Advertisement

DH-575-01 MACHINE LEARNING-ENABLED MULTIMODAL FUSION OF INTRA-ATRIAL AND BODY SURFACE SIGNALS IN PREDICTION OF ATRIAL FIBRILLATION ABLATION OUTCOMES

      Background

      Machine learning (ML) is a promising approach to personalize atrial fibrillation (AF) management strategies for patients after catheter ablation. Prior studies applied classical ML methods to clinical scores, and none have leveraged intracardiac electrograms (EGM) or 12-lead electrocardiograms (ECG) for outcome prediction.

      Objective

      We aimed to show that (a) ML models trained on EGM or ECG can better predict patient outcomes after AF ablation than existing clinical scores and (b) fusion of EGM, ECG, and clinical features can further improve the prediction performance.

      Methods

      Consecutive patients who underwent catheter ablation between 2015-2017 with panoramic left atrial EGM prior to ablation and clinical follow-up for at least one year following ablation were included. A convolutional neural network (CNN) and a fusion framework were developed for predicting 1-year AF recurrence after catheter ablation from EGM, ECG, and clinical features. The models were trained and validated using 10-fold cross-validation.

      Results

      156 patients (64.5±10.5 years, 74% male, 42% paroxysmal) were analyzed. Using EGM alone, the CNN achieved an Area Under the Receiver Operating Characteristics Curve (AUC) of 0.73, outperforming existing APPLE (AUC=0.63) and CHA2DS2-VASc scores (AUC=0.62). Similarly using 12-lead ECG alone, the CNN achieved an AUC of 0.77. Combining EGM, ECG, and clinical features, the fusion model achieved an AUC of 0.87, outperforming single and dual modality models.