AI Can Spot Heart Disease and Stroke Risk From Routine Mammograms, Study Finds

Researchers have used artificial intelligence to analyze routine mammograms and identify women with coronary heart disease, high blood pressure, or prior stroke, according to findings presented on 2026-08-27 at the European Society of Cardiology's annual congress in Munich (The Guardian).

Dr Viana Copeland from Tel Aviv University presented the results. The Israeli study examined 97,364 mammograms from 29,921 women with an average age of 54. Within the group, 16% had high blood pressure, 2.5% had coronary heart disease, and 2.5% had experienced a stroke.

The machine-learning model — a type of AI that finds patterns in data without being explicitly programmed for each task — identified women who had suffered a stroke from mammogram data alone 86% of the time. It was 79% reliable at distinguishing women with high blood pressure and 78% reliable for coronary heart disease. The model maintained these results regardless of age or whether a woman also had cancer (The Guardian).

Copeland said cardiovascular disease is consistently underdiagnosed and undertreated despite being the leading cause of death in women worldwide. Elena Arbelo, an expert member of the European Society of Cardiology communication committee, called the findings compelling. Dr Sonya Babu-Narayan, a consultant cardiologist and clinical director of the British Heart Foundation, welcomed the development (The Guardian).

The ESC findings build on a body of research examining the automated detection of breast arterial calcifications (BACs) — calcium deposits in the arteries of the breast — as a proxy for cardiovascular risk. A peer-reviewed study published in the Journal of Medical Internet Research (JMIR) reported that AI-quantified BAC can predict heart disease risk in women. The JMIR study noted that mammography can potentially screen women for cardiovascular disease risk without additional radiation exposure, costs, or doctor visits (JMIR).

The European Society of Cardiology previously stated that AI can assess the build-up of calcium deposits in breast arteries from standard X-ray mammography scans to predict serious or fatal heart disease risk (ESC). The George Institute for Global Health developed a machine learning model using mammograms to predict cardiovascular risk. Research from Emory University published in May 2026 found that AI-assessed mammograms could reveal cardiovascular risks for heart attacks, heart failure, and stroke (Emory University).

Earlier work established the technical foundations. A 2017 study by J. Wang investigated the feasibility of automated BAC detection in mammograms for coronary artery disease risk assessment. A 2022 article by D.A. Adedinsewo stated that AI algorithms can identify women who already carry a diagnosis of overt cardiovascular disease and may benefit from more intensive guideline-directed medical therapy. A 2025 review by S. Chidurala stated that incorporating machine-learning-based BAC detection into routine mammography could improve cardiovascular risk stratification, especially in women.

More recent efforts have continued to develop the underlying technology. A 2026 study by J.Y. Barraclough developed and tested a deep learning algorithm for cardiovascular risk prediction based on routine mammography images. The commercial ecosystem has also expanded; startup Genexia developed AI to diagnose coronary artery disease risk during a mammogram, aiming to use explainable AI for early detection. Separately, an AI-based technology analyzing mammograms to predict personalized five-year breast cancer risk received FDA breakthrough device designation.

The broader context here is the clinical gap in female cardiovascular health. By leveraging an existing screening infrastructure already used by millions of women, AI-assisted mammography could provide a dual-purpose diagnostic tool. The technology targets a physiological marker, breast arterial calcification, that is visible on standard scans and is linked to systemic cardiovascular disease.

For clinicians and health systems, the appeal lies in operational efficiency. Integrating predictive models into existing mammography workflows could yield cardiovascular risk stratification without incremental resource burden. However, translating high reliability figures from retrospective studies — which look back at existing data — into prospective clinical utility requires further validation. Whether these models can improve patient outcomes through earlier intervention, and how they integrate with existing guideline-directed medical therapy pathways, are the next questions for implementation.