Abstract
This study investigates the network mechanisms of temporal lobe epilepsy (TLE) using MEG data, focusing on directed connectivity networks across different frequency bands. Unlike previous studies that primarily localize epileptogenic zones, this research aims to explore whole-brain network differences between left TLE (lTLE), right TLE (rTLE), and healthy controls (HCs). MEG data from 13 lTLE patients, 21 rTLE patients, and 14 HCs were source-reconstructed to 116 brain regions (AAL116). Directed Transfer Function (DTF) was used to construct directed connectivity networks, followed by networks and graph-theoretical analyses. The results indicate that, compared to HCs, TLE subjects exhibited a significant increase in average connectivity strength in the Low Gamma band. The connectivity patterns across frequency bands in TLE patients were found to be unstable. Both HC and TLE subjects demonstrated left hemisphere lateralization. In the mid-to-low frequency bands, TLE subjects showed increases in global clustering coefficient (GCC), global characteristic path length (GCPL), and local efficiency (LE) compared to HCs, which is attributed to enhanced synchronization between local brain regions in TLE subjects.
Similar content being viewed by others
Introduction
Temporal lobe epilepsy (TLE) is the most common form of drug-resistant epilepsy in adults, which can be treated with curative or palliative interventions targeting specific regions through surgery or neuromodulation. Previously, the origin of such focal seizures was considered to be anatomically isolated. However, an increasing number of recent studies suggest that TLE is a network disorder involving disrupted large-scale brain networks1,2,3, replacing the classical concept of the “epileptogenic zone”. Brain network construction methods provide researchers with powerful tools to explore brain connectivity, thereby revealing the functions of different brain regions and their complex cortical communications, contributing to a better understanding of TLE4. Currently, brain network construction can be subdivided into structural connectivity (SC), functional connectivity (FC), and effective connectivity (EC)4. SC describes the physical connections through anatomically existing nerve fiber bundles, but this approach has the limitation of not providing functional information. FC methods calculate the synchronous activity between brain regions, which is represented by neural activities that are temporally correlated. EC reflects the direct regulation and influence of activities between brain regions, providing causal and directional information compared to functional connectivity.
The development of non-invasive neurodata acquisition techniques, combined with powerful network modeling tools, offers a more accurate characterization and quantification of network disruption patterns in TLE. Researchers commonly use neuroimaging data (derived from magnetic resonance imaging (MRI) techniques) and neurophysiological data (from magnetoencephalography (MEG) and scalp electroencephalography (EEG) recordings) for constructing brain networks5. For example, an EEG study by Coito et al. found a significant reduction in connectivity within regions consistent with the default mode network (DMN) in patients with TLE, and distinct network patterns between TLE subjects and healthy controls (HCs). The strongest connections were observed the ipsilateral hippocampus in TLE patients and the posterior cingulate cortex in HCs, respectively6. In a separate study, Englot et al. used fMRI to identify widespread reductions in resting-state FC in epilepsy patients, particularly involving the lateral fissure, posterior temporoparietal junction, and orbitofrontal cortex7. Kudo et al. constructed brain networks using MEG data, revealing significant alterations in alpha band information flow in the frontal and occipital regions of TLE patients compared to HCs8. Furthermore, Carboni et al. employed EEG data to build brain networks and calculated global efficiency, demonstrating a significant increase in the global efficiency of resting-state networks in TLE patients compared to HCs, especially in the somatomotor network, ventral attention network, and DMN9. Connectivity studies based on functional MRI (fMRI) typically rely on correlation analysis to construct FC networks, due to the low temporal resolution of fMRI and the fluctuations of the hemodynamic response10. In contrast, EEG and MEG provide more direct measurements of neuronal activity and offer higher temporal resolution, making them well-suited for constructing EC networks. Among these, MEG has become a crucial electrophysiological source imaging modality in the study of TLE network mechanisms due to its non-invasive nature, superior spatiotemporal resolution, and higher number of channels11,12. EC networks can be constructed using various methods, with Granger Causality Analysis (GCA) being one of the most commonly used approaches.
The Directed Transfer Function (DTF) is a frequency-domain method based on GCA13,14,15, widely used in neuroscience for constructing and analyzing EC brain networks. This method can infer directional connections between multivariate time series and accurately estimate the information flow between different brain regions, thereby providing deeper insights into the information transfer pathways within brain networks. Previous studies have demonstrated that DTF exhibits strong robustness in handling nonlinear signals16,17. To explore the network mechanisms of TLE, researchers have extensively employed the DTF method to establish directed connectivity networks in TLE patients using multimodal data. These networks have been applied to various aspects such as the localization of epileptic foci18,19,20,21,22, epilepsy diagnosis and prediction23,24, and the analysis of differences in brain network connectivity25,26,27. Currently, the studies utilizing DTF analysis on MEG data have focused primarily on the localization of epileptic foci. For example, Dai et al. used the DTF method to analyze source signals from five cases of drug-resistant epilepsy to locate the primary sources and propagation pathways of seizures, providing valuable references for clinical assessment and treatment of epilepsy patients19. Sohrabpour et al. explored individual network nodes using DTF and Partial Directed Coherence (PDC) methods on simulated signals. Additionally, they applied the adaptive Directed Transfer Function (aDTF) method to analyze two epilepsy patients (one with parietal lobe epilepsy and one with temporal lobe epilepsy) to explore EC within brain networks of foci regions20. The application of the DTF method to MEG data has yielded excellent results in epilepsy localization studies, underscoring the effectiveness of DTF in investigating TLE networks using MEG data. Notably, there is a research gap in constructing whole-brain EC networks with the DTF method based on MEG data to compare network differences between TLE subjects and HCs. This gap exists in two aspects: First, current studies lack network analysis of high-frequency components; Second, they mainly focus on regions of interest (ROIs) in cortical areas, which is due to the challenges of observing deep brain structures with MEG. However, several recent MEG studies have shown that MEG can detect source activity in subcortical regions28,29,30, suggesting that the construction of whole-brain EC networks is feasible.
