Epilepsia, 48(10):1883–1894, 2007
Blackwell Publishing, Inc.
C ?2007 International League Against Epilepsy
Analysis of Initial Slow Waves (ISWs) at the Seizure Onset
in Patients with Drug Resistant Temporal Lobe Epilepsy
∗§Anatol Bragin, ?Pieter Claeys, ?Kristl Vonck, ?Dirk Van Roost,∗§Charles Wilson,
¶?Paul Boon, and∗†‡§Jerome Engel Jr.
Departments of∗Neurology, †Neurobiology, and ‡Psychiatry and Biobehavioral Sciences, §Brain Research Institute,
David Geffen School of Medicine at UCLA, Los Angeles, California, U.S.A.; ¶Departments of Neurology
and ?Neurosurgery, Ghent University Hospital, Ghent, Belgium
slow waves (ISWs) at seizure onset in patients with refractory
temporal lobe epilepsy. ISWs are a specific type of ictal EEG
by low voltage fast activity.
Methods: Investigations were carried out on 14 patients from
the UCLA hospital (USA) and 10 from the Ghent University
localization of the epileptogenic zone.
in the Ghent group were analyzed. Fourteen UCLA and seven
Ghent patients had ISWs at seizure onset. The duration of ISWs
0.2 to 1.4 mV. ISWs in three of 14 UCLA patients (30% of
seizures) had a consistent positive polarity at the deepest con-
tacts that were located in the amygdala, hippocampus, or en-
torhinal cortex and reversed polarity outside of these brain areas
or neocortex (ISWs2). All ISWs from the seven Ghent patients
electrodes at the cortical surface. ISWs1 were associated with
EEG spikes at the onset and on increase in amplitude of 10–20
Hz sinusoidal activity. In contrast, ISWs2 were associated with
suppression of EEG amplitude, an increase in frequency in the
range of 20–50 Hz, and did not have EEG spikes at the onset.
Multiunit neuronal activity showed strong synchronization of
neuronal discharges during interictal spikes, but multiunit syn-
chronization was not obvious during ISWs2.
Conclusion: The existence of EEG spikes and phase reversal
with ISWs1 indicates this type of seizure may be triggered by
hypersynchronous neuronal discharges; however, seizures with
ISWs2 at the onset may be triggered by different mechanisms,
perhapsnonneuronal. KeyWords: Epilepsy—Seizureonset—
Localization of the seizure generating zone is a challeng-
ing problem in epileptology. Distinct patterns of electri-
cal activity at the onset of a seizure can indicate unique
triggering mechanisms and identification of the source of
Understanding the mechanisms that trigger ictal activity
could lead to the development of more appropriate drugs
to prevent their occurrence.
Among various EEG patterns that occur at the onset
of ictal activity, paroxysmal slow waves with a duration
of 1 s and longer are commonly encountered in patients
with neocortical (Bragin et al., 2005; Hughes et al., 2005;
Kobayashi et al., 2005) and temporal lobe (Bragin et al.,
Accepted March 7, 2007.
Address correspondence and reprint requests to Anatol Bragin, 710
Westwood Plaza, Department of Neurology, David Geffen School
of Medicine at UCLA, Los Angeles, CA 90095, U.S.A. E-mail:
in patients with temporal lobe epilepsy and found that
their voltage depth profiles differ from voltage depth pro-
onset seizures (Bragin et al., 2005). ISWs in temporal
lobe epilepsy are associated with low voltage fast onset
seizures, which are commonly nonlocalizable. Patients
with this type of seizures are often not candidates for re-
sective surgical treatment.
The source and neuronal mechanisms of these waves
are unknown. Studying their voltage depth profiles could
help to identify their generators and to understand their
role in seizure initiation.
The goal of this study was to further analyze the spa-
tial characteristics of ISWs in patients with drug resistant
temporal lobe epilepsy using implanted depth electrodes,
and investigate possible mechanisms triggering this elec-
this study: (1) patients from UCLA hospital (USA) with
1884A. BRAGIN ET AL.
orthogonal depth electrode placements, perpendicular to
the lateral surface of the temporal lobe, which permitted
construction of voltage depth profiles, and (2) patients
from Ghent University Hospital (Belgium) with depth
electrodes inserted into mesial temporal structures from
over the temporal and frontal lobes, which permitted ob-
servation of electrical correlates of ISWs on the cortical
The presurgical protocol at UCLA Hospital has been
described previously (Engel, 1993; Fried et al., 1999). Pa-
tients with medically intractable complex partial seizures
were implanted with intracerebral clinical depth elec-
trodes for the localization of the area of seizure onset.
Prior to depth electrode implantation, patients gave their
informed consent for participation in this research study
under the approval of the Internal Review Board of the
UCLA Office for Protection of Research Subjects. Data
analysis was carried out on all 36 patients who were sub-
jected to depth electrodes implantation for localization of
the epileptogenic zone during the period 1999–2003.
Clinical depth electrodes-microelectrodes
The depth electrode configuration consisted of MRI-
compatible, flexible, 1.25-mm diameter polyurethane
probes (AdTech, Racine, WI, U.S.A.) with seven 1.5-mm
ter, with the exception of the two contacts nearest the tip,
which were separated by 3.0 mm on center. Each depth
electrode contained a semi-rigid internal stylet to facil-
itate the accurate placement of the electrode tip in the
selected brain area. After surgical placement within the
target structure, the stylet was removed, leaving the can-
nula lumen open for introduction of the microelectrodes.
