Cardiac MRI traits linked to cardiomyopathy mutations in phenotype naive carriers

Abstract Hypertrophic cardiomyopathy (HCM) and dilated cardiomyopathy (DCM) are prevalent forms of cardiomyopathies, characterised by changes in cardiac structure and function, associated with an increased risk of atrial fibrillation (AF), heart failure (HF), and sudden cardiac death (SCD). Due to these potential life-threatening consequences close monitoring of genetic mutation carriers of HCM and DCM, is important. However, disease onset in these subjects is highly variable, with a substantial subset of people only developing disease late-in-life, or not at all, making risk stratification challenging. We set out to identify cardiac MRI (CMR) based imaging biomarkers of HCM and DCM carriership in phenotype naïve individuals, which were benchmarked against associations with AF and HF. CMR and whole genome sequencing (n: 42,723) data were sourced from a UK biobank sample free of heart failure at the time of CMR measurements. A previously validated deep learning(DL) network was used to derive 22 CMR traits(1). Adjusted cox regression models were used to determine the CMR associations with the onset of heart failure and atrial fibrillation, providing empirical validation of the DL CMR network. Univariable and multivariable generalised linear models were employed to identify associations with genetic carriership conditional on cardiac risk factors. 14 and 19 CMR measurements were associated with the time until HF (274 cases), AF (704 cases), respectively. These variables included left-ventricle (LV) ejection fraction (EF), LV end systolic volume (ESV), Right-ventricular (RV) EF and right atrial (RA) EF. Using WGS data we identified 154 people with an HCM mutation and 220 with a DCM mutation, where HCM carriership associated with five and DCM with four CMR traits. This included RA-EF (OR 1.26 95%CI 1.10; 1.45), and RV-EF (OR 1.41 95%CI 1.23;1,63) for HCM, and LV-EF (OR 0.72 95%CI 0.62; 0.84), and RV stroke volume (OR 0.75 95CI 0.62; 0.91) for DCM. By combining deep learning of CMR with large sample size WGS we were able to identify cardiac imaging biomarkers of HCM and DCM carriership in apparently healthy subjects. This could prove relevant for early identification and monitoring of phenotype naïve people.