Multi-channel Electrocardiogram Classification by semantic segmantation methods

Lingfeng LIU, Nan Jiang, Qin Xia
East China Jiaotong University


Abstract

In this paper, we proposed a semantic segmentation neural network design with squeeze-net and bidirectional Long Short-Term Memory cells to realize a heart beat division based multi-channel electrocardiogram automated classification system. Focal loss is introduced to alleviate the label imbalancing. The training of the neural network is also enhanced with random K-fold technique.