Computational Pathology · owkin
her2-challenge-2026
A computational pathology dataset for predicting breast cancer HER2 status from whole-slide histology images.
What her2-challenge-2026 is
her2-challenge-2026 is a computational pathology benchmark dataset developed by Owkin for an imaging data challenge linked to their 2026 Spring medical imaging internship. It provides haematoxylin and eosin (H&E) stained whole-slide images designed for predicting human epidermal growth factor receptor 2 (HER2) status in breast cancer tissue.
Evaluating HER2 status is a standard step in breast cancer assessment, traditionally determined through laboratory staining methods such as immunohistochemistry or in situ hybridisation. This dataset allows researchers to evaluate whether artificial intelligence models can infer HER2 scores directly from standard digitised histology slides, supporting methodological exploration in digital pathology and automated biomarker scoring.
How it was made
The dataset provides digitised H&E-stained whole-slide breast cancer images alongside associated scoring annotations. It is distributed as a single compressed archive, HER2_CHALLENGE_FOR_INTERNSHIP_2026.zip, accompanied by a Jupyter notebook (baseline.ipynb) providing a reference implementation for feature extraction and image classification tasks.
Key facts
- Category
- Computational Pathology
- Image type
- H&E-stained whole-slide images
- Task categories
- Image classification, feature extraction
- Distribution format
- ZIP archive with Jupyter notebook (baseline.ipynb)
What you can use it for
- HER2 Status Prediction
- Training and evaluating deep learning classifiers to predict breast cancer HER2 status directly from H&E-stained whole-slide images.
- Histopathology Feature Extraction
- Developing and benchmarking feature representation methods on whole-slide pathology images in oncology.
- Competitive Algorithm Benchmarking
- Comparing model performance against a provided baseline script in the context of the Owkin 2026 Spring medical imaging challenge.
Explore the data
No queryable preview
This dataset is not available in a queryable form — usually an unusual file format or a repository too large to convert. The original source may still be readable.Strengths and limits
Good at
- Provides whole-slide images targeted specifically at automated HER2 scoring in breast cancer.
- Includes an official baseline notebook to standardise and facilitate experimental setup.
- Enables benchmarking across both feature extraction and classification pipelines.
Watch out for
- Data is packaged in a single large archive file rather than streamed slide by slide.
Questions
Can I look inside her2-challenge-2026 without downloading it?
Not here: it is not available in a queryable form. Its description and statistics are still shown.
What licence is her2-challenge-2026 under?
Its authors declare cc-by-nc-sa-4.0. Read the terms on the original page before reusing the data.
What imaging modality does this dataset contain?
It contains haematoxylin and eosin (H&E) stained whole-slide histology images of breast cancer tissue.
What tasks can be performed with this dataset?
The primary tasks are image classification and feature extraction focused on determining HER2 status.
Is starter code provided with the benchmark?
Yes, a baseline script is included in the repository as a Jupyter notebook named baseline.ipynb.
Written from her2-challenge-2026's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.