Electrical source imaging for presurgical focus localization in epilepsy patients with normal MRI


Address correspondence to Dr. Verena Brodbeck, Functional Brain Mapping Laboratory, 4 Rue Gabrielle-Perret-Gentil, CH-1211 Geneva, Switzerland. E-mail: verena.brodbeck@unige.ch


Purpose:  Patients with magnetic resonance (MR)–negative focal epilepsy (MRN-E) have less favorable surgical outcomes (between 40% and 70%) compared to those in whom an MRI lesion guides the site of surgical intervention (60–90%). Patients with extratemporal MRN-E have the worst outcome (around 50% chance of seizure freedom). We studied whether electroencephalography (EEG) source imaging (ESI) of interictal epileptic activity can contribute to the identification of the epileptic focus in patients with normal MRI.

Methods:  We carried out ESI in 10 operated patients with nonlesional MRI and a postsurgical follow-up of at least 1 year. Five of the 10 patients had extratemporal lobe epilepsy. Evaluation comprised surface and intracranial EEG monitoring of ictal and interictal events, structural MRI, [18F]fluorodeoxyglucose positron emission tomography (FDG-PET), ictal and interictal perfusion single photon emission computed tomography (SPECT) scans. Eight of the 10 patients also underwent intracranial monitoring.

Results:  ESI correctly localized the epileptic focus within the resection margins in 8 of 10 patients, 9 of whom experienced favorable postsurgical outcomes.

Discussion:  The results highlight the diagnostic value of ESI and encourage broadening its application to patients with MRN-E. If the surface EEG contains fairly localized spikes, ESI contributes to the presurgical decision process.

Patients with intractable focal epilepsy in whom magnetic resonance imaging (MRI) does not show evidence of an underlying lesion are thought to be less favorable candidates for surgery (Berkovic et al., 1995; Tonini et al., 2004), and are, therefore, less frequently referred for surgical treatment than patients with discrete MRI lesions (Siegel et al., 2001). The rate of patients with postoperative seizure freedom (complete or almost, corresponding to Engel class I or II) in MR-negative epilepsy (MRN-E) ranges between 40% and 70%, whereas the rate reaches between 60% and 90% in patients with MRI lesions. Patients with MR-negative extratemporal lobe epilepsy (MRN-ETLE) have the least promising outcome (about 50% seizure freedom).

Lesions are considered a reliable guide for the definition of the epileptogenic zone. In patients with MRN-E this guide is absent and the success of surgery has to rely on other methods that are able to identify the epileptic area. Several studies have examined the predictors of postoperative favorable outcome in MRN-E (Holmes et al., 2000; Cukiert et al., 2001; Siegel et al., 2001; Lee et al., 2003; Chapman et al., 2005; Alarcon et al., 2006; Madhavan & Kuzniecky, 2007; Jayakar et al., 2008). The value of careful analysis of ictal and interictal EEG, together with the semiology, is underscored in several studies, with the concordance of seizure semiology, ictal, and interictal EEG (scalp or intracranial) being related to a better outcome than discordance (Blume et al., 2004). Frequent and prolonged focal EEG abnormalities are known to be good prognostic factors. Dysplasias are not always visible on the MRI and are often related to rhythmic or continuous discharges on the scalp or intracranial EEG. Because resected tissue of patients with MRN-E often contains dysplastic lesions (McGonigal et al., 2007), focus localization using the EEG is a promising complement to other brain imaging techniques, especially in cases in which a brain lesion might be suspected even if the MRI result is normal.

Analysis of interictal EEG seems to be particularly important for the surgical success of patients with MRN-E (Blume et al., 2004), as shown for patients with MRN-TLE. This may be due to the fact that temporal foci are easily localized by visual analysis, even with standard electrode numbers (i.e., 19–32), compared to MRN-ETLE. Given the difficulty for ETLE focus localization, complementary tools such as electric source imaging (ESI) may improve the prognosis of MRN-ETLE patients.

