Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 138

4.1.6.5 EEG in monitoring and anesthesia

4.1.6.5.1 Monitoring brain injury by quantitative EEG The monitoring tech-
nique of brain injury (which may be, e.g., a result of cardiac arrest) is based on the
hypothesis that the injury causes unpredictable changes in the statistical distribution
of EEG, which reduce the information content of the signal. The impairment of brain
function decreases its ability to generate complex electrophysiological activity, lead-
ing to a reduction of the EEG entropy. Therefore the measure which is commonly
used in monitoring brain injury is entropy (equation 2.28, Sect. 2.3.1.2 ).
Since the EEG of an injured brain is usually non-stationary, the time dependent
entropy based on sliding window technique, was introduced, e.g., [Bezerianos et al.,
2003]. In this technique the entropy is calculated within each window and its time
course is followed.
The alternative measure called information quantity (IQ) is based on discrete
wavelet transform DWT. In this approach the DWT is performed within each moving
window and IQ is calculated from probability distribution of wavelet coefficients. In
this way subband information quantity ( SIQ ) may be found [Tong et al., 2004].
The probability p n (

is obtained from the probability distribution of the wavelet
coefficients in the k th subband; m is the bin number in the probability histogram.
Subband information quantity in k th

m )

subband and in the epoch n may be expressed

as [Thakor et al., 2009]:

M
m = 1 p n ( m ) log 2 p n ( m )

SIQ k ( n )= − (4.4)

In this way the time evolution in the subbands may be followed. It was found that SIQ
depends on frequency band [Thakor et al., 2009]. Entropy or SIQ are early markers
of brain injury and measure of recovery after ischemic incident.

4.1.6.5.2 Monitoring of EEG during anesthesia EEG monitoring is a routine
aid to patient care in the operating room. It is used to reliably assess the anes-
thetic response of patients undergoing surgery and predict whether they respond to
verbal commands and whether they are forming memories, which may cause unin-
tentional unpleasant recall of intraoperative events. The algorithms working in both
domains—time and frequency—were developed for evaluation of EEG during anes-
thesia. In this state patients may develop a pattern of EEG activity characterized by
alternating periods of normal to high or very low voltage activity. The phenomeno-
logical parameter in time domain called burst suppression ratio (BSR) was intro-
duced to quantify these changes [Rampil, 2009]. BSR is calculated as the fraction
of the epoch length where EEG meets the criterion: occurrence of epoch longer than
0.5 s with the amplitude not exceeding 5 μ V.

Spectral analysis is used in anesthesia monitoring as well. Specific parameters
based on calculation of spectral density function were introduced, namely: median
power frequency (MPF)—frequency which bisects the spectrum, with half the power
above and the other half below—and spectral edge frequency (SEF) the highest fre-

±