Ictal EEG-fMRI in localization of epileptogenic area in patients with refractory neocortical focal epilepsy




To evaluate the usefulness of ictal electroencephalography (EEG)–combined functional magnetic resonance imaging ( MRI) (EEG-fMRI) in localizing epileptogenic zone in refractory neocortical focal epilepsy.


From the EEG-fMRI database of our institution including 62 adult patients, 14 (age 18–46 years) experienced some ictal event during the test. Data were segmented into 10-s blocks, and the results were analyzed by contrasting each block to the contiguous 10-s block from the onset of seizure onward, in all cases. In seizures lasting >10 s (five cases), a supplementary analysis was performed, contrasting each block to a baseline condition, in the framework of the general linear model (GLM) of analysis. Regions of activations were compared to results from the different techniques performed during presurgical evaluation, such as SISCOM, positron emission tomography (PET), and invasive subdural EEG monitoring.

Key Findings

Regarding the structural MRI findings, nine cases presented some lesion, with blood oxygen level– dependent (BOLD) signal activation placed in the same location in eight of them (89%). SISCOM studies were performed in 11 patients; 5 were concordant with the increase in BOLD signal in a sublobar level, whereas in 3 cases the concordance was in a lobar level. Eleven patients underwent PET studies, being also concordant in a sublobar level in four of them and in a lobar level in four additional cases. Finally, invasive EEG evaluation was performed in three patients and all of them had the seizure-onset zone in the initial area of BOLD activation.


This study adds relevant information to support the integration of EEG-fMRI in the multidisciplinary presurgical workup in patients with refractory epilepsy.

The presurgical workup of patients with refractory epilepsy, including long-term video electroencephalography (EEG) monitoring, neuropsychological assessment, structural magnetic resonance imaging (MRI), functional neuroimaging testing such as ictal single-photon emission computed tomography (SPECT), subtraction ictal SPECT coregistered to MRI (SISCOM), and [18F] fluorodeoxyglucose positron emission tomography (FDG-PET), and in some cases invasive electrophysiologic recordings, aims to establish an hypothesis regarding the localization of the epileptogenic zone, the resection or complete disconnection of which is the ultimate goal of epilepsy surgery (Lüders et al., 1987; Jayakar, 1999; Rosenow & Lüders, 2001). Since the late 1990s, the coregistration of electroencephalography and functional magnetic resonance imaging (EEG-fMRI) has emerged as a noninvasive imaging technique (Ives et al., 1993; Hamandi et al., 2004; Gotman et al., 2006; Laufs & Duncan, 2007; Grova et al., 2008), based on blood oxygen level–dependent (BOLD) signal, which is closely related to changes in cerebral blood flow that accompany focal cortical activation. Continuous or spike-triggered EEG-fMRI has been applied to series of adults and children (De Tiège et al., 2007), also with the intention of inferring the location of the epileptogenic zone.

Many studies using EEG-fMRI have been focused on the noninvasive identification of the irritative zone (Al-Asmi et al., 2003; Hamandi et al., 2004; Gotman et al., 2006; Salek-Hadaddi et al., 2006; Thornton et al., 2010a; Donaire et al., 2013), detecting BOLD signal changes in 40–80% of patients and revealing some degree of concordance with the presumed epileptic generator in 50–100% of the cases. However, there are few large case series reporting patients with ictal EEG-fMRI, taking into account the intrinsic difficulty in performing this test during the ictal period (Jackson et al., 1994; Detre et al., 1995; Krings et al., 2000; Kubota et al., 2000; Salek-Hadaddi et al., 2002; Mórocz et al., 2003; Salek-Hadaddi et al., 2003; Federico et al., 2005; Archer et al., 2006; Di Bonaventura et al., 2006; Kobayashi et al., 2006; Liu et al., 2008; Tyvaert et al., 2008; Donaire et al., 2009a,b; Salek-Haddadi et al., 2009; Thornton et al., 2010b; Fernández et al., 2011; Chaudhary et al., 2012). Most of the studies correlated the BOLD activations in the ictal EEG-fMRI with the results from structural neuroimaging tests (Kobayashi et al., 2006; Tyvaert et al., 2008) or with invasive examinations (Thornton et al., 2010b; Chaudhary et al., 2012). However, few studies compared that with SPECT/SISCOM activations, PET areas of hypometabolism, invasive EEG evaluation, or postsurgery outcome (Donaire et al., 2009a,b; Fernández et al., 2011; Laufs et al., 2011; Donaire et al., 2013). Overall, these reports showed good correlation between ictal BOLD changes with other clinical, imaging, or surgery results.

