Man Plus Machine: A Randomized Control Trial of Artificial Intelligence Including the Impact of Adjunctive Polyp Detection Techniques

Abstract

BACKGROUND AND AIMS: Computer-aided polyp detection tools (CADe) utilizing artificial intelligence (AI) have been shown to demonstrate benefit with improved polyp detection during colonoscopy. Questions remain around the impact of CADe when combined with additional techniques that improve polyp detection such as lengthening withdrawal time, cecal and rectal retroflexion, dynamic position change, and narrow band imaging (NBI) use. METHODS: A single-center prospective, randomized control trial comparing ENDO-AID AI module to conventional colonoscopy was conducted between October 11, 2023 and March 16, 2024 at Waitakere Hospital. Additional techniques to improve polyp detection were recorded but left to the discretion of participating 26 endoscopists. RESULTS: Seven hundred seventy-six patients (mean ± SD age, 61.2 ± 13.0 years; 344 females) were recruited, 383 patients allocated to AI and 393 to control. Position change was used in 43%, antispasmodic in 25%, distal cap in 25%, NBI in 21%. Overall, univariate analysis demonstrated a nonsignificant trend towards higher adenoma detection rate (ADR) in the CADe than control group (63.4% vs. 57.3%, p = 0.08). AI was most effective in the screening cohort (ADR 79% vs. 68%, average polyp rate 3.7 vs. 2.8 p < 0.05). Multivariable analysis demonstrated CADe was independently associated with increased adenoma detection rate (odds ratio [OR], 1.38; 95% confidence interval [CI], 1.01-1.89; p = 0.042), as was use of NBI, OR 2.00; (95% CI: 1.23-3.25; p = 0.006) and increased withdrawal time, OR 1.11; (95% CI: 1.08-1.15; p < 0.001). CONCLUSION: ADR was increased by CADe in a cohort of high detectors and was further augmented by traditional techniques known to be beneficial. It is important to incorporate traditional techniques with CADe to maximize ADR.

Ähnliche Arbeiten