Abnormal functional brain network in epilepsy patients with focal cortical dysplasia

Epilepsy Research (2014) 108, 1618—1626

journal homepage: www.elsevier.com/locate/epilepsyres

Abnormal functional brain network in epilepsy patients with focal cortical dysplasia Woorim Jeong a,b,1, Seung-Hyun Jin a,c,1, Museong Kim a, June Sic Kim a,d, Chun Kee Chung a,b,c,e,∗ a

MEG Center, Department of Neurosurgery, Seoul National University Hospital, Seoul, South Korea Interdisciplinary Program in Neuroscience, Seoul National University College of Natural Science, Seoul, South Korea c Neuroscience Research Institute, Seoul National University Medical Research Center, Seoul, South Korea d Research Center for Sensory Organs, Seoul National University, Seoul, South Korea e Department of Brain and Cognitive Sciences, Seoul National University College of Natural Sciences, Seoul, South Korea b

Received 14 May 2014; received in revised form 18 August 2014; accepted 6 September 2014 Available online 16 September 2014

KEYWORDS Epilepsy; Functional connectivity; Resting state; Focal cortical dysplasia; Magnetoencephalography

Summary Purpose: Focal cortical dysplasia (FCD) is the second most common pathological entity in surgically treated neocortical focal epilepsy. Despite the recent increase of interest in network approaches derived from graph theory on epilepsy, resting state network analysis of the FCD brain has not been adequately investigated. In this study, we investigated the difference in the resting state functional network between epilepsy patients with FCD and healthy controls using whole-brain magnetoencephalography (MEG) recordings. Methods: Global mutual information (MIglob ) and global efficiency (Eglob ) were calculated for theta (4—7 Hz), alpha (8—12 Hz), beta (13—30 Hz), and gamma (31—45 Hz) bands in 35 epilepsy patients with FCD and 23 healthy controls. Results: Resting state FCD brains had stronger functional connectivity (MIglob ) in the beta and gamma bands and higher functional efficiency (Eglob ) in the beta and gamma bands than those of the controls (p < 0.05). The MIglob and Eglob values of FCD type I and II brains in the beta band were higher than those of healthy control brains (p < 0.05). In the gamma band, the values of FCD type II brains were higher than those of control and FCD type I brains (p < 0.05).

∗ Corresponding author at: Department of Neurosurgery, Seoul National University Hospital, 101 Daehak-ro, Jongno-gu, Seoul 110-744, South Korea. Tel.: +82 2 2072 2352; fax: +82 2 2072 0806. E-mail addresses: [email protected] (W. Jeong), [email protected] (S.-H. Jin), [email protected] (M. Kim), [email protected] (J.S. Kim), [email protected], [email protected] (C.K. Chung). 1 These authors contributed equally to the manuscript.

http://dx.doi.org/10.1016/j.eplepsyres.2014.09.006 0920-1211/© 2014 Elsevier B.V. All rights reserved.

Abnormal functional brain network in focal cortical dysplasia

1619

Conclusions: FCD brains had increased functional connectivity in the beta and gamma frequency bands at the resting state compared with those in healthy controls. In addition, patients exhibited different network characteristics depending on the type of FCD. The resting state network analysis could be useful in a clinical setting because we observed network differences even when there was no prominent interictal spike activity. © 2014 Elsevier B.V. All rights reserved.

Introduction Focal cortical dysplasia (FCD) is the second most common pathological entity in surgically treated neocortical focal epilepsy (Chung et al., 2005). Conventional clinical electrophysiological examination is insufficient to unravel the characteristics of FCD that are associated with less favorable surgical outcome when compared with those of epilepsy associated with hippocampal sclerosis or tumor. Recently, functional connectivity and further network analyses have been used to reveal the intrinsic properties of the epileptic network (Engel et al., 2013; Horstmann et al., 2010; Liao et al., 2010; Morgan and Soltesz, 2008; Varotto et al., 2012; Wilke et al., 2011). Studies of focal epilepsies have reported enhanced connectivity in the region of the ictal onset zone, thereby revealing the existence of highly interconnected nodes that may play a crucial role in the onset and propagation of ictal activity (Morgan and Soltesz, 2008; Wilke et al., 2011). However, instead of including the entire brain, previous approaches covered only part of the whole brain, and such approaches may not grasp the full complexity of the brain as a network. Several previous studies utilized functional magnetic resonance imaging (fMRI) in network analyses of medial temporal lobe epilepsy (Liao et al., 2010; Pereira et al., 2010). However, with fMRI modality, it is difficult to investigate the dynamics of frequency bands above 0.1 Hz, which may play an important role in epileptogenesis. In this respect, electrophysiological studies, such as electrocorticography (ECoG), electroencephalography (EEG), and magnetoencephalography (MEG), have advantages over fMRI. Recently, one study using ECoG investigated network characteristics in type II FCD patients (Varotto et al., 2012). Those authors reported that the interictal functional connectivity was higher in the seizure onset zone than in the non-involved zone in the gamma band. However, ECoG covers only part of the whole brain; moreover, it was not possible to study healthy controls in that study due to the method’s invasiveness. Although one network study covered the whole brain with various frequency bands, that study compared only 5 healthy controls and 5 non-focal, absence seizure patients (Chavez et al., 2010). To the best of our knowledge, there are no reports concerning whole-brain resting state network studies that compare focal epilepsy patients with FCD and healthy controls in various frequency bands. Unraveling such differences between FCD and healthy brains would enhance our knowledge of epileptogenesis in FCD. In this study, we investigated differences in the resting state functional network between patients with FCD and healthy controls by using whole-brain MEG signals. Global mutual information (MIglob ) as a measure of functional connectivity and global efficiency (Eglob )

as a measure of network efficiency were used to compare the global resting state functional network between the two groups.

