Content uploaded by Klaus Lehnertz
Author content
All content in this area was uploaded by Klaus Lehnertz on Apr 26, 2019
Content may be subject to copyright.
Epilepsia,
41(7):81
1-817,
2000
Lippincott
Williams
&
Wilkins, Inc., Baltimore
0
International League Against Epilepsy
Clinical Research
Spatial Distribution
of
Neuronal Complexity
Loss
in
Neocortical Lesional Epilepsies
G.
Widman,
*K.
Lehnertz,
TH.
Urbach,
and
*C.
E.
Elger
Clinic
,for
Neurology, University Hospital Essen; *Department
of
Epileptology and fDepartment
of
Neuroradiology,
Medical Center,
University
of
Bonn,
Germany
Summary:
Purpose:
Nonlinear EEG analysis is valuable in
characterizing the spatiotemporal dynamics of the epilepto-
genic process in mesial temporal lobe epilepsy. We examined
the ability of the measure neuronal complexity
loss (L*)
to
characterize the primary epileptogenic area of neocortical le-
sional epilepsies during the interictal state.
Methods:
Spatial distribution of
L* (L*
map) was extracted
from electrocorticograms (n
=
52)
recorded during presurgical
assessment via subdural 64-contact grid electrodes covering
lesions in either frontal, parietal, or temporal neocortex in
15
patients. The exact location of recording contacts on the brain
surface was identified by matching a postimplant lateral x-ray
of the skull with
a
postoperatively obtained sagittal MRI scan.
Reprojecting
L*
maps onto the subject’s brain surface allowed
us to compare the spatial distribution of
L*
with the resection
range of the extended lesionectomy.
Results:
In each
of
the six patients who became seizure-free,
maximum values
of
L*
were restricted to recording sites coin-
ciding with the area of resection. In contrast,
L*
maps of most
patients who had no benefit from the resection indicated a more
widespread extent
or
the existence of additional, probably au-
tonomous, foci. The mean
of
Id*
values obtained from record-
ing sites outside the area of resection correctly distinguished
13
patients (86.7
%)
with respect to seizure outcome.
Conclu,sion,s:
Relevant information obtained from long-
lasting interictal electrocorticographic recordings can be com-
pressed to
a
single
L*
map that contributes to a spatial charac-
terization of the primary epileptogenic area. In neocortical
lesional epilepsies,
L*
allows
for
identification and character-
ization of epileptogenic activity and thus provides an additional
diagnostic tool for presurgical assessment.
Key
Words:
Neo-
cortical epilepsy-Primary epileptogenic area-Neuronal com-
plexity loss-Effective correlation dimension-ECoG.
In neocortical lesional epilepsies
(NLE)
seizure con-
trol can be achieved by surgical removal of the lesion if
the primary epileptogenic area is included
(1-5).
Usu-
ally, invasive electrophysiological techniques that record
the region of seizure origin and interictal spiking define
the extent of the area to be resected. In addition to struc-
tural imaging techniques such as magnetic resonance im-
aging
(MRI)
or cranial computer tomography (CCT),
functional techniques such
as
positron emission tomog-
raphy or single photon emission computed tomography
(6-8)
as
well as noninvasive recordings of brain electro-
magnetic activity
(9)
are considered the basis to define
the primary epileptogenic area. However, in NLE, this
area is not necessarily identical with the lesion zone
(lo),
although both areas are related in the most patients
(1 1).
Thus,
the current standard for the localization and de-
Accepted February
4,
2000.
Address correspondence and reprint requests
to
Guido Widman
at
Clinic
for
Neurology, University
Hospital
Essen, Hufelandstr.
55,
45
122
Essen, Germany. E-mail: Guido.Widman @uni-essen.de
marcation of the primary epileptogenic area is
to
record
the patient’s spontaneous habitual seizures using chronic
electrocorticography (ECoG). This method, however, in-
cludes an additional risk for the patient and, depending
on the individual seizure frequency, represents
a
time-
consuming and cost-intensive procedure for the presur-
gical work-up. In addition, the relevance of electro-
graphical features as enhanced low frequencies or spike
densities used for the interictal identification of the pri-
mary epileptogenic area is still matter of discussion
(12).