ESI is based on EEG data and allows identifying the underlying source of a given potential. Its clinical utility has been demonstrated mainly in lesional epilepsy, where it localized correctly in 74–90% of all cases (Michel et al., 2004a; Sperli et al., 2006; Ding et al., 2007a; Leijten & Huiskamp, 2008; Brodbeck et al., 2009).

Here we test the ability of ESI to identify the epileptic focus in patients with normal MRI in whom surgery had been performed, in order to determine its usefulness also in this difficult patient group.



We searched our database for patients who had been referred to our department for surgical workup since the year 2000 for the following inclusion criteria:

  • 1. Intractable unifocal epilepsy

  • 2. No structural defect detected by recently performed high-resolution MRI

  • 3. Epileptic spikes detectable in surface EEG

  • 4. Underwent surgery and a follow-up of at least 1 year

A total of 10 patients with MRN-E met the inclusion criteria.

The clinical characteristics of these 10 patients are summarized in Table 1. The age at examination ranged from 2.8 to 57.1 years (mean 23.7, median 20.5), with age at epilepsy onset between 0.3 and 18 years (mean 8.7, median 8.0). Seven patients were female. Based upon prior interictal EEG and seizure histories, the location of the epileptic focus was described as temporal in five, frontal in two, and temporooccipital, parietooccipital, and frontotemporal one patient each.

Table 1.   Patient characteristics
Patient no.GenderAge at evaluationAge at onsetFocus pre-workupIntracranial electrodes + position (total n of electrodes)Seizures during evaluationType of surgeryHistopathologyFollow-upElectrodes ESINo. of Spikes
  1. For all patients, the age at presurgical evaluation, the age at epilepsy onset (both in years), the presumed focus localization before evaluation, the placement and number of electrodes in the invasive phase of the work-up (not performed in Patients 2 and 6), the type of resection performed, the histopathologic finding, and the duration of the follow-up period are listed. The two last columns contain the numbers of electrodes and the numbers of epileptic spikes in the interictal surface EEG that were subjected to the ESI (EEG source imaging) analysis. Five patients underwent high-resolution recordings with 128 or 256 electrodes. F, female; M, male; F, frontal; C, central; T, temporal; P, parietal; O, occipital; para, parasagital; L, left; R, right; Hemi, hemisphere; ant, anterior; post, posterior; orb, orbital; bas, basal; sup, superior; lat, lateral; Amyg, amygdala; Hipp, hippocampus; interhemisph, interhemisphereal; inf, inferior; insul, insular.

 1F8.93FC LGrid: Hemi L F, T, P
Strips: interhemispheric ant + post (76)
16Parietal parasagittal cortectomy LDiffuse subpial and focal intracortical gliosis5.03113
 2M6.55T L40Anterior temporal lobectomy sparing temporomesial structures LCortical dysplasia6.32128
 3M22.818FT LDepth: F (orb, ant, cing, SMA) L, T (Amyg/Hipp) L (7)9Anterior temporal lobectomyDiffuse gliosis8.712830
 4F18.210F LGrid: Hemi L F, T, interhemisph
Strips: F orb, T bas, P sup (124)
11Prefrontal lobectomy LDiffuse gliosis5.312868
 5M17.913T LStrips: F lat L
Depth: F orb L, Amyg/Hipp R + L (58)
4Anterior temporal lobectomy sparing temporomesial structures LDiffuse gliosis1.025630
 6F57.16T L18Anterior temporal lobectomy LMicroheterotopies3.72174
 7F2.80.3PO LGrid: TPO L
Strips: interhemsph P, PO L, T lat L (82)
26Parietooccipital cortectomy LCortical dysplasia2.51961
 8F39.516TO LGrid: FP,T L
Strips: T L (bas), P (sup, inf) L (100)
7Temporal lobectomy and basal parietal cortectomy LDiffuse gliosis7.812814
 9F27.614T RGrid: interhemisph
Depth: F orb R, Amyg/Hipp R + L (64)
7Anterior temporal lobectomy RCortical dysplasia1.425649
10F35.51.3Para FRGrids: F(parasg, lat) R; depth: Hipp R + insul R
Strips: interhemisph ant + post, T lat R (90)
40Polar frontal lobectomy RCortical dysplasia, heterotopies + gliosis1.53131

