita Y, Mizuta H. T1ρ and T2 Mapping Reveal the in Vivo Extracellular Matrix of Articular Cartilage. Journal of Magnetic Resonance Imaging. 2012;35(1):147-55.

References

157

195. Gold GE, Burstein D, Dardzinski B, Lang P, Boada F, Mosher T. MRI of Articular Cartilage in Oa: Novel Pulse Sequences and Compositional/Functional Markers. Osteoarthritis and Cartilage. 2006;14, Supplement 1:76-86.

196. Burstein D, Gray ML, Hartman AL, Gipe R, Foy BD. Diffusion of Small Solutes in Cartilage as Measured by Nuclear Magnetic Resonance (NMR) Spectroscopy and Imaging. Journal of Orthopaedic Research. 1993;11(4):465-78.

197. Filidoro L, Dietrich O, Weber J, Rauch E, Oerther T, Wick M, Reiser MF, Glaser C. High-Resolution Diffusion Tensor Imaging of Human Patellar Cartilage: Feasibility and Preliminary Findings. Magnetic Resonance in Medicine. 2005;53(5):993-8.

198. Wheaton AJ, Casey FL, Gougoutas AJ, Dodge GR, Borthakur A, Lonner JH, Schumacher HR, Reddy R. Correlation of T1ρ with Fixed Charge Density in Cartilage. Journal of Magnetic Resonance Imaging. 2004;20(3):519-25.

199. Blumenkrantz G, Majumdar S. Quantitative Magnetic Resonance Imaging of Articular Cartilage in Osteoarthritis. European Cells and Materials. 2007;15(13):76-86.

200. Stahl R, Luke A, Li X, Carballido-Gamio J, Ma CB, Majumdar S, Link T. T1rho, T2 and Focal Knee Cartilage Abnormalities in Physically Active and Sedentary Healthy Subjects Versus Early Oa Patients—a 3.0-Tesla MRI Study. European Radiology. 2009;19(1):132-43.

201. Regatte RR, Akella SVS, Wheaton AJ, Lech G, Borthakur A, Kneeland JB, Reddy R. 3d-T1ρ-Relaxation Mapping of Articular Cartilage: In Vivo Assessment of Early Degenerative Changes in Symptomatic Osteoarthritic Subjects1. Academic Radiology. 2004;11(7):741-9.

202. Li X, Han ET, Busse RF, Majumdar S. In Vivo T1ρ Mapping in Cartilage Using 3d Magnetization-Prepared Angle-Modulated Partitioned K-Space Spoiled Gradient Echo Snapshots (3d Mapss). Magnetic Resonance in Medicine. 2008;59(2):298-307.

203. Wheaton AJ, Borthakur A, Shapiro EM, Regatte RR, Akella SVS, Kneeland JB, Reddy R. Proteoglycan Loss in Human Knee Cartilage: Quantitation with Sodium Mr Imaging—Feasibility Study. Radiology. 2004;231(3):900-5.

204. Borthakur A, Shapiro EM, Beers J, Kudchodkar S, Kneeland JB, Reddy R. Sensitivity of MRI to Proteoglycan Depletion in Cartilage: Comparison of Sodium and Proton MRI. Osteoarthritis and Cartilage. 2000;8(4):288-93.

205. Wang L, Wu Y, Chang G, Oesingmann N, Schweitzer ME, Jerschow A, Regatte RR. Rapid Isotropic 3d-Sodium MRI of the Knee Joint in Vivo at 7T. Journal of Magnetic Resonance Imaging. 2009;30(3):606-14.

206. McKenzie CA, Williams A, Prasad PV, Burstein D. Three-Dimensional Delayed Gadolinium-Enhanced MRI of Cartilage (dGEMRIC) at 1.5T and 3.0T. Journal of Magnetic Resonance Imaging. 2006;24(4):928-33.

207. Intema F, Hazewinkel HAW, Gouwens D, Bijlsma JWJ, Weinans H, Lafeber FPJG, Mastbergen SC. In Early Oa, Thinning of the Subchondral Plate Is Directly Related to Cartilage Damage: Results from a Canine ACLT-Meniscectomy Model. Osteoarthritis and Cartilage. 2010;18(5):691-8.

208. Hayami T, Pickarski M, Zhuo Y, Wesolowski GA, Rodan GA, Duong LT. Characterization of Articular Cartilage and Subchondral Bone Changes in the Rat Anterior Cruciate Ligament Transection and Meniscectomized Models of Osteoarthritis. Bone. 2006;38(2):234-43.

209. Goldring MB. Potential Mechanisms of Ptoa: Inflammation. In: Olson MDAS, Guilak PF, editors. Post-Traumatic Arthritis: Pathogenesis, Diagnosis and Management. Boston, MA: Springer US; 2015. p. 201-9.

References

158

210. Zanetti M, Bruder E, Romero J, Holder J. Bone Marrow Edema Pattern in Osteoarthritic Knees: Correlation between MRI Imaging and Histological Findings Radiology. 2000;215(3):835-40.

211. Peterfy C, Guermazi A, Zaim S, Tirman P, Miaux Y, White D, Kothari M, Lu Y, Fye K, Zhao S, Genant H. Whole-Organ Magnetic Resonance Imaging Score (Worms) of the Knee in Osteoarthritis. Osteoarthritis and Cartilage. 2004;12(3):177-90.

212. Li X, Ma B, Bolbos R, Stahl R, Lozano J, Zuo J, Lin K, Link T, Safran M, Majumdar S. Quantitative Assessment of Bone Marrow Edema-Like Lesion and Overlying Cartilage in Knees with Osteoarthritis and Anterior Cruciate Ligament Tear Using Mr Imaging and Spectroscopic Imaging at 3 Tesla. Journal of Magnetic Resonance Imaging. 2008;28(2):453-61.

213. Roemer F, Bohndorf K. Long-Term Osseous Sequelae after Acute Trauma of the Knee Joint Evaluated by MRI. Skeletal Radiology. 2002;31(11):615-23.

214. Marieb EN, Hoehn K. Human Anatomy & Physiology: Pearson Education; 2007.

215. Standring S. Gray's Anatomy E-Book: The Anatomical Basis of Clinical Practice: Elsevier Health Sciences; 2015.

216. Lohmander L, Ostenberg A, Englund M, Roos H. High Prevalence of Knee Osteoarthritis, Pain, and Functional Limitations in Female Soccer Players Twelve Years after Anterior Cruciate Ligament Injury. Arthritis and Rheumatism. 2004;50(10):3145-52.

217. Gillquist J, Messner K. Anterior Cruciate Ligament Reconstruction and the Long-Term Incidence of Gonarthrosis. Sports Medicine. 1999;27(3):143-56.

218. Felson D, Lawrence R, Dieppe P, Hirsch R, Helmick C, Jordan J, Kington R, Lane N, Nevitt M, Zhang Y, Sowers M, McAlindon T, Spector T, Poole A, Yanovski S, Ateshian G, Sharma L, Buckwalter J, Brandt K, Fries J. Osteoarthritis: New Insights. Part 1: The Disease and Its Risk Factors. Annals of Internal Medicine. 2000;133(8):635-46.

219. Cerejo R, Dunlop D, Cahue S, Channin D, Song J, Sharma L. The Influence of Alignment on Risk of Knee Osteoarthritis Progression According to Baseline Stage of Disease. Arthritis and Rheumatism. 2002;46(10):2632-36.

220. Kurusawa H, Fukubayashi T, Nakajima H. Load-Bearing Mode of the Knee Joint: Physical Behavior of Knee Joint with or without Menisci. Clinical Orthopaedics and Related Research. 1980:283-90.

221. Braun H, Gold G. Diagnosis of Osteoarthritis: Imaging. Bone. 2012;51(2):278-88.

222. Song Y, Greve J, Carter D, Giori N. Menisectomy Alters the Dynamic Deformational Behavior and Cumulative Strain of Tibial Articular Cartilage in Knee Joints Subjected to Cyclic Loads. Osteoarthritis and Cartilage. 2008;16(12):1545-54.

223. Englund M, Guermazi A, Roemer F, Yang M, Zhang Y, Nevitt M, Lynch J, Lewis C, Torner J, Felson D. Meniscal Pathology on MRI Increases the Risk for Both Incident and Enlarging Subchondral Bone Marrow Lesions of the Knee: The Most Study. Annals of the Rheumatic Diseases. 2010;69(10):1796-802.

224. Crema M, Guermazi A, Li L, Nogueira-Barbosa M, Marra M, Roemer F, Eckstein F, Le Graverand M, Wyman B, Hunter D. The Association of Prevalent Medial Meniscal Pathology with Cartilage Loss in the Medial Tibiofemoral Compartment over a 2-Year Period. Osteoarthritis Cartilage. 2010;18(3):336-43.

225. Helms C. The Meniscus: Recent Advances in Mr Imaging of the Knee. American Journal of Roentgenology. 2002;179(5):1115-22.

References

159

226. Hayashi D, Roemer F, Katur A, Felson D, Yang S, Alomran F, Guermazi A. Imaging of Synovitis in Osteoarthritis: Current Status and Outlook. Seminars in Arthritis and Rheumatism. 2011;41(2):116-30.

227. Liu L, Ishijima M, Futami I, Kaneko H, Kubota M, Kawasaki T, Matsumoto T, Kurihara H, Ning L, Xu Z, Ikeda H, Takazawa Y, Saita Y, Kimura Y, Xu S, Kaneko K, Kurosawa H. Correlation between Synovitis Detected on Enhanced-Magnetic Resonance Imaging and a Histological Analysis with a Patient-Oriented Outcome Measure for Japanese Patients with End-Stage Knee Osteoarthritis Receiving Joint Replacement Surgery. Clinical Rheumatology. 2010;29(10):1185-90.

228. Myers S, Flusser D, Brandt K, Heck D. Prevalence of Cartilage Shards in Synovium and Their Association with Synovitis in Patients with Early and End Stage Osteoarthritis. The Journal of Rheumatology. 1992;19(8):1247-51.

229. Guermazi A, Burstein D, Conaghan P, Eckstein F, MP HLG-G, Keen H, Roemer F. Imaging in Osteoarthritis. Rheumatic Disease Clinics in North America. 2008;34(3):645-87.

230. Lassere M, Bird P. Measurement of Rheumatoid Arthritis Disease Activity and Damage Using Magnetic Resonance Imaging. Truth and Discrimination: Does MRI Make the Grade? Journal of Rheumatology. 2001;28(5):1151-57.

231. Boyd NF, Guo H, Martin LJ, Sun L, Stone J, Fishell E, Jong RA, Hislop G, Chiarelli A, Minkin S, Yaffe MJ. Mammographic Density and the Risk and Detection of Breast Cancer. New England Journal of Medicine. 2007;356(3):227-36.

232. McCormack VA, dos Santos Silva I. Breast Density and Parenchymal Patterns as Markers of Breast Cancer Risk: A Meta-Analysis. Cancer Epidemiology Biomarkers and Prevention Biomarkers. 2006;15(6):1159-69.

