CRAN, UMR 7039, Lorraine University, Vandœuvre-lès-Nancy Cedex, France
CNRS, CRAN, UMR 7039, Vandœuvre-lès-Nancy Cedex, France
Address correspondence to Laurent Koessler, CRAN, UMR 7039, CNRS-Université de Lorraine, 2, Avenue de la forêt de Haye, Vandoeuvre-lès-Nancy Cedex 54516, France. E-mail: email@example.com
Delineation of the epileptogenic zone (EZ) in refractory epilepsy related to malformations of cortical development (MCDs) often requires intracranial electroencephalography (EEG) recordings, especially in cases of negative magnetic resonance imaging (MRI) or discordant MRI and video-EEG findings. It is therefore crucial to promote the development of noninvasive methods such as electrical source imaging (ESI). We aimed to (1) analyze the localization concordance of ESI derived from interictal discharges and EZ estimated by stereo-EEG (SEEG); (2) compare the concordance of ESI, MRI, and electroclinical correlations (ECCs) with SEEG-EZ; and (3) assess ESI added value in the EZ localization.
We prospectively analyzed 28 consecutive patients undergoing presurgical investigation for MCD-related refractory epilepsy in 2009–2012. ESI derived from 64-channel scalp EEG was interpreted with blinding to, and subsequently compared with, SEEG-estimated EZ. Anatomic concordance of ESI with SEEG-EZ was compared with that of video-EEG and MRI. We further assessed ESI added value to ECC and MRI.
Twelve patients (43%) had temporal and 16 (57%) had extratemporal epilepsy. MRI was negative in 11 (39%) and revealed a cortical malformation in 17 (61%). ESI was fully concordant with the EZ in 10 (36%) and partly concordant in 15 (53%). ECC presented a full and partial concordance with EZ in 11% and 82% of cases, respectively, and MRI in 11% and 46%, respectively. Of 11 patients with negative MRI, ESI was fully concordant with the EZ in 7 (64%) and partly concordant in 4 (36%). ESI correctly confirmed restricted or added localizations to ECC and MRI in 12 (43%) of 28 patients and in 8 (73%) of 11 patients with negative MRI.
ESI contributes to estimating the EZ in MCD-related epilepsy. The added value of ESI to ECC is particularly high in patients with MCD and negative MRI, who represent the most challenging cases for epilepsy surgery.
A PowerPoint slide summarizing this article is available for download in the Supporting Information section here.
Estelle Rikir is an epileptologist at the University Hospital, Liège, Belgium.
Malformations of cortical development (MCDs) encompass a wide spectrum of congenital cortical structural abnormalities. They constitute one of the main causes of neocortical epilepsy. Drug resistance is almost the rule and was recently estimated to concern 85% of patients with MCDs in a tertiary epilepsy center. In those cases, surgical treatment provides the best results to achieve long-term seizure freedom, and full resection of the epileptogenic zone (EZ) remains the main predictor of seizure freedom after surgery.[3-5] The EZ may be restricted to the malformation, extend to a network of remote cortical areas,[6, 7] or involve only part of the malformation.[7, 8] Therefore, precise estimation of the EZ associated with MCDs still often requires intracranial electroencephalography (iEEG)[3, 5] or intracerebral EEG (stereo-electroencephalography, SEEG) recordings.[6, 7] SEEG targets in MCDs are currently defined on the basis of electroclinical correlations (ECCs) and magnetic resonance imaging (MRI).
There is a growing body of studies that have assessed the clinical value of additional imaging tools such as electric source imaging or magnetic source imaging (ESI, MSI) derived from interictal discharges (IIDs) in the context of presurgical evaluation.[10-17]
However, none of these studies specifically addressed the added value of these methods in the context of MCD-related refractory epilepsy. Past studies have focused on focal cortical dysplasia (FCD) or polymicrogyria (PMG), and were either retrospective or not systematically validated against surgical outcomes or invasive EEG recordings.[18-20]
The primary purpose of this study was to prospectively evaluate the sublobar localization concordance of ESI derived from IIDs and EZ estimated by SEEG in MCDs. The secondary purpose of our study was to compare the concordance of ESI, MRI, and ECCs with SEEG-estimated EZ, and to ultimately assess the added value of ESI in the EZ localization of MCD-related drug-resistant partial epilepsy.