The presurgical workup comprised long-term video-EEG monitoring, high-resolution MRI, [18F]fluorodeoxyglucose positron emission tomography (FDG-PET), ictal and interictal single photon emission computed tomography (SPECT), and the subtraction of the two SPECT modalities coregistered to MRI (Subtraction Ictal SPECT Co-registered to MRI, SISCOM). Eight patients underwent a second evaluation phase with invasive recordings.

EEG surface recordings

Long-term video-EEG recording was performed in all patients over the course of several days, using standard EEG with 19, 21, or 31 electrodes mounted according to the international 10–20 montage (Coherence; Deltamed SA, Paris, France). Impedances were kept below 10 kOhm, and the EEG was sampled at 256 Hz with a 0.1–120 Hz bandpass filter and recorded using a vertex reference. The long-term EEG recordings captured between 4 and 40 seizures across all patients (mean 17.8, median 13.5).

In five patients, spikes from these standard EEG recordings during wake were used for the ESI analysis. In the other five patients, a separate EEG recording session with high-resolution EEG was performed using 128- or 256-channel nets with interconnected electrodes (Electrical Geodesic, Inc., Eugene, OR, U.S.A.). For these recordings, impedances were below 20 kOhm, the sampling rate was set to 500 Hz, the bandpass filter to 0.1–100 Hz, and the EEG was recorded using a vertex reference [see (Michel et al., 2004a) for details of the protocol]. The 30-min high-resolution recordings were performed at rest and alert states in a calm surrounding. The number of electrodes that was available for ESI is given in Table 1 for each patient.

EEG source imaging

Spike selection and averaging

The offline analysis was performed by one of the authors (VB) experienced in reading clinical EEG, who was blind to the clinical data at the time of EEG analysis. First, isolated interictal spikes, that is, without any other discharges within ±500 ms of the spike, were selected and marked. All patients had one dominant spike type with invariable morphology and maximal amplitude at the same electrode. These dominant spikes were visually identified and marked at the exact time point of maximal negativity on the same electrode trace. Spikes with the most dominant and similar voltage map distribution were chosen. Subsequently, the spatial correlation between the averaged map and the maps of each single spike at the time point of the Global Field Power peak were calculated, and spikes with <80% correlation were discarded (Michel et al., 2004a). In the remaining single spike epochs, electrodes containing artifacts were interpolated using a three-dimensional spline interpolation algorithm (Perrin et al., 1989) to keep the number of electrodes the same for all epochs. Spikes were then aligned to the global field power peak and averaged over epochs of ±500 ms around this peak.

For the localization of the electrical activity in the brain we used the local autoregressive average (LAURA) algorithm, a distributed linear inverse solution (Grave de Peralta Menendez et al., 2001, 2004). LAURA relies upon the weighted minimum norm algorithm with additional biophysical constraints (see Michel et al., 2004b for review). For the calculation of LAURA inverse solution, the solution space is restricted to the gray matter of the individual patient’s brain using a semiautomatic segmentation tool based upon the realistic head model SMAC method (Spinelli et al., 2000; Phillips et al., 2005). This method first transforms the individual MRI to the best fitting sphere using homogeneous transformation operators. It then determines a regular grid of approximately 4,000 solution points in the gray matter of this spherical MRI and computes the lead field matrix using the known analytical solution for a spherical head model with three shells of different conductivities as defined by Ary et al. (1981). Standard electrode position coordinates were used for all patients, which were derived from the mean of positions measured in 20 healthy subjects with a three-dimensional digitizer (3D) that were projected to a best fitting sphere. The coregistration between the spherical MRI and the spherical electrode positions was achieved by marking anatomic landmarks (nasion, inion, left and right preauricular points, and Cz) on the individual 3D MRI and matching them to the corresponding electrode positions.

