UNIVERSITY OF OULU P .O. Box 8000 F I -90014 UNIVERSITY OF OULU FINLAND

A C T A U N I V E R S I T A T I S O U L U E N S I S

Professor Esa Hohtola

University Lecturer Santeri Palviainen

Postdoctoral research fellow Sanna Taskila

Professor Olli Vuolteenaho

University Lecturer Veli-Matti Ulvinen

Director Sinikka Eskelinen

Professor Jari Juga

University Lecturer Anu Soikkeli

Professor Olli Vuolteenaho

Publications Editor Kirsti Nurkkala

ISBN 978-952-62-0763-6 (Paperback)ISBN 978-952-62-0764-3 (PDF)ISSN 0355-3221 (Print)ISSN 1796-2234 (Online)

U N I V E R S I TAT I S O U L U E N S I S

MEDICA

ACTAD

D 1286

ACTA

Jari Rautiainen

OULU 2015

D 1286

Jari Rautiainen

NOVEL MAGNETIC RESONANCE IMAGING TECHNIQUES FOR ARTICULAR CARTILAGE AND SUBCHONDRAL BONESTUDIES ON MRI RELAXOMETRY AND SHORT ECHO TIME IMAGING

UNIVERSITY OF OULU GRADUATE SCHOOL;UNIVERSITY OF OULU, FACULTY OF MEDICINE;MEDICAL RESEARCH CENTER OULU

A C T A U N I V E R S I T A T I S O U L U E N S I SD M e d i c a 1 2 8 6

JARI RAUTIAINEN

NOVEL MAGNETIC RESONANCE IMAGING TECHNIQUES FOR ARTICULAR CARTILAGE AND SUBCHONDRAL BONEStudies on MRI relaxometry and short echo time imaging

Academic dissertation to be presented with the assent of theDoctoral Training Committee of Health and Biosciences ofthe University of Oulu for public defence in AuditoriumMD100, Mediteknia Building of the University of EasternFinland (Yliopistonranta 1B, Kuopio), on 20 March 2015, at12 noon

UNIVERSITY OF OULU, OULU 2015

Copyright © 2015Acta Univ. Oul. D 1286, 2015

Supervised byProfessor Miika NieminenDocent Mikko Nissi

Reviewed byAssociate Professor Xiaojuan LiAssociate Professor Jiang Du

ISBN 978-952-62-0763-6 (Paperback)ISBN 978-952-62-0764-3 (PDF)

ISSN 0355-3221 (Printed)ISSN 1796-2234 (Online)

Cover DesignRaimo Ahonen

JUVENES PRINTTAMPERE 2015

OpponentAssociate Professor Bernard J. Dardzinski

Rautiainen, Jari, Novel magnetic resonance imaging techniques for articularcartilage and subchondral bone. Studies on MRI relaxometry and short echo timeimagingUniversity of Oulu Graduate School; University of Oulu, Faculty of Medicine; MedicalResearch Center OuluActa Univ. Oul. D 1286, 2015University of Oulu, P.O. Box 8000, FI-90014 University of Oulu, Finland

Abstract

Osteoarthritis is a common joint disease in the adult population characterized by degeneration andlimited repair capability of articular cartilage. Cartilage degeneration is also associated withremodeling and sclerosis of the subchondral bone. The relative contributions of these processesare not fully understood partly because of a lack of proper biomarkers for the diagnosis and follow-up of the disease progress. Current diagnostic methods are too insensitive or subjective for reliableand early diagnosis of osteoarthritis.

Novel magnetic resonance imaging (MRI) techniques can be used for indirect assessment ofcartilage constituents. In addition, the imaging of the subchondral bone and osteochondraljunction has become feasible, which has been very difficult with conventional MRI methods dueto the very fast signal (T2) decay in those tissues. Quantitative MRI of cartilage involvesmeasurement of relaxation time parameters (T1, T2, T1ρ, T2ρ, TRAFF, T1sat) and investigation ofhow the parameters reflect the properties of the cartilage. SWIFT (Sweep imaging with Fouriertransform) is an MRI technique capable of imaging short T2 structures, enabling assessment of theosteochondral junction as well as the subchondral bone. The aim of this work was to investigateosteoarthritic changes occurring in articular cartilage and subchondral bone in experimentalanimal models of osteoarthritis and degenerated human specimens with quantitative MRImeasurements and SWIFT imaging at 9.4 T. Moreover, the feasibility of the different quantitativeMRI techniques in the assessment of osteoarthritic changes was studied. For reference,biomechanical, quantitative and qualitative histology, biochemical and micro-computedtomography measurements were conducted.

Novel adiabatic T1ρ and T2ρ relaxation time parameters were more sensitive than conventionalmethods (T1, T2) or continuous wave T1ρ to both early degenerative changes in a rabbit model ofcartilage degeneration and to spontaneous degeneration of human cartilage. SWIFT enabledassessment of the structural properties of the subchondral bone. Moreover, signal changes relatedto the degree of osteoarthritis were detected at the cartilage–bone interface with SWIFT. Thesemethods may provide new tools for improving the diagnosis of early osteoarthritis. Further studiesare needed to investigate the clinical feasibility of these MRI techniques.

Keywords: articular cartilage, calcified cartilage, magnetic resonance imaging,osteoarthritis, relaxation time, subchondral bone, sweep imaging with Fourier transform

Rautiainen, Jari, Nivelruston ja rustonalaisen luun magneettikuvaus. Relaksaatio-aikakartoitus ja lyhyen kaikuajan kuvantaminenOulun yliopiston tutkijakoulu; Oulun yliopisto, Lääketieteellinen tiedekunta; Medical ResearchCenter OuluActa Univ. Oul. D 1286, 2015Oulun yliopisto, PL 8000, 90014 Oulun yliopisto

Tiivistelmä

Nivelrikko on yleinen nivelsairaus, johon ei tällä hetkellä ole parannuskeinoa. Nivelrikossanivelrusto rappeutuu asteittain aiheuttaen lopulta kudoksen tuhoutumisen; samanaikaisesti myösrustonalaisessa luussa sekä rusto–luurajapinnassa tapahtuu muutoksia. Nivelrustokudoksenuusiutumiskyky on erittäin heikko, joten vauriot olisi hyvä havaita varhaisessa vaiheessa. Nivel-rikon syntymekanismeja ei kuitenkaan vielä täysin ymmärretä, koska nykyiset seurantamenetel-mät eivät ole riittävän tarkkoja taudin etenemisen tarkkailemiseksi.

Magneettikuvausteknologian kehitys mahdollistaa kuitenkin nivelruston koostumuksen tar-kastelun nivelen ulkopuolelta. Kvantitatiivisen magneettikuvauksen avulla (T1, T2, T1ρ, T2ρ,TRAFF, T1sat relaksaatioaikaparametrit) voidaan mitata epäsuorasti nivelruston koostumusta jarakennetta. SWIFT (Sweep Imaging with Fourier Transform) on uusi lyhyen kaikuajan magneet-tikuvausmenetelmä, joka mahdollistaa myös erittäin lyhyen T2-relaksaatioajan omaavien kudos-ten, kuten rusto–luurajapinnan sekä rustonalaisen luun kuvaamisen. Työssä tutkittiin nivelrikos-sa syntyviä muutoksia nivelrustossa, rusto–luurajapinnassa ja rustonalaisessa luussa erilaisissaeläinmalleissa sekä ihmiskudoksessa. Lisäksi työssä tarkasteltiin eri magneettikuvaustekniikoi-den herkkyyttä havaita muutokset näissä kudoksissa. Nivelruston ja rustonalaisen luun ominai-suuksien määrittelemiseen käytettiin verrokkimenetelminä biomekaanisia, histologisia sekä bio-kemiallisia mittauksia ja tietokonetomografiakuvauksia.

Kvantitatiivisista magneettikuvausparametreista adiabaattinen T1ρ sekä adiabaattinen T2ρosoittautuivat herkimmiksi havaitsemaan muutoksia varhaisen nivelrikon eläinmallissa sekäihmiskudosnäytteissä. SWIFT-magneettikuvaustekniikka puolestaan mahdollisti sekä rustonalai-sen luukudoksen tilan arvioinnin että nivelrikossa tapahtuvien muutosten havainnoinnin rus-to–luurajapinnassa. Nämä menetelmät voivat mahdollistaa nivelrikon diagnosoinnin taudin var-haisessa vaiheessa, mutta lisätutkimuksia vaaditaan menetelmien soveltuvuudesta kliiniseenkäyttöön.

Asiasanat: kalkkeutunut rusto, magneettikuvaus, nivelrikko, nivelrusto, relaksaatioaika,rustonalainen luu, sweep imaging with Fourier transform

Acknowledgements

This study was carried out in the Quantitative Imaging of Cartilage research group,Department of Diagnostic Radiology, University of Oulu, during 2011–2015 in a veryclose collaboration with the Biophysics of Bone and Cartilage (BBC) research group,Department of Applied Physics, University of Eastern Finland, Kuopio.

I owe my deepest gratitude to my principal supervisor Professor Miika Nieminen,Ph.D., for the guidance and sharing his expertise during the thesis project. I greatlyadmire his positive attitude and dedication towards science and other things in life, andit has been a true privilege to work in his top-class research group. I am grateful to mysecondary supervisor, Docent Mikko Nissi, Ph.D., who was the first one to introduceme to the field of cartilage imaging and MRI research. His continuous support withhumorous touch and hands-on help has been invaluable for which I am very grateful.

I am also thankful to Professor Jukka Jurvelin, Ph.D., and Professor Juha Töyräs,Ph.D., for allowing me to work in the BBC group. It has been an honor to witness theirenthusiasm for scientific work and the way of leading and conducting research.

I would like to thank the official pre-examiners of this thesis, Associate ProfessorXiaojuan Li, Ph.D., and Associate Professor Jiang Du, Ph.D. for their constructivecriticism and feedback that greatly improved the quality of this thesis. Mrs. MaaritRissanen is acknowledged for the linguistic revision.

I would like to express my gratitude to Professor Jutta Ellermann, M.D., Ph.D.,from the Center of Magnetic Resonance Research, University of Minnesota for fruitfulcollaboration during these years, which was essential for my research.

I want to thank Associate Professor Simo Saarakkala, Ph.D., the chairperson of mythesis follow-up group, for significant contributions to my studies. I am also indebted toother co-authors of my studies: Timo Liimatainen, Ph.D., Lauri Lehto, M.Sc., WalterHerzog, Ph.D., Rami Korhonen, Ph.D., Virpi Tiitu, Ph.D., Hertta Pulkkinen, Ph.D., OutiKiekara, M.D., Mikko Finnilä, Ph.D., Olli-Matti-Aho, M.D., Petri Lehenkari, M.D.,Ph.D., Anne Brünott, D.V.M., D.E.C.V.S, René van Weeren, D.V.M., Ph.D., D.E.C.V.S.,Harold Brommer, D.V.M., Ph.D., D.E.C.V.S., Peter Brama, D.V.M., M.B.A., Ph.D.,and Ilkka Kiviranta, M.D., Ph.D. for their efforts. Particularly, I want to express mygratitude to Elli-Noora Salo, B.Sc, for introducing me to the laboratory practices in thebeginning of this thesis project, and cooperation in MRI research.

7

I want to thank the staff of the Department of Applied Physics, especially all theformer and current members of the BBC group. It has been a great pleasure to work andstudy in a friendly and supporting atmosphere with such a talented people, to name justa few: Sami Väänänen, Ph.D., Mika Mononen, Ph.D., Pia Puhakka, M.Sc., Satu Inkinen,M.Sc., Janne Mäkelä, M.Sc., Mikael Turunen, Ph.D., Juuso Honkanen, M.Sc., JukkaLiukkonen, M.Sc., Lasse Räsänen, M.Sc., Petri Tanska, M.Sc., Roope Lasanen, M.Sc.,Elvis Danso, M.Sc., Cristina Florea, Ph.D., Katariina Nissinen, Ph.D., Hanna Isaksson,Ph.D., Antti Aula, Ph.D., Janne Karjalainen, Ph.D., Markus Malo, Ph.D., Lassi Rieppo,Ph.D., Tuomas Virén, Ph.D. and Harri Kokkonen, Ph.D. Discussions and fun momentsduring the coffee breaks, sports and other off-duty activities with them kept my spiritup and helped me finish this thesis. Colleagues at the University of Oulu and OuluUniversity Hospital are acknowledged for the collaboration.

I would like to thank Professor Olli Gröhn, Ph.D., from Biomedical NMR group ofA.I. Virtanen Institute, University of Eastern Finland for providing the MRI facilities atuse for my research. I also wish to thank the rest of the group for the fruitful researchcollaboration and friendly companion during the conference trips.

Finally, I want express my gratitude to all my friends for the support and sharedmoments outside of the work. I am also deeply grateful for my brother Pasi, and parentsRitva and Matti for their continuous and endless support throughout my entire life.

For financial support Academy of Finland (grants 128603, 260321), University ofOulu Graduate School, Medical Research Center Oulu Doctoral Programme, Universityof Oulu, National Doctoral Programme for Musculoskeletal Disorders and Biomaterials(TBDP) and Foundation for Advanced Technology of Eastern Finland are acknowledged.

Kuopio, March 2015

Jari Rautiainen

8

Abbreviations

ACLT Anterior cruciate ligament transectionAFP Adiabatic full passageAHP Adiabatic half passageAUC Area under curveBV/TV Bone volume-to-tissue volume ratioBW BandwidthCPMG Carr-Purcell-Meiboom-GillCT Computed tomographyCTRL ContralateralCW Continuous waveDD Digital densitometryDE Double echoDESS Double echo steady stateDW Dry weightdGEMRIC Delayed gadolinium enhanced MRI of cartilageECM Extracellular matrixFCD Fixed charge densityFID Free induction decayFLASH Fast low angle shotFOV Field of viewFSE Fast spin-echoFT Fourier transformFTIRI Fourier transform infrared imagingGAG GlycosaminoglycanGd GadoliniumGRE Gradient echoHSn Hyperbolic secantIR Inversion recoveryMRI Magnetic resonance imagingMT Magnetization transferMTR Magnetization transfer ratio

9

NMR Nuclear magnetic resonanceOA OsteoarthritisOARSI Osteoarthritis Research Society InternationalPG ProteoglycanPI Parallelism indexPLM Polarized light microscopyRAFF Relaxation along a fictitious fieldRF RadiofrequencyRFR Rotating frame of referenceROC Receiver operator characteristicROI Region of interestSAR Specific absorption ratioSNR Signal-to-noise ratioSE Spin-echoSD Standard deviationSL Spin-lockSR Saturation recoverySWIFT Sweep imaging with Fourier transformT1 Longitudinal relaxation timeT1Gd T1 in the presence of gadoliniumT2 Transverse relaxation timeTb.N Trabecular numberTb.Sp Trabecular separationTb.Th Trabecular thicknessTE Time-to-echoTI Time-to-inversionTKA Total knee arthroplastyTR Time-to-repetitionT1ρ Longitudinal relaxation time in RFRT2ρ Transverse relaxation time in RFRUTE Ultrashort echo timeVOI Volume of interestWW Wet weightZTE Zero echo time

10

List of original publications

This thesis is based on the following articles, which are referred to in the text by theirRoman numerals (I–III):

I Rautiainen J, Nissi MJ, Liimatainen T, Herzog W, Korhonen RK & Nieminen MT. (2014)Adiabatic Rotating Frame Relaxation of MRI Reveals Early Cartilage Degeneration in aRabbit Model of Anterior Cruciate Ligament Transection. Osteoarthritis Cartilage 22(10):1444–1452.

II Rautiainen J, Nissi MJ, Salo EN, Tiitu V, Finnila MA, Aho OM, Saarakkala S, Lehenkari P,Ellermann J & Nieminen MT. (2014) Multiparametric MRI assessment of human articularcartilage degeneration: Correlation with quantitative histology and mechanical properties.Magn Reson Med. In press.

III Rautiainen J, Lehto LJ, Tiitu V, Kiekara O, Pulkkinen H, Brunott A, van Weeren R, BrommerH, Brama PA, Ellermann J, Kiviranta I, Nieminen MT & Nissi MJ. (2013) Osteochondralrepair: evaluation with sweep imaging with fourier transform in an equine model. Radiology269(1): 113–121.

The thesis also contains previously unpublished data.

11

12

Contents

AbstractTiivistelmäAcknowledgements 7Abbreviations 9List of original publications 11Contents 131 Introduction 152 Articular cartilage and subchondral bone 17

2.1 Structure and composition of articular cartilage . . . . . . . . . . . . . . . . . . . . . . . . . 172.2 Mechanical properties of articular cartilage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.3 Structure and composition of subchondral bone . . . . . . . . . . . . . . . . . . . . . . . . . 202.4 Osteoarthritis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3 Magnetic Resonance Imaging 233.1 Nuclear magnetic resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.2 Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.2.1 T1 relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.2.2 T2 relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.2.3 Rotating frame relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.2.4 Magnetization transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.3 MRI pulse sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.3.1 Ultrashort echo time sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.4 Quantitative MRI of articular cartilage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.5 MRI of subchondral bone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4 Aims of the thesis 395 Materials and methods 41

5.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415.1.1 Rabbit samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425.1.2 Human samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425.1.3 Equine samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

5.2 MRI measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425.2.1 T1 measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435.2.2 T2 measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

13

5.2.3 Rotating frame relaxation measurements . . . . . . . . . . . . . . . . . . . . . . . . . 445.2.4 Magnetization transfer measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 445.2.5 Gradient echo imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445.2.6 SWIFT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5.3 Reference methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455.3.1 Micro-computed tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455.3.2 Biomechanical testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455.3.3 Microscopic techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465.3.4 Histological evaluations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475.3.5 Morphologic assessment of cartilage repair . . . . . . . . . . . . . . . . . . . . . . .475.3.6 Biochemical analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

5.4 Data-analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485.4.1 Relaxation time and microscopic ROI analyses . . . . . . . . . . . . . . . . . . . 485.4.2 Assessment of short T2 spins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .495.4.3 Estimation of structural bone parameters . . . . . . . . . . . . . . . . . . . . . . . . . 505.4.4 Assessment of signal-to-noise ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515.4.5 Measurement of cartilage thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5.5 Statistical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526 Results 55

6.1 Quantitative MRI of cartilage degeneration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 556.1.1 Reference methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 556.1.2 Relaxation time measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 576.1.3 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 616.1.4 Association with structure and function . . . . . . . . . . . . . . . . . . . . . . . . . . 62

6.2 Evaluation of osteochondral structures with short T2 imaging . . . . . . . . . . . . . 666.2.1 SWIFT imaging of osteochondral repair in equine model . . . . . . . . . . 666.2.2 SWIFT signal at the cartilage-bone interface . . . . . . . . . . . . . . . . . . . . . 69

7 Discussion 717.1 Quantitative MRI of degenerated articular cartilage . . . . . . . . . . . . . . . . . . . . . . 717.2 MRI of short T2 components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 787.3 Future prospects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

8 Summary and conclusions 83References 85Original publications 97

14

1 Introduction

Articular cartilage is a highly specialized tissue that enables virtually frictionlessmovement of the articulating joints and softens peak loads generated during locomotion[1–3]. The tissue has poor regenerative properties due to the absence of blood vessels[1, 3]. However, articular cartilage is very durable as it can remain fully functional fordecades or the entire life of an individual. Osteoarthritis (OA) is the most commonjoint disease, the knee joint being one of the most frequently affected sites [4, 5]. Theprevalence of OA is increasing in the elderly population but the pathophysiology ofOA is still unclear, partly owing to its complex structure and the interactions of tissueconstituents [6]. OA is generally considered as cartilage disease although all jointcompartments are affected. It is characterized by progressive degeneration of articularcartilage which starts at the molecular level [4, 7], coupled with sclerosis of subchondralbone and changes at the cartilage–bone interface [8, 9]. In the end stage of the diseasecartilage has been lost leaving only the bony surfaces articulating against each other,which causes pain, joint stiffness and disability [4]. The total costs of OA to economiesdue to medical expenses and early retirements are enormous [10], approximated inFinland to be about EUR 1 billion annually [11].

In the final stage of knee OA the only proper treatment is total knee arthroplasty(TKA) which removes pain and partly restores mobility [12]. Also other more conser-vative surgical repair techniques and disease modifying drugs have been developedwithout succeeding to fully restore normal joint function [4]. The reason for this is partlyexplained by the lack of sensitive biomarkers for the early diagnosis and follow-up ofthe disease progress or monitoring of the treatment outcome. OA diagnostics is currentlybased on radiography of the joint-space width or clinical magnetic resonance imaging(MRI), which are too insensitive for reliable and early diagnosis of the disease [13].Consequently, there is a need for new more sensitive diagnostic imaging methods forassessment of cartilage degeneration.

The development of MRI technology has provided new promising tools for theevaluation of the status of the joint. Noninvasive assessment of cartilage constituents canbe implemented by means of quantitative MRI relaxometry which involves measurementof tissue specific relaxation time parameters [14–17]. Direct imaging of bone is insteadvery difficult with conventional MRI techniques due to very fast signal decay from

15

calcified structures [18, 19]. Sweep Imaging with Fourier Transformation (SWIFT) is anMRI technique in the ultrashort echo time (UTE) family, capable of capturing a signalfrom objects having a broad range of relaxation times [20, 21], including tissues withextremely short T2 relaxation time such as subchondral bone.

The aim of this thesis was to study the feasibility of a number of novel MRItechniques for the assessment of degenerative changes in articular cartilage, subchon-dral bone and the cartilage–bone interface. Quantitative MRI and SWIFT methodswere tested in animal models of OA and in human tissue samples suffering varyingdegrees of degeneration. Furthermore, quantitative MRI techniques were compared toreference methods to study the sensitivity of the methods to the cartilage constituents,histopathological grade and biomechanical properties of cartilage.

16

2 Articular cartilage and subchondral bone

Articular cartilage is a connective tissue that covers the articulating surfaces in thesynovial joints [1–3]. In a knee joint, approximately 2-4 mm thick [22] articular cartilagecovers femoral condyles, tibial plateau and patella (Fig 1). It has a highly anisotropicstructure with depth-dependent amount and organization of the tissue constituents[3, 23].

Tibia

Femur

Meniscus

Anterior cruciate

ligament

Patella

Articular

cartilage

Femoral

condyles

Posterior cruciate

ligament

Zone of

calcified

cartilage

Subchondral

bone plate

Trabecular bone

Deep

zone

Transitional

zone

Tangential

zone

A B C

Fig 1. Anterior (A) and cross-sectional (B) view of a knee joint. Cross-section of articularcartilage (C) reveals depth-dependent orientation of the collagen fibrils.

2.1 Structure and composition of articular cartilage

Articular cartilage is composed of cells and an extracellular matrix (ECM), whichincludes tissue water and a macromolecular network [2, 3]. Extracellular water comprisesup to 80% of the tissue’s wet weight, the rest consisting mostly of large macromoleculessuch as collagen, proteoglycans (PG) and non-collagenous proteins [2, 3]. Cartilagecells (chondrocytes) contribute to only about 1% of the tissue volume. The interactionof the interstitial water with the ECM components determines the mechanical propertiesof the tissue [2, 3].

Type II collagen is the most abundant type found in cartilage accounting for about95% of all collagens [2, 3]. Collagens are organized into a network of fibrils that providesthe tissue with its tensile strength properties and attach the large macromolecules andwater to the matrix. In total, collagen fibrils contribute approximately 60% of the dry

17

weight of cartilage. The orientation of the collagen fibrils varies with the depth of thetissue [24], and is the main determinant of different zones of cartilage.

