Sequence to Sequence ECG Cardiac Rhythm Classification Using Convolutional Recurrent Neural Networks

Abstract

This paper proposes a novel deep learning architecture involving combinations of Convolutional Neural Networks (CNN) layers and Recurrent neural networks (RNN) layers that can be used to perform segmentation and classification of 5 cardiac rhythms based on ECG recordings. The algorithm is developed in a sequence to sequence setting where the input is a sequence of five second ECG signal sliding windows and the output is a sequence of cardiac rhythm labels. The novel architecture processes as input both the spectrograms of the ECG signal as well as the heartbeats' signal waveform. Additionally, we are able to train the model in the presence of label noise. The model's performance and generalizability is verified on an external database different from the one we used to train. Experimental result shows this approach can achieve an average F1 scores of 0.89 (averaged across 5 classes). The proposed model also achieves comparable classification performance to existing state-of-the-art approach with considerably less number of training parameters.

Fig. 1.

Overview diagram of the proposed model

Fig. 2.

Diagram of the Convolutional Layers used to process each heart beat’s signal

Fig. 3.

Diagram of the convolutional layers used to process a spectrogram of a 5s ECG signal segment

Fig. 4.

Normalized Confusion Matrix for Model A: Co-teaching