Stool detection in colonoscopy videos.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Gastroenterology

Background:

• Colonoscopy is vital for colorectal cancer (CRC) and polyp detection, contributing to reduced CRC mortality.
• Inadequate bowel preparation remains a significant challenge, impacting colonoscopy's diagnostic accuracy in 10-75% of cases.
• Accurate assessment of bowel cleansing quality is crucial for effective CRC screening.

Purpose of the Study:

• To develop and evaluate a machine learning model for automated stool detection in colonoscopy videos.
• To improve the accuracy of colonoscopy by objectively assessing bowel preparation quality.

Main Methods:

• A machine learning approach utilizing a support vector machine (SVM) classifier.
• Classification of colonoscopy video frames based on extracted color features to detect stool presence.
• Digitization of colonoscopy video files for image analysis.

Main Results:

• The proposed machine learning method demonstrated high accuracy in classifying stool images.
• The SVM classifier effectively utilized color features for precise stool detection.
• This automated method offers a promising solution for objective bowel preparation assessment.

Conclusions:

• Machine learning, specifically SVM with color features, can accurately detect stool in colonoscopy videos.
• This technology has the potential to enhance colonoscopy quality and improve colorectal cancer detection rates.
• Automated stool detection can standardize the assessment of bowel preparation, overcoming current limitations.