Long-term evaluation of synchronization between scalp EEG signals in partial epilepsy
ABSTRACT For the anticipation of epileptic seizures synchronization between signals from intracranial recorded EEG has been studied. Here, we present our first findings for scalp EEG. For 3 epilepsy patients 85 hours of scalp EEG were analyzed. After determining the instantaneous phase with the Hilbert transform, the level of synchrony was calculated for all possible electrode pairs within 4 defined groups. This was done for 14 frequency bands of 2Hz between 1 and 50Hz. As well during sleep as in the awake state we found a particular behavior of the synchrony levels in the pathological hemisphere. Furthermore, we found a similar sleep-wake cycle for the two temporal lobe epilepsy patients, not seen in the case of the patient with extra temporal lobe epilepsy. These results show an altered brain dynamics for epilepsy patients, which can give information on the localization of the epilepsy.
LONG-TERM EVALUTION OF SYCHRONIZATION BETWEEN SCALP EEG
SIGNALS IN PARTIAL EPILEPSY
Elly Gysels1,2, Michel Le Van Quyen3, Jacques Martinerie3, Paul Boon2, Kristl Vonck2,
Ignace Lemahieu1 , Rik Van de Walle1
1 Department of Electronics and Information Systems, Ghent University, Sint-Pietersnieuwstraat 41,
B-9000 Ghent, Belgium
2 Reference Center for Refractory Epilepsy, Department of Neurology, Ghent University Hospital,
De Pintelaan 185, B-9000 Ghent, Belgium
3 Neurodynamics Group, LENA-CNRS UPR640, Hôpital de la Salpêtrière, 47 Boulevard de l’Hôpital,
75634 Paris Cedex 13, France
For the anticipation of epileptic seizures synchronization
between signals from intracranial recorded EEG has been
studied. Here, we present our first findings for scalp
For 3 epilepsy patients 85 hours of scalp EEG were
analyzed. After determining the instantaneous phase with
the Hilbert transform, the level of synchrony was
calculated for all possible electrode pairs within 4 defined
groups. This was done for 14 frequency bands of 2Hz
between 1 and 50Hz. As well during sleep as in the
awake state we found a particular behavior of the
synchrony levels in the pathological hemisphere.
Furthermore, we found a similar sleep-wake cycle for the
two temporal lobe epilepsy patients, not seen in the case
of the patient with extra temporal lobe epilepsy. These
results show an altered brain dynamics for epilepsy
patients, which can give information on the localization of
The quality of life of epilepsy patients could be improved
significantly if their seizures can be anticipated. Several
methods, linear and non-linear [1, 2], have been
introduced for the study of anticipation of epileptic
seizures. Recent publications [2, 3, 4, 5, 6, 7] have drawn
the attention to synchronization phenomena between
different brain regions. Most of these studies were
performed on intracranial recorded EEG.
This study investigates the phase synchronization
between signals recorded at different electrodes on the
scalp. Some authors  suggest that the epileptic seizure
may result from a process of changes that starts long
before. Therefore we consider very long time segments:
about 85 hours per patient.
The EEG signals are recorded from electrodes placed on
the scalp according to the standard 10-20 International
electrode placement system. Subsequently, they are
sampled at 200Hz and filtered in 14 different frequency
bands of 2Hz between 1 and 50Hz. The data are analyzed
using a moving window technique. The time series are
divided in non-overlapping
corresponding to 3000 samples at the given sampling rate.
To investigate the phase synchronization of two
EEG signals, we firstly determine the instantaneous phase
of each signal following the analytic signal approach:
windows of 15s,
where s~(t) is the Hilbert transform of the original signal
‘p.v.’ denotes the Cauchy principal value.
Subsequently, we calculate the phase difference
between two channels for each time sample. We use the
‘Phase Locking Value’ [8, 9] to characterize the stability
of the phase-differences across the trials:
where f is the frequency, t the time, Ntrial the total number
of samples in the considered time window and x and y are
two EEG-channels for which the phase locking is studied.
