ECG Denoising based on Total Variation Denoising and Deep Learning Convolutional Neural Networks

Electrocardiogram (ECG) signals serve an important role in identifying heart problems, but they are usually distorted by noise, preventing accurate interpretation. This work provides a novel method for denoising ECG signals using a hybrid methodology that combines deep learning and signal processing techniques. Our method entails decomposing noisy ECG signals at different Signal-to-Noise Ratio (SNR) levels, training a Convolutional Neural Network (CNN) model to learn the mapping from Discrete Wavelet Transform (DWT) approximation coefficient of a noisy signal to that of a clean signal at each SNR level, and then applying a Total Variation Denoising (TVD) transform to further refine the denoised results. We compared standalone CNN-based denoising, Total Variation Denoising (TVD) transform, and our suggested approach. Experimental results from benchmark ECG datasets show that our proposed method outperforms the standalone CNN and TV denoising methods in terms of denoising efficacy. Our technique preserves important ECG features while efficiently reducing noise aberrations, demonstrating its potential to improve diagnostic accuracy and patient care outcomes in clinical settings.