Evaluation efficacy and accuracy of a real-time computer-aided polyp detection system during colonoscopy: a...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Gastroenterology

Background:

• Colorectal cancer (CRC) is a leading cause of cancer death globally.
• Colonoscopy is the primary screening tool for CRC, but has a ~30% miss rate for polyps.
• AI-assisted systems are being developed to improve colonoscopy detection rates.

Purpose of the Study:

• To assess the effectiveness and accuracy of a real-time AI-assisted polyp detection system during colonoscopy.
• To compare polyp detection rates and diagnostic accuracy between AI-assisted and standard colonoscopy.

Main Methods:

• A randomized controlled trial involving 390 patients (aged 40-75) undergoing colonoscopy.
• Participants were assigned to either an AI-assisted (computer-aided detection - CADe) group or a standard (physician diagnosis) control group.
• Efficacy metrics included polyp per colonoscopy (PPC), adenoma per colonoscopy (APC), adenoma detection rate (ADR), and polyp detection rate (PDR). Accuracy was measured by sensitivity and specificity against expert video review.

Main Results:

• The CADe group showed significantly higher polyp detection rates (PDR) (67.18% vs. 56.92%).
• More small polyps (≤5mm) were detected in the CADe group (61.03% vs. 56.92%).
• The CADe group demonstrated improved sensitivity (95.82% vs. 77.53% in PPS dataset) with high specificity (100% in PPS dataset).

Conclusions:

• Real-time AI-assisted polyp detection enhances colonoscopy efficacy.
• The system significantly boosts adenoma and polyp detection rates.
• The AI system achieves high accuracy with excellent sensitivity and specificity, improving CRC screening.