The use of artificial intelligence (AI) applied to the analysis of mammograms could serve to identify different common cardiovascular diseases (CVD), according to a study from the Chaim Sheba Medical Center of Tel Aviv University, in Ramat Gan (Israel), which will be presented at the ESC 2026 Congress in Munich (Germany).
Viana Copeland, a doctor at the Chaim Sheba Medical Center of Tel Aviv University, emphasizes the need to develop new screening strategies for cardiovascular diseases (CVD) in the female population.
"Despite being the leading cause of death in women worldwide, CVDs are underdiagnosed and treated insufficiently. A common finding in our medical center, and around the world, is that when women seek medical attention, their CVD is already advanced," she explains.
The researcher also recalls that "many women undergo routine breast cancer screening, even when they have not sought attention for cardiovascular symptoms. We investigated whether AI could help mammography fulfill an additional function in this group: the early detection of CVD, which would allow for the implementation of preventive strategies."
In this retrospective cohort study, data were collected from 29,921 women who underwent 97,364 mammograms, with an average age of 54 years.
From electronic medical records, prescriptions, and imaging test results, information was obtained on the presence of three common CVDs (hypertension, ischemic heart disease, or coronary artery disease) and stroke. The observed prevalence was 16% for hypertension, 2.5% for ischemic heart disease, and 2.5% for stroke.
A deep learning model was designed to recognize patterns in the mammograms of women with hypertension, ischemic heart disease, or stroke. Its performance in differentiating between women with and without each cardiovascular pathology was measured using the area under the ROC curve (AUROC), which ranges from 0.5 (random detection) to 1.0 (perfect discrimination).
The initial model showed solid behavior, with an AUROC of 0.79 for hypertension, 0.78 for ischemic heart disease, and 0.86 for stroke, with stable results when stratified by age and by the presence of cancer.
Doctor Copeland emphasizes that, "given that mammography is already widely used, analyzing the same images to obtain cardiovascular information could offer a scalable approach without the need for an additional imaging exam. Furthermore, mammography reaches many middle-aged women, an important period to recognize and address cardiovascular risk."
The scientific team is now continuing to refine the model to increase its accuracy and reduce both false positives and false negatives, and plans to study whether mammograms can also contribute to the detection of other cardiovascular diseases.