Our objective was to analyze the usefulness of ictal EEG-fMRI in the presurgical workup of patients with drug-resistant neocortical focal epilepsy, comparing the results with the standardized complementary tests currently performed (electroclinical data, neuroimaging and functional studies—PET and SISCOM, invasive EEG studies, and surgical outcome if available), to localize the seizure-onset zone (SOZ). For that purpose, we applied a method previously described by our group (Donaire et al., 2009a,b) to a series of 14 patients with drug-resistant neocortical focal epilepsy, who experienced a total of 17 seizures during EEG-fMRI recordings.


We selected 14 patients from the EEG-fMRI database of our institution (2006–2008, including a total of 62 adults), with neocortical epilepsy who had experienced some ictal event during the recording. We did not include mesial temporal lobe epilepsy in order to make the sample more homogeneous, considering that some patients with mesial temporal lobe epilepsy may not be well reflected on the scalp EEG (Le Van Quyen et al., 2001). All patients presented EEG changes that progressed in frequency, amplitude, and distribution in a similar way to seizures recorded during prolonged video-EEG monitoring, in some cases without clinical symptoms (subclinical seizures). As part of their presurgical evaluation, they underwent a directed clinical history, neurologic examination, long-term video-EEG monitoring, neuropsychological assessment, structural MRI, SISCOM, PET, and invasive-EEG evaluation with subdural electrodes (if needed). Patients were required to have predictable seizures to be considered candidates for ictal-fMRI, excluding those who usually presented seizures associated with significant movements. This study was approved by the research ethics committee of our institution, and all procedures were performed after obtaining the patient's informed consent.

Acquisition of fMRI data

fMRI was acquired on a Signa 1.5-Tesla General Electric Magnetic Resonance Scanner (GE Medical Systems, Milwaukee, WI, U.S.A.) in 12 patients and on a 3-Tesla Siemens Magnetic Resonance Scanner (3T Magnetom Trio; Siemens, Erlangen, Germany) in two patients. First, a structural high-resolution axial three-dimensional T1-weighted image was acquired using a fast spoiled gradient-recalled (FSPGR) acquisition sequence. Therefore, functional images were acquired as a series of single-shot gradient-echo planar imaging (EPI) volumes providing T2-weighted BOLD contrast (repetition time/echo time [TR/TE], 2,000/34 msec for 1.5T and 2,000/16 msec for 3T scan; field of view [FOV], 24 × 24 cm, 64 × 64 pixel matrix for 1.5T and 128 × 128 for 3T scan; slice thickness, 5 mm for 1.5T and 3 mm for 3T scan; gap, 1.5 mm; 20 axial slices per scan for 1.5T and 40 axial slices per scan on 3T). Continuous whole-brain EPI-BOLD volumes were acquired over runs of 11 min 20 s, corresponding to 340 scans, until at least one clinical seizure was registered. A maximum of four runs were scheduled per patient. Antiepileptic treatment was reduced 24 h before recording to increase the likelihood of seizures inside the scanner. The EEG signal was continuously acquired inside the MRI scanner using 27 MRI-compatible electrodes (10/20 montage). Data were transmitted from a BrainAmp amplifier (sampling rate, 5 kHz; Brain Products, Munich, Germany) via a fiberoptic cable to the EEG recorder located outside the scanner room. EEG and electrocardiography (ECG) signals recorded during MRI were postprocessed using Vision Analyzer software (Brain Products), and cardioballistic and gradient artifacts were removed. EEG recordings were independently reviewed by three different epileptologists (I.M, A.D, A.S). All patients were under permanent supervision (I.M, A.D) during EEG-fMRI register to record the onset and the end of ictal semiology and to ensure that recorded seizures were similar to their typical seizures. Seizure onset and seizure end were therefore determined by simultaneous EEG recordings and/or clinical correlates.

