A Novel Approach to Fetal ECG Extraction Using Temporal Convolutional Encoder-Decoder Network (TCED-Net).

Area of Science:

• Biomedical Engineering
• Signal Processing
• Maternal-Fetal Medicine

Background:

• Extracting fetal electrocardiogram (fECG) signals from maternal abdominal recordings is challenging due to interference from the maternal ECG.
• Accurate fetal ECG analysis is crucial for monitoring fetal well-being and detecting potential complications.

Purpose of the Study:

• To develop and validate a novel method for extracting weak fetal ECG signals from mixed abdominal ECG recordings.
• To provide a basis for accurate fetal heart rate estimation and fetal ECG morphology analysis.

Main Methods:

• A temporal convolutional encoder-decoder network (TCED-Net) was trained to model the nonlinear transmission of maternal ECG from chest to abdomen.
• The maternal ECG component in the abdominal signal was estimated and subtracted to isolate the fetal ECG component.

Main Results:

• The proposed TCED-Net approach achieved superior performance on the FECGSYN dataset, with an F1 score of 98.94%, MSE of 0.18, and qSNR of 8.30.
• On the NI-FECG dataset, the method effectively suppressed maternal ECG components, yielding clearer and more accurate fetal ECG signals.
• Compared to existing algorithms, the proposed method demonstrated superior ability in extracting clearer fetal ECG signals.

Conclusions:

• The developed TCED-Net based method offers a significant advancement in fetal ECG signal extraction.
• This technique holds substantial application value for effective fetal health monitoring during pregnancy.
• The improved fetal ECG extraction facilitates more reliable fetal heart rate estimation and morphological analysis.