Efficiency of Real-time Computer-aided Polyp Detection during Surveillance Colonoscopy: A Pilot Study

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Computer-aided polyp detection (CAD) using artificial intelligence (AI) has shown potential for improving adenoma identification during colonoscopy.
• However, its real-world effectiveness in routine surveillance colonoscopies remains under investigation.

Purpose of the Study:

• This study aimed to evaluate the practical utility of CAD during regular surveillance colonoscopies.
• The research specifically assessed CAD's effectiveness in identifying adenomas in patients undergoing follow-up colonoscopies.

Main Methods:

• The study included consecutive patients undergoing surveillance colonoscopy with CAD between January and March 2023, who had at least two previous colonoscopies in the past three years.
• Clinicopathological findings of lesions detected by CAD were analyzed.
• Detection performance was sub-analyzed based on endoscopist experience and the presence of diminutive adenomas (≤5 mm).

Main Results:

• Out of 78 patients, CAD identified 46 adenomas in 28 patients; no carcinomas were detected.
• Newly detected adenomas were found in 34.1% of patients with untreated diminutive adenomas and 38.2% of patients with no prior adenomas.
• Endoscopist expertise did not significantly influence the detection results.

Conclusions:

• Computer-aided detection (CAD) demonstrates potential value in identifying adenomas during surveillance colonoscopies.
• CAD may be particularly beneficial for patients with a history of untreated diminutive adenomas or previous endoscopic resections.