A 9-contact laminar microelectrode array with 500-µm
intertip spacing was inserted into the depth electrode, so
that the microelectrode recordings were obtained from an
area spanning 4.0 mm beyond the tip of the depth elec-
trode (Bragin et al., 2002). The ninth wire is the shortest
(most proximal) and is uninsulated for 1.0 mm to provide
a recording reference for the other eigth microwires that
were insulated except for the tip.
Depth electrode implantation and recording procedure
MRI guided stereotactic procedure was used for depth
electrode implantation under general anesthesia (Engel,
1993). From four to eight orthogonal depth electrodes
were implanted in each hemisphere, with the majority
placed in hippocampal and parahippocampal structures.
Depth electrodes were connected to the video-EEG
monitoring system (128-channel digital video-EEG,
antiepileptic drugs were gradually tapered until habitual
seizures were recorded in all patients. Electrographic cor-
relates of seizures were recorded for 7–21 days during
and 70 Hz). Identification of the epileptogenic region was
based upon results of electrographic seizure recordings,
neuroimaging, and other clinical factors (Engel, 1996).
Localization of the recording microelectrode positions
The following approaches were used in combination to
increase the precision of microelectrode localization. Im-
mediately after implantation, anterior-posterior (AP) and
lateral skull x-rays were initially used to determine the
electrode position. During the period the depth electrodes
were in place MRI and CT scans were used for localiza-
tion of clinical depth electrode contacts, but skull x-rays
electrode removal, another MRI was obtained to visualize
the electrode tracks.
Postoperative CT and preoperative MRI images with
tion of the clinical electrode contacts. For microelectrode
localization x-ray images were adjusted to the same scale
as the postdeplant MRI using Adobe Photoshop and the
electrodes, were superimposed on the tracks of the depth
electrodes to identify the position of the shortest and the
longest microelectrodes. The distance between the short-
est and the longest microelectrodes was a constant 4.0
mm, so that when they were superimposed on the MRI,
the position of each microelectrode tip spaced at 500 µm
based on references of mesial temporal lobe anatomy by
Duvernoy (1998) and Amaral and Insausti (1990).
Ghent University Hospital Patients
All patients who were initially selected for deep brain
stimulation treatment between 1997 and 2000 were in-
cluded in this study. The presurgical protocol at Ghent
University Hospital has been described previously (Boon
et al., 1996). Prior to depth electrode implantation, pa-
tients gave their informed consent for participation in this
research. The study protocol was approved by the Ethics
Committee of Ghent University Hospital.
Clinical depth and grid electrodes
The depth electrode configuration consists of MRI-
compatible, flexible, 1.25-mm diameter polyurethane
probes. Four contact depth electrodes (Model 3387,
Medtronic, Minneapolis, MN, U.S.A.) with 1.5-mm wide
platinum contacts, each separated by 1.5 mm on center
were used in this group of patients. Each electrode con-
tained a semirigid internal stylet to facilitate the accurate
placement of the electrode tip in the selected brain area.
Grid electrodes consisted of a silastic sheath with embed-
ded platinum (5 mm in diameter) electrodes with 10-mm
center to center interelectrode distance (AdTech).
Epilepsia, Vol. 48, No. 10, 2007
ANALYSIS OF ISWs AT SEIZURE ONSET 1885
Depth and grid electrode implantation and recording
The surgical procedure included implantation of multi-
strip electrodes. During an MRI guided stereotactic pro-
cedure under general anesthesia, depth electrodes were
implanted in each hemisphere through two occipital burr
holes. Electrodes on each side were placed into the amyg-
dala and hippocampus areas. After surgical placement
within the target structure, the stylet was removed. In
all patients additional subdural grids and/or strips were
placed over the temporal and/or frontal neocortex through
craniotomy or burr holes, depending on the results of
the presurgical evaluation. All intracranial electrode con-
tacts were connected to a video-EEG monitoring system
(128-channel digital video-EEG, Grass-Telefactor). Sub-
habitual seizures were recorded in all patients. Electro-
graphic seizure onsets were recorded during 7–21 days
of telemetry monitoring (frequency band 0.1 Hz and 70
Hz). Identification of the epileptogenic region was based
primarily upon results of electrographic seizure record-
ings, neuroimaging, and other clinical factors (Engel,
Localization of the recording electrode positions
The precise location of the intracranial electrode con-
tacts was assessed by performing an MRI using an
MPRAGE sequence (1.5 T Magnetom, Siemens, Erlan-
gen, Germany) while electrodes were in place.
troencephalographers for electroclinical correlation and
identification of the seizure onset zone and MRI image
analysis and final conclusion regarding the existence of
hippocampal atrophy was made by certified neuroradiol-
ogists, for both UCLA and Ghent patients.
Averaging of seizure onset was performed using Data-
positivity “down” to the positivity “up.” Seizures were
aligned at the peak of the slow wave and concatenated
in a separate file before the averaging procedure. Voltage
versus depth profile analysis was done by measuring the
the values against the distance between electrodes.
