doi:10.1093/brain/awh491
Brain (2005), 128, 1556–1570
Electrocorticographic high gamma activity versus
electrical cortical stimulation mapping of naming
Alon Sinai,1 Christopher W. Bowers,1 Ciprian M. Crainiceanu,2 Dana Boatman,1,3
Barry Gordon,1 Ronald P. Lesser,1,4 Frederick A. Lenz4 and Nathan E. Crone1
Departments of 1Neurology, 3Otolaryngology and 4Neurosurgery, The Johns Hopkins University School
of Medicine and 2Department of Biostatistics, The Johns Hopkins Bloomberg School of Public Health,
Baltimore, MD, USA
Correspondence to: Nathan E. Crone, MD, The Johns Hopkins Hospital, 600 N. Wolfe Street, Meyer 2–147,
Baltimore, MD 21287, USA
E-mail: [email protected]
Subdural electrocorticographic (ECoG) recordings in patients undergoing epilepsy surgery have shown that
functional activation is associated with event-related broadband gamma activity in a higher frequency range
(>70 Hz) than previously studied in human scalp EEG. To investigate the utility of this high gamma activity
(HGA) for mapping language cortex, we compared its neuroanatomical distribution with functional maps
derived from electrical cortical stimulation (ECS), which remains the gold standard for predicting functional
impairment after surgery for epilepsy, tumours or vascular malformations. Thirteen patients had undergone
subdural electrode implantation for the surgical management of intractable epilepsy. Subdural ECoG signals
were recorded while each patient verbally named sequentially presented line drawings of objects, and estimates
of event-related HGA (80–100 Hz) were made at each recording site. Routine clinical ECS mapping used a
subset of the same naming stimuli at each cortical site. If ECS disrupted mouth-related motor function, i.e. if it
affected the mouth, lips or tongue, naming could not be tested with ECS at the same cortical site. Because
naming during ECoG involved these muscles of articulation, the sensitivity and specificity of ECoG HGA were
estimated relative to both ECS-induced impairments of naming and ECS disruption of mouth-related motor
function. When these estimates were made separately for 12 electrode sites per patient (the average number
with significant HGA), the specificity of ECoG HGA with respect to ECS was 78% for naming and 81% for mouthrelated motor function, and equivalent sensitivities were 38% and 46%, respectively. When ECS maps of naming
and mouth-related motor function were combined, the specificity and sensitivity of ECoG HGA with respect to
ECS were 84% and 43%, respectively. This study indicates that event-related ECoG HGA during confrontation
naming predicts ECS interference with naming and mouth-related motor function with good specificity but
relatively low sensitivity. Its favourable specificity suggests that ECoG HGA can be used to construct a
preliminary functional map that may help identify cortical sites of lower priority for ECS mapping. Passive
recordings of ECoG gamma activity may be done simultaneously at all electrode sites without the risk of afterdischarges associated with ECS mapping, which must be done sequentially at pairs of electrodes. We discuss the
relative merits of these two functional mapping techniques.
Keywords: electrical cortical stimulation; electrocorticography; functional mapping; gamma band; language
Abbreviations: ECoG = electrocorticography; ECS = electrical cortical stimulation; fMRI = functional MRI;
HGA = high gamma activity; MEG = magnetoencephalography
Received September 23, 2004. Revised February 3, 2005. Accepted March 1, 2005. Advance Access publication April 7, 2005
Introduction
When assessing the risk of neurological impairment following
surgery for intractable epilepsy, cerebral neoplasms or vascular malformations, interindividual variability in the details of
#
functional anatomy, particularly with respect to language
(Ojemann et al., 1989, 2003; Branco et al., 2003; Miglioretti
and Boatman, 2003) often make it necessary to map cortical
The Author (2005). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: [email protected]
ECoG gamma versus cortical stimulation maps
function in each patient. Despite continuing advances in
functional neuroimaging, electrical cortical stimulation
(ECS) mapping is still the only method in widespread clinical
use for reversibly inducing a functional lesion. Its utility for
mapping essential language cortex has been demonstrated
in a large number of patients (Ojemann et al., 1989) and it
is widely considered the gold standard for predicting postoperative functional impairment. ECS mapping can be
done intraoperatively or extraoperatively through surgically
implanted subdural electrodes. Implanted electrodes also
allow ictal recordings of the epileptogenic zone in patients
with intractable epilepsy and provide more time for language
mapping with a comprehensive battery of tests (Lesser et al.,
1987, 1994).
When subdural electrodes are implanted for the surgical
management of intractable epilepsy, continuous electrocorticography (ECoG) is performed through them to detect the
area of seizure onset. These passive recordings may also detect
cortical activity associated with functional brain activation
and could therefore be used to map cortical function.
ECoG functional mapping, whether intra- or extraoperative,
would have several potential advantages over ECS mapping.
While ECS mapping must be done at one cortical site at a
time, ECoG mapping could be done at all cortical sites at once,
reducing the time needed for mapping and allowing more
widespread and comprehensive mapping of cortical function.
In addition, ECoG mapping would carry no risk of the afterdischarges that are often associated with ECS mapping and
that occasionally lead to seizures (Lesser et al., 1984; Blume
et al., 2004), sometimes preventing further mapping of at-risk
cortical sites. However, the success of ECoG mapping would
depend on the ability of electrophysiological indices of cortical activation to identify cortical tissue that is necessary for
normal function.
EEG activity in the gamma band (>30 Hz) has been a
particularly promising index of cortical activation for both
theoretical and empirical reasons. Its theoretical significance
stems from animal studies that have implicated gamma oscillations in the dynamic representation of information in distributed cortical neuronal networks (Singer and Gray, 1995;
Gray, 1999; Engel and Singer, 2001). Event-related gamma
activity has also been observed in human scalp EEG during
auditory (Pantev, 1995), visual (Tallon-Baudry and Bertrand,
1999) and motor (Pfurtscheller et al., 1993) tasks. Most of
these responses have been observed in relatively low gamma
frequencies near 40 Hz. However, subdural ECoG recordings
in patients with epilepsy have also revealed activity in a higher
gamma frequency range beginning at 70–80 Hz (Crone et al.,
1998b) and including a broad range of higher frequencies that
may extend up to 150 Hz (Crone et al., 2001b; Ray et al.,
2003). This broadband ‘high gamma’ activity (HGA) has been
observed during a variety of functional activation tasks,
including self-paced and visually cued limb movements
(Crone et al., 1998b; Ohara et al., 2000; Pfurtscheller et al.,
2003; Leuthardt et al., 2004), auditory discrimination (Crone
et al., 2001b) and word production tasks (Crone et al., 2001a).
Brain (2005), 128, 1556–1570
1557
ECoG studies have also shown that, compared with eventrelated activity in other frequency bands, event-related HGA
typically occurs in spatial and temporal patterns that are more
consistent with functional anatomy and the results of ECS
(Crone et al., 1998a, b, 2001b; Crone and Hao, 2002a;
Pfurtscheller et al., 2003). For example, whereas alpha desynchronization is observed in a broad distribution over bilateral
sensorimotor cortex during unilateral limb movements, HGA
occurs in a more discrete distribution over the limb’s contralateral motor representation (Pfurtscheller et al., 2003), and
HGA has a temporal profile that more closely corresponds
with the timing of stimuli and responses (Crone et al.,
1998b, 2001a, b). In ECoG recordings of auditory association
cortex in dominant superior temporal gyrus, HGA was greater
and more widespread during speech discrimination than during tone discrimination, whereas auditory evoked responses
were not appreciably different, suggesting that HGA provided a
better index of the cortical processing demands of the tasks
(Crone et al., 2001b). In addition, although ECoG power in
high gamma frequencies was consistently augmented during
cortical activation, the power in lower gamma frequencies
around the traditional 40 Hz band was sometimes increased
and at other times decreased, varying not only across patients,
but also across recording sites within patients.
Because of its spatial, temporal and functional response
characteristics, HGA appears to be a promising index of
task-related cortical activation with potential applications
in functional mapping, particularly in patients undergoing
surgery for intractable epilepsy. However, like the BOLD
response of functional MRI (fMRI), ECoG gamma activity
is a measure of cortical activation and could theoretically
overestimate cortical tissue that is necessary for function. It
is therefore important to compare HGA with ECS, and ultimately with the results of surgical resection. Similar comparisons have been made between fMRI and ECS maps of
language cortex in comparably sized groups of patients
(FitzGerald et al., 1997; Pouratian et al., 2002; Rutten et al.,
2002; Roux et al., 2003). In this paper we compared maps of
language according to event-related ECoG HGA with those
according to ECS.
