UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

In vivo assessments of patellofemoral kinematics and contact areas with applications to osteoarthritis McWalter, Emily Jane 2010

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

Item Metadata

Download

Media
24-ubc_2011_spring_mcwalter_emily.pdf [ 4.26MB ]
Metadata
JSON: 24-1.0071575.json
JSON-LD: 24-1.0071575-ld.json
RDF/XML (Pretty): 24-1.0071575-rdf.xml
RDF/JSON: 24-1.0071575-rdf.json
Turtle: 24-1.0071575-turtle.txt
N-Triples: 24-1.0071575-rdf-ntriples.txt
Original Record: 24-1.0071575-source.json
Full Text
24-1.0071575-fulltext.txt
Citation
24-1.0071575.ris

Full Text

 IN VIVO ASSESSMENTS OF PATELLOFEMORAL KINEMATICS AND CONTACT AREAS WITH APPLICATIONS TO OSTEOARTHRITIS  by  Emily Jane McWalter  BSc Mech Eng, Queen’s University, 2002 MASc, The University of British Columbia, 2004     A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY  in  The Faculty of Graduate Studies (Mechanical Engineering)     THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) December 2010  © Emily Jane McWalter, 2010 ABSTRACT Mechanical-based treatment strategies for patellofemoral osteoarthritis (OA) have had limited success.  This is likely because the magnitude of mechanical change required to improve clinical symptoms has not been quantified because, until recently, the tools required to do so were not available.  The aim of this thesis was to develop and characterize MRI-based assessments of in vivo joint mechanics (three-dimensional patellar kinematics and contact areas) that can be used in studies of patellofemoral OA.  Three studies of three-dimensional patellar kinematics were carried out.  Study 1 examined the effect of load on kinematic measurements.  The results showed that increased load caused patellae to flex, tilt medially and translate proximally and posteriorly (p<0.05).  In Study 2, the need to carry out a full kinematic assessment through the range of knee flexion was assessed.  The results showed that a single, widely used measure of patellar position and orientation was an inadequate surrogate marker and a full kinematic assessment was required.  In Study 3, the effect of a patellofemoral brace on kinematics was examined in patients with patellofemoral OA.  The brace caused the patellae to extend, spin externally, tilt medially and translate distally, medially and posteriorly (p<0.05).  In Study 4, a method of assessing contact areas using an MRI scan of less than a minute was developed and yielded measurements with errors similar to methods employing significantly longer scans.  This method can also be used in series with the kinematic method, allowing kinematics and contact areas to be assessed simultaneously.  In Study 5, a simple, patient specific, kinematics-driven multibody model to predict contact area was developed and validated.  The model showed good agreement with direct measures of contact area from MRI but was sensitive to the proximity threshold value used to define contact and to the kinematic input data.  This model may be useful in studies where direct measures of contact area are not possible.  The MRI-based tools for assessing patellofemoral joint mechanics in vivo characterized and validated in this thesis can potentially be used to identify the magnitude of mechanical change required to improve symptoms in patients with patellofemoral OA.  ii PREFACE A version of Chapter 2 has been published. McWalter, E.J., MacIntyre, N.J., Cibere, J., Wilson, D.R. (2010) A single measure of patellar kinematics is an inadequate surrogate marker for patterns of three-dimensional kinematics in healthy knees. The Knee 17(135-140).  I was responsible for designing the study, collecting a portion of the data, analyzing a portion of the data, carrying out the statistical analysis and writing the manuscript.  Copyright permission to include this work in the thesis was obtained from The Knee.  Ethical approval was provided by the University of British Columbia Human Ethics Board under certificate C05-0166 (Appendix B) for a portion of the study volunteers, the remaining volunteers fell under ethical approval from the collaborating institution, Queen’s University. A version of Chapter 3 has been published. McWalter, E.J., Hunter, D.J. and Wilson, D.J. (2010) The effect of load magnitude on three-dimensional patellar kinematics in vivo. Journal of Biomechanics 43:1890-1897.  I was responsible for designing the study, collecting the data, analyzing the data, carrying out the statistical analysis and writing the manuscript.  Copyright permission to include this work in the thesis was obtained from the Journal of Biomechanics.  Ethical approval was provided by the University of British Columbia Human Ethics Board under certificate C05-0166 (Appendix B). A version of Chapter 4 has been submitted.  McWalter, E.J., Hunter, D.J., Harvey, W.F., McCree, P., Hirko K.A., Felson, D.T., Wilson, D.R. (submitted July 2010) The effect of a patellar brace on three-dimensional patellar kinematics in patients with lateral patellofemoral osteoarthritis. Osteoarthritis and Cartilage.  Submission number OAC 5824.  I was responsible for helping design the study, training the research assistant to collect the data, analyzing the data, carrying out the statistical analysis and writing the manuscript.  Ethical approval was not required from the University of British Columbia but was obtained at the collaborating institution, Boston University School of Medicine. A version of Chapter 5 is being prepared for submission.  McWalter, E.J., O’Kane, C., Wilson, D.R., A rapid, validated MRI scan for assessing patellar cartilage contact areas in vivo.  I was responsible for designing the study, collecting the data, writing the software, analyzing the data, carrying out the statistical analysis and writing the manuscript.  Ethical approval was provided by the University of British Columbia Human Ethics Board under certificates C05-0140 and C05-0166 (Appendix B).  iii TABLE OF CONTENTS ABSTRACT ................................................................................................................................... ii PREFACE ..................................................................................................................................... iii TABLE OF CONTENTS............................................................................................................. iv LIST OF TABLES....................................................................................................................... vii LIST OF FIGURES...................................................................................................................... ix LIST OF ABBREVIATIONS..................................................................................................... xii ACKNOWLEDGEMENTS....................................................................................................... xiii 1 Introduction............................................................................................................................. 1 1.1 Overview......................................................................................................................... 1 1.2 The Patellofemoral Joint ................................................................................................. 2 1.2.1 Reference Planes and Directions........................................................................ 3 1.2.2 Anatomy............................................................................................................. 4 1.2.2.1 Bones and Joints.................................................................................. 4 1.2.2.2 Articular Cartilage............................................................................... 5 1.2.2.3 Muscles ............................................................................................... 6 1.2.2.4 Tendons and Ligaments ...................................................................... 7 1.2.2.5 Other Passive Soft Tissues .................................................................. 7 1.2.3 Function ............................................................................................................. 8 1.2.3.1 Forces Acting on the Patella ............................................................... 8 1.2.3.2 Response of Cartilage to Load .......................................................... 10 1.3 Patellofemoral Osteoarthritis ........................................................................................ 11 1.3.1 Epidemiology ................................................................................................... 11 1.3.2 Clinical Definitions .......................................................................................... 12 1.3.3 Mechanical Factors .......................................................................................... 13 1.3.4 Mechanical-based Treatment Strategies .......................................................... 16 1.3.5 Summary .......................................................................................................... 18 1.4 Three-dimensional Patellar Kinematics ........................................................................ 18 1.4.1 Clinical Motions............................................................................................... 19 1.4.2 Coordinate Systems and Calculation of Kinematics ........................................ 21 1.4.3 Measurement Systems...................................................................................... 25 1.4.4 Loading ............................................................................................................ 27 1.4.4.1 Ex Vivo ............................................................................................. 27 1.4.4.2 In Vivo .............................................................................................. 29 1.4.5 Range of Motion .............................................................................................. 30 1.4.6 In Vivo Validation ............................................................................................ 30 1.4.7 Ex Vivo and In Vivo Study Results................................................................... 33 1.4.8 Summary .......................................................................................................... 36 1.5 Patellofemoral Joint Contact Areas............................................................................... 36 1.5.1 Measurement Systems...................................................................................... 37 1.5.1.1 Ex Vivo ............................................................................................. 37  iv 1.5.1.2 In Vivo .............................................................................................. 41 1.5.2 Experimental Considerations ........................................................................... 46 1.5.3 Ex Vivo and In Vivo Study Findings ................................................................ 47 1.5.4 Summary .......................................................................................................... 50 1.6 Patellofemoral Joint Models ......................................................................................... 50 1.6.1 Platform............................................................................................................ 52 1.6.2 Inputs................................................................................................................ 52 1.6.3 Validation......................................................................................................... 53 1.6.4 Study Findings Using Models.......................................................................... 54 1.6.5 Summary .......................................................................................................... 55 1.7 Thesis Objectives and Research Questions................................................................... 55 2 Surrogate Markers of Kinematics ....................................................................................... 59 2.1 Introduction................................................................................................................... 59 2.2 Methods......................................................................................................................... 61 2.2.1 Subjects ............................................................................................................ 61 2.2.2 Image Acquisition ............................................................................................ 61 2.2.3 Three-dimensional Patellar Kinematic Analysis.............................................. 62 2.2.4 Statistical Analysis........................................................................................... 66 2.3 Results........................................................................................................................... 66 2.4 Discussion ..................................................................................................................... 70 3 Loading & Kinematics.......................................................................................................... 73 3.1 Introduction................................................................................................................... 73 3.2 Methods......................................................................................................................... 74 3.2.1 Subjects ............................................................................................................ 74 3.2.2 Image Acquisition and Kinematic Analysis..................................................... 74 3.2.3 Statistical Analysis........................................................................................... 74 3.3 Results........................................................................................................................... 75 3.4 Discussion ..................................................................................................................... 80 4 Patellofemoral Bracing & Kinematics in Osteoarthritis ................................................... 84 4.1 Introduction................................................................................................................... 84 4.2 Methods......................................................................................................................... 85 4.2.1 Subjects ............................................................................................................ 85 4.2.2 Patellar Brace ................................................................................................... 86 4.2.3 Three-Dimensional Patellar Kinematic Analysis............................................. 86 4.2.4 Statistical Analysis........................................................................................... 87 4.3 Results........................................................................................................................... 87 4.4 Discussion ..................................................................................................................... 93 5 MRI-based Assessment of Contact Area: Development & Validation ............................ 97 5.1 Introduction................................................................................................................... 97 5.2 Methods......................................................................................................................... 99 5.2.1 Specimens Preparation ..................................................................................... 99 5.2.2 Experimental Procedure................................................................................. 100 5.2.3 Contact Area Analysis ................................................................................... 102 5.2.4 Repeatability .................................................................................................. 105 5.3 Results......................................................................................................................... 106 5.4 Discussion ................................................................................................................... 112  v 6 Model Development & Validation..................................................................................... 118 6.1 Introduction................................................................................................................. 118 6.2 Methods....................................................................................................................... 119 6.2.1 Model Development....................................................................................... 119 6.2.2 Model Validation ........................................................................................... 123 6.3 Results......................................................................................................................... 123 6.3.1 Proximity Threshold Sensitivity .................................................................... 126 6.3.2 Kinematic Input Sensitivity ........................................................................... 129 6.4 Discussion ................................................................................................................... 134 7 Integrated Discussion.......................................................................................................... 138 7.1 Summary of Findings.................................................................................................. 138 7.2 General Discussion ..................................................................................................... 139 7.2.1 Why are In Vivo Assessments of Patellofemoral Joint Mechanics Important? ...................................................................................................... 139 7.2.2 How do the In Vivo Assessments of Patellofemoral Joint Mechanics Used in this Thesis Compare to Others in the Literature? ...................................... 142 7.2.2.1 Three-dimensional Patellar Kinematics .......................................... 142 7.2.2.2 Contact Area ................................................................................... 144 7.2.3 Do Statistically Significant Differences Relate to Clinically Important Changes in Three-dimensional Patellar Kinematics? .................................... 145 7.2.4 Are Patellofemoral Joint Models Useful? ...................................................... 147 7.2.5 Why should more Sophisticated Measures of Joint Mechanics be used in Studies of Patellofemoral OA? ...................................................................... 148 7.3 Strengths and Limitations ........................................................................................... 149 7.4 Contributions............................................................................................................... 151 7.5 Future Work ................................................................................................................ 152 7.6 Conclusion .................................................................................................................. 154 References .................................................................................................................................. 155 Appendix A: Glossary of Terms................................................................................................ 174 Appendix B: Ethics Certificates ................................................................................................ 177 Appendix C: Loading Rig .......................................................................................................... 179 Appendix D: Details of Kinematic Analysis............................................................................. 185 Appendix E: Linear Hierarchical Random-effects Models .................................................... 190 Appendix F: Raw Data from Loading Study........................................................................... 192 Appendix G: Additional Information on Parent Bracing Study............................................ 199 Appendix H: Dye Leaching Test ............................................................................................... 201 Appendix I: Sensitivity of Centroids to Proximity Threshold................................................ 202 Appendix J: Sensitivity of Centroid to Kinematics Input ...................................................... 208  vi LIST OF TABLES Table 1-1: Landmarks used to create anatomical coordinate systems in the patella and femur........... 22 Table 1-2: Division of force applied to quadriceps muscles based on PCSA. ..................................... 28 Table 1-3: Errors in three-dimensional patellar kinematic assessment. ............................................... 32 Table 1-4: Advantages and disadvantages of different ex vivo measures of contact area. ................... 38 Table 1-5 Comparison of MRI-based methods of measuring patellofemoral contact area. ................. 43 Table 1-6: Summary of computational models in the literature. .......................................................... 51 Table 2-1: T1-weighted MRI sequence parameters for the high- and low-resolution scans. ............... 62 Table 3-1: Coefficients (standard error) for each kinematic parameter’s linear hierarchical random-effects model.............................................................................................................. 77 Table 3-2: Comparison between the results for 15% BW load in the present study and two other studies of three-dimensional patellar kinematics using sequential static methods [104,134]. ................................................................................................................................ 81 Table 4-1: Coefficients (confidence interval) for the hierarchical random-effects models for rotations. .................................................................................................................................. 89 Table 4-2: Coefficients (confidence interval) for the hierarchical random-effects models for translations. ............................................................................................................................. 90 Table 5-1: T1-weighted MRI sequence parameters. .......................................................................... 101 Table 5-2: Error of the MRI-based assessment of contact area as compared to the reference standard. ................................................................................................................................ 107 Table 5-3: Mean ± standard deviation difference in distances (mm) between MRI and reference standard centroids (3 dye trials) and percentage overlap of MRI- and reference standard areas. ....................................................................................................... 108 Table 5-4: Intra-reader error expressed as the mean standard deviation and the mean standard deviation ± standard deviation (mm2). .................................................................................. 109 Table 6-1: Errors of model estimates as compared to the delineated sagittal MRI-based measures of contact area........................................................................................................ 124  vii Table 6-2: Errors of model estimates as compared to the dye-based measures of contact area. ........ 124 Table 6-3: Mean difference in centroid position as compared to the sagittal delineated MRI- based reference standard. ...................................................................................................... 125 Table 6-4: Mean difference in centroid position as compared to the dye-based reference standard. ................................................................................................................................ 125   viii LIST OF FIGURES Figure 1-1 : The knee. ............................................................................................................................ 3 Figure 1-2: Anatomical planes. .............................................................................................................. 4 Figure 1-3: Posterior surface of the patella (left) and trochlear groove (right) of the femur.................. 5 Figure 1-4: Cross-section of articular cartilage structure. ...................................................................... 5 Figure 1-5: Quadriceps muscles. ............................................................................................................ 6 Figure 1-6: The retinaculum and the iliotibial band. .............................................................................. 8 Figure 1-7: Extensor mechanism with and without a patella. ................................................................ 9 Figure 1-8: Creep and relaxation behaviour of articular cartilage........................................................ 11 Figure 1-9: Varus and valgus tibiofemoral alignment (left) and patellar malalignment (right). .......... 14 Figure 1-10:  Two-dimensional assessments of patellar alignment...................................................... 15 Figure 1-11: Between study variation in patellar tilt (in vivo) and rotation (in vivo and ex vivo). ........................................................................................................................................ 19 Figure 1-12: Patellar motions. .............................................................................................................. 20 Figure 1-13: Anatomical coordinate systems of femur and patella. ..................................................... 23 Figure 1-14: Joint Coordinate System modified for the patellofemoral joint. ..................................... 24 Figure 1-15: In vivo three-dimensional patellar kinematics from MRI in normal individuals............. 35 Figure 1-16: Contact area in the bovine tibiofemoral joint. ................................................................. 41 Figure 1-17: Axial and sagittal MRI slice directions for patellofemoral contact area assessment. .............................................................................................................................. 42 Figure 1-18: Effect of time on contact area measurement of a cadaveric femoral condyle against glass............................................................................................................................. 47 Figure 1-19: Summary of patellofemoral contact areas reported in the literature for ex vivo (above) and in vivo (below) studies......................................................................................... 48 Figure 1-20: Patellar contact area patterns ex vivo. .............................................................................. 49 Figure 1-21: Contact area patterns in vivo............................................................................................ 49  ix Figure 1-22: MRI vs contact areas estimated by model. ...................................................................... 54 Figure 1-23: Overview of thesis studies. .............................................................................................. 58 Figure 2-1: Subject positioned in MRI scanner for a low-resolution, loaded MRI scans. ................... 62 Figure 2-2: Anatomical landmarks identified on MRI images............................................................. 64 Figure 2-3: Illustration of the modified Joint Coordinate System........................................................ 65 Figure 2-4: Three-dimensional patellar kinematic parameters............................................................. 65 Figure 2-5: Regression line fits describing the relationship between the surrogate marker and pattern of rotation (flexion, spin and tilt). ............................................................................... 68 Figure 2-6: Regression line fits describing the relationship between the surrogate marker and pattern of translation (proximal, lateral and anterior).............................................................. 69 Figure 3-1: Hierarchical random-effects model (REM) and raw data for rotations. ............................ 78 Figure 3-2: Hierarchical random-effects model (REM) and raw data for translations......................... 79 Figure 4-1: The patellofemoral brace (Bio Skin Q, Cropper Medical Inc., Ashland, OR, USA) evaluated in this study. ............................................................................................................ 86 Figure 4-2: Rotational results as a function of knee flexion for patellar a) flexion, b) spin and c) tilt. ....................................................................................................................................... 91 Figure 4-3: Translation results as a function of knee flexion for a) proximal, b) lateral and c) anterior translation................................................................................................................... 92 Figure 5-1: Schematic of cadaver knee specimen in MRI-safe loading apparatus positioned in the MRI scanner. ................................................................................................................... 102 Figure 5-2: Example of axial (top) and sagittal (bottom) scans delineated (left) and segmented (right)..................................................................................................................................... 103 Figure 5-3: Scale of areas to orient reader to magnitude of errors. .................................................... 107 Figure 5-4: Centroids of the MRI and three trials of the reference standard contact areas. ............... 108 Figure 5-5: Percentage overlap between MRI-based and reference standard-based assessment of contact area. ...................................................................................................................... 109 Figure 5-6: Contact area intra-subject repeatability results presented for three trials for two subjects. ................................................................................................................................. 110  x Figure 5-7: Contact area contours and centroids for three trials in two subjects................................ 111 Figure 6-1: Femoral cartilage original contour data (yellow) and mesh (blue).................................. 120 Figure 6-2: Bone contours (blue and red), femoral cartilage mesh (turquoise) and patellar cartilage contours (magenta) in loaded kinematic position. .................................................. 121 Figure 6-3: Proximity analysis for one slice....................................................................................... 122 Figure 6-4: Femoral cartilage mesh (blue), patellar cartilage contours (yellow) and periphery of contact contour (magenta)................................................................................................. 122 Figure 6-5: Model (yellow) and the MRI-based (red) and mean dye-based (blue) reference standard of contact region in the four cadaver specimens..................................................... 126 Figure 6-6: Proximity threshold sensitivity for Specimen 1............................................................... 127 Figure 6-7: Proximity threshold sensitivity for Specimen 2............................................................... 127 Figure 6-8: Proximity threshold sensitivity for Specimen 3............................................................... 128 Figure 6-9: Proximity threshold sensitivity for Specimen 4............................................................... 128 Figure 6-10: Sensitivity of patellar cartilage contact area to patellar flexion..................................... 130 Figure 6-11: Sensitivity of patellar cartilage contact area to proximal/distal translation................... 130 Figure 6-12: Sensitivity of patellar cartilage contact area to anterior/posterior translation. .............. 131 Figure 6-13: Sensitivity of patellar cartilage contact area to patellar tilt. .......................................... 131 Figure 6-14: Sensitivity of patellar cartilage contact area to patellar medial/lateral translation. ....... 132 Figure 6-15: Sensitivity of patellar cartilage contact area to patellar spin. ........................................ 132 Figure 6-16: Contact contours and centroids for Specimen 2 for changes in each of the 6 kinematic parameters............................................................................................................. 133 Figure 6-17: Sensitivity of proximal/distal centroid position to proximal translation. ...................... 134 Figure 7-1: Patellar tilt as a function of tibiofemoral flexion............................................................. 141     xi LIST OF ABBREVIATIONS ANOVA Analysis of Variance BMI Body Mass Index BW Bodyweight CI Confidence Interval CT Commuted Tomography CV% Percentage Coefficient of Variation EMG Electromyography ICP Iterative Closest Points ISB International Society of Biomechanics MRI Magnetic Resonance Imaging NURBS Non-uniform, rational, basis spline OA Osteoarthritis OR Odds Ratio PCSA Physiologic Cross-sectional Area REM Random-effects Model RSA Roentgen Stereophotogrammetric Analysis   xii ACKNOWLEDGEMENTS First I would like to thank my supervisor, Dr. David Wilson, for his support and encouragement over the years and for the long hours he spent editing abstracts, manuscripts, scholarship applications and, of course, this thesis.  I appreciate all that he has taught me about research and academia, in particular, the importance of the ‘big picture’.  I know I leave here a better writer, a better thinker and as a result, a better researcher.  I would also like to thank my committee members Dr. Tom Oxland, Dr. Jolanda Cibere and Dr. Elizabeth Croft.  I feel privileged to have had the opportunity to learn from Dr. Oxland, over my years at UBC; on countless occasions our discussions and his advice have been invaluable. I thank Dr. Cibere for always challenging me to think about my work from a clinical point of view and for the meticulous job she has always done reviewing my work.  Finally, I thank Dr. Croft for her enthusiasm and her unique point of view on my work; I thoroughly enjoyed the challenge of her class and having her as a committee member. I have been fortunate to be a part of a lab of bright, intelligent and enthusiastic young researchers without whom I would not have found coming to work everyday quite so appealing!  In particular I would like to thank Claire Jones for her technical advice, unwavering support and shoulder to cry on, JD Johnston for sharing his Matlab code, providing exceptional technical advice and entertaining Skype conversations and Lindsay Nettlefold for our endless discussions about what should go in a literature review and our weekly PhD club ‘self-help’ meetings.  I would also like to thank Heather Macdonald, Carolyne Albert and Angela Kedgley for showing me that it is possible to finish a PhD; you ladies are inspiring!  Finally, to all the other lab members, who are too numerous to name, thank you for making this lab a great place to work. I would like to thank the team at the UBC MRI Research Centre for their technical support, in particular Paul Hamill, Trudy Harris, Linda James and Burkhart Maedler, and also the team at Boston University Center for Biomedical Imaging, in particular Kelley Erb and Kevin Hallock.  I would like to thank the patellofemoral bracing study team at the New England Baptist Hospital in Boston, in particular Dr. David Hunter, Kelly Hirko, Dr. William Harvey, Paula Cree and Dr. David Felson.  I would also like to thank Dr. Penny Brasher for statistical advice.  I would like to acknowledge Colm O’Kane for his insight on the work in Chapter 5, as well as Laura Given and Katharine Wilson for helping in data collection. I would like to thank the study participants who volunteered their time and the anonymous study participants who generously donated their bodies to science; without all of you these studies would not be possible.  xiii  xiv I would also like to acknowledge funding from the Natural Sciences and Engineering Research Council, the Canadian Institute for Health Research, the Canadian Arthritis Network, the Arthritis Society and the Michael Smith Foundation for Health Research To my family, I would like to thank you for supporting me through 25 years of school; I would not have been able to make it this far without you!  Thanks Mum for our weekly chats that always cheered me up and your unwavering encouragement and support, thanks Dad for showing me the value of hard work and commitment and thanks Richard for inspiring me to take chances (even though I seem do it in a different way than you do). And finally, I would like to thank Steve for always believing in me even when I didn’t, for making me laugh every day and for taking me on a date to Temaki every Friday night!  Your love and support has meant more than you know.    1 Introduction 1.1 Overview The patellofemoral joint is a crucial but insufficiently studied component of the knee.  Its importance is often overshadowed by the tibiofemoral joint, whose role in facilitating knee flexion and extension during daily activities such as standing, bending and walking is easily understood. The patellofemoral joint plays an important mechanical role in the extensor mechanism, a system comprised of bones and soft tissues that drives knee extension.  Specifically, the patella increases the moment arm of the extensor mechanism thereby decreasing the amount of quadriceps force required for knee extension.  Although we know the patellofemoral joint is mechanically important to the overall function of the knee, relatively little is know about patellofemoral joint mechanics.  One reason for this is that the patella is technically difficult to study due to its small size and its ability to move in three dimensions.  There are several diseases that affect the patellofemoral joint which are believed to be mechanical in origin, such as osteoarthritis (OA).  It is therefore essential that we have a fundamental understanding of normal patellofemoral joint mechanics in order to identify and correct the aberrant patterns that lead to joint disease. Patellofemoral joint mechanics have been studied both ex vivo and in vivo.  While ex vivo models provide a useful platform for measuring kinematics, forces, contact pressures and contact areas, it is unclear whether these studies adequately represent in vivo joint mechanics.  Therefore, several groups have developed methods using imaging technologies, such as bi-planar radiography, computed tomography (CT) and magnetic resonance imaging (MRI), to assess patellofemoral joint mechanics in vivo.  MRI is particularly well suited for studying joint mechanics because study participants are not exposed to ionizing radiation and soft tissues, such as articular cartilage, are visible; this is not the case with the radiography-based methods.  Currently, methods exist to assess three-dimensional patellofemoral joint kinematics and articular cartilage surface contact areas in vivo with MRI; however, many of these methods have not been rigorously validated.  Further method development, validation and characterization is required to describe normal and aberrant patterns of patellofemoral kinematics and contact areas fully with MRI. Patellofemoral OA is a common disease whose risk factors and treatment strategies are primarily mechanical in nature.  Obesity [1], varus/valgus alignment (bow-legs/knock-knees) [2,3] and planar patellar alignment [2-5] have all been associated with patellofemoral OA.  Although these factors suggest that mechanics play a role in disease onset and progression, they are surrogate  1 measures of mechanics. We do not know how patellofemoral loads change with pathology or how they are affected by treatment.  Treatment strategies for patellofemoral OA also focus on correcting mechanics; however, often they are not successful in relieving symptoms or arresting disease progression.  One reason for this is that we don’t know what magnitude of mechanical correction is required to restore normal joint mechanics or improve patient symptoms.  This is likely because until recently the tools required to measure patellofemoral joint mechanics directly in vivo have not been available.  Therefore, in vivo imaging tools are essential for gaining an understanding of local patellofemoral joint mechanics in individuals with OA and for developing and evaluating treatment strategies that aim to correct mechanics. The aim of this chapter is to provide an overview of patellofemoral joint mechanics and the mechanical basis of OA risk factors and treatments.  First, a brief background of anatomical terminology and patellofemoral joint anatomy will be provided and the current understanding of mechanical risk factors and treatment strategies for patellofemoral OA will be outlined.  Next, a critical review of methods of assessing ex vivo and in vivo kinematics and contact areas, as well as of computational models of the patellofemoral joint, will be provided.  Finally, the chapter concludes with defining the objectives and scope of this thesis. 1.2 The Patellofemoral Joint The primary functions of a joint are to transmit load and allow movement between body segments.  The knee is a complex joint composed of two articulations, the tibiofemoral joint (between the femur and the tibia) and the patellofemoral joint (between the femur and the patella) (Figure 1-1). This section describes patellofemoral joint anatomy and function, using anatomically defined reference systems and terminology (for Glossary of Terms see Appendix A).  2  Figure 1-1 : The knee. The articulation between the femur and the patella form the patellofemoral joint.  Reprinted from Hughston Health Alert, Volume 10, Number 1 – Winter 1998 with permission from The Hughston Foundation.   1.2.1 Reference Planes and Directions Anatomically defined reference planes (Figure 1-2) are used to describe joint anatomy and movements in a clinically relevant manner.  This terminology will be used throughout the thesis and is of particular importance when defining MRI imaging planes.  Three reference planes are used: the transverse (or axial), the frontal (or coronal), and the sagittal.  The transverse plane divides the structure into proximal (towards the head) and distal (towards the feet) regions.  The frontal plane divides the structure into anterior (towards the front) and posterior (towards the back) regions. The sagittal plane divides the structure into left and right regions. In patellofemoral joint imaging, the term axial is most commonly used to describe the transverse plane, therefore this convention will be used throughout this thesis.  Structures are also identified using their position with respect to the sagittal midline; structures towards the midline are in a medial position while structures away from the midline are in a lateral position. There is also a convention to describe anatomical motions.  Flexion and extension are decreases and increases in the angle between two segments, respectively.  Abduction is a movement away from the midline, while adduction is a movement towards it.  Internal and external rotation is rotation about the long axis of the bone towards the midline and away from the midline, respectively.  3  Figure 1-2: Anatomical planes. Coronal (or frontal), sagittal and transverse (or axial) planes of the body. Image reprinted from Wikimedia Commons.  1.2.2 Anatomy The patellofemoral joint is comprised of bones, articular cartilage, muscles, tendons, ligaments and other soft tissues. 1.2.2.1 Bones and Joints The patellofemoral joint is the articulation of the posterior side of the patella with the anterior side of the distal femur.  The patella is a small, inverted tear-drop shaped bone with a complex shaped articular surface. The articular surface is divided into medial and lateral facets by a ridge that runs vertically along its length.  Most patellae also have an ‘odd facet’ which is found on the most medial edge of the patella, divided by a medial vertical ridge (Figure 1-3).  There is also a less prominent horizontal ridge which runs from the lateral edge to the central vertical ridge and then extends in a medio-distal direction. There is substantial variation in the proportions of the medial and lateral facets in normal patellae and, as such, three main types have been characterized [6].  Type 1 patellae have medial and lateral facets of approximately equal size, type 2 patellae have slightly larger lateral facets and type 3 patellae have markedly larger lateral facets.  The anterior region of the distal femur with which the patella articulates is often referred to as the trochlea.  The trochlea is a groove that is quite shallow at its proximal end but deepens distally.  When the patella articulates with the trochlea it begins in the shallow region when the leg is extended and then flexes into the deeper region with tibiofemoral flexion.  4 Lateral facetMedial facet Odd facet Apex Proximal Distal Figure 1-3: Posterior surface of the patella (left) and trochlear groove (right) of the femur. Anatomical features of the patella and femur with particular attention to articulating surfaces.  Graphics adapted from Gray’s Anatomy 1918 (copyright expired).  1.2.2.2 Articular Cartilage Articular cartilage provides a smooth, frictionless bearing surface for joint articulation. Articular cartilage is a viscoelastic material composed of water (68-85%) and proteins (collagen and proteoglycans).  The arrangement and density of the collagen fibres and proteoglycans vary over the depth of the cartilage, forming three distinct zones (Figure 1-4).  In the superficial zone (the articular surface) the collagen fibres are oriented tangentially to the surface, in the middle zone orientation is random and in the deep zone the fibres are perpendicular to the bone surface.  The articular cartilage is nourished and lubricated by synovial fluid sealed within a joint capsule (synovial membrane).  At the patellofemoral joint, cartilage covers the proximal three quarters of the patellar articular surface and the entire trochlea.  Patellar articular cartilage is amongst the thickest in the body, at up to 6 mm thick in the central region, while trochlear cartilage is approximately 3 mm thick [7].  Superficial Zone Middle Zone Deep Zone Articular Surface Bone Figure 1-4: Cross-section of articular cartilage structure. Collagen fibres are oriented tangentially in the superficial zone, randomly in the middle zone and radially in the deep zone.  5 1.2.2.3 Muscles Muscles are the actuators of biological joints.  The quadriceps muscles are the most important to the patellofemoral joint.  As the name implies, the quadriceps group is composed of four muscles: the rectus femoris, vastus intermedius, vastus lateralis and vastus medialis (Figure 1-5).  The rectus femoris originates on the pelvis, crosses the hip and inserts on the proximal margin of the patella. Since it is biarticular it facilitates both hip flexion and knee extension.  The vastus intermedius originates along the anterior surface of the femoral shaft and also inserts on the proximal margin of the patella.  The vastus intermedius lies underneath the rectus femoris.   The vastus lateralis originates on the greater trochanter of the femur and inserts on the proximal-lateral margin of the patella.  The vastus medialis originates the medial side of the femoral neck and inserts on the proximal-medial margin of the patella.  The vastus lateralis and medialis are often also divided into their longus and obliquus components because the orientation of the muscle fibres, termed pennation angle, changes from being in an oblique direction at their distal end to an axial direction at their proximal end.  The primary purpose of this muscle group is to extend the knee.   Figure 1-5: Quadriceps muscles. The vastus medialis, lateralis and intermedius and the rectus femoris form the quadriceps muscle group.  Image reprinted from: http://www.mendmeshop.com/_img/quadricep-muscles.jpg, with permission from Mr. Darren Cole, President, Mend Me Shop.   6 1.2.2.4 Tendons and Ligaments Tendons and ligaments play an important role in providing joint stability.  Tendons attach muscle to bones and are composed primarily of highly aligned collagen fibres.  Because they transmit large forces they are very strong.  Ligaments attach bone to bone and play a stabilizing role in the joint.  Ligaments are primarily composed of aligned collagen fibres, but in comparison to tendons the collagen content is slightly lower and the fibres are more randomly oriented.  At the patellofemoral joint, the distinction between tendons and ligaments is not straightforward.  This is because the patella is a sesamoid bone (a bone embedded in a tendon).  Therefore, the tendons for each of the four quadriceps muscles do not technically insert on the patella, but rather merge into one tendon and insert on the tibia at the tibial tubercle.  As a result, it is not surprising that there is some variation in the terminology used to describe the tendon and ligament attachments at the patellofemoral joint.  The portion of the quadriceps tendon proximal to the patella is referred to as the quadriceps tendon, while the portion distal to the patella is referred to as either the patellar tendon or the patellar ligament.  The term patellar ligament will be used throughout this thesis. 1.2.2.5 Other Passive Soft Tissues The patella is also attached to the tibia and femur with passive soft tissue structures on its medial and lateral margins.  The retinaculum is a thin sheath of tissue that is an extension of the vastus lateralis and medialis muscles.  It attaches the patella to the collateral ligaments of the tibiofemoral joint and the tibial condyles.  The medial patellofemoral ligament originates at the medial femoral epicondyle and inserts on the superior part of the medial patella.  The iliotibial band is a thin fibrous tissue that extends from the lateral condyle of the tibia, along the lateral margin of the patella to the iliac crest (pelvic bone).  To summarize, the medial margin of the patella has attachments to the retinaculum and the patellofemoral ligament and the lateral margin of the patella has attachments to the retinaculum and the iliotibial band.  7  Figure 1.2.3 has been removed because permission to use this figure in the thesis was not granted by the illustrator.  The figure was an anterior view of the knee showing the patella, the patellar ligament, the articular capsule, the medial and lateral retinaculum, the vastus medialis obliquus, the vastus lateralis, the iliotibial band and the quadriceps tendon.  Images of these structures can be found in anatomy textbooks that include detailed knee anatomy. Figure 1-6: The retinaculum and the iliotibial band.  1.2.3 Function The tissues described in the previous section form the extensor mechanism of the knee.  The patella plays a key role in optimizing the mechanical function of the extensor mechanism.  In this section the role of the patellofemoral joint within the extensor mechanism will be highlighted, with particular attention to the loads acting on the patella and the response of patellofemoral cartilage to load. 1.2.3.1 Forces Acting on the Patella The quadriceps muscle group drives the extensor mechanism.  The patella increases the moment arm of the extensor mechanism and therefore reduces the quadriceps force required to extend the knee.  One could imagine the system without the patella (Figure 1-7).  In this system the quadriceps tendon would sit within the femoral trochlea and insert directly into the tibia at the tibial tuberosity.  For illustrative purposes, the moment arm of the patellofemoral joint can be described as the perpendicular distance from the axis of rotation (a medial/lateral directed axis) to the quadriceps  8 force vector.  The location of the centre of rotation is within the distal femur and its location likely changes with knee flexion.  It is clear from the figure below that the moment arm in the system without a patella is shorter than the one with it (d1<d2) and therefore more force is required to extend the leg [8].  Also, the frictional properties between the tendon and trochlear cartilage would differ from the very low friction cartilage-cartilage scenario of the patellofemoral joint, which could also increase the amount of force required to extend the knee. Fq-patella Fq+patella FR Fq-patella > Fq+patella = Centre of Rotation d1 d2 d1 < d2 therefore Fp  Figure 1-7: Extensor mechanism with and without a patella. Sagittal (left) and (axial) views of knee. The images without the patella highlight the mechanical advantage the patella has in the extensor mechanism.  The patella reduces the force required to extend the knee against a given flexion moment by increasing the moment arm of the system.   The primary forces acting on the patellofemoral joint are those of the quadriceps tendon (Fq), the patellar ligament (Fp) and the joint reaction force (FR).  The weight of the patella is usually neglected in the force analysis because it is very small compared to Fq, Fp and FR.  Frictional forces are also usually ignored because cartilage has a very low coefficient of friction (less than 0.018) [9]. Cadaver studies and inverse kinematic analyses have shown that the Fq required to extend the leg, in both open chain and closed chain loading scenarios, increases with knee flexion [10].  It has also been shown that the ratio of Fq to Fp changes with knee flexion [10,11].  At flexion angles less than approximately 45° the force in the patellar ligament is greater than in the quadriceps tendon; at angles greater than 45° the opposite is true. This indicates that the patella is not simply acting to change the direction of the Fq but also the direction and location of FR in a knee flexion angle dependent manner. This is likely due to the changing location of the contact regions with knee flexion (the contact region migrates proximally along the patella with knee flexion, which will be discussed in greater detail in  9 subsequent sections) [12] and also possibly the length of the moment arm [8].  Currently, the actual magnitude of Fq during activities of daily living is unknown because it is currently impossible to measure muscle force directly.  However, using inverse dynamic analysis, physiologic cross sectional muscle areas and electromyography (EMG) these forces have been estimated.  FR has been estimated to be between 0.5 and 7.6 times BW, depending on the nature of the activity, with the lower range being for level walking and the upper range for deep knee bends [13]. It is unclear how the exclusion of the forces applied by the passive tissues of the medial and lateral margins (retinaculum, iliotibial band, medial patellofemoral ligament) would affect joint mechanics in the above model.  Passive tissues have been shown to influence mechanics ex vivo [14- 18].  Although these tissues cannot actively load the patella, they appear to be important in passively resisting medial and lateral patellar motion as they are stretched, thereby stabilizing the joint. However, these forces are likely small compared to Fq, Fp and Fr and likely wouldn’t greatly alter the findings discussed above. 1.2.3.2 Response of Cartilage to Load One role of articular cartilage is to support and distribute load through the joint.  Articular cartilage displays a viscoelastic response to applied loads due to its fluid and solid matrix phases.  It has been shown that the response of cartilage to compressive, shear and tensile loads is dependent on the biphasic nature [19].  When a compressive load is applied to the cartilage, cartilage deforms and it is primarily the fluid phase that supports the load due to the decreased volume and the low permeability of the cartilage surface.  