Materials and methods Patients with FCD Initially included in this study were 64 patients with intractable epilepsy and histologically proven FCD who underwent MEG examination before surgery between 2005 and 2011 at Seoul National University Hospital. Patients were screened for additional exclusion criteria: younger than 18 years of age at the time of surgery, FCD type III, which is associated with other pathologies, MEG recording after their first surgery for epilepsy, and a post-surgery follow-up period of less than two years. As a result, 35 patients (mean age at surgery ± SD = 30.0 ± 8.6 years; 18 males) were included in the study. Fig. 1 shows the flow chart of patient selection. Surgical outcome was classified according to the epilepsy surgery outcome classification system of the International League Against Epilepsy (Wieser et al., 2001).

Control subjects Twenty-three healthy subjects (mean age ± SD = 24.7 ± 4.1 years; 12 males) were recruited as controls. There were no statistically significant differences in the gender mix between the patient and healthy subject groups (p = 0.956). Because there were a significant difference in ages between patients and healthy subjects (p = 0.003), we used age as a covariate in the statistical analysis. They had no history of neurologic or psychiatric disorders. The control subjects’ MEG data were included in a previous study (Jin et al., 2011). All patients and healthy participants provided written informed consent. The study was approved by the institutional review board of Seoul National University Hospital (IRB H-0607-029-178).

MEG Spontaneous magnetic activities were recorded inside a magnetically shielded room by using a 306-channel, whole-head MEG system (version 2.2, VectorViewTM , Elekta Neuromag Oy, Helsinki, Finland). The MEG sensors were arranged at 102 locations at which triplets consisting of two orthogonal planar gradiometers and one magnetometer were included. Data were collected for approximately 60 min (four 15 min sessions) from patients with their eyes closed. For healthy participants, MEG signals were recorded

1620

W. Jeong et al.

Figure 1

Flow chart of patient selection. The patients included in the study are shown in the gray box.

in the eyes-closed and eyes-open resting states for 2 min each, as described in our previous study (Jin et al., 2011). In the present study, data obtained from the eyes-closed resting states were used. Patients and healthy participants were instructed to stay relaxed and asked not to think about anything while in a supine position. Because the epochs from MEG recordings of patients and healthy participants were selected from the raw data recorded in the same supine position with the same eyes-closed state under the same instructions, the epochs selected from patients were comparable enough to those from healthy participants. Data were collected at a 600 Hz sampling rate with an analog 0.1—200 Hz filter. A bipolar electro-oculogram and an electrocardiogram were simultaneously recorded to monitor eye movement and cardiac artifacts. The temporal signal space separation algorithm (Taulu and Hari, 2009) implemented in the Maxfilter Software (Elekta Neuromag Oy, Helsinki, Finland) was applied to reduce environmental and biological noise before undertaking further analysis. Five epochs of 10 s of data were selected for connectivity analysis by visual inspection. Specifically, each epoch was selected from a spike-free duration in patients to maintain the same resting state as healthy participants for further analysis. Five epochs of 10 s of spike-free data were selected for connectivity analysis by visual inspection. In the present study, MEG signals used for the analysis refer to those acquired from 102 magnetometer sensors.

Magnetic resonance imaging All patients were examined using either a GE 1.5 T or 3 T MRI system (GE Horizon Echospeed; GE Healthcare, Little Chalfont, UK) or a Siemens 1.5 T scanner (MAGNETOM Avanto; Siemens, Erlangen, Germany). Our standard MRI protocol included T2 and fluid attenuated inversion recovery (FLAIR) axial, T2 and FLAIR oblique coronal, fast inversion recovery with myelin suppression (FIRMS), and three-dimensional (3D) gradient echo coronal T1 images with whole brain coverage. The 3D gradient echo T1 images were reconstructed

to a slice thickness of 1 mm, while the T2 images were acquired by using a 3-mm thickness with a 1-mm inter-slice gap. Preoperative MRI results were reviewed separately and confirmed by two neuroradiologists specializing in epilepsy and blinded to seizure focus.

Surgical outcome and lobar distribution Surgical procedures have been described in a previous report (Chung et al., 2005). We evaluated surgical outcome after a follow-up period of at least 1 year. Surgical outcome was classified according to the ILAE outcome classification (Wieser et al., 2001). The seizure onset zone was determined on the basis of results of presurgical evaluations including ictal scalp VEEG, MEG, PET, ictal and interictal SPECT, intracranial EEG, and clinical semiology. For the purpose of analysis, we classified seizure onset zone into two groups: temporal lobe and extratemporal lobe.