Since naming is the task most often used for mapping
language with both intraoperative and extraoperative ECS,
we studied ECoG HGA during naming and compared its
spatial distribution with ECS maps of naming. In addition,
because naming is routinely performed aloud during ECS
mapping to measure the effect of stimulation, and therefore
involves muscles of articulation, we compared ECoG gamma
maps of naming with ECS maps of mouth-related motor
function.
Methods
Patients and clinical setting
Thirteen patients (see Table 1) were enrolled after surgical implantation of subdural electrode grids. Two patient had arteriovenous
malformations, and 12 patients had intractable epilepsy. Patients
1558
Brain (2005), 128, 1556–1570
A. Sinai et al.
Table 1 Patient characteristics
Patient no.
Age (years)
Handedness
Gender
Age at seizure onset
Hemispheric dominance
for language (IAP)
Full scale IQ
1
2
3
4
5
6
7
8
9
10
11
12
13
26
47
16
47
35
18
30
32
46
20
38
45
40
Right
Right
Right
Right
Right
Right
Right
Right
Right
Left
Right
Right
Right
Male
Female
Female
Male
Female
Male
Male
Male
Male
Female
Female
Male
Male
9 months
39 years
3 years
8 years
6 months
1 year
15 years
16 years
39 years
15 years
NA‡
Neonatal
35 years
Left
Left
Left
Left
Bilateral
–†
Left
Left
Left
Left
–§
Left
Left
92
116
91
–*
106
109
110
97
113
114
87
81
114
*IQ was not tested; patient completed 12 years of education and owned his business. †IAP test was not administered. ‡Patient had
an arteriovenous malformation with no history of seizures. §IAP was not administered; ECS results indicated left hemisphere dominance.
IAP = intracarotid amobarbital procedure.
were recruited for this study if subdural electrode arrays were surgically implanted for clinical purposes over the language-dominant
hemisphere (left in all patients). Patients with full-scale IQ below 80
or significant preoperative impairment of naming function were
excluded. The electrode grids were implanted for better localization
of the seizure focus and/or for functional mapping of language and
motor cortex to prevent postoperative neurological impairments.
Patients were admitted to the Johns Hopkins Epilepsy Monitoring
Unit one day following electrode implantation and remained under
continuous video and EEG surveillance for 6–14 days before a second
surgery to remove the electrodes and perform a resection of the
seizure focus, if possible. The protocol used for this study was
approved by the Johns Hopkins Hospital Institutional Review
Board in compliance with the Declaration of Helsinki. Patients
volunteered to participate and gave consent for research ECoG
recordings in accordance with this protocol.
Subdural electrode placement and
localization
The subdural electrode arrays consisted of 1.5-mm-thick, soft Silastic
sheets embedded with platinum-iridium electrodes (4-mm diameter,
2.3-mm diameter exposed surface) that were equally spaced with
1 cm centre-to-centre distances (Adtech, Racine, WI, USA). The
location of these electrodes with respect to underlying cortical
gyral anatomy was determined by coregistration of preimplantation volumetric brain MRI (1- to 1.8-mm coronal slice
thickness) with post-implantation volumetric brain CT (1-mm
axial slice thickness) according to anatomic fiducials (Curry, Compumedics Neuroscan, El Paso, TX, USA). Electrode positions derived
from post-implantation CT were then displayed with a brain surface
rendering derived from the pre-implantation MRI.
Electrical cortical stimulation
All patients underwent functional ECS mapping of motor and language cortex according to routine clinical procedures (Lesser et al.,
1987). ECS testing was done in 2-h blocks, twice a day, for 2–4 days.
ECS mapping utilized constant current electrical stimulation
between pairs of adjacent electrodes using a Grass S-88 or S-12
cortical stimulator (Grass-Telefactor/Astro-Med, Inc., West
Warwick, RI, USA) with 1- to 5-s trains of 50 Hz, 0.3 ms, alternating
polarity square-wave pulses, starting with a stimulus intensity of
1 mA and increasing in 0.5- to 1.0-mA increments up to a maximum
of 15 mA. Stimulus intensity at each electrode pair was individualized
according to the highest amperage below 15 mA that did not produce
after-discharges (Lesser et al., 1984). While stimulus intensity was
adjusted, each patient reported any unusual or involuntary sensations or movements, and once a maximum stimulus intensity was
reached, disruption of motor function was detected by observing the
patient during voluntary movements in turn of the tongue, bilateral
fingers and bilateral toes.
If ECS interfered with voluntary movement or produced involuntary movement or unpleasant sensations in one or more
body parts, ECS mapping of language function was usually not
performed for that electrode pair. At some electrode pairs these
sensory or motor effects were not sufficiently uncomfortable or
distracting to interfere with ECS mapping of language, but this
was the exception rather than the rule. ECS interference with
speech-related motor function, i.e. involving mouth, lips or tongue,
was usually too distracting to proceed with ECS testing of language
function. Even if ECS was tolerable to the patient, ECS mapping
of language at that electrode pair was usually skipped because the
cortical region was already considered unresectable due to the risk of
motor impairment.
ECS mapping of language was performed with a battery of tasks
probing expressive and receptive language function, including
picture naming, sentence comprehension (modified Token Test),
paragraph reading and spontaneous speech. Picture naming was
performed with 84 pictures of objects derived from the Boston
Naming Test (Goodglass et al., 1983). Each object was depicted
by a black and white line drawing on paper or on a computer
monitor. ECS current was begun immediately before stimulus
onset and lasted no longer than 5 s. Naming errors could consist
of absent or delayed responses, incorrect responses, or paraphasias.
ECS-induced functional impairment was confirmed only if there was
an error induced by ECS current during at least two stimulus trains
and if the errors during ECS stimulation could be clearly differentiated from the patient’s baseline performance on the same task
without ECS stimulation.
ECoG gamma versus cortical stimulation maps
Subdural ECoG recordings
ECoG signals were amplified (5 · 1000) and filtered (1–300 Hz,
6 dB/octave) using Grass amplifiers (Model 12A5; GrassTelefactor/Astro-Med, Inc.). All ECoG recordings were made
with a referential montage using a single subdural reference electrode chosen for its relative inactivity and greatest distance from
electrodes recording from the areas of interest. ECoG recordings in
two patients were limited to a maximum of 128 channels; otherwise,
recordings were limited to a maximum of 64 ECoG channels. All
ECoG signals were remontaged to an average reference to obtain
reference-independent topographic maps of spectral measures
(Crone et al., 1998a). The amplified ECoG signals were digitally
recorded (sampling rate = 1 kHz) in parallel with experimental
markers of stimulus onset and offset. All ECoG recordings were
made after the patients had recovered from surgical implantation
of subdural electrodes and only if the patients had not had a seizure
in the previous 2 h.
During ECoG recordings, picture naming stimuli were sequentially presented to the patients one at a time on a video monitor.
These stimuli consisted of 84 objects (black and white drawings)
from which stimuli for ECS mapping were derived. The order of
stimulus presentation was randomized for each ECoG recording.
Patients were asked to fixate on a black dot on a white background
between stimulus presentations, and the onset of each visual stimulus
began 400 ms after disappearance of the fixation point, at least 3 s
following the patient’s response to the previous stimulus. Vocal
responses were monitored with a voice trigger. Picture stimuli
were replaced with the fixation point once a vocal response was
detected. Trials were excluded from analysis if the patient’s responses
were absent, delayed or incorrect, or if the patient was interrupted or
distracted during the stimulus-response trial or in the 1-s timeframe
preceding it. Trials that were contaminated with epileptiform activity
were also excluded from further analysis.
ECoG signal analysis
ECoG recordings during the picture naming task were analysed for
event-related changes in the ECoG power spectrum, focusing particularly on event-related HGA. A detailed description of the ECoG
signal analyses used in this study has been published elsewhere
(Crone et al., 2001b). In brief, after subtracting the average evoked
potential, ECoG signals were filtered in the high gamma
(80–100 Hz) band, then squared and averaged across trials. The
resulting time series were segmented into a series of 100-ms epochs
with 50% overlap, spanning 1 s before and 3 s after stimulus onset.
Statistical analyses of ECoG gamma power have also been described
in detail elsewhere (Crone et al., 1998a). In brief, estimates of ECoG
gamma power were compared between post-stimulus epochs and a
pool of pre-stimulus (baseline) epochs. A separate mixed effects
regression model was fitted for each channel using the change in
power from baseline to each post-stimulus epoch as the dependent
variable and post-stimulus epoch latency as the independent variable. The mean values and confidence limits were then backtransformed to give the geometric mean percentage change from
baseline and 95% confidence limits. An event-related change in
gamma power from baseline was considered statistically significant
if the confidence intervals did not include zero for at least four
consecutive epochs. In bootstrap simulations using samples of
detrended energy estimates this criterion was met in no more
than three out of 1000 samples.