Defining the accuracy of ESI focus localization

The average maps of each patient’s spikes were analyzed in source space at the time point of 50% rising phase (50% global field power peak). It has been demonstrated that the ESI of the rising phase of the average spike most reliably represents the actual source of the epileptic activity, whereas the spike peak involves areas of propagation (Lantz et al., 2003b; Ray et al., 2007). In all patients who were seizure free after surgery (N = 9), the focus localization was assumed to be within the epileptogenic zone, when the ESI maximum fell within the resected area. Even though the extension of resected area is most likely to overestimate the epileptic zone in most patients, seizure freedom following the resection can be taken as proof of correct localization on a sublobar level.

For 7 of the 10 patients, a postoperative MRI was available with a quality sufficient to superimpose the MRI with the ESI-identified focus. For the three patients in whom no postoperative MRI was available (2, 4, and 9), the ESI result is shown in the preoperative MRI with the resection zones approximated by the dashed line based on the operation report (see Fig. 1).

Figure 1.

Electroencephalography (EEG) source imaging (ESI) results. Localization of the epileptic activity superimposed on each patient’s individual magnetic resonance imaging (MRI) (green volumes). For patients 2, 4, and 9 no postoperative MRI was available; the resection margins are indicated approximately by the red dashed lines. For the other seven patients, the postoperative MRI is displayed; the resection areas are filled in black. (A) Shows the eight patients in whom ESI suggested the epileptic focus within the resected zone, and (B) the two patients in whom the ESI source was outside the resection zone. Note that patient 1 was the only patient in our series who had no improvement in seizure rate after surgery. Note also that the indicated ESI result (green volumes) corresponds to the maximum of activation and its distribution. The size of the ESI volume is threshold dependent and, therefore, does not provide information on the strength of the source.


All patients had baseline MRI scans as part of the presurgical evaluation, acquired with a 1.5 Tesla (n = 6, patients 1–4, 6, and 8) or 3 Tesla (n = 4, patients 5, 7, 9, and 10) Eclipse scanner (Picker Inc. Cleveland, OH, U.S.A.). The MRI was performed according to a standardized seizure protocol [coronal T2-weighted fast spin-echo (TR 3,092, TE 11/100, voxel size 0.9 × 0.9 × 9.6 mm), coronal and axial fluid-attenuated inversion recovery (FLAIR; TR 11,000, TE 140, TI 2,800, voxel size 0.45 × 0.45 × 6 mm), sagittal 3D gradient echo T1 (TR 12, TE 4, voxel size 0.98 × 0.98 mm2, thickness 1 mm), and diffusion sequences].

All MRI examinations were reviewed by two neuroradiologists, before and after the presurgical work-up, with special attention to the most likely region of seizure onset.

Nuclear neuroimaging

Six patients underwent conventional FDG-PET. The emission study started 30 min after intravenous injection of a mean activity of 132 ± 127 MBq (Megabequerel) of FDG. The data sets were acquired in the 3D mode on an ECAT ART PET scanner (CTI PET Systems, Knoxville, TN, U.S.A.). Following the installation of a state-of-the-art PET/CT (computed tomography) scanner (Biograph 64; Siemens medical solutions, Erlangen, Germany), another three patients underwent FDG-PET/CT with a similar protocol. Reconstruction of the raw data was performed using a standardized filtered backprojection algorithm (3DRP). One patient (Patient 1) did not have a PET scan.

All patients, with the exception of Patient 5, underwent ictal and interictal SPECT. For ictal SPECT the injection of the tracer during the seizure was verified by review of the video-EEG recording, and was not considered ictal if the injection was carried out after the EEG seizure had stopped. Experienced specialists in nuclear medicine interpreted both the PET and the ictal and interictal SPECT investigations [for details on the nuclear neuroimaging methods see (Kurian et al., 2007; Sperli et al., 2006)].

