IDDF2024-ABS-0053 Artificial intelligence-based gastrointestinal endoscopy quality control system in colonoscopy: a feasibility study

<h3>Background</h3> To evaluate the application of an artificial intelligence-based gastrointestinal endoscopy quality control system in monitoring the quality of colonoscopy. <h3>Methods</h3> A total of 3,974 patients involving colonoscopy procedures between November 23, 2023, and March 10, 2024, were enrolled in our study. Two Endoscopic equipment with Artificial Intelligence (AI) were utilized consecutively for 1987 cases and systematically recorded all colonoscopy case information. A control group was established by selecting a consecutive series of 1987 cases from non-AI-assisted colonoscopies during the same time frame. Quality control data were exported at weeks 1, 4, and 15 from AI device, analyzing biopsy rates, positive diagnostic rates, and pathological diagnosis. The primary outcomes included colonoscopy completion rate, withdrawal time, polyp detection rate, adenoma detection rate, and adequate bowel preparation rate defined as Boston score ≤ 6, with each bowel segment scoring ≥ 2 points. <h3>Results</h3> The colonoscopy completion rate was 92.6% in the AI group and 91.2% in the control group (P&gt;0.05). There was also no statistical significance in the rate of withdrawal time ≥ 6min (91.7% vs 89.7%) and adequate bowel preparation rate (87.5% vs 85.2%) (P&gt;0.05), though those in the AI group were slightly higher than the control group. The AI group detected 85 additional cases of colorectal polyps and 65 additional cases of colorectal adenomas compared to the control group. The polyp detection rate and adenoma detection rate were statistically higher in the AI group (P&lt;0.05). <h3>Conclusions</h3> Endoscopic equipment with Artificial Intelligence (AI) helps quality control and auxiliary diagnosis in colonoscopy. It promptly identifies high-risk lesions and potential tumors, thereby assisting physicians in conducting the colonoscopy in more detail. Moreover, it standardizes colonoscopy procedures of endoscopists by reminding fields of view of blind spots and withdrawal time.