Availability and use of computer-aided detection during colonoscopy: A real-world observational study at an Australian tertiary center

Abstract

BACKGROUND

One of the main aims of colonoscopy is to detect and remove precancerous polyps. Multiple studies have shown that computer-aided detection (CADe) technology enhances key metrics, including adenoma detection rate (ADR), among endoscopists with low baseline detection rates. However, these findings largely come from controlled prospective studies where CADe is systematically used and endoscopists are aware they are being monitored for such metrics, bringing inherent biases. In Australia, CADe implementation is not yet standard practice, and its availability varies across endoscopy centers. We hypothesized that greater endoscopist use of CADe would be associated with higher ADR in real-world clinical settings.

AIM

To evaluate how varying levels of endoscopist use of CADe affect adenoma detection and other colonoscopy quality metrics in a tertiary Australian center, where CADe was available for all elective procedures from 2023.

METHODS

A single-center retrospective cohort study was conducted at a tertiary Australian center after introduction of the Olympus Endo-AID^® CADe module in July 2023, available for all elective procedures. Colonoscopy reports from six months before and after implementation were reviewed. Endoscopists were grouped by observed CADe usage. The primary outcome was change in ADR by group. Secondary outcomes included sessile serrated lesion detection rate (SSL-DR), adenomas per patient (APP), and sessile serrated lesions per patient (SPP).

RESULTS

Seven endoscopists performed 636 pre-CADe and 386 post-CADe colonoscopies. Two endoscopists used CADe 100% of the time, four used it 50%-99%, and one did not use CADe. No endoscopists used CADe 1%-50% of the time. ADR significantly improved from 29% to 41.9% in the 50%-99% group (odds ratio 1.77, 95%CI: 1.13-2.75, P = 0.01). No ADR change was observed in the 100% group, which had a baseline ADR above 60%, although APP increased from 0.90 to 2.08 (relative risk 2.31, 95%CI: 1.97-2.73, P < 0.0001). SSL-DR and SPP were not significantly affected by CADe.

CONCLUSION

In this real-world study, the availability of CADe was associated with an uptake by the majority of the endoscopists and led to significant improvement in ADR even when not being used in all procedures. Similarly to previous studies, no such benefit was observed for endoscopists who had a high baseline ADR. However, endoscopists with high baseline ADR did improve their APP after introduction of CADe. In addition, CADe availability did not improve SSL-DR across the cohort. This real-world significant increase from moderate baseline ADR reinforces its benefit when adopted in routine clinical practice.

Keywords: Computer-aided detection; Adenoma detection rate; Colonoscopy; Artificial intelligence; Endoscopy; Screening, surveillance; Colorectal neoplasm; Polyp; Adenoma

Core Tip: This study evaluates the real-world impact of varying computer-aided detection (CADe) usage frequency on colonoscopy quality metrics in a tertiary Australian endoscopy center. Endoscopists were stratified by CADe usage, revealing that even partial use (50%-99% of procedures) significantly improved adenoma detection rate (ADR). High-frequency users maintained high ADR, but total adenomas per procedure increased, highlighting additional quality benefits. The findings emphasize that CADe effectiveness depends not only on technology availability but also on endoscopist level of usage, offering insights into optimizing implementation and improving patient outcomes through enhanced polyp detection and risk stratification.