AI-based automated bleeding monitoring in conventional and robot-assisted laparoscopic surgery: a systematic review

Artificial intelligence has emerged as a promising approach for improving the detection and management of intraoperative bleeding during conventional and robotic-assisted laparoscopic surgery, where delayed recognition of hemorrhage can lead to increased morbidity and procedural complexity. This review synthesizes current evidence on the use of artificial intelligence for intraoperative bleeding monitoring, with a particular focus on performance, feasibility, and clinical integration. A systematic review was conducted in accordance with PRISMA 2020 guidelines. Comprehensive searches of PubMed, Scopus, Web of Science, IEEE Xplore, Embase, and grey literature identified studies published between 2016 and 2025 that applied artificial intelligence to bleeding prediction, detection, localization, tracking, and quantitative blood-loss estimation during conventional and robotic-assisted laparoscopic surgery. Data relating to study design, model architectures, evaluation metrics, latency, and integration feasibility were extracted and summarized narratively. Across the included studies, artificial intelligence models demonstrated high detection accuracy in predominantly single-centre, retrospective, or simulation-based settings, with several approaches reporting promising real-time feasibility under controlled experimental conditions. Emerging work also explored bleeding source tracking, blood-loss estimation, and early integration into surgical workflows. However, most studies relied on small, single-center datasets and retrospective validation, limiting generalizability and clinical translation. Overall, artificial intelligence–based bleeding monitoring in conventional and robotic-assisted laparoscopic surgery shows increasing technical maturity and potential, though largely unvalidated in prospective clinical settings. Future research should prioritize large, multi-institutional datasets, prospective clinical evaluation, and optimized low-latency deployment within real surgical workflows to support safe and effective intraoperative use.