Semi-Automatic Prostate Segmentation of MR Images Based on Flow Orientation
Maryam Samiee, Gabriel Thomas, Reza Fazel-Rezai · 2006
Prostate cancer is one of the leading causes of death in men. Accurate segmentation of prostate magnetic resonance imagery allows for the maximum volume of the prostate to be considered in diagnosis, using magnetic resonance spectroscopy, and in treatment using intensity modulated radiotherapy. In this work a semi-automatic method which segments the prostate on magnetic resonance images is presented. This algorithm evaluates the curve of the prostate in an orientation based frame work. Having computed the edge direction of each individual pixel in the region surrounding the prostate, and considering the location of four points of this region that have been previously selected by the user, a statistical average mask of size 3times3 follows the direction of a channel of low intensity pixels around the prostate using prior knowledge of the shape of the object. The initial results show promising results when comparing this method to the segmentation done by an expert