Bio-inspired optimization applied to synthetic ECG models for generating cardiac arrhythmias

Abstract

The Electrocardiogram (ECG) is an essential and straightforward procedure used to detect cardiac abnormalities. With the development of deep learning algorithms, the focus of many works is on the design of automatic ECG disorders detection algorithms, and many of them have achieved very high accuracy. The main challenge is still the same: clinical ECG datasets are expensive, and there are no sufficient expert annotations. This work introduces a new methodology for generating synthetic ECG signals using a mathematical model and a bio-inspired algorithm to estimate the model’s parameters. The objective is to expand the original model to reproduce different cardiac arrhythmias using ECG records as a reference. The model’s parameters are different for each selected ECG wave and are obtained by minimizing the difference between the ECG recordings and the synthetic ones. The results show that the proposed methodology is able to estimate the synthetic ECG model parameters using Gaussian functions for each ECG wave and is highly adaptable to different cardiac diseases.