Brain (2005), 128, 1556–1570
1559
Comparison of ECS and ECoG
functional maps
All electrode sites at which ECS affected cortical function were
counted separately for (i) sites where ECS interfered with picture
naming, and (ii) sites where ECS interfered with or produced involuntary movements of the mouth, lips or tongue, i.e. movements
involving muscles of articulation. Although ECS was performed at
pairs of adjacent electrodes, each electrode site was counted separately in order to compare ECS results with maps of HGA, which
were derived from referential ECoG recordings of individual electrodes. Therefore, we reported and discussed all results in terms of
individual electrode sites. We referred to electrode sites at which
there was statistically significant HGA as HGA(+) and those at
which there was not as HGA(), and we referred to electrode
sites at which ECS affected task performance as ECS(+) and those
at which it did not as ECS(). In accordance with routine clinical
interpretation of ECS, if an individual electrode was common to two
pairs of electrodes where ECS was performed, that electrode was
deemed ECS() if any of these pairs did not affect task performance.
Otherwise, the electrode was ECS(+) for all tasks whose performance
was affected by ECS at pairs that included the electrode. This interpretation is based on the understanding that ECS may impair the
function of cortex underlying each of the electrodes in a pair, as well
as cortex between the electrode pair (Nathan et al., 1993).
All comparisons of HGA and ECS functional maps were based on
the assumption that ECS was the ‘gold standard’ for localization of
cortical function. We calculated descriptive statistics of the sensitivity
and specificity of HGA in relation to ECS in a manner analogous to
the way a diagnostic marker for disease (e.g. prostate-specific antigen) is tested against a definitive disease diagnosis (e.g. biopsyproven prostate cancer). The sensitivity and specificity of HGA
were calculated only for sites where both HGA and ECS were tested.
HGA sensitivity was calculated as the percentage of sites that were
both HGA(+) and ECS(+) among all ECS(+) sites, and HGA specificity was calculated as the percentage of sites that were both
HGA() and ECS() among all ECS() sites. These measures
were calculated separately for electrode sites where ECS interfered
with picture naming and mouth-related motor function. Because
ECS interference with mouth-related motor function often precluded ECS testing of naming, HGA specificity and sensitivity
were also calculated for the subset of ECS-tested electrode sites
that were tested for naming, excluding electrode sites where naming
was not tested because of interference with mouth-related motor
function. Confidence intervals (P < 0.05) were calculated for the
sensitivity and specificity estimates under the assumption that, conditional on ECS, the findings of HGA were independent, identically
distributed, Bernoulli random variables.
We noticed a considerable interindividual variability amongst
patients in (i) the number of ‘active’ sites, i.e. HGA(+) with eventrelated gamma activity according to our criteria, (ii) the magnitude
of statistically significant event-related gamma activity, and (iii) the
spatial distribution of HGA(+) sites. In order to explore the selection
criteria for HGA(+) sites that would yield the optimal sensitivity and
specificity in relation to ECS functional maps, we systematically
changed our criteria for determining HGA(+) sites in two different
ways (Table 2). The first was the threshold magnitude of eventrelated HGA at which we selected electrodes as HGA(+). We studied
the effect of varying this threshold in three categories of HGA(+) sites
defined by (i) all significant sites (no threshold magnitude), (ii) only
sites at which the magnitude of significant event-related HGA was
1560
Brain (2005), 128, 1556–1570
A. Sinai et al.
Table 2 Sensitivity/specificity of ECoG HGA in relation to ECS maps: effects of different selection criteria
for HGA(+) sites
Sensitivity (%)
Mouth*
HGA magnitude threshold†
>50%
28.9
>30%
45.8
No threshold
49.4
HGA(+) sites‡
1
8.4
2
13.3
3
18.1
4
21.7
5
25.3
6
30.1
7
34.9
8
41.0
9
44.6
10
45.8
11
45.8
12
45.8
13
47.0
14
47.0
15
48.2
16
48.2
17
49.4
18
49.4
Specificity (%)
Naming
Mouth + naming
Mouth*
Naming
Mouth + naming
25.0
38.3
43.3
27.3
42.7
46.9
83.6
73.6
70.4
82.2
71.3
68.3
85.2
75.9
73.0
3.3
6.7
10.0
15.0
16.7
18.3
21.7
26.7
30.0
36.7
38.3
38.3
38.3
38.3
40.0
41.7
41.7
41.7
6.3
10.5
14.7
18.9
21.7
25.2
29.4
35.0
38.5
42.0
42.7
42.7
43.4
43.4
44.8
45.5
46.2
46.2
98.4
96.0
93.8
91.6
89.5
88.1
86.8
85.7
84.6
83.0
81.7
80.6
79.8
79.0
78.4
77.6
77.1
76.5
97.2
94.4
91.9
89.8
87.3
85.3
83.5
82.0
80.7
79.9
78.9
77.9
76.9
76.1
75.6
75.1
74.4
73.9
98.7
96.5
94.5
92.9
90.7
89.4
88.4
88.1
87.5
86.8
85.5
84.2
83.3
82.3
82.0
81.4
80.7
80.1
*ECS interference with mouth-related motor function. †Threshold for designating HGA(+) sites according to percentage
increase of HGA (power in 80–100 Hz band) over baseline. ‡The maximum number of HGA(+) electrode sites
allowed for calculations of sensitivity/specificity.
>30% above baseline, and (iii) only sites with significant eventrelated HGA >50% above baseline (Table 2). The second way by
which we varied our criteria for HGA(+) sites was to limit the total
number of HGA(+) sites to be allowed for each patient in our calculations of HGA sensitivity and specificity. This was done to reduce the
influence of individual differences in the magnitude (and therefore
the statistical significance) of HGA and the number of HGA(+) electrode sites, without arbitrarily selecting a threshold of HGA magnitude. In each patient the HGA(+) electrode sites were ordered
according to the magnitude of statistically significant event-related
HGA, and the additive specificity and sensitivity of HGA were calculated for one to 18 of the sites with the greatest event-related gamma
magnitude (Table 2). Note that some patients had <18 HGA(+) sites
and therefore did not contribute to the calculations beyond their total
number of HGA(+) sites.
A lateral view of electrode positions over the left hemisphere
was taken from each patient and normalized to a single schematic
lateral brain image. All ECS(+) sites for naming and mouth-related
motor function, and HGA(+) sites up to a limit of 12 sites per patient,
were displayed in separate groups in the common coordinates of the
schematic brain (Fig. 1), and the coordinates of the electrode sites
were measured to evaluate the spatial distributions for each group
of sites.
To further investigate the association between HGA and ECS
functional maps, a x2-test was performed to test the null hypothesis
that HGA results were independent of ECS results. For each electrode
site studied we obtained one of four possible outcomes for the two
mapping methods (ECS, HGA): (+,+), (+,), (,+) or (,). In this
coding, for example, (+,) corresponded to a site that was ECS(+)
and HGA(), i.e. ECS induced impairment without event-related
gamma activity. Given the paired nature of ECS and HGA testing at
each electrode site, a 2 · 2 table was constructed in which each cell
contained the number of observed pairs of (ECS,HGA) results. The
corresponding 2 · 2 table was then used in a chi-square test of the
independence of ECS versus HGA maps. The contribution of each of
the individual cells was also calculated separately to show the largest
cell contributions to the overall test statistic.
ECS and ECoG functional maps and
postoperative outcome
In the 10 patients who underwent surgical resection, the extent of the
resection and its relationship with the results of ECS and ECoG
mapping were determined from (i) preoperative and operative
notes and/or (ii) postoperative neuroimaging (available in four
patients). Postoperative outcome was assessed with formal cognitive
testing in seven patients who underwent surgical resection. Naming
performance was assessed with the Boston Naming Test (Goodglass
et al., 1983). Verbal fluency was assessed with the FAS test
(Tombaugh et al., 1999). Verbal memory was assessed with the
Rey Auditory Verbal Learning Test (Western Psychological Services,
Los Angeles, CA, USA). Outcome information in the remainder was
drawn from informal clinical assessments found in patients’ postoperative medical records.
