Semantic Segmentation of Knee Articular Cartilage

The main objective of this thesis is to identify an automatic knee articular cartilage tissue (ACT) segmentation method using existing deep learning-based methods and develop a novel segmentation framework by employing conventional machine learning techniques as a component to improve segmentation accuracy. The following properties should be integrated into such ACT segmentation framework: (1) implementation of a deep learning model to perform automatic segmentation of knee joint, (2) incorporation of the shape information to capture shape variations of the knees within a population of healthy and pathologic subjects, and (3) extraction of more sophisticated ACT-specific feature sets from the corresponding knee ACT regions.