Rationale & Question: Epicardial adipose tissue (EAT) drives cardiovascular risk through its functional state, not just volume. Clinical imaging assesses only anatomy. We hypothesize that multi-modal integration of cardiac MRI radiomics and plasma proteomics/metabolomics can non-invasively capture EAT’s pro-inflammatory phenotype and improve risk prediction.
Objectives:Develop a machine learning model that fuses CMR radiomics and plasma multi-omics to create a novel “EAT Functional Risk Score.”Identify key proteins, metabolites, and biological pathways (e.g., IL-6 signalling) defining high-risk EAT.Validate the score’s independent, incremental value for predicting major adverse cardiovascular events.
Intended Research: Using the UK Biobank cohort, we will segment EAT from CMR Dixon scans and extract radiomic features. These will be integrated with proteomic/metabolomic data via machine learning to derive the risk score and phenotypes. Differential omics analysis will reveal underlying molecular mechanisms. The score’s predictive power for incident events (e.g., heart failure) will be prospectively validated against traditional risk factors.
Value: This project will translate EAT assessment from anatomy to function, yielding an actionable biomarker for precise risk stratification and mechanistic insights into heart-fat crosstalk.
