Histo-Miner Extracts Tissue Features From cSCC
Sancéré et al. (published Jan 21, 2026) present Histo-Miner, a deep-learning pipeline for H&E whole-slide images of cutaneous squamous cell carcinoma, trained on datasets with 47,392 annotated nuclei across five cell types in 21 WSIs and tumor-region segmentation in 144 WSIs. Models achieve nucleus mPQ 0.569, macro F1 0.832 and tumor mIoU 0.907, and the pipeline predicts immunotherapy response on 45 patients with mean AUC 0.755±0.091.
Key Points
- 1Generated annotated datasets: 47,392 nuclei across five cell types and 144 tumor-segmented whole-slide images
- 2Demonstrated robust models: nucleus mPQ 0.569, classification macro-F1 0.832, tumor segmentation mIoU 0.907
- 3Predicts immunotherapy response for 45 patients with mean AUC 0.755 ±0.091, highlighting immune spatial biomarkers
Scoring Rationale
Strong dataset, validated models and clinical use-case justify high score, limited to cSCC-specific cohort and moderate sample size.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
