Cardiometabolic Imaging, Adipose Tissue, and Vascular Inflammation
We study how adipose tissue — perivascular, epicardial, and systemic — interacts with the cardiovascular system and how these interactions can be measured noninvasively from routine imaging. Using CT-based radiomics and machine learning, our work has established adipose tissue phenotyping as a scalable readout of cardiometabolic risk and early cardiovascular disease.
Digital Multimodality Biomarkers for Valvular Heart Disease
We develop AI-enabled digital biomarkers for aortic stenosis that capture complementary structural, myocardial, and hemodynamic remodeling patterns from routine clinical data. This work aims to enable earlier phenotyping of biologically active disease and improve surveillance and timing of intervention.
Scalable AI-Enabled Screening for Underrecognized Cardiomyopathies
We develop AI tools using ECG and echocardiography to detect transthyretin amyloid cardiomyopathy and hypertrophic cardiomyopathy before overt clinical presentation. These approaches are designed for scalable use across health systems and real-world care settings.
Precision Clinical Trials and Digital Phenomapping
We learn signatures of individualized treatment responses — integrating EHR, imaging, and unstructured clinical data into machine learning frameworks to change how we conduct, design, and interpret cardiovascular clinical trials.
Early Risk Stratification of Heart Failure and Cardiotoxicity
We use AI applied to ECG and echocardiographic data to identify early myocardial injury, heart failure risk, and therapy-related cardiotoxicity. The goal is to support scalable surveillance and earlier preventive intervention in high-risk populations.
Best practices in Cardiovascular AI
We contribute frameworks and checklists for the rigorous development, evaluation, and real-world deployment of cardiovascular AI — emphasizing transparency, equity, and clinical impact over model performance alone.
Featured Works
