104 S
that precedes 3-Hz SWC in patients with absence epilepsy sampled EEG data; two sequences of spindle appear in Fig.
[22]. The investigation was carried out by a simultaneous 1A. We excluded the spindle oscillation with a preceding
analysis of far-field evoked potentials brainstem auditory large negative EEG transient K-complex or vertex sharp
evoked potentials, BAEPs with reference to the paroxys- transient from the analysis. An example of the spindle
mal discharges SWC. oscillation with K-complex is also shown in Fig. 1A
Here, we investigated the brainstem function over the marked with ‘1’. K-complexes can be detected easily,
course of a spindle oscillation before and during with a because they show regular phase reversals between C3–Cz
similar analytic method in humans. Probably, spindle and Cz–C4 according to the distribution pattern in the
oscillations are more suitable candidates for the analysis EEG field [6].
because of their periodic characteristics rather than in- The method for obtaining sequential BAEPs over one
frequent EEG events such as epileptic discharges. spindle oscillation was as follows.
i The EEG data just before one spindle oscillation were selected with a series of 16 segments in a manner that
2. Materials and methods the last segment in the series positioned the first spindle
wave red rectangles in Fig. 1B, ‘onset’. BAEPs were 2.1. Data acquisition
calculated by averaging these EEG data in Cz–A1 and Cz–A2 160 average counts; 16 segments in 10 sequences
A full description for the recording and the analyzing of spindle, and termed as ‘onset’.
device appeared in our previous study [22]. Six-channel ii The average area for ‘onset’ consisting of 16
sleep EEGs Cz–A1 and Cz–A2 for BAEP recording, 1.5 segments was shifted backward by 2 segments red rectan-
Hz–1.5 kHz for a band-pass filter; T3–C3, C3–Cz, Cz–C4 gles in Fig. 1B, ‘B1’ in 10 sequences of spindle, then the
and C4–T4 for EEG monitoring, 1.5–100 Hz for a band- EEG data was averaged. As a segment was defined as a
pass filter were recorded together with one channel of 512-point EEG data with a fixed position of one trigger
continuously delivered acoustic trigger signal 80 dB signal 20 times s repetition rate, the average area was
binaural single rectangular pulse with a 0.1-ms width, 20 shifted backward 0.1 s by this procedure. The calculated
times s of repetition rate from a human subject through a BAEPs were termed as ‘B1’. This averaging procedure
multi-modal EEG device. Twelve male volunteers aged was repeated until the BAEPs termed as ‘B18’ were
from 20 to 27 years were gathered for this study. The obtained.
procedure was approved by the ethical committee of iii The average area for ‘onset’ was shifted forward by
Hokkaido University School of Medicine. The EEG data 2 segments red rectangles in Fig. 1B, ‘A1’, then the EEG
were temporarily stored on a magnetic tape through a PCM data was averaged in 10 sequences of spindle. The
digital data recorder. The recording lasted until a volunteer calculated BAEPs were termed as ‘A1’. This averaging
fell asleep to stage II according to the sleep-stage scoring procedure was repeated until the BAEPs termed as ‘A16’
where sleep spindles appeared constantly [29]. All EEG were obtained.
electrodes were shielded, and the electrode resistance was Following the above procedures, two series left, Cz–
carefully kept below 5 kV throughout the recording. While A1; right, Cz–A2 of sequentially changing BAEPs, with a
the EEG data were replayed, the periods with stable time shift of 0.1 s, were obtained as the spindle wave
spindle oscillations were selected and sampled 10 kHz for gradually emerged in the EEGs. The time range for the
sampling frequency through a personal computer PC analysis spanned from 21.860.4 s ‘B18’; median6range
equipped with an A D converting board. The sampled to 1.660.4 s ‘A16’ through ‘onset’ 060.4 s. The range
EEG data were transferred to the PC memory with a is also shown in Fig. 1A upper horizontal bar with vertical
capacity to store approximately 240 s EEG data for seven limits; from the top segment in ‘B18’ to the last segment in
channels. All procedures for the analysis were carried out ‘A16’.
on the PC, which were programmed by the author S.K.. Since individual evoked potentials of BAEPs are fine
A part of an example of EEGs with spindle oscillations electrical activities embedded in a relatively large EEG
is shown in Fig. 1A. A 512-point EEG data 51.2 ms fluctuation refer to Section 4, we applied a method for
duration with a fixed position of a trigger signal is subtracting the trends before the average procedure Fig.
designated as a ‘segment’ in the following discussion. The 1C. A 7-order polynomial curve red tracings in Fig. 1C
analysis window for BAEPs was designed to cover 128- was fit to the raw EEG data Cz–A1raw and Cz–A2raw
point EEG data 12.8 ms duration within a segment. A in Fig. 1C within the range of BAEP analysis window
magnified example of a segment is shown in Fig. 1C. 128 points by the method of least square [4]. The
miniature waves so-called ‘embryo waves’ for two major 2.2. Sequential analysis of BAEPs over the course of
BAEP components wave-III and -V were discernible in spindle oscillation
the subtracted EEG data filled and open circles on Cz– A1sub in Fig. 1C even in one trial of acoustic trigger
Ten sequences of spindle were selected by viewing the signal.
S . Kohsaka et al. Brain Research 882 2000 103 –111
105
Fig. 1. A method of sequential analysis of BAEPs. Upward positive in both Cz–A1 and Cz–A2 for the analysis of BAEPs, and upward negative in the other EEGs. A A part of an example of sleep EEGs with two sequences of spindle is shown. The acoustic trigger signal trigger was delivered
continuously 20 times s throughout the recording. A K-complex preceding the spindle oscillation, marked with ‘1’, was easily detected by its phase reversal between C3–Cz and Cz–C4. Spindle oscillations with preceding K-complexes were excluded from the analysis. A horizontal bar with limits shows
the range for background EEG analysis refer to text. B Three series ‘B1’, ‘onset’ and ‘A1’ are aligned by expanding the EEG data with a horizontal bar labeled with ‘b’ in A. By shifting the average area a sequence of 16 segments, red rectangles backward ‘B1’, then forward ‘A1’ from the area of
‘onset’, sequentially changing BAEPs were averaged in 10 sequences of spindle. The average area was shifted by 2 segments 0.1 s, 1 20 repetition rate32 segment shift. C A segment 512 points labeled with ‘c’ in B is magnified. To reduce large fluctuations in raw EEG data Cz–A1raw and Cz–A2
raw, the trends were subtracted by fitting a 7-order polynomial curve red tracings, offset by 22.5 mV for recognition within the range of BAEP analysis window 128 points. The ‘embryo waves’ for BAEP components filled circle for wave-III, open circle for wave-V were discernible even in one trial of
acoustic signal in the subtracted EEG data Cz–A1sub and Cz–A2sub.
2.3. Background EEG analysis C3–Cz, Cz–C4 and C4–T4 were again averaged across
the subject grand-average. To correlate the BAEPs’ change with the corresponding
EEG activity, background EEGs were also averaged within the range from the top segment in ‘B18’ to the last
3. Results