Partial sparse shape constrained sector-driven bladder wall segmentation

Bladder wall segmentation from Magnetic Resonance (MR) images plays a crucial role in clinical applications. Level set-based methods are often used to extract the bladder boundaries. When suffering from the fuzzy boundaries, it often results in confused and leaking boundaries. It has been proved that an accurate shape prior can generate an effective force to address these problems. However, the shape prior estimation for the bladder is difficult due to the complex shape variations. Moreover, how to constrain the level set is another challenge. In this paper, we first propose a partial sparse shape model to construct a robust shape prior. Specifically, the partial reliable contour is encoded by the corresponding partial shape dictionary and decoded on the complete shape dictionary to obtain a complete reliable shape prior. Second, we propose a novel sector-driven level set model for locally constraining the evolution to address the problems caused by fuzzy boundaries. Our method was validated on 167 T2 FSE MR images acquired from 15 different patients, better results were obtained compared to the state-of-the-art methods.