Independent component analysis algorithms for non-invasive fetal electrocardiography

Area of Science:

• Biomedical Engineering
• Signal Processing
• Maternal-Fetal Medicine

Background:

• Independent Component Analysis (ICA) is widely used for non-invasive fetal electrocardiogram (NI-fECG) processing, often combined with adaptive algorithms.
• Numerous ICA variants exist, but their suitability for NI-fECG extraction remains unclear.
• Objective evaluation of different ICA methods is crucial for optimizing NI-fECG signal processing.

Purpose of the Study:

• To objectively evaluate 11 variants of ICA methods combined with an adaptive fast transversal filter (FTF).
• To determine the most suitable ICA-FTF combination for extracting NI-fECG signals.
• To assess the performance based on QRS complex detection accuracy and computational speed.

Main Methods:

• Tested 11 ICA variants paired with an adaptive FTF on real-world Labour and Pregnancy datasets.
• Evaluated method efficiency using accuracy (ACC), sensitivity (SE), positive predictive value (PPV), and F1 score.
• Measured computation time to assess the performance-to-speed ratio.

Main Results:

• The combination of FastICA and FTF achieved the highest performance: ACC = 83.72%, SE = 92.13%, PPV = 90.16%, F1 = 91.14%.
• FastICA ranked sixth in speed (0.452 s computation time) but offered the best performance-to-speed ratio.
• The best performing method requires only abdominal signal acquisition, eliminating the need for maternal chest reference signals.

Conclusions:

• The combination of FastICA and adaptive FTF is a highly promising approach for NI-fECG extraction.
• This method offers accurate QRS complex detection with efficient computational performance.
• Future devices could utilize this technique with simplified signal acquisition from the maternal abdomen.