Time course reconstruction of fetal cardiac signals from fMCG: independent component analysis versus adaptive maternal beat subtraction

Abstract

M-mode and pulsed Doppler echocardiography, cardiotocography and transabdominal fetal ECG are available in clinical practice to monitor fetal cardiac activity during advancing gestation, but none of these methods allows the direct measurement of morphological and temporal parameters for fetal rhythm assessment. Fetal magnetocardiograms (fMCGs) are noninvasive recordings of magnetic field variations associated with electrical activity of the fetal heart obtained with superconducting sensors positioned over the maternal abdomen inside a shielded room. Because of maternal cardiac activity, fMCGs are contaminated by maternal components that need to be eliminated to reconstruct fetal cardiac traces. The aim of the present work was to use two methods working in the time domain, an independent component analysis algorithm (FastICA) and an adaptive maternal beat subtraction technique (AMBS), for the retrieval of fetal cardiac signals from fMCGs. Detection rates of both methods were calculated, and FastICA and AMBS performances were compared in the context of clinical applications by estimating several temporal and morphological characteristics of the retrieved fetal traces, such as the shape and duration P-QRS-T waves, arrhythmic beat detection and classification, and noise reduction. Quantitative and qualitative comparison produced figures that always suggested that FastICA was superior to AMBS from the perspective of clinical use of the recovered fetal signals.

Simultaneous monitoring of separate fetal magnetocardiographic signals in twin pregnancy

“…The SNR of the reconstructed fetal signals could be further improved with a smoothing procedure based on a Kaiser filter working with a moving window that maintained the original time resolution (Comani et al 2004d).…”

“…Consequently, fMCG observations need to be processed in order to reconstruct uncontaminated fetal signals that may be suitable for clinical application. To this aim several techniques have been used, such as the correlation method, which subtracts all maternal QRS complexes from fMCG using a spatio-temporal correlation function (Abraham-Fuchs et al 1990), an eigenvector-based spatial filtering, which isolates the signal from the interference identifying their respective subspaces (Chen et al 2001) or the adaptive maternal beat subtraction procedure, which iteratively eliminates the average maternal signal from fMCG recordings (Comani et al 2004d). However, in the majority of cases averaged signals need to be calculated in order to characterize cardiac waves.…”

“…A new procedure, using neural networks, has been proposed by the authors; it is based on an ICA algorithm (FastICA), and it has been successfully applied to reconstruct the time course of fetal magnetocardiographic signals in singleton pregnancies (Comani et al 2004d. The retrieved fetal traces were unaffected by maternal interference, and background noise was strongly reduced, so that a high signal-to-noise ratio (SNR) could be attained.…”

“…Although various groups have reported on dependable fetal signal extraction in singleton pregnancy (van Leeuwen et al 1999, Wakai et al 2000, 2003, Kandori et al 2003, Menendez et al 2001, Hosono et al 2002, Quartero et al 2002, Comani et al 2004a, 2004b, 2004d, 2004e, Grimm et al 2003, Horigome et al 2001, Kahler et al 2002, Lowery et al 2003, Peters et al 2001, only one case of successful separation of fetal magnetocardiographic signals in twin pregnancy has been reported until present; fMCG was recorded using two distinct MCG sensor arrays positioned over the abdominal projections of the twins' hearts previously obtained with ultrasound examination, and the fetal traces were averaged in time before being analyzed (Costa Monteiro et al 2001).…”

Maternal ECG removal from non-invasive fetal ECG recordings

“…Several techniques to extract the fECG from abdominal recordings have been proposed. Commonly, methods based on independent component analysis (ICA) and linear prediction (LP) are used [5] [6]. ICA determines the linear transformation that maximizes the statistical independency of the transformed signals [7].…”