EEG spectral changes underlying BOLD responses contralateral to spikes in patients with focal epilepsy


Address correspondence to Eliane Kobayashi, M.D., Ph.D., Montreal Neurological Institute, McGill University, 3801 University Street, Montreal, Quebec, Canada, H3A 2B4. E-mail:


Purpose: Simultaneous electroencephalogram and functional magnetic resonance imaging (EEG–fMRI) in patients with focal epilepsy and unilateral spikes often shows positive blood oxygenation level–dependent (BOLD) responses (activations), not only ipsilateral but also contralateral to the spikes. We aimed to investigate whether minimal EEG changes could underlie these contralateral BOLD responses by using EEG spectral analysis.

Methods: We studied 19 patients with focal epilepsy and unilateral spikes. According to the pattern of BOLD activation, patients were divided into Group 1 (ipsi- and contralateral to the spikes) or Group 2 (only ipsilateral). EEG from outside the scanner was used to mark spikes similar to those recorded in the scanner. Epochs of 640 ms before and after the peak of the spikes were chosen as baseline and spike epochs. Spectral analysis was performed in referential montage (FCz reference), and differences between baselines and spikes were analyzed by paired t-test.

Results: Significant EEG changes in electrodes contralateral to the spikes were seen in 9 of 10 patients in Group 1 and in only 2 of 10 patients in Group 2 (one patient had two types of spikes that were analyzed separately). Spectral changes were seen in delta and/or theta bands in all patients except one (in Group 1) who had changes in all bands.

Discussion: Significant contralateral EEG changes occurred in 90% of contralateral BOLD activations and in only 20% of patients without contralateral BOLD responses. The reason why these changes predominate in lower frequencies rather than in higher frequencies is unclear. These spectral changes in areas corresponding to contralateral activations might reflect poorly synchronized but possibly intense neuronal activity.

Focal epilepsy is the most frequent type of epilepsy in adults (Hauser et al., 1993). Although focal epileptiform discharges on the electroencephalogram (EEG) are characteristic of focal epilepsy, many functional imaging studies used on a clinical basis, such as positron emission tomography (PET) (Hammers et al., 2002; Chassoux et al., 2004) and magnetic resonance spectroscopy (MRS) (Li et al., 2000; Mueller et al., 2004), frequently disclose widespread abnormalities.

Most imaging techniques do not have sufficient temporal resolution to assess changes directly related to the spikes, which occur within a window of milliseconds duration. Combined recording of EEG and functional magnetic resonance imaging (EEG–fMRI) is a unique method that allows us to image the whole brain at the time of epileptic spikes (Gotman et al., 2006; Gotman, 2008). It has been demonstrated that EEG–fMRI can identify positive and negative changes in the blood oxygenation level–dependent (BOLD) signal related to interictal spikes (Bagshaw et al., 2004; Kobayashi et al., 2006; Salek-Haddadi et al., 2006; Zijlmans et al., 2007).

Previous EEG–fMRI studies in patients with focal epilepsies demonstrated that BOLD activations are frequently seen in regions that are spatially concordant with the localization of the EEG spikes (Bagshaw et al., 2004; Kobayashi et al., 2006; Salek-Haddadi et al., 2006; Zijlmans et al., 2007). This is in agreement with the premise that increased neuronal activity in the focus leads to an increase in the BOLD signal. However, BOLD activations often involved additionally the regions contralateral and homologous to the spike topography (Kobayashi et al., 2006), where no EEG changes could be identified. Moreover, we have found in a few patients responses only contralateral and homologous to the spike topography, with no ipsilateral involvement (Kobayashi et al., 2006). This phenomenon has not yet been investigated, and the possibility of subtle EEG changes that occur simultaneously or by propagation cannot be ruled out.

Because fMRI assesses the epileptic network at the time of spike occurrence, further characterization of the neuronal correlates of BOLD response may help to interpret this phenomenon. Finding associated EEG abnormalities may help confirm that the BOLD responses are not artifactual. Moreover, these findings may relate to neuropsychological deficits or other clinical aspects of epilepsy in an individual patient.

