Automatic Design of Window Operators for the Segmentation of the Prostate Gland in Magnetic Resonance Images

Abstract

W-operators are nonlinear image operators that are translation invariant and locally defined inside a finite spatial window. In this work, we consider the problem of automatic design of W-operators for the segmentation of magnetic resonance (MR) volumes as a problem of classifier design. We propose to segment the objects of interest in an MR volume by classifying each pixel of its slices as either part of the objects of interest or background. The classifiers used here are the artificial feed-forward neural networks. The proposed method is applied to the segmentation of the two main regions of the prostate gland: the peripheral zone and the central gland. Performance evaluation was carried out on the volumes of the Prostate-3T collection of the NCI-ISBI 2013 Challenge. The results obtained show the suitability of our approach as a marker detector of the prostate gland.