Multiunit activity was extracted from raw data by a
high pass 600-Hz filtering procedure. Units were sepa-
old of two standard deviations from the baseline. Activity
be multiunit activity. Rate histograms were used for anal-
ysis of unit activity. We did not use spike cluster software
for discrimination of single units from multiunit activity
because it requires analysis of a large number of spikes
acquired under stable conditions. The transition from the
interictal to the ictal states is a rapidly changing process
and silence of others. This could create a bias for the clus-
Chi-square or Fisher’s exact tests were used for statisti-
cal analysis. Significance level for all analyses was set at
α = 0.05
At UCLA, seizure associated slow waves (ISWs) at the
low voltage fast (LVF) seizure onset were found in 14 out
of 36 patients (39%). In the other 22 patients ISWs were
absent or slow waves did occur later during seizure de-
velopment. These patients are subjects of other studies.
Fig. 1A illustrates a seizure with the ISW at the onset. In
dala, hippocampus, entorhinal, and orbitofrontal cortices,
and approximately 0.5 s later it appeared in the right ho-
motopic structures. The spatiotemporal characteristics of
ISWs were different from recorded EEG interictal spikes
(Fig. 1B), which were more localized and had shorter du-
ration. A total of 61 seizures were analyzed and results
are summarized in the Table 1. Duration of ISWs varied
amplitude varied from 0.1 to 1.4 mV (mean 0.6 mV ± 0.4
The clinical characteristics of patients with depth and
grid electrodes from Ghent University Hospital are pre-
sented in Table 2. Forty seizures from 10 patients were
taken for analysis. None of them had hippocampal atro-
the deep brain stimulation protocol. ISWs at seizure onset
were observed in 7 of 10 patients and 82% of 33 seizures.
The amplitude and duration of ISWs varied from 0.1 to
1.2 mV (mean 0.5 mV ± 0.3 SD) and from 0.4 to 3.5 s,
(mean 1.1 ± 0.6 SD) and were not significantly differ-
ent from those in the UCLA patients. Fig. 2 illustrates a
typical seizure from a Ghent hospital patient whose ISWs
had maximum amplitude and negative polarity in the left
amygdala (LA) and hippocampus (LH).
Spatial distribution of the ISWs
According to the terminology used at UCLA, clinical
deep target areas (such as hippocampus and amygdala).
Electrode contacts numbered 3, 4, 5 are usually located
in some patients the most superficial electrode contact
within neocortex was number 6 (see Figs. 1 and 3) or in
some patients number 7 (see Fig. 4).
In three out of 14 UCLA patients all ISWs had max-
imum amplitude at the deepest contacts located in the
amygdala, hippocampus, or entorhinal cortex and were
Epilepsia, Vol. 48, No. 10, 2007
1886 A. BRAGIN ET AL.
FIG. 1. (A) An example of ISWs at the
beginning of a seizure (UCLA patient
369). The black or grey blocks of tracings
represent an array of orthogonal record-
ing sites in one clinical electrode with
recording sites numbered from 1 to 6,
where number 1 is the deepest record-
ing site located in the brain area indi-
ficial located in the neocortex (see also
Fig. 4A for explanation). Dashed lines in-
dicate the peak of the ISW slope, the
value of which was used to plot volt-
age depth profile graphs in Fig. 5A. Here
and in all other figures positivity is up,
high pass filter is 0.1 Hz, and low pass
filter is 70 Hz. (B) An example of in-
terictal activity recorded 1 h before the
seizure shown in the part A. Abbrevia-
tions here and in the following figures: LA
and RA—left and right amygdala; LMH
and RMH-–left and right hippocampus;
LOF and ROF-–left and right orbitofrontal
positive in relation to the reference electrode located in
ment toward the neocortex (Fig. 3). Voltage depth profiles
of these ISWs are presented in Fig. 5A where ISWs are
positive in the depth electrodes located in the amygdala
and hippocampus. Their amplitude sharply decreased and
changed polarity at contact 2. The ISWs recorded from
TABLE 1. Characteristics of UCLA patients with initial slow waves seizure onset
Epileptogenic zone on the
basis of clinical analysis Patient IDHip atrophyNo. of SZsurgerySZ outcome Am HipECOF
LTL & RTL
Most SZ from R lat post Hipp,
1 from LOF
Left Sup.Temp. Gyrus
LAH, LPG, LEC
L Mesial Temp
LA mTL ectomy
gyrus; LOF, left orbitofrontal cortex; RATL, left anterior temporal lobe.
the entorhinal cortex electrode did not show phase rever-
sal and the amplitude did not change significantly toward
In 11 of 14 UCLA patients ISWs at all seizure onsets
had negative polarity in relation to the reference electrode
located on the skull (CZ) and did not show phase rever-
sal of the ISWs across any electrode contacts (Fig. 1A).
Epilepsia, Vol. 48, No. 10, 2007
ANALYSIS OF ISWs AT SEIZURE ONSET 1887
TABLE 2. Characteristics of Ghent patients with and without initial slow waves at the seizure
Polarity of initial phase
DepthPatient IDHip atrophy No. of SZISW Surface
Total 40Yes = 33 No = 7Yes = 33N = 33n = 33
Voltage versus depth profile analysis showed that in all
seizures in seven of the 11 patients the amplitude of ISWs
was smallest in the deep brain areas and increased toward
the superficial contacts located in the neocortex (Fig. 5B).