In the case of extratemporal neocortical lesions, only
67%
of the patients benefit from surgery
(1
3)
despite the
immense presurgical work-up, including unambiguous
identification of morphological lesions.
Nonlinear time series analyses (ref.
14
and references
therein) have been repeatedly applied to electroencepha-
lographic recordings of epilepsy patients. Extracting
nonlinear measures such
as
the correlation dimension or
the largest Lyapunov exponent has yielded promising
results in characterizing the epileptic disorder
(15-20).
Concerning the spatiotemporal dynamics of the epilep-
811
812
G.
WIDMAN ET AL.
togenic process, we interpreted temporary transitions
from high- to low-dimensional system states, neuronal
complexity loss
(L*),
as representing synchronization
phenomena of the activity of neurons involved. These
transitions are of high relevance for the lateralization of
the primary epileptogenic area in patients with mesial
temporal lobe epilepsy
(1
8,20). In addition, the neuronal
complexity loss proved sensitive for the investigation of
anticonvulsant drug effects on the epileptogenic process
(21).
Moreover, nonlinear measures showed that specific
and long-lasting changes in the preictal period were in-
dicative of impending seizures (22-25).
However, these findings were mainly based on analy-
ses of recordings in patients with mesial temporal lobe
epilepsy. To prove an extended applicability of nonlinear
time series analysis beyond this well-defined syndrome,
we applied these methods to recordings of patients suf-
fering from NLE of different origin. We specifically ex-
amined whether the
L*
extracted from interictal
ECoG
recordings allows spatial characterization of the primary
epileptogenic area in neocortical lesional epilepsies.
METHODS
Patient data
A total of
15
patients (nine women and six men; mean
age,
30
k
10
years; range, 9-46) with a mean age at onset
of epilepsy of
15
f
9 years (range,
3-38)
were included
in the study (Table
1).
Each patient was diagnosed as
having
a
pharmaco-resistant lesional neocortical epilepsy
of
temporal or extratemporal origin. In each case, non-
invasive diagnostical methods such
as
MRI, positron
emission tomography, single photon emission computed
tomography, or surface
EEG
could not accomplish the
concordant localization of the primary epileptogenic area
or its exact demarcation from those regions of the brain
whose resection would lead to intolerable neurological
deficits. Thus, the implantation of a subdural 64-contact
grid electrode covering the lesion was necessary (follow-
ing the Bonn protocol for presurgical assessment
(26),
approved by the local medical ethics committee). In all
patients, the grid (intercontact distance,
10
mm) was
supplemented by subdural strip electrodes. In five pa-
tients, diagnostic findings of the noninvasive phase in-
dicated an involvement of mesial temporal structures,
necessitating the implantation of additional intrahippo-
campal depth electrodes. In
12
patients, presurgical
work-up led to an extended lesionectomy. In the remain-
ing three patients, the latter was restricted because of
nearby functionally relevant areas. Postoperatively, all
patients have been in the care of our out-patient depart-
ment for more than
1
year. Informed consent was ob-
tained from all patients.
For evaluation of results, the patients were divided
into two groups according to their outcomes
(1
3):
group
A comprised six patients who were seizure-free, corre-
sponding to the outcome-class
1A,
and group
B
com-
prised nine patients in whom the resection did not lead to
complete seizure control (classes 2A-4B).
Recording techniques
ECoG
and stereoelectroencephalograms were re-
corded during the invasive phase
of
presurgical assess-
ment using an average common reference
(18).
The sig-
nal was bandpass-filtered
(0.53-85
Hz;
12
dB/octave)
and, after 12-bit
A/D
conversion, was digitally stored at
a sampling frequency of
173
Hz.
Estimation
of
L*
A total of
52
artifact-free
ECoG
recordings (mean du-
ration,
35
min) of the interictal state from the awake
patient (2-7 recordings per patient) were analyzed using
a
moving-window analysis of an effective correlation
dimension
DPtf.
Details of the methods applied have
TABLE
1.