Proteoglycans are large molecules consisting of protein core and glycosaminoglycan(GAG) side chains [2, 3]. Negatively charged GAGs create the fixed charge density(FCD) of cartilage [23], attracting positive charges and repelling other negativelycharged molecules. In articular cartilage, PGs are mostly large aggregating proteoglycanmonomers or aggrecans and smaller PGs, which make up approximately 25 to 35%of the dry weight of cartilage [3]. The interfibrillar space of the ECM is primarilyfilled with aggrecans that are associated with hyaluronic acid and link proteins forminglarge proteoglycan aggregates [25]. During mechanical loading aggregates prevent thedisplacement of PGs within the ECM [3].

Chondrocytes, the only type of cells found in the cartilage, are spheroidal in shapeand they do not form cell-to-cell contacts within the ECM [3]. Chondrocytes maintaincartilage ECM by synthesizing and degrading cartilage molecules [2, 3]. Due tothe avascularity of cartilage, chondrocytes derive their nutrition from the synovialfluid through diffusion. The activity of cells is dependent on the mechanical stressesexperienced during loading [26].

The structure of articular cartilage is highly anisotropic. Although the structurevaries smoothly, three histological layers with varying composition and organizationof tissue constituents are generally distinguishable [2, 3]. The layers are identifiedaccording to the organization of the collagen fibril network (Figure 1). The superficialzone is about 5% of the cartilage thickness, the transitional zone 20% and the radial ordeep zone 75% in mature cartilage [27]. The composition, organization, mechanicalproperties of the matrix and cell morphology vary according to the depth from thesurface of the articular cartilage [2, 3, 23].

The superficial zone is the thinnest zone of articular cartilage [2, 3]. It has higherconcentrations of collagen and water and a lower concentration of PGs relative to otherzones. The superficial zone has higher tensile stiffness and strength than deeper zones asthe orientation of the collagen fibrils in the superficial zone is parallel to the cartilagesurface. Chondrocytes in the superficial zone are more flattened in shape compared toother zones.

Collagen fibrils in the transitional zone are more randomly organized, changingfrom parallel to surface to perpendicular to the surface in the deeper cartilage [2, 3]. Thetransitional zone has a higher amount of PGs but lower concentrations of water andcollagen than the superficial zone.

18

The deep zone or radial zone has the highest concentration of proteoglycans andthe lowest concentration of water [2, 3]. Chondrocytes are organized in columnsperpendicular to the joint surface. Similarly, the primary orientation of the collagenfibrils in the deep cartilage is perpendicular to the cartilage surface. Eventually, thecollagen fibrils reach tidemark, the thin 5-µm thick [9], irregular boundary between thecalcified and uncalcified cartilage layers.

The zone of calcified cartilage is a mineralized layer at the osteochondral junction,which integrates the uncalcified cartilage to the subchondral bone and transmits forcesacross the cartilage-bone interface [9]. The thickness of calcified cartilage variesbetween 20 and 250 µm in healthy human subjects [28]. The calcified cartilage ismainly composed of type II collagen, type X collagen, GAG and chondrocytes [9].However, chondrocytes in normal calcified cartilage are quiescent and their density ismuch lower than in the overlying articular cartilage. In addition, while normal articularcartilage is avascular, the zone of calcified cartilage contains vascular channels thatoriginate from the subchondral bone [9]. In healthy subjects, the tidemark limits thevascular invasion and calcification to the calcified cartilage layer. The mechanicalstrength of the calcified cartilage is lower than that of the subchondral bone, though it isstill much stiffer than articular cartilage [29].

2.2 Mechanical properties of articular cartilage

The mechanical properties of articular cartilage are mainly determined by the interactionof collagens and PGs with the interstitial water. PGs are mainly responsible for thecompressive stiffness (equilibrium modulus), while the collagen fibril network andwater control the dynamic properties [30, 31].

Mechanical parameters can be determined with stress-relaxation, creep or dynamicloading schemes. In the stress-relaxation test, the time-dependent response of the forceis tracked after application of an instantaneous load, allowing the tissue to recover to itsequilibrium state. In the creep test, the deformation of the tissue is investigated duringconstant loading. In dynamic loading, the force is cyclically applied for several periodswith a certain frequency [32].

Different loading schemes can be implemented in confined, unconfined or indentationgeometry [33]. In unconfined compression, a cylindrical cartilage sample detached fromthe subchondral bone is compressed between two plates allowing free fluid flow fromthe sides of the sample during mechanical testing. In confined compression, the cartilage

19

sample is placed in a chamber allowing axial fluid flow through a permeable piston. Inindentation, cartilage is loaded with an indenter perpendicular to the cartilage surface. Itdoes not require removal of the cartilage from the subchondral bone. The diameter ofthe indenter should be over four times larger than width of the sample preventing theedges of the sample from affecting the measurement [34].

Young’s modulus for elastic material is given by

E =σ

ε, (1)

where σ and ε are the axial stress and strain, respectively. Young’s modulus is typicallydetermined from the slope of a stress–strain plot of a step-wise stress-relaxation test. Forindentation geometry, Young’s modulus is obtained by applying a correction factor thattakes into account the size and shape of the indenter and the sample size:

E =σ

ε

πa(1−ν2)

2hκ, (2)

where a is the radius of the plane-ended indenter, ν is Poisson’s ratio describing thecompressibility of the sample, h is the thickness of the cartilage and κ is a theoreticalscaling factor [35]. The dynamic modulus is given by

Ed =√

E21 +E2

2 =σd

ε, (3)

where E1 is the storage modulus, proportional to the elastically stored energy in thetissue, E2 is the loss modulus that describes the viscous energy dissipated in the loadingprocess and σd is the dynamic stress [32]. The dynamic modulus is typically severaltimes higher than the elastic modulus.

2.3 Structure and composition of subchondral bone

Bone is a calcified tissue that provides support and protection for organs and enablesmovement of the limbs. Furthermore, it serves as a storage for minerals such as calciumand phosphorus. Bone undergoes remodeling throughout the lifetime of an individual byreformation and resorption of bone matrix. Morphologically two separate types of bonecan be identified in the musculoskeletal system. Hard and dense cortical bone covers theouter layers of bones while the inner parts are more porous trabecular or cancellousbone filled with bone marrow [36, 37]. Subchondral bone is the bone located in thesynovial joints under the zone of calcified cartilage, separated by the boundary known as

20

the cement line [37]. The cortical endplate of an articulating bone is also referred to asthe subchondral bone plate.

Bone is a composite material that consists of cells, water, organic and inorganicphase [36]. The mineralized inorganic phase is mainly composed of hydroxyapatitecrystals while the organic phase consists mostly of type I collagen, which is produced byosteoblasts, cells that synthesize new unmineralized bone. Osteoblasts differentiate intoosteocytes (bone maintaining cells) after they become surrounded by the bone matrixthey produce. Hydroxyapatite and the organic phase together form the mineralized bonematrix. Osteoclasts are large multinucleated cells that are capable of resorbing the bonematrix produced by osteoclasts.

The cortical bone consists of cylindrical structures called osteons [36]. Osteons arecomposed of bone lamellae that surround Haversian canals that include blood vesselsand nerves. Vascular channels from the cortical bone can extent also into calcifiedcartilage [37]. Trabecular bone consists of a network of trabeculae, which are connectedplate-like structures of bone, aligned principally in the direction of mechanical loading.As it does not contain osteons as the cortical bone, trabecular bone is nourished bybone marrow. Trabecular bone is sensitive to mechanical stresses and is continuouslyremodeled due to altering loads. The cortical bone is also remodeled, although the rateis significantly higher in the trabecular bone. This adaptive response of bone to externalforces is known as Wolff’s law. The medullar cavities between the trabeculae are filledwith yellow and red bone marrow. The yellow bone marrow consists mainly of fat whilethe red marrow is responsible for the production of new blood cells.

2.4 Osteoarthritis

Osteoarthritis, also referred to as degenerative joint disease, may develop idiopathicallyor due to injury, inflammation or a developmental disorder [4, 5]. OA is most commonlydiagnosed in hand, knee, hip and spine joints. Typical clinical symptoms are joint painand stiffness which are often associated with late-stage OA. The pathophysiology ofOA involves progressive degeneration of cartilage accompanied by subchondral bonesclerosis that ultimately results in loss of cartilage tissue [4, 5]. In addition, other jointcomponents, such as menisci, ligaments and synovial fluid are also affected. The diseaseprogress can be roughly divided into three overlapping stages.

The earliest signs of OA occur in the ECM of cartilage at the molecular level [4, 5, 7].The PG content of cartilage decreases, minor fibrillation may be shown at the cartilage

21

surface and the water content increases. In addition, the integrity of the collagennetwork is disrupted causing swelling of the tissue. Together these changes increase thepermeability of cartilage, leaving the tissue susceptible to additional damage as thestiffness of the cartilage is reduced. In the early phase, remodeling of the subchondralbone increases the thickness of the subchondral bone plate.

Chondrocytes detect and react to degenerative changes in the tissue by initiatingthe cellular repair response [4]. The cells degrade damaged matrix componentsenzymatically and synthesize new ECM macromolecules. The repair response may lastfor years, even stabilizing the tissue. However, if the degradation process acceleratesand breaks the balance, the tissue is vulnerable to further damage.

Failure in the repair response leads to further disorganization of the collagen network,loss of PGs and fibrillation of the cartilage surface [4, 5]. Progressive loss of tissue leadsto release of cartilage pieces into joint space, decreasing the thickness of the cartilage.This leads to further thickening of the subchondral bone plate and trabeculae, narrowingthe joint space. In addition, osteophytes or cartilaginous tissue may be formed into jointmargins. In the cartilage–bone interface, thickening of the zone of calcified cartilage,duplication, and vascular invasion of the tidemark are detected, which further contributesto the thinning of cartilage tissue [9, 38]. Ultimately cartilage tissue is lost, leaving onlythe dense surface of the subchondral cortical bone.

22

3 Magnetic Resonance Imaging

The following section covers the essential theoretical aspects of MRI which is based onthe physical phenomenon of nuclear magnetic resonance (NMR) [39–41].

3.1 Nuclear magnetic resonance

Spin or nuclear spin angular momentum is an intrinsic quantized physical property ofall nuclei [41]. A nucleus with an odd atomic mass number has a nonzero spin andcan interact with an external magnetic field, which is the basis of NMR. Nuclei havemagnetic dipole moments which align themselves along external magnetic field (B0). Ahydrogen atom 1H, a proton has a spin number of 1/2 and it has two possible energystates, parallel (α-state) and anti-parallel (β -state) to the B0 field. The energy differencebetween these states depends on the magnitude of the B0

∆E =−γ h̄B0, (4)

where γ is the gyromagnetic constant (42.58 MHz/T for 1H), h̄ is the Planck’s constantdivided by 2π . Spins aligned parallel to the B0 have slightly lower energy and it is morefavored state at equilibrium compared to the anti-parallel orientation. The number ofspins at different energy states is given by the Boltzmann distribution

Nα

Nβ

= e∆EkT , (5)

where T is temperature and k is the Boltzmann constant. The excess of the spins parallelto the static magnetic field creates a net magnetization (polarization) vector M0 whichis oriented to the direction of B0 (z direction in the laboratory frame xyz coordinatesystem) at equilibrium [41]. In the external magnetic field, the magnetic moments arenot stationary but precess around the B0. The precession frequency ω0, also calledLarmor frequency is defined as

ω0 =−γB0. (6)

The Larmor frequency of 1H, for example, at 9.4 T is 400.252 MHz. The resonancefrequency is also affected by the molecular environment in which the hydrogen atomsare located. In biological tissues, hydrogen atoms in large fat molecules are partly

23

shielded by the electron clouds from neighboring atoms, reducing the strength of theexternal magnetic field experienced by the protons. A water molecule has a more simplestructure and the shielding effect is much weaker. The difference in the resonancefrequencies (chemical shift) between water and fat is 3.5 parts per million, approximately1.4 kHz at 9.4 T [41].

The equilibrium state can be disturbed by transmitting radiofrequency (RF) energyto the system at the Larmor resonance frequency. The RF pulse creates an additionaloscillating magnetic field, noted as B1, perpendicular to the B0 field. For simplicity,the behavior of the M0 is analyzed in the rotating frame of reference (RFR), in whichthe precessing net magnetization appears stationary as the frame rotates at the Larmorfrequency [41]. According to equation (6), the magnetization M0 precesses also aboutthe RF field, which in effect causes the M0 to tilt away from the z axis as long as theRF pulse is applied. If the RF pulse contains off-resonance components, the resultingeffective field (Be f f ) is defined as

Be f f =√

B21 +∆B2, (7)

where ∆B is the field due to the difference between the resonance and rotating framefrequencies. The magnetization follows the effective field during RF irradiation.

3.2 Relaxation

The application of RF pulses at resonance frequency to the spin system at the equilibriumstate rotates the net magnetization M0 from the z axis towards the xy-plane (transverseplane). The transverse component of the net magnetization can be detected with an RFcoil orthogonal to the xy-plane. This NMR signal induced in the coil is called the freeinduction decay (FID), emitted by the system after the application of the RF pulse, as thesystem returns to the equilibrium state through different relaxation processes [42, 43].

3.2.1 T1 relaxation

The spin-lattice or longitudinal relaxation time constant, T1 characterizes the recovery ofthe longitudinal magnetization back to the equilibrium state. The molecular motion ofthe spins creates local magnetic field fluctuations, thus spins in different molecular envi-ronments experience slightly different magnetic fields. These dipole-dipole interactions,which are the main relaxation mechanisms in tissues, causes spins to exchange energy

24

with their surroundings (lattice), until the thermal equilibrium is reached. Spectraldensity is a quantity that describes the frequency distribution of the molecular motions

J(ω) =τc

1+ω2τ2c, (8)

where τc is the correlation time that characterizes the time scale of the moleculartumbling. The dependency of T1 on the correlation time is described as:

1T1

∝ B2xyJ(ω), (9)

where Bxy is the field present in the xy-plane. Thus, T1 relaxation is the most effective,when magnetic field fluctuations occur near the Larmor frequency.

90° 90° 90° 90°

Mz

Inversion recovery Saturation recovery

180° 90° 180° 90°

Mz

TR TR

TI

Fig 2. Inversion and saturation recovery sequences.

One of the most common ways to measure T1 is to use the inversion-recovery (IR)sequence (Fig. 2). The magnetization is inverted with a 180◦ inversion pulse followedby recovery of the net magnetization with T1 relaxation rate. After a certain time period(time-to-inversion, TI) a 90◦ pulse is applied, which flips the longitudinal magnetizationto the transverse plane for measurement. The magnetization follows the equation

M = M0(1−2e−T I/T1). (10)

T1 can be estimated by repeating the experiment with several TI values, howeverthe repetition time (TR) between the inversion pulses must be long enough for fullrecovery of the longitudinal magnetization. T1 can also be determined with a saturationrecovery (SR) sequence that utilizes repeated 90◦ pulses (Fig. 2). The behavior of themagnetization resembles that of the IR sequence (10):

25

M = M0(1− e−T R/T1). (11)

Thus, signal is dependent on TR. In saturation recovery, the magnetization reaches asteady state and enough repetitions must be allowed to stabilize the steady state beforemeasuring the signal.

3.2.2 T2 relaxation

Spin-spin or transverse relaxation time, T2 characterizes the decay of the transversemagnetization that results from the energy exchange between the spins. Magnetic fieldfluctuations caused by molecular tumblings and static magnetic field inhomogeneitiesresult in a loss of phase coherence between the spins due to the slight variations in theprecession frequencies. No energy is lost to the surroundings. In addition, T2 relaxationis enhanced by diffusion as spins may change place and experience different localmagnetic fields. T2 is always equal or faster than T1 relaxation. T ∗2 is a time constant thattakes into account all the effects contributing to the decay of transverse magnetizationand is defined as

1T ∗2

=1T2

+1T ′2

, (12)

where T ′2 captures the relaxation due to magnetic field inhomogeneities. T2 is alsodependent on the correlation time:

1T2

∝ B2z τc, (13)

where Bz is the field present in the z-axis. Transverse relaxation is more effective at slowtumbling rates.

In the ideal case, the transverse magnetization (FID signal) follows T2 decay but dueto field inhomogeneities the decay is characterized by the T ∗2 time constant. The phasecoherence of the spins can be refocused in a simple spin-echo (SE) experiment whichstarts with a 90◦ RF pulse that flips the magnetization to the transverse plane (Fig. 3).The dephasing of the spins is reversed with a refocusing 180◦ inversion pulse and after acertain time period (time-to-echo, TE) a signal called the spin echo or the Hahn echo[44] is produced, as the spins refocus. However, the refocusing is not complete due todiffusion. The magnetization follows exponential decay

26

M = M0e−T E/T2 . (14)

The measured signal intensity depends on the echo time, and thus the T2 relaxationtime can be determined by repeated measurements at different TE values. Insensitivityto diffusion can be obtained with the Carr-Purcell-Meiboom-Gill (CPMG), sequencewhich utilizes repeated refocusing pulses separated by an inter-pulse delay that form atrain of echoes (Fig. 3).

< 90°

Double echo Gradient echo

180°90° 180°

90°

Spin echo CPMG

180°90°

Mxy

180° 180° 180° 180°

Mxy

Mxy Mxy

TE

G

Fig 3. Spin-echo, double echo, CPMG and gradient echo sequences. In the spin-echo exper-iment, the echo is formed after refocusing inversion pulse. Gradient echo is formed throughdephasing of the spins with negative gradient followed by rephasing with positive gradient.

The gradient echo (GRE) experiment utilizes gradient magnetic fields instead of RFpulses for the refocusing of the spins. A negative gradient is applied after the initialexcitation pulse (typically less than 90◦), which causes rapid dephasing of the transversemagnetization (Fig. 3). Reversal of the first gradient causes rephasing of the spins whichproduces a gradient echo after TE. The signal follows a T ∗2 decay

M = M0e−T E/T ∗2 . (15)

27

Usually shorter echo times can be achieved with GRE compared with SE basedtechniques.

3.2.3 Rotating frame relaxation

Relaxation processes may occur also in the so-called rotating frame of reference whichcan be studied with spin-lock (SL) methods [45]. Effectively spin-locking experimentsprobe relaxation at very low magnetic fields, utilizing the higher signal-to-noise ratio(SNR) of high magnetic fields. In a classical spin-locking experiment, spins are lockedto the transverse plane with a continuous wave (CW) RF pulse B1SL after the initial 90◦

tip. The magnetic moments start to precess around the spin-lock field at ω = -γB1SL.During the spin-lock pulse, magnetization relaxes exponentially with a longitudinalrelaxation time constant in the rotating frame, T1ρ (’T1 rho’), along the applied B1SL

field [46]. Thus, T1ρ approximates to T1 at very low magnetic fields. The magnetizationduring SL follows the equation

M = M0e−TSL/T1ρ , (16)

where TSL is the duration of the spin-lock pulse. For the quantification of T1ρ , TSL isvaried in repeated measurements. T1ρ is dependent on the molecular motion:

1Tρ

= K

(32

τc

1+4ω2e f f τ2

c+

52

τc

1+ω20 τ2

c+

τc

1+4ω20 τ2

c

), (17)

where K is a constant including information on the local field fluctuation and ωe f f is theeffective SL field, which is equal to γB1SL in the case of on-resonance excitation [47].The equation (17) shows that T1ρ is sensitive to slow molecular motion. T1ρ dispersion,the SL measurement with varying B1SL provides information about the frequencydistribution of the molecular motions. At low spin-lock power, the T1ρ relaxationtime approaches T2 relaxation, although T1ρ is always greater than T2. Spin-lockingexperiments can also be performed with off-resonance pulses with lower RF pulse poweras the magnetization follows and relaxes along the effective RF field, defined by theequation (7).

Another way to induce and quantify T1ρ relaxation is to use adiabatic RF pulsetrains (Fig. 4) by varying the number of pulses [48]. Adiabatic RF pulses cover a broadrange of resonance frequencies by utilizing time dependent frequency and amplitudemodulation [49]. The spin-locking can also be achieved with adiabatic pulses as the

28

magnetization follows the effective field during the adiabatic sweep, assuming that theso-called adiabatic condition is fulfilled which requires that the effective field is muchlarger than the rate of change of its orientation. Hyperbolic secant (HSn) pulses arecommon types of adiabatic pulses for performing an adiabatic full passage pulse (AFP)(180◦ flip) or adiabatic half-passage (AHP) (90◦) frequency sweeps. By changing thestretching factor n (= 1, 4, 8. . .), relaxation rates can also be altered for the dispersionmeasurement. The advantage of adiabatic pulses is that the RF power can be distributedsequentially in time while still covering a broad frequency range with a clean profilecompared with hard pulses. Furthermore, they are relatively insensitive to B0 or B1

inhomogeneities [49].

freq.

MT

RAFF

Adiabatic T1ρ

Adiabatic T2ρ

RF irradiation

off-resonance

+AHP

HS1 AFP

-AHP

180°

Fig 4. Pulse sequences for adiabatic T1ρ , adiabatic T2ρ , RAFF and inversion-prepared MTmethods.

The transverse relaxation component in RFR is called T2ρ , which characterizesthe relaxation of magnetization perpendicular to B1SL [50]. If the RF pulse is appliedperpendicular to the magnetization, relaxation is dominated by the T2ρ time constant.In an adiabatic T2ρ measurement, the magnetization is flipped to the transverse planewith an AHP pulse which is followed by an AFP pulse train, similar to that in theadiabatic T1ρ experiment (Fig. 4). After the pulse train, an inverse AHP pulse brings themagnetization back to the z axis.

Relaxation along a fictitious field (RAFF) is another RFR relaxation component thatworks in the doubly rotating frame or higher rank rotating frame [51–53]. The relaxationtime TRAFF is different from the T1ρ or T2ρ constants. In the RAFF experiment, the

29

frequency swept sine and cosine amplitude and frequency modulated pulses violate theadiabatic condition generating a fictitious magnetic field. TRAFF relaxation occurs alongthe fictitious field that can be utilized for spin-locking. The magnetization follows theequation

M± = M±0e−t/TRAFF −MSSe−t/TRAFF , (18)

where M±0 are the initial magnetizations before the application of sine/cosine pulses(with and without inversion), and MSS is the steady-state magnetization [51].

3.2.4 Magnetization transfer

Magnetization transfer (MT) loosely means exchange of the magnetization betweenthe free and bound water pools [41, 54]. With MT experiments, it is possible to studythe interactions between the macromolecular and free water pools. Bound protons areassociated with macromolecules and have very short transverse relaxation time. Freeor more mobile water molecules have a much narrower resonance peak and a longerT2 relaxation time. The bound water pool has a broader resonance peak and thereforeit can be irradiated with an off-resonance RF pulse (off-resonance to the narrow freewater peak). As a result, the saturated magnetization of the macromolecular spinswill move into the free pool through exchange, reducing the free water signal. Thestrength, duration and frequency offset of the saturation RF pulse affect the amount ofsignal reduction. During an MT experiment, the free water pool also experiences somesaturation, which has to be taken into account in the analysis of MT effects. A commonway to quantify the MT effect is to calculate the magnetization transfer ratio (MT R)

MT R = 1− Msat

M0, (19)

where Msat and M0 are the magnetizations of the mobile protons with and withoutsaturation, respectively. Another possible way to quantify the MT effect is to measure T1

during off-resonance saturation (T1sat) (Fig. 4) [55]. The magnetization follows thebehavior described in equation (18) [55]. In conventional MT experiments, the steadystate is difficult to achieve as the duration of the saturation pulse is typically limiteddue to RF energy or pulse duration limits, causing instability to signal fitting. With aninversion preparation, the dynamic range of the MT measurement and signal fitting canbe improved.

