An elliptical level set method for automatic TRUS prostate image segmentation

TitleAn elliptical level set method for automatic TRUS prostate image segmentation
Publication TypeConference Paper
Year of Publication2006
AuthorsKachouie, N., N. Nezamoddin, P. Fieguth, and S. Rahnamayan
Conference NameIEEE International Symposium on Signal Processing and Information Technology
Keywordsautomatic level set prostate segmentation, automatic TRUS prostate image segmentation, biomedical ultrasonics, cancer, classification method, elliptical level set method, image classification, image segmentation, medical image processing, prostate cancer diagnosis, speckle noise, transrectal ultrasound, velocity function, weak prostate boundaries

One of the most important tasks in prostate cancer diagnosis and treatment is segmentation of transrectal ultrasound (TRUS) prostate images. Due to the large volumes of TRUS prostate images, automatic segmentation systems are mandatory. Weak prostate boundaries, speckle noise and the short range of gray levels make the task more challenging and difficult. Deformable models have been considered as an effective approach for semi-automatic prostate segmentation. However the main problem toward a fully automatic segmentation system using deformable models is initialization of seed or control points. In this paper an automatic level set prostate segmentation is presented. A classification method is employed to locate the approximate location of the prostate which is used to initiate the proposed elliptical level set contour. The deformations of the level set are guided by a velocity function which is derived using the TRUS prostate image histogram