Artificial intelligence-assisted estimation of superior mesenteric artery dissection extent during robotic pancreatoduodenectomy.

BACKGROUND

Superior mesenteric artery (SMA) dissection is an important oncological and technical aspect of pancreatoduodenectomy, but its assessment remains subjective. This study evaluated the feasibility of an artificial intelligence (AI)-assisted approach to estimate SMA dissection extent during robotic pancreatoduodenectomy (RPD).

METHODS

Intraoperative RPD videos (October 2019-July 2025) were used to develop and validate an AI-assisted anatomic recognition model. The model identified SMA-related structures and estimated circumferential SMA exposure as a continuous dissection score. Four expert hepatopancreatobiliary surgeons rated SMA dissection extent using standardized still images, with or without AI overlay, blinded to clinical data. Agreement between AI-derived scores and surgeon ratings, surgeon confidence, and postoperative outcomes were evaluated.

RESULTS

Seventy-eight RPD videos were analyzed, including 4059 annotated images from 49 videos used for model development. AI-derived scores correlated strongly with surgeon ratings (Spearman ρ = 0.767, p < 0.001), with excellent inter-surgeon agreement (ICC = 0.81). Surgeon confidence increased with AI overlay visualization (p < 0.001). Higher AI-derived dissection scores were more frequent in pancreatic ductal adenocarcinoma.

CONCLUSION

AI-assisted analysis generated automated estimates of SMA dissection extent during RPD that correlate with expert surgeon assessment. These findings support surgical video analysis as a tool to approximate technique and enable standardized evaluation of SMA dissection.