Artificial Intelligence‐Based Recognition of the Prostatic Capsule During Nerve‐Sparing Robot‐Assisted Radical Prostatectomy

OBJECTIVES: Artificial intelligence has diverse applications, especially in perioperative surgery. Intraoperative recognition of anatomical structures affects surgical decision-making, surgical progress, and outcomes. Nerve-sparing robot-assisted radical prostatectomy is performed to preserve postoperative function; however, recognition of the layers that need to be preserved still depends largely on surgeon experience. Therefore, we developed a convolutional neural network-based deep-learning model to assist in recognition of the prostatic capsule. METHODS: Videos of 18 nerve-sparing robot-assisted radical prostatectomy procedures performed between 2023 and 2025 were used; 13 cases were assigned to the training set and five to the evaluation set. Specimens from our institution were confirmed by pathological evaluation using the nerve-sparing classification, whereas external cases were evaluated solely by video review. Still images were manually annotated for model training, and frames randomly extracted from five independent evaluation videos were used for quantitative evaluation. Seven board-certified urologists experienced in robotic surgery performed qualitative assessment using structured questionnaires. RESULTS: The median Dice score and Intersection over Union were 0.621 (interquartile range: 0.399-0.687) and 0.451 (interquartile range: 0.250-0.524), respectively. For assisting with nerve-sparing layer recognition, the mean concordance and effectiveness score were 3.6 and 3.1 (out of 5), respectively, in the qualitative evaluation of the still images. In the video assessment, the mean surgeon-reported stress level and effectiveness score were 3.6 and 2.9 (out of 5), respectively. CONCLUSIONS: The artificial intelligence-based model aided in prostatic capsule recognition; this may make the nerve-sparing technique safer and could be associated with better postoperative outcomes.