Therefore, this study aims to investigate whole-brain EC network differences across various frequency bands during the resting state. Specifically, the DTF method will be used to construct whole-brain EC networks from MEG source signal data of subjects with left TLE (lTLE), right TLE (rTLE), and HCs, with frequency-specific analysis. A comprehensive network analysis (average connectivity strength, brain region outflow, connectivity patterns, directionality and lateralization) and graph-theoretical analysis will be conducted. The main contributions of this work are two aspects: First, we utilize MEG data from all brain regions to construct EC brain networks, providing insights into EC differences in subcortical structures. Second, we construct EC brain networks across six frequency bands, offering an analysis of high-frequency EC network (high gamma) differences between TLE and HCs. Figure 1 illustrates the overall workflow of this study, while Table 1 presents the full names of the brain regions corresponding to the abbreviations used in the article.
Workflow for Analyzing Directed Connectivity Brain Networks Across Frequency Bands Using MEG Data.
Results
Average connectivity strength
Figure 2 shows the average connectivity strength of the directed brain networks across different frequency bands for HCs, lTLE, and rTLE subjects. The results indicate that in the Low Gamma frequency band, the mean connectivity strength in both the lTLE and rTLE groups is significantly higher than that in the HC group. No significant differences were observed in the other frequency bands.
Average connectivity strength across different frequency bands for the HCs, lTLE, and rTLE groups. The data represent the mean connectivity values of the normalized DTF-directed brain networks. The x-axis indicates different frequency bands, while the y-axis shows the average connectivity strength values. The box represents the interquartile range (from the first to the third quartile), and the line inside the box indicates the median. The whiskers denote the range of normal values excluding outliers, and each dot represents an individual subject. The SPSS Mann-Whitney U non-parametric test with Holm correction was used to assess the statistical significance of differences between the HCs and the lTLE and rTLE groups, where * indicates \(q < 0.05\), ** indicates \(q < 0.01\), and *** indicates \(q < 0.001\).
Connectivity patterns
Figure 3 and Table 2 present the connectivity patterns of the HCs, lTLE, and rTLE groups across different frequency bands. The results indicate differences in connectivity patterns between HCs and lTLE/rTLE subjects across various frequency bands. In each group, several brain regions exhibited higher connectivity strength and degree centrality (DC) values than other regions, acting as hubs for information transmission. Specifically, for HCs, the ACG.L region serves as a hub in the Delta, Theta, and Alpha bands, while in the Beta band, the hub shifts to both the ACG.L and ORBsupmed.R regions. In the high-frequency bands (Low Gamma and High Gamma), the ORBsupmed.R region plays a crucial hub role. For lTLE subjects, the THA.L and PAL.R regions serve as hubs in the Delta band, while in the Theta band, ACG.L, ORBsupmed.R, and THA.L jointly function as hub regions. In the Alpha and Beta bands, the hub role of THA.L weakens, while it becomes dominant in the mid-to-high frequency bands (Low Gamma and High Gamma). For rTLE subjects, the changes in hub regions are relatively stable compared to lTLE, with THA.L as the hub in the low frequencies (Delta, Theta) and DCG.L in the mid frequencies (Alpha, Beta, Low Gamma), while in the High Gamma band, STG.L becomes the dominant hub region.
Connectivity patterns across different frequency bands for the three groups of subjects. Each row represents a subject group, and each column represents a specific frequency band. The hub region is highlighted within the red rectangular box. The figure displays only the top 20 connections from the normalized DTF average connectivity matrix. The visualization was generated using the BrainNet Viewer toolbox on the Matlab platform.
It is noteworthy that the connectivity patterns of HC subjects exhibit minimal variation across different frequency bands, with the hub regions remaining relatively stable. In contrast, lTLE and rTLE subjects show variations in connectivity patterns across frequency bands.
Total outflow
Figure 4 shows the total outflow of brain regions based on non-normalized DTF across different frequency bands for HC, lTLE, and rTLE subjects. The brain regions PreCG.L, PreCG.R, HIP.L, HIP.R, AMYG.R, PoCG.R, ANG.L, ANG.R, PCL.L, and PCL.L exhibited relatively higher total outflow across all three groups. Mann-Whitney U non-parametric tests indicated that, for lTLE subjects, there were no significant differences in the total outflow of most brain regions compared to HCs in the Theta and Alpha bands. However, in all other frequency bands, the total outflow of all brain regions showed significant differences compared to HC subjects (\(q < 0.05\)). For rTLE subjects, the total outflow of all brain regions significantly differed from HCs across all frequency bands (\(q < 0.05\)).
Total outflow of brain regions for HCs, lTLE, and rTLE groups across different frequency bands. This data is obtained by averaging the second dimension of the non-normalized 3D DTF (Directed Transfer Function) connectivity matrix for each group. The x-axis represents different brain regions, while the y-axis represents the corresponding DTF outflow values for those regions.
Directionality index
Figures 5 and 6 illustrate the directionality index (DI) for specific brain regions across different frequency bands in the three participant groups. The results indicate differences in directionality between HCs and TLE patients, with the largest difference in DI between HCs and lTLE patients occurring in the Alpha frequency band, and between HCs and rTLE patients in the Delta frequency band. Specifically, in lTLE participants, directionality was higher than in HCs in brain regions such as ORBmid.R, OLF.L, REC.R, INS.L, INS.R, SOG.R, ITG.L, ITG.R, CRBL10.L, and Vermis8, while it was lower in regions such as SFGdor.R, IFGoperc.L, SFGmed.R, PCG.R, CUN.L, SPG.L, CRBL6.L, and Vermis9. For rTLE participants, directionality was higher than in HC participants in regions including ORBmid.R, OLF.L, REC.L, INS.L, INS.R, AMYG.L, IOG.L, ITG.R, CRBL8.L, CRBL8.R, CRBL9.L, and CRBL9.R, and lower in regions such as SFGdor.L, SFGdor.R, MFG.L, IFGoperc.L, CUN.L, LING.R, SPG.L, SPG.R, PCUN.L, CRBL6.L, Vermis7, Vermis9, and Vermis10. These findings are based on full-frequency band analyses. To further explore the DI value differences between HC and TLE subjects, we conducted additional clustering experiments. The methods and results of the clustering experiments are presented in Supplementary Manuscript Sections S1.1 Cluster analysis method and S2.1 Cluster analysis results.
DI for HC, lTLE, and rTLE participants across different frequency bands. Figure displays the heatmaps of DI values for each group across different frequency bands. The color bar transitions from blue to red, indicating the directionality of the DI values: red represents a positive DI value (with a maximum of 1), blue represents a negative DI value (with a minimum of -1), and white represents a DI value of 0, indicating no difference.
