3-D heart segmentation on CT images using superellipsoid model fitting

Model based segmentation is one of the best methods for segmentation of the heart in CT and MRI images. Because of basic characteristics of the heart shape, it can be represented by such models like ellipse and superellipse. Among model based methods elliptic and superelliptic methods are more computationally expensive. But when prior knowledge of the heart shape is available the parameter estimations become reliable. In this paper we propose a 3-D semi-automatic heart segmentation using transverse axis slices. The approach consists of fitting superellipse to each slice and making a 3-D superellipsoid model of the heart. We used an iterative method over a set of given data as partial data. Moreover partial data for obtaining prior knowledge of the heart shape are used that makes our method computationally less expensive. Results show that the addition of partial data increase robustness of a fitting. Also this method can segment hidden parts of heart in transverse axis CT images.