Clinical data ofpatients
Age at Follow-up
ID“
Outcome” Age onset (months) Side Region Histology
P-01 IA
30
19
20
Right Parietotemporal Contusion
P-02 1A
29
25
25
Left Frontal Astrocytoma WHO
I
P-04
1A 21
15
40 Left
Temporal Ganglioglioma WHO
1
P-05 IA
43
13
30
Right Parietal Cortical malformation
P-06 IA
9
3
17
Left Occipitotemporal Glioneuronal harmatoma
P-07
2A
26
17
53 Left Parietotemporal Astrocytoma WHO 2
P-08 2B 18 13
68
Left Temporal Ganglioglioma WHO
1
P-09 2B
42
24
22
Right Temporal Contusion
P-I0
3A 28 18
30
Right Frontoparietal Gliosis
P-11
4A 32
4
52 Right Frontoparietd Contusion
P-12
4A
39
8
73
Left Temporal Gliosis
P-13
4A 34
19
35
Left Frontal Cystic malformation
P-14
4A
24 4
41
Left Parietal Ganglioglioma WHO
1
P-15 4B 30
9 30
Left Parietal Oligodendroglioma WHO 2
P-03 1A 46 38 28 Left Frontal Astrocytoma WHO
2-3
a
ID,
identification number of each patient.
’
Postoperative seizure status according to
(13).
&ikpsiu,
Vol.
41,
No.
7,
2000
L*
MAPS
IN
NEOCORTICAL EPILEPSIES
813
been described elsewhere
(18,23).
Briefly, data sets were
segmented into consecutive and normalized epochs of
30
sec duration
(N
=
5,190
data points) for which quasis-
tationarity may be assumed
(27).
After digitally low-
pass-filtering the data (cutoff frequency,
40
Hz; 4th order
Butterworth characteristic), correlation sums
C,
(r) and
their local derivatives
C',
(r)
were calculated for em-
bedding dimensions m
=
1
to
30
using an optimized
Grassberger-Procaccia algorithm
(28,29).
The range of
possible values for the hypersphere radius r was selected
to match the resolution of the
A/D
converter. The time
FIG.
1.
A:
Representative ex-
amples of dimension profiles ex-
tracted from an interictal
ECoG
segment (duration,
20
min). Upper
rows are recording sites (82 and
C2)
within the primary epilepto-
genic area; lower rows are remote
recording sites
(B8
and
C8).
Pro-
files were smoothed for better vi-
sualization using a three-point
moving average. B:
L*
value for
each recording site
of
the
64-
contact subdural grid electrode. In
this example, maximum values
are confined
to
recording sites co-
inciding with the area of resection
(gray bars). C: Projection of elec-
trode contacts (obtained from a
postimplant lateral x-ray expo-
sure) onto a postoperatively ob-
tained sagittal
MRI
scan. The area
of resection is marked black and
encircled white.
5
0
10
5
Deff
0
*
10
5
0
10
5
0
delay for the phase space reconstruction
(30)
was set to
one sampling point, and a Theiler cutoff of five sampling
points was used to limit autocorrelative effects
(31).
A
reliable estimation of
D2eff
requires the existence of
a
"scaling region" of
a
sufficient length. However,
in
the
analysis of electroencephalographic data, this require-
ment is hardly fulfilled in the strict sense. We therefore
defined
a
"quasiscaling" that had to meet the following
weaker requirements: starting from the one-dimensional
embedding (m
=
l),
the upper bound r, of the quasis-
caling was defined
as
the r value at which
C'
,(r) exceeds
t
-1
1,,,,1/,//11,
l""1"'~
I""
0
5
10
15
20
t
[min]
Epilepsiu,
Vol.
41,
NO.
7,
2000
814
G.
WIDMAN
ET
AL.
0.95. The lower bound rl was reached when CZ5(r)
>
1.05 C’25(ru) for decreasing r.
D;If
was assigned the
average of C’zs (r) values within the range [rl, rJ, pro-
vided the interval [r,, r,] was long enough (at least one
octave, i.e., at least five consecutive C’2s(r) values, con-
tributed to the average) and
DZCff
was lower than a theo-
retical resolution limit (32) of 210g1,,(N)
=
7.43.
In
all
other cases,
Dgff
was set to a fixed value of
D,,
=
10.
The time-independent measure
L*
was extracted from
the resulting
Dzeff
profiles (Fig.
1A).
L*
represents the
area between the
DZcff
profile and the
D,,
line, normal-
ized with respect to the observation time
(18).