Materials and Methods
Thirty consecutive patients with MCDs (26 from the University Hospital Nancy) were selected among a larger cohort of 85 patients with drug-resistant partial epilepsy undergoing SEEG and prospectively enrolled between October 2009 and March 2012 in the multicenter National Clinical Research Project PHRC 2009-17-05, Clinical trial NCT 01090934 (Nancy, Marseille, Reims). We selected patients who were older than15 years of age and had MRI or electroclinical findings consistent with MCDs. We excluded patients with a contraindication to SEEG. All patients with MCD-related refractory epilepsy who were undergoing presurgical investigations during this period were enrolled in the study, since all underwent an SEEG in order to delineate the EZ, to perform functional mapping and therapeutic thermocoagulation when indicated. This study was approved by the local ethics committee and all patients provided written informed consent.
Noninvasive evaluation included comprehensive medical history, neurologic examination, long-term video-EEG recordings, high-resolution MRI, and neuropsychological evaluation in all cases, as well as optional interictal positron emission tomography (PET) and/or interictal/ictal single photon emission computed tomography (SPECT). We analyzed the ECCs, taking into account the interictal and ictal EEG findings as well as the ictal semiology, combining both interdependent modalities toward dynamic spatiotemporal hypotheses.[9, 21] The sublobar localization of the presumed EZ according to ECC was consensually classified (LGM, JPV, and ER in Nancy; MG and FB in Marseille) in at least one of 18 predefined regions: ventral-medial prefrontal, dorsal-medial prefrontal, ventral-lateral prefrontal, dorsal-lateral prefrontal, medial premotor, lateral premotor, medial central, lateral central, medial-anterior temporal, lateral-anterior temporal, medial posterior temporal, lateral-posterior temporal, medial parietal, lateral parietal, medial occipital, lateral occipital, operculoinsular and temporo-parietooccipital junction (Figs. 1-4).
Structural MRI was acquired with a 1.5 or 3 Tesla Signa General Electric Medical System (GE Healthcare, Milwaukee, WI, U.S.A.) according to a standardized epilepsy protocol. MRI scans were reviewed in a multidisciplinary case management conference by experienced epileptologists and neuroradiologists as to assess (1) the presence of MCDs and (2) the classification of their sublobar localization (Figs. 1-4).
64-Channel scalp video-EEG recordings and ESI
EEG was recorded from 64 scalp-taped electrodes, placed according to the 10/10 system.[23, 24] The localization of all 64 electrodes and of three fiducials (nasion and right and left tragi) was performed prior to acquisition either with a three-dimensional (3D) digitizer system (3Space Fastrak; Polhemus, Colchester, VT, U.S.A.) or with an automated detection and labeling system of MRI-visible sensors (ALLES).
Electrode-skin impedance was below 5 kΩ. EEG was recorded with a 1,024 Hz sampling rate and a 0.53–400 Hz band-pass filter (University Hospital Nancy: Micromed, Mogliano Veneto, Italy; University Hospital Marseille: Deltamed, San Carlos, CA, U.S.A.). The Fpz electrode generally provided the reference, except for frontal lobe cases, where the Oz electrode was used. Hyperventilation trials combined with antiepileptic drug (AED) tapering served to activate IIDs and seizures. Video-EEG was recorded for 4 days, 24 hours/day, in order to study interictal and ictal discharges. Two hours were selected for ESI after careful visual analysis of the full recording according to the following criteria: (1) absence of artifacts, (2) presence of calm wakefulness, and (3) presence of interictal discharges representative of all IID types.
Interictal discharge detection and analysis
Interictal spikes (<70 msec) and sharp waves (<200 msec) were visually identified and marked in an average reference montage by one of four experienced epileptologists (ER, LGM, JPV, or MG) according to the following established criteria: (1) paroxysmal occurrence, (2) abrupt change in polarity, (3) duration <200 msec, and (4) scalp topography consistent with a physiologic field. Temporal windows of analysis (max 100 msec) were defined around the IID and centered at the time of maximal negativity on the electrode trace with the highest amplitude (Advanced Signal Analysis/ASA software, Enschede, The Netherlands).[11, 14] All 64 electrode traces were then superimposed to ascertain that the signal to noise ratio (SNR), defined as the highest IID amplitude divided by the highest background activity amplitude, was >2.5. For each patient, an average of 15 single IIDs were categorized according to their respective topography and morphology, and individually analyzed for source localization (range 1–4 IID types, average of 2 IID types by patient). We chose to analyze individual IIDs as opposed to averaged IIDs because of the risk of merging IID with comparable scalp cartography from different sources that is inherent in averaging.