233. Yaghjyan L, Colditz GA, Collins LC, Schnitt SJ, Rosner B, Vachon C, Tamimi RM. Mammographic Breast Density and Subsequent Risk of Breast Cancer in Postmenopausal Women According to Tumor Characteristics. JNCI Journal of the National Cancer Institute. 2011;103(15):1179-89.

234. Shawky MS, Ricciardelli C, Lord M, Whitelock J, Ferro V, Britt K, Thompson EW. Proteoglycans: Potential Agents in Mammographic Density and the Associated Breast Cancer Risk. Journal of Mammary Gland Biology and Neoplasia. 2015;20(3):121-31.

235. Boyd N, Byng J, Jong R, Fishell E, Little L, Miller A, Lockwood G, Tritchler D, Yaffe MJ. Quantitative Classification of Mammographic Densities and Breast Cancer Risk: Results from the Canadian National Breast Screening Study. JNCI: Journal of the National Cancer Institute. 1995;87(9):670-5.

236. American College of Radiology. Breast Imaging Reporting and Data System. American College of Radiology, Reston, 1998.

237. Berg WA, Campassi C, Langenberg P, Sexton MJ. Breast Imaging Reporting and Data System. American Journal of Roentgenology. 2000;174(6):1769-77.

238. Lin SJ, Cawson J, Hill P, Haviv I, Jenkins M, Hopper JL, Southey MC, Campbell IG, Thompson EW. Image-Guided Sampling Reveals Increased Stroma and Lower Glandular Complexity in Mammographically Dense Breast Tissue. Breast Cancer Research and Treatment. 2011;128(2):505-16.

239. American Cancer Society. American Cancer Society Recommendations for the Early Detection of Breast Cancer: American cancer society; 2017 [Available from:

References

160

https://www.cancer.org/cancer/breast-cancer/screening-tests-and-early-detection/ american-cancer-society-recommendations-for-the-early-detection-of-breast-cancer.html].

240. Chistiakov DA, Voronova NV, Chistiakov PA. Genetic Variations in DNA Repair Genes, Radiosensitivity to Cancer and Susceptibility to Acute Tissue Reactions in Radiotherapy-Treated Cancer Patients. Acta Oncologica. 2008;47(5):809-24.

241. Bernstein JL, Teraoka SN, John EM, Andrulis IL, Knight JA, Lapinski R, Olson ER, Wolitzer AL, Seminara D, Whittemore AS, Concannon P. The Chek2*1100delc Allelic Variant and Risk of Breast Cancer: Screening Results from the Breast Cancer Family Registry. Cancer Epidemiology Biomarkers and Prevention Biomarkers. 2006;15(2):348-52.

242. Hogge JP, Palmer CH, Muller CC, Little ST, Smith DC, Fatouros PP, Paredes ESd. Quality Assurance in Mammography: Artifact Analysis. RadioGraphics. 1999;19(2):503-22.

243. Carney PA, Miglioretti DL, Yankaskas BC, et al. Individual and Combined Effects of Age, Breast Density, and Hormone Replacement Therapy Use on the Accuracy of Screening Mammography. Annals of Internal Medicine. 2003;138(3):168-75.

244. Buist DSM, Porter PL, Lehman C, Taplin SH, White E. Factors Contributing to Mammography Failure in Women Aged 40–49 Years. Journal of the National Cancer Institute. 2004;96(19):1432-40.

245. Houssami NL, Sarah J; Ciatto, Stefano. Breast Cancer Screening: Emerging Role of New Imaging Techniques as Adjuncts to Mammography. Medical Journal of Australia. 2009;190(9):493-8.

246. Hugo HJ, Tourell MC, O’Gorman PM, Paige AE, Wellard RM, Lloyd T, Momot KI, Thompson EW. Looking Beyond the Mammogram to Assess Mammographic Density: A Narrative Review. Biomedical Spectroscopy and Imaging. 2018(Preprint):1-18.

247. McDonald ES, Schopp J, Peacock S, Olson ML, DeMartini WB, Rahbar H, Lehman CD, Partridge SC. Diffusion-Weighted MRI: Association between Patient Characteristics and Apparent Diffusion Coefficients of Normal Breast Fibroglandular Tissue at 3 Tesla. American Journal of Roentgenology. 2014;202(5):W496-W502.

248. Hawley JR, Kalra P, Mo X, Raterman B, Yee LD, Kolipaka A. Quantification of Breast Stiffness Using MR Elastography at 3 Tesla with a Soft Sternal Driver: A Reproducibility Study. Journal of Magnetic Resonance Imaging. 2017;45(5):1379-84.

249. Wu S, Weinstein SP, Conant EF, Kontos D. Automated Fibroglandular Tissue Segmentation and Volumetric Density Estimation in Breast MRI Using an Atlas-Aided Fuzzy C-Means Method. Medical Physics. 2013;40(12):122302.

250. Nie K, Chen J-H, Chan S, Chau M-KI, Yu HJ, Bahri S, Tseng T, Nalcioglu O, Su M-Y. Development of a Quantitative Method for Analysis of Breast Density Based on Three-Dimensional Breast MRI. Medical Physics. 2008;35(12):5253-62.

251. Wang J, Azziz A, Fan B, Malkov S, Klifa C, Newitt D, Yitta S, Hylton N, Kerlikowske K, Shepherd JA. Agreement of Mammographic Measures of Volumetric Breast Density to MRI. PLOS ONE. 2013;8(12):e81653.

252. Magritek. NMR-MOUSE One-Sided NMR 2015 [Available from: http://www.magritek.com/products/nmr-mouse].

253. Kiselev VG, Novikov DS. Transverse NMR Relaxation in Biological Tissues. NeuroImage. 2018;15(182):149-68.

http://www.magritek.com/products/nmr-mouse

References

161

254. Eitel I, Friedrich MG. T2-Weighted Cardiovascular Magnetic Resonance in Acute Cardiac Disease. Journal of Cardiovascular Magnetic Resonance. 2011;13(1):13.

255. Montant P, Sigovan M, Revel D, Douek P. Mr Imaging Assessment of Myocardial Edema with T2 Mapping. Diagnostic and Interventional Imaging. 2015;96(9):885-90.

256. Johnston JH, Kim HK, Merrow AC, Laor T, Serai S, Horn PS, Kim DH, Wong BL. Quantitative Skeletal Muscle MRI: Part 1, Derived T2 Fat Map in Differentiation between Boys with Duchenne Muscular Dystrophy and Healthy Boys. American Journal of Roentgenology. 2015;205(2):W207-W15.

257. Mankodi A, Azzabou N, Bulea T, Reyngoudt H, Shimellis H, Ren Y, Kim E, Fischbeck KH, Carlier PG. Skeletal Muscle Water T2 as a Biomarker of Disease Status and Exercise Effects in Patients with Duchenne Muscular Dystrophy. Neuromuscular Disorders. 2017;27(8):705-14.

258. Ploutz-Snyder LL, Nyren S, Cooper TG, Potchen EJ, Meyer RA. Different Effects of Exercise and Edema on T2 Relaxation in Skeletal Muscle. Magnetic Resonance in Medicine. 1997;37(5):676-82.

259. LeBlanc A, Lin C, Shackelford L, Sinitsyn V, Evans H, Belichenko O, Schenkman B, Kozlovskaya I, Oganov V, Bakulin A, Hedrick T, Feeback D. Muscle Volume, MRI Relaxation Times (T2), and Body Composition after Spaceflight. Journal of Applied Physiology. 2000;89(6):2158-64.

260. Bakshi R, Shaikh ZA, Janardhan V. MRI T2 Shortening (‘Black T2’) in Multiple Sclerosis: Frequency, Location, and Clinical Correlation. NeuroReport. 2000;11(1):15-21.

261. Ordidge RJ, Gorell JM, Deniau JC, Knight RA, Helpern JA. Assessment of Relative Brain Iron Concentrations Using T2-Weighted and T2*-Weighted MRI at 3 Tesla. Magnetic Resonance in Medicine. 1994;32(3):335-41.

262. Rocca MA, Ceccarelli A, Falini A, Colombo B, Tortorella P, Bernasconi L, Comi G, Scotti G, Filippi M. Brain Gray Matter Changes in Migraine Patients with T2-Visible Lesions: A 3-T MRI Study. Stroke. 2006;37(7):1765-70.

263. Hirano H, Yokoyama S, Yunoue S, Yonezawa H, Yatsushiro K, Yoshioka T, Hanaya R, Tokimura H, Arita K. MRI T2 Hypointensity of Metastatic Brain Tumors from Gastric and Colonic Cancers. International Journal of Clinical Oncology. 2014;19(4):643-8.

264. Prastawa M, Bullitt E, Ho S, Gerig G. A Brain Tumor Segmentation Framework Based on Outlier Detection. Medical Image Analysis. 2004;8(3):275-83.

265. Mosher TJ, Liu Y, Torok CM. Functional Cartilage MRI T2 Mapping: Evaluating the Effect of Age and Training on Knee Cartilage Response to Running. Osteoarthritis and Cartilage. 2010;18(3):358-64.

266. Weinreb JH, Sheth C, Apostolakos J, McCarthy M-B, Barden B, Cote MP, Mazzocca AD. Tendon Structure, Disease, and Imaging. Muscles, Ligaments and Tendons Journal. 2014;4(1):66-73.

267. Guerrero P, Li X, Patel K, Brown M, Busconi B. Medial Patellofemoral Ligament Injury Patterns and Associated Pathology in Lateral Patella Dislocation: An MRI Study. BMC Sports Science, Medicine and Rehabilitation. 2009;1(1):17.

268. Yao L, Magalnick M, Wilson M, Lipsky P, Goldbach-Mansky R. Periarticular Bone Findings in Rheumatoid Arthritis: T2-Weighted Versus Contrast-Enhanced T1-Weighted MRI. American Journal of Roentgenology. 2006;187(2):358-63.

References

162

269. Abdellaoui A, Iyengar S, Freeman S. Imaging in Prostate Cancer. Future oncology. 2011;7(5):679-91.

270. Haider MA, van der Kwast TH, Tanguay J, Evans AJ, Hashmi A-T, Lockwood G, Trachtenberg J. Combined T2-Weighted and Diffusion-Weighted MRI for Localization of Prostate Cancer. American Journal of Roentgenology. 2007;189(2):323-8.

271. Mow VC, Huiskes R. Basic Orthopaedic Biomechanics & Mechano-Biology: Lippincott Williams & Wilkins; 2005.

272. Dawson TJ, Taylor CR. Energetic Cost of Locomotion in Kangaroos. Nature. 1973;246(5431):313-4.

273. Fuss FK. Fibrous Tissue on the Tibia Plateau of the Kangaroo. A Theory on the Pressure Absorption of Joint Surfaces. The Anatomical Record. 1994;238(3):297-303.

274. Alexander RM, Vernon A. The Mechanics of Hopping by Kangaroos (Macropodidae). Journal of Zoology. 1975;177(2):265-303.