ECoG gamma versus cortical stimulation maps
Brain (2005), 128, 1556–1570
1561
Fig. 1 Spatial distribution of ECoG and ECS maps from all 13 patients in common coordinates. (A) Electrode sites where both ECoG and
ECS mapping was performed. (B) Electrode sites with event-related ECoG HGA during naming, limited to the 12 sites per patient with
greatest HGA magnitude. (C) Electrode sites where ECS interfered with or produced involuntary movements of mouth, lips or tongue.
(D) Electrode sites where ECS interfered with naming.
Results
Maps of ECS and ECoG HGA
The electrode sites tested with both ECS and ECoG HGA are
illustrated in Fig. 1A for all patients in common coordinates.
The number of electrode sites (average 6 SD) tested for ECoG
HGA (60.6 6 18.5) was greater than the number tested with
ECS (44.2 6 18.7) in 11 of the 13 patients, resulting in many
HGA(+) results at sites that were not tested by ECS. On the
other hand, most sites tested with ECS were also tested for
HGA. This may reflect the fact that ECoG recording could be
performed in parallel on all subdural electrodes at once. In
contrast, ECS testing was carried out sequentially on pairs of
electrodes and was therefore more time consuming, requiring
prioritization of electrode sites for ECS testing.
Results from both ECS and ECoG HGA were obtained in
an average of 34.9 6 13.1 electrode sites per patient. Of these
sites, ECS interfered with naming in an average of 7.6 6 5.7
sites. ECS interfered with mouth-related motor function in an
average of 8.6 6 7.2 sites. The average number of all HGA(+)
sites in each patient was 11.6 6 10.1. Figures 2–4 illustrate the
neuroanatomical relationships between the results of ECS and
ECoG HGA in three different patients. Note the differences
between these patients in the cortical regions covered by
subdural electrodes, determined solely by clinical considerations. The plots of event-related ECoG HGA in these patients
illustrate the variability of this activity across patients and
across electrode sites within patients. This variability led us
to investigate the effect of different thresholds of HGA on our
sensitivity/specificity estimates (see below). These patients
also illustrate that there were electrodes where ECoG was
recorded, but ECS mapping was not performed, in some
cases because of stimulation-induced pain (yellow bars in
Figs 3 and 4).
The neuroanatomical distribution of HGA(+) and ECS(+)
sites exhibited a good deal of interindividual variability. This
is evident in the grouped data illustrated in common topographical coordinates (Fig. 1B–D). However, the anatomical
variability of sites that were ECS(+) for mouth-related motor
function appeared to be less than that of sites that were
ECS(+) for naming. The SDs of the distances from the
mean location of HGA(+) and ECS(+) sites were computed,
and the SD for ECS(+) mouth-related motor sites was about
51% of the SD for ECS(+) naming sites.
The distribution of sites that were ECS(+) for mouthrelated motor function was concentrated near posterior
inferior frontal gyrus and inferior paracentral cortex in the
frontal-parietal operculum (Fig. 1C). Sites that were ECS(+)
for naming were most concentrated near the superior temporal gyrus and tended to exclude the aforementioned regions
where ECS commonly disrupted mouth-related motor function. This was likely due in large part to the fact that naming
was not routinely tested at sites that were ECS(+) for mouthrelated motor function. The distribution of sites that were
HGA(+) during naming appeared to encompass both regions
1562
Brain (2005), 128, 1556–1570
A. Sinai et al.
Fig. 2 Comparison of event-related ECoG HGA with ECS maps in patient 3. White circles denote electrode sites where ECoG was
recorded. Yellow plots show the magnitude of HGA as a percentage change (y-axis) with respect to baseline. The onset of the pictured
object to be named occurred 400 ms after disappearance of a fixation point at 0 s (x-axis). Coloured bars join electrode pairs where ECS
mapping was performed, colour-coded for the occurrence and types of functional effects. Note that ECS was not performed at some
electrode sites where ECoG was recorded.
Fig. 3 Comparison of event-related ECoG HGA with ECS maps in patient 11, as in Fig. 2. Yellow bars indicate stimulation-induced pain,
precluding functional mapping with ECS.
ECoG gamma versus cortical stimulation maps
Brain (2005), 128, 1556–1570
1563
Fig. 4 Comparison of event-related ECoG HGA with ECS maps in patient 4, as in Figs 2 and 3. The extent of the temporal lobectomy
is indicated by the shaded area.
with mouth-related ECS(+) sites and regions with sites
ECS(+) for naming. Estimates of the sensitivity and specificity
of ECoG HGA were therefore also made by combining sites
that were ECS(+) for mouth-related motor function with
those that were ECS(+) for naming (see below).
Sensitivity and specificity of ECoG HGA
The sensitivity and specificity of ECoG HGA were calculated
only for sites where both ECS and ECoG mapping were performed (Fig. 1A and Table 2). Calculations for all 454 of these
sites across the entire group of patients yielded a decreasing
HGA specificity as the maximum number of allowed HGA(+)
sites per patient increased and as the threshold gamma magnitude for HGA(+) sites was reduced; an opposite trend was
observed for estimates of HGA sensitivity. When the average
number of HGA(+) sites across patients, i.e. 12 sites, was used
as the maximum number of allowed HGA(+) sites for each
patient, the specificity of HGA was 77.9 6 3.5% (95% confidence interval) relative to ECS() sites for naming and
80.6 6 3.1% relative to ECS() sites for mouth-related
motor function. The same number of allowed HGA(+)
sites yielded a sensitivity of 38.3 6 3.5% relative to ECS(+)
naming sites and 45.8 6 4.0% relative to ECS(+) sites for
mouth-related motor function.
When the criterion for HGA(+) sites was set at different
thresholds for the magnitude of event-related gamma activity
during naming, the estimates of HGA sensitivity and specificity varied in a manner similar to the effects of varying the
maximum number of allowed HGA(+) sites (Table 2). When
1564
Brain (2005), 128, 1556–1570
A. Sinai et al.
the HGA(+) criterion required the magnitude of significant
event-related HGA to exceed a threshold of 50% above baseline, the estimated specificity of HGA was 82.2 6 3.1% relative
to ECS() naming sites and 83.6 6 3.1% relative to ECS()
sites for mouth-related motor function sites. The equivalent
HGA sensitivities were 25 6 3.2% for ECS(+) naming sites
and 28.9 6 3.3% for mouth-related ECS(+) sites.
The sensitivity and specificity of HGA were also calculated
after combining ECS sites for naming and mouth-related
motor function. This was done because the distribution of
sites that were HGA(+) during naming appeared to encompass both regions with mouth-related ECS(+) sites and
regions with sites that were ECS(+) for naming or language
function. Also, because naming was usually not tested if a site
was ECS(+) for mouth-related motor function, sites that were
ECS(+) for naming were almost mutually exclusive of sites
that were ECS(+) for mouth-related motor function. Since the
naming task used for ECoG recordings (and for ECS mapping) required verbal responses, presumably activating cortical sites responsible for mouth-related motor function, we
estimated the sensitivity and specificity of ECoG HGA based
on the assumption that if sites that were ECS(+) for mouthrelated motor function could have been tested for naming,
naming would have been impaired as well. This change in
our calculations resulted in HGA sensitivity estimates that
were intermediate between those calculated separately for
ECS maps of mouth-related motor function and naming,
but the combined specificity was higher than the separate
calculations (Table 2). The sensitivity and specificity of
HGA were also calculated only at sites where ECS naming
was tested, excluding the sites where naming was not tested
with ECS because of interference with mouth-related motor
function. The sensitivities, e.g. 42.7 6 3.7 for 12 HGA(+) sites,
were similar to those calculated from all ECS-tested sites,
whereas the specificities, e.g. 84.2 6 3.4 for 12 HGA(+)
sites, were higher than those calculated for ECS(+) naming
sites among all ECS-tested sites (Table 2).
The null hypothesis of independence between ECS and
ECoG gamma maps was rejected for mouth-related motor
function (P < 0.001) and for naming (P < 0.01) after limiting
the allowable numbers of HGA(+) electrodes per patient. The
null hypothesis was rejected for naming when the number of
allowed HGA(+) sites was 10 or more, whereas it was rejected
for mouth-related motor function and for the combined
mouth and naming sites regardless of the number of
HGA(+) sites allowed. Evaluating the contribution to the
x2 statistic of each individual cell revealed that the cell contributing the most was that of electrode sites that were both
HGA(+) and ECS(+), regardless of the function that was
tested with ECS.
