Radiomic phenotype of peri-coronary adipose tissue as a potential non-invasive imaging tool for detecting atrial...

Epicardial adipose tissue (EAT) contributes to atrial fibrillation (AF). We sought to explore the role of fat attention index (FAI), volume, and fat radiomic profile (FRP) of peri-coronary artery adipose tissue (PCAT) on coronary computed tomography angiography (CCTA) in determining the presence of AF and differentiating its types.

This study enrolled 300 patients who underwent CCTA retrospectively and divided them into AF (n = 137) and non-AF (n = 163) groups. The imaging parameters of FAI, volume, and FRP were excavated and measured after PCAT segmentation. Every coronary artery extracted 853 radiomics and a total of 2559 radiomics were collected. Significant and relevant FRP was screened by random forest algorithm based on machine learning, and then 3 models-VF (FAI and volume), FRP, and FRPC (FRP and clinical factors)-were then compared. Among AF individuals, the FRP and FRPC scores of persistent AF (PerAF, n = 44) and paroxysmal AF (PAF, n = 93) were compared with boxplot.

In the test cohort, FRP score demonstrated excellent distinctive ability in identifying AF, with an area under the curve (AUC) of 0.89, compared with the model incorporating FAI and volume (AUC = 0.83). The FRPC model, which combined FRP with clinical factors, showed an improved AUC of 0.98. Among AF types, FRP and FRPC scores are significantly higher in the PerAF than PAF patients (P < .001) and 20 most contributive features were selected in identifying AF.

Textural radiomic features derived from PCAT on coronary CTA detect micro-pathophysiological information associated with AF, which may help identify and differentiate AF and provide a hopeful imaging target.

The analysis of epicardial tissue around coronary arteries helps identify and differentiate atrial fibrillation and its types. Fat radiomic profiles derived from peri-coronary arteries fat could provide a non-invasive tool for atrial fibrillation.