Prostate biomedical images segmentation and classification by using U-NET CNN model

Prostate cancer is one of the most widespread types of cancer among men. The successful treatment of prostate cancer is based on accurate diagnosis. Gleason grading patterns system is one of the most efficient methods in diagnosing histological biopsies of prostate images by pathologists. Automatic detection and segmentation of the prostate on histological Gleason grading system is still the most powerful prognostic tool. In this paper, we propose a powerful deep convolutional neural network (CNN) technique called U-Net module to predict the prostate Gleason score based on tissue microarray (TMA) images. We developed a U-Net model for object semantic segmentation, where the goal is to precisely label each pixel in an image as being part of a given object (foreground) or not (background). Our proposed U-Net model of prostate segmentation achieved a mean test accuracy of 96%. The model achieved a mean Dice index coefficient (DI) of 0.56 and a mean IOU of 0.95 that show how close the output segments are to the corresponding lesions in the ground truth maps.