Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 195

respond to QRS and high frequency noise. As a threshold for QRS separation the
3-D ellipsoid was defined, which made it possible to determine the fiducial points
of QRS and perform the subtraction of mECG from the original (not filtered) signal.
In the second stage, the denoising was performed by means of the PCA application
to wavelet coefficients (6-levels Daubechies) followed by reconstruction. The proce-
dure of denoising is shown in Figure 4.41 (b). Then in the denoised signal R peaks
of fECG not coincident with maternal R peaks were identified. In order to detect the
superimposed peaks, a histogram of R-R intervals was constructed and the intervals
of double length were considered in the procedure of finding the lost fetal R peaks.
This information was used to construct the final fHRV series.
The above described method was tested on simulated signals with various signal to
noise (SNR) ratios, generated according to [Sameni et al., 2007] and on experimental
signals from the University of Nottingham database [Pieri et al., 2001]. Simulated
data helped to improve the parameters of the ellipsoid used for thresholding, which
resulted in better sensitivity values. The results for experimental time series also
gave high values (around 90%) of selectivity. The described approach was compared
with other methods: time-frequency based methodology [Karvounis et al., 2007],
parabolic fitting [Zhang et al., 2006], template matching [Gibson et al., 1997], fast
ICA algorithm [Hyvarinen, 1999]. The accuracy for real data showed the best results
for the above described method and time-frequency method.

4.2.4 Magnetocardiogram and fetal magnetocardiogram

4.2.4.1 Magnetocardiogram

The magnetic field produced by a heart is around 50 pT, which is more than an
order of magnitude bigger than the one generated by the brain (around 1 pT), which
makes MCG easier to record than MEG. The first magnetocardiogram (MCG) was
measured almost 50 years ago by Baule and Mc Fe by means of a gradiometer using a
set of copper coils with several millions of windings. The era of magnetocardiology
was opened by the introduction of SQUID. At first single sensors were used. At
present commercially available devices include multiple MCG leads and the custom
made systems have up to 128 sensors (the description of the magnetic field sensors
may be found in Sect. 4.1.4).
One of the advantages of MCG is the lack of direct contact of the sensors with
the skin which allows for fast screening of patients and is useful for people with
burns. The same currents generate MCG and ECG signals and their time evolution
is similar, but not identical. Tangential components of the fields are attenuated in the
ECG. Contrary, the ideal magnetic lead is sensitive only to tangential components
of electric sources and hence should be particularly responsive to abnormalities in
activation, since normal activation sources are primarily radial. Another advantage
of MCG is connected with the fact that magnetic permeability of the tissue is that of
free space. This allows to record by means of MCG activity of the posterior side of
the heart. In ECG measurement this activity is strongly attenuated by the resistivity
of the lungs.