275. Xia Y, Momot KI, Chen Z, Chen CT, Kahn D, Badar F. Introduction to Cartilage. In: Xia Y, Momot KI, editors. Biophysics and Biochemistry of Cartilage by NMR and MRI. Cambridge: Royal Society of Chemistry; 2016. p. 3-43.

276. He B, Wu JP, Kirk TB, Carrino JA, Xiang C, Xu JK. High-Resolution Measurements of the Multilayer Ultra-Structure of Articular Cartilage and Their Translational Potential. Arthritis Research & Therapy. 2014;16(2):205

277. Mononen ME, Mikkola MT, Julkunen P, Ojala R, Nieminen MT, Jurvelin JS, Korhonen RK. Effect of Superficial Collagen Patterns and Fibrillation of Femoral Articular Cartilage on Knee Joint Mechanics-a 3d Finite Element Analysis. Journal Of Biomechanics. 2012;45(3):579-87.

278. Nickien M, Thambyah A, Broom N. How Changes in Fibril-Level Organization Correlate with the Macrolevel Behavior of Articular Cartilage. Wiley Interdisciplinary Reviews-Systems Biology and Medicine. 2013;5(4):495-509.

279. Korhonen RK, Laasanen MS, Toyras J, Lappalainen R, Helminen HJ, Jurvelin JS. Fibril Reinforced Poroelastic Model Predicts Specifically Mechanical Behavior of Normal, Proteoglycan Depleted and Collagen Degraded Articular Cartilage. Journal of Biomechanics. 2003;36(9):1373-9.

280. Laasanen MS, Toyras J, Korhonen RK, Rieppo J, Saarakkala S, Nieminen MT, Hirvonen J, Jurvelin JS. Biomechanical Properties of Knee Articular Cartilage. Biorheology. 2003;40(1-3):133-40.

281. Menezes NM, Gray ML, Hartke JR, Burstein D. T-2 and T-1, MRI in Articular Cartilage Systems. Magnetic Resonance in Medicine. 2004;51(3):503-9.

282. Brama PAJ, Tekoppele JM, Bank RA, Barneveld A, Van Weeren PR. Functional Adaptation of Equine Articular Cartilage: The Formation of Regional Biochemical Characteristics up to Age One Year. Equine Veterinary Journal. 2000;32(3):217-21.

283. Brama PAJ, TeKoppele JM, Bank RA, Barneveld A, van Weeren PR. Development of Biochemical Heterogeneity of Articular Cartilage: Influences of Age and Exercise. Equine Veterinary Journal. 2002;34(3):265-9.

284. Eyre DR, Brickleyparsons DM, Glimcher MJ. Predominance of Type-1 Collagen at Surface of Avian Articular Cartilage. Febs Letters. 1978;85(2):259-63.

285. Benjamin M, Evans EJ. Fibrocartilage. Journal of Anatomy. 1990;171:1-15.

References

163

286. Kurkijarvi JE, Mattila L, Ojala RO, Vasara AI, Jurvelin JS, Kiviranta I, Nieminen MT. Evaluation of Cartilage Repair in the Distal Femur after Autologous Chondrocyte Transplantation Using T-2 Relaxation Time and dGEMRIC. Osteoarthritis and Cartilage. 2007;15(4):372-8.

287. Tadimalla S, Momot KI. Effect of Partial H2O-D2O Replacement on the Anisotropy of Transverse Proton Spin Relaxation in Bovine Articular Cartilage. PLOS ONE. 2014;9(12):e115288.

288. Fullerton GD. The Magic Angle Effect in NMR and MRI of Cartilage. Biophysics and Biochemistry of Cartilage by NMR and MRI. Cambridge, UK: Royal Society of Chemistry; 2016. p. 109-44.

289. Eliav U, Navon G. A Study of Dipolar Interactions and Dynamic Processes of Water Molecules in Tendon by H-1 and H-2 Homonuclear and Heteronuclear Multiple-Quantum-Filtered NMR Spectroscopy. Journal of Magnetic Resonance. 1999;137(2):295-310.

290. Tourell MC, Momot KI. Molecular Dynamics of a Hydrated Collagen Peptide: Insights into Rotational Motion and Residence Times of Single-Water Bridges in Collagen. Journal of Physical Chemistry B. 2016;120(49):12432-43.

291. Navon G, Eliav U, Demco DE, Bluemich B. Study of Order and Dynamic Processes in Tendon by NMR and MRI. Journal of Magnetic Resonance Imaging. 2007;25(2):362-80.

292. de Visser SK, Crawford RW, Pope JM. Structural Adaptations in Compressed Articular Cartilage Measured by Diffusion Tensor Imaging. Osteoarthritis and Cartilage. 2008;16(1):83-9.

293. Meder R, de Visser SK, Bowden JC, Bostrom T, Pope JM. Diffusion Tensor Imaging of Articular Cartilage as a Measure of Tissue Microstructure. Osteoarthritis and Cartilage. 2006;14(9):875-81.

294. Bradley JV. Distribution-Free Statistical Tests. Englewood Cliffs, N.J.: Prentice-Hall; 1968.

295. He B, Wu JP, Xu JK, Day RE, Kirk TB. Microstructural and Compositional Features of the Fibrous and Hyaline Cartilage on the Medial Tibial Plateau Imply a Unique Role for the Hopping Locomotion of Kangaroo. PLOS ONE. 2013;8(9):e74303.

296. LeRoux MA, Arokoski J, Vail TP, Guilak F, Hyttinen MM, Kiviranta I, Setton LA. Simultaneous Changes in the Mechanical Properties, Quantitative Collagen Organization, and Proteoglycan Concentration of Articular Cartilage Following Canine Meniscectomy. Journal of Orthopaedic Research. 2000;18(3):383-92.

297. Sophia Fox AJ, Bedi A, Rodeo SA. The Basic Science of Articular Cartilage: Structure, Composition, and Function. Sports Health. 2009;1(6):461-8.

298. Kurosawa H, Fukubayashi T, Nakajima H. Load-Bearing Mode of the Knee Joint: Physical Behavior of Knee Joint with or without Menisci. Clinical Orthopaedics and Related Research. 1980;149:283-90.

299. Tourell MC, Kirkwood M, Pearcy MJ, Momot KI, Little JP. Load-Induced Changes in the Diffusion Tensor of Ovine Anulus Fibrosus: A Pilot MRI Study. Journal of Magnetic Resonance Imaging. 2017;45(6):1723-35.

300. Meng QG, An SQ, Damion RA, Jin ZM, Wilcox R, Fisher J, Jones A. The Effect of Collagen Fibril Orientation on the Biphasic Mechanics of Articular Cartilage. Journal of the Mechanical Behavior of Biomedical Materials. 2017;65:439-53.

References

164

301. Wellard R, Ravasio JP, Guesne S, Bell C, Oloyede A, Tevelen G, Pope J, Momot K. Simultaneous Magnetic Resonance Imaging and Consolidation Measurement of Articular Cartilage. Sensors. 2014;14(5):7940.

302. Raya JG, Arnoldi AP, Weber DL, Filidoro L, Dietrich O, Adam-Neumair S, Mutzel E, Melkus G, Putz R, Reiser MF, Jakob PM, Glaser C. Ultra-High Field Diffusion Tensor Imaging of Articular Cartilage Correlated with Histology and Scanning Electron Microscopy. Magnetic Resonance Materials In Physics Biology And Medicine. 2011;24(4):247-58.

303. Pierce DM, Trobin W, Raya JG, Trattnig S, Bischof H, Glaser C, Holzapfel GA. DT-MRI Based Computation of Collagen Fiber Deformation in Human Articular Cartilage: A Feasibility Study. Annals of Biomedical Engineering. 2010;38(7):2447-63

304. Neu CP, Walton JH. Displacement Encoding for the Measurement of Cartilage Deformation. Magnetic Resonance in Medicine. 2008;59(1):149-55.

305. Chang NJ, Lam CF, Lin CC, Chen WL, Li CF, Lin YT, Yeh ML. Transplantation of Autologous Endothelial Progenitor Cells in Porous Plga Scaffolds Create a Microenvironment for the Regeneration of Hyaline Cartilage in Rabbits. Osteoarthritis and Cartilage. 2013;21(10):1613-22.

306. Huey DJ, Hu JC, Athanasiou KA. Unlike Bone, Cartilage Regeneration Remains Elusive. Science. 2012;338(6109):917-21.

307. Huo CW, Waltham M, Khoo C, Fox SB, Hill P, Chen S, Chew GL, Price JT, Nguyen CH, Williams ED, Henderson M, Thompson EW, Britt KL. Mammographically Dense Human Breast Tissue Stimulates MCF10DCIS.com Progression to Invasive Lesions and Metastasis. Breast Cancer Research. 2016;18:106.

308. Wolfe JN. Breast Patterns as an Index of Risk for Developing Breast Cancer. American Journal of Roentgenology. 1976;126(6):1130-7.

309. Sprague BL, Gangnon RE, Burt V, Trentham-Dietz A, Hampton JM, Wellman RD, Kerlikowske K, Miglioretti DL. Prevalence of Mammographically Dense Breasts in the United States. J Natl Cancer Inst. 2014;106(10).

310. Huo CW, Chew G, Hill P, Huang D, Ingman W, Hodson L, Brown KA, Magenau A, Allam AH, McGhee E, Timpson P, Henderson MA, Thompson EW, Britt K. High Mammographic Density Is Associated with an Increase in Stromal Collagen and Immune Cells within the Mammary Epithelium. Breast Cancer Research. 2015;17(1):79.

311. Mirette H, Caroline D. Is Mammographic Density a Biomarker to Study the Molecular Causes of Breast Cancer? In: Uchiyama N, editor. Mammography - Recent Advances: INTECH Open Access Publisher; 2012.

312. Britt K, Ingman W, Huo C, Chew G, Thompson E. The Pathobiology of Mammographic Density. Journal of Cancer Biology and Research. 2014;2(1):1021.

313. Evans DG, Astley S, Stavrinos P, Harkness E, Donnelly LS, Dawe S, Jacob I, Harvie M, Cuzick J, Brentnall A. Improvement in Risk Prediction, Early Detection and Prevention of Breast Cancer in the Nhs Breast Screening Programme and Family History Clinics: A Dual Cohort Study. 2016.

314. Blend R, Rideout D, Kaizer L, Shannon P, Tudor-Robers B, Boyd N. Parenchymal Patterns of the Breast Defined by Real Time Ultrasound. European Journal of Cancer Prevention. 1995;4(4):293-8.

315. Glide C, Duric N, Littrup P. Novel Approach to Evaluating Breast Density Utilizing Ultrasound Tomography. Medical Physics. 2007;34(2):744-53.

References

165

316. Maskarinec G, Morimoto Y, Laguaña MB, Novotny R, Guerrero RTL. Bioimpedence to Assess Breast Density as a Risk Factor for Breast Cancer in Adult Women and Adolescent Girls. Asian Pacific journal of cancer prevention. 2016;17(1):65-71.

317. Simick MJ, R; Wilson, B; Lilge, L. Non-Ionizing near-Infrared Radiation Transillumination Spectroscopy for Breast Tissue Density and Assessment of Breast Cancer Risk. Journal of Biomedical Optics. 2004;9(4):794-803.

