A routine mammogram appointment. No cardiac symptoms. No referral. Just a standard screening — and the scan quietly flags elevated stroke risk before a single doctor has noticed anything wrong. That’s the promise researchers presented at ESC Congress 2026 in Munich, the world’s largest cardiology conference. The underlying problem it addresses is blunt: heart disease kills more women than anything else, yet medicine has historically treated it as a man’s condition.
One Test, Two Jobs
Israeli researchers repurposed 97,364 existing mammogram scans to detect three major cardiovascular conditions — without a single extra test.
Buried in 14 years of routine breast scans, a deep-learning model found what routine cardiac check-ups were missing. Researchers at Chaim Sheba Medical Center and Tel Aviv University trained AI on 97,364 mammograms from 29,921 women — average age 54 — collected between 2011 and 2025. Sixteen percent of those women had hypertension; 2.5% had coronary heart disease; 2.5% had experienced a stroke.
The AI’s discriminatory accuracy, measured on a scale where 0.5 is a coin flip and 1.0 is perfect:
- Stroke risk: AUROC 0.86 — roughly 86% discriminatory ability
- Hypertension: AUROC 0.79
- Coronary heart disease: AUROC 0.78
- Performance held consistent across age groups and cancer status
Turning an existing screening programme into a cardiovascular early-warning system is exactly the efficiency play healthcare needs. As study presenter Dr. Viana Copeland put it, mammography “could potentially offer a scalable approach without requiring an additional imaging examination.” Midlife is precisely when cardiovascular risk accelerates and interventions matter most — and mammography already reaches millions of women at that window.
From Research to Reality – The Gap That Remains
Compelling results, cautious experts, and a long road before this reaches your doctor’s report.
Separate research published in the European Heart Journal found AI could quantify breast arterial calcification — calcium buildup in breast arteries visible on existing scans — as an independent predictor of serious cardiovascular events. Women with moderate calcification faced 70% higher risk; severe calcification correlated with two-to-three times the risk. A large Australian study of 49,196 women matched established clinical risk calculators with a concordance index of 0.72, using only age and mammogram features. Think of it like Shazam identifying a song from ambient sound: the signal was already in the room — the AI just learned to hear it.
Dr. Elena Arbelo of the European Society of Cardiology called the findings “compelling,” adding that mammograms may one day “offer a window on to cardiovascular health” — before cautioning that the challenge now is moving “from experimentation to clinical implementation.” Dr. Sonya Babu-Narayan of the British Heart Foundation framed the stakes plainly: heart disease mythology still positions it as a man’s disease, leaving women “unaware, unheard, underdiagnosed, undertreated.”
For women who already schedule mammograms, the unresolved questions are real:
- validation across diverse populations
- integration into clinical workflows
- regulatory clearance
- protecting patients from anxiety triggered by false positives
The technology suggests something genuinely promising. Clinical adoption, however, requires the broader healthcare system to close a significant gap — and that work is only beginning.
