Magnetic source imaging and ictal SPECT in MRI-negative neocortical epilepsies: Additional value and comparison with intracranial EEG


Address correspondence to Dr. med. Felix Schneider, Department of Neurology, Epilepsy Center, University of Greifswald, Sauerbruchstrasse, 17489 Greifswald, Germany. E-mail:


Purpose:  To investigate the utility of magnetic source imaging (MSI) and ictal single photon emission computed tomography (SPECT), each compared with intracranial electroencephalography (EEG) (ICEEG), to localize the epileptogenic zone (EZ) and predict epilepsy surgery outcome in patients with nonlesional neocortical focal epilepsy.

Methods:  Studied were 14 consecutive patients with nonlesional neocortical epilepsy who underwent presurgical evaluation including ICEEG, positive MSI, and localizing subtraction Ictal SPECT coregistered to MRI (SISCOM) analysis. Follow-up after epilepsy surgery was ≥24 months. ICEEG, MSI, and SPECT results were classified using a sublobar classification.

Key Findings:  Of 14 patients, 6 (42.9%) became seizure-free after surgery. Sublobar ICEEG focus was completely resected in 11 patients; 5 (45.5%) of them became seizure- free. Concordance of ICEEG and MSI and complete focus resection was found in 5 (35.7%) patients; 80% of them became seizure-free. Sublobar ICEEG-MSI concordance and complete focus resection significantly increased the chance of seizure freedom after epilepsy surgery (p = 0.038). In contrast, of the 6 patients (42.9%) with concordant ICEEG and SISCOM and complete focus resection, only 66.7% became seizure-free (p = 0.138). Assuming concordant results, the additive value to ICEEG alone for localizing the EZ is higher with ICEEG-MSI (odds ratio 14) compared to ICEEG-SISCOM (odds ratio 6).

Significance:  This study shows that combination of MSI and/or SISCOM with ICEEG is useful in the presurgical evaluation of patients with nonlesional neocortical epilepsy. Concordant test results of either MSI or SISCOM with ICEEG provide useful additive information for that provided by ICEEG alone to localize the EZ in this most challenging group of patients. When sublobar concordance with ICEEG is observed, MSI is more advantageous compared to SISCOM in predicting seizure-free epilepsy surgery outcome.

The epileptogenic zone (EZ) is the area of the cortex that is indispensable for the generation of the patient’s seizures; resection of this area will by definition render the patient seizure free (Rosenow & Lueders, 2001). Invasive and noninvasive electroencephalography (EEG) is considered as the primary tool in evaluation of surgical candidates to localize the EZ (Knowlton et al., 2008a). Identification of an underlying lesion on magnetic resonance imaging (MRI) concordant with electrophysiologic findings has the highest predictive value for a favorable surgical outcome (Ferrier et al., 1999; Hong et al., 2002; Stavem et al., 2004; Jeha et al., 2006; Wieshmann et al., 2008; Téllez-Zenteno et al., 2010). Despite advances in structural and functional MRI, 20–30% of patients with temporal lobe epilepsy (TLE) and 20–40% of those with extratemporal lobe epilepsy (ETLE) show no evidence of an underlying lesion on MRI (MRI-negative) (Kutsy, 1999; Hong et al., 2002; Carne et al., 2004; Téllez-Zenteno et al., 2010). In patients undergoing evaluation for surgical treatment of their epilepsy, the absence of a lesion on MRI is the worst prognostic sign for surgical treatment (Berkovic et al., 1995; Tonini et al., 2004), especially when an extratemporal focus is suspected (Brodbeck et al., 2010). Previous outcome studies in patients with nonlesional ETLE revealed a seizure-free outcome in 24–42% (Van Ness, 1992; Zentner et al., 1996; Blume et al., 2004; Cohen-Gadol et al., 2006; Wetjen et al., 2009). Patients with neocortical TLE are considered to have slightly better outcome; prior investigation showed a seizure-free rate of 54.8% (Lee et al., 2005).

The identification of the EZ in patients with nonlesional neocortical epilepsy is a great challenge. Although a recent study, which specifically focused on stereoelectroencephalography (SEEG), showed that SEEG is equally effective in MRI-negative and lesional MRI cases to localize the EZ (McGonigal et al., 2007), a multimodal noninvasive evaluation is indispensible to make the surgical decision and to determine the site and size of surgical resection (McGonigal et al., 2007; Knowlton et al., 2008a,b). Magnetoencephalography (MEG) and single photon emission computed tomography (SPECT) are noninvasive techniques that have shown promise in the presurgical evaluation of focal epilepsies. The aim of this study was to evaluate the localization agreement of both MEG and SPECT in patients with MRI-negative epilepsy, each compared with intracranial EEG (ICEEG) as the current “gold standard.” We also aim to investigate if a multimodal approach contributes to better localization of the EZ and prediction of surgical outcome in patients with consecutive nonlesional neocortical epilepsy.

