Introduction: The manual creation of operative reports is a tedious documentation task that increases administrative burden among surgeons, which is a primary driver of physician burnout. Using pre-templated notes may reduce this burden but can lead to documentation inaccuracies. Additionally, surgical operative reports are subjective and lack transparency. We aim to develop an artificial intelligence (AI) tool for automatically generating operative reports based on surgical video alone. Methods: A previously developed computer vision AI algorithm was employed to automatically detect major steps of robotic radical prostatectomy, including pelvic lymph node dissection, Space of Retzius dissection, anterior bladder neck transection, posterior bladder neck transection, seminal vesicle and posterior dissection, lateral/pedicle and apical dissection, urethral transection, vesicourethral anastomosis, and final inspection/extraction. Each step was mapped to pre-specified text, which was then compiled into a narrative operative report based on AI recognition of surgical steps. Accuracy of the AI-generated operative reports was assessed by comparing to operative reports documented in the medical record (human). All discrepancies between AI and human operative reports were adjudicated by independent video review performed by a fellowship-trained urologic oncologist. Results: A total of 117 cases from a single tertiary referral center were included. There was concordance between human and AI operative reports in 107 cases, suggesting that the AI reproduces major components of the human operative report with 91.5% accuracy. Discrepancies between human and AI operative reports were identified in 10 cases, of which 9 were clinically significant. These included 8 discrepancies in lymph node dissection and 1 discrepancy in anterior bladder neck transection. Upon expert video review, the human was inaccurate in 3 discrepancies (2.6%), while the AI was inaccurate in 6 discrepancies (5.1%). Conclusions: To our knowledge, this is the first report of AI-powered automated creation of operative reports, which achieve high accuracy as compared to human operative reports for major surgical steps. This novel tool has potential to reduce documentation burden, improve operative report accuracy, promote surgical transparency, and decrease subjectivity in surgical documentation. SOURCE OF Funding: None
