A dataset of mammography images with area-based breast density values, breast area, and dense tissue segmentation masks

Abstract

A new dataset is presented to propel research in automated breast density estimation, a crucial factor in mammogram interpretation. Mammography, a low-dose X-ray technique for breast cancer screening, can be affected by breast density. Dense tissue appears white on mammograms, potentially obscuring tumors. This dataset, built upon the public VinDr-Mammo dataset, offers 745 mammogram images (including training and test sets) along with expert-radiologist annotations for both the entire breast and dense tissue regions. Researchers can leverage this dataset for multiple purposes: training deep learning models for automated breast density analysis, refining segmentation methods for accurate delineation of breast tissue, and benchmarking existing and novel breast density estimation algorithms. This resource holds promise for improving breast cancer screening through advancements in automated breast density analysis.