End-to-end trained encoder–decoder convolutional neural network for fetal electrocardiogram signal denoising

Abstract

published version features the final layout of the paper including the volume, issue and page numbers. Link to publication General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the "Taverne" license above, please follow below link for the End User

Multi-Channel Fetal ECG Denoising With Deep Convolutional Neural Networks

“…However, the network suppressed some morphological characteristics in cases there was not sufficient content for denoising i.e., when most signal channels were severely corrupted. The multi-channel network outperformed the single-channel (21) in cases of low SNR of the input signals, while for SNR more than 11 dB the single-channel network exhibited slightly better performance. This behavior could be anticipated.…”

“…As mentioned in the introduction and also in (21), the shortcoming of denoising single-channel fetal ECG with a convolutional network is that the network can output signals that look as if they were ideally denoised, but that can have "fake" waves that can differ both in location and polarity when compared to the actual ECG waves. This happens mostly when the quality of the input signals is relatively low and the network, not having enough signal information, reconstructs a clean signal from unreliable information in the encoded latent space.…”

“…Our method was evaluated in comparison with 3 other ECG denoising methods. The first method is the single-channel CNN denoising network, where each fetal ECG channel is denoised separately (21). The second method is a wavelet denoising algorithm that removes the noise by thresholding the detail coefficients after the signal decomposition.…”

“…However, they did not extract the fetal ECG from the preprocessed signals to ensure that it is not suppressed by the network. The authors were the first to propose a deep convolutional encoder-decoder network for postprocessing non-invasive single-channel fetal ECG (20,21), achieving a substantial quality improvement of the noisy signals. The method tackled some of the shortcomings of the state-ofthe-art non-invasive fetal ECG postprocessing methods, since it can preserve beat-to-beat morphological variations and does not require prior knowledge about the location of the fetal pulses.…”

A semi-analytical solution and AI-based reconstruction algorithms for magnetic particle tracking

“…Vincent et al [32] constructed a denoising auto-encoder (DAE) neural network, aiming to find robust representations of features from noisy input data. Subsequent works are dedicated in optimizing deep neural network (DNN) structures to achieve better performance in handling complex noise and interference [33][34][35][36] [37]. Another popular trend is to utilize RNN to preserve historical information and temporal coherence while denoising, which is effective when handling sequential data, e.g., time series [38].…”