# Radiomic Assessment of Epicardial Adipose Tissue for the Prediction of Non-Calcified Coronary Atherosclerotic Plaques

## TL;DR

This study explores how fat around the heart can predict dangerous artery plaques using imaging and machine learning.

## Contribution

The study introduces radiomic analysis of epicardial adipose tissue as a novel biomarker for predicting severe coronary plaques.

## Key findings

- EAT features correlated significantly with non-calcified plaques causing severe stenosis.
- An EML model achieved high sensitivity and accuracy in predicting plaque severity.
- EAT evaluation via CCTA shows clinical utility for patients with low to intermediate cardiovascular risk.

## Abstract

Epicardial adipose tissue (EAT) has previously been associated with coronary artery calcium scores, an increased burden of coronary artery disease (CAD), and features of plaque instability. These associations are likely mediated by endocrine and paracrine signaling from bioactive molecules secreted by EAT, which may contribute to coronary atherosclerosis. EAT can be non-invasively quantified on images obtained during coronary computed tomography angiography (CCTA). This study aimed to evaluate the potential association between EAT and non-calcified coronary plaques with severe stenosis using radiomic methodology. Materials and Methods: A total of 128 consecutive patients undergoing CCTA—both with and without contrast—for known or suspected CAD were retrospectively analyzed. EAT features were extracted from contrast scans. Coronary artery plaque features were evaluated using Coronary Artery Disease-Reporting and Data System (CAD-RADS). Results: EAT features showed a statistically significant positive correlation with non-calcified coronary plaques with severe grades of stenosis (CAD-RADS > 4). The Ensemble Machine Learning (EML) model combined with coronary plaque data showed a sensitivity of 1.00 and a specificity of 0.93, with a negative predictive value of 1.00 and a positive predictive value of 0.85, and an accuracy of 0.95 (95% CI: 0.9221–1) in internal validation. Conclusions: EAT may represent a novel imaging biomarker associated with the presence of actionable coronary plaques. Radiomic texture analysis of EAT could enhance the non-invasive prediction of coronary stenoses. These preliminary findings support the clinical utility of EAT evaluation via CCTA in patients with low to intermediate cardiovascular risk.