fMRI analysis

fMRI images were processed and analyzed using sequential fMRI analysis (Donaire et al., 2009a) by means of seqfMRI, an in-house SPM8 (Institute of Neurology, University College London, United Kingdom) toolbox running on MATLAB (Mathworks, Inc, Natick, MA, U.S.A.). Two types of analyses were performed. The first one, called Continuous Forward (CF), consists of comparing two consecutive blocks within the analysis windows. The second way, called Blocked Base period (BB), consists of comparing the activation period with a fixed rest period previously defined. The CF method gives the relative variations of consecutive blocks, being useful to analyze the sequential evolution of the BOLD signal. It was applied by default to all cases. On the other hand, the BB method shows the absolute variations with respect to the resting period; it was also applied in five patients in whom seizure duration exceeded 10 s (to avoid a possible cancellation of BOLD signal) and with a well-established resting period without interictal discharges, according to the simultaneous EEG information in each case (13 s at least). Finally, we also analyzed interictal activity in 10 of 14 patients (in the rest four patients, it was not possible due to the presence of nearly continuous interictal activity, complicating the analysis).

Preprocessing steps involved realignment of fMRI images for motion correction, normalization to the Montreal Neurological Institute (MNI) template for activation area labeling, and smoothing with a Gaussian kernel of 8 mm. Relevant parameters of the sequential analysis used in this work were an analysis window size of 44 s (22 scans), split in 10 s activation blocks (five scans). For each seizure, the preictal (EEG changes without clinical seizures) and ictal phases (from clinical seizure onset to seizure end as indicated by the patient) were analyzed. We did not include the late-ictal period as it has demonstrated a lower localizing value in other studies (Thornton et al., 2010b; Chaudhary et al., 2012).

The statistical significance threshold was set to a p < 0.05, corrected for family wise-error, minimum cluster size of 50 voxels.

Resulting statistical maps were overlaid on the patient-normalized T1-weighted structural MRI. The visual evaluation of hemodynamic changes occurring along the seizures was performed by neurology and neuroradiology experts (A.S., N.B.) to determine, according to the EEG information, the SOZ and its propagation. The examiners were blind to the results of other tests, except the semiologic and EEG data, and the structural imaging, which were indispensable to conduct the technique.

The anatomic location of the cerebral-activated areas was determined using the AAL toolbox of the SPM8 (Tzourio-Mazoyer et al., 2002).

Invasive studies and surgical outcome evaluation

Three patients underwent invasive EEG evaluation with subdural electrodes. The mean number of electrodes used was 66 (64–70). Comparison was performed with the anatomic location of the initial contacts involved in seizure generation. In addition, surgical outcome was classified with Engel's classification of postoperative outcome (Engel & Jerome, 1993).

Data analysis

The first activation cluster in fMRI with a minimum of five contiguous voxels with a ǀtǀ > 3 (corresponding to p < 0.05 corrected for the multiple comparisons resulting from the number of voxels in the brain), time-locked to the electrographic seizure onset, sustained in time (repeated in at least three correlative comparisons), and whose spatial distribution evolved in agreement with the previous electroclinical information concerning the video-EEG monitoring (AS, IM), was considered the SOZ. Later on, the neurologist (AS) identified the seizure period in the EEG and the neuroradiologist (NB) identified the seizure period in the fMRI. Where both coincided, a simultaneous EEG-fMRI activation was registered. Finally, these activation zones were compared with the results determined by other diagnostic methods (AS, JS, NB, SR), including long-term video-EEG, structural MRI, SISCOM, PET, and invasive EEG evaluation and postoperative outcome, if available.

The degree of concordance was determined as follows: (1) discordant when there was a different location of SOZ determined by BOLD signal analysis, and the other complementary tests (e.g., presumed SOZ placed in a different hemisphere than the onset of seizures determined by video-EEG, the location of the structural lesions or the initial activations in SISCOM, or else PET revealing occipital hypometabolism with BOLD activation in frontal regions); (2) partially concordant: in the case of video-EEG, the epileptiform activity was diffuse but maximum on the regions of BOLD activations; for structural imaging, SISCOM, PET, and subdural electrodes, the BOLD activations were coincident in a lobar, but not at a sublobar level; and (3) totally concordant, for video-EEG, BOLD activations were located over the same regions than focal epileptiform activity; for structural imaging, SISCOM, PET, and subdural-EEG, the areas of initial activation, were coincident with the increase in BOLD signal in a sublobar level.


Simultaneous EEG-fMRI recordings were well tolerated by all patients, and none of them suffered from injuries or damages as a result of motor seizures inside the scanner. Clinical characteristics and the results of the preoperative studies, including ictal-fMRI findings, are included in Table 1.

