HER2-IHC-40x: High-Resolution Histopathology Image Dataset for HER2 Scoring in Breast Cancer

HER2-IHC-40x: High-Resolution Histopathology Datasets for HER2 IHC Scoring Overview This dataset contains high-resolution histopathological images of HER2-stained breast cancer tissue sections. Designed for deep learning-based HER2 scoring, the dataset includes two variants: HER2-IHC-40x: Patches extracted after splitting WSIs. HER2-IHC-40x-WSI: Patches extracted before splitting. Each image patch is categorized into one of four HER2 classes (0, 1+, 2+, 3+), based on staining intensity. Dataset Contents Dataset Variants 1. HER2-IHC-40x WSI-based 80-20 split before patch extraction. 107 WSIs → 9940 patches (8093 train / 1847 test) 2. HER2-IHC-40x-WSI Patch-based 80-20 split after patch extraction. 107 WSIs → 10,997 patches (8897 train / 2200 test) Folder Structure HER2-IHC-40x/ ├── WSI/ # Original Whole Slide Images (.svs) ├── ROI/ # Expert-annotated tumor regions (.png) ├── Patches/ # 1024x1024 image patches, labeled 0, 1+, 2+, 3+ ├── Train/ # 80% training set └── Test/ # 20% test set HER2-IHC-40x-WSI/ ├── Patches/ ├── Train/ └── Test/ HER2 Class Definitions | HER2 Score | Description |----------|----------------------------------------------------------------------------- | 0 | No observable staining | 1+ | Weak/incomplete membrane staining in ≤10% tumor cells | 2+ | Moderate circumferential staining in >10% tumor cells (Equivocal) | 3+ | Strong circumferential staining in >10% tumor cells (Positive) Preprocessing & Quality Control ROI Selection: Manual annotation by expert pathologists using Cytomine. Color Histogram Filtering: Removed non-tumor/low-quality patches using HSV filtering. Normalization: Intensity normalization across all patches. Patch Extraction: Adaptive 1024×1024 extraction using sliding window method. Usage This dataset is suitable for: HER2 scoring automation using deep learning Explainable AI (Grad-CAM, attention models) Color normalization and domain adaptation Model benchmarking and generalization research Dataset Statistics HER2-IHC-40x (WSI Split) | HER2 Class | WSIs | ROIs | Patches | |----------|------|----- -|---------| | 0 | 23 | 429 | 3789 | | 1+ | 26 | 131 | 2153 | | 2+ | 27 | 483 | 634 | | 3+ | 31 | 156 | 3364 | | Total | 107 | 1199 | 9940 | HER2-IHC-40x (Patch Split) | HER2 Class | WSIs | ROIs | Patches | |----------|-----|--------|---------| | 0 | 23 | 429 | 3789 | | 1+ | 26 | 131 | 2689 | | 2+ | 27 | 483 | 1131 | | 3+ | 31 | 156 | 3388 | | Total | 107 | 1199 | 10,997 | Citation If you use this dataset, please cite: ```bibtex @dataset{nabi2025her2, author = {Md Serajun Nabi and Mohammad Faizal Ahmad Fauzi and Hezerul Bin Abdul Karim and Phaik Leng Cheah and Seow Fan Chiew and Lai Meng Looi}, title = {HER2-IHC-40x and HER2-IHC-40x-WSI: High-Resolution Histopathology Dataset for HER2 IHC Scoring in Breast Cancer}, year = 2025, publisher = {Zenodo}, doi = {10.5281/zenodo.15179608}, url = {https://zenodo.org/record/xxxxxxx} } This dataset is part of the research article: **"Enhancing HER2 IHC Scoring Using HRNet and SwinT with Cross-Dataset Generalization"** Authors: Md Serajun Nabi, Mohammad Faizal Ahmad Fauzi, Hezerul Bin Abdul Karim, et al. (Preprint server or journal details to be confirmed.) Search the paper for detail data description: **HER2-IHC-40x: High-Resolution Histopathology Datasets for HER2 IHC Scoring** The color histogram code source:* https://github.com/seraju77/HER2-IHC-40x-data-preprocessing.git *