Objectives and Rationale

Our objective was to investigate whether subtle EEG changes identified by spectral analysis were associated with BOLD responses contralateral and homologous to the EEG focus in patients with focal epilepsy. We aimed to compare patients with unilateral spikes and bilateral activations with patients who had only an ipsilateral response to the spikes.



We retrospectively investigated patients from the EEG–fMRI database at the Montreal Neurological Institute, who had been evaluated between November 2003 and February 2008.

Patients were selected based on the following criteria: history of focal seizures and unilateral spikes (including those with bilateral independent spikes, which were analyzed separately as different spike types), with no generalized discharges.

A total of 19 patients and 20 EEG–fMRI studies were included (one patient had two types of spikes that were analyzed separately). They were eight males and 11 females, whose mean age at EEG–fMRI studies was 33.2 years (range 11–55 years). Mean age at seizure onset was 15.8 years (range 9 months to 50 years), and mean duration of epilepsy was 16.5 years (range 5–47 years). Two patients experienced febrile seizures in childhood. Seizure frequency ranged from 1 per year to 30 per month. Based on electroclinical localization and neuroimaging findings, there were 11 cases of temporal lobe epilepsy, 4 cases of frontal lobe epilepsy (including one frontocentral), 2 cases of frontotemporal epilepsy, and 2 cases of temporoparietal epilepsy.

MRI investigation showed that six patients had unilateral hippocampal atrophy (HA), four had malformations of cortical development, two had cavernomas, one had a posttraumatic lesion, one had left frontal and parietal white matter abnormality (subcortical), and one had right amygdala atrophy. The remaining four showed no clear MRI abnormalities. Clinical details are shown in Table 1.

Table 1.   Clinical data of the patientsa
Patient numberGroupAge/age at onset (years)/sexHabitual Seizure typesInterictal EEGMRIElectroclinical Localization Surgery/outcome
  1. L, left; R, right; B, bilateral; T, temporal; F, frontal; P, parietal; C, central; O, occipital; I, insular.

  2. WMA, white matter abnormality; MCD, malformation of cortical development; HA, hippocampal atrophy; AA, amygdala atrophy; PHN, periventricular nodular heterotopia; CPS, complex partial seizure; GTCS, secondary generalized tonic–clonic seizure; SPS, simple partial seizure.

  3. aDiagnosis is based on neuroimaging findings and electroclinical localization according to the International League Against Epilepsy (ILAE) criteria.

 1131/27/FCPSL T and FLHAL TYes/seizure-free
 2124/13/MCPSLTNegativeL FT-
 312/16/FSPSL F and TL F, P WMAL FCYes/50% improvement
 4115/9/MCPSLFNegativeL FYes/50% improvement
 9143/26/FCPSLTR trigone and L atrium PNHLT-
101, 220/14/FSPS, CPSRT and R TPL TO PNHR TPYes/seizure-free
11234/0.75/MCPS, GTCSLTLHALTYes/seizure-free
12236/16/FSPSLTL F and T PNHL TP-
13255/50/FCPS, GTCSLTLHALTYes/seizure-free
14242/35/FCPSLTLT cavernomaLTYes/80% improvement
15251/4/MSPSL CPMCD in LFCLFYes/lost follow-up
17238/23/FCPS, GTCSRTBF lesionR FT-
18217/5/MCPS, GTCSRTMild LHARTNo improvement
19239/28/MCPSLTLT cavernomaLTYes/seizure-free

Based on the spatial distribution of BOLD activations, patients were divided into two groups (Fig. 1):

Figure 1.

Patients were grouped by activation patterns (in orange). Group 1, activations ipsilateral and contralateral to the spikes. Group 2, only ipsilateral activation.

  • • Group 1 (bilateral activation): activation was seen within or very close to the area anatomically related to the spike (ipsilateral) and involved at the same time the contralateral homologous region.
  • • Group 2 (ipsilateral activation): activation occurred only ipsilateral to the spikes.

Patients were selected regardless of the presence of additional areas of activation at a distance from these two target regions.