Volume conduction parameters
We chose to construct realistic head models for each patient because of evidence for lower ESI accuracy with spheric head models derived from a template MRI.[10, 29] Realistic head models derived from an additional MRI sequence (3D BRAVO T1-weighted sequence with pixel size of 1.25 mm2, slice thickness of 1.25 mm without gaps between slices, 252 slices, matrix 192 × 192, field of view of 23 cm, Signa 1.5 Tesla, GE Healthcare). Coregistration of EEG and MRI data derived from the identification of the same three fiducials (nasion and right and left tragi). We performed a semiautomatic segmentation by ASA software that consisted of the identification of three isoconductivity compartments (scalp, skull, and intracranial space), with the skull estimated by the dilation of the intracranial space. We generated a realistic individual head model by the boundary element method (BEM), which describes each of three individual surfaces by triangulation using about 1,700–2,000 nodes per model. We subsequently calculated an electric matrix with a conductivity of 0.008 S/m for the skull and 0.33 S/m for the brain and scalp (conductivity ratio of skull to scalp = 1/40).
Inverse problem methods
Intracerebral sources of each selected IID were modeled by both equivalent current dipole (ECD) and distributed source methods.[31, 32] The ECD analysis was performed over the full duration of the selected temporal window with a moving dipole, involving the calculation of a new dipole localization, orientation, and amplitude that best reproduces the measured electric field for each millisecond, as well as with a rotating dipole, involving the calculation of a unique dipole localization across the time window of analysis. The localization of the moving dipole was considered optimal at the time point where goodness of fit (GOF), reflecting the percentage of EEG data explained by the model, was maximal, generally corresponding to the IID peak with maximal amplitude ratio. Stability of the source over the course of the IID ascending phase and peak was also assessed. No noise floor regularization was done with ECD models. Noise was estimated by the SNR ratio.
We further applied a MultipleSignal Classification Method (MUSIC) that uses a 3D dipole grid (10 mm) model placed in the brain volume combined with the principal component analysis (PCA) method and the standardized low-resolution brain electromagnetic tomography (sLORETA) procedure that relies on a distributed source model[31, 32] to provide a 3D activity distribution over time. For MUSIC, noise floor regularization was performed using the PCA decomposition on the time window of analysis and the selection of the eigenvectors (dipoles), which explain 95% of the signal subspace. For sLORETA, we used the interval window for which we computed the inverse solution to estimate the noise floor. Noise was assumed to be independent in each sensor and taken into account using regularization parameter. In our study, regularization was estimated via Generalized Cross Validation based on “leave one out” method.[14, 32]
Determination of the irritative zone by ESI
Two experienced epileptologists (LGM and JPV in Nancy; MG and FB in Marseille) prospectively and independently interpreted the ESI results in the individual anatomic space for each selected IID and each patient in order to localize the irritative zone (IZ). In the event of discordance between the two interpreters, further joint analysis led to consensus.
The anatomic localization of each source was obtained from the coordinates of moving and rotating dipoles with GOF >90%, of the equivalent dipoles explaining >95% of the signal (eigenvector decomposition) using MUSIC, and of the dipoles with the highest magnitude using sLORETA. For each IID type and each source model, only reproducible localizations were considered relevant. In case of discordance between source models, we solely considered the models that yielded the most reproducible solutions for each IID type. These reproducible anatomic localizations were then classified according to the same 18 predefined sublobar localizations used for ECC and MRI (Figs. 1-4). In the event of multifocal IIDs, each corresponding and reproducible source was anatomically classified. There were 1–4 (mean 2) IID types per patient. This step was performed several months prior to SEEG and therefore epileptologists were blinded to its results.
SEEG recordings and analysis
Intracerebral depth electrodes (Nancy: Dixi Medical, Besançon, France; Marseille: Alcis, Besançon, France) consisting of 5–15 contiguous contacts (length 2 mm, interval 1.5 mm) were stereotactically placed under general anesthesia. A postsurgical computerized tomography (CT) scan performed to rule out hemorrhage was fused with the presurgical MRI to determine depth electrode positions.
SEEG was recorded 20 hours a day for 5–7 days under the same conditions as 64-channel scalp EEG. Spontaneous and electrically induced seizures were analyzed by one of four experienced epileptologists (ER, LGM, JPV, or MG) to estimate the EZ, defined as “the anatomical location of the site of the beginning and of the primary organization of the epileptic discharge.” SEEG-estimated EZ was then classified according to the same 18 predefined sublobar localizations applied to ECC, MRI, and ESI (Figs. 1-4).