Estimation of functional connectivity and network efficiency Mutual information (MI), which quantifies the shared information between two time series based on information theory, was calculated to obtain the functional connectivity matrix for four frequency bands corresponding to the classical EEG bands: theta (4—7 Hz), alpha (8—12 Hz), beta (13—30 Hz), and gamma (31—45 Hz). MI values of bandpassfiltered time series have been used to measure functional connectivity strengths between MEG sensors in previous studies (Bassett et al., 2009; David et al., 2004; Jin et al., 2011). MI was calculated using the following equation: MI

= MIXY = MIYX = MI(X(t), Y (t)) =−



p(X(t), Y (t)) log

X(t),Y (t)

p(X(t), Y (t)) p(X(t))p(Y (t))

Abnormal functional brain network in focal cortical dysplasia where X(t) and Y(t) denote the two bandpass filtered time series obtained from pairs of 102 sensors, and p(X(t),Y(t)) is the joint probability density function (PDF) between X(t) and Y(t). Following the approach used in our previous studies (Jin et al., 2011, 2013), 32 bins were selected for application to the 4096 samples used to construct the PDF. Weighted graphs can indicate the strength of connections (Reijneveld et al., 2007). According to previous studies (Barrat et al., 2004; Barthelemy et al., 2005; Newman, 2004; Onnela et al., 2005; Park et al., 2004; Stam et al., 2009), weighted graphs can be more accurate models of real networks. Weighted graph analysis avoids an issue related to the selection of an appropriate threshold (van Wijk et al., 2010). From these bases, we used a weighted MI. Pairs of nodes closer than 40 mm (Garcia Dominguez et al., 2005) were excluded in the present study, as in our previous work (Jin et al., 2013), to minimize spurious correlations between MEG signals (Chavez et al., 2010; de Pasquale et al., 2012). After the MI matrix calculation for each epoch, the five MI matrices were averaged at each frequency band and normalized by the band’s maximum value. To yield the global MI value per subject, the normalized MI was averaged across all possible sensors at each frequency band and denoted as MIglob . It should be emphasized that because MIglob is the sum of all weights in the network, which is equivalent to the total weighted degree, MIglob is regarded as a measure of the strength of functional connectivity indicating how strongly the network nodes are connected within the network. Note that the group averaged MI values at each node and each frequency band were calculated to create a topographical map of the MI values at each node and denoted as MInodal (Fig. 2). The links between two nodes, i and j, are associated with the connection weights wst . The wst is the normalized connection weight between s and t, equivalent to the normalized MI by the maximum value, 0 ≤ wst ≤ 1 for all nodes (Rubinov and Sporns, 2010). The shortest weighted path length ofthe path from node i to node j was calculated i,j as dw = wst ∈gw f(wst ), where f is an inverse of the weight i→j

to length and gwi→j is the shortest weighted path between two nodes i and j (Rubinov and Sporns, 2010). Global efficiency (Eglob ), one of measures of functional integration, is a measure of parallel information transfer in the whole network and was initially proposed to express the capacity of networks to facilitate information exchange (Latora and Marchiori, 2001). Global efficiency is calculated as an average of nodal efficiency (Enodal ) at each node (Achard and Bullmore, 2007), which represents the communication efficiency at each node in the network (Wang et al., 2009). Enodal was derived from the following equation for the node i:

Enodal (i) =

1  1 w n−1 di,j jεN,j = / i

1621 Brain Connectivity Toolbox released on December, 4, 2012 (http://www.brain-connectivity-toolbox.net/).

Statistical analysis We applied the Shapiro—Wilk test to determine the type of distribution of the variables. When a variable was normally distributed, we used parametric tests such as Student’s t test or analysis of covariance (ANCOVA) with age as a covariate. Otherwise, a Rank ANCOVA was applied. Because the MIglob values of the beta and gamma band were not normally distributed, we used a rank ANCOVA in the statistical analysis for these two values. For categorical variables, we used the 2 test or, where appropriate, Fisher’s exact test. Statistics were calculated using SPSS 19.0 software (IBM, Armonk, NY, USA) with the significance level set at 0.05 and application of a Bonferroni correction for multiple testing to sub-groups. Data are presented as the mean ± SD.

Results Strength of functional connectivity The difference in the MIglob values, as a measure of the strength of functional network connectivity, between FCD patients and controls were determined for each band. In the theta and alpha bands, there were no significant differences between FCD patients and control values. In the beta band, the MIglob value of patients was significantly greater than that of controls (p = 0.001, Fig. 2A). In the gamma band, FCD patients had higher MIglob values than controls (p = 0.021, Fig. 2A). In addition to the MIglob values of FCD patients and controls, the topographic distributions of the group averaged MInodal values at each frequency band are shown in Fig. 2B. The topographic maps provide some sense of how MInodal values were distributed at each frequency band in each group. Because FCD patients have warmer colors in topographic maps of the beta and gamma bands, FCD patients have stronger functional connectivity than those in the controls, which was statistically confirmed as shown in Fig. 2A.

Efficiency of functional network Similar to the MIglob results, there were no significant differences in the theta and alpha bands between FCD patients and controls in Eglob values. In the beta band, the Eglob values for FCD patients were significantly higher than those for controls (p = 0.001, Fig. 2C). Similarly, in the gamma band FCD patients had greater Eglob values than those in the controls (p = 0.003, Fig. 2C). Topographical maps show that FCD patients have higher overall Eglob values than those in the controls in both the beta and gamma bands (Fig. 2D).

