Novel stochastic framework for accurate segmentation of prostate in dynamic contrast enhanced MRI

Prostate segmentation is an essential step in developing any non-invasive Computer-Assisted Diagnostic (CAD) system for the early diagnosis of prostate cancer using Dynamic Contrast Enhancement Magnetic Resonance Images (DCE-MRI). In this paper we propose a novel approach for segmenting the prostate region from DCE-MRI based on using a graph cut framework to optimize a new energy function consists of three descriptors: (i) 1^st -order visual appearance descriptors of the DCE-MRI; (ii) a spatially invariant 2^nd -order homogeneity descriptor, and (iii) a prostate shape descriptor. The shape prior is learned from a subset of co-aligned training images. The visual appearances are described with marginal gray level distributions obtained by separating their mixture over the image. The spatial interactions between the prostate pixels are modeled by a 2^nd -order translation and rotation invariant Markov-Gibbs random field of object / background labels with analytically estimated potentials. Experiments with prostate DCE-MR images confirm robustness and accuracy of the proposed approach.