AI Quantifies Breast Arterial Calcification Predicting CVD Risk

A retrospective cohort study of 123,762 women, published in the European Heart Journal (Dapamede et al.), applied an artificial intelligence algorithm to routine mammograms to quantify breast arterial calcification (BAC) and assessed the association between BAC severity and subsequent major adverse cardiovascular events (MACE), as reported in JMIR and press releases from the European Society of Cardiology and the American College of Cardiology. According to those reports, women with mild BAC had about a 30% higher risk of serious cardiovascular events compared with no BAC; moderate BAC carried more than a 70% higher risk; and severe BAC was associated with roughly 2-3x the risk, per ESC and Medical Xpress coverage. The study cohort excluded women with known cardiovascular disease at baseline, per the published report and affiliated press materials.

Per the published study and accompanying summaries, the investigators used a deep-learning image analysis pipeline to detect and quantify arterial calcification visible on standard X-ray mammography; the algorithm produced categorical severity labels (absent, mild, moderate, severe) that were correlated with longitudinal cardiovascular outcomes. The American College of Cardiology noted the potential scale advantage because roughly 40 million mammograms are performed yearly in the United States, allowing risk screening without additional radiation exposure, blood tests, or clinic visits, as described in the ACC press release.

Reusing existing imaging for secondary screening echoes prior work in AI-enabled opportunistic biomarkers (for example, coronary calcium on chest CT and bone density from CT/mammography). Companies and research teams pursuing similar secondary-use workflows typically confront dataset shift, label heterogeneity, and the need for calibrated risk outputs before clinical integration. Observers will note that retrospective association is an important first step but does not itself demonstrate prospective clinical benefit or established integration into care pathways.

For ML practitioners building medical imaging tools, this study highlights two operational advantages-large existing image volumes and the possibility of passive, low-cost risk stratification-and two common challenges: securing diverse multisite data for generalization testing and producing explainable, auditable outputs that clinicians can trust. Regulatory and reimbursement pathways for AI that repurpose screening tests remain evolving; press materials frame the approach as a screening adjunct rather than a diagnostic replacement.

observers should follow prospective validation efforts, external performance across different mammography hardware and populations, linkage studies that integrate BAC quantification with standard clinical risk scores, and any premarket regulatory filings or guideline statements from cardiology and radiology societies. Successful clinical adoption will likely require demonstration that AI-driven BAC reporting changes clinician behavior or patient outcomes in randomized or implementation studies.