Deep multitask noise estimation for ECG denoising

The electrocardiogram (ECG) is one of the most commonly used tools for assessing various cardiac conditions. This is achieved by analyzing the signal of the electrical activity of the heart, however, these signals are often corrupted by various types of noise, complicating the process of effectively diagnosing patients. This thesis addresses the challenge of denoising ECG signals from common noises like baseline wander (BW), muscular artifacts (MA) and powerline interference (PLI) by introducing a novel deep learning methodology that is based on two key elements: noise estimation and multitask learning (MTL). With noise estimation, the conventional approach of estimating the clean ECG signal from the noisy one is inverted, instead, the model focuses on directly estimating the noises present in the signal. After estimating such noises, they will be subtracted from the noisy ECG to perform denoising. This estimation will be analyzed in different experiments, either by using multiple models, one for each noise source independently, or by applying MTL, which is an approach that leverages learning multiple related tasks at the same time, aiming to enhance the model’s performance and efficiency. These experiments, performed using publicly available data, indicate that the estimation of noises like BW and PLI outperforms the outcomes of tested open-source digital filters. While the multitask approach does not yield substantial differences in noise estimation performance compared to training individual models, it increases efficiency and reduces model storage requirements due to the partial sharing of weights across different tasks. Although the presented methodology does not achieve the denoising results of more complex state-of-the-art models, it outperforms the conventional ECG denoising task employing the same architecture. This work contributes to the field of biomedical signal processing by offering, to my knowledge, a novel approach and laying the groundwork for further refinement and application of noise estimation techniques for ECG denoising.