Answers are grounded in this paper and cite their sources.
Background Biliopancreatic endoscopic ultrasound (EUS) is highly operator-dependent, and adherence to standardized examination pathways may vary in routine practice. Evidence on whether real-time artificial intelligence (AI) assistance improves adherence to standardized biliopancreatic EUS examination in real-world settings remains limited. Methods We conducted a retrospective comparative cohort study at a tertiary referral center to evaluate the real-world implementation of a real-time AI assistance system during biliopancreatic EUS. Examinations from the pre-AI period were compared with those from the post-AI period after a washout interval. Propensity score matching and interrupted time-series analysis were used to reduce baseline imbalance and assess temporal changes. The primary outcome was adherence to the standardized examination protocol, operationalized as total documentation completeness across eight predefined anatomical stations. Secondary outcomes included station-level adherence and lesion detection, with prespecified evaluation of pancreatic masses smaller than 2 cm. Results Among 1,465 eligible examinations, 854 matched examinations (427 per group) were analyzed. Adherence to the standardized eight-station examination was higher in the post-AI group, reflected by greater total documentation completeness (89.93% vs 81.29%, adjusted absolute difference 8.65%, 95% confidence interval [CI] 6.58–10.71, P < 0.001). Detection of pancreatic masses smaller than 2 cm was also higher after AI implementation (11.48% vs 6.56%, adjusted odds ratio 1.82, 95% CI 1.13–3.01, P = 0.016). Conclusions In this real-world comparative study, real-time AI assistance was associated with greater adherence to a standardized biliopancreatic EUS examination protocol and a higher observed detection rate of pancreatic masses smaller than 2 cm. These findings suggest that real-time AI assistance may facilitate more standardized biliopancreatic EUS examination in routine practice.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: