Introduction Withdrawal time (WT) has emerged as a critical quality measure in colonoscopy for colorectal cancer screening. Due to the high variability in calculating WT, recent works have explored the use of artificial intelligence (AI) to standardize this process, but prospective validation remains elusive. Methods This prospective, superiority trial compared the accuracy of AI-assisted WT calculation to that of physicians during routine colonoscopy from December 2023 to March 2024. The gold standard was obtained via manual, frame-by-frame annotation of the examination video recordings. The AI also automatically generated an image report, which was qualitatively assessed by four endoscopists. ClinicalTrials.gov NCT06094270 Results In total, 126 patients recruited from December 2023 to March 2024 were analyzed. The proposed AI demonstrated a significantly lower mean absolute error (MAE) (2.2 minutes) in estimating WT compared to physicians (4.2 minutes; p<0.001). This is attributed to examinations containing endoscopic interventions where the AI had significantly lower MAE compared to physicians (2.1 vs 5.2; p<0.001). The MAE was comparable in the absence of interventions (2.3 vs 2.3; p=0.519). High-quality image reports were generated by the AI, with 97% of assessments indicating satisfaction with the timeline representation and 81% expressing overall satisfaction. Conclusion Our study demonstrates the superiority of an AI system in calculating WT during colonoscopy compared to physicians, providing significant improvements, especially in examinations containing interventions. This work demonstrates the promise of AI in streamlining clinical workflows.
Kafetzis et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: