Epilepsia, 46(5):669–676, 2005
Blackwell Publishing, Inc.
C ?2005 International League Against Epilepsy
Intracranial EEG Substrates of Scalp EEG Interictal Spikes
James X. Tao, Amit Ray, Susan Hawes-Ebersole, and John S. Ebersole
Department of Neurology, Adult Epilepsy Center, University of Chicago, Chicago, Illinois, U.S.A.
Summary: Purpose: To determine the area of cortical genera-
tors of scalp EEG interictal spikes, such as those in the temporal
Methods: We recorded simultaneously 26 channels of scalp
EEG with subtemporal supplementary electrodes and 46 to 98
channels of intracranial EEG in 16 surgery candidates with tem-
poral lobe epilepsy. Cerebral discharges with and without scalp
EEG correlates were identified, and the area of cortical sources
was estimated from the number of electrode contacts demon-
strating concurrent depolarization.
Results: We reviewed ∼600 interictal spikes recorded with
intracranial EEG. Only a very few of these cortical spikes were
associated with scalp recognizable potentials; 90% of cortical
whereas only 10% of cortical spikes having <10 cm2of source
area produced scalp potentials. Intracranial spikes with <6 cm2
of area were never associated with scalp EEG spikes.
Conclusions: Cerebral sources of scalp EEG spikes are
larger than commonly thought. Synchronous or at least tem-
porally overlapping activation of 10–20 cm2of gyral cortex
is common. The attenuating property of the skull may actu-
ally serve a useful role in filtering out all but the most sig-
nificant interictal discharges that can recruit substantial sur-
roundingcortex. KeyWords: IntracranialEEG—ScalpEEG—
Interictal spikes—Cortical sources—Subdural electrodes—
Spike generator—Temporal lobe epilepsy.
Interictal epileptiform discharges from scalp EEG
play an important role in the lateralization and local-
ization of epileptogenic foci in the presurgical evalu-
ation (1–3). Scalp interictal spikes may even provide
more reliable localization than ictal discharges because
seizure propagation has often occurred by the time ic-
tal rhythms are recorded by scalp electrodes. In addi-
tion, muscle and movement artifacts commonly obscure
seizure rhythms (4,5). Other technologies currently used
cephalograpy (MEG) and functional magnetic resonance
imaging (fMRI), also are largely dependent on the analy-
sis of interictal spikes (6,7).
EEG spikes, their underlying cerebral substrate is mostly
speculative. The area of cerebral sources of scalp EEG
spikes has seldom been confirmed directly. Misconcep-
tions are therefore prevalent regarding the extent of cere-
bral activation required to generate those that are record-
able at the scalp. In 1965, Cooper et al. (8) proposed that
6 cm2of synchronized cortical activity was probably nec-
essary. This figure has gained widespread acceptance.
Their model, however, used in vitro measurements of a
Accepted December 26, 2004.
Address correspondence and reprint requests to Dr. J. Tao at De-
partment of Neurology, Adult Epilepsy Center, University of Chicago,
piece of fresh cadaver skull, a pulse generator connected
to saline-soaked cotton balls placed on the inside of the
skull, an artificial dura made from a polyethylene sheet,
and EEG recording electrodes on the exterior surface of
punched into the polyethythene sheet when the artificial
outside of the skull. Not only was this an artificial model,
but measurements also were made in the absence of any
background activity. Thus 6 cm2may not accurately re-
flect the necessary size of cortical sources for scalp EEG
spikes in clinical recordings.
ing in epilepsy surgery candidates. Such in vivo measure-
EEG spikes. The goal of the present investigation was to
determine the area of cortical sources that generate scalp-
visual inspection of the EEG.
MATERIALS AND METHODS
Simultaneous cortical and scalp EEG recordings were
ing intracranial EEG monitoring. Standard subdural elec-
trode strips and grids were used. Each strip consisted of
670 J. X. TAO ET AL.
patient’s skull after subdural electrode placement. Note minimal
skull defects over the left temporal region. A: anterior; P: posterior.
