Use of an AI Image Recognition-Based Smartphone App in Personalized Bowel Preparation Before Colonoscopy: Prospective Randomized Controlled Trial.

Abstract

Background: Sufficient bowel preparation is critical for increasing the quality of colonoscopy. However, current bowel preparation guidelines have limitations. We used a constructed and validated convolutional neural network model to assist patients in assessing the adequacy of bowel preparation and guide the use of laxatives based on individual variations.

Objective: This research intended to explore the influence of AI-guided bowel preparation on patients' use of laxatives and bowel cleansing quality.

Methods: We conducted a prospective randomized controlled trial that included patients who had a colonoscopy at the First Affiliated Hospital of Zhengzhou University and used sodium phosphate solution for bowel cleansing. Included individuals were randomly categorized into 2 groups: the AI-guided group, which used an AI program to assess the adequacy of bowel cleansing and stopped the use of laxatives once the assessment was passed, and the control group, which evaluated bowel preparation adequacy based on standard practices and used the recommended doses of laxatives. The primary outcomes were the amount of laxatives used and the rate of adequate bowel cleansing based on the Boston Bowel Preparation Scale (BBPS). Other outcomes included overall BBPS scores, BBPS scores by colon region, adenoma detection rate (ADR), polyp detection rate, and advanced ADR, as well as adverse reactions and compliance.

Results: This study included 513 patients (AI-guided group: n=255, 49.7%; control group: n=258, 50.3%). The volume of bowel-cleansing solution consumed by patients in the AI-guided group was 1245.5 (SD 377.2) mL, significantly lower than the 1586.2 (SD 50.4) mL in the control group (P<.001). Reduced use of laxatives was independently predicted by younger age, lower BMI, and higher defecation frequency. The rate of adequate bowel preparation in the AI group was not inferior to that in the control group (238/255, 93.3% vs 229/258, 88.8%, respectively; 1-sided P<.001 for noninferiority). No significant differences were observed in segmental or total BBPS scores, ADR, polyp detection rate, and advanced ADR between the 2 groups. Additionally, the medication adherence rate in the AI-guided group was significantly higher than in the control group (247/255, 96.9% vs 236/258, 91.5%, respectively; P=.009), whereas the incidence of adverse reactions was notably lower (5/255, 2% vs 15/258, 5.8%, respectively; P=.02).

Conclusions: Using the AI-driven bowel preparation program to assist patients in personalized bowel preparation can reduce unnecessary use of laxatives while achieving results comparable to those of standard practices.

Trial registration: ClinicalTrials.gov NCT06610630; https://clinicaltrials.gov/study/NCT06610630.