Deepath-SCC: a deep learning model for accurate tissue origin identification in squamous cell carcinoma - PubMed
3 hours ago
- #squamous cell carcinoma
- #digital pathology
- #deep learning
- Deepath-SCC is a deep learning model designed to identify the tissue origin of squamous cell carcinoma (SCC) from hematoxylin and eosin-stained whole-slide images.
- The model was trained and validated on a retrospective cohort of 4,217 whole slide images covering nasopharyngeal, head and neck/esophageal, lung, cervical, and urothelial carcinomas.
- In internal testing, Deepath-SCC achieved an accuracy of 91.2% and a micro AUROC of 0.986, with high-confidence predictions (similarity score ≥0.7914) boosting accuracy to 96.2% for primary SCCs and 94.4% for metastatic SCCs.
- External validation showed an accuracy of 86.1% and an AUROC of 0.972, supporting its feasibility for clinical use.
- Deepath-SCC offers an efficient, cost-effective computational approach to complement existing diagnostic workflows, especially in challenging or resource-limited settings.