30

3.3 MRI pulse sequences

For imaging purposes, the NMR signal must be localized using linearly changingmagnetic field gradients [41]. Frequency and phase encoding, as well as slice selectiongradients are used for mapping of spatial frequencies of the image and the data isgathered into the k-space, which is the raw MRI data matrix. Fourier transform of thek-space produces the final magnetic resonance image. Before spatial localization, themagnetization can be prepared and set to carry e.g. T1, T2, T1ρ or other weighting (Fig.4). The final image can be encoded (readout) with a basic sequence such as SE or GRE.However, they are not generally feasible in their simplest form because of too longacquisition times. Fast spin echo (FSE) or turbo FLASH (fast low angle shot, (GRE))sequences allow collection of multiple echoes within a single TR period by applyingevenly spaced 180◦ or gradient pulses that form an echo train and drastically speed upthe acquisition. With two refocusing pulses, the acquisition is named dual or doubleecho (DE) sequence (Fig. 3).

3.3.1 Ultrashort echo time sequences

UTE MRI techniques are able to capture signal from very short T2/T ∗2 components[18, 56, 57]. In conventional MRI, typical echo times are a few milliseconds whereas inUTE, the echo times are in the range of 10 µs. For imaging of short T2 species withbound water protons, such as deep regions of cartilage and subchondral bone, an echotime of a few milliseconds is already enough for almost all of the signal to disappear dueto fast T2 relaxation, making signal detection impossible. Although the sequence familyis termed UTE here, UTE also refers to particular ’UTE’ sequences.

In the UTE sequence, the signal acquisition starts after the excitation RF pulse,during ramping up of the gradients (Fig. 5). TE is defined as the time between themidpoint of the RF pulse and the start of the acquisition and is limited only by theability of the MRI device to switch between transmit and receive modes. In addition,the length of the RF pulse is a trade-off between bandwidth (BW) and RF power, asshorter pulses require a higher power to maintain the same bandwidth. Before UTE dataacquisition, preparation pulses can be embedded into the sequence. Long inversionpulses may be used for nulling of the long T2 components, for example fat or free watersignals. UTE generally utilizes radial acquisition although spiral trajectories for k-spacesampling have been introduced [58]. For reconstruction of non-Cartesian data, the

31

k-space data is typically converted into a Cartesian grid and reconstructed by 2D or 3DFourier transformation (FT).

SWIFT

G

UTE

G

TE

RF

acquisition

acquisition

RFZTE

G

acquisition

RF

amp.

freq.

Fig 5. Ultrashort echo time (UTE), zero echo time (ZTE) and SWIFT pulse sequences.

Zero echo time (ZTE) is another technique in the UTE family in which readoutgradients are already ramped up before the RF excitation [59, 60]. A short, high BWrectangular pulse is immediately followed by data acquisition (Fig. 5) resulting inalmost zero TE. In ZTE experiments, gradients are stepped gradually instead of rampingthem up and down, which produces silent scans with short TR. The short delay, a deadtime between excitation and acquisition is mostly caused by the ability of the hardwareto switch between transmit and receive modes. This delay, causing an undersampling ofthe center of the k-space has to be taken into account in the image reconstruction byacquisition oversampling [59].

SWIFT is a 3D sequence that allows a nearly zero TE, of a few µs, due to interleavedexcitation and signal acquisition [20, 21, 61, 62]. The pulse sequence of the SWIFTcontains repeated broadband frequency sweep HSn pulses, which are divided into shortsegments when the RF power is on (Fig. 5). The FID signal is collected between theexcitation pulse segments. Similar to ZTE, the signal acquisition is done in the presenceof a frequency encoding gradient, which is stepped in small increments after eachHSn pulse. Each gradient step collects a frequency encoded radial projection (alsocalled view or spoke), from different angle in a spherical 3D volume. The acquisitionpoints are typically distributed on a spiral trajectory in the k-space. SWIFT images canbe reconstructed using a 3D back-projection algorithm or Cartesian gridding of the

32

k-space data and subsequent 3D Fourier transform [63]. Due to the stepped nature of thegradients, SWIFT is also a silent sequence [20].

Before the final image reconstruction, the collected FID signals have to be processedwith the so-called cross-correlation method due to their contamination with the frequencycomponents of the transmit RF pulse [20, 21, 61, 62]. The measured response r(t) ofthe spin system to the RF pulse x(t) is considered as linear and they are related throughconvolution of the unit impulse response h(t) (i.e. the true FID) and x(t). Convolutionin the time domain corresponds to a complex multiplication in the frequency domain ofthe Fourier transformed signals:

R(ω) = H(ω)X(ω), (20)

The frequency spectrum of the system H(ω) is obtained using correlation, which can beperformed in the frequency domain:

H(ω) =R(ω)X∗(ω)

|X(ω)|2, (21)

where |X(ω)| is the modulus of X(ω). As a result, all contaminant signals are in theimaginary part of H(ω) and the real part of H(ω) corresponds to the spin density profileM0 of the system. The signal intensity of SWIFT depends only on T1 and the protondensity as

M = M01− e−T R/T1

1− e−T R/T1cos(θ)sin(θ), (22)

where θ is the nominal flip angle [20]. Thus, T1 can be measured with SWIFT byvarying the flip angle [64, 65]. The maximum signal intensity with the chosen TR isobtained with an optimum flip angle (so-called Ernst angle) of θopt :

cos(θopt) = e−T R/T1 , (23)

If the flip angle is much lower than the optimum flip angle, the T1-weighting becomesnegligible and the image contrast depends mainly on the proton density M0 [20].

Fat and water signals can also be selectively suppressed with SWIFT by embeddingsaturation pulses between excitation pulses at a selected resonance at appropriateintervals [66]. SWIFT acquisition can be also modified to carry T1ρ or RAFF weighting[67, 68].

33

3.4 Quantitative MRI of articular cartilage

MRI provides means for non-invasive assessment of articular cartilage. Several differentquantitative MRI methods have been proposed and studied in cartilage for assessment ofcartilage constituents and degeneration [13–17]. In addition, semi-quantitative scoringsystems have also been developed for the evaluation of the status of the joints [13]. Thescoring systems are based on the evaluation of morphologic changes of cartilage, such asdepth of lesions, thickness and volume of cartilage. Morphological evaluation is usuallybased on gradient echo or double echo steady state (DESS) imaging [13]. However,the volume and the thickness of cartilage varies between individuals and morphologicchanges are not always evident in the early stages of OA. Thus, quantitative MRI ofcartilage is potentially more sensitive to early OA changes than semi-quantitative MRIevaluation.

T2 relaxation time mapping is one of the first quantitative MRI methods applied forthe assessment of cartilage. The structure and integrity of the collagen fibril networkand tissue water content are the main components affecting T2 relaxation time in thecartilage [69–83]. In the normal articular cartilage, T2 maps or T2-weighted imagesdemonstrate a laminar appearance, corresponding to different histological cartilagezones [69–74, 77, 79]. The laminar appearance is related to the dipolar coupling of thespins associated with the collagen fibrils, which reduces the signal intensity and has thefollowing orientation dependence, also known as the magic angle effect:

1T2

∝ 3cos2(θ)−1, (24)

where θ is the angle between B0 and the vector joining the two nuclei [84]. The signalintensity is greatest at the minima of this term, namely at 54.7◦, the ’magic angle’. Theeffect is also seen in other highly organized joint structures like menisci and tendons. T2

elevation due to cartilage degeneration, associated with increased tissue water contentand disorganization of the collagen network has been reported in numerous studies in

vitro and in vivo [75, 78, 81–83, 85–87].Delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) is a technique that uses

gadolinium-based anionic contrast agent (Gd-DTPA2−) for the assessment of negativelycharged GAGs or the FCD of cartilage. Several studies have validated and shown strongassociations of the dGEMRIC method with the PG content of cartilage, or with cartilagedegeneration [80, 86, 88–97]. The accumulation of the contrast agent is increased in the

34

regions where GAG content has decreased, reducing also the T1 relaxation time. Thus,there is an inverse relationship between the concentration of the contrast agent and theGAG content of the cartilage, which can be assessed by T1 relaxation time mapping.Estimation of the concentration of the contrast agent requires measurement of T1 in theabsence of gadolinium, or T1 can also be assumed to be constant [98]. In vivo dGEMRICmeasurement requires intravenous or intra-articular injection of the gadolinium followedby delay to allow the contrast agent to diffuse into the cartilage [99].

Pre-contrast T1 relaxation time has not shown orientational dependence in cartilage[70, 100]. T1 relaxation time profiles have shown decreasing trends from the surface tothe deep cartilage, similar to the distribution of water in the cartilage [86, 101]. Indeed,a strong correlation has been found between T1 and the water content of cartilage [101].T1 has also shown significant elevation with cartilage degeneration [86].

The MT effect in cartilage is dominated by the collagen macromolecules [102–105],although the PGs may also provide some contributions [102]. It has been reportedthat the MT R parameter is not sensitive enough for monitoring changes in damagedcartilage after repair procedures [106]. However, T1sat relaxation time measured withthe inversion prepared MT protocol has demonstrated sensitivity to trypsin-induced PGdepletion in bovine cartilage [107]. Chemical exchange saturation transfer (CEST), aspecific type of MT technique, exploits the dependence between the frequency of theoff-resonance saturation and the extent of the MT effect [108]. Certain frequencies areassociated with specific molecules such as GAG, which contain exchangeable protons.The GAG content of articular cartilage has been successfully investigated with the CEST(gagCEST) method [109–111].

T1ρ has been proposed as a more sensitive biomarker for early cartilage degenerationthan conventional T1 or T2 mapping due to the high sensitivity of RFR relaxation to slowmacromolecular motion. Studies with CW-T1ρ have associated the elevation of T1ρ

with cartilage degeneration, specifically to GAG loss and changes in FCD of cartilage[112–117, 117–128]. However, CW-T1ρ has also been shown to be dependent on themagic angle effect (less than T2), suggesting some contribution of the collagen networkto the relaxation [129–131]. Furthermore, T1ρ has been found to be more sensitivethan T2 in revealing cartilage degeneration [124, 132–134]. CW-T1ρ measurements in

vivo are typically performed at spin-lock frequencies around 500 Hz due to RF power(specific absorption ratio, SAR) limitations. T1ρ dispersion studies in cartilage havedemonstrated the laminar appearance in T1ρ weighted images or relaxation time mapsat low spin-lock frequencies when T1ρ approximates T2, but with reduced effect with

35

increasing frequency [79, 115]. There are potentially several relaxation mechanisms thatcontribute simultaneously to T1ρ relaxation in cartilage, including dipolar interactionand chemical exchange between protons on the GAG side chains and extracellular water.However, the relative contributions of these relaxation mechanisms are not yet fullyunderstood [45, 79].

Adiabatic T1ρ , adiabatic T2ρ and RAFF methods have not yet been widely appliedin cartilage research. The advantage of the methods relates to the relaxed RF powerrequirements compared to CW type SL methods [49, 51]. Furthermore, adiabatic pulsesare more insensitive to B0 or B1 field variations [49]. In addition, adiabatic T1ρ hasbeen shown to be more insensitive to sample orientation than CW-T1ρ [130]. In theinitial study, adiabatic T1ρ and RAFF were sensitive to trypsin-induced degenerationin bovine articular cartilage [107]. Although studies on adiabatic T2ρ in cartilage arelacking, its potential as a biomarker has been demonstrated in other tissues, for examplein the assessment of brain pathologies [135]. Theoretical descriptions of adiabaticT2ρ and RAFF relaxation have shown evidence of dependence on both dipolar andexchange-related relaxation channels, although the relative contributions depend on thetissue investigated [48, 50, 51].

In addition to the quantitative methods listed above, diffusion weighted imaging(DWI) techniques have also been applied for the assessment of articular cartilage[103, 136–140]. DWI utilizes diffusion gradients in the pulse sequence, which causes anon-refocusing phase accumulation in diffusing spins. By varying the gradient strength,an apparent diffusion coefficient (ADC) can be determined. ADC has been suggestedas indicator for early cartilage degeneration [103]. Apart from 1H imaging, sodium(23Na) MRI is a potential method for quantitative assessment of cartilage as sodiumions are associated with the GAG molecules of PGs, nearly directly related to the FCDof cartilage. The method has been validated to show differences in the GAG content,however, 23Na MRI has limitations due to a poor signal-to-noise ratio (SNR) and specialhardware requirements [45, 141–143].

3.5 MRI of subchondral bone

Bone has much lower proton density in comparison with cartilage and due to thecalcified nature of the tissue, it has a very short T2 relaxation time, at a few hundred µs[19, 144]. Consequently, direct imaging of bone with conventional methods is extremelydifficult as the majority of the signal decays before data acquisition is possible. Free or

36

relatively mobile water protons are located in the small pores such as the Haversiancanals of the cortical bone whereas bound protons are associated with collagen [19, 144].Thus, bone contains different proton pools that affect its T2 relaxation time.

MRI techniques capable of imaging extremely short T2s such as UTE, ZTE andSWIFT enable the direct assessment of bone tissue properties [56, 145–148]. TheNMR signal of the cortical bone has been shown to correlate with the mechanicalproperties of the bone [149]. In addition, the T ∗2 relaxation time of the cortical bonehas been quantified using UTE which allows echo times down to approximately 8 µs[145, 150]. Furthermore, UTE has been used for quantification of the cortical bonewater concentration and the ratio of long and short T2 image signal intensities, whichhave been correlated with bone porosity and bone mineral density, respectively [151].

Traditionally the trabecular bone microstructure has been assessed indirectly byimaging the fatty bone marrow, which has much higher proton density than bone. Inprevious studies, strong linear correlations between high-resolution MRI and micro-computed tomography (micro-CT) have been reported [81, 152–156]. High resolutionZTE imaging has demonstrated imaging of bone microstructure with high correspon-dence to histology [147]. SWIFT imaging has been initially applied for bone and dentalimaging [20, 66, 148, 157]. SWIFT allowed depiction of tumor invasion into corticaland trabecular bone while using water and fat suppression schemes [152].

UTE studies of the osteochondral junction have shown a high intensity signalband in the cartilage-bone interface which likely originates from the zone of calcifiedcartilage and the deepest regions of uncalcified cartilage [57, 158–162]. The dualinversion recovery UTE sequence allows acquisition of images with and without long T2

suppression, which difference (subtraction) enhances short T2 components in the image[159, 160].

37

38

4 Aims of the thesis

The specific aims of the study were:

1. To investigate the sensitivity of novel quantitative MRI relaxation time parameters forassessment of cartilage degeneration in a rabbit model of early cartilage degeneration

2. To evaluate cartilage degeneration in osteoarthritic human osteochondral specimenswith novel quantitative MRI techniques

3. To investigate the feasibility of the SWIFT MRI technique for the assessment ofspontaneous repair of osteochondral lesions in an equine model

4. To evaluate the cartilage–bone interface in osteoarthritic human specimens with theSWIFT MRI technique

39

40

5 Materials and methods

This thesis consists of three independent studies (I-III) and unpublished data. Allmaterials and methods used in the studies are summarized in Table 1.

Table 1. The summary of materials and methods in studies I-III and in unpublished data.

Study Material MRI parameters Reference methodsI Rabbit femoral

condyles,ACLT (n = 8),contralateral(n = 8)joints

T1ρ,adiab., T2ρ,adiab., CW-T1ρ ,TRAFF , T2 (DE & CPMG),T1sat

Histology (Safranin-O stain-ing, DD, PLM, Mankin grad-ing)

II Humanosteochondralsamples (tibia),early OA(n = 5),advanced OA(n = 9)

T1ρ,adiab., T2ρ,adiab., CW-T1ρ ,TRAFF , T2 (DE and SE), T1sat ,T1, T1Gd

Histology (Safranin-O stain-ing, DD, PLM, OARSI grad-ing), FTIRI

III Equineosteochondral(n = 5),chondral(n = 5)lesions

SWIFT (with fat and watersuppression schemes), FSE,gradient echo sequence

Micro-CT, histology(Safranin-O staining)

Unpublished Humanosteochondralsamples (tibia),(n = 16)

SWIFT (with fat and watersuppression and 3 kHzoff-resonance saturationschemes)

Micro-CT, histology (Mas-son’s trichrome and Safranin-O stainings, OARSI grading)

5.1 Materials

In studies I-III and in unpublished data, three different sample sets and species (rabbit,human and equine) were used. The following sections include detailed description ofthe materials and preparation procedures.

41

5.1.1 Rabbit samples

In study I, unilateral anterior cruciate ligament transection (ACLT) was performed to theleft knee of skeletally mature New Zealand White rabbits (Oryctolagus cuniculus, age14 months, n = 8). The right knees were used as a contralateral (CTRL) control group.Four weeks after the ACLT, the knee joint were harvested and frozen for storage at-20◦C. In the sample preparation, the joints were first thawed to room temperature inphosphate buffered saline (PBS) which was followed by removal of the distal ends ofthe femurs including the femoral condyles of the joints. The animal experiments werecarried out according to the guidelines of the Canadian Council on Animal Care andwere approved by the committee on Animal Ethics at the University of Calgary.

5.1.2 Human samples

In study II, human osteochondral plugs of 6 mm in diameter (n = 14) were collectedfrom the tibial plateaus of TKA patients. The samples were stored at -20◦C. Theexperiments were approved by the Ethical Committee of the Northern OstrobothniaHospital District, Oulu, Finland (191/2000). The same samples were used in theunpublished dataset which included two additional specimens (n = 16).

5.1.3 Equine samples

In study III, chondral (n= 5) and osteochondral lesions (n= 5) of 6 mm in diameter weresurgically created on the fourth carpal bone of the intercarpal joints of 2-year-old horses(n = 7). Osteochondral lesions penetrated to a depth of 3–4 mm in the subchondralbone while chondral lesions were restricted to the articular cartilage. The animals weresacrificed after spontaneous healing of the lesions for 12 months. Osteochondral plugsof 14 mm in diameter containing the lesion and adjacent tissue were collected, markedwith a landmark cut followed by storage at -20◦C. The procedures were approved by theUtrecht University Animal Ethics Committee.

5.2 MRI measurements

All MR studies were conducted in a 9.4 T vertical bore magnet (Oxford 400 NMR,Oxford Instruments, Witney, England) and by using Varian DirectDrive (Varian, Palo

42

Alto, California, USA) console (VnmrJ 3.1A). In all imaging experiments, a 19-mmquadrature RF volume transceiver (RAPID Biomedical, Rimpar, Germany) was used.The samples were placed inside a Teflon test tube filled with perfluoropolyether tominimize susceptibility effects from the tissue–air interface and for a 1H signal freebackground. In studies I and II, a magnetization preparation block was coupled to anFSE readout (T R = 5 s, T Ee f f = 5 ms, echo train length = 4, 256×128 matrix size, 62.5µm depth-wise resolution, FOV 16×16 mm, 1-mm slice thickness). In study I, slices(two slices in total) were positioned to the lateral and medial femoral condyles and thereadout was modified accordingly to multi-slice FSE imaging acquiring both of theslices immediately after the preparation pulses. Slices were positioned at the sites ofbiomechanical testing, in study I to the load-bearing cartilage region at the apexes of thefemoral condyles. The resolution over the depth of the cartilage in each slice was 62.5µm for all relaxation parameters. The orientation of the samples was controlled byplacing the cartilage surface of the samples parallel to B0 in study I and, perpendicularto B0 in studies II and III.

5.2.1 T1 measurements

In study II, the dGEMRIC technique was applied for the assessment of cartilage degen-eration. Pre-contrast T1-mapping was performed before immersion of the specimens into1 mM Gd-DTPA2− (MagnevistT M , Bayer Healthcare, Montville, NJ) with a saturationrecovery FSE sequence (T E = 10 ms, T R = 80, 160, 320, 640, 1280, 2560, 5120 ms).Post-contrast T1-maps (T1Gd) were acquired at 24 hours after equilibration.

5.2.2 T2 measurements

In studies I and II, T2-mapping was performed with an adiabatic double echo technique(DE-T2). Two HS1-AFP pulses were placed between AHP pulses (T E = 4, 8, 16, 32, 64,128 ms, T R = 5 s). For comparison of different T2 mapping methods, T2 relaxation timewas also measured with CPMG sequence (CPMG-T2, train of hard 180◦ pulses, pulseduration = 115 µs, inter-pulse delay = 1 ms) in study I, and with spin echo preparation(SE-T 2) in study II (T E = 4, 8, 16, 32, 64, 128 ms, T R = 5 s in both studies).

In study III, the FSE sequence (T E = 3.5 ms; T R = 2.5 s, 19×19 mm2 FOV, 256×128matrix size, 75 µm depth-wise resolution) was used for acquiring a single 1-mm thickslice covering the lesion site of the equine specimens.

43

5.2.3 Rotating frame relaxation measurements

In studies I and II, adiabatic T1ρ was measured with a train of 0, 4, 8, 12 and 24 HS1AFP pulses (pulse duration = 4.5 ms, γB1,max = 2.5 kHz) [163]. Measurement ofadiabatic T2ρ was performed by embedding the same AFP pulse train as in adiabatic T1ρ

experiment between AHP pulses (pulse duration = 4.5 ms, γB1,max = 2.5 kHz) [50].TRAFF was acquired with a train of 0, 2, 4 and 6 sine/cosine modulated RAFF pulses(pulse duration = 9 ms, γB1,max = 625 Hz) [51].

In study I, on-resonance CW-T1ρ was measured using hard CW spin-lock pulseswith durations of 0, 10, 20, 40, 80 and 160 ms (γB1 = 1 kHz) embedded between HS1AHP pulses [164]. T1ρ dispersion was investigated in study II with similar CW pulsesetup at spin-lock frequencies of 125, 250, 500 and 1000 Hz.

5.2.4 Magnetization transfer measurements

As an MT experiment, T1sat -mapping was performed in studies I and II. T1sat relaxationtime was measured using 10 kHz off-resonance CW hard pulse irradiation (γB1 = 250Hz) with durations of 0.1, 0.3, 0.8, 2.0 and 4.0 s (study I), both with and without aninversion preparation to increase the dynamic range [55]. In study II, the CW irradiationpulse durations were 0.1, 0.2, 0.4, 0.8, 1.6, 3.5 and 7.0 s.

5.2.5 Gradient echo imaging

In study III, a 2D gradient echo sequence (T E = 1.33 ms, T R = 15 ms, 20◦ flip angle,30×30 mm2 FOV, 256×256 matrix size, 117 µm in-plane resolution, imaging time 30seconds) was used for acquiring a single 1-mm thick slice covering the lesion site of theequine specimens.

5.2.6 SWIFT

In study III, SWIFT MR imaging was performed with 144 000 views, 353 mm3 FOV,BW = 62.5 kHz, 256 complex points, 137 µm isotropic resolution and a nominal flipangle of approximately 5◦, which was optimized for each sample. Fat saturation wasimplemented by placing an HS4 inversion pulse (BW = 1 kHz) centered on the fatresonance frequency after every 16 views [66]. Suppression of the water spins was done

44

similarly at the water resonance frequency. In addition, images without saturation wereacquired with identical timing parameters. The imaging time was approximately 40minutes per SWIFT dataset.

The samples of study II including 2 additional samples were also imaged withSWIFT MRI (unpublished data). SWIFT imaging was performed with 96000 views,FOV 403 mm3, BW = 62.5 kHz, 384 complex points, 104 µm isotropic resolution and aflip angle of approximately 5-7◦. SWIFT images without suppression schemes and, withfat and water saturation were obtained similarly as described for study III. Furthermore,to suppress short T2 spins, saturation pulses were centered 3 kHz off-resonance to thewater resonance frequency. The scan time was approximately 30 minutes per singleSWIFT dataset.

