Cited by 4Colorectal Cancer Screening and DetectionAI in cancer detectionGenetic factors in colorectal cancer

The Augmented Colonoscopy With Computer-Aided Polyp Characterization Study: Prospective Study Comparing the Diagnostic Reliability of Optical Diagnosis of Trainees With Experts Without Artificial Intelligence

Bernhofer, Sebastian, Prosenz, Julian, Duller, Christine, Venturi, David, Maieron, Andreas·2026·American Journal of Gastroenterology

Abstract

INTRODUCTION: Optical diagnosis (OD) is an essential part of a high-quality colonoscopy, but highly experience-dependent. Artificial intelligence (AI) in the form of computer-aided diagnosis (CADx) may bridge the gap between trainee endoscopists and experts. The aim of this study was to evaluate the diagnostic reliability of OD of trainee endoscopists with the help of AI compared with experts. METHODS: This prospective, observational study included patients undergoing trainee-performed CADx-supported (GI Genius) colonoscopy. Resected polyps were recorded and video-reviewed without CADx information by experts. The primary outcome was the negative predictive value (NPV) for adenomatous histology of diminutive (≤5 mm) rectosigmoid polyps of trainees vs experts and CADx output alone. Secondary outcomes were the NPV for rectosigmoid polyps of any size and sensitivities and specificities of adenomas in the entire colon. RESULTS: Overall, 630 polyps were resected in 225 patients (48.9% male, mean age 63.8 (SD 12.7) years). In the rectosigmoid, 252 lesions (40%) were found, 223 (88.5%) of which were ≤5 mm. The NPV for diminutive rectosigmoid polyps of trainees using CADx was 90.2% (95% confidence interval [CI] 0.85–0.94), NPV of the experts without CADx was 90.3% (95% CI 0.84–0.94). There was no statistically significant difference in NPV between these 2 groups. The NPV of CADx alone was 93.2% (95% CI 0.88–0.97). The NPV for rectosigmoid polyps of any sizes were 90.1% (95% CI 0.85–0.94) for trainees, 90.4% (95% CI 0.85–0.95) for experts, and 93.4% (95% CI 0.88–0.97) for CADx alone. DISCUSSION: OD of rectosigmoid polyps by trainee endoscopists with CADx is highly accurate, fulfilling PIVI 2 “diagnose-and-leave” strategy.

Integrity signals

Registry

Crossref · 10.14309/ajg.0000000000003558

Cites

42 of 42 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

This paper’s full text isn’t openly available, so its claims and self-check weren’t run. The registry and citation checks still apply.

Cites

Flagged first, then by citations

Cited by

Still being checked.

Authors
Bernhofer, S., Prosenz, J., Duller, C., Venturi, D., & Maieron, A.
Affiliations
  • Universitätsklinikum St. Pölten (AT)
  • Paracelsus Medical University (AT)
  • Johannes Kepler University of Linz (AT)
Corresponding author
  • Bernhofer, Sebastian
  • Prosenz, Julian
Venue
American Journal of Gastroenterology
Published
2025-05-28
Volume
121
Issue
3
Pages
661-670
ISSN
  • 0002-9270 (ISSN-L)
  • 1572-0241
OpenAlex
W4410821521
Cited by
4
References
42
Citations by year
2026: 1 · 2025: 3
FWCI
1.66
Type
journal article
Language
en
Open access
closed
Locations
  • The American Journal of Gastroenterology (publishedVersion)
  • PubMed (publishedVersion)
Topics
  • Health Sciences › Medicine › Oncology › Colorectal Cancer Screening and Detection
  • Physical Sciences › Computer Science › Artificial Intelligence › AI in cancer detection
  • Health Sciences › Medicine › Pathology and Forensic Medicine › Genetic factors in colorectal cancer
Keywords
computer-aided diagnosis, colonoscopy, polyp characterization, adenomatous polyps, trainee endoscopists, endoscopist experience, negative predictive value, GI Genius, colorectal cancer screening, diagnostic accuracy, artificial intelligence in endoscopy, adenoma detection rate, endoscopic submucosal dissection
MeSH terms
Adenoma, Aged, Artificial Intelligence, Clinical Competence, Colonic Polyps, Colonoscopy, Diagnosis, Computer-Assisted, Female, Humans, Male, Middle Aged, Predictive Value of Tests, Prospective Studies, Sensitivity and Specificity, Colorectal Neoplasms, Reproducibility of Results

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review