TIGER: ROI-Level Annotations for Breast Cancer Whole-Slide Images

195 whole-slide images of breast cancer biopsies and resections, with regions of interest and manual annotations released at 0.5 um/px magnification. The dataset contains ROI-level annotations of tissue compartments and cells from Her2+ and Triple Negative breast cancer cases. Data originates from Radboud University Medical Center, Jules Bordet Institute, and the TCGA-BRCA archive, released by Mart van Rijthoven of Radboud University Nijmegen in conjunction with the TIGER challenge.

Use Cases

• Training cell detection models based on bounding box annotations in COCO format.
• Developing tissue compartment segmentation algorithms based on pixel-wise class label PNGs.
• Benchmarking multi-source medical image analysis pipelines on Her2+ and TNBC breast cancer data.

Strengths

• Includes annotations from a panel of board-certified breast pathologists for the RUMC and JB subsets.
• Provides data from 195 whole-slide images across three distinct sources (TCGA, RUMC, JB).
• Released in standard computer vision formats (PNG, COCO) to lower the barrier for entry.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Radboud University Medical Center (Netherlands), Jules Bordet Institute (Belgium), TCGA-BRCA archive.

Collection Method

Regions of interest were cropped from whole-slide images and manually annotated; cell annotations were adapted from public BCSS and NuCLS datasets for the TCGA subset.

Geography

Netherlands, Belgium, and cases from the TCGA-BRCA archive.