The SWIFT images were reconstructed using custom-made LabVIEW software [63].For each suppression scheme, image acquisition was performed at the water resonancefrequency. During image reconstruction, water suppression images (fat images) wereadjusted for correct frequency shifts to focus the signal of fat.

5.3 Reference methods

5.3.1 Micro-computed tomography

In study III, a high-resolution micro-computed tomography (micro-CT) desktop scanner(Skyscan 1172, Bruker microCT, Kontich, Belgium) was used for the assessment of themicrostructure of the subchondral bone in equine specimens. Datasets with 20 µmisotropic resolution were acquired using 100 kV tube voltage. Similarly, the humanspecimens (n = 16) were imaged with the micro-CT at 28 µm resolution (unpublisheddata). The micro-CT images were reconstructed using NRecon software (BrukermicroCT).

5.3.2 Biomechanical testing

In studies I-II, the biomechanical properties of cartilage were determined using indenta-tion testing with a plane-ended indenter of 1 mm in diameter. Before the mechanicaltesting, the thickness of the cartilage was measured with a high-resolution ultra-soundsystem [165]. Equilibrium elastic modulus (Eeq) was obtained from stress-relaxationtests (3×5% step in study I, 4×5% step in study II, 15 min relaxation after each step).

45

Dynamic moduli (Edyn) were determined from sinusoidal loading tests (1 Hz frequency,amplitudes of 4% and 1% of cartilage thickness in studies I and II, respectively).

5.3.3 Microscopic techniques

In studies I and II, anisotropy of the collagen fibrils and PG content of articularcartilage were determined with polarized light microscopy (PLM) [166] and digitaldensitometry (DD) [167, 168], respectively. Microscopic sections were obtained atthe sites of MR imaging and biomechanical testing (four sections/condyle in study Iand three sections/sample in study II). Unstained 5 µm thick sections were used inPLM measurement to obtain the parallelism index (PI), an indicator for collagen fibrilanisotropy. PI was determined by measuring the signal intensity of the light at sevendifferent polarizer settings at 15◦ steps (0-90◦). The signal intensity change followsa sinusoidal function and using least square fitting, maximum and minimum signalintensities can be determined from the equation. PI is determined as

PI =Imax− Imin

Imax + Imin, (25)

where Imax and Imin are maximum and minimum signal intensities, respectively. A PIvalue 0 indicates random arrangement while PI of 1 indicates fibrils running parallel toeach other.

For DD, 3 µm thick microscopic sections were stained with Safranin-O, which is acationic dye that binds stoichiometrically to negatively charged GAG side chains of thePGs [169]. According to the Beer-Lambert law, optical density (OD) or absorbance (A)is linearly proportional to the concentration of the absorbing molecule

A =− lnII0

= εcl, (26)

where I and I0 are the intensities of the light transmitted through and entering the sample,respectively, ε is the molecular absorption coefficient, c is the concentration of theabsorbing molecule and l is the thickness of the sample. Thus, the amount of Safranin-Ostaining is linearly proportional to the PG content of articular cartilage.

In study II, Fourier transform infrared imaging (FTIRI) was used for quantification ofcollagen and PG contents of the cartilage. In infrared (IR) spectroscopy, a spectrum of IRlight is directed through a sample which is absorbed by chemical bonds in molecules at

46

specific wavelengths. Therefore, the IR absorption is related to the chemical compositionof a sample.

The human samples were imaged using Hyperion 3000 FTIR Microscope (BrukerInc, Germany). The spectral resolution and pixel size were 4 cm−1 and 25 µm,respectively. The spectral region of 950–1720 cm−1 was analyzed and an offset baselinecorrection was performed for all spectra prior to analysis. Collagen and PG contentswere estimated by integrating over the area of the amide I absorbance peak (1585–1720cm−1) and the carbohydrate region (985–1140 cm−1), respectively [170, 171].

5.3.4 Histological evaluations

In studies I and II, Safranin-O stained microscopic sections were histopathologicallygraded with Mankin [172] and Osteoarthritis Research Society International (OARSI)grading [173] systems, respectively. In Manking grading, the following criteria areevaluated: cartilage structure (0–6), cell population (0–3), distribution of Safranin-Ostain (0–4) and tidemark integrity (0–1), with zero indicating normal appearance. Thescale in OARSI grading is 0–6, which indicates the depth of progression of osteoarthriticchanges in the cartilage. Grade 0 represents intact normal cartilage and grade 6 completeloss of articular cartilage with bone deformation. In both studies, grading was performedby three independent observers and grades were averaged.

5.3.5 Morphologic assessment of cartilage repair

In study III, the specimens underwent overall qualitative morphologic assessment usingspecific criteria for articular cartilage, subchondral bone and cartilage-bone interface.The assessment of the MRI images included evaluation of the integrity of articularcartilaginous surfaces as well as the signal intensity within the cartilage and the repairtissue. Furthermore, the areas of chondral and osteochondral lesions were investigatedto evaluate the degree of repair outcome, ranging from completely filled defect withcartilage repair tissue to no repair tissue with exposed subchondral bone. In addition, thecontinuity and/or discontinuity between the native and repair tissue was assessed at thecartilage-bone interface. In the subchondral bone, sclerosis, integrity and bone marrowedema were evaluated at the lesion sites and the adjacent native tissue.

47

5.3.6 Biochemical analyses

In study II, the total water content of cartilage was determined biochemically. First, thearticular cartilage was detached from the subchondral bone, followed by measurementof the wet weight (WW) of the cartilage. The measurement was repeated three timesand averaged. Finally, specimens were freeze-dried, re-weighted (DW) and the watercontent was determined as [165]

(1− DW

WW

)×100%. (27)

5.4 Data-analysis

For further analysis of the results, the specimens in different studies were divided intogroups using the following criteria. In study I, samples were divided into ACLT (n = 8)and CTRL (n = 8) groups. In addition, medial and lateral condyles were also treated asseparate groups. In study II, the samples were divided into two groups according toOARSI grading with grade = 1.5 as cut-off point [174]. The samples with an OARSIgrade less than 1.5 were considered as the early OA group (n = 5) and samples with anOARSI grade higher than 1.5 as the advanced OA group (n = 9). In study III, specimenswith chondral (n = 5) and osteochondral lesion (n = 5) were considered as separategroups.

5.4.1 Relaxation time and microscopic ROI analyses

MRI relaxation time maps were calculated using mono-exponential two-parameter fittingon a pixel-by-pixel basis for T1, T2, CW-T1ρ , T1ρ,adiab. and T2ρ,adiab. measurements. ForTRAFF and T1sat (inversion-prepared MT) maps, a three-parameter mono-exponentialfitting with steady-state was applied [51, 55] (equation (18)). In studies I and II, 3 mmwide full-thickness regions of interest (ROI) were selected from cartilage at the site ofbiomechanical testing for determination of mean ± standard deviation (SD) values foreach relaxation time parameter. Furthermore, in study I the full-thickness cartilage ROIwas split into superficial and deep ROIs for zonal analysis. In study II, full-thicknessROIs were used for the determination of mean depth-wise profiles of relaxation timesin cartilage, which were interpolated to a normalized length of 10 points due to thevariable thickness of the cartilage. The thicknesses of the full-thickness ROIs were

48

compared against the thicknesses derived from microscopy images to avoid inclusionof non-cartilage material in the ROI. Similar depth-wise profiles were calculated forthe PLM, DD and FTIRI measurements. All data and ROI analysis was performedusing Aedes software (http://aedes.uef.fi) and custom routines written in MATLAB(MATLAB R2010a, Math-Works, Natick, MA).

5.4.2 Assessment of short T2 spins

To evaluate the cartilage–bone interface, images of short T2 components were obtainedby subtracting SWIFT images with and without 3 kHz off-resonance suppression pulses(Fig 6). By subtracting SWIFT images with and without water suppression (fat blurreddue to chemical shift), both long and short T2 components are visualized, while thesignal of fat is subtracted in the process. Combination of fat and water suppression (fatin focus) SWIFT images results in a lack of short T2 components but both fat and watersignals are in focus in the image.

49

Difference SWIFT images

Combined SWIFT images

SWIFT

(no saturation)

- long T2s

- short T2s

- fat (blurred)

- water (focused)

SWIFT

(fat saturation)

- long T2s

- short T2s

- fat

- water (focused)

SWIFT

(water saturation)

- long T2s

- short T2s

- fat (focused)

- water

SWIFT

(fat + water image)

- long T2s

- short T2s

- fat (focused)

- water (focused)

SWIFT

(+3 kHz saturation)

- long T2s

- short T2s

- fat (blurred)

- water (focused)

SWIFT

(water saturation)

- long T2s

- short T2s

- fat (blurred)

- water (focused)

SWIFT

(long + short T2s)

- long T2s

- short T2s

- fat

- water (focused)

SWIFT

(short T2 image)

- long T2s

- short T2s

- fat

- water (focused)

Fig 6. The summary of signal components in different SWIFT images with or without signalsuppression schemes. The images can be further subtracted or combined to obtain imagesof the short T2 component or to improve contrast between subchondral bone and marrow.

5.4.3 Estimation of structural bone parameters

In addition to qualitative assessment of the equine specimens in study III, structuralbone parameters were determined from SWIFT and micro-CT data using CTAn software(Skyscan). Bone volume-to-tissue volume ratio (BV/TV), trabecular thickness (Tb.Th),trabecular separation (Tb.Sp) and trabecular number (Tb.N) parameters were calculatedfrom cylindrical volumes of interest (VOI) (2×3 mm3) that were placed in the trabecularbone of the lesion site and adjacent bone, in the same locations for SWIFT and micro-CTdata. Prior to the analysis, the resolution of SWIFT and micro-CT images was matchedby downsampling the micro-CT data to 137 µm resolution. Furthermore, the orientationsof the datasets were co-registered using axial rotation with the help of surgical landmarks.

50

To improve contrast between marrow and bone, combination images of fat and watersaturated SWIFT data were calculated (Fig 6). In addition, the gray scale values in theSWIFT data were inverted to obtain similar contrast compared to micro-CT images.Datasets were binarized with the global thresholding method [175], with both modalitieshaving their own separate values (Fig 7).

Micro-CT Micro-CT

(downsampled)

SWIFT (inverted

water + fat image)

Fig 7. SWIFT (combined fat and water saturation data with inverted contrast) and micro-CTimages with original 20 µm and downsampled to 137 µm resolution show the structure ofbone from similar locations. Binarized images were used for calculating structural boneparameters.

5.4.4 Assessment of signal-to-noise ratio

In study III, signal-to-noise ratio was determined for FSE, GRE and SWIFT images(without saturation, fat and water saturation data). Orientation parameters of the FSEsequence were used to calculate matching slices from the SWIFT data co-registered withthe micro-CT data. Also the field of view of SWIFT and GRE sections was matched toFSE data. SNR was defined as the ratio of mean signal of tissue ROI to the SD of thebackground ROI. Signal ROIs were determined for articular cartilage, subchondral boneplate and trabecular bone. For a comparison between the MRI sequences, SNRs were

51

corrected with the voxel size and square root of the imaging time. Furthermore, thevalues were normalized to SNR of cartilage in the FSE image.

5.4.5 Measurement of cartilage thickness

The thickness of articular cartilage in the equine specimens was determined frommatched SWIFT (without saturation), FSE, GRE and Safranin-O stained sections(average of 9 slices). The thickness in the lesion site and adjacent intact cartilage wasobtained as an average value from 2 mm wide ROIs that were manually segmented usinga custom made Aedes plugin. Finally, group mean thickness values were determined forchondral and osteochondral lesion samples.

5.5 Statistical analysis

In study I, the non-parametric Kruskal-Wallis post-hoc test was used for investigatingdifferences in MRI relaxation time and biomechanical parameters between femoralcondyles in the ACLT and CTRL joints. Pearson’s linear correlation analysis was usedfor studying relationships between the MRI findings and biomechanical properties, PGcontent (optical density) and collagen fibril anisotropy (PI), by pooling the ACLT andCTRL groups. Full-thickness ROI values were used in the correlation analysis betweenMRI and biomechanical parameters and superficial ROIs between MRI and histology.

In study II, differences in the MR imaging parameters and reference methods betweenearly and advanced OA groups were evaluated with the Mann-Whitney U-test. As instudy I, Pearson’s linear correlation analysis was performed for pooled data betweenMRI relaxation time parameters (full-thickness ROI) and biomechanical parameters,water content of cartilage, optical density, PI, OARSI grade and FTIR-derived PG andcollagen contents.

For studies I and II, relative changes (%) in MRI relaxation times between CTRLand ACLT, and between early OA and advanced OA groups were calculated as

mean(degenerated)−mean(re f erence)mean(re f erence)

×100%. (28)

Moreover, the ability of the MRI parameters to differentiate between the two groupswas investigated using receiver operating characteristics (ROC) analysis. The area under

52

the curve (AUC) values (0–1) are an indicator of the sensitivity and specificity of amethod, 1 indicating 100% sensitivity and specificity and 0.5 a random guess.

In study III, Pearson linear correlation coefficients between SWIFT and micro-CTderived structural bone parameters were calculated. Furthermore, Wilcoxon signedrank tests were performed to study differences between modalities in the thicknesses ofarticular cartilage.

All statistical analyses were performed with SPSS Statistics 19.0 (IBM, Armonk,NY) and MATLAB R2010a. Values of P < 0.05 were considered indicative of astatistically significant difference.

53

54

6 Results

6.1 Quantitative MRI of cartilage degeneration

6.1.1 Reference methods

In study I, ACLT resulted in significantly reduced PG content in the superficial part ofthe cartilage in the medial and lateral femoral condyles of the rabbit joints in comparisonto the contralateral joints (Fig. 8). Full-thickness cartilage defects were not found in theACLT joints. Collagen fibril anisotropy (PI) and Eeq were also significantly reduced inthe lateral compartment of the ACLT joint (Table 2). In addition, Eeq and Edyn were alsosignificantly reduced in the medial compartment of the ACLT joint. There was also asignificant difference in Mankin grades in both condyles between ACLT and CTRLgroups.

Opti

cal

Den

sity

(au

) 2.0

1.5

1.0

0.2 0.4 0.6 0.8 01.0 1.0

0.5

0 0.2 0.4 0.6 0.8

ACLT

CTRL

ACLT

CTRLOpti

cal

Den

sity

(au

)

0.5

1.0

2.0

1.5

1.0

0.9

0.8

0.7

0.6

1.0

0.9

0.8

0.7

0.6

Anis

otr

opy (

au)

Anis

otr

opy (

au)

0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0

ACLT ACLT

CTRL CTRL

Normalized cartilage depth Normalized cartilage depth

Normalized cartilage depth Normalized cartilage depth

Lateral condyle Medial condyle

PI (

a.u.)

Opti

cal

den

sity

(a.

u.)

Fig 8. Normalized depth-wise profiles (0 indicates cartilage surface, 1 cartilage-bone inter-face) of PG content (optical density) and collagen fibril anisotropy (PI) of rabbit articularcartilage in lateral and medial femoral condyles in ACLT and CTRL group as determined byDD and PLM, respectively. Grey background indicates a statistically significant difference(Kruskal-Wallis test, p < 0.05)

In study II, OARSI grades of the human cartilage specimens ranged from 0.9 to 1.5and 1.6 to 4.1 in the early and advanced OA groups, respectively (Table 2). The relative

55

Table 2. Mean values ± SD of biomechanical parameters histopathological grades for rabbitand human tissue specimens in different groups. Mankin and OARSI grading systems wereused for rabbit and human specimens, respectively.

Parameter Rabbit Rabbit Rabbit Rabbit Human Humanmedial medial lateral lateral early OA advancedcondyle condyle condyle condyle (n = 5) OACTRL ACLT CTRL ACLT (n = 9)(n = 8) (n = 8) (n = 8) (n = 8)

Eeq (MPa) 1.3 ± 0.5 0.6 ± 0.42 1.1 ± 0.51 0.6 ± 0.42 1.2 ± 0.3 0.2 ± 0.32

Edyn (MPa) 3.5 ± 1.2 2.0 ± 0.62 3.4 ± 1.4 2.9 ± 1.3 6.8 ± 1.7 1.9 ± 2.32

Mankin grade 1.4 ± 1.1 4.8 ± 2.62 0.9 ± 1.0 4.5 ± 4.12 – –OARSI grade – – – – 1.3 ± 0.3 3.2 ± 1.0Kruskal-Wallis post-hoc test (rabbits), Mann-Whitney U-test (humans) 1 p < 0.05, 2 p < 0.01

water content of cartilage showed an increasing trend (early OA 74.3± 2.5%, advancedOA 77.2± 4.6%), while Eeq and Edyn were significantly reduced in the advanced OAgroup. There were no statistically significant differences in FTIRI-derived collagen orPG contents (amide I and carbohydrate absorptions, respectively) while DD-derived PGcontent (optical density) and collagen anisotropy were significantly lower in the deepcartilage of the advanced OA group in comparison with early OA group (Fig. 9).

56

0 10.5

2000

300

0

0

0

10 0 0

0 1200 300

0 10.5 0 10.5 0 10.5

0 10.5 0 10.5 0 10.5 0 10.5

0 10.5 0 10.5 0 10.5 0 10.5

0 10.5 0 10.5 0 10.5 0 10.50

200

100

200

0

150

100

50

200

100

1000

600

1

0.81

2

3

5

10

15

30

50

80

40 1600

600

500

400

0

150

100

50

0

150

100

50

0

150

100

50

0

150

100

50

80

40

t(ms) t(ms) t(ms)

a.u. PI

T1ρ, adiab. T2ρ, adiab. TRAFF T1sat

DE-T2 SE-T2 T1 T1Gd

CW-T1ρ 1 kHz CW-T1ρ 500 Hz CW-T1ρ 250 Hz CW-T1ρ 125 Hz

FTIR - Amide I FTIR - Carbohydrate DD Collagen anisotropy

Normalized cartilage depth Normalized cartilage depth Normalized cartilage depth Normalized cartilage depth

a.u.a.u.

t(ms)

---- OARSI grade < 1.5 (early OA)

---- OARSI grade > 1.5 (advanced OA)

Fig 9. Mean depth-wise normalized MRI relaxation time parameter (T1ρ,adiab., T2ρ,adiab., CW-T1ρ (γB1 = 125−1000 Hz), TRAFF, DE-T2, T2, T1, T1Gd, T1sat) and reference method (FTIR amideI and carbohydrate absorptions, DD (PG content), collagen ansitropy (PI)) profiles in humanspecimens in early and advanced OA groups. Depth value 0 indicates cartilage surface andgrey background a statistically significant difference (Mann-Whitney U-test, p < 0.05)

6.1.2 Relaxation time measurements

In study I, there was a statistically significant increase in adiabatic T1ρ and adiabatic T2ρ

relaxation time parameters in the superficial part of cartilage of both femoral condyles inthe ACLT joints (Table 3). In the lateral compartment, CW-T1ρ and DE-T2 were alsosignificantly longer in the ACLT group.

57

Table 3. Mean MRI relaxation time (T1ρ,adiab., T2ρ,adiab., CW-T1ρ (γB1 = 1 kHz), TRAFF, DE-T2,CPMG-T2, T1sat) values ±SD in the superficial cartilage ROI of rabbit femoral condyles inACLT and CTRL groups.

Parameter Medial Medial Lateral LateralCTRL ACLT CTRL ACLT(n = 8) (n = 8) (n = 8) (n = 8)

T1ρ,adiab. 79.9 ± 6.0 102.3 ± 32.51 75.3 ± 4.4 109.8 ± 61.51

T2ρ,adiab. 26.3 ± 2.9 37.4 ± 20.41 26.8 ± 2.5 44.9 ± 31.11

CW-T1ρ , γB1 = 1 kHz 21.6 ± 8.2 33.2 ± 14.9 19.9 ± 5.2 31.7 ± 10.61

TRAFF 18.5 ± 7.6 19.3 ± 4.5 15.7 ± 3.5 16.4 ± 3.6DE-T2 20.6 ± 3.2 26.2 ± 4.1 22.0 ± 3.8 36.4 ± 22.11

CPMG-T2 29.9 ± 2.8 39.3 ± 21.0 30.6 ± 5.5 57.6 ± 50.5T1sat 338.6 ± 46.7 409.4 ± 103.1 345.6 ± 46.8 420.3 ± 175.4Kruskal-Wallis post-hoc test 1 p < 0.05

0 120

3 mm

CPMG-TDE-TT T CW-T TRAFF T

Safranin-O

1sat221ρ, adiab 2ρ, adiab 1ρ

3 mm

ACLT CTRL

ACLT

CTRL

MRI

t (ms) 60030 30 300 0 0 000 40 60

Fig 10. Representative MRI relaxation time maps of T1ρ,adiab., T2ρ,adiab., CW-T1ρ (γB1 = 1 kHz),TRAFF, DE-T2, CPMG-T2, T1sat parameters of from rabbit lateral femoral condyle of ACLT andCTRL groups.

In full-thickness analysis of rabbit cartilage, adiabatic T1ρ and adiabatic T2ρ showeda statistically significant prolongation in both medial and lateral condyles in the ACLTgroup compared to the CTRL joints (Table 4). Moreover, DE-T2 was significantly longerin the lateral compartment of the ACLT group. The prolongation in relaxation times isalso evident in the representative relaxation time maps from the lateral femoral condyle(Fig. 10). There were no statistically significant differences in the deep cartilage. InCW-T1ρ and TRAFF relaxation time maps, banding artifacts were shown in cartilage thatoriginate most likely due to B0 field inhomogeneities.

58

Tabl

e4.

Sum

mar

yof

the

aver

age

MR

Ire

laxa

tion

times

inth

efu

ll-th

ickn

ess

RO

I(T

1ρ,a

diab

.,T

2ρ,a

diab

.,C

W-T

1ρ(γ

B1=

125−

1000

Hz)

,T

RA

FF,

DE

-T2,

CP

MG

-T2,

SE

-T2,

T1,

T1G

d,T

1sat

)val

ues±

SD

(inm

s)in

rabb

itan

dhu

man

cart

ilage

atdi

ffer

ents

ites

and

grou

ps.

Par

amet

er(m

s)R

abbi

tR

abbi

tR

abbi

tR

abbi

tH

uman

Hum

anm

edia

lm

edia

lla

tera

lla

tera

lea

rlyO

Aad

vanc

edO

Aco

ndyl

eco

ndyl

eco

ndyl

eco

ndyl

e(n

=5)

(n=

9)C

TRL

AC

LTC

TRL

AC

LT(n

=8)

(n=

8)(n

=8)

(n=

8)T 1

ρ,a

diab.

75.6±

5.6

88.6±

15.6

171

.9±

5.1

94.5±

35.5

111

6.1±

10.5

153.

0±

23.8

2

T 2ρ,a

diab.

21.8±

1.6

27.8±

10.5

123

.0±

2.5

35.8±

22.1

132

.7±

4.2

53.2±

11.6

2

CW

-T1ρ

,γB

1=

1kH

z19

.7±

7.3

28.7±

12.8

18.9±

4.6

27.2±

10.5

61.2±

5.5

80.2±

21.3

CW

-T1ρ

,γB

1=

0.5

kHz

––

––

45.8±

4.1

69.3±

15.0

2

CW

-T1ρ

,γB

1=

0.25

kHz

––

––

35.0±

6.1

54.7±

12.1

2

CW

-T1ρ

,γB

1=

0.12

5kH

z–

––

–25

.0±

7.0

40.2±

13.2

1

T RA

FF

15.9±

5.5

16.5±

3.2

13.8±

2.9

14.1±

2.4

31.1±

8.3

50.6±

11.8

2

DE

-T2

18.1±

2.7

23.0±

10.1

19.3±

3.1

29.6±

14.4

120

.9±

3.9

34.8±

8.32

CP

MG

-T2

25.8±

2.5

29.9±

11.5

25.5±

4.1

41.6±

26.3

––

SE

-T2

––

––

20.0±

4.0

32.0±

8.81

T 1–

––

–13

90.4±

86.9

1566

.7±

21.0

1

T 1G

d–

––

–48

2.1±

53.2

434.