ECS and ECoG functional maps and
postoperative outcome
To compare how well ECS and ECoG functional maps predicted postoperative functional outcome, it would be necessary to know precisely which ECS(+) and HGA(+) sites were
resected. However, this information was not available for
many of the patients that were studied with both mapping
techniques (Table 3). Three patients did not undergo a
Table 3 Functional mapping and functional outcome of surgery
Patient
1
2
3
4
5
6
7
8
9
10
11
12
13
Surgical procedure
(extent of resection, if known)
[complications, if any]
Postoperative
MRI
ECS(+) sites
resected
HGA(+) sites
resected
Post-operative cognitive outcome
LTL + AH (5 cm)
No resection
No resection
LTL + AH (Fig. 4)
LTL + AH
LTL + AH (4.5 cm)
–
NA
NA
+
–
–
–
NA
NA
–
–
+
–
NA
NA
+†
+ /
+ /
Left frontal lobectomy
No resection
LTL + H (5–5.5 cm)
LTL only (3.5–4 cm)
Left temporal AVM resection
[perioperative infarct of left
temporal-parietal junction]
LTL + AH
–
NA
+
+
– (CT)
+ /
NA
+ /
–
+‡
–
NA
–
–
+‡
# verbal memory*
NA
NA
Impaired naming and verbal memory
No clinical change (no formal testing)
Transient mild expressive dysphasia
(no formal testing)
# verbal memory, naming unchanged
NA
# naming, # verbal memory
No change
Aphasia, right hemiparesis and hemianopia
(no formal testing)
–
+ /
Left frontal AVM resection
[venous thrombosis and infarct]
–
+ /
–
+ /
Initial word-finding difficulties resolved
within 4 months, naming unchanged,
# verbal memory
Impaired verbal fluency, naming unchanged
*Decrement without impairment, post-op cognitive testing done only two days post-op, no further testing done. †HGA + site not included
in calculations of sensitivity/specificity. ‡Multiple sites involved in perioperative infarct. # = Decrement in performance on formal testing,
without impairment; + / = Insufficient information to verify whether (+) electrodes resected; LTL = left temporal lobectomy;
AH = amygdalohippocampectomy, H = hippocampectomy without amygdalectomy; AVM = arteriovenous malformation.
ECoG gamma versus cortical stimulation maps
resection because their seizure focus appeared to be located
within functionally critical cortex. Postoperative neuroimaging was either not done, or the films were not available, in all
but four patients. Although medical records describing the
extent of resection sometimes contained sufficient information to determine whether ECS(+) or HGA(+) sites were
resected, there were several patients in whom this information
was indeterminate (‘+/’ in Table 3).
In two patients, attempts to resect an arteriovenous malformation (AVM) were complicated by perioperative infarcts
extending beyond the planned margins of resection. In Patient
11 (Table 3) the perioperative infarct involved a large portion
of the left temporal-parietal junction and included several
cortical sites that were ECS(+) and/or HGA(+). This patient
had a severe postoperative aphasia, as well as a right hemiparesis and a right homonymous hemianopsia. However, the
large extent of this patient’s lesion makes it difficult to draw
conclusions about the predictive validity of ECS or ECoG
event-related HGA.
Patient 6 suffered a mild transient expressive dysphasia
following a left temporal resection that definitely included
ECS(+) sites; however, HGA(+) sites may have also been
resected. In contrast, Patient 4 suffered severe and persistent
naming and verbal memory impairments following resective
surgery that included at least one HGA(+) site, but no ECS(+)
sites. This was confirmed by coregistration of the subdural
electrode coordinates from a post-implantation/pre-resection
CT scan with a postoperative volumetric MRI demonstrating
the extent of resection (Fig. 4). However, a large portion of the
resected temporal lobe was not mapped, either because it was
not covered with electrodes (no ECoG or ECS) or because of
stimulation-induced pain (no ECS).
Of the two patients in whom it could be definitively said
that no ECS(+) or HGA(+) sites were resected, one had a
decrement (but no impairment) of verbal memory and no
change in naming performance. Cognitive testing in the other
patient revealed no change in naming or memory.
In the remaining four patients with resections (patients 5, 7,
9 and 12), there was insufficient information about the extent
of resection (e.g. no postoperative neuroimaging) and its
relationship to the subdural electrodes that were mapped,
to determine definitively whether ECS(+) or HGA(+) sites
were resected. Postoperative outcome in these patients was
mixed, ranging from no clinical impairment (without formal
testing) in patient 5, to decrements in naming and/or verbal
memory (without impairment) on formal testing in patients
7, 9 and 12.
Discussion
General considerations
An ideal evaluation of the utility of ECoG HGA for functional
mapping would consist of its comparison with postoperative
neuropsychological outcome following the surgical resection
of ECoG-mapped cortical tissue. Although ECoG mapping
Brain (2005), 128, 1556–1570
1565
would be validated in part by the absence of impairment
following resection of HGA() cortical tissue, an equally
important validation would consist of impairment following
resection of HGA(+) cortex. However, assuming the surgical
plan cannot yet take ECoG maps into account because their
clinical utility is still being investigated, such a finding is likely
to occur only if (i) ECS is discordant with ECoG, i.e. ECS()/
HGA(+), (ii) ECS mapping cannot be done because of
stimulation-induced pain (Lesser et al., 1985) or excessive
afterdischarges (Lesser et al., 1984, 1999; Blume et al.,
2004), or (iii) ECS-induced impairments are disregarded
because of an overriding concern for postoperative seizure
control or tumor resection. Given our findings that HGA(+)
tissue is often ECS(+), we expect that HGA(+) tissue will only
rarely be resected and that a large number of patients will be
required to make this critical test possible.
The data available from this study are too limited to draw
definitive conclusions about the relationship between postoperative cognitive outcome and whether HGA(+) tissue was
resected. Nevertheless, HGA(+) tissue was either resected
(patient 4) or destroyed by perioperative infarction (patient
11, during resection of an AVM) in two patients, and both had
postoperative language impairment. In patient 11, many
ECS(+) sites were also involved in the perioperative infarct.
In patient 4, ECS was negative in all tested electrode sites that
were resected; however, there were several electrode sites that
were not tested with ECS because of pain, and a good portion
of the resected basal temporal neocortex was not covered by
electrodes and thus was not mapped with either ECoG or ECS.
In four patients (patients 7, 9, 10 and 12), all the cortical sites
that were resected were HGA(). Patients 7 and 10 had no
postoperative change in naming performance, patient 9
demonstrated a decrease (without impairment) in naming
and patient 12 experienced transient word-finding difficulties,
but was unchanged in naming performance on subsequent
formal testing. It could not be determined with certainty
whether ECS(+) sites were included in the resection in the
latter two patients. These results, though provocative, are still
insufficient to demonstrate the predictive ability of ECoG
HGA, and also underline the methodological difficulties of
comparing either ECoG or ECS with postoperative outcome.
Even when the extent of resection is well-defined (patient 4),
cortical regions that are not covered with electrodes or cannot
be mapped are nevertheless often included in the resection.
Since a comprehensive evaluation of ECoG mapping against
postoperative cognitive outcome is still forthcoming, we used
the best available alternative, i.e. ECS mapping, to evaluate the
utility of ECoG mapping using event-related HGA as an index
of functional activation.
The widespread clinical use of ECS is based on numerous
studies (for a recent review see Ojemann, 2003) that have
supported its utility in preventing postoperative deficits.
ECS is currently considered the gold standard for functional
mapping in humans, against which newer mapping procedures, notably fMRI, have been measured (FitzGerald et al.,
1997; Simos et al., 1999; Pouratian et al., 2002; Rutten et al.,
1566
Brain (2005), 128, 1556–1570
2002; Roux et al., 2003). However, ECS is not an absolute gold
standard, but is rather the best method that is currently available. Concerns about the accuracy of ECS have arisen from
several sources. For example, ECS maps of sensorimotor cortex have demonstrated motor responses in cortex up to 4.7 cm
anterior and 3.4 cm posterior to the central sulcus (Nii et al.,
1996), yet postoperative motor deficits usually do not occur
without lesions of relatively circumscribed and somatotopically specific regions in and around the banks of the central
sulcus. Likewise, ECS of basal temporal cortex, including
anterior fusiform gyrus and inferior temporal gyrus, often
interferes with a variety of language tasks, but these same
language tasks are usually unimpaired following its resection
(Luders et al., 1991; Krauss et al., 1996). These observations
and others (see below) have suggested that ECS may sometimes overestimate the cortical territory that is critical to
function and thus may underestimate the territory that is
safe for resection. For this reason it seems prudent for the
time being to compare all functional mapping techniques,
including ECS, with one another in order to understand
the different perspectives that each method provides on cortical function, and to improve the utility of each for predicting
postoperative deficits.
