Synthesis of Realistic ECG using Generative Adversarial Networks

Abstract

Access to medical data is highly restricted due to its sensitive nature, preventing communities from using this data for research or clinical training. Common methods of de-identification implemented to enable the sharing of data are sometimes inadequate to protect the individuals contained in the data. For our research, we investigate the ability of generative adversarial networks (GANs) to produce realistic medical time series data which can be used without concerns over privacy. The aim is to generate synthetic ECG signals representative of normal ECG waveforms. GANs have been used successfully to generate good quality synthetic time series and have been shown to prevent re-identification of individual records. In this work, a range of GAN architectures are developed to generate synthetic sine waves and synthetic ECG. Two evaluation metrics are then used to quantitatively assess how suitable the synthetic data is for real world applications such as clinical training and data analysis. Finally, we discuss the privacy concerns associated with sharing synthetic data produced by GANs and test their ability to withstand a simple membership inference attack. For the first time we both quantitatively and qualitatively demonstrate that GAN architecture can successfully generate time series signals that are not only structurally similar to the training sets but also diverse in nature across generated samples. We also report on their ability to withstand a simple membership inference attack, protecting the privacy of the training set. Keywords generative adversarial networks • time series synthesis • health data

ECG synthesis with Neural ODE and GAN models

“…The Discriminator for this GAN model is a four-layer 2-dimensional convolutional neural network and a Minibatch discrimination layer. The discriminator model is similar to the discriminator model used in [15]. The discriminator model also has noise added to the gradient of the optimiser to preserve privacy.…”

“…(i) Generate Normal Sinus ECG and (ii) Generate Arrhythmia ECG. The proposed model is evaluated against state art Neural Network used for ECG Synthesis as described in [15] and NeuralCDE.…”

“…Among different deep learning models, GANs are leading the research for ECG data generation. [14][15][16][17] shows how GANs generate ECG data. Among these works, [16,17] shows domain-specific knowledge such as personalisation or external simulation can improve the performance of GANs model.…”

“…But training personalised data with ECG data is time-consuming and needs multiple environments for training. On the other hand, [14,15] adopt the hybrid model concept to design the architecture for GAN. GANs described in [14,15] dynamical equations of motion for ECG signal can be described as ordinary differential equations [18].…”

“…On the other hand, [14,15] adopt the hybrid model concept to design the architecture for GAN. GANs described in [14,15] dynamical equations of motion for ECG signal can be described as ordinary differential equations [18]. Therefore, ODE can model ECG signal as a function of continuous-time.…”