When cartilage is loaded in shear, which doesn’t occur in isolation in vivo but at the bone-cartilage interface which is constrained during compression, the response is flow independent because it is the collagen network that carries the load.  Finally, cartilage in tension responds with a combination of fluid flowing due to change in volume and collagen fibres stretching.  The viscoelastic (creep and relaxation) response of articular cartilage to compressive load (Figure 1-8) has been characterized ex vivo in indentation (creep) and confined compression (stress relaxation) experiments [20].  Equilibrium is reached for creep when the majority of the fluid has exuded from the cartilage therefore the solid matrix is carrying the load (which does not occur in vivo) and for stress relaxation when fluid pressure gradients within the cartilage have equalized.  When a joint is loaded in vivo, cartilage deforms by 2.4-8.6% and takes more than 90 minutes to return to its pre-loading state [21], highlighting the length of time required for cartilage to return to unloaded equilibrium in vivo.  Further, all synovial joints in the body experience similar cartilage contact stresses; ex vivo studies have found average contact stresses to be between 0.2 and 2 MPa and peak contact stresses of 2 to 10 MPa [22].  This highlights the ability of cartilage to  10 distribute load over the articular surface (and also throughout the tissue) to an optimized physiologic level. ε σ time time Creep (σ=const) Stress Relaxation (ε =const) equilibrium equilibrium  Figure 1-8: Creep and relaxation behaviour of articular cartilage.  1.3 Patellofemoral Osteoarthritis OA is a degenerative joint disease that causes pain, stiffness and loss of mobility.  It is characterized by the development of osteophytes (bony protrusions on the joint margins), cartilage degeneration, bone marrow lesions (abnormalities in bone marrow), synovitis (inflammation of the synovial membrane), joint laxity and muscle weakness.  It is not surprising that these changes in structure result in changes in mechanical function of the joint such as difficulties walking, rising from a chair or climbing stairs.  It has also been hypothesised that mechanics play a role in the onset of the disease: ‘OA is best defined as a failed repair of joint damage that has been caused by excessive mechanical stress.’ [23].  In this section, patellofemoral OA will be discussed in terms of epidemiological data, clinical definitions, mechanical risk factors and mechanical-based treatment strategies. 1.3.1 Epidemiology 1 in 10 Canadians suffer from OA [24].  OA commonly affects the knee joint [25], with an estimated prevalence between 19.2% [26]  and 27.8% [27] in individuals over 45 years of age [28], and the patellofemoral joint is involved in 50% of all knee OA cases [1].  Lateral patellar facet disease is more prevalent than medial patellar facet disease, with lateral disease accounting for between 78% [29] and 89% [30] of all cases.  A longitudinal study also found that lateral patellofemoral OA progression is more common than medial progression [31].  Patellofemoral OA  11 can occur in isolation or in combination with tibiofemoral OA.  In a radiographic population-based study of individuals with knee pain, 40% had combined knee OA, 24% had isolated patellofemoral OA, 4% had isolated tibiofemoral OA while the remainder did not have OA [32].  Further, isolated patellofemoral OA has also been associated with pain, stiffness and functional limitation [33]. Together, these findings suggest that patellofemoral OA is a major clinical problem for which pain is a significant clinical symptom. 1.3.2 Clinical Definitions Defining OA has been a challenge for researchers because there is both a structural and symptomatic component to the disease.  Practically this means that we can define OA from anatomical changes that can be visualized using various imaging modalities (radiography, MRI) and also from symptoms experienced by the individual that can be assessed by physical exam and questionnaires.  Radiography is the most commonly used tool to assess structural changes in OA, while questionnaires are most commonly used to assess symptoms. The radiographic quantification of OA consists of identifying and scoring two particular structural changes: joint space narrowing and presence of osteophytes.  Protocols have been developed to characterize and score the magnitude of these changes from axial (or sometimes sagittal) weightbearing radiographs [34,35].  Since cartilage is not visible on radiographs, joint space narrowing (a surrogate measure of cartilage thinning), is used.  Joint space narrowing is the narrowing of the space between the bone surfaces using the assumption that articular cartilage occupies the entire space between the bones.  Joint space narrowing and osteophytes are scored according to atlases associated with the protocols.  One limitation of radiographic assessments is that focal defects, which are a feature of early disease, cannot be identified.  Radiography is therefore not ideal for identifying individuals in the early stages of the disease [36]. MRI assessment tools are also becoming more widely used in order to identify and classify other structural features of OA.  This is because with MRI it is possible to visualize many tissues in addition to bone.  Methods of assessing cartilage morphology (such as thickness, volume and surface area) have been validated [37-40] and used extensively to study OA.  Biochemical assessments of cartilage composition, which can detect early changes in cartilage composition, can also be carried out using MRI [41].  Since OA is a disease of the entire joint, MRI scoring systems that encompass tissues beyond cartilage have been developed [42,43].  These scoring systems use semi-quantitative assessments of factors such as cartilage thinning, bone marrow lesions, bone cysts, bone attrition, osteophytes, meniscal damage, ligament damage, effusion and synovitis.  These broader scoring  12 systems provide a more complete description of structural changes in OA than can be obtained with radiography. Validated questionnaires are often used to assess OA symptoms such as pain and function.  The Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) is the most commonly used questionnaire and it has been validated for hip and knee OA [44].  The questions are divided into the subsections of pain, stiffness and difficulty carrying out daily activities.  The Knee Injury and Osteoarthritis Outcome Score (KOOS), which is an extension of the WOMAC, is a validated, knee specific questionnaire that has also been used to assess OA symptoms [45].  In this questionnaire, five dimensions (pain, symptoms, activities of daily living function, sports and recreation function and quality of life) are assessed.  Questionnaires are easy to administer and are very important for describing the impact of disease on the individual’s daily life; however, they are subjective and not specific to the patellofemoral joint. The American College of Rheumatology recommends that a combination of structural and symptomatic assessments be used to describe the severity of OA [46].  However, in practice it is difficult to summarize this information into a single descriptor of OA severity and therefore, even when they are both assessed in a study, they are used independently for correlations.  Radiographic [34] and symptomatic definitions [46] remain the most common in OA studies; however, new semi- quantitative scoring systems are becoming more widely used. 1.3.3 Mechanical Factors Several mechanical factors have been associated with radiographic and/or MRI-based definitions of patellofemoral OA such as obesity [1,47,48], varus/valgus tibiofemoral alignment (Figure 1-9)  [29,31,49] , patellar alignment (Figure 1-9)  [2-5], knee height [50] and meniscectomy [51].  Although, apart from patellar alignment, these are not direct measures of patellofemoral mechanics these associations suggest that mechanics play a role in patellofemoral OA.  13  aligned malaligned Figure 1-9: Varus and valgus tibiofemoral alignment (left) and patellar malalignment (right). Patellar malalignment is most often assessed in the axial plane and can be in the medial or lateral direction, the malaligned case above is an example of lateral malalignment. FM=mechanical axis of the femur, TM=mechanical axis of the tibia, LBA=load bearing axis, HKA=hip-knee-ankle angle.  Varus/valgus image reprinted with permission from the Journal of Rheumatology, 34(9), Frontal plane knee alignment: a call for standardized measurement, Cooke et al, 2007, p1797 [52].   The evidence to support the relationships between surrogate measures of patellofemoral joint mechanics and OA is quite strong.  Obesity has been identified as a risk factor of both patellofemoral OA and tibiofemoral OA [1,47].  In fact, obesity puts individuals at greater risk of radiographic patellofemoral OA than radiographic tibiofemoral OA (odds ratios (OR)=3.5, 7 and 1.9 for isolated patellofemoral, combined and tibiofemoral OA, respectively) [1].  Increased knee height, measured from the ground to the femoral condyles when sitting, was associated with radiographic patellofemoral OA in men (OR=1.7) and was associated with symptomatic patellofemoral OA in women (OR=2.2) [50].  Varus tibiofemoral malalignment increases the odds of medial compartment radiographic patellofemoral OA progression (OR=1.85) and valgus tibiofemoral malalignment increases the odds of lateral compartment patellofemoral OA progression (OR=1.64) [31]. Meniscectomy also increases risk (OR=2.6) of developing radiographic patellofemoral OA [51]. Patellar alignment, assessed in a single axial plane, is associated with patellofemoral OA disease features and progression, defined using radiography or MRI.  Lateral alignment in the axial plane, expressed as lateral tilt angle (greater angle is medial tilt) and bisect offset (greater offset is lateral translation) (Figure 1-10), has been associated with lateral joint space narrowing (OR=0.10 and 8.26, respectively) [4], lateral osteophytes (OR=0.29 and 3.07, respectively) [4], lateral cartilage loss (OR=0.3 and 3.4, respectively) [3] and lateral bone marrow lesions (OR=0.1 and 3.2, respectively) [3]. Medial alignment, expressed again as lateral tilt angle and bisect offset, was  14 associated with medial joint space narrowing (OR=2.85 and 0.189, respectively) [4]. In a study of OA progression, medial displacement of the patella increased the risk of medial patellofemoral joint space narrowing progression (OR=2.2), while it was protective of lateral patellofemoral joint space narrowing progression (OR=0.4) and increasing lateral tilt was protective of medial patellofemoral joint space narrowing progression (OR=0.2) [2].  It has also been shown that patella alta (a high riding patella) is associated with lateral joint space narrowing (OR=2.77) [4], lateral osteophytes (OR=1.67) [4], medial and lateral cartilage loss (OR=2.0 and 2.0, respectively) [3] and lateral bone marrow lesions (OR=2.5) [3].  All ORs reported represent those of the highest quartile.   Figure 1-10:  Two-dimensional assessments of patellar alignment. Patellar tilt angle (LPT1) and bisect offset (BO).  The LPT1 is the angle between the line connecting the posterior condyles and the line along the lateral facet.  The BO is the percentage of the patella that lies lateral to the line through the deepest part of the trochlear groove that is perpendicular to the line connecting the posterior condyles.   Image modified with  permission from Kalichman et al. 2007 [4] under the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0).  It has been hypothesised that the association between joint alignment and OA is due to loading increases and altered loading patterns in the joint.  It is clear that increased load (obesity, a complex systemic issue) and changes to patterns of loading (knee height, varus/valgus tibiofemoral alignment, meniscus damage, patellar alignment) are related to OA.  It is not clear, however, if it is peak load, load distribution or load history that is most important in the disease process [22].  While cartilage contact stresses have been shown to be greater around cartilage defects [53], this gradient has been shown to disappear over time in an animal study [54]  indicating the joint is able to adapt and that peak loads or stresses may not be the most important factor for disease progression.  Other studies have found that strain rate was related to articular cartilage cell activity.  Greater strain rates were associated with greater damage [55,56].  These findings suggest that loading rate could be very important in OA disease; however, further work is required in a broader OA model that includes other tissues and other measures of joint damage.  These findings highlight the importance of loading and the continued study of joint mechanics in OA.  15 1.3.4 Mechanical-based Treatment Strategies Treatment strategies for patellofemoral OA have focussed on unloading the lateral facet.  This stems from an early hypothesis that OA is caused by an imbalance in stress distribution (referred to as hyper-pressure by the authors) between compartments [57].  Treatment options range from conservative (physiotherapy, bracing and taping) to operative but the evidence to support the effectiveness of these treatments is limited.  A systematic review of treatment options for isolated patellofemoral OA deemed that the current evidence to support any of the treatment options was weak to moderate for all but one study [58].  The one study deemed to be of ‘high’ quality found that a physiotherapy intervention was not an effective treatment strategy for patellofemoral OA, since patient outcomes between the intervention and control group did not differ at the 12 month mark [59]. Studies were deemed to be of moderate or low quality due to issues such as small sample size, short length of intervention or follow-up and methodological assessments of outcomes; however, some of the treatments did show reductions in symptoms and therefore merit further study. Physiotherapy, including muscle strengthening and mobilization, is often prescribed in cases of patellofemoral OA but the effectiveness of this strategy is unclear.  Only one study, mentioned above, has examined the effect of a 10 week physiotherapy intervention [59].  This intervention included patellar taping, 7 muscle strengthening exercises (focussing on quadriceps muscle), posture correction and footware advice.  While pain decreased and quadriceps strength increased 10 weeks after completion of the intervention, these results were not maintained at the 1 year mark.  It is not clear if the reduction in pain would continue if a long-term physiotherapy intervention was carried out.  A second group has proposed a protocol for a targeted physiotherapy intervention that includes quadriceps muscle retraining, quadriceps and hip muscle strengthening, patellar taping, manual patellofemoral joint mobilization and education [60].  The results of this study have yet to be published.  The effect of a physiotherapy intervention on patellar alignment has not been assessed. Patellar taping is one treatment strategy that appears to merit further study.  The aim of patellar taping is to position the tape in such a way that the patella is translated medially.  It is hypothesized that the medialization of the patella will unload the lateral compartment and will reduce pain and arrest OA progression.  One small randomized controlled trial of 14 patients with isolated patellofemoral OA found that medial taping reduced pain, assessed daily on a visual analogue scale, by 25% [61].  Patellar taping was also effective at reducing pain in study populations that included individuals with isolated or combined patellofemoral OA [62-64].  Challenges of patellar taping include skin irritation caused by the adhesive, the technical skill required to correctly apply the tape and loosening of the tape during activity.  16 Patellar medialization braces are commonly prescribed by physiotherapists, family physicians and rheumatologists as an alternative to taping.  With bracing, skin irritation is not an issue, individuals can be easily instructed on how to correctly don and doff the brace and the brace can be tightened if it loosens with activity.  However, there is currently no evidence to support this intervention in the patellofemoral OA population.  Bracing has been effective at reducing pain in the patellofemoral pain population; however, this population is generally much younger and does not have degenerative joint changes [65,66].  It is unclear whether a patellar medialization brace can translate the patella to the same extent as taping. Surgical interventions have also been used to treat patellofemoral OA with variable success. These interventions aim to correct joint mechanics through different means such as lateral retinacular release, tibial tubercle osteotomy, tibiofemoral joint replacement and patellofemoral joint replacement [67]. Surgical strategies reduce pain in some individuals [68-73] but others experience no improvement [74,75].  Currently, there is not enough evidence to support surgical interventions [58] and ideally a surgical intervention would resolve the underlying mechanical problem and not just reduce pain.  Further, the actual mechanical changes induced by these interventions have not been studied. Most studies to date have measured the success of the mechanical-based treatment strategies with respect to their ability to reduce pain; however, knowing the magnitude of mechanical correction required to reduce pain is essential for designing and evaluating treatment strategies.  Only one study has assessed mechanical correction as an outcome measure.  This study examined the effect of patellar taping on pain and patellar alignment in the axial plane [64].  The authors found that the patellar tape did cause a medial translation of the patella and did reduce pain during a squat exercise. One limitation of this study was that the assessment was carried out immediately after taping, therefore it is unclear if the medialization was maintained as the tape loosened throughout the day.  A cadaver study examined the effect of patellar bracing on load distribution and kinematics and found that the brace reduced contact stress by 10% and caused the patella to move proximally [76]. However, since we do not know the magnitude of mechanical correction required to reduce pain in vivo it is difficult to know if this reduction in contact stress and change in kinematics is clinically significant. It is clear that a better understanding of mechanical treatment strategies of patellofemoral OA are required.  It is crucial to assess mechanical changes before and after intervention in order to identify the magnitude of the mechanical change and the mechanism by which these interventions reduce pain.  17 1.3.5 Summary Although patellofemoral OA is a major health care problem, mechanical-based risk factors of patellofemoral OA have been identified and mechanical-based treatment strategies are prescribed in cases of patellofemoral OA, relatively little is understood about what magnitude of mechanical correction is required to alleviate patient symptoms.  Knowing the magnitude of the mechanical correction required is essential for improving and evaluating current treatment strategies, therefore direct measurements of patellofemoral joint mechanics are necessary. 1.4 Three-dimensional Patellar Kinematics The primary motion of the patella is flexion/extension; however, it does rotate and translate in three dimensions.  While the rotations and translations of the patella outside of the sagittal plane are small they may be of substantial clinical importance.  Patellar kinematics have been studied ex vivo [14-18,77-96] and in vivo [97-100] using motion tracking systems that require markers to be rigidly fixed to bones.  More recently, non-invasive, in vivo, imaging-based methods have been developed using MRI and bi-planar radiography [101-106].  Results for normal three-dimensional patellar kinematics vary greatly between studies for both ex vivo and in vivo measurements [107].  For example, medial, lateral and neutral patterns of patellar tilt and spin (rotation) have been reported (Figure 1-11).  It is very difficult to identify exactly why such large differences have been seen because there are many factors that can influence the results.  The first and most fundamental factor in any kinematic analysis is defining the coordinate systems to be used to determine the attitude and position of the target body.  Different definitions of local coordinate systems and methods used to calculate the orientation and position likely contribute substantially to the differences observed in Figure 1-11.  When carrying out a kinematic analysis in a biomechanical system (ex vivo or in vivo) additional factors influence the calculated kinematic results such as the type of measurement system used for tracking, the applied loads and the constraints applied to the joint.  In this section factors affecting three-dimensional patellar kinematic assessments and errors associated with the various methods will be discussed.  18  Figure 1-11: Between study variation in patellar tilt (in vivo) and rotation (in vivo and ex vivo). Each line on the graphs represents the results from a different study.  Reprinted from Katchburian et al, Measurement of patellar Tracking: Assessment and Analysis of the Literature, Clinical Orthopaedics and Related Research, 2003, 412:241-59 with permission from Wolters Kluwer Health  [107].   1.4.1 Clinical Motions It is useful to describe the three-dimensional motion of the patella using clinically relevant descriptions based on the anatomical planes described in Section 1.2.1 (Figure 1-12).  There is some variation in the literature in the terminology used to describe patellar motions.  One reason for this is that clinical motions are most easily understood when they are between two long bones, and therefore some clarification of these definitions is required at the patellofemoral joint.  Patellar flexion and extension is the rotation in the sagittal plane; inferior and superior tilt is also sometimes used to describe this motion.  Tilt refers to rotation of the patella in the axial plane and is described as medial or lateral in direction.   While internal/external rotation is more commonly used in other joints to describe motion in the axial plane, the term tilt originates from radiographic assessments of the axial position of the patella and therefore will be used for consistency.  Internal and external spin refers to abduction and adduction of the patella in the frontal plane.  Spin is often referred to as rotation in the literature but since ‘rotation’ is too general, the term spin will be used in this thesis.  Translations are defined according to anatomical directions (proximal/distal, medial/lateral and anterior/posterior).  19  Figure 1-12: Patellar motions. Pr=proximal, D=distal, M=medial, L=lateral, P=posterior, A=anterior.  The arrows describe the positive directions to be used throughout the thesis.  Reprinted from The Knee, 17, McWalter et al, A single measure of patellar kinematics is an inadequate surrogate marker for patterns of three-dimensional kinematics in healthy knees, 135-140, 2010 with permission from Elsevier [108].  While three-dimensional assessments of kinematics are standard in the ex vivo literature, many in vivo assessments are not truly three-dimensional, although they often claim to be.  In fact there are only 8 groups who have developed three-dimensional in vivo methods in the natural knee (3 MRI based [102,104,109], 4 bi-planar radiography based [101,103,106,110] and 1 surface mounted patellar clamp [97]).  The remainder of the assessments of patellar tracking use three-dimensional images (using CT or MRI); however, the actual assessment of kinematics consists of measuring clinical motions in a single imaging slice.  Most of the two-dimensional assessments include measures of lateral translation and tilt which are described using planar measurements such as bisect offset, patellar tilt angle and patellar displacement (Figure 1-10).  These measures originate from assessments of alignment made from axial (skyline) radiographs and have carried over into the three- dimensional imaging literature.  Because these methods don’t fully describe the motion of the patella in three dimensions, the remainder of this section will focus on three-dimensional analyses. Henceforth, when patellar kinematics are referred to, three-dimensional assessment is assumed.  20 1.4.2 Coordinate Systems and Calculation of Kinematics In biomechanics two different types of coordinate systems have been used to describe motion: finite helical axis [111] and body-fixed axes [112,113].  In the finite helical axis approach, movement of a body from a reference position to a current position can be described as a single rotation about and translation along the helical axis.  Advantages of this approach are that it provides a simple description of rotation and translation simultaneously and does not suffer from singularities; however, it does not describe motion in the standard clinical terms described in the previous section, which makes communicating findings to a broader audience difficult.  The finite helical axis approach has never been used at the patellofemoral joint and therefore will not be discussed further.  In the fixed- body axis approach, coordinate systems must be assigned to each rigid body and motions are then assessed using various conventions based on these systems.  The coordinate systems can be defined using anatomical landmarks which then allow motion to be described in clinically relevant terms. Fixed-body axis approaches will therefore be the focus of the remainder of this discussion. In order to use a fixed-body axis approach, axes must be assigned to the femur and patella. This may sound like a straightforward task; however, currently there is no standard for patellofemoral joint axis assignment.  As a result there is substantial variation in the landmarks used to assign coordinate systems in the literature (Table 1-1).  The most common method of assigning the coordinate systems is to identify anatomical landmarks that serve to create an origin, a flexion axis and a long axis (Figure 1-13).  The third, anteriorly directed axis is the cross-product of the flexion and long axes.  The cross-product of the third and long axes is then determined to create a flexion axis that is orthogonal to the other two (i.e. an orthogonal coordinate system).  The anatomical landmarks used to create the systems are either directly digitized (in the case of ex vivo assessments only) or identified on an image and then related to the motion tracking coordinate system.  If the landmarks are identified directly or on a radiograph or CT image, then the landmarks correspond to the cortical bone surface landmark.  If the landmarks are identified on an MRI then trabecular bone landmarks are used, since cortical bone appears as signal void on an MRI.  Trabecular bone is the porous bone at the centre of the patella, while cortical bone is a dense layer of bone at the surface.  Further, if imaging techniques are used then the slice in which these landmarks are selected must be specified.     21 Table 1-1: Landmarks used to create anatomical coordinate systems in the patella and femur.          Femur           Patella  Flexion Axis  1. line joining medial and lateral epicondyles [86,89,95,114,115] 2. line joining medial and lateral posterior points of condyles [79,102,109] 3. line joining centres of spheres fit to femoral condyles [14,116] 4. plane touching posterior condyles [78] 5. parallel to tibial coordinate system at full extension [93]   1. line joining medial and lateral points [14,116] 2. line joining posterior point and lateral point [102,109] 3. parallel to femoral axis at 90° [78] 4. parallel to femoral axis at full extension [114] 5. parallel to tibial coordinate system at full extension [93] 6. perpendicular to long and third [115] Long Axis 1. mechanical (origin to centre of femoral head) [114,115] 2. anatomical (origin along centre of femoral shaft) [14,79,86,89,116] 3. anatomical (between two points along femoral shaft) [117] 4. trochlear (line along deepest portion of trochlear groove) [78] 5. parallel to tibial coordinate system at full extension [93]  1. line joining proximal and distal points [79,109] 2. posterior flat edge [118] 3. perpendicular to flexion and third  [116] 4. parallel to femoral axis at 90° [78] 5. parallel to femoral axis at full extension [114] 6. parallel to tibial coordinate system at full extension [93] 7. between origin and distal point [115] 8. line connecting insertions of quadriceps and patellar tendons[14] Third Axis 1. perpendicular to flexion and long axes 2. perpendicular to flexion/extension and long [14,114,115] 3. parallel to tibial coordinate system at full extension [93] 1. perpendicular to flexion and long axes [14] 2. cross product of medial lateral axis and vector from inferior apex along anterior surface [116] 3. parallel to femoral axis at 90° [78] 4. parallel to femoral axis at full extension [114] 5. parallel to tibial coordinate system at full extension [93] 6. perpendicular to plane containing medial, lateral and distal points [115]  Origin 1. deepest point of the trochlear groove where it meets the intercondylar notch [90,102] 2. most proximal point of the intercondylar notch [102] 3. most posterior point on the sulcus groove in the axial plane [109] 4. proximal point of intercondylar notch along to anatomical axis [78] 5. centre of the intercondylar notch [93] 6. midpoint between medial and lateral epicondyles [17,114] 7. midpoint of centres of spheres fit to femoral condyles [14] 8. intersection between the screw axis of the femur (helical axis) and the sagittal plane bisecting the knee [87] 9. centre of the posterior cruciate ligament insertion [89]  1. posterior point on axial midslice [102,109] 2. midpoint of line joining medial and lateral points of patella [115,116] 3. centre of the patella [78,79,89-91,114] 4. centre of the patella using most proximal, distal, medial and lateral points [79] 5. centre of the patellar ridge [93] 6. projection of centre of patella onto line connecting insertions of quadriceps and patellar tendons [14] 7. geometric centre [17] 8. aligned with femur at full extension [114]   22 Lp Lf Tp Tf Fp Ff  Figure 1-13: Anatomical coordinate systems of femur and patella. F=flexion axis, L=long axis, T=third axis, subscripts f=femur, p=patella.  Using these anatomically assigned coordinate systems the relative position and attitude of the patella with respect to the femur can be defined.  Attitude has generally been calculated using either Cardan angles or a modified Joint Coordinate System technique [113].  When Cardan angles are used the order of the rotations must be specified.  The International Society of Biomechanics (ISB) has developed a standardization document that suggests the zyx order (for the patellofemoral joint this would be flexion, tilt, spin) should be used when determining the attitude of a rigid body with respect to the global reference frame [119].  This is a general convention for all joints.  However, the effect of this order has not been characterized at the patellofemoral joint and therefore it is not clear if this order provides the best representation of the clinically described rotations.  In this system, position is defined with reference to the coordinate system fixed in the femur.  The same ISB document suggests that the Joint Coordinate System, also referred to as floating axis, approach should be used for determining relative orientation of two rigid bodies [119].  This convention is most often used at the patellofemoral joint but was originally developed for the tibiofemoral joint [113].  The Joint Coordinate System can be likened to a three-cylinder open-chain system (Figure 1-14).  Using this convention, a unique coordinate system is defined as the flexion axis of the fixed body (e1), the long axis of the moving body (e3) and a third, floating axis (e2) which is orthogonal to e1 and e3. Applying this to the patellofemoral joint, the femoral flexion axis is e1 and the patellar long axis is e3.  The Joint Coordinate System is the global coordinate system: patellar flexion is about the femoral flexion axis (e1), patellar tilt is about the patellar long axis (e3), patellar spin is about the third, floating axis (e2). The advantage of this type of coordinate system is that the results are independent of rotation order.  One study explored the effect of defining e3 as the anteriorly directed axis instead of the long axis of the moving body and found that the largest difference between the two coordinate systems was observed in rotations about e1.  Only very small differences were observed in the other  23 two rotations as long as these rotations were small (which is the case at the patellofemoral joint) [112].  These authors proposed that e3 should be the long axis of the moving body in order for e2 to be perpendicular to e1 and therefore represent a true ab/adduction (or spin in the case of the patella). This would not necessarily be the case if e3 was defined as the anteriorly directed axis.  When the Joint Coordinate System is used, position can be described as the distance between origins along the axes described by the convention, along the anatomically based femoral axes or along the anatomically based patellar axes (using original patellar position as a reference). Tilt FlexionSpin e1 e3 e2  Figure 1-14: Joint Coordinate System modified for the patellofemoral joint. Reprinted  from Journal of Biomedical Engineering, 14(4), Hefzy et al, Effects of tibial rotations on patellar tracking and patello-femoral contact areas, 329-43, 1992 with permission from Elsevier [93].   The choice of anatomical landmarks used to create the coordinate systems and the method used to calculate patellar position and attitude likely explain some of the variability observed in the literature.  One study examined the effect of choosing different anatomical landmarks on kinematics using the Joint Coordinate System approach in one cadaver specimen [117].  This study found only small differences in flexion and medial/lateral translation (less than 1° and 1 mm, respectively) when using an epicondylar-based femoral flexion axis and posterior condyle-based femoral flexion axis. However, large differences were found in patellar tilt when different femoral long axes were used. For example, differences of up to 9° of tilt were found between assessments using a femoral long axis defined between the origin and the centre of the midshaft and one defined between two points along the shaft.  The location of the origin is also a likely source of variation because several axes and all translations are defined by these landmarks.  Often studies do not describe the locations of the origins well, for example the description ‘centre of the patella’ has been used to describe the patellar origin; however, it is not clear how this centre is calculated and if it refers to the centre of the patellar surface  24 or the centre of the bone itself.  Also, sometimes the origin of the patella is not defined using anatomical landmarks but rather by defining a ‘zero’ position at full extension [90,114] or at 90° [78], which allows relative motion to be evaluated but does not allow for comparison between specimens or participants because the ‘zero’ position likely differs between individuals.  Following this, it is likely that the data presented in Figure 1-11 are zeroed for comparison purposes, but it is also likely that there is actually variability in this zero value as a result of coordinate system assignment. Patellar kinematic data are usually plotted as a function of tibiofemoral flexion and therefore differences in tibiofemoral angle calculation between studies may also account for variability in the literature. Tibiofemoral angle can be calculated using the techniques described for patellofemoral kinematics. Virtually all ex vivo studies but only some in vivo studies used this method.  In vivo, standard goniometers are often used for positioning during imaging.  Some studies then use this goniometer based measure of flexion angle while others use the angle calculated from the image analysis.  Standard goniometers can have an error of up to ±10°; therefore, it is likely that, if this angle is used for comparison between subjects, an additional tibiofemoral flexion error is introduced. Plotting patellar kinematics as a function of calculated tibiofemoral angle will greatly reduce the tibiofemoral flexion error; however, errors will still exist as a result of anatomical axis assignment and kinematic calculation methods. 1.4.3 Measurement Systems Ideally joint kinematics would always be assessed during continuous movement.  However, it has frequently been necessary to estimate kinematics from measurements of position and orientation taken at sequential static poses through the range of motion.  Early ex vivo studies measured patellar kinematics at sequential static poses over the range of knee flexion using techniques such as casting, photography, manual grids, three-dimensional electromechanical goniometers and roentgen stereophotogrammetry (RSA) [78,82,85,90,92,96,120].  When optical and magnetic dynamic tracking systems became available they were quickly adopted for kinematic assessment and are still the most widely used method in ex vivo studies [79,80,82,86-89,91,95,114,115].  These systems allow markers, which are rigidly affixed to the bones, to be tracked with up to 0.1 mm system accuracy (manufacturer quoted error).  Motion tracking systems have also been used in vivo; however, rigidly affixing the markers to bones is an invasive procedure and only three studies have used this method [98-100].  All of these studies had a small number of study participants and therefore results currently exist for only 6 normal individuals and 2 individuals with patellofemoral pain.  However, results for just one individual are available from peer reviewed literature [98], the remainder of the data are  25 contained in thesis documents and therefore are not widely available.  Due to the invasive nature of this technique, it is not useful for longitudinal or large scale cross-sectional studies. Non-invasive techniques for assessing kinematics in vivo have also been developed.  One group developed a patient specific patellar surface clamp that is tracked using an optoelectronic system [97].  It has been used to assess patellar kinematics during gait.  This method is limited by the range of flexion that can be studied (full extension to 20° of knee flexion) and the fact that the presence of the clamp may affect kinematics.  Several imaging techniques have also been developed. As in the first ex vivo assessments, most of the in vivo assessments are carried out using a sequential static pose measurement technique [102-104,116].  These techniques have used markerless bi-planar radiography, [103], RSA [106] and MRI [102,104,116].  Kinematics are assessed by creating bone models from the images and assigning anatomical coordinate systems to the models.  These methods are estimates of kinematics since images are not acquired during continuous knee flexion.  There are two techniques that assess patellar kinematics in vivo during continuous knee flexion. The first is an MRI based method that employs a scan sequence originally developed for imaging blood flow [109]. It is a combination of two MRI techniques: cine MRI and phase-contrast MRI.  In cine MRI, images are acquired at constant intervals throughout periodic movement, in this case repetitive flexion and extension, and then averaged to create a set of anatomical images over the range of flexion angles in a single sagittal plane.  In phase-contrast MRI, phase information is used to determine the x, y and z velocities of each pixel.  By combining the two techniques, cine phase-contrast MRI describes the velocity of each pixel in the set of anatomical images and by integrating between knee angles orientation of the patella relative to the femur can be assessed.  This technique is limited because a large number of periodic motion cycles are required to obtain images of sufficient quality, images are acquired in a single sagittal slice that must be carefully selected to minimize out of plane motion errors and the range of motion is limited to between full extension and 40° of knee flexion.  The final method has employed a custom built dynamic bi-planar radiography system [101].  In this method images are acquired at 60 frames/s (shutter speed 1/500 s) while the participant carries out a lunge motion.  Kinematics can be assessed in a similar manner to the sequential static pose techniques, therefore no additional markers are required as in traditional RSA.  This particular technique is limited because it requires that the participant be exposed to ionizing radiation and it does not use a commercially available imaging system.  An overall limitation of imaging-based patellar kinematic assessments is that they are very expensive in terms of both monetary cost (system cost and hourly fees) and analysis time.  This may limit their applicability to large scale studies and diagnosis of disease.  The errors associated with the in vivo methods will be compared in Section 1.4.6.  26 It is currently not clear if differences exist between kinematics assessed using sequential static and continuous methods because this has not been examined to date.  One two-dimensional study of kinematics in the axial plane compared an unloaded, sequential static assessment to a continuous assessment using MRI and found differences in patellar tilt angle and bisect offset (Figure 1-10) [121].  These differences may be due to differences in inertial forces between the assessments; hysteresis has been found between flexion and extension data in ex vivo continous studies [79]. However, these differences could also be attributed to other factors such as planar versus three- dimensional measurement or loading (to be discussed in the following section). 1.4.4 Loading The method used to load the patellofemoral joint likely has a large effect on the results of kinematic analysis.  In ex vivo studies, knee specimens are mounted in a loading apparatus and loads of varying magnitude and direction are applied, most often through the quadriceps tendons [14-18,77- 96].  In in vivo studies, individuals are required to carry out a loading task, such walking, lunging or statically resisting a load, in which muscle recruitment is a physiological process that is automatic [97,98,100-106].  In this section, the methods used to load the patellofemoral joint both ex vivo and in vivo will be examined and the effect of loading methods on kinematics will be discussed. 1.4.4.1 Ex Vivo Two distinct types of loading apparatus have been used to date.  The first simulates an open- chain leg extension [14,15,17,18,78-80,82,85,88-90,92,93,96,122,123], while the second simulates a closed-chain weightbearing squat [86,87,91,94,95,124,125].  In leg extension type rigs, one segment (femur or tibia) is fixed, while the other is free.  The free segment is unloaded [14,15,17,82,93] or flexion moments [78,79,81,93,123] are applied.  The magnitude of the flexion moment is not often specified because is applied to resist a particular load applied to the quadriceps.  By leaving the tibial end unloaded or by applying a flexion moment (which is usually a simple applied weight or a stationary structure to resist applied quadriceps force), the tibia is unconstrained.  In the squat type rigs, knees are attached to the rig such that the patellofemoral joint has 6 degrees of freedom.  Muscle loading is also simulated.  The fidelity of this movement (as compared to a normal in vivo situation) will depend on the forces applied to the system through the apparatus attachments and simulated muscle loads.  Most studies utilize leg extension type rigs likely because of their ease in design and mounting; however, the squat type rigs are more representative of weightbearing activity. The magnitude and direction of quadriceps loading are of particular importance in ex vivo studies.  Quadriceps loading has been represented as a single resultant line of action along the femoral  27 shaft [15,17,87,88,91-94,124],  by dividing the load between each individual quadriceps muscle [85,86,90,95,96], by further dividing the vastus medialis into its longus and obliquus components (due to differences in muscle pennation angle) [14,18,78,79,81,82,85] and by combining the vastus intermedius and rectus femoris [14,18,89,95].  Other studies have included the iliotibial band [14,86,95] and the hamstrings [14].  The angle of the loading line of action must be specified.  This has been done using a previous anatomical study [126] in which angles are set relative to the femoral axis [86,96], using variations of this anatomical study’s results [14,16,18,79,81,122], using origins and insertions of muscles based on the descriptions in Gray’s Anatomy [127] and applying them by scaling in a specimen specific manner [12,14,78,82], and by using the pennation angle of the specimen measured during dissection [89,90].  Some studies did not specify the method used [85,95]. Load magnitudes must also be specified.  Applied loads have ranged from approximately 20N to 1400N with most studies falling in the 175N to 400N range.  These load magnitudes are often limited by constraints imposed by the loading system and the strength of muscle attachment clamps.  Joint loads range between 0.5 and 7.6 times BW for different simulated activities [13]; the same study estimated that this was the result of a quadriceps load between 0.5 and 3.5 times BW.  If we take an 80 kg male (785 N) as an example, the loads applied in cadaver studies are likely similar to those experienced during walking.  In cases where multiple loading vectors are used to model the quadriceps muscle, the total load must be divided amongst them.  This is most often done using the ratios based on physiologic cross-sectional area (PCSA) previously determined in two studies [12,122] (Table 1-2).  Table 1-2: Division of force applied to quadriceps muscles based on PCSA. Author Rectus Femoris Vastus Intermedius Vastus Lateralis Vastus Medius Ahmed 1983 [12] 14.3% 14.3% 35.7% 14.3% (L) 21.4% (O)  Farahmand 1998 [122] 15% 20% 40% 15% (L) 10% (O)   There has been limited study of the effect of direction of quadriceps loading on patellar kinematics.  One study examined the effect of a changing quadriceps angle (q-angle) using a single vector model to describe quadriceps force and found patterns of lateral translation and tilt when the vector was shifted laterally and medial translation and tilt when the vector was translated medially  28 [87].  This result is not surprising.  Another study compared using a single loading vector to using 3 loading vectors (vastus intermedius + rectus femoris orientated along shaft, vastus lateralis 35° from shaft and vastus medialis 40° from shaft) and found that the single loading vector condition caused greater patellar flexion, internal spin and lateral translation [95].  Varying load ratios between multiple quadriceps vector loading scenarios has also resulted in differences in kinematics [82,86,89], as has the addition of iliotibial band [14,18] and hamstrings [14] loads.  Although it is not surprising that varying loading lines of action alters kinematics, it is currently not clear which loading system most closely represents the in vivo case. 1.4.4.2 In Vivo The in vivo loading methods vary greatly depending on the measurement system used.  Ideally we would study patellofemoral kinematics under physiologic loading conditions, such as walking, rising from a chair, climbing stairs or squatting.  It practice, this is only possible using invasive methods because the imaging-based measurement systems are restricted by the size of the imaging field.  For example, in bi-planar radiography, the patellofemoral joint must be centred between the two beam lines within a region of usually less than 20 cm by 20 cm and therefore the most common motion assessed is a lunge because the knee remains approximately in the same position throughout motion.  The imaging field is restricted even further in MRI because of the machine bore size and because the participant is lying down.  When using the sequential static approach, axial loads are applied to the participant’s foot through pedal systems creating closed chain loading [102,104]. Applied loads have ranged from 80 to 150 N in studies using these systems.  These loads are much lower than what would be applied during daily activities (for example the ground reaction force during a single leg stance in an 80kg individual is approximately 870 N); however, because of the differences in body position (supine and upright) it is not clear how the loads generated by the quadriceps muscles compare.  It is likely that muscle recruitment and activation patterns will differ between the scenarios.  When using cine phase-contrast MRI, a repetitive flexion-extension motion is carried out during imaging, creating open chain loading.  Sometimes an additional free weight is attached to the shank [105]; however, often extension is resisted by only the weight of the shank [118,128].  The magnitude of these dynamic loads is much lower than those of daily activities and it is not clear how different muscle recruitment and activation are from physiologic motion. It is currently not clear how patellar kinematics are affected by differences between the supine loading simulations and upright physiologic loading.  We can gain some insight from a two- dimensional patellar kinematic study that examined differences between a supine, dynamic, open- chain flexion-extension scenario with no externally applied load in a closed bore MRI and an upright,  29 dynamic, weightbearing, closed-chain leg extension scenario in an open bore MRI [129].  The study found that the patellae were positioned more laterally in early angles of flexion and tilted laterally in later angles of knee flexion for the supine loading scenario as compared to the upright.  It is likely that differences in supine and upright loading scenarios exist in the static case also.  It is not clear how load magnitude and position affect three-dimensional patellar kinematics in vivo. 1.4.5 Range of Motion The knee range of motion achievable depends entirely on the configuration of the measurement system.  Range of motion referred to here is that of tibiofemoral flexion, because usually patellar kinematic results are presented as a function of tibiofemoral flexion.  Most ex vivo studies examine kinematics between full extension and 120° of knee flexion.  It has been noted by some authors that ex vivo assessments at less than 10° of knee flexion are not repeatable [78].  The range of motion assessed using in vivo techniques is much more variable across studies.  The patellar clamp can only be used to study the first 20° of knee flexion.  The methods that assess kinematics during a lunge have a range of approximately 90° of tibiofemoral flexion.  The range of motion that can be studied with MRI is much more limited.  The sequential static pose and continuous assessments of kinematics have ranges of motion of approximately 50° and 40°, respectively.  The range of motion achievable using a standard MRI is obviously not ideal.  In the foreseeable future, open MRI systems, which allow imaging over the entire range of knee flexion during weightbearing, will be available.  One study has shown that it is possible to assess three-dimensional patellar kinematics with these systems [130].  While it is important to study kinematics over the entire range of flexion, early angles of knee flexion have always been considered important because there is likely more variation in kinematics when the patella is travelling in the shallower, less constrained, proximal portion of the trochlear groove. 1.4.6 In Vivo Validation There has been limited validation of in vivo kinematic assessment methods.  A complete validation would consist of assessments of error in terms of agreement with a reference standard and measures of repeatability (intra- and inter-subject, intra- and inter-reader).  Errors in agreement and repeatability should be sufficiently small to measure clinically relevant changes (which are likely on the order of 1-2 mm or degrees).  In the literature, reference standards for patellofemoral kinematics have been defined as symmetrical phantoms [104,131], theoretical simulation scenarios [109,110] and cadaver knee specimens in conjunction with RSA [101-103].  The errors associated with the validated methods are shown in Table 1-3.  Sheehan et al. used a single plane motion phantom to validate their  30 cine phase-contrast MRI method and assessed intra-subject repeatability.  This phantom is not representative of the three-dimensional motion of the patellofemoral joint.  Fellows et al. reported agreement errors by using 3 cadaver specimens and RSA as a reference standard at 9 different tibiofemoral flexion angles.  Because the MRI- and RSA-based assessments were carried out sequentially, not simultaneously, the 9 angles differed slightly between assessments.  To facilitate a comparison between methods, splines were fit to each set of data and differences were compared at 1° increments along the curves.  Two challenges arise from not having kinematic measures at identical angles of tibiofemoral flexion.  First, the spline interpolation represents an estimation of patellar orientation and position between measured datapoints and second, error may have been introduced when the specimen was repositioned between the RSA and MRI-based assessments.  It is likely that the reported agreement errors are a ‘worst case scenario’.  Intra-subject and inter-reader repeatability and registration error (the latter not shown in the table) were also reported [132].  The studies by Patel et al. and Shin et al. were carried out by the same group [104,116].  The former reported error with reference to a static phantom, while the latter reported intra-reader repeatability; however, there were methodological differences between studies and so it is not clear if these two error assessments are interchangeable between studies.  Nha et al. reported error with reference to one cadaver specimen and an RSA method using ceramic beads (tantalum beads are standard) [103].  The authors did not state whether the reader was blinded to the results of the ceramic RSA analysis.  In order to be blinded, the beads would have to be removed from the images in a manner that the reader could not detect.  This is important because, if he or she was not blinded it is possible to orient the models to minimize the difference between the markerless and RSA methods.  The agreement error reported is far superior to what other groups carrying out markerless assessments report using bi-planar radiography in other, less technically challenging joints such as the tibiofemoral joint [110], and rivals their ceramic bead RSA reference standard (for which they report accuracy to be less than 0.1 mm in translation and 0.1° for rotation).  Further detail would be required to fully interpret these agreement error results.  Bey et al. reported the error of agreement for their dynamic, markerless bi- planar radiograph method as compared to their dynamic traditional RSA system for 3 cadaver specimens [101].  Finally, Lin et al. reported error of their patellar clamp with reference to direct motion tracking of the patella in one cadaver specimen [133].  The authors also state that there was good agreement between the clamp and fluoroscopic imaging in two subjects; however, no quantitative data were presented.    