Seeing the Aging Heart: Multimodal AI Quantifies Cardiac Biological Aging from Angiography, Echocardiography, and ECG

Cardiac biological aging results in vascular, structural, and electrical changes that account for age-related cardiovascular disease. Using techniques such as deep neural networks, Artificial Intelligence (AI) based analysis of signals such as ECGs and echocardiograms has revealed that patients age at different rates, and that differences between AI-estimated age and chronological age ("age gaps") are associated with long-term survival. However, so far, cardiac biological aging has focused on single modality estimation and in addition has not yet been evaluated using coronary angiograms. In this study, we first developed a cardiac age estimation model using coronary angiograms. Then, in a group of 1,345 patients who had an echocardiogram and ECG within one month of their angiogram, we examined how patient survival was associated with age gaps in each modality in isolation (using previously developed models for echocardiograms and ECGs) and then across the three modalities combined. For the average across the three modalities, we observed a hazard ratio (HR) of 2.24 (95% confidence interval: 1.71-2.92) per unit increase in the age gap, which was a marked increase compared to each modality on its own (HR of 1.63, 1.54, and 1.24 for angiograms, echocardiograms, and ECGs respectively). This result demonstrates that the predictive value of AI-estimated cardiac age compounds with additional inputs. While angiograms are not practical for routine monitoring, they serve as proof of concept that richer vascular imaging can enhance biological age prediction. As interventions targeting aging emerge, we will need objective tools to measure their impact. Multi-modal cardiac age may provide a scalable, interpretable marker of cardiovascular aging and possibly even rejuvenation.