ESI and SEEG estimated EZ sublobar concordance
The ESI and SEEG estimated EZ sublobar concordance was assessed and classified as fully concordant, partly concordant, or discordant for each patient. Full concordance corresponded with a complete matching between ESI and EZ sublobar localizations (ESI=EZ). Partial concordance corresponded with a partial matching between ESI and EZ sublobar localizations and encompassed three different conditions: (1) ESI pointed to the full EZ as well as to additional sublobar localizations (ESI > EZ); (2) ESI pointed only to some EZ sublobar localizations (ESI < EZ); (3) ESI pointed to some EZ sublobar localizations and to additional localizations outside the EZ (ESI><EZ). ESI and EZ were discordant if they had no common sublobar localization (ESI≠EZ; Figs. 2-4). In case of multifocal sources corresponding to multifocal IIDs, if all sources were included in but did not fully match the EZ, ESI was considered partially concordant. If none of the sources corresponded to the EZ, they were considered discordant.
Comparison of ECC, MRI, and ESI sublobar concordance with SEEG estimated EZ
We further assessed the sublobar concordance of ECC and MRI with SEEG-estimated EZ according to the same definitions as for ESI (Fig. 1). Cases with negative MRI were considered discordant. The sublobar concordance of ESI, MRI, and ECC were thus compared.
ESI added value
In case of sublobar concordance of ESI with the localizations resulting from ECC and MRI, ESI was considered to confirm these localizations. In case of sublobar discordance of ESI and the localizations resulting from ECC and MRI, ESI could either restrict or add localizations. ESI was considered to have an added value compared to ECC and MRI in cases where it correctly confirmed, restricted, or added valid localizations (using SEEG as reference method) to those obtained from ECC and MRI.
We enrolled 30 patients (11 female) with mean age 28 years at inclusion, including 4 from the University Hospital Marseille and 26 from the University Hospital Nancy. Two patients were excluded due to failure to record seizures in SEEG.
Twelve (43%) of 28 patients had temporal lobe epilepsy (TLE) and 16 (57%) had extratemporal epilepsy (ETLE): 12 (43%) frontal lobe epilepsy (FLE) and 4 (14%) posterior epilepsy, arising from the occipitotemporal or parietal regions.
MRI was negative in 11 patients (39%) and showed a lesion suggestive of MCD in 17 (61%), including FCD in 7 cases, dysembryoplastic neuroepithelial tumor (DNT) in 4, PMG with or without schizencephaly (SCZ) in 3, ganglioglioma in 2, and Bourneville tuberous sclerosis (TS) in a single case (Table 1).
Table 1. Clinical, neurophysiologic, neuroradiologic, and surgical data in the studied population
ESI and EZ were fully concordant in 10 patients (36%), including 7 MRI-negative cases (Figs. 2 and 3), partly concordant in 15 (53%) (Fig. 4), and discordant in 3 (11%). Among the partly concordant cases: (1) ESI was entirely included into the EZ in 3 (ESI < EZ); (2) ESI entirely included the EZ in 6 (ESI > EZ); and (3) ESI and EZ partially overlapped in six patients (ESI >< EZ).
In the 11 MRI-negative patients, ESI was fully concordant with EZ in 7 and partly concordant in 4 patients. ESI was never discordant with EZ in this subgroup. Regarding the 17 patients with MRI evidence of MCD, ESI was fully concordant with the EZ in 3 patients, partly concordant in 11, and discordant in 3. Overall, the full concordance rate amounted to 64% for MRI-negative patients and 18% for patients with MRI evidence of MCD.
In the 12 TLE patients, ESI was fully concordant with EZ in 3 (25%), partly concordant in 8 (67%), and discordant in 1 (8%). Regarding the 16 ETLE patients, ESI was fully concordant in 7 (44%), partly concordant in 7 (44%), and discordant in 2 (12%) (Table 2; see also Table S1).
Table 2. Concordance of ECC, MRI, and ESI with SEEG-estimated EZ
ECC MRI sublobar concordance with SEEG-estimated EZ
ECC and EZ were fully concordant in 3 (11%), partly concordant in 23 (82%), and discordant in 2 patients (7%) (Table 2; see also Table S1). MRI and EZ were fully concordant in 3 (11%), partly concordant in 13 (46%), and discordant in 12 patients (43%), including 11 MRI-negative (Table 2; see also Table S1).
