AI-Enabled Echocardiographic Biomarkers and Real-World Data for Predicting Cancer Therapy-Related Cardiac Dysfunction

About This Grant

Abstract: Cancer survivors have up to a 42% higher risk of cardiac dysfunction than non-cancer populations, in part due to sequelae of cardiotoxic cancer therapies—collectively termed cancer therapy–related cardiac dysfunction (CTRCD). Yet current surveillance strategies suffer from low sensitivity, poor reproducibility, and lack of personalization. This proposal aims to transform cardio-oncology care by integrating FDA-cleared artificial intelligence (AI) echocardiographic biomarkers with real-world electronic health record (EHR) data to improve early detection and risk stratification of CTRCD. Leveraging a diverse cohort of over 34,000 patients treated with cardiotoxic therapies within an integrated healthcare system with comprehensive data, we will apply validated AI models to archived echocardiograms to identify subclinical cardiac dysfunction and develop predictive models that guide personalized surveillance. Specific aims include: (1) evaluating AI-derived biomarkers for early and standardized detection of CTRCD, and (2) developing and validating multimodal predictive models combining AI-echo and EHR data. This work will generate regulatory-grade evidence to inform FDA guidance, improve clinical outcomes, and establish a scalable framework for cardio-oncology surveillance in real-world settings.