Diagnostic performance of artificial intelligence to identify deeply invasive colorectal cancer on non-magnified plain endoscopic images
Source check: Source checked. Not independent scientific confirmation.
Finding: A fine-tuned ResNet-50 computer-aided diagnosis system classified deeply invasive T1b colorectal cancer from non-magnified endoscopic images. In a third-hospital test set of 44 lesions, it outperformed novice readers on specificity but was slightly below expert readers. The authors did not test real-time colonoscopy.
Technology: A pretrained ResNet-50 CNN is fine-tuned to estimate the probability that an endoscopic image depicts deeply invasive T1b colorectal cancer.
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A pretrained ResNet-50 CNN is fine-tuned to estimate the probability that an endoscopic image depicts deeply invasive T1b colorectal cancer.
- Test and result
- Third-hospital lesion-level evaluation reported AUC 0.888; at the study threshold, specificity was 87%, sensitivity 81% and accuracy 84%. The cohort contained 44 lesions.
- Limitation / still unknown
- Small retrospective test set; performance was below experts and no real-time or clinical-outcome evaluation was performed.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7508661/ · https://pubmed.ncbi.nlm.nih.gov/33015336/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Primary source full text, including Methods and Results; first result-bearing version checked against primary publication metadata..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.