CCRA: A colon cleanliness rating algorithm based on colonoscopy video analysis

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Gastroenterology

Background:

• Manual colon cleanliness assessment during colonoscopy is subjective and prone to inconsistency.
• Accurate bowel preparation quality is crucial for effective colonoscopy and polyp detection.

Purpose of the Study:

• To develop an automated Colon Cleanliness Rating Algorithm (CCRA) for objective colon cleanliness evaluation.
• To address the inter-observer variability associated with manual bowel preparation quality assessment.

Main Methods:

• Intercepting images from colonoscopy videos for analysis.
• Utilizing U-Net for semantic segmentation of stool areas within colonoscopy images.
• Calculating colon cleanliness by comparing segmented stool area to a standard proportion.

Main Results:

• The U-Net model achieved high performance with 97.02% pixel accuracy and 90.95% F1-Score.
• The CCRA demonstrated accuracy ranging from 92.45% to 99.275% in rating colon cleanliness.
• The algorithm provides rapid and precise colon cleanliness ratings.

Conclusions:

• The proposed CCRA offers a reliable, automated solution for colon cleanliness assessment.
• This AI-driven approach enhances objectivity and efficiency in evaluating bowel preparation quality.
• CCRA eliminates the need for manual inspection, improving diagnostic consistency.