A novel multi-step adaptive Kalman filtering method based on dynamic noise estimation for FECG extraction

Area of Science:

• Biomedical Engineering
• Maternal-Fetal Medicine
• Signal Processing

Background:

• Increasing maternal age elevates risks of fetal abnormalities and abortion.
• Current fetal monitoring via Doppler ultrasound is limited for continuous, precise assessment.
• Fetal electrocardiogram (FECG) is crucial for monitoring fetal health but difficult to extract from abdominal signals (AECG) due to noise.

Purpose of the Study:

• To develop a novel and robust method for extracting FECG from AECG.
• To overcome limitations of existing algorithms in noisy environments.
• To enhance the accuracy and reliability of fetal health monitoring.

Main Methods:

• Utilized three adaptive Kalman filters (KF) for FECG extraction.
• Employed the Expectation Maximization algorithm to whiten noise and estimate parameters.
• Introduced a 'measurement time difference' pseudo-measurement and adaptively updated noise covariance matrices (R and Q) using forgetting factors.

Main Results:

• The proposed method demonstrated superior FECG extraction accuracy and quality compared to three other algorithms.
• Validation performed on diverse datasets: FECGSYN, ADFECGDB, FECGDARHA, and custom-collected AECG data.
• Successful extraction from simulated, clinical, and real-world low-cost hardware data.

Conclusions:

• The novel adaptive Kalman filter approach significantly improves FECG extraction from AECG.
• Accurate FECG assessment aids obstetricians in better evaluating fetal physiological status.
• Enhanced monitoring facilitates informed clinical decisions, improving maternal and fetal health outcomes.