Maternal ECG-Guided Neural Network for Improved Fetal Electrocardiogram Extraction.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Maternal-Fetal Medicine

Background:

• Acquiring non-invasive fetal electrocardiograms (fECG) faces challenges due to maternal signal interference and hardware limitations.
• Existing signal processing algorithms struggle to effectively separate maternal and fetal ECG signals.
• Accurate fetal monitoring is crucial for assessing fetal well-being during pregnancy.

Purpose of the Study:

• To evaluate and improve signal-processing algorithms for non-invasive fetal ECG extraction.
• To propose a novel learning-based method integrating maternal ECG as guidance for fetal ECG signal extraction.
• To enhance the accuracy of transabdominal fetal ECG signal acquisition.

Main Methods:

• Development of a learning-based signal processing approach.
• Integration of maternal electrocardiogram (mECG) signals as input for neural network training.
• Comparison of a network trained with mECG guidance versus one trained solely on abdominal ECG data.

Main Results:

• The proposed learning-based method incorporating maternal ECG as guidance demonstrated superior performance.
• Neural networks trained with integrated maternal ECG input outperformed those trained only on abdominal ECG.
• Maternal ECG serves as an effective prior for attenuating maternal signals in abdominal ECG recordings.

Conclusions:

• Leveraging maternal ECG as guidance is a promising strategy for improving non-invasive fetal ECG extraction.
• The proposed method offers a more accurate and reliable approach to transabdominal fetal ECG signal acquisition.
• This advancement has the potential to enhance prenatal monitoring and fetal health assessment.