DI for HCs, lTLE, and rTLE participants across different frequency bands. Figure shows the difference maps obtained by calculating the differences in DI values between lTLE and rTLE participants compared to HCs across various frequency bands. The color bar transitions from blue to red, indicating the directionality of the DI values: red represents a positive DI value (with a maximum of 1), blue represents a negative DI value (with a minimum of -1), and white represents a DI value of 0, indicating no difference.
Lateralization index
Figure 7a shows the lateralization index (LI) for specific brain regions across different frequency bands for the three participant groups. The results indicate that certain regions, such as IFGoperc, OLF, CAL, and PCL, exhibit left-hemisphere lateralization, while regions like ACG, AMYG, SOG, POCG, and TPOsup display right-hemisphere lateralization. Notably, compared to HCs, both lTLE and rTLE participants show lateralization differences in the MFG, IFGtriang, CAL, FFG, ANG, PCUN, and CRBL6 regions across multiple frequency bands. Additionally, Figure 7b compares the overall brain lateralization across different frequency bands for the three participant groups. The results show that all three groups exhibit left-hemisphere lateralization, and there are no significant differences in the lateralization indices between the groups.
Lateralization Index (LI) for HCs, lTLE, and rTLE groups across different frequency bands. Part a displays the LI values for specific brain regions in each group across various frequency bands. The asterisk (*) indicates the top 10 LI differences obtained by subtracting the LI values of lTLE and rTLE patients from those of HCs for each frequency band, leading to a total of 120 asterisks. Part b shows the overall brain lateralization index for HCs, lTLE, and rTLE patient groups across different frequency bands.
Graph theoretical analysis
Comparing the comprehensive global topological parameters of networks between HCs and lTLE and rTLE participants (Fig. 8), the results indicate that, in the Delta, Theta, Beta, and Low Gamma bands, the global clustering coefficient (GCC) of both lTLE and rTLE subjects was significantly higher than that of HCs. In contrast, in the High Gamma band, the GCC was lower for both TLE groups compared to HC subjects. In the Alpha band, only lTLE subjects showed significantly higher GCC compared to HCs. The results for global characteristic path length (GCPL) and local efficiency (LE) were similar to those of GCC. For the LE attribute, TLE subjects had significantly higher values than HCs in the mid-to-low frequency bands (Delta, Theta, Alpha, Beta, Low Gamma), while the High Gamma band showed the opposite pattern. For the GCPL attribute, TLE subjects had significantly higher values than HCs in the mid-to-low frequencies (Delta, Beta, Low Gamma), and in the Theta band, only rTLE showed a significant difference compared to HCs. As with GCC, the High Gamma band exhibited the opposite pattern. The results for the global efficiency (GE) attribute differed from the other three attributes. In the mid frequencies (Beta, Low Gamma), TLE subjects had significantly lower GE values compared to HC subjects, while the High Gamma band showed the opposite trend.
Four global topological parameters of brain functional networks in HCs, lTLE, and rTLE patients. The SPSS Mann-Whitney U test with Holm correction was used to assess significant differences between HCs and the rTLE and lTLE groups, where * indicates \(q < 0.05\), ** indicates \(q < 0.01\), and *** indicates \(q < 0.001\).
Discussion
Average connectivity strength
Our study shows that both lTLE and rTLE subjects exhibit significantly higher average connectivity strength in the Low Gamma frequency band compared to HCs, although differences in average connectivity strength were observed in the mid- and low-frequency bands, these differences did not reach statistical significance. This finding is consistent with previous studies, such as a MEG study by Wang et al., which used a phase lag index to construct functional brain networks. They found that in the Low Gamma and alpha frequency bands, TLE subjects had significantly higher average connectivity strength than HC subjects, and that in high-frequency bands, the average functional connectivity in epilepsy patients was more strongly affected by epilepsy than in lower frequency bands31. Similarly, Niso et al. used MEG data from 30 epilepsy patients and 15 HCs to construct functional connectivity through phase-locking value (PLV). They found that epilepsy patients exhibited significantly higher connectivity patterns across all frequency bands compared to HCs32. Therefore, our results further support the conclusion of Wang et al., namely that in high-frequency bands, epilepsy patients’ average functional connectivity is more strongly affected by epilepsy than in lower-frequency bands. The increased functional connectivity is associated with a functional imbalance caused by epileptic activity. The primary reason for this increased connectivity may be the epilepsy-induced disruption of theta oscillations in the septohippocampal system, along with the damage to septal and hippocampal GABAergic cells, which contributes to the abnormalities in theta activity33.
Connectivity patterns
We found that the brain connectivity patterns in HCs were relatively stable across different frequency bands, with ACG.L and ORBsupmed.R serving as hub regions in various bands. This finding is consistent with previous studies34,35. These brain regions play a crucial role in the transmission and integration of information within the brain. For example, the ACG is known to be important in emotional processing and cognitive control36, while ORBsupmed is associated with higher cognitive functions and emotional regulation.
In contrast, the connectivity patterns in both lTLE and rTLE subjects exhibited greater variability across different frequency bands. In the TLE group, THA.L, DCG.L, and STG.L served as hub regions in different frequency bands. Previous studies have indicated that the THA is a central node in seizure propagation37,38, while the DCG and STG are associated with seizure propagation and maintenance39. This pattern of variability is consistent with the pathological mechanisms of epilepsy, where seizures lead to brain network reorganization and abnormal synchronization, resulting in significant changes in brain network connectivity patterns40. These changes further contribute to the poorer performance of TLE subjects in cognitive tasks compared to HCs41.
Total outflow
The PCL.L brain region exhibits the highest total outflow across all groups, which is not surprising considering its role as a hub in the functional connectivity network. Previous studies have shown that the hub regions of the functional connectivity network are primarily located in the medial and lateral frontal and parietal lobes (including the PCL region) as well as the lateral frontal lobe42, indicating that these hub regions play crucial roles in information transmission and integration43,44.
