Electrocardiogram Classification by Modified EfficientNet with Data Augmentation

Abstract

The standard 12-ECG are widely used in the diagnosis of arrhythmias and other cardiac disorders. Early and correct diagnosis of cardiac abnormalities can improve treatment results. However, manual interpretation of the electrocardiogram (ECG) is time-consuming and difficult to be scaled. Thus, automatic detection and classification of cardiac abnormalities can assist physicians in the diagnosis of the growing number of ECGs recorded. In recent years, deep neural networks (DNNs) have shown significant improvement in variety of tasks, including ECG classification. In this study, we attempt to classify 12-ECG PhysioNet/Computing in Cardiology Challenge 2020 data using DNN model. We adopt EfficientNet model, which achieved state-of-the-art result with ImageNet classification task, and modify model for ECG classification. During the training, we adopt data augmentation for ECG to improve the robustness of the model. With training data we achieve score of 0.585 using cross validation, relative improvement of 7.73% over model without data augmentation. We achieved a score of 0.456, but were not ranked due to omissions in the submission (Team name: NN-MIH).