31 Table 1-3: Errors in three-dimensional patellar kinematic assessment. Author Modality Agreement Error Repeatability Error   Translation (mm) Rotation (°) Reference Standard Translation (mm) Rotation (°) Type Sheehan 1998 [131] MRI - dynamic 0.55 (in-plane) 1.48 (out of plane) - Phantom - 2.9 Intra-subject (3 subjects, 2 times)  Fellows 2005 [102,132] MRI – seqential static pose 0.88 1.02 3 cadavers using RSA 0.81 / 0.68 1.40 / 2.14 Intra-subject (3 subjects, 4 times) / Inter-reader (1 subject, 3 readers)  Patel 2003 [104] MRI– seqential static pose  1.8 3 Phantom - - - Shin 2009 [116] MRI– seqential static pose  - - - 1.76 CV% 4.15 CV% Intra-reader (1 knee, 6 times) Nha 2008 [103] Radiography (bi-planar) – seqential static pose  0.09 0.13 1 cadaver using Ceramic beads, RSA - - - Bey 2008 [101] Radiography (bi-planar) – dynamic  0.395 0.877 3 cadavers using RSA - - - Lin 2004 [133] Patellar clamp - dynamic 0.26 0.38 1 cadaver using optoelectric system - - -   It is clear from Table 1-3 that further work is required by most groups to fully characterize the errors associated with measuring patellar kinematics using the various methods.  The use of phantoms and theoretical situations as reference standards is not ideal because the errors that may result from imaging tissues are not included.  Imaging is based on relative densities of tissues (radiography) or the amount of water in the tissue (MRI); by removing this variable from the assessment a true measure of overall error in the method cannot be obtained.  Also, in the case of motion phantoms, it is  32 not clear if the simulated motion adequately represents patellar motion.  Cadaver assessments also have limitations because it is not clear if cadaveric tissues are representative of live tissues.  Minimal work has been done to characterize the repeatability of the kinematic assessment methods.  Ideally repeatability would be assessed in several ways: repositioning error (intra-subject repeatability), within reader error (intra-reader repeatability) and between reader error (inter-reader repeatability). Differences in methodological errors likely also contribute to the variability observed in the literature. 1.4.7 Ex Vivo and In Vivo Study Results The variation observed in kinematic results throughout the literature is likely a result of the various factors discussed throughout this section [107].  While it is very difficult to compare results due to methodological differences, it is important to identify whether any consistent patterns have emerged for any parameter or between types of studies.  In this section, the overall trends that have emerged from the ex vivo and in vivo literature will be summarized and an in depth comparison of MRI-based results will be carried out, because this is particularly relevant to the work of this thesis. Flexion: The patella flexes at a rate of approximately 0.6° per degree of tibiofemoral flexion and this finding is generally consistent between ex vivo [18,78,79,82,85,88,90,91,115] and in vivo [98,103,118,133-135] studies of normal joint kinematics.  There is, however, usually an offset in the measured value for patellar flexion between studies which is likely due to differences in coordinate system assignment.  An example of this offset can be observed in Figure 1-15, where the rate of flexion is similar between studies but the magnitude differs up to approximately 15° [104,118,134]. Spin: No consistent pattern of spin has emerged from the literature.  Patterns of increasing internal spin [77,78,82,92,103], increasing external spin [18,85,87,89,90,94,96,104,134], constant internal spin [91,97], constant external spin [88], neutral spin [98,115,118] or variable spin [14,79] with knee flexion have been found in normal knees in both ex vivo and in vivo studies.  The results from the in vivo MRI studies are also not consistent (Figure 1-15).  This inconsistency for spin may originate from methodological differences; however, it must also be noted that the error in this measure is often large [102]. It is possible that no ‘normal’ pattern of patellar spin exists and that these differences are inherent to the normal population. Tilt: Most studies to date have found a pattern of increasing lateral tilt [18,79,91,94,98,106,136] or constant lateral tilt [82,87,88,97] with knee flexion in normal knees. However, the in vivo MRI studies and two ex vivo studies have shown patterns of medial tilt in early angles of flexion transitioning to lateral tilt above approximately 20° of knee flexion (Figure 1-15) [85,89,104,118,134].  Some ex vivo studies have shown patterns of medial tilt with knee flexion  33 [77,78,92,96].  It appears that the patella tilts laterally after it engages the deeper, more constrained portion of the trochlear groove between 20-30° of knee flexion.  Prior to this point, the line of action of the quadriceps muscle likely plays a larger role in patellar tilt and this may explain why increasing, constant and medial tilt have been observed in early angles of knee flexion since quadriceps loads and directions have varied between studies. Proximal translation: Proximal translation is not often reported, but in the studies that do, the predominant pattern is distal translation with knee flexion both ex vivo [14,86,88,106] and in vivo [97,104,118,134].  Two studies found patterns of proximal translation in early angles of knee flexion and then distal translation in later angles [91,136], this is likely due to differences in the locations of the origins of the femoral and patellar coordinate systems relative to the location at which the patella begins to flex over the femoral condyles, thereby causing the translation of the origin to appear to be proximal then distal. Lateral translation: Most studies have shown constant [15,88,91,106] or increasing lateral translation [79,82,85-87,89,90,94,98] with knee flexion.  The MRI-based studies all show a pattern of increasing medial translation to approximately 25° and then a pattern of increasing lateral translation [104,118,134] (Figure 1-15); this pattern was also observed in one ex vivo [18] and one in vivo bi- planar radiography study [103].  Patterns of medial translation have also been observed [14,77,78,96,97,137].  These results are similar to those reported for tilt and therefore, again, these findings may indicate that medial-lateral patellar position is variable in early flexion angles where the trochlear groove is shallow, likely again because of differences in quadriceps loading. Anterior translation: Anterior translation is another parameter that is not often reported; however, slight posterior translation with knee flexion is the most commonly reported pattern [88,97,104,106,134].  We would expect the patella to translate posteriorly as it follows the path of the trochlea with knee flexion, although a slight anterior translation in early knee flexion would not be surprising given the shape of the trochlea.  The study by Seisler et al. showed a slight anterior translation with knee flexion, as did one ex vivo study [91].  Similar to results for proximal translation, this variability is likely due in part to the assignment of the origins of the femoral and patellar coordinate systems.  In the ex vivo study, the origin was the base of a transducer positioned at the centre of the patella and all translations were relative to the position at full extension with no applied load.  No details were provided about how the centre was determined but this could be why this difference in pattern was observed.  In the study by Seisler et al. the femoral origin was at the distal end of the trochlear groove and due to the shape of the femoral groove and condyles, anterior translation would be expected when employing this coordinate system.  34  -40 -30 -20 -10 0 10 20 30 40 -10 0 10 20 30 40 50 60 Pr ox im al Tr an sla tio n (m m ) Tibiofemoral Flexion (degrees) sheehan MacIntyre patel -20 -10 0 10 20 30 40 -10 0 10 20 30 40 50 60 Pa te lla r F lex io n (d eg re es ) Tibiofemoral Flexion (degrees) sheehan MacIntyre patel -5 -4 -3 -2 -1 0 1 2 3 4 5 -10 0 10 20 30 40 50 60 Pa te lla r S pi n (d eg re es ) Tibiofemoral Flexion (degrees) sheehan MacIntyre patel -5 -4 -3 -2 -1 0 1 2 3 4 5 -10 0 10 20 30 40 50 60 La te ra l T ra ns lat io n (m m ) Tibiofemoral Flexion (degrees) sheehan MacIntyre patel -5 0 5 10 15 20 -10 0 10 20 30 40 50 60 Pa te lla r T ilt  (d eg re es ) Tibiofemoral Flexion (degrees) Seisler MacIntyre Patel -5 0 5 10 15 20 25 30 -10 0 10 20 30 40 50 60 An te rio r T ra ns lat io n (m m ) Tibiofemoral Flexion (degrees) Seisler MacIntyre Patel Pr ox im al Tr an sla tio n (m m ) Pa te lla r F lex io n (d eg re es ) Pa te lla r S pi n (d eg re es ) La te ra l T ra ns lat io n (m m ) Pa te lla r T ilt  (d eg re es ) An te rio r T ra ns lat io n (m m )  Figure 1-15: In vivo three-dimensional patellar kinematics from MRI in normal individuals. Results presented for studies by Seisler 2007 [118], MacIntyre 2006 [134] and Patel 2003 [104] for normal individuals.   Differences observed within the MRI-based assessments in normal individuals (Figure 1-15) are possibly due to differences in coordinate system assignment.  Seisler et al. defined the origin of the patellar and femoral coordinate systems in the axial slice containing the epicondyles when the knee was extended; the patellar origin was at the most posterior point of the patella and the femoral origin was at the depth of the trochlear groove in this slice [118].  MacIntyre et al defined the origins of the coordinate systems as the most posterior point of the patellar axial midslice for the patella and the most proximal point of the intercondylar notch for the femur [134].  Patel et al appear to have ‘zeroed’ their data; however, no details of the coordinate systems used were provided in the  35 manuscript [104].  Seisler et al and MacIntyre et al use similar definitions for the anatomical axes; however, the patellar long axis was defined slightly differently (the line along the flat posterior edge of the patella and the line connecting the proximal and distal patellar point in the sagittal midslice, respectively).  Further, Seisler et al calculated Euler angles according to the ISB standard (flexion, tilt, spin) while MacIntyre et al used a modified Joint Coordinate System [102,112].  Differences in magnitude for translations (which is particularly evident in anterior translation) are likely due to assignment of the origin of the coordinate systems.  However, the patterns of kinematics agree reasonably well between methods. 1.4.8 Summary In this section, several different factors that may be responsible for the inconsistencies in the three-dimensional patellar kinematic literature both ex vivo and in vivo were highlighted.  In particular, the definition of coordinate systems, the measurement technique used, the calculation techniques used, the range of flexion studied and the loads prescribed are all potential sources of differences.  Loading warrants further study, because it likely has the greatest effect in the raw data results (for example, coordinate systems or calculation method could be changed post hoc if need be and the range of motion studied is a limitation of the measurement tools available).  In terms of loading, in vivo assessments have several advantages over ex vivo measures, the most important being the ability to apply loads in a physiologic manner (no estimates of line of action or magnitudes of applied load are required).  The question of the effect of load magnitude has not been sufficiently addressed in the in vivo three-dimensional patellar kinematic literature. Although in vivo assessments are more representative of kinematics during daily activities, they are time intensive and costly.  This may limit their applicability in large clinical studies. Therefore, another area that warrants further study is determining whether a full assessment of kinematics over a range of flexion angles is required to adequately represent patterns of kinematics. 1.5 Patellofemoral Joint Contact Areas Patellofemoral joint contact areas are fundamental to the understanding of load transmission patterns through the joint and local degeneration in OA.  Ideally we would measure contact areas, contact loads, contact stresses and tissue stresses; however, currently the only in vivo measure possible is contact area.  Direct measures of contact area can only be made with MRI.  Estimates of joint surface interactions have been made using bone surface proximity maps from bi-planar radiography [138]; however, the relationship between these maps and actual cartilage contact areas has not been assessed and so will not be discussed further.  Contact area measurements are useful  36 because they provide information about where loads are transmitted and they allow for a computation of average contact stress (if other assumptions are also made).  Therefore, this section will focus on the direct measurement of contact area both ex vivo and in vivo.  First background on the forces acting on the patella and the response of cartilage to load will be provided.  Next, the tools used to measure contact area and experimental considerations will be discussed.  Finally, ex vivo and in vivo study results will be summarized. 1.5.1 Measurement Systems 1.5.1.1 Ex Vivo Several techniques have been developed to assess contact areas ex vivo, all of which have inherent limitations.  As such, there is no clear consensus on which technique is a reference standard (such as RSA for kinematics).  In this section, the current methods used to experimentally measure contact areas ex vivo, which include dying, casting, pressure sensitive film, dynamic pressure measurement systems and stereophotogrammetry, and their inherent strengths and limitations will be discussed (Table 1-4). The first studies of contact areas were carried out using dye staining techniques.  In this technique, the joint is loaded and dye is introduced into the joint; the areas devoid of dye represent contact. Alternatively, the entire patella is painted, the joint is loaded and then a contact print is created on the femoral surface.  The latter technique has been used only once at the patellofemoral joint  [139].  The dye technique has been used in several different joints, including the hip [140,141], the ankle [142-144], the tibiofemoral joint [145,146], the elbow joint [147], and the patellofemoral joint [124,139].  The dyes used in these techniques have been permanent [124,139-141] or reversible [142,144].  When the permanent techniques are used, a limited number of positions can be assessed by using a series of different coloured dyes [124,139].  There are 2 reversible techniques, one uses a series of chemicals and saline for staining and washing [142] while the other uses carbon particles suspended in oil that is sufficient to create a print which can then be wiped off [144].  For all techniques, once the contact prints are visible on the cartilage surface, they must be interpreted. Studies have qualitatively presented patterns using photographs of the contact prints [124,142,146] or quantified contact area by placing a ruler in the photograph (two-dimensional projection) [139,143,145] or mesh grid on the surface [140].  With quantification tools currently available area measurements could be greatly improved by using current technologies.  Error in this type assessment would be related to the analysis tools used.   37 Table 1-4: Advantages and disadvantages of different ex vivo measures of contact area. Method Advantages Disadvantages  Dye Staining  • Clear description of contact border • immediate visibility of contact area • no additional material introduced into the joint • minimal equipment required  • joint capsule must be opened • areas of non-contact within area of contact is not defined • potential for dye leaching • when quantification carried out using scaled photographs the area is a two dimensional projection which does not reflect the three dimensional topology  Casting • permanent impression of the area, that can be re-quantified at a later date • minimal equipment required • joint capsule must be opened • additional material is introduced into the joint space • location of contact on the contact surface unknown without fiducials • creep or relaxation of the cartilage during curing  Pressure Sensitive Film • high resolution of area assessment • repeated trials possible • agreement with a reference standard and repeatability errors have been quantified (generally underestimates area) • joint capsule must be opened • the deterioration of colour density over time • foreign material is introduced into joint • flat film must be fit to a three dimensional surface by cutting (thereby releasing dye along the cut) or folding (may overestimate area) • location of contact on the contact surface unknown without fiducials  Dynamic Pressure Measurement System • dynamic measurements (throughout flexion if desired) • instant measurements • agreement with a reference standard and repeatability errors have been quantified • repeated trials possible • joint capsule must be opened • resistive sensels ‘leak’ into neighbouring sensels when overloaded • resistive sensels give signal when sensor is bent, even without applied load • capacitive sensor resolution inadequate • capacitive sensor thick and likely to have greater influence on contact • location of contact on the contact surface unknown without fiducial  Stereophotogrammetry • joint capsule need not be opened (until after the assessment to create the geometric model) • location of the contact area on the joint surface is known from model  • requires specialized equipment (calibration frame, custom software) • assessment doesn’t take into consideration the deformation behaviour of cartilage    38 Casting is another early technique used to assess contact area.  In this technique casting material, such as silicone rubber [92,120,145,148] or acrylic cement [149], is introduced into the joint space, the joint is loaded, casting material redistributes to areas of non-contact, the casting material is allowed to set and a cast impression is obtained.  Similar to dye staining assessments, cast impressions of contact area have been assessed qualitatively [92,120,145], by projecting the cast area onto photographic paper and measuring using a polar planimeter [149], or by flattening the cast onto a flat surface, tracing the area with a digitizer and calculating the area (no detail provided on this calculation) [148].  Again, error and resolution of this assessment is based on analysis tools used and has not been quantified to date. Pressure sensitive film has very often been used to quantify contact area, although as its name suggests it is also used to measure pressure distribution over a surface [12,82,86,95,145,150-153]. The most commonly used film is Fuji Prescale film (Fujifilm, Tokyo, Japan) which consists of small coloured fluid filled packets that burst upon exposure to pressure.  Greater pressures will cause more capsules to burst and cause deeper colours.  Magnitudes of pressure are determined using a colour calibration curve.  The smallest measurable area is 0.1 mm2 and the manufacturer cites accuracy of pressure measurement to be 10%, but details on the reference standard are not provided.  One group uses a custom film [150], which has layers of a plastic substrate, an enamel paint and a nylon screen; when the film is loaded in compression, the nylon screen leaves an impression and area can be quantified.  The error of this sensor again would depend on quantification technique, which was not specified.  These films range between 0.25 mm (Fuji film) and 0.285 mm (custom film) in thickness. Dynamic pressure measurement systems are becoming more widely used [76,88,154-157]. Again, they are designed to assess contact stress; however, they use the load applied to a sensing unit of known, finite size to assess contact stress, therefore area measurements are possible.  There are two main types of dynamic pressure measurement system: resistive and capacitive.  When loads are applied to the sensor matrix the resistance or capacitance of the sensing unit is altered and can be calibrated to applied load.  Both systems quote manufacturer accuracies of less than 10% in compression.  The area of the sensing unit is as small as 0.6 mm x 0.6 mm for the resistive systems and 2.5 mm by 2.5 mm for the capacitive systems.  The thickness of the sensors is as thin as 0.1 mm for the resistive systems and 1 mm for the capacitive system.  One independent validation study of a resistive system found errors in contact area to range from 5 to 27% using a round indenter of known size on a flat surface as the reference standard.  These authors improved this error by filtering out all pressures more than 2 standard deviations from the mean and found that smaller areas had much larger errors [158].  Another independent study of a resistive system found repeatability of contact  39 area at the patellofemoral joint to be 7.2% for static loads and 3.0% for dynamic loads [159]. Agreement with a reference standard and repeatability errors of resistive sensors has also been assessed in the facet joints of the spine and were found to be between 18-56% (as compared to a known applied load) and 4-10%, respectively [160].  Results were found to be dependent on calibration protocol and the authors concluded that this type of sensor is not suitable for small joints. One paper compared the resistive and capacitive sensors and found that the pressure measurement obtained using the capacitive sensors showed better agreement with pressure measurements obtained by applying a known load over a known area (-3 to 5% for capacitive, -12 to 20% for resistive). Resistive sensors showed better agreement with the known area measurements (less than 6% for capacitive, less than 2% for resistive) which is not surprising as the resolution of the capacitive sensor is lower [161]. Stereophotogrammetry is also used to assess contact area ex vivo; however, it is not a direct measure but a prediction based on geometric models [145,162,163].  Contact area measurement is based on a proximity analysis that is carried out after determining the relative positions of precise geometric bone models.  The relative positions of the models are determined using a calibration frame and two stereograms (photographs) in convergent directions that include both the frame and the two bodies.  Qualitatively, this method has been shown to agree with dye stains, casting and pressure sensitive film [145]; however, quantitative results have not been assessed. As mentioned, none of these techniques has emerged as a reference standard; however, agreement between techniques has been assessed [145,147].  Ateshian et al. provided the most comprehensive comparison.  In this article dye staining, casting, Fuji film and stereophotogrammetry were assessed in a bovine glenohumeral joint and a bovine lateral tibiofemoral joint (without meniscus).  The results of the four analyses were superimposed in a single image, allowing for ease of comparison.  Agreement was assessed visually and it was found that the results between measures were similar in size and location (Figure 1-16).  Another study compared casting, staining and Fuji film in the elbow and qualitatively found very similar results between casting and staining but did not obtain presentable results with Fuji film because of the large amount of curvature and the small area at the elbow [147].  Because of these results, any of these methods would be suitable for use in studies, but if dye staining or casting is used, three-dimensional measurements should be made, which is now possible with newer methods of assessing surface topology and digitization.  The resistive pressure measurement systems have also been compared to Fuji film in total knee arthroplasty components, and contact areas have been shown to be 11-36% lower with Fuji film than the resistive  40 sensor [164].  This is not surprising because in Fuji film areas loaded below a given threshold will not produce a stain.   Figure 1-16: Contact area in the bovine tibiofemoral joint. Results for dye staining, casting (silicone rubber), pressure sensitive film (Fuji film) and stereophotogrammetry analysis.  Reprinted from Journal of Biomechanics, 27(1), Ateshian et al, A stereophotogrammetric method for determining in situ contact areas in diarthroidal joints, and a comparison with other methods, 111-124, 1994, with permission from Elsevier  [145].   1.5.1.2 In Vivo In vivo measurements of contact area are also a substantial challenge.  MRI has the capacity to image cartilage and therefore it has been the method of choice when measuring contact areas in vivo. There is one group, however, that has used a capacitive pressure transducer to measure radioulnarcarpal areas in vivo under local anaesthetic [157]; this technique is not practical for many joints, including the patellofemoral joint, because of the difficulty in access, the extreme invasiveness of inserting a sensor into the joint space, the ethics approval required, the risk of infection and the recruitment of volunteers, to name but a few reasons.  Therefore, the focus of this section will be MRI-based in vivo assessments of contact area. Several groups have developed methods of assessing contact area in vivo from MRI [104,116,165-176].  Table 1-5 provides a comprehensive overview of all of the studies of patellofemoral contact area to date.  All of these methods have used a fast low angle shot (FLASH) MRI sequence with fat suppression, which is optimized to view cartilage, apart from one that also used a dual echo steady state (DESS) sequence [169].  These scans have been acquired using traditional scanners (at 1.5 and 3.0T) and open configuration scanners (upright and horizontal configurations at 0.5T).  When traditional scanners are used, the leg must be loaded during the MRI  41 scan and this is done using the methods described in Section 1.4.4.2.  In the upright open bore configuration, subjects bear 45% BW through the leg with the aid of an MRI-safe backrest [165,166]. In the horizontal open bore configuration, the knee is loaded by applying an extension moment to the shank with lines and pulleys [168,169,172] .  Groups have used both axial and sagittal scans to assess contact area (Table 1-5); however, there is no literature to indicate which scan direction is most appropriate for assessing contact areas.  A sagittal scan appears to provide more information about the contact periphery than an axial scan (Figure 1-17).  MRI Slices  Figure 1-17: Axial and sagittal MRI slice directions for patellofemoral contact area assessment. The sagittal scan (right) provides more information about contact area periphery than the axial scan (left) for comparable slice spacing.   42 Table 1-5 Comparison of MRI-based methods of measuring patellofemoral contact area. *These studies refer to Brechter et al. 2003 as the validation of the technique; however, the validation uses a considerably longer scan that is sagittal rather than axial. Agreement refers to agreement with a reference standard. NR=not reported  Author MR System Plane Resolution (mm) Time Analysis method Loading Validation Heino-Brechter 2003 [167] 1.5 T closed bore Sagittal 0.39x0.39x2 11 min Manual delineation of region with no distinct separation between cartilage plates using straight line segments, multiplied by thickness  Unknown magnitude, compressive force applied directly to anterior side of patella Agreement: CV 13% with reference to area measures from 5 cadaver specimens using Fuji film Repeatability: Intra-subject CV 2.3% (3 times in 1 specimen)  Salsich 2003* Powers 2004* Ward 2004* Ward 2007* [171,173,174,176]  1.5 T closed bore Axial 0.40x0.40 x1 or x2 39 s Manual delineation of region with no distinct separation between cartilage plates using curvilinear line, multiplied by thickness  25% BW, in supine position Agreement: NR Repeatability: Intra-reader repeatability standard error 1.3mm2 (10 subjects, 2 assessments each) Patel 2003 [104] 1.5 T closed bore Axial 0.28x0.28x1 Not reported b-spline fit along selected points of contact, splines integrated across slices to obtain total area  13.61 Kg (133.5N), in supine position Agreement: NR Repeatability: Inter-reader <11.4%, Intra-reader <6.1% assessed in the tibiofemoral joint  Shin 2009 [116] 3T closed bore Sagittal 0.31x0.31x1.5 4min 40s Manual spline fitting, interpolation between slices 125 N, in supine position Agreement: NR Repeatability: NR  Nakagawa 2003 [170] 1.5 T closed bore Axial NRxNRx2 6min 30s Manual delineation, sprain interpolation algorithm, number of pixels counted  None, passive and active deep knee flexion Agreement: NR Repeatability: Intra-reader 0.21 cm2 Table continued on following page.  43   44 Table 5-1 continued Author MR System Plane Resolution (mm) Time Analysis method Loading Validation Gold 2004 Besier 2005 [165,166] 0.5 T vertical open bore Sagittal 0.78 x 0.78x2 2min 13s Manual delineation of ‘gray-on- gray’ pixels, multiplied by slice thickness 45% BW, in standing position Agreement: CV 3% with reference to area assessed using a gelatine- doped urethane phantom Repeatability: Intra-observer CV 3.0%, Inter-observer 7% (3 times by 3 observers in 6 subjects)  Von Eisenhart- Rothe 2004 Hinterwimmer 2004 [169,172] 0.2 T horizontal open bore Sagittal 0.86x0.86x1.9 4min 26s Patellar and femoral cartilage segmented manually, outline of each surface expanded by one pixel, voxels counted  10 Nm torque to shank, while lying on side Agreement: NR Repeatability: Intra-subject 8.3% (same knee imaged 6 times at 30° of flexion by 1 observer) Hinterwimmer 2005 [168] 1.5 T closed bore, 0.2T horizontal open bore  Sagittal Closed: 0.31x0.31xNR Open: 0.86 x0.86xNR  Closed: 9min 28s Open: 4 min 26s And 4 min 59s  Patellar and femoral cartilage segmented manually, outline of each surface expanded by one pixel, voxels counted None Agreement: NR Repeatability: Intra-subject  8.3% (same knee imaged 6 times at 30° of flexion by 1 observer) for open FLASH sequence Connolly 2009 [175] 3T closed bore Sagittal 0.63x0.63x3 2min 28s Segmenting surface, reconstructing surface using a thin plate spline, proximity analysis to denote contact  10-13 N, in supine position NR   Once the MRI scans are acquired, they must be processed to calculate contact areas.  Several different techniques have been employed to do this.  The simplest of the techniques is to delineate contact manually in a slice-by-slice manner, calculate the length of the contact line and multiply by slice thickness [165-167,171,173,174,176].  Instead of simply multiplying by slice thickness, some groups have carried out linear [104,116] or higher order [170] interpolation between slices and summed the areas of discrete patches (created using points along the interpolated lines).  A second technique that has been used involves segmenting the patellar and femoral cartilage plates separately and then either expanding one surface by a pixel and defining overlap as contact [168,169,172] or carrying out a proximity analysis [175].  In the former case area is then calculated by multiplying by slice thickness while in the latter it is based on fitting a thin-plate spline surface and carrying out a proximity analysis.  It is likely that carrying out an interpolation between slices or fitting a surface yields a better estimate of contact area than multiplying by slice thickness, which will likely underestimate contact area; however, no study has examined this to date. Validation of MRI-based techniques has been limited (Table 1-5).  There are only two studies that have examined agreement with a reference standard: one using cadaver specimens [167] and one using a phantom [166].  In the cadaver study, pressure sensitive film was used as the reference standard.  The film was placed between the patellofemoral contact surfaces, and the knee was loaded though the quadriceps tendon and by applying direct posterior pressure to the patella using rubber tubing.  A sagittal MRI scan (11 minutes in length) was acquired with the film in place and the process was repeated 3 times in each of the 6 cadaver knee specimens.  Agreement error was expressed as the percentage coefficient of variation (CV%) and found to be 13% and the repeatability of the MRI- and film-based assessments were 2.3% and 3.2% respectively.  This assessment was limited because the scan was likely too long for loaded in vivo assessment.  Also, the magnitude of the applied load was not measured and therefore it is difficult to determine whether it was consistent between trials and specimens.  Finally, subsequent studies by this group that reference this validation use a much shorter axial scan sequence (39 s) [171,173,174,176]; each sequence should undergo its own validation.  The second validation study used a hemispherical phantom whose contact surfaces were made of gelatine-doped urethane, which simulate cartilage relaxation times.  The phantom had low and high contact area settings with known areas.  Agreement with the reference was 3.0% (expressed as CV%).  These errors are likely much smaller than what would be observed in a joint due to the circular shape of the contact area and the fact that the readers were not blind to this shape. Intra-observer, inter-observer and intra-subject repeatability has also been assessed in vivo for the MRI based assessments (Table 1-5).  One study examined inter-observer repeatability in vivo and  45 found the CV% to be 7.0% for 3 readers assessing results for 6 individuals at 30° of knee flexion in both loaded and unloaded conditions [166].  Three studies have assessed intra-observer repeatability and found it to be 0.21 cm2 (which was about 6% of the area measured in that study) [170], 1.3 mm2 (which was about 1.3% of smallest measurement) [171] and 3.0% (expressed as CV%) [166].  One study has examined the intra-subject repeatability (6 trials in 1 subject at 30° of knee flexion) and found it to be 8.3% [172].  It is clear that since no study has examined all of these measures of repeatability that further characterization of these methods is required. 1.5.2 Experimental Considerations Many of the factors that affect measures of patellar kinematics, such as loading, range of motion and tibiofemoral angle, are equally applicable to measures of contact area (Section 1.4). There are, however, some factors that are important to contact area assessments, in particular, the viscoelastic nature of cartilage.  While it can be argued that kinematics are also affected by cartilage deformation, contact area assessments generally take longer than kinematic assessments (in particular with MRI in vivo) and therefore are of greater concern. The influence of the viscoelastic behaviour of cartilage is dependent on the type of measurement tool being used.  For the methods in which the joint capsule is opened (dye staining, casting, pressure sensitive film, dynamic pressure measurement systems) cartilage will respond differently than if the joint capsule were left intact (stereophotogrammetry and MRI).  The main reason for this is that the properties of fluid flow will likely be altered when the joint capsule is unsealed.  It is likely that contact area will be overestimated in this case because the cartilage will deform more due to exudation of fluid.  This hypothesis is supported by one study that examined the effect of time on contact area [149].  When a cadaveric femoral condyle was compressed against a piece of glass, contact area increased gradually with time (Figure 1-18), for example at approximately the 1 minute mark (which may represent the quickest print that can be determined experimentally) and the 10 minute mark (which may represent the cure time of the casting material) contact area measures were 0.6 and 0.9 cm2, respectively.  In most contact area assessments, the compressive loads are applied for much longer duration than would be expected for most daily activities, with dynamic measurements being the exception, therefore loading must be considered when interpreting results.  46  Figure 1-18: Effect of time on contact area measurement of a cadaveric femoral condyle against glass. Reprinted from Journal of Biomechanics, 10(4), Seedhom and Tsubuku, A technique for the study of contact between visco-elastic bodies with special reference to the patello-femoral joint, 253-260, 1977, with permission from Elsevier [149].  Another important experimental consideration is reporting the detail of the contact area assessment.  Most studies to date have not provided sufficient detail on size and location of contact areas.  As previously mentioned, qualitative and quantitative measures of contact area have been reported.  Qualitative photographs of contact patches provide useful information about the location and size of the contact as it changes with knee flexion.  They are limited because the area projection does not represent the three-dimensional topology of the surface.  Quantitative assessments of contact area are useful for data synthesis within and between studies.  They are limited because usually information about contact location is not provided.  Ideally both area and location would be quantified.  This is a challenge for the casting, measurement film and dynamic pressure measurement system techniques because an additional step of registration would be required since the measure is not made directly on the cartilage surface.  Dye staining, stereophotogrammetry and MRI provide enough information to measure both simultaneously (with additional processing steps).  The coordinates of the contact centroid have been reported on occasion [104,116,175]; however, these values must be reported in a useful coordinate system (such as the one used in kinematic assessment). 1.5.3 Ex Vivo and In Vivo Study Findings Contact area assessments have been carried out in vivo and ex vivo in order to study normal contact area patterns.  Similar to kinematics, there was considerable variability in magnitudes of contact area between studies both ex vivo and in vivo in the literature (Figure 1-19).  The ex vivo studies show that generally contact areas increase with flexion until approximately 90° of  47 tibiofemoral flexion and then they decrease.  This pattern is likely similar for in vivo studies; however, currently there are very few data above 60°, with just one study reporting data between 60° and 90° [172] and one between 90° and 140° [170]; this is due to the size constraints of closed bore MRI scanners.  From the few data that are available it is possible that contact areas also decrease after 90° in vivo.  There is also limited information about the location of contact areas on the patella. Several ex vivo studies have shown that contact areas migrate proximally with knee flexion [12,14,92,93,124,152] (Figure 1-20).  One of these studies tracked the location of the contact area centroid, which provides a useful quantitative measure of contact area migration with knee flexion [14].  There are limited in vivo data about contact area locations, although, presumably these could be determined from the same scans acquired to assess contact areas.  Of the two studies that did provide a measure of location one provided just an example [175] and the other showed results only for full extension and 40° of knee flexion [116] (Figure 1-21).  Figure 1-19: Summary of patellofemoral contact areas reported in the literature for ex vivo (above) and in vivo (below) studies.  48  lateral medial proximal distal  Figure 1-20: Patellar contact area patterns ex vivo. Reprinted from the Journal of Biomechanical Engineering, 105, In-vitro measurement of static pressure distribution in synovial joints-part II: retropatellar surface,  Ahmed et al, 1983,with permission from ASME [12].    Figure 1-21: Contact area patterns in vivo. Left: Example of contact area determined using stereophotogrammetry at an unspecified angle.  Reprinted from Journal of Biomechanics, 42(16), Differences in patellofemoral contact mechanics associated with patellofemoral pain syndrome, Connolly et al, 2802-7, 2009 with permission from Elsevier [175]. Right: An example of contact area at full extension and 40° of flexion. Reprinted from Arthroscopy, 25(11), Shin et al, Three-dimensional in vivo patellofemoral kinematics and contact area of anterior cruciate ligament- deficient and -reconstructed subjects using magnetic resonance imaging, 1214-23, 2009, with permission from Elsevier [116].    49 1.5.4 Summary It is clear from the literature that further validation of contact area assessments from MRI is required.  In particular, each the MRI sequences should be validated individually.  Further, the location of the contact area on to the patella should also be quantified to fully characterize contact areas. 1.6 Patellofemoral Joint Models There are a limited number of computational models that have been developed for the patellofemoral joint.  While two dimensional models have been useful to estimate contact mechanics in the sagittal [177,178] and axial [179] planes, three-dimensional models are required to adequately describe contact mechanics, kinematics and loads at the patellofemoral joint (Table 1-6) in normal individuals, in individuals with joint disease and in studies of treatment interventions.  In this section three-dimensional models will be compared in terms of the output, the platform used, the inputs required and the validation.  Areas in which the clinical relevance of models can be improved, will be identified.                50 Table 1-6: Summary of computational models in the literature. Kinematics refers to both patellofemoral and tibiofemoral kinematics unless explicitly stated.  Author Model Type Input Output  Hirokawa 1991 [180]  Multibody  Bone geometry Tissue material properties  Kinematics Forces (contact, patellar ligament) Contact Stress  Hefzy 1993 [181] Multibody  Bone geometry  Kinematics Contact points (1 medial, 1 lateral) Ratios of force (contact and patellar ligament to quadriceps)  Heegard 1995 [84] FEM Bone geometry Tissue material properties Contact Stress Contact Area Force (patellar ligament) Bone stress  Kwak 2000 [182] Multibody Bone geometry Quadriceps loads Tissue material properties Kinematics Contact stress Contact Area  Caruntu 2004 [183] Multibody Bone geometry Quadriceps load Tissue material properties  Kinematics Force (contact, ACL, PCL) Elias 2004 [184] Multibody  Bone geometry Kinematics Quadriceps load Tissue material properties  Contact stress Centre of pressure Contact force Mesfar 2005 [185] FEM Quadriceps load Tissue Material properties Kinematics Contact Area Force (contact, patellar ligament, ACL, PCL)  Besier 2005 [186] FEM Bone geometry Kinematics Muscle forces (from  EMG-driven musculoskeletal model) Tissue material properties  Cartilage contact stress Contact Area Cartilage shear stress Cartilage Hydrostatic Pressure Fernandez 2008 [187] FEM Bone geometry Quadriceps load Tissue material properties Tibiofemoral kinematics Kinematics Contact stress Contact Area Force (patellar ligament, contact)     51 1.6.1 Platform In general, two types of computational models have been developed at the patellofemoral joint: finite element models (FEM) [84,185-187] and multibody models [180-183,188].  With FEM it is possible to predict stresses throughout all of the tissues included in the model; however, in order to do so the tissues must be modelled according to correct geometry and tissue properties, which can be challenging because of inherent variability between individuals (therefore generalized models may not be useful) and the complexity of anatomical tissues (the viscoelastic behaviour of cartilage, in particular).  Multibody models have been created using numerous types of governing equations and theories.  Discrete element analysis [188], Hertzian elastic theory [180] and proximity (or overlap) analysis [182,183] have been used to predict contact stresses.  Euler angles [180,182,183] and the Joint Coordinate System [181] have been used to estimate kinematics.  Newtonian static equilibrium equations [182,188] or equations of motion [183] have been used to calculate forces and moments. Several groups have highlighted the importance of being able to create these models in a patient specific manner in order to capture the inherent variability between individuals [184,186,187]. 1.6.2 Inputs Bone geometry, quadriceps load, tissue material properties and kinematics have all been used as inputs for patellofemoral models (Table 1-6). The geometry of the models is obtained from cadaver or human subject data.  While some models aim to be patient specific [184,186,187], others are meant to be generalized models.  The geometry of the bones themselves has come from CT [84],  MRI [186,187], casts [180] and digitizing the bone surface [182,183] from which anatomical landmarks important for tendon and ligament insertions could be determined.  Cartilage geometry has either been determined from MRI [186,187] or prescribed a thickness of 2mm [183], 5mm [182,188] or an undisclosed valued [84,185].  All models also include the patellar tendon and the quadriceps tendon.  Each has been described using a single line of action [181,183,188] or multiple lines of action [84,180,182,185-187].  Other tissues that are sometimes modelled include the joint capsule [188], the medial and lateral collateral and the anterior and posterior cruciate ligaments [183,185] and  the medial and lateral patellofemoral ligament [185]. Tissue properties must be prescribed in FEM and multibody models (depending on the governing equations).  Bone is modelled as a rigid body [180,185,186] or according to experimental bone properties [84].  Cartilage has been modelled as a linear elastic material [84,186,187].  The elastic modulus of cartilage has been defined between 2 and 60 MPa [84,182-187] and Poisson’s ratio  52 has been defined as 0.3 [180] and 0.45-0.47 [84,182-187].  These values are assigned based on experimental findings in the literature.  All of these models have simplified the viscoelastic behaviour of cartilage.  Although it is often argued that most daily activities are carried out at a rate of less than 0.1 Hz (such as walking)[186], loading rates at heel strike can be up to 400 Hz [189] and clearly the viscoelastic nature of cartilage will be of importance in this case.  Tendons and ligaments have been modelled as tension-only elements with stiffness between 300 and 5000 N/mm [180,182,186,187]. Quadriceps loads have been assigned as single resultant forces [181,183,184,187] or forces of the individual quadriceps muscles [84,180,182,185,186].  Some studies have not explicitly defined a load magnitude and instead have used ratios of quadriceps tendon load to patellar ligament load based on either data from the literature [184] or as an unknown in the governing equation [181]. Musculoskeletal models have also been used to estimate quadriceps loads for input into the patellofemoral joint model [186,187]. One group used patient specific EMG driven musculoskeletal models to predict forces in the four individual quadriceps muscles [165], while another used a generalized model, and therefore was subject independent [187].  Many other groups use forces from cadaver studies or the literature [84,180,183,185] or are not specified [182].  The total quadriceps forces used have had a general range of 40 to 800 N with the exception of one study that used loads up to 2700 N [187].  The lower end of the scale is similar to loads that would be applied in cadaver studies or during activities such as walking [13] and the higher end represents loads estimated from musculoskeletal models [187].  When the quadriceps loads have been distributed between components of the quadriceps muscle group, PCSA and EMG data have been used to determine ratios as discussed in Section 1.4.4.1.  When kinematics are required as an input parameter they are calculated using MRI [186], fluoroscopy [187] or motion tracking from cadaver studies [184]. 1.6.3 Validation The validation of computational patellofemoral models to date has been limited.  They have been validated by comparing to experimental results from ex vivo [84,180-182,184] and in vivo [186] studies.  Ideally all the output parameters should be validated, but in practice this has not occurred to date.  Kinematic output from cadaver tests has been used to validate models and agreement has been shown to be quite good (however, this obviously does not make sense for kinematics-driven models). Errors were reported as 1.05° for rotations and 0.43mm for translation in one study based on agreement with results from 6 cadaver specimens [182] and 0.5° for rotations and 0.5mm for translations in another study based on agreement with results from two cadaver specimens [84].  53 Since these models also predict contact stresses, contact forces and patellar ligament forces, ideally these parameters would also be validated.  For kinematics-driven models [184,186], contact parameters were used for validation.  One study compared computational results to experimental results of 4 cadavers using a resistive dynamic pressure measurement system and found that the force ratio between patellofemoral compartments correlated well (r2=0.79); however, contact stress profiles did not correspond as well from visual assessment [184].  Another study validated just the contact area component of the output and found agreement to be 2.3% [186] to 5% [190] with measures of contact area obtained from MRI.  The study that found 5% agreement also visually displayed agreement of the areas (Figure 1-22).  The limited validation of the models to date indicates that more work is required in this area.  Currently, the tools exist to validate kinematics, contact areas and contact stresses quantitatively ex vivo and kinematics and contact areas quantitatively in vivo, therefore, a rigorous validation of a model would include quantitative comparisons to ex vivo and in vivo results.  Figure 1-22: MRI vs contact areas estimated by model. Gradient scale indicates pressures predicted from the model, black line indicates the in vivo MRI-based assessment of contact area. Reprinted from Journal of Orthopaedic Research, 26(12), Besier et al, The influence of femoral internal and external rotatin on cartilage stresses within the patellofemoral joint, 1627-35, 2008, with permission from John Wiley and Sons [190].  1.6.4 Study Findings Using Models Models have seldom been used to answer clinical questions; however, the few studies that have employed models examined the effect of changes in Q-angle [184], quadriceps load [185], quadriceps load distribution [191] and femoral internal and external rotation [190].  The Q-angle study showed  54 that when Q-Angle was increased the percentage force in the lateral compartment also increased; however, the results were not consistent in all specimens and varied from experimentally obtained results [184].  It has also been shown that by increasing quadriceps load, ACL, patellar tendon and contact forces and areas increase [185], which is not unexpected.  Load distribution was also studied based on EMG-based distributions and PCSA distributions and this study found that the EMG-based distributions decreased the lateral tilt and lateral rotation moments applied to the patella, while mean and maximum contact stresses were similar [191].  One study using the patient specific, EMG driven model studied the effect of femoral internal and external rotation in models based on 16 healthy volunteers [190].  It found that the effects of rotation on stress were not consistent across all subjects; however, in 75% of subjects an external rotation of 15° increased shear stress by 10% or more.  Other than this finding the results were variable.  These findings highlight the complexity of modelling the patellofemoral joint, even when patient specific modelling is carried out.  More complete validation of many of the models is required before they can be used to answer clinically relevant research questions relating to their output measures. 1.6.5 Summary Modelling the patellofemoral joint is a difficult task and the applicability of these models in answering clinical questions is uncertain due to the time intensive nature of modelling and the great variability between individuals.  Further, it is not clear if complex models are currently required because many simple research questions have yet to be answered in terms of characterization of normal patellofemoral joint mechanics. 1.7 Thesis Objectives and Research Questions The aim of this thesis was to develop and characterize in vivo assessments of three-dimensional patellar kinematics and patellofemoral contact areas that can be used in studies of patellofemoral OA. There currently exists no validated tool to assess three-dimensional kinematics and contact areas in vivo simultaneously.  It is critical that simultaneous measurements be made in order to understand the relationship between the two parameters and to understand how mechanical changes relate to patient symptoms.  While patellar kinematics can be measured in vivo using various different techniques (Section 1.4.3), contact areas can only be measured using MRI.  Therefore, in order to measure kinematics and contact areas simultaneously, MRI-based assessment measures must be used. Throughout this introductory chapter, summary sections have identified several areas in which improvements were required in the areas of kinematic, contact area and OA assessments.  While it is outside the scope of this thesis to address all of these matters, the most important ones will be  55 addressed.  To this end, this thesis is divided into three phases (Figure 1-23).  Phase 1 is a study of kinematics, Phase 2 is a study of articular cartilage contact areas and Phase 3 is the integration of kinematics and contact area measures. Phase 1: Kinematics  Phase 1 includes the characterization and implementation of a method of assessing three- dimensional kinematics in vivo using a sequential static pose approach that has been developed previously by our research group.  The method has been rigorously validated [102,132], making it distinct from other methodologies validated to date (Table 1-3).  Three studies of kinematics were carried out, the first two were characterization studies and the last was an implementation study. Study 1: ‘Does a single surrogate marker of three-dimensional patellar kinematics adequately represent the pattern of patellar kinematics over a range of knee flexion angles?’ Rationale:  Two-dimensional patellar alignment assessed from skyline radiographs at a single angle of knee flexion is the most widely used measure of mechanics in clinical studies.   In vivo assessments of three-dimensional patellar kinematics over a range of knee flexion angles provide a better description of joint mechanics but are time consuming and expensive and therefore it may be difficult to apply this technique in large scale clinical studies.  It is possible that a single measure of three-dimensional patellar alignment (akin to the two-dimensional measure used to describe patellar alignment in the OA literature), may be sufficient to describe patellar kinematics. Study 2: ‘What is the effect of load magnitude on in vivo three-dimensional patellar kinematics?’ Rationale: Joint loading affects two dimensional patellar kinematic results [129] (Section 1.4.4); however, the effect of loading on the critical out-of-plane motions has not been assessed because it has never been studied in three-dimensions.  This is important for comparing results between studies and prescribing loading conditions in future studies. Study 3: ‘What is the effect of a patellofemoral brace on three-dimensional patellar kinematics in individuals with radiographic lateral patellofemoral OA?’ Rationale: Mechanical treatment strategies, such as bracing, are often prescribed to individuals with patellofemoral OA.  However, the magnitude of mechanical change required to reduce patient symptoms has not been assessed.  This information is important for evaluating the clinical success of treatment strategies and for developing new treatment strategies.     56 Phase 2: Contact Area Study 4: ‘The aim of this study was to develop and validate a method of assessing patellofemoral articular cartilage contact areas using MRI that can be integrated into the method of assessing three- dimensional patellar kinematics developed previously.’ Rationale: No well validated protocol for assessing patellofemoral joint contact area simultaneously with three-dimensional patellar kinematics has been reported in the literature to date.  Assessments of kinematics are essential because so many treatment strategies for patellofemoral joint disease focus of correcting tracking with the underlying goal of correcting contact mechanics; therefore, having a validated measure is required for treatment development and evaluation.  