Automated colonoscopy withdrawal phase duration estimation using cecum detection and surgical tasks classification

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Colorectal cancer is a leading global cancer, often preventable by removing polyps during colonoscopy.
• Missed polyp detection during colonoscopy remains a significant challenge, impacting cancer prevention.
• Accurate measurement of the colonoscopy withdrawal time is essential for effective polyp detection.

Purpose of the Study:

• To develop an automated method for precise colonoscopy withdrawal time estimation.
• To accurately detect the cecum and classify colonoscopy procedure phases.
• To enhance polyp detection rates by optimizing the exploration phase duration.

Main Methods:

• Utilized a Resnet model for automated cecum detection and colonoscopy phase classification.
• Trained the Resnet model using two public and one private dataset comprising 96 full colonoscopy procedures.
• Evaluated the method on 19 independent testing procedures.

Main Results:

• The automated system accurately estimated withdrawal time in 18 out of 19 testing procedures.
• Achieved a mean error of 5.52 seconds per minute in withdrawal time estimation.
• Demonstrated the feasibility of automatic phase classification for precise time measurement.

Conclusions:

• Automated cecum detection and phase classification can reliably determine colonoscopy withdrawal time.
• This method offers a significant improvement over manual time measurement, enhancing procedural efficiency.
• Potential to improve polyp detection rates and reduce colorectal cancer incidence through optimized colonoscopy protocols.