With the aforementioned caveats in mind, we compared
maps of naming according to event-related ECoG HGA
(80–100 Hz) with ECS maps of the same or related functions. The results of this study suggest that ECoG HGA is a
reasonably specific but relatively insensitive screening test for
predicting ECS-associated interference with naming. Our
findings are not consistent with the widespread assumption
that mapping methods based on cortical activation, e.g.
fMRI, PET, EEG and MEG, will detect a larger area of
activation than the area in which ECS and other lesions
impair function. Maps based upon the effects of lesions,
as with ECS, are assumed to include only those cortical
regions without which function would not be possible.
Activation-based mapping methods are assumed to also
identify regions that are participating but are not necessarily
critical for function (Sergent, 1994). These assumptions have
been supported by three studies comparing fMRI with ECS
(FitzGerald et al., 1997; Pouratian et al., 2002; Rutten et al.,
2002). In these studies, fMRI activations predicted ECSinduced impairment with relatively high sensitivity but
low specificity. However, these studies used batteries of
four to five fMRI language tasks in comparisons with intraoperative ECS maps. fMRI maps of multiple language tasks
were compared with intraoperative ECS maps of naming
alone in two of these studies (Pouratian et al., 2002; Rutten
et al., 2002), and in the remaining study (FitzGerald et al.,
1997) intraoperative ECS mapping was guided by preoperative fMRI maps. A recent study comparing fMRI with ECS
in 14 patients undergoing surgery for brain tumours (Roux
et al., 2003) found that fMRI performed with naming alone
had relatively low sensitivity but high specificity for
detecting intraoperative ECS(+) cortical sites, similar to
our findings.
A. Sinai et al.
Potential explanations for the discrepancies between ECoG
and ECS maps, particularly the low apparent sensitivity of
HGA, can be divided into (i) the methodological limitations
of our comparison of ECoG with ECS, (ii) the limitations of
HGA as a functional mapping tool, and (iii) the limitations of
ECS as a gold standard for comparison.
Methodological limitations of comparing
ECoG and ECS
Fundamental differences between ECoG and ECS complicate
any comparison of their functional maps. ECoG yields an
estimate of cortical activation, the magnitude of which varies
over time and correlates temporally with stimuli and
responses, and by inference with cortical processing (Crone
et al., 1998b, 2001a, b). There is no a priori knowledge of what
magnitude of HGA is indicative of cortical processing, or
more fundamentally, of how much cortical processing is
necessary and sufficient for function. Like fMRI and other
activation-based methods, ECoG mapping requires the setting of a threshold that must be arbitrarily or empirically
derived. In contrast, ECS typically yields a result that is essentially all-or-none and does not provide information about the
amount or timing of cortical processing that was disrupted,
except under experimental conditions (Hart et al., 1998).
Because of the all-or-none, time-invariant nature of ECS, a
significant amount of data reduction was required of our
ECoG data to compare it with ECS, and this may have affected
our results in ways that are yet to be discovered.
Another difference between the two methods arose from
the fact that ECS mapping of language was not usually performed in electrodes where it interfered with mouth-related
motor function. This required us to use ECS interference with
mouth-related motor function as a surrogate for ECS interference with motor activity specific to speech production. However, among the electrode sites that were ECS(+) for mouthrelated motor function there may have been some sites that
were not critical for speech production. Although silent naming could have been used for ECoG mapping, it is not routinely
used for ECS because of the need to verify the presence or
absence of a response and therefore would have presented
other problems for comparisons of the two methods.
Our comparison of ECoG and ECS was facilitated by the
fact that both methods used the same subdural electrodes, and
therefore the same coordinate space, making coregistration of
their maps more straightforward than comparisons of other
methods. However, the stimulating current of ECS was passed
between pairs of electrodes and may have affected cortical
tissue that was not within the field of view of ECoG recordings. Likewise, even if ECS exerted its effect through cortical
tissue underlying only one of the pair of electrodes, both
electrodes could have been considered ECS(+). Both of
these sources of inaccuracy could have had the effect of artifactually reducing the apparent sensitivity of ECoG HGA. In
addition, ECoG was recorded from many electrodes that were
not tested with ECS. These electrode sites could not be
ECoG gamma versus cortical stimulation maps
included in our calculations of HGA sensitivity/specificity or
of the overall association of the two methods, and might have
changed our estimates if they could have been. There were also
many regions that were simply not covered by electrodes and
for which neither method could be used, and because clinical
circumstances necessarily dictated the placement of subdural
electrodes, there was a good deal of variability across patients
in the number and location of electrodes.
Potential limitations of ECoG HGA
In spite of the necessary limitations of our methodology for
comparison, our estimates of HGA sensitivity with respect to
ECS were lower than expected and suggested that HGA might
not have detected activation of some cortical regions critical
to function. One potential explanation for this would be that
our subdural ECoG recording apparatus did not record all the
HGA associated with cortical activation. Given the low amplitude of gamma activity in macroelectrode recordings such as
subdural ECoG, this is an important consideration. Based
upon the common observation that the amplitudes of field
potential oscillations logarithmically decline with increasing
frequency, it is generally accepted that higher frequency oscillations are generated by coherent neuronal assemblies with
smaller or more dispersed neuronal aggregates (Menon et al.,
1996). ECS could potentially interfere with processing in cortical regions that are too small to generate enough gamma
activity to be recorded by our apparatus. In addition, it is
possible that the 1-cm centre-to-centre spacing of the
implanted electrode array provided insufficient spatial sampling to capture all the gamma activity that was present. ECS
may have interfered with function in a larger region of cortex,
particularly between electrodes, than the field of view of
ECoG, and it is possible that the gamma activity generated
by these regions simply ‘fell between the cracks’. Future studies with more densely spaced electrode arrays may help
determine whether this was a likely explanation for our findings, and in turn may improve the sensitivity of HGA.
It is also possible that the high gamma band we used was
not optimal for comparison with ECS maps. Our choice of
this band was based upon previous studies of event-related
ECoG gamma activity in both low (40 Hz) and high
(>70 Hz) gamma bands (Crone et al., 1998b, 2001b; Ohara
et al., 2000; Pfurtscheller et al., 2003). These studies found that
HGA was more consistently present in cortical regions where
functional activation was expected. Based upon these studies
HGA appeared to be a good candidate for a simple, general
index of cortical activation. However, the functional significance of HGA, as well as its relationship to event-related
changes in other frequency bands, including the traditional
40 Hz band, is still a topic of active investigation (Pfurtscheller
and Lopes da Silva, 1999b; Crone and Hao, 2002b; Kaiser and
Lutzenberger, 2003; Herrmann et al., 2004). These considerations underline the fact that ECoG spectral mapping is yet a
new method and that like fMRI, many of its parameters will
require adjustment for optimal performance.
Brain (2005), 128, 1556–1570
1567
Potential limitations of ECS
It is also important to consider the potential limitations of the
ECS gold standard against which we compared ECoG HGA. It
is possible that ECS is less specific than assumed, and conversely, that ECoG HGA is more sensitive than it appears
upon comparison with ECS. Previous studies of ECS functional mapping of motor cortex (Nii et al., 1996) and language
cortex (Luders et al., 1991; Krauss et al., 1996) have suggested
that it may sometimes overestimate functionally critical cortex and underestimate cortex that is safe for resection. In
the present study, the grouped data (Fig. 1B–D) showed
that ECoG gamma had a more restricted distribution over
perisylvian cortical regions known from lesion studies to be
most critical for naming and other language functions
(Gordon, 1997).
The specificity of ECS with respect to surgical outcome
could potentially be reduced if cortex that is essential for
function were deactivated by ECS in cortex that is itself
not essential but has strong functional connectivity with
essential cortex. This effect could take place by transynaptic
interference with normal network activity in essential cortex
and/or by diaschisis from deactivation of functionally interconnected cortex. In a recent case report (Ishitobi et al., 2000),
ECS of basal temporal language cortex produced aphasic
symptoms in association with intra-stimulus remote discharges in posterior superior temporal gyrus. Resection of
ECS(+) sites in basal temporal cortex did not produce language deficits. Although the ECS current itself is likely concentrated within several millimeters of the pair of stimulating
electrodes (Nathan et al., 1993), studies of ECS-induced afterdischarges have shown that they often spread to electrodes
outside the immediate site of stimulation current (Lesser et al.,
1984; Motamedi et al., 2002; Blume et al., 2004). Not only
does this cast doubt on the reliability of ECS when it is associated with afterdischarges, but it also suggests a more general
potential for ECS effects outside the stimulating current field.