The PLV is a normalized index that expresses the
constancy of the phase difference over the considered
time window. Two channels are synchronized during
some time period if the PLV for this time window is close
Optimization of the algorithm allows computing in
real time. The PLV of 50 electrode pairs is obtained in
1.14s for a 15s window with a standard PC running
The scalp EEG is recorded from 27 electrodes. This
results in 351 possible electrode pairs. Calculating
synchronies for all pairs
representation and does not give a clear view on the brain
dynamics. Therefore we defined 4 groups of electrodes
(see figure 1), each of which is studied in detail. Per
group one reference is chosen and subtracted from the
other electrodes in the group. The PLV is calculated as a
function of time for the 10 pairs in each group. Finally,
the PLV-curves are compared with each other.
causes problems for
Figure 1. Position of the 27 electrodes on the scalp and the
definition of the groups (dashed line) and references (bold line)
considered to study phase synchrony.
3. EPILEPSY PATIENTS
All patients were included in an extensive presurgical
evaluation protocol (long term video-EEG monitoring,
FDG-PET, optimum MRI, neuropsychological evaluation)
at Ghent University Hospital . Based on these
examinations we found that the first patient (female) has
left medial temporal lobe epilepsy; the second one
(female) has right temporal lobe epilepsy. The third
patient’s (male) epileptic focus is suspected to be located
outside the temporal lobe. The hemisphere of seizure
onset is uncertain in his case.
The time course of the synchrony levels for different
groups and in different frequency bands shows a clear
distinction between wake and sleep. Moreover, it is
asymmetric when we compare the groups on the left and
right hand side of the scalp. We observed some particular
phenomena in different frequency ranges. They were
reproducible over the 3 days and nights the patient was
In the case of the patient with right temporal lobe
epilepsy, the electrode pairs F8-FT10, F8-T10, FT10-T10
and P8-TP10 showed an increased level of synchrony
during sleep, as compared to the other electrode pairs in
this group. Equally, an increase in synchrony was
observed for the corresponding pairs on the left hand side
of the scalp (F7-FT9, F7-T9, FT9-T9, P7-TP9) for the
patient with left medial temporal lobe epilepsy and for the
patient with extratemporal lobe epilepsy. The increased
synchrony of these 4 electrode pairs in the pathological
hemisphere during sleep was observed in the α-frequency
At each time instance the Phase Locking Values of
the 10 curves were averaged to obtain a mean value for
the group. In the awake state, the resulting curve in the
β-frequency range showed a greater mean for the
pathological hemisphere than for the healthy hemisphere
(see figure 2). Also, the pair which is least synchronized
on the pathological side is more synchronized than the
least synchronized pair on the healthy side.
Figure 2. Average level of local synchrony for a 10-hours
period in the frequency band 21-23 Hz during wake (left) and
sleep (right) for the patient with left medial temporal lobe
epilepsy. White corresponds to higher values of PLV.
On the front of the scalp we observed a particular
sleep-wake cycle (see figure 3) in the two temporal lobe
epilepsy patients. While the synchrony levels were spread
when the patients were awake, they converged during
sleep. For some pairs the PLV decreased with the
transition from wake to sleep, while other pairs showed an
increased PLV. This pattern was not detected for the
patient with extra temporal lobe epilepsy.
Figure 3. Time course (about 17 hours) of the 10 PLV curves in
the frontal group in the 41-43Hz-frequency band for the patient
with left medial temporal lobe, right temporal lobe and extra
temporal lobe epilepsy respectively. For the patients with
temporal lobe epilepsy wake (dashed line) and sleep (full line)
can be clearly distinguished: when the patient is asleep the levels
of phase synchrony are closer to each other.
To our knowledge the changing phase synchronization
between EEG signals throughout the sleep-wake cycle of
epilepsy patients has not yet been reported in literature.
Most studies on phase synchronization between
signals recorded from different brain areas of epilepsy
patients have been performed on intracranial recorded
EEG. Usually only short time segments (a couple of
hours) have been considered.
While intracranial recorded EEG is usually of better
quality, containing less noise and artifacts, scalp EEG
recordings have the advantage of being non-invasive and
better covering the brain region. The fact that the whole
monitoring period is studied and that the evolution of
phase synchronization between electrode pairs are
compared with each other, opens new perspectives on the
brain dynamics of epilepsy patients.
For the frequency bands in the higher frequency
range spatial averaging within the groups defined on the
left and right hand side of the scalp revealed a greater
synchronization for interictal recorded EEG during wake
on the side of the epileptogenic focus. The higher the
frequency, the more pronounced this difference is. This
result is consistent with the findings of Mormann  who
found a higher level of synchronization in the pathological
The capability of determining the side of the
epileptogenic focus without necessity of recording actual
seizure activity has been reported in [3, 11] for
intracranial recorded EEG and in  for REM sleep
recordings. Such an evaluation of scalp EEG would be of
even higher diagnostical value.