We retrospectively analyzed clinical profiles, video-EEG (noninvasive and invasive), MEG and SPECT data, surgical procedure and size of surgical resection, and pathologic findings as to their relation to postsurgical seizure outcomes in a cohort of patients with nonlesional neocortical focal epilepsy.



The Cleveland Clinic Institutional Review Board approved this study. Data from patients who underwent presurgical evaluation, including MEG and both ictal and interictal SPECT studies using SISCOM (subtraction ictal SPECT coregistered to MRI), at the Cleveland Clinic Epilepsy Center between February 2008 and August 2010 were studied retrospectively. Of 178 patients screened, 54 (30.3%) had negative preoperative MRI (1.5 T or 3 T, epilepsy protocol). We further screened patients based on the following additional inclusion criteria: (1) diagnosis of medically refractory neocortical epilepsy based on electroclinical findings from video-EEG monitoring, (2) localizing SISCOM as determined by nuclear medicine physician; (3) positive MEG (at least five spikes detected on MEG) as determined by trained epileptologists/magnetoencephalographers, and (4) resective epilepsy surgery was performed with a postoperative follow-up of ≥24 months. The predominant reason for exclusion from the study was a positive MRI result and previous surgery. Fourteen patients met all the criteria above and were analyzed in this study. Surgical decision and determination of the site and size of surgical resection was made at an interdisciplinary epilepsy patient management conference. In all patients, postoperative MRI investigations were performed. Preoperative MEG and SPECT results were superimposed on the postoperative MRI to evaluate the exact extent of their location as compared to surgical resection.

Noninvasive and invasive video-EEG monitoring

Routine EEG and noninvasive video-EEG monitoring were performed (international 10–20 system of electrode placement including anterior temporal and sphenoidal electrodes if appropriate) (Silverman, 1960). In cases of invasive video-EEG monitoring using ICEEG, the electrodes (subdural grid electrodes, depths electrodes, and/or SEEG electrodes) were placed according to the consensus hypothesis of epilepsy localization based on the findings of the presurgical evaluation. ICEEG electrode placement was confirmed by computerized tomography (CT) performed after electrode implantation.

MEG/Magnetic source imaging (MSI)

We performed an approximately 50-min resting-state recording of spontaneous MEG activity on all patients (minimum 3, 10-min recordings; maximum 12, 10-min recordings) using a whole-head, 306-channel Neuromag system (Elekta, Helsinki, Finland). Continuous head position monitoring was performed using five head position indicator (HPI) coils. EEG was recorded simultaneously. Sampling frequency of the MEG and EEG was 1,000 Hz. Data were subsequently filtered with 0.1 Hz high-pass and 333 Hz low-pass filters. Spike analysis using Neuromag software was performed on data segments (approximately 200 msec duration) that contained epileptiform discharges. This analysis was performed without prior knowledge of the clinical history. The location, orientation, and strength of the dipole sources that best fit the measured magnetic fields were calculated with standard Neuromag software using nonlinear least squares algorithms to fit the magnetic fields predicted by a single equivalent current dipole model, an established model that has been widely used in clinical MEG (Stefan et al., 2003; Pataraia et al., 2004). A minimum number of five spikes was required for a localization result. The MEG image was then coregistered to the preoperative MRI to evaluate the anatomic localization.


SPECT was performed during video-EEG monitoring using 99m-Tc-ECD (99m-Tc labeled ethylcysteinate diethylester) at a dose of 25–40 mCi (adults, body weight of 70 kg). Radiotracer was injected by specially trained nurses who were at bedside at seizure onset. In ictal studies, the radiotracer was injected immediately after either clinical or electroencephalographic seizure onset. Seizure onset and ending times as well as the radiotracer injection times (time from the beginning of injection up to the complete depression of the syringe plunger) were retrospectively calculated by reviewing the video-EEG recordings. A minimum of 15 s between the end of radiotracer injection and the end of the seizure was required to avoid postictal images. Interictal injections were performed after a seizure-free period of ≥24 h. Scans were obtained immediately after injection using a triple-head gamma camera with low energy ultra high-resolution parallel collimators (triad XLT 20; Trionix Research Laboratory, Twinsburg, OH, U.S.A.). The used step angle was 3 degrees; 120 views per detector head were acquired for 14 s. Images were reconstructed (e.soft system; Syngo Software Siemens, Medical Systems, Erlangen, Germany) using a 128 × 128 matrix iterative ordered subsets-expectation maximization (OS-EM) reconstruction (using OS-EM 3D [three-dimensional], 6 iterations, 12 subsets) with an 8-mm Gaussian post filter. Attenuation correction was performed using Chang attenuation correction with an attenuation coefficient of 0.15 (with scatter correction) or 0.11 (without scatter correction). Images were reconstructed in six transaxial sections with an isotropic voxel size of 2.54 mm.

