Heart Podcast Predicting cardiovascular events from routine mammograms using machine learning
Nov 4, 2025
Dr. Jennifer Barraclough, a Sydney-based interventional cardiologist and cardiovascular researcher at the George Institute for Global Health, discusses groundbreaking research on using mammograms to predict cardiovascular risks. She explains how her team's deep learning model utilizes mammogram images and patient age, opening doors to better cardiovascular disease prevention in women. Barraclough highlights the advantages over traditional methods and emphasizes the importance of flagging high-risk women for early intervention.
AI Snips
Chapters
Transcript
Episode notes
Cardio Risk Often Overlooked In Women
- Cardiovascular disease is the leading killer of women yet remains under-recognised and under-prioritised.
- Mammography is a timely touchpoint to reach peri- and postmenopausal women for cardiovascular prevention.
Whole Breast Architecture Adds Signal
- Breast arterial calcification gives an incomplete picture of cardiovascular risk because it omits key risk associations.
- Whole breast architecture may capture additional predictive signals beyond calcification alone.
Deep Survival Model From Routine Mammograms
- The model converts routine bilateral mammograms into a continuous time-to-event cardiovascular risk score using a deep survival network.
- It combines automated deep features and handcrafted radiomic features to optimise prediction.