EEG–fMRI data analysis

EEG–fMRI acquisition and data processing have been described in detail in previously published papers (Bagshaw et al., 2004; Kobayashi et al., 2006). In summary, EEGs recorded inside the scanner were filtered offline, for visual identification of spikes, which were marked according to spatial distribution and morphology. Each type of spike from each subject constituted one EEG–fMRI study, generating t-maps of the BOLD response. In the present study, only activations (positive BOLD responses) have been evaluated.

EEG spectral analysis

Although it is possible to successfully remove the gradient artifact from the EEG recorded inside the MRI scanner to identify the spikes, residuals might be still present in the signal and might affect spectral analysis. Therefore, for the spectral analysis we decided to use EEGs from video-EEG monitoring (not recorded during fMRI scanning). Only one patient did not have such an investigation, and we used a prolonged EEG recording acquired immediately after the EEG–fMRI session. Telemetry EEG was recorded by 21 electrodes using the 10–20 system, in addition to zygomatic electrodes and inferior temporal electrodes placed according to the 10–10 system. The interval between telemetry EEGs and EEG–fMRI investigations ranged from 1 to 9 days. The EEG of the patient who had a prolonged recording immediately after the EEG–fMRI study included 44 electrodes placed on the scalp according to the 10–10 system.

Spikes with spatial distribution and morphology similar to those in the EEG–fMRI session were marked in the telemetry EEG. For each patient, a minimum of 20 spikes was marked.

An EEG segment of 640 ms before and after the peak of the marked spikes was used as baseline and spike epoch, respectively. The order of magnitude of the length (0.5–1 s) was selected because we hypothesized that this is a reasonable duration for which one could expect the effect of a spike, itself lasting around 100 ms, to last. The specific value of 640 ms comes from the convenience of using a number of samples that is a power of 2 for the calculation of the fast Fourier transform. Given the 200 Hz sampling rate, 128 samples correspond to 640 ms. Fluctuations in the level of alertness of the subject result in important fluctuations of EEG baseline. To compare the post-spike segment to a baseline segment that is most likely to be in the same state, we selected the baseline segment just before the spike. In some patients we had to select a baseline epoch that was farther from the spike due to the presence of repeated spikes or bursts of slow waves occurring with the spike under consideration. The referential montage (FCz) was used for spectral analysis, which included five frequency bands: delta (0.1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), and gamma (30–70 Hz). For each study, we selected for spectral analysis one electrode contralateral and homologous to the region of the ipsilateral BOLD response (Fig. 2). The differences between baselines and spikes for each study were analyzed by paired t-test (spike minus baseline) at a p ≤ 0.05. Analysis was performed using Matlab (Mathworks, Natick, MA, U.S.A.).

Figure 2.

Example of electrode selection for spectral analysis. (A) Activations in bilateral homologous regions, in a patient with left temporal spikes (Patient 9, Table 2). Activations are close to electrodes T3 and T4, and the T4 electrode was chosen for spectral analysis. (B) Ipsilateral activation only, in a patient with right temporal spikes (Patient 18, Table 3). Activation is adjacent to electrode F8, and the F7 electrode was, therefore, selected for spectral analysis. To note: maximum BOLD response in these two patients appears maximum in the suprasylvian rather than in the temporal regions. This same pattern has been demonstrated in a series of patients with temporal lobe epilepsy patients who have studied (Kobayashi et al., 2006; Epilepsia), and this has been confirmed in an interindividual analysis as the common pattern of activation found in patients with temporal lobe epilepsy (Kobayashi E, Grova C, Duveau F, Gotman J., unpubl. ms.).


One patient had two types of spikes, which corresponded, respectively, to Group 1 and Group 2. Therefore, a total of 20 spectral studies in 19 patients were performed: 10 in Group 1 and 10 in Group 2. The mean number of spikes in the spectral analysis per study was 30.4 (range 21–48) in Group 1 and 23.9 (range 20–32) in Group 2. We averaged all spikes from each patient: the average spike voltage in Group 1 was 74 μV and in Group 2 was 73 μV (t = 0.071; p = 0.94; d.f. = 18).

