Colorectal mucosal exposure area assessment based on artificial intelligence: a multi-center, prospective, observational study

Background: This study proposed a new quality control indicator, cumulative colorectal mucosal exposure area (CCMEA) for colonoscopy to assess mucosal exposure, constructed CCMEA system based on deep learning. and validated the indicator in a multi-center prospective observational study. Methods: The CCMEA system worked based on ResNet50 and UNet++. Determine the CCMEA threshold for qualified colonoscopies based on international colonoscopy quality criteria, the 25% adenoma detection rate (ADR). A multi-center prospective observational study was conducted to evaluate the system and threshold in clinical practice. Based on the threshold, patients were divided into the qualified and unqualified group, with ADR and other lesion detection rates were compared between the two groups. Results: A total of 510 participants underwent colonoscopy were evaluated, with grouping into qualified (n=270) and unqualified (n=240) groups based on the CCMEA qualification threshold of 2000.. ADR was 39.53 percentage points higher in the qualified group than in the unqualified group (53.70% vs 14.17%, aOR=7.97, 95%CI 4.97-12.80, p<0.001), especially for lesions ≤5 mm (42.22% vs. 10.00%, aOR=6.85, 95%CI 4.09-11.46, p<0.001). The qualified group also had a significantly greater polyp detection rate (89.63% vs. 40.00%, aOR=13.05, 95%CI 7.81-21.82, p<0.001) and higher mean number of both adenomas (1.00 vs 0.17, aIRR=5.94, 95%CI 4.32-8.36, p<0.001) and polyps (5.79 vs 1.32, aIRR=3.99, 95%CI 3.52-4.52, p<0.001). Conclusions: The CCMEA qualified group, based on a CCMEA threshold of 2000, shows higher ADR than the unqualified group, indicating CCMEA can be a promising colonoscopy quality indicator.

Publikationsverlauf

Eingereicht: 24. März 2025

Angenommen nach Revision: 29. August 2025

Accepted Manuscript online:
03. September 2025