ISWs was maximal at contacts 4–5 and sharply decreased
at more superficial contacts (Figs. 4 and 5C). Identifica-
tion of the location of the recording sites in these patients
crease in amplitude of ISWs occurred in contacts located
in the gray matter of the neocortex and the maximum am-
plitude of ISWs was in the white matter (Fig. 5C). We
never observed a change in polarity of ISWs at the border
between gray and white matter, while some EEG inter-
ictal spikes did reverse polarity between gray and white
matter. There was also no phase reversal when the track
of the clinical electrode crossed other gyri of the neocor-
tex (compare the location of the left amygdala and left
posterior hippocampus contacts in Fig. 4C, D).
were recorded, there was negative polarity in depth elec-
patients. As indicated in the methods section depth elec-
trodes in this group were implanted through two occip-
ital burr holes and recording sites were oriented along
the saggital axis; in this orientation ISWs showed ampli-
tude decrement outside of hippocampus and amygdala.
At the same time a positive polarity wave was observed at
the level of subdural grid or strip contacts (see Fig. 2, grid
contacts 1–20), which were located on the temporal part
of the neocortex.
Electrographic patterns associated with ISWs
There was a difference in electrographic patterns ac-
companying ISWs with and without phase reversal. ISWs
with phase reversal were associated with an increase in
amplitude of 10–20 Hz sinusoidal activity (Fig. 6A). This
activity remained for 5–10 s and than decreased to a fre-
quency of 2–5 Hz. In the three UCLA patients with ISWs
ISWs (Figs. 3 and 7).
In contrast, ISWs for all UCLA patients without phase
reversal and all ISWs in the Ghent group were associated
with a suppression of EEG amplitude and an increase in
the frequency of EEG activity to 20–50 Hz (Fig. 6B). The
highest increase in frequency was observed at the peak
of the ISWs. The frequency decreased on the descending
part of the ISWs to 2–5 Hz (Fig. 6B).
There was no correlation between the type of ISW oc-
currence and either the presence of hippocampal atrophy
or occurrence or seizure outcome following surgery
reversed ISWs to make much of this observation.
Extracellular unit activity during ISWs
Recording of extracellular multiunit activity requires
sampling at least 10 kHz. Existing technical abilities did
not allow continuous 24 h recordings from these patients
to captures all seizures with such a high sampling rate.
We were able to record seizures with high sampling rate
in only four patients, all of whom had, ISWs at seizure
onset that were without phase reversal, and microelec-
trodes located in either entorhinal cortex (three patients)
or hippocampus (one patient).
In 48 microelectrodes multiunit activity was recorded
during transition from interictal to ictal activity. An in-
crease in the frequency of discharges of multiunit ac-
tivity indicates an increase in the rate of discharges of
individual neurons and an increase in synchrony of dis-
charges within the recorded population of neurons. In the
majority of microelectrodes (n = 34, 71%), neurons ei-
ther did not change their firing rate or decreased their
firing rate at the ISWs onset. Significant increases in the
frequency of unit discharges were observed in 14 (29%)
microelectrodes. In five of 14 microelectrodes (one in
Epilepsia, Vol. 48, No. 10, 2007
1888A. BRAGIN ET AL.
FIG. 2. (A) An average of ISWs from 3 seizures (Ghent patient
no. 209) recorded from occipitally placed depth electrodes in the
amygdala and hippocampus and surface grid electrodes. Notice
reversed polarity between depth contacts and grid electrodes lo-
cated in the left neocortex illustrated by the scheme. (B) A draw-
ing of brain with arrangement of depth electrodes and grids. LA
and RA— left and right amygdala; LH and RH—left and right hip-
hippocampus and four in entorhinal cortex (from two
of spikes at the ascending part and peak of ISWs. In all
cases the changes in neuron discharges occurred with a
FIG. 3. An example of an ISW at seizure onset with phase re-
versal in UCLA patient no. 366. The ISW is positive in the first
recording site of left hippocampus (LAH) and left amygdale (LA)
and negative in all other electrodes. Notice the interictal spikes
occurring at the onset of the ISW, the increase in amplitude of
EEG at the peak of the ISW and the occurrence of regular 10–
15 Hz activity. There is no ISW reversal in LEC. Abbreviations as
in Fig. 2. Dashed lines show the shape of ISWs after low pass
(2 Hz) filtering.
delay of 30–150 ms after onset of ISWs. In nine of 14
microelectrodes (three in hippocampus and 11 in entorhi-
nal cortex (from three depth electrodes in three patients))
during the descending part of ISWs. These different pat-
An example of multiunit activity during EEG interictal
spikes and ISWs is shown in Fig. 8. This is the same
patient as shown in Fig. 4. Unit activity was recorded
from an array of seven microelectrodes implanted into
Epilepsia, Vol. 48, No. 10, 2007
ANALYSIS OF ISWs AT SEIZURE ONSET1889
FIG. 4. The location of the recording sites
of different clinical electrodes illustrated by
x-ray images (A, B) and MRI (C) taken in
the plan indicated by dashed line in part
“A.” (D) An example of seizure recorded in
this patient (UCLA no. 339). Numbers in
part “A” indicate the shortest distance be-
tween orbitofrontal cortex and posterior hip-
pocampus and entorhinal cortex. Symbol
part of the brain. Abbreviations: LAH-–left
anterior hippocampus; LEC-–left entorhinal
cortex; LPH-–left posterior hippocampus;
LA—left amygdala; LOF—left orbitofrontal
cortex. Notice that the maximum negative
deflection occurs initially (dashed line) in
recording sites 4–5 in LPH, LA, and LEC. All
of these recording sites are located in the
white matter (Fig. 4C). There is no visible
deflection in LAH at this point. The positive
and second negative waves (star) are syn-
chronous in three areas: LA, LAH, and LEC.
