Automatic Detection and Classification of 12-lead ECGs Using a Deep Neural Network

The objective of the PhysioNet/Computing in Cardiology Challenge 2020 is to identify clinical diagnoses from 12-lead ECG recordings. We developed an end-to-end deep neural network model to classify 27 scored clinical diagnosis from Electrocardiogram (ECG). The Squeeze and Excitation (SE) layer, which can explicitly model channel-interdependencies within modules and selectively enhance useful features and suppress less useful ones, and ResNet are integrated into a deep neural network, which is called SE-ResNet34 in our paper. We use the one- dimensional convolution to extract the features among different 12-lead ECG channels and the convolution network is a standard 34-layers ResNet. Finally, we also concatenate some demographic features from the ECGs and the deep features from the SE-ResNet34 to identify clinical diagnosis. The evaluation metrics is calculated, which assigns different weights to different classes, according to the similarity between different classes. Our team named PALab ranked 10 out of 41 teams in the official ranking and achieved a challenge validation score of 0.653 and full test score of 0.359. If confirmed in clinical settings, this approach could reduce the rate of misdiagnosed computerized ECG interpretations and improve the efficiency of expert human ECG interpretation.