Further investigation is needed to identify different
processes that give rise to epileptic seizures. This may
allow short or long term anticipation of seizures.
It may be suggested that the hemisphere showing a higher
level of synchrony for the awake state, as well as pairs of
fronto temporal and posterotemporal derivations showing
an increased synchrony during sleep, indicate the
pathological hemisphere. A particular sleep-wake pattern
of converging and diverging levels of synchrony for the
frontal pairs may suggest epilepsy from temporal origin.
These results show an altered brain dynamics for
patients with different types of epilepsy. Study of this
dynamics by examining phase synchronies as a function
of time for different regions on the scalp may give
information on the pathological hemisphere and the type
of epilepsy. Further studies are needed to eventually
confirm these preliminary results.
 M. Le Van Quyen, J. Martinerie, V. Navarro, M. Baulac, and
F.J. Varela, "Characterizing Neurodynamic Changes Before
Seizures," Journal of Clinical Neurophysiology, Lippincott
Williams & Wilkins, Inc. Philadelphia, 18(3): pp. 191-208,
 K. Lehnertz, "Non-linear Time Series Analysis of
Intracranial EEG Recordings in Patients with Epilepsy – an
Overview," International Journal of Psychophysiology, Elsevier,
34: pp. 45-52, 1999.
 F. Mormann, K. Lehnertz, P. David, and C.E. Elger, "Mean
Phase Coherence as a Measure for Phase Synchronization and its
Application to the EEG of Epilepsy Patients," Physica D,
Elsevier, 144: pp. 358-369, 2000.
 M. Palus, V. Komarek, T. Prochazka, Z. Hrncir, and K.
Sterbova, “Synchronization and Information Flow in EEGs of
Epileptic Patients,” IEEE engineering in medicine and biology,
pp. 65-71, 2001.
 C.E. Elger, and K. Lehnertz, “Can Chaos Analyses Be a Tool
to Predict Epileptic Seizures,” Epilepsia, Blackwell Science,
Inc., Malden, 42: suppl. 7: pp. 3, 2001.
 F.J. Varela, “The Preictal State: Dynamic Neural Changes
Preceding Seizure Onset,” Epilepsia, Blackwell Science, Inc.,
Malden, 42: suppl. 7: pp. 3, 2001.
 F.H. Lopes das Silva, P. Suffczynski, J. Parra, W. Blanes, D.
Velis, and S. Kalitzin, “Epilepsies as Dynamic Diseases of Brain
Seizures,” Epilepsia, Blackwell Science, Inc., Malden, 42: suppl.
7: pp. 3, 2001.
 J.P. Lachaux, E. Rodriguez, J. Martinerie, and F.J. Varela,
“Measuring Phase Synchrony in Brain Signals,” Human Brain
Mapping, 8: pp. 194-208, 1999.
 M. LeVan Quyen, J. Foucher, J.P. Lachaux, E. Rodriguez, A.
Lutz, J. Martinerie, and F.J. Varela, “Comparison of Hilbert
Transform and Wavelet Methods for the Analysis of Neuronal
Synchrony,” Journal of Neuroscience Methods, 111: pp. 83-98,
 P. Boon, T. Vandekerckhove, E. Achten, E. Thiery, L.
Goossens, K. Vonck, M. D’Havé, G. Van Hoey, B. Vanrumste,
B. Legros, L. Defreyne, and J. De Reuck, “Epilepsy Surgery in
Belgium, the Experience in Gent,” Acta Neurologica Belgica,
99: pp. 256-265, 1999.
 K. Lehnertz, and C.E. Elger, “Spatio-temporal Dynamics of
the Primary Epileptogenic Area in Temporal Lobe Epilepsy
Characterized by Neuronal Complexity Loss,” Electroenceph.
clin. Neurophysiol., Elsevier, 95: pp. 108-117, 1995.
 B.A. Malow, and M.S. Aldrich, “Localizing Value of Rapid
Eye Movement Sleep in Temporal Lobe Epilepsy,” Sleep
Medicine, Elsevier, 1: pp. 57-60, 2000.
Basic Models and the