Utilising preoperative pericoronary adipose tissue radiomics to predict improvements in European heart rhythm association symptom scores postatrial fibrillation ablation.

Aim: Atrial fibrillation (AF) recurrence after catheter ablation is clinically challenging; the predictive potential of pericoronary adipose tissue (PCAT) radiomics for symptom improvement remains underexplored. We developed a PCAT radiomics model utilising preablation cardiac computed tomography angiography (CTA) to predict symptom improvement among patients with postoperative AF recurrence.

Materials And Methods: We included 146 patients who experienced AF recurrence after their first radiofrequency ablation procedure. Patients were divided into improvement (n=48) and nonimprovement (n=98) groups based on preoperative and postoperative European Heart Rhythm Association (EHRA) symptom scores, and into training (n=103) and validation (n=43) cohorts (7:3 ratio). In total, 5,064 PCAT radiomics features were automatically extracted from cardiac CTA images taken within 3 days preoperatively. Feature selection (logistic regression analysis, maximum-correlation minimum-redundancy, and genetic algorithms) and classification (logistic regression [LR], random forest [RF], and support vector machine [SVM]) methods were employed to construct the PCAT radiomics predictive model. Its predictive performance was evaluated using receiver operating characteristic curves, calibration, and decision curve analysis (DCA).

Results: Five PCAT radiomics features were associated with EHRA symptom score improvement. The area under the curve (AUC) values of the LR, RF, and SVM models were 0.637, 0.858, and 0.756 in the training cohort, and 0.680, 0.812, and 0.751 in the validation cohort, respectively. The RF model had the highest AUC values. Calibration and DCA indicated good clinical efficacy of the radiomics model.

Conclusion: The RF model based on preoperative PCAT radiomics features predicts EHRA symptom score improvement in patients with postablation AF recurrence.