An artificial intelligence system for qualified mucosal observation time during colonoscopic withdrawal.
prospective_cohort · Level II
Where this comes from
- Record sourced from PubMed, PMID 41254127.
- Also identified by DOI 10.1038/s41746-025-02067-2 and PMC identifier 12627440.
- Licence recorded as CC BY-NC-ND.
- Because redistribution is not established, this page shows the abstract only. Follow the links below for the full text.
Abstract
Colonoscopic withdrawal time is crucial for achieving a high adenoma detection rate (ADR) and reducing post-colonoscopy colorectal cancer risk. Enhanced qualified mucosal observation improves ADR, but manual quantification of qualified mucosal observation time (QMOT) in routine is challenging. We developed an artificial intelligence (AI) system, QAMaster, for automatic QMOT calculation during colonoscopy withdrawal. QAMaster comprises two models: Model I for image quality analysis (trained with 57,235 images from 64 patients) and Model II for anatomical landmark identification (trained with 7712 images from 3013 patients). Patients were stratified by QMOT, and ADR was compared. The areas under the curve (AUC) of Model I were 0.980-0.991, and Model II were 0.977-0.997. Among 482 patients, ADR was 36.54% (57/156) vs. 19.94% (65/326) in high-QMOT group (≥90 s) vs. low-QMOT group (<90 s) (adjusted OR 2.02; 95% CI 1.23-3.33). QAMaster provides a promising tool for assessing colonoscopy withdrawal quality.