A Shallow Convolutional Neural Network Predicts Prognosis of Lung Cancer Patients in Multi-Institutional CT-Image Data
Source check: Source checked. Not independent scientific confirmation.
Finding: LungNet uses pretreatment CT images to predict non-small-cell lung cancer survival across four medical-center cohorts, with concordance indices of 0.58–0.62. Transfer learning improved nodule-classification AUC from 0.82 to 0.85; these retrospective predictions do not demonstrate improved clinical outcomes.
Technology: LungNet, a shallow three-dimensional convolutional neural network with transfer learning
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A shallow 3D CNN extracts prognostic features from pretreatment CT scans and predicts survival; transfer learning applies the model to nodule malignancy classification.
- Test and result
- Four independent NSCLC survival cohorts (n=129, 185, 311, 84) had C-indices 0.62, 0.62, 0.62, and 0.58; transfer learning reached AUC 0.85 for nodule malignancy classification versus 0.82 from scratch.
- Limitation / still unknown
- All evidence is retrospective prediction; the study does not show improved patient outcomes or clinical utility.
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/PMC8008967/
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 full-text Methods and Results; bibliographic date and exact-result catalog comparison 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 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.