An Energy-based Algorithm for Automatic Prostate Segmentation in TRUS Images

Abstract

Determining prostate’s boundary in TRUS (Transrectal Ultra Sound) images is very necessary (so essential) in lots of treatment methods of prostate cancer. The presence of strong speckle noise, Weakness edges and shadow artifacts limits the effectiveness of classical segmentation schemes. In this paper a new energy-based method for automatic prostate segmentation in TRUS images is presented. This method involves three main stages. These stages are as follows, preprocessing step (edge preserving noise reduction and smoothing), inside point finding step and prostate segmentation respectively. In the first stage, Speckle reduction achieved by using stick filter and top-hat transform has been implemented for smoothing. A feed forward neural network has been use in second stage to find a point inside prostate object. Finally using inside point of achieved of before stage and active contour algorithm, extracts boundaries of prostate in last stage. A number of experiments are conducted to validate this method. One of the characteristics of this method is to detect the prostate boundary with Normalized Area Error (NAE) of lower than 4.8%.