Contact areas provide a more direct measure of load transmission through the joint and may be more closely related to patient symptoms.  A hybrid assessment is therefore desirable. Phase 3: Kinematics and Contact Area Study 5: ‘The aim of this study was to develop and validate a simple, patient specific, kinematics- driven, computationally inexpensive multibody model of the patellofemoral joint to predict contact areas that is appropriate for use in clinical studies and to assess its sensitivity to kinematic inputs.’ Rationale: There are situations in which direct assessments of contact area are not possible, such as assessments of kinematics using bi-planar radiography and in clinical studies of patellofemoral joint disease where joint loading must be kept to a minimum.  In these situations, a simple estimate of contact area would be desirable.  Further, it is unclear how such estimates are influenced by the input parameters.  57 Phase 1: Kinematics Phase 2: Contact Area Study 2: Effect of Loading Study 1: Surrogate Marker Study 3: Bracing in OA Study 4: Development and Validation Phase 3: Kinematics and Contact Area Study 5: Integration and Modelling  Figure 1-23: Overview of thesis studies.  58 2 Surrogate Markers of Kinematics1  SYNOPSIS: Patellar alignment, assessed from skyline radiographs at a single knee flexion angle, is most often used to describe mechanics in studies of patellofemoral OA. However, it is not clear if this description is adequate.  Therefore, the aim of this study was to determine if a single surrogate marker of three-dimensional patellar kinematics sufficiently describes patterns of patellar kinematics over a range of knee flexion angles.  Again, the method of assessing three-dimensional patellar kinematics validated previously by our group was used.  2.1 Introduction Disorders of the patellofemoral joint are prevalent in the population, with the patellofemoral joint involved in half of all knee OA cases [1] and patellofemoral pain syndrome affecting 25% of the population [192-194].  Abnormal patellar kinematics is thought to contribute to the onset and progression of patellofemoral disorders.  One possible reason for this is that abnormal kinematics disrupts the pattern of force transmission through the joint.  Since patellar kinematics is difficult to measure, particularly in vivo, the relationship between patellar kinematics and disease onset and progression is not well understood. Patellar position and orientation have been assessed in two-dimensions (axial plane) at a single angle or sequential angles of knee flexion using radiographs, computed tomography (CT) or magnetic resonance imaging (MRI) both statically [195-198] and dynamically [65,121,199,200].  Parameters such as congruence angle [196], lateral displacement and tilt (Figure 1-10) [195] are the most  1  A version of Chapter 2 has been published. McWalter, E.J., MacIntyre, N.J., Cibere, J., Wilson, D.R. (2010) A single measure of patellar kinematics is an inadequate surrogate marker for patterns of three-dimensional kinematics in healthy knees. The Knee, 17(135-140).  doi:10.1016/j.knee.2009.08.001 Additional detail in the methods section has been published.  McWalter, E.J., Hunter, D.J. and Wilson, D.J. (2010) The effect of load magnitude on three-dimensional patellar kinematics in vivo. Journal of Biomechanics, 43:1890-1897. doi:10.1016/j.jbiomech.2010.03.027  59 common measures of patellar position in the axial plane.  These parameters are often used in research studies to quantify malalignment in patients with patellofemoral disorders.  For example, a recent study found that static patellar malalignment (displacement and tilt), measured radiographically at 30- 40° of weightbearing flexion, was associated with radiographic patellofemoral OA progression [2,3]. Another study, using a dynamic MRI-based assessment from 0° to 60° of flexion, found greater lateral patellar translation and greater lateral patellar tilt in subjects with patellofemoral pain as compared to controls [65].  The two-dimensional methods of analysis have been successful at describing malalignment in the patient population; however, there are four other components of motion which cannot be described using these methods. In an effort to capture all components of patellar position, patellar kinematics has also been assessed in three dimensions using MRI-based methods [102,104,105].  The three-dimensional methods have been used to study patients with patellofemoral disorders.  Differences in three- dimensional kinematic parameters were found between subjects with varus and valgus tibiofemoral malalignment and knee OA [201] and between patients with clinical signs of malalignment and patellofemoral pain and matched controls [128,134].  The results of studies to date suggest that additional, useful information is obtained by doing three-dimensional assessments; however, these methods require substantially more imaging and analysis time than two-dimensional methods. As such, their application in large research studies or as a diagnostic or treatment planning tools may not be feasible. It is not clear whether the additional imaging and analysis time required to measure three- dimensional patellar kinematics over a range of knee flexion angles is essential to characterize the pattern of kinematics of any given subject.  It is possible that one measurement of position and orientation at one angle of knee flexion provides an adequate surrogate marker of patellar kinematics through the range of knee flexion. Such a surrogate marker would be clinically very useful.  An example of a widely used surrogate maker in musculoskeletal research is the measurement of bone mineral content from duel-energy X-ray absorptiometry (DXA) to predict fracture risk [202,203]. While the type of surrogate marker proposed here varies from one used to predict disease, it would still need to predict a large portion of the variance in kinematics over a range of flexion angles.  A strong marker is commonly thought of as one that predicts a minimum of 70% of the variance in pattern [204]; for example bone mineral content from DXA predicts at least 85% of the variance in proximal femur failure load [205]. In this study we will require a surrogate marker to meet this criterion for all 6 kinematic parameters.  Therefore, we answered the question: Can a single static  60 measure of three-dimensional patellar kinematics provide a surrogate marker for three-dimensional patellar kinematics over a range of flexion angles? 2.2 Methods 2.2.1 Subjects Asymptomatic subjects were selected from a database of participants with available three- dimensional patellar kinematic data. All subjects had undergone assessments of three-dimensional patellar kinematics using a sequential static position MRI-based method [102,132] while participating in previous studies conducted by our group [134,206,207].  The subjects had no history of knee injury, pathology or pain. Subjects who could not undergo MRI were excluded from their respective studies. Institutional ethical approval was obtained and all participants gave informed consent. 2.2.2 Image Acquisition For each subject, one high-resolution and six low-resolution loaded MR scans, required to assess three-dimensional patellar kinematics, were acquired using a 1.5 T scanner (Genesis-Signa, General Electric, Waukesha, USA) or a 3.0T scanner (Intera, Phillips, Eindhoven, Netherlands).  The high-resolution scan was acquired in the sagittal plane (Table 2-1) with the subject’s knee in a relaxed position.  Six quick, low-resolution, loaded sagittal scans (Figure 2-1) were acquired over a range of flexion angles (approximately 0, 10, 20, 30, 40 and 50°, positioned in this order using a standard goniometer).  To load the knee, the subject placed his or her foot on a plate attached to a custom designed loading rig (Figure 2-1 and Appendix C).  The knee was positioned at the desired knee flexion angle and the prescribed load was released against the subject’s foot.  Applied loads during the low-resolution scans were between 80 and 152 N (this creates an estimated external knee moment of between 10 and 40 Nm over the range of flexion, which is similar to gait (Appendix C)).  The subject was trained to maintain his or her knee flexion position using the quadriceps muscle group. The MRI data collection was carried in accordance with a standardized protocol by one of four trained experimenters.       61 Table 2-1: T1-weighted MRI sequence parameters for the high- and low-resolution scans. Reprinted from Journal of Biomechanics, 43, McWalter et al, The effect of load magnitude on three- dimensional patellar kinematics in vivo, 1890-1897, 2010 with permission from Elsevier [207].  Parameter High-resolution, Unloaded Low-Resolution, Loaded In-plane resolution 0.586 mm 1.25 mm Field of view 300 mm 320 mm Slice separation 2 mm 7 mm Matrix size 512 x 512 256 x 256 Repetition Time 360 ms 307 ms Echo Time 10.0 ms 6.2 ms Flip Angle 90° 90° Scan time 9 min 10 s 34 s MRI Coil Knee Body  0, 15 or 30% BW  Figure 2-1: Subject positioned in MRI scanner for a low-resolution, loaded MRI scans. The subject’s foot was positioned on the custom-designed loading rig and the knee was positioned at the prescribed flexion angle by the research assistant using foam triangles.  The prescribed load (0%, 15% or 30% BW) was applied to the loading apparatus and the subject was instructed to maintain the leg positioning using primarily the quadriceps muscles.  Reprinted from Journal of Biomechanics, 43, McWalter et al, The effect of load magnitude on three-dimensional patellar kinematics in vivo, 1890-1897, 2010 with permission from Elsevier [207].  2.2.3 Three-dimensional Patellar Kinematic Analysis  The MRI scans were analyzed according to a validated method (Appendix D) [102,132] by one of three trained experimenters.  Briefly, patient specific bone models of the femur, tibia and patella were created from the high-resolution MRI scans by segmenting each bone in a slice-by-slice manner using a seed growing technique with manual correction in the Image Edit module of Analyze 8.1 (Analyze Direct Inc, Overland Park, KS, USA).  The seed growing tool uses a connected component analysis algorithm which requires the user to identify a pixel in the desired region and to  62 set a threshold value.  The region is defined as all pixels connected to the ‘seed’ within the specified threshold.  Models were created from the segmented data using the Adapt/Deform Module in Analyze.  Anatomical axes were then assigned to each bone model.  Origins of the femoral, tibial and patellar coordinate systems were defined as the most proximal point of the trochlea (Figure 2-2 D), the most proximal point of the medial intercondylar eminence (Figure 2-2 E) and the most posterior point on the axial mid-slice (Figure 2-2 C), respectively.  The direction of the flexion axis for each bone was determined using anatomical landmarks (Figure 2-2 A-C).  The direction of the long axis was defined from the origin of the coordinate system to the centroid of the proximal femoral shaft and distal tibial shaft for the femur and tibia and as the line connecting the proximal and distal points on the sagittal midslice for the patella (Figure 2-2 F).  The third axis was orthogonal to the other two. Orthogonal anatomical coordinate systems were created using the long axis, the third axis and a flexion axis that was the cross product of the third and the long axis.  Positive directions for all coordinate systems were proximal, anterior, and lateral.  Next, bone contours were created from the low-resolution MRI scans by segmenting in the same manner as for the bone models.  The bone models were registered to the contours using an Iterative Closest Points (ICP) algorithm [208] custom written in Matlab (The MathWorks, Natick, MA, USA).  Finally, three-dimensional patellar attitude and position (flexion, spin, and tilt; proximal, lateral, and anterior translation) were calculated using a modified Joint Coordinate System convention [102,112], which is a coordinate system defined as the flexion axis of the femur (e1), the long axis of the patella (e3) and a third, floating axis (e2) that is perpendicular to e1 and e3 (Figure 2-3 and Figure 2-4).  Tibiofemoral flexion was assessed similarly. The mean error of this method, assessed using RSA analysis, is less that 1.02° for spin and tilt and less that 0.88 mm for translations [102].  The intra-subject repeatability, defined as the mean standard deviation of 4 trials in 3 subjects, is less than 1.04° for spin and tilt and less than 0.81 mm for translations [132].  63 Femoral Landmarks Tibial Landmarks Patellar Landmarks Posterior point – lateral condyle Posterior point – medial condyle Proximal point of trochlea (femoral origin) Posterior point – medial plateau Posterior point – lateral plateau Proximal point medial eminence (tibial origin) Posterior point - axial midslice (patellar origin) Lateral point – axial midslice Distal point – sagittal midslice Proximal point – sagittal midslice A B C D E F  Figure 2-2: Anatomical landmarks identified on MRI images. The anatomical landmarks used to create the femoral coordinate system are displayed in panels A and D. The femoral origin is the most proximal point of the intercondylar notch defined, in the sagittal plane, and the direction of the flexion axis is defined using the most posterior points of the medial and lateral condyles in the axial slice containing the origin.  The anatomical landmarks used to create the tibial coordinate system are displayed in panels B and E. The tibial origin is the most proximal point of the medial intercondylar eminence, defined in the sagittal plane, and the direction of the flexion axis is defined using the most posterior medial and lateral points in the axial slice containing the most superior point of the fibula. The anatomical landmarks used to create the patellar coordinate system are displayed in panels C and F. The patellar origin is the most posterior point on the axial midslice. The direction of the flexion axis is defined using the origin and the most lateral point on the axial midslice. The direction of the long axis is defined using the most proximal and distal points on the sagittal midslice.  Reprinted from Journal of Biomechanics, 43, McWalter et al, The effect of load magnitude on three-dimensional patellar kinematics in vivo, 1890-1897, 2010 with permission from Elsevier [207].     64 Spin Flexion Tilt e2 Lp=e3 Ff=e1  Figure 2-3: Illustration of the modified Joint Coordinate System. The coordinate system consists of the femoral flexion axis (e1), the patellar long axis (e3) and a third axis (e2) orthogonal to e1 and e3.  The highlighted rotational parameters are those reported in the present study. Reprinted from Journal of Biomechanics, 43, McWalter et al, The effect of load magnitude on three- dimensional patellar kinematics in vivo, 1890-1897, 2010 with permission from Elsevier [207].   Figure 2-4: Three-dimensional patellar kinematic parameters. Reprinted from Journal of Biomechanics, 43, McWalter et al, The effect of load magnitude on three- dimensional patellar kinematics in vivo, 1890-1897, 2010 with permission from Elsevier [207].  65 2.2.4 Statistical Analysis We used the results from our kinematic analysis to determine whether a surrogate marker (patellar alignment at one position) could predict patterns of three-dimensional patellar kinematics. We defined the surrogate marker as the static measurement of three-dimensional patellar alignment at 30º of knee flexion.  Therefore, there were six surrogate markers, one corresponding to each patellar parameter (Figure 2-4).  This knee flexion angle was chosen for the surrogate marker because it is commonly used to obtain skyline radiographs of the patella from which patellar alignment is measured [34], data were available at this angle in all subjects and it was in the mid-range of the flexion angles assessed.  To define the pattern of three-dimensional patellar kinematics, we fit a linear least-squares line to the graph of kinematic parameter versus knee flexion for each subject. This yielded 6 lines per subject corresponding to the 6 patellar kinematic parameters. Kinematic pattern was defined as the slope of the least-squares fit line, which described the rate and direction of change of the kinematic parameter with knee flexion.  We determined whether the surrogate marker could predict patterns of three-dimensional patellar kinematic parameters using a regression model. Quadratic terms were included in the model when significant.  To assess the fit of these models we used the coefficient of determination.  All statistical analysis was carried out using Stata (StataCorp LP, College Station, TX, USA). 2.3 Results Data for forty subjects (29 male, 11 female, mean age 28.6 ± 8.7 years) was available in the database.  Thirty subjects were originally recruited from a military population [134,206] and ten were originally recruited from a university community [209].  The mean knee flexion angle for the surrogate marker was 28.7 ±3.2º. Using regression models, we found statistically significant relationships of the single measure at 30° knee flexion (surrogate marker) with the patellar pattern from full extension to 45° knee flexion for four of the six parameters measured, including patellar flexion, proximal translation, spin and anterior translation. We found no statistically significant relationship of the surrogate marker with the patellar pattern for patellar tilt or lateral translation. Patellar flexion: A linear regression model showed that patellar flexion at 30° knee flexion predicted 26% of the variance in the pattern of patellar flexion from full extension to 45° of knee flexion (p < 0.001). The model showed that when the patellae were in a greater angle of flexion at 30° of knee flexion, they flexed at a greater rate over the range of knee flexion (Figure 2-5).  66 Patellar Tilt: There was no relationship between the patellar tilt at 30° of knee flexion and pattern of patellar tilt over the range of knee flexion (R2 = 0.00, p = 0.867) (Figure 2-5). Patellar Spin: Using a quadratic regression model, we found that patellar spin at 30° of knee flexion predicted 27% of the variance in the pattern of patellar spin over the range of knee flexion (p = 0.003).  The model for patellar spin showed that when the surrogate marker was less than 2° of internal spin, the patellae spun internally at a relatively constant rate of 0.1° per degree of knee flexion, but when the surrogate marker was greater than 2° of internal spin, the rate of internal spin with knee flexion increased (Figure 2-5). Proximal translation: A linear regression model showed that proximal patellar translation at 30° knee flexion predicted 11% of the variance in the pattern of proximal patellar translation over the range of knee flexion (p = 0.037). The model showed that when the patellae were in a more proximal position at 30 ° of knee flexion, they translated distally at a greater rate over the range of knee flexion (Figure 2-6). Lateral Translation: There was no relationship between lateral translation at 30° of knee flexion and pattern of lateral translation over the range of knee flexion (R2 = 0.00, p = 0.698) (Figure 2-6). Anterior Translation: Using a quadratic regression model, we found that anterior translation at 30° of knee flexion predicted 39% of the variance in the pattern of anterior translation over the range of knee flexion (p < 0.001).  The model for anterior translation showed that when the surrogate marker was in a more anterior position, the patellae translated posteriorly at a lesser rate (Figure 2-6).            67 y = 0.007x + 0.597 R2 = 0.26 0.4 0.5 0.6 0.7 0.8 0.9 -15 -10 -5 0 5 10 15 20 25 Patellar Flexion (degrees) at 30 Degrees of Knee Flexion Pa tte rn  o f P at ell ar  F lex io n (d eg re es /d eg re e o f k ne e f lex io n) flexion flexionextension  y = 0.001x + 0.034 R2 = 0.00 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 -5 0 5 10 15 20 Patellar Tilt (degrees) at 30 Degrees of Knee Flexion Pa tte rn  o f P at ell ar  T ilt (d eg re es /d eg re e o f k ne e f lex io n) medial mediallateral y = 0.002x2 + 0.004x + 0.060 R2 = 0.27 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 -8 -6 -4 -2 0 2 4 6 8 10 Patellar Spin (degrees) at 30 Degrees of Knee Flexion Pa tte rn  o f P at ell ar  S pi n (d eg re es /d eg re e o f k ne e f lex io n) internal internalexternal  Figure 2-5: Regression line fits describing the relationship between the surrogate marker and pattern of rotation (flexion, spin and tilt). The surrogate marker is the measurement of the patellar rotation at 30° of knee flexion for each individual. The pattern of patellar rotation is the slope of the linear least-squares fit line for each individual’s values of patellar rotation versus knee flexion.  A positive pattern indicates patellar flexion, internal spin or medial tilt and a negative pattern indicates patellar extension, external spin or lateral tilt.  Dashed lines indicate the 95% confidence intervals.  Reprinted from The Knee, 17, McWalter et al, A single measure of patellar kinematics is an inadequate surrogate marker for patterns of three-dimensional kinematics in healthy knees, 135-140, 2010 with permission from Elsevier [108].  68 y = -0.005x - 0.583 R2 = 0.11 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -5 0 5 10 15 20 25 Proximal Translation (mm) at 30 Degrees of Knee Flexion Pa tte rn  o f P ro xim al Tr an sla tio n (m m /d eg re e o f k ne e f lex io n) proximal proximaldistal  y = -0.002x - 0.033 R2 = 0.00 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 -10 -8 -6 -4 -2 0 2 4 6 Lateral Translation (mm) at 30 Degrees of Knee Flexion Pa tte rn  o f L at er al Tr an sla tio n (m m /d eg re e o f k ne e f lex io n) lateral lateralmedial  y = -0.001x2 + 0.080x - 1.341 R2 = 0.39 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 15 20 25 30 35 40 Anterior Translation (mm) at 30 Degrees of Knee Flexion Pa tte rn  o f A nt er io r T ra ns lat io n (m m /d eg re e o f k ne e f lex io n) anterior anteriorposterior  Figure 2-6: Regression line fits describing the relationship between the surrogate marker and pattern of translation (proximal, lateral and anterior). The surrogate marker is the measurement of the translation at 30° of knee flexion for each individual.  The pattern of translation is the slope of the linear least-squares fit line for each individual’s values of translation versus knee flexion.  A positive pattern indicates a proximal, lateral or anterior translation and a negative pattern indicates a distal, medial or posterior translation. Dashed lines indicate the 95% confidence intervals. Reprinted from The Knee, 17, McWalter et al, A single measure of patellar kinematics is an inadequate surrogate marker for patterns of three-dimensional kinematics in healthy knees, 135-140, 2010 with permission from Elsevier [108].  69 2.4 Discussion We have shown that the surrogate markers of patellar measurements at 30° knee flexion described less than 39% of the variance in patterns of patellar flexion, spin, proximal translation and anterior translation and did not describe patterns of patellar tilt or lateral translation.  Since patterns of patellar tilt and lateral translation could not be adequately described by the surrogate markers and only a portion of the variance was described for other parameters, three-dimensional patellar kinematics should be optimally assessed over a range of knee flexion angles. It is possible that the surrogate marker did not predict kinematic parameters because of the inter-individual differences inherent in three-dimensional patellar kinematic data, the particular angle chosen for the surrogate marker and the dependent variables used in the regression model.  Large inter-individual differences, described as the standard deviation of the average slope (pattern), of up to 0.17° and 0.10 mm per degree of knee flexion were observed in this study and this is consistent with differences found previously in normal subjects using a dynamic method [118].  Even for parameters where a portion of the variance was predicted by the surrogate marker, many subjects did not follow the prevailing patterns. For example, in two subjects with an anterior patellar position of 24 mm at 30º of knee flexion (surrogate marker angle), the rate of anterior translation was -0.3 mm per degree of knee flexion in one subject, while it was -0.05 mm per degree of knee flexion in the other.  Measuring patellar alignment at a different knee flexion angle may yield a more predictive surrogate marker; however, the angle chosen is consistent with clinical assessments of patellar malalignment made in two dimensions from a skyline radiograph.  Our regression model assumed that patterns of three-dimensional patellar kinematics were dependent on only one measure of patellar position.  In reality, kinematics are likely affected by soft tissue properties, bone geometry, muscle pull and attachment, gender, weight and clinical characteristics (varus/valgus malalignment, joint laxity, muscle strength, etc).  Further, in cases of diseased joints, structural changes and patient symptoms may also affect kinematics.  A surrogate marker that includes these measures may be better at predicting patterns of patellar kinematics. Interestingly, the surrogate marker did not predict patterns of patellar tilt or lateral translation, which are the parameters commonly studied in two-dimensional analyses in the axial plane.  This finding may explain why alignment differences in patellar tilt and lateral translation were not observed between subjects with patellofemoral pain syndrome and normal subjects in radiographic studies at one angle of knee flexion in the axial plane [210].  However, it should also be noted that when these same groups and parameters were studied, using both a dynamic kinematic method in the axial plane [65] and a sequential static kinematic method in three-dimensions [134], differences were  70 observed between patients and normals.  These contrasting findings suggest that it is important to assess kinematics over a range of knee flexion angles and that there may be differences in kinematics between the static and dynamic measures.  Further, the findings of the current study are not surprising from a biomechanical point of view.  Patterns of patellar tilt and lateral translation are influenced by muscle lines of action and soft tissue properties, in particular before the patella engages the trochlear groove (between 20 and 30° of knee flexion), and by bony geometry, after the patella engages the trochlear groove.  Therefore, it is not unexpected that the surrogate marker did not adequately predict patterns over a range of flexion angles. The overall patterns of three-dimensional patellar kinematics in normal subjects are similar to those reported by other groups using both sequential static [104] and dynamic [118] methods. Comparisons to the other sequential static method are most suitable in this case [104].  In that study, kinematics were assessed in 10 normal subjects between approximately -10° and 60° of knee flexion using random effects models.  Patellar spin and anterior translation were modelled linearly with slope or pattern values of 0.07891° per degree of knee flexion and 0.05180 mm per degree of knee flexion, respectively, which are comparable to the mean patterns (slope) of kinematics found in our study. Patellar tilt, proximal translation and lateral translation were modelled quadratically.  One possible reason for the difference in the model between this study and the present study (all linear at the subject level) was that most of the curvature was seen above 35°, in particular for tilt and lateral translation, and therefore may not have been captured by the data of the present study. Comparisons to measurements of kinematics made using the dynamic method are also useful [118] and in that study knee flexion angles from 1° to  44° were assessed and patterns were described as the mean value of the parameter for each knee flexion angle at 1° increments.  By qualitative comparison, patterns of patellar flexion, spin, proximal translation and anterior translation are similar to those in the current study, while patterns of tilt and lateral translation are different. The strengths of this study lie in the fact that our measurement method for assessing three- dimensional patellar kinematics has been well validated [102,132]. Both the variability between subjects observed in this population and the changes in parameters through the range of flexion are large relative to the errors in the measurements. To our knowledge, we are the first group to study the necessity for studying three-dimensional patellar kinematics over a range of flexion angles, which is important since the data collection and analysis procedure is time consuming and expensive.  A limitation of this study is that it employed a sequential static method of assessing three-dimensional patellar kinematics; therefore, the results may not necessarily be extended to assessments of three- dimensional patellar kinematics using dynamic methods.  Another limitation is that we chose to  71 describe patterns of kinematics with a linear fit because the physical interpretation of the slope and intercept of the linear model is easily comprehensible.  Initially, quadratic models were also fit to each subject’s kinematic data; however, no relationships were seen between the surrogate marker and pattern of slope or pattern of the quadratic coefficient for any parameter and therefore this analysis was not presented in the current study.  The lack of relationship with the quadratic model may be due to the small number of data points being fit (five or six) and the lack of an existence of a clear quadratic pattern in the data.  Finally, this study is an assessment of asymptomatic subjects and therefore the results cannot necessarily be extended to subjects with patellofemoral disorders. We found that patellar measurements at 30° knee flexion described a portion of the variance in patterns of patellar flexion, spin, proximal translation and anterior translation but not in patterns of patellar tilt or lateral translation measured from full extension to 45° of knee flexion. Therefore, the surrogate marker does not capture the full pattern of three-dimensional patellar kinematics in asymptomatic adults. As a result, three-dimensional assessments of patellar kinematics over a range of knee flexion angles are preferable to evaluate patellar malalignment.   72 3 Loading & Kinematics2  SYNOPSIS: This chapter examines the effect of load magnitude on three-dimensional patellar kinematics using a method validated previously by our group.  This study is the first to characterize the effect of load on patellar kinematics in three dimensions.  3.1 Introduction In vivo assessments of patellofemoral joint kinematics are useful when studying patellofemoral joint disease; however, substantial variability in normal patellar kinematics has been reported even between MRI-based methods (Figure 1-15) [107] and within studies (Chapter 2).  One possible reason for this is that the amount of load applied to the joint during assessment differs between protocols. Because one of a joint’s primary functions is to transmit load, ideally measurements of joint mechanics, including kinematics, should be made under physiological loading conditions.  This has been difficult to achieve in practice.   With fluoroscopy it is possible to assess patellar kinematics during activities such as a lunge.  Two methods have been validated tod date [101,103]; however, they have not yet been used to study patellofemoral joint disease and ionizing radiation is required. Upright, open configuration MRI systems have been used to assess patellar kinematics in three dimensions using sequential static methods [130] and in two dimensions continuously [65]. However, these upright, open MRI systems are not widely available and have low field strengths, limiting their applicability. It is likely that patellar kinematics will be affected by the amount of load applied by the quadriceps muscle, therefore when assessments of patellar kinematics are made using closed-bore MRI systems, which are readily available, a physiologic loading condition must be simulated.  In these systems, the type of loading task possible is limited by the size of the MRI bore and the position  2 A version of Chapter 3 has been published. McWalter, E.J., Hunter, D.J. and Wilson, D.J. (2010) The effect of load magnitude on three-dimensional patellar kinematics in vivo. Journal of Biomechanics, 43:1890-1897. doi:10.1016/j.jbiomech.2010.03.027  73 of the subject (most often supine).  When kinematics were assessed using three-dimensional continuous methods, the knee was flexed and extended repeatedly (open chain); extension was resisted by the weight of the shank [118] or by a small (34N) externally applied load [105].  When kinematics were assessed using three-dimensional sequential static methods an external axial load was applied through the foot with custom designed pedal systems (closed chain) [102,104,116]. These applied loads (80 to 152 N) typically must be lower than those experienced during daily activities due to the position of the subject.  It is unclear to what extent differences in these load magnitudes affect measures of three-dimensional patellar kinematics. To date, no study has assessed three-dimensional patellar kinematics under different prescribed loading conditions.  However, differences have been observed in two-dimensional patellar kinematics between a supine dynamic and an upright, weight-bearing dynamic loading condition in individuals with patellofemoral pain [211] and also a supine dynamic and unloaded static case in normal individuals [121].  Differences in kinematics in the supine, loaded, sequential static assessments have not been quantified. Therefore, we asked the research question: What is the effect of load magnitude on three-dimensional patellar kinematics in a closed-bore MRI scanner? 3.2 Methods 3.2.1 Subjects Ten normal subjects were recruited from a university community to participate in this study (4 female, 6 male, 34.3 ± 6.5 yrs, 70.6 ± 15.6 kg).  Subjects with history of knee pain, injury, OA or surgery or contraindication to MRI were excluded from the study. Institutional ethics board approval was obtained and each subject provided informed consent. 3.2.2 Image Acquisition and Kinematic Analysis Three-dimensional patellar kinematics were assessed at 0% BW (BW) load (no load), 15% BW load and 30% BW load using a sequential static, MRI-based method which has been validated previously by our group [102,132] and described in detail in Sections 2.2.2 and 2.2.3 and Appendix C and D. 3.2.3 Statistical Analysis  We tested the null hypothesis that there was no difference between the load levels for each three-dimensional patellar kinematic parameter (Figure 2-4) using hierarchical random-effects models (Appendix E).  Briefly, these models are the weighted average of maximum likelihood estimation fits  74 to each individual’s kinematic data.  Individuals who have fewer datapoints or who are further from the mean have less weight.  This type of model takes into consideration the correlation between data points at the subject level, thereby increasing the power of the statistical test.  This is not the case, for example with a repeated measures analysis of variance (ANOVA) where differences would be assessed at discrete angles of knee flexion.  Further, the repeated measures ANVOA is not appropriate because these data do not follow a randomized block design (we did not randomize the order of knee angles during data collection) and because of error in measuring knee flexion angle with a goniometer, data were not available at the same flexion angles in all individuals. In the present study, models were fit as a function of knee flexion according to the following equation: anglekneelevelloadlevelloadanglekneeanglekneey _*_*_*_*_* 43 2 210 βββββ ++++= where y is any one of the six kinematic quantities.  The quadratic (β2) and the load/knee angle interaction (β4) coefficients were included in the model when significantly different from zero (p<0.05).  Knee angle is a continuous variable, while load level is a discrete variable.  The knee flexion angles used to create the model were those calculated using the modified joint coordinate system, not those measured from the goniometer. 3.3 Results Increased loading changed patellar flexion, tilt, proximal translation and anterior translation significantly (Table 3-1).  There were no statistically significant differences for patellar spin or lateral translation (Figure 3-1 B, Figure 3-2 B and Table 3-1).  In this study group the 15% bodyweight loads ranged between 77 and 140 N, while the 30% BW loads ranged between 154 and 281 N.  Samples of raw results for select individuals can be found in Appendix F. Increasing the load to 30% BW reduced the rate of patellar flexion with knee flexion as compared to the 0% and 15% BW condition (Figure 3-1 A).  Both the β2 (quadratic) and β4 (load_level*knee_angle interaction) coefficients were significant and therefore included in this model.  The difference in the β4 coefficient, which indicates a difference in slope, was -0.075 °/° of knee flexion (95% Confidence Interval (CI): -0.111 to -0.040) between the 0% and 30% BW loads (p<0.001) and -0.053°/° of knee flexion (CI: -0.090 to -0.017) between the 15% and 30% BW loads (p=0.004). The patellae maintained approximately 10 degrees of medial tilt load through the range of knee flexion at 30% BW while they tilted laterally with knee flexion at 0% and 15% BW loads  75  76 (Figure 3-1 C).   The β2 coefficient was not significant and therefore not included in this model.  A significant difference in the β4 coefficient of 0.066°/° of knee flexion (CI: 0.012 to 0.121) was observed between the 0% and 30% BW loads (p=0.017) and of 0.057°/° of knee flexion (CI: -0.002 to 0.112) between the 15% and 30% BW loads (p=0.043).  The patellae were in a more proximal position at 30% BW load than at 0% BW or 15% BW load (Figure 3-2 C).  The β2 coefficient was significant and therefore included in this model, while the β4 coefficient was not.  The difference in the load_level (β3) coefficient was 0.64mm (CI: 0.16 to 1.13) between the 0% and 30% BW loads (p=0.010) and 0.70mm (CI: 0.22 to 1.19) between the 15% and 30% BW load (p=0.007). The patellae were in a more posterior position at 30% BW load than at 0% BW or 15% BW loads (Figure 3-2 C).  The β2 was significant and therefore included in the model, while the β4 coefficient was not. The difference in the β3 coefficient was -0.59mm (CI: -0.88 to -0.30) between the 0% and 30% BW loads (p<0.001) and -0.32mm (CI: -0.62 to -0.03) between the 15% and 30% BW load (p=0.029). Table 3-1: Coefficients (standard error) for each kinematic parameter’s linear hierarchical random-effects model. The following equation describes the model used:  .   anglekneelevelloadlevelloadanglekneeanglekneey _*_*4_*3 2_*2_*10 βββββ ++++= Quadratic (knee_angle2) and load_level x knee_angle interaction terms were included in the model when significant. NS = not significant, *=p<0.05, **p<0.01, ***p<0.001.  Reprinted from Journal of Biomechanics, 43, McWalter et al, The effect of load magnitude on three-dimensional patellar kinematics in vivo, 1890- 1897, 2010 with permission from Elsevier [207].   Kinematic Parameter  β0 β1 β2 β3  (difference between groups) β4 (difference between groups)   0% vs 15%  0% vs 30% 15% vs 30% 0% vs 15% 0% vs 30% 15% vs 30% Flexion  -5.13 *** (1.02)  0.757 *** (0.264) -0.0012 ** (0.0005) -0.56 (0.47) 0.08 (0.47) 0.64 (0.47) -0.022 (0.018) -0.075 *** (0.018) -0.053 *** (0.018) Spin  -5.61 *** (0.85)  0.129 *** (0.025) -0.0012 * (0.0005) 0.14 (0.36) -0.14 (0.36) -0.28 (0.36) NS NS NS Tilt  9.44 *** (2.02)  -0.072 * (0.030) NS 0.32 (0.74) 0.64 (0.72) 0.32 (0.72) 0.009 (0.027) 0.066 * (0.027) 0.057 * (0.028) Proximal Translation  24.20 *** (1.61)  -0.656 *** (0.024) 0.0015 *** (0.0004) -0.06 (0.25) 0.64 * (0.25) 0.70 ** (0.25) NS NS NS Lateral Translation  -2.46 *** (0.76)  -0.023 (0.023) 0.0013 *** (0.0004) -0.42 (0.25) 0.01 (0.25) 0.43 (0.25) NS NS NS Anterior Translation  28.06 *** (1.10)  0.070 ** (0.026) -0.0063 *** (0.0002) -0.26 (0.15) -0.59 *** (0.15) -0.32 * (0.15) NS NS NS  77   Figure 3-1: Hierarchical random-effects model (REM) and raw data for rotations. Specifically, patellar flexion (A), spin (B) and tilt (C) at 0%, 15% and 30% BW load.  The model describes the weighted group average at each loading level. The load_level*knee_angle interaction term (β4) for the 30% BW load was significantly different from the 0% and 15% BW loads for both flexion and tilt. Reprinted from Journal of Biomechanics, 43, McWalter et al, The effect of load magnitude on three-dimensional patellar kinematics in vivo, 1890-1897, 2010 with permission from Elsevier [207].  78  Figure 3-2: Hierarchical random-effects model (REM) and raw data for translations. Specifically, proximal, lateral and anterior translation at 0%, 15% and 30% BW load.  The model describes the weighted group average at each loading level.  The load_level interaction term (β3) for the 30% BW load was significantly different from the 0% and 15% BW loads for both proximal and anterior translation. Reprinted from Journal of Biomechanics, 43, McWalter et al, The effect of load magnitude on three-dimensional patellar kinematics in vivo, 1890-1897, 2010 with permission from Elsevier [207].  79  3.4 Discussion  We assessed whether applied load changes in vivo patellar kinematics measured using MRI. A 30% BW applied load caused the patella to translate proximally and posteriorly with knee flexion relative to the lower load cases and caused the patella to flex and tilt medially at a lower rate than in the lower load cases.  However, it should be noted that the differences observed were small, especially for proximal and posterior translation.  We observed that under increasing applied axial load, the patellae were more extended only at greater angles of knee flexion; this difference was up to 5°.  The general trends for all loading conditions are consistent with findings from other sequential static in vivo three-dimensional patellar kinematic studies of normal subjects; for loads of 15% BW (mean 104N) we observed a rate of patellar flexion, which was similar to other studies that used axial loads of 133N [104] and 152N [134] (Table 3-2).  However, the difference observed at greater angles of flexion is somewhat surprising.  We might expect the increased applied load to cause a constant increase in patellar extension over the entire range of knee flexion, in which case we would observe 3 parallel lines with constant offset.  It is unclear why we did not observe this expected pattern; however, we speculate that it may be due in part to changes in the equilibrium position of the patella with knee flexion.  For example, we know that the force ratio between the patellar ligament and quadriceps tendon decreases with knee flexion [10,11], the location of the patellofemoral reaction force moves proximally with knee flexion [12,124] and the angle between the patellar ligament and the tibia decreases with knee flexion [212].  It is possible that increases in quadriceps force at greater load levels (15% and 30% BW load) also alters the equilibrium position in a flexion dependent manner; however, additional biomechanical measures, such as contact areas and quadriceps and patellar ligament forces, would be required to confirm this hypothesis. The findings for patellar tilt differed from what has been observed previously, while results for patellar spin were similar.  The findings of increasing lateral tilt with knee flexion at 0% and 15% BW loads differed from others who reported a quadratic pattern of tilt with knee flexion using sequential static methods [104,134] (Table 3-2).  Our finding that tilt did not change with knee flexion for 30% BW load was surprising because most studies (sequential static and continuous, invasive and imaging-based methodologies) to date have reported increasing lateral tilt with knee flexion [107]; however, this is the first time this load magnitude has been studied in this type of supine loading scenario.  Patterns of internal patellar spin with knee flexion were observed in this and previous studies [104,134]; however, a quadratic model, as compared to a linear model, was unique to this study.  Our y-intercept for spin contrasted with those from Patel et al. and MacIntyre et al. (Table  80  3-2).   The contrast with Patel et al. may be due to differences in assigning the anatomical coordinate systems or the method of angle calculation.  The differences observed between the current work and the work by MacIntyre et al. are surprising since the same loading, imaging and analysis methodology was used; however, there may have been differences in the fitness level of study populations (military vs. university community populations).  Table 3-2: Comparison between the results for 15% BW load in the present study and two other studies of three-dimensional patellar kinematics using sequential static methods [104,134]. Results from the study by MacIntyre et al are based on models centred at 19° of knee flexion, therefore the values for the y-intercept (knee flexion = 0) are calculated by substituting knee_flexion=-19 into the equation of the model obtained from the manuscript. NS=not significant. Reprinted from Journal of Biomechanics, 43, McWalter et al, The effect of load magnitude on three-dimensional patellar kinematics in vivo, 1890-1897, 2010 with permission from Elsevier [207].   y-intercept (knee flexion = 0) (° or mm) Slope coefficient (° or mm /° knee flexion) Quadratic coefficient (° or mm /° knee flexion2)  Present Study MacIntyre et al. Patel et al. Present Study MacIntyre et al. Patel et al. Present Study MacIntyre et al. Patel et al. Flexion  -5.69 -12.344 -4.0794 0.735 0.601 -0.7590 -0.0012 -0.001 0.003942 Spin  -5.47 -0.458 NS 0.129 0.059 -0.0789 -0.0012 NS NS Tilt  9.76 6.825 -1.6785 0.081 0.143 -0.1870 NS -0.006 0.003403 Proximal Translation 24.14 35.609 1.9653 -0.656 -0.704 0.3100 0.0015 0.004 0.003474 Lateral Translation -2.88 1.221 -1.2148 -0.023 -0.085 -0.1463 0.0013 0.003 0.002682 Anterior Translation 27.80 25.172 NS 0.070 0.013 -0.0518 -0.0063 -0.006 NS   Our finding of small proximal and posterior translation with increasing load highlights the interaction of tissues in the extensor mechanism.  It is not surprising that the patellae were pulled proximally with a higher applied load because it is likely that more quadriceps force was required to maintain the prescribed knee angle.  It is also not surprising that the patellae translated posteriorly with higher applied load, since this likely led to higher contact force which caused further cartilage compression.  The patterns of translation observed in this study were also similar to the results of previous studies using comparable methods at similar loads [104,134].  A quadratic model with a pattern of distal patellar translation with knee flexion was observed in all three studies (Table 3-2). The reported coefficients all correspond to an inferior translation of 35-40mm of translation over 60  81  degrees of knee flexion.  The results for lateral translation were similar in both pattern and magnitude between the studies (Table 3-2).  The patterns were parabolic with the patella in a more lateral position in early knee flexion, translating medially and then moving laterally again through flexion. In all three studies, the patellae translated posteriorly with knee flexion.  Quadratic models were fit in the current and MacIntyre et al. studies (Table 3-2), while Patel et al. reported a linear pattern of posterior translation.  The quadratic models had a range of 15 mm anterior/posterior position with knee flexion while for the linear model it was 3mm.  The differences in y-intercepts reported are likely due to assignment of the origin of the coordinate systems [104], in particular for proximal translation because a proximally located origin would result in less anterior/posterior motion with knee flexion while a distally located origin would result in considerably more translation in this plane. Some differences were observed between the results of this study and those found using a continuous method in normal subjects [118].  Our finding of increasing lateral tilt with knee flexion at 0% and 15% BW loads differed from the continuous method which found a medial tilt with knee flexion.  However, both methods showed that the patellae translated distally with flexion but in the current study the patellae were positioned 5mm more distally over the coincidental range of knee flexion (0° to 40°).  Patterns of patellar flexion were also similar. Lateral translation with knee flexion was similar in both pattern and magnitude with differences of only approximately 2mm over the coincidental range of knee flexion.  The pattern of a relatively constant neutral spin and the slightly increasing pattern of anterior translation (range 6.8-11.3mm from 0-40° of knee flexion) found using the continuous method was different from what was observed in this study. The differences that were observed were likely due to anatomical axes assignment or the continuous nature of the method. The findings of our study may also be useful in determining what loads should be prescribed when measuring three-dimensional patellar kinematics in patient populations.  Since patients with patellofemoral joint disease experience pain and sometimes muscle weakness, lower loads must be applied.  The 30% BW load was difficult for some normal subjects and therefore would likely be too difficult for participants with joint disease or injury.  These load magnitudes may seem low compared to activities of daily living, for example, the single leg stance phase of gait applies approximately 100% BW load to the knee.  Relationships between axial loading of the patellofemoral joint in a supine and standing position have not been assessed to date.  Since this study was of normal subjects, it is unclear if similar trends would be observed in patient populations.  It is possible that differences would be larger if joint instability existed. This study was the first to assess differences in three-dimensional patellar kinematics at different load magnitudes.  A strength of this study is that the kinematic measurement method was  82  made using a well validated method [102,132].   The method has the advantage of assessing three- dimensional patellar kinematics using a closed-chain loading, which is more similar to load bearing activities of daily living than the dynamic three-dimensional assessments which use open-chain loading with the weight of the shank [109] or a small additional applied weight [105].  Also, this study provides context for comparing patterns of kinematics between studies using sequential static methods but different prescribed loads.  One limitation of this study is that the method used is not continuous.  Previously, differences have been observed in two-dimensional patellar kinematics between the unloaded sequential static and continuous cases [121] but no study has examined differences between the loaded sequential static and continuous cases.  As discussed above, some agreement was observed between three-dimensional sequential static and continuous methods. Another limitation is that, due to the size of the MRI bore and the imaging field of view, the maximum flexion angle possible is between approximately 45° and 60°, depending on the size of the subject.  Ideally we would collect kinematic data over the entire range of knee flexion.  Further, alignment parameters such as leg axis, patella height and femoral rotation were not assessed in the healthy volunteers and therefore it is possible that such parameters influenced the hierarchical models; however, since the comparisons made were within subject the differences observed with load were likely not affected substantially.  Also, this study evaluated kinematics using a bone model, cartilage was not considered.  Therefore changes in cartilage thickness and contact with load are unknown.  Finally, the 30% BW loading task was difficult to carry out for some subjects, in particular at greater angles of knee flexion.  As a result, some scans had motion artefact and were re-acquired. In this study, we found that increased loading caused significantly different models to be fit for patellar flexion, tilt, proximal translation and anterior translation but not for spin or lateral translation.  However, these differences were relatively small, especially for proximal and anterior translation, and it is unclear if they represent clinically relevant differences.  Nonetheless, knowing how load affects kinematics is essential, especially in clinical studies, so that we know what portion of observed differences might be attributed to differences in load. Therefore, these findings are useful for designing studies of three-dimensional patellar kinematics and for comparing results between studies with different applied loads.   83  4 Patellofemoral Bracing & Kinematics in Osteoarthritis3  SYNOPSIS: Patellofemoral bracing is a commonly prescribed treatment strategy for patellofemoral OA which aims to correct patellofemoral joint mechanics.  It is thought that this correction will reduce patient symptoms. However, it is unclear what magnitude of kinematic correction is required to reduce pain. Therefore the aim of this Chapter was to assess the effect of a patellofemoral brace on three-dimensional patellar kinematics in individuals with patellofemoral OA.  Again, the method of assessing patellar kinematics validated previously by our group was used.  This was the first study of patellofemoral kinematics in individuals with patellofemoral OA.  4.1 Introduction Knee OA is symptomatic in 12.1% of Americans over 60 years of age [213], and the patellofemoral joint is involved in 50% of all radiographic knee OA cases in either an isolated form or combined with tibiofemoral OA [1].  Despite the prevalence of patellofemoral OA, the patellofemoral joint has received relatively little attention in the OA literature and there are relatively few treatment options for individuals with patellofemoral OA.  Because patellar malalignment is associated with radiographic patellofemoral OA progression and symptoms [2] and with magnetic resonance imaging (MRI) measures of cartilage loss and bone marrow lesions [3], it is hypothesized that correcting patellar malalignment using strategies such as bracing or taping may arrest OA symptoms and progression.  Patellar taping reduced pain in patients with symptomatic generalized knee OA [62,63] and with clinical and radiographic patellofemoral OA [61,64] and reduced lateral patellar tilt and displacement in the latter group [64].  3 A version of Chapter 4 has been submitted for publication.  McWalter, E.J., Hunter, D.J., Harvey, W.F., McCree, P., Hirko K.A., Felson, D.T., Wilson, D.R. (submitted July 2010) The effect of a patellar brace on three-dimensional patellar kinematics in patients with lateral patellofemoral osteoarthritis. Osteoarthritis and Cartilage.  Submission number OAC 5824.  84  The mechanical aim of a patellofemoral brace is to alter patellofemoral kinematics, specifically to cause a medial translation of the patella thereby unloading the lateral compartment.  Bracing is considered suitable for individuals with lateral patellar tracking, isolated lateral patellofemoral OA or bi-compartmental patellofemoral OA.  Bracing is a commonly used treatment strategy because of its relatively low cost and ease of implementation.  Further, it does not have some of the limitations of taping including loosening, skin irritation due to the adhesive and the technical skill requirement for application.  However, it is unclear whether a patellar brace is capable of changing patellofemoral joint kinematics in subjects with patellofemoral OA. Although patellar kinematics have not been assessed in the patellofemoral OA population, the effect of a patellofemoral brace has been assessed in individuals with patellofemoral pain.  