Group 1 (bilateral activation)

Nine of the 10 studies showed spectral changes in both delta and theta bands (n = 3), in delta band only (n = 3), in theta band only (n = 2), or in delta, theta, alpha, and beta bands (n = 1). See Table 2 for details.

Table 2.   Summary of spectral analysis results in Group 1
Patient numberEEG spikeNumber of spikesArea of activationElectrode analyzedBand(s) with changesp-value
  1. L, left; R, right; B, bilateral; T, temporal; F, frontal; P, parietal; C, central; O, occipital; I, insular.

 1LT48B I  T4Theta0.0013
 2L FT26B TFF8Delta0.0243
 3L FC21B T IT4Delta, theta≤0.0053
 4L F22B F F4Delta, theta, alpha, beta≤0.0085
 5RT38B T T3Delta0.0015
 6RF25B F F3Theta0.0200
 7LT38B I  T4Delta, theta≤0.0009
 8RT38B T F9Delta, theta≤0.0191
 9LT21B TFT4Delta0.0187
10R T27B P P3None 

The only patient (no. 10, in Table 2) in this group who did not show spectral changes had T4–T6 spikes but strong bilateral homologous BOLD activation in the parietal lobes. We, therefore, selected P3 as the electrode on which to perform spectral analysis in the contralateral region.

Group 2 (ipsilateral activation only)

Only 2 of the 10 studies (from patients 11 and 12 in Table 3) showed spectral changes, in delta band (n = 1) or in theta band (n = 1). See Table 3 for details.

Table 3.   Summary of spectral analysis results in Group 2
Patient numberEEG spikeNumber of spikesArea of activationElectrode analyzedBand(s) with changesp-value
  1. L, left; R, right; B, bilateral; T, temporal; F, frontal; P, parietal; C, central; O, occipital; I, insular.

11LT23L TT10Theta0.0131
12L TP20L T, IT4Delta0.0356
10R TP32R TOT5None 
13LT24L TT4None 
15LF26L CC4None 
16RT23R ITT9None 
17R FT22R TT5None 
18RT21R TFF7None 
19LT27L TF8None 

To exclude the possibility that, in these two patients who had spectral changes, there were activations just below the statistical threshold of the t maps, we looked back at the maps with a lower threshold. No activations could be seen in the contralateral region, even when using a lower threshold.

To evaluate whether this lack of spectral change in Group 2 was due to suboptimal selection of the contralateral electrode, we further investigated whether in these studies with no spectral changes, the four electrodes located around the originally chosen one showed any spectral change. No changes could be seen in this additional investigation.


In this study we aimed to assess the origin of the contralateral BOLD changes that has been observed in spikes that appear unilateral on EEG. We could demonstrate that this corresponds to a subtle EEG change and, therefore, confirm that genuine contralateral changes can occur in spikes that appear unilateral. Although BOLD responses at a distance from the focus seem to occur frequently (Bagshaw et al., 2004; Kobayashi et al., 2006; Salek-Haddadi et al., 2006; Zijlmans et al., 2007), contralateral homologous responses to unilateral spikes seem to be observed in a larger proportion than other brain regions (Kobayashi et al., 2006). The possibility of propagation of epileptic activity, effect at a distance from the spike, or simultaneous changes in neuronal activity could underlie this phenomenon. In our patient group, spikes were strictly unilateral and no contralateral spike could be seen by visual inspection of the EEG. We found that 90% of contralateral homologous activations were associated with minimal but significant EEG changes, mostly in delta or theta bands. Such abnormalities occurred in only 20% of studies with activation only ipsilateral to the spikes.

In summary, in the majority of our studies we found (1) contralateral homologous BOLD activation with concomitant spectral EEG changes; and (2) no contralateral spectral changes in most cases where there was no contralateral BOLD response. These two scenarios corroborate our working hypothesis. However, we also found in a much smaller proportion of studies: (3) contralateral homologous activation without concomitant spectral changes, seen in one study; and (4) spectral changes contralateral to the spikes, despite only ipsilateral activation, seen in two studies.