Table 1. Summary of clinical characteristics and results of the noninvasive testing
PatientAgeGenderSzSz inside the fMRI and duration (s)Scalp-EEGStructural MRIBOLD activations in EEG-fMRISISCOMPETContributions of EEG-fMRI in the presurgical workup
  1. Sz, seizure; BAT, bilateral asymmetric tonic; R, right; L, left; Gen, generalized; 2G, secondary generalization; F, frontal; T, temporal; P, parietal; O, occipital; max, maximum; FCD, focal cortical dysplasia; t, time of isotope injection; SOZ, seizure-onset zone.

130Axial tonicAxial tonic sz (12)Mesial FFT Subcortical band heterotopia + P dysplasiaR opercular gyrus and bilateral PR inferior F→anterior cingulate, corpus callosum (t:8s)More precise delimitation of SOZ
240BAT szSubclinical (11)Gen (max bifrontal)L F porencephalic cystL middle and inferior F gyri, adjacent to the surgical cavity→R frontal cortexL middle and inferior F gyri (t:26s)L middle and inferior F gyriConcordance with other tests
318R arm somatosensory aura→R arm tonic sz→BAT szSubclinical (10)L mesial FNormalL angular gyrusL angular gyrusConcordance with other tests
446L arm tonic szL arm tonic sz (4)R mesial FNormalR precuneus gyrusR inferior P, R postcentral gyri (t:28s)R inferior P, R postcentral gyriConcordance with other tests
533Nonspecific aura→BAT sz→R arm tonic szAura→BAT sz→R arm tonic sz (13)Gen (max bifrontal)L hemispheric porencephalic cystL inferior T gyrus in the anterior pole and L middle T gyrus adjacent to the cystL inferior T, middle T, fusiform, hippocampus, parahippocampal gyri. (t:42s)Concordance with other tests
627Loss of consciousnessLoss of consciousness (6)R T PR rolandic FCDPosterior region of the R superior T gyrus→R Supramarginal (inferior P)R middle temporal gyrusMore precise delimitation of SOZ
727L arm somatosensory aura→L arm clonic szSubclinical (9)Bilateral mesial FP (max on the R)NormalR posterior mild T→R Inferior T gyrusR anterior T, R anterior cingulate (t:4s)NormalMore precise delimitation of SOZ
830Tonic sz→R clonic szSubclinical (11)L FL superior and medium F gyrus FCDL inferior F gyrus (anterior region)L middle and superior F gyri (t:26s)L inferior, middle and superior F gyriMore precise delimitation of SOZ
935Abdominal, psychic and visual auras→Loss of consciousnessAbdominal, psychic and visual auras (7)Gen max FT (R predominance)R basal T cavernous hemangioma with associated dysplasiaR precentral, postcentral, middle and inferior F and R inferior P gyriR inferior and superior P and supramarginal gyri (t:28s)L middle and inferior F, L inferior, middle and superior TPointing-out SOZ in non-conclusive patients
1029Ictal speech→Versive to the LSubclinical (10)Gen max bifrontal (R predominance)NormalR precentral gyrus and R insular cortexL postcentral (t:45s, postictal)L postcentralDiscordance with other tests
1137Visual aura→L face somatosensory aura→AutomatismsVisual→L face somatosensory aura (8)R hemispheric (FT predominance)R T FCDR middle and superior T gyrus (posterior region)R middle and superior T→R T pole, inferior T and R hippocampus (t:26s)R inferior, middle and superior T, hippocampus, fusiform gyri and R insulaMore precise delimitation of SOZ
1234Versive (to the R)→BAT with R arm clonic szBAT with R arm clonic sz (20)L mesial FNormalL superior F gyrus (mesial predominance)Inconclusive (t:46s, postictal)L superior and middle F gyrusMore precise delimitation of SOZ
1330BAT sz→R arm clonic szSubclinical (14)L hemispheric (max mesial F)L FTP porencephalic cystL postcentral gyrus (adjacent to the cyst)L postcentral and precentral gyri, R anterior superior F (t:15s)Diffuse hypoperfusion, involving all L P, L superior F and precentral gyriConcordance with other tests
1423Versive to the RSubclinical (10)Bifrontal max LL F FCDL orbitofrontal gyrusMore precise delimitation of SOZ

The medium age was 31.6 (18–46) years; there were eight females and six males. The mean duration of epilepsy was 20.4 (14–35) years. Seven patients experienced at least one of their typical clinical seizures inside the fMRI scanner, consisting of an axial tonic (1), partial motor (4, 5, and 12), auras (9 and 11), and dialeptic seizures (6). The remaining seven patients presented at least one typical subclinical seizure. Mean duration of recorded seizures was 10 (4–20 s). As mentioned earlier, all of them presented an electrographic pattern similar to the long term video-EEG register.

