Fully Automated Stain Quantification Framework for IHC Whole Slide Images in Breast Cancer

Introduction Immunohistochemistry (IHC) plays a crucial role in breast cancer diagnosis, treatment selection, and research. However, manual scoring of IHC whole slide images (WSIs) is time-consuming and suffers from inter- and intra-observer variability. Methods To help address these challenges, we present and publicly release a fully automated, compartment-specific (ie, tumor and stroma) H-scoring framework for IHC analysis. The framework consists of three deep learning modules: tumor–stroma segmentation, nuclei segmentation, and H-score estimation for tumor and stroma. It processes WSIs in minutes, delivering consistent and reproducible H-scores with accuracy comparable to expert pathologists. The modular design also allows flexibility for use in other IHC tasks such as cellularity quantification, and supports configuration options to balance accuracy and computational efficiency. Results Fine-tuned on 87 expert-annotated patches, the framework achieved a Spearman's rank correlation ( ρ ) in internal validation of 0.84 (95% confidence interval [CI]: 0.77–0.89) across 100 expert-annotated WSIs, outperforming state-of-the-art ( ρ = 0.78, 95% CI: 0.68–0.85) and matching the inter-observer variability between two expert pathologists ( ρ = 0.84, 95% CI: 0.63–0.94). In external validation, it achieved 86% accuracy in HER2 classification (0–3+) and a mean absolute error of 21 ± 10 (range: [5–46]) in CD73 scoring, where ground truth H-scores were all 0. Conclusion The framework achieves agreement comparable to that of expert pathologists, underscoring its clinical utility in providing reproducible IHC scores that can reduce diagnostic variability and support consistent treatment decisions. The code is available at https://github.com/YinTuo/AutoIHC .