BreAst Cancer Histology (BACH) Dataset contains Hematoxylin and eosin (H&E) stained breast histology microscopy images. Images are labelled as normal, benign, in situ carcinoma, or invasive carcinoma based on the predominant cancer type. The annotation was performed by two medical experts and images where there was disagreement were discarded.
Use Cases
- Classify breast tissue images into normal, benign, in situ, or invasive carcinoma categories based on H&E staining.
- Train image classification models on expert-annotated medical histology data.
- Benchmark machine learning models for the Grand Challenge on Breast Cancer Histology images.
Strengths
- Annotations were performed by two medical experts.
- Images where annotators disagreed were discarded, suggesting a focus on label consensus.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- 1aurent on Hugging Face
- Collection Method
- Microscopy images of H&E stained breast histology.
- Freshness
- Last updated 2024-05-25 17:11:58; freshness should be verified.