6±

98.5

T 1sa

t32

5.3±

34.1

374.

5±

66.1

330.

1±

34.0

381.

5±

116.

744

4.7±

37.9

582.

8±

96.4

1

Kru

skal

-Wal

lispo

st-h

octe

st(r

abbi

ts),

Man

n-W

hitn

eyU

-test

(hum

ans)

1p<

0.05

,2p<

0.01

59

In study II on spontaneously degenerated human cartilage obtained from kneereplacement surgeries, all quantitative MRI parameters except T1,Gd and CW-T1ρ (γB1

= 1 kHz) showed a statistically significant difference between early and advancedOA groups in full-thickness cartilage ROI analysis (Table 4), which is also evident inthe representative relaxation time maps (Fig. 11). Zonal analysis revealed significantprolongation of the relaxation times, especially in the superficial part of the cartilage(Fig. 11).

In the early OA group, T2 relaxation time maps demonstrated the typical tri-laminarappearance which was also evident in adiabatic T2ρ , TRAFF and CW-T1ρ maps withspin-lock frequencies below 1 kHz (Fig. 11). T1ρ dispersion was demonstrated atCW-T1ρ maps with different spin-locking fields, the T2-like appearance shown mostevidently at γB1 = 125 Hz. In the adiabatic T1ρ , CW-T1ρ (γB1 = 1 kHz) and pre-contrastT1 maps relaxation times showed a decreasing trend towards deep cartilage.

60

TRAFF

Safranin-O

T1sat

T1Gd

T1ρ, adiab.CW-T1ρ

1 kHzT2ρ, adiab.

ms

OARSI grade 0.9

OARSI grade 3.7

ms ms ms ms ms ms ms

ms ms ms ms00 0 0 0 0 0.6

0 0000000

1

CW-T1ρ

500 Hz

CW-T1ρ

250 Hz

CW-T1ρ

125 Hz

DE-T2 SE-T2 T1

PG content

Low High

Amide I Carbohydrate PI

a.u a.u

Safranin-OAmide I Carbohydrate PI

200 100 100 80 80 100 100 800

ms ms ms ms ms ms ms ms0 0000000 200 100 100 80 80 100 100 800

60 60 2000 600 80 15

ms ms ms ms00 0 0 0 0 0.6 1 PG content

Low High

a.u a.u60 60 2000 600 80 15

TRAFF T1satT1ρ, adiab.

CW-T1ρ

1 kHzT2ρ, adiab.

CW-T1ρ

500 Hz

CW-T1ρ

250 Hz

CW-T1ρ

125 Hz

T1GdDE-T2 SE-T2 T1

Fig 11. Representative MRI relaxation time maps of T1ρ,adiab., T2ρ,adiab., CW-T1ρ (γB1

= 125−1000 Hz), TRAFF, DE-T2, SE-T2, T1, T1Gd, T1sat parameters for human cartilage spec-imens in early and advanced OA groups. Reference methods (FTIR amide I (collagen) andcarbohydrate (PG) absorptions, DD (PG content), collagen anisotropy (PI)) reveal superficialPG loss in the early OA specimen. The sample with more advanced degeneration showssurface fibrillation, loss of both PGs and collagen as well as disorganization of the collagenfibril network.

6.1.3 Sensitivity analysis

The highest relative changes in the rabbit ACLT model of study I were obtained withadiabatic T2ρ and T2 measured with DE and CPMG sequences in the lateral femoral

61

condyle, approximately 50–60% (Table 5). For adiabatic T1ρ and CW-T1ρ the changeswere in the range of 30–40%, respectively. TRAFF and T1sat were prolonged only by2% and 16% in the lateral femoral condyle, respectively. In the human specimens ofstudy II, the prolongation of adiabatic T2ρ , TRAFF , DE-T2 and SE-T2 and CW-T1ρ (γB1

= 125−500 Hz) was about 60% while the change in adiabatic T1ρ , CW-T1ρ (γB1 = 1kHz) and T1sat was about 30%. All parameters, except T1Gd , which was shortened,experienced prolongation as expected.

Considering both studies I and II, ROC-analysis showed the highest AUC valuesfor adiabatic T1ρ and adiabatic T2ρ parameters (Table 5). In study II, high, over 0.8AUC-values were obtained also for CW-T1ρ , TRAFF , DE-T2, SE-T2, T1 and T1sat .

Table 5. Summary of the relative changes (%) in the relaxation time parameters (T1ρ,adiab.,T2ρ,adiab., CW-T1ρ (γB1 = 125−1000 Hz), TRAFF, DE-T2, CPMG-T2, SE-T2, T1, T1Gd, T1sat) andthe AUC values for rabbit and human specimens.

Relative change (%) AUC valueParameter Rabbit Rabbit Human Rabbit Rabbit Human

lateral medial lateral medialcondyle condyle condyle condyle

T1ρ,adiab. 31% 17% 32% 0.88 0.78 0.98T2ρ,adiab. 55% 28% 63% 0.80 0.77 1.00CW-T1ρ , γB1 = 1 kHz 44% 45% 31% 0.84 0.72 0.82CW-T1ρ , γB1 = 0.5 kHz – – 52% – – 0.96CW-T1ρ , γB1 = 0.25 kHz – – 56% – – 0.98CW-T1ρ , γB1 = 0.125 kHz – – 61% – – 0.91TRAFF 2% 4% 63% 0.58 0.61 1.00DE-T2 53% 27% 67% 0.77 0.58 0.98CPMG-T2 63% 16% – 0.83 0.66 –SE-T2 – – 60% – – 0.87T1 – – 13% – – 0.87T1Gd – – -10% – – 0.64T1sat 16% 15% 31 % 0.61 0.77 0.87

6.1.4 Association with structure and function

In study I, there was moderate negative correlation between quantitative MRI parametersand equilibrium modulus (Table 6). While TRAFF was not significantly correlated withEeq, there was a significant but weak correlation to dynamic modulus. Also CW-T1ρ

and T1sat were significantly correlated to Edyn. There were no significant correlations

62

between MRI parameters and PG content or collagen anisotropy as determined from DDand PLM measurements, respectively.

In study II, all quantitative MRI parameters, except T1Gd were significantly correlatedwith histopathological OARSI grades and biomechanical parameters (Eeq and Edyn),particularly adiabatic T1ρ , adiabatic T2ρ , CW-T1ρ (γB1 = 250− 500 Hz), TRAFF andT1sat (Table 6). There were significant correlations between the water content of cartilageand adiabatic T1ρ , CW-T1ρ (γB1 = 250− 500 Hz) and T1sat . The same parameterscorrelated also with optical density while collagen anisotropy correlated with DE-T2 andCW-T1ρ (γB1 = 125 Hz). Collagen content as represented by FTIR-derived amide Iabsorption correlated significantly with T1 relaxation time while carbohydrate region,representing PG content, correlated with CW-T1ρ (γB1 = 250 Hz), DE-T2, T2, andTRAFF . Furthermore, Eeq and Edyn correlated significantly with OARSI grade, while Eeq

also correlated with optical density and Edyn with water content of cartilage (Table 7).There was also a significant correlation between Eeq and FTIR-derived PG content, andbetween optical density and OARSI grade.

63

Table6.S

umm

aryofthe

linearcorrelationcoefficients

between

MR

Irelaxationtim

eparam

eters(T

1ρ,adiab

. ,T2

ρ,adiab

. ,CW

-T1

ρ(γB

1=

125−1000

Hz),T

RA

FF ,DE

-T2 ,C

PM

G-T

2 ,SE

-T2 ,T

1 ,T1G

d ,T1sat )

andreference

methods

(Eeq ,E

dyn ,OD

,PI,O

AR

SI,w

atercontent

(%),am

ideI(collagen)

andcarbohydrate

(PG

)absorptions).

Param

eterS

peciesE

eqE

dynO

DP

IO

AR

SI

Water

FTIRFTIR

gradecontent(%

)A

mide

IC

arbohydrateT

1ρ,adiab.

Hum

an-0.80

2-0.78

2-0.59

1-0.24

0.812

0.621

-0.22-0.32

Rabbit

-0.391

-0.301

-0.290.13

––

––

T2

ρ,adiab.

Hum

an-0.82

2-0.71

2-0.61

1-0.38

0.852

0.52-0.25

-0.50R

abbit-0.42

1-0.26

-0.290.13

––

––

CW

-T1

ρ ,γB

1=

1kH

zH

uman

-0.641

-0.712

-0.32-0.30

0.631

0.66-0.03

-0.23R

abbit-0.38

1-0.44

1-0.25

0.22–

––

–C

W-T

1ρ ,

γB1

=0.5

kHz

Hum

an-0.81

2-0.76

2-0.60

1-0.28

0.782

0.601

-0.30-0.47

CW

-T1

ρ ,γB

1=

0.25kH

zH

uman

-0.802

-0.722

-0.591

-0.320.80

20.59

1-0.34

-0.541

CW

-T1

ρ ,γB

1=

0.125kH

zH

uman

-0.672

-0.621

-0.34-0.54

10.68

20.48

-0.27-0.46

TR

AF

FH

uman

-0.742

-0.591

-0.50-0.39

0.742

0.40-0.37

-0.541

Rabbit

-0.04-0.35

1-0.31

0.14–

––

–D

E-T

2H

uman

-0.732

-0.571

-0.32-0.54

10.69

20.47

-0.38-0.56

1

Rabbit

-0.471

-0.221

-0.220.10

––

––

CP

MG

-T2

Rabbit

-0.431

-0.23-0.28

0.12–

––

–S

E-T

2H

uman

-0.651

-0.49-0.24

-0.520.59

10.31

1-0.44

-0.581

T1

Hum

an-0.71

2-0.68

2-0.34

-0.270.64

10.45

-0.601

-0.48T

1Gd

Hum

an0.48

0.520.45

-0.09-0.40

-0.460.16

0.28T

1satH

uman

-0.762

-0.812

-0.531

-0.320.80

20.57

1-0.42

-0.47R

abbit-0.48

1-0.411

-0.200.13

––

––

1p<

0.05,2p

<0.01

64

Tabl

e7.

Sum

mar

yof

the

linea

rco

rrel

atio

nco

effic

ient

sbe

twee

nbi

omec

hani

calp

aram

eter

s(E

eqan

dE

dyn)

and

othe

rre

fere

nce

met

hods

(OD

,P

I,O

AR

SI,

wat

erco

nten

t(%

),am

ide

I(co

llage

n)an

dca

rboh

ydra

te(P

G)

abso

rptio

ns).

Inad

ditio

n,ot

her

refe

renc

em

etho

dsar

eco

rrel

ated

with

allo

ther

para

met

ers.

Par

amet

erS

peci

esE

eqE

dyn

OD

PI

OA

RS

IW

ater

FTIR

FTIR

grad

eco

nten

t(%

)A

mid

eI

Car

bohy

drat

eE

eqH

uman

1.00

0.92

20.

611

0.36

-0.9

02-0

.44

0.38

0.55

1

Rab

bit

1.00

0.63

20.

25-0

.30

––

––

Edy

nH

uman

0.92

21.

000.

460.

43-0

.902

-0.5

910.

310.

43R

abbi

t0.

632

1.00

0.39

1-0

.16

––

––

OD

Hum

an0.

611

0.46

1.00

-0.2

0-0

.551

0.1

0.12

0.26

Rab

bit

0.25

0.39

11.

00-0

.07

––

––

PI

Hum

an0.

360.

43-0

.20

1.00

-0.4

3-0

.26

-0.1

8-0

.06

Rab

bit

-0.3

0-0

.16

-0.0

71.

00–

––

–O

AR

SI

Hum

an-0

.902

-0.9

02-0

.551

-0.4

31.

000.

571

-0.2

0-0

.38

grad

eW

ater

Hum

an-0

.44

-0.5

910.

10-0

.26

0.57

11.

00-0

.30

-0.3

4co

nten

t%FT

IRH

uman

0.38

0.31

0.12

-0.1

8-0

.20

-0.3

01.

000.

842

Am

ide

IFT

IRH

uman

0.55

10.

430.

26-0

.06

-0.3

8-0

.34

0.84

21.

00C

arbo

hydr

ate

1p<

0.05

,2p<

0.01

65

6.2 Evaluation of osteochondral structures with short T2 imaging

6.2.1 SWIFT imaging of osteochondral repair in equine model

In study III, spontaneously healed osteochondral and chondral lesions in equine jointsresulted in different healing outcomes according to histological evaluation. All chondrallesions (n = 5) demonstrated poor cartilage repair (Fig. 12) while four osteochondrallesions showed healing. One sample with osteochondral lesion healed poorly by fillingthe lesion only with fibrous tissue that reached deep into the subchondral bone (Fig. 13).

A

B

C

D

E G

HF

Fig 12. Chondral lesion in equine joint with poor cartilage repair after spontaneous healingof 12 months demonstrated by SWIFT (A. non-saturated, B. fat-saturation, C. combinationimage of fat and water saturation images, D. water saturation) and micro-CT, Safranin-Ostained microscopy section, FSE and GRE images. With SWIFT, the cartilage lesion andthinning of the subchondral bone plate is clearly depicted (white arrow). In addition, thesignal intensity in the subchondral bone is higher in SWIFT images in comparison withconventional FSE and GRE techniques.

SWIFT MRI enabled a differentiation of spontaneous healing outcomes of osteo-chondral and chondral lesions by simultaneous assessment of articular cartilage andsubchondral bone. SWIFT produced high signal intensity throughout cartilage andsubchondral bone, both with or without fat saturation (Fig. 12 & 13). In cartilage, thesignal intensity remained constant at the lesion sites and adjacent intact tissue. Bonemarrow was clearly visualized in water saturated SWIFT data and by combining it withthe fat saturated image, the contrast in the subchondral bone was improved allowing acalculation of the bone parameters. Combined SWIFT images demonstrated strikingvisual similarities between SWIFT and micro-CT (Fig. 12 & 13). The correlation ofBV/TV and Tb.Th parameters between SWIFT and micro-CT derived values was high

66

while Tb.Sp and Tb.N parameters correlated only moderately with SWIFT derivedvalues (Table 8). However, an overestimation was noted in SWIFT-derived BV/TV andTb.Th values and an underestimation of Tb.Sp values.

A

B

C

D

E G

HF

Fig 13. Osteochondral lesion in equine joint with poor repair after spontaneous healingof 12 months. SWIFT (A. non-saturated, B. fat-saturation, C. combination image of fat andwater saturation images, D. water saturation) and micro-CT, Safranin-O stained microscopysection, FSE and GRE images reveal fibrous tissue penetrating deep into the subchondralbone (white arrows). With SWIFT, the fibrous tissue is separated from bone marrow withfat and water suppression schemes (black arrows). In addition, the signal intensity in thesubchondral bone is higher in SWIFT images in comparison with conventional FSE and GREtechniques.

Table 8. Linear correlation coefficients of microstructural bone parameters between SWIFTand micro-CT derived values.

Parameter Linear correlation coefficientBone volume/tissue volume ratio R = 0.832

Trabecular separation R = 0.792

Trabecular thickness R = 0.682

Trabecular number R = 0.491

1 p < 0.05, 2 p < 0.01

In the osteochondral sample with poor healing and fibrous infiltration, SWIFTdepicted a high intensity signal from fibrous tissue that corresponded with the void in themicro-CT image at the lesion site (Fig. 13). The fibrous tissue was further differentiatedfrom marrow by applying fat saturation to the SWIFT image. Thinning and sclerosis ofthe subchondral bone plate as well as cartilage defects were also clearly detected at thelesion site (Fig. 12 & 13). At the cartilage–bone interface, a high intensity signal band,typical of the osteochondral interface with UTE sequences, was not present. In SWIFT

67

without any signal suppression schemes, the microstructure of the trabecular bone wasblurred due to the chemical shift of fat. By applying fat or water saturation pulses, thesignal contributed by marrow and water protons was efficiently removed, respectively(Fig. 12 & 13).

The normalized signal-to-noise ratio in SWIFT was high in articular cartilage,subchondral bone plate and trabecular bone in comparison to matched FSE images,except with water saturation, as expected (Table 9). With respect to GRE, the SNR ofcartilage was comparable with that of SWIFT, but was reduced to one fourth in thesubchondral bone plate. The SNR of trabecular bone was high in SWIFT datasetscompared to FSE or GRE with a clear reduction when fat saturation was used withSWIFT.

Table 9. The normalized signal-to-noise ratios in SWIFT (without any saturation pulses, withfat and water saturation), FSE and GRE images in articular cartilage, subchondral bone plateand trabecular bone.

Normalized signal-to-noise ratioSequence Articular cartilage Subchondral bone plate Trabecular boneSWIFT 1.8 0.9 1.2SWIFT fat sat. 1.7 0.7 0.8SWIFT water sat. 0.4 0.5 1.3GRE 1.8 0.2 0.6FSE 1.0 0.1 0.9

The thicknesses of articular cartilage at the lesion sites and adjacent tissue measuredfrom SWIFT images were comparable to those obtained from GRE or microscopyimages (Table 10). In FSE images, the signal from the deep regions of cartilage waslacking, resulting in a trend towards thinner cartilage at the adjacent intact cartilage,however, this was not statistically significantly as compared to other modalities.

Table 10. The thickness of articular cartilage±SD (mm) at lesion sites and adjacent intactcartilage of equine specimens as determined from SWIFT, GRE, FSE and microscopy images

Chondral lesion (n = 5) Osteochondral lesion (n = 5)Modality Intact (mm) Lesion site (mm) Intact (mm) Lesion site (mm)SWIFT 0.71 ± 0.18 0.55 ± 0.30 0.66 ± 0.06 1.40 ± 0.41GRE 0.65 ± 0.17 0.62 ± 0.24 0.59 ± 0.06 1.36 ± 0.32FSE 0.47 ± 0.10 0.55 ± 0.28 0.48 ± 0.09 1.33 ± 0.37Microscopy 0.69 ± 0.04 0.48 ± 0.25 0.69 ± 0.12 1.00 ± 0.20The differences between the modalities were studied using Wilcoxon signed rank test.Statistically significant differences were not found.

68

6.2.2 SWIFT signal at the cartilage-bone interface

In SWIFT subtraction images of human specimens, a high-intensity signal was observedat the osteochondral junction in relatively intact samples with mild OA and OARSIgrades (Fig. 14 A & B, white arrows). The bright line appeared in the region of thecalcified cartilage, which was shown in microscopy sections stained with Masson’strichrome and Safranin-O. The samples with low OARSI grade demonstrated normalsubchondral bone in the micro-CT images, while the Safranin-O stained sectionsrevealed superficial PG loss in the cartilage.

With the advancement of OA, the high intensity signal band was lacking in theSWIFT images which included both long and short T2 components (Fig. 14 C-E). TheSWIFT images with only short T2 components showed thinning or diffuse thickeningof the high intensity signal band. The samples with OARSI grades higher than 3.0demonstrated sclerosis of subchondral bone, increased roughness of the zone ofcalcified cartilage, duplication of tidemarks and PG loss, which were quite visible in themicroscopy and micro-CT images. Furthermore, fibrillation of the cartilage surface wasalso seen in SWIFT images. However, similarly to study III, the signal intensity insidearticular cartilage did not show significant alterations with the progression of cartilagedegeneration.

69

A

B

D

Masson's trichrome Micro-CT

OARSI = 1.6

OARSI = 2.7

OARSI= 3.7

OARSI = 4.1

Safranin-OSWIFT(short T2)

OARSI= 3.3C

E

SWIFT (long + short T2)

Fig 14. SWIFT subtraction images of the human cartilage–bone interface reveal signalchanges during the progression of osteoarthritis. Masson’s trichrome, Safranin-O stain-ing and micro-CT images reveal PG loss, sclerosis of the subchondral bone, duplication oftidemarks and thickening of the zone of calcified cartilage. In SWIFT, a high-intensity signal(white arrows) is observed at the osteochondral junction in relatively intact cartilage sam-ples. During advancement of OA, loss, thinning or diffuse thickening of the high-intensitysignal was observed in SWIFT difference images (black arrows).

70

7 Discussion

In this thesis, novel MRI techniques were tested on animal models of OA and humanarticular cartilage and subchondral bone. With quantitative MRI relaxation time mapping,the biochemical status of cartilage may be assessed indirectly and noninvasively.Moreover, SWIFT MRI enables imaging of tissues with very short T2 relaxation time,which are found in deep cartilage and bone. Assessment of those structures is verydifficult with conventional MRI. Thus, quantitative MRI techniques may advanceknowledge about OA pathology and provide research tools for development and follow-up of new treatments for OA. In addition, application of the novel MRI methods inclinical decision-making may help to identify patients with early stage OA. The mainfindings of the study are discussed in more detail in the following sections.

7.1 Quantitative MRI of degenerated articular cartilage

The comparison of the different relaxation parameters has previously been difficult asonly a few studies have systematically compared several quantitative MRI parameterswithin the same OA model [105, 107, 176, 177]. In the present study, the sensitivity ofmultiple quantitative MRI relaxation time parameters for the assessment of cartilagedegeneration was investigated in two OA models. Conventional T1 and T2 relaxationtime mapping as well as novel rotating frame relaxation parameters (CW-T1ρ , adiabaticT1ρ , adiabatic T2ρ , RAFF) and an MT experiment (T1sat) were tested in a rabbit ACLTmodel of early cartilage degeneration 4 weeks post-transection. Degenerative changeswere also investigated with quantitative MRI in osteoarthritic human specimens withvarying degrees of degeneration. The ultimate aim was to sensitively detect degenerativechanges with quantitative MRI in the biochemical composition of cartilage at early stageof the disease.

To evaluate the MRI findings, the degeneration of articular cartilage in studies I andII was investigated with quantitative and qualitative histology and biomechanical andbiochemical measurements. In both studies, degenerative changes were demonstratedby diminished mechanical properties of cartilage in the degenerated sample groupsas compared to controls, and confirmed with quantitative and qualitative histology.In study I, DD and PLM measurements revealed PG depletion and disruption of thecollagen network limited to the superficial part of the cartilage, suggesting early cartilage

71

degeneration in the femoral condyles of ACL transected rabbits. Similar degenerativechanges in cartilage have also been reported in previous studies with the ACLT rabbitmodel four weeks post-ACLT [126, 178, 179]. These spatially limited degenerativechanges further emphasize the importance of the zonal evaluation of articular cartilagewith reference methods or quantitative MRI. No full-thickness cartilage erosion wasobserved in the present or previous works [126, 178, 179] which is relevant whenstudying early OA changes. To compare with humans, the anatomy of the rabbit kneejoint and post-ACLT disease progression are similar, although the rate of OA progressionis much faster in rabbits [7].

In study II, human osteochondral specimens with a wide range of degenerationwere collected from the tibiae of TKA patients. Although healthy (OARSI grade 0)cartilage was not found, the samples with an OARSI score less than 1.5 represent verymild degeneration [173]. Early OA samples demonstrated PG loss in the superficialcartilage, however without statistical significance in comparison with the advanced OAgroup. The advanced OA specimens (OARSI grade > 1.5) showed more severe PG loss,surface fibrillation, and disruption of the collagen fibril network, which is consistentwith previous work on human specimens [2, 4, 174]. As there was a considerablevariation among the human samples, it may have resulted in a higher spread of the MRIrelaxation and reference parameter values and thus resulted in improved correlationsbetween them compared to study I. However in study I, the changes in the PG contentand collagen fibril anisotropy were very small and limited only to the most superficialpart of cartilage, leading to some averaging in the ROI analysis, which may potentiallyexplain the difference between the two studies.