Surgical outcome and lobar distribution where N is the set of all nodes, and n is the number of the nodes (here, 102). As mentioned, the Enodal values at each sensor were averaged to estimate Eglob . Now, Eglob is indicative of the global network efficiency. In this study, the analyses were performed by using the

Nine patients were class I, 3 were class II, 5 were class III, 15 were class IV, 1 was class V, and 2 were class VI. Lobar distribution was temporal in 18 patients, frontal in 15, parietal in 1, and occipital in 1 patient. The MIglob and Eglob values

1622

W. Jeong et al.

Figure 2 Group averaged MIglob (A) and Eglob (C) values of control and FCD subjects (*p < 0.05), and topographic maps of the group averaged MInodal (A) and Enodal (B) values at each node and each frequency band: theta (4—7 Hz), alpha (8—12 Hz), beta (13—30 Hz), and gamma (31—45 Hz). In (B) and (D), points on each head model indicated the 102 MEG sensor locations. Note that the same scale bars indicating group averaged MInodal (B) and Enodal (D) values are used for each frequency band, but they are different from the other frequency bands. Nodes with higher group averaged MInodal (B) or Enodal (D) are presented in a warmer color. In the beta and the gamma bands, FCD patients (lower panel in B and D) show a warmer distribution in topographic maps than that in controls (upper panel in B and D) in both MInodal (A) and Enodal (B).

were not significantly different between favorable (n = 17; ILAE 1—3) and unfavorable (n = 18; ILAE 4—6) surgical outcome groups (MIglob : theta, p = 0.646; alpha, p = 0.649; beta, p = 0.765; gamma, p = 0.992; Eglob : theta, p = 0.629; alpha, p = 0.780; beta, p = 0.359; gamma, p = 0.850). In addition, there were no statistically significant differences in MIglob and Eglob values between temporal lobe epilepsy (n = 18) and extratemporal lobe epilepsy (n = 17) groups (MIglob : theta, p = 0.036; alpha, p = 0.636; beta, p = 0.188; gamma, p = 0.111; Eglob : theta, p = 0.283; alpha, p = 0.502; beta, p = 0.060; gamma, p = 0.095).

Pathology and MRI We classified patients’ FCD type according to Blumcke’s criteria (Blumcke et al., 2011). There were 27 patients with FCD type IA, 1 with type IB, and 7 with type IIB. Because there was only one patient classified as type IB, we combined types IA and IB into a single group for statistical analysis. As a result, differences in MIglob and Eglob values were assessed among healthy controls, FCD type I, and FCD type II. With regard to MIglob , no significant differences were found in the theta and alpha bands. The MIglob values in the beta (p = 0.002) and gamma (p = 0.004) bands were significantly different among the three sub-groups. In the beta band, the MIglob values for the FCD type I (p = 0.024) and FCD type II (p = 0.003) patients were higher than that for the controls. Regarding the gamma band MIglob values, that

for the FCD type II patients was higher than those for the controls (p = 0.006) and the FCD type I patients (p = 0.039). Regarding the comparison of Eglob values, no significant differences were found in the theta and alpha bands. However, Eglob values in the beta (p = 0.002) and gamma (p = 0.001) bands were significantly different among subgroups. Post hoc analysis revealed that beta band Eglob values for FCD type I (p = 0.012) and FCD type II (p = 0.006) were higher than that for the controls. The gamma band Eglob value for FCD type II patients was higher than that for the controls (p < 0.001) and the FCD type I patients (p = 0.031). Fig. 3 presents a bar plot of the MIglob and Eglob values at each frequency band within the three sub-groups. Twenty (57%) patients showed no abnormalities in MRI results. The FCD type I patients had a higher likelihood of normal MRI than FCD type II patients (p = 0.027, Fisher’s exact test). Moreover, MIglob and Eglob values were not significantly different between the presence and absence of MRI abnormality groups.

Discussion In our whole-brain MEG study, we revealed that the functional connectivity and network efficiency values in the resting state beta and gamma bands are higher in epilepsy patients with FCD than in healthy controls. In addition, patients showed different network characteristics depending on the type of FCD. FCD type I and II brains in the

Abnormal functional brain network in focal cortical dysplasia

1623

Figure 3 Group averaged MIglob and Eglob values of control, FCD type I, and FCD type II subjects. Rank ANCOVA was applied for the beta and gamma band of MIglob values. Otherwise, we used ANCOVA with age as a covariate for statistical analysis (*p < 0.05).

beta band showed higher functional connectivity and network efficiency values than those of control brains. In the gamma band, the functional connectivity and network efficiency values of FCD type II brains were higher than those of control and of FCD type I brains.

Increased functional network in FCD patients Various types of measures can used for functional connectivity estimation such as correlation, coherence, phase leg index and nonlinear synchronization. Among them, we chose MI in the present study. MI can quantify the shared information between two time series based on information theory, and capture both linear and nonlinear relationships between time series (Pereda et al., 2005). MI is a relatively sensitive way to identify frequency-specific functional connectivity compared to cross-correlation, generalized synchronization, and phase synchronization (David et al., 2004). Because MI is not an amplitude-dependent measure, it may be more robust for the estimation of the changes in brain electrical activity than a linear method of spectral power analysis that depends on amplitude, and it is suited to measure changes in synchronization of different neuronal electrical activities (Frasch et al., 2007; Stam, 2005). In EEG and MEG studies, MI has been used to evaluate brain connectivity (Alonso et al., 2010; Bassett et al., 2009; Chen et al., 2008; David et al., 2004; Jin et al., 2010, 2011, 2012). In those previous studies, a higher MIglob value was interpreted as an indication of strong functional coupling among brain regions. To assess network topology, we used the widely reported ‘‘efficiency’’ measure, which enables investigators to measure the efficiency of a network from a global perspective. The Eglob measure is an average of Enodal values and is calculated using the inverse of the harmonic mean of the shortest weighted path length of the path to another node. By definition, when nodes within a functional network are associated with short path lengths among those nodes, that is, when a node can be easily connected to the rest of the nodes in the network, the network’s Eglob value will be enhanced. In this study, patients presented enhanced functional network strength and network efficiency in the beta and gamma frequency bands.

