Smartphone application for artificial intelligence-based evaluation of stool state during bowel preparation before...

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Colonoscopy (CS) is crucial for detecting precancerous lesions.
• Adequate bowel preparation (BP) is essential for high-quality CS.
• Objective stool state evaluation during BP is challenging.

Purpose of the Study:

• To develop a smartphone application (app) with an AI model for evaluating stool state during BP.
• To assess if the app improves CS quality by ensuring adequate BP.

Main Methods:

• Collected and categorized stool images (grades 1-3) for AI model development.
• Constructed and validated an AI model for stool state assessment.
• Conducted a prospective study to evaluate CS quality in patients using the app.

Main Results:

• The AI model achieved high accuracy rates for stool state evaluation (90.2% for grade 1, 89.3% for grade 3).
• In the prospective study (106 patients), 99.0% achieved a Boston Bowel Preparation Scale (BBPS) score of ≥6.
• The app facilitated successful bowel preparation for high-quality CS.

Conclusions:

• The developed AI-powered app effectively evaluates stool status during BP.
• The app contributes to achieving high-quality colonoscopies by ensuring adequate bowel preparation.
• This technology has the potential to enhance CS quality in clinical practice.