Radiomic features of peri-left atrial epicardial adipose tissue and atrial fibrillation recurrence after ablation.

OBJECTIVES

This study aimed to establish a prediction model that incorporates the radiomic features of epicardial adipose tissue (EAT) to predict atrial fibrillation (AF) recurrence after ablation.

METHODS

We prospectively enrolled patients with AF who underwent pulmonary CT venography before ablation therapy at two hospitals (470 patients in the internal cohort and 81 in the external cohort) between June 2018 and December 2019. Stepwise regression was used to identify clinically relevant factors, including quantitative EAT and left atrial (LA)-EAT measurements (model 1). The random forest algorithm was used to select the radiomic features of EAT and LA-EAT. A radiomics model predicting AF recurrence within 1 year after ablation was developed using these features (model 2). Subsequently, logistic regression was used to integrate radiomic features with clinical data (model 3).

RESULTS

In total, 551 patients were enrolled (median age: 66 years, IQR: 60-72 years; 340 men), with 145 experiencing AF recurrence within 1 year. Model 2, based on LA-EAT radiomic features, demonstrated significantly better performance than model 1 (clinical predictive factors and LA-EAT volume) for predicting AF recurrence (areas under the curve (AUC): 0.737 vs 0.584 in the external validation cohort). Model 3 exhibited the highest performance (AUC=0.790 in the external validation cohort, sensitivity value=0.800). Additionally, the combined model provided the highest net clinical benefit within a threshold probability range of 0.2-0.4.

CONCLUSIONS

The LA-EAT radiomics model along with LA-EAT volume and clinical risk factors exhibited the highest predictive performance for AF recurrence following ablation therapy.