SISCOM analysis was performed in the context of noninvasive presurgical evaluation data (Kaiboriboon et al., 2002). Ictal and interictal SPECT images were coregistered using an automatic registration algorithm based on maximal mutual information. In each patient, interictal SPECT images were normalized and subtracted from the ictal ones and afterwards smoothed using a 3D-Gaussian smoothing kernel. The results were transformed into a Z score using the mean and the standard deviation (SD) of the differences in all brain voxels. The mean image was then coregistered to the preoperative MRI to evaluate the anatomic localization. For clinical interpretation, a Z score of 2 was used. The criteria used to determine localization included SISCOM image intensity, location of hyperperfusion, and number of regions of hyperperfusion. Taking into account these criteria, the most likely cluster in the context of the presurgical evaluation was chosen.

Postoperative seizure outcome

Epilepsy surgery outcome was assessed for a period of ≥24 months as part of the regular epilepsy program follow-up. For practical purposes we classified the postoperative outcome into two groups according to Engel’s classification (Engel et al., 1993): class I (Engel’s class Ia, completely seizure-free); class 2 (Engel’s class Ib–IV, not seizure-free).

Analysis and statistics

All test results were subclassified as “sublobar,”“lobar,”“multilobar,” or “nonlocalizing.” Sublobar regions were defined as follows: frontopolar, frontal superior/mesial, frontal inferior, parietal superior/mesial, parietal inferior, occipital mesial, occipital lateral, temporomesial, temporolateral, and temporopolar. Lobar concordance was defined as localization in either the same anatomic lobe, or within the perirolandic (central lobe), opercular, or insular region.

We performed intertest agreement and agreement of single tests as well as test combinations with site of epilepsy surgery resection (SR), defined by postoperative MRI. All cases were classified with respect to concordance as shown in Table 1.

Table 1.   Intertest agreement (MSI and ICEEG) and agreement of single tests and test combinations with site of epilepsy surgery resection (SR), defined by postoperative MRI
ClassificationIntertest agreement or single-test agreement with SRClassificationTwo-test agreement with SRClassificationThree-test agreement with SR
  1. aOverlapping but one of both tests is more widespread.

  2. bNo epileptic discharges, contralateral results, or one or both tests nonlocalizing.

ASublobar concordance1Sublobar or lobar test concordance, matching with SRISublobar test concordance, matching with SR
BConcordance plusa2Sublobar or lobar on test 1 and matching SR, test 2 overlappingIIConclusively lobar test concordance, matching with SR
CNot overlapping but localized ipsilateral3Sublobar or lobar on test 2 and matching SR, test 1 overlappingIIINonlobar concordance, nonlocalizing test results, or not matching with SR
DNot overlapping and not localizingb4At least one test not overlapping with SRIV

Results were considered as concordant if the MSI- and/or SISCOM-defined region indicated the same sublobar or lobar region as ICEEG (Knowlton et al., 2008a; Seo et al., 2011) (Fig. 1). We considered results as congruent with the SR, if the resected area determined by postoperative MRI completely contained the focus assessed in the preoperative MSI, SISCOM, and/or ICEEG. The concordance groups were then reviewed to assess their correlation with surgical outcome. Sensitivity, specificity, predictive values, and odds ratios (ORs) for each test and test combination to predict epilepsy surgery outcome were calculated. SPSS 20.0 (IBM Co., Armonk, NY, U.S.A.) was used for statistical processing of the data. The statistical methods were descriptive statistics with frequency analysis and cross tab analysis, as well as mean and SD calculation in parametric data. Statistical significance was assessed using Spearman correlation (SC) and Fisher’s exact test, with a significance defined as a probability (p) value of ≤0.05.

Figure 1.

Results of SISCOM and MEG with all of the identified dipoles (yellow) superimposed on each patient’s individual MRI. (A) Sublobar classified MSI and SISCOM, concordance; outcome class 1 (Patient 1). (B) Multilobar classified MSI and SISCOM, nonconcordance; outcome class 2 (Patient 9).


Of the 54 surgical candidates who had negative MRI and underwent MEG and SPECT, we assessed the percentage of unsuccessful studies in each modality (Table 2). Ten patients (18.5%) had MEG results that were not clinically helpful or conclusive, including eight normal studies (i.e., no epileptiform discharges detected), one study with only one sharp transient, and one study with epileptiform-appearing discharges whose clinical relevance remained unclear. In the same group of 54 patients, 11 (20.4%) had inconclusive SPECT, including four studies in which ictal isotope injection was attempted but not achieved, one study with no seizure (i.e., no ictal SPECT), five studies where no dominant foci could be determined despite early injection, and two studies of late injection that confounded the interpretation of hyperperfusion.