A three-dimensional reconstruction (CURRY 4.5) of a
four to eight platinum disk electrodes spaced 10 mm be-
tween centers. The disks were embedded in a 0.7-mm-
thick Silastic strip and had an exposed surface diameter
of 2.3 mm. Grid electrodes (4 × 8) contained 32 plat-
center-to-center distance. In our standard implantation, a
4 × 8 grid was positioned to encompass the anterior and
midtemporal regions. Two strips were placed to record
from the anterior and inferior temporal tip. Using sterile
procedures, standard EEG electrodes [International 10–
20 plus supplementary subtemporal electrodes: F9, T9,
M1 (mastoid), F10, T10, M2] were applied with collo-
dion to the scalp of the implanted patients. Occasionally
standard positions because of surgical scars and soft tis-
sue swelling. These electrodes were usually near the “C”-
shaped scalp incision in frontocentral and superior tem-
electrodes in the same patient as in Fig.
1 obtained from coregistration of post-
extensive electrode coverage of the left
anterior and infero-lateral temporal cor-
tex. Standard subdural electrodes em-
ployed in this study including one 4 ×
8 left mid temporal grid (LMT1-32), one
1 × 8 left anterior temporal strip (LAT1-
8), and one 1 × 4 left inferior temporal
strip (LIT1-4). The important electrodes
that are necessary for identification of all
subdural electrodes are indicated in the
3-D visualization of subdural
A volumetric computer tomography (CT) scan of the
head with 1.0-mm axial slices was obtained after implan-
electrode locations (CURRY, Compumedics/Neuroscan
El Paso, TX, U.S.A.). Patients with significant skull de-
EEG by visual inspection, were excluded from this study.
The positions of subdural electrodes also were coregis-
trated with a presurgical volumetric magnetic resonance
image (MRI) by using mutually identifiable surface fidu-
cials to provide 3-D visualization of electrode locations
relative to cortical anatomy (Fig. 2).
Continuous recordings of simultaneous scalp EEG (26
channels) and intracranial EEG (46 to 98 channels) were
The system input range was 2.0 mV, and data were digi-
tized with a 12-bit analog-to-digital converter (an ampli-
tude resolution of 0.448 µV). EEG data were digitized
at 200 Hz and bandpass filtered (0.3 to 70 Hz). All data
scalp electrode CPZ. Additional digital filters were used
with scalp data as necessary to minimize artifact and im-
prove the signal-to-noise ratio.
Cortical interictal spikes that had a source extent com-
pletely within or nearly within the coverage of intracra-
nial electrodes were selected for analysis. Accordingly,
most of the selected cortical spikes originated from the
lateral and inferolateral aspect of the anterior to midtem-
poral lobe, which was covered by the subdural electrode
grid. Although cortical spikes in patients with tempo-
ral lobe epilepsy were commonly observed in the tem-
poral tip region (11,12), they were not usually selected
for the study because their source area was often not re-
solved sufficiently with the implanted subdural strip elec-
trodes. We defined scalp EEG interictal spikes as “recog-
nizable” when they possessed an amplitude ≥50% higher
than the EEG background and a discrete voltage field
and produced clear disruption of the EEG background.
Epilepsia, Vol. 46, No. 5, 2005
EEG SUBSTRATES OF INTERICTAL SPIKES671
ing. Note that only two of the intracranial spikes (labeled 1 and 2) generate recognizable scalp interictal potentials. LOF: left orbital frontal;
LAT: left anterior frontal; LIT: left inferior temporal; LMT: left mid temporal.
A: Intracranial EEG recording demonstrates a heterogeneous population of interictal spikes. B: Simultaneous scalp EEG record-
Intracranial spikes associated with scalp spikes were al-
ways easily discernable from ongoing activity, had a high
tacts. The gyral area of scalp spike sources was esti-
demonstrating concurrent depolarization.