Numbers above the arrows in part A indi-
cate the linear distances between recording
sites. Dashed circle in part A and dashed
box in part B outline the square occupied by
LA, LAH, and EC.
entorhinal cortex. Neuronal discharges recorded with all
(Fig. 8A). However, during ISWs the same population of
neurons showed much weaker increase in the frequency
and synchrony of neuronal discharges. (Fig. 8B, C). The
population of neurons recorded in microelectrodes no.
1 and 2 generated bursts of discharges at the ascending
part of ISW and then became silent for several seconds.
The population of neurons recorded with microelectrode
the neurons recorded with the other five microelectrodes
either decreased their frequency of discharge or did not
dle eventually became involved in the hypersynchronous
population spikes (not shown).
characterized by a slow wave at the clinical seizure onset
followed by low voltage fast activity. Two types of ISWs
were seen in our experiments. The first type of ISWs has
positive polarity at the deep electrodes and shows phase
reversal (ISWs1). The second type of ISWs has negative
polarity at the deep electrodes and has maximum am-
plitude in neocortical gray matter or in the underlying
white matter adjacent to neocortex (ISWs2). ISWs2 do
not show a phase reversal across depth electrode contacts
the most superficial of which was within neocortex, but
do show a phase reversal on grid electrodes positioned on
the surface of the brain. Although we did not have the
data for voltage depth profile analysis of the Ghent pa-
tients, the associated EEG characteristics of ISWs resem-
bled these of ISWs2 from UCLA patients. Because these
ISWs had negative polarity in depth electrodes and posi-
tive in grid electrodes, it is possible that the phase reversal
The main difference between ISWs 1 and 2 is that
the first are always localized in deep brain structures,
while the second are not localized involving broad ar-
eas of neocortex from orbitofrontal to temporal areas as
events is that ISWs1 are accompanied by EEG spikes and
an increase in amplitude of 8–20 Hz activity (alpha–beta
range), while ISWs2 are not accompanied by EEG spikes
and show a decrease in amplitude of EEG and occurrence
of 30 Hz and higher activity (gamma range) with the fre-
quency decrease while seizure develops. These data in-
dicate that ISWs1 and ISWs2 are generated by different
networks and suggest that the occurrence of ISWs1 in a
Epilepsia, Vol. 48, No. 10, 2007
1890A. BRAGIN ET AL.
FIG. 5. Voltage depth profiles of ISWs in three patients. (A) From
the seizure shown in the Fig. 3; (B) from seizure shown in the Fig.
1A. (C) From the seizure shown in Fig. 5D. Dashed boxes indicate
the recording sites in the areas indicated near the box. See details
in the text. Abbreviations: left and right amygdala (LA, RA), mid
hippocampus (LMH, RMH), and orbitofrontal cortex (LOF, ROF).
specific brain area could be a prominent indicator for lo-
calization of epileptogenic zone.
One explanation for the absence of a phase reversal
in the ISWs2 of UCLA patients could be that electrode
contacts did not pass through the source of the recorded
event located somewhere in the deep area of the brain.
pattern is in neocortex.
The relation of ISWs to other known slow waves as-
sociated with seizures is unclear. ISWs may be similar to
the DC shift described in several studies (Goldring, 1963;
FIG. 6. Electrographic patterns associated with the ISW revealed
by high pass filtering and power spectral analysis. (A) An exam-
ple of a seizure onset with an ISW with phase reversal. Notice
an increase in amplitude and frequency of 10–20 Hz electrical
activity. White line-–EEG recorded with 0.1 Hz–70 Hz frequency
band. Yellow line–the same activity high pass (3 Hz) filtered and
amplified five times. Color contours indicate frequency bands su-
perimposed on the ISW. (B) An example of an ISW without phase
activity is decreases during the ISW while the power of frequen-
cies around 30 Hz is increases at the peak of the ISW with re-
placement of activity in the frequency range of 12–20 Hz. Theta
frequencies (2–5 Hz) become dominant again during seizure de-
velopment. Notice the interictal spikes at the onset of the ISW.
Ikeda et al., 1996, 1997; Vanhatalo et al., 2003, 2005;
Voipio et al., 2003). However, the duration of DC shift in
these studies usually varied from several seconds to sev-
eral tens of seconds, while the mean duration of ISWs
is 2.3 s. The difference in the duration and shape of the
ISWs and DC shift cannot be explained by the effect of
the AC coupled amplifiers used in our experiments. The
0.1 and 70 Hz, and the shape of waves with duration of
several seconds did not change as a result of change in
If we assume that EEG spikes preceding ISWs1 reflect
hypersynchronous discharges of pathological clusters of
principal cells (Ayala et al., 1973; Buzsaki et al., 1991),
they might be the trigger of local seizures. Our experi-
ments confirmed existing data (de Curtis and Avanzini,
2001) that synchrony of neuronal discharges increases
during interictal spikes (Fig. 8A). ISWs following these
EEG spikes at seizure onset could reflect the occur-
rence of massive IPSPs generated by a wide interneuronal
Epilepsia, Vol. 48, No. 10, 2007
ANALYSIS OF ISWs AT SEIZURE ONSET1891
FIG. 7. EEG spikes associated with an ISW at seizure onset in
UCLA patient no. 351. EEG spikes occurred both in the right and
LAH) and are accompanied by an ISW with phase reversal in the
LAH. Abbreviations as in Fig. 1.
of pathologically interconnected neurons (PIN clusters),
viously suggested this mechanism of seizure generation
(Bragin et al., 1999, 2000, 2004).
