Background Suboptimal human - artificial intelligence (AI) interaction is a potential roadblock for implementation of AI in clinical practice. We aimed to evaluate interaction between AI and endoscopists in optical diagnosis of colorectal carcinoma (CRC). Methods International endoscopists were invited to diagnose colorectal lesions online. After a pretest of 15 videos, the diagnosis could be adjusted in the next 35 videos after receiving a simulated AI diagnosis. Histology was the gold standard. Endoscopist's diagnoses before and after AI were compared using the signal detection theory. Inappropriate AI response meant decreased diagnostic performance by the endoscopist after AI. Results In 2730 diagnoses, the 78 participants reached diagnostic accuracies of 71.4% before and 73.7% after AI (p<0.001). Before AI, endoscopists missed only few true CRCs (mean hit rate 0.86), at the cost of a mean false alarm rate of 0.35. Inappropriate AI response was observed in 28.2% of the participants. After AI, most endoscopists (41.0%) shifted towards a more conservative attitude, tending to diagnose a lesion as non-CRC. Conclusions Inappropriate AI response is an issue in human-AI interaction for the optical diagnosis of CRCs. Before using AI, endoscopists could inform themselves on their own and AI's optical diagnostic performance to possibly improve interaction.

Background and Aims Endoscopic and radiologic staging of early rectal cancer (ERC) remains suboptimal, contributing to inaccurate treatment allocation and the risk of under- and overtreatment. This study aimed to evaluate the feasibility and safety of the Pocket-detection method with Real-time Intraprocedural Muscle-retracting sign Evaluation (PRIME) resection algorithm for ERC. Patients and Methods We retrospectively analysed a prospective registry of rectal lesions managed according to the PRIME algorithm between August 2023 and October 2025. MRS status guided adaptive dissection plane selection: endoscopic submucosal dissection (ESD) for MRS-; endoscopic intermuscular dissection (EID) or full-thickness resection (knife-assisted [kFTR] or device-assisted [FTRD]) for MRS+ lesions. Primary outcomes were technical success and R0 rate of endoscopic resection. The secondary outcomes were the technical success of deep margin optical diagnosis and adverse events. Results 47 lesions were included. MRS prevalence was 34.0% (16/47). Technical success of endoscopic resection was 97.9% (46/47, 95% CI, 88.7%-99.9%). ESD was performed in 31 lesions, EID in 9, kFTR in 5 and FTRD in 1. R0 resection was achieved in 89.1% (41/46, 95% CI, 76.4%-96.4%) and R0 vertical margin in 95.7% (44/46, 95% CI, 85.2%-99.5%). Technical success of deep-margin optical diagnosis was 100% (47/47, 95% CI 92.5-100.0). Non-severe (AGREE classification grade I-IIIa) adverse events occurred in 10.6% patients (5/47, 95% CI, 3.5%-23.1%). Conclusion The MRS-guided adaptive endoscopic resection algorithm (PRIME) was feasible and safe for ERC achieving high R0 rates. These findings support validation in larger studies with longer follow-up.