AI ‐Assisted Recognition of Actual Ureteral and Pancreatic Injuries in Laparoscopic Colorectal Surgery: A Randomized Video Review Study

Le résumé fourni par la source

ABSTRACT Aim Artificial Intelligence (AI)‐assisted anatomical recognition may help surgeons recognize intraoperative adverse events missed during surgery. We evaluated whether AI‐enhanced videos improved recognition of ureteral and pancreatic injuries. Methods In this two‐institution randomized video‐review study using historical laparoscopic colorectal surgery videos, 150 surgeons reviewed either EUREKA‐generated AI‐enhanced videos or unenhanced videos. Five videos were reviewed: three actual injury videos, including unrecognized injuries, and two clinically comparable no‐injury control videos. The primary outcome was correct injured‐organ recognition. Secondary outcomes included recognition within the predefined injury time‐code window and correct no‐adverse‐event judgment. Results Seventy‐five surgeons were assigned to each group, with no significant baseline differences. AI‐enhanced videos improved ureteral injury recognition in Video A (40.0% vs. 22.7%; p = 0.022) and Video C (50.7% vs. 28.0%; p < 0.01). Time‐code window recognition was greater for Video A (20.0% vs. 5.3%; p < 0.01) and Video C (18.7% vs. 4.0%; p < 0.01). Pancreatic injury recognition in Video D did not differ overall (77.3% vs. 70.7%; p = 0.352) but was greater among surgeons without advanced laparoscopic certification (81.3% vs. 63.3%; p = 0.048). Correct no‐adverse‐event judgments did not differ between groups. Conclusion AI‐generated anatomical highlighting improved surgeons' recognition of actual ureteral injuries under controlled video‐review conditions without increasing incorrect adverse‐event judgments. These findings provide proof‐of‐concept evidence that AI‐generated overlays may support timely recognition of clinically important intraoperative adverse events. Prospective intraoperative studies are needed to determine whether this recognition benefit facilitates earlier management, changes operative decision‐making, reduces complications, or improves patient outcomes. Trial Registration UMIN‐CTR; UMIN ID: UMIN000059439

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.