Atrial Fibrillation Detection Algorithm Based on Graph Convolution Network

Atrial fibrillation (AF) is a common type of arrhythmia with high incidence and risk, it difficult to monitor. Deep learning-based algorithms for AF detection have made preliminary progress, Recurrent Convolution Neural Networks (RCNN) being widely used detection. The RCNN algorithm combines ECG morphology rhythm information, where Network (CNN) mainly processes morphological features (RNN) deals information. However, problem this method that direct splicing fusion cannot fully exploit the information both because they different representative spaces. Therefore, paper proposes novel based on graph convolution network, named AF-GCN. First, we transform heartbeats into structure, called heartbeat diagrams. node are each heartbeat, edge linked according time interval between heartbeats. Moreover, calculate weight similarity signal can be unifiedly expressed in one heartbeat. Then, using Graph (GCN), rhythmic effectively extracted, modeled, fused network modal.We evaluated our AF-GCN two popular recognized benchmarks, MIT-BIH AFDB Physionet Challenge 2017 datasets. results show outperforms other existing methods.