Table 2.   Patient disposition flow chart showing the numbers of MRI-negative surgical candidates with positive or negative MEG/SPECT tests
  1. +: localizing (positive) test result; −: inconclusive (negative) test result.

54 surgical candidates with negative MRI 
 MEG+ and SPECT+
  Surgery19 (14/19 included due to follow-up)
  No surgery16
 MEG− and SPECT+
  No surgery6
 MEG+ and SPECT−
  No surgery7
 MEG− and SPECT−
  No surgery1

Details of all 14 patients that fulfilled our inclusion and exclusion criteria are given in Table 3.

Table 3.   The clinical data of all patients included in the study
Pat. no.AgeSexHandED (Y)Invasive procedure (n elec/n cont)ICEEG onsetEF classMSI focusMSI classSISCOM focusSISCOM classInj. time (s)EpilepsySurgeryICEEG/MSI conc.ICEEG/MSI/SR conc.ICEEG/SISCOM conc.ICEEG/SISCOM/SR conc.ICEEG/MSI/SISCOM/SR conc.PathologyOutcome
  1. ED (y), epilepsy duration (years); EF, ICEEG focus; SR, site of surgical resection; Conc, concordance; M, male; F, female; Invasive procedure (n electrodes/n contacts); SEEG, stereotactic EEG; SDG, subdural grid electrode; DE, depth electrode; S, sublobar; L, lobar; M, multilobar; f, frontal; t, temporal; p, parietal; o, occipital; tm, temporomesial; tl, temporolateral; aTL, anterior temporal lobe; sup, superior; inf, inferior; operc, opercular; SFG, superior frontal gyrus; MFG, medial frontal gyrus; IFG, inferior frontal gyrus; STG, superior temporal gyrus; post, posterior; L, left; R, right; FLE, frontal lobe epilepsy; nTLE, neocortical temporal lobe epilepsy; PLE, parietal lobe epilepsy; FCD, focal cortical dysplasia (Palmini classification); outcome (24-months follow-up): 1: seizure-free (class 1); 2: not seizure-free (class 2); for concordance classification (A–D and I–IV) see Table 1.

142MR31SDG 5/124
DE 4/28
243MR25SDG 7/164
DE 7/62
L tlSL t – pML tlS10nTLEL tl and polarB2A11FCD IA1
353FR37SEEG 16/156R f (operc, insula)SR f (operc, insula)SR f (operc, insula)S18FLER f (operc, insula)A1A11FCD IA1
440MR10SEEG 14/152R p + L tmML > R t-p-o; R tlMR tl, R p (inf)M16PLER sup p lobuleD4C43Gliosis1
526MR11SDG 6/130
DE 3/24
R f (operc, insula)SR f (operc, insula)SR t polar; R ant insula; R f basalM27FLER f (operc, insula)B2A12Gliosis2
622MR8SDG 7/144
DE 2/16
R t basalSR t basal; R SFGMR tl S 17nTLER aTL + tmB2C21Gliosis2
726MR18SEEG 15/158L p supSL t-p (inf)-oML tlS19PLEL sup p lobuleC4C43FCD IA2
811MR9SDG 4/96
DE 6/50
L p (inf)SL p (inf)SBilat fM30PLEL post insula; p opercA1C44FCD IA1
914FA5SEEG 12/128L perirolandicLL MFG; precentralSL t basalS120PLEL p (postcentral)C4C43Gliosis2
1026MR2SEEG 9/82R t basalSR t basalSR SFG; R ant p; R IFGM10nTLER aTL + tm + o basalA1A13FCD IA2
1110FR2SDG 10/212
DE 2/16
1245FR20SEEG 13/142L SFGSR > L p; L t; R tML o mesial; L SFGM7FLEL SFGD4D22FCD IA2
1314MR2SEEG 15/152R t and p opercLR tl > L t postMR STGS14nTLER STG, p operc, post insulaB2B31Gliosis2
1430ML18SEEG 11/100R f opercSR f opercSR f opercS19FLER f opercA1A11FCD IIB1

The mean age of all included patients was 28.7 years (SD ± 13.9, range 10–53), the mean duration of epilepsy 14.2 years (SD ± 11.1, range 2–37). Six patients (42.9%) became seizure-free. Pathologic findings (Table 3) included gliosis and focal cortical dysplasia (Palmini et al., 2004). However, neither epilepsy category (SC: p = 0.796) nor pathologic findings (SC: p = 0.228) were significantly correlated with the epilepsy surgery outcome.