In addition, several brain regions in TLE patients, such as HIP and AMYG, exhibit significantly higher total outflow compared to healthy controls. These brain regions, which show significant differences in total outflow, are closely related to emotional processing, memory, and the foci and propagation of epileptic seizures45. The total outflow results are consistent with previous studies. Verhoeven et al. estimated whole-brain directed functional connectivity using resting-state high-density EEG recordings from 20 left TLE patients, 20 right TLE patients, and 35 HCs. Their findings indicated that outflow from the left and right hippocampus was the most important for diagnosis46.
Directionality index and lateralization index
The results indicate that although there are differences in the DI values of brain regions between HCs and TLE subjects, no significant findings were observed. Additionally, lateralization differences were found in some brain regions across the three groups, with a general trend of left hemisphere dominance. Left hemisphere lateralization is typically associated with language function and is a common phenomenon in healthy individuals47. There is no definitive conclusion regarding the lateralization in TLE subjects. For example, a study by Coito et al., using EEG data from 20 lTLE, 20 rTLE, and 20 HCs, found that both HCs and lTLE subjects exhibited left hemisphere lateralization, whereas rTLE subjects displayed right hemisphere lateralization6. In contrast, a study by Carpentier et al. using fMRI found that rTLE subjects still exhibited significant left hemisphere lateralization during language comprehension and production tasks, despite the epilepsy originating from the right hemisphere48. The lack of change in lateralization in rTLE subjects may be due to the recruitment of contralateral homologous regions, which may lead to the development of contralateral homologous functions, including left-hemisphere dominance for language and compensatory cognitive functions49,50,51. However, the reasons for the differences in results remain unclear, and further research with a larger sample size and more extensive investigation of local brain regions is needed to address this issue comprehensively.
Topological properties
Our study demonstrates significant differences in global topological properties between HCs and TLE subjects. In the middle and low-frequency bands (delta, theta, alpha, beta), our findings are largely consistent with previous studies. For example, Li et al. evaluated the brain network properties of 20 left and 23 right MRI-negative TLE patients and 22 HCs using magnetoencephalographic recordings in six main frequency bands. They found that the TLE groups exhibited significantly greater values in the GCC and GCPL compared to HCs in the theta band52. Similarly, a study by Bartolomei et al. using EEG reported a similar phenomenon, with mesial TLE patients showing significantly increased network GCC and GCPL attributes compared to non-mesial TLE patients53. This increase in network GCC and GCPL properties in TLE subjects in the middle and low-frequency bands is widespread54,55,56,57. Wang et al. reported a significant decrease in GE in TLE subjects in a study with 26 TLE subjects and 25 HCs58. Similarly, Yasuda et al. conducted a cross-sectional study involving 86 left TLE subjects, 70 right TLE subjects, and 116 HCs, finding that TLE subjects had significantly reduced network GE and significantly increased LE attributes compared to HCs57. This observation may be explained by enhanced synchronization between local brain regions in TLE subjects52, which leads to the formation of tighter local connections in brain networks in the middle and low-frequency bands, creating more segregated functional modules. The decrease in GE reflects cognitive impairments in TLE subjects, as a reduction in global efficiency is typically associated with a weakening of the brain’s overall functional integration capacity59,60.
However, in the High Gamma frequency band, our results show that both lTLE and rTLE subjects exhibited significant decreases in GCC, GCPL, and LE, while GE significantly increased. This is in contrast to the network property results observed in the middle and low-frequency bands. These frequency-dependent changes in brain network topology may be associated with high-frequency activity during seizures (such as epileptic spikes), which often involve widespread cross-regional networks61.
Limitations
This study has several limitations. First, due to constraints such as a limited number of subjects, data collection costs, and environmental conditions during the actual data acquisition process, the dataset size is relatively small. In the future, efforts should be directed toward acquiring more clinical data. Secondly, this study constructed static directed brain networks; however, the connections between brain regions are inherently unstable. Future research could benefit from constructing dynamic directed brain networks to provide a more dynamic perspective for network analysis. Third, the beamforming map should exhibit the topography of local maxima, and the underlying generator may span across two ROIs. Therefore, in subsequent experiments, additional validation should be performed during the source reconstruction process of MEG data to ensure the accuracy of the representative channel for each ROI.
Methods
Data acquisition
Nanjing Brain Hospital affiliated with Nanjing Medical University in China recruited 48 subjects for head MEG and structural MRI (sMRI) scans. The data collection process was conducted following written consent from the participants. During the MEG data acquisition, the subjects remained awake with their eyes closed and in a supine position. EEG data were simultaneously recorded during the scanning process. The head position is measured at the beginning and end of the scan. Data are excluded if the head movement exceeds 5 mm during recording. The 48 subjects included 14 HCs (Male: Female = 6:8; Age [years]: 27.90±6.97) and 34 TLE subjects (Male: Female = 23:11; Age [years]: 27.00±5.63; Course [years]: 9.29±4.95; Age at onset [years]: 17.7±7.19), providing high-quality MEG and sMRI modality data. The TLE group included 13 subjects with lTLE (Male: Female = 9:4; Age [years]: 26.08±5.54; Course [years]: 7.46±3.23; Age at onset [years]: 18.62±6.65) and 21 subjects with rTLE (Male: Female = 14:7; Age [years]: 27.57±5.74; Course [years]: 10.43±5.54; Age at onset [years]: 17.14±7.61). The quality of the MEG modality data was assessed by two experienced experts in the MEG field. HCs had no history of psychiatric disorders or sMRI abnormalities, while all TLE subjects were diagnosed with drug-resistant TLE with no significant structural MRI lesions. There were no significant differences in age and gender among the HCs, lTLE, and rTLE groups, and all subjects were right-handed. Clinical data for the TLE subjects are provided in Supplementary Table 1 (Participation’s information).
MEG modality data were acquired using a whole-head MEG system with 275 channels (VSM Med Tech Systems Inc., Coquitlam, BC, Canada). The original sampling frequency of the MEG data was 4 kHz, and synthetic 3rd-order gradiometry was applied during the MEG data acquisition process. Although each subject underwent 20 data acquisitions, with each scan lasting 120 seconds, we selected one acquisition for analysis. This segment was carefully chosen to minimize the presence of seizure-like signals, which helps in defining the interictal period. sMRI data were collected using a GE Signa NV/i 1.5 T scanner. Informed written consent was obtained from all subjects prior to the data acquisition process.
