Monitoring Neurological Injury by qEEG - Quantitative EEG Analysis Methods and Clinical Applications - page 176

Neurologic deficit score: 46 Neurologic deficit score: 74

36

36

34

31

2

5

14

0.50

17

0.22

0.067
0

0.064

50 100 150 0 50

Time (minutes)

100

Time (minutes)

Isoelectric phrase Fast progression Slow progression

Figure 7.4 IQ characteristic comparison of poor and good outcomes after 7-minute injury. The
small figure inside each figure is a compressed EEG. We quantify IQ evolution from various perspec-
tives mainly in three different phases: isoelectric phase just after CA injury, fast increase, and slow
increase phases.

Indeed, there is a very good relationship between the IQ levels obtained and the
eventual outcome of the animal as assessed by the neurological deficit scoring (NDS)
evaluation [39, 43, 53, 54]. A low NDS value reflects a poor outcome and a high
NDS a better outcome. As seen in Figure 7.4, the IQ level recovery takes place faster
and equilibrates to a higher level for the animal with the greater NDS. What we dis-
covered is that the recovery patterns are quite distinctive, with periods of
isoelectricity, fast progression, and slow progression. In addition, in the poor out-
come case, there is a period of spiking and bursting, while in the good outcome case
there is a rapid progression to a fused, more continuous EEG.

7.3.2 Subband Information Quantity
Although IQ is a good measure of EEG signals, it has the limitation that EEG recov-
ery in each clinical band ( ) is not characterized [55]. Therefore, we extend

the IQ analysis method and propose another measure that separately calculates IQ
in different subbands (

δ , θ , α , β , γ

)? This subband method, SIQ, is similar to IQ but
separately estimates the probability in each subband. The probability that pm
n

δ , θ , α , β , γ

()in
the k th subband for that the sampled EEG belongs to the interval I m is the ratio
between the number of the samples found within interval I m and the total number of
samples in the k th subband. Using pm
n

k

( , SIQ k ( n ) in k th subband is defined as

k

M
=− ∑

() () ()

k k k

IQn pm log 2 pm (7.7)

n n

m 1

Thus, we can now evaluate the evolution of SIQ for the whole EEG data { s ( i ), for

i 1, …, N }. Figure 7.5 clearly indicates that recovery differs among subbands. The

subband analysis of signal trends might lead to better stratification of injury and
recovery and identification of unique features within each subband. This wavelet

=