By simultaneously stimulating and recording through subdural electrodes, Matsumoto et al. (2004) recently found
that single pulses at intensities typical for ECS produced
cortico-cortical-evoked potentials. These potentials were
observed in basal temporal cortex after stimulating anterior
and posterior language areas, suggesting strong functional
connectivity between these areas. Likewise, evoked potentials
were elicited in anterior language areas by stimulation of
posterior language areas and vice versa. All these findings
suggest that ECS could have effects outside the immediate
area of stimulation. Such a possibility must be considered in
the clinical utilization of ECS and in its use as a gold standard
for comparison with other functional mapping techniques.
Potential clinical uses for ECoG HGA in
preoperative language mapping
The ideal methodology for functional mapping would have
sufficient sensitivity to find all essential language regions and
enough specificity to maximize the resection of diseased
1568
Brain (2005), 128, 1556–1570
tissue. Assuming that ECS meets these criteria, the methodology of ECoG mapping with HGA used in this study appears to
have good specificity but insufficient sensitivity to be considered a potential replacement for ECS. However, the favourable specificity of ECoG HGA suggests that in the clinical
setting of patients undergoing epilepsy surgery, HGA could
be useful for choosing cortical sites of lower priority for ECS
mapping. That is, if a cortical site were HGA(+) according to
the criteria we used in this study, it could reasonably be
expected to be ECS(+) as well. However, if a site were
HGA(), it could still be ECS(+) and should therefore be
given priority for ECS mapping, particularly if it lies in the
path of the planned resection. ECoG could therefore provide a
preliminary functional map from all implanted subdural electrodes before ECS is carried out sequentially at each cortical
site. Similar ECoG methodology could also be used in the
operating room in conjunction with ECS mapping. A functional map derived from ECoG gamma activity as defined in
this paper would necessarily be understood to complement,
but not replace the ECS map.
ECoG mapping could also be useful when ECS is not feasible. ECS sometimes produces excessive afterdischarges or
even clinical seizures, which can interfere with localization of
the focus responsible for spontaneous seizures (Lesser et al.,
1984, 1999; Blume et al., 2004). ECS may also be halted by
stimulus-induced pain, presumably through stimulation of
trigeminal afferents in dura mater and larger blood vessels
(Lesser et al., 1985). Under these circumstances, however, it
would be especially important to remember the potential
limitations of ECoG, i.e. HGA() sites might still be ECS(+).
ECoG mapping could also be used in conjunction with
preoperative fMRI. Like ECoG, fMRI has not yet been
accepted as a replacement for ECS. Direct comparisons of
fMRI with intraoperative ECS have yielded contradicting conclusions (Pouratian et al., 2002; Roux et al., 2003). Although
comparisons of ECoG or ECS with fMRI are complicated by
the coregistration of functional maps that are topographic
versus those that are tomographic, respectively, it seems prudent to obtain as much complementary information as possible given the varying strengths and limitations of these
different methodologies. Greater experience with all available
techniques may also yield additional insights into their respective roles in human functional brain mapping.
Future studies
There are a number of ways in which the utility of ECoG
spectral mapping might be improved in the future. As already
mentioned, the sensitivity of HGA could be improved by
increasing the spatial sampling rate of the ECoG sensors,
i.e. by reducing interelectrode distances. This would greatly
increase the number of recording electrodes and amplifiers to
record them, a technical challenge that should be easily met by
emerging technology. In addition, the threshold for determining if sites are HGA(+) could be varied according to the
individual considerations of each patient and perhaps each
A. Sinai et al.
cortical region, including such factors as the distance of a site
from the proposed resection and other data such as the results
of the preoperative neuropsychological testing or the presence
of a lesion on MRI. The use of additional or different language
tasks might also help in identifying essential language cortex.
To simplify our comparison of ECoG with ECS we did not
use the temporal information that is available from ECoG.
However, the timing of cortical activation is a dimension of
ECoG mapping that should not be underestimated. Previous
studies have shown that the latency of ECoG HGA varies
according to response latency (Crone et al., 1998b, 2001a),
suggesting that HGA can be used to estimate the timing of
cortical activation. The timing of cortical activation may in
turn be used to infer the stage of associated information
processing at the site where HGA is observed, analogous to
the interpretation of the latency of event-related potentials
and fields (Nobre and McCarthy, 1995; Levelt et al., 1998).
Inferences about the stage of processing that takes place at a
cortical site may provide more detailed information about
its function, and thus about the risks associated with its
resection.
Although prior studies have suggested that HGA is a good
candidate for a general index of cortical activation, it is possible that the performance of ECoG spectral mapping might
also be improved with the use of other ECoG frequency bands,
or perhaps a combination of different frequency bands with
complementary response properties. For example, both scalp
EEG and subdural ECoG studies have suggested that eventrelated desynchronization in the alpha band may be a more
sensitive but less specific index of functional cortical activation than the gamma band (Crone et al., 1998a, b;
Pfurtscheller and Lopes da Silva, 1999a; Crone and Hao,
2002a; Pfurtscheller et al., 2003). Nevertheless, its apparent
high sensitivity could improve the sensitivity of ECoG mapping, assuming the contribution of this index can be weighted
appropriately in relation to that of the gamma band. Likewise,
a different, perhaps broader, definition of the gamma band
might yield different results. Previous ECoG recordings suggested that, across patients and recording sites, the 80–100 Hz
band is most consistently reactive during functional cortical
activation (Crone et al., 2001b). However, these studies also
observed ECoG high gamma responses extending upward to
frequencies 150 Hz (Ray et al., 2003). Systematic exploration of the effects of different frequency bands on the sensitivity/specificity of ECoG spectral maps were beyond the
scope of this study but may eventually yield better diagnostic
results. More importantly, a better understanding of the functional significance of event-related changes in the energy of
different frequency bands may allow these electrophysiological phenomena to be used more effectively for both clinical
and research purposes.
Because all currently available methods for functional mapping appear to have their unique strengths and limitations,
further studies will be necessary to determine how they should
be used to guide clinical decision-making. The utility of ECoG
mapping, using event-related gamma activity and/or other
ECoG gamma versus cortical stimulation maps
indices of cortical activation, will ultimately depend on its
ability to predict postoperative functional impairments, but
further studies involving many more patients will be required
to assess the predictive power of ECoG maps from this standpoint. Until then ECoG mapping may provide functional
maps that complement those of other available methods,
namely ECS and fMRI, in patients undergoing cortical resection for epilepsy and other brain diseases.
Acknowledgements
The authors would like to thank Robert Webber, PhD,
Lei Hao, PhD, Jason Hamner, Muralikrishna Indukuri and
Darryl Jackson for technical assistance. This study was supported by NINDS R01 NS040596 (N.E.C.). The work of D.B.
was supported by NIDCD R01 DC005645. The work of B.G.
was supported by R01 NS26553 and R01 NS29973. The work
of F.A.L. was supported by R01 NS38493 and R01 NS40059.
References
Blume WT, Jones DC, Pathak P. Properties of after-discharges from cortical
electrical stimulation in focal epilepsies. Clin Neurophysiol 2004; 115:
982–9.
Branco DM, Coelho TM, Branco BM, Schmidt L, Calcagnotto ME,
Portuguez M, et al. Functional variability of the human cortical motor
map: electrical stimulation findings in perirolandic epilepsy surgery. J Clin
Neurophysiol 2003; 20: 17–25.
Crone NE, Hao L. Functional dynamics of spoken and signed word production: a case study using electrocorticographic spectral analysis. Aphasiology
2002a; 16: 903–26.
Crone NE, Hao L. The functional significance of event-related spectral
changes (ERD/ERS) from the perspective of electrocorticography. In:
Reisin RC, Nuwer MR, Hallett M, Medina C, editors. Advances in clinical
neurophysiology (Suppl to Clin Neurophysiol. Vol. 54). Amsterdam:
Elsevier Science; 2002b. p. 435–42.
Crone NE, Miglioretti DL, Gordon B, Sieracki JM, Wilson MT, Uematsu S,
et al. Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. I. Alpha and beta event-related desynchronization. Brain 1998a; 121: 2271–99.
Crone NE, Miglioretti DL, Gordon B, Lesser RP. Functional mapping of
human sensorimotor cortex with electrocorticographic spectral analysis.
II. Event-related synchronization in the gamma band. Brain 1998b; 121:
2301–15.
Crone NE, Hao L, Hart J Jr, Boatman D, Lesser RP, Irizarry R, et al. Electrocorticographic gamma activity during word production in spoken and sign
language. Neurology 2001a; 57: 2045–53.