ESI added value
SEEG validated the confirmation of localizations by ESI in one of two cases, the restriction in 11 of 15. In the subgroup of 11 MRI-negative patients, ESI correctly restricted ECC in 8 cases. Regarding the 17 patients with MRI evidence for MCD, ESI correctly confirmed ECC and MRI in one case, correctly restricted ECC and MRI in 3.
ESI presented an added value in 12 (43%) of 28 patients. ESI added value was higher in the MRI-negative subgroup, where it correctly restricted sublobar localizations in 8 (73%) of 11 patients. In contrast, this added value was lower in the MRI-positive subgroup (4 of 17: 23%; Table 2). ESI added value was inferior in patients with temporal (4 of 12: 33%) compared to extratemporal (eight of 16: 50%) epilepsy.
The purposes of this study were to prospectively evaluate the sublobar concordance of ESI and the SEEG estimated EZ in MCD-related epilepsy and to further assess its added value to video-EEG and MRI.
The strengths of our study entail (1) the prospective design and (2) the application of a uniform ESI methodology with 64-channel EEG recordings and realistic head models in all patients.
The overall sublobar concordance of ESI with the EZ is consistent with the previously reported ESI concordance rates of 84–94%.[10, 17] Only three discordant cases were observed including one case with no detectable IID in scalp EEG (patient 28). In the two remaining cases, EZ localized in the lateral part of the right temporal lobe (case 5), or in the left lateral prefrontal area (case 12). In both cases, the EZ localizations over the convexity render a nonobservability of IID sources in scalp EEG improbable[11, 12] and rather suggest a selection bias of IID (Table 1).
The relatively low rate of full concordance of ESI in our study (36%) may be related to the use of IIDs for ESI analysis, whereas the estimation of the EZ relied primarily on the analysis of ictal discharges.
We identified an average of 2 IID types for each patient (range 1–4) corresponding to distinct sources. The partially concordant sources extending beyond the EZ corresponded to the SEEG-defined propagation zone in 10 of 12 cases. Therefore, our results suggest that the interictal sources located outside of the EZ mainly reflected the overlapping of the IZ with both the propagation and the epileptogenic zone.[9, 21] We chose to consider the source with the highest GOF that corresponded to the IID peak in most cases. Sources corresponding to IID peak have been presumed to reflect propagation, whereas those related to IID rising phase have been presumed to depict the EZ more reliably.[34, 35] However, an analysis systematically restricted to the rising phase would have required averaging IIDs in order to increase their SNR, entailing the risk of merging IID with comparable scalp cartography from different sources.
Moreover, in our cohort, 75–100% of individually analyzed IIDs per patient showed a stable source over the superior part of the ascending phase (50% to the peak) with both moving dipole and sLORETA models. This suggests that the IID sources localized in the SEEG-defined propagation zone reflected propagated interictal spikes rather than the intrinsic spatiotemporal dynamic of individual spikes. The relatively low rate of full concordance might also be related to the choice of the reference method. We chose to rely on the SEEG estimation of the EZ because it is physiologically meaningful in the following: (1) validating another electrophysiological investigation, (2) differentiating the epileptogenic and propagation zones, and (3) including cases with a surgical contraindication for functional reasons, as was the case in five patients.
Beyond the issue of the ESI accuracy, it is crucial to determine the ESI added value that represents the cases where ESI correctly confirmed, restricted or added valid localizations (using SEEG as reference method) to those obtained from ECC and MRI.
ESI overall concordance (89%) was comparable to that of ECC (93%) and much higher than that of MRI (57%) because of the high rate of MRI-negative patients in our study. This can be primarily attributed to the focus on MCDs in our study. FCDs constitute the most frequent etiology of MRI-negative refractory neocortical epilepsy in adults. In addition, the rate of ESI full concordance was much higher than that of ECC and MRI (36% vs. 11% and 11%, respectively). Indeed, ESI correctly focused or reinforced hypothesis derived from ECC and MRI in 43% of all patients, meaning that ESI may facilitate ranking various hypotheses derived from video-EEG and MRI.
ESI-added value was higher for ELTE (50%) than for TLE (33%). This might be attributed to the use of 64 electrodes according to an adapted 10/10 international system that did not provide an optimal exploration of the basal temporal region.
