← Back to Breast Atlas

Datasets

Breast imaging datasets for AI training, validation, and clinical research. Praxium datasets are available under license; public datasets are listed for reference.

Request sample + pricing
DatasetOriginYearStudiesImagesModalityAnnotation typeFinding typeFormatLicense
VinDr-Mammo
Vietnam20225,00020,000FFDMLesion-level bounding boxes + exam-/breast-level BI-RADS & densityBreast-level BI-RADS + density + lesion bounding boxes (mass/calcification/asymmetry/distortion)DICOMPhysioNet Restricted Health Data License 1.5.0
PRAXIUMGeorgia — Screening cohort

Proprietary. Licensed through Praxium.

Georgia2022–20254,347~17,400FFDMBreast-level + finding-level (BI-RADS, density)Screening mammography findings, BI-RADS + densityDICOMPraxium license
PRAXIUMGermany — Screening cohort

Proprietary. Licensed through Praxium.

Germany2020–20252,074~8,300FFDMBreast-level + finding-level (BI-RADS, density, location, pathology)Screening mammography findings, pathology-verifiedDICOMPraxium license
CMMD
China20211,7755,202FFDMExam-/patient-level labels only (no lesion localization)Biopsy-proven abnormalities (masses/calcifications) + molecular subtype metadataDICOMCC BY 4.0
CBIS-DDSM
United States20171,56610,239SFMLesion-level ROI masks + lesion attributes + pathology/BI-RADS labelsMasses and calcifications (ROI/mask annotations available)DICOMCC BY 3.0
KAU-BCMD

Low resolution, unpaired JPG images.

Saudi Arabia20211,4165,662FFDMExam-/image-level BI-RADS + breast density (no lesion localization)BI-RADS–based case labels (benign/malignant/normal)JPGCC0 (Public Domain)
Iraq-Sulaymaniyah

Also includes 9,685 augmented images.

Iraq2024745745FFDMImage-level binary label only (cancer vs non-cancer)Cancer vs non-cancer classificationDICOMCC BY 4.0
CDD-CESM

Includes radiology manual annotations + reports + structured XLSX.

Egypt2019–20213262,006CESMLesion-/case-level descriptors + BI-RADS + radiology reports + annotation dataCESM low-energy + subtracted images; categorized findings + annotations + reportsDICOMCC BY 4.0
DMID

Includes DICOM and TIFF folders.

India2023127510FFDMExam-/image-level BI-RADS + density + abnormality type + lesion ROI masks + reportsMultiple findings (calcifications, masses, asymmetry, distortion, normal) + ROI masks + reportsDICOMResearch & education only
INbreast
Portugal2012115410FFDMLesion-level polygon contours (XML)Masses, calcifications, asymmetries, distortions (lesion contours in XML)DICOMNot explicitly specified
KIOS — Micro-Calcifications
Republic of Cyprus2021100400FFDMLesion-level micro-calcification annotations + BI-RADS (sequential prior/recent)Micro-calcifications (BI-RADS; lesion-level precise annotations)DICOMCC BY
KIOS — Masses
Republic of Cyprus2021100400FFDMLesion-level mass annotations + benign/suspicious labels (sequential prior/recent)Masses (sequential mammograms; lesion annotations)DICOMCC BY
PRAXIUMUkraine — Diagnostic centers

Proprietary. Available on request.

Ukraine2025–2026On requestOn requestFFDMCustom labeling availableCurated diagnostic cohortsDICOMPraxium license
13datasets
17,691total studies
71,274total images