Regarding the preoperative structural MRI findings; five of them were considered normal, five revealed malformations of cortical development (MCDs), four cases of focal cortical dysplasia (FCD) and one case of subcortical band heterotopia; three presented a porencephalic cyst; and one patient presented a cavernous angioma.

CF method was applied by default to all cases, being able to determine significant BOLD activations in 11 patients (79%). BB method was also applied to those cases showing both a long (>10 s) ictal event and a well-established resting period to get complementary information (five patients). This method was able to show significant BOLD activations in four patients (80%), in the three cases with negative CF, and in the additional patient with positive CF, showing similar activation areas than with the CF method. Overall, all 14 patients showed significant changes in BOLD signal. Quantitative analysis yielded a mean maximum t-value of 10.27 (6.17–14.77) in the 14 patients. Ten patients showed activations with peak-level and clusters, whereas four patients showed only clusters activations. Indeed, we analyzed the degree of concordance in BOLD activation signals between ictal and interictal EEG in 10 patients. Therefore, interictal activations were superimposed to the ictal maps in six patients (60%) and discordant or absent in four patients (40%).

We also investigated the concordance of the ictal increase in BOLD signal in relation to the location of the SOZ suggested by different tests (Table 2). Regarding the video-EEG, BOLD data were concordant in all patients. From the total of nine lesional patients, eight (88.89%) presented the BOLD area of activation in the same area than the lesion in the structural MRI. In addition, SISCOM studies were performed in 11 patients—mean time of isotope injection after seizure onset 26.72 (4–46 s); 8 (73%) were partially or totally concordant with the increase in BOLD signal. Eleven patients underwent PET studies, showing some degree of concordance also in eight of them (73%). Invasive EEG evaluation was performed in three patients (2, 11, 12), and in all of them the electrodes involved in seizure onset were located around the area of initial BOLD activation (100%). Finally, five patients were operated on, with a medium follow-up after surgery of 2.4 (2–3.5 years). One patient achieved the Ia group of Engel's classification outcome, one patient the Ib, and the three additional patients the IIIa group. Both patients within I group showed a good concordance between areas of BOLD activation and electrodes implied in the SOZ. By contrast, two of the three patients within the IIIa group underwent palliative surgery, because of the implication of the eloquent cortex in the SOZ determined both by EEG-fMRI and invasive evaluation. The remaining patient within IIIa (9) did not show a good concordance. The results of the EEG-fMRI, invasive subdural-EEG studies, type of surgery performed, pathology, and postoperative outcome are summarized in Table 3. In particular, patient 2 (Fig. 1), who previously had undergone a tailored resection of the left prefrontal area, a subdural grid of 6 × 8 electrodes was placed in the left lateral frontal lobe, and another grid of 2 × 8 was placed in the left mesial frontal lobe. The SOZ was localized over the superior and middle frontal gyri, around the supplementary motor area, overlapping with the EEG-fMRI initial activations (left superior frontal gyrus adjacent to the surgical cavity, propagating to the right frontal cortex). Epilepsy surgery consisted of a left mesial and lateral premotor and prefrontal corticectomy, but it did not include all the electrodes implied in the SOZ in order to preserve the motor function. The pathology was consistent with active chronic encephalitis. After 2 years of follow-up, the patient has noticed a 50% of reduction in seizure frequency and intensity. Patient 11 (Fig. 2) was evaluated with a subdural grid (8 × 8) over the right lateral temporal, inferior-posterior frontal, inferior parietal and anterior occipital lobes, and also a 6-electrode strip over the right mesial temporal region. The SOZ was estimated to be located over the posterior middle temporal gyrus, coinciding with the areas of ictal BOLD activation (posterior part of the right middle and superior temporal gyrus), leading to a tailored resection of the affected cortex. After 2 and a half years of follow-up, the patient is completely seizure-free. Finally, in patient 12, an 8 × 8 subdural grid was placed over the left frontal lobe, concluding that the SOZ was located in the precentral and middle frontal gyri, including the primary and secondary motor areas. The EEG-fMRI initial activations were placed slightly above (left superior frontal gyrus with a mesial predominance). To preserve the motor function, palliative surgery was conducted with an incomplete tailored resection, and after 1 year remaining asymptomatic his seizures returned, although less frequently (seizure reduction of 50%). Both patients 11 and 12 presented a type I focal cortical dysplasia (FCD) in the pathology samples.

