Biomedical Engineering Reference
In-Depth Information
two scenarios leading to decrease of power measured on the EEG level. First, the
simplest, is that in the whole patch of cortex the oscillations in a given frequency
band disappeared; perhaps the frequency of oscillation had shifted to some other
frequency band. The second scenario is that, during the baseline time, an ensemble
of neural populations oscillates synchronously in a given rhythm. Then during the
information processing or mental activity the ensemble splits into many smaller en-
sembles, each oscillating with its own frequency and phase. In the superposition of
their activity we would observe a decrease of power of the rhythm in respect to that
which was present during baseline time.
In the following paragraphs we shall review methods developed for quantification
of ERD/ERS. In order to determine the statistically significant ERD/ERS values one
needs an appropriate number of repetitions of the experiment.
4.1.7.3.2 Classical frequency band methods
The standard method for evalua-
tion of ERD/ERS was proposed by Pfurtscheller and Aranibar [Pfurtscheller and
Aranibar, 1979]. It can be described by the following algorithm:
•
Let
x
f
(
t
;
i
)
be the band-pass filtered signal in trial
i
∈{
1
,...,
N
}
, and the base-
line time
t
b
∈{
t
1
,...,
t
k
}
•
Square the filtered signal samples to obtain instantaneous band-power
x
f
(
S
f
(
t
;
i
)=
t
;
i
)
(4.18)
sometimes the mean of the data across trials is subtracted before squaring to
remove the ERP component, giving so called inter-trial variance:
2
S
f
(
t
;
i
)=(
x
f
(
t
;
i
)
−
x
(
t
,
f
))
(4.19)
•
Compute averaged band-power by averaging instantaneous band-power across
trials:
N
i
=
1
S
f
(
t
;
i
)
1
N
S
f
(
t
)=
(4.20)
•
Average the power in the baseline time
1
N
t
∈
t
b
S
f
R
f
=
(
t
)
(4.21)
•
Compute the relative power change
S
f
(
t
)
−
R
f
ERD
/
ERS
(
t
)=
·
100%
(4.22)
R
f
•
Smooth the relative power change by moving average or low-pass filtering or
complex demodulation using the Hilbert transform.
Illustration of the above algorithm is shown in
Figure 4.19
Search WWH ::
Custom Search