Figure 4

(a) Illustration of ROI definition process: First the first and last intact cortical slice in the imaging plane are found (proximal blue, distal red). These shapes are then interpolated across the defect, providing an estimation of the original periosteum. Rays are then cast from the image boundaries along normals to the major axis of the fragments. Rays which intersect the cortical fragments are classified as being members of FP, while rays which intersects only virtual periosteal surface (red) are classified as being in DP. The DC volume is then determined as the space not occupied by FC, FP or DP. (b) The pipeline for FE analysis: initially the load on the femur was determined using load estimation on intact contralateral femurs. These loads were then applied to a coarse FE-model which determined the boundary conditions for the in vivo micro-CT model. (c) The association between mineralised tissue formation and mechanical strain in the soft tissue, micro-FE analysis is used to calculate the strains in the soft tissue, using the time-lapsed in vivo micro-CT images as ground truth ROC analysis is performed determining how effective mechanical strain is as a predictor. (d) Association of remodelling events with mechanical strain. Using the micro-FE results for the mineralised tissue a set of thresholds are swept through the strain space classifying formation and resorption events. Using the time-lapsed in vivo micro-CT images as ground truth, a correct classification rate for each set of thresholds is determined.