On-Device Maternal-Fetal Heart Monitoring From Abdominal ECG Using a NEO-Based Adaptive Predictor

Area of Science:

• Biomedical Engineering
• Signal Processing
• Wearable Technology

Background:

• Non-invasive fetal ECG (FECG) systems typically require external processing, limiting their use in long-term, battery-powered applications.
• Current systems face challenges with high data transmission rates for continuous monitoring.

Purpose of the Study:

• To develop a hardware-efficient, on-device signal processing architecture for real-time fetal and maternal heart rate extraction.
• To eliminate the need for high-data-rate wireless streaming in wearable FECG monitoring.

Main Methods:

• Implemented parallel maternal/fetal band-pass filtering.
• Utilized a nonlinear energy operator-based adaptive predictor for robust R-peak identification.
• Developed a fully on-device signal processing architecture.

Main Results:

• Achieved high fetal R-peak detection performance with an average F1-score >96.0% on public datasets.
• Reduced data transmission by >99.9% by outputting RR intervals instead of raw signals.
• Demonstrated robust real-time R-peak identification for both maternal and fetal signals.

Conclusions:

• The proposed on-device system is feasible for continuous wearable FECG monitoring.
• Eliminating external computation and high-rate wireless transmission enhances practicality for long-term use.
• This approach significantly reduces power consumption and data requirements for fetal monitoring devices.