Crone NE, Boatman D, Gordon B, Hao L. Induced electrocorticographic
gamma activity during auditory perception. Brazier Award-winning article,
2001. Clin Neurophysiol 2001b; 112: 565–82.
Engel AK, Singer W. Temporal binding and the neural correlates of sensory
awareness. Trends Cogn Sci 2001; 5: 16–25.
FitzGerald DB, Cosgrove GR, Ronner S, Jiang H, Buchbinder BR,
Belliveau JW, et al. Location of language in the cortex: a comparison
between functional MR imaging and electrocortical stimulation. Am J
Neuroradiol 1997; 18: 1529–39.
Goodglass H, Kaplan E, Weintraub S. Boston Naming Test. Philadelphia: Lea
and Febiger; 1983.
Gordon B. Models of naming. In: Goodglass H, Wingfield A, editors. Anomia.
San Diego: Academic Press; 1997. p. 31–64.
Gray CM. The temporal correlation hypothesis of visual feature integration:
still alive and well. Neuron 1999; 24: 31–47, 111–25.
Brain (2005), 128, 1556–1570
1569
Hart J Jr, Crone NE, Lesser RP, Sieracki J, Miglioretti DL, Hall C, et al.
Temporal dynamics of verbal object comprehension. Proc Natl Acad Sci
USA 1998; 95: 6498–503.
Herrmann CS, Munk MH, Engel AK. Cognitive functions of gamma-band
activity: memory match and utilization. Trends Cogn Sci 2004; 8: 347–55.
Ishitobi M, Nakasato N, Suzuki K, Nagamatsu K, Shamoto H, Yoshimoto T.
Remote discharges in the posterior language area during basal temporal
stimulation. Neuroreport 2000; 11: 2997–3000.
Kaiser J, Lutzenberger W. Induced gamma-band activity and human brain
function. Neuroscientist 2003; 9: 475–84.
Krauss GL, Fisher R, Plate C, Hart J, Uematsu S, Gordon B, et al. Cognitive
effects of resecting basal temporal language areas. Epilepsia 1996; 37:
476–83.
Lesser RP, Luders H, Klem G, Dinner DS, Morris HH, Hahn J. Cortical
afterdischarge and functional response thresholds: results of extraoperative
testing. Epilepsia 1984; 25: 615–21.
Lesser RP, Luders H, Klem G, Dinner DS, Morris HH, Hahn J. Ipsilateral
trigeminal sensory responses to cortical stimulation by subdural electrodes.
Neurology 1985; 35: 1760–3.
Lesser RP, Luders H, Klem G, Dinner DS, Morris HH, Hahn JF, et al. Extraoperative cortical functional localization in patients with epilepsy. J Clin
Neurophysiol 1987; 4: 27–53.
Lesser R, Gordon B, Uematsu S. Electrical stimulation and language. J Clin
Neurophysiol 1994; 11: 191–204.
Lesser RP, Kim SH, Beyderman L, Miglioretti DL, Webber WR, Bare M, et al.
Brief bursts of pulse stimulation terminate afterdischarges caused by cortical stimulation. Neurology 1999; 53: 2073–81.
Leuthardt EC, Schalk G, Wolpaw JR, Ojemann JG, Moran DW. A braincomputer interface using electrocorticographic signals in humans. J Neural
Eng 2004; 1: 63–71.
Levelt WJ, Praamstra P, Meyer AS, Helenius P, Salmelin R. An MEG study of
picture naming. J Cogn Neurosci 1998; 10: 553–67.
Luders H, Lesser RP, Hahn J, Dinner DS, Morris HH, Wyllie E, et al. Basal
temporal language area. Brain 1991; 114: 743–54.
Matsumoto R, Nair DR, LaPresto E, Najm I, Bingaman W, Shibasaki H, et al.
Functional connectivity in the human language system: a cortico-cortical
evoked potential study. Brain 2004; 127: 2316–30.
Menon V, Freeman WJ, Cutillo BA, Desmond JE, Ward MF, Bressler SL, et al.
Spatio-temporal correlations in human gamma band electrocorticograms.
Electroencephalogr Clin Neurophysiol 1996; 98: 89–102.
Miglioretti DL, Boatman D. Modeling variability in cortical representations of human complex sound perception. Exp Brain Res 2003; 153:
382–7.
Motamedi GK, Lesser RP, Miglioretti DL, Mizuno-Matsumoto Y, Gordon B,
Webber WR, et al. Optimizing parameters for terminating cortical afterdischarges with pulse stimulation. Epilepsia 2002; 43: 836–46.
Nathan SS, Sinha SR, Gordon B, Lesser RP, Thakor NV. Determination of
current density distributions generated by electrical stimulation of the
human cerebral cortex. Electroencephalogr Clin Neurophysiol 1993; 86:
183–92.
Nii Y, Uematsu S, Lesser RP, Gordon B. Does the central sulcus divide motor
and sensory functions? Cortical mapping of human hand areas as revealed
by electrical stimulation through subdural grid electrodes. Neurology 1996;
46: 360–7.
Nobre AC, McCarthy G. Language-related field potentials in the anteriormedial temporal lobe: II. Effects of word type and semantic priming.
J Neurosci 1995; 15: 1090–8.
Ohara S, Ikeda A, Kunieda T, Yazawa S, Baba K, Nagamine T, et al.
Movement-related change of electrocorticographic activity in human supplementary motor area proper. Brain 2000; 123: 1203–15.
Ojemann GA. The neurobiology of language and verbal memory: observations from awake neurosurgery. Int J Psychophysiol 2003; 48: 141–6.
Ojemann G, Ojemann J, Lettich E, Berger M. Cortical language localization in
left, dominant hemisphere. An electrical stimulation mapping investigation
in 117 patients. J Neurosurg 1989; 71: 316–26.
Pantev C. Evoked and induced gamma-band activity of the human cortex.
Brain Topogr 1995; 7: 321–30.
1570
Brain (2005), 128, 1556–1570
Pfurtscheller G, Lopes da Silva FH. Event-related EEG/MEG synchronization
and desynchronization: basic principles. Clin Neurophysiol 1999a; 110:
1842–57.
Pfurtscheller G, Lopes da Silva FH. Functional meaning of event-related
desynchronization (ERD) and synchronization (ERS). In: Pfurtscheller
G, Lopes da Silva FH, editors. Event-related desynchronization. Vol. 6.
Amsterdam: Elsevier Science; 1999b. p. 51–65.
Pfurtscheller G, Neuper C, Kalcher J. 40-Hz oscillations during motor behavior in man. Neurosci Lett 1993; 164: 179–82.
Pfurtscheller G, Graimann B, Huggins JE, Levine SP, Schuh LA. Spatiotemporal patterns of beta desynchronization and gamma synchronization in
corticographic data during self-paced movement. Clin Neurophysiol 2003;
114: 1226–36.
Pouratian N, Bookheimer SY, Rex DE, Martin NA, Toga AW. Utility of
preoperative functional magnetic resonance imaging for identifying language cortices in patients with vascular malformations. J Neurosurg 2002;
97: 21–32.
Ray S, Jouny CC, Crone NE, Boatman D, Thakor NV, Franaszczuk PJ. Human
ECoG analysis during speech perception using matching pursuit: a comparison between stochastic and dyadic dictionaries. IEEE Trans Biomed
Eng 2003; 50: 1371–3.
A. Sinai et al.
Roux FE, Boulanouar K, Lotterie JA, Mejdoubi M, LeSage JP, Berry I.
Language functional magnetic resonance imaging in preoperative assessment of language areas: correlation with direct cortical stimulation.
Neurosurgery 2003; 52: 1335–45, discussion 45–7.
Rutten GJ, Ramsey NF, van Rijen PC, Noordmans HJ, van Veelen CW.
Development of a functional magnetic resonance imaging protocol for
intraoperative localization of critical temporoparietal language areas.
Ann Neurol 2002; 51: 350–60.
Sergent J. Brain imaging studies of cognitive functions. Trends Neurosci 1994;
17: 221–7.
Simos PG, Papanicolaou AC, Breier JI, Wheless JW, Constantinou JE,
Gormley WB, et al. Localization of language-specific cortex by using magnetic source imaging and electrical stimulation mapping. J Neurosurg 1999;
91: 787–96.
Singer W, Gray CM. Visual feature integration and the temporal correlation
hypothesis. Annu Rev Neurosci 1995; 18: 555–86.
Tallon-Baudry C, Bertrand O. Oscillatory gamma activity in humans and its
role in object representation. Trends Cogn Sci 1999; 3: 151–62.
Tombaugh TN, Kozak J, Rees L. Normative data stratified by age and education for two measures of verbal fluency: FAS and animal naming. Arch Clin
Neuropsychol 1999; 14: 167–77.