Recently, the utilization of noninvasive modalities to investigate fast frequency components in relation to determining an epileptogenic focus has been reported. In one previous study, authors determined the epileptic zone of refractory partial seizure by localizing the sources of MEG spike-locked power increases in the beta and gamma bands (12—55 Hz) (Guggisberg et al., 2008). Other MEG studies also proposed the usefulness of measuring high gamma oscillations (>50 Hz), which were highly associated with epileptic networks (Rampp et al., 2010; Xiang et al., 2009). However, none of those studies included a single pathology of FCD, used recordings of the resting state without using interictal spike data, compared patient data with that from a healthy control group, or analyzed multiple frequency bands. Therefore, it has been difficult to generalize the roles of the beta and gamma bands in the FCD patient group. Similar to the approach used in this study, Varotto and colleagues used graph theory in their network analysis of FCD patients and revealed a significantly different connectivity pattern, primarily in the gamma band, which distinguished the epileptogenic zone from other cortical regions (Varotto et al., 2012). However, they could not compare the FCD results with those from healthy brains because they employed invasive stereo-EEG methods. To date, no reports have confirmed that the epileptic FCD brain has different properties with the healthy brain in the beta and gamma bands, which have higher frequencies than the conventional interictal spike frequency. Several EEG and MEG studies have reported an increased functional connectivity in the theta band in epilepsy patients with tumor (Bartolomei et al., 2006; Bosma et al., 2009; Douw et al., 2010b; van Dellen et al., 2012a) or without available pathologic information (Douw et al., 2010a; Horstmann et al., 2010). Because differences in methodology, etiology, or duration of disease can affect interictal networks (van Diessen et al., 2013), our observations might be representative of different global network features of the epileptic FCD brain. The present study demonstrated for the first time that, compared to the healthy brain, the epileptic FCD brain has enhanced global network features in the beta and gamma frequency bands, results that may usher in studies on fast activities in FCD. It is also notable that we used resting state

1624 data independent from interictal spike data, which implies the possibility of using beta and gamma band data in the interictal state, even when there is no prominent interictal spike activity.

FCD type According to the previous study, epileptic patients with other pathologies such as low-grade gliomas or non-glial lesions showed different network characteristics with FCD (van Dellen et al., 2012b). Patients with those pathologies showed decreased theta band network synchronizability compared to that in healthy controls, while we observed increased network values in the beta and gamma bands in FCD. These results may suggest that the FCD brain has different electrophysiological epileptogenic mechanisms from those of other pathologies. Palmini and colleagues reported continuous interictal fast activity recorded directly from cortical dysplastic lesions during intraoperative electrocorticography suggesting an intrinsic epileptogenicity of FCD (Palmini et al., 1995). Although we could not detect evidence of interictal fast activity in our data, the increased functional connectivity values in the beta and gamma frequency bands may reflect its intrinsic epileptogenicity. Because an epileptogenic zone is responsible for the generation and propagation of epileptic seizure, we can speculate the presence of a route of easy electrophysiological access to the rest of the brain network through the epileptogenic zone in FCD brains. Easy access among brain regions can be reflected in the decreased path length within the brain network equivalent to the strong functional connectivity, and in turn, the increased network efficiency from the network perspective. Of the frequency bands that showed increased functional connectivity in the present study, the beta band is more likely to be related to the general FCD pathology, and may be more likely to differentiate the pathologic brain from the healthy one. We observed higher MIglob and Eglob values in FCD than control brains, but there were no differences between type I and type II FCD brains in the beta band. Conversely, increased functional connectivity in the gamma band seems specifically related to FCD type II. In the gamma band, the MIglob and Eglob values of type II brains were higher than those of healthy controls and FCD type I brains, but no significant difference was found between FCD type I and healthy control brains. The observed differences may connote those particular structural elements of type II FCD, for example, dysmorphic neurons or balloon cells (Blumcke et al., 2011), explain the increased functional connectivity in the gamma band. Our speculation about the role of the gamma band functional connectivity in FCD type II is supported by the study of Varotto and colleagues, in which they detected increased functional connectivity only in the gamma frequency band (Varotto et al., 2012), which may be because they only included patients with type II FCD in their study.

Surgical outcome and MRI findings with respect to functional connectivity Previous studies have reported that, compared to FCD type II, FCD type I is more likely to be associated with no lesion on

W. Jeong et al. MRI (Kim et al., 2009; Krsek et al., 2008), and the presence of an MRI evident lesion is a predictor of good surgical seizure outcome (Chung et al., 2005; Kim et al., 2009; Rowland et al., 2012). In the present study, even though we did not detect any significant differences related to MRI lesion or FCD type with respect to surgical outcome, FCD type I patients had a higher likelihood of no abnormality in MRI than that in FCD type II patients (p = 0.021). The discrepancy between previous studies and our results with respect to the surgical outcome may be attributable to small sample sizes. The location of MRI evident lesions, especially in a temporal location, is also an important predictor of good surgical outcome (Rowland et al., 2012); however, in our study, 11 of 15 patients who showed lesions on MRI had their lesion in an extratemporal location and 13 of 20 who showed normal MRI findings had their ictal onset zone in the temporal lobe. This difference in lesion location may also have affected the surgical outcome in the present study, resulting in no detectable difference in outcome depending on the presence of MRI abnormality. Regarding the relationship of functional connectivity values with MRI abnormality, we found no difference in MIglob and Eglob values according to the presence of MRI abnormality. Varotto and his colleagues suggested that advanced signal processing techniques aimed at studying synchronization and characterization of brain networks could substantially improve the pre-surgical evaluation of patients with focal epilepsy, even in cases without an associated anatomically detectable lesion (Varotto et al., 2012). However, they only included patients who had an MRI lesion, and the presence of dysplasia was established by means of MRI in accordance with the current radiological criteria for a diagnosis of type II FCD. In the present study, we included FCD patients diagnosed by pathology, regardless of the presence of an MRI lesion. The results show that FCD patients had increased functional connectivity regardless of the presence of a lesion on MRI, supporting the suggestion that advanced network analysis could improve the presurgical evaluation of patients with focal epilepsy, especially in cases without notable MRI lesions.