Subtraction of ictal versus interictal SPECT

For SISCOM analysis (O’Brien et al., 1998) the nuclear medicine images of interictal and ictal SPECT were resampled in an isotropic set and then coregistered with an anatomic framework using AIR 5.2 software (Woods et al., 1998). Once the interictal and ictal SPECT images were coregistered with MRI, the two sets of images were normalized prior to subtraction.

Invasive EEG recordings

Intracranial electrodes were implanted in eight patients. Intracranial recordings were not considered in two patients because ictal and interictal EEG, PET, and ictal SPECT/SISCOM were concordant, that is, in favor of a left temporal origin. In the remaining patients the electrode locations were chosen according to the individually suspected regions based upon the presurgical workups (see Table 1). The number of intracranial electrodes ranged from 36–124 (mean 79, median 80). Three patients received a combination of subdural and depth electrodes, one patient had only depth electrodes, and four patients had only subdural electrodes (for details see Table 1).


Fig. 1 shows all patients’ MRI with the superimposed ESI source volume in green. Note that the indicated ESI result corresponds to the maximum of activation and its distribution. However, to make the localization result visible and to avoid the impression of a point-like source, we set the amplitude threshold in such a way that a certain area around the maximum is displayed. Although this threshold setting is fully arbitrary and might discard additional secondary sources, it limits the data presentation to the area of maximal activity, which was the only criterion for the analysis. It is important to note that the extent of the area provides no information on the strength of the source.

Thirteen to 74 spikes were selected in each patient’s EEG (average 40, median 31), and the mean signal-to-noise ratio was 4.06 (2.73/7.58 min/max).

The identified maximum source of the epileptic activity fell within the epileptogenic zone in 8 of the 10 patients (Fig. 1A). The two patients in whom the ESI localized outside the resected zone are patients 1 and 7 (Fig. 1B). In patient 1 the ESI localized a source adjacent to the resected area. This patient was the only one of the 10 operated patients who experienced an unfavorable postsurgical outcome. His seizures remained unchanged in type and number after surgery. In this patient, we used left parasagittal sharp waves for localization, as those were the only focal abnormalities in the EEG. In a second evaluation 4 years later, left frontotemporal spikes were found. Without further information from follow-up tests and/or surgery at the time of publication, it is unclear whether the seizures in this patient originate from the ESI-identified focus near to the resected area or from the later-identified frontotemporal focus.

The second patient (Patient 7) in whom the estimated source was not detected within the resected area, was seizure-free after surgery. Intracranial recordings identified two foci, that is, an interhemispheric parietooccipital and an inferior parietal foci, whereas ESI identified only the focus in the inferior parietal lobe. Two explanations are possible. The missed interhemispheric focus could be due to the low electrode number at disposition in this very young child (19 electrodes (Lantz et al., 2003a). Another explanation lies in the interhemispheric localization. Such deep, low-amplitude discharges frequently cannot be detected at the scalp, even with a high number of electrodes. Interestingly, the interictal discharges from the inferior parietal region persisted in the postoperative (scalp) EEG, indicating that this was not an erroneous “ghost source.” Fortunately, the latter, recorded from several control EEGs, did not result in postoperative seizure recurrence (2 years follow-up). Overall, ESI correctly localized the focus within the resected area in 8 of the 10 patients. One of the two patients in whom ESI localized outside the resected area experienced no change in seizure frequency following surgery.

Comparison of ESI with other imaging modalities

The results of all examinations are summarized in Table 2. The most reliable focus localization was derived from ictal intracranial EEG. It indicated the correct location in all patients with positive postsurgical outcomes who underwent invasive recordings (= eight patients, except patients 2 and 6). In patient 1 (the only patient with Engel class IV outcome) the interictal intracranial EEG also indicated the focus to be within the subsequently resected area.