318. Simick MK, Lilge L. Optical Transillumination Spectroscopy to Quantify Parenchymal Tissue Density: An Indicator for Breast Cancer Risk. The British Journal of Radiology. 2005;78(935):1009-17.

319. Rössler E, Mattea C, Saarakkala S, Lehenkari P, Finnilä M, Rieppo L, Karhula S, Nieminen MT, Stapf S. Correlations of Low‐Field NMR and Variable‐Field NMR Parameters with Osteoarthritis in Human Articular Cartilage under Load. NMR in Biomedicine. 2017;30(8):e3738.

320. Bergman E, Sarda Y, Ritz N, Sabo E, Navon G, Bergman R, Nevo U. In Vivo Assessment of Aged Human Skin with a Unilateral NMR Scanner. NMR in Biomedicine. 2015;28(6):656-66.

321. Bansal R, Hao X, Liu F, Xu D, Liu J, Peterson BS. The Effects of Changing Water Content, Relaxation Times, and Tissue Contrast on Tissue Segmentation and Measures of Cortical Anatomy in Mr Images. Magnetic Resonance Imaging. 2013;31(10):1709-30.

322. Hashido T, Saito S. Quantitative T1, T2, and T2* Mapping and Semi-Quantitative Neuromelanin-Sensitive Magnetic Resonance Imaging of the Human Midbrain. PLOS ONE. 2016;11(10):e0165160.

323. Welsch GH, Mamisch TC, Marlovits S, Glaser C, Friedrich K, Hennig FF, Salomonowitz E, Trattnig S. Quantitative T2 Mapping During Follow-up after Matrix-Associated Autologous Chondrocyte Transplantation (Mact): Full-Thickness and Zonal Evaluation to Visualize the Maturation of Cartilage Repair Tissue. Journal of Orthopaedic Research. 2009;27(7):957-63.

324. Apprich S, Welsch GH, Mamisch TC, Szomolanyi P, Mayerhoefer M, Pinker K, Trattnig S. Detection of Degenerative Cartilage Disease: Comparison of High-Resolution Morphological Mr and Quantitative T2 Mapping at 3.0 Tesla. Osteoarthritis and Cartilage. 2010;18(9):1211-7.

325. Blümich B, Perlo J, Casanova F. Mobile Single-Sided NMR. Progress in Nuclear Magnetic Resonance Spectroscopy. 2008;52(4):197-269.

326. Graham SJ, Stanchev PL, Bronskill MJ. Criteria for Analysis of Multicomponent Tissue T2 Relaxation Data. Magnetic Resonance in Medicine. 1996;35(3):370-8.

327. Sirrs SM, Laule C, Mädler B, Brief EE, Tahir SA, Bishop C, MacKay AL. Normal-Appearing White Matter in Patients with Phenylketonuria: Water Content, Myelin Water Fraction, and Metabolite Concentrations. Radiology. 2007;242(1):236-43.

328. Schad LR, Brix G, Zuna I, Härle W, Lorenz WJ, Semmler W. Multiexponential Proton Spin-Spin Relaxation in Mr Imaging of Human Brain Tumors. Journal of Computer Assisted Tomography. 1989;13(4):577-87.

329. Chew GL, Huang D, Lin SJ, Huo C, Blick T, Henderson MA, Hill P, Cawson J, Morrison WA, Campbell IG, Hopper JL, Southey MC, Haviv I, Thompson EW. High and Low Mammographic Density Human Breast Tissues Maintain Histological Differential in Murine Tissue Engineering Chambers. Breast Cancer Research and Treatment. 2012;135(1):177-87.

References

166

330. Provencher SW, editor A General-Purpose Constrained Regularization Method for Inverting Photon Correlation Data. Schulz-DuBois EO (eds) Photon Correlation Techniques in Fluid Mechanics; 1983; Berlin, Heidelberg: Springer Series in Optical Sciences.

331. Song YQ, Venkataramanan L, Hürlimann MD, Flaum M, Frulla P, Straley C. T1–T2 Correlation Spectra Obtained Using a Fast Two-Dimensional Laplace Inversion. Journal of Magnetic Resonance. 2002;154(2):261-8.

332. Mathur-De Vre R. The NMR Studies of Water in Biological Systems. Progress in Biophysics and Molecular Biology. 1980;35:103-34.

333. Koenig S, Brown R. Relaxometry of Tissue. In: Grant D, Harris R, editors. Encyclopedia of Nuclear Magnetic Resonance. New York: Wiley; 1996. p. 4108-20.

334. Cameron IL, Ord VA, Fullerton GD. Characterization of Proton NMR Relaxation Times in Normal and Pathological Tissues by Correlation with Other Tissue Parameters. Magnetic Resonance Imaging. 1984;2(2):97-106.

335. Song YQ. Categories of Coherence Pathways for the Cpmg Sequence. Journal of Magnetic Resonance. 2002;157(1):82-91.

336. Bashyam A, Li M, Cima MJ. Design and Experimental Validation of Unilateral Linear Halbach Magnet Arrays for Single-Sided Magnetic Resonance. Journal of Magnetic Resonance. 2018;292:36-43.

337. Watzlaw J, Glöggler S, Blümich B, Mokwa W, Schnakenberg U. Stacked Planar Micro Coils for Single-Sided NMR Applications. Journal of Magnetic Resonance. 2013;230:176-85.

338. Oligschläger D, Lehmkuhl S, Watzlaw J, Benders S, De Boever E, Rehorn C, Vossel M, Schnakenberg U, Blümich B. Miniaturized Multi-Coil Arrays for Functional Planar Imaging with a Single-Sided NMR Sensor. Journal of Magnetic Resonance. 2015;254:10-8.

339. Utsuzawa S, Fukushima E. Unilateral NMR with a Barrel Magnet. Journal of Magnetic Resonance. 2017;282:104-13.

340. Wang Q, Gao R, Liu S. Topology Optimization Based Design of Unilateral NMR for Generating a Remote Homogeneous Field. Journal of Magnetic Resonance. 2017;279:51-9.

341. Momot KI, Kuchel PW. PFG NMR Diffusion Experiments for Complex Systems. Concepts in Magnetic Resonance Part A: An Educational Journal. 2006;28(4):249-69.

342. Afara I, Prasadam I, Crawford R, Xiao Y, Oloyede A. Non-Destructive Evaluation of Articular Cartilage Defects Using near-Infrared (Nir) Spectroscopy in Osteoarthritic Rat Models and Its Direct Relation to Mankin Score. Osteoarthritis and Cartilage. 2012;20(11):1367-73.

343. Afara IO, Prasadam I, Crawford R, Xiao Y, Oloyede A. Near Infrared (NIR) Absorption Spectra Correlates with Subchondral Bone Micro-Ct Parameters in Osteoarthritic Rat Models. Bone. 2013;53(2):350-7.

344. Prasadam I, Mao X, Wang Y, Shi W, Crawford R, Xiao Y. Inhibition of P38 Pathway Leads to Oa-Like Changes in a Rat Animal Model. Rheumatology. 2012;51(2):813-23.

345. Bolbos R, Benoit-Cattin H, Langlois JB, Chomel A, Chereul E, Odet C, Janier M, Pastoureau P, Beuf O. Measurement of Knee Cartilage Thickness Using MRI: A Reproducibility Study in a Meniscectomized Guinea Pig Model of Osteoarthritis. NMR in Biomedicine. 2008;21(4):366-75.

346. Ali TS, Prasadam I, Xiao Y, Momot KI. Progression of Post-Traumatic Osteoarthritis in Rat Meniscectomy Models: Comprehensive Monitoring Using MRI. Scientific Reports. 2018;8(1):6861.

References

167

347. Brown TD, Johnston RC, Saltzman CL, Marsh JL, Buckwalter JA. Posttraumatic Osteoarthritis: A First Estimate of Incidence, Prevalence, and Burden of Disease. Journal of Orthopaedic Trauma. 2006;20(10):739-44.

348. Dedrick DK, Goldstein SA, Brandt K, O'Connor B, Goulet RW, Albrecht M. A Longitudinal Study of Subchondral Plate and Trabecular Bone in Cruciate-Deficient Dogs with Osteoarthritis Followed up for 54 Months. Arthritis and Rheumatism. 1993;36(10):1460-7.

349. Liu L, Ishijima M, Futami I, Kaneko H, Kubota M, Kawasaki T, Matsumoto T, Kurihara H, Ning L, Xu Z, Ikeda H, Takazawa Y, Saita Y, Kimura Y, Xu S, Kaneko K, Kurosawa H. Correlation between Synovitis Detected on Enhanced-Magnetic Resonance Imaging and a Histological Analysis with a Patient-Oriented Outcome Measure for Japanese Patients with End-Stage Knee Osteoarthritis Receiving Joint Replacement Surgery. Clinical Rheumatology. 2010;29(10):1185-90.

350. Sofat N, Ejindu V, Kiely P. What Makes Osteoarthritis Painful? The Evidence for Local and Central Pain Processing. Rheumatology. 2011;50(12):2157-65.

351. Inoue R, Ishibashi Y, Tsuda E, Yamamoto Y, Matsuzaka M, Takahashi I, Danjo K, Umeda T, Nakaji S, Toh S. Knee Osteoarthritis, Knee Joint Pain and Aging in Relation to Increasing Serum Hyaluronan Level in the Japanese Population. Osteoarthritis and Cartilage.19(1):51-7.

352. Goddard N, Gosling P. Intra-Articular Fluid Pressure and Pain in Osteoarthritis of the Hip. Bone & Joint Journal. 1988;70-B(1):52-5.

353. Watson PJ, Carpenter TA, Hall LD, Tyler JA. Cartilage Swelling and Loss in a Spontaneous Model of Osteoarthritis Visualized by Magnetic Resonance Imaging. Osteoarthritis and Cartilage. 1996;4(3):197-207.

354. Pelletier JP, Raynauld JP, Abram F, Haraoui B, Choquette D, Martel-Pelletier J. A New Non-Invasive Method to Assess Synovitis Severity in Relation to Symptoms and Cartilage Volume Loss in Knee Osteoarthritis Patients Using MRI. Osteoarthritis and Cartilage.16:S8-13.

355. Loeuille D, Rat AC, Goebel JC, Champigneulle J, Blum A, Netter P, Gillet P, Chary-Valckenaere I. Magnetic Resonance Imaging in Osteoarthritis: Which Method Best Reflects Synovial Membrane Inflammation?: Correlations with Clinical, Macroscopic and Microscopic Features. Osteoarthritis and Cartilage. 2009;17(9):1186-92.

356. Nieminen MT, Töyräs J, Laasanen MS, Silvennoinen J, Helminen HJ, Jurvelin JS. Prediction of Biomechanical Properties of Articular Cartilage with Quantitative Magnetic Resonance Imaging. Journal of Biomechanics. 2004;37(3):321-8.

