Predicting cartilage morphology from bone using a statistical shape model

We present a statistical shape model (SSM) for predicting cartilage morphology from bone, which may enable the rapid generation of cartilage morphology from sparse imaging data. A SSM was trained on 25 segmented MRIs of the knee. The bones of the SSM were fitted to raw segmented data from one test subject outside of the training set and one test subject within the training set to predict the cartilage mesh. Cartilage predictions were reasonable in load bearing areas but poor at the boundaries. Predictions were highly accurate within the training set, suggesting that increased training will improve prediction capabilities.