As previously observed in patients implanted for
epilepsy surgery, most cortical interictal discharges
recorded from subdural or depth electrodes were not ev-
ident in the scalp EEG (9,10). Cortical interictal spikes
recorded from these patients with temporal lobe epilepsy
were heterogeneous in source location, area, synchrony,
and amplitude. Cortical source area appeared to be the
most significant variable in determining whether spikes
were recordable from the scalp. Only a small fraction
of cortical spikes had sufficient extent (Fig. 3). Cortical
spikes with ≤6 cm2of source area did not produce rec-
ognizable scalp potentials (Fig. 4). Cortical spikes asso-
ciated with 6–10 cm2of synchronous cerebral depolar-
ization rarely generated scalp-recordable EEG interictal
spikes (Fig. 5), whereas spike sources having an area of
≥10 cm2commonly resulted in recognizable scalp poten-
tials (Fig. 6). Furthermore, prominent scalp spikes were
often associated with the activation of 30 cm2of cortex
(Fig. 7), which is >70% of temporal lobe gyral cortex.
Among the nearly 600 intracranial interictal spikes re-
viewed, 90% of those having a cortical source area of
≥10 cm2generated recognizable scalp spikes, whereas
only 10% of intracranial spikes with <10 cm2of cortical
area did so.
Cortical generators of EEG spikes produce three-
ing the cerebral sources of scalp EEG potentials has been
challenging (14,15). Inherent ambiguities exist in EEG
source-modeling techniques attempting this inverse solu-
tion. These include not only the effects of an irregularly
shaped skull that distorts and attenuates the EEG signal,
but also our lack of appreciation for the extent of cerebral
sources of scalp EEG.
An early animal study by Delucchi et al. (16) showed
that the scalp acts as a spatial averager of electrical ac-
tivity and transmits only those components that are com-
mon to and synchronous over relatively large areas of the
cortex. Cooper et al. (8) proposed that 6 cm2of cortical
Epilepsia, Vol. 46, No. 5, 2005
672 J. X. TAO ET AL.
evident from this cortical source. Neither is there an organized scalp voltage field associated with the intracranial spike (C). D illustrates
the subdural electrodes recording a negative depolarization (black) during the spike. Note that the spike source area is approximately 6
cm2. LOF: left orbital frontal; LAT: left anterior frontal; LIT: left inferior temporal; LMT: left mid temporal.
Simultaneous intracranial (A) and scalp (B) EEG recording of a left temporal spike (indicated by arrow). No scalp potential is
activity was probably necessary to produce scalp-
recordable potentials. However, as explained previously,
the in vitro design of their study was hardly comparable
to the generation of human EEG. Additionally, measure-
of EEG background activity. Because visual recognition
ground, a 6-cm2source would not be likely to produce a
recognizable scalp spike, even if their estimate for neces-
sary source area was accurate.
Our results demonstrate that cortical sources of scalp
EEG spikes are larger than commonly thought. At least
activity is usually necessary to produce scalp-recordable
are common substrates for prominent scalp spikes. These
observations add to our understanding of the cortical ori-
gin of scalp EEG. They have significant implications for
both basic and clinical science in defining epileptogenic
foci based on scalp EEG spikes.
Several noninvasive techniques are attempting to char-
acterize and depict graphically the extent of cortical in-
volvement in epileptiform spikes. Among these are EEG-
triggered fMRI (17,18) and a variety of extended-source
Epilepsia, Vol. 46, No. 5, 2005
EEG SUBSTRATES OF INTERICTAL SPIKES673
nizable associated scalp potential. This potential does have an appropriate scalp voltage field (C). The approximate area of the cortical
spike source is 10 cm2(D). Active electrodes: black. LOF: left orbital frontal; LAT: left anterior frontal; LIT: left inferior temporal; LMT: left
Simultaneous intracranial (A) and scalp (B) EEG recording of a left temporal spike (indicated by arrow). Note the barely recog-
models of EEG (19–22). Most of these methods apply
statistical thresholds to their results to limit the extent of
the spike source model, yet at present none uses a priori
knowledge concerning the likely area of these sources.