357. Loeuille D, Gonord P, Guingamp C, Gillet P, Blum A, Sauzade M, Netter P. In Vitro Magnetic Resonance Microimaging of Experimental Osteoarthritis in the Rat Knee Joint. Journal of Rheumatology. 1997;24(1):133-9.

358. Ding M, Odgaard A, Hvid I. Changes in the Three-Dimensional Microstructure of Human Tibial Cancellous Bone in Early Osteoarthritis. Journal of Bone & Joint Surgery, British Volume. 2003;85(6):906-12.

359. Loeser RF, Goldring SR, Scanzello CR, Goldring MB. Osteoarthritis: A Disease of the Joint as an Organ. Arthritis & Rheumatism. 2012;64(6):1697-707.

360. Mankin HJ, Dorfman H, Lippiello L, Zarins A. Biochemical and Metabolic Abnormalities in Articular Cartilage from Osteo-Arthritic Human Hips Ii. Correlation of Morphology with Biochemicaland Metabolic Data. The Journal of Bone & Joint Surgery. 1971;53(3):523-37.

References

168

361. van der Sluijs JA, Geesink RGT, van der Linden AJ, Bulstra SK, Kuyer R, Drukker J. The Reliability of the Mankin Score for Osteoarthritis. Journal of Orthopaedic Research. 1992;10(1):58-61.

362. Schmitz N, Laverty S, Kraus VB, Aigner T. Basic Methods in Histopathology of Joint Tissues. Osteoarthritis and Cartilage. 2010;18(Supplement 3):S113-6.

363. Krause W, Pope M, Johnson R, Wilder D. Mechanical Changes in the Knee after Meniscectomy. The Journal of Bone & Joint Surgery. 1976;58(5):599-604.

364. Little CB, Fosang AJ. Is Cartilage Matrix Breakdown an Appropriate Therapeutic Target in Osteoarthritis - Insights from Studies of Aggrecan and Collagen Proteolysis? Current Drug Targets. 2010;11(5):561-75.

365. Liebl H, Joseph G, Nevitt MC, Singh N, Heilmeier U, Subburaj K, Jungmann PM, McCulloch CE, Lynch JA, Lane NE. Early T2 Changes Predict Onset of Radiographic Knee Osteoarthritis: Data from the Osteoarthritis Initiative. Annals of the Rheumatic Diseases. 2015;74(7):1353-9.

366. Pauli C, Bae WC, Lee M, Lotz M, Bydder GM, D’Lima DL, Chung CB, Du J. Ultrashort–Echo Time Mr Imaging of the Patella with Bicomponent Analysis: Correlation with Histopathologic and Polarized Light Microscopic Findings. Radiology. 2012;264(2):484-93.

367. Reiter DA, Roque RA, Lin PC, Irrechukwu O, Doty S, Longo D, Pleshko N, Spencer RG. Mapping Proteoglycan-Bound Water in Cartilage: Improved Specificity of Matrix Assessment Using Multiexponential Transverse Relaxation Analysis. Magnetic Resonance in Medicine. 2011;65(2):377-84.

368. Nissi MJ, Töyräs J, Laasanen MS, Rieppo J, Saarakkala S, Lappalainen R, Jurvelin JS, Nieminen MT. Proteoglycan and Collagen Sensitive MRI Evaluation of Normal and Degenerated Articular Cartilage. Journal of Orthopaedic Research. 2004;22(3):557-64.

369. Yao W, Qu N, Lu Z, Yang S. The Application of T1 and T2 Relaxation Time and Magnetization Transfer Ratios to the Early Diagnosis of Patellar Cartilage Osteoarthritis. Skeletal Radiology. 2009;38(11):1055-62.

370. Keenan KE, Besier TF, Pauly JM, Han E, Rosenberg J, Smith RL, Delp SL, Beaupre GS, Gold GE. Prediction of Glycosaminoglycan Content in Human Cartilage by Age, T1ρ and T2 MRI. Osteoarthritis and Cartilage. 2011;19(2):171-9.

371. Wong CS, Yan CH, Gong NJ, Li T, Chan Q, Chu YC. Imaging Biomarker with T1ρ and T2 Mappings in Osteoarthritis – in Vivo Human Articular Cartilage Study. European Journal of Radiology. 2013;82(4):647-50.

372. Messner K, Fahlgren A, Ross I, Andersson B. Simultaneous Changes in Bone Mineral Density and Articular Cartilage in a Rabbit Meniscectomy Model of Knee Osteoarthrosis. Osteoarthritis and Cartilage. 2000;8(3):197-206.

373. Prasadam I, Mao X, Shi W, Crawford R, Xiao Y. Combination of Mek-Erk Inhibitor and Hyaluronic Acid Has a Synergistic Effect on Anti-Hypertrophic and Pro-Chondrogenic Activities in Osteoarthritis Treatment. Journal of Molecular Medicine. 2013;91(3):369-80.

374. Mann HB. Nonparametric Tests against Trend. Econometrica. 1945;13(3):245-59.

375. Kendall MG. Rank Correlation Methods. Oxford, England: Griffin; 1948.

376. Spearman C. The Proof and Measurement of Association between Two Things. International Journal of Epidemiology. 2010;39(5):1137-50.

References

169

377. Guermazi A, Alizai H, Crema MD, Trattnig S, Regatte RR, Roemer FW. Compositional MRI Techniques for Evaluation of Cartilage Degeneration in Osteoarthritis. Osteoarthritis and Cartilage. 2015;23(10):1639-53.

378. Mow V. Fundamentals of Articular Cartilage and Meniscus Biomechanics. Articular cartilage and knee joint function. 1990:1-18.

379. Weiss C, Rosenberg L, Helfet AJ. An Ultrastructural Study of Normal Young Adult Human Articular Cartilage. Journal of Bone and Joint Surgery. 1968;50(4):663-74.

Appendix 1

170

Appendix 1: Supporting Information for Chapter 4

_____________________________________________________________________

SUPPORTING INFORMATION

FIGURE S1 Comparison of slice mammograms of a A, fresh and B, frozen breast tissue slice.

The two images are of the same physical slice; image A was obtained from the fresh slice

immediately after excision; image B was obtained from the frozen slice following a 1‐ year 9‐month storage at –80⁰C. The slice shown was not used in the main part of this study but is

representative of the breast tissue slices used. Freezing-and-thawing cycle causes slight changes

in the topography of the sample and local nonuniformity of the sample thickness; any areas thus

affected were avoided when selecting the measurement regions. The red circles show the HMD

and LMD regions of interest (ROIs) selected by the radiologist to match the same topographical

features in the fresh and frozen sample. The areas of the ROIs were A, 20.4 mm2 (LMD) and

3.8 mm2 (HMD); B, 13.2 mm2 (LMD) and 7.5 mm2 (HMD). The absorbed dose per unit mass

was A, 2452 ± 41 Gy (LMD) and 3052 ± 79 Gy (HMD); B, 2477 ± 76 Gy (LMD) and 3089 ±

137 Gy (HMD). The absorbed doses are similar between the fresh and the frozen sample,

indicating that freezing and prolonged storage at –80⁰C do not have a significant effect on the

distribution of the mammographic density of the sample. HMD, high mammographic density;

LMD, low mammographic density

Appendix 1

171

FIGURE S2 Effect of the ILT regularization parameter α on the computed ILT spectra: A, The

main plot is a representative CPMG dataset with n = 4000 echoes. Each sample point

corresponds to one echo integrated from −8 s to +8 s from the echo centre. The SNR value is

18, which is representative of the remaining data sets. The inset shows the plot of χ2 versus the

regularization parameter for a wide range of α values (see section 2.4 in the main text). This

plot is approximately L‐shaped. The corner of the “L”, which was selected after visual

inspection as the point of the apparent maximum of the second derivative of the plot,

corresponds to the optimal range of α in the ILT. The circled points labelled b, c, and d in the

inset correspond to the values of α used to compute the ILT spectra in panels B, C, and D,

respectively. B, An underregularized ILT spectrum computed with α set too low. This makes

the ILT smooth the physical features of the T2 spectrum as well as the noise; the resulting

oversmoothed spectrum does not reliably distinguish between the fat and water T2 peaks). A

properly regularized ILT spectrum with the in the optimal range. This spectrum reliably

distinguishes between the fat and water T2 peaks without introducing spurious peaks). An

overregularized ILT spectrum with the α set too high, making the ILT overly sensitive to noise

and resulting in the introduction of spurious T2 peaks. HMD, high mammographic density; ILT,

inverse Laplace transform; LMD, low mammographic density

Appendix 1

172

Table S1: The most probable T2 value, AF and gmT2 computed from both water and fat

peaks of the T2 distributions measured from excised HMD regions before and after H2O-

D2O replacement. The individual distributions represent measurements at a specific

depth within a given slice: Patient 1-Slice 1-Depth 1 (P1-S1-D1), Patient 1-Slice 1-



D2), Patient 2-Slice 1-Depth 1 (P2-S1-D1), Patient 3-Slice 1-Depth 1 (P3-S1-D1) and

Patient 3-Slice 1-Depth 2 (P3-S1-D2).

Excised HMD Region

Water peak Fat peak

Most

probable

T2 (ms)

AF (%) gmT2 (ms)

Most

probable T2

(ms)

AF (%) gmT2

(ms)

P1-S1-D1 8.7 26.49 9.42 81.1 72.99 77.94

P1-S1-D2 9.55 27.99 9.45 81.1 71.67 83.39

P1-S2-D1 7.92 22.42 10.03 73.90 75.78 74.87

P1-S2-D2 10.5 22.95 10.87 81.1 76.74 81.23

P1-S3-D1 12.6 50.95 12.76 73.9 48.80 81.02

P1-S3-D2 9.55 40.06 9.57 73.9 59.91 76.35

P2-S1-D1 8.70 46.96 8.87 67.3 53.04 67.27

P3-S1-D1 16.70 65.71 13.35 97.70 34.29 94.19

P3-S1-D2 11.50 55.48 12.66 89.00 44.03 88.13

Excised HMD Region after H2O-D2O Replacement

Water peak Fat peak

Most

probable

T2 (ms)

AF (%) gmT2 (ms)

Most

probable T2

(ms)

AF (%) gmT2

(ms)

P1-S1-D1 - - - 73.9 97.66 74.86

P1-S1-D2 - - - 81.1 99.80 75.69

P1-S2-D1 - - - 67.3 95.13 72.93

P1-S2-D1 - - - 73.9 99.40 76.86

P1-S3-D2 - - - 81.1 97.74 64.93

P1-S3-D2 - - - 89 98.63 65.95

P2-S1-D1 4.98 9.57 5.01 61.4 90.43 67.11

P3-S1-D1 - - - 81.1 98.08 61.47

P3-S1-D2 - - - 67.3 96.24 69.56

Appendix 1

173


peaks of the T2 distributions measured from excised LMD regions before and after H2O-

D2O replacement. The individual distributions represent measurements at a specific

depth within a given slice: Patient 1-Slice 1-Depth 1 (P1-S1-D1), Patient 1-Slice 1-



D2), Patient 2-Slice 1-Depth 1 (P2-S1-D1), Patient 3-Slice 1-Depth 1 (P3-S1-D1) and

Patient 3-Slice 1-Depth 2 (P3-S1-D2).

