Robust adaptive fetal heart rate estimation for single-channel abdominal ECG recording

• Published in: 2012 5th International Conference on Biomedical Engineering and Informatics pp. 587 - 591
• Main Authors: Li, Guojun, Zeng, Xiaoping, Zhou, Xiaona, Liu, Qilie
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2012
• Subjects: Adaptation models; Dynamical space model; ECG model; Electrocardiography; Fetal heart rate; Fetal heart rate (FHR); Filtering; Interference; Kalman filters; Mathematical models; Monitoring; Noise; Recording; Single Channel; Strong tracking Kalman filter
• DOI: 10.1109/BMEI.2012.6513193
• ISBN: 9781467311830; 1467311839

Although doppler ultrasound is now routinely applied to the detection of fetal heart rate (FHR) for the assessment of perinatal fetal well-being, it is unsuitable for long-term at-home monitoring due to its transmission of ultrasound beam directly to the fetal heart and the requirement of frequent repositioning the ultrasound probe. As a non-invasive diagnostic tool, fetal Electrocardiogram (FECG) recorded on the maternal abdominal wall enables long-term at-home monitoring during both pregnancy and delivery. However, the abdominal FECG possesses a very low signal-to-noise ratio (SNR), in which the maternal ECG (MECG) represents an predominant source of interference. This study focuses on the detection of FHR from abdominal ECG recordings for the application of portable household fetal monitoring. Considering the MECG exhibiting apparently morphological characteristics in time domain, the MECG is first described with a nonlinear dynamical space model. An adaptive strong tracking Kalman filtering algorithm is then proposed to deal with the time-varying impulsive noise like FECG and electromyogram and the nonlinearity of the dynamical model in the approaching of MECG from a single channel abdominal ECG recording. Finally, the FHR can be efficiently obtained by subtracting MECG from the abdominal recoding. Experimental results on both synthetic and realistic abdominal ECG recordings show that the proposed adaptive robust modelbased FHR detection method outperforms the other singlechannel model-based methods and is suitable for the application of portable at-home FHR monitoring due to its low computational complexity and its simple lead configuration.