Electrocardiogram generation with a bidirectional LSTM-CNN generative adversarial network

Fei Zhu,Bairong Shen,Quan Liu,Yuchen Fu,Fei Ye

doi:10.1038/s41598-019-42516-z

Fei Zhu, Bairong Shen + Show 3 more

Open Access

https://doi.org/10.1038/s41598-019-42516-z

Copy DOI

Journal: Scientific Reports	Publication Date: May 1, 2019
Citations: 145	License type: open-access

Affiliation: Soochow University, Sichuan University

Abstract

Heart disease is a malignant threat to human health. Electrocardiogram (ECG) tests are used to help diagnose heart disease by recording the heart’s activity. However, automated medical-aided diagnosis with computers usually requires a large volume of labeled clinical data without patients' privacy to train the model, which is an empirical problem that still needs to be solved. To address this problem, we propose a generative adversarial network (GAN), which is composed of a bidirectional long short-term memory(LSTM) and convolutional neural network(CNN), referred as BiLSTM-CNN,to generate synthetic ECG data that agree with existing clinical data so that the features of patients with heart disease can be retained. The model includes a generator and a discriminator, where the generator employs the two layers of the BiLSTM networks and the discriminator is based on convolutional neural networks. The 48 ECG records from individuals of the MIT-BIH database were used to train the model. We compared the performance of our model with two other generative models, the recurrent neural network autoencoder(RNN-AE) and the recurrent neural network variational autoencoder (RNN-VAE). The results showed that the loss function of our model converged to zero the fastest. We also evaluated the loss of the discriminator of GANs with different combinations of generator and discriminator. The results indicated that BiLSTM-CNN GAN could generate ECG data with high morphological similarity to real ECG recordings.

Highlights

Cardiovascular diseases are the leading cause of death throughout the world
When machine learning approaches are applied to personalized medicine research, such as personalized heart disease research, the ECGs are often categorized based on the personal features of the patients, such as their gender and age
We build up two layers of bidirectional long short-term memory (BiLSTM) networks[12], which has the advantage of selectively retaining the history information and current information

Summary

Introduction

Cardiovascular diseases are the leading cause of death throughout the world. Approximately 32.1% of the annual global deaths reported in 2015 were related with cardiovascular diseases[1]. Many machine learning techniques have been applied to medical-aided diagnosis, such as support vector machines[4], decision trees[5], random conditional fields[6], and recently developed deep learning methods[7] Most of these methods require large amounts of labeled data for training the model, which is an empirical problem that still needs to be solved. Cao et al designed an ECG system for generating conventional 12-lead signals[10] Most of these ECG generation methods are dependent on mathematical models to create artificial ECGs, and they are not suitable for extracting patterns from existing ECG data obtained from patients in order to generate ECG data that match the distributions of real ECGs. The generative adversarial network (GAN) proposed by Goodfellow in 2014 is a type of deep neural network that comprises a generator and a discriminator[11]. The results indicated that our model worked better than the other two methods, the deep recurrent neural network-autoencoder (RNN-AE)[14] and the RNN-variational autoencoder (RNN-VAE)[15]

Methods

Results

Conclusion