Prostate cancer segmentation from multiparametric MRI based on fuzzy Bayesian model

Many studies have shown that multiparametric magnetic resonance imaging (MRI), which combines MR spectroscopic imaging (MRSI), T2 weighted MRI, diffusion weighted imaging (DWI) and dynamic contrast enhanced (DCE) MRI, leads to more accurate cancerous tissue localization for prostate cancer patients. However, manual delineation with multiparametric MRI datasets requires a high level of expertise, is a labor-intensive procedure and prone to inter- and intra-observer variability. In this paper, we present an automatic prostate cancer segmentation method based on fuzzy information fusion of multiparametric MRI. In this method, fuzzy c-means clustering (FCM) is first used to obtain fuzzy information related to cancerous tissue shown on each kind of MRI data. Then, an adaptive fuzzy fusion operator based on Bayesian model with a Gibbs penalty term is designed to fuse fuzzy sets obtained by FCM and produces a membership degree map for the region of interest. Based on this map, a decision of cancer regions can be made. In this study, datasets from biopsy-confirmed prostate cancer patients are used to test this method. Experimental results have shown that the proposed method can well localize cancerous regions not only in peripheral zones but also in transition zones of the prostates.