Feasibility of a Stool State Check App Using AI During Bowel Preparation Before Colonoscopy: Multicenter Prospective Study (SCAN Study).

Background: Optimal bowel preparation (BP) is crucial for a successful colonoscopy. Although multiple factors influence BP quality, including patient adherence to laxatives and dietary instructions, the stool state during BP should be properly evaluated to perform a colonoscopy of sufficient quality. Therefore, we developed a smartphone app to evaluate a patient's stool state during BP and a viewer to enable real-time monitoring by medical staff. Objective: This study aimed to assess the feasibility of performing colonoscopies of appropriate quality using the app-and-viewer system. Methods: This prospective observational study was conducted between November 2022 and December 2023, involving patients scheduled for colonoscopy at 10 Japanese institutions, comprising 6 tertiary hospitals, 3 regional general hospitals, and 1 community-based clinic. Patients who (1) underwent a colonoscopy at participating institutions, (2) were aged between 20 and 70 years, and (3) owned smartphones compatible with Android or iOS were included in the study. The patients downloaded the app on their smartphones and captured images of their stools during BP, while the medical staff reviewed the evaluation of the stools by the app via the viewer system. The primary end point was defined as the proportion of patients with a Boston Bowel Preparation Scale (BBPS) score of ≥6 among those who successfully used the app. Secondary end points included mean BBPS score, rate of an excellent BBPS score (≥8), adenoma detection rate, cecal intubation rate, and withdrawal time in negative colonoscopy. Additionally, we evaluated the usability of the app, medical staff workload burden with the app, and viewer usage via questionnaire surveys. Results: A total of 343 patients were enrolled, and 326 were ultimately included in the analysis. Overall, 99.1% (323/326, 95% CI 97.3%-99.8%) of the patients achieved the primary end point. The mean BBPS score was 8.5 (SD 1.0), and the proportion of excellent BBPS scores was 87.4% (285/326). The adenoma detection rate, cecal intubation rate, and mean withdrawal time in negative colonoscopy were 46.9% (153/326, 95% CI 41.4%-52.5%), 99.7% (325/326, 95% CI 98.3%-99.9%), and 10.7 (SD 5.9) minutes, respectively. In the questionnaire survey, 98.5% (321/326) of the patients reported that the tutorial was easy to understand, 96.0% (313/326) found stool image capture easy, and 87.8% (286/326) reported reduced anxiety regarding BP. Furthermore, 90.5% (295/326) of the patients indicated that they would like to use the app again for future colonoscopies. Among medical staff, 92.5% (62/67) considered the viewer system necessary, 89.6% (60/67) found it easy to use, and 89.6% (60/67) reported a reduction in workload burden. Conclusions: AI-based stool state assessment using the app and the viewer during BP was feasible across diverse BP methods and clinical environments. Favorable BP outcomes and high usability among patients and medical staff support the potential use of this approach in real-world colonoscopy practice.