Observations (1) and (2) suggest that although we do not identify any clear EEG transients the electrodes contralateral to the spikes, subtle EEG changes do exist in underlying regions of BOLD activation. One explanation for these changes in delta and theta bands could be that some spikes are followed by a slow wave, which could have a wider distribution than the spike itself, including the contralateral region. However, slow waves following the averaged spikes for each patient were equally identified in both groups (seven studies in each group), thus not supporting such a hypothesis to explain spectral changes observed in Group 1.

One could consider whether the frequency of secondary generalized tonic–clonic seizures (GTCS) could influence the presence of contralateral spread and BOLD response in Group 1. However, there were more patients in Group 2 (n = 6) than in Group 1 (n = 4) with history of GTCS, and only one patient in Group 1 was still having frequent GTCS at the time of scanning (3–4 per month). Therefore, frequency of GTCS does not explain the difference between the groups.

The fact that we have not identified any spectral changes when further exploring the electrodes surrounding the contralateral region in Group 2, suggests that the presence or absence of such changes is not diffuse in the contralateral hemisphere. This supports the idea that if the spikes do have an effect at a distance and this can cause contralateral homologous activation, this finding is not related to a widespread nonspecific effect of the spikes.

The effect at a distance of the spikes could, therefore, occur as propagation of the epileptic activity, following the underlying structure of the epileptic network and involving the mirror region of the main activated area. Although the underlying path for this possible propagation could be transcallosal in at least some of these patients, we have no means of proving it. Propagation speed cannot be assessed by such a low temporal resolution method as the hemodynamic response measured by the BOLD signal. If propagation indeed occurs, it does not result in activity that is sufficiently synchronous to be seen in the scalp EEG, since no clear discharge can be seen in these contralateral electrodes.

A possible explanation for observation (3) is related to the fact that, in this patient, the spikes from which the BOLD maps originated and which were used to trigger the spectral analysis, had right temporal topography, whereas the bilateral activation and, therefore, the EEG electrode analyzed were in parietal regions. This could suggest that both ipsilateral and contralateral parietal responses correspond to an effect at a distance from the original spike. However, this left parietal activation had a high t value, and we cannot explain why such robust hemodynamic response does not appear associated with any underlying EEG change.

In situation (4), we have excluded the presence of contralateral homologous activation just below threshold in these areas showing EEG spectral changes, by looking at the BOLD maps with a lower threshold. One possible explanation is that in these two studies, the spectral changes that were seen were not sufficient to cause any detectable BOLD changes. This could be compared with patients who have a good number of spikes during an EEG–fMRI acquisition, but for whom no BOLD responses can be detected in the t maps. The possibility of a variation of the neurovascular coupling in these specific regions cannot be ruled out.

In this study, we have not looked at deactivations. This was due to the fact that the number of EEG–fMRI studies that had ipsilateral only or bilateral deactivations to unilateral spikes, was much lower and not sufficient to provide statistical power to the analysis. In addition, the significance of deactivations is much less understood, and these uncertainties could make any analysis much more speculative than activations, which are known to be related to increased neuronal activity. We do acknowledge, however, that the evaluation of these negative BOLD responses would also be interesting.

We have demonstrated that epileptic spikes that appear unilateral are often accompanied by a contralateral change in the form of a BOLD activation associated with an EEG change; other spikes that appear similarly unilateral are not accompanied by either BOLD change or EEG change. The EEG–fMRI method was thus able to identify the spikes that have a contralateral effect, and this change could be confirmed with EEG analysis. Spikes that have a contralateral effect may have a more deleterious effect on cognition or long-term epileptogenesis than spikes that remain totally unilateral.


Study supported by operating grant MOP-38079 from the Canadian Institutes of Health Research. JMY is supported by China scholarship Council. LT is supported by a postdoctoral fellowship from the Savoy Foundation for Epilepsy. EK is supported by the Milken Family Foundation through the Early Career Physician Scientist Award from the American Epilepsy Society.

We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Disclosure: None of the authors has any conflict of interest to disclose.