Intracranial EEG

ICEEG focus was classified as sublobar in 11 patients (78.6%) and completely resected in all of them (Table 3). Five patients (45.5%) became seizure-free following surgery. The remaining three patients (21.4%) had lobar (n = 2, not seizure-free) and multilobar (n = 1, seizure-free) ICEEG. Under consideration of the small sample size, there was no correlation between sublobar ICEEG results and their complete resection with the epilepsy surgery outcome (SC: p = 0.865).

Magnetic source imaging

MSI focus was classified as sublobar in eight patients (57.1%). Four patients (50%) became seizure-free; in all of them the MSI focus was completely resected. Seizure-free rate in patients with complete sublobar MSI focus resection was 66.7%, compared to 25% in patients not matching these criteria. Based on these numbers, no significant correlation was found between seizure-free outcome and sublobar MSI results/complete MSI focus resection (SC: p = 0.569).


The mean radiotracer injection time was 24.1 s (range 7–120 s) after electroencephalographic seizure onset; there was no postictal SPECT study among our patients. All SPECT studies were performed during seizures that are representative of the patient’s epilepsy (i.e., patient’s typical seizure). SISCOM focus was classified as sublobar in eight patients (57.1%) and completely resected in six of them; four of the eight patients (50%) became seizure-free. If sublobar SISCOM focus was completely resected, 66.7% became seizure-free, compared to 25% seizure-free rate in patients not matching these criteria. Neither sublobar SISCOM results (SC: p = 0.569) nor complete SISCOM focus resection (SC: p = 0.138) was significantly correlated with the surgical outcome.

MSI and ICEEG coanalysis

Five patients (35.7%) showed sublobar concordant ICEEG and MSI results (Table 3). In all of them the indicated focus was completely resected; four of them (80%) became seizure-free. In cases not matching these criteria, only 22.2% (two of nine patients) became seizure-free. Sublobar concordance of ICEEG and MSI, which leads to complete resection of both foci, significantly increases the chance of seizure freedom after epilepsy surgery (SC: p = 0.038; cross tab analysis of complete or incomplete/failed focus resection and surgical outcome).

SISCOM and ICEEG co-analysis

Six patients (42.9%) had concordant ICEEG and SISCOM results; the indicated focus was completely resected in all of them; four patients (66.7%) became seizure-free. Only two of eight patients (25%) without concordant ICEEG and SISCOM results became seizure-free. No correlation was found between concordant and completely resected ICEEG and SISCOM results and a seizure-free epilepsy surgery outcome (SC: p = 0.138; cross tab analysis of complete or incomplete/failed focus resection and surgical outcome).

MSI, SISCOM, and ICEEG co-analysis

Sublobar concordance of ICEEG, MSI, and SISCOM was found in three patients (21.4%); in all of them the indicated focus was completely resected, which led to seizure freedom in all. In cases not matching these criteria only 3 of 11 patients (27.3%) became seizure-free. Sublobar three-test concordance and complete focus resection significantly increases the chance of seizure freedom after epilepsy surgery (SC: p = 0.022).

ICEEG had the greatest proportion of sublobar test results (78.6%), followed by MSI (57.1%) and SISCOM (57.1%). Sublobar ICEEG-MSI concordance was seen mostly in patients with frontal lobe epilepsy (FLE) and patients with neocortical TLE (nTLE) (each n = 2), followed by patients with parietal lobe epilepsy (PLE) (n = 1). Concordant ICEEG and SISCOM results were most frequent in nTLE (n = 4) followed by FLE (n = 2).

Diagnostic values

Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and OR for each diagnostic test alone and in combination as measured against epilepsy surgery outcome (seizure freedom) is listed in Table 4.

Table 4.   Diagnostic measures for EZ localization based on the epilepsy surgery outcome
TestSens (95% CI)Spec (95% CI)PPV (95% CI)NPV (95% CI)OR (95% CI)p-Valueap-Valueb
  1. Sens, sensitivity; Spec, specificity; PPV, positive predictive value; NPV, negative predictive value; OR, odds ratio; CI, confidence interval; EZ, epileptogenic zone.

  2. aSpearman correlation.

  3. bFisher's exact test.