Data preprocessing
Figure 9 illustrates the preprocessing pipeline. As shown in the figure, MEG data underwent stringent quality control, with abnormal channels replaced by the average signal from adjacent channels. Specifically, the first step involved identifying and removing bad channels. We calculated the correlation between each channel and its neighboring channels and compared it to a correlation threshold. Channels with correlation values lower than three standard deviations below the mean correlation of all channels were marked as bad channels and excluded. The second step involved identifying and removing bad segments. For each channel, we computed the Z-score of every data point relative to the statistical characteristics of the entire time series. Data points with Z-scores exceeding the threshold (20) were removed, and the remaining segments were stitched together to ensure signal continuity and smoothness. Subsequently, a high-pass filter with a cutoff of 0.5 Hz was applied to remove baseline drift, and a 50 Hz notch filter was used to eliminate power line interference. Independent Component Analysis (ICA) was employed to remove artifacts electromyography (EMG), electrooculography (EOG), electrocardiography (ECG)62, and large spikes. The artifact-corrected MEG data were then resampled to 500 Hz, and the signal was segmented into 2-second epochs. Each epoch was visually inspected, and a relatively stable segment of the signal was selected to represent the spike-free MEG signal for the subject. T1 MRI data are co-registered based on MEG fiducial points (nasion, preauricular points, and cranial landmarks). The above data preprocessing was carried out using the FieldTrip 2019 toolbox on the MATLAB 2018b platform.
Workflow for Data Preprocessing.
The preprocessed MEG data remained as time series for each channel. To reconstruct the MEG data to brain regions, we constructed head and source models. These models were built using T1 MRI data: The head model was constructed by segmenting and extracting the inner surface of the skull from the T1 MRI data, while the source model was based on slices of the T1 MRI data to describe dipole positions and orientations derived from cortical surface information. It is a volumetric grid model with a regular resolution of 8 mm. Linear constraint minimum variance (LCMV) beamformer was then used for reconstruction63,64,65,66, over 0-250 Hz frequency band, and the whole brain was parcellated into 116 regions using the Anatomical Automatic Labeling (AAL) template67,68,69. By calculating the total power of each time series across all time points, the channel with the highest total power is selected as the region of interest (ROI) signal for each ROI. After acquiring the ROI signals, a matrix of time series for brain regions, measuring 116 (number of brain regions) by 1000 (time points), was obtained in preparation for the subsequent construction of the connectivity matrix. The head and source models were constructed using the FreeSurfer and HCP Workbench tools.
Directed transfer function
In this study, the DTF method was used to assess the directional connectivity between the time series obtained from source reconstruction, reflecting the information flow between brain regions. The computation of DTF was performed using the econectome toolbox ( https://www.nitrc.org/projects/econnectome/) on the MATLAB 2022b platform. For each subject, both non-normalized and normalized 3D DTF coefficient matrices were obtained (\(116 \times 116 \times f\), with frequency ranging from 1 to 250 Hz, with a frequency step of 1 Hz), representing the information flow from one brain region to another at specific frequency bands. To assess the outflow of information from brain regions across the entire frequency range, we calculated the mean value of the second dimension of the non-normalized DTF matrix for each subject. This was followed by a group average across all subjects to obtain the total outflow of information from brain regions across all frequencies.
Additionally, we extracted 3D DTF coefficient matrices for six frequency bands from both non-normalized and normalized DTF matrices: Delta (1-4 Hz), Theta (4-8 Hz), Alpha (8-12 Hz), Beta (12-30 Hz), Low Gamma (30-80 Hz), and High Gamma (80-250 Hz). By averaging across the third dimension for each frequency band, we generated a \(116 \times 116\) DTF connectivity matrix for each participant, representing the average information flow in that frequency band. To investigate group differences between lTLE, rTLE, and HCs, we computed group averages of the DTF connectivity matrices for each frequency band, resulting in group-specific DTF average connectivity matrices for lTLE, rTLE, and HCs. For calculating average connectivity strength, we determined the mean value of the average connectivity matrix for each subject across each frequency band, providing a single DTF strength metric per participant.
Normalized DTF was used to estimate relative connectivity strength and overall network characteristics, such as average connectivity strength and topological properties. Non-normalized DTF was used to reflect actual connectivity strength, such as node outflow and inflow.
The network metrics used in this study represent concepts related to different neurophysiological aspects. The average connection strength of the network reflects the overall level of information flow in the brain. The total outflow value of a single brain region indicates the amount of information transmitted from that region to others, while the inflow value represents the opposite, reflecting the amount of information received by that region.
Directional and lateralization metrics
To gain a deeper understanding of the directional flow of information between brain regions, we introduced the DI, which aims to quantify the directional preference of information transfer between specific brain regions. The formula for calculating DI is as follows:
Where \(\text {ROI}_{\text {outflow}}\) represents the total amount of information flowing from the brain region to other brain regions, while \(\text {ROI}_{\text {inflow}}\) represents the total amount of information flowing into the brain region from other regions. By calculating the DI, we can quantitatively assess the role of each brain region in the overall brain network in terms of information transfer. For example, a positive DI value indicates that the brain region primarily acts as a source of information output, while a negative DI value suggests that the brain region primarily receives information from other regions. The DI values are calculated based on the non-normalized group-average DTF matrices. To quantitatively analyze the differences in driving force between the left and right hemispheres among the groups and to further investigate the distribution of driving force in specific brain regions, we calculated the LI for both the whole brain and individual brain regions:
\(\text {sum}(\text {LH})\) represents the total outflow of information from the left hemisphere, while \(\text {sum}(\text {RH})\) represents the total outflow from the right hemisphere. The \(\text {LI}_{\text {sum}}\) is the LI for the total outflow of the entire brain. For specific brain regions, \(\text {LH}_{\text {ROI}}\) denotes the total outflow from a particular region in the left hemisphere, and \(\text {RH}_{\text {ROI}}\) denotes the total outflow from the corresponding region in the right hemisphere. The \(\text {LI}_{\text {ROI}}\) represents the LI for the total outflow of an individual brain region. The calculation of \(\text {LI}_{\text {sum}}\) is based on the non-normalized DTF average matrices, while the \(\text {LI}_{\text {ROI}}\) is calculated from the non-normalized group DTF average connectivity matrices.
