A novel method to automatically segment the pelvic floor muscles in MR images

Abstract

A method is proposed to segment the levator ani muscles in the axial view of MR images of pelvic cavity. The proposed algorithm is based on the main ideas of Chan-Veses model. A computational region and an initial contour are defined through the boundary of rectum. Intensity variances of the regions inside and outside the moving contour are used as the main external forces. The average intensity of the region inside the moving contour in the C-V model is replaced by the priori intensity of levator ani to guide the segmentation. Additionally, a shape influence field is proposed to handle the problem caused by the similar appearances of levator ani muscles and the fiber tissues. Experiments are illustrated to show the necessity of adding the initial influences and the effects of the proposed method.