1. Combining the strengths of UMIST and The Victoria University of Manchester Knees Project Meeting 28/11/2005 Tomos G. Williams Imaging Science and Biomedical Engineering University of Manchester

2. Combining the strengths of UMIST and The Victoria University of Manchester Agenda 16:00 – 16:15Project review 16:15 – 17:15Publications –Methodology paper illustrated with CP77 for TMI: Final review. –OARSI poster: Review draft poster. –ISBI abstract: Status, issues, timelines. –CP78 paper(s): Plans following CP78 investigators meeting. 17:15 – 17:20Issues/tasks outstanding. 17:20 – 17:30Current plans –TGW: Plans 1H06. –RM: AZ OA imaging situation. –CJT/TGW: Other ISBE OA activities. –iMorphics: Current interests/plans. 17:30 – 18:00Next steps –Brainstorm possible next steps: What? Who? Where? How? –Building on discussion at 31-AUG-05 Joints Imaging Group. –Possible exploitation paths? –Potential projects of interest for AZ EI-GPT strategic review of OA imaging

3. Combining the strengths of UMIST and The Victoria University of Manchester Project Review General project aims –To develop methodology and tools for focal analysis of bone and articular cartilage quantity and quality in MR images of the knee. –To apply these tools to the previously acquired CP77 (normal) and CP78 (disease progression) data sets. –To deliver tools to enable future exploitation of these techniques in AstraZeneca clinical study programmes.

4. Combining the strengths of UMIST and The Victoria University of Manchester Achievements Triangulated surface representation from parallel slice segmentations Application and development of MDL automatic model building algorithm Bone-referenced, corresponded cartilage thickness maps –Normal range (CP77) –OA Disease Progression (CP78) Application of Active Shape Models to automatic segmentation of bones Identified cartilage loss in sub-regions of the joint –Small cohort, short time-period –Imperfect cartilage segmentations

5. Combining the strengths of UMIST and The Victoria University of Manchester VCB Bias Experiment %CP77 VC - TOSCA CP77 VC - VCB CP78 VC - VCB Femur-3.693.375.91 Patella-3.803.616.26 Lat Tibia-4.032.184.38 Med Tibia-5.342.786.23 Cartilage surface constructed from mean bone shape –Average inner and outer thicknesses Cartilage Volume, VC, measured using Gauss’ theorem VCB measured for mean bone shapes with decreasing number of vertices Mean bone shape with constructed cartilage (8198 faces). VC = 9261

6. Combining the strengths of UMIST and The Victoria University of Manchester VCB Bias Experiment Results Sampling Points VC-VCB (%) 25824.40 10267.29 40982.98 163860.28 Cartilage under-sampled at similar ratio to actual surfaces results in similar underestimation. Reduced tessellated mean shape (258 points) with cartilage. Thickness map and facet volumes. VCB: 7247 VC - VCB: 2014 = 24.40%

7. Combining the strengths of UMIST and The Victoria University of Manchester TMI Figures “The current cost for color reproduction is a flat printing fee of US$1,045.00 plus US$125.00 per piece of color artwork.”

8. Combining the strengths of UMIST and The Victoria University of Manchester CP78 Results Summary VCB AV (ml)ThCtAB Av (mm) Comp Mean StdDevP-value Mean StdDevP-value F-0.2740.6140.019-0.0340.0950.056 Trm F-0.214 0.3440.002-0.0560.1080.008 Trm TrF-0.035 0.1660.245-0.0220.1060.263 Trm LF-0.059 0.2290.164-0.0560.2410.263 Trm MF-0.120 0.2050.015-0.1150.2100.015 Trm P-0.002 0.1670.951-0.0100.2040.793 Trm LT-0.002 0.1280.946-0.0060.1730.853 Trm MT-0.040 0.1740.214-0.0480.2380.267

9. Combining the strengths of UMIST and The Victoria University of Manchester CP78 Analyses Subgroup –Medial vs. Lateral –Upper and lower tertiles Correlation with soluble biomarkers

10. Combining the strengths of UMIST and The Victoria University of Manchester Issues and Tasks Outstanding Assess (non-normal) distribution of cartilage thickness (across subjects), and possible transformations to approximate normality and/or stabilise variance. Investigate differences between the CP78 (Diseased) and CP77 (Normal) bone shapes. Assess feasibility of identifying points corresponding to CP77 (normal) bone surface model on subsequent bone segmentation. (Pre-existing bone segmentations.) Proposed analyses Abnormality and lesion proximity hypotheses Cartilage signal intensity mapping Data and Technology Transfer

11. Combining the strengths of UMIST and The Victoria University of Manchester CP78 Thickness Analysis Test What?Large ROI e.g. Femoral Cartilage Small Anatomical ROI e.g. Central Medial Tibial Plateau Statistical Approach Volume Analysis 1 Analysis 2 (ROI specification tool) Univariate tests “Has the cartilage volume changed?” Overall Thickness Map Analysis 4 (work in progress) Analysis 3Multivariate Tests “Has the pattern of cartilage thickness changed? Thickness at each Location Analysis 5Analysis 6Multiple Comparison Tests “Where has the thickness changed?”

12. Combining the strengths of UMIST and The Victoria University of Manchester Disease Progression Hypotheses 1.Anatomical –Simultaneous analysis of thickness change at each and every correspondence point. 2.Abnormality –Assess thickness change in regions where cartilage is thin or thick at baseline (in comparison with CP77). 3.Lesion proximity –Assess thickness changes in regions surrounding a lesion in the cartilage at baseline. 4.Opposition –Assess thickness changes at locations which articulate with regions that are abnormal at baseline.

13. Combining the strengths of UMIST and The Victoria University of Manchester Immediate/Future Work-plan Publications –TMI to Publication –CP78 analyses Outstanding issues –Difference between CP77 and CP78 bone shapes –Original analysis –Abnormal and lesion proximity analysis –Difference maps multivariate analysis –Data and technology transfer Application to OAI Expert defined regions of interest Improved MDL optimisation –Adoption of revised code Steve Gandy’s aged diurnal study Smooth surface representation Automatic cartilage segmentation Charles Hutchinson –Multicentre comparison of OA Initiative protocol –Per-patient analysis Imorphics Suggestions –Utilise bone model –Single T1 scan analysis