Automatic Real-time Segmentation of the Prostate from Magnetic Resonance and Trans-rectal Ultrasound Images using Deep Learning

This thesis applies novel deep learning techniques to automatically segment the prostate and its interior zones from magnetic resonance imaging (MRI) and trans-rectal ultrasound (TRUS) images to a high level of accuracy. This research will contribute to the development of 3D/augmented reality-guided tools for more accurate prostate biopsies and treatment. The general contributions of the thesis are twofold: an improvement in the semantic segmentation of objects in images, and its application to prostate MRI and TRUS images.