A shape guided C-V model to segment the levator ani muscle in axial magnetic resonance images

Abstract

This paper proposes a modified Chan-Vese model to segment levator ani muscles from axial magnetic resonance (MR) images. Intensity variances of the foreground and the background are used as the main segmentation clues. As in most cases the boundary of the rectum can be successfully segmented in axial MR images, it is assumed as a priori and is used to define the region of interest and the initial contour. In order to handle the complex influences of the connective tissues, a shape influence field is formed based on the shape information of the rectum and is integrated into the Chan-Vese model. Several segmentation examples are presented to show the effectiveness of the proposed method and the necessity of adding the shape influences.