Information Technology Reference
In-Depth Information
Table 4.1: Clinical data of probable CJD patients
Patient
Gender
Age at onset
Disease onset
Clinical presentation
Original EEG report
1
M
73 y/o
6
Memory impairment
PLED, DBS 7 Hz
2
M
74 y/o
9
Memory impairment
FIRDA, DBS 7-8 Hz
3
M
85 y/o
4
Memory impairment
PLED, DBS 6-7 Hz
4
M
52 y/o
5
Memory impairment
DBS 4-5 Hz
5
M 80 y/o 3 Memory impairment Periodic epileptiform
Disease onset time: weeks before the first admission. DBS: diffuse background slowing
6, 9, 4, 5, 3 weeks, respectively, before the first EEG recording. The EEGs were ac-
quired using a 19-channel Nicolet EEG system (digitized at 250 Hz) with Ag/AgCl
surface electrodes, which were placed based on the configuration of the international
10-20 system (Fig. 4.1c). We used the referential montage, rather than the bipolar
or standard EEG, because the EEG signals can be expressed as
X
Ref
so
that the mixing matrix can be obtained directly from FastICA (the
Ref
term was
eliminated in the zero-mean preprocessing of FastICA). The use of bipolar montage
would make the recovery of the mixing matrix much more difficult since the bipo-
lar EEG signals are formulated as
X
=
AS
−
=(
A
1
−
A
2
)
S
with the additional constrain
A1
. Five-minute EEG recording was clipped for each subject,
which was bandpass filtered between 0.5 and 10 Hz prior to the ICA process. In this
study, the infinite impulse response (IIR) digital filter was designed based on the
Butterworth magnitude response:
(
i
,
j
)=
A2
(
i
,
j
+
1
)
1
P
L
(
Ω
)
=
2
L
,
1
≤
L
,
(4.1)
√
1
+
Ω
where
was the analog frequency and
L
was the order of the normalized low-pass
analog filter [21]. Furthermore, the associated s-plane poles were given by
Ω
exp
j
(
2
k
+
L
−
1
)
π
s
k
=
,
1
≤
k
≤
2
L
.
(4.2)
2
L
The bandpass filtering of the EEG was performed by the 6th-order high-pass fil-
ter followed by the 16th-order low-pass filter, which were implemented using MAT-
LAB build-in functions.
Figure 4.2b displays a 15-s waveform of the 17-channel EEG (excluding two
referential electrodes, Ref1 and Ref2) from one patient. We selected several time
points at which the negative peaks or positive peaks (Figs. 4.2b, 4.3b, and 4.4b)
are in conjunction with the corresponding topographic maps (Figs. 4.2c, 4.3c, and
4.4c) which may possess some physiological meanings. However, due to the mix-
ture of source signals, such as disease-related waveforms, environmental noises, and
eye-movement artifacts, the disease-related compartments can be barely discerned
either from the waveforms or from the topographic maps. It should be noted that
the 15-s time windows showed in the Figs. 4.2, 4.3, and 4.4 were selected merely to