FIG. 8. Different patterns of unit activity during EEG interictal
spikes and seizure onset with ISW in UCLA patient no. 339. (A
Perievent time histogram of multiunit discharges from the seven
microelectrodes represented in “B” before, during, and after an
EEG interictal spike. (B) fragment of the seizure shown in Fig. 5
recorded with clinical depth electrodes. “wb”-–raw data recorded
in the frequency band 0.1 Hz and 10 kHz. “bp”—the same data af-
ter band pass filtering (2–80 Hz) and additional magnification. (C)
Multiunit activity at the seizure onset. Lines 1a and 7a show mul-
tiunit activity recorded with two microelectrodes after high pass
filtering (600 Hz). Horizontal dashed lines indicated the level of
separation of unit discharges from the noise. Activity after num-
bers 1–7 illustrates multiunit discharges recorded with different
microelectrodes from the bundle of microelectrodes within this
area. (D) Perievent time histogram of multiunit discharges from all
seven microelectrodes in relation seizure onset and ISW. Notice
that the onset of ISW (vertical dashed line) precedes all changes
in neuron activity. See details in the text.
Epilepsia, Vol. 48, No. 10, 2007
1892A. BRAGIN ET AL.
FIG. 9. Illustration of a hypothetical sequence of neuronal net-
at the onset of ISW. The EEG spike is a field potentials of sum-
mated population spikes generated by principal cells involved in
pathologically interconnected neuron clusters (PIN clusters). The
their frequency of discharge. The ISW occurs as a result of feed-
back discharges of the inhibitory network in response to the initial
PIN cluster discharges and reflects IPSPs on the somata of prin-
cipal cells. At the termination of this massive inhibition principal
cells rebound simultaneously resulting in ictal generation.
Each cluster generates bursts of population spikes and
these synchronous spikes are followed by generalized
feedback inhibition in large areas of neocortex or even
rebound from global inhibition (Fig. 9). Occurrence of
seizures as a result of rebound from strong inhibition has
also been proposed by several other authors (Mody, 1998;
Coulter, 1999; Andre et al., 2001; Dudek et al., 2002;
Timofeev et al., 2002; Wendling et al., 2002; Freund,
2003; Mody, 2005). The existence of beta and gamma
activity on the tail of ISWs supports this hypothesis be-
cause this type of activity is believed to reflect synchro-
nization of inhibitory interneuron networks (Bragin et al.,
1995; Buzsaki et al., 1995; Buzsaki and Chrobak, 1995;
Whittington et al., 1995; Traub et al., 1996; Chrobak and
Buzsaki, 1998; Traub et al., 1998; Csicsvari et al., 2003).
Mechanisms generating ISWs2 are less clear. As men-
tioned above, there are three major features of ISWs2 in
addition to the absence of phase reversal: (1) The vol-
ume of tissue showing ISWs2 is very large. (2) ISWs2
are accompanied by a suppression of EEG amplitude and
ronal discharges and EEG spikes are absent prior to or at
the moment when ISWs2 occur.
The maximum amplitude of ISWs2 in the neocortex of
UCLA patients and their phase reversal on the grid elec-
trodes located on the surface of the neocortex of Ghent
patients suggests that its neuronal generator is located
FIG. 10. Hypothetical mechanism of seizure occurrence as a re-
sult of impairment of glial buffering function. The ISW reflects re-
lease of glutamate in the extracellular space. The interneuronal
network responds first to the change in glutamate concentration
due to its lower threshold for action potential generation. Gamma
activity at the descending part of the ISW reflects an increase in
firing rate of the interneuronal network suppressing discharges of
the principal cell network. Principal cells recover simultaneously
after initial inhibition forming pathological synchronous ictal gen-
within neocortex, however, absence of phase reversal be-
tween white and gray matter in UCLA patients renders
this hypothesis questionable.
all recording electrodes. Another explanation for absence
of phase reversal of ISWs2 at the level of neocortex is that
these slow waves could reflect nonsynaptic events. Sev-
eral publications indicate the existence of potentials that
have no clear dipole (Gross et al., 2000; Merlet and Got-
(Voipio et al., 2003; Nita et al., 2004). Perhaps ISWs2 are
generated by a diffuse glial network due to impairment of
2002; Borges et al., 2003; Eid et al., 2004; Huang et al.,
Spencer and coworkers showed loss of glutamine syn-
thetase activity and impairment of glutamate–glutamine
cycle in the human epileptogenic hippocampus, which
could be a possible mechanism for raised extracellular
glutamate in mesial temporal lobe epilepsy (Petroff et al.,
2002). Glutamate reuptake mechanism may be dampened
in the epileptogenic zone, which could contribute to pro-
gression of spontaneous seizure activity (Gorter et al.,
Epilepsia, Vol. 48, No. 10, 2007
ANALYSIS OF ISWs AT SEIZURE ONSET1893
may release glutamate into the extracellular space (Tian
et al., 2005). It has also been shown that changes in extra-
cellular environment, such as osmolarity, effect the firing
frequency of interneurons but not pyramidal cells (Bara-
ban et al., 1997). Microinjection of 1 mm3of glutamate
in rats with chronic seizures causes negative slow wave
at the place of injection resembling ISWs at the onset
of spontaneous seizures (Bragin et al unpublished obser-
vations). When the level of extracellular glutamate rises,
interneurons, which have a lower threshold for spike gen-
eration than pyramidal cells, (Alonso and Klink, 1993;
Fricker et al., 1999; McBain et al., 1999) increase their
beta and gamma activity on the descending part of ISWs.