ICEEG0.883 (0.453–1.214)0.25 (−0.133–0.633)0.455 (0.079–0.83)0.667 (−0.014–1.347)  1.67 (0.115–24.256)0.8650.46
MSI0.667 (0.186–1.148)0.5 (0.058–0.942)0.5 (0.058–0.942)0.667 (0.186–1.148)2 (0.224–17.894)0.5690.35
SISCOM0.667 (0.186–1.148)0.5 (0.058–0.942)0.5 (0.058–0.942)0.667 (0.186–1.148)2 (0.224–17.894)0.5690.35
ICEEG + MSI0.667 (0.186–1.148)0.875 (0.583–1.167)0.8 (0.353–1.247)0.778 (0.431–1.124)14 (0.944–207.598)0.0380.06
ICEEG + SISCOM0.667 (0.186–1.148)0.75 (0.367–1.133)0.667 (0.186–1.148)0.75 (0.367–1.133)6 (0.582–61.842)0.1380.14
ICEEG + MSI + SISCOM0.5 (−0.01–1.01)110.727 (0.392–1.063)  00.0220.05

With regard to single test results, a sublobar ICEEG result has the highest sensitivity (88.3%); specificity was higher in MSI (50%) and SISCOM (50%). Sublobar concordance of ICEEG and MSI has the highest specificity (87.5%), PPV (80%), and NPV (77.8%) compared to both concordant ICEEG-SISCOM results and any single test alone. Although only marginally significant, sublobar concordance of ICEEG and MSI has the highest OR for seizure freedom after surgery (OR 14, 95% confidence interval [CI] 0.944–207.598, Fisher’s exact test: p = 0.05994). Concordance of ICEEG and SISCOM shows inferior results (OR 6, 95% [CI] 0.582–61.842, Fisher exact test: p = 0.13986) but is still superior to any single test alone. Notwithstanding the small number of patients, three-test concordance (ICEEG, MSI, and SISCOM) has the highest specificity (100%) and PPV (100%).

The proportion of test concordance with sublobar ICEEG was greater for MSI compared to SISCOM (MSI: κ 0.432, 95% CI −0.224–1.089; SISCOM: κ 0.222, 95% CI −0.369–0.813).


The goal of this retrospective study was to assess and compare the additive diagnostic value of both MSI and SISCOM in patients with MRI-negative neocortical epilepsy who require ICEEG evaluation for ultimate localization of their epilepsy. We also investigated if a multimodal approach contributes to better localize the EZ and predict surgical outcome. Although several recent studies on MSI, SPECT, and ICEEG in patients with focal epilepsy have been published (Knowlton et al., 2006, 2008a,b, 2009), no prior study specifically focused on nonlesional neocortical epilepsies. To our knowledge, this is the largest study to examine and compare the utility of both diagnostic modalities compared to ICEEG, the current “gold standard” in presurgical evaluation, in this most challenging group of patients.

The most important finding from our study is that sublobar concordance with ICEEG of either MSI or SISCOM is superior to ICEEG alone in localizing the EZ. When sublobar concordant with ICEEG, MSI has the highest PPV and OR for accurate localization of the EZ based on seizure-free outcome; complete resection of both foci correlates with a favorable surgical outcome (SC: p = 0.038). Assuming concordant results, the additive value to ICEEG alone is higher for ICEEG-MSI (OR 14) compared to ICEEG-SISCOM (OR 6) for localizing the EZ. Several prior studies reported complete ICEEG focus resection to be the most important predictor of a favorable surgical outcome (Stefan et al., 2000; Ossenblok et al., 2007; Kim et al., 2010). In contrast, our results show that, assuming complete focus resection, test combinations of ICEEG-MSI and, to a lesser degree, ICEEG-SISCOM are superior to ICEEG alone. When sublobar concordant with ICEEG, notably MSI, but also SISCOM provides useful additive information to that provided by ICEEG alone for localizing the EZ and predicting the epilepsy surgery outcome, and, therefore, can potentially increase the clinician’s ability and confidence to recommend localized resection based on subsequent ICEEG results. One should note that only nonlesional patients who underwent epilepsy surgery were studied; therefore, a potential selection bias toward patients with a more localized focus has to be regarded, confirmed by a relatively high percentage of patients with a sublobar ICEEG, and consequently a presumably more favorable epilepsy surgery outcome. This potential selection bias has to be considered in the interpretation of our results.

Test concordance in presurgical evaluation is essential for predicting the epilepsy surgery outcome (Lee et al., 2005). Lower concordance rates of MSI, SISCOM, and ICEEG in neocortical epilepsies compared to mesial TLE cases were reported in several prior studies to have lower sensitivities (Won et al., 1999; Hwang et al., 2001; Meyer et al., 2001). Other authors reported a high proportion of concordance with ICEEG of both MSI and SPECT in partial epilepsies including nonlesional neocortical cases (Stefan et al., 2000; Pataraia et al., 2002; Baumgartner, 2004; Knowlton et al., 2008a,b). A recent study, which did not specifically focus on nonlesional neocortical epilepsies, showed that MSI had the highest concordance rate with ICEEG compared to SPECT and positron emission tomography (PET) (Knowlton et al., 2008a). Although κ agreement results are relatively low, our results still show a slightly better κ score of concordance with ICEEG for MSI (κ = 0.432, moderate agreement) compared to SISCOM (κ = 0.222, fair agreement). In a previous study, SPECT had been shown to be able to localize the EZ, even when EEG and MRI have been nonlocalizing (Hong et al., 2002). In our study, the mean radiotracer injection delay after seizure onset was 24.1 s; consequently, fast seizure propagation may be one reason for the lower κ scores for SISCOM agreement with ICEEG in our study. Confirming our results, Seo et al. (2011) also reported a slightly better concordance of ICEEG-MSI compared to ICEEG-SISCOM in a cohort of children with nonlesional epilepsy. The presence of concordance between different diagnostic modalities may indicate convergence of different underlying physiologic mechanisms of the localized lesion, which conclusively increases the possibility that the localized lesion is an EZ (Zentner et al., 1996; Knowlton, 2006).