Table 2.   Results of noninvasive and invasive examinations
Patient no.Engel classsEEG intersEEG ictESIiEEG ictiEEG interPETSPECT interSPECT ictSISCOM
  1. Engel classification: I = free of disabling seizures, II = rare disabling seizures, IV = no worthwhile improvement; Signs: ✓ = correctly localizing, Ø = incorrectly localizing, loc+ = correctly localizing but also showing other areas, N = normal result, – = not performed.

  2. The percentages given in the lowest row are calculated as the number of cases in which the examination had correctly localized the focus within the resection zone, divided by the number of successfully operated patients in whom the examination had been performed (n = 9 for ESI, n = 6 for iEEG, n = 9 for PET, and n = 8 for SPECT and SISCOM).

  3. EEG, electroencephalography; ESI, EEG source imaging; sEEG, surface EEG; iEEG, intracranial EEG; inter, interictal; ict, ictal; PET, positron emission tomography; C, central; F, frontal; O, occipital; P, parietal; T, temporal; parasag, parasagittal; ant, anterior; R, right; L, left.

 1IVCP parasagCP LØØloc+Ø
 2IT ant LT ant LN
 3IFT + FC LFT Lloc+loc+loc+Ø
 6IT LT Lloc+NØ
 7IPO LPO LØloc+loc+loc+
 8IIT P + Tant LTP LØloc+loc+
 9IT RT ant R

Ictal intracranial EEG correctly localized the focus in seven of eight patients with invasive recordings and positive outcome. In patient 7 the intracranial interictal recordings identified a second focus. The ESI result was concordant with this second focus. However, because this second focus did not generate seizures and the patient was seizure-free following resection in a remote area, we score this patient’s ESI result incorrect.

FDG-PET was performed in all nine patients with positive postsurgical outcome, but indicated a normal result in two (patient 4, class 2 outcome; patient 10, class 1 outcome). It indicated the hypometabolic area unambiguously with respect to the resected zone in five patients. Ictal SPECT alone (performed in nine patients) localized within the resected zone in five patients; however, in three of these five patients, additional areas with focally increased perfusion were shown. This was also the case for patient 1 whose outcome after surgery was unfavorable. Subtraction analysis (SISCOM) did not increase the yield, and correctly localized in only four patients.

Surgery and histopathology

The resections performed in each patient are listed in Table 1. Two patients had a multilobar resection comprising the temporal and frontal lobe in one patient and the parietal and occipital lobe in the second. The remaining eight patients had unilobar resection (temporal in five, frontal in two, and parietal in one).

The postsurgical follow-up periods were at least 12 months in all patients (range 1–8.7 years, mean 4.3, median 4.4). Histopathologic results of the resected tissue showed diffuse gliosis in four patients; cortical dysplasia in three; microheterotopias in one; diffuse and focal gliosis in one; and a combination of heterotopias, gliosis, and cortical dysplasia in the remaining patient.


Despite major advances in anatomic and functional cerebral imaging, MRN-E still represents the most difficult scenario for presurgical evaluation and eventually surgical treatment, especially if an extratemporal lobe focus is suspected. Previous studies of postsurgical outcome of patients with MRN-E revealed positive outcome between 33% and 83% for MRN-ETLE and between 41% and 93% for MRN-TLE, with an average of approximately 60% for both groups, but slightly better for MRN-TLE than MRN-ETLE (Siegel et al., 2001; Blume et al., 2004). In the series of Blume et al. (2004), only 24% became seizure-free. The remaining had no change or a worsening of seizures post-operatively. A large pediatric series of more than 100 children and adolescents with MRN-E reported that 38–44% of all patients remained seizure-free during up to 10 years of follow-up (Jayakar et al., 2008). Therefore, according to the literature, ≥40–50% of all patients with MRN-E will not become seizure-free and/or have little benefit from surgery, despite invasive chronic EEG monitoring in most cases. It is thus obvious that further tools are needed to determine precisely and unambiguously the site of the epileptic focus.