Excised LMD Region

Water peak Fat peak

Most

probable T2

(ms)

AF

(%) gmT2 (ms)

Most

probable T2

(ms)

AF (%) gmT2

(ms)

P1-S1-D1 11.5 7.80 12.46 73.9 92.20 73.97

P1-S1-D2 10.5 10.28 15.61 73.9 89.65 73.09

P1-S2-D1 11.5 9.61 8.92 73.9 88.78 74.22

P1-S2-D2 10.5 7.93 9.81 73.9 91.65 75.85

P1-S3-D1 12.6 12.04 12.58 73.9 84.99 73.82

P1-S3-D2 10.5 13.57 10.79 81.1 86.15 83.65

P2-S1-D1 16.7 22.04 13.18 81.1 77.95 73.58

P3-S1-D1 13.8 14.11 13.97 81.1 85.89 77.94

P3-S1-D2 16.7 8.72 17.18 81.1 89.34 75.11

Excised LMD Region after H2O-D2O Replacement

Water peak Fat peak

Most

prob

able

T2

(ms)

AF (%) gmT2 (ms)

Most

probable T2

(ms)

AF (%) gmT2 (ms)

P1-S1-D1 - - - 81.1 96.19 80.77

P1-S1-D2 - - - 81.1 99.78 79.19

P1-S2-D1 - - - 67.3 99.38 75.74

P1-S2-D2 - - - 61.4 100 77.03

P1-S3-D1 - - - 67.3 99.09 70.10

P1-S3-D2 - - - 61.4 99.99 75.94

P2-S1-D1 - - - 97.7 100 63.25

P3-S1-D1 - - - 61.4 99.99 74.98

P3-S1-D2 - - - 67.3 100 76.43

Appendix 1

174


peaks of the T2 distributions measured from HMD and LMD regions of full-slice

samples. The individual distributions represent measurements at a specific depth within

a given slice: Patient 1-Slice 1-Depth 1 (P1-S1-D1), Patient 1-Slice 1-Depth 2 (P1-S1-

D2), Patient 1-Slice 2-Depth 1 (P1-S2-D1), Patient 1-Slice 2-Depth 2 (P1-S2-D2),

Patient 1-Slice 3-Depth 1 (P1-S3-D1), Patient 1-Slice 3-Depth 2 (P1-S3-D2), Patient

2-Slice 1-Depth 1 (P2-S1-D1), Patient 3-Slice 1-Depth 1 (P3-S1-D1) and Patient 3-

Slice 1-Depth 2 (P3-S1-D2).

HMD Region in Full Breast Slice

Water peak Fat peak

Most

probable

T2 (ms)

AF

(%) gmT2 (ms)

Most

probable T2

(ms)

AF (%) gmT2

(ms)

P1-S1-D1 8.70 29.84 9.17 61.40 70.15 75.98

P1-S1-D2 8.70 25.88 7.99 81.10 74.11 71.12

P1-S2-D1 9.55 49.79 9.97 73.90 49.32 79.42

P1-S2-D2 11.50 43.25 11.42 89.00 56.20 81.70

P1-S3-D1 10.50 44.11 11.13 81.10 55.04 69.75

P1-S3-D2 11.50 40.33 13.01 81.10 58.70 79.64

P2-S1-D1* 10.50 79.96 15.33 118.00 20.04 116.10

P3-S1-D1 11.50 22.29 9.83 61.40 77.71 66.32

P3-S1-D2 8.70 35.79 8.51 73.90 62.42 76.21

LMD Region in Full Breast Slice

Water peak Fat peak

Most

probable

T2 (ms)

AF (%) gmT2

(ms)

Most

probable T2

(ms)

AF (%) gmT2

(ms)

P1-S1-D1 11.50 5.80 11.50 61.40 94.19 73.71

P1-S1-D2 - - - 89.00 97.09 73.29

P1-S2-D1 8.70 6.43 8.50 67.30 91.68 74.43

P1-S2-D2 - - - 81.10 99.80 70.17

P1-S3-D1 - - - 55.90 94.09 66.67

P1-S3-D2 6.58 14.14 6.32 73.90 84.99 80.57

P2-S1-D1 - - - 89.00 100 55.32

P3-S1-D1 7.92 8.71 7.66 61.40 91.28 74.17

P3-S1-D2 - - - 97.7 99.97 71.12

* No inflection point was present in the curve of χ2that was computed for this sample.

This T2 relaxation distribution (χ2 = 1.15) was excluded from Figure 6 and Figure 7.

Appendix 2

175

Appendix 2: Preliminary Investigation for Chapter 5

_____________________________________________________________________

This section presents the methods that were used to obtain T1, T2 and T2*

weighted MR echoes and measure corresponding parametric maps by scanning three

rat knee joints. It also presents the preliminary results obtained from the MRI. The

imaging was conducted as part of a larger study that aimed to establish a MRI only

protocol for whole knee joint evaluation [97]. The particular objective of this section

was to identify a MRI sequence that allows imaging of all tissues of a whole rat knee

joint within a practical time frame and permits the measurement of MR parameters that

are sensitive to the tissues of the knee joint.

A2.1 Methods

A2.1.1 Development of Rat OA Model

A rat PTOA model was chosen for this study (presented in Chapter 5). Animal

ethics approval for this project was granted by the Queensland University of

Technology (QUT) and the Prince Charles Hospital Ethics Committees (QUT Ethics

approval number: 0900001134). All methods were carried out in accordance with the

relevant guidelines and regulations of QUT. A total of 30 Male Wistar Kyoto rats (11-

12 weeks old, 300-350 grams weight) were purchased from the Medical Engineering

Research Facility (MERF, Brisbane, Australia) and housed in controlled day-night

cycle (light/dark, 12/12 h) and controlled temperature (23 ± 1 ˚C).

The control group consisted 6 rats that did not undergo surgery (CTRL joint).

PTOA was induced in the remaining 24 rats by complete removal of medial meniscus

or by complete medial meniscectomy on the right hind knee joint (MSX joint). No

surgery was carried out on the contralateral joints of the left hind knee (CLAT joint).

The rats were allowed to walk freely in the cage after surgery. Pain killers

(Buprenorphine 0.05 mg/kg) and antibiotics (Gentamycin 5 mg/kg) were given to the

rats to ease the pain and discomfort. During the next 8 weeks following surgery, 3 rats

Appendix 2

176

were sacrificed every week (week 1 – week 8) and 6 knee samples were harvested: 3

samples from the MSX joints and 3 samples from the CLAT joints. The samples were

kept frozen (-20⁰C) after accrual, defrosted prior to MRI experiments and kept at room

temperature during the scanning.

Each whole-joint knee sample extended from the middle of femoral diaphysis

to the middle of tibial diaphysis (Figure 1 of [97]). The muscle and skin surrounding

the knee joint were left intact in order to maintain an anatomically realistic

environment. The three CLAT joints harvested at week 4 were preliminary scanned

using three different MRI sequences in order to identify the suitable sequence for whole

knee joint evaluation. These CLAT joints were also scanned using varying imaging

parameters in order to standardise the imaging protocol that will be used for the

assessment of whole joint PTOA [97]. This section presents the MRI methods used for

imaging week-4 CLAT joints and the preliminary results obtained from the same joints.

A2.1.2 MRI Protocol

MR images were acquired at room temperature using a Bruker Avance NMR

spectrometer (Bruker, Germany) at 7 T using 1.5 T m-1 (150 G cm-1) triple-axis gradient

set, a Micro2.5 microimaging probe and a 25 mm radiofrequency (RF) birdcage 1H

resonator coil. Each knee joint sample was hydrated for 2 hours in 0.01 M phosphate

buffered saline (Sigma-Aldrich, USA) in order to maintain physiological osmotic

conditions in the tissues imaged. Each sample was placed in a 25 mm NMR tube. Two

Teflon plugs, each with flat surfaces and a hole at the centre, secured the position of the

sample as shown in Figure 1. The top part of femur and the bottom part of tibia were

inserted in the holes of the Teflon plugs. The axis of the limb was positioned

approximately parallel to the NMR tube axis and the static magnetic field (B0). By

repetitive imaging of the CLAT joints using various imaging parameters, and by

analysing the resulting MR images, this phase attempted to determine the optimum

choices for slice thickness, slice location, field of view (FOV) and image resolution.

The initial Proton Density (PD) weighted MRI of whole CLAT joints showed

that: 1) ligaments, subchondral bone, and cortical bone had very low signal intensity;

2) the volume or the cross section of synovial fluid was very small; and 3) the MR

signal obtained from the meniscal region was only sufficient for structural information.

Appendix 2

177

Therefore, for quantitative MRI, the imaging slices were selected in order to obtain

clear cross sections of cartilages, growth plates, menisci, and subchondral bones of the

femoral and tibial epiphyses. MRI imaging planes were specified in the axial, coronal,

and sagittal orientation as shown in Figure 1 (a), (b), and (c), respectively. Multiple

slices were imaged at each orientation to cover the whole joint area as shown by the

dotted lines. Although Figure 1 shows the ideal scenario for sample placement, the

actual joints were surrounded by soft tissues including ligaments, fat pads and muscles

and therefore the knee joints were never as straight as shown in the figure. Slight

angular adjustments were necessary while specifying slices for MRI. Figure 2 shows

T2 weighted images obtained along the axial, coronal, and sagittal orientations.

Figure 1. Cartoon sketch of a mouse knee joint fixed by two Teflon plugs in a 25 mm NMR

tube. The sample is immersed in PBS. Imaging planes are shown by dotted lines along axial

(a), coronal (b), and sagittal (c) orientations. The schematic outline of the knee in the inset is

reproduced from https://en.wikipedia.org/wiki/Knee#/media/File:Knee_skeleton_lateral_ante-

-rior_views.svg in accordance with the terms of the CC BY 2.5 license.

The optimum slice thickness was determined by examining the imaging

outcomes with varying slice thicknesses. For this, MSME sequence was used with TE

= 6 ms, 20 echoes, TR = 2 s, 30x30 mm FOV, 256x256 image matrix, and 8 averages.

Slice thicknesses of 0.25 mm - 1 mm were specified with increments of 0.25 mm.

PBS

PBS

PBS

Appendix 2

178

Examples are shown in Figure 3, where a – d present the second echoes (TE = 12 ms)

of a coronal slice with thicknesses of 0.25 mm, 0.5 mm, 0.75 mm, and 1 mm. The train

of echoes were checked for signal intensity by the image viewer of ParaVision. A

minimum slice thickness of 0.5 mm preserved sufficient amount of signals for good

SNR. To keep partial voluming effect at the minimum, slice thickness of 0.5 mm was

chosen for the remaining MRI experiments of this project.

Figure 2. First echoes obtained by MSME sequence (TE = 6 ms) in axial (a), coronal (b), and

sagittal (c) orientation.

MR images were obtained with varying FOVs ranging from 20 mm x 20 mm to

40 mm x 40 mm. The diameter of the NMR tube was 25 mm for rat knee samples.