This could reflect attempts of an interneuronal network to
occur later, either as a result of the remaining high con-
centration of glutamate, or as a result of rebound from the
Seizure activity would develop from the subsequent syn-
chronization of principal neurons in large brain networks
responsible for clinical manifestations (Fig. 10). Further
experiments are required to confirm or reject this hypoth-
Alonso A, Klink R. (1993) Differential electroresponsiveness of stellate
of Neurophysiology 70:128–143.
Amaral D, Insausti R. (1990) Hippocampal Formation. In Paxinos G
(Ed) The human nervous system. Academic Press, San Diego, pp.
Amzica F, Massimini M, Manfridi A. (2002) Spatial buffering during
slow and paroxysmal sleep oscillations in cortical networks of glial
cells in vivo. Journal of Neuroscience 22:1042–1053.
Andre V, Marescaux C, Nehlig A, Fritschy JM. (2001) Alterations of
hippocampal GAbaergic system contribute to development of spon-
taneous recurrent seizures in the rat lithium-pilocarpine model of
temporal lobe epilepsy. Hippocampus 11:452–468.
Ayala GF, Dichter M, Gumnit RJ, Matsumoto H, Spencer WA. (1973)
Genesis of epileptic interictal spikes. New knowledge of cortical
paroxysms. Brain Research 52:1–17.
Baraban SC, Bellingham MC, Berger AJ, Schwartzkroin PA. (1997)
Osmolarity modulates K+ channel function on rat hippocampal in-
terneurons but not CA1 pyramidal neurons. The Journal of Physiol-
Boon P, Vandekerckhove T, Calliauw L, Achten E, De Reuck J, Thiery
E, Caemaert J, Desomer A, Drieghe C, Vanbelleghem H, Vonck K,
Defreyne L, Van Duyse A. (1996) Epilepsy surgery in Belgium, the
Flemish experience. Acta Neurologica Belgica 96:6–18.
Borges K, Gearing M, McDermott DL, Smith AB, Almonte AG, Wainer
BH, Dingledine R. (2003) Neuronal and glial pathological changes
Bragin A, Jando G, Nadasdy Z, Hetke J, Wise K, Buzsaki G. (1995)
Gamma (40-100 Hz) oscillation in the hippocampus of the behaving
rat. Journal of Neuroscience 15:47–60.
Bragin A, Engel J Jr, Wilson CL, Vizentin E, Mathern GW. (1999) Elec-
trophysiologic analysis of a chronic seizure model after unilateral
hippocampal KA injection. Epilepsia 40:1210–1221.
development of a network of pathologically interconnected neuron
clusters: a hypothesis. Epilepsia 41:S144–152.
tic brain: entorhinal cortex. Annals of Neurology 52:407–415.
Bragin A, Wilson CL, Almajano J, Mody I, Engel J Jr. (2004) High-
frequency oscillations after status epilepticus: epileptogenesis and
seizure genesis. Epilepsia 45:1017–1023.
Bragin A, Wilson C, Fields T, Fried I, Engel JJ. (2005) Analysis of
seizure onset on the basis of wideband EEG recordings. Epilepsia
ion in Neurobiology 5:504–510.
Buzsaki G, Hsu M, Slamka C, Gage FH, Horvath Z. (1991) Emergence
and propagation of interictal spikes in the subcortically denervated
hippocampus. Hippocampus 1:163–180.
Buzsaki G, Bragin A, Chrobak J, Nadasdy Z, Sik A, Ylinen A. (1995)
to memory trace formation. InBuzsaki G, Llinas RR, Singer W,
Berthoz A, Christen Y (Eds) Temporal coding in the brain. Singer,
Berlin, pp. 145–172.
Chrobak JJ, Buzsaki G. (1998) Gamma oscillations in the entorhinal
cortex of the freely behaving rat. Journal of Neuroscience 18:388–
bic system after status epilepticus. Epilepsia 40:S23–33.
Crino PB, Jin H, Shumate MD, Robinson MB, Coulter DA, Brooks-
Kayal AR. (2002) Increased expression of the neuronal gluta-
mate transporter (EAAT3/EAAC1) in hippocampal and neocortical
epilepsy. Epilepsia 43:211–218.
Csicsvari J, Jamieson B, Wise KD, Buzsaki G. (2003) Mechanisms of
gamma oscillations in the hippocampus of the behaving rat. Neuron
D’Ambrosio R, Maris DO, Grady MS, Winn HR, Janigro D. (1999)
Impaired K(+) homeostasis and altered electrophysiological prop-
erties of post-traumatic hippocampal glia. Journal of Neuroscience
Progress in Neurobiology 63:541–567.
Dudek FE, Hellier JL, Williams PA, Ferraro DJ, Staley KJ. (2002) The
course of cellular alterations associated with the development of
spontaneous seizures after status epilepticus. Progress in Brain Re-
Duvernoy HM. (1998) The human hippocampus. Springer, New York.
Eid T, Thomas MJ, Spencer DD, Runden-Pran E, Lai JC, Malthankar
GV, Kim JH, Danbolt NC, Ottersen OP, de Lanerolle NC. (2004)
Loss of glutamine synthetase in the human epileptogenic hippocam-
pus: possible mechanism for raised extracellular glutamate in mesial
temporal lobe epilepsy. Lancet 363:28–37.