For each
recording contact of the grid, means of
L*
values
(F)
were obtained by averaging results obtained for all
ECoG
data sets under consideration. Because absolute
values of
DPrs
and
L*
depend on recording, filtering, and
calculation parameters, only relative differences can be
investigated. The resulting spatial distribution
(L*
maps) was normalized (i.e., mapped to
to,]]),
thus al-
lowing for comparison of results obtained from different
patients (Fig.
lB).
Matching
L*
maps and area resected
To relate the spatial distribution of to the area
resected, the location of grid electrodes was determined
with respect to anatomical landmarks.
A
lateral x-ray
exposure of the skull was performed during presurgical
assessment with the patient in
a
supine position. The film
was placed ipsilateral to the side of the implanted grid,
and the focus film distance was set to
1
50 cm to approxi-
mate a parallel path of x-rays. To identify the area of
resection, an MRI was performed
56
months after the
lesionectomy (sagittal
T1
-weighted spin echo images;
slice thickness
5
mm; interslice gap 0.5 mm). The x-ray
exposure depicting the grid was matched with the medial
sagittal MRI slice using the boundary of the calvarium,
the sella turcica, and the odontoid process
as
anatomical
landmarks. The hemisphere onto which the grid had been
implanted was segmented from cerebrospinal fluid and
scar tissue in all sagittal MRI sections using a semiau-
tomatic algorithm. Resulting brain slices were superim-
posed and projected on the medial MRT slice. Finally,
markers of the electrode contacts were projected onto the
surface of the brain (Fig.
IC).
Electrode contacts were split into two groups accord-
ing to their location with respect to the area resected. On
average,
19
contacts per patient
(+lo;
range, 7-32) were
identified within or on the boundary of the area resected
(C,,)
in group
A;
in group
B,
these values amounted
to
15
per patient (27; range, 6-25). The number
of
contacts
outside the area resected
(Gout)
was
44
(kll; range, 32-
57) for group
A
and
48
(+8;
range, 38-58) for group
B.
The differences between patient groups were nonsignif-
icant for both Ci, and C<,ut (Kolmogorov-Smirnov
Z
test).
Statistical analyses were performed on
Lati,,
and
L*oul
values, representing the mean of
F
values within and
outside the area resected, respectively.
RESULTS
L*
maps in relation to the area resected
Fig. 2 depicts typical examples of spatial
F
distribu-
tion. In all group
A
patients, maximum values and
those above the
80%
quantile were restricted to record-
ing sites covering the area of resection (Fig.
2A).
In four
group
B
patients, maximum values were also inside
FIG.
2.
L*maps projected onto the brainsurface
for
patients
4
(A),
10
(B), and 13
(C)
(Table 1) with a
50%
transparency
to
preserve
underlying brain structures. Normalized
L”
values were encoded using a color scale as shown below. Pixels between contacts were
interpolated using a two-dimensional second-order spline function (33) for better visualization
of
results.
Epikpsia,
Vol.
41,
No.
7,
2000
L*
MAPS
IN
NEOCORTICAL EPILEPSIES
815
the area
of
resection. In two of these patients, values
above the
80%
quantile were also found outside the area
of resection; whereas in two patients, maximum val-
ues were found at recording sites at the edge
of
the grid,
suggesting an insufficient coverage
of
the primary epi-
leptogenic area.
In
the five remaining group
B
patients,
maximum values were found at recording sites away
from the area
of
resection (Fig.
2B,
C).
Fig.
3
depicts the distribution of
L*
and the resected
area in relation to the grid electrode for each patient. For
comparison,
a
summary of ECoG-related findings of the
presurgical work-up, including electrocorticographical
FIG.
3.
L*
maps, extent of resection, and ECoG-related findings of the presurgical work-up for all patients. Upper left square
IS
the
distribution
of
L*;
upper right square is the resected area (black boxes) in relation
to
the grid electrode. Gray boxes indicate eloquent brain
areas as defined by electrical stimulation. Brown boxes indicate an overlap of these two areas. Lower two squares are a summary of
ECoG activity obtained during presurgical work. Left: ictal state; black boxes, seizure onset; gray boxes, seizure spread, excluding
secondary generalized seizures; brown boxes, overlap of the two activity areas. Right: interictal state; black boxes, epilepsy specific focal
activity; gray boxes, unspecific focal activity; brown boxes, overlap of the two activity areas. In patient
1
and
14,
no seizures occurred
during presurgical assessment. It should be noted that these plots present findings of ictal and interictal ECoG activity only. Other
evaluation techniques (e.g., seizure semiology or imaging) contribute as well
to
the definition of the area
of
resection
(26).