Graph theoretical analysis
Graph theory provides a robust framework for quantitatively describing brain nodes and their connections. We extracted four global topological properties and one local topological property from the normalized DTF average connectivity matrix for each subject. The four global topological properties are: GCC, GCPL, GE, and LE, and the local topological property is DC. These topological properties reflect different aspects of the network’s capabilities. The extraction of these topological properties was performed using the Gretna toolbox (https://www.nitrc.org/projects/gretna/) on the MATLAB 2022b platform.
Connectivity patterns
The identification of hub regions is determined by both DC and the strongest 20 connections. If a single brain region appears more than 12 times in the strongest 20 connections and is among the top 5 brain regions in terms of DC, it is considered the unique hub region for that group in the given frequency band. If the two brain regions with the highest frequency of occurrence are both within the top 5 brain regions for DC, and the sum of their frequencies in the strongest 20 connections exceeds 12, then both of these brain regions are considered hub regions. If the frequencies in the strongest 20 connections are more widely distributed across 4 or more brain regions, starting with the brain region with the highest frequency among the top 20 strongest connections, the frequencies were summed sequentially until the cumulative frequency exceeded 12. The brain regions contributing to this cumulative frequency were identified as hub regions. The threshold of a frequency of 12 was chosen because it represents 60% of the top 20 strongest connections. The DC and the top 20 connections related to the hub region selection process for the three groups of subjects across different frequency bands are presented in Supplementary Table 2 (Hub region selection process).
Statistical analysis
All data requiring statistical analysis were first tested for normality using the Kolmogorov-Smirnov (K-S) test. This non-parametric method assesses whether sample data follow a normal distribution. For data meeting the normality assumption, independent (unpaired) t-tests were conducted. The t-test assumes normally distributed data, ensuring that the test statistic follows a t-distribution, which is reliable for small sample sizes. For data that did not meet the normality assumption, the Mann-Whitney U test was performed. This test does not rely on specific distribution assumptions and is applicable when data significantly deviate from normality. The Holm correction was applied to the multiple comparisons of the above t-tests and Mann-Whitney U tests to control the family-wise error rate70. All statistical analyses were conducted using MATLAB 2022b software.
Ethics statement
The clinical data acquisition of MRI and MEG for the Beijing Nova Program, led by Chunlan Yang, was conducted according to the guidelines of the Declaration of Helsinki, and all procedures involving human subjects were approved by the Beijing University of Technology Ethics Committee (Approval No. 20230484469). This approval was granted on December 1st, 2023, and falls within the scope of science and technology ethics. The ethical review certificate has also been submitted to and received approval from the Beijing Municipal Science & Technology Commission and the Administrative Commission of Zhongguancun Science Park.
Conclusion
This study aimed to investigate differences among HCs, lTLE, and rTLE groups across various frequency bands. Normalized DTF was employed to calculate and analyze overall network characteristics, including metrics such as average connectivity strength, connectivity patterns, and global properties, which emphasize relative connectivity strength and overall network organization. Non-normalized DTF was used to compute node-level metrics, including out-degree, LI, and DI, providing a more direct reflection of actual connectivity strengths.
The results revealed that, compared to HCs, TLE patients exhibited a significant increase in average connectivity strength within the Low Gamma band. Additionally, the connectivity patterns across frequency bands in TLE patients were found to be unstable. Both HC and TLE groups demonstrated left hemisphere lateralization. In the mid-to-low frequency bands, TLE patients showed increases in global clustering coefficient, global path length, and local efficiency compared to HCs. These findings suggest enhanced synchronization between local brain regions in TLE patients.
Data availability
The datasets used and analysed during the current study available from the corresponding author on reasonable request.
References
Caciagli, L. et al. Thalamus and focal to bilateral seizures: a multiscale cognitive imaging study. Neurology 95, e2427–e2441. https://doi.org/10.1212/WNL.0000000000010645 (2020).
Girardi-Schappo, M. et al. Altered communication dynamics reflect cognitive deficits in temporal lobe epilepsy. Epilepsia 62, 1022–1033. https://doi.org/10.1111/epi.16864 (2021).
Richardson, M. P. Large scale brain models of epilepsy: Dynamics meets connectomics. J. Neurol. Neurosurg. Psychiatry 83, 1238–1248. https://doi.org/10.1136/jnnp-2011-301944 (2012).
Cao, J. et al. Brain functional and effective connectivity based on electroencephalography recordings: A review. Hum. Brain Mapp. 43, 860–879. https://doi.org/10.1002/hbm.25683 (2022).
Cao, M. et al. Dynamical network models from eeg and meg for epilepsy surgery-a quantitative approach. Front. Neurol. 13, 837893. https://doi.org/10.3389/fneur.2022.837893 (2022).
Coito, A. et al. Altered directed functional connectivity in temporal lobe epilepsy in the absence of interictal spikes: A high density eeg study. Epilepsia 57, 402–411. https://doi.org/10.1111/epi.13308 (2016).
Englot, D. J. et al. Global and regional functional connectivity maps of neural oscillations in focal epilepsy. Brain 138, 2249–2262. https://doi.org/10.1093/brain/awv130 (2015).
Kudo, K. et al. Magnetoencephalography imaging reveals abnormal information flow in temporal lobe epilepsy. Brain Connect. 12, 362–373. https://doi.org/10.1089/brain.2020.0989 (2022).
Carboni, M. et al. Abnormal directed connectivity of resting state networks in focal epilepsy. NeuroImage Clin. 27, 102336. https://doi.org/10.1016/j.nicl.2020.102336 (2020).
Smith, S. M. et al. Network modelling methods for fmri. Neuroimage 54, 875–891. https://doi.org/10.1016/j.neuroimage.2010.08.063 (2011).
Bagić, A. I. et al. The 10 common evidence-supported indications for meg in epilepsy surgery: An illustrated compendium. J. Clin. Neurophysiol. 37, 483–497. https://doi.org/10.1097/WNP.0000000000000726 (2020).
Rampp, S. et al. Magnetoencephalography for epileptic focus localization in a series of 1000 cases. Brain 142, 3059–3071. https://doi.org/10.1093/brain/awz231 (2019).
Van Mierlo, P., Höller, Y., Focke, N. K. & Vulliemoz, S. Network perspectives on epilepsy using eeg/meg source connectivity. Front. Neurol. 10, 721. https://doi.org/10.3389/fneur.2019.00721 (2019).
