A Dual Attention-Based Autoencoder Model for Fetal ECG Extraction From Abdominal Signals

Abstract

Fetal electrocardiogram (FECG) signals contain important information about the conditions of the fetus during pregnancy. Currently, pure FECG signals can only be obtained through an invasive acquisition process which is life-threatening to both mother and fetus. In this study, single-channel ECG signals from the mother's abdomen are analysed with the aim of extracting the clean FECG waveform. This is a challenging task due to the very low amplitude of the FECG, various noises involved in the signal acquisition, and the overlap of R waves. To address this problem, we propose a novel convolutional auto-encoder network architecture to learn and extract the FECG patterns. The proposed model is equipped with a dual attention model, composed of squeeze-and-excitation and channel-wise modules, in the encoder and decoder blocks, respectively. It also benefits from a bidirectional long short-term memory (LSTM) layer. This combination allows the proposed network to accurately attend to and extract FECG signals from abdominal data. Three well-established datasets are considered in our experiments. The obtained results of FECG extraction are promising and confirm the effectiveness of using attention modules within the deep learning model. The results also suggest that the proposed auto-encoder network can accurately extract the fetal ECG signals where no information about maternal ECG is available.

Enhancing Fetal Electrocardiogram Signal Extraction Accuracy through a CycleGAN Utilizing Combined CNN–BiLSTM Architecture

“…Notably, models leveraging the CycleGAN as a foundational framework outperform other models, underscoring the high-quality extraction of FECG signals by the GAN. [24] 0.061 ± 0.006 0.019 ± 0.005 90.69 ± 0.17 AEDL [35] 0.059 ± 0.002 0.018 ± 0.003 92.09 ± 0.22 CSGSA-Net [36] 0.057 ± 0.003 0.016 ± 0.002 92.27 ± 0.33 CycleGAN [27] 0.042 ± 0.008 0.011 ± 0.004 92.71 ± 0.29 CAA-CycleGAN [28] 0.024 ± 0.003 0.007 ± 0.002 95.34 ± 0.12 this work 0.019 ± 0.004 0.006 ± 0.002 98.01 ± 0.26 Finally, Figure 10 illustrates an image comprising a unit circle (depicted in red) alongside a 3D trajectory (depicted in blue) generated based on data from ADFECGDB r01. As the trajectory approaches one of the P-QRS-T waves, the 3D trajectory exhibits vertical movement, with the limit ring oscillating up and down.…”

Artificial Intelligence and Machine Learning in Electronic Fetal Monitoring

“…The LSTM combined with the autoencoder (AE) model was tested on real data by Ghonchi et al [112]. AE is a multilayer symmetric network containing an encoder and decoder that was able to improve the feature representation of the input data.…”

A Non-Invasive Fetal QRS Complex Detection Method Based on a Multi-Feature Fusion Neural Network

“…With significant advancements in deep learning technologies, numerous studies have used neural networks to extract fetal ECG signals without separating the maternal ECG [22][23]. Zhong et al employed a one-dimensional convolutional neural network (CNN) to detect fetal QRS complexes without removing the maternal ECG signal, aiming for a similar goal as this study, achieving an accuracy of 75.33%, a recall rate of 80.54%, and an F-1 score of 77.85% [24].…”