Cardiovascular risk from comprehensive evaluation of the CT calcium score exam

Cardiovascular risk from comprehensive evaluation of the CT calcium score exam Summary Using a comprehensive machine learning analysis of coronary artery calcifications and thoracic fat depots in CT calcium score images, we will predict future major adverse cardiovascular events. Improved characteriza- tion of cardiovascular risk will advance knowledge of cardiometabolic disease phenotypes and support clinical therapeutic decision-making and patient counseling for improved adherence. With improved risk prediction and identification of high-risk phenotypes, there will be an opportunity to guide precision preventive therapies, where guidance is needed given the cost and side effects associated with some of these therapies. The Agat- ston calcium score is the leading predictor of a future major adverse cardiovascular event, better than any oth- er single assessment. Epicardial fat volume and HU values are independent risk factors. We will combine analyses of fat and calcifications in CT calcium score exams in an unprecedented way. For coronary calcifica- tions, pathological observations and preliminary results suggest that examining other features (calcium-omics) can improve prediction as compared to whole-heart Agatston. Our assessments will characterize small, spotty, low-density calcifications, providing a better surrogate of disease vulnerability than Agatston, which is numeri- cally dominated by large, likely stable, calcifications. In addition to fat volumes, we will examine quantitative texture and shape features (fat-omics). Elevated HU values and tissue textures are indicative of fat inflamma- tion. All these observations suggest significant value in a combined fat-omics and calcium-omics analysis. We will use large archives of CT calcium score exams from different sites, including University Hospitals of Cleve- land, which is an institution with the largest no-charge CT calcium scoring program (>13,000 scans per year). These big data repositories provide a unique machine-learning opportunity. Numerous technical innovations are planned, including novel features, data representations, and machine learning approaches. In addition to clinical risk prediction, our CT calcium score analyses will dovetail in the future with many research interests, including the role of genes, metabolomics, co-morbidities (e.g., diabetes and psoriasis), socio-economic status, and cardio-oncology on cardiovascular risk.