Parametric Shape Modeling Using Deformable Superellipses for Segmentation of Medical Image

Medical image segmentation is a challenging task due to speckle noise, poor contrast, and especially absent boundary segments. One popular approach has been the use of deformable models. A parametric shape modeling using deformable superellipses for segmentation of medical image was proposed. Superellipses can represent the smooth and symmetrical contour perfectly, Most of anatomical structures having such characters can be approached by a superellipse. Prior shape information been got from a statistical modal analysis of a training set. This information was used to restrict the transformation of the surperellipse parameters. We used Hybrid Genetic Optimization Algorithm to find the optimal superellipse parameters, and snake algorithm embed with these shape information to segment the image. With shape guidance, this algorithm is less sensitive to initial contour placement and more robust even in the presence of large boundary gaps. The experiments show the efficiency of this method.