One study of two-dimensional patellar kinematics in loaded, sequential static postures reported that lateral translation was reduced when the brace was applied in subjects with patellofemoral pain [171].  In a more recent study of two-dimensional patellar kinematics, assessed during continuous flexion in weightbearing using an open-bore MRI scanner, the brace reduced lateral tilt and translation in subjects with patellofemoral pain [65].  It is not clear whether the same changes in kinematics observed in the patellofemoral pain population would also be observed in the patellofemoral OA population with its attendant structural changes, such as the degeneration of articular cartilage or presence of osteophytes which may constrain patellar motion.  Further, it is likely that a brace will not be able to create the same magnitude of medial translation and tilt as patellar taping [64].  It is also unclear how a patellar brace affects patellar kinematics in the other 4 components of motion (patellar flexion, spin, anterior translation and proximal translation).  Therefore, we asked the research question: How does a patellofemoral brace affect three-dimensional patellar kinematics in individuals with radiographic lateral patellofemoral OA? 4.2 Methods 4.2.1 Subjects Twenty subjects with symptomatic radiographic lateral patellofemoral OA were recruited to participate in this study.  This group was a subset of participants enrolled in a randomized crossover trial examining the efficacy of an intervention brace and a control brace for reducing knee pain (Appendix G).  There was a 6 week washout period between treatment phases.  The kinematic assessment for the current study took place approximately 2 weeks into the phase of the study where the active experimental brace was worn.  Subjects were included in this study if they had knee pain on most days of the past month and either isolated patellofemoral OA or patellofemoral OA with  85  concomitant tibiofemoral OA.  Radiographic patellofemoral OA was defined as a grade 2 or above osteophyte (0-3 scale) or joint space narrowing of greater than 1 (0-3 scale) with a concurrent grade 1 osteophyte on a skyline radiograph in accordance with the OARSI atlas.  Tibiofemoral OA was defined as a Kellgren/Lawrence grade 2 or above from a posteroanterior radiograph.   Subjects with concomitant tibiofemoral OA were only included if the symptoms of pain location or physical examination findings were consistent with patellofemoral disease as the predominant source of symptoms, as determined by the examining physician (DJH).   These include, but are not limited to, pain during stair climbing, pain while rising from a chair, anterior knee pain and tenderness with patellar mobilization.   The study was approved by the Boston University Medical Center and University of British Columbia institutional review board and all subjects gave informed consent. 4.2.2 Patellar Brace The patellar brace used in this study was a standard sleeve with a T-strap (Bio Skin Q, Cropper Medical Inc., Ashland, OR, USA) (Figure 4-1).  The T-strap was positioned to create a medial translation of the patella.  A trained research assistant (KH) instructed participants on correctly donning and doffing the brace.  Figure 4-1: The patellofemoral brace (Bio Skin Q, Cropper Medical Inc., Ashland, OR, USA) evaluated in this study.  4.2.3 Three-Dimensional Patellar Kinematic Analysis We assessed three-dimensional patellar kinematics in each subject using a validated, non- invasive, sequential static, MRI-based method described in detail in Sections 2.2.2 and 2.2.3 and Appendices C and D [102,132].   Kinematics were assessed under 4 different conditions: 1) no load,  86  no brace 2) 15% BW load, no brace 3) no load, brace 4) 15% BW, brace.  Images for conditions 1 and 2 were acquired first; the subject then donned the brace and images were acquired for conditions 3 and 4.  We chose a load of 15% BW because we found previously that this load is achievable by study participants with knee OA.  We tested a no load condition in the study to determine whether kinematic changes created by the brace were consistent at different loads.  For these data, each kinematic parameter was plotted as a function of knee flexion (calculated from the kinematic analysis, not from goniometer measure) for each of the 4 conditions.  The calculated knee flexion data were scaled so that the calculated and goniometer measures of knee flexion angle at the 10° no brace, no load condition knee were equal (i.e. the data were centered at 10° of knee flexion). 4.2.4 Statistical Analysis We tested the null hypothesis that there was no difference between the bracing conditions at no load and 15% BW load for each patellar kinematic parameter (Figure 2-4) using hierarchical linear random-effects models (Section 2.2.4 and Appendix E).  In the hierarchical structure of the model, subject was level 1 and condition was level 2.   The model assumes a within subject correlation for the kinematic data.  The model is a weighted average of each subject’s kinematic data that takes into consideration any missing data and the variance from the mean.  The models in this study took the following form: anglekneeconditionconditionanglekneeanglekneey _***_*_* 43 2 210 βββββ ++++= where ‘y’ was the kinematic parameter, ‘knee_angle’ was the knee flexion angle and ‘condition’ was 1) no load, no brace 2) 15% BW load, no brace 3) no load, with brace, and  4) 15% BW, with brace. The random intercept term (β3) was always included in the model while the quadratic (β2) and random slope (β4) terms were included only when significant (p<0.05).  Knee angle is a continuous variable, and condition is a discrete variable.  Statistically significant differences were defined as p<0.05.  All statistical analysis was carried out using packaged software (Stata 10, StataCorp LP, College Station, TX, USA). 4.3 Results Nineteen subjects (14 female, 5 male, 62.4 ± 9.9 years, 86.6 ± 18.9 kg) with lateral patellofemoral OA participated in this study.  The twentieth subject was excluded from the analysis due to motion artifact in the high-resolution MRI scan which made patellar kinematic assessment impossible.  Of the 19 remaining subjects, 6 had concomitant tibiofemoral OA and 6 had concomitant medial patellofemoral OA and tibiofemoral OA.  The subjects had been wearing the brace for  87  16.0±6.8 (mean ± standard deviation) days when the study was carried out.  The subjects wore the brace for an average of 5.1 ±1.9 hours per day. The brace caused the patellae to extend in early angles of knee flexion but the difference was not maintained in greater angles of knee flexion (Figure 4-2 A, Table 4-1).  The slope (β4) varied between conditions for flexion; the braced condition had greater slopes than the no brace condition at both load levels (both p<0.001) (Table 4-1).  The brace caused the patellae to spin externally by 0.62° for the loaded condition (p=0.012) throughout the range of knee flexion studied (Figure 4-2 B, Table 4-1).  The trend was similar for the unloaded condition but the results were not statistically significant.  The brace caused a medial tilt of the patella of 1.11° for the unloaded condition (p<0.001) and 0.85° for the loaded condition (p=0.004) throughout the range of knee flexion studied (Figure 1-1Figure 4-2 C, Table 4-1). The brace caused the patellae to translate distally by 0.67 mm for the unloaded condition (p=0.001) and 1.09 mm for the loaded condition (p<0.001) throughout the range of knee flexion studied (Figure 4-3 A, Table 4-2).  Medial translations of 0.23 mm for the unloaded condition (p<0.001) and 0.46 mm for the loaded condition (p<0.001) were also found throughout the range of knee flexion studied (Figure 4-3 B, Table 4-2).  Finally, the brace caused a posterior translation of 0.51 mm for the unloaded condition (p<0.001) and 0.47 mm for the loaded condition (p<0.001) throughout the range of knee flexion studied (Figure 4-3 C, Table 4-2). The quadratic coefficient (β2) was significant for patellar flexion, tilt, lateral translation and anterior translation and therefore included in those models.  The random slope coefficient (β4) was significant for patellar flexion therefore included in that model. Loading did not affect kinematics for any parameter except lateral translation in which the patella translated laterally by 0.22 mm (p=0.42) with applied load for the no brace condition (Figure 4-3 B, Table 4-2).         88  Table 4-1: Coefficients (confidence interval) for the hierarchical random-effects models for rotations. The quadratic and random slope terms (β2 and β4, respectively) were included in the model when significant. NS = not significant, *=p<0.05, **p<0.01, ***p<0.001.  Coefficients Flexion Spin Tilt β0 (y-intercept) -12.83*** (-15.43, -10.23) -5.83*** (-7.03, -4.63) 7.60** (2.72, 12.48) β1 (slope) 0.459*** (0.385, 0.534) 0.089*** (0.072, 0.105) 0.131** (0.052, 0.211) β2 (quadratic) 0.0041*** (0.0033, 0.0050) NS -0.0016** (-0.0025, -0.0006) β3: (difference in random intercept between conditions) 0% load: no brace vs brace -3.88*** (-4.82, -2.94) -0.43 (-0.90, 0.04) 1.11*** (0.54, 1.68) 15% load: no brace vs brace -3.60*** (-4.51, -2.68) -0.62* (-1.10, -0.15) 0.85** (0.27, 1.43) No brace: 0% load vs 15% load -0.23 (-1.15, 0.69) 0.02 (-0.46, 0.04) -0.06 (-0.64, 0.51) Brace: 0% load vs 15% load 0.05 (0.98, -0.88) -0.18 (-0.65, 0.29) -0.32 (-0.89, 0.25) β4: (difference in random slope between conditions) 0% load: no brace vs brace 0.003 (-0.029, 0.036) NS NS 15% load: no brace vs brace 0.062*** (0.030, 0.094) NS NS No brace: 0% load vs 15% load 0.060*** (0.028, 0.092) NS NS Brace: 0% load vs 15% load 0.002 (-0.030, 0.033) NS NS                      89  Table 4-2: Coefficients (confidence interval) for the hierarchical random-effects models for translations. The quadratic term (β2) was included in the model when significant. The random slope term (β4) was not significant for any translation and therefore not included.  NS = not significant, *=p<0.05, **p<0.01, ***p<0.001.  Coefficients Proximal Lateral Anterior β0 (y-intercept) 31.77*** (28.64, 34.90) 0.54 (-1.37, 2.45) 27.50*** (26.21, 28.78) β1 (slope) -0.531*** (-0.574, -0.489) -0.013 (-0.047, 0.022) 0.072*** (0.034, 0.11) β2 (quadratic) NS 0.0007*** (0.0003, 0.0010) -0.0036*** (-0.0039, -0.0033) β3: (difference in random intercept between conditions) 0% load: no brace vs brace -0.67** (-1.05, 0.29) -0.23* (-0.44, -0.02) -0.51*** (-0.68, -0.33) 15% load: no brace vs brace -1.09*** (-1.48, -0.69) -0.46*** (-0.68, -0.25) -0.47*** (-0.64, -0.29) No brace: 0% load vs 15% load 0.31 (-0.07, 0.69) 0.22* (0.01, 0.44) -0.06 (-0.23, 0.12) Brace: 0% load vs 15% load -0.11 (-0.49, 0.28) -0.01 (-0.22, 0.20) -0.02 (-0.16, 0.19)   90  Figure 4-2: Rotational results as a function of knee flexion for patellar a) flexion, b) spin and c) tilt. The lines indicate the results of the hierarchical linear random-effects model for each condition 1) no load, no brace, 2) 15% BW load, no brace, 3) no load, with brace, 4) 15% BW load, with brace.  The raw data points are included and are colored according to condition.  The data were scaled to 10° of knee flexion.  91   Figure 4-3: Translation results as a function of knee flexion for a) proximal, b) lateral and c) anterior translation. The lines indicate the results of the hierarchical linear random-effects model for each condition 1) no load, no brace, 2) 15% BW load, no brace, 3) no load, with brace, 4) 15% BW load, with brace.  The raw data points are included and are colored according to condition.  The data were scaled to 10° of knee flexion.  92  4.4 Discussion We assessed whether wearing a patellofemoral brace altered three-dimensional patellar kinematics in subjects with lateral symptomatic radiographic patellofemoral OA.  The application of the brace caused a consistent external spin, medial tilt, proximal translation, medial translation and posterior translation of the patellae over the range of knee flexion studied.  It also caused the patellae to extend in early angles of knee flexion, but this difference was not maintained in greater flexion. These findings were consistent for the unloaded and loaded conditions, although the magnitudes of the differences varied slightly.  The brace accomplished its mechanical aim of translating the patellae medially; however, the magnitude of this translation was, on average, quite small (0.46 mm for 15% BW load condition).  Since this is the first study to examine the effect of bracing in subjects with patellofemoral OA, there are no directly comparable results in the literature.  However, our results are generally consistent with studies of patellofemoral bracing in subjects with patellofemoral pain and patellar subluxation and dislocation but the magnitudes of the differences are smaller in the present study.  In these studies, two-dimensional assessments in the axial plane were made using a measure of medial translation termed the bisect offset (the percentage of the width of the patella that is lateral to the deepest part of the trochlear groove [214]) and mediolateral patellar tilt angle (the angle between the line across the maximum width of the patella and the line joining the posterior femoral condyles [200]) or patellar tilt angle (the angle between the line parallel with the lateral patellar facet and a line joining the posterior femoral condyles [215]).  In a study of patellofemoral pain, two different braces caused medial translations of 0.9 and 1.4 mm, respectively, (bisect offset decrease of  2.4% and 3.6%, standardized to mean patellar width of 38.8mm) when assessing kinematics with MRI at loaded sequential static postures [171].  In a second study of patellofemoral pain the brace caused a 1.6 mm medial translation (bisect offset 4%) when assessing kinematics in weightbearing flexion in an upright, open-bore MRI scanner [65].  In the study of patellar subluxation and dislocation a medial translation the brace caused a medial translation of 0.4 mm (bisect offset approximately 1%) when assessing kinematics with MRI in continuous flexion, resisting the weight of the shank [215].  Medial translations were smaller in the present study (0.23 to 0.46 mm) which may be due to differences in the OA joint anatomy, the age and strength of the patients and the definition of patellar translation.  In these same studies the brace caused the patellae to tilt medially by approximately 1.5° [171] and 3° [65] (mediolateral patellar tilt angle) and 1.5° [215] (the patellar tilt angle).  The magnitude of medial tilt was smaller in the present study (between 0.79 and 1.11°).  Some of the differences are likely the result of differences in the definition of tilt.  In the present study, tilt was based on a coordinate  93  system defined in three dimensions, in contrast with the two-dimensional assessments used in the other studies.  Differences could also be due to variations in the design of the patellofemoral brace itself. The patellofemoral brace also caused small changes in patellar flexion, spin and distal translation, which is not its intended purpose.  Our finding that bracing extended the patella in early knee flexion may be the result of the patellae translating proximally in early flexion, which may have shifted the proximal end of the patella above the brace’s patellar cutout and caused the patella to extend.  Further, the brace caused the patellae to translate distally and spin externally, which may be due to the restriction of patellar motion that the cutout imposes.  Finally, although small, the trend of posterior translation when the knee brace was worn was not surprising because, as others have speculated, the brace may cause the patella to be located more firmly in the trochlear groove, which is supported by the finding that bracing increases cartilage contact area [171].  Because this is the first three-dimensional assessment of bracing, comparison to the patellofemoral pain/bracing literature is not possible for these parameters.  Further, it is unclear if the small changes in these additional parameters created by the brace are clinically important in terms of the biomechanical aim of unloading the lateral compartment of the patella and/or reducing patient symptoms. It is not surprising that a quadratic model provided the best fit for most of the kinematic parameters because previous studies of normal subjects have shown similar findings [104,134].  In particular, these studies showed that flexion, tilt, proximal translation and lateral translation follow a quadratic pattern.  In the present study proximal translation was found to be linear and this difference may be due to differences in study population.  There is variation in results for anterior translation with one study finding a linear pattern [104] and the other a quadratic pattern [134]; the present study had a quadratic pattern.  The previous studies and the present study found linear patterns of spin. Patellofemoral bracing produces similar kinematic results as patellar taping; however magnitudes were smaller.  A study of the effect of taping on patellar malalignment in patients with patellofemoral OA found that the tape caused the patellae to tilt medially by 3.57° but did not change the bisect offset [64].  It also found a decrease of 2.94% (or 1.1 mm, again using the comparative patellar width of 38.8 mm) in lateral displacement, which in this case was defined as the percentage of the patella lateral to the most anterior portion of the lateral condyle.  The tape caused much greater tilt than observed in the present study and our measure of medial translation falls between the two measures used in the taping study.  This highlights the difficulty in comparing results for medial translation between methods of assessment and this may be because they are sensitive to patellar type [6] and coordinate system assignment [117].  Differences may be due to differing aims of the tape and  94  the brace.  The tape was positioned to translate and tilt the patella medially, tilt the patella superior and unload the infrapatellar fat pad.  This aim is more complex than the medial translation applied by the brace.  Differences may also be due to the cross-sectional design of the taping study; alignment was assessed immediately after tape application therefore the effect of loosening was not considered. There were no differences in kinematics between the two load levels for the braced condition and only for lateral translation for the no brace condition; however, this difference was very small. We have previously shown that there was no difference in kinematics between no load and 15% BW loads in normal subjects [207];  in the present study we have shown that this finding is also true in individuals with lateral patellofemoral OA.  The small but statistically significant increase in lateral translation (0.22 mm) may be due to a laterally directed line of action of the active quadriceps muscles.  It is possible that individuals with lateral patellofemoral OA display a more lateral patellar tracking than normal individuals and it appears that the quadriceps muscle contraction contributes to this pattern. We observed large variability in kinematic data between individuals, suggesting that factors other than load may influence kinematics.  Possible factors include patellar type, varus/valgus alignment, joint laxity, muscle strength, radiographic or MRI-based OA grade, WOMAC or KOOS score and clinical symptoms to name but a few.  We did not have these data available in this study and therefore these factors were not controlled for in the statistical model. It may be possible to reduce the variability in the data in future studies by controlling for these variables in the statistical model or by studying a more homogenous subgroup.  A large sample size and a well clinically defined population would be required to determine which of these factors, if any, influence patterns of three-dimensional patellar kinematics in a consistent manner. It is not clear whether the small changes in kinematics caused by the brace in this study reflect changes to the mechanical environment sufficient to produce clinical benefits.  Patellar taping reduced pain in individuals with knee and patellofemoral OA [61-64]; however, kinematics were not assessed in those studies.  Therefore, it remains uncertain what magnitude of correction is required to reduce pain.  The results from the main cross-over study from which this subset was drawn found that the patellar brace did not reduce pain in this lateral patellofemoral OA population (Appendix G).  It is possible that the brace is not able to apply the same amount of mechanical correction as taping which may explain the difference in pain results.  However, the subject groups differed between these two studies; the subjects taping study had bilateral patellofemoral OA.  The kinematic differences for medial tilt and translation in the present study are smaller than those in the patellofemoral pain studies [65,171,215], which is another indication that bracing affects the patellofemoral OA and pain  95  populations differently.  It is possible that the degenerative joint changes associated with OA impede the performance of the brace. This study was the first to assess the effect of a patellofemoral brace on three-dimensional patellar kinematics in subjects with patellofemoral OA.  Strengths of this study include the carefully defined radiographic lateral patellofemoral OA population studied and the use of a validated method to assess three-dimensional patellar kinematics.  A limitation of the study is that a sequential static method for assessing kinematics was used.  Ideally we would assess patellar kinematics in normal activity, but to date no such method of assessment exists.  Differences have been observed in two- dimensional patellar kinematics between unloaded, sequential static positions and continuous movement in resisted extension (open-chain) [121], and between continuous resisted extension and continuous weightbearing flexion [129].  Further, the 15% BW load used in this study is less than that experienced during dynamic activity; however, it was a load that all the participants with knee pain could tolerate and maintain throughout MRI scanning.  It is likely that differences exist between sequential static and continuous methods and between resisted extension and simulated or actual weightbearing in three dimensions.  Another limitation is that the patients wore the brace on average for 16 days prior to the kinematic assessment and as a result there may have been a learning effect, in particular for the no brace condition.  Finally, we did not randomize the order in which each condition was studied because of the additional time that would have been required for the kinematic assessment.  However, we chose to study the no brace condition prior to the brace condition because this order is aligned with the application of this treatment strategy. We found that bracing produced small but statistically significant differences in three- dimensional patellar kinematics in subjects with radiographic lateral patellofemoral OA.  The brace caused the patellae to spin externally, tilt medially and translate distally, medially and posteriorly through the entire range of knee flexion studied but most of these changes were small.  The brace also extended the patellae in early angles of knee flexion, however, again this difference was small.  These trends were similar for both loading conditions.  Although the brace achieved the mechanical aim of translating the patellae medially it remains unclear if this small medial translation is able to adequately correct patellar kinematics in patients with radiographic lateral patellofemoral OA.  96  5  MRI-based Assessment of Contact Area: Development & Validation4  SYNOPSIS: Treatment strategies for patellofemoral OA have focussed on correcting kinematics with the assumption that cartilage contact mechanics are also corrected as a result.  Until kinematics and contact areas can be assessed simultaneously in vivo, it will be unclear if this is the case.  There currently exists no fully validated method to assess these two parameters simultaneously and therefore the aim of this chapter was to develop and validate a rapid method to assess contact areas that can be used in series with the method of assessing kinematics validated previously by our group.  The effect of scan direction and calculation method on contact area measures was also assessed.  5.1 Introduction It is currently possible to assess three-dimensional patellar kinematics and patellar contact areas in vivo using MRI.  However, apart from two studies [104,116], these quantities are most often measured independently of one another.  It is particularly important to assess patellofemoral kinematics in cases of patellofemoral OA because treatment strategies generally focus on correcting patellar tracking.  The aim of this correction is to restore an imbalance of load distribution between the medial and lateral patellar facets.  A simultaneous measure of in vivo contact area would allow assessment of whether the mechanical correction completed this aim.  Further, it is possible that contact area is more closely associated with patient symptoms. In order to develop and evaluate treatment strategies, it is imperative that these two quantities be measured simultaneously.  However, the loaded MRI scans required to assess contact area are generally quite long and validation has been limited.  Since individuals with patellofemoral joint pathologies, such as OA, may have difficulty  4 A version of Chapter 5 is being prepared for submission.  McWalter, E.J., O’Kane, C., Wilson, D.R., A short, validated MRI scan for assessing patellar cartilage contact areas in vivo.  97  carrying out the loading task, there is a need to develop and validate a short MRI-based method of assessing contact areas in vivo so that a simultaneous measure can be carried out.  Quantification of contact area from MRI has been carried out using both axial and sagittal scans and various different calculation methods.  The length of these scans ranges in time from 39 seconds [171,173,174,176] to 11 minutes [167], with most studies being in the 2 to 5 minute range [165,166,168,169,172,175].  Most assessments have been carried out using MRI scanners with field strengths of 1.5 Tesla or lower [104,165,166,168-174,176,216].  Throughout the scans, the knee is usually loaded.  This has been carried out using a vertical open configuration scanner in which the participant is in an upright position bearing 45% of their body weight through the study knee [165,166], using a horizontal open configuration scanner in which the participant is lying on his or her side resisting a 10 Nm torsional load applied to the shank with a pulley system [168,169,172] or using traditional closed bore scanners in which the participant in a supine position and axial load applied to the foot (range from 10% of maximum isometric contraction (10.5 to 13 N) [175] to 25% BW [171,173,174,176]).  In all cases, the participant must remain still for the duration of the scan, therefore the applied loads must be low enough to be maintained by the subject without substantial tremor.  Shorter scans are therefore desirable.  The applied loads are typically lower than those experienced during regular activity.  To assess contact area, first cartilage contact must be determined in each MRI slice; this has been carried out by delineating contact area directly [104,116,165- 167,170,171,173,174] or by defining contact as the intersection of segmented cartilage plates that have been dilated by a pixel [168,169,172].  Finally, contact area must be calculated by multiplying in-slice contact by slice thickness [165-169,171-174,176], interpolating between slices [104,116,170] or surface fitting [175].  Most groups report only contact area, while two studies have also reported contact centroid location with respect to the patella [116,175]. Limited data exist on the measurement errors associated with these methods.  Errors were reported to be 3% (CV%) in a phantom-based validation study [166] and 13% in a cadaver-based validation study [167], both using sagittal scans.  Errors in assessments from axial scans have not been assessed.  Various types of repeatability have been quantified.  Intra-subject repeatability was reported to be 8.3% (CV%) in human subjects [172].  Intra-reader repeatability was reported to be 3% (CV%) [166] and 1.3 mm2 [171] and 21 mm2 [170] (expressed as the standard error).  Inter-reader repeatability was reported to be 7% [166] and 11.4% [135] (CV%).  The intra-MRI scan repeatability, found by repeating MRI scans three times in one cadaver specimen was reported to be 3.2% (CV%) [167]. It is important that each method be validated independently because of differences in the MRI scan sequences and calculation method used.  98   It is currently not possible to assess accuracy in vivo because we lack an agreed upon reference standard.  Cadaver specimens provide a useful model for validating MRI based contact area measurements. They are advantageous because tissue structures and joint geometry are similar between the validation and actual in vivo measurements.  Also, numerous techniques exist to assess joint contact areas in cadaver specimens such as dye staining [124,139,142,144], casting [92,145,148,149], pressure sensitive film [12,82,86,95,151-153], stereophotogrammetry [145,162,163] and dynamic pressure measurement systems [76,88,154-157].  Agreement between these techniques has been shown to be very good by visual observation of registered areas [145]. These techniques can be used to validate the MRI-based measures of contact area.  There currently exists no rapid, fully validated protocol for assessment of patellofemoral contact areas in vivo.  A scan less than one minute is required so that individuals with patellofemoral OA are able to complete the loading task.  With 3 Tesla MRI scanners, this shorter scan time is possible.  Further, variability exists between methods of calculating contact areas from MRI in terms of the length and direction of scan used, the calculation methods and the level of validation.  It is clear that further development and characterization of MRI-based measures of contact area are required. Therefore the aims of this study were: 1. To develop an axial and a sagittal MRI scan sequence of less than 1 minute to be used to assess patellofemoral contact area during loaded knee flexion 2. To develop an analysis tool that allows for calculation of contact area magnitude and location on the patellar surface 3. To determine the agreement of contact area assessments from axial and sagittal MRI scans with reference to a dye staining technique in a cadaver model 4. To determine the intra-reader and intra-subject repeatability of the technique 5.2 Methods 5.2.1 Specimens Preparation  Six cadaver knee specimens, with no history of knee surgery, paralysis, lower limb abnormalities or multiple sclerosis, were used in this study.  Specimens were thawed at room temperature over night.  Soft tissues were then removed from the proximal and distal ends of the specimen, exposing approximately 12 cm of the femoral shaft and 8 cm of the tibial shaft.  A longitudinal incision was made from the distal end of the patella to the proximal end of the specimen along the femoral shaft and the skin and fat layers were resected.  The quadriceps tendon was dissected and the patellofemoral joint capsule was released along the proximal, medial and lateral  99  margins of the patella.  The femoral shaft was potted 10 cm from the proximal end in a custom poly vinyl chloride piping based connector using standard dental stone.  A length of non-stretch, ultra high molecular weight polyethylene line (Spectra, Honeywell, Morristown, NJ, USA) was sutured to the quadriceps tendon and another was attached to the tibia, 2 cm from the distal end.  The fat and skin layer was then sutured closed.  This was carried out in order to maintain a normal tissue layer structure for the MRI imaging to follow. 5.2.2 Experimental Procedure  Two unloaded sagittal MRI scans (Intera, Philips, the Netherlands), one optimized for bone and the other for cartilage, of each specimen were acquired without moving the specimen between scans (Table 5-1).  The specimen was then mounted in a custom designed loading apparatus (Figure 5-1).  The femoral end was connected rigidly to the apparatus using a nylon bolt.  The line connected to the quadriceps tendon ran through a pulley system and was locked down at the prescribed knee flexion angle using a plastic one-way rope anchor.  The line connected to the tibia also ran through a pulley system and was loaded with a 60 N load (MR safe sandbags) to ensure patellofemoral contact. Once loaded, three additional MRI scans were acquired (Table 5-1): 1) sagittal low resolution TI weighted spin echo sequence, optimized for bone 2) axial T1 weighted, three-dimensional fast field echo sequence, optimized for cartilage and 3) sagittal T1 weighted, three-dimensional fast field echo sequence, optimized for cartilage.  The specimens were in the same position for these three scans. The MRI bed was slowly retracted from the MRI scanner, ensuring that the loading setup was not altered.  The sutures attaching the skin and fat layers were released using an MRI-safe scalpel, allowing for visualization of the loaded quadriceps tendon and the patella.  Food grade dye (Blue Neon, Club House, McCormick Canada, London, ON) was applied liberally to the region surrounding the patellofemoral contact using a paintbrush.  The region of the patella devoid of dye indicated regions of patellofemoral contact.  The excess dye was carefully removed using absorbent cotton pads.  The load was released and the quadriceps tendon was pulled distally exposing the articular surfaces of the patella and the femur. The specimen was transported from the MRI centre to the biomechanics laboratory for assessment of contact area from the dye-based reference standard.  Four fiducial markers were rigidly attached to the margins of the patella.  The fiducial markers and the periphery of the contact area were digitized using an optical tracking system (Optotrak Certus, Northern Digital Inc, Waterloo, ON). This process was repeated 3 times.  The manufacturer quoted accuracy of the system is 0.15 mm in three dimensions.  The surface topology of the patella was determined using a three-dimensional non-  100  contact digitizer (Vivid 9i, Konica-Minolta, Osaka, Japan).  The manufacturer quoted accuracy of this system is 0.05 mm and the precision is 0.008 mm.  In practical application, one independent validation study found the agreement of the results from this particular scanner when scanning 10 masks created from human faces with reference to a tactile surface scanner to be 0.08 mm [217]. The repeatability was reported as 0.003 mm based on scanning each of the 10 masks twice [217].  Table 5-1: T1-weighted MRI sequence parameters. Indicates the reconstructed in plane resolution, the loaded cartilage scans were acquired at 0.4 x 0.4 mm. MS = multi-shot, SE=spin echo, 3D=three dimensional, FFE=fast field echo   Parameter Unloaded Bone Sagittal Loaded Bone Sagittal Unloaded Cartilage Sagittal Loaded Cartilage Sagittal Loaded Cartilage Axial Scan Sequence MS SE MS SE 3D FFE 3D FFE 3D FFE In-plane Resolution 0.60 mm 1.25 mm 0.60 mm 0.3125 mm* 0.3125 mm* Field of View 300 mm 320 mm 300 mm 180 mm 180 mm Slice Thickness 2 mm 2 mm 2 mm 2 mm 2 mm Slice Separation 2 mm 7 mm 2 mm 2 mm 2 mm Matrix Size 512 x 512 256 x 256 512 x 512 576 x 576 576 x 576 Repetition Time 360 ms 306 ms 12 ms 11 ms 14 ms Echo Time 10.0 ms 6.2 ms 4.3 ms 3.9 ms 4.6 ms Flip Angle 90° 90° 15° 10° 10° Scan Time 7 min 44 s 35 s 4 min 53 s 58 s 52 s MRI Coil Knee Body Knee Body Body    101  M Figure 5-1: Schematic of cadaver knee specimen in MRI-safe loading apparatus positioned in the MRI scanner.  5.2.3 Contact Area Analysis  The MRI-based assessment of contact area was carried out using the axial and sagittal scans. Two different methods were used to identify regions of in-slice contact using commercially available image processing software (Analyze 8.0, Analyze Direct, Overland Park, KS, USA).  In the first method, ‘delineation’, contact (defined as no visible separation between cartilage plates) was delineated directly in a slice-by-slice manner using the Image Edit module (Figure 5-2).  The second method, ’intersection’, was a multi-step process.  First, the patellar and femoral trochlear cartilage was segmented individually in a slice-by-slice manner, also using the Image Edit module (Figure 5-2).  Next, the segmented cartilage maps were converted into binary images and the femoral trochlear surface was dilated by 1 pixel using the Morphology module.  Finally, intersection between the patellar cartilage and dilated femoral trochlear cartilage was defined as the Boolean intersection using the Image Calculator module.  Contact regions defined on the axial and sagittal images using the delineation and intersection methods were imported into custom written software (Matlab, The Mathworks, Natick, MA, USA).  Splines were fit to the in-slice contact data.  Area was then calculated in three different ways: 1) by multiplying the length of the in-plane spline by the slice thickness (2mm in this case), 2) by carrying out a linear interpolation between splines in adjacent slices and summing the area of the triangle faces created by the interpolation, and 3) by fitting splines between slices and summing the area of the triangle faces created by the interpolation.  The area of  102  the three-dimensional triangle was calculated by determining half the magnitude of the vector cross product (half the area of the parallelogram defined by two edges of the triangle).  The number of points sampled along the in-plane and between-plane splines was determined in a case specific manner and was defined as the number of points required for the area calculation to change by less than 0.01mm2. The number of points sampled along the splines was consistent within each direction (for example all in-plane splines would be sampled at 400 points and all between-plane splines would be sampled at 11 points).   Figure 5-2: Example of axial (top) and sagittal (bottom) scans delineated (left) and segmented (right).  103  The assessment of contact area from the dye-based reference standard was carried out using custom written software (Matlab, The Mathworks, Natick, MA, USA).  The digitized contact area was registered to the patellar surface using the laser scanner data, using the fiducial markers and an ICP algorithm.  The registered data were imported into commercial laser scanning/solid modelling software (Rapidform XOR, INUS Technology Inc., Seoul, Sourth Korea).  The laser scanner data was processed to create a surface mesh.  This process included, filling holes, finding defects in the mesh, smoothing the outer boundary and optimizing the mesh, all using tools built into the commercial software.  This type of processing is standard for laser scanner data and is similar to that used in the validation paper discussed in Section 5.2.2 [217].  A non-uniform, rational, basis spline (NURBS) surface was fit to the optimized mesh; the difference between the mesh and the NURBS surface was never more than 0.1mm.  A spline was fit to the dye contour, which was projected onto an oblique plane containing the four posterior points on the NURBS surface.  The projected profile of the dye contour was extruded through the NURBS surface, which was then used to cut the extrusion.  The result was a NURBS surface describing the contact area.  Area of the contact surface was calculated using a built in tool.  This process was carried out for all 3 sets of dye-based contours. In order to determine the position of the MRI-based contact areas relative to the reference standard, the data were plotted in the anatomical patellar coordinate system used in the kinematic assessment.  In order to accomplish this, the bone and cartilage from the unloaded MRI scans and the loaded bone scan were segmented in a slice-by-slice manner using the same technique used for the kinematic analysis described in Section 2.2.3.  The patellar coordinate system was defined on the unloaded bone model and therefore all data were registered to this model.  The unloaded patellar cartilage was already expressed in this coordinate system because the images have the same dimensions and the specimens were not moved between scans.  The contact area from the reference standard was plotted in the patellar coordinate system by registering the laser scanned cartilage surface to the unloaded MRI cartilage surface using an ICP algorithm.  Plotting the MRI-based contact area in the patellar coordinate system required three steps.  First, the bone was segmented from the loaded bone scan in a slice-by-slice manner and contours were created.  Next, the loaded bone contour and the loaded contact area, which were created from images of different size and resolution, were transformed into the same coordinate system using a homogenous transformation matrix created using the Image Patient Position (coordinate of upper left hand corner of the image) and Image Orientation Patient (direction cosines) tags from the digital imaging and communication in Medicine (DICOM).  Finally, the bone contour was registered to the bone model using an ICP, in a similar manner as would be done in the kinematic analysis described in Appendix D.  The MRI-based and reference standard contact areas were plotted in the patellar coordinate system and contact  104  centroids were calculated for each (1 MRI and 3 trials of reference standard).  The distance between the MRI-based contact centroid and the three reference standard contact centroids was calculated. The data were imported into the laser scanning/solid modelling software and the MRI-based contact area was projected onto the laser scanned surface in the manner described above. The percentage overlap between the MRI-based contact area and each reference standard contact area was defined as the area of the Boolean intersection of the two areas divided by the area of the reference standard contact area.  The error of the MRI based assessment of contact area as compared to the reference standard was defined as the mean absolute and percent difference between MRI-based assessment and the mean of the three reference standard assessments.  The distance between the MRI and reference standard centroids and the mean percentage overlap of the MRI and reference standard areas were also calculated. 5.2.4 Repeatability The intra-reader repeatability of the axial and sagittal MRI-based assessment of contact area was determined by repeating the delineation and intersection analysis 3 times for all datasets.  Intra- reader repeatability was expressed as the mean standard deviation of the assessment. Intra-subject repeatability of the MRI-based measure of contact area was determined in two normal subjects.  Unloaded bone and cartilage scans were acquired with the knee coil (Table 5-1). Loaded sagittal bone and cartilage scans were acquired at six different angles of knee flexion between full extension and approximately 45°.  The knee was loaded to 15% BW using a custom designed loading device (as was used in the kinematic analysis).  The bone and cartilage scans were acquired in series without a break and therefore the total loading time was 1 minute 35 seconds (including pre- scan) for each flexion angle.  Two additional sets of 6 loaded scans were acquired with 10 minute break between sets during which the participant left the scanner.  Only sagittal cartilage images were acquired because they were determined to show better agreement with the reference standard and had better intra-reader repeatability (see results section to follow).  Contact area was assessed at each angle of knee flexion using the delineation method described above.  Contact area was plotted as a function of knee flexion and splines were fit to each repeated set.  Repeatability was expressed as the mean standard deviation of contact area sampled at 1° increments along the fitted spline for the coincidental range of knee flexion.  105  5.3 Results Complete datasets were obtained for four specimens (3 right, 1 left, 2 female, mean age 70.0±5.5 years), and only these results are reported.  The sagittal MRI scans consistently showed better agreement with the reference area than the axial scans; mean error for the three calculation types ranged from 47.7 to 68.4 mm2 for the sagittal scans and from 61.4 to 175.6 mm2 for the axial scans (Table 5-2).  For the sagittal scans, the delineation method consistently showed equal or better agreement with the reference than the intersection method; errors ranged from 47.4 to 64.1 mm2 for delineation and from 59.4 to 68.4 mm2 for intersection (Table 5-2).  For ease of visualizing the magnitudes of these errors a scale is provided (Figure 5-3).  The registration error between the digitized contact area contour and the cartilage surface topology determined from the laser scan was 0.19 mm (±0.12mm) (expressed as the mean absolute difference between point cloud models). The mean distance between MRI-based centroid (calculated from the delineation results from the sagittal scan) and the reference standard centroids was 2.4±1.1 mm (Table 5-3, Figure 5-4) and the mean percentage overlap between contact patches obtained using the two methods was 82.6±12.4% (Table 5-3, Figure 5-5).  Since the centroid results were plotted in the anatomically based patellar coordinate system (Figure 5-4), it can be said that the MRI-based centroids were positioned medial and proximal to the mean reference standard centroid.  The registration error of the MRI- based cartilage surface model (created from the unloaded scan) to the laser scan based cartilage surface model (expressed as the mean absolute difference between models) required to assess overlap was 0.41 mm (±0.37mm). For intra-reader MRI-based assessments, the delineation method was more repeatable than the intersection method (Table 5-4); mean standard deviations ranged from 8.4 to 15.4 mm2 for the delineation method and from 20.8 to 26.2 mm2 for the intersection method.  Overall, assessments from sagittal scans were more repeatable than axial; mean standard deviations ranged from 8.4 to 26.2 mm2 for the axial scans and 9.1 to 22.9 mm2 for the sagittal scans.  Further, the results using the delineation method for the sagittal scans were the most repeatable between calculation methods, with mean standard deviations of 9.4, 9.1 and 9.3 mm2 for the multiplication by slice thickness, the linear interpolation and the spline interpolation methods, respectively. The intra-subject repeatability, assessed in 2 subjects (1 female, age 31, 1 male, age 29), expressed as the mean standard deviation at 1° along the spline, was 38.2±19.7 mm2, 39.9±23.0 mm2, and 39.8±23.1 mm2, using the multiplication by slice thickness, the linear interpolation and the spline interpolation methods, respectively (Figure 5-6).  Generally, contact areas increased with knee flexion  106  (Figure 5-7).  Contact centroids translated proximally and medially with knee flexion; however, all centroids remained in the lateral compartment of the patella (lateral to the origin of the patellar coordinate system).  Table 5-2: Error of the MRI-based assessment of contact area as compared to the reference standard. Error is expressed as mean absolute difference (and ±standard deviation for totals) (mm2) and percentage difference (in brackets).  Scan Direction Specimen Multiply by Slice Thickness Linear Interpolation Spline Interpolation    Delineated Intersection Delineated Intersection Delineated Intersection Axial  1 8.0 (1.8%) 39.5 (8.9%) 100.0 (22.5%) 126.5 (28.4%) 87.9 (19.8%) 123.6 (27.8%)  2 13.9 (4.1%) 23.0 (6.7%) 84.7 (24.8%) 101.2 (29.7%) 76.5 (22.4%) 97.6 (28.6%)  3 170.1 (35.7%) 145.7 (30.6%) 354.2 (74.4%) 355.9 (74.8%) 344.5 (72.4%) 353.7 (74.3%)  4 53.7 (13.6%) 34.0 (8.6%) 153.2 (38.9%) 119.1 (30.3%) 142.1 (36.1%) 105.6 (26.9%)  mean±SD 61.4±75.2 (13.8%) 60.5±57.2 (13.7%) 173.0±124.3 (40.2%) 175.6±120.6 (40.8%) 162.7±124.7 (37.7%) 170.1±122.9 (39.4%)  Sagittal  1 49.8 (11.2%) 93.4 (21.0%) 43.2 (9.7%) 89.1 (20.0%) 43.5 (9.8%) 93.1 (20.9%)  2 8.5 (2.5%) 25.1 (7.4%) 15.0 (4.4%) 0.6 (0.2%) 15.5 (4.6%) 2.3 (0.7%)  3 98.7 (20.7%) 71.6 (15.0%) 129.9 (27.3%) 104.9 (22.0%) 130.1 (27.3%) 99.7 (20.9%)  4 33.6 (17.1%) 47.3 (12.0%) 66.0 (16.8%) 79.2 (20.1%) 67.3 (17.1%) 74.8 (19.0%)  mean±SD 47.7±38.1 (10.7%) 59.4±29.6 (13.9%) 63.5±48.9 (14.5%) 68.4±46.4 (15.6%) 64.1±48.8 (14.7%) 67.5±44.7 (15.4%)    1 mm2 25 mm2 50 mm2 75 mm2 100 mm2 125 mm2  Figure 5-3: Scale of areas to orient reader to magnitude of errors.  107  Table 5-3: Mean ± standard deviation difference in distances (mm) between MRI and reference standard centroids (3 dye trials) and percentage overlap of MRI- and reference standard areas. Differences between MRI and reference standard centroids are expressed using the anatomically based coordinate system used for kinematic analysis.  Specimen Centroid Distance (mm) % Overlap  Δanterior Δlateral Δproximal length 1 -0.3± 0.1 0.8±0.0 -3.8±0.4 3.9±0.4 64.4±0.8 2 0.48±0.4 -1.0±1.1 -1.0±0.6 1.8±0.6 87.5±4.0 3 0.7±0.1 0.8±0.3 -2.5±0.2 2.7±0.2 82.4±1.7 4 0.5±0.2 0.1±0.6 1.2 ±0.4 1.4 ±0.4 93.2±2.0 Overall mean 0.4±0.4 0.2±0.9 -1.5±2.1 2.4±1.1 82.6±12.4  LL LL M M MM (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) Specimen 1 Specimen 2 Specimen 3 Specimen 4  Figure 5-4: Centroids of the MRI and three trials of the reference standard contact areas. The blue lines represent the 3 reference standard trials while the blue marker represents the mean reference standard centroid.  The red line and marker represent the MRI based assessment of contact using the sagittal scan and the delineation method.  M = medial, L = lateral.  108  Specimen 1 Specimen 2  Specimen 3 Specimen 4 Figure 5-5: Percentage overlap between MRI-based and reference standard-based assessment of contact area. Purple area represents overlap, blue line represents MRI-based contact area, black line represents reference standard contact area (1 representative trial).   Table 5-4: Intra-reader error expressed as the mean standard deviation and the mean standard deviation ± standard deviation (mm2). Scan Direction Specimen Multiply by Slice Thickness Linear interpolation Spline Interpolation   Delineation Intersection Delineation Intersection Delineation Intersection Axial  1 5.4 39.6 7.4 46.4 7.2 46.5  2 9.2 7.7 9.8 4.7 9.9 3.9  3 13.8 15.9 36.5 5.4 36.7 6.7  4 5.0 20.1 7.8 47.0 6.0 47.5  mean 8.4±4.1 20.8±13.1 15.4±14.1 25.9±24.1 15.0±14.6 26.2±24.1  Sagittal  1 15.0 20.8 11.0 20.0 10.6 19.7  2 7.6 31.8 7.9 333.6 8.0 32.2  3 10.0 21.3 11.4 20.3 11.3 20.3  4 4.9 13.0 6.3 17.8 7.1 15.7  mean 9.4±4.3 21.7±7.7 9.1±2.5 22.9±7.1 9.3±2.0 22.0±7.1  109  0 100 200 300 400 500 600 -15 -5 5 15 25 35 45 55 65 Tibiofemoral Flexion (degrees) C on ta ct  A re a (m m ^2 ) S01 - Trial 1 S02 - Trial 1 S01 - Trial 2 S02 - Trial 2 S01 - Trial 3 S02 - Trial 3  Figure 5-6: Contact area intra-subject repeatability results presented for three trials for two subjects. Repeatability was assessed over the coincidental range of the three trials only.  Raw data are denoted by markers, splines are denoted by lines.  This graph uses the spline interpolation area calculation technique; however, results were similar for all techniques.  110  L L L L L L M M M M M M (mm) (mm) (mm)(mm) Subject 1 Subject 2 Trial 1 Trial 2 Trial 3 Trial 1 Trial 2 Trial 3  Figure 5-7: Contact area contours and centroids for three trials in two subjects. Contours and centroids were colour coded with respect to the knee flexion angle at which they were assessed, which differed between trials because of the inherent error in the goniometer used for positioning and the use of the calculated flexion angle in the analysis.   111  5.4 Discussion In this study, short axial and sagittal MRI scans were developed to assess articular cartilage contact areas in vivo during loaded knee flexion.   The area calculations from the sagittal scans consistently showed better agreement with the dye-based, cadaver model reference standard than the axial scans.  The direct delineation method showed better agreement than the intersection method. The intra-reader repeatability results also showed that the sagittal scans using the delineation method was most precise.  Intra-subject repeatability was assessed from the sagittal scans and determined to be similar to that reported previously [168].  Contact area centroid locations were also determined to be a useful for describing the location of the centroid with respect to the patella, when plotted in an anatomically based coordinate system.  These results of this study suggest that sagittal scans and the delineation method should be used to assess patellofemoral contact areas and area locations in vivo during loaded knee flexion using MRI.  The errors reported for the sagittal scans were similar to results reported in the literature; however, the scan used in this study was substantially shorter.  A cadaver-based validation study that reported an error of 13% (expressed as CV%) used a sagittal scan and a delineation and multiplication by slice thickness technique to calculate contact area [167].  However, the scan sequence used was 11 minutes long.  This study also reported an average percentage difference 10.9%, which is similar to the average percentage difference for similar techniques in the present study (10.7%).  A validation study that utilized a contact phantom and a scan sequence of 2 minutes 13 seconds had superior agreement with the reference (CV%=3%); however, this is to be expected due to the symmetry of the phantom (circular contact) and the difference in the signal obtained from the contact and surrounding material (the boundary was clearly discernable).  The intra-reader repeatability found in the present study (less than 9.4 mm2 for the sagittal scans) is in the range of what has been reported previously (1.3 mm2 with an axial scan 39 seconds in length [171] and 21 mm2 with an axial scan 6 minutes and 30 seconds in length [170]).  The intra-subject repeatability was found to be 8.3% in 1 human subject with a sagittal scan 4 minutes and 26 seconds in length [172].  For direct comparison the CV% for intra-subject error can also be calculated in the present study (expressed as the mean CV% sampled at 1° increments along the spline) and are slightly higher in the present study (between 10.4% and 11.3% for the three calculation techniques).  This may be due to the fact that the knee flexion angle used was based on a goniometer measure, not a calculation and that only 30° and 90° angles were assessed [172].  The spline fitting technique appears to cause the data to diverge at certain points and therefore may overestimate the actual repeatability error.  112   The error of agreement with the reference standard reported in the present study had quite a large standard deviation, highlighting some inconsistencies in results between specimens.  Specimen 2 displayed the best agreement for all calculation techniques using the sagittal scans and the best or second best agreement for the axial scans.  Specimen 3 had the largest errors.  This specimen was very large and had thick cartilage with signs of degeneration and therefore discerning the cartilage periphery was noticeably more difficult.  For specimens 3 and 4, the intersection method sometimes showed better agreement with the reference than the delineation method; however, since the delineation assessment was shown to be more repeatable than the intersection method using the delineation method is still suggested. The differences between MRI contact area centroid and reference standard centroid locations was small, while differences in contact patch overlap were larger than expected.  The largest errors in the centroid location results were in the proximal-distal direction, with three of the four specimens having MRI-based centroids positioned more proximal than the centroids from the reference standard. Not surprisingly, the contact patch overlap erred in the same direction in the same three specimens.  It would be reasonable to find that the MRI-based area was consistently overestimated around the entire periphery due to the resolution of the image; however, this was not the case.  