Although our series of 10 patients with MRN-E is small, we feel that it provided encouraging results. In 8 of ten patients with MRN-E (and 9 with favorable postoperative outcome), ESI correctly localized the epileptic focus within the resected area. In one of the remaining two patients (Patient 7), the ESI focus localization was confirmed by the intracranial recordings as an independent interictal second focus. In the other patient (Patient 1), the ESI localization had to be considered as incorrect (similar to the intracranial EEG), as the patient did not become seizure free postoperatively.

ESI has proven its utility in several studies regarding mainly patients with lesional epilepsy. Lantz et al. (1997) demonstrated the possibility to correctly retrieve temporal lobe spike sources with ESI, as verified by simultaneously recorded intracranial electrodes. Reanalyzing the same data with a more advanced source- and head model revealed clear separation of different spike locations within the mesial and lateral temporal lobe (Lantz et al., 1997). A more recent study by Zumsteg et al. (2005) confirmed that it was possible to localize mesial temporal lobe spikes with ESI, as verified with foramen ovale electrodes. They found the correct source in 14 of 19 different patterns of intracranial mesial temporal lobe discharges.

Sperli et al. (2006) performed ESI analysis in 30 successfully operated pediatric epilepsy patients, of which 17 had ETLE. By statistically comparing the electrical source before and at the negative amplitude peak of the spike (following the statistical parametric method used for functional MRI analysis), significant activity was correctly localized in the resected area in all 17 ETLE cases and in 10 of the 13 TLE cases. The poorer outcome for the TLE patients was well explained by the fact that the usual clinical electrode montages did not cover the basal brain areas sufficiently. In Michel et al. 2004a) ESI was based on 128 channel recordings with sufficient coverage of the basal temporal areas. Thirty-two patients with unambiguous focus identification were analyzed. All 17 TLE patients and 13 of the 15 ETLE cases were correctly localized. Similar accuracy of high-resolution ESI was demonstrated in a recent study by (Brodbeck et al., 2009) in patients with large brain lesions, showing correct localization in 11 of 12 patients with postoperative seizure freedom. Therefore, like any other imaging technique, more sophisticated systems provide more precise results, being it higher field strength in MRI, or more electrodes in ESI (Lantz et al., 2003a; Michel et al., 2004b). Most promising in this respect are the recent progresses in combining ESI and functional MRI (fMRI) (Seeck et al., 1998; Benar et al., 2006). This was demonstrated in two recent studies that compared ESI analysis of interictal spikes recorded within the MR scanner with the concurrent fMRI blood oxygen level dependent (BOLD) responses (Groening et al., 2009; Vulliemoz et al., 2009). Both studies showed that ESI distinguishes the temporal course of the BOLD response into areas of BOLD responses related to initiation from those related to propagation of the epileptic activity.

The major limitation of ESI is the inability to determine the extent of the epileptogenic zone. At present, this is not possible with ESI given that there is no absolute threshold of the strength of the signal. Instead, the displayed size depends on subjective thresholding set by the investigator. This is similar to fMRI, PET, or SPECT imaging in the absence of quantitative or semiquantitative analysis. There currently exists no appropriate algorithm that allows the clear demarcation of the epileptogenic zone based on ESI alone. The statistical mapping approaches described earlier (Zumsteg et al., 2005; Sperli et al., 2006) provide additional information, but this remains to be shown in combined intracranial and scalp recordings. There is evidence that PET is more powerful for the delineation of the extent of the epileptogenic zone (Carne et al., 2004), given the possibility of (semi-)quantitative analysis. Nevertheless, intracranial EEG recording still remains the most important tool in this respect, particularly in patients with MRN-E.

