Medical Image Segmentation Based on Graph Cut with Shape Prior

Abstract

This paper deals with a graph-based image segmentation and its improvement by using the information about the shape of the object for creating specific graph architecture (template). Improved method allows the cut to prefer more complicated structures, especially when the image contains a lot of noise and the object is hardly indistinguishable from the background. Algorithm was tested on simulated data and real CT and MRI images of vertebra and brain in 2D. Method was also extended to 3D further purposes.