Key facts
- An AI model can detect heart disease, high blood pressure, and stroke from mammograms.
- The AI achieved 86% accuracy in identifying stroke survivors from mammograms.
- The approach could allow breast cancer screening to also flag cardiovascular issues.
- The study analyzed over 97,000 mammograms from nearly 30,000 women.
- Cardiovascular disease is the leading cause of death in women globally.
Researchers have developed an artificial intelligence model capable of detecting cardiovascular diseases, including heart disease, high blood pressure, and stroke, by analyzing routine mammograms. This breakthrough, presented at the European Society of Cardiology's annual congress, could transform breast cancer screening into a dual-purpose tool for identifying women at risk of heart conditions.
The study, conducted by doctors in Israel, examined over 97,000 mammograms from nearly 30,000 women. The AI model demonstrated significant accuracy, identifying stroke survivors 86% of the time, high blood pressure 79% of the time, and coronary heart disease 78% of the time, based solely on the mammogram images. These results were consistent across different ages and whether or not the women had cancer.
Cardiovascular disease (CVD) is the leading cause of death for women globally and is often underdiagnosed or detected late. Dr. Viana Copeland of Tel Aviv University highlighted that many women attend regular breast cancer screenings, making mammography a potentially scalable method for assessing heart health without additional imaging. This is particularly important as midlife is a critical period for addressing cardiovascular risk.
Experts, including Elena Arbelo from the European Society of Cardiology and Dr. Sonya Babu-Narayan from the British Heart Foundation, welcomed the findings, emphasizing the potential for earlier detection and prevention of CVD in women. The next steps involve refining the AI model's accuracy, reducing false results, and expanding its ability to detect a wider range of heart conditions before clinical implementation.