Because intracranial electrodes pick up near-field electrical activity only, the optimal placement is crucial and has to be well determined prior to implantation (Sperling, 1997). In a recent review on outcome after epilepsy surgery, the need for intracranial EEG was identified as a negative prognostic factor (Tellez-Zenteno et al., 2005). Certainly, one reason is the incomplete coverage of the epileptogenic zone due to the absence of visual guidance by the lesion. The present results, combined with the results in this growing body of research indicate that ESI can improve the yield of intracranial EEG monitoring by guiding the sites of implantation.

Another limiting aspect of this and other ESI studies is the fact that they rely on interictal activity. Because the irritative zone is not in all cases concordant with the seizure onset zone (Hirsch et al., 1991; Alarcon et al., 1994), comparison of interictal and ictal activity is mandatory in the presurgical work-up (Pillai & Sperling, 2006). However, in cases of concordant localization, the interictal EEG discharges seem to be more indicative of the affected brain area than the ictal data (Blume et al., 2004; Ray et al., 2007). Using simultaneous intracranial and scalp EEG recordings, Ray et al. (2007) showed that the cortical source area at the rising phase of a spike is more discrete and focal than the area at seizure onset, indicating that the seizure onset zone, recorded from the scalp EEG, might extend beyond the true epileptogenic zone. However, increasing evidence indicates this is correct only if the rising phase of the spike is analyzed. ESI of the spike peak may indicate propagation of the electrical activity within the epileptic network (Lantz et al., 2003b). Unfortunately, the relatively low signal-to-noise ratio at the beginning of the spike makes source localization often difficult and requires averaging of a large number of spikes.

High frequency activity in the gamma range often precedes the spike and might be more focal than the spike itself (Medvedev, 2002). Therefore, spike analysis in the frequency domain focusing on high frequency bands appears to be an important area for future studies. Other approaches of ictal and interictal source localization based on multivariate autoregressive procedures and Granger causality estimation in the inverse space (Ding et al., 2007b; Wilke et al., 2009) could provide additional tools to determine the epileptogenic zone in cases with fast and complex propagation of the epileptic activity. Such methods allow studying connectivity patterns in the epileptic networks within which seizures and spikes propagate. Given the fact that high-resolution long-term EEG monitoring systems are now readily available (Holmes et al., 2008), concomitant ESI analysis of spikes and seizures using novel spatiotemporal analysis procedure is possible.

Despite a large body of literature demonstrating ESI’s focal localization capability, the clinical integration of the technique has been relatively slow. This may be due to the fact that the literature abounds with reports on the technical aspects of ESI that are largely nonintuitive and intractably complex for nonmathematicians. Although ESI analysis does indeed require a certain level of experience in handling the software and knowledge of its pitfalls (as is the case for PET or functional MRI), the most crucial points remain the technically well-performed EEG recording and the correct selection of epileptic spikes that are subjected to the analysis. With increasing availability of computational power and sophisticated analysis software, ESI is now an easily accessible technique to image neuronal activity. The routine use of ESI in presurgical focus localization should be the logical next step (Leijten & Huiskamp, 2008). In their review of ESI in clinical use, Plummer et al. (2008) aptly stated that “the full possibilities of ESI have probably not yet been explored in clinical practice except for specialized centers.”

Although patients with MRN-E have a much higher chance of seizure-freedom than 10 or 20 years ago, these patients are most frequently not referred for further evaluation. Application of ESI to these cases in particular may provide another localizing element that increases the likelihood of surgery. It is of note that in five of our nine patients with positive postoperative outcomes, the histopathologic report indicated the presence of developmental abnormalities known to be both epileptogenic and invisible in the MRI.


The data were analyzed with the Cartool software (http://brainmapping.unige.ch/Cartool.php), which is developed by Denis Brunet from the Functional Brain Mapping Laboratory, Geneva, supported by the Center for Biomedical Imaging (CIBM), Geneva and Lausanne, Switzerland. The work was also supported by the Swiss National Science Foundation by the grant No 320030-122073, No. 3200-113766 (M.S.), No. 3200-111783 (C.M.), and SPUM 33CM30-124089. We thank Stacey Crane for kind revision of the manuscript.

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.


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