Simulation of ECG for Cardiac Diseases Using Generative Adversarial Networks

Every year, the fundamental technology related to deep learning evolves. Recently, remarkable progress has been made not only in the fields of classification and regression, but also in the field of generation. To date, various models have been proposed for generative models using deep learning, including generative adversarial networks (GAN) and variational auto-encoder (VAE). In this study, we attempted to simulate an electrocardiogram (ECG) using GAN. In addition, ECG may have various states simultaneously, such as AV block and WPW syndrome. Therefore, we propose a method for generating ECGs that considers the fact that multiple states exist simultaneously. The generated ECG validated the basic elements that compose an ECG, such as R and T waves. We demonstrated that AI system can be applied to numerical simulations of bio-signals such as time sequences measured by 3D motion capture, ECGs, and electrogastrograms (EGGs). Furthermore, we conducted experiments to study the effects of stereoscopic video clips on the elderly.