67,512 3D mammography images from 15,149 patients, annotated with cancer classifications and lesion bounding boxes by the Oregon-Massachusetts Mammography Database. The dataset includes metadata on image view, laterality, pixel spacing, and confidence scores for regions of interest. Haehn, Daniel published this collection, which was last updated on July 15, 2026.
Use Cases
- Training lesion detection models based on the provided bounding box coordinates (Coords).
- Developing image classification algorithms based on the cancer classification labels (Label).
- Analyzing the impact of different image windowing settings (WindowCenter, WindowWidth) on model performance.
- Validating model predictions using the provided confidence scores (Score) for regions of interest.
Strengths
- Large scale with 67,512 images and 15,149 patients.
- Includes cancer annotations with 374 cancer-positive cases.
- Provides detailed metadata per image, including view (CC/MLO), laterality, and pixel spacing.
- Contains structured region-of-interest data with coordinates and confidence scores.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Data may reflect geographic or institutional bias inherent to the OMAMA-DB source.
Provenance
- Source
- The Oregon-Massachusetts Mammography Database (OMAMA-DB)
- Freshness
- Last updated 2026-07-15 15:15:34; freshness should be verified.