Aliasing or wrap over artefact was observed when the FOV was equal or smaller than

the imaging object. Figure 4a shows the wrap over artefact identified on a sagittal slice

for a FOV of 20 mm x 20 mm. There was no limitation on increasing the FOV size,

however, the regions outside the sample did not contain any meaningful information.

Therefore, 25mm x 25 mm FOV was chosen for the sagittal and axial MR slices.

However, in case of the coronal slices, the sample was positioned close to the centre of

the FOV and the wrap over effect on the sides of FOV did not influence the MR

measurements obtained from the sample. Therefore, 20 mm x 20 mm FOV was chosen

for the coronal MR slices.

The resolution of a MR image is determined by the choice of FOV and the size

of image matrix. Isotropic voxel size is beneficial for image processing. A small voxel

size is beneficial for imaging tissues without partial voluming effect. On the other hand,

too small a voxel may result in poor SNR and longer acquisition time. Figure 5 shows

the second echo of a MSME sequence of a sagittal slice with 128x128, 256x256, and

512x512 image matrices. With 8 averages, acquisition of 128x128 image matrix

Appendix 2

179

required 34 minutes, 256x256 image matrix required 1 hour and 8 minutes, and

512x512 image matrix required 2 hour and 16 minutes. Considering these, the matrix

size 256x256 was chosen for this project.

Figure 3. Second echoes obtained by a MSME sequence (TE = 12 ms) of a coronal slice with

thickness of 0.25 mm (a), 0.5 mm (b), 0.75 mm (c), and 1 mm (d).

Figure 4. Second echo of MSME sequence (TE = 12 ms) of a sagittal slice with FOV of 20x20

mm (a), 25x25 mm (b), and 30x30 mm (c).

A good signal to noise ratio (SNR) is essential for quantitative MRI. The two

options for increasing the SNR are: 1) to enhance the signal itself or 2) to reduce the

noise. For a fixed voxel size, the signal intensity depend on the imaging sequence.

Appendix 2

180

However, noise can be reduced by increasing the signal averages. Considering the

imaging time, 8 averages were chosen for T2 and T2* weighted sequences and 4

averages were chosen for T1 weighted sequence. The SNR was computed by taking the

ratio of the mean pixel intensity in a region of interest (ROI) within the sample to the

noise amplitude in a ROI of the background air (noise-only region). The noise

amplitude was computed as the square root of the sum of the squared mean and the

squared standard deviation of the signal in a noise-only region in a magnitude image

(noise and noise, respectively):

2 2

noise noiseNoise (1)

This measurement was repeated for ROIs in cartilage, muscle and tibial

epiphysis.

Figure 5. Second echo of MSME sequence of a sagittal slice with image matrix of 128x128

(a), 256x256 (b), and 512x512 (c).

A2.1.3 Scanning by MRI and Image Processing

The CLAT joints were imaged using T1, T2 and T2* weighted MRI sequences in

the sagittal, coronal and axial orientations. Using custom designed MATLAB codes,

2D parametric maps (quantitative maps of R1, T1, R2, T2, R2* and T2

*) were computed

for all MRI slices. After the MRI experiments, the MR data were processed by custom

designed codes written in MATLAB (Mathworks Inc., version 2012) to compute

relaxations time / rate maps. Individual scripts were written for fitting the T1, T2 and

T2* weighted decays to the corresponding mono-exponential mathematical models of

relaxation decays. Bi-exponential fitting was not considered at this stage. Curve fitting

models were defined for unconstrained fitting by non‐linear least squares method with

optimization based on trust-region algorithm. The goodness of fit was checked by

Appendix 2

181

boundary conditions and Runs test at the 5% significance level. The voxels outside the

knee joint area were excluded from analysis.

A2.1.3.1 T1 weighted Imaging

Image Acquisition

T1 weighted images were obtained by MSMEVTR sequence provided by

Bruker, which is a modified version of MSME with fixed TE and variable TR. This

sequence is based on the concept of saturation recovery. In this sequence, the imaging

parameters are chosen to produces contrast mainly based on the T1 characteristics of

tissues by de-emphasizing the T2 contributions. This can be accomplished by using

relatively short TR to maximise the difference in longitudinal relaxation during the

return to equilibrium, and a short TE to minimise T2 dependency during signal

acquisition. For this study, the TE was specified at 4.24 ms and 1 echo was generated

for each TR. A list of TR was selected ranging from 300 ms to 13,000 ms. The TR values

were spaced to maintain equal distribution of echo peaks across the rise, neck, and

plateau of the longitudinal relaxation curve. The shortest TR was placed approximately

right after the decay while the longest TR was at the signal plateau. With sequential

ordering, 3 slices were imaged simultaneously during each sequence. The T1 weighted

sequences requires long acquisition time which may risk the biodegradation process for

samples. Therefore, only one CLAT sample went through the T1 weighted imaging only

in one orientation.

Relaxation Mapping

The MSMEVTR sequence acquires one echo for every TR for a set of different

TR values. By default, ParaVision computes the echo for the longest TR first and repeats

is for all imaging slices, then acquires the echo for the second longest TR and so on.

For relaxation mapping, the echoes were re-organised so that for each voxel of the

imaging slice, the signal intensities follow the T1 weighted exponential curve as shown

in Figure 6b. This decay can be described by the equation below.

𝑆 = 𝑆0( 1 − 𝑒−(𝑅1𝑡)) + 𝑆𝑜𝑓𝑓𝑠𝑒𝑡 (2)

Here, S is the signal amplitude measured at time t (t = TE at variable TR values),

S0 is the full signal intensity at time 0, R1 is longitudinal relaxation rate constant (R1 =

Appendix 2

182

1 𝑇1⁄ ) and S0ffset is the magnitude of noise. With measured S and known t, the value for

S0, T1 and Soffset are obtained by least-square iterative fitting of the above equation to

the measured decay. The fit results were checked for the following boundary

conditions: 1) S0 is positive; 2) S0 is less than the double of maximum signal intensity;

3) R1 is positive and 0.0001 < R1 < 1; 4) S0ffset is positive and less than one fifth of S0,

and 5) the fit was achieved within 100 iterations. The voxel results that failed to fulfil

these conditions were rejected and R1 values of those voxels were forced to null. Then

the residuals were computed for each fit. The residuals went through Run test for

randomness at the 5% significance level. The residuals (computed from the voxel fits)

that passed this test with z value less than 1.96 were accepted and the associated R1 was

plotted in the 2D relaxation map of R1. The R1 was nulled if the associated residuals

failed the Run test of randomness. Figure 7 presents a flowchart describing this

procedure. This test confirmed that the data was free from systematic data fluctuation

due to technical errors.

Spatial maps of R1 and T1 were obtained by repeating this procedure in a voxel-

by-voxel basis for the entire MR imaging slice. Figure 6 shows the first echo of a

sagittal slice with the longest TR, where a voxel in tibial epiphysis is highlighted. The

signal intensities measured from that particular voxel at variable TRs are shown by blue

dots and the resultant fit is shown in red in Figure 6b. The residuals of this fit are shown

in Figure 6c. After checking the boundary conditions and the results of the Run test, a

R1 relaxation map was computed including all the voxels in the knee joint area. This

process was repeated for all MR images obtained by T1 weighted sequences.

Figure 6. First echo of MSMEVTR sequence of a sagittal slice, the voxel selected for analysis

is highlighted in yellow (a), signal magnitude measured at 22 echo peaks (for 22 different values

of TR) shown in blue and mathematically calculated fit shown in red (b), and the fitting residuals

of the fit in b (c).

Appendix 2

183

Figure 7. The data fitting method for T1 weighted decay. This method was repeated for the data

acquired from every voxel of an imaging plane. The mathematical model was defined for

unconstrained fitting by non‐linear least squares method with optimization based on trust-

region algorithm.

A2.1.3.2 T2 weighted Imaging

Image Acquisition

T2 weighted images were obtained by MSME pulse sequence provided by

Bruker. This is a spin echo [140] pulse sequence designed based on typical CPMG [41,

139]. In this sequence, the initial 90˚ excitation pulse is followed by multiple 180˚

refocussing pulses at regular intervals, which generates multiple spin echoes. The

amplitudes of the echoes follow the envelope of the T2 weighted transverse relaxation

decay. Here, Sinc3 shaped pulses were used for both excitation and refocusing pulses.

Using ParaVision, the signals from bones and cartilages were observed to reach the

Fit measured signal S to

𝑆 = 𝑆0( 1 − 𝑒−(𝑅1𝑡)) + 𝑆𝑜𝑓𝑓𝑠𝑒𝑡

Converge in 100

iterations?

Compute boundary conditions

Realistic fit?

Run test on fit residuals

Random?

Map R1 on relaxation map

Discard the result, R1 = 0

Y

N

Y

Y

N

N

Appendix 2

184

noise floor after 200 ms. Two sets of parameters were initially chosen for MSME

sequence: 25 echoes with 8 ms TE and 8 ms spacing and 32 echoes with 6 ms TE and

6 ms spacing. The second choice (TE = 6 ms and 6 ms spacing) was found to be more

effective for curve fitting and therefore was chosen for T2 imaging. The TR was set to

2 s. Multiple slices were imaged using the MSME sequence with sequential ordering

of slices and a train of T2 weighted echoes were obtained for each slice.

Relaxation Mapping

The T2 weighted relaxation decay can be described by the following equation:

𝑆 = 𝑆0 𝑒−𝑡𝑅2 + 𝑆𝑜𝑓𝑓𝑠𝑒𝑡 (3).

Here, R2 is the transverse relaxation rate constant (1 𝑇2⁄ ), S is the signal

amplitude measured at time t (t = n x TE, n = 1, 2, 3, ...), S0 is the full signal intensity

at time 0 and S0ffset is the magnitude of noise. The above equation was fitted to the

measured values of S by non-linear least square fitting with maximum 100 iterations.

Figure 8 shows the signal measured from a voxel of a T2 weighted image and the

exponential decay curve model that was fitted to the measured data points.

Figure 8. The first echo of MSME sequence of a coronal slice (TE = 6 ms), the voxel selected

for analysis is highlighted in yellow (a), signal magnitude at 25 echo peaks shown in blue and

mathematically calculated fit shown in red (b), and the residuals of the calculated fit (c).

The fit results were checked for the following boundary conditions: 1) S0 is

positive; 2) S0 is less than the double of maximum signal intensity; 3) R2 is positive and

0.0001 < R2 < 1; 4) S0ffset is positive and less than one fifth of S0, and 5) the fit was

achieved within 100 iterations. The voxel results that failed to fulfil these conditions

were rejected and R2 values of those voxels were forced to null. Then the residuals were

computed for each fit. The residuals went through Run test for randomness at the 5%

significance level. The residuals (computed from the voxel fits) that passed this test

with z value less than 1.96 were accepted and the associated R2 was plotted in the 2D

Appendix 2

185

relaxation map of R2. The R2 was nulled if the associated residuals failed the Run test

of randomness.