While this was surprising, there are several potential reasons for this.  First, it is possible that the registration error (up to 0.4 mm) between the laser-based surface and the MRI-based surface accounts for some of this difference.  Another source of this error could originate in cadaver model.  MRI signal intensity is dependent on the water (hydrogen) content of the tissue being imaged.  It is not clear if the distribution of water within the cartilage changes when the joint capsule is opened; a pressure differential between the cartilage and its surroundings is created, likely changing the water distribution.  Although attempts were made to keep the cartilage hydrated prior to MRI assessment by spraying the surface with saline, once the dye was introduced into the joint, the patella was allowed to dry in order to avoid dye leaching.  Further, the water distribution likely changed with time when the cartilage was subjected to load (as was shown previously [149]) and therefore some of the error may result from actual differences in area.  It should also be noted that there was some visible fibrillation of the patellar articular surface of Specimen 3.  The dye appeared darker in fibrillated regions, suggesting that staining and leaching behaviour may have differed in these regions.  This may also explain some of the differences in the contact patch overlap data.  It is unclear how the fibrillated cartilage appears on MRI, but it is reasonable to expect the water content to differ from healthy cartilage and therefore may appear different in the image.  Finally, every attempt was made to not alter the loading setup while the loading rig containing the specimen was retracted from the MRI  113  scanner; however, it is possible that the suspended load moved slightly, which would have had a minor effect on contact area.  While the results show that the multiplication by slice thickness technique provides the best agreement with the reference standard, the linear and spline interpolation techniques may provide a better description of the actual cartilage surface topology.  The difference in error between calculation methods was more apparent when measurements were made from the axial scans (13.8, 40.2 and 37.7% for the multiplication by slice thickness, the linear interpolation and spline interpolation for the delineation method) as compared to the sagittal scans (8.5, 14.5 and 14.7%, respectively in the same order).  Theoretically, the multiplication by thickness technique will underestimate the contact area when compared to a linear interpolation method.  This can be best visualized in two dimensions as the contact area analysis using the adjacent rather than the hypotenuse of the right-angled triangle created between slices.  The greater differences in the axial results are likely a result of the greater change in topology in the slice direction than in the sagittal plane and the considerably less information available to describe the periphery (Figure 1-17).  From the quantitative values of area, the MRI assessment was consistently greater than the reference standard; however, from visual inspection and the contact patch overlap data it is clear that the overestimation was not consistent around the entire periphery.  Some of the overestimation of the MRI-based assessment was therefore likely moderated by the underestimation of the multiplication by slice thickness technique.  Nonetheless, the multiplication by slice technique will not provide an adequate description of the surface topology; however, this seems to be of greater concern in axial assessments.  The dye technique was the most suitable reference standard for the present study.  While the various ex vivo techniques of assessing contact area (dye staining, casting, pressure sensitive film and stereophotogrammetry) have shown agreement previously [145], there were several reasons that the dye staining technique was chosen.  First, it was important that no additional material be introduced into the joint capsule that might affect the MRI scan or the properties of the mating surfaces; the aim was to keep the ex vivo scenario as close to the in vivo scenario as possible, therefore the pressure sensitive film was not selected. Dynamic pressure measurement systems were not chosen for similar reasons, and because the electronics could not be used due to the MRI magnetic field.  Next, the MRI centre has scent restrictions (one of the employees has a scent sensitivity) and therefore it was not possible to use the casting technique because an odour is produced during curing.  Further, additional data collection and analysis would be required to relate locations of contact areas between the cast and the MRI.  Finally, the stereophotogrammetry requires special equipment and the loading setup to be altered and therefore was not useable in this study.  The dye technique also has several advantages,  114  highlighted in Table 1-4, including that very little equipment is required to create the print, the equipment required is MRI compatible, characterization of the area does not need to be done immediately (it was possible to return to the lab to complete the assessment), the loading setup did not need to be altered between assessments and finally that the position of the contact area with respect to the patella was known (allowing for calculation of contact area overlap and centroid locations). The patterns of proximal translation and increase in contact area with knee flexion observed in the intra-subject repeatability results are similar to those reported in the literature previously.  In both the in vivo and ex vivo literature, the reported contact area migrates proximally with knee flexion [12,14,92,93,116,124,152,175].  This was generally the case for the intra-subject repeatability analysis.  Subject 2 also displayed some tendo-femoral contact (contact area visible above the top end of the patella), which does not usually occur until 120° of knee flexion in cadaver studies [152].  On the MRI scan it appeared as contact according to the definition described in the methods section and therefore was included.  This was not entirely surprising because it appears as though this participant had patella baja (a distally located patella).  Even near full extension the contact areas were positioned quite proximally on the patella, especially compared to Subject 1.  The results also suggest that Subject 1 has substantial hyperextension of the knee, although this is not necessarily the case since flexion here is described between the anatomical, not the mechanical axes.  Correction would be possible if the hip and ankle were scanned also but this is outside the scope of this study and is not of great concern because coordinate systems were defined in the same manner for all participants.  The variation between trials may also be due to measurement errors (up to 65 mm2); however, it is also possible that the patella was positioned differently between trials due to potential differences in muscle recruitment and fatigue.  The primary strength of this study is that the sagittal MRI scan developed is the fastest scan to be validated to date and the errors of the contact area assessment were not compromised.  This scan can be used in series with the scan sequence required for kinematic assessment while still maintaining a total imaging time of 1 minute and thirty seconds.  Other methods that have also assessed kinematics simultaneously require scans of between 4 and 7 minutes (although these methods haven’t been well validated) [104,116,169,172] which is a problem because some participants, especially those with patellofemoral disorders, may not be able to maintain the loaded position for this length of time.  One limitation of the study was that, ideally, the joint capsule would not be opened because as a result the contact areas measured are likely larger than those that would be experienced in vivo; however, since this is a validation study it is not necessary that the contact area be biofidelic.  It is  115  sufficient that the areas are the same between the reference standard and MRI-based assessments. There was, however, a time lapse of approximately 5 minutes between the MRI-based and reference standard assessments and therefore there may be small differences in absolute contact area between assessments [149].  Further, once the dye was applied, the cartilage surface was dried and there was an additional time lapse for transportation of the specimen from the MRI centre to the biomechanics laboratory (approximately 20 minute drive plus loading and unloading time), therefore the contact area digitization and the surface laser scan data were acquired a minimum of 1 hour after dye application.  Because of the time lapse involved in transport, one concern during the study design phase was the potential for dye leaching.  A dye leaching test, in which contact contours were delineated at 10 minute intervals over a two hour period and then plotted in a temporal manner according to a colour spectrum map, was shown qualitatively to be not a concern (Appendix H). Further even in regions where dye leaching was suspected the leached dye was a much lighter colour and the dark contour of the original boundary was still evident.  There was also error associated with digitizing the periphery of the dye-based contact area.  The end of the digitizing probe is a 1 mm diameter ball bearing which had to be lightly placed on the cartilage surface so as not to compress the cartilage.  An attempt was made to minimize this error by using the average of 3 digitization trials. The quality of the laser scan obtained was also dependent on several factors.  First, the dye stain was much darker in regions of cartilage fibrillation which appeared as holes in the laser scanned surface because of its lack of reflectance.  Second, the contact regions devoid of dye were quite shiny, also affecting the quality of the laser scan.  To account for these differences in reflectance, a dusting of talcum powder was applied to the cartilage surface creating a more uniform reflectance and colour across the cartilage surface.  This allowed scans free of holes to be created while maintaining surface topology.  This is a common technique used in laser scanning.  The registration of the digitized area to the laser surface was carried out using the fiducials markers as an initial position estimate.  This was done using a single point on the top of each fiducial and may not perfectly align with the point chosen from the laser scanner point cloud.  However, the fiducials points were included in the ICP assessment and therefore were allowed to find their optimal position based on the algorithm.  A sensitivity analysis showed that variation in choosing these points did not affect the convergence to the correct solution.  Another limitation was that, for the delineation method, it was sometimes difficult to distinguish between cartilage and surrounding tissues (joint capsule, fat, other small pieces of tissue that may have migrated to the region as a result of dissection), in particular at the proximal, distal, medial and lateral margins of the patella.  This would be expected this to cause an overestimation of the contact area. Finally, intra-subject repeatability was assessed in only two subjects because the imaging took in excess of 3 and a half hours to complete.  This was therefore  116  quite onerous for the participants and fatigue may contribute to differences between trials.  Intra- subject repeatability was only assessed in healthy subjects and the applicability of these measures to diseased or injured participants is not clear. This protocol yields assessments of contact area in less than a minute that have errors similar to those made using scans many times longer.  Further, this sequence can be used in series with kinematic scans to assess contact areas and kinematics in vivo using MRI.  117  6 Model Development & Validation  SYNOPSIS: Because it is not always possible to assess contact areas directly in vivo, there is a need for a simple tool that predicts contact area from kinematic data.  The aim of the chapter was to develop and validate such a tool.  The sensitivity of this tool to the kinematic input was also assessed.  6.1 Introduction  Ideally, patellar kinematics and contact areas would always be assessed simultaneously in vivo; however, there are some situations in which direct measures of contact area are not possible.  In these cases an estimate of contact area from kinematic data would be useful; for example, in studies where kinematics are assessed during continuous motion using bi-planar radiography [101] or in clinical studies where joint loading tasks must be kept to a minimum because study participants have patellofemoral joint disease, such as OA.  Therefore, there is a need for a simple, patient specific tool that can predict contact areas in a manner that does not require a great amount of expertise, additional analysis, additional imaging or additional computational time.  With three-dimensional models of the patellofemoral joint it is possible to make predictions of contact area [84,180-187]; however, current models suffer from several limitations such as a lack of validation and the requirement of extensive analysis and computation time.  As such, these models have seldom been used to answer clinical research questions, as discussed in Section 1.6.4.  Models developed to date fall within the broad categories of finite element models [84,185-187] and multibody models [180-184].  An advantage of multibody models is that they are computationally less expensive and therefore more practical for use in answering clinical questions.  Further, because of the inherent variability between individuals in joint geometry, structure and function, many groups have chosen to apply their models in a patient specific manner. Validation of predictions of contact area from these models has been limited.  The only group to validate contact area predictions found the errors to be 2.3% [186] and 5% [190] using a patient specific, kinematic and EMG driven finite element model.  The reference standard used in this case was an in vivo MRI-based measure assessed during weightbearing flexion using an open-bore MRI scanner [166].  This study suggests that it is  118  possible to predict contact areas using a computational model, but this model requires sophisticated inputs and considerable expertise to run.  Simple models have been used to predict contact area in the tibiofemoral [138] and patellofemoral [93] joints.  The tibiofemoral joint model is a kinematics-driven model in which the proximity of the bones is assessed; cartilage is not included [138].  The authors describe this proximity quantity as a functional joint space, which originates from the idea of measuring joint space narrowing in OA (Section 1.3.2).  However, it is not clear how bone proximity relates to cartilage contact areas.  At the patellofemoral joint, a model that predicts contact area by calculating the overlap of cartilage surfaces has been developed [93].  In this model, cartilage surface geometry is obtained from cadaver specimens by digitizing the articular surface at 5 mm intervals and fitting a surface to the digitized points.  The surfaces are positioned according to kinematic data, distances of the femoral surface normal to the patellar surface at discrete points are calculated and contact is considered to be within a particular threshold.  It is not clear if the overlap adequately represents cartilage contact because the model was not validated.  Further, this model cannot be applied in vivo.  There is currently a need for a simple model to predict contact areas from in vivo kinematic data; therefore, the aim of this chapter was two-fold: 1. To develop and validate a simple, patient specific, kinematics-driven, computationally inexpensive multibody model of the patellofemoral joint to predict contact areas that is appropriate for use in clinical studies 2. To understand the sensitivity of the model to kinematic input data 6.2 Methods 6.2.1 Model Development  A simple patient specific, kinematics-driven multibody model of the patellofemoral joint was developed to predict articular cartilage contact areas and centroids using custom written software (Matlab, The Mathworks, Natick, MA, USA).  The model required inputs of cartilage surface geometry, bone geometry and three-dimensional patellar kinematics.  Patellar and femoral cartilage surface geometry was obtained by segmenting the cartilage surface in a slice-by-slice manner from unloaded sagittal MRI cartilage images (Table 5-1) using the image edit module in commercially available image processing software (Analyze 8.1, AnalyzeDirect, Overland Park, KS, USA). Cartilage contours were extracted using the surface extractor module.  Bone geometry (models) and three-dimensional patellar kinematic data were obtained using the method described in Section 2.2.3.  119  In addition to the standard assessment of kinematics, bone contours were extracted using the segmented bone model data from the unloaded bone scans.  To prepare the cartilage surfaces for use in the model, a surface mesh was fit to the femoral cartilage contours and splines were fit to the patellar cartilage contours.  A fine surface mesh was fit to the femoral cartilage surface according to the same procedure used to mesh the contact surfaces for the area calculation in Section 5.2.3.  Specifically, splines were fit to the in-plane data and sampled at 1000 points along the length and then splines were fit between slices and samples at 40 points between each slice (Figure 6-1).  In-plane splines were fit to the patellar cartilage contours and sampled at 1000 points along the spline; no between plane splines were fit to the patellar cartilage data in order to be consistent with Chapter 5 in which contact areas are calculated from MRI contact contours.  Figure 6-1: Femoral cartilage original contour data (yellow) and mesh (blue).  The femoral cartilage mesh and the patellar cartilage contours were registered to their respective bone models.  Since the unloaded bone and cartilage scans were acquired with the knee in the same position, the relationship between these tissues is known and therefore no registration is required.  However, the surface extractor module of the image processing program provides the contour information in terms of pixel and slice numbers and the bone model information in terms of scaled coordinates.  Therefore, registration between the bone model and cartilage contour was required.  This was accomplished by registering bone contours (that are in the same space as the cartilage contours) to the bone models using an ICP algorithm [208].  The homogeneous transformation matrix between the bone contours in the original and the registered position was  120  determined using a least squares based algorithm [218].  This homogeneous transformation matrix was then applied to transform the femur cartilage mesh and the patellar cartilage contours to the bone model space.  An in vivo three-dimensional patellar kinematic analysis was carried out [102,132], the results were input into the model and cartilage contact areas in the loaded joint position were determined.  During the kinematic analysis homogeneous transformation matrices were obtained to transform the bone models to the bone contours describing bone positions in the loaded configuration (Section 2.2.3).  This same homogenous transformation matrix was used to transform the femur cartilage mesh and the patellar cartilage contours to the loaded position.  To determine the overlap of the cartilage surfaces, a proximity analysis was carried out in a slice-by-slice manner (Figure 6-2). For each point in the patellar cartilage contour slice, the minimum Euclidean distance to the femoral mesh was determined.  All points on the surface with a distance less than the proximity threshold (to be discussed in the next section) were considered to be in contact.  The periphery of the contact area was defined as the most proximal and most distal point meeting the proximity threshold requirement of the contour (Figure 6-3).  The final contact area included all points in the sagittal plane between the proximal and distal contact points in order for regions of extreme overlap to be accounted for (Figure 6-3), resulting in contact contours similar to those that would be obtained by delineating contact from a loaded MRI scan directly (Figure 6-4).  Contact area (using the multiplication by slice thickness, linear interpolation and spline interpolation techniques) and centroids (expressed in the anatomical patellar coordinate system) were calculated using the same method described in Section 5.2.3.  Figure 6-2: Bone contours (blue and red), femoral cartilage mesh (turquoise) and patellar cartilage contours (magenta) in loaded kinematic position.   121  Threshold Proximal Contact Point Distal Contact Point All Contact  Figure 6-3: Proximity analysis for one slice. The red line represents the proximity threshold value.  The most proximal and distal points that cross the proximity threshold are considered the contact contour boundaries.  All points on the slice between the proximal and distal boundary points are considered contact.  Figure 6-4: Femoral cartilage mesh (blue), patellar cartilage contours (yellow) and periphery of contact contour (magenta).   122  6.2.2 Model Validation  Four models were created and validated using the ex vivo cadaver specimen data and contact area results from Chapter 5.  The MRI- and dye-based assessments of contact area determined in Section 5.3 were both used for reference comparisons of model predictions.  Differences between the prediction of contact area from the model and the reference standards (MRI- and mean dye-based assessments of contact area) were expressed as the mean absolute and percentage difference. Centroids were also calculated and expressed as the mean absolute difference from the reference standards.  Sensitivity of the model to proximity threshold value and to errors in the kinematic input data was assessed.  The value of the proximity threshold required to obtain an adequate prediction of contact area from the model was unknown at the outset of the validation study.  Therefore, a sensitivity analysis was required in which the proximity threshold value was incrementally changed and the contact area was re-calculated.  Optimal proximity threshold values were then determined. Sensitivity of the model to kinematic input was also assessed.  For each kinematic parameter (flexion, tilt, and spin; proximal, lateral and anterior translation) the position of the loaded patella was changed by 1° or 1 mm increments between ± 3° or 3mm from the imaged position in a manner that was physiologically possible (for example, there could be no femoral bone-patellar bone contact).  This range was chosen because errors in rotation can be up to 3° (Table 1-3) and for consistency 3 mm was chosen for translations.  The modified Joint Coordinate System was used to define the orientation and the anatomical femoral coordinate system was used to define position.  Contact area and centroid were predicted from the model at each orientation and position and plotted as a function of distance from the imaged position for each kinematic parameter. 6.3 Results The percentage absolute differences in predictions of contact area obtained from the model were between 3.8% and 3.9% compared to the sagittal, delineated MRI-based contact area measure and between 4.1% and 11.6% compared to the dye-based contact area measure (Table 6-1 and Table 6-2, respectively, and Figure 6-5).  The proximity threshold sensitivity analysis determined that thresholds of 1.1, 0.06, 0.5 and 0.5 mm be used for specimens 1 through 4, respectively (for more detail, see below).  The errors were smaller when comparisons were made to the MRI-based reference standard because proximity thresholds were optimized for this case.  The original imaged position of the models was 18.5, 28.4, 27.4 and 30.2° of tibiofemoral flexion for Specimens 1 through 4, respectively.  123  The position of the centroids estimated from the model was within 2.1 mm of the reference standard results (Table 6-3 and Table 6-4, respectively, and Figure 6-5).  The difference in proximal/distal position between the model estimate and the reference standards was the largest (1.5 and 1.6 mm for the MRI- and dye-based reference standards, respectively), while the difference in anterior/posterior position was smallest (0.3 and 0.2 mm for the MRI- and dye-based reference standards, respectively).  Table 6-1: Errors of model estimates as compared to the delineated sagittal MRI-based measures of contact area.  Multiplication by thickness Linear Interpolation Spline Interpolation Specimen 1 0.4 mm2 (0.1%) 19.8 mm2  (4.9%) 16.3  mm2 (4.1%) Specimen 2 5.3 mm2 (1.6%) 0.5 mm2  (0.1%) 4.2 mm2 (1.2%) Specimen 3 90.3 mm2 (15.7%) 50.0 mm2  (8.3%) 54.2 mm2 (8.9%) Specimen 4 23.5 mm2 (5.5%) 10.2 mm2 (3.6%) 6.9 mm2 (1.5%) Mean absolute difference 29.9  ± 41.5 mm2 (5.7%) 20.1 ± 21.4 mm2 (3.8%) 20.4 ± 23.1 mm2 (3.9%)  Table 6-2: Errors of model estimates as compared to the dye-based measures of contact area.  Multiplication by thickness Linear Interpolation Spline Interpolation Specimen 1 49.4 mm2 (11.1%) 23.3 mm2  (5.2%) 27.2  mm2 (6.1%) Specimen 2 3.1 mm2 (0.9%) 14.5 mm2  (4.3%) 11.4 mm2 (3.3%) Specimen 3 8.4 mm2 (1.8%) 79.9 mm2  (16.8%) 75.9 mm2 (15.9%) Specimen 4 10.1 mm2 (2.6%) 76.2 mm2 (19.4%) 74.2 mm2 (18.9%) Mean absolute difference 17.8 ± 21.3 mm2 (4.1%) 48.5 ± 34.4 mm2 (11.6%) 47.2 ± 32.8 mm2 (11.1%)       124  Table 6-3: Mean difference in centroid position as compared to the sagittal delineated MRI-based reference standard.  Δanterior (mm) Δlateral (mm) Δproximal (mm) Length (mm) Specimen 1 0.3 0.5 1.6 1.7 Specimen 2 -0.5 0.5 -1.7 1.8 Specimen 3 -0.4 1.5 1.5 2.1 Specimen 4 0.0 -0.1 -1.5 1.5 Mean absolute difference 0.3 ±0.2 0.6 ± 0.6 1.5 ± 0.1 1.8 ± 0.3  Table 6-4: Mean difference in centroid position as compared to the dye-based reference standard.  Δanterior (mm) Δlateral (mm) Δanterior (mm) Length (mm) Specimen 1 -0.0 -1.3 2.2 2.6 Specimen 2 -0.0 0.5 2.7 2.8 Specimen 3 -0.3 -2.3 1.0 2.5 Specimen 4 -0.5 -0.1 0.3 0.6 Mean absolute difference 0.2 ± 0.2 1.1 ± 1.0 1.6 ± 1.1 2.1 ± 1.0    125  (mm) (mm) Specimen 1 Specimen 2 Specimen 3 Specimen 4  Figure 6-5: Model (yellow) and the MRI-based (red) and mean dye-based (blue) reference standard of contact region in the four cadaver specimens. Note: centroid for mean dye-based assessment is not visible because it is directly behind the other centroids (Table 6-3 and Table 6-4).  6.3.1 Proximity Threshold Sensitivity The contact area predicted is sensitive to proximity threshold value (Figures 6-6 to 6-9). Threshold values appear to be patient specific.  Contact area centroids were less sensitive to threshold value (Appendix I).  126  0 50 100 150 200 250 300 350 400 450 500 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) ar ea  (m m ^2 ) Model-Spline Interpolation Model-Linear Interpolation Model- Multiply by Thickness MRI-Spline Interpolation MRI -Linear Interpolation MRI-Multiply by Thickness Mean Dye Figure 6-6: Proximity threshold sensitivity for Specimen 1. 0 100 200 300 400 500 600 700 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) ar ea  (m m ^2 ) Model-Spline Interpolation Model-Linear Interpolation Model-Multiply by Thickness MRI-Spline Interpolation MRI-Linear Interpolation MRI-Multiply by Thickness Mean Dye  Figure 6-7: Proximity threshold sensitivity for Specimen 2.  127  0 100 200 300 400 500 600 700 800 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) ar ea  (m m ^2 ) Model-Spline Interpolation Model-Linear Interpolation Model- Multiply by Thickness MRI-Spline Interpolation MRI -Linear Interpolation MRI-Multiply by Thickness Mean Dye Figure 6-8: Proximity threshold sensitivity for Specimen 3. 0 100 200 300 400 500 600 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) ar ea  (m m ^2 ) Model-Spline Interpolation Model-Linear Interpolation Model- Multiply by Thickness MRI-Spline Interpolation MRI -Linear Interpolation MRI-Multiply by Thickness Mean Dye Figure 6-9: Proximity threshold sensitivity for Specimen 4.     128  6.3.2 Kinematic Input Sensitivity Contact area estimates from the model were sensitive to kinematic input, in particular for patellar flexion and proximal and anterior translation.  Contact area increased by approximately 90 mm2 per degree of patellar flexion (Figure 6-10), increased by approximately 80 mm2 per mm of proximal translation (Figure 6-11) and decreased by approximately 170 mm2 per mm of anterior translation (Figure 6-12).  Contact area decreased with medial tilt at a rate between 10 and 50 mm2 per degree of medial tilt in Specimens 1-3 but in Specimen 4 there was a slight peak in contact area at the imaged position (Figure 6-13).  When patellar position was varied in the medial-lateral direction, a slight peak was observed for Specimens 2-4 (Figure 6-14).  Specimen 1 displayed increasing area with lateral translation.  Finally, contact area remained relatively constant for changes in patellar spin in Specimens 1, 3 and 4 while for Specimen 2 contact area decreased at a rate of approximately 17 mm2 per degree of internal spin (Figure 6-15).  Area calculations for Figures 6-10 to 6-15 were carried out using the spline interpolation method; similar results were obtained using linear interpolation or multiplication by slice thickness methods.  A representative figure of contact areas and centroids is presented for Specimen 2 (Figure 6-16). Contact area centroid positions were not as sensitive to kinematic input, except for proximal translation.   For proximal-distal translation the centroid migrated distally at a rate of approximately 1.0 mm per mm of proximal patellar translation (Figure 6-17). All other results were either relatively constant or varied with no observable pattern (Appendix J).  129  0 100 200 300 400 500 600 700 800 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) ar ea  (m m ^2 ) Specimen 1 Specimen 2 Specimen 3 Specimen 4 Extension Flexion  Figure 6-10: Sensitivity of patellar cartilage contact area to patellar flexion. The spline interpolation method was used for area calculation.   0 100 200 300 400 500 600 700 800 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) ar ea  (m m ^2 ) Specimen 1 Specimen 2 Specimen 3 Specimen 4 Distal Proximal  Figure 6-11: Sensitivity of patellar cartilage contact area to proximal/distal translation. The spline interpolation method was used for area calculation.   130  -100 0 100 200 300 400 500 600 700 800 900 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) ar ea  (m m ^2 ) Specimen 1 Specimen 2 Specimen 3 Specimen 4 Posterior Anterior   Figure 6-12: Sensitivity of patellar cartilage contact area to anterior/posterior translation. The spline interpolation method was used for area calculation.   0 100 200 300 400 500 600 700 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) ar ea  (m m ^2 ) Specimen 1 Specimen 2 Specimen 3 Specimen 4 Lateral Medial  Figure 6-13: Sensitivity of patellar cartilage contact area to patellar tilt. The spline interpolation method was used for area calculation.   131  0 100 200 300 400 500 600 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) ar ea  (m m ^2 ) Specimen 1 Specimen 2 Specimen 3 Specimen 4 Medial Lateral  Figure 6-14: Sensitivity of patellar cartilage contact area to patellar medial/lateral translation. The spline interpolation method was used for area calculation.   0 100 200 300 400 500 600 700 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) ar ea  (m m ^2 ) Specimen 1 Specimen 2 Specimen 3 Specimen 4 External Internal  Figure 6-15: Sensitivity of patellar cartilage contact area to patellar spin. The spline interpolation method was used for area calculation.  132  (mm) (mm) LM LM LM LM LM LM Flexion Spin Tilt Anterior Translation Lateral Translation Proximal Translation  Figure 6-16: Contact contours and centroids for Specimen 2 for changes in each of the 6 kinematic parameters. The positive values of the legend correspond to flexion, internal spin, medial tilt, and proximal, lateral and anterior translation.    133  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) pr ox im al /d is ta l c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Distal Proximal Distal Proximal  Figure 6-17: Sensitivity of proximal/distal centroid position to proximal translation.  6.4 Discussion  In this study, a simple, patient specific, kinematics-driven multibody computational model for predicting patellofemoral articular contact areas was developed and validated.  The model was able to yield adequate predictions contact area if the correct proximity threshold was used.  It appears, however, that the threshold value is patient/specimen specific and that the contact area prediction is sensitive to the value of this threshold.  The contact area measurement was also sensitive to the kinematic input required by the model, in particular for patellar flexion, posterior translation and anterior translation which varied between 80 and 170 mm2 per degree or mm change in the kinematic parameter.  Centroid locations were less sensitive to the proximity threshold or the kinematic input data, except for the kinematic input of proximal translation.  The validation results from this study compare well to the one finite element model that has been validated for contact area.  In that model, contact area estimates were compared to measured values from in vivo MRI assessments during weightbearing using an open-bore MRI [166].  The errors in the model were assessed in two different studies and found to be 2.3% in a study of one healthy subject [16] and 5% in a study of 16 healthy subjects [18].  The errors in area prediction of the present study using the multiplication by slice thickness technique were similar to the validated  134  finite element model which also uses the multiplication by slice thickness technique (5.7%); however, errors using the linear interpolation and the spline interpolation techniques showed better agreement (3.8 and 3.9%, respectively).  The present study is the first to compare contact areas assessed using a model to contact areas assessed using an ex vivo reference standard, therefore there is no available literature for comparison.  It is not clear if the threshold value should be optimized for agreement with in vivo assessments (MRI-based reference) or with the ex vivo assessments (dye-based reference).  In this study, thresholds were optimized for agreement with the MRI-based reference standard.  Since the proximity threshold values varied between specimens, it is clear that a method must be developed to determine the patient specific threshold if this model is to be used in vivo.   Possible reasons for the threshold value variation include patellar size and type, cartilage morphology, quality of the cartilage (and therefore the ability to obtain a good MRI image from which to extract the cartilage morphology), resolution of the unloaded cartilage MRI scans (due to variation in patellar size between individuals), error propagation as a result of the registration of the bone contours to bone models and magnitude of load applied (the same load was applied to all knees regardless of size).  It is possible that a characteristic equation to calculate the appropriate proximity threshold could be developed using some or all of these variables.  Another, more straightforward solution would be to determine the proximity threshold value experimentally.  For example, kinematic and contact area data could be acquired at a single knee flexion angle using the technique described in Chapter 5 and the patient specific proximity threshold could then be determined.  This proximity threshold value could then be used to predict contact areas from the remainder of the kinematic data (so that contact area scans are not required at all knee flexion angles) thereby reducing the amount of time the participant is required to maintain the loaded flexion angle during scanning.  It is not clear if the previous study that used a proximity threshold suffered from the same variation in proximity threshold [93]. The proximity threshold in that study was referred to as ‘μ’; however, the value of ‘μ’ was never explicitly stated.  The results show that the model is sensitive to the kinematic input, highlighting the need for valid kinematic data.  Errors in the kinematics method used in this thesis are on the order of 1° and 1 mm, which for patellar flexion and proximal and anterior translation would equate to a very large error in contact area measure (up to 140 mm2).   However, other MRI-based methods have greater errors (Table 1-3) and therefore the kinematic inputs would not be sufficiently robust for use in this model.  Kinematic assessments made with bi-planar radiography tend to have smaller errors than MRI-based measures (Table 1-3) and therefore, they are likely more suitable for this type of model. However, an MRI scan would be required to create the cartilage surface models.  Also, one of the  135  methods for determining the patient specific proximity threshold would be required.  This study was the first to examine the sensitivity of contact area to kinematic inputs.  It is possible that other models suffer from the same sensitivity problems. Many of the findings observed by incrementally changing the position of the models in three- dimensions were not surprising.  For example, it is intuitive that patellar contact areas increase as the patella flexes, translates posteriorly and translates proximally by a small amount if the starting position is between 20 and 30°.  However, the magnitude of this change was larger than expected. Also, it is not surprising that three of the four specimens reached a peak in area for medial-lateral translation; this peak likely represents the position in which the patellar and femoral mating surfaces are optimally aligned.  Specimen 1 did not display a peak; it is possible that the optimal position for this specimen was outside of the range of values studied.  One might also expect a similar pattern for patellar tilt; however, this was only observed in Specimen 4.  This phenomenon may not be captured in tilt because of the potential for contralateral compartment ‘lift off’ during this motion; since the medial compartment is generally smaller than the lateral, the overall contact area will never be as large as when there is contact in both compartments.  Finally, contact areas remained relatively constant with patellar spin, with the exception of in Specimen 2 whose contact area decreased by approximately 100 mm2 over the range of angles studied.  These different patterns may be a result of differences in joint geometry. Although the aim of this study was to validate the model, the kinematic input sensitivity analysis can also be discussed in terms of the effect of small changes in kinematics on contact areas. Differences smaller than 1° or 1 mm have been observed between subject groups [134] or study conditions (Chapter 3 and 4).  It appears that small changes in kinematics may be important because of the proportionally greater changes in contact area that may result (in particular for patellar flexion, proximal translation and anterior translation).  A larger contact area is likely preferable because the joint load is distributed over a larger area thereby decreasing contact stresses.  Most mechanical-based interventions used to treat patellofemoral disorders to date, such as bracing and taping, have aimed to correct patellofemoral medial-lateral position and tilt.  However, the results of the sensitivity analysis suggest that greater changes in contact area will result if kinematics are altered slightly by flexing the patella or translating it proximally or posteriorly.  It is possible that interventions that alter kinematics in directions other than in tilt and medial-lateral translation may also be useful and appear to warrant further exploration.  This may be particularly useful in studying and treating individuals with patella alta or baja.  It must be noted, however, that the positions and orientations applied to the model are  136  not necessarily physiologic because the patellae are likely not in equilibrium.  In vivo these changes in kinematics would likely not occur in isolation, as they were studied here.  Strengths of this model include that it is relatively simple, computationally inexpensive and validated making it suitable for use to answer clinically motivated questions.  The only additional image analysis required, as compared to the standard kinematic analysis used throughout this thesis, is the delineation of the cartilage surfaces from the unloaded MRI scans.  However, the model is limited by the need for a patient specific proximity threshold and valid kinematic input data.  The sensitivity of the kinematic input was only assessed between 18 and 30°; it is not clear if sensitivity differs with knee flexion angle.  However, this was the first study of model sensitivity to kinematic input and a strength is that this limitation has been characterized.  Currently, only contact area predictions are possible; however, with further development other parameters may be predicted.  For example, if the magnitude of overlap of the cartilage plates were to be assessed, cartilage strains could be determined and contact stresses could be predicted using material properties and governing equations for deformation obtained from the literature.  This would require further validation using a pressure measurement system.  However, this is not a straightforward task due to the viscoelastic nature of cartilage.  This added complexity would also greatly increase the computation time required and therefore may not meet the requirement of this model to be computationally inexpensive and suitable for large clinical studies.  Also, in this study, the resolution of the unloaded cartilage MRI scan was 0.6 mm; it is possible that the error in the model may be reduced by using a higher resolution unloaded scan to create the cartilage contours.  Another limitation is that additional error was introduced into the system by having to register the bone contours to the bone models.  Further work could be done to eliminate this registration step by, for example, exploring different output methods in the surface extractor model or using a different commercial image processing package.  This study showed that it is possible to predict cartilage contact areas and centroids using a simple, patient specific, kinematics driven multibody model if a patient specific proximity threshold is used and if valid kinematic input data is available.  This model is suitable for use in studies where such kinematic data are available, such as in studies that employ bi-planar radiography, and in studies which require an estimate of contact area but direct assessments cannot be made because the participants are unable to complete the loading task.  In some cases, however, a direct analysis of contact area using the method developed in Chapter 5 may be more suitable and may yield a better estimate of area.  137  7 Integrated Discussion The aim of this chapter is to place this thesis into the context of the wider body of literature.  In the first section the findings of each research chapter will be summarized.  Next, in the general discussion section, the findings will be placed in context of the overarching themes of the research. Finally, the limitations of the thesis as a whole will be discussed, particular contributions will be highlighted, areas of potential future research will be identified and an overall conclusion for the work will be provided. 7.1 Summary of Findings A single measure of patellar alignment does not provide a surrogate marker of three- dimensional patellar kinematics: Since three-dimensional assessments of patellar kinematics are time consuming and expensive, the aim of Chapter 2 was to determine if a full assessment was required or if a surrogate marker provided an adequate description of kinematic behaviour at other flexion angles. Since patterns could not be predicted for all kinematic parameters, these data showed that a full assessment of kinematics is required.  This information was essential before using the method in clinical studies.  Further, these results put studies that use patellar alignment as a measure of mechanics into perspective. Measures of three-dimensional patellar kinematics are dependent on the magnitude of the applied load: The aim of Chapter 3 was to assess the effect of load magnitude on three-dimensional patellar kinematics because studies to date have used various loading protocols but the effect of this variation on findings was unclear.  The results of the study showed statistically significant differences in some patellar kinematic parameters (flexion, tilt, proximal translation and anterior translation) between the 0% and 30% load levels and the 15% and 30% load levels.  This study was the first to examine the effect of load magnitude on three-dimensional patellar kinematics in vivo.  These results are useful for interpreting the results of previous studies and for designing future studies. A patellofemoral brace alters three-dimensional patellar kinematics:  It was unclear if a patellofemoral brace would be able to alter kinematics in individuals with patellofemoral OA; therefore the aim of Chapter 4 was to determine the effect of a patellofemoral brace on three- dimensional patellar kinematics in this population.  Statistically significant differences in all kinematic parameters were observed when the brace was applied.  This study was the first to assess three-dimensional patellar kinematics in the patellofemoral OA population.  Since the parent study, of which these participants were a subset, found no improvement in pain (Appendix G), it appears as  138  though the magnitude of the kinematic change induced by the brace was insufficient to relieve patient symptoms. Contact areas can be assessed in vivo using a rapid MRI sequence:  For many applications, it is essential that kinematics and contact areas be assessed simultaneously, therefore the aim of Chapter 5 was to develop and validate a rapid MRI-based method of assessing patellofemoral joint contact areas in vivo that could be used in series with the current kinematic method. In this study, MRI scan sequences and area calculation methods were developed.  The errors in the method were determined to be similar to methods that require scans an order of magnitude longer.  Therefore, this method can be used to measure kinematics and contact areas simultaneously in individuals with patellofemoral joint disease, such as OA, because the scans are sufficiently short for the participants to complete a near physiological loading task. Contact area predictions from a simple, patient specific, kinematics-driven multibody model agree well with direct measures from MRI: There are several situations in which a simple prediction of contact areas from kinematics is required, such as in studies of kinematics using bi-planar radiography or in studies where joint loading time must be kept to a minimum.  Therefore the aim of Chapter 6 was to develop and validate a patient specific, kinematics-driven multibody model that could predict patellofemoral contact areas and assess its sensitivity to kinematics input.  The contact areas predicted from the model, based on determining the proximity of the patellar and femoral cartilage surfaces, were found to agree with direct MRI measures of contact area.  However, it was sensitive to proximity threshold and the kinematic input data.  This model can be used in situations where a patient specific proximity threshold can be determined and valid kinematic data are available. 7.2 General Discussion 7.2.1 Why are In Vivo Assessments of Patellofemoral Joint Mechanics Important? Ideally joint mechanics would always be assessed during activities of daily living; however, practically this is not possible.  In vivo, ex vivo and model-based assessments of joint mechanics all attempt to best replicate an individual’s normal joint function.  However, in vivo assessments have several advantages over ex vivo measures and modelling.  These advantages include taking into consideration the inherent variability in joint behaviour between individuals, the ability to relate mechanics to clinical symptoms and the ability to evaluate mechanical-based treatment strategies.  The patellofemoral joint is a site of substantial individual variation in joint kinematics, even in normal individuals without joint pathology.  This suggests that a broader range of normal  139  mechanics exists at the patellofemoral joint as compared to other joints such as the tibiofemoral joint, where it has been shown that patterns of gait do not vary substantially between individuals [219]. Because there are so many variables that likely contribute to patterns of patellofemoral joint mechanics, when ex vivo studies or models are used, many assumptions and simplifications must be made.  When employing in vivo assessments fewer assumptions have to be made and therefore results likely most closely resemble joint mechanics experienced during normal daily activities.  The inter- subject variability in three-dimensional patellar kinematics has been described time and time again in the literature [107] and throughout this thesis (Chapters 2-4).  Using some raw data from Chapter 2 as an example, patellar tilt in normal, healthy individuals ranged from 10 degrees of lateral tilt to 20 degrees of medial tilt and patterns of increasing, decreasing, constant and variable tilt with tibiofemoral flexion were all observed (Figure 7-1).  Variability has similarly been seen in other kinematic parameters as discussed in Section 1.4.7.  Variability in measures of contact area between studies was highlighted in Section 1.5.3; however, raw normative data are limited.  From the results of the four cadaver specimens and the two healthy subjects used in the validation in Chapter 5, variability was present.  Specifically contact areas between 150 and 800 mm2 were observed.  To truly examine the variation in the parameter, normalization to patellar size or overall patellar cartilage surface may be required.  Variation in kinematics and contact areas also suggests variation in other mechanical parameters, such as contact stress.  Some of the observed variation in joint mechanics likely originates in the structure and function of the extensor mechanism.  Firstly, differences in patellar type and trochlear shape, discussed in Section 1.2.2.1 [6], likely contribute to some of the differences observed in kinematics and contact areas.  It is possible that different patellar types have different patterns of patellar kinematics.  Also, differences in cartilage coverage of the patella and trochlea and the individual cartilage material properties may contribute to differences in kinematics and contact area.  Secondly, differences in gender, weight, varus/valgus malalignment, joint laxity, quadriceps muscle line of action, quadriceps tendon insertions on the patella, patellar ligament insertion on the tibia (i.e. location of the tibial tuberosity) and insertion of the medial and lateral passive structures may also contribute to differences in mechanics.  Also, in individuals with patellofemoral joint disease, features such as cartilage thinning, joint effusion, bone marrow lesions, pain and stiffness, to name but a few, could account for some variability.  Finally, the function of other joints in the lower limb, such as the ankle, tibiofemoral joint and hip, may also contribute to differences in mechanics due to differences in loads being transmitted through the system.  It is possible that variability in patellofemoral joint kinematics and contact areas in the normal and disease population could be reduced by studying subpopulations.  140  Figure 7-1: Patellar tilt as a function of tibiofemoral flexion. These data were from the 40 normal healthy individuals studied in Chapter 2.   The most important reason to assess joint mechanics in vivo is the opportunity to relate mechanical changes to clinical symptoms and to perform longitudinal assessments to study disease process or to design and evaluate treatments.  This is particularly important at the patellofemoral joint since many patellofemoral joint disorders, such as OA, are thought to be mechanical in origin. Mechanical risk factors of patellofemoral OA have been established (Section 1.3.3), however, these are most often surrogate measures of joint mechanics.  In vivo assessments of joint mechanics would facilitate a better understanding of how these surrogate measures affect kinematics and contact areas. Further, longitudinal studies are possible and therefore changes in joint mechanics throughout the disease process could be captured.  Also, since many intervention strategies focus on altering kinematics (using treatments such as bracing, taping and physiotherapy as discussed in Section 1.3.4) it is essential to determine what magnitude of mechanical change is required to improve clinical symptoms.  Mechanical-based treatment strategies have had limited success [58] and one possible reason for this is that the magnitude of change required to improve symptoms remains unknown.  In Chapter 4, the effect of a bracing intervention on kinematics was examined; however, in the portion of the study presented in this thesis, pain was not considered.  In the parent study, of which the data in Chapter 4 were a subset, no statistically significant reduction in clinical symptoms was found when patients donned the patellofemoral medialization brace as compared to the control brace.  This finding therefore suggests that the magnitude of change in kinematics that the brace provided was insufficient.  Taping has had some success at reducing pain in cases of patellofemoral and generalized knee OA [61-64], therefore by designing a study of three-dimensional patellar kinematics with a  141  taping intervention the magnitude of mechanical change required to reduce patient symptoms could be determined.  Further, once these data are available, an improved brace could be designed to create the required kinematic change.  These proposed studies must be conducted in vivo, once again highlighting the need for development, characterization and improvement of in vivo assessments of joint mechanics. 7.2.2 How do the In Vivo Assessments of Patellofemoral Joint Mechanics Used in this Thesis Compare to Others in the Literature? 7.2.2.1 Three-dimensional Patellar Kinematics The validated method of assessing three-dimensional patellar kinematics used in this thesis [102,132] has been used to study several patellofemoral joint diseases, such as patellofemoral OA (Chapter 4), general knee OA [220] and patellofemoral pain syndrome [134].  As discussed in Sections 1.4.4.2 and 1.4.6, several other methods of assessing three-dimensional patellar kinematics in vivo have been developed.  Most of these are imaging based and can be compared to the method used in this thesis.  The main areas in which the methods will be compared are validation, loading task, the participant’s experience and overall assessment and analysis time. The kinematics method used in this thesis has been validated more rigorously than any other method in the literature; the error of agreement with a RSA reference standard in a cadaver model has been quantified and the intra-subject and inter-reader repeatability errors have also been assessed. From Table 1.3, it is clear that validation of most techniques has been limited (there are many empty boxes in the table).  The only other method that has quantified errors in terms of agreement with a reference and repeatability was the cine phase-contrast MRI technique [131].  However, a motion phantom was used as the reference standard.  Errors in human subjects may be greater because of variability in the repeated flexion/extension motion, which is particularly important because the method relies on averaging images over a number of cycles of motion.  The motion phantom did include a piece of femoral bone, so the study did make an attempt to consider errors associated with imaging bone.  