Ic4FECG: A New Index for Automatic Selection of the Most Relevant Independent Component in Noninvasive Fetal Electrocardiography

Independent Component Analysis is often used to extract fetal ECG (FECG) from maternal abdominal signals, but choosing the correct independent component (IC) has traditionally been empirical and subjective. This study introduces Ic4FECG, a quantitative index for automatically selecting the most relevant IC. The index is based on the assumption, supported by the literature, that the typical Fetal Heart Rate (FHR) is around 140bpm, (RR interval of 428ms), and that deviations reflect noise or maternal contamination. Ic4FECG is defined as Ic4FECG = ||(428ms-μFRR)×σFRR||, where μFRR and σFRR are mean and standard deviation of the fetal RR interval series. Using 36 maternal abdominal recordings from the “NInFEA” database, maternal interference was first reduced with PCA, assuming FECG lies in the lowest 5% variance. ICA then decomposed the residual into 20 ICs. Fetal R-peaks were detected in each IC, and Ic4FECG was computed. The IC with the lowest Ic4FECG was selected, and its FHR (FHRIC) was compared with ultrasound-derived FHR (FHRDUS). Results showed strong agreement with FHRIC = 140 ± 9 bpm, and FHRDUS = 141 ± 8 bpm, and significant correlation (ρ = 0.75, p < 10-8). Ic4FECG appears to be a potentially useful tool for automated selection of the most relevant IC in FECG analysis.