Mylène Jansen

CT subchondral bone changes after KJD 381 18 Afterwards, wxRegSurf v18 (Cambridge University Engineering Department, Cambridge, UK, in-house developed software freely available at http://mi.eng.cam.ac.uk/~ahg/wxRegSurf/ ) was used for registration of all femur and tibia surfaces to a canonical femur and tibia respectively, to allow combining and comparing of surface objects from multiple scans. The vertex by vertex displacement data to the canonical surfaces of each individual scan was saved, and used to visually explore the bone shape changes between time points. Figure 1 : Example segmentation from 1 CT slice. (A) Axial CT slice; (B) thresholding of bone (pink); (C) final semi-automatic segmentation of this slice (yellow line), excluding osteophytes. Only patients for whom baseline and at least 1 of the 2 follow-up time points available were included for analysis. Since KJD has previously shown significant results mostly in the patients’ most affected compartment (MAC), patients were separated into 2 groups based predominantly medial compartmental OA and predominantly lateral compartmental OA, defining their most affected compartment for all analyses. Statistical analyses MATLAB R2020a and the SurfStat MATLAB package (https://www.math.mcgill.ca/keith/ surfstat/, optimized for this specific application by Graham Treece of the University of Cambridge) were used for whole-bone, vertex-wise data analysis and visualization. Average cortical bone thickness and trabecular density were displayed for each time point separately, by averaging data of all available patients at each time point. Statistical parametric mapping (SPM) was used for statistical analysis, which uses all subjects’ values at each vertex for statistical testing and delivers vertex-wise p- value corrections for multiple comparisons at a set corrected p- value threshold. 16 SPM with 2-tailed F -tests were used to calculate changes over time against a null hypothesis of no change. 18 In all cases, a p- value <0.05 was considered statistically significant. Although measurement and analysis of the bony parameters are performed for the whole bone surfaces, in this study we focus attention on the subchondral cortical bone plate and trabecular density.