DOI: 10.1161/circ.148.suppl_1.17603 ISSN: 0009-7322

Abstract 17603: Novel Cardiac Dynamics Modeling of Clinical Cardiac MRIs Can Identify Who Has Clonal Hematopoiesis of Indeterminate Potential With High Accuracy

Emanuele Zappala, James Cross, Anthos Christofides, Carlos Matute Martinez, Ritujith Jayakrishnan, Yunju Im, Ana Ferrigno Guajardo, Shawn Ahn, Alokkumar Jha, Stephanie Halene, Jennifer VanOudenhove, David Van Dijk, Jennifer M Kwan
  • Physiology (medical)
  • Cardiology and Cardiovascular Medicine

Background: Clonal hematopoiesis of indeterminate potential (CHIP) has been shown to be an independent risk factor for heart disease, heart failure and is associated with aging. CHIP is diagnosed by DNA sequencing of blood. It is unclear whether imaging can help predict who has CHIP, which would then be able to help stratify patients at risk for adverse cardiovascular events.

Hypothesis: We investigated whether use of cardiac dynamics modeling on routine cardiac MRI (CMR) would be able to help 1) predict who has CHIP and 2) predict age of the patient

Methods: Whole exome sequencing of peripheral blood DNA and subequent analysis on our somatic variant pipeline using MUTECT2 was used to identify those who had CHIP. Using dynamical modeling based on neural integral equations, dynamic imaging assessment of the cine images of the CMRs were used to identify CHIP and age of the myocardium using both supervised and unsupervised approaches.

Results: We identified 98 unique patients with both CMR and CHIP status and found that 40 (41%) had CHIP. Majority were female 65 (66%) and there was no significant difference in median age between CHIP vs no CHIP patients (69, IQR 62-78 vs 65, IQR 54-72). There were no significant differences in co-morbidities between those with CHIP or no CHIP, including CAD, diabetes, hypertension, heart failure. Dynamic imaging was able to predict CHIP with 74% accuracy (Figure) with an unsupervised approach and above 90% accuracy with a supervised approach. In addition, this technique was able to accurately predict the age of the patient with accuracy within 1 year of the patient’s age (standard deviation of 0.0012 years) at the time of the MRI, using a supervised approach.

Conclusions: Cardiac dynamics modeling is a promising deep learning technique that can both accurately classify CHIP and age. Ongoing work shows promise in predicting the outcome of cardiomyopathy using similar techniques. Ongoing validation is underway using a large cohort from UK Biobank.

More from our Archive