Exploring the Clinical Value of an AI-Assisted Patient Self-Assessment App for Bowel Preparation: A Multicenter Study

Sun Yat-sen University

Status: Not yet enrolling

Conditions: Bowel Preparation for Colonoscopy; Colonoscopy

Treatments: Other: The app group

Study type: Interventional

Funder types: Other

Identifiers

NCT07337694

2025ZSLYEC-747

Details and patient eligibility

About

The quality of bowel preparation hinges on how well patients follow the prep-drug regimen, so intensive education is essential. Phone calls, texts, short videos, and mini-programs have all been shown to boost compliance and improve prep quality. Still, we also need a way to spot-early-those patients who are likely to prep poorly so we can step in with a rescue plan.

In our pilot work the investigators built an AI-assisted mini-program that lets patients photograph their effluent and get an instant quality read-out. The single-center RCT showed excellent performance. Because these findings came from one center, the investigators are now launching a multicenter study to test the tool more broadly. Patients will use the AI mini-program at home; if the algorithm predicts inadequate prep it will prompt them to come in early or alert staff so the investigators can initiate a rescue protocol and, ultimately, improve bowel-cleansing quality.

Enrollment: 524 estimated patients

Sex: All

Ages: 18+ years old

Volunteers: Accepts Healthy Volunteers

Inclusion criteria

1. Age ≥ 18 years
2. Undergoing colonoscopy at a participating site
3. Able to defecate in a toilet (or in a setting where stool characteristics can be observed) after taking the bowel-prep solution
4. Proficient in using a smartphone
5. Willing to participate voluntarily

Exclusion criteria

• American Society of Anesthesiologists (ASA) class III or IV;

  • Gastric-outlet or intestinal obstruction; ③ Active gastrointestinal bleeding; ④ Enterostomy (colostomy/ileostomy); ⑤ Status post total colectomy.

Trial design

Primary purpose: Other

Allocation: Randomized

Interventional model: Parallel Assignment

Masking: Single Blind

---

524 participants in 2 patient groups

The App group

Experimental group

Description:

After taking the bowel-prep solution, patients in the app group are instructed to photograph their final stool in the toilet and upload the image through our smartphone app. The AI algorithm immediately scores the prep quality. If the image is rated "poor" or "inadequate," the app displays an alert advising the patient to come to the hospital early or contact clinical staff; clinical staffs then review the photo and decide whether a rescue preparation is needed. The standard rescue was an additional packet of polyethylene glycol. If the image is rated "adequate," the patient is told the prep is acceptable and should proceed to the appointment, where clinical staffs will use the uploaded photo to confirm eligibility for colonoscopy.

Treatment:

Other: The app group

The control group

No Intervention group

Description:

The control group followed the conventional procedure: before the examination, clinical staffs asked questions and relied on the patient's verbal description of their stool to judge the adequacy of bowel preparation. If the prep was deemed inadequate, the standard rescue was an additional packet of polyethylene glycol.