AI analysis of mammograms predicts cardiovascular risk, study finds

A deep learning algorithm trained on routine mammograms can predict a woman’s risk of major cardiovascular events, researchers reported, raising the possibility that breast screening could also identify risk of heart attack, heart failure and stroke without additional tests.

Scientists in Australia developed and tested the model using images from 49,196 women enrolled in the Victoria Lifepool cohort registry. Over an average follow-up of almost nine years, 2,383 women experienced a heart attack, 731 had heart failure and 656 had a stroke. The study, published in the journal Heart, found the algorithm performed comparably to traditional risk calculators that use age and clinical variables.

The algorithm used mammographic features such as breast arterial calcification and tissue density, which the researchers said are associated with cardiovascular risk. One advantage highlighted was that the mammography-based model did not require additional history-taking or integration of medical-record data, potentially making cardiovascular risk assessment more widely accessible during routine breast screening appointments.

The cohort’s average age was 59. About one-third of participants were taking medication for high cholesterol and 27% were being treated for high blood pressure. The researchers reported that the dual-use approach — screening for breast cancer and assessing cardiovascular risk from the same mammogram images — performed “just as well as individual screening” in trials and could be a cost-effective strategy to broaden cardiovascular screening among women.

Routine mammography practice varies by country. In the United Kingdom, for example, women aged 50 to 71 are typically invited for screening every three years, during which two X-rays are taken of each breast to look for signs of cancer. The Australian study’s authors noted that many women undergo mammography in midlife, a period when cardiovascular risk rises, making the imaging an opportune moment to identify elevated cardiometabolic risk.

The published paper states that a deep learning algorithm using routine mammograms and age “shows promise as a cardiovascular risk prediction tool.” The investigators cautioned that using mammography images to predict cardiovascular risk is a novel application, while observing that machine learning models for cardiovascular risk prediction are increasingly being explored.

The researchers did not propose immediate changes to screening programmes and described the findings as a step toward investigation of broader implementation. They indicated further validation and evaluation of clinical pathways would be needed before the approach could be adopted in routine practice. The study adds to a growing body of research testing whether existing diagnostic images can be repurposed with artificial intelligence to provide additional, clinically relevant information without extra procedures.