Automatic Segmentation of Prostate from Multiparametric MR Images Using Hidden Features and Deformable Model

This paper proposes a hidden feature learning technique to segment prostate from Multiparametric Magnetic Resonance Imaging. In image guided intervention an automatic segmentation of prostate is an essential task. The automated segmentation of prostate is challenging process due to blurred prostate boundaries and extensive variations in prostate shape among the subject population. To defeat these challenges, we used image patches to construct probabilistic map. Atlas based segmentation is used to minimize an energy function to label the patches as the prostate or background. The latent features are determined from prostate MR image by using sparse auto encoder. Deformable model segmentation is deployed to perform final segmentation by integrating sparse patch matching method. The performance of presented method is enormously tested on the dataset that contains T2-weighted prostate MR images of 184 subjects. The performance indices like dice similarity coefficient (DSC), and mean absolute surface distance (MASD) are used to evaluate the performance of our method by considering manual groundtruth delineated by experienced radiologist. DSC obtained by our algorithm is 87.4% +/- 4.6%, and MASD of 1.80 mm. The preliminary results demonstrate that the learned features are more competent in MR prostate segmentation.