The importance of resecting the MSI/SISCOM-defined region to obtain seizure freedom has been described previously (O’Brien et al., 2000; Otsubo et al., 2001; Fischer et al., 2005; RamachandranNair et al., 2007; Knowlton et al., 2009). Tightly clustered spike dipoles on MEG have been shown to be associated with a favorable epilepsy surgery outcome (Oishi et al., 2006). Conclusively, sublobar MSI results may reflect the presence of a tight cluster of spike dipoles. Consistent with our findings that sublobar ICEEG-MSI concordance and complete focus resection correlates with a seizure-free surgical outcome (SC: p = 0.038), two recent studies have reported that concordance of ICEEG with MEG results increases the predictive value for a seizure-free surgical outcome in patients with nonlesional neocortical epilepsy (Zhang et al., 2011; Schneider et al., 2012). With regard to SISCOM, our results indicate that, although not significant, the presence of positive SISCOM results concordant with ICEEG and complete resection of both foci may have prognostic implications, forecasting a more favorable epilepsy surgery outcome (seizure-free rate 66.7%, p = 0.138). Taken together, in patients with nonlesional neocortical epilepsy both positive MSI and SISCOM studies may indicate a higher chance of a localized ICEEG result. Therefore, both diagnostic modalities provide additive and not redundant localizing information for that provided by ICEEG alone, even if ICEEG is localizing.

Our screening procedure enabled us to examine the percentage of “unsuccessful” studies in MSI and SPECT in a nonlesional population. MSI has long been accused of primarily localizing interictal events, that is, due to the relatively short recording time in the MEG suite, one may get only a snapshot of the brain activities in the recorded time window. This issue leads to a certain number of normal MEG studies (18.5% in our cohort). One should be equally aware of the fact that unsuccessful ictal SPECT can also happen, and in our same cohort it is 20.4%, a higher percentage than negative MEG. Unsuccessful ictal SPECT is frequently due to clinical and logistical practicality of this test, for example, late injection leading to incomprehensible hyperperfusion patterns, failure of injection, or, in rare cases, no seizures captured.

With regard to the diagnostic values of each modality alone, we did not observe significant differences between ICEEG, MSI, and SISCOM. Although limited to nonlesional neocortical epilepsies, our findings of sensitivity and PPV are highly consistent with those of previously reported larger series on nonlesional epilepsies (Ahnlide et al., 2007; Knowlton et al., 2008a,b). ICEEG has the highest sensitivity for localizing the EZ based on epilepsy surgery outcome; specificity and PPV are higher with MSI and SISCOM. These differences, however, are complementary from a diagnostic standpoint, as PPV is increased in both ICEEG-MSI and ICEEG-SISCOM combinations. Compared to the PPV, NPV is relatively high in all single tests and test combinations, indicating that a negative test result might be associated with a higher chance of nonlocalizing results in the other tests and a poor epilepsy surgery outcome, respectively. Compared to ICEEG alone, test combinations increase specificity; however, specificity was highest in ICEEG-MSI (87.5%) compared to ICEEG-SISCOM (75%). One excellent, recent, prospective study has reported that diagnostic tests that are used to discriminate among surgical candidates regarding who will and will not be localized with ICEEG and benefit from surgery, should trend toward high specificity more than sensitivity (Knowlton et al., 2008a). Our findings suggest that, in addition to ICEEG as the current “gold standard,” primarily MSI but also SISCOM increases the accuracy of ICEEG to localize the EZ and, therefore, provide clinicians with valuable additional information to predict the epilepsy surgery outcome. Three-test concordance has the highest specificity for localizing the EZ and is significantly associated with a seizure-free outcome (SC: p = 0.022). Taken together, our results provide statistical validation that the combination of electrophysiologic (invasive and noninvasive) and imaging (structural and functional) modalities provides clinicians with valuable additional information compared to ICEEG alone, and, therefore, can lead to better hypotheses about the putative EZ in this challenging group of patients. Multimodal approach, therefore, has to be considered as an essential part of the regular protocol in presurgical evaluation of patients with nonlesional neocortical epilepsy considered for epilepsy surgery. Nevertheless, the number of our patients showing sublobar MSI and SISCOM results, however, is relatively small, yielding high specificity and relatively low sensitivity. This has to be considered when interpreting our results.

