iMerit, Segmed, and Advocate Health have released an open-source, annotated digital breast tomosynthesis (DBT) dataset designed to support the development of AI tools for breast cancer detection.
The dataset includes imaging studies from 558 female patients with biopsy-confirmed diagnoses, comprising 271 malignant (48.5%) and 287 benign (51.5%) cases. Images were interpreted by U.S. board-certified, Mammography Quality Standards Act (MQSA)-certified radiologists and segmented by breast imaging specialists, the firms said. The average tumor size in the dataset is 1.34 cm, with approximately 85% of lesions smaller than 2 cm, meaning the dataset can train AI models on early-stage findings.
Data is provided in DICOM and Nearly Raw Raster Data (NRRD) formats, including segmentation images and JavaScript Object Notation (JSON) classifications, and is fully deidentified in compliance with HIPAA and General Data Protection Regulation (GDPR) standards, the company said. The dataset is available for free download to registered users at imerit.net.