A2.1.3.3 T2* weighted Imaging

Image Acquisition

Multiple Gradient Echo (MGE) pulse sequence by Bruker was applied for T2*

weighted imaging. In a perfectly homogeneous magnetic field, the transverse relaxation

process for a spin system with single spectral component is represented by Eqn 2. Due

to the distortion of the main magnetic field upon the placement of an imaging object,

the local magnetic fields experienced by the 1H fluctuate, leading to the loss of phase

coherence between the protons. For a spin system with a large number of isochromats,

the protons quickly go out of phase owing to the differences in the precessing speeds.

The magnetic moments are cancelled by spin-spin interaction and the transverse

magnetization decays with time constant T2* < T2. A gradient echo pulse sequence

applies a small flip angle followed by magnetic gradients that first dephase and then

rephase the isochromats generating echoes of the T2* weighted decaying signal. For

this project, sinc-3 shaped pulse with 30˚ flip angle was applied first. Then 12 echoes

were generated and measured starting from 4 ms with 6 ms increments. The TR was set

to 2 s. This sequence computed a 128x256 image matrix, which was then remapped

onto 256x256 grid. Multiple slices were imaged simultaneously with interlaced

ordering.

Relaxation Mapping

For gradient echoes, the relaxation rate constant is R2*. The amplitudes of the

gradient echoes generated by MGE sequence follow the envelope of an exponential

curve as shown here.

𝑆 = 𝑆0 𝑒−𝑡𝑅2

∗+ 𝑆𝑜𝑓𝑓𝑠𝑒𝑡 (4)

Here, the echo amplitudes measured at 12 time points are represented by S, S0

is the full signal intensity at time 0 and S0ffset is the magnitude of noise. The above

equation was fitted to the measured data points by non-linear least square method. The

fit results were checked for the following boundary conditions: 1) S0 is positive; 2) S0

is less than the double of maximum signal intensity; 3) R2* is positive and 0.0001 < R2

*

Appendix 2

186

< 1; 4) S0ffset is positive and less than one fifth of S0, and 5) the fit was achieved within

100 iterations.

Figure 9a shows the first gradient echo of a coronal slice. A voxel location is

highlighted in yellow. The signal intensity observed at that voxel and the resultant fit is

shown in Figure 9b. The residuals, the differences between the actual measurements

and the fit are shown in Figure 9c. After checking the results for the boundary

conditions mentioned above, a R2* relaxation map was computed with the R2

* values

measured from individual voxels.

Figure 9. The first echo obtained by a MGE sequence of a coronal slice (a), the voxel selected

for analysis is highlighted in yellow, signal magnitude at 12 echo peaks shown in blue and the

mathematically calculated fit shown in red (b), and the residuals of the calculated fit (c).

A2.2 Results

A2.2.1 Quantitative T1 Analysis

Figure 10 shows a T1 weighted sagittal slice obtained by using the MSMEVTR

sequence. It also shows the quantitative R1 map computed from the T1 weighted echoes

obtained from the same imaging slice. It was observed that many pixels (in R1 map)

corresponding to the cartilage region were black or nulled. Misfit of the voxel decay or

poor quality of the MR signal are the probable reasons for such results. Additionally,

the tissues of the knee joint could not be identified or delineated based on the pixel R1

values of the quantitative R1 map. Similar results were obtained from all quantitative

R1 maps that were computed from T1 weighted echoes.

The acquisition of T1 weighted MRI with good resolution (98 x 98 µm pixel)

required several hours (> 4 hours). The soft tissues of the knee joint samples were partly

degraded during the period of image acquisition. Therefore, among the T1 weighted

echoes obtained from the same sample, the structure and the composition of the imaged

Appendix 2

187

tissues were not consistent. This may have resulted in misfit for the longitudinal

relaxation decays. In order to avoid degradation of the sample, it was necessary to

reduce the scan time. However, for using the same imaging sequence, the imaging time

can only be reduced by decreasing the image resolution or by reducing the number of

averages in signal acquisition. These changes may result in significant increase in the

partial voluming artefact and in low SNR. Considering these issues, T1 weighted

imaging was discontinued and not used for imaging of the CTRL, MSX and the

remaining CLAT joints.

Figure 10. The First echo obtained by MSMEVTR sequence, the region selected for analysis

is outlined by a blue rectangle (a), the R1 relaxation rate map computed from 22 echoes (b),

and the results of the Run test results where 0 (black) = pass, 1 (white) = fail (c).

A2.2.2 Quantitative T2 Analysis

Figure 11 shows the first T2 weighted echo (TE = 6 ms) of a coronal slice

measured by a MSME sequence, the R2 relaxation map computed from the same

imaging slice and the results of the Run test computed for the R2 map, where 0 (black)

indicates pass and 1 (white) indicates fail. It was observed that the voxels that failed

boundary condition or Run test were randomly distributed across the imaging plane. If

a cluster of such voxels was observed, the fitting was procedure was repeated for bi-

exponential fitting.

In comparison to the T1 and T2* weighted MRI, the T2 weighted echoes obtained

at the coronal plane showed the clearest cross section of the cartilage and the bones of

the knee joint. Therefore, the T2 weighted echoes were used for identifying the tissues

Appendix 2

188

of the knee joint. Automatic segmentation of the knee joint components by edge

detection and region growing algorithms were attempted. Figure 12 shows a

quantitative R2 map and the result of edge detection on the same map. The edges

between the tissue types were not continuous. For the same reason, the region growing

algorithm also failed to automatically segment the components. Therefore, the voxel

selection and region of interest (ROI) drawing algorithms were implemented for

selecting the knee joint tissues for further analysis.

Figure 11. The first echo obtained by MSME sequence (TE = 6ms) along the coronal plane,

the region selected for analysis is outlined by a blue rectangle (a), the R2 relaxation map

computed from 25 echoes (b), and the results of Run test where 0 (black) = pass, 1 (white) =

fail (c).

Figure 12. R2 relaxation map of a coronal slice (a) and the result from edge detection (b).

Next, the quantitative T2 values that corresponded to the regions in cartilage,

growth plate, and trabecular bone of epiphysis, metaphysis, and diaphysis were

identified and plotted. Figure 13 shows the highlighted pixels in the AC layer of a T2

Appendix 2

189

weighted echo, which were selected for quantitative T2 analysis, and the T2 distribution

computed from the corresponding voxels. There, the longitudinal axis presented T2

times in milliseconds. Although the horizontal axis had no scale, the data points were

distributed by the ‘jitter’ method to minimise overlapping. Figure 14 shows a closed

ROI drawn in the tibial epiphysis. This region was specified on a T2 weighted echo for

quantitative T2 analysis. Then a T2 distribution was computed from the corresponding

voxels within the selected region. Figure 15 shows the trabecular bone regions specified

in the metaphysis and diaphysis of tibia and the T2 distribution measured from the

voxels corresponding to the selected region. Following the same procedure, T2

distributions were computed from multiple tissues and regions of CLAT joints as shown

in Figure 16.

Figure 13. Voxels of AC are highlighted in purple (a) and the T2 distribution computed from

the data of corresponding voxels (b). Here the longitudinal axis presents T2 times in

milliseconds while the horizontal axis has no scale. ‘Jitter’ method has been used for

distributing the points to minimise overlaps.

Appendix 2

190

Figure 14. In a T2 weighted echo (TE = 6ms), a region is outlined in the tibial epiphysis by a

closed ROI (a). The T2 distribution computed from the voxel T2 measurements obtained from

voxels within the outlined region (b).

Figure 15. In a T2 weighted echo (TE = 6ms), trabecular region is outlined in the tibial

metaphysis and diaphysis by a closed ROI (a). The T2 distribution computed from the voxel T2

measurements obtained from voxels within the outlined region (b).

Appendix 2

191

Figure 16. T2 distributions computed from tissues of two CLAT joints.

A2.2.3 Quantitative T2* Analysis

Figure 17 shows the first T2* weighted echo obtained by a MGE sequence, the

R2* relaxation map computed from the T2

* weighted echoes of the same imaging slice,

and the results of Run test for voxel based fitting where 0 (black) indicates pass and 1

(white) indicates fail. It was observed that the voxel intensities of T2* weighted echoes

were appropriate for identifying cartilage and measuring cartilage thickness. However,

voxel T2* measurements were not sensitive to the distribution of 1H in tissues. Unlike

quantitative T2 analysis, the measurements obtained by quantitative T2* analysis were

not sensitive to the water micro-environment in the knee joint tissues and the tissues

could not be identified or classified based on the voxel T2* measurements.

Appendix 2

192

Figure 17. The first echo obtained by a MGE sequence, the region selected for analysis is

outlined by a blue rectangle (a), the relaxation map of R2* computed from 12 echoes (b), and

the results of Run test where 0 = pass, 1 = fail (c).

A2.3 Conclusions

This section have presented the methods of the preliminary MRI experiments

and the image processing that were used to investigate the tissues of three whole knee

joints of rats. It was part of a larger study that aimed to establish a MRI only protocol

for quantitative evaluation of whole joint PTOA in rat knee joints [97]. The results

obtained from this section were used to determine the suitable imaging protocol and

analysis procedure for scanning and analysing the remaining rat knee joints for the

above mentioned study. The results presented in this section have demonstrated that T1

weighted imaging was incompatible for whole knee joint imaging in this case due to

the long time required for imaging at a good resolution. At the same time, T2* weighted

imaging was quick but the parametric T2* maps were not sensitive to the variations in

the water content among different tissues of the knee joint. Nevertheless, it was possible

to obtain sufficient T2 weighted echoes for quantitative T2 analysis approximately in 1

hour. The T2 weighted echoes allowed measurement of cartilage thickness while the

voxel T2 measurements were sensitive to the water content of the tissues of knee joint.

Therefore, considering the issues mentioned above, only T2 weighted images were

acquired for all control joints, all joints that were subjected to meniscectomy and the

remaining contralateral joints. In addition, considering the attainable resolution and the

accessibility to the tissues of knee joint (described in previous sections), T2 imaging

Appendix 2

193

was performed along the coronal plane with 20 x 20 mm FOV and 256 x 256 image

matrix. These transverse relaxation based MR images were then analysed to identify

the developmental pathway of PTOA [97].

Appendix 2

194

A2.4 References

1. Ali TS, Prasadam I, Xiao Y, Momot KI. Progression of Post-Traumatic

Osteoarthritis in rat meniscectomy models: Comprehensive monitoring using MRI.

Scientific Reports. 2018;8(1):6861.

2. Hahn E. Spin echoes. Physical Review. 1950:80(4):580-94.

3. Carr H, Purcell E. Effects of diffussion on free precession in nuclear magnetic

resonance experiments. Physical Review. 1954:94(3):630-38.

4. Meiboom S, Gill D. Modified Spin‐Echo Method for Measuring Nuclear

Relaxation Times. Review of Scientific Instruments. 1958;29(8):4.

Documents

Transverse Relaxation based Magnetic Resonance Techniques ... · Magnetic resonance imaging (MRI) is a medical imaging technique that allows non-invasive assessment of the microstructure