Prior studies reported that, although considered the “gold standard” to map the ictal onset zone, ICEEG still has the weakness of limited spatial sampling, especially in nonlesional neocortical cases (Knowlton et al., 2008a,b; Seo et al., 2011). One should note that our study did not focus on the question of whether any of the tests, alone or in combination, is able to replace ICEEG. In fact, several prior studies reported that preimplantation MSI and SISCOM results increase the chance that the seizure onset zone is sampled, when patients undergo ICEEG (Knowlton et al., 2008a, 2009; Sutherling et al., 2008; Seo et al., 2011). Preimplantation MSI and SISCOM can be used to guide ICEEG electrode placement (Lee et al., 2003, 2006; Sutherling et al., 2008; Knowlton et al., 2009), which was performed in all of our patients. Conversely, when ICEEG electrodes are not placed at the proper locations, ictal and interictal events may represent propagated activity, but not the true seizure onset zone. In contrast, both MEG and ictal SPECT usually provide a global view of ictal and interictal sources. It is likely that utility of MEG and SISCOM in our patients may have influenced planning of electrode placement. Therefore, our ICEEG results cannot be regarded as entirely independent of both MSI and SISCOM, a bias that has to be considered as a limitation of this study.

Several limitations of our study need to be considered. First, this is a single-center study with a relatively small group of 14 patients, and therefore the possibility of type 2 statistical errors that limit statistical validity should be considered, particularly with regard to further subgroup analyzes. To compare our MSI and SPECT results with those of large future study groups of nonlesional patients considered for epilepsy surgery and especially of those with a favorable seizure outcome, would likely strengthen the validity of our findings and should be the object of future studies. Second, this study compares diagnostic modalities with different underlying mechanisms: ICEEG and MEG are electrophysiologic investigations, whereas SPECT as an imaging modality aims to map brain perfusion changes during an actual seizure (Knowlton, 2006). EZ localization in our study was based on ictal ICEEG patterns (ictal onset zone), interictal dipoles in MEG (irritative zone), and blood flow changes ideally during a seizure. This limits the comparability of the diagnostic modalities. Particularly for SPECT, both late isotope injection and an unswayable delay of 30–60 s (Cho et al., 2010) until the radiotracer reaches the cerebral circuit increase the risk of showing propagated epileptic activities, which may lead to erroneous SISCOM interpretation. In our study 11 patients (78.6%) had a radiotracer injection time of <20 s (mean 24.1 s) after seizure onset; however, in one patient the injection time was 120 s. Although classified as sublobar, SISCOM may show propagated activity. In future studies injection time should therefore be taken into consideration as an inclusion criterion. Third, our concordance grading system is qualitative, not quantitative. Therefore, the sensibility of this grading system to evaluate concordance among MEG, SISCOM, ICEEG, and SR may be reduced. Finally, one has to note that all three modalities were analyzed differently. MEG analysis was performed without prior knowledge of the clinical history; SPECT was performed in the context of other noninvasive presurgical evaluation data; ICEEG implantation and analysis could both be influenced by MEG and SPECT, as well as other presurgical data. This limitation is inevitable due to the retrospective design of this single center study, which may be overcome with future properly designed prospective studies specifically looking at nonlesional patients.

Despite the above limitations, this study shows that test combinations of MSI and/or SISCOM with ICEEG are useful in the presurgical evaluation of patients with nonlesional neocortical epilepsy. Concordant test results of either MSI or SISCOM with ICEEG provide useful additive information for that provided by ICEEG alone to localize the EZ and to determine the site of surgical resection in this very specific group of patients. Despite the relatively small sample size, our results suggest that if both MSI and SISCOM are concordant with ICEEG on a sublobar level, MSI appears to be advantageous compared to SISCOM in predicting seizure-free surgical outcome. Our results strongly suggest that, in cases where ICEEG is localizing, a multimodal approach should still be part of the regular protocol in presurgical evaluation. Our future goals include comparison of MSI/SISCOM/ICEEG results in a larger group of nonlesional patients considered for epilepsy surgery. In addition, prospective studies will need to be undertaken in this challenging patient population to test the replicability of our observations.


We gratefully acknowledge the assistance of Professor Dr. med. Dr. h.c. C. Kessler, Professor Dr. med. U. Runge, and Dr. rer. nat. P. Kolyschkow, University of Greifswald, Germany.


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.


This study received no industrial, governmental, or institutional funding or sponsorship.