Independent component analysis: fetal signal reconstruction from magnetocardiographic recordings

Abstract

Independent component analysis (ICA) was used for the processing of cardiological signals obtained by means of fetal magnetocardiography (fMCG), a technique allowing the non-invasive recording of the weak magnetic field variations associated to the electrical activity of the fetal heart. Purpose of the present work was to verify whether a computational-light ICA algorithm (FastICA), tailored to the characteristics of fMCG, could reconstruct reliable signals of the fetal cardiac activity during the last gestational trimester, when good electrophysiological traces are difficult to obtain although being extremely important for clinical diagnosis of severe fetal dysrhythmias. Several combinations of input recordings and output components were examined in order to assess the best configuration to successfully use FastICA. The reconstructed traces were compared with those obtained with deterministic techniques already used for this purpose, and they showed to be stable and reliable, unaffected by overlapped maternal and fetal beats and suitable for clinical applications.

Linear and nonlinear measures of fetal heart rate patterns evaluated on very short fetal magnetocardiograms

“…FastICA separates signals that originate from statistically independent sources by minimizing their mutual information (Hyvarinen 1999). The fetal and maternal cardiac signals, as well as other physiological and interference signals embedded in the fMCG recordings, are considered statistically independent, and FastICA was demonstrated to have the best performance in fMCG applications (Comani et al 2004, Hild et al 2007). Matched filtering is a temporal filtering technique in which a template of the interfering signal is subtracted from the recording whenever the interfering heartbeat is detected.…”

Characterization of Fetal Arrhythmias by Means of Fetal Magnetocardiography in Three Cases of Difficult Ultrasonographic Imaging

“…Raw fMCG recordings were processed to eliminate the maternal signal and to reconstruct the time course of the electrical activity of the fetal heart only; independent component analysis (ICA) was used for this purpose 7,8,17–23 . ICA permits a completely blind signal sources separation starting from n mixed input signals x j (t) and assuming that they are a linear instantaneous mixture of m independent components s 1 (t), where m must be at most equal to n .…”

“…FMCG traces are morphologically similar to ECG, but they are undistorted by the insulating properties of vernix caseosa and have a signal‐to‐noise ratio (SNR) that increases with the advance of gestation 2,6 . After adequate signal processing, the detection of fetal cardiac signals, appropriate for the estimation of time intervals and P‐QRS‐T waves morphology, increases from 85% at 26 weeks to 100% at 36 weeks 7,8 . This fact is particularly interesting from the perspective of the fetal arrhythmias characterization 9–14 …”

Clinical application of magnetocardiography

“…However, after signal averaging triggered on the QRS peak of the FMCG itself, it is possible to identify even the P-wave on the amplified averaged signal, to calculate MF maps and to localize fetal ventricular activity distribution (FIGURE 11). Further improvement of the signal quality is achieved with the Independent Component Analysis (ICA) approach to reconstruct the fetal signal [211][212][213], making it possible to measure cardiac intervals beat-to-beat and to evaluate heart rate variability parameters, according to the recommended standard for FMCG [206]. As those recordings were performed with a system designed for adult MCG mapping, improved sensitivity could be expected with the development of a sensor geometry tailored to FMCG applications.…”