Segmentation of female pelvic organs in axial magnetic resonance images using coupled geometric deformable models

Abstract

The segmentation of pelvic structures in magnetic resonance (MR) images of the female pelvic cavity is a challenging task. This paper proposes the use of three novel geometric deformable models to segment the bladder, vagina and rectum in axial MR images. The different imaging appearances and prior shape knowledge are combined into a level set framework as segmentation cues. The movements of the contours are coupled with each other based on interactive information, and the organ boundaries can be segmented simultaneously. With the region-based external forces defined, the proposed algorithms are robust against noise and partial volume effect.