In studies I and II, T2 relaxation time was mapped using the adiabatic DE technique.For comparison T2-mapping was also performed with CPMG and SE preparationsin studies I and II, respectively. DE-T2 was sensitive to cartilage degeneration byshowing significant prolongation only in the lateral femoral condyle of the rabbits,which however showed more severe degeneration. In study I, CMPG-T2 also showed atrend towards elevated values but without statistical significance. T2-relaxation timemaps of intact articular cartilage typically show a laminar appearance associated withthe histological cartilage zones, which has been reported in many animal and humanstudies [69–74, 77, 79, 180]. In study I, however, rabbit cartilage was relatively thin,making it difficult to detect the laminar structure in T2-maps while in human cartilagespecimens both T2-methods demonstrated the typical tri-laminar structure in the earlyOA group. Consistent with previous studies, a significant elevation due to cartilage

72

degeneration was noted with both DE-T2 and SE-T2 methods in the advanced OA groupof human specimens [75, 78, 81–83, 85–87, 181].

T2 relaxation time has been reported to be sensitive to both collagen content andorientation, as well as to water content of the cartilage [69–83]. In study I, collagenfibril anisotropy (PI) was increased in the lateral condyle and was also detected withDE-T2 by an increase in the relaxation time. However, the correlation of DE-T2 orCPMG-T2 with PI or Edyn was not statistically significant in study I. In the humansample study, only DE-T2 correlated significantly with both PI and Edyn parameters.There were also significant correlations between T2 methods and Eeq, which havealso been reported previously [73]. Similarly, T2 methods correlated moderately withFTIR-derived PG-content. DE-T2 and SE-T2 also correlated significantly with theOARSI grade, the correlation being stronger with DE-T2. The reason for the bettersensitivity of DE-T2 might originate from improved robustness of the adiabatic inversionpulses in comparison to the hard pulse CPMG or SE-based T2, improving estimation ofthe relaxation time.

The dGEMRIC technique is a well-established method for evaluating the GAGcontent of cartilage, which has been demonstrated in vitro and in vivo [80, 86, 88–97].In the present study, the T1 relaxation time was mapped after 24-hour equilibrationof the human specimens in gadopentetate contrast agent. The advanced OA samplesshowed significant PG depletion in the deep cartilage and there was a decreasing trendin T1Gd relaxation time in all cartilage zones in the advanced OA group in comparisonto the early OA group. However, the changes in T1Gd between the groups were notstatistically significant. Moreover, there were no significant correlations between T1Gd

and any of the reference methods including that between T1Gd and PG content, althoughthe depth-wise profiles of T1Gd and PG content appeared similar in both groups. Inaddition, the T1Gd relaxation time maps also resembled FTIRI and Safranin-O maps.Similarly, in a previous study the dGEMRIC technique could not distinguish betweenearly hip OA and reference cartilage [182]. T1 relaxation time was also mapped priorto immersion into contrast agent and was significantly elevated in the advanced OAgroup. Previously, pre-contrast T1 relaxation time has been linked to the water contentof cartilage [101] and in this study it demonstrated significant correlations with T1 andEeq, Edyn and OARSI grade.

MRI relaxation parameters in the rotating frame of reference such as T1ρ , T2ρ

and TRAFF utilize spin-locking RF pulses and the relaxation processes are sensitiveto magnetic field fluctuations in the kHz range [48, 51]. Thus, the parameters are

73

sensitive to a broad range of slow macromolecular motion, which are relevant in manypathologies such as OA. T1ρ relaxation time has been the most intensively studiedRFR parameter for cartilage and it has been suggested as biomarker for early cartilagedegeneration. Furthermore, it has been associated with the PG content of cartilage[45, 104, 112–114, 116, 117, 120, 123, 124, 127, 132]. Traditionally T1ρ has beenmeasured with CW RF pulses, although RFR relaxation measurement can also beimplemented with adiabatic type RF pulses.

In the present study, CW-T1ρ detected early cartilage degeneration in the lateralfemoral condyle in the rabbit ACLT model and was also significantly elevated inadvanced OA human samples. Similar to the present findings, CW-T1ρ was significantlyelevated at three weeks in the lateral femoral condyles in a previous study with therabbit ACLT model [126]. At 6 weeks post-ACLT, both compartments had significantlyhigher T1ρ values compared to the control joint, although there was no clear quantitativehistological evidence of the progression of OA [126]. In previous clinical studies at3 T, CW-T1ρ with spin-lock power γB1 = 500 Hz was significantly increased in kneeOA patients compared with healthy controls as well as in patients with post-traumaticcartilage degeneration 1 and 2 years after ACL reconstruction [123, 124, 127]. Similarlyat 1.5 T, CW-T1ρ (γB1 = 500 Hz) was elevated due to arthroscopically confirmedcartilage degeneration in the knee joint [118].

The CW-T1ρ dispersion phenomenon was evident in study II, showing features of atri-laminar appearance at spin-locking powers less than 1 kHz. In the CW-T1ρ (γB1 = 1kHz) map, relaxation times were decreasing from surface towards deep cartilage. Inaddition, PG content as measured with DD was correlated significantly with CW-T1ρ atγB1 = 250-500 Hz while at γB1 = 125 Hz CW-T1ρ correlated significantly with collagenfibril anisotropy. These findings are consistent with reports associating T1ρ to the PGcontent of cartilage as well as with the relaxation theory, as the T1ρ relaxation timetheoretically approaches T2 with decreasing spin-locking frequency [45]. However,at 1 kHz spin-locking field, one would also expect significant correlation with thePG content, which was not found in the present study. Possible tissue erosion of PGdepleted superficial cartilage in the advanced OA samples (study II) may also affectthe correlation and explain this unexpected finding. With respect to other referenceparameters, CW-T1ρ correlated significantly with Eeq, Edyn, water content and OARSIgrade, the correlations being highest at spin-lock frequencies of 250 and 500 Hz. Incontrast, a previous study reported no significant correlation of histological Mankingrade with T1ρ or T2 on osteoarthritic human specimens. However, the OARSI grading

74

system has a larger dynamic range for early OA grades in comparison with the Mankingrading [173], which may partly explain the discrepancy.

Of the rotating frame MRI contrasts, adiabatic T1ρ and adiabatic T2ρ have not yetbeen widely tested in cartilage. Previously adiabatic T1ρ has been found to be sensitiveto trypsin induced PG depletion in bovine articular cartilage [107]. In studies I and II,both adiabatic T1ρ and adiabatic T2ρ were sensitive to degeneration in both medial andlateral condyles of rabbits as well as degeneration in human specimens, as demonstratedby significant elevation of the relaxation times. Adiabatic T1ρ relaxation times weredecreasing towards deep cartilage while a laminar appearance was shown in adiabaticT2ρ maps in the early OA specimens. In study II, both relaxation parameters were highlycorrelated with Eeq, Edyn and OARSI grade. Moreover, significant correlations werefound with PG content and there was also a significant relationship between adiabaticT1ρ and water content of cartilage. On the other hand, there were no association betweenPG content or collagen anisotropy and adiabatic relaxation parameters in study I. Thedecrease in the PG content and changes in collagen fibril anisotropy were small andlimited only to the superficial part of cartilage, which may have led to averaging in theROI analysis.

RAFF is an attracting RFR parameter with low SAR properties [51–53], which hasbeen shown to be a potential biomarker for cartilage degeneration [107]. In the presentstudy, RAFF was not sensitive to early cartilage degeneration in rabbit specimens whilein study II there were significantly prolonged relaxation times in the advanced OAgroup of human specimens. In addition, RAFF correlated significantly with Eeq, Edyn,OARSI grade and the FTIR-derived PG content in study II. The sample orientation withrespect to B0 was different between the studies which might at least partially explain thedifference in the findings. Furthermore, the cartilage degeneration induced by trypsin inthe previous study was much more severe than the post-ACLT degeneration in the rabbitspecimens.

In the present study, the MT effect in cartilage was investigated with an inversionprepared MT experiment and measurement of the T1sat , the relaxation time during off-resonance saturation. With inversion preparation, the dynamic range of the measurementis increased, improving the parameter mapping and thereby reducing MT pulse lengthcompared to conventional ways of measuring MT [55]. T1sat showed an elevated trend instudy I, but without statistical significance. In study II, T1sat was significantly elevated inthe advanced OA group, similar to a previous study on trypsin digested bovine cartilagein comparison to control groups [107]. Furthermore, T1sat was highly correlated with the

75

Eeq, Edyn and OARSI grade and there were also significant correlations with water andPG contents. In previous studies, MT in cartilage has been linked to changes in thecollagen content and to interaction of mobile water with collagen or collagen-boundwater [102]. As in the present study, some studies have reported minor contributions ofPGs to MT in cartilage [183, 184].

The presented multiparametric measurement of quantitative MRI parameters is notfeasible in a clinical setting, for example, due to imaging time limitations and thusthe most sensitive MRI parameters should be identified. In studies I and II, the ROCanalysis was performed to investigate the accuracy of the quantitative MRI parametersto separate between the two sample groups. In previous clinical studies, CW-T1ρ

has been reported as a more sensitive biomarker for cartilage degeneration than T2

[123, 132, 133, 185]. This has also been demonstrated with the ROC analysis [133, 134],which is in agreement with the present results. In the present study, the AUC values ofadiabatic T1ρ , adiabatic T2ρ and CW-T1ρ relaxation parameters were high (0.8-1) in boththe rabbit ACLT model and osteoarthritic human specimens in the full-thickness ROIanalysis.

In studies I and II, the highest relative changes ranging from 50 to 60% were obtainedfor CW-T1ρ , adiabatic T2ρ and T2 relaxation parameters in the full-thickness analysis ofcartilage. In study II, RAFF was also elevated over 60% in the advanced OA group.For adiabatic T1ρ the change was approximately 30% in both studies. The relativechanges of adiabatic T1ρ , RAFF and T1sat are similar to those found in a previous study[107]. The present results together with the ROC analysis indicate a high sensitivity andspecificity of adiabatic RFR methods for cartilage degeneration.

In the present work, the relatively small number of samples in studies may havelimited the statistical power of comparisons between the groups. Furthermore, com-pletely intact cartilage specimens were not studied in study I or II. In the ACLT rabbitmodel, the contralateral knees likely do not represent fully intact tissue [186] andobtaining intact human cartilage specimens is typically difficult. However, the MRItechniques could separate between the sample groups and with healthy control tissues thechanges could have been further emphasized. The MRI experiments were conducted ina non-physiological environment at very high 9.4 T B0 field strength, and consequentlythe MRI techniques investigated require clinical validation studies. In study I, anotherlimitation was the relatively thin cartilage on rabbit femoral condyles, approximately 10imaging pixels, which is comparable, however, to the resolution achieved in humanstudies on clinical scanners. The advanced OA samples in study II may have experienced

76

tissue erosion which may have an effect on the zonal analysis of cartilage specimens.In study I, the cartilage degeneration was much milder and the tissue erosion was notso significant. Moreover, relaxation times in the superficial cartilage may have beeninfluenced by small amounts of residual PBS solution since specimens in studies I and IIwere thawed in PBS prior to experiments, which may have led to longer relaxationtimes than expected. Nonetheless, perfluoropolyether was used to minimize this effect.Furthermore, the thicknesses of the full-thickness ROIs in study II were comparedagainst the thicknesses derived from microscopy images. Although the segmentationof cartilage ROIs was done manually, the intra- and inter-operator variation was notinvestigated in the present study. However, the segmentation of the ex vivo specimens isquite straightforward and it was done consistently for each sample using the criteriaspecified in each study.

The comparison between CW-T1ρ and adiabatic T1ρ was limited since the SAR-values were not matched. With adiabatic pulses the peak spin-lock power was 2.5kHz which equals approximately to 1090 Hz RMS power, which in turn is close to theCW-T1ρ experiment (γB1 = 1 kHz) in studies I and II. In clinical human studies, the useof CW-T1ρ has been limited to relatively low spin lock powers (γB1 ≤ 500 Hz) due toSAR limitations, especially at high magnetic field strengths. Adiabatic T1ρ and T2ρ

measured using HSn pulse trains also require relatively high peak RF pulse power tomaintain adiabatic condition, which may limit their usefulness at high B0. However,both techniques have been successfully used for in vivo human brain imaging at 4 T[51, 135].

MRI relaxation time parameters have different sensitivities to the sample orientationwith respect to the main magnetic field. In the present work, specimens were placedconsistently in each study such that differences between samples due to the orientationwere minimized. T2 relaxation time is known to be susceptible to the magic angle effect,i.e. sensitive to the orientation of the collagen network with respect to the B0 magneticfield. In addition, the RFR parameters have also been reported to be sensitive to themagic angle effect in articular cartilage. In initial studies, T1ρ has been shown to be lesssensitive to sample orientation than T2, with adiabatic T1ρ being even less sensitive thanCW-T1ρ [129–131].

In summary, the systematic comparison of different quantitative MRI relaxationtechniques revealed that rotating frame MRI relaxation parameters, namely T1ρ and T2ρ ,which are sensitive to slow macromolecular motion, efficiently detected degeneration ina rabbit model of early OA and in spontaneously degenerated human articular cartilage.

77

Furthermore, T1ρ , T2ρ and also inversion prepared MT and RAFF were highly correlatedto the histopathological grade and biomechanical status of the tissue in study II. AlthoughRAFF was not sensitive to degeneration in the ACLT rabbit model, the techniquecan be implemented in a number of different configurations and higher rank rotatingframes [53]. After optimizing the technique it may become a potential biomarker forcartilage degeneration, especially owing to its low SAR properties compared to the otherRFR techniques. Furthermore, the method has already been implemented for imagingthe human brain at 4 T with promising results [51, 52]. However, banding artifactsmost likely caused by B0 inhomogeneities in study I in CW-T1ρ as well as in RAFFmaps [187] were absent in the adiabatic T1ρ or adiabatic T2ρ data, demonstrating theadvantages of using adiabatic pulses that are less sensitive to B1 and B0 field variations.The present findings are also in agreement with another recent investigation in which theadiabatic T1ρ was found to be the most sensitive parameter to trypsin-induced changes inbovine articular cartilage structure [107]. The adiabatic RF pulses cover a broad rangeof resonance frequencies [48, 49], which may be the reason for improved sensitivity ofthe adiabatic T1ρ method as compared to the CW preparation. In addition, T1ρ in generalhas been reported to be a more sensitive biomarker for cartilage degeneration and lesssensitive to the magic angle effect [129, 130] than T2. In the future, clinical studies withRFR techniques employing adiabatic RF pulses may be preferable due to increasedflexibility for dealing with the SAR constraints and reduced B0 and B1 inhomogeneityissues.

7.2 MRI of short T2 components

In the present study, the feasibility of the SWIFT MRI technique for assessment ofspontaneous osteochondral repair in the equine model and osteoarthritic changes in thehuman cartilage–bone interface was investigated. Assessment of tissues with very shortT2 relaxation time such as subchondral bone, or the zone of calcified cartilage require anMRI method, that is capable of capturing T2 species in the sub-millisecond range. TheSWIFT method utilizes interleaved excitation and signal acquisition [20, 21], capturingvirtually all of the signals which are undetectable with conventional SE or GRE basedsequences [18, 19].

The feasibility of SWIFT for evaluating features of articular cartilage and subchon-dral bone was demonstrated in study III. More specifically, the healing outcomes of theequine specimens were detected using SWIFT with fat and water suppression schemes.

78

Spontaneous cartilage repair was more successful in osteochondral lesions as comparedto chondral lesions, which was also shown in the SWIFT images. Osteochondrallesions penetrate into subchondral bone, which releases stem cells that are capable ofdifferentiation and production of new cartilaginous tissue and bone while spontaneousrepair of chondral-only lesions is much more inefficient [4, 188–191]. With the fatand water suppression schemes of SWIFT, fatty bone marrow could be separated fromfibrous tissue that infiltrated into the subchondral bone. In both equine and humancartilage, SWIFT without any contrast modifications demonstrated constant signalintensities in non-calcified cartilage in both degenerated tissue and intact cartilage,making the technique useful, for example, in morphological studies. Magnetizationpreparation schemes in SWIFT could also be used for relaxation time mapping (T1,T1ρ , TRAFF ) [64, 65, 67, 68], whiche provides a new perspective in the quantitativeassessment of articular cartilage and bone.

In SWIFT difference images of relatively intact human cartilage (unpublished data),a band of high intensity signal was detected at the cartilage-bone interface, in the zone ofcalcified cartilage. A similar high-intensity signal at the osteochondral junction has alsobeen found in UTE studies that utilize subtraction images of different TEs [159–161].However, the same feature was not seen in equine cartilage (study III), possibly due toresolution as equine cartilage was much thinner than human cartilage in the presentstudy, or due to actual physiologic differences.

With the advancement of cartilage degeneration, SWIFT difference images demon-strated loss or thinning as well as diffuse thickening of the high intensity signal atthe cartilage–bone interface, which were associated with sclerosis of the subchondralbone, thickening of calcified cartilage and duplication of tidemarks, which are typicalfeatures in advanced OA [4, 9, 38]. The changes in the high intensity signal band are inagreement with recent UTE studies [161] and have previously been associated withincreased vascular invasion of cartilage due to degeneration [162]. The source of thehigh intensity signal seems to correspond to the zone of calcified cartilage and/or deepregions in articular cartilage. Furthermore, the signal intensity changes likely reflect OAin some way, although the mechanisms are not fully understood.

Structural bone parameters derived from SWIFT combination images (fat + waterimage) correlated highly with those derived from micro-CT reference measurements,demonstrating the advantage of SWIFT in the imaging of calcified structures with avery short transverse relaxation time. In previous studies with high-resolution MRI,high correlations have also been found between MRI and micro-CT derived structural

79

bone parameters [81, 152, 153, 155]. It has been reported that the voxel size shouldbe less than 1003 µm3 for a reliable estimation of the bone parameters and detectionof osteoporotic changes [175]. Although SWIFT overestimated the absolute boneparameter values, the findings are in agreement with previous studies [153, 156].

In study III, the SNRs of SWIFT, GRE and FSE images were compared. Due tothe superior T2 sensitivity of SWIFT in comparison to GRE and FSE techniques, theSNR was highest for SWIFT throughout the osteochondral specimens, especially in thesubchondral cortical bone. Similar findings were observed in the human specimensalthough absolute SNRs were not calculated. As deep layers of cartilage have a short T2

relaxation time, intact cartilage appeared thicker in the SWIFT images in comparisonto FSE imaging. This is because the longer TE of the FSE sequence results in poorervisibility of the deep cartilage layers and especially of bone. At lesion sites, thedifferences in cartilage thicknesses were minor, probably due to increased water contentand resulting longer T2 relaxation times.

Previously, the SWIFT technique has been applied to the imaging of musculoskeletaltissues in only a few studies [20, 65, 148, 157]. The present findings indicate that SWIFTcould be applied in the three-dimensional morphometric assessment of articular cartilageand subchondral bone, thereby not being limited to only monitoring osteoarthriticchanges. With a further reduction in imaging time, three-dimensional relaxation timemapping with SWIFT may become useful for evaluating the status of the whole joint.Besides bone and cartilage, SWIFT could also be applied to other joint componentwith short T2s such as menisci, ligaments or tendons. Furthermore, evaluation of theosteochondral junction could be useful in verifying the existence of calcified cartilageafter cartilage repair. However, further in vitro and in vivo studies are required fortesting its clinical applicability. Although SWIFT has moderate requirements on peakRF power, the interleaved excitation and signal acquisition scheme requires very fasttransmit/receive circuits to yield practically zero effective TE. In terms of achievableecho times and scanning times, SWIFT is comparable to UTE and ZTE sequences [57].One important advantage of SWIFT regarding the clinical implementation is the nearlysilent MRI scans [20, 148], due to incremental changes in gradients, which are morecomfortable for patients. In summary, SWIFT enabled the assessment of spontaneousosteochondral repair in an equine model and was sensitive to osteoarthritic changes inthe human cartilage–bone interface.

80

7.3 Future prospects

In future, the intention is to perform clinical measurements using quantitative MRImethods, particularly with adiabatic T1ρ and adiabatic T2ρ to verify the potentialindicated by the ex vivo results. In vivo studies with RFR techniques employing adiabaticRF pulses are preferable due to SAR properties and insensitivity to B0 and B1 fieldinhomogeneities. Adiabatic T1ρ dispersion in cartilage could be studied by modifyingthe contrast using different modulation functions with HSn pulses. In addition, RFRtechniques should be compared to conventional T2 measurements. Development ofmagnetization preparation schemes for SWIFT may provide new aspects for quantitativeMRI studies.

As all joint compartments are affected in OA, SWIFT should be applied to other shortT2 tissues and their pathologies such as menisci, tendons and ligaments. Furthermore,signal changes in subchondral bone and marrow with pathologies are to be quantifiedand investigated. Future studies should also compare SWIFT with UTE and ZTEsequences, for example, in terms of SNR and T2 sensitivity. Investigation of the originof the high intensity signal at the cartilage–bone interface in SWIFT images and themechanisms of signal change in OA could potentially advance knowledge about thedisease pathology.

81

82

8 Summary and conclusions

This thesis can be summarized into the following findings:

1. Adiabatic T1ρ and T2ρ relaxation time parameters were the most sensitive to bothvery early degenerative changes in a rabbit model of early cartilage degeneration anddegeneration in human cartilage

2. Rotating frame relaxation parameters and an inversion-prepared MT experimentwere highly correlated with biomechanical parameters and the histopathologicalOARSI grade of degenerated human cartilage specimens

3. The SWIFT MRI technique enabled three-dimensional assessment of the struc-tural properties of subchondral bone as well as the assessment of the outcome ofspontaneous osteochondral repair

4. SWIFT MRI was sensitive to osteoarthritic tissue changes in the zone of calcifiedcartilage and subchondral bone, which were associated with the degree of OA.

83

84

References

1. Hunter W (1995) Of the structure and disease of articulating cartilages. 1743. Clin OrthopRelat Res (317)(317): 3–6.

2. Mow VC, Ratcliffe A & Poole AR (1992) Cartilage and diarthrodial joints as paradigms forhierarchical materials and structures. Biomaterials 13(2): 67–97.

3. Buckwalter JA & Mankin HJ (1998) Articular cartilage: tissue design and chondrocyte-matrix interactions. Instr Course Lect 47: 477–486.

4. Buckwalter JA & Mankin HJ (1998) Articular cartilage: degeneration and osteoarthritis,repair, regeneration, and transplantation. Instr Course Lect 47: 487–504.

5. Goldring MB & Goldring SR (2010) Articular cartilage and subchondral bone in thepathogenesis of osteoarthritis. Ann N Y Acad Sci 1192: 230–237.

6. Buckwalter JA & Martin JA (2006) Osteoarthritis. Adv Drug Deliv Rev 58(2): 150–167.7. Madry H, Luyten FP & Facchini A (2012) Biological aspects of early osteoarthritis. Knee

Surg Sports Traumatol Arthrosc 20(3): 407–422.8. Li G, Yin J, Gao J, Cheng TS, Pavlos NJ, Zhang C & Zheng MH (2013) Subchondral bone

in osteoarthritis: insight into risk factors and microstructural changes. Arthritis Res Ther15(6): 223.

9. Hoemann CD, Lafantaisie-Favreau CH, Lascau-Coman V, Chen G & Guzman-Morales J(2012) The cartilage-bone interface. J Knee Surg 25(2): 85–97.

