TL;DR
A machine-learning model trained on nearly 100,000 breast scans identified women with cardiovascular disease from the mammogram alone. Prior stroke was detected 86% of the time, with hypertension and coronary heart disease close behind at 79% and 78%. The findings went before the annual congress of the ESC, the European Society of Cardiology, meeting in Munich.
What the study did
Doctors in Israel worked through 97,364 scans belonging to 29,921 women whose average age was 54, then cross-referenced medical records to establish who actually had which condition. Hypertension turned up in roughly one woman in six; coronary heart disease and stroke history each appeared in one in forty. The model’s discrimination held steady across age bands, and a cancer diagnosis made no difference to it either.
Presenting the work, Dr Viana Copeland of Tel Aviv University made the practical argument rather than the technical one. Women who develop cardiovascular disease frequently present only once it is advanced, yet many of those same women turn up reliably for breast screening. Reading the scan twice costs no additional imaging.
Elena Arbelo, who sits on the ESC’s communication committee, called the results compelling while drawing the obvious line: moving from experiment to clinical use requires accuracy and reliability that has not yet been established. The team is still working to cut false results and widen the range of conditions covered.
Why this matters in Britain
The NHS Breast Screening Programme invites women aged 50 to 71 every three years, which is exactly the midlife window in which cardiovascular risk becomes worth acting on and is routinely missed. The infrastructure — the machines, the invitations, the attendance habit — is already paid for.
The British Heart Foundation’s clinical director, consultant cardiologist Sonya Babu-Narayan, notes that heart disease is still widely miscast as a male problem, which leaves women underdiagnosed and thin on the ground in trials. A screening programme women already attend is an unusually direct way around that.
Looking forward
The obstacle is commissioning, not computation. Resultsense has covered several NHS deployments this month — chest X-ray triage in East Grinstead, AI-guided clot scanning at Chelsea and Westminster — and in each the AI supports a pathway that already exists. Dual-purpose mammography does something different: it produces an incidental cardiovascular finding inside a cancer programme, which raises questions of who receives that result, which budget follows it, and which clinician is responsible for acting on it.