Bulten W, Pinckaers H, van Boven H, et al. - Researchers undertook this retrospective study to test the utility of deep learning to perform automated Gleason grading of prostate biopsies. From patients at the Radboud University Medical Center, randomly picked biopsies, sampled by the biopsy Gleason score, were used to build a deep-learning system to grade prostate biopsies after the Gleason grading standard. The purpose of constructing this system was to know about individual glands, assign Gleason growth patterns, and ascertain the biopsy-level grade. A total of 5,759 biopsies were obtained from 1,243 patients. A high agreement with the reference standard was afforded by the developed system, and it scored highly at clinical decision thresholds: benign vs malignant, grade group of 2 or more, and grade group of 3 or more. Overall, compared with pathologists, a similar performance for Gleason grading was achieved with the automated deep-learning system developed by the experts, and the possible potential contribution of this system to prostate cancer diagnosis was also suggested. The potential applications of this system could be assisting pathologists by screening biopsies, enabling second opinions on grade group, and presenting quantitative measurements of volume percentages.