Artificial Intelligence-Assisted Colonoscopy for Detection of Colon Polyps: a Prospective, Randomized Cohort Study

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Colorectal cancer (CRC) prevention relies on effective polyp detection during colonoscopy.
• Real-time automatic polyp detection systems utilize deep learning to aid endoscopists.
• Improving polyp detection rate (PDR) is crucial for CRC prevention.

Purpose of the Study:

• To evaluate if a high-performance, real-time AI polyp detection system can enhance PDR in a clinical setting.
• To assess the impact of AI assistance on polyp detection during colonoscopies.

Main Methods:

• 150 patients underwent same-day, back-to-back colonoscopies (traditional vs. AI-assisted) in a randomized order.
• Experienced endoscopists (>3000 colonoscopies) performed the procedures.
• The primary outcome measured was the polyp detection rate (PDR). ClinicalTrials.gov Identifier: NCT047126265.

Main Results:

• AI-assisted colonoscopy significantly increased the overall PDR (38.7% vs. 34.0%, p < 0.001).
• AI notably improved the detection of small polyps (< 6 mm) (91 vs. 69, p < 0.001).
• No significant difference in detecting larger polyps was observed.

Conclusions:

• Real-time AI polyp detection systems can boost PDR, particularly for diminutive polyps.
• Larger sample sizes are required in future studies to validate these findings.
• AI shows promise in enhancing colonoscopic polyp detection efficacy.