Organ shape plays an important role in various clinical practices such as segmentation. Effective modeling of shape priors is challenging because: (1) shape variation is complex and cannot always be modeled by a parametric probability distribution; (2) a shape instance derived from image appearance cues (input shape) may have gross errors; and (3) local details of the input shape are difﬁcult to preserve if they are not statistically signiﬁcant in the training data. We propose Sparse Shape Composition model (SSC) to deal with these three challenges in a uniﬁed framework. In our method, a sparse set of shapes in the shape repository is selected and composed together to infer/reﬁne an input shape. The a priori information is thus implicitly incorporated on-the-ﬂy. It is formulated as a sparse learning problem, and is extensively validated on several medical applications, including 2D lung localization in X-ray images and 3D liver segmentation in low-dose CT scans.
Project page and code: http://www.research.rutgers.edu/~shaoting/research/siemens2010/project.htm