Document resource
Background Adequate bowel preparation is essential to ensure diagnostic accuracy and safety. We developed an artificial intelligence (AI)-driven smartphone app, ‘BPEasy’, which allows patients to evaluate their bowel preparation status and receive suggestions. A randomized controlled trial was conducted to assess the app’s effectiveness.Methods We collected stool images to develop the AI model and verified it in internal validation. The AI-based prediction model was integrated into a smartphone app. A randomized controlled study was conducted from June 2024 to May 2025, which was registered at the Chinese Clinical Trial Registry (ChiCTR2300071917). Outpatients aged 18-75 years old were chosen and randomized into control and AI-driven app groups.Results A total of 500 patients were enrolled; 39 patients cancelled their colonoscopy. Ultimately, 461 patients (229 in the control group and 232 in the app group) participated in the study. The rate of adequate bowel preparation (97% VS 90%, P = 0.002) and excellent bowel preparation (86.2% vs. 59.8%, P < 0.001) was significantly higher in the app group. The mean BBPS score (8.2 ±0.8 VS 7.8 ±1.2, P < 0.001) and BBPS scores for each colon segment were also higher in the AI-driven app group. No significant differences were observed in cecal intubation rate, PDR, and ADR. Patient satisfaction was high, with 95.5% of app users willing to recommend it to others.Conclusions The BPEasy app significantly improved bowel preparation quality for colonoscopy, enhancing diagnostic accuracy and patient satisfaction. Future multicenter studies and cost-effectiveness analyses are warranted to further validate its clinical utility.