Table 2. Degree of concordance between seizure-related significant BOLD-activation signals in fMRI sequences and the putative seizure onset zone (SOZ) identified by different testing
 Discordant (%)Concordant in a lobar level (%)Concordant in a sublobar level (%)
  1. p, patients.

video-EEG (14 p)05 (36)9 (64)
Structural MRI (9 p)01 (11)8 (89)
SISCOM (11 p)3 (27)3 (27)5 (45)
PET (11 p)3 (27)4 (36)4 (36)
Invasive EEG (3 p)003 (100)
Table 3. Summary of invasive subdural-EEG evaluation, type of surgery performed, pathology, Engel's classification of postoperative outcome and follow-up since surgery
PatientEEG-fMRISubdural EEG evaluationType of surgeryPathologyEngel's classificationFollow-up since surgery (years)
  1. NP, nonperformed; L, left; R, right; HSV, herpes simplex virus; FCD, focal cortical dysplasia.

2L superior F gyrus)→R frontal cortexL superior and middle F gyri, SMAL mesial and lateral prefrontal corticotomy (palliative)Active chronic encephalitis HSV-1IIIa2
9R pre central and middle frontal gyriNPAnteromesial temporal lobectomyCavernous angiomaIIIa2
11R superior and middle T gyrus (posterior region)R middle T gyri (posterior region)Tailored resectionFCD type IIa2.5
12L superior F gyrusL pre central and middle F gyriIncomplete tailored resection (palliative)FCD type IIIIa3.5
13L post central gyrusNPHemispherotomyReactive gliosis of astrocytesIb2
Figure 1.

The findings of different testing in patient 2. (A) EEG during the simultaneous fMRI recording: Run of epileptic discharges over the frontal lobes, maximum on the left. (B) fMRI: Consecutive activations in BOLD signal: Left frontal BOLD activation, adjacent to the surgical cavity (superior frontal gyrus), propagating to the right frontal cortex. (C) SISCOM: Left frontal activation; (D) PET: Left frontal hypometabolism; and (E) subdural-EEG evaluation: Seizure-onset zone placed over the superior and middle left frontal gyri.

Figure 2.

The findings of different testing in patient 11. (A) EEG during the simultaneous fMRI recording: Run of spike and spike-and-wave discharges over the right hemisphere (frontotemporal predominance). (B) fMRI: BOLD activation of the posterior region of the right middle superior temporal gyrus. (C) SISCOM: Right temporal hyperperfusion; (D) PET: Right lateral temporal hypometabolism; and (E) subdural-EEG evaluation: Seizure-onset zone placed over the posterior middle temporal gyrus.

Figure 3.

The significant contribution of EEG-fMRI in patient 9, in which presumably SOZ was pointed out in a previously inconclusive patient. (A) fMRI: BOLD activations in right precentral and middle frontal gyri. (B) SISCOM: Hyperperfusion in right inferior and superior parietal and supramarginal gyri. (C) PET: Hypometabolism in left middle and inferior frontal, left inferior, middle and superior temporal gyri.

Two additional patients underwent surgery without previous invasive monitoring. In patient 9 a resection of the cavernous angioma associated with right anteromesial temporal lobectomy resulted in a significant reduction (>75%) after 2 years of follow-up. Lastly, patient 13 underwent a left parietal lobectomy and inferior precentral gyrus corticotomy, including the posterior edge of a congenital porencephalic cyst. Because of the recurrence of seizures after palliative surgery, a hemispherectomy was finally performed.

To assess the clinical relevance of this technique, ictal EEG-fMRI provided clinically additional information in 8 (57%) cases, mainly contributing to delimitate the area of SOZ more precisely than other non-invasive tests, or else pointing out this area in nonlesional or nonconclusive patients (Table 1 and Fig. 3).