10. Jr DR, Koenig L, Dall TM, Gallo P, Narzikul A, Parvizi J & Tongue J (2013) The direct andindirect costs to society of treatment for end-stage knee osteoarthritis. J Bone Joint Surg Am95(16): 1473–1480.

11. Heliövaara M, Slätis P & Paavolainen P (2008) Nivelrikon esiintyvyys ja kustannukset.Duodecim 124(16): 1869–1874.

12. Harris WH & Sledge CB (1990) Total hip and total knee replacement. N Engl J Med323(12): 801–807.

13. Roemer FW, Crema MD, Trattnig S & Guermazi A (2011) Advances in imaging ofosteoarthritis cartilage. Radiology 260(2): 332–354.

14. Gold GE, Burstein D, Dardzinski B, Lang P, Boada F & Mosher T (2006) Mri of articularcartilage in oa: novel pulse sequences and compositional/functional markers. OsteoarthritisCartilage 14 Suppl A: A76–86.

15. Nieminen MT, Nissi MJ, Mattila L & Kiviranta I (2012) Evaluation of chondral repair usingquantitative mri. J Magn Reson Imaging 36(6): 1287–1299.

16. Li X & Majumdar S (2013) Quantitative mri of articular cartilage and its clinical applications.J Magn Reson Imaging 38(5): 991–1008.

17. Oei EH, van Tiel J, Robinson WH & Gold GE (2014) Quantitative radiologic imagingtechniques for articular cartilage composition: toward early diagnosis and development ofdisease-modifying therapeutics for osteoarthritis. Arthritis Care Res 66(8): 1129–1141.

18. Gatehouse PD & Bydder GM (2003) Magnetic resonance imaging of short t2 components intissue. Clin Radiol 58(1): 1–19.

19. Horch RA, Nyman JS, Gochberg DF, Dortch RD & Does MD (2010) Characterization of 1hnmr signal in human cortical bone for magnetic resonance imaging. Magn Reson Med64(3): 680–687.

85

20. Idiyatullin D, Corum C, Park JY & Garwood M (2006) Fast and quiet mri using a sweptradiofrequency. J Magn Reson 181(2): 342–349.

21. Idiyatullin D, Corum C, Moeller S & Garwood M (2008) Gapped pulses for frequency-sweptmri. J Magn Reson 193(2): 267–273.

22. Hall FM & Wyshak G (1980) Thickness of articular cartilage in the normal knee. J BoneJoint Surg Am 62(3): 408–413.

23. Maroudas A & Venn M (1977) Chemical composition and swelling of normal and os-teoarthrotic femoral head cartilage. ii. swelling. Ann Rheum Dis 36(5): 399–406.

24. Benninghoff A (1925) Form und bau der gelenkknorpel in ihren beziehungen zur funktion.Zeitschrift für Zellforschung und Mikroskopische Anatomie 2(5): 783–862.

25. Roughley PJ (2006) The structure and function of cartilage proteoglycans. Eur Cell Mater12: 92–101.

26. Urban JP (1994) The chondrocyte: a cell under pressure. British journal of rheumatology33(10): 901–908.

27. Keenan KE, Besier TF, Pauly JM, Han E, Rosenberg J, Smith RL, Delp SL, Beaupre GS &Gold GE (2011) Prediction of glycosaminoglycan content in human cartilage by age, t1rhoand t2 mri. Osteoarthritis Cartilage 19(2): 171–179.

28. Wang F, Ying Z, Duan X, Tan H, Yang B, Guo L, Chen G, Dai G, Ma Z & Yang L (2009)Histomorphometric analysis of adult articular calcified cartilage zone. J Struct Biol 168(3):359–365.

29. Oegema TR, Carpenter RJ, Hofmeister F & Thompson RC (1997) The interaction of thezone of calcified cartilage and subchondral bone in osteoarthritis. Microsc Res Tech 37(4):324–332.

30. Laasanen MS, Toyras J, Korhonen RK, Rieppo J, Saarakkala S, Nieminen MT, Hirvonen J& Jurvelin JS (2003) Biomechanical properties of knee articular cartilage. Biorheology40(1-3): 133–140.

31. Korhonen RK, Laasanen MS, Toyras J, Lappalainen R, Helminen HJ & Jurvelin JS(2003) Fibril reinforced poroelastic model predicts specifically mechanical behavior ofnormal, proteoglycan depleted and collagen degraded articular cartilage. J Biomech 36(9):1373–1379.

32. Hayes WC & Bodine AJ (1978) Flow-independent viscoelastic properties of articularcartilage matrix. J Biomech 11(8-9): 407–419.

33. Korhonen RK, Laasanen MS, Toyras J, Rieppo J, Hirvonen J, Helminen HJ & JurvelinJS (2002) Comparison of the equilibrium response of articular cartilage in unconfinedcompression, confined compression and indentation. J Biomech 35(7): 903–909.

34. Spilker RL, Suh JK & Mow VC (1992) A finite element analysis of the indentationstress-relaxation response of linear biphasic articular cartilage. J Biomech Eng 114(2):191–201.

35. Hayes WC, Keer LM, Herrmann G & Mockros LF (1972) A mathematical analysis forindentation tests of articular cartilage. J Biomech 5(5): 541–551.

36. Buckwalter JA, Glimcher MJ, Cooper RR & Recker R (1996) Bone biology. ii: Formation,form, modeling, remodeling, and regulation of cell function. Instr Course Lect 45: 387–399.

37. Madry H, van Dijk CN & Mueller-Gerbl M (2010) The basic science of the subchondralbone. Knee Surg Sports Traumatol Arthrosc 18(4): 419–433.

38. Thambyah A & Broom N (2007) On how degeneration influences load-bearing in thecartilage-bone system: a microstructural and micromechanical study. Osteoarthritis Cartilage

86

15(12): 1410–1423.39. Bloch F (1946) Nuclear induction. Phys Rev 70(7-8): 460–474.40. Bloch F, Hansen WW & Packard M (1946) The nuclear induction experiment. Phys Rev

70(7-8): 474–485.41. McRobbie DW, Moore EA, Graves MJ & Prince MR (eds.) (2006) MRI From Picture to

Proton Second edition. Cambridge University Press, New York, United States.42. Bloembergen N, Purcell EM & Pound RV (1948) Relaxation effects in nuclear magnetic

resonance absorption. Phys Rev 73(7): 679–712.43. Hahn EL (1950) Nuclear induction due to free larmor precession. Phys Rev 77(2): 297–298.44. Hahn EL (1950) Spin echoes. Phys Rev 80(4): 580–594.45. Borthakur A, Mellon E, Niyogi S, Witschey W, Kneeland JB & Reddy R (2006) Sodium

and t1rho mri for molecular and diagnostic imaging of articular cartilage. NMR Biomed19(7): 781–821.

46. Sepponen RE, Pohjonen JA, Sipponen JT & Tanttu JI (1985) A method for t1 rho imaging. JComput Assist Tomogr 9(6): 1007–1011.

47. Jones GP (1966) Spin-lattice relaxation in the rotating frame: Weak-collision case. Phys.Rev. 148: 332–335.

48. Mangia S, Liimatainen T, Garwood M & Michaeli S (2009) Rotating frame relaxationduring adiabatic pulses vs. conventional spin lock: simulations and experimental results at 4t. Magn Reson Imaging 27(8): 1074–1087.

49. Garwood M & DelaBarre L (2001) The return of the frequency sweep: designing adiabaticpulses for contemporary nmr. J Magn Reson 153(2): 155–177.

50. Michaeli S, Sorce DJ, Idiyatullin D, Ugurbil K & Garwood M (2004) Transverse relaxationin the rotating frame induced by chemical exchange. J Magn Reson 169(2): 293–299.

51. Liimatainen T, Sorce DJ, O’Connell R, Garwood M & Michaeli S (2010) Mri contrast fromrelaxation along a fictitious field (raff). Magn Reson Med 64(4): 983–994.

52. Liimatainen T, Mangia S, Ling W, Ellermann J, Sorce DJ, Garwood M & Michaeli S (2011)Relaxation dispersion in mri induced by fictitious magnetic fields. J Magn Reson 209(2):269–276.

53. Liimatainen T, Hakkarainen H, Mangia S, Huttunen JM, Storino C, Idiyatullin D, Sorce D,Garwood M & Michaeli S (2014) Mri contrasts in high rank rotating frames. Magn ResonMed .

54. Henkelman RM, Stanisz GJ & Graham SJ (2001) Magnetization transfer in mri: a review.NMR Biomed 14(2): 57–64.

55. Mangia S, Martino FD, Liimatainen T, Garwood M & Michaeli S (2011) Magnetizationtransfer using inversion recovery during off-resonance irradiation. Magn Reson Imaging29(10): 1346–1350.

56. Robson MD, Gatehouse PD, Bydder M & Bydder GM (2003) Magnetic resonance: anintroduction to ultrashort te (ute) imaging. J Comput Assist Tomogr 27(6): 825–846.

57. Du J, Bydder M, Takahashi AM, Carl M, Chung CB & Bydder GM (2011) Short t2 contrastwith three-dimensional ultrashort echo time imaging. Magn Reson Imaging 29(4): 470–482.

58. Du J, Bydder M, Takahashi AM & Chung CB (2008) Two-dimensional ultrashort echo timeimaging using a spiral trajectory. Magn Reson Imaging 26(3): 304–312.

59. Weiger M, Hennel F & Pruessmann KP (2010) Sweep mri with algebraic reconstruction.Magn Reson Med 64(6): 1685–1695.

87

60. Weiger M, Pruessmann KP & Hennel F (2011) Mri with zero echo time: hard versus sweeppulse excitation. Magn Reson Med 66(2): 379–389.

61. Garwood M, Idiyatullin D, Corum CA, Chamberlain R, Moeller S, Kobayashi N, Lehto LJ,Zhang J, O’Connell R, Tesch M, Nissi MJ, Ellermann J & Nixdorf DR (2012) CapturingSignals from Fast-relaxing Spins with Frequency-Swept MRI: SWIFT. Encyclopedia ofMagnetic Resonance. Wiley.

62. Garwood M (2013) Mri of fast-relaxing spins. J Magn Reson 229: 49–54.63. Corum C, Moeller S, Idiyatullin D & Garwood M (2007) Signal processing and image

reconstruction for swift. Proc Intl Soc Mag Reson Med 15: 1669.64. Wang L, Corum CA, Idiyatullin D, Garwood M & Zhao Q (2013) T(1) estimation for

aqueous iron oxide nanoparticle suspensions using a variable flip angle swift sequence.Magn Reson Med 70(2): 341–347.

65. Nissi MJ, Lehto LJ, Corum CA, Idiyatullin D, Ellermann JM, Grohn OH & Nieminen MT(2014) Measurement of t relaxation time of osteochondral specimens using vfa-swift. MagnReson Med .

66. Ellermann J, Idiyatullin D, Corum C, Moeller S, Ling W & Garwood M (2009) Theultrastructure of bone using selective saturation in swift at 9.4 t. Proc Intl Soc Mag ResonMed 17: 849.

67. Zhang J, Nissi MJ, Idiyatullin D, Michaeli S, Garwood M & Ellermann J (2013) Swift withmagnetization preparation: Signal partitioning and 3d measurement of adiabatic t1rho inosteochondral specimen. Proc. Intl. Soc. Mag. Reson. Med. 21: 764.

68. Liimatainen T, Idiyatullin D, Zhang J, Hakkarainen H, Mangia S, Garwood M & Michaeli S(2014) Maximizing tissue contrast in mri using raffn preparation with swift and spin echoreadouts. Proc Intl Soc Mag Reson Med 22: 341.

69. Xia Y, Farquhar T, Burton-Wurster N & Lust G (1997) Origin of cartilage laminae in mri. JMagn Reson Imaging 7(5): 887–894.

70. Xia Y (1998) Relaxation anisotropy in cartilage by nmr microscopy (mumri) at 14-micromresolution. Magn Reson Med 39(6): 941–949.

71. Xia Y, Moody JB & Alhadlaq H (2002) Orientational dependence of t2 relaxation in articularcartilage: A microscopic mri (micromri) study. Magn Reson Med 48(3): 460–469.

72. Nieminen MT, Rieppo J, Toyras J, Hakumaki JM, Silvennoinen J, Hyttinen MM, HelminenHJ & Jurvelin JS (2001) T2 relaxation reveals spatial collagen architecture in articularcartilage: a comparative quantitative mri and polarized light microscopic study. MagnReson Med 46(3): 487–493.

73. Nissi MJ, Rieppo J, Toyras J, Laasanen MS, Kiviranta I, Nieminen MT & Jurvelin JS (2007)Estimation of mechanical properties of articular cartilage with mri - dgemric, t2 and t1imaging in different species with variable stages of maturation. Osteoarthritis Cartilage15(10): 1141–1148.

74. Nissi MJ, Rieppo J, Toyras J, Laasanen MS, Kiviranta I, Jurvelin JS & Nieminen MT (2006)T(2) relaxation time mapping reveals age- and species-related diversity of collagen networkarchitecture in articular cartilage. Osteoarthritis Cartilage 14(12): 1265–1271.

75. Mosher TJ & Dardzinski BJ (2004) Cartilage mri t2 relaxation time mapping: overview andapplications. Semin Muskuloskelet Radiol 8(4): 355–368.

76. Mosher TJ, Smith H, Dardzinski BJ, Schmithorst VJ & Smith MB (2001) Mr imaging andt2 mapping of femoral cartilage: in vivo determination of the magic angle effect. AJR Am JRoentgenol 177(3): 665–669.

88

77. Smith HE, Mosher TJ, Dardzinski BJ, Collins BG, Collins CM, Yang QX, Schmithorst VJ& Smith MB (2001) Spatial variation in cartilage t2 of the knee. J Magn Reson Imaging14(1): 50–55.

78. Dardzinski BJ, Mosher TJ, Li S, Slyke MAV & Smith MB (1997) Spatial variation of t2 inhuman articular cartilage. Radiology 205(2): 546–550.

79. Mlynarik V, Szomolanyi P, Toffanin R, Vittur F & Trattnig S (2004) Transverse relaxationmechanisms in articular cartilage. J Magn Reson 169(2): 300–307.

80. Lammentausta E, Kiviranta P, Nissi MJ, Laasanen MS, Kiviranta I, Nieminen MT & JurvelinJS (2006) T2 relaxation time and delayed gadolinium-enhanced mri of cartilage (dgemric) ofhuman patellar cartilage at 1.5 t and 9.4 t: Relationships with tissue mechanical properties. JOrthop Res 24(3): 366–374.

81. Lammentausta E, Kiviranta P, Toyras J, Hyttinen MM, Kiviranta I, Nieminen MT & JurvelinJS (2007) Quantitative mri of parallel changes of articular cartilage and underlying trabecularbone in degeneration. Osteoarthritis Cartilage 15(10): 1149–1157.

82. Liess C, Lusse S, Karger N, Heller M & Gluer CC (2002) Detection of changes in cartilagewater content using mri t2-mapping in vivo. Osteoarthritis Cartilage 10(12): 907–913.

83. Lusse S, Claassen H, Gehrke T, Hassenpflug J, Schunke M, Heller M & Gluer CC (2000)Evaluation of water content by spatially resolved transverse relaxation times of humanarticular cartilage. Magn Reson Imaging 18(4): 423–430.

84. Erickson SJ, Prost RW & Timins ME (1993) The "magic angle" effect: background physicsand clinical relevance. Radiology 188(1): 23–25.

85. Nieminen MT, Toyras J, Rieppo J, Hakumaki JM, Silvennoinen J, Helminen HJ & JurvelinJS (2000) Quantitative mr microscopy of enzymatically degraded articular cartilage. MagnReson Med 43(5): 676–681.

86. Nissi MJ, Toyras J, Laasanen MS, Rieppo J, Saarakkala S, Lappalainen R, Jurvelin JS &Nieminen MT (2004) Proteoglycan and collagen sensitive mri evaluation of normal anddegenerated articular cartilage. J Orthop Res 22(3): 557–564.

87. Kijowski R, Blankenbaker DG, Rio AMD, Baer GS & Graf BK (2013) Evaluation of thearticular cartilage of the knee joint: Value of adding a t2 mapping sequence to a routine mrimaging protocol. Radiology 267(2): 503–513.

88. Bashir A, Gray ML & Burstein D (1996) Gd-dtpa2- as a measure of cartilage degradation.Magn Reson Med 36(5): 665–673.

89. Bashir A, Gray ML, Boutin RD & Burstein D (1997) Glycosaminoglycan in articularcartilage: in vivo assessment with delayed gd(dtpa)(2-)-enhanced mr imaging. Radiology205(2): 551–558.

90. Bashir A, Gray ML, Hartke J & Burstein D (1999) Nondestructive imaging of humancartilage glycosaminoglycan concentration by mri. Magn Reson Med 41(5): 857–865.

91. Williams A, Gillis A, McKenzie C, Po B, Sharma L, Micheli L, McKeon B & Burstein D(2004) Glycosaminoglycan distribution in cartilage as determined by delayed gadolinium-enhanced mri of cartilage (dgemric): potential clinical applications. AJR Am J Roentgenol182(1): 167–172.

92. Williams A, Mikulis B, Krishnan N, Gray M, McKenzie C & Burstein D (2007) Suitabilityof t(1gd) as the dgemric index at 1.5t and 3.0t. Magn Reson Med 58(4): 830–834.

93. Williams A, Sharma L, McKenzie CA, Prasad PV & Burstein D (2005) Delayed gadolinium-enhanced magnetic resonance imaging of cartilage in knee osteoarthritis: findings at differentradiographic stages of disease and relationship to malalignment. Arthritis Rheum 52(11):

89

3528–3535.94. Tiderius CJ, Olsson LE, Nyquist F & Dahlberg L (2005) Cartilage glycosaminoglycan loss

in the acute phase after an anterior cruciate ligament injury: delayed gadolinium-enhancedmagnetic resonance imaging of cartilage and synovial fluid analysis. Arthritis Rheum 52(1):120–127.

95. Tiderius CJ, Olsson LE, Leander P, Ekberg O & Dahlberg L (2003) Delayed gadolinium-enhanced mri of cartilage (dgemric) in early knee osteoarthritis. Magn Reson Med 49(3):488–492.

96. Nieminen MT, Rieppo J, Silvennoinen J, Toyras J, Hakumaki JM, Hyttinen MM, HelminenHJ & Jurvelin JS (2002) Spatial assessment of articular cartilage proteoglycans withgd-dtpa-enhanced t1 imaging. Magn Reson Med 48(4): 640–648.

97. Salo EN, Nissi MJ, Kulmala KA, Tiitu V, Toyras J & Nieminen MT (2012) Diffusion ofgd-dtpa(2)(-) into articular cartilage. Osteoarthritis Cartilage 20(2): 117–126.

98. Li W, Du H, Scheidegger R, Wu Y & Prasad PV (2009) Value of precontrast t1 for dgemricof native articular cartilage. J Magn Reson Imaging 29(2): 494–497.

99. Tiderius CJ, Olsson LE, de Verdier H, Leander P, Ekberg O & Dahlberg L (2001) Gd-dtpa2)-enhanced mri of femoral knee cartilage: a dose-response study in healthy volunteers. MagnReson Med 46(6): 1067–1071.

100. Mlynarik V, Degrassi A, Toffanin R, Vittur F, Cova M & Pozzi-Mucelli RS (1996) In-vestigation of laminar appearance of articular cartilage by means of magnetic resonancemicroscopy. Magn Reson Imaging 14(4): 435–442.

101. Berberat JE, Nissi MJ, Jurvelin JS & Nieminen MT (2009) Assessment of interstitial watercontent of articular cartilage with t1 relaxation. Magn Reson Imaging 27(5): 727–732.

102. Gray ML, Burstein D, Lesperance LM & Gehrke L (1995) Magnetization transfer incartilage and its constituent macromolecules. Magn Reson Med 34(3): 319–325.

103. Mlynarik V, Sulzbacher I, Bittsansky M, Fuiko R & Trattnig S (2003) Investigation ofapparent diffusion constant as an indicator of early degenerative disease in articular cartilage.J Magn Reson Imaging 17(4): 440–444.

104. Regatte RR, Akella SV & Reddy R (2005) Depth-dependent proton magnetization transferin articular cartilage. J Magn Reson Imaging 22(2): 318–323.

105. Stanisz GJ, Odrobina EE, Pun J, Escaravage M, Graham SJ, Bronskill MJ & HenkelmanRM (2005) T1, t2 relaxation and magnetization transfer in tissue at 3t. Magn Reson Med54(3): 507–512.

106. Palmieri F, Keyzer FD, Maes F & Breuseghem IV (2006) Magnetization transfer analysis ofcartilage repair tissue: a preliminary study. Skeletal Radiol 35(12): 903–908.

107. Ellermann J, Ling W, Nissi MJ, Arendt E, Carlson CS, Garwood M, Michaeli S & Mangia S(2013) Mri rotating frame relaxation measurements for articular cartilage assessment. MagnReson Imaging 31(9): 1537–1543.

108. Liu G, Song X, Chan KW & McMahon MT (2013) Nuts and bolts of chemical exchangesaturation transfer mri. NMR Biomed 26(7): 810–828.

109. Ling W, Regatte RR, Navon G & Jerschow A (2008) Assessment of glycosaminogly-can concentration in vivo by chemical exchange-dependent saturation transfer (gagcest).Proceedings of the National Academy of Sciences 105(7): 2266–2270.

110. Schmitt B, Zbyn S, Stelzeneder D, Jellus V, Paul D, Lauer L, Bachert P & Trattnig S (2011)Cartilage quality assessment by using glycosaminoglycan chemical exchange saturationtransfer and (23)na mr imaging at 7 t. Radiology 260(1): 257–264.

90

111. Singh A, Haris M, Cai K, Kassey VB, Kogan F, Reddy D, Hariharan H & Reddy R (2012)Chemical exchange saturation transfer magnetic resonance imaging of human knee cartilageat 3 t and 7 t. Magn Reson Med 68(2): 588–594.

112. Duvvuri U, Reddy R, Patel SD, Kaufman JH, Kneeland JB & Leigh JS (1997) T1rho-relaxation in articular cartilage: effects of enzymatic degradation. Magn Reson Med 38(6):863–867.

113. Duvvuri U, Kudchodkar S, Reddy R & Leigh JS (2002) T(1rho) relaxation can assesslongitudinal proteoglycan loss from articular cartilage in vitro. Osteoarthritis Cartilage10(11): 838–844.

114. Duvvuri U, Charagundla SR, Kudchodkar SB, Kaufman JH, Kneeland JB, Rizi R, Leigh JS& Reddy R (2001) Human knee: in vivo t1(rho)-weighted mr imaging at 1.5 t–preliminaryexperience. Radiology 220(3): 822–826.

115. Duvvuri U, Goldberg AD, Kranz JK, Hoang L, Reddy R, Wehrli FW, Wand AJ, EnglanderSW & Leigh JS (2001) Water magnetic relaxation dispersion in biological systems: thecontribution of proton exchange and implications for the noninvasive detection of cartilagedegradation. Proceedings of the National Academy of Sciences 98(22): 12479–12484.

116. Regatte RR, Akella SV, Wheaton AJ, Borthakur A, Kneeland JB & Reddy R (2003) T 1rho-relaxation mapping of human femoral-tibial cartilage in vivo. J Magn Reson Imaging18(3): 336–341.

117. Regatte RR, Akella SV, Wheaton AJ, Lech G, Borthakur A, Kneeland JB & Reddy R (2004)3d-t1rho-relaxation mapping of articular cartilage: in vivo assessment of early degenerativechanges in symptomatic osteoarthritic subjects. Adad Radioly 11(7): 741–749.

