ASW-Net: Adaptive Spectral Wavelet Network for Accurate Fetal ECG Extraction

Area of Science:

• Biomedical Engineering
• Signal Processing
• Fetal Monitoring

Background:

• Noninvasive fetal electrocardiography (FECG) is vital for assessing fetal well-being.
• Extracting FECG from abdominal ECG (AECG) is challenging due to noise and maternal ECG interference.
• Existing methods struggle with signal complexity and aliasing in both time and frequency domains.

Purpose of the Study:

• To develop an advanced deep learning network for accurate FECG signal extraction.
• To address the challenges of noise and maternal ECG interference in noninvasive FECG monitoring.
• To improve computational efficiency and performance in FECG extraction.

Main Methods:

• An Adaptive Spectral Wavelet Network (ASW-Net) was proposed for FECG extraction.
• The network utilizes an adaptive spectral wavelet module for efficient frequency-domain component extraction.
• A residual attention module enhances FECG signal distinction from noise, followed by an inverse spectral wavelet module for reconstruction.

Main Results:

• The ASW-Net demonstrated superior performance in FECG extraction compared to state-of-the-art methods.
• Experiments on benchmark datasets validated the effectiveness of the proposed network.
• The method successfully addressed noise and maternal ECG interference challenges.

Conclusions:

• The ASW-Net provides a robust and efficient solution for noninvasive FECG extraction.
• This advancement holds significant potential for improved fetal health monitoring.
• The proposed network architecture offers a promising direction for complex biomedical signal processing.