AI-Assisted Colonoscopy Improves Polyp Detection and Diagnostic Accuracy

Despite its established role in colorectal cancer (CRC) screening, colonoscopy is not without limitations. In fact, studies estimate that approximately 30% of adenomas may be missed during routine procedures. In a prospective, multicenter randomized trial, researchers evaluated whether a real-time computer-aided detection (CADe) system could help address this gap, and the results suggest it can.

Study Design

This parallel-controlled trial enrolled 390 patients aged 40 to 75 undergoing colonoscopy for CRC screening or clinical diagnosis. Participants were randomly assigned to physician diagnosis or diagnosis supported by an AI-based image recognition software. The CADe system was designed to integrate directly with endoscopy platforms via common video outputs, enabling real-time detection of polyps.

The trial’s endpoints focused on both detection efficacy, measured by polyp per colonoscopy (PPC), adenoma per colonoscopy (APC), adenoma detection rate (ADR), and polyp detection rate (PDR), as well as diagnostic accuracy, including sensitivity and specificity relative to expert-reviewed video footage.

Detection Improved, Especially for Smaller Polyps

In terms of the findings, the CADe-supported colonoscopies showed a marked improvement in polyp detection metrics. Polyp detection rate increased from 56.92% in the control group to 67.18% in the CADe group.

Notably, the system appeared particularly effective at identifying diminutive polyps (≤5 mm) as detection in this subgroup rose from 56.92% to 61.03%.

Diagnostic Accuracy Also Favored AI Support

Beyond quantity of detection, the CADe system also demonstrated strong diagnostic performance. In the full analysis set (FAS), sensitivity and specificity reached 95.19% and 98.44%, respectively. Meanwhile, in the per-protocol set (PPS), sensitivity was even higher at 95.82% compared with 77.53% in the control group, and both groups maintained 100% specificity.

These figures underscore the potential of AI tools to increase lesion recognition while maintaining high diagnostic accuracy, which is particularly relevant for improving detection without increasing false positives.

Balancing Benefits with Study Design Parameters

While the trial was prospective, randomized, and multicenter, participants were limited to patients aged 40–75 undergoing colonoscopy for screening (risk score ≥4) or clinical diagnosis.

Additionally, diagnostic accuracy was assessed by comparing real-time system performance to retrospective expert video review, which served as the study’s gold standard.

Clinical Takeaway: AI Is an Adjunct, Not Replacement

Given the demonstrated potential to improve adenoma and polyp detection rates without compromising diagnostic specificity, this study supports the use of real-time CADe systems as an effective adjunct to colonoscopy. And as AI continues to evolve in endoscopic practice, its integration could help optimize detection consistency across varying procedural contexts.

Reference:
Xu X, Ba L, Lin L, et al. Evaluation efficacy and accuracy of a real-time computer-aided polyp detection system during colonoscopy: a prospective, multicentric, randomized, parallel-controlled study trial. Surg Endosc. 2025;39(11):7417-7427. doi:10.1007/s00464-025-12080-x