Engel J Jr. (1993) Intracerebral recordings: organization of the human
Freund TF. (2003) Interneuron diversity series: rhythm and mood in
perisomatic inhibition. Trends in Neurosciences 26:489–495.
Fricker D, Verheugen JA, Miles R. (1999) Cell-attached measurements
Fried I, Wilson CL, Maidment NT, Engel J, Behnke E, Fields TA, Mac-
in neurosurgical patients. Technical note. Journal of Neurosurgery
discharge. InBrazier MAB (Ed) Brain function. UCLA forum med.
sci. University of California Press, Los Angeles, pp. 215–236.
Gorter JA, Van Vliet EA, Proper EA, De Graan PN, Ghijsen WE, Lopes
Da Silva FH, Aronica E. (2002) Glutamate transporters alterations
in the reorganizing dentate gyrus are associated with progressive
seizure activity in chronic epileptic rats. Journal Of Comparative
Epilepsia, Vol. 48, No. 10, 2007
1894A. BRAGIN ET AL. Download full-text
the epileptic focus and hand area in central epilepsy: combining
dipole models and anatomical landmarks. Journal of Neurosurgery
Huang YH, Sinha SR, Tanaka K, Rothstein JD, Bergles DE. (2004)
Astrocyte glutamate transporters regulate metabotropic glutamate
receptor-mediated excitation of hippocampal interneurons. Journal
of Neuroscience 24:4551–4559.
Hughes J, Fino JJ, Patel K. (2005) A newly described ictal pattern:
the initial ictal slow shift. Clinical EEG and Neuroscience 36:161–
Ikeda A, Terada K, Mikuni N, Burgess RC, Comair Y, Taki W, Hamano
T, Kimura J, Luders HO, Shibasaki H. (1996) Subdural recording of
ictal DC shifts in neocortical seizures in humans. Epilepsia 37:662–
Ikeda A, Yazawa S, Kunieda T, Araki K, Aoki T, Hattori H, Taki W,
Shibasaki H. (1997) Scalp-recorded, ictal focal DC shift in a patient
with tonic seizure. Epilepsia 38:1350–1354.
Kobayashi K, Oka M, Inoue T, Ogino T, Yoshinaga H, Ohtsuka Y.
tic spasms. Epilepsia 46:1098–1105.
McBain CJ, Freund TF, Mody I. (1999) Glutamatergic synapses onto
hippocampal interneurons: precision timing without lasting plastic-
ity. Trends in Neuroscience 22:228–235.
Merlet I, Gotman J. (2001) Dipole modeling of scalp electroencephalo-
ical Neurophysiology 112:414–430.
Mody I. (1998) Interneurons and the ghost of the sea. Nature Neuro-
Mody I. (2005) Aspects of the homeostaic plasticity of GABA
A receptor-mediated inhibition. Journal of Physiology 562:37–
Nonneuronal origin of CO2-related DC EEG shifts: an in vivo study
in the cat. Journal of Neurophysiology 92:1011–1022.
Petroff OA, Errante LD, Rothman DL, Kim JH, Spencer DD. (2002)
Glutamate-glutamine cycling in the epileptic human hippocampus.
Sepkuty JP, Cohen AS, Eccles C, Rafiq A, Behar K, Ganel R, Coulter
DA, Rothstein JD. (2002) A neuronal glutamate transporter con-
tributes to neurotransmitter GABA synthesis and epilepsy. Journal
of Neuroscience 22:6372–6379.
Tian GF, Azmi H, Takano T, Xu Q, Peng W, Lin J, Oberheim N, Lou N,
Wang X, Zielke HR, Kang J, Nedergaard M. (2005) An astrocytic
basis of epilepsy. Nature Medicine 11:973–981.
inhibition and the activity of fast-spiking neurons during cortical
spike-wave electrographic seizures. Neuroscience 114:1115–1132.
Analysis of gamma rhythms in the rat hippocampus in vitro and in
vivo. Journal of Physiology 493: 471–484.
GR. (1998) Gamma-frequency oscillations: a neuronal population
phenomenon, regulated by synaptic and intrinsic cellular processes,
and inducing synaptic plasticity. Progress in Neurobiology 55:563–
Vanhatalo S, Tallgren P, Becker C, Holmes MD, Miller JW, Kaila K,
Voipio J. (2003) Scalp-recorded slow EEG responses generated in
Vanhatalo S, Voipio J, Kaila K. (2005) Full-band EEG (FbEEG): an
emerging standard in electroencephalography. Clinical Neurophysi-
scale DC shifts in the human scalp EEG: evidence for a nonneuronal
generator. Journal of Neurophysiology 89:2208–2214.
Wendling F, Bartolomei F, Bellanger JJ, Chauvel P. (2002) Epileptic
fast activity can be explained by a model of impaired GABAergic
dendritic inhibition. European Journal of Neuroscience 15:1499–
Whittington M, Traub R, Jefferys J. (1995) Synchronized oscillations in
interneuron networks driven by metabotrophic receptor activiation.
Zhou J, Sutherland ML. (2004) Glutamate transporter cluster formation
in astrocytic processes regulates glutamate uptake activity. Journal
of Neuroscience 24:6301–6306.
Epilepsia, Vol. 48, No. 10, 2007