Electrocardiogram
193
Since ECG has a large statistical basis and rarely perfectly reproducible, the
direct analytical schemes based merely on sine and cosine transformations are not
adequate. As a result, it is often the case that raw ECG recordings, assuming ergodicity, and the correlation function are first estimated and then the power spectrum
of the signal is used for the frequency analysis. The details of this process were
discussed in Part I of the book.
Several frequency effects measured by the electrodes are not necessarily associated with the true frequency spectrum of the cellular activity. Many have been
caused by different sources of noise such as the breathing signals and the electromyography (EMG) of other skeletal muscles. Such noises are often filtered in the
frequency domain. For instance, a notch filter at the main power frequency is always
applied to filter out equipment and capacitive noise.
As mentioned earlier, one of the obvious skeletal motion artifacts are the breath
signals, which have both the chest and diaphragm muscles involved. Both groups of
muscles are large muscles, and the EMG resulting from the muscular excitation produces large signal amplitudes. Since the breathing motion is often a regular process
and has much slower rate than the heart rate, it may be easily distinguished. However,
the electrode motion resulting from the breathing action has a modulation effect on
the spectrum content of the measured signal. An effective method of filtering these
types of electric interference signals out is by triggering the measurement on the
motion that is causing the electric noise. In addition, the inhale and exhale process
can be monitored by flow probes and subtracted from the detected ECG signal.
9.5.3 WAVELET-DOMAIN ANALYSIS
Since action potentials are mainly stochastic in nature, wavelet analysis of a single
action potential may not provide the reliable data required for accurate diagnosis.
However, when observing a repetitive signal (such as ECG) as a resultant of the
summation of many action potentials, the wavelet-domain features can identify the
relative contributions of the higher frequencies (lower scales).
The wavelet features used in the analysis of ECG often detect the existence of a
scaled or shifted version of a typical pattern or wave. Wavelet decomposition using
mother wavelet resembles the general shape of the QRS complex, which reveals the
location, the amplitude, and the scaling of the QRS pattern quantitatively. Wavelet
analysis is also performed using Daubeches and Coiflet wavelets.
A typical application of the wavelet analysis is the separation of the mother’s and the
baby’s ECG. As mentioned earlier, the waveform of the fetal ECG is similar to that of
the adult ECG in the wavelet transform (WT) domain, except for the scale of the signal.
The wavelet decomposition of the observed signal can effectively separate the mother’s
ECG from the baby’s, simply because the two ECG signals reside on different scales.
9.6 SUMMARY
In this chapter, the function and structure of the heart as well as the origin of the
ECG signal are briefly described. This chapter also introduces a number of cardiovascular diseases that can be diagnosed using ECG. The processing methods commonly used for processing of ECG are also covered in this chapter.
193
Since ECG has a large statistical basis and rarely perfectly reproducible, the
direct analytical schemes based merely on sine and cosine transformations are not
adequate. As a result, it is often the case that raw ECG recordings, assuming ergodicity, and the correlation function are first estimated and then the power spectrum
of the signal is used for the frequency analysis. The details of this process were
discussed in Part I of the book.
Several frequency effects measured by the electrodes are not necessarily associated with the true frequency spectrum of the cellular activity. Many have been
caused by different sources of noise such as the breathing signals and the electromyography (EMG) of other skeletal muscles. Such noises are often filtered in the
frequency domain. For instance, a notch filter at the main power frequency is always
applied to filter out equipment and capacitive noise.
As mentioned earlier, one of the obvious skeletal motion artifacts are the breath
signals, which have both the chest and diaphragm muscles involved. Both groups of
muscles are large muscles, and the EMG resulting from the muscular excitation produces large signal amplitudes. Since the breathing motion is often a regular process
and has much slower rate than the heart rate, it may be easily distinguished. However,
the electrode motion resulting from the breathing action has a modulation effect on
the spectrum content of the measured signal. An effective method of filtering these
types of electric interference signals out is by triggering the measurement on the
motion that is causing the electric noise. In addition, the inhale and exhale process
can be monitored by flow probes and subtracted from the detected ECG signal.
9.5.3 WAVELET-DOMAIN ANALYSIS
Since action potentials are mainly stochastic in nature, wavelet analysis of a single
action potential may not provide the reliable data required for accurate diagnosis.
However, when observing a repetitive signal (such as ECG) as a resultant of the
summation of many action potentials, the wavelet-domain features can identify the
relative contributions of the higher frequencies (lower scales).
The wavelet features used in the analysis of ECG often detect the existence of a
scaled or shifted version of a typical pattern or wave. Wavelet decomposition using
mother wavelet resembles the general shape of the QRS complex, which reveals the
location, the amplitude, and the scaling of the QRS pattern quantitatively. Wavelet
analysis is also performed using Daubeches and Coiflet wavelets.
A typical application of the wavelet analysis is the separation of the mother’s and the
baby’s ECG. As mentioned earlier, the waveform of the fetal ECG is similar to that of
the adult ECG in the wavelet transform (WT) domain, except for the scale of the signal.
The wavelet decomposition of the observed signal can effectively separate the mother’s
ECG from the baby’s, simply because the two ECG signals reside on different scales.
9.6 SUMMARY
In this chapter, the function and structure of the heart as well as the origin of the
ECG signal are briefly described. This chapter also introduces a number of cardiovascular diseases that can be diagnosed using ECG. The processing methods commonly used for processing of ECG are also covered in this chapter.