PO-03-092 MACHINE LEARNING ANALYSIS OF THE PREDICTORS OF AF RECURRENCE POST-ABLATION IN PATIENTS UNDERGOING PRE-ABLATION MRI

Catheter ablation, a widely performed rhythm control strategy for atrial fibrillation (AF), has variable results in the complete suppression of arrhythmia recurrence, especially obese patients. Structural remodeling left atrium (LA), evidenced by LA enlargement and fibrosis, is well-known substrate AF. Epicardial adipose tissue (EAT) thought to be crucial factor mediating cardiovascular disease obesity, it can result electrical heterogeneity local conduction block forming We aimed apply machine learning approach assess predictors AF recurrence following catheter ablation patients undergoing pre-ablation cardiac magnetic resonance (CMR). 101 underwent CMR prior ablation. Fibrotic was assessed using late gadolinium enhancement sequence. EAT novel fat-water separation 3D Dixon Panel A shows posterior anterior views representative model showing fibrotic EAT. Patients were followed ambulatory monitoring 12-lead ECGs. random forest classifier with 5x stratified cross-validation trained predict post-ablation SHapley Additive exPlanations (SHAP) analysis used feature importance prediction recurrence. During an average follow-up period 1 year, occurred 31 (30.7%) had higher index (20.7 [16.9, 30.4] vs 13.7 [10.5, 20.1] mL/m2, p<0.001), volume (66 [52.6, 77.5] 49.9 [37.7, 61.8] p=0.001). There trend toward fibrosis burden who (18.6% 16.1%, p=0.076). The predicted AUC 0.67 ± 0.09 (Panel B). SHAP revealed that overall population, three most important features index, C). enlargement, EAT, markers clear ties pathophysiology, are after Future studies testing value imaging-based failure calculator, as well elucidate mechanism failure, required.