Further, the intra-subject repeatability of the cine phase-contrast method is poorer than that of the method used in this thesis (2.9° and 1.4°, respectively), likely because of errors associated with out of plane motion of the technique.  Markerless bi-planar radiography has smaller errors than MRI-based techniques, even when used dynamically [101].  This is not surprising since radiography provides a better image of bone at a higher resolution; however, it has yet to be applied in a patellofemoral joint disease population.  142  It is not clear whether the joint loads experienced as a result of the loading task used in this thesis are representative of those experienced during activities of daily living.  Similar methods of applying static axial loads to the foot have been used by other groups [104,116], but this type of loading protocol may be limited because the inertial loads present during activity are not considered. The continuous cine phase-contrast MRI method does account for inertial loading but does not employ an additional external load to the system [131]; therefore extension is resisted by only the weight of the shank.  However, the resultant joint loads are likely still very small compared to those experienced during daily activity.  Using the bi-planar radiographic methods, full weightbearing in an upright position is possible and the loading tasks have used sequential static poses during a simulated lunge [103] and a continuous lunge [101].  It is clear that the loading scenarios used for the bi-planar radiography techniques are more representative of daily weightbearing activity.  A lunge is well suited for bi-planar radiography assessments because the knee stays in a relatively fixed position throughout the motion which is required because of the small imaging volume; however, it is likely that this is an extreme example of joint loads compared to most activities of daily living.  Finally, it should be noted that the bi-planar radiography-based methods allow a greater range of knee flexion to be assessed than all MRI-based methods.  Ideally patellar kinematics would be assessed over the entire range of tibiofemoral flexion. The kinematics data collection protocol used in this thesis is suitable for studies of patellofemoral OA because the loaded MRI scan is only 30 seconds in length.  For the OA population (and potentially other patient populations), the data collection protocol must be sufficiently short to ensure that all participants are able carry out the required loading task.  The OA subset most often studied have clinical symptoms of OA (because this is how they are recruited) and this limits the types of activities the participants can carry out.  For example, the 15% BW loading task was manageable by all patellofemoral OA study participants (Chapter 4); however, it did become increasingly difficult for some of the participants with knee flexion.  This was evidenced by comments made by study participants and the increase in number of scans with motion artefact at deeper flexion angles.  It was determined, after carrying out the loading study (Chapter 3), that the 30% BW loading level was difficult even for some young healthy participants at deeper knee flexion angles and therefore it was decided that this magnitude of loading would not be suitable for individuals with patellofemoral OA.  The sequential static pose lunge or the continuous lunge loading task used in the bi-planar radiography based methods would likely be very difficult for individuals with patellofemoral OA.  It is likely that, if these methods were to be applied in the OA population (which they have not been to date), then the loading task would likely have to be modified.  Another concern might be with the overall length of the data collection protocol.  For example, in Chapter 4,  143  the patellofemoral OA patients underwent a 1.5 hour session in the MRI scanner with just one break. While the length of time spent loading the joint was very short (12 x 30 seconds), the entire process took a long period of time; however, the participants did tolerate the session well.  The cine phase- contrast MRI method likely doesn’t require the same amount of time to carry out, nor do the bi-planar radiography based methods. The analysis portion of the kinematics method used in this thesis is very time intensive. Segmentation of the bones from the MRI images accounts for a large part of the analysis time; segmentation of the bone models takes between 1 and 1.5 hours and segmentation of the loaded contours 30 to 45 minutes each.  The other sequential static MRI-based methods are likely equally time intensive in terms of segmentation [104,116].  It is also likely that the segmentation and registration process required for bi-planar radiography-based assessments is similarly time consuming [101,103].  Less segmentation appears to be required for the cine phase-contrast MRI based method, because only a single slice sagittal slice is used, but since the theory behind this analysis is different from the other techniques (pixel velocities are used to assess kinematics as opposed to shape based registration) it is unknown whether it is similarly time consuming. 7.2.2.2 Contact Area The MRI-based method of assessing patellar contact areas developed in Chapter 5 is the most rigorously validated method to date and requires a much shorter scan than the other validated techniques used in the literature.  Other features of the methods which should be compared are: scan direction, analysis technique, image resolution and joint loading.  Most of these comparisons were made in the discussion section of Chapter 5; however, additional discussion will be provided here. To briefly summarize the comparisons made in Chapter 5, the method developed in this thesis has errors similar to other validated methods in the literature.  The MRI scan sequences developed provided the first comparison of contact area assessments obtained from axial and sagittal scans.  The findings of this thesis show that assessments of patellofemoral contact areas from axial scans show worse agreement with the reference than those from sagittal scans and therefore results in the literature using axial scans should be interpreted with care.  Differences in analysis details were also compared. While most studies multiply the in-plane line of contact by slice thickness to obtain a contact area measure, it is likely that interpolating between slices (using linear or spline interpolation) provides a better description of the patellar surface. The resolution of the images used to assess contact area in this thesis was comparable to the highest used in the literature to date [104,116,167,168]; however, again the scans were significantly  144  shorter.  It is likely that the resolution of the image used for assessment contributes to errors in contact area assessment.  In MRI, a balance must be obtained between scan length and resolution and in this thesis scan length was considered to be very important (a scan less than 1 minute in length was mandated).  The scans developed in this thesis were collected at a resolution of 0.4 mm x 0.4 mm but were then interpolated to 0.3125 mm x 0.3125 mm (this is a standard interpolation done by most MRI scanners).  This resolution is the same as that of the validated technique of Heino-Brechter et al. [167]; however, that scan took 11 minutes.  The resolution used in the phantom validation and in subsequent studies by that group was 0.78 x 0.78 mm [165,166].  The resolution of the scan sequence developed in this thesis was therefore similar to or better than others used in the literature.  Of course, as MRI technology advances, fast images with higher resolution may become available.  These would be desired to better isolate regions of contact. The sequential static loading protocol employed in this thesis was similar to those used in all studies of contact area carried out to date because it is currently not possible to assess contact areas during continuous knee flexion.  Most studies have used an axial applied load [116,169,171,173-176], such as the one used in the present thesis, others have applied a torsional load [169,172] or no load [168,170] and finally one group employed an upright weightbearing sequential static loading protocol using an open-bore MRI scanner [165,166].  To date, the effect of load on contact area has not been assessed; therefore it is difficult to interpret study results when different loading protocols are employed.  Ideally, contact areas would be assessed during continuous motion; however, it does not appear that this will be possible for many years. 7.2.3 Do Statistically Significant Differences Relate to Clinically Important Changes in Three-dimensional Patellar Kinematics? In Chapters 3 and 4, statistically significant differences in three-dimensional patellar kinematics were observed with load and when participants with patellofemoral OA donned a patellofemoral brace as compared to unloaded and no brace conditions, respectively; however, as mentioned in Sections 3.4 and 4.4, it is not clear if these differences are of clinical importance.  From the literature we can glean some information on what magnitudes of change might be clinically important.  For example, in the patellofemoral pain population differences in two-dimensional patellar kinematics (0.9 to 1.6 mm for bisect offset and 1.5 to 3° for tilt) and a reduction in pain was observed when a brace was donned [65,171] and differences in three-dimensional patellar kinematics (for example the pain group’s patella were positioned approximately 2.25 mm more lateral) were observed when the pain population were compared to normal individuals [134].  In the patellofemoral OA population, the only data available come from a taping study where differences in two-  145  dimensional axial alignment in a single slice were observed when the tape was applied (1 mm for lateral displacement and 3.5° for tilt); taping also reduced pain in this study [64].  These data provide a reference for expected changes and also the magnitude of change that may be required to reduce pain.  The differences observed in the present study were, for the most part, smaller than this. The majority of the data available to provide context of differences in patellar kinematics comes from the two-dimensional alignment or kinematic literature; these differences must be interpreted with care because the measures of alignment have not been validated.  Two different methods of measuring the medial-lateral position have been used; bisect offset and lateral displacement.  The former refers to the proportion of the patella lateral to the trochlear groove while the latter refers to the proportion of the patella lateral to the lateral epicondyle.  Patellar tilt has also been used and is defined as the angle between the femoral condyles and the lateral compartment of the patella.  These definitions were designed to optimize potential differences in these parameters. For example, in the three-dimensional method used in this thesis, medial-lateral translations are expressed as the position along the flexion axis of the femur, however, bisect offset and lateral displacement may reflect displacement in an oblique direction. Therefore, bisect offset and lateral displacement may be more representative of a resultant displacement than a medial-lateral translation. The differences in definitions of coordinate systems make it difficult to compare results between the two- and three-dimensional cases and may explain why smaller differences were observed in the bracing study (Chapter 4).  Validation of the two-dimensional measurements would be possible using a phantom or cadaver specimen, an x-y table (or similar) and an imaging system, but this has not been done to date.  It should be noted that these unvalidated measures of two-dimensional kinematics are widely published. The differences observed in Chapters 3 and 4 both represent important contributions to the clinical literature.  In Chapter 3, differences in tilt between the unloaded and 30% BW load condition were up to 4°, which, in terms of previous literature, is a clinically important change.  However, one might argue that in the case of the loading study, any statistically significant difference may be important because it has the potential to affect results in future clinical studies; for example, if load is not controlled, a portion of observed differences in kinematics could be explained by differences in load, thereby confounding study results.  Therefore understanding the effects of load, regardless of size, is important. In Chapter 4, differences of up to 1° in tilt and 0.46 mm in translation were observed.  These differences are smaller than what the above studies suggest is clinically relevant (although, the bottom end of this spectrum is not well defined).  However, these are the first three- dimensional patellar kinematic data in the patellofemoral OA population, therefore, even though they  146  do not appear to represent a clinically important change (Appendix G), they are essential for designing future studies in this population. 7.2.4 Are Patellofemoral Joint Models Useful? Patellofemoral joint models are difficult to develop, validate and employ; however, they may be useful for answering research questions that cannot be addressed using in vivo or ex vivo studies. The various strengths and limitations of current patellofemoral models are discussed in Section 1.6. This section argues that models should be patient specific and be validated for all output parameters (if possible).  While there are particular instances in which models are the best means of answering a research question, the scenarios in which they are applied should be carefully considered. Cartilage material properties were not considered in the model developed in Chapter 6. Other models have included cartilage material properties but have defined it as a linear elastic material, which we know to be an over simplification (Section 1.2.3.2).  This is likely a source of error in contact stress estimates observed in the literature, which is evidenced by the mediocre agreement (only presented qualitatively) between the model-based and ex vivo-based measures of contact stress found in the validation study by Elias et al. [184]. It is unclear if a more sophisticated description of cartilage material properties would improve the agreement between model and ex vivo measures of contact stress. However, adding another level of complexity to a model will increase the computational time required and may further limit its applicability in a patient specific manner, which may be a crucial limitation.  Modelling contact stress is therefore likely only worth the computational time if the estimates have sufficiently small errors, especially since clinically significant stress thresholds are unclear. The model developed in this thesis was designed to be applied in a patient specific manner in order to account for the inherent variability in joint kinematics observed between individuals.  One patient specific finite element model [186] was applied to study the effect of femoral internal and external rotation on patellar mechanics [190].  This study found large variation in cartilage stress between subjects when internal and external rotations were applied.  This study could not have been carried out in vivo because inducing an internal or external rotation in a controlled manner in a study participant is not possible.  This question likely could have equally been answered in an ex vivo model; however, the inherent limitations in terms of joint loading and the number of measurement tools that would have been required to produce the same number of outputs as that particular model would have been great.  Therefore, in this case the patient specific finite element model does yield a credible answer to the research question.  Further, it should be noted that this model is validated with  147  in vivo assessments of contact area which is distinct from other models in the literature, which all use ex vivo measures for validation.  In that study, 16 models were created from in vivo data obtained from normal study participants highlighting that even though a modelling framework was used, the number of participants required to answer the research question was still quite large.  In this instance the model was used as a clinical research tool in a patient specific manner (instead of a general model) which is likely the forum in which patellofemoral models will be useful. The model developed and validated in Chapter 6 can only be used to estimate contact areas and is sensitive to proximity threshold and kinematic input data.  Chapter 6 represents the first time that the sensitivity of a model to kinematic input was examined and it is not known if other models suffer from this same limitation. It is clear that the development of a patient specific model that adequately predicts numerous output parameters (such as contact area, contact stress, cartilage strain, tendon and ligament loads) is very involved. While the development of such a model was outside the scope of the present work, there are many research questions for which a complex model would be useful and suitable.  However, at this point in time there appears to be a greater need for a simple model that can be applied in a patient specific manner in large clinical studies or studies where in vivo contact area assessments are not possible. 7.2.5 Why should more Sophisticated Measures of Joint Mechanics be used in Studies of Patellofemoral OA? Without elucidating the direct relationship between patient symptoms and patellofemoral joint mechanics, mechanical-based treatment strategies will not be successful; again highlighting the need for a validated method of simultaneously measuring kinematics and contact areas in vivo, such as the one developed in Chapter 5.  Surrogate or simplistic measures of patellofemoral joint mechanics have traditionally been used in clinical studies of patellofemoral OA.  Although more sophisticated in vivo measures of three-dimensional patellar kinematics and contact areas now exist, they have not often been used in clinical studies.  This may be why mechanical-based treatment strategies have had limited success (Section 1.3.4), such as the bracing study in Chapter 4.  Section 1.3.3 detailed the surrogate (obesity, varus/valgus tibiofemoral alignment, knee height) or simple (static alignment at one knee flexion angle in the axial plane) measures of mechanics used in patellofemoral OA research; this section also highlighted that these measures only suggest that mechanics are associated with patellofemoral OA.  One study in particular identified that patellar alignment was associated with patellofemoral OA disease progression [2].  Patellar alignment (specifically bisect offset and tilt) in this study was assessed from a skyline radiograph of the patellofemoral joint acquired at a single angle between 30 and 40° of knee flexion. However, in Chapter 2, it was determined that patellar  148  alignment assessed at one knee flexion angle (30° in this case) did not provide a surrogate marker of three-dimensional patellar kinematics over the range of knee flexion angles.  Therefore, it is possible that the results of the patellofemoral OA progression study might have been different if the radiograph had been acquired and alignment had been assessed at a different knee flexion angle.  The study in Chapter 4 represented the first time more sophisticated measures of joint kinematics were used in a clinical study of the patellofemoral OA population, demonstrating that it is possible to incorporate these measures into a clinical study (a cross-over trial in this case) even if it was in just a subset of subjects.  There is a trend in the literature to move towards direct measures of patellofemoral mechanics in studies of patellofemoral OA, such as the taping study that used a simple measures of patellar alignment [64].  However, in the bracing study of Chapter 4 and the taping study [64], differences in three-dimensional patellar kinematics and patellar alignment, respectively, were quite small highlighting the need for kinematic assessments with small measurement errors because small changes in patellar kinematics may prove to be very important in terms of improving patient symptoms. 7.3 Strengths and Limitations While strengths and limitations specific to each chapter were detailed in each of their respective discussion sections, there are some strengths and limitations associated with the general approaches taken to answer research questions and to develop research tools throughout this thesis. While MRI is particularly useful for studying patellofemoral joint kinematics and contact areas because cartilage imaging is possible and no ionizing radiation is required, in some aspects it is not ideal.  Kinematic assessments made with MRI have larger errors than those made with bi-planar radiography systems.  With MRI, the range of motion studied for both the kinematic and contact area assessments is limited by the size of the bore of the scanner.  With the introduction of open-MRI systems (like the one currently being installed at UBC) and wide bore systems, greater ranges of motion will be possible.  Further, open systems allow imaging in weightbearing.  Ideally, we would assess joint kinematics and contact areas during a continuous motion, but the kinematics method currently validated by our group does not allow for this.  A method of assessing three-dimensional patellar kinematics during continuous flexion with MRI is currently being developed by our group. Continuous assessments of contact area are likely still some years away.  Since MRI is the only method available to assess contact area directly, it is practical to assess kinematics using the same tool, especially since one of the major strengths of this work is the development of a well validated method to assess three-dimensional patellar kinematics and contact areas simultaneously.  149  The results in this thesis are greatly dependent on MRI image quality and subjective segmentation, and these are likely a source of error in the kinematic and contact area analyses.  The first and most challenging step in MRI image processing is segmentation of the necessary tissues from the raw images.  During the kinematic analysis, trabecular bone is segmented.  This task is reasonably straightforward for the higher quality scans used to create the bone models; however, for the lower resolution, loaded scans this task becomes more difficult because the boundaries between tissues are not as clear.  This is likely one of the major sources of error accounted for in the validation results.  During the contact area analysis, cartilage segmentation was required.  Cartilage segmentation is much more challenging than bone segmentation because the segmentor is required to distinguish between cartilage and surrounding tissues such as the joint capsule which can be further complicated by possible structural changes, such as cartilage degeneration, in the case of patellofemoral OA.  This type of analysis requires training and many rules must be followed to minimize intra-reader repeatability.  Therefore it is important that the contact area analysis be carried out by an experienced reader trained in cartilage segmentation.  For the studies in this thesis, training was obtained from one of the leading groups in this field [37,38,40,221].  The loaded cartilage scans used for the contact area analysis are of a lower quality than those that would be used for morphological assessments and therefore may be a source of some of the disagreement between the dye-based and MRI-based assessments found in Chapter 5. Another source of error is the extraction of the segmented data from the image processing programs.  In the commercial image processing software used (Analyze 8.0, Analyze Direct, Overland Park, KS, USA), this can be carried out in two ways.  The first is to directly export the contours which are described as pixel and slice numbers.  This method was used for the loaded bone and cartilage scans.  The second method is to create a surface and this was carried out using the Adapt/Deform tool in the Surface Extractor module of Analyze.  The nodes of this surface create the three-dimensional point-cloud model.  There is likely error associated with this step; however, the Analyze settings are such that the surface closely fits the raw data, although no measure of this is available in the software. Image registration was used in several instances throughout the kinematic and contact area analysis and is another potential source of error.  The tool used for image registration was the ICP algorithm [208].  This algorithm is based on, as its name suggests, finding the position of the target dataset relative to the reference dataset in which the mean difference between points in the two datasets is at a minimum.  Finding this correct position depends on the quality of the initial position estimate input into the algorithm.  Therefore, in the present thesis a pre-registration step that was  150  visually verified was used to ensure that the ICP converges on the correct solution.  This algorithm is dependent on the quality of the data being input, which in this case included the contours and surfaces created from the segmented data or laser scan data.  Because of the nature of this discretized input data, points in one dataset do not necessarily exactly correspond to points in the second dataset.  This will introduce some error into the system.  However, the registration errors obtained throughout the thesis have been quantified [132] and likely slightly inflated due to the discrete nature of the input data.  Other registration methods do exist which use other types of information, such as image intensity, to carry out the registration.  However, these methods are better suited for radiographic images because the bone boundaries are clear.  The registration method used in this thesis is likely sufficient due to the type of data that can be extracted from the MRI images, although it is a potential area for improvement in the future. It is clear that MRI has several inherent limitations in its application to in vivo assessments of joint mechanics; however, it also has several advantages.  MRI technology is continually advancing and as such so will the quality of the in vivo assessments that utilize it.  Also, MRI can be optimized to image many different tissues and therefore there is great potential, with sequence development, to study other tissues in the extensor mechanism such as tendon, ligament and muscle.  Further, it may even be possible, if resolution continues to improve, to calculate cartilage strain from MR images. This has already been done at the 7 Tesla magnet level in a cadaver hip joint [222].  Finally, MRI is ideal for in vivo assessments of joint mechanics because participants are not exposed to ionizing radiation. Finally, due to the time intensive nature of the kinematic analysis, the number of participants studied was relatively small.  Further, detailed clinical information was not collected.  For these reasons, it was not possible to control for other parameters that may affect kinematics such as gender, weight, varus/valgus tibiofemoral alignment, joint laxity or features of joint disease (in the patellofemoral OA cases).  It is likely that subpopulations would need to be defined to reduce the variability observed in the kinematic results. 7.4 Contributions This thesis addressed many different aspects of in vivo assessments of joint mechanics, from method development to characterization to application; as such many novel contributions were made: 1. Chapter 2 was the first study to show that measures of patellofemoral joint alignment at a single knee flexion angle (30° of knee flexion) were not sufficient to describe three- dimensional patellar kinematic patterns over the range of knee flexion.  This finding  151  highlights that measures of patellar alignment in a single plane assessed from radiographs are not sufficient to describe patterns of patellar tracking, which is particularly important to the clinical patellofemoral OA literature. 2. Chapter 3 demonstrated the importance of imaging loaded joints, finding that three- dimensional patellar kinematics varied with applied load.  It was the first study to assess three-dimensional patellar kinematics at varying load levels, which is beneficial for interpreting study results and designing future studies. This study also provided additional normal three-dimensional patellar kinematic data to the literature. 3. Chapter 4 represents the first time three-dimensional patellar kinematics were assessed in the patellofemoral OA population and the first time that the effect of a mechanical-based treatment strategy for patellofemoral OA was evaluated using three-dimensional measures. The findings of the study were important because they highlight that patellofemoral bracing creates changes in kinematics which may not be sufficient to reduce patient symptoms in individuals with patellofemoral OA. 4. In Chapter 5, an MRI-based contact area assessment method was developed and validated; the errors in the method are similar to those of methods that require scans that are many times longer.  Chapter 3 showed that joint kinematics should be assessed under load; this scan is sufficiently short to be used in series with the loaded kinematics scan, therefore three- dimensional patellar kinematics and contact areas can now be assessed simultaneously at the same knee flexion angle under load.  Further, it was the first study to compare axial and sagittal assessment of patellofemoral contact area, finally determining that sagittal measures yield results with smaller errors, and it was the first study to examine the effect of contact area calculation method. 5. In Chapter 6 a simple, patient specific, kinematics-driven multibody model of the patellofemoral joint to predict cartilage contact areas was developed and validated.  This validation was the first to show the sensitivity of models to kinematic input data.  This model could be useful in studies where direct measures of contact area are not possible, such as studies assessing kinematics using bi-planar radiography or studies of individuals who may not be able to carry out the loading task. 7.5 Future Work The work in this thesis has highlighted the need for additional work in several different aspects of method development, characterization and application:  152  1. In Table 1.1, the variation in methods of assigning anatomically-based coordinate systems to the patellofemoral joint was highlighted.  It is clear that standardization of these coordinate systems is required. 2. In Chapters 2 and 3, variation in normal patterns of three-dimensional patellar kinematics was highlighted.  Since quite a large database of normal kinematic data and three-dimensional models now exists, it may be possible to determine if some of this variation can be attributed to patella type.  However, it may be that clinical heterogeneity in normal individuals must also be considered.  For example, screening for factors such as varus/valgus alignment, muscle strength and patellar laxity or dividing individuals into subgroups based on gender or weight may also decrease variability. 3. Chapter 4 highlighted that patellar bracing did not create a sufficiently large kinematic change to reduce pain in individuals with patellofemoral OA.  Therefore, a study is needed to determine the amount of mechanical correction required to reduce patient symptoms.  This magnitude of change might be best assessed through a longitudinal taping intervention since taping has been shown to reduce pain cross-sectionally; however, in those studies three- dimensional patellar kinematics were not quantified.  Further, the addition of a contact area assessment would likely elucidate the relationship between patellofemoral OA symptoms and joint mechanics and this is now possible with the method developed in Chapter 5. 4. In Chapter 5, repeatability was assessed in normal individuals only.  Before using the contact area assessment in the patellofemoral OA study a repeatability study particular to this population should be carried out. 5. In Chapter 6, a simple model to predict contact areas was developed.  Additional work could be done to create a characteristic equation for calculating the patient specific proximity threshold values. 6. All of the studies of patellofemoral joint mechanics and OA carried out to date have been cross-sectional.  It may be useful to assess changes in patellofemoral kinematics and contact areas longitudinally in this population in order to better understand the relationship between mechanics and the disease process. 7. The University of British Columbia is in the process of installing an open configuration MRI scanner.  The kinematic and contact area scans developed in this thesis should be adapted for this scanner in order to make measurements during weightbearing.  153  7.6 Conclusion Patellofemoral joint diseases, such as OA, are thought to be mechanical in origin and as such many treatment strategies have focused on correcting joint mechanics in an effort to relieve patient symptoms.  However, the success of these treatment strategies has been variable.  This thesis describes investigations of a number of important factors influencing mechanical assessments of the patellofemoral joint.  The work in this thesis showed that measures of alignment are insufficient surrogates of full measurements of three-dimensional patellar kinematics, which may explain why studies using alignment as an outcome measure may not capture all individuals with aberrant patterns of patellar mechanics and therefore the relationship between mechanics and patellofemoral OA remains unclear.  Further, it was shown that a patellofemoral brace resulted in only small changes in kinematics in individuals with patellofemoral OA; in some individuals these changes were smaller than the measurement error of most current methods highlighting the need for measures of patellofemoral joint kinematics and contact areas with small measurement errors in order to relate mechanics to disease progression and treatment.  This thesis also showed that loading must be controlled and representative of physiological loading during assessments of kinematics for results to be compared between individuals within studies.  A fully validated method for simultaneously measuring patellofemoral kinematics and contact areas was developed; with this method it will be possible to elucidate the relationship between patellar tracking and contact area which may provide insight into the limited success of mechanical-based treatment strategies to date.  Finally, a method of predicting patellofemoral contact areas from kinematics was validated; although the model showed that predictions of contact areas are sensitive to small errors in kinematics.  These findings demonstrate that errors in kinematic assessments must be minimized and demonstrate the need for rapid, valid, simultaneous measures of kinematics and contact area in order to be of value for relating mechanics to disease progression and for evaluating mechanical-based treatment strategies of patellofemoral joint disease, such as OA.  154  References 1. McAlindon T, Zhang Y, Hannan M, Naimark A, Weissman B, Castelli W, et al. Are risk factors for patellofemoral and tibiofemoral knee osteoarthritis different? J Rheumatol 1996; 23: 332-337. 2. Hunter DJ, Zhang YQ, Niu JB, Felson DT, Kwoh K, Newman A, et al. Patella malalignment, pain and patellofemoral progression: the Health ABC Study. Osteoarthritis Cartilage 2007; 15: 1120-1127. 3. Kalichman L, Zhang Y, Niu J, Goggins J, Gale D, Felson DT, et al. The association between patellar alignment and patellofemoral joint osteoarthritis features--an MRI study. Rheumatology (Oxford) 2007; 46: 1303-1308. 4. Kalichman L, Zhang Y, Niu J, Goggins J, Gale D, Zhu Y, et al. The association between patellar alignment on magnetic resonance imaging and radiographic manifestations of knee osteoarthritis. Arthritis Res Ther 2007; 9: R26. 5. Kalichman L, Zhu Y, Zhang Y, Niu J, Gale D, Felson DT, et al. The association between patella alignment and knee pain and function: an MRI study in persons with symptomatic knee osteoarthritis. Osteoarthritis Cartilage 2007; 15: 1235-1240. 6. Wiberg G. Roentgenographic and anatomic studies on the femoropatellar joint, with special reference to chondromalacia patellae. Acta Orthop Scand 1941; 12: 319-340. 7. Draper CE, Besier TF, Gold GE, Fredericson M, Fiene A, Beaupre GS, et al. Is cartilage thickness different in young subjects with and without patellofemoral pain? Osteoarthritis Cartilage 2006; 14: 931-937. 8. Kaufer H. Mechanical function of the patella. J Bone Joint Surg Am 1971; 53: 1551-1560. 9. Schmidt TA, Sah RL. Effect of synovial fluid on boundary lubrication of articular cartilage. Osteoarthritis Cartilage 2007; 15: 35-47. 10. Ahmed AM, Burke DL, Hyder A. Force analysis of the patellar mechanism. J Orthop Res 1987; 5: 69-85. 11. Huberti HH, Hayes WC, Stone JL, Shybut GT. Force ratios in the quadriceps tendon and ligamentum patellae. J Orthop Res 1984; 2: 49-54.  155  12. Ahmed AM, Burke DL, Yu A. In-vitro measurement of static pressure distribution in synovial joints--Part II: Retropatellar surface. J Biomech Eng 1983; 105: 226-236. 13. Reilly DT, Martens M. Experimental analysis of the quadriceps muscle force and patello- femoral joint reaction force for various activities. Acta Orthop Scand 1972; 43: 126-137. 14. Kwak SD, Ahmad CS, Gardner TR, Grelsamer RP, Henry JH, Blankevoort L, et al. Hamstrings and iliotibial band forces affect knee kinematics and contact pattern. J Orthop Res 2000; 18: 101-108. 15. Sandmeier RH, Burks RT, Bachus KN, Billings A. The effect of reconstruction of the medial patellofemoral ligament on patellar tracking. Am J Sports Med 2000; 28: 345-349. 16. Farahmand F, Naghi Tahmasbi M, Amis A. The contribution of the medial retinaculum and quadriceps muscles to patellar lateral stability--an in-vitro study. Knee 2004; 11: 89-94. 17. Ostermeier S, Holst M, Hurschler C, Windhagen H, Stukenborg-Colsman C. Dynamic measurement of patellofemoral kinematics and contact pressure after lateral retinacular release: an in vitro study. Knee Surg Sports Traumatol Arthrosc 2007; 15: 547-554. 18. Merican AM, Amis AA. Iliotibial band tension affects patellofemoral and tibiofemoral kinematics. J Biomech 2009; 42: 1539-1546. 19. Mankin HJ, Mow VC, Buckwalter JA, Ianotti JP, Ratcliffe A. Articular Cartilage Structure, Composition, and Function. In: Orthopaedic Basic Science, Buckwalter JA, Einhorn TA, Simon SR, 2nd Edition, American Academy of Orthopaedic Surgeons 2000:458-465. 20. Mow VC, Kuei SC, Lai WM, Armstrong CG. Biphasic creep and stress relaxation of articular cartilage in compression? Theory and experiments. J Biomech Eng 1980; 102: 73-84. 21. Eckstein F, Tieschky M, Faber S, Englmeier KH, Reiser M. Functional analysis of articular cartilage deformation, recovery, and fluid flow following dynamic exercise in vivo. Anat Embryol (Berl) 1999; 200: 419-424. 22. Brand RA. Joint contact stress: a reasonable surrogate for biological processes? Iowa Orthop J 2005; 25: 82-94. 23. Brandt KD, Dieppe P, Radin EL. Etiopathogenesis of osteoarthritis. Rheum Dis Clin North Am 2008; 34: 531-559. 24. Types of Arthritis - Osteoarthritis. 2010. The Arthritis Society of Canada. www.arthritis.ca  156  25. Oliveria SA, Felson DT, Reed JI, Cirillo PA, Walker AM. Incidence of symptomatic hand, hip, and knee osteoarthritis among patients in a health maintenance organization. Arthritis Rheum 1995; 38: 1134-1141. 26. Felson DT, Naimark A, Anderson J, Kazis L, Castelli W, Meenan RF. The prevalence of knee osteoarthritis in the elderly. The Framingham Osteoarthritis Study. Arthritis Rheum 1987; 30: 914-918. 27. Jordan JM, Helmick CG, Renner JB, Luta G, Dragomir AD, Woodard J, et al. Prevalence of knee symptoms and radiographic and symptomatic knee osteoarthritis in African Americans and Caucasians: the Johnston County Osteoarthritis Project. J Rheumatol 2007; 34: 172-180. 28. Lawrence RC, Felson DT, Helmick CG, Arnold LM, Choi H, Deyo RA, et al. Estimates of the prevalence of arthritis and other rheumatic conditions in the United States. Part II. Arthritis Rheum 2008; 58: 26-35. 29. Elahi S, Cahue S, Felson DT, Engelman L, Sharma L. The association between varus-valgus alignment and patellofemoral osteoarthritis. Arthritis Rheum 2000; 43: 1874-1880. 30. Iwano T, Kurosawa H, Tokuyama H, Hoshikawa Y. Roentgenographic and clinical findings of patellofemoral osteoarthrosis. With special reference to its relationship to femorotibial osteoarthrosis and etiologic factors. Clin Orthop 1990: 190-197. 31. Cahue S, Dunlop D, Hayes K, Song J, Torres L, Sharma L. Varus-valgus alignment in the progression of patellofemoral osteoarthritis. Arthritis Rheum 2004; 50: 2184-2190. 32. Duncan RC, Hay EM, Saklatvala J, Croft PR. Prevalence of radiographic osteoarthritis--it all depends on your point of view. Rheumatology (Oxford) 2006; 45: 757-760. 33. Duncan R, Peat G, Thomas E, Wood L, Hay E, Croft P. Does isolated patellofemoral osteoarthritis matter? Osteoarthritis Cartilage 2009; 17: 1151-1155. 34. Buckland-Wright C. Protocols for precise radio-anatomical positioning of the tibiofemoral and patellofemoral compartments of the knee. Osteoarthritis Cartilage 1995; 3 Suppl A: 71- 80. 35. Altman RD, Gold GE. Atlas of individual radiographic features in osteoarthritis, revised. Osteoarthritis Cartilage 2007; 15 Suppl A: A1-56. 36. Cibere J. Do we need radiographs to diagnose osteoarthritis? Best Pract Res Clin Rheumatol 2006; 20: 27-38.  157  37. Eckstein F, Adam C, Sittek H, Becker C, Milz S, Schulte E, et al. Non-invasive determination of cartilage thickness throughout joint surfaces using magnetic resonance imaging. J Biomech 1997; 30: 285-289. 38. Eckstein F, Schnier M, Haubner M, Priebsch J, Glaser C, Englmeier KH, et al. Accuracy of cartilage volume and thickness measurements with magnetic resonance imaging. Clin Orthop 1998: 137-148. 39. Eckstein F, Sittek H, Milz S, Schulte E, Kiefer B, Reiser M, et al. The potential of magnetic resonance imaging (MRI) for quantifying articular cartilage thickness -- a methodological study. Clin Biomech (Bristol, Avon) 1995; 10: 434-440. 40. Eckstein F, Westhoff J, Sittek H, Maag KP, Haubner M, Faber S, et al. In vivo reproducibility of three-dimensional cartilage volume and thickness measurements with MR imaging. AJR Am J Roentgenol 1998; 170: 593-597. 41. Burstein D, Gray M, Mosher T, Dardzinski B. Measures of molecular composition and structure in osteoarthritis. Radiol Clin North Am 2009; 47: 675-686. 42. Peterfy CG, Guermazi A, Zaim S, Tirman PF, Miaux Y, White D, et al. Whole-Organ Magnetic Resonance Imaging Score (WORMS) of the knee in osteoarthritis. Osteoarthritis Cartilage 2004; 12: 177-190. 43. Hunter DJ, Lo GH, Gale D, Grainger AJ, Guermazi A, Conaghan PG. The reliability of a new scoring system for knee osteoarthritis MRI and the validity of bone marrow lesion assessment: BLOKS (Boston Leeds Osteoarthritis Knee Score). Ann Rheum Dis 2008; 67: 206-211. 44. Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt LW. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol 1988; 15: 1833-1840. 45. Roos EM, Lohmander LS. The Knee injury and Osteoarthritis Outcome Score (KOOS): from joint injury to osteoarthritis. Health Qual Life Outcomes 2003; 1: 64. 46. Altman R, Asch E, Bloch D, Bole G, Borenstein D, Brandt K, et al. Development of criteria for the classification and reporting of osteoarthritis. Classification of osteoarthritis of the knee. Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association. Arthritis Rheum 1986; 29: 1039-1049.  158  47. Cicuttini FM, Spector T, Baker J. Risk factors for osteoarthritis in the tibiofemoral and patellofemoral joints of the knee. J Rheumatol 1997; 24: 1164-1167. 48. Cooper C, McAlindon T, Snow S, Vines K, Young P, Kirwan J, et al. Mechanical and constitutional risk factors for symptomatic knee osteoarthritis: differences between medial tibiofemoral and patellofemoral disease. J Rheumatol 1994; 21: 307-313. 49. Harrison MM, Cooke TD, Fisher SB, Griffin MP. Patterns of knee arthrosis and patellar subluxation. Clin Orthop Relat Res 1994: 56-63. 50. Hunter DJ, Niu J, Zhang Y, Nevitt MC, Xu L, Lui LY, et al. Knee height, knee pain, and knee osteoarthritis: the Beijing Osteoarthritis Study. Arthritis Rheum 2005; 52: 1418-1423. 51. Englund M, Lohmander LS. Patellofemoral osteoarthritis coexistent with tibiofemoral osteoarthritis in a meniscectomy population. Ann Rheum Dis 2005; 64: 1721-1726. 52. Cooke TD, Sled EA, Scudamore RA. Frontal plane knee alignment: a call for standardized measurement. J Rheumatol 2007; 34: 1796-1801. 53. Brown TD, Pope DF, Hale JE, Buckwalter JA, Brand RA. Effects of osteochondral defect size on cartilage contact stress. J Orthop Res 1991; 9: 559-567. 54. Nelson BH, Anderson DD, Brand RA, Brown TD. Effect of osteochondral defects on articular cartilage. Contact pressures studied in dog knees. Acta Orthop Scand 1988; 59: 574- 579. 55. Elder SH, Goldstein SA, Kimura JH, Soslowsky LJ, Spengler DM. Chondrocyte differentiation is modulated by frequency and duration of cyclic compressive loading. Ann Biomed Eng 2001; 29: 476-482. 56. Quinn TM, Allen RG, Schalet BJ, Perumbuli P, Hunziker EB. Matrix and cell injury due to sub-impact loading of adult bovine articular cartilage explants: effects of strain rate and peak stress. J Orthop Res 2001; 19: 242-249. 57. Ficat P, Ficat C, Bailleux A. [External hypertension syndrome of the patella.  Its significance in the recognition of arthrosis]. Rev Chir Orthop Reparatrice Appar Mot 1975; 61: 39-59. 58. van Jonbergen HP, Poolman RW, van Kampen A. Isolated patellofemoral osteoarthritis. Acta Orthop 2010; 81: 199-205.  159  59. Quilty B, Tucker M, Campbell R, Dieppe P. Physiotherapy, including quadriceps exercises and patellar taping, for knee osteoarthritis with predominant patello-femoral joint involvement: randomized controlled trial. J Rheumatol 2003; 30: 1311-1317. 60. Crossley KM, Vicenzino B, Pandy MG, Schache AG, Hinman RS. Targeted physiotherapy for patellofemoral joint osteoarthritis: a protocol for a randomised, single-blind controlled trial. BMC Musculoskelet Disord 2008; 9: 122. 61. Cushnaghan J, McCarthy C, Dieppe P. Taping the patella medially: a new treatment for osteoarthritis of the knee joint? BMJ 1994; 308: 753-755. 62. Hinman RS, Bennell KL, Crossley KM, McConnell J. Immediate effects of adhesive tape on pain and disability in individuals with knee osteoarthritis. Rheumatology (Oxford) 2003; 42: 865-869. 63. Hinman RS, Crossley KM, McConnell J, Bennell KL. Efficacy of knee tape in the management of osteoarthritis of the knee: blinded randomised controlled trial. BMJ 2003; 327: 135. 64. Crossley KM, Marino GP, Macilquham MD, Schache AG, Hinman RS. Can patellar tape reduce the patellar malalignment and pain associated with patellofemoral osteoarthritis? Arthritis Rheum 2009; 61: 1719-1725. 65. Draper CE, Besier TF, Santos JM, Jennings F, Fredericson M, Gold GE, et al. Using real- time MRI to quantify altered joint kinematics in subjects with patellofemoral pain and to evaluate the effects of a patellar brace or sleeve on joint motion. J Orthop Res 2009; 27: 571- 577. 66. Warden SJ, Hinman RS, Watson MA, Jr., Avin KG, Bialocerkowski AE, Crossley KM. Patellar taping and bracing for the treatment of chronic knee pain: a systematic review and meta-analysis. Arthritis Rheum 2008; 59: 73-83. 67. Saleh KJ, Arendt EA, Eldridge J, Fulkerson JP, Minas T, Mulhall KJ. Symposium. Operative treatment of patellofemoral arthritis. J Bone Joint Surg Am 2005; 87: 659-671. 68. Becker R, Ropke M, Krull A, Musahl V, Nebelung W. Surgical treatment of isolated patellofemoral osteoarthritis. Clin Orthop Relat Res 2008; 466: 443-449. 69. Torga Spak R, Teitge RA. Fresh osteochondral allografts for patellofemoral arthritis: long- term followup. Clin Orthop Relat Res 2006; 444: 193-200. 70. Beltran JE. Resection arthroplasty of the patella. J Bone Joint Surg Br 1987; 69: 604-607.  160  71. Carofino BC, Fulkerson JP. Anteromedialization of the tibial tubercle for patellofemoral arthritis in patients > 50 years. J Knee Surg 2008; 21: 101-105. 72. Aderinto J, Cobb AG. Lateral release for patellofemoral arthritis. Arthroscopy 2002; 18: 399- 403. 73. Alemdaroglu KB, Cimen O, Aydogan NH, Atlihan D, Iltar S. Early results of arthroscopic lateral retinacular release in patellofemoral osteoarthritis. Knee 2008; 15: 451-455. 74. Yercan HS, Ait Si Selmi T, Neyret P. The treatment of patellofemoral osteoarthritis with partial lateral facetectomy. Clin Orthop Relat Res 2005; 436: 14-19. 75. Paulos LE, O'Connor DL, Karistinos A. Partial lateral patellar facetectomy for treatment of arthritis due to lateral patellar compression syndrome. Arthroscopy 2008; 24: 547-553. 76. Bohnsack M, Halcour A, Klages P, Wilharm A, Ostermeier S, Ruhmann O, et al. The influence of patellar bracing on patellar and knee load-distribution and kinematics: an experimental cadaver study. Knee Surg Sports Traumatol Arthrosc 2008; 16: 135-141. 77. Ahmed AM, Duncan NA. Correlation of patellar tracking pattern with trochlear and retropatellar surface topographies. J Biomech Eng 2000; 122: 652-660. 78. Ahmed AM, Duncan NA, Tanzer M. In vitro measurement of the tracking pattern of the human patella. J Biomech Eng 1999; 121: 222-228. 79. Amis AA, Senavongse W, Bull AM. Patellofemoral kinematics during knee flexion- extension: an in vitro study. J Orthop Res 2006; 24: 2201-2211. 80. Belvedere C, Leardini A, Ensini A, Bianchi L, Catani F, Giannini S. Three-dimensional patellar motion at the natural knee during passive flexion/extension. An in vitro study. J Orthop Res 2009; 27: 1426-1431. 81. Farahmand F, Tahmasbi MN, Amis AA. Lateral force-displacement behaviour of the human patella and its variation with knee flexion--a biomechanical study in vitro. J Biomech 1998; 31: 1147-1152. 82. Goh JC, Lee PY, Bose K. A cadaver study of the function of the oblique part of vastus medialis. J Bone Joint Surg Br 1995; 77: 225-231. 83. Grelsamer RP, Weinstein CH. Applied biomechanics of the patella. Clin Orthop 2001: 9-14. 84. Heegaard J, Leyvraz PF, Curnier A, Rakotomanana L, Huiskes R. The biomechanics of the human patella during passive knee flexion. J Biomech 1995; 28: 1265-1279.  161  85. Heegaard J, Leyvraz PF, Van Kampen A, Rakotomanana L, Rubin PJ, Blankevoort L. Influence of soft structures on patellar three-dimensional tracking. Clin Orthop Relat Res 1994: 235-243. 86. Lee TQ, Sandusky MD, Adeli A, McMahon PJ. Effects of simulated vastus medialis strength variation on patellofemoral joint biomechanics in human cadaver knees. J Rehabil Res Dev 2002; 39: 429-438. 87. Mizuno Y, Kumagai M, Mattessich SM, Elias JJ, Ramrattan N, Cosgarea AJ, et al. Q-angle influences tibiofemoral and patellofemoral kinematics. J Orthop Res 2001; 19: 834-840. 88. Omori G, Koga Y, Bechtold JE, Gustilo RB, Nakabe N, Sasagawa K, et al. Contact pressure and three-dimensional tracking of unresurfaced patella in total knee arthroplasty. Knee 1997; 4: 15-21. 89. Sakai N, Luo ZP, Rand JA, An KN. The influence of weakness in the vastus medialis oblique muscle on the patellofemoral joint: an in vitro biomechanical study. Clin Biomech (Bristol, Avon) 2000; 15: 335-339. 90. van Kampen A, Huiskes R. The three-dimensional tracking pattern of the human patella. J Orthop Res 1990; 8: 372-382. 91. Zavatsky AB, Oppold PT, Price AJ. Simultaneous in vitro measurement of patellofemoral kinematics and forces. J Biomech Eng 2004; 126: 351-356. 92. Fujikawa K, Seedhom BB, Wright V. Biomechanics of the patello-femoral joint. Part I: A study of the contact and the congruity of the patello-femoral compartment and movement of the patella. Eng Med 1983; 12: 3-11. 93. Hefzy MS, Jackson WT, Saddemi SR, Hsieh YF. Effects of tibial rotations on patellar tracking and patello-femoral contact areas. J Biomed Eng 1992; 14: 329-343. 94. Nagamine R, Otani T, White SE, McCarthy DS, Whiteside LA. Patellar tracking measurement in the normal knee. J Orthop Res 1995; 13: 115-122. 95. Powers CM, Lilley JC, Lee TQ. The effects of axial and multi-plane loading of the extensor mechanism on the patellofemoral joint. Clin Biomech (Bristol, Avon) 1998; 13: 616-624. 96. Reider B, Marshall JL, Ring B. Patellar tracking. Clin Orthop Relat Res 1981: 143-148.  162  97. Lin F, Makhsous M, Chang AH, Hendrix RW, Zhang LQ. In vivo and noninvasive six degrees of freedom patellar tracking during voluntary knee movement. Clin Biomech (Bristol, Avon) 2003; 18: 401-409. 98. Koh TJ, Grabiner MD, De Swart RJ. In vivo tracking of the human patella. J Biomech 1992; 25: 637-643. 99. McClay IS, Cavanagh PR, Kalenak A, Sommer Iii HJ. 3-Dimensional kinematics of the patellofemoral joint during running. Journal of Biomechanics 1992; 25: 669-669. 100. Lafortune MA. The use of intra-cortical pins to measure the motion of the knee joint during walking. 1984. PhD Thesis. Pennsylvania State University 101. Bey MJ, Kline SK, Tashman S, Zauel R. Accuracy of biplane x-ray imaging combined with model-based tracking for measuring in-vivo patellofemoral joint motion. J Orthop Surg Res 2008; 3:38. 102. Fellows RA, Hill NA, Gill HS, MacIntyre NJ, Harrison MM, Ellis RE, et al. Magnetic resonance imaging for in vivo assessment of three-dimensional patellar tracking. J Biomech 2005; 38: 1643-1652. 103. Nha KW, Papannagari R, Gill TJ, Van de Velde SK, Freiberg AA, Rubash HE, et al. In vivo patellar tracking: clinical motions and patellofemoral indices. J Orthop Res 2008; 26: 1067- 1074. 104. Patel VV, Hall K, Ries M, Lindsey C, Ozhinsky E, Lu Y, et al. Magnetic resonance imaging of patellofemoral kinematics with weight-bearing. J Bone Joint Surg Am 2003; 85-A: 2419- 2424. 105. Sheehan FT, Drace JE. Quantitative MR measures of three-dimensional patellar kinematics as a research and diagnostic tool. Med Sci Sports Exerc 1999; 31: 1399-1405. 106. Veress SA, Lippert FG, Hou MC, Takamoto T. Patellar tracking patterns measurement by analytical x-ray photogrammetry. J Biomech 1979; 12: 639-650. 107. Katchburian MV, Bull AM, Shih YF, Heatley FW, Amis AA. Measurement of patellar tracking: assessment and analysis of the literature. Clin Orthop Relat Res 2003: 241-259. 108. McWalter EJ, Macintyre NJ, Cibere J, Wilson DR. A single measure of patellar kinematics is an inadequate surrogate marker for patterns of three-dimensional kinematics in healthy knees. Knee 2010; 17: 135-140.  163  109. Sheehan FT, Zajac FE, Drace JE. In vivo tracking of the human patella using cine phase contrast magnetic resonance imaging. J Biomech Eng 1999; 121: 650-656. 110. Fregly BJ, Rahman HA, Banks SA. Theoretical accuracy of model-based shape matching for measuring natural knee kinematics with single-plane fluoroscopy. J Biomech Eng 2005; 127: 692-699. 111. Woltring HJ. Representation and Calculation of 3-D Joint Movement. Human Movement Science 1991; 10: 603-616. 112. Cole GK, Nigg BM, Ronsky JL, Yeadon MR. Application of the joint coordinate system to three-dimensional joint attitude and movement representation: a standardization proposal. J Biomech Eng 1993; 115: 344-349. 113. Grood ES, Suntay WJ. A joint coordinate system for the clinical description of three- dimensional motions: application to the knee. J Biomech Eng 1983; 105: 136-144. 114. Hsu HC, Luo ZP, Rand JA, An KN. Influence of patellar thickness on patellar tracking and patellofemoral contact characteristics after total knee arthroplasty. J Arthroplasty 1996; 11: 69-80. 115. Belvedere C, Catani F, Ensini A, Moctezuma de la Barrera JL, Leardini A. Patellar tracking during total knee arthroplasty: an in vitro feasibility study. Knee Surg Sports Traumatol Arthrosc 2007; 15: 985-993. 116. Shin CS, Carpenter RD, Majumdar S, Ma CB. Three-dimensional in vivo patellofemoral kinematics and contact area of anterior cruciate ligament-deficient and -reconstructed subjects using magnetic resonance imaging. Arthroscopy 2009; 25: 1214-1223. 117. Bull AM, Katchburian MV, Shih YF, Amis AA. Standardisation of the description of patellofemoral motion and comparison between different techniques. Knee Surg Sports Traumatol Arthrosc 2002; 10: 184-193. 118. Seisler AR, Sheehan FT. Normative three-dimensional patellofemoral and tibiofemoral kinematics: a dynamic, in vivo study. IEEE Trans Biomed Eng 2007; 54: 1333-1341. 