Limitations and future study To utilize MEG functional connectivity analysis in the clinical field, unraveling the relationship between altered functional connectivity and the epileptogenic zone would be necessary. In this study, we looked at functional connectivity at the whole-brain level; thus, we were only able to describe an increase in functional connectivity in the whole brain of patients with FCD over that in our control subjects. Therefore, we cannot say whether the specific location of FCD is related to brain areas that have greater functional connectivity than that in other areas, or whether the location of the ictal onset zone affects the distribution of the altered functional connectivity. Thus, the reconstruction of the MEG signal to the source level needs further study to investigate whether any specific regional source of increased functional connectivity is related to FCD lesion location and/or the ictal onset zone. The influence of spikes on the functional

Abnormal functional brain network in focal cortical dysplasia connectivity can be further investigated. We recruited epilepsy patients with a single pathology of FCD. Thus, results of studies into differences in functional networks among epilepsy patients with different pathologies, such as those with hippocampal sclerosis, would be of great interest. Also considering the poor surgical outcome of patients with FCD, it would be interesting to determine whether functional connectivity is related to surgical outcome. However, to study the relationship between functional connectivity and surgical outcome, we need to include enough patients to ensure statistical power. Additionally, we need to stratify the known risk factors, such as extra-temporal involvement, type I FCD, and no lesion on MRI (Chung et al., 2005; Rowland et al., 2012).

Conclusions The present study revealed, for the first time, a difference in resting state functional connectivity between patients with FCD and healthy controls by analyzing whole-brain MEG data. The results indicate that global functional connectivity in the beta and gamma bands of FCD patients in the resting state is stronger and has higher efficiency than that in control subjects. We suggest that the beta band is more likely related to the general FCD pathology; on the other hand, increased functional connectivity in the gamma band seems specifically related to FCD type II. The network differences were observed even in the resting state without any prominent interictal spike activity in our study. This result indicates that the future usage of the resting state network in a clinical setting could be useful as a subsidizing tool with other electrophysiological evaluation tools. Our report on global network differences may work as a base in the studies of the FCD epileptic brain network in the future.

Acknowledgments This research was partly supported by a grant of the Korean Health Technology R&D Project, the Ministry of Health and Welfare, Republic of Korea (Grant no. HI11C1360), and by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning (20090081342).

References Achard, S., Bullmore, E., 2007. Efficiency and cost of economical brain functional networks. PLoS Comput. Biol. 3, e17. Alonso, J.F., Ma˜ nanas, M.A., Romero, S., Hoyer, D., Riba, J., Barbanoj, M.J., 2010. Drug effect on EEG connectivity assessed by linear and nonlinear couplings. Hum. Brain Mapp. 31 (3), 487—497. Barrat, A., Barthelemy, M., Pastor-Satorras, R., Vespignani, A., 2004. The architecture of complex weighted networks. Proc. Natl. Acad. Sci. U.S.A. 101, 3747—3752. Barthelemy, M., Barrat, A., Pastor-Satorras, R., Vespignani, A., 2005. Dynamical patterns of epidemic outbreaks in complex heterogeneous networks. J. Theor. Biol. 235, 275—288. Bartolomei, F., Bosma, I., Klein, M., Baayen, J.C., Reijneveld, J.C., Postma, T.J., Heimans, J.J., van Dijk, B.W., de Munck, J.C., de Jongh, A., Cover, K.S., Stam, C.J., 2006. Disturbed