The added value of ESI was varied according to the presence or absence of an MRI-visible lesion (23% vs. 73%). Although lower than in MRI-negative patients, the ESI-added value in MRI-positive subgroup was still meaningful in selected cases. In two of three cases with regional PMG (Table 1; see also Table S1), ESI contributed to correctly focus the hypothesis to the epileptogenic part of the malformation in one case (case 2) and correctly pointed to a localization outside the visible lesion in another (case 4). Moreover, in cases of ECC and MRI discrepancy, ESI contributed to determine the epileptogenicity of the MRI-detectable lesion (case 18; Fig. 3) or correctly localized the EZ outside of the visible lesion (case 4). In cases of more limited visible malformations such as FCD and DNT, the EZ may have either a focal lesion-centered or a network organization extending beyond the lesion. In these cases, functional neuroimaging tools such as ESI or MSI can contribute to identify these lesion-centered or network organizations and to unravel the complex relation between lesion and EZ with sources either clustered in the vicinity of the dysplasia or scattered in remote cortical areas.
The added value of ESI was especially relevant in MRI-negative patients (73%), in whom the EZ localization relies primarily on ECC. This higher accuracy most probably reflects the overlapping between irritative and epileptogenic zones observed in FCD that constituted the etiologic substrate of the MRI-negative subgroup. In this subgroup, ESI primarily helped to rank the hypothesis and could thus contribute to appropriately focus the anatomic targets and guide the placement of intracerebral electrodes.
This higher accuracy of ESI in MRI-negative patients with MCD-related epilepsy is particularly important because this subgroup of refractory epilepsy patients has less favorable postsurgical outcomes, especially with respect to extratemporal epilepsy.
Finally, there are some inherent limitations related to our study design. The first potential limitation concerns the smaller sample size, due to our prospective study design and the focus on MCDs, compared to a recent ESI study. Our sample of 28 patients with MCD-related refractory epilepsy represented 33% of a larger prospective cohort that included 85 patients with refractory epilepsy of variable etiology. This is comparable to the 38% MCD rate in this previous study assessing the diagnostic value of ESI regardless of epilepsy substrate. All patients with MCD-related refractory epilepsy undergoing presurgical investigations during this 29-month period underwent SEEG and were enrolled in this study. Altogether, this supports the absence of a selection bias toward cases of noncongruent or noncontributive video-EEG and/or MRI. Moreover, the sample size of our prospective study (n = 28) far exceeds the average sample size (n = 14) of previous retrospective studies assessing the contribution of ESI or MSI in MCD-related epilepsy work-up.[18-20] The second limitation is related to the use of 64-channel compared to 128- or 256-channel EEG recordings that allow a superior spatial sampling. These 64-channel EEG recordings with scalp-taped electrodes, as applied in our study, facilitate long-term sampling. Furthermore, a previous study comparing 31-, 64-, 128-, and 256-channel EEG recordings demonstrated that the most crucial step in increasing source-localization accuracy was related to the increase from 31 to 64 electrodes. Therefore, 64-channel EEG recordings were performed as a viable compromise between dense array and clinical practice. The third limitation is related to the sampling bias of SEEG resulting from the partial coverage of the cortical surface. SEEG estimation of the EZ was chosen as the reference method, as opposed to surgical volume, because it was physiologically meaningful in the following: (1) validating another electrophysiologic investigation, (2) differentiating the zone of seizure initiation from the zone of propagation, and (3) including cases with a surgical contraindication for functional reasons.
A previous retrospective study showed that PET may be useful in the identification of relevant SEEG targets and improve surgical outcome in MCD. In the current study we did not assess the ESI added value compared to interictal PET because PET data were not prospectively collected and blindly analyzed. A previous study comparing ESI to PET in temporal lobe epilepsy suggests that ESI may present an added value to PET findings in MCD, by allowing hypometabolic areas to be distinguished related to initiation of IIDs from propagation areas. This issue will have to be specifically addressed in a dedicated prospective study.
In conclusion, our prospective study showed that ESI combining data derived from 64-channel EEG recordings and from the individual anatomic MRI dataset had a higher concordance with SEEG-estimated EZ and a clinically relevant added value compared to scalp video-EEG and MRI for EZ estimation in MCD. ESI accuracy was particularly high in MRI-negative patients, who represent the most challenging subset of refractory epilepsy in term of EZ localization. Our results strongly suggest that ESI should improve the presurgical noninvasive evaluation of MCD-related partial refractory epilepsy. ESI provides valuable information regarding EZ localization that may improve the diagnostic yield of SEEG by identifying relevant SEEG targets and thus enhance the sampling of the suspected epileptogenic cortical regions.
This study was supported by the French Ministry of Health (PHRC 17-05, 2009). Estelle Rikir was supported by a grant from the Medical Council of the CHU of Liège, Belgium.
None of the authors has any conflict of interest to disclose. 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.