This study shows that ictal EEG-fMRI is a useful tool in the noninvasive localization of SOZ in selected patients with symptomatic and cryptogenic drug-resistant neocortical focal epilepsy. Both methods, applied in a complementary manner, were able to demonstrate BOLD signal activations in all patients (11/14 with CF method and 4/5 with BB method), and in three of them colocalization could be confirmed by invasive EEG monitoring. All patients showed some degree of concordance with previous video-EEG monitoring. Within the nine patients with a lesion on the structural MRI, only one patient (9), was not concordant, but we hypothesize that the presence of calcium deposits in the cavernous angioma could disrupt the BOLD signal.

Up to now, only two large case series of ictal EEG-fMRI have been reported (Thornton et al., 2010b; Chaudhary et al., 2012), including 9 and 20 patients, respectively. Both of them compared BOLD signal activations triggered by proven typical seizures mainly with the SOZ demonstrated by invasive evaluation (eight and six patients with intracranial EEG recordings, respectively) with a high degree of concordance in a sublobar level. In contrast, our study provides data of comparison between BOLD activations and noninvasive studies, demonstrating also a good concordance, either in a lobar or a sublobar level. As mentioned, only a few studies have compared functional neuroimaging tests with ictal EEG-fMRI (Donaire et al., 2009a,b; Fernández et al., 2011; Laufs et al., 2011; Donaire et al., 2013), supporting the integration of this technique in the presurgical workup of patients with refractory epilepsy. Otherwise, in previous studies (Thornton et al., 2010b; Chaudhary et al., 2012), seizures were divided in different phases, demonstrating that the degree of concordance was significantly better for the phases corresponding with the earliest observable changes in EEG (p < 0.05). For that reason we have considered only these initial stages, analyzing seizures as a whole and not fragmented.

A short number of studies with ictal EEG-fMRI have included patients with MCD, all of them comparing with structural imaging. Therefore, BOLD signal increase was detected in the right angular gyrus in a patient with a nodular heterotopia in the same location, and presenting with simultaneous brief repetitive electrographic seizures (Kobayashi et al., 2006). Otherwise, in eight patients with different types of MCD, the relationship between interictal and ictal EEG-fMRI zones was analyzed, finding that dysplastic and heterotopic cortex of band heterotopia were involved in the interictal and seizure processes, whereas nodular heterotopias did not generate seizures (Tyvaert et al., 2008). In contrast with our study, the four anatomic regions affected by FCD were involved in the maximum area of BOLD activations, and in the case of the heterotopia, the BOLD maximum signal was placed over the areas of greatest epileptiform activity.

In conclusion, the application of ictal EEG-fMRI in patients with focal refractory epilepsy seems to be reliable, even in the difficult cases of nonlesional epilepsy. However, at the present moment, it has significant limitations to be used in the routine presurgical evaluation: It can only be applied to patients whose seizures are associated with none or minimal head movement; on the other hand, hemodynamic activation and deactivation patterns are sometimes difficult to interpret. Our study is limited by the fact that only five patients underwent surgery, and in the case of the three patients who had invasive EEG monitoring (considered the goal standard technique to locate SOZ), two had partial resections to spare eloquent cortex. Therefore, our data do not support the association of this technique with the postsurgical outcome, even if these findings must be interpreted cautiously, given the small sample and the complexity of patients. So we do not have a definitive proof of the accuracy of the technique; however, we feel that the high rate of concordance with the functional neuroimaging tests usually performed during presurgical evaluation support its use in selected patients, where it may help to decide the resection or plan the placement of invasive electrodes.

Further studies will help to elucidate the exact role of this technique in the evaluation of drug-resistant epilepsy.


This report was supported by “Fondo de Investigación Sanitaria” PI 050052 (Spain) and by “Fundació la Marató de TV3 Catalunya” PI 060910 (Spain).


Dr. Sierra and Dr. Maestro were sponsored by “Fundació la Marató de TV3 Catalunya” PI 060910 (Spain). They performed the EEG-fMRI process with the support of “Fondo de Investigación Sanitaria” PI 050052 (Spain). Dr. Donaire, Dr. Falcón, Dr. Carreño, Dr. Aparicio, Dr. Rubí, Dr. Setoain, Dr. Rumià, Dr. Pintor, Dr. Boget, and Dr Bargalló report no disclosures. 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.