1625 functional connectivity in brain tumour patients: evaluation by graph analysis of synchronization matrices. Clin. Neurophysiol. 117, 2039—2049. Bassett, D.S., Bullmore, E.T., Meyer-Lindenberg, A., Apud, J.A., Weinberger, D.R., Coppola, R., 2009. Cognitive fitness of costefficient brain functional networks. Proc. Natl. Acad. Sci. U.S.A. 106, 11747—11752. Blumcke, I., Thom, M., Aronica, E., Armstrong, D.D., Vinters, H.V., Palmini, A., Jacques, T.S., Avanzini, G., Barkovich, A.J., Battaglia, G., Becker, A., Cepeda, C., Cendes, F., Colombo, N., Crino, P., Cross, J.H., Delalande, O., Dubeau, F., Duncan, J., Guerrini, R., Kahane, P., Mathern, G., Najm, I., Ozkara, C., Raybaud, C., Represa, A., Roper, S.N., Salamon, N., SchulzeBonhage, A., Tassi, L., Vezzani, A., Spreafico, R., 2011. The clinicopathologic spectrum of focal cortical dysplasias: a consensus classification proposed by an ad hoc Task Force of the ILAE Diagnostic Methods Commission. Epilepsia 52, 158—174. Bosma, I., Reijneveld, J.C., Klein, M., Douw, L., van Dijk, B.W., Heimans, J.J., Stam, C.J., 2009. Disturbed functional brain networks and neurocognitive function in low-grade glioma patients: a graph theoretical analysis of resting-state MEG. Nonlinear Biomed. Phys. 3, 9. Chavez, M., Valencia, M., Navarro, V., Latora, V., Martinerie, J., 2010. Functional modularity of background activities in normal and epileptic brain networks. Phys. Rev. Lett. 104, 118701. Chen, C.C., Hsieh, J.C., Wu, Y.Z., Lee, P.L., Chen, S.S., Niddam, D.M., Yeh, T.C., Wu, Y.T., 2008. Mutualinformation-based approach for neural connectivity during self-paced finger lifting task. Hum. Brain Mapp. 29, 265—280. Chung, C.K., Lee, S.K., Kim, K.J., 2005. Surgical outcome of epilepsy caused by cortical dysplasia. Epilepsia 46 (Suppl. 1), 25—29. David, O., Cosmelli, D., Friston, K.J., 2004. Evaluation of different measures of functional connectivity using a neural mass model. Neuroimage 21, 659—673. de Pasquale, F., Della Penna, S., Snyder, A.Z., Marzetti, L., Pizzella, V., Romani, G.L., Corbetta, M., 2012. A cortical core for dynamic integration of functional networks in the resting human brain. Neuron 74, 753—764. Douw, L., de Groot, M., van Dellen, E., Heimans, J.J., Ronner, H.E., Stam, C.J., Reijneveld, J.C., 2010a. ‘Functional connectivity’ is a sensitive predictor of epilepsy diagnosis after the first seizure. PLoS ONE 5, e10839. Douw, L., van Dellen, E., de Groot, M., Heimans, J.J., Klein, M., Stam, C.J., Reijneveld, J.C., 2010b. Epilepsy is related to theta band brain connectivity and network topology in brain tumor patients. BMC Neurosci. 11, 103. Engel Jr., J., Thompson, P.M., Stern, J.M., Staba, R.J., Bragin, A., Mody, I., 2013. Connectomics and epilepsy. Curr. Opin. Neurol. 26, 186—194. Frasch, M.G., Walter, B., Friedrich, H., Hoyer, D., Eiselt, M., Bauer, R., 2007. Detecting the signature of reticulothalamocortical communication in cerebrocortical electrical activity. Clin. Neurophysiol. 118, 1969—1979. Garcia Dominguez, L., Wennberg, R.A., Gaetz, W., Cheyne, D., Snead 3rd, O.C., Perez Velazquez, J.L., 2005. Enhanced synchrony in epileptiform activity? Local versus distant phase synchronization in generalized seizures. J. Neurosci. 25, 8077—8084. Guggisberg, A.G., Kirsch, H.E., Mantle, M.M., Barbaro, N.M., Nagarajan, S.S., 2008. Fast oscillations associated with interictal spikes localize the epileptogenic zone in patients with partial epilepsy. Neuroimage 39, 661—668. Horstmann, M.T., Bialonski, S., Noennig, N., Mai, H., Prusseit, J., Wellmer, J., Hinrichs, H., Lehnertz, K., 2010. State dependent properties of epileptic brain networks: comparative graphtheoretical analyses of simultaneously recorded EEG and MEG. Clin. Neurophysiol. 121, 172—185.

1626 Jin, S.H., Jeong, W., Seol, J., Kwon, J., Chung, C.K., 2013. Functional cortical hubs in the eyes-closed resting human brain from an electrophysiological perspective using magnetoencephalography. PLOS ONE 8, e68192. Jin, S.H., Lin, P., Hallett, M., 2010. Linear and nonlinear information flow based on time-delayed mutual information method and its application to corticomuscular interaction. Clin. Neurophysiol. 121 (3), 392—401. Jin, S.H., Lin, P., Hallett, M., 2012. Reorganization of brain functional small-world networks during finger movements. Hum. Brain Mapp. 33 (4), 861—872. Jin, S.H., Seol, J., Kim, J.S., Chung, C.K., 2011. How reliable are the functional connectivity networks of MEG in resting states? J. Neurophysiol. 106, 2888—2895. Kim, D.W., Lee, S.K., Chu, K., Park, K.I., Lee, S.Y., Lee, C.H., Chung, C.K., Choe, G., Kim, J.Y., 2009. Predictors of surgical outcome and pathologic considerations in focal cortical dysplasia. Neurology 72, 211—216. Krsek, P., Maton, B., Korman, B., Pacheco-Jacome, E., Jayakar, P., Dunoyer, C., Rey, G., Morrison, G., Ragheb, J., Vinters, H.V., Resnick, T., Duchowny, M., 2008. Different features of histopathological subtypes of pediatric focal cortical dysplasia. Ann. Neurol. 63, 758—769. Latora, V., Marchiori, M., 2001. Efficient behavior of small-world networks. Phys. Rev. Lett. 87, 198701. Liao, W., Zhang, Z., Pan, Z., Mantini, D., Ding, J., Duan, X., Luo, C., Lu, G., Chen, H., 2010. Altered functional connectivity and small-world in mesial temporal lobe epilepsy. PLoS ONE 5, e8525. Morgan, R.J., Soltesz, I., 2008. Nonrandom connectivity of the epileptic dentate gyrus predicts a major role for neuronal hubs in seizures. Proc. Natl. Acad. Sci. U.S.A. 105, 6179—6184. Newman, M.E., 2004. Analysis of weighted networks. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 70, 056131. Onnela, J.P., Saramaki, J., Kertesz, J., Kaski, K., 2005. Intensity and coherence of motifs in weighted complex networks. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 71, 065103. Palmini, A., Gambardella, A., Andermann, F., Dubeau, F., Dacosta, J.C., Olivier, A., Tampieri, D., Gloor, P., Quesney, F., Andermann, E., Paglioli, E., Pagliolineto, E., Coutinho, L., Leblanc, R., Kim, H.I., 1995. Intrinsic epileptogenicity of human dysplastic cortex as suggested by corticography and surgical results. Ann. Neurol. 37, 476—487. Park, K., Lai, Y.C., Ye, N., 2004. Characterization of weighted complex networks. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 70, 026109. Pereda, E., Quian Quiroga, R., Bhattacharya, J., 2005. Nonlinear multivariate analysis of neurophysiological signals. Prog. Neurobiol. 77, 1—37. Pereira, F.R., Alessio, A., Sercheli, M.S., Pedro, T., Bilevicius, E., Rondina, J.M., Ozelo, H.F., Castellano, G., Covolan, R.J., Damasceno, B.P., Cendes, F., 2010. Asymmetrical hippocampal connectivity in mesial temporal lobe epilepsy: evidence from resting state fMRI. BMC Neurosci. 11, 66. Rampp, S., Kaltenhauser, M., Weigel, D., Buchfelder, M., Ingmar Blumcke, I., Dorfler, A., Stefan, H., 2010. MEG correlates of epileptic high gamma oscillations in invasive EEG. Epilepsia 51, 1638—1642.