Our data show that most scalp-recordable, temporal lobe
spikes originate from sources between 10 and 30 cm2in
gyral area. We have no reason to suspect that sources of
scalp EEG spikes from other areas of convexity cortex
would be substantially different. Thus publications de-
picting significantly smaller or larger cortical sources for
scalp EEG spikes are likely to be in error. Source models
using the findings of this investigation to constrain so-
lutions to a 10- to 30-cm2size should prove to be more
Identifying the source area required for spikes originat-
ing from different temporal lobe regions was beyond the
scope of this investigation. Given our customary place-
ment of the subdural grid electrode, we chose to study
spikes arising from the antero- and inferolateral temporal
cortex. These sources resulted in a spike voltage field that
was principally radial in orientation, which provides the
our results define the lower limit of source area for scalp
spikes. It is likely that other temporal and extratempo-
ral sources producing principally tangential voltage fields
would require an even larger area to result in recognizable
scalp EEG potentials.
Thus given the attenuating property of the skull, most
cortical spikes with extent of <10 cm2do not produce a
recognizable scalp EEG potential. However, this may ac-
tually serve a useful role in filtering out all but the most
significant spike sources that can recruit substantial sur-
rounding cortex. Spikes recorded with intracranial EEG
are numerous, variable, and commonly arise from multi-
focus. Accordingly, interictal activity is often considered
to have little clinical importance. However, the select sub-
population of scalp-recordable spikes may have more lo-
intracranial brethren (23).
Epilepsia, Vol. 46, No. 5, 2005
674 J. X. TAO ET AL.
associated scalp potential and temporal scalp voltage field (C). The approximate area of the cortical spike source is 13 cm2(D). Active
electrodes: black. LOF: left orbital frontal; LAT: left anterior frontal; LIT: left inferior temporal; LMT: left mid temporal.
Simultaneous intracranial (A) and scalp (B) EEG recording of a left temporal spike (indicated by arrow). Note the distinct
Although simultaneous scalp and intracranial record-
ings provide the most reliable data for assessing the rela-
tions between cortical sources and scalp EEG fields, this
technique involves two potential technical confounds. A
“breach effect” associated with the skull defect created
for subdural electrode placement could have led to more
prominent EEG potentials and an underestimation of the
necessary cortical source area in normal conditions. In
our study, the skull defect was minimal (see Fig. 1), and
the bone flap was replaced after electrode implantation.
This defect, a linear fissure <3 mm in width, was usu-
ally located above the midtemporal region and thus above
the most active anterior and inferior temporal scalp elec-
trodes. Localized EEG amplitude enhancement also was
not observed visually or with voltage-field topography in
those patients selected for study. Conversely, the Silastic
membrane of the subdural grid could have had an atten-
uating effect on the voltage field, leading to an overes-
timation of the cortical source area needed to generate
scalp-recordable spikes. This too is unlikely to have been
a significant factor because appreciable amplitude asym-
metry in EEG from the two temporal areas was not ob-
served, and any additional attenuating effects of the Silas-
that of the skull. Furthermore, similar findings of neces-
sary source area were obtained in a preliminary study that
difficult to determine precisely, if anything, these two
Epilepsia, Vol. 46, No. 5, 2005
EEG SUBSTRATES OF INTERICTAL SPIKES675
associated scalp potential and temporal scalp voltage field (C). The approximate area of the cortical spike source is 25 to 30 cm2(D).
Active electrodes: black. LOF: left orbital frontal; LAT: left anterior frontal; LIT: left inferior temporal; LMT: left mid temporal.
Simultaneous intracranial (A) and scalp (B) EEG recording of a left temporal spike (indicated by arrow). Note the prominent
possible confounds tend to offset one another. Regard-
less of these technical limitations, we believe that our in
the source area of scalp EEG spikes than previous in vitro
of frontotemporal epileptic foci. Neurology 1991;41:1425–33.
