Denoising ECG signals and their analysis using Hybrid Deep learning model

Abstract

Proactive illness diagnosis with AI and related technologies has been an intriguing and productive field in the last ten years. Cardiovascular illnesses are among the medical conditions that need regular monitoring. Arrhythmia, a type of coronary heart disease, is frequently observed by clinicians using electrocardiography (ECG). In humans, an ECG records electrical activity and cardiac rhythm. In recent decades, there has been a substantial surge in the use of neural networks to detect cardiovascular abnormalities. It has been shown that using the denoised signal as compared to the raw input signal increases the probability of better identification of arrhythmias. In this paper, rigorous, three-step preprocessing is done to improve classification accuracy. Firstly, denoising is done using a wavelet transform, then, for baseline artifact filtering, five filters have been applied to ECG signals, and lastly, an R peak is detected. A hybrid (CNN+LSTM) model is implemented to automate arrhythmia categorization on a denoised ECG signal. Comparative analysis demonstrates that the suggested model outperforms contemporary models in terms of various performance factors.