W. Jeong et al. Reijneveld, J.C., Ponten, S.C., Berendse, H.W., Stam, C.J., 2007. The application of graph theoretical analysis to complex networks in the brain. Clin. Neurophysiol. 118, 2317—2331. Rowland, N.C., Englot, D.J., Cage, T.A., Sughrue, M.E., Barbaro, N.M., Chang, E.F., 2012. A meta-analysis of predictors of seizure freedom in the surgical management of focal cortical dysplasia. J. Neurosurg. 116, 1035—1041. Rubinov, M., Sporns, O., 2010. Complex network measures of brain connectivity: uses and interpretations. Neuroimage 52, 1059—1069. Stam, C.J., 2005. Nonlinear dynamical analysis of EEG and MEG: review of an emerging field. Clin. Neurophysiol. 116, 2266—2301. Stam, C.J., de Haan, W., Daffertshofer, A., Jones, B.F., Manshanden, I., van Cappellen van Walsum, A.M., Montez, T., Verbunt, J.P., de Munck, J.C., van Dijk, B.W., Berendse, H.W., Scheltens, P., 2009. Graph theoretical analysis of magnetoencephalographic functional connectivity in Alzheimer’s disease. Brain 132, 213—224. Taulu, S., Hari, R., 2009. Removal of magnetoencephalographic artifacts with temporal signal-space separation: demonstration with single-trial auditory-evoked responses. Hum. Brain Mapp. 30, 1524—1534. van Dellen, E., Douw, L., Hillebrand, A., Ris-Hilgersom, I.H., Schoonheim, M.M., Baayen, J.C., De Witt Hamer, P.C., Velis, D.N., Klein, M., Heimans, J.J., Stam, C.J., Reijneveld, J.C., 2012a. MEG network differences between low- and high-grade glioma related to epilepsy and cognition. PLoS ONE 7, e50122. van Dellen, E., Douw, L., Hillebrand, A., Ris-Hilgersom, I.H.M., Schoonheim, M.M., Baayen, J.C., Hamer, P.C.D.W., Velis, D.N., Klein, M., Heimans, J.J., Stam, C.J., Reijneveld, J.C., 2012b. MEG network differences between low- and high-grade glioma related to epilepsy and cognition. PLoS ONE, 7. van Diessen, E., Diederen, S.J., Braun, K.P., Jansen, F.E., Stam, C.J., 2013. Functional and structural brain networks in epilepsy: what have we learned. Epilepsia 54, 1855—1865. van Wijk, B.C., Stam, C.J., Daffertshofer, A., 2010. Comparing brain networks of different size and connectivity density using graph theory. PLoS ONE 5, e13701. Varotto, G., Tassi, L., Franceschetti, S., Spreafico, R., Panzica, F., 2012. Epileptogenic networks of type II focal cortical dysplasia: a stereo-EEG study. Neuroimage 61, 591—598. Wang, L., Zhu, C., He, Y., Zang, Y., Cao, Q., Zhang, H., Zhong, Q., Wang, Y., 2009. Altered small-world brain functional networks in children with attention-deficit/hyperactivity disorder. Hum. Brain Mapp. 30, 638—649. Wieser, H.G., Blume, W.T., Fish, D., Goldensohn, E., Hufnagel, A., King, D., Sperling, M.R., Luders, H., Pedley, T.A., Commission on Neurosurgery of the International League Against, E, 2001. ILAE Commission Report. Proposal for a new classification of outcome with respect to epileptic seizures following epilepsy surgery. Epilepsia 42, 282—286. Wilke, C., Worrell, G., He, B., 2011. Graph analysis of epileptogenic networks in human partial epilepsy. Epilepsia 52, 84—93. Xiang, J., Liu, Y., Wang, Y., Kirtman, E.G., Kotecha, R., Chen, Y., Huo, X., Fujiwara, H., Hemasilpin, N., Lee, K., Mangano, F.T., Leach, J., Jones, B., DeGrauw, T., Rose, D., 2009. Frequency and spatial characteristics of high-frequency neuromagnetic signals in childhood epilepsy. Epileptic Disord. 11, 113—125.