Automated Principal Component Analysis of cartilage thickness, T2 times and bone shape as interpretable model for osteoarthritis progression

In this study we used fully automatic segmentations of the bones and the cartilage of the knee to perform a principal component analysis of cartilage thickness, T2-relaxation times and bone shapes in the Osteoarthritis Initiative Dataset. We extracted and visualized the principal components which significantly contributed to a model for prediction of osteoarthritis progression to (i) show that besides mean values of these parameters also their distribution is crucial for osteoarthritis progression and (ii) prove the feasibility of visual interpretation of these components for better understanding factors associated with the disease.