A deep-learning model trained on 97,364 mammography exams from 29,921 women distinguished ischaemic heart disease on routine breast screening with an AUROC of 0.78, alongside hypertension at 0.79 and stroke at 0.86. In a retrospective study set for presentation at the European Society of Cardiology congress, performance was consistent across age groups and cancer status, including the roughly 18% of women with breast cancer.

Why It Matters To Oncology

• Breast imaging may yield dual-use clinical signals: cancer detection plus cardiovascular risk stratification from the same exam.

• That matters for oncology because cardiometabolic disease can shape treatment selection, survivorship planning and cardio-oncology monitoring.

• For drug discovery teams, the work underscores how AI can extract non-obvious phenotypes from standard imaging datasets, potentially expanding biomarker discovery beyond tumor-specific endpoints.

The Financials

• The potential economic appeal is scalability: no additional imaging exam is required if cardiovascular readouts can be layered onto existing mammography workflows.

• Earlier detection of HTN, ischaemic heart disease and stroke risk could reduce downstream costs tied to late presentation and acute events.

• Still, any real value proposition will depend on prospective validation, workflow integration and acceptable false-positive and false-negative rates.

What They're Saying

• Study author Viana Copeland said analyzing mammograms for cardiovascular information could offer a scalable approach because the imaging is already widely used.

• Copeland also noted mammography reaches many women in midlife, an important window for recognizing cardiovascular risk.

• ESC commentator Elena Arbelo said a mammogram may eventually offer a window onto cardiovascular health, but the field must now prove accuracy and reliability before clinical implementation.

What's Next

• The investigators are working to improve model accuracy and reduce false positives and false negatives.

• They also plan to test whether mammograms can reveal other cardiovascular conditions beyond HTN, ischaemic heart disease and stroke.

• Key next steps for clinicians will be prospective validation, external replication and defining how such outputs should be acted on in breast screening and oncology settings.