Evaluating Deep-Learning Based Quantification of Breast Arterial Calcification on Mammography for Cardiovascular Risk Assessment

The study focused on developing a deep learning model for automated detection and quantification of breast arterial calcification (BAC) on mammograms. The research included 202,006 women without prior cardiovascular events who underwent screening mammography. The model achieved high accuracy with AUROC of 0.97 and correlation of 0.961 with manual measurement. During a median follow-up of 7.5 years, 3.8% of women developed major adverse cardiovascular events. Five-year event incidence increased from 1.5% in women without BAC to 6.9% in those with high BAC burden. Although BAC alone showed modest predictive ability (AUROC 0.661 for 5 years), its combination with the PREVENT score provided only minimal improvement. The BAC model represents a feasible and accurate method for identifying women at elevated cardiovascular risk.