Epllr.psiu,
Vd.
41,
No
7,
2000
816
G.
WIDMAN
ET
AL.
mapping of eloquent areas and ictal and interictal
ECoG
activity obtained by visual analysis, is shown.
Ally
interpretation of these findings should account for
the location of high values in relation to functionally
relevant areas. In two group B patients (patients 14 and
IS),
electrical cortical stimulation determined the func-
tional constraints on the resection. In these patients, the
primary epileptogenic area was found close to either mo-
tor cortex or Broca’s area. Evaluation of maps of the
seven remaining group B patients showed high values
in functionally relevant areas in four
of
the patients (pa-
tients
7,
12,
13,
and 14).
Estimating the spatial extent
of
the primary
epileptogenic area
Multivariate analysis of variance showed significant
effects of seizure outcome on mean values obtained
from inside or outside the area resected (p
<
0.01).
Post
hoc univariate analysis of variance using
L*in
and
L*,,,,
values
as
dependent variables yielded significant differ-
ences between the two outcome groups for
L*otri
values
(Fig. 4). Discriminant analysis using
L*o,Lt
values al-
lowed all but two patients (patient
8
and
9)
to be distin-
guished correctly. In one
of
these cases, the highest
L*
values were restricted to few contacts at the edge of the
grid (Fig.
2).
DISCUSSION
It
is likely that the primary epileptogenic area
in
NLE
is closely related to a lesion
(11).
The demarcation of
epileptogenic tissue can be simple, as in the case of solid,
benign tumors that are well defined by structural imaging
techniques. Other lesions, e.g., dysplasias, may be more
diffuse, and imaging techniques can fail to detect such
alterations. Thus, identification and demarcation of
a
brain area that generates seizures requires a close spatial
sampling of ictal and interictal epileptiform events.
Resection of
a
lesion along with the surrounding brain
1
.o
0.8
a
e
0.6
2
0.4
E
0.2
0.0
0,
C
**
in
out
recording site
FIG.
4.
Group means and standard deviations of the neuronal
complexity
loss
t*in
and
L*,,,
at recording sites within and outside
the resection range for seizure outcome groups A (filled bars) and
B
(empty bars). Asterisks denote significant intergroup differ-
ences (p
<
0.005,
Mann-Whitney U test).
tissue may not guarantee complete seizure control
(1
1).
Postoperative persistence
of
seizures may occur for sev-
eral reasons. Known problems caused by the surgery
include epileptogenicity of the scar or a restricted extent
of resection (e.g., to avoid neurological deficits). Sei-
zures may also persist due to specific properties
of
the
epileptic process, such as secondary epileptogenesis that
may generate additional foci
(34).
Thus, although different methodologies are available
to localize the primary epileptogenic area and demarcate
it from functionally relevant areas
(3),
an unequivocal
characterization
of
the spatial extent of the epileptogenic
process requires additional information. Despite the few
patients investigated
so
far, our results suggest that rel-
evant information can be obtained from nonlinear time
series analysis of the interictally recorded electrocortico-
gram in
NLE.
By taking advantage of surgical outcome
as the most important validation criterion, we showed
that circumscribed areas
of
maximum
L*
matched the
resected area in patients who benefited from surgery. In
contrast, multiple and widespread areas
of
increased
L*
were found in all but two patients who continue suffering
from seizures despite an extended lesionectomy. These
areas were predominantly located apart from the area
resected and, in some cases, close to or within function-
ally relevant area. This suggests the existence of addi-
tional, possibly autonomous foci that were either not
identified or not sufficiently considered by conventional
methods used in the presurgical work-up
(26).
It is
ob-
vious that the presented method cannot explain persis-
tence of seizures in all cases, especially not in those cases
with an insufficient coverage of epileptogenic brain ar-
eas. In summary, our results show that the nonlinear
measure neuronal complexity loss provides additional
information for
a
spatial characterization of the primary
epileptogenic area in neocortical lesional epilepsy, even
during the interictal state. Nonlinear time series analysis
can thus contribute to an improvement of the presurgical
assessment. Future studies should include more complex
cases (e.g., nonlesional neocortical epilepsy or involve-
ment
of
mesial temporal structures in neocortical epilep-
sies) as well as investigations
of
the applicability of the
proposed method in acute electrocorticography
.