Kaminski, M. J. & Blinowska, K. J. A new method of the description of the information flow in the brain structures. Biol. Cybern. 65, 203–210. https://doi.org/10.1007/BF00198091 (1991).
Kamiński, M., Ding, M., Truccolo, W. A. & Bressler, S. L. Evaluating causal relations in neural systems: Granger causality, directed transfer function and statistical assessment of significance. Biol. Cybern. 85, 145–157. https://doi.org/10.1007/s004220000235 (2001).
Babiloni, F. et al. Estimation of the cortical functional connectivity with the multimodal integration of high-resolution eeg and fmri data by directed transfer function. Neuroimage 24, 118–131. https://doi.org/10.1016/j.neuroimage.2004.09.036 (2005).
Blinowska, K. J. Review of the methods of determination of directed connectivity from multichannel data. Med. Biol. Eng. Comput. 49, 521–529. https://doi.org/10.1007/s11517-011-0739-x (2011).
Staljanssens, W. et al. Seizure onset zone localization from ictal high-density eeg in refractory focal epilepsy. Brain Topogr. 30, 257–271. https://doi.org/10.1007/s10548-016-0537-8 (2017).
Dai, Y., Zhang, W., Dickens, D. L. & He, B. Source connectivity analysis from meg and its application to epilepsy source localization. Brain Topogr. 25, 157–166. https://doi.org/10.1007/s10548-011-0211-0 (2012).
Sohrabpour, A., Ye, S., Worrell, G. A., Zhang, W. & He, B. Noninvasive electromagnetic source imaging and granger causality analysis: an electrophysiological connectome (econnectome) approach. IEEE Trans. Biomed. Eng. 63, 2474–2487. https://doi.org/10.1109/TBME.2016.2616474 (2016).
Yedurkar, D. P., Metkar, S. P. & Stephan, T. Multiresolution directed transfer function approach for segment-wise seizure classification of epileptic eeg signal. Cogn. Neurodyn. 18, 301–315. https://doi.org/10.1007/s11571-021-09773-z (2024).
Narasimhan, S. et al. Seizure-onset regions demonstrate high inward directed connectivity during resting-state: An seeg study in focal epilepsy. Epilepsia 61, 2534–2544. https://doi.org/10.1111/epi.16686 (2020).
Hejazi, M. & Motie Nasrabadi, A. Prediction of epilepsy seizure from multi-channel electroencephalogram by effective connectivity analysis using granger causality and directed transfer function methods. Cogn. Neurodyn. 13, 461–473. https://doi.org/10.1007/s11571-019-09534-z (2019).
Sun, Q., Liu, Y. & Li, S. Weighted directed graph-based automatic seizure detection with effective brain connectivity for eeg signals. SIViP 18, 899–909. https://doi.org/10.1007/s11760-023-02816-4 (2024).
Ren, Y. et al. Functional brain network mechanism of executive control dysfunction in temporal lobe epilepsy. BMC Neurol. 20, 1–13. https://doi.org/10.1186/s12883-020-01711-6 (2020).
Li, M. & Zhang, N. A dynamic directed transfer function for brain functional network-based feature extraction. Brain Inform. 9, 7. https://doi.org/10.1186/s40708-022-00154-8 (2022).
Han, L., Song, X. & Li, C. Dynamic analysis of epileptic causal brain networks based on directional transfer function. Sheng wu yi xue Gong Cheng xue za zhi= J. Biomed. Eng.= Shengwu Yixue Gongchengxue Zazhi 39, 1082–1088. https://doi.org/10.7507/1001-5515.202202022 (2022).
Sauer, A. et al. A meg study of visual repetition priming in schizophrenia: Evidence for impaired high-frequency oscillations and event-related fields in thalamo-occipital cortices. Front. Psych. 11, 561973. https://doi.org/10.3389/fpsyt.2020.561973 (2020).
Grent-‘t Jong, T., et al. Acute ketamine dysregulates task-related gamma-band oscillations in thalamo-cortical circuits in schizophrenia. Brain 141, 2511–2526. https://doi.org/10.1093/brain/awy175 (2018).
Sauer, A. et al. Spectral and phase-coherence correlates of impaired auditory mismatch negativity (mmn) in schizophrenia: A meg study. Schizophr. Res. 261, 60–71. https://doi.org/10.1016/j.schres.2023.08.033 (2023).
Wang, B. & Meng, L. Functional brain network alterations in epilepsy: A magnetoencephalography study. Epilepsy Res. 126, 62–69. https://doi.org/10.1016/j.eplepsyres.2016.06.014 (2016).
Niso, G. et al. What graph theory actually tells us about resting state interictal meg epileptic activity. NeuroImage Clin. 8, 503–515. https://doi.org/10.1016/j.nicl.2015.05.008 (2015).
Kitchigina, V. et al. Disturbances of septohippocampal theta oscillations in the epileptic brain: reasons and consequences. Exp. Neurol. 247, 314–327. https://doi.org/10.1016/j.expneurol.2013.01.029 (2013).
Buckner, R. L. et al. Cortical hubs revealed by intrinsic functional connectivity: mapping, assessment of stability, and relation to alzheimer’s disease. J. Neurosci. 29, 1860–1873. https://doi.org/10.1523/JNEUROSCI.5062-08.2009 (2009).
Tomasi, D. & Volkow, N. D. Functional connectivity hubs in the human brain. Neuroimage 57, 908–917. https://doi.org/10.1016/j.neuroimage.2011.05.024 (2011).
Bush, G., Luu, P. & Posner, M. I. Cognitive and emotional influences in anterior cingulate cortex. Trends Cogn. Sci. 4, 215–222. https://doi.org/10.1016/S1364-6613(00)01483-2 (2000).
Bernabei, J. M., Litt, B. & Cajigas, I. Thalamic stereo-eeg in epilepsy surgery: where do we stand?. Brain 146, 2663–2665. https://doi.org/10.1093/brain/awad178 (2023).
Wu, T. Q. et al. Multisite thalamic recordings to characterize seizure propagation in the human brain. Brain 146, 2792–2802. https://doi.org/10.1093/brain/awad121 (2023).
Khambhati, A. N. et al. Dynamic network drivers of seizure generation, propagation and termination in human neocortical epilepsy. PLoS Comput. Biol. 11, e1004608. https://doi.org/10.1371/journal.pcbi.1004608 (2015).
