Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 221

4.5 Acoustic signals
4.5.1 Phonocardiogram
Phonocardiogram (PCG) is a recording of the acoustic signal produced by the me-
chanical action of the heart. The main frequency range of normal PCG is in the range
of 10-35 kHz; however in case of artificial heart valves frequencies up to 50 kHz may
be recorded. Heart sounds are generated by the vibration of the heart valves during
their opening and closure and by the vibration of the myocardium and the associated
structures. The sound generated by a human heart during a cardiac cycle consists of
two dominant components called first heart sound, S1, and the second heart sound,
S2. The S1 sound corresponds to the QRS complex of the ECG and the second heart
sound, S2, follows the systolic pause in the normal cardiac cycle. In S1 two compo-
nents, M1 and T1, may be distinguished which are due respectively to the closure of
the mitral and of the tricuspid valve. The components of S2 are due to the closure
of the aortic valve (A2) and pulmonary valve (P2). Additionally noise-like sounds
called murmurs caused by the turbulence of the blood flow may be produced. They
appear mostly during abnormal action of a heart. Examples of normal and patholog-
ical PCG are shown in Figure 4.59.
The technique of listening to the heart sounds, called auscultation, has been used
for diagnostic purposes since the 16 th century. However, the human ear is not very
well suited to recognize the several events of short duration occurring in very small
intervals of time, especially since they occur in the low frequency range, where the
sensitivity of a human ear is not very good. Therefore signal processing techniques
are very helpful in PCG analysis.
One of the first steps in PCG analysis is its segmentation. Usually ECG signal
is used for this purpose, however the segmentation may be based on the PCG signal
itself taking into account its time domain features and identification of S1 and S2 [Ari
et al., 2008]. The recognition of S1 and S2 without referring to ECG was reported
by [Ning et al., 2009], who used wavelet transform. PCG signal was found to be a
convenient reference for establishing heart rate in fMRI studies, since PCG, contrary
to ECG, is not disturbed by the electromagnetic fields of the equipment [Becker et al.,
2010].
The PCG signal is non-stationary and consists of short transients of changing
frequency. Hence time-frequency methods are the appropriate tools in its analysis.
The early works on PCG analysis were based mainly on spectrograms calculated by
the Fourier transform. Later the wavelet analysis was introduced to PCG analysis.
The usefulness of different kinds of wavelets (Daubechies, Symlet, biorthogonal) for
PCG reconstruction was tested in [Debbal and Bereksi-Reguig, 2008]. It was found
that the best results were obtained by means of Daubechies db7 wavelet. The authors
reported that the error of reconstruction can be used as a discriminatory parameter
in classifying the pathological severity of the PCG. Daubechies wavelets were also
used in the study of [Ning et al., 2009] for finding the offset of the systolic murmur,