Acknowledgment:
We thank our neurosurgical colleagues
J.
Schramm, M.D.,
J.
Zentner, M.D., D. Van
Roost,
M.D., and
E.
Behrens, M.D., who implanted the intracranial electrodes and
performed the lesionectomies, as well as M. Kurthen,
M.D.
and
W. Burr, Ph.D., for helpful comments. This study was
sup-
ported by the Deutsche Forschungsgemeinschaft (grant no.
EL
122/3-2).
REFERENCES
I.
Salanova
V,
van Ness
PC,
Andermann
F.
Frontal, parietal, and
occipital epilepsies. In: Wyllie
E,
ed.
The
treafment
of
epilepsy:
principle&
and practice.
2nd edition. Baltimore: Williams
&
Wilkins,
1996:423-3
1.
Epilepspsia,
Vol.
41,
No.
7,
2000
L*
MAPS
IN
NEOCORTICAL EPILEPSIES
817
2.
3.
4.
5.
6.
7
8
9
10
11
12
13
14
15
16
Zentner
J,
Hufnagel A, Ostertun B, et al. Surgical treatment of
extratemporal epilepsy: clinical, radiological, and histopathologi-
cal findings in 60 patients.
Epilepsia
199637: 1072-80.
Comair YC, Choi HY, van Ness PC. Neocortical resections.
In:
Engel
1
Jr, Pedley TA,
eds.
Epilepsy: a comprehensive textbook.
Philadelphia: Lippincott-Raven, 1997:
18
19-28.
Fried
I,
Cascino GD. Lesionectomy. In: Engel J Jr, Pedley
TA,
eds.
Epilepsy:
a
comprehensive textbook.
Philadelphia: Lippincott-
Raven, 1997:1841-50.
Tran TA, Spencer
SS,
Spencer DD. Epilepsy: Medical and surgical
outcome.
In:
Swash M, ed.
Outcomes
in
neurological and
neuro-
surgical disorders.
New York: Cambridge University Press, 1998:
407-40.
Hajek M, Antonini A, Leenders KL, Wieser HG. Mesiobasal ver-
sus lateral temporal lobe epilepsy: metabolic differences in the
temporal lobe shown by interictal
IRF-FDG
positron emission to-
mography.
Neurology
1993;43:79-86.
Wieser HG. PET and SPECT in epilepsy.
Eur Neurol
1994;34
(suppl 1):58-62.
Duncan JS. Imaging and epilepsy.
Brain
1997;120:339-77.
Nakasato N, Levesque MF, Barth
DS,
Baumgartner
C,
Rogers RL,
Sutherling WW. Comparisons
of
MEG, EEG, and ECoG source
localization in neocortical partial epilepsy in humans.
Electr-oen-
cephalogr
Clin
Neurophy,siol
1994;91:171-8.
Luders
HO,
Awad
1.
Conceptual considerations.
In:
Liiders HO,
ed.
Epilep.sy surgery.
New York: Raven Press, 1991
:5
1-62,
Wieser
HC.
Epilepsy surgery.
Baillieres
Clin
Neurol
1996;5:849-
7c
ronal complexity loss.
Electroencephalogr
Clin
Neurophysiol
1995;95:108-17.
19. Casdagli MC, Iasemidis LD, Savit RS, Gilmore RL, Roper SN,
Sackellares JC. Non-linearity in invasive EEG recordings from
patients with temporal lobe epilepsy.
Electroencephnlugr
Clin
Neurophysiol
1997; 102:98-
105.
20. Weber
9,
Lehnertz
K,
Elger CE, Wieser HG. Neuronal complexity
loss in interictal EEG recorded with foramen ovale electrodes pre-
dicts side
of
primary epileptogenic area in temporal lobe epilepsy:
a replication study.
Epilepsia
1998;39:922-7.
21. Lehnertz
K,
Elger CE. Neuronal complexity
loss
in temporal lobe
epilepsy: effects of carbamazepine
on
the dynamics of the epilep-
togenic focus.
Electroencephalogr
Clin
Neurophysiol
1997;
103:
376-80.
