Relaxed Conditional Statistical Shape Models and Their Application to Non-contrast Liver Segmentation

This paper proposes a novel conditional statistical shape model (SSM) that allows a relaxed conditional term. The method is based on the selection formula and allows a seamless transition between the non-conditional SSM and the conventional conditional SSM. Unlike a conventional conditional SSM, the relaxed conditional SSM can take the reliability of the condition into account. Organ shapes estimated by the proposed SSM were used as shape priors for Graph Cut based segmentation. Results for liver shape estimation and subsequent liver segmentation show the benefit of the proposed model over conventional conditional SSMs.