118. Witschey WR, Borthakur A, Fenty M, Kneeland BJ, Lonner JH, McArdle EL, SochorM & Reddy R (2010) T1rho mri quantification of arthroscopically confirmed cartilagedegeneration. Magn Reson Med 63(5): 1376–1382.

119. Wheaton AJ, Borthakur A, Kneeland JB, Regatte RR, Akella SV & Reddy R (2004) In vivoquantification of t1rho using a multislice spin-lock pulse sequence. Magn Reson Med 52(6):1453–1458.

120. Wheaton AJ, Casey FL, Gougoutas AJ, Dodge GR, Borthakur A, Lonner JH, SchumacherHR & Reddy R (2004) Correlation of t1rho with fixed charge density in cartilage. J MagnReson Imaging 20(3): 519–525.

121. Wheaton AJ, Dodge GR, Elliott DM, Nicoll SB & Reddy R (2005) Quantification ofcartilage biomechanical and biochemical properties via t1rho magnetic resonance imaging.Magn Reson Med 54(5): 1087–1093.

122. Wheaton AJ, Dodge GR, Borthakur A, Kneeland JB, Schumacher HR & Reddy R (2005)Detection of changes in articular cartilage proteoglycan by t(1rho) magnetic resonanceimaging. J Orthop Res 23(1): 102–108.

123. Li X, Ma CB, Link TM, Castillo DD, Blumenkrantz G, Lozano J, Carballido-Gamio J,Ries M & Majumdar S (2007) In vivo t(1rho) and t(2) mapping of articular cartilage inosteoarthritis of the knee using 3 t mri. Osteoarthritis Cartilage 15(7): 789–797.

124. Li X, Kuo D, Theologis A, Carballido-Gamio J, Stehling C, Link TM, Ma CB & MajumdarS (2011) Cartilage in anterior cruciate ligament-reconstructed knees: Mr imaging t1rho andt2–initial experience with 1-year follow-up. Radiology 258(2): 505–514.

125. Li X, Cheng J, Lin K, Saadat E, Bolbos RI, Jobke B, Ries MD, Horvai A, Link TM &Majumdar S (2011) Quantitative mri using t1rho and t2 in human osteoarthritic cartilagespecimens: correlation with biochemical measurements and histology. Magn Reson Imaging

91

29(3): 324–334.126. Lozano J, Saadat E, Li X, Majumdar S & Ma CB (2009) Magnetic resonance t(1 rho)

imaging of osteoarthritis: a rabbit acl transection model. Magn Reson Imaging 27(5):611–616.

127. Su F, Hilton JF, Nardo L, Wu S, Liang F, Link TM, Ma CB & Li X (2013) Cartilagemorphology and t1rho and t2 quantification in acl-reconstructed knees: a 2-year follow-up.Osteoarthritis Cartilage 21(8): 1058–1067.

128. Souza RB, Kumar D, Calixto N, Singh J, Schooler J, Subburaj K, Li X, Link TM &Majumdar S (2014) Response of knee cartilage t1rho and t2 relaxation times to in vivomechanical loading in individuals with and without knee osteoarthritis. OsteoarthritisCartilage 22(10): 1367–1376.

129. Wang N & Xia Y (2012) Depth and orientational dependencies of mri t(2) and t(1rho)sensitivities towards trypsin degradation and gd-dtpa(2-) presence in articular cartilage atmicroscopic resolution. Magn Reson Imaging 30(3): 361–370.

130. Nissi MJ, Mangia S, Michaeli S & Nieminen MT (2013) Orientation anisotropy of rotatingframe and t2 relaxation parameters in articular cartilage. Proc Intl Soc Mag Reson Med 21:3552.

131. Du J, Pauli C, Li S, Statum S, Bae W, Chang E & Chung CB (2014) The angular dependenceof t1ρ relaxation in normal and abnormal patellae with histological correlation. Proc IntlSoc Mag Reson Med 22: 148.

132. Regatte RR, Akella SV, Lonner JH, Kneeland JB & Reddy R (2006) T1rho relaxationmapping in human osteoarthritis (oa) cartilage: comparison of t1rho with t2. J Magn ResonImaging 23(4): 547–553.

133. Takayama Y, Hatakenaka M, Tsushima H, Okazaki K, Yoshiura T, Yonezawa M, NishikawaK, Iwamoto Y & Honda H (2013) T1rho is superior to t2 mapping for the evaluation ofarticular cartilage denaturalization with osteoarthritis: Radiological-pathological correlationafter total knee arthroplasty. Eur J Radiol 82(4): e192–8.

134. Nishioka H, Hirose J, Nakamura E, Okamoto N, Karasugi T, Taniwaki T, Okada T, YamashitaY & Mizuta H (2013) Detecting icrs grade 1 cartilage lesions in anterior cruciate ligamentinjury using t1rho and t2 mapping. Eur J Radiol 82(9): 1499–1505.

135. Michaeli S, Oz G, Sorce DJ, Garwood M, Ugurbil K, Majestic S & Tuite P (2007)Assessment of brain iron and neuronal integrity in patients with parkinson’s disease usingnovel mri contrasts. Mov Disord 22(3): 334–340.

136. Xia Y, Farquhar T, Burton-Wurster N, Vernier-Singer M, Lust G & Jelinski LW (1995)Self-diffusion monitors degraded cartilage. Arch Biochem Biophys 323(2): 323–328.

137. Xia Y, Farquhar T, Burton-Wurster N, Ray E & Jelinski LW (1994) Diffusion and relaxationmapping of cartilage-bone plugs and excised disks using microscopic magnetic resonanceimaging. Magn Reson Med 31(3): 273–282.

138. de Visser SK, Bowden JC, Wentrup-Byrne E, Rintoul L, Bostrom T, Pope JM & Momot KI(2008) Anisotropy of collagen fibre alignment in bovine cartilage: comparison of polarisedlight microscopy and spatially resolved diffusion-tensor measurements. OsteoarthritisCartilage 16(6): 689–697.

139. Meder R, de Visser SK, Bowden JC, Bostrom T & Pope JM (2006) Diffusion tensor imagingof articular cartilage as a measure of tissue microstructure. Osteoarthritis Cartilage 14(9):875–881.

92

140. Filidoro L, Dietrich O, Weber J, Rauch E, Oerther T, Wick M, Reiser MF & Glaser C(2005) High-resolution diffusion tensor imaging of human patellar cartilage: feasibility andpreliminary findings. Magn Reson Med 53(5): 993–998.

141. Wheaton AJ, Borthakur A, Dodge GR, Kneeland JB, Schumacher HR & Reddy R (2004)Sodium magnetic resonance imaging of proteoglycan depletion in an in vivo model ofosteoarthritis. Adad Radiol 11(1): 21–28.

142. Trattnig S, Welsch GH, Juras V, Szomolanyi P, Mayerhoefer ME, Stelzeneder D, MamischTC, Bieri O, Scheffler K & Zbyn S (2010) 23na mr imaging at 7 t after knee matrix-associated autologous chondrocyte transplantation preliminary results. Radiology 257(1):175–184.

143. Madelin G, Babb J, Xia D, Chang G, Krasnokutsky S, Abramson SB, Jerschow A & RegatteRR (2013) Articular cartilage: evaluation with fluid-suppressed 7.0-t sodium mr imaging insubjects with and subjects without osteoarthritis. Radiology 268(2): 481–491.

144. Wehrli FW & Fernandez-Seara MA (2005) Nuclear magnetic resonance studies of bonewater. Annals of Biomedical Engineering 33(1): 79–86.

145. Du J, Carl M, Bydder M, Takahashi A, Chung CB & Bydder GM (2010) Qualitative andquantitative ultrashort echo time (ute) imaging of cortical bone. J Magn Reson 207(2):304–311.

146. Wehrli FW (2013) Magnetic resonance of calcified tissues. J Magn Reson 229: 35–48.147. Weiger M, Stampanoni M & Pruessmann KP (2013) Direct depiction of bone microstructure

using mri with zero echo time. Bone 54(1): 44–47.148. Kendi AT, Khariwala SS, Zhang J, Idiyatullin DS, Corum CA, Michaeli S, Pambuccian SE,

Garwood M & Yueh B (2011) Transformation in mandibular imaging with sweep imagingwith fourier transform magnetic resonance imaging. Arch Otolaryngol Head Neck Surg137(9): 916–919.

149. Horch RA, Gochberg DF, Nyman JS & Does MD (2011) Non-invasive predictors ofhuman cortical bone mechanical properties: T(2)-discriminated h nmr compared with highresolution x-ray. PloS one 6(1): e16359.

150. Krug R, Larson PE, Wang C, Burghardt AJ, Kelley DA, Link TM, Zhang X, Vigneron DB &Majumdar S (2011) Ultrashort echo time mri of cortical bone at 7 tesla field strength: Afeasibility study. J Magn Reson Imaging .

151. Li C, Seifert AC, Rad HS, Bhagat YA, Rajapakse CS, Sun W, Lam SC & Wehrli FW (2014)Cortical bone water concentration: dependence of mr imaging measures on age and porevolume fraction. Radiology 272(3): 796–806.

152. Wehrli FW, Song HK, Saha PK & Wright AC (2006) Quantitative mri for the assessment ofbone structure and function. NMR Biomed 19(7): 731–764.

153. Majumdar S, Newitt D, Mathur A, Osman D, Gies A, Chiu E, Lotz J, Kinney J & GenantH (1996) Magnetic resonance imaging of trabecular bone structure in the distal radius:relationship with x-ray tomographic microscopy and biomechanics .

154. Majumdar S & Genant HK (1997) High resolution magnetic resonance imaging of trabecularstructure. Eur Radiol 7 Suppl 2: S51–5.

155. Majumdar S, Kothari M, Augat P, Newitt DC, Link TM, Lin JC, Lang T, Lu Y & GenantHK (1998) High-resolution magnetic resonance imaging: three-dimensional trabecular bonearchitecture and biomechanical properties. Bone 22(5): 445–454.

156. Krug R, Carballido-Gamio J, Burghardt AJ, Kazakia G, Hyun BH, Jobke B, BanerjeeS, Huber M, Link TM & Majumdar S (2008) Assessment of trabecular bone structure

93

comparing magnetic resonance imaging at 3 tesla with high-resolution peripheral quantitativecomputed tomography ex vivo and in vivo .

157. Idiyatullin D, Corum C, Moeller S, Prasad HS, Garwood M & Nixdorf DR (2011) Dentalmagnetic resonance imaging: making the invisible visible. J Endod 37(6): 745–752.

158. Du J, Takahashi AM, Bae WC, Chung CB & Bydder GM (2010) Dual inversion recovery,ultrashort echo time (dir ute) imaging: creating high contrast for short-t(2) species. MagnReson Med 63(2): 447–455.

159. Bae WC, Dwek JR, Znamirowski R, Statum SM, Hermida JC, D’Lima DD, Sah RL, Du J &Chung CB (2010) Ultrashort echo time mr imaging of osteochondral junction of the knee at3 t: identification of anatomic structures contributing to signal intensity. Radiology 254(3):837–845.

160. Du J, Carl M, Bae WC, Statum S, Chang EY, Bydder GM & Chung CB (2013) Dualinversion recovery ultrashort echo time (dir-ute) imaging and quantification of the zone ofcalcified cartilage (zcc). Osteoarthritis Cartilage 21(1): 77–85.

161. Bae WC, Biswas R, Chen K, Chang EY & Chung CB (2014) Ute mri of the osteochondraljunction. Curr Radiol Rep 2(2): 35.

162. Bae WC, Statum S, Biswas R, D’Lima DD, Chang EY & Chung CB (2014) Ute mrmorphology and histopathology of the osteochondral junction of the knee. Proc Intl SocMag Reson Med 22: 153.

163. Michaeli S, Sorce DJ, Jr CSS, Ugurbil K & Garwood M (2006) T1rho mri contrast in thehuman brain: modulation of the longitudinal rotating frame relaxation shutter-speed duringan adiabatic rf pulse. J Magn Reson 181(1): 135–147.

164. Jokivarsi KT, Niskanen JP, Michaeli S, Grohn HI, Garwood M, Kauppinen RA & Grohn OH(2009) Quantitative assessment of water pools by t 1 rho and t 2 rho mri in acute cerebralischemia of the rat. J Cereb Blood Flow Metab 29(1): 206–216.

165. Toyras J, Laasanen MS, Saarakkala S, Lammi MJ, Rieppo J, Kurkijarvi J, Lappalainen R &Jurvelin JS (2003) Speed of sound in normal and degenerated bovine articular cartilage.Ultrasound Med Biol 29(3): 447–454.

166. Rieppo J, Hallikainen J, Jurvelin JS, Kiviranta I, Helminen HJ & Hyttinen MM (2008)Practical considerations in the use of polarized light microscopy in the analysis of thecollagen network in articular cartilage. Microsc Res Tech 71(4): 279–287.

167. Kiviranta I, Jurvelin J, Tammi M, Saamanen AM & Helminen HJ (1985) Microspectropho-tometric quantitation of glycosaminoglycans in articular cartilage sections stained withsafranin o. Histochemistry 82(3): 249–255.

168. Kiraly K, Lammi M, Arokoski J, Lapvetelainen T, Tammi M, Helminen H & Kiviranta I(1996) Safranin o reduces loss of glycosaminoglycans from bovine articular cartilage duringhistological specimen preparation. The Histochemical journal 28(2): 99–107.

169. Rosenberg L (1971) Chemical basis for the histological use of safranin o in the study ofarticular cartilage. J.Bone Joint Surg.Am. 53(1): 69–82.

170. Camacho NP, West P, Torzilli PA & Mendelsohn R (2001) Ftir microscopic imaging ofcollagen and proteoglycan in bovine cartilage. Biopolymers 62(1): 1–8.

171. Rieppo L, Saarakkala S, Narhi T, Helminen HJ, Jurvelin JS & Rieppo J (2012) Applicationof second derivative spectroscopy for increasing molecular specificity of fourier transforminfrared spectroscopic imaging of articular cartilage. Osteoarthritis Cartilage 20(5): 451–459.

172. Mankin HJ, Dorfman H, Lippiello L & Zarins A (1971) Biochemical and metabolicabnormalities in articular cartilage from osteo-arthritic human hips. ii. correlation of

94

morphology with biochemical and metabolic data. J Bone Joint Surg Am 53(3): 523–537.173. Pritzker KP, Gay S, Jimenez SA, Ostergaard K, Pelletier JP, Revell PA, Salter D & van den

Berg WB (2006) Osteoarthritis cartilage histopathology: grading and staging. OsteoarthritisCartilage 14(1): 13–29.

174. Saarakkala S, Julkunen P, Kiviranta P, Makitalo J, Jurvelin JS & Korhonen RK (2010)Depth-wise progression of osteoarthritis in human articular cartilage: investigation ofcomposition, structure and biomechanics. Osteoarthritis Cartilage 18(1): 73–81.

175. Isaksson H, Toyras J, Hakulinen M, Aula AS, Tamminen I, Julkunen P, Kroger H & JurvelinJS (2011) Structural parameters of normal and osteoporotic human trabecular bone areaffected differently by microct image resolution .

176. Lin PC, Reiter DA & Spencer RG (2009) Classification of degraded cartilage throughmultiparametric mri analysis. J Magn Reson 201(1): 61–71.

177. Lukas VA, Fishbein KW, Reiter DA, Lin PC, Schneider E & Spencer RG (2014) Sensitivityand specificity of univariate mri analysis of experimentally degraded cartilage under clinicalimaging conditions. J Magn Reson Imaging .

178. Yoshioka M, Coutts RD, Amiel D & Hacker SA (1996) Characterization of a model ofosteoarthritis in the rabbit knee. Osteoarthritis Cartilage 4(2): 87–98.

179. Batiste DL, Kirkley A, Laverty S, Thain LM, Spouge AR & Holdsworth DW (2004) Exvivo characterization of articular cartilage and bone lesions in a rabbit acl transection modelof osteoarthritis using mri and micro-ct. Osteoarthritis Cartilage 12(12): 986–996.

180. Xia Y, Moody JB, Burton-Wurster N & Lust G (2001) Quantitative in situ correlationbetween microscopic mri and polarized light microscopy studies of articular cartilage.Osteoarthritis Cartilage 9(5): 393–406.

181. Mosher TJ, Dardzinski BJ & Smith MB (2000) Human articular cartilage: influence ofaging and early symptomatic degeneration on the spatial variation of t2–preliminary findingsat 3 t. Radiology 214(1): 259–266.

182. Stubendorff JJ, Lammentausta E, Struglics A, Lindberg L, Heinegard D & Dahlberg LE(2012) Is cartilage sgag content related to early changes in cartilage disease? implicationsfor interpretation of dgemric. Osteoarthritis Cartilage 20(5): 396–404.

183. Lattanzio PJ, Marshall KW, Damyanovich AZ & Peemoeller H (2000) Macromolecule andwater magnetization exchange modeling in articular cartilage. Magn Reson Med 44(6):840–851.

184. Stikov N, Keenan KE, Pauly JM, Smith RL, Dougherty RF & Gold GE (2011) Cross-relaxation imaging of human articular cartilage. Magn Reson Med 66(3): 725–734.

185. Stahl R, Luke A, Li X, Carballido-Gamio J, Ma CB, Majumdar S & Link TM (2009) T1rho,t2 and focal knee cartilage abnormalities in physically active and sedentary healthy subjectsversus early oa patients–a 3.0-tesla mri study. Eur Radiol 19(1): 132–143.

186. Makela JT, Rezaeian ZS, Mikkonen S, Madden R, Han SK, Jurvelin JS, Herzog W &Korhonen RK (2014) Site-dependent changes in structure and function of lapine articularcartilage 4 weeks after anterior cruciate ligament transection. Osteoarthritis Cartilage .

187. Witschey WR, Borthakur A, Elliott MA, Mellon E, Niyogi S, Wallman DJ, Wang C &Reddy R (2007) Artifacts in t1 rho-weighted imaging: compensation for b(1) and b(0) fieldimperfections. J Magn Reson 186(1): 75–85.

188. Hunziker EB (2002) Articular cartilage repair: basic science and clinical progress. a reviewof the current status and prospects. Osteoarthritis Cartilage 10(6): 432–463.

95

189. Hunziker EB & Rosenberg LC (1996) Repair of partial-thickness defects in articularcartilage: cell recruitment from the synovial membrane. J Bone Joint Surg Am 78(5):721–733.

190. McIlwraith CW, Fortier LA, Frisbie DD & Nixon AJ (2011) Equine models of articularcartilage repair. Cartilage 2(4): 317–326.

191. Desjardins MR & Hurtig MB (1990) Cartilage healing: A review with emphasis on theequine model. Can Vet J 31(8): 565–572.

96

Original publications

I Rautiainen J, Nissi MJ, Liimatainen T, Herzog W, Korhonen RK & Nieminen MT. (2014)Adiabatic Rotating Frame Relaxation of MRI Reveals Early Cartilage Degeneration in aRabbit Model of Anterior Cruciate Ligament Transection. Osteoarthritis Cartilage 22(10):1444–1452.

II Rautiainen J, Nissi MJ, Salo EN, Tiitu V, Finnila MA, Aho OM, Saarakkala S, Lehenkari P,Ellermann J & Nieminen MT. (2014) Multiparametric MRI assessment of human articularcartilage degeneration: Correlation with quantitative histology and mechanical properties.Magn Reson Med. In press.

III Rautiainen J, Lehto LJ, Tiitu V, Kiekara O, Pulkkinen H, Brunott A, van Weeren R, BrommerH, Brama PA, Ellermann J, Kiviranta I, Nieminen MT & Nissi MJ. (2013) Osteochondralrepair: evaluation with sweep imaging with fourier transform in an equine model. Radiology269(1): 113–121.

Reprinted with permission from Elsevier (I), John Wiley & Sons, Inc. (II) and RSNA(III).

Original publications are not included in the electronic version of the dissertation.

97

98

A C T A U N I V E R S I T A T I S O U L U E N S I S

Book orders:Granum: Virtual book storehttp://granum.uta.fi/granum/

S E R I E S D M E D I C A

1270. Silvola, Anna-Sofia (2014) Effect of treatment of severe malocclusion and relatedfactors on oral health-related quality of life

1271. Mäkelä, Tuomas (2014) Systemic transplantation of bone marrow stromal cells :an experimental animal study of biodistribution and tissue targeting

1272. Junttila, Sanna (2014) Studies of kidney induction in vitro using gene expressionprofiling and novel tissue manipulation technique

1273. Häkli, Sanna (2014) Childhood hearing impairment in northern Finland :prevalence, aetiology and additional disabilities

1274. Lehto, Liisa (2014) Interactive two-step training and management strategy forimprovement of the quality of point-of-care testing by nurses : implementation ofthe strategy in blood glucose measurement

1275. Harjunen, Vanessa (2014) Skin stem cells and tumor growth : functions ofcollagen XVIII in hair follicle cycling and skin cancer, and Bmx tyrosine kinase intumor angiogenesis

1276. Savela, Salla (2014) Physical activity in midlife and health-related quality of life,frailty, telomere length and mortality in old age

1277. Laitala, Anu (2014) Hypoxia-inducible factor prolyl 4-hydroxylases regulatingerythropoiesis, and hypoxia-inducible lysyl oxidase regulating skeletal muscledevelopment during embryogenesis

1278. Persson, Maria (2014) 3D woven scaffolds for bone tissue engineering

1279. Kallankari, Hanna (2014) Perinatal factors as predictors of brain damage andneurodevelopmental outcome : study of children born very preterm

1280. Pietilä, Ilkka (2015) The role of Dkk1 and Wnt5a in mammalian kidneydevelopment and disease

1281. Komulainen, Tuomas (2015) Disturbances in mitochondrial DNA maintenance inneuromuscular disorders and valproate-induced liver toxicity

1282. Nguyen, Van Dat (2015) Mechanisms and applications of disulfide bond formation

1283. Orajärvi, Marko (2015) Effect of estrogen and dietary loading on rat condylarcartilage

1284. Bujtár, Péter (2015) Biomechanical investigation of the mandible, a related donorsite and reconstructions for optimal load-bearing

1285. Ylimäki, Eeva-Leena (2015) Ohjausintervention vaikuttavuus elintapoihin jaelintapamuutokseen sitoutumiseen

UNIVERSITY OF OULU P .O. Box 8000 F I -90014 UNIVERSITY OF OULU FINLAND

A C T A U N I V E R S I T A T I S O U L U E N S I S

Professor Esa Hohtola

University Lecturer Santeri Palviainen

Postdoctoral research fellow Sanna Taskila

Professor Olli Vuolteenaho

University Lecturer Veli-Matti Ulvinen

Director Sinikka Eskelinen

Professor Jari Juga

University Lecturer Anu Soikkeli

Professor Olli Vuolteenaho

Publications Editor Kirsti Nurkkala

ISBN 978-952-62-0763-6 (Paperback)ISBN 978-952-62-0764-3 (PDF)ISSN 0355-3221 (Print)ISSN 1796-2234 (Online)

U N I V E R S I TAT I S O U L U E N S I S

MEDICA

ACTAD

D 1286

ACTA

Jari Rautiainen

OULU 2015

D 1286

Jari Rautiainen

NOVEL MAGNETIC RESONANCE IMAGING TECHNIQUES FOR ARTICULAR CARTILAGE AND SUBCHONDRAL BONESTUDIES ON MRI RELAXOMETRY AND SHORT ECHO TIME IMAGING

UNIVERSITY OF OULU GRADUATE SCHOOL;UNIVERSITY OF OULU, FACULTY OF MEDICINE;MEDICAL RESEARCH CENTER OULU