119. Wu G, Cavanagh PR. ISB recommendations for standardization in the reporting of kinematic data. J Biomech 1995; 28: 1257-1261. 120. Fujikawa K, Seedhom BB, Wright V. Biomechanics of the patello-femoral joint. Part II: A study of the effect of simulated femoro-tibial varus deformity on the congruity of the patello- femoral compartment and movement of the patella. Eng Med 1983; 12: 13-21.  164  121. Brossmann J, Muhle C, Schroder C, Melchert UH, Bull CC, Spielmann RP, et al. Patellar tracking patterns during active and passive knee extension: evaluation with motion-triggered cine MR imaging. Radiology 1993; 187: 205-212. 122. Farahmand F, Senavongse W, Amis AA. Quantitative study of the quadriceps muscles and trochlear groove geometry related to instability of the patellofemoral joint. J Orthop Res 1998; 16: 136-143. 123. Hungerford DS, Barry M. Biomechanics of the patellofemoral joint. Clin Orthop 1979: 9-15. 124. Goodfellow J, Hungerford DS, Zindel M. Patello-femoral joint mechanics and pathology. 1. Functional anatomy of the patello-femoral joint. J Bone Joint Surg Br 1976; 58: 287-290. 125. Singerman R, Berilla J, Kotzar G, Daly J, Davy DT. A six-degree-of-freedom transducer for in vitro measurement of patellofemoral contact forces. J Biomech 1994; 27: 233-238. 126. Lieb FJ, Perry J. Quadriceps function. An anatomical and mechanical study using amputated limbs. J Bone Joint Surg Am 1968; 50: 1535-1548. 127. Warwick R, Williams PL. Gray's Anatomy. 35 Edition. Edinburgh, Longman 1973. 128. Sheehan FT, Derasari A, Brindle TJ, Alter KE. Understanding patellofemoral pain with maltracking in the presence of joint laxity: complete 3D in vivo patellofemoral and tibiofemoral kinematics. J Orthop Res 2009; 27: 561-570. 129. Draper CE, Besier TF, Santos JM, Fredericson M, Beaupre GS, Delp SL, et al. Patellofemoral kinematic differences exist between high-load and low-load conditions in patients with patellofemoral pain. In: American Society of Biomechanics Annual Meeting. State College, PA, USA 2009. 130. McWalter EJ, Wilson DC, Kacher DF, Hunter DJ, Wilson DR. Three-Dimensional Patellar Kinematics in Weightbearing Flexion Using Open MRI. In: 5th World Congress of Biomechanics, vol. 5 (6740). Munich 2006. 131. Sheehan FT, Zajac FE, Drace JE. Using cine phase contrast magnetic resonance imaging to non-invasively study in vivo knee dynamics. J Biomech 1998; 31: 21-26. 132. Fellows RA, Hill NA, Macintyre NJ, Harrison MM, Ellis RE, Wilson DR. Repeatability of a novel technique for in vivo measurement of three-dimensional patellar tracking using magnetic resonance imaging. J Magn Reson Imaging 2005; 22: 145-153.  165  133. Lin F, Wang G, Koh JL, Hendrix RW, Zhang LQ. In vivo and noninvasive three-dimensional patellar tracking induced by individual heads of quadriceps. Med Sci Sports Exerc 2004; 36: 93-101. 134. MacIntyre NJ, Hill NA, Fellows RA, Ellis RE, Wilson DR. Patellofemoral Joint Kinematics in Individuals with and without Patellofemoral Pain Syndrome. J Bone Joint Surg Am 2006; 88: 2596-2605. 135. Patel VV, Hall K, Ries M, Lotz J, Ozhinsky E, Lindsey C, et al. A three-dimensional MRI analysis of knee kinematics. J Orthop Res 2004; 22: 283-292. 136. Li G, Papannagari R, Nha KW, Defrate LE, Gill TJ, Rubash HE. The coupled motion of the femur and patella during in vivo weightbearing knee flexion. J Biomech Eng 2007; 129: 937- 943. 137. Asano T, Akagi M, Koike K, Nakamura T. In vivo three-dimensional patellar tracking on the femur. Clin Orthop 2003: 222-232. 138. Anderst WJ, Tashman S. A method to estimate in vivo dynamic articular surface interaction. J Biomech 2003; 36: 1291-1299. 139. Matthews LS, Sonstegard DA, Henke JA. Load bearing characteristics of the patello-femoral joint. Acta Orthop Scand 1977; 48: 511-516. 140. Greenwald AS, Haynes DW. Weight-bearing areas in the human hip joint. J Bone Joint Surg Br 1972; 54: 157-163. 141. Greenwald AS, Nelson CL, Jr. Relationship of degenerative arthritis to weight-bearing areas in the human hip joint. Surg Forum 1972; 23: 463-464. 142. Black JD, Matejczyk MB, Greenwald AS. Reversible cartilage staining technique for defining articular weight-bearing surfaces. Clin Orthop Relat Res 1981: 265-267. 143. Ramsey PL, Hamilton W. Changes in tibiotalar area of contact caused by lateral talar shift. J Bone Joint Surg Am 1976; 58: 356-357. 144. Yao JQ, Seedhom BB. A new technique for measuring contact areas in human joints--the '3S technique'. Proc Inst Mech Eng H 1991; 205: 69-72. 145. Ateshian GA, Kwak SD, Soslowsky LJ, Mow VC. A stereophotogrammetric method for determining in situ contact areas in diarthrodial joints, and a comparison with other methods. J Biomech 1994; 27: 111-124.  166  146. Harding ML, Harding L, Goodfellow JW. A preliminary report of a simple rig to aid study of the functional anatomy of the cadaver human knee joint. J Biomech 1977; 10: 517-523. 147. Stormont TJ, An KN, Morrey BF, Chao EY. Elbow joint contact study: comparison of techniques. J Biomech 1985; 18: 329-336. 148. Momose T, Nakatsuchi Y, Saitoh S. Contact area of the trapeziometacarpal joint. J Hand Surg Am 1999; 24: 491-495. 149. Seedhom BB, Tsubuku M. A technique for the study of contact between visco-elastic bodies with special reference to the patello-femoral joint. J Biomech 1977; 10: 253-260. 150. Ahmed AM. A pressure distribution transducer for in-vitro static measurements in synovial joints. J Biomech Eng 1983; 105: 309-314. 151. Hille E, Schulitz KP, Henrichs C, SchneiderT. Pressure and contract-surface measurements within the femoropatellar joint and their variations following lateral release. Arch Orthop Trauma Surg 1985; 104: 275-282. 152. Huberti HH, Hayes WC. Patellofemoral contact pressures. The influence of q-angle and tendofemoral contact. J Bone Joint Surg Am 1984; 66: 715-724. 153. Huberti HH, Hayes WC. Contact pressures in chondromalacia patellae and the effects of capsular reconstructive procedures. J Orthop Res 1988; 6: 499-508. 154. Goudakos IG, Konig C, Schottle PB, Taylor WR, Singh NB, Roberts I, et al. Stair climbing results in more challenging patellofemoral contact mechanics and kinematics than walking at early knee flexion under physiological-like quadriceps loading. J Biomech 2009; 42: 2590- 2596. 155. Luyckx T, Didden K, Vandenneucker H, Labey L, Innocenti B, Bellemans J. Is there a biomechanical explanation for anterior knee pain in patients with patella alta?: influence of patellar height on patellofemoral contact force, contact area and contact pressure. J Bone Joint Surg Br 2009; 91: 344-350. 156. Melegari TM, Parks BG, Matthews LS. Patellofemoral contact area and pressure after medial patellofemoral ligament reconstruction. Am J Sports Med 2008; 36: 747-752. 157. Rikli DA, Honigmann P, Babst R, Cristalli A, Morlock MM, Mittlmeier T. Intra-articular pressure measurement in the radioulnocarpal joint using a novel sensor: in vitro and in vivo results. J Hand Surg Am 2007; 32: 67-75.  167  158. Drewniak EI, Crisco JJ, Spenciner DB, Fleming BC. Accuracy of circular contact area measurements with thin-film pressure sensors. J Biomech 2007; 40: 2569-2572. 159. Wilson DR, Apreleva MV, Eichler MJ, Harrold FR. Accuracy and repeatability of a pressure measurement system in the patellofemoral joint. J Biomech 2003; 36: 1909-1915. 160. Wilson DC, Niosi CA, Zhu QA, Oxland TR, Wilson DR. Accuracy and repeatability of a new method for measuring facet loads in the lumbar spine. J Biomech 2006; 39: 348-353. 161. Martinelli L, Hurschler C, Rosenbaum D. Comparison of capacitive versus resistive joint contact stress sensors. Clin Orthop Relat Res 2006; 447: 214-220. 162. Scherrer PK, Hillberry BM, Vansickle DC. Determining the Invivo Areas of Contact in the Canine Shoulder. Journal of Biomechanical Engineering-Transactions of the Asme 1979; 101: 271-278. 163. Soslowsky LJ, Flatow EL, Bigliani LU, Pawluk RJ, Ateshian GA, Mow VC. Quantitation of in situ contact areas at the glenohumeral joint: a biomechanical study. J Orthop Res 1992; 10: 524-534. 164. Harris ML, Morberg P, Bruce WJ, Walsh WR. An improved method for measuring tibiofemoral contact areas in total knee arthroplasty: a comparison of K-scan sensor and Fuji film. J Biomech 1999; 32: 951-958. 165. Besier TF, Draper CE, Gold GE, Beaupre GS, Delp SL. Patellofemoral joint contact area increases with knee flexion and weight-bearing. J Orthop Res 2005; 23: 345-350. 166. Gold GE, Besier TF, Draper CE, Asakawa DS, Delp SL, Beaupre GS. Weight-bearing MRI of patellofemoral joint cartilage contact area. J Magn Reson Imaging 2004; 20: 526-530. 167. Heino Brechter J, Powers CM, Terk MR, Ward SR, Lee TQ. Quantification of patellofemoral joint contact area using magnetic resonance imaging. Magn Reson Imaging 2003; 21: 955- 959. 168. Hinterwimmer S, Gotthardt M, von Eisenhart-Rothe R, Sauerland S, Siebert M, Vogl T, et al. In vivo contact areas of the knee in patients with patellar subluxation. J Biomech 2005; 38: 2095-2101. 169. Hinterwimmer S, von Eisenhart-Rothe R, Siebert M, Welsch F, Vogl T, Graichen H. Patella kinematics and patello-femoral contact areas in patients with genu varum and mild osteoarthritis. Clin Biomech (Bristol, Avon) 2004; 19: 704-710.  168  170. Nakagawa S, Kadoya Y, Kobayashi A, Tatsumi I, Nishida N, Yamano Y. Kinematics of the patella in deep flexion. Analysis with magnetic resonance imaging. J Bone Joint Surg Am 2003; 85-A: 1238-1242. 171. Powers CM, Ward SR, Chan LD, Chen YJ, Terk MR. The effect of bracing on patella alignment and patellofemoral joint contact area. Med Sci Sports Exerc 2004; 36: 1226-1232. 172. von Eisenhart-Rothe R, Siebert M, Bringmann C, Vogl T, Englmeier KH, Graichen H. A new in vivo technique for determination of 3D kinematics and contact areas of the patello-femoral and tibio-femoral joint. J Biomech 2004; 37: 927-934. 173. Ward SR, Powers CM. The influence of patella alta on patellofemoral joint stress during normal and fast walking. Clin Biomech (Bristol, Avon) 2004; 19: 1040-1047. 174. Ward SR, Terk MR, Powers CM. Patella alta: association with patellofemoral alignment and changes in contact area during weight-bearing. J Bone Joint Surg Am 2007; 89: 1749-1755. 175. Connolly KD, Ronsky JL, Westover LM, Kupper JC, Frayne R. Differences in patellofemoral contact mechanics associated with patellofemoral pain syndrome. J Biomech 2009; 42: 2802- 2807. 176. Salsich GB, Ward SR, Terk MR, Powers CM. In vivo assessment of patellofemoral joint contact area in individuals who are pain free. Clin Orthop Relat Res 2003: 277-284. 177. Gill HS, O'Connor JJ. Biarticulating two-dimensional computer model of the human patellofemoral joint. Clin Biomech (Bristol, Avon) 1996; 11: 81-89. 178. van Eijden TM, Kouwenhoven E, Verburg J, Weijs WA. A mathematical model of the patellofemoral joint. J Biomech 1986; 19: 219-229. 179. Jafaril A, Farahmand F, Meghdari A. The effects of trochlear groove geometry on patellofemoral joint stability--a computer model study. Proc Inst Mech Eng H 2008; 222: 75- 88. 180. Hirokawa S. Three-dimensional mathematical model analysis of the patellofemoral joint. J Biomech 1991; 24: 659-671. 181. Hefzy MS, Yang H. A three-dimensional anatomical model of the human patello-femoral joint, for the determination of patello-femoral motions and contact characteristics. J Biomed Eng 1993; 15: 289-302.  169  182. Kwak SD, Blankevoort L, Ateshian GA. A Mathematical Formulation for 3D Quasi-Static Multibody Models of Diarthrodial Joints. Comput Methods Biomech Biomed Engin 2000; 3: 41-64. 183. Caruntu DI, Hefzy MS. 3-D anatomically based dynamic modeling of the human knee to include tibio-femoral and patello-femoral joints. J Biomech Eng 2004; 126: 44-53. 184. Elias JJ, Wilson DR, Adamson R, Cosgarea AJ. Evaluation of a computational model used to predict the patellofemoral contact pressure distribution. J Biomech 2004; 37: 295-302. 185. Mesfar W, Shirazi-Adl A. Biomechanics of the knee joint in flexion under various quadriceps forces. Knee 2005; 12: 424-434. 186. Besier TF, Gold GE, Beaupre GS, Delp SL. A modeling framework to estimate patellofemoral joint cartilage stress in vivo. Med Sci Sports Exerc 2005; 37: 1924-1930. 187. Fernandez JW, Akbarshahi M, Kim HJ, Pandy MG. Integrating modelling, motion capture and x-ray fluoroscopy to investigate patellofemoral function during dynamic activity. Comput Methods Biomech Biomed Engin 2008; 11: 41-53. 188. Elias JJ, Cosgarea AJ. Computational modeling: an alternative approach for investigating patellofemoral mechanics. Sports Med Arthrosc 2007; 15: 89-94. 189. Gillespie KA, Dickey JP. Determination of the effectiveness of materials in attenuating high frequency shock during gait using filterbank analysis. Clin Biomech (Bristol, Avon) 2003; 18: 50-59. 190. Besier TF, Gold GE, Delp SL, Fredericson M, Beaupre GS. The influence of femoral internal and external rotation on cartilage stresses within the patellofemoral joint. J Orthop Res 2008; 26: 1627-1635. 191. Elias JJ, Bratton DR, Weinstein DM, Cosgarea AJ. Comparing two estimations of the quadriceps force distribution for use during patellofemoral simulation. J Biomech 2006; 39: 865-872. 192. Devereaux MD, Lachmann SM. Patello-femoral arthralgia in athletes attending a Sports Injury Clinic. Br J Sports Med 1984; 18: 18-21. 193. Hahn T, Foldspang A. Prevalent knee pain and sport. Scand J Soc Med 1998; 26: 44-52. 194. Malek M, Mangine R. Patellofemoral pain syndromes: A comprehensive and conservative approach. J Orthop Sports Phys Ther. 1981; 2: 108-116.  170  195. Laurin CA, Levesque HP, Dussault R, Labelle H, Peides JP. The abnormal lateral patellofemoral angle: a diagnostic roentgenographic sign of recurrent patellar subluxation. J Bone Joint Surg Am 1978; 60: 55-60. 196. Merchant AC, Mercer RL, Jacobsen RH, Cool CR. Roentgenographic analysis of patellofemoral congruence. J Bone Joint Surg Am 1974; 56: 1391-1396. 197. Martinez S, Korobkin M, Fondren FB, Hedlund LW, Goldner JL. Diagnosis of patellofemoral malalignment by computed tomography. J Comput Assist Tomogr 1983; 7: 1050-1053. 198. Muellner T, Funovics M, Nikolic A, Metz V, Schabus R, Vecsei V. Patellar alignment evaluated by MRI. Acta Orthop Scand 1998; 69: 489-492. 199. Shellock FG, Stone KR, Crues JV. Development and clinical application of kinematic MRI of the patellofemoral joint using an extremity MR system. Med Sci Sports Exerc 1999; 31: 788- 791. 200. Powers CM, Shellock FG, Pfaff M. Quantification of patellar tracking using kinematic MRI. J Magn Reson Imaging 1998; 8: 724-732. 201. McWalter EJ, Cibere J, MacIntyre NJ, Nicolaou S, Schulzer M, Wilson DR. Relationship between varus-valgus alignment and patellar kinematics in individuals with knee osteoarthritis. J Bone Joint Surg Am 2007; 89: 2723-2731. 202. Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ 1996; 312: 1254-1259. 203. Johnell O, Kanis JA, Oden A, Johansson H, De Laet C, Delmas P, et al. Predictive value of BMD for hip and other fractures. J Bone Miner Res 2005; 20: 1185-1194. 204. Cohen J. Statistical Power Analysis for Behavioral Sciences. 2nd Edition. Hillsdale, NJ, L. Erlbaum Associates 1988. 205. Bouxsein ML, Coan BS, Lee SC. Prediction of the strength of the elderly proximal femur by bone mineral density and quantitative ultrasound measurements of the heel and tibia. Bone 1999; 25: 49-54. 206. MacIntyre NJ, McKnight EK, Day A, Wilson DR. Consistency of patellar spin, tilt and lateral translation side-to-side and over a 1 year period in healthy young males. J Biomech 2008; 41: 3094-3096.  171  207. McWalter EJ, Hunter DJ, Wilson DR. The effect of load magnitude on three-dimensional patellar kinematics in vivo. J Biomech 2010; 43: 1890-1897. 208. Besl PJ, McKay ND. A method for registration of 3-D shapes. IEEE Transaction on Pattern Analysis and Machine Intelligence 1992; 14: 239-255. 209. McWalter EJ, Hunter DJ, Wilson DR. Effect of Load on Three-dimensional Patellar Kinematics in Healthy Individuals. In: Orthopaedic Research Society Transactions, vol. 33(1205). San Francisco 2008. 210. Laprade J, Culham E. Radiographic measures in subjects who are asymptomatic and subjects with patellofemoral pain syndrome. Clin Orthop Relat Res 2003: 172-182. 211. Draper CE, Besier TF, Fredericson M, Santos JM, Beaupre GS, Delp SL, et al. Differences in patellofemoral kinematics between weight-bearing and non-weight-bearing conditions in patients with patellofemoral pain. J Orthop Res 2010. 212. van Eijden TM, de Boer W, Weijs WA. The orientation of the distal part of the quadriceps femoris muscle as a function of the knee flexion-extension angle. J Biomech 1985; 18: 803- 809. 213. Dillon CF, Rasch EK, Gu Q, Hirsch R. Prevalence of knee osteoarthritis in the United States: arthritis data from the Third National Health and Nutrition Examination Survey 1991-94. J Rheumatol 2006; 33: 2271-2279. 214. Stanford W, Phelan J, Kathol MH, Rooholamini SA, el-Khoury GY, Palutsis GR, et al. Patellofemoral joint motion: evaluation by ultrafast computed tomography. Skeletal Radiol 1988; 17: 487-492. 215. Muhle C, Brinkmann G, Skaf A, Heller M, Resnick D. Effect of a patellar realignment brace on patients with patellar subluxation and dislocation. Evaluation with kinematic magnetic resonance imaging. Am J Sports Med 1999; 27: 350-353. 216. Heino Brechter J, Powers CM. Patellofemoral stress during walking in persons with and without patellofemoral pain. Med Sci Sports Exerc 2002; 34: 1582-1593. 217. Boehnen C, Flynn P. Accuracy of 3D scanning technologies in a face scanning scenario. Fifth International Conference on 3-D Digital Imaging and Modeling, Proceedings 2005: 310-317. 218. Veldpaus FE, Woltring HJ, Dortmans LJ. A least-squares algorithm for the equiform transformation from spatial marker co-ordinates. J Biomech 1988; 21: 45-54.  172  219. Lamoreux LW. Kinematic measurements in the study of human walking. Bull Prosthet Res 1971; 10: 3-84. 220. McWalter EJ. Links between three-dimensional patellar kinematics, cartilage morphology and varus/valgus alignment in early knee osteoarthritis. 2004. MASc Thesis. University of British Columbia 221. Eckstein F, Heudorfer L, Faber SC, Burgkart R, Englmeier KH, Reiser M. Long-term and resegmentation precision of quantitative cartilage MR imaging (qMRI). Osteoarthritis Cartilage 2002; 10: 922-928. 222. Greaves LL, Gilbart MK, Yung A, Kozlowski P, Wilson DR. Deformation and recovery of cartilage in the intact hip under physiological loads using 7T MRI. J Biomech 2009; 42: 349- 354. 223. Karrholm J. Roentgen stereophotogrammetry. Review of orthopedic applications. Acta Orthop Scand 1989; 60: 491-503. 224. Andriacchi TP, Johnson TS, Hurwitz DE, Natarajan RN. Musculoskeletal Dynamics, Locomotion, and Clinical Applications. In: Basic Orthopaedic Biomechanics and Mechano- biology, Mow VC, Huiskes R, 3rd Edition, Philadelphia, Lippincott Williams & Wilkins 2005.    173  Appendix A: Glossary of Terms Abduction Movement away from the midline; opposite of adduction. Adduction Movement towards the midline; opposite of abduction. Anterior Towards the front of the body; opposite of posterior. Articular Cartilage A viscoelastic material that covers the articulating surface of joints and distributes load. Axial The plane that divides the body into proximal (upper) and distal (lower) portions.  Also known as transverse. Cartilage See articular cartilage. Computed Tomography (CT) A three-dimensional radiography-based method of acquiring anatomical images. Condyle A round bony projection at a joint. Contact Area In this case refers to articular cartilage contact area.  The area of contact between two articular cartilage surfaces when a joint is loaded. Coronal The plane that divides the body into anterior and posterior.  Also known as frontal. Distal Towards the feet; opposite of distal. Epicondyle An additional bony projection ‘on the condyle’. Extension Increase in the angle between two segments; opposite of flexion. Ex vivo ‘Out of living’.  In this case describes experiments done in cadaver specimens. Femur The thigh bone. Fiducial Reference markers visible with multiple data collection modalities for use in registration. Flexion Decrease in the angle between two segments; opposite of extension. Frontal The plane that divides the body into anterior and posterior.  Also known as coronal. Goniometer A tool for measuring angles between two segments.  174  In vivo ‘Within the living’.  In this case refers to experiments done using live volunteers or patients. Kinematics The study of motion of bodies or systems without considering forces. Lateral Away from the sagittal midline; opposite of lateral. Ligament Fibrous tissue that attaches bone to bone. Magnetic Resonance Imaging (MRI) A three-dimensional medial imaging technique that employs magnetic fields and radiofrequency pulses to image anatomical structures. Medial Towards the sagittal midline; opposite of lateral. Patella An inverted tear-drop shaped bone in the leg.  Also known as the knee-cap. Patellofemoral Pertaining to the joint between the patella and the femur. Phantom An object used to simulate biological tissues in order to validate or calibrate imaging techniques. Photogrammetry A method of determining geometric properties of an object using photographs. Posterior Towards the back of the body; opposite of anterior. Proximal Towards the top of the body; opposite of distal. Radiograph A two-dimensional image of an anatomical structure created using X- rays. Roentgen Stereophotogrammetric Analysis (RSA) A method in which accurate three-dimensional positions of objects can be determined using radiographs [223]. Sagittal The plane that divides the body into left and right. Spin Used in reference to the patella.  A rotation about an anteriorly directed axis. Tendon Fibrous tissue that attaches bone to muscle. Tibia The shin bone. Tilt Used in reference to the patella.  A rotation about a proximally or distally directed axis. Transverse Divides the body into upper and lower portions. Also known as axial. Tuberosity An outgrowth on a bone.  175  Valgus ‘Knock-knees’. Varus ‘Bow-legs’.    176  Appendix B: Ethics Certificates   177   178  Appendix C: Loading Rig The loading rig used in this thesis was designed and constructed as a part of my MASc thesis project [220].  The loading system consists of a foot plate attached to a loading platform that hinges on an inversed U-shaped frame (Figure C-1); therefore, the foot plate system is free to rotate about the crossbar of the frame.  When in use, the pedal is positioned so that it is at approximately 10° from vertical and a prescribed load is placed on the loading platform (Figure C-1).  The participant’s foot is placed on the loading plate and the load is released; the participant is instructed to maintain his or her leg position so that the angle of the pedal does not change.  The amount of load required to create a particular axial load was determined experimentally using a single axis load cell.  The load was assumed to act at 15 cm from the bottom of the pedal for the purpose of this calibration.  This is approximately at the centre of the foot.  From this an estimate of shear load could be determined from the free body diagram (Figure C-2) and the equation in the rig loading section below.  For example, for an individual who weighs 80 kg, the shear loads associated with an axial load of approximately 15% BW (118 N) was approximately 60 N and for 30% BW (236 N) was approximately 120 N. These loads applied by the rig can then be used to estimate the intersegmental forces and moments at the knee joint (external forces and moments).  The segment lengths and the weight of the shank in the sample calculation were based on an example of a sagittal plane kinetic gait analysis found in a prominent textbook chapter [224]; this allows for direct comparison of intersegmental forces and moments at the knee.  In the section on intersegmental loads below, a schematic of the system, free-body diagrams of the foot and the shank and equilibrium equations are shown (Figures C-3 to C-5).  These equations can be used to estimate intersegmental forces and moments at the different load levels for different knee flexion angles (Table C-1).  For example, the external moments at the knee ranged from 10 Nm for the 15% BW load at 45° to 84 Nm for the 30% BW load at full extension.  In the textbook example, the external moment at the knee was 43 Nm when ground reaction forces of 700 N in the vertical direction and 150 N in the horizontal direction were applied (textbook Figure 3-6 and Table 3-1) [224].  Therefore, the external moments in the sagittal plane range from approximately half to double of what would be experienced during gait.  The external moments at the knee created by the loading rig at the 15 and 30% BW loads applied to the supine loading scenario create are similar to those experienced during gait; however, this is a measure of resultant external moment only and does not tell us about internal contact or muscle loads which would have to be estimated using many assumption and simplifications and a reduction or optimization algorithm.  179   M   Figure C-1: Photographs of MRI safe knee loading rig. The participant places foot on foot plate and load is placed on the loading platform.    180  Loads Applied to Foot by Rig Fweight Fsystem Ffoot-axial Ffoot-shear +’ve directions x y d1 d2 d3  θ Figure C-2: Free body diagram of the loading system. The red cross indicates the location about which the moments were taken.  D1, d2, and d3 are distances measured from the system.  The weight of the loading system acts at the point that moments were taken about.    Sum of forces and moments about the origin:  ∑ = 0M      ( ) ( )32)(32)(1 sincoscossin0 ddFddFdF shearfootaxialfootw ×+×−+×+×−×= θθθθ  The following assumptions and approximations were made to use this equation: à moments were taken about the pivot point of the loading pedal (red cross on diagram below) à d1=0.3 m, d2=0.1 m, d3=0.2 m, θ=10° à the weight of the system acts at the pivot point à the pivot point is frictionless  181  Analysis of Intersegmental Loads  Figure C-3: Schematic of foot, shank and rig.     V1 Fankle-x Fankle-y Mankle Frig-ax Frig-sh θ1 h1  +’ve directions x y  Figure C-4: Forces applied by loading rig and intersegmental forces and moments at the ankle.   Equations of equilibrium for foot (moments about ankle):  ∑ = 0xF       xankleshrigaxrig FFF −−− −−= 11 sincos0 θθ ∑ = 0yF       yankleshrigaxrig FFF −−− −+= 11 cossin0 θθ ∑ = 0M      ankleshrigshrigaxrigaxrig MhFvFhFvF +×−×+×−×−= −−−− 11111111 cossinsincos0 θθθθ  Assumptions:  à rig-ax and rig-sh are the axial and shear loads applied to the foot by the loading rig  à weight of foot is negligible  à h1=0.16 sinθ1, v1=0.16 cosθ1, θ1=10°  182  Fknee-x Fknee-y Mknee FmgFankle-x Fankle-y Mankle h2 V2 h3 θ2  +’ve directions x y  Figure C-5: Intersegmental forces and moments at the ankle and knee and weight of shank..   Equations of equilibrium for shank (moments about knee):  ∑ = 0xF       xkneexankle FF −− −=0 ∑ = 0yF       ykneemgyankle FFF −− −−=0 ∑ = 0M      kneeanklemgyanklexankle MMhFhFvF +−×+×−×= −− 3220  Assumptions:  à the forces applied to the posterior side of the shank at the knee are negligible compared to the intersegmental forces and therefore not included in the analysis  à h2 = 0.32 cosθ2, v2 = 0.32 sinθ2, h3 = 0.15 cosθ2, θ2 = 0, 15, 30 or 45°            183   Table C-1: Intersegmental knee joint loads and moments for 15% and 30% BW loading conditions.  Knee Flexion Angle 15% BW Load 30% BW Load  Fknee-x (N) Fknee-y (N) Mknee (Nm) Fknee-x (N) Fknee-y (N) Mknee (Nm) 0° 106 52 40 212 131 84 15° 106 52 31 212 131 65 30° 106 52 20 212 131 44 45° 106 52 10 212 131 23   184  Appendix D: Details of Kinematic Analysis This appendix provides further detail of the MRI-based three-dimensional patellar kinematic assessment developed and validated previously by our group [102,132]. Segmentation: Bone is segmented from the MRI images in a slice-by-slice using commercially available software (Analyze 8.0, Analyze Direct, Overland Park, KS, USA) (Figure D-1).  Bone models are created using the Adapt/Deform tool in the Surface Extractor module and bone contours were created using the Contour tool in the same module (Figure D-1).   Figure D-1: Example of segmented bone from a single MRI slice (left), of a bone model created from the high-resolution, unloaded bone MRI scan (centre) and of bone contours created from a low-resolution, loaded bone MRI scan (right).  Registration:  All bone contours, at all loaded knee flexion angles are registered to the bone models (Figure D-2).  Figure D-2: Registration of bone contour to bone model using an ICP algorithm [208].  185  The ICP algorithm developed by Besl and McKay 1992 [208] was used to transform the model set to the location of the contour set by minimizing the Euclidean distance between the sets using the following method: à The iteration was initialized by setting the error between the model and contour sets to be large and defining the homogeneous transformation matrix to be a 4 x 4 identity matrix. à The iteration began by determining the shortest Euclidean distance between each point in contour set and the model set; a new closest points contour set was created by selecting the point in the model to which the minimum distance corresponds. à The homogeneous transformation matrix between the original contour set and the new closest points contour set was determined using a least-squares based algorithm by Veldpaus et al. 1988 [218].  This algorithm was developed for the application of determining the transformation matrices between bone positions during motion determined using imaging methods but is also suitable for the present application.  The algorithm determines the homogeneous transformation matrix between positions as follows: à The homogeneous transformation matrix was determined using approximations of the translation vector and the rotation matrix that minimize the following unweighted least- squares function: ( ) ( )( ) ( )( )∑ = ⎥⎦ ⎤⎢⎣ ⎡ −−−−×−−−−= m i ii T ii aaHrapaaHrapm Hrf 1 ˆˆˆˆˆˆ1ˆ,ˆ Where: r̂  is the approximation of the translation vector        Ĥ  is the approximation of the rotation matrix       a is the position vector of the centre of the first point cloud       ai is the position vector of the individual point in the first point cloud      is the position vector of the centre of the second contour point cloud set as compared to the first given by pi=a+r+R(ai-a) where R is RTR=I ip̂ à While a detailed account of the algorithm used to solve this equation can be found in the publication by Veldpaus et al 1988 (including the FORTRAN code), briefly, an average matrix was created from the average vectors of the centre of each point cloud; the Newton- Raphson method is used to determine the invariants of the average matrix (convergence  186  tolerance set to 10E-10); the invariants were used to calculate the rotation matrix and the transformation vector; and the homogeneous transformation matrix is assembled. à The error in the method was determined by using the homogeneous transformation matrix to transform the original contour set to the location of the closest points contour set; the distance between points in the model set and the transformed contour set was determined and the mean distance between points was considered the error. à The homogeneous transformation matrix was used to transform the model set to the dataset. à The difference in error between the error of the previous iteration (or initial value) and the present iteration was calculated.  The iteration continued while the shapematch error changes by less than a prescribed value between iterations (0.00001 in this thesis). à The overall homogeneous transformation matrix from the original model position to the final registered position was calculated (using the method described above). The ICP algorithm is very good at finding the local minimum of the system; however, there is the possibility for it to converge on an incorrect solution.  This is highly dependent on how the model and contour sets are initially aligned.  Therefore, a precomputation step was added immediately prior to the ICP registration step in order ensure the initial positions are close to the final solution.  This was carried out by manually selecting anatomical landmarks that were visible in the model and contour sets and running a quick ICP (25 iterations). This step ensures that the ICP will converge on the correct solution.  Anatomical axes assignment: The anatomical landmarks used to create the anatomical coordinate systems are shown in Figure 2-2.  These landmarks were used to create the coordinate systems as follows:        187  Coordinate System Origin/Axis Description Femur  Origin Most proximal point in the intercondylar notch, defined in the sagittal plane  Long Origin to centroid of proximal femoral shaft  Third Cross product of the long axis with the vector that joins the posterior condylar points  Flexion Cross product of the long and the third axis  Patella  Origin Most posterior point on axial midslice  Long The vector that joins the proximal and distal points on the sagittal midslice  Third Cross product of the long axis and the vector that joins the origin and the most lateral point in the axial midslice  Flexion Cross product of the long and the third axis Tibia  Origin Most proximal point of the medial intercondylar eminence  Long Origin to centroid of distal tibial shaft  Third Cross product of the long axis with the vector that joins the posterior points on in the axial slice containing the proximal point of the fibula  Flexion Cross product of the long and the third axis  Kinematics:  Three-dimensional patellar kinematics were calculated at each position using the anatomical axes described above and the final homogeneous transformation matrix estimated in the registration step.  The following procedure was followed for each loaded knee flexion position: à The anatomically based coordinate systems were transformed from the bone model to the contour set using the homogeneous transformation matrix. à Patellar position as compared to the femoral coordinate system was calculated using the following method:  188  o The vector between the origin of the transformed patellar and femoral coordinate systems was found: originfemurdtransformeoriginpatelladtransformevectororalpatellofem _____ −= o The dot products of the patellofemoral vector and each of the vectors that make up the femoral coordinate system were used to determine the proximal, lateral and anterior translations. thirdfemurvectororalpatellofemcomponentanterior flexionfemurvectororalpatellofemcomponentlateral longfemurvectororalpatellofemcomponentproximal ___ ___ ___ •= •= •=  à Patellar attitude was calculated in a side dependent manner (left or right) based on the modified Joint Coordinate System described in Sections 1.4.2 and 2.2.3 using the following method: o The joint coordinate system was created with:  312 __3 __1 eee axislongpatellare axisflexionfemure ×= = = o Patellar flexion occurs about e1 and was calculated as follows: ( )thirdfemurea pi flexionpatellar _2cos180_ •∗= o Patellar spin occurs about e2 and was calculated as follows: ( ) ⎟⎠ ⎞⎜⎝ ⎛ •× ×∗= longpatella eflexionfemurnorm eflexionfemur a pi spinpatellar _ 2_ 2_ cos 180 _ o Patellar tilt occurs about e3 and was calculated as follows: ( )thirdpatellarea pi tiltpatellar _2cos180_ •∗=  189  Appendix E: Linear Hierarchical Random-effects Models Rationale for use: Linear hierarchical random-effects models (also known as generalized linear mixed models, generalized linear models, linear mixed models or linear multilevel models) are suitable for data that cannot be analyzed using more traditional statistical methods, such as repeated measures ANOVA, because of the inherent nature of the data.  Such cases include: longitudinal data, unbalanced repeated measures data, correlated data and nested data.  Repeated measures ANOVA assumes a randomized block design (the order in which the data is collected is randomized), sphericity (all pair-wise comparisons have the same variance) and compound symmetry (homogeneity of variance-covariance matrix).  The kinematic data acquired in this thesis has a nested structure (flexion angles within individuals within conditions) and the repeated measures are unbalanced for some individuals (due to missing data at some flexion angles).  Further, data were acquired at tibiofemoral flexion angles in ascending order and therefore the assumption of randomized block design is violated.  Finally, a standard, hand-held goniometer (error of up to ±10°) was used to position the participant’s knee at the prescribed tibiofemoral flexion angle; however, errors in tibiofemoral flexion angles could be reduced by calculating them from the image data using the Joint Coordinate System (Appendix D).  As a result, these angles were not consistent between participants making a test such as a repeated measures ANOVA impossible without binning or interpolation between data points.  Advantages of using linear hierarchical random effects models are that they take into consideration the within subject correlation of the data at different tibiofemoral flexion angles and participants with missing data can still be used in the analysis but the contribution of these data are weighted accordingly towards the overall mean. Description of Modelling Method and Interpretation of Results:  Hierarchical linear random-effects models are similar to regression models however the final model is a weighted average of individual subject data.  First, regression models are fit to each individual’s kinematic data (dependent variable: patellar flexion, spin, tilt, proximal translation, lateral translation, anterior translation; independent variable: tibiofemoral flexion).  This is level 1 of the model.  The results of these individual regression models are pooled into a single model based on a weighted average of the level 1 result.  Weighting is based on the number of data points for that individual in level 1 and the distance of the level 1 data from the pooled mean (for example, an individual with missing data or with data further from the mean would contribute less to the pooled mean).  The pooling of individual data within a single condition is level 2 of the model.  Using the  190  data structure of Chapter 3 as an example, to create a hierarchical linear random-effects model, the statistical package (Stata 10, StataCorp LP, College Station, TX, USA) fits regression models to each individual’s kinematic data (for each parameter) at all three load levels; the final model is a weighted average of the data for each of the three loading conditions. The structure of the model to be fitted must be decided a prior.  There are various considerations such as the form of the model (linear, quadratic, etc) and whether the variables should be fixed or random.  In this study, the quadratic model below was tested first. anglekneeconditionconditionanglekneeanglekneey _***_*_* 43 2 210 βββββ ++++= If the quadratic term (β2) was significant it was included in the model, if not it was removed.  This choice of model was based on previous models of three-dimensional patellar kinematic data carried out by our group [134,201] and other groups [104].  In this model, the conditions were fixed in all cases (only the specified conditions were studied) as were the tibiofemoral flexion angles.  Subject was a random variable because these individuals represent a sample of the larger population.  Slope (β4) and intercept (β3) were also random variables (allowed to vary between conditions).  If the random slope term was not statistically significant it was not included in the model (ie was then a fixed variable).  Examples of a fixed and random slope with a random intercept can be seen in below (Figure D-1).  Results from the model were considered significant if p<0.05 (two-tailed test).  Tibiofemoral Flexion K in em at ic  P ar am et er K in em at ic  P ar am et er   Tibiofemoral Flexion K in em at ic  P ar am et er K in em at ic  P ar am et er Figure E-1: Example of a fixed slope and a random intercept (left) and a random slope and random intercept (right) between two conditions. When the slope is fixed, significant differences in the y-intercept term indicate an offset between groups, when slope is random, significant differences in the slope term indicate a difference in patterns between groups.    191  Appendix F: Raw Data from Loading Study This appendix provides examples of raw data from the loading study.  Two representative data sets will be shown for each kinematic parameter. Patellar Flexion Subject 2 -10 -5 0 5 10 15 20 25 30 35 -10 0 10 20 30 40 50 60 Tibiofemoral Flexion (degrees) Pa te lla r F le xi on  (d eg re es ) 0% BW load 15% BW load 30% BW load Figure F-1: Raw loading study data for patellar flexion for subject 2.   192  Subject 8 -10 -5 0 5 10 15 20 25 30 35 -5 0 5 10 15 20 25 30 35 40 45 Tibiofemoral Flexion (degrees) Pa te lla r F le xi on  (d eg re es ) 0% BW load 15% BW load 30% BW load Figure F-2: Raw loading study data for patellar flexion for subject 8.  Patellar Spin Subject 2 -10 -8 -6 -4 -2 0 2 4 6 -10 0 10 20 30 40 50 60 Tibiofemoral Flexion (degrees) Pa te lla r S pi n (d eg re es ) 0% BW load 15% BW load 30% BW load Figure F-3: Raw loading study data for patellar spin for subject 2.  193  Subject 9 -10 -8 -6 -4 -2 0 2 4 6 -10 0 10 20 30 40 50 Tibiofemoral Flexion (degrees) Pa te lla r S pi n (d eg re es ) 0% BW load 15% BW load 30% BW load Figure F-4: Raw loading study data for patellar spin for subject 9.  Patellar Tilt Subject 1 -10 -5 0 5 10 15 20 25 30 0 5 10 15 20 25 30 35 40 45 50 Tibiofemoral Flexion (degrees) Pa te lla r T ilt  (d eg re es ) 0% BW load 15% BW load 30% BW load Figure F-5: Raw loading study data for patellar tilt for subject 1.  194  Subject 6 -10 -5 0 5 10 15 20 25 30 -10 0 10 20 30 40 50 60 Tibiofemoral Flexion (degrees) Pa te lla r T ilt  (d eg re es ) 0% BW load 15% BW load 30% BW load Figure F-6: Raw loading study data for patellar tilt for subject 6.  Proximal Patellar Translation Subject 4 0 5 10 15 20 25 30 35 40 -15 -5 5 15 25 35 45 Tibiofemoral Flexion (degrees) Pr ox im al  T ra ns la tio n (m m ) 0% BW load 15% BW load 30% BW load Figure F-7: Raw loading study data for proximal patellar translation for subject 4.  195  Subject 10 -10 -5 0 5 10 15 20 25 30 35 40 -15 -5 5 15 25 35 45 Tibiofemoral Flexion (degrees) Pr ox im al  T ra ns la tio n (m m ) 0% BW load 15% BW load 30% BW load Figure F-8: Raw loading study data for proximal patellar translation for subject 10.  Lateral Patellar Translation Subject 10 -10 -8 -6 -4 -2 0 2 4 6 -15 -5 5 15 25 35 45 Tibiofemoral Flexion (degrees) La te ra l T ra ns la tio n (m m ) 0% BW load 15% BW load 30% BW load Figure F-9: Raw loading study data for lateral patellar translation for subject 10.  196  Subject 5 -10 -8 -6 -4 -2 0 2 4 6 -15 -5 5 15 25 35 45 55 Tibiofemoral Flexion (degrees) La te ra l T ra ns la tio n (m m ) 0% BW load 15% BW load 30% BW load Figure F-10: Raw loading study data for lateral patellar translation for subject 5.  Anterior Patellar Translation Subject 5 10 15 20 25 30 35 40 -15 -5 5 15 25 35 45 55 Tibiofemoral Flexion (degrees) A nt er io r T ra ns la tio n (m m ) 0% BW load 15% BW load 30% BW load Figure F-11: Raw loading study data for anterior patellar translation for subject 5.  197  Subject 9 10 15 20 25 30 35 40 -15 -5 5 15 25 35 45 55 Tibiofemoral Flexion (degrees) A nt er io r T ra ns la tio n (m m ) 0% BW load 15% BW load 30% BW load Figure F-12: Raw loading study data for anterior patellar translation for subject 9.   198  Appendix G: Additional Information on Parent Bracing Study Study Title: ‘A Randomized Trial of Patellofemoral Bracing for Treatment of Patellofemoral Osteoarthritis5’ This appendix summarizes the research question, methods and main findings of the parent study of Chapter 4.  This work has been submitted for publication to Osteoarthritis and Cartilage and is currently in the revision stage (as of September 30, 2010). Study population: 592 individuals were screened by telephone for participation in this study; 145 were eligible for a screening visit.  80 individuals were determined to be eligible for this study and participated.  These individuals had radiographic lateral patellofemoral OA (63 females, 17 males, mean age 61 years, mean body mass index 28 kg/m2); 40 individuals had isolated patellofemoral OA and 40 individuals had concomitant tibiofemoral and patellofemoral OA.  Participants also had knee pain, aching or stiffness on most days of the past month and a definite osteophyte in the patellofemoral joint. Methods: In this double blind, randomized crossover trial, two treatments were tested: 1) an active treatment in which the BioSkin Q patellofemoral brace with medialization strap (Cropper Inc, Ashland, OR, USA); and 2) a control treatment in which the BioSkin Q patellofemoral brace was used without the medialization strap.  The trial was 18 weeks in length and consisted of 6 weeks of each active and control treatments and a 6 week washout period between treatments.  Patients were randomized to one of two groups.  Group 1 received the active treatment first; group 2 received the control treatment first.  The primary outcome was change in knee pain that was assessed during each treatment period (active and control) using a visual analog scale (1-100).  The WOMAC pain, function and stiffness scores were also included as secondary outcomes.  The relationship between the intervention and the visual analog pain score was determined using a linear regression model (based on an unstructured correlation matrix for within participant observations and a generalized estimating equation).  The relationship between the intervention and the secondary outcomes were assessed in a  5 The work summarized in this appendix is based on the study: A Randomized Trial of Patellofemoral Bracing for Treatment of Patellofemoral Osteoarthritis by D.J. Hunter, W.F. Harvey, K.D. Gross, D. Felson, P. McCree, L. Li, K.A. Hirko, B. Zhang, K. Bennell which is currently in the review process at Osteoarthritis and Cartilage   199  similar manner.  The advantage of a crossover trial is that the individual acts as his or her own control. Results: The regression model examining the main effect of the intervention on visual analog pain score showed no statistical significant effect of treatment (-0.68 visual analog scale units, p=0.81), nor did any of the secondary outcomes (0.11 WOMAC pain, p=0.77; -0.02 WOMAC function, p=0.98; -0.11 WOMAC stiffness, p=0.61). Clinical Significance: The patellofemoral medialization brace tested appears to reduce patient pain in a clinical or statistically significant manner.  200  Appendix H: Dye Leaching Test The effect of dye leaching on contact contour analysis was assessed in a qualitative manner.  A c-clamp with a circular face 13 mm in diameter was clamped to the cartilage surface of the patella, thereby loading the cartilage.  Food grade dye (Yellow, Club House, McCormick Canada, London, ON) was applied liberally to the patella in the region surrounding the clamp.  The clamp was removed and the contact contour (contour of the region devoid of dye) was digitized using an optical tracking system (Optotrak Certus, Northern Digital Inc, Waterloo, ON).  Digitization was repeated at 10 minute increments over a 2 hour period.  Digitized points were plotted over the colour spectrum using custom software (Matlab, The Mathworks, Natick, MA, USA).  Results qualitatively show that the error in digitization is greater than the error in leaching.  Had leaching been an issue a progressive colour shift would be evident; however, this was not the case (Figure B-1).  The maximum distance between digitized contours was approximately 1.5 mm which is not surprising since the tip of the digitizing probe is a ball bearing 1mm in diameter.  Therefore, in the Chapter 5, three measurements were taken and averaged in order to get a better measure of the contact contour boundary. (mm) (mm)  Figure H-1:  Dye leaching over time. The legend colour spectrum varies with time, therefore had there been an effect of dye leaching over time this pattern would be apparent by a shift in the colour spectrum (green at the outer margin, transitioning to blue then red as the contact boundaries moved inward).  201  Appendix I: Sensitivity of Centroids to Proximity Threshold 0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) an te rio r/p os te rio r c oo rd in at e o f c en tro id  (m m ) Model MRI Mean Dye Anterior Posterior  Figure I-1: Proximity threshold sensitivity of anterior/posterior position of centroid for Specimen 1. 0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) m ed ial /la te ra l c oo rd in at e o f c en tro id  (m m ) Model MRI Mean Dye Lateral Medial  Figure I-2: Proximity threshold sensitivity of medial/lateral position of centroid for Specimen 1.   202  -6 -4 -2 0 2 4 6 8 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) pr ox im al/ di st al co or di na te  o f c en tro id  (m m ) Model MRI Mean Dye Proximal Distal  Figure I-3: Proximity threshold sensitivity of proximal/distal position of centroid for Specimen 1. -2 0 2 4 6 8 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) an te rio r/p os te rio r c oo rd in at e o f c en tro id  (m m ) Model MRI Mean Dye Anterior Posterior  Figure I-4: Proximity threshold sensitivity of anterior/posterior position of centroid for Specimen 2.  203  0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) m ed ial /la te ra l c oo rd in at e o f c en tro id  (m m ) Model MRI Mean Dye Lateral Medial  Figure I-5: Proximity threshold sensitivity of medial/lateral position of centroid for Specimen 2. 0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) pr ox im al/ di st al co or di na te  o f c en tro id  (m m ) Model MRI Mean Dye Proximal Distal  Figure I-6: Proximity threshold sensitivity of proximal/distal position of centroid for Specimen 2.  204  0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) an te rio r/p os te rio r c oo rd in at e o f c en tro id  (m m ) Model MRI Mean Dye Anterior Posterior  Figure I-7: Proximity threshold sensitivity of anterior/posterior position of centroid for Specimen 3. 0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) m ed ial /la te ra l c oo rd in at e o f c en tro id  (m m ) Model MRI Mean Dye Lateral Medial  Figure I-8: Proximity threshold sensitivity of medial/lateral position of centroid for Specimen 3.  205  0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) pr ox im al/ di st al co or di na te  o f c en tro id  (m m ) Model MRI Mean Dye Proximal Distal  Figure I-9: Proximity threshold sensitivity of proximal/distal position of centroid for Specimen 3.  0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) an te rio r/p os te rio r c oo rd in at e o f c en tro id  (m m ) Model MRI Mean Dye Anterior Posterior  Figure I-10: Proximity threshold sensitivity of anterior/posterior position of centroid for Specimen 4.      206  0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) m ed ial /la te ra l c oo rd in at e o f c en tro id  (m m ) Model MRI Mean Dye Lateral Medial  Figure I-11: Proximity threshold sensitivity of medial/lateral position of centroid for Specimen 4. 0 1 2 3 4 5 6 7 8 9 10 0 0.2 0.4 0.6 0.8 1 1.2 1.4 threshold (mm) pr ox im al/ di st al co or di na te  o f c en tro id  (m m ) Model MRI Mean Dye Proximal Distal  Figure I-12: Proximity threshold sensitivity of proximal/distal position of centroid for Specimen 4.    207  Appendix J: Sensitivity of Centroid to Kinematics Input -1 1 3 5 7 9 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) an te rio r/p os te rio r c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Extension Flexion Anterior Posterior  Figure J-1:  Sensitivity of model predicted anterior/posterior centroid position to patellar flexion.   -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) m ed ia l/l at er al  c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Extension Flexion Medial Lateral  Figure J-2:  Sensitivity of model predicted medial/lateral centroid position to patellar flexion.  208  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) pr ox im al /d is ta l c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Extension Flexion Distal Proximal  Figure J-3:  Sensitivity of model predicted proximal/distal centroid position to patellar flexion.  -1 0 1 2 3 4 5 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) an te rio r/p os te rio r c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 External Internal Anterior Posterior  Figure J-4:  Sensitivity of model predicted anterior/posterior centroid position to patellar spin.  209  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) m ed ia l/l at er al  c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Medial Lateral External Internal  Figure J-5:  Sensitivity of model predicted medial/lateral centroid position to patellar spin.  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) pr ox im al /d is ta l c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Distal Proximal External Internal   Figure J-6:  Sensitivity of model predicted proximal/distal centroid position to patellar spin.  210  -1 0 1 2 3 4 5 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) an te rio r/p os te rio r c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Lateral Medial Anterior Posterior  Figure J-7:  Sensitivity of model predicted anterior/posterior centroid position to patellar tilt.  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) m ed ia l/l at er al  c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Medial Lateral Lateral Medial  Figure J-8: Sensitivity of model predicted medial/lateral centroid position to patellar tilt.   211  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (degrees) pr ox im al /d is ta l c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Distal Proximal Lateral Medial  Figure J-9: Sensitivity of model predicted proximal/distal centroid position to patellar tilt.   -1 0 1 2 3 4 5 6 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) an te rio r/p os te rio r c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Distal Proximal Anterior Posterior  Figure J-10:  Sensitivity of model predicted anterior/posterior centroid position to proximal patellar translation.   212  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) m ed ia l/l at er al  c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Distal Proximal Medial Lateral  Figure J-11:  Sensitivity of model predicted medial/lateral centroid position to proximal patellar translation.  -1 0 1 2 3 4 5 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) an te rio r/p os te rio r c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Medial Lateral Anterior Posterior  Figure J-12:  Sensitivity of model predicted anterior/posterior centroid position to lateral patellar translation.  213  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) m ed ia l/l at er al  c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Medial Lateral Medial Lateral  Figure J-13:  Sensitivity of model predicted medial/lateral centroid position to lateral patellar translation.  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) pr ox im al /d is ta l c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Distal Proximal Medial Lateral  Figure J-14:  Sensitivity of model predicted proximal/distal centroid position to lateral patellar translation.  214  -1 0 1 2 3 4 5 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) an te rio r/p os te rio r c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Posterior Anterior Anterior Posterior  Figure J-15:  Sensitivity of model predicted anterior/posterior centroid position to anterior patellar translation.  -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) m ed ia l/l at er al  c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Medial Lateral Posterior Anterior  Figure J-16:  Sensitivity of model predicted medial/lateral centroid position to anterior patellar translation.  215   216 -15 -10 -5 0 5 10 15 -4 -3 -2 -1 0 1 2 3 4 distance from imaged position (mm) pr ox im al /d is ta l c oo rd in at e of  c en tr oi d (m m ) Specimen 1 Speciemn 2 Specimen 3 Specimen 4 Distal Proximal Posterior Anterior   Figure J-17:  Sensitivity of model predicted proximal/distal centroid position to anterior patellar translation. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0071575/manifest

Comment

Related Items