A Hybrid GCN-LSTM Model for Ventricular Arrhythmia Classification Based on ECG Pattern Similarity
Accurate differentiation between Ventricular Tachycardia (VT) and Ventricular Fibrillation (VF) is essential in the field of cardiology. Recent advancements in deep learning have facilitated automated arrhythmia recognition, surpassing traditional electrocardiogram (ECG) methods that depend on manual feature extraction. Building on our previous work, which emphasized the importance of identifying patterns of regularity, we have developed a model that merges Graph Convolutional Networks (GCN) with Long Short-Term Memory (LSTM) networks. This GCN-LSTM model employs a trainable weighted ϵ-neighborhood graph to capture the similarity among time series within ECG segments. This approach has demonstrated substantial improvement in the classification of VT, VF, and non-ventricular rhythms.