ACC

ACC 2025: AI mammograms unveil hidden heart risks

March 26, 2025

AI-powered mammograms can predict future cardiovascular events such as MI, stroke, and heart failure by analyzing breast arterial calcifications (BAC) visible on mammogram images. Researchers will share findings on this dual-purpose screening approach at the upcoming American College of Cardiology's Annual Scientific Session.

Dr. Theo Dapamede and colleagues from Emory Healthcare and Mayo Clinic trained a deep-learning AI model to segment calcified vessels in mammogram images and calculate the future risk of cardiovascular events using EHR data. Unlike traditional mammograms, where BAC is visible but not quantified or reported by radiologists, this AI model automatically analyzes BAC and translates it into a cardiovascular risk score. The model's robustness was enhanced by a large dataset of over 56,000 patients' images and health records from Emory Healthcare between 2013 and 2020, with at least five years of follow-up data.

The AI model effectively characterized patients' cardiovascular risk as low, moderate, or severe based on mammogram images. The rate of serious cardiovascular events increased with BAC level in women <60 years and those aged 60-80, but not in those >80. Women with the highest BAC levels had significantly lower five-year event-free survival rates compared with those with the lowest levels, highlighting the tool's potential for early heart disease risk detection in younger women.

Source:

(2025, March 20). American College of Cardiology. AI-Powered Mammograms: A New Window Into Heart Health. [Press release]. https://www.acc.org/About-ACC/Press-Releases/2025/03/20/09/58/AI-Powered-Mammograms