Effectiveness of artificial intelligence's system ArtInCol in diagnostic of colorectal neoplasia during colonoscopy: results of multicenter randomised clinical trial

AIM: to evaluate the effectiveness of the Russian artificial intelligence system ArtInCol during routine colonoscopy. PATIENTS AND METHODS: from August to December 2024 a multicenter randomized trial was done and included 4 medical institutions and 1,128 patients. The patients were randomized into colonoscopy groups without AI (n = 547) and colonoscopy group using the ArtInCol artificial intelligence system (n = 581). The data was analyzed according to the “intention-to-treat” and «per protocol» types, with the primary endpoint being the frequency of detection of adenomas. RESULTS: the randomized groups were homogenous in all analyzed variables. When comparing the primary end-point, the detection rate of adenomas (ADR) in the studied group of AI-assisted colonoscopy was 47.2% (95% CI: 43.1–51.2), compared with 41.3% (95% CI: 37.3–45.5) without AI, the effect value was 5.9%, p = 0.048. The average number of detected adenomas was 0.97 (95% CI: 0.85–1.09), versus 0.79 (95% CI: 0.67–0.92) in the control group, which is a statistically significant difference (p = 0.01). CONCLUSION: the study confirm the hypothesis of the effectiveness of the AI — ArtInCol system in order to improve the quality of neoplasm detection during colonoscopy. An increase in the detection rate of adenomas by 5.9% was recorded.