2. Gilliam F, Bowling S, Bilir E, et al. Association of combined MRI,
interictal EEG, and ictal EEG results with outcome and pathology
after lobectomy. Epilepsia 1997;38:1315–20.
3. Blume WT, Holloway GM, Samuel Wiebe. Temporal epileptogen-
esis: localizing value of scalp and subdural interictal and ictal EEG
data. Epilepsia 2000;42;508–14.
4. Foldvary N, Klem G, Hammel MS, et al. The localizing value of
ictal EEG in focal epilepsy. Neurology 2001;57:2022–8.
5. Pacia SV, Ebersole JS. Intracranial EEG substrates of scalp ic-
tal patterns from temporal lobe foci. Epilepsia 1997;38:642–
6. Sakamoto S, Takami T, Hara M, et al. Interictal patterns of cere-
bral glucose metabolism, perfusion, and magnetic field in mesial
temporal lobe epilepsy. Epilepsia 2003;44:1196–206.
7. Al-Asmi A, Bennar CG, Gross DW, et al. fMRI activation in con-
tinuous and spike-triggered EEG-fMRI studies of epileptic spikes.
8. Cooper R, Winter AL, Crow HJ, et al. Comparison of subcorti-
cal, cortical and scalp activity using chronically indwelling elec-
9. Abraham K, Ajmone-Marsan C. Patterns of cortical discharges and
their relationship to routine scalp electroencephalography. Elec-
troencephalogr Clin Neurophysiol 1958;10:447–61
10. Goldensohn ES. Neurophysiologic substrates of EEG activity. In:
Klass DW, Daly DD, eds. Current practice of clinical electroen-
cephalogrphy. New York: Raven Press, 1979:421–39.
11. Hamer HM, Najm I, Mohamed A, et al. Interictal epileptiform dis-
charges in temporal lobe epilepsy due to hippocampal sclerosis ver-
sus medial temporal lobe tumors. Epilepsia 1999;40:1261–8.
12. L¨ uders H, Wyllie E, Skipper G, et al. Basal temporal subdural elec-
13. Ebersole JS. Cortical generators and EEG voltage fields. Current
Practice in Clinical Electroencephalography. Philadelphia: Lippin-
cott Williams & Wilkins, 2002:12–15.
trocorticography: application of volume conductor theory to elec-
troencephalography. J Clin Neurophysiol 1985;2:327–54.
15. Ebersole JS. Defining epileptogenic foci: past, present and future. J
Clin Neurophysiol 1997;14:470–83.
Epilepsia, Vol. 46, No. 5, 2005
676J. X. TAO ET AL.
16. DeLucchi MR, Garoutte B, Ird RB. The scalp as an electroen-
cephalographic averager. Electroencephalogr Clin Neurophysiol
17. Gotman J, Benar CG, Dubeau F. Combing EEG and fMRI in
18. Hamandi K, Salek-Haddadi A, Fish DR, et al. EEG/function MRI
in epilepsy: the Queen Square experience. J Clin Neurophysiol
19. Hamalainen MS, Ilmoniemi RJ. Interpreting magnetic fields of
the brain—minimum norm estimates. Med Biol Eng Comput
in the brain. Int J Psychophysiol 1994;18:49–65.
21. Fuchs M, Wagner M, Kohler T, Wischmann HA. Linear and
non-linear current density reconstructions. J Clin Neurophysiol
22. Grave de Peralta R, Gonzalez S, Lantz G, et al. Non-invasive local-
ization of electromagnetic epileptic activity, I. Method descriptions
and simulations. Brain Topogr 2001;14:131–7.
23. Ray A, Tao JX, Hawes-Ebersole S, et al. Do scalp EEG spikes
accurately localize the seizure onset zone? A simultaneous scalp
and intracranial study. Neurology 2003;60(suppl a):357–8.
Epilepsia, Vol. 46, No. 5, 2005