22. Elger CE, Lehnertz
K.
Seizure prediction by noii-linear time series
analysis
of
brain electrical activity.
Eur
J
Neurosci
1998;
10:786-9.
23. Lehnertz
K,
Elger CE. Can epileptic seizures be predicted? Evi-
dence from nonlinear time series analyses of brain electrical ac-
tivity.
Phys Rev Lett
1998;80:5019-23.
24. Martinerie J, Adam C, Le Van Quyen M, et al. Epileptic seizures
can be anticipated by non-linear analysis,
Nut Med
1998;4: 1173-6.
25. Moser HR, Weber
B,
Wieser HG, Meier PF. Electroencephalo-
grams in epilepsy: analysis and seizure prediction within thc
framework
of
Lyapunov theory.
Physica
D
2000;
In
press.
26. Presurgical evaluation protocols.
In:
Engel J Jr, ed.
Surgical freat-
ment ofthe
epilepsies. 2nd
edition.
New
York
Raven
Press Ltd.,
I993:740-2.
I
J.
27. Lopes da Silva FH. EEG analysis: Theory and practice.
In:
Nie-
dermayer E, Lopes da Silva FH, eds.
Electroencephalography,
basic
principles,
clinical applications and related
fields.
Balti-
more, MD: Urban and Schwarzenberg, 19872371.
28. Grassberger P, Procaccia
I.
Measuring the strangeness of strange
attractors.
Physica
D
1983;9: 189-208.
29. Widman
G,
Lehnertz
K,
Jansen P, Meyer W, Burr
W,
Eker CE.
A
fast general purpose algorithm for the computation
of
auto- and
cross-correlation integrals from single channel data.
Physica
D
1998;121:65-74.
30.
Takens
F.
Detecting strange attractors in turbulence.
In:
Rand DA,
Young LS, eds.
Lecture notes
on
mathematics.
New York:
Springer,l981:366-81.
31.
Theiler J. Spurious dimension from correlation algorithms applied
to limited time series data.
Phys
Rev
A
1986;34:2427-32.
32.
Ruelle D. Deterministic chaos: the science and the fiction.
Proc
R
Tran
TA,
Spencer
SS,
Javidan M, Pacia
S,
Marks D, Spencer DD.
Significance of spikes recorded
on
intraoperative electrocorticog-
raphy in patients with brain tumor and epilepsy.
Epilepsia
1997;
38:1132-9.
Engel
Jr,
Van
Ness pC,
Rasmussen
TB,
ojemann
LM,
Outcome
with respect to epileptic seizures.
In:
Engel
J
Jr, ed.
Surgical treat-
?[
the
eI,ilepsies,
2nd edition,
N~~
York:
R~~~~
press Ltd,
1993509-22.
Kano
H,
Schreiber T, eds.
Nonlinear time
series
analysis.
Cam-
bridge: Cambridge University Press, 1997.
Iasemidis
LD,
sackellares JC, zaveri HP, wi1liams
WJ,
Phase
space topography and the Lyapunov exponent of electrocortico-
grams in partial seizures.
Brain
Topogr
1990;2: 187-201,
pijn jp,
van
N~~~~~~
J, ~~~~t
A,
L~~~~
da silva
FH,
Chaos
or
noise in EEG signals: dependence
on
state and brain site.
Electro-
encenhulowr
Clin
Neuroohvsial 199 1:79:37
1-81.
17. Pi,in.JP, felis DN, van
be;
Heyden MJ, DeGoede J, van Veelen
CW, Lopes da Silva FH. Nonlinear dynamics of epileptic seizures
on
basis of intracranial EEG recordings.
Brain
Topogr 1997;9:
249-70.
logr
Clin
Neurophysiol
1987;66:75-81.
18. Lehnertz
K,
Elger CE. Spatiotemporal dynamics of the primary
epileptogenic area in temporal lobe epilepsy characterized by neu-
Sac
Lond
1990;A427:241-8.
33.
Perrin
F,
Pernier J, Bertrand
0,
Giard MM, Echallier JF. Mapping
of scalp potentials by surface spline interpolation.
Electroencepha-
34. Morrell
F.
Secondary epileptogenesis in man.
Arch Neurol
1985;
42:
3
18-35.
Epilepsia,
Vol.
41, No.
7,
2000
