Computer vision applications in vascular surgery: From preoperative mapping to intraoperative flow assessment

Artificial intelligence (AI) and computer vision (CV) are increasingly applied in vascular surgery for preoperative mapping and intraoperative flow/perfusion assessment, with potential to enhance safety and efficiency. However, the extent of their clinical validation and generalizability remains unclear. To synthesize current evidence on AI/CV applications in vascular surgery, focusing on quantitative clinical and technical outcomes, and to identify methodological gaps hindering widespread adoption. A narrative review was conducted by PubMed/MEDLINE, Scopus and complemented with ELICIT, Web of Science, and IEEE Xplore searches. Eligible studies (last 5 years, English/Spanish) included narrative/systematic reviews, clinical trials, observational, and experimental studies with clinical evaluation. Applications considered were preoperative vascular mapping, intraoperative flow/perfusion assessment, and anastomosis assistance/verification. Data extracted included study design, CV/AI technique, sample size, outcomes (diagnostic accuracy, time reduction, radiation/contrast use, complication rates), and performance metrics. Ten studies met inclusion criteria: 7 addressed preoperative mapping, 5 intraoperative flow/perfusion, and 4 anastomosis assistance (some covering multiple domains). Reported benefits included reductions in radiation exposure (up to 48%), contrast use (up to 38%), and operative time (up to 10%), as well as high diagnostic accuracy (median AUROC 0.88) and real-time guidance capabilities. Limitations included predominance of retrospective, single-center designs, scarce external validation (<5% of studies), reliance on single data modalities, and suboptimal adherence to reporting standards. Artificial intelligence and computer vision tools show promise in improving procedural efficiency and safety in vascular surgery, but robust multicenter, prospectively validated studies are needed to confirm their clinical value and enable broad implementation.