Abstract 12631: Recognizing and Predicting Drug-Induced Type I Brugada Pattern Using ECG-Based Deep Learning

Abstract

Introduction: Brugada syndrome (BrS) has been associated with sudden cardiac death in otherwise healthy subjects and drug-induced BrS accounts for 55-70% of all BrS patients. Aims: This study aims to develop a deep convolutional neural network and evaluate its performance in recognizing and predicting BrS diagnosis. Methods: Consecutive patients that underwent ajmaline testing for BrS following a standardized protocol were included. ECG tracings from baseline and during ajmaline were transformed using wavelet analysis and a DCNN was trained to recognize and predict BrS type I pattern. The resultant network is referred to as BrS-Net. Results: A total of 1188 patients were included of which 361 patients (30.3%) developed BrS type I pattern during ajmaline infusion. When trained and evaluated on ECG tracings during ajmaline, BrS-Net recognized a BrS type I pattern with an AUC-ROC of 0.945 (0.921 - 0.969). The best predictive performance was obtained when predictions for all ECG leads were combined using a machine learning (ML) technique, Figure 1. When trained and evaluated on ECG tracings at baseline, before ajmaline, BrS-Net predicted the development of a BrS type I pattern during ajmaline with an AUC-ROC of 0.805 (0.736 - 0.845). The best predictive performance was obtained when predictions for all ECG leads were combined using a ML technique, Figure 2. Conclusions: BrS-Net, a deep convolutional neural network, can identify BrS type I pattern with high performance. BrS-Net can predict from baseline ECG the eventual development of a BrS type I pattern after ajmaline with good performance in an unselected population.