AI Science Tracker
EN

Abdominal-X-ray network classifies recorded colorectal-cancer status

Tri-Service General Hospital; National Defense Medical Center · 31 Oct 2024 · medicine

Discovery

Abdominal-X-ray network classifies recorded colorectal-cancer status

Source check: Source checked. Not independent scientific confirmation.

Finding: A DenseNet-121 analyzed abdominal X-rays against recorded colorectal-cancer labels in Taiwanese hospital data. It achieved AUC 0.738 in 16,875 internal-validation patients and 0.656 in 15,876 patients from a community hospital; external sensitivity was 47.7% at the selected cutoff. Cancer was uncommon, registry records could miss diagnoses elsewhere, and the one-year follow-up/label timing limits claims about detection at the image date. This retrospective study does not establish effective screening or earlier diagnosis.

Technology: DenseNet-121 convolutional neural network

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A 121-layer DenseNet learns radiographic patterns associated with recorded colorectal-cancer labels and returns a case probability for a single abdominal radiograph.
Test and result
Patient-separated development, tuning and validation cohorts, one randomly selected KUB radiograph per person, and registry/EMR colorectal-cancer labels. The threshold chosen in the tuning set was reused in both validation cohorts; external AUC .656 and sensitivity 47.7% were lower than internal results.
Limitation / still unknown
Prospective screening performance, time gained before diagnosis and clinical benefit remain unestablished. Prior cancer was excluded and one-year follow-up used, but the precise diagnosis-to-image timing is not fully specified. Low case prevalence, distal-tumor imbalance and incomplete capture of outside-hospital diagnoses limit interpretation.
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/PMC12092315/

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: 21 Nov 2026. Review scope: Primary relevant cohort/label, DenseNet operation and threshold Methods, validation Results and limitations; supplementary architecture/training appendix not independently read. Exact primary identity/date and fresh primary PubMed correction/retraction relationships checked..

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 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open in the full catalog for filters, charts and related discoveries. Browse medicine discoveries.