Zhong, L. et al. Temporal and spatial dynamic propagation of electroencephalogram by combining power spectral and synchronization in childhood absence epilepsy. Front. Neuroinform. 16, 962466. https://doi.org/10.3389/fninf.2022.962466 (2022).
Corriveau, A., Yoo, K., Kwon, Y. H., Chun, M. M. & Rosenberg, M. D. Functional connectome stability and optimality are markers of cognitive performance. Cereb. Cortex 33, 5025–5041. https://doi.org/10.1093/cercor/bhac396 (2023).
Liang, X., Zou, Q., He, Y. & Yang, Y. Coupling of functional connectivity and regional cerebral blood flow reveals a physiological basis for network hubs of the human brain. Proc. Natl. Acad. Sci. 110, 1929–1934. https://doi.org/10.1073/pnas.1214900110 (2013).
Van den Heuvel, M. P. & Sporns, O. Network hubs in the human brain. Trends Cogn. Sci. 17, 683–696. https://doi.org/10.1016/j.tics.2013.09.012 (2013).
Liao, X., Vasilakos, A. V. & He, Y. Small-world human brain networks: perspectives and challenges. Neurosci. Biobehav. Rev. 77, 286–300. https://doi.org/10.1016/j.neubiorev.2017.03.018 (2017).
Gao, Y., Su, Q., Liang, L., Yan, H. & Zhang, F. Temporal lobe dysfunction in neuropsychiatric disorder, https://doi.org/10.3389/fpsyt.2022.1077398 (2022).
Verhoeven, T. et al. Automated diagnosis of temporal lobe epilepsy in the absence of interictal spikes. NeuroImage Clin. 17, 10–15. https://doi.org/10.1016/j.nicl.2017.09.021 (2018).
Labache, L., Ge, T., Yeo, B. T. & Holmes, A. J. Language network lateralization is reflected throughout the macroscale functional organization of cortex. Nat. Commun. 14, 3405. https://doi.org/10.1038/s41467-023-39131-y (2023).
Carpentier, A. et al. Functional mri of language processing: dependence on input modality and temporal lobe epilepsy. Epilepsia 42, 1241–1254. https://doi.org/10.1046/j.1528-1157.2001.35500.x (2001).
Springer, J. A. et al. Language dominance in neurologically normal and epilepsy subjects: a functional mri study. Brain 122, 2033–2046. https://doi.org/10.1093/brain/122.11.2033 (1999).
Pugh, K. R. et al. Functional neuroimaging studies of reading and reading disability (developmental dyslexia). Ment. Retard. Dev. Disabil. Res. Rev. 6, 207–213. (2000).
Shaywitz, S. E. et al. Functional disruption in the organization of the brain for reading in dyslexia. Proc. Natl. Acad. Sci. 95, 2636–2641. https://doi.org/10.1073/pnas.95.5.2636 (1998).
Li, Y. et al. Evaluation of brain network properties in patients with mri-negative temporal lobe epilepsy: an meg study. Brain Topogr. 34, 618–631. https://doi.org/10.1007/s10548-021-00856-y (2021).
Bartolomei, F., Bettus, G., Stam, C. & Guye, M. Interictal network properties in mesial temporal lobe epilepsy: a graph theoretical study from intracerebral recordings. Clin. Neurophysiol. 124, 2345–2353. https://doi.org/10.1016/j.clinph.2013.06.003 (2013).
Ponten, S., Bartolomei, F. & Stam, C. Small-world networks and epilepsy: graph theoretical analysis of intracerebrally recorded mesial temporal lobe seizures. Clin. Neurophysiol. 118, 918–927. https://doi.org/10.1016/j.clinph.2006.12.002 (2007).
Bernhardt, B. C., Chen, Z., He, Y., Evans, A. C. & Bernasconi, N. Graph-theoretical analysis reveals disrupted small-world organization of cortical thickness correlation networks in temporal lobe epilepsy. Cereb. Cortex 21, 2147–2157. https://doi.org/10.1093/cercor/bhq291 (2011).
Bernhardt, B. C., Bernasconi, N., Hong, S.-J., Dery, S. & Bernasconi, A. Subregional mesiotemporal network topology is altered in temporal lobe epilepsy. Cereb. Cortex 26, 3237–3248. https://doi.org/10.1093/cercor/bhv166 (2015).
Yasuda, C. L. et al. Aberrant topological patterns of brain structural network in temporal lobe epilepsy. Epilepsia 56, 1992–2002. https://doi.org/10.1111/epi.13225 (2015).
Wang, J. et al. Graph theoretical analysis reveals disrupted topological properties of whole brain functional networks in temporal lobe epilepsy. Clin. Neurophysiol. 125, 1744–1756. https://doi.org/10.1016/j.clinph.2013.12.120 (2014).
Cohen, J. R. & D’Esposito, M. The segregation and integration of distinct brain networks and their relationship to cognition. J. Neurosci. 36, 12083–12094. https://doi.org/10.1523/JNEUROSCI.2965-15.2016 (2016).
Wang, B. et al. Dynamic reconfiguration of functional brain networks supporting response inhibition in a stop-signal task. Brain Imaging Behav. 14, 2500–2511. https://doi.org/10.1007/s11682-019-00203-7 (2020).
Smith, E. H. et al. Dual mechanisms of ictal high frequency oscillations in human rhythmic onset seizures. Sci. Rep. 10, 19166. https://doi.org/10.1038/s41598-020-76138-7 (2020).
Jung, T.-P. et al. Extended ica removes artifacts from electroencephalographic recordings. Adv. Neural Inf. Process. Syst.10 (1997).
Sekihara, K., Nagarajan, S. S., Poeppel, D. & Marantz, A. Performance of an meg adaptive-beamformer technique in the presence of correlated neural activities: effects on signal intensity and time-course estimates. IEEE Trans. Biomed. Eng. 49, 1534–1546. https://doi.org/10.1109/TBME.2002.805485 (2002).
Hillebrand, A. & Barnes, G. R. The use of anatomical constraints with meg beamformers. Neuroimage 20, 2302–2313. https://doi.org/10.1016/j.neuroimage.2003.07.031 (2003).
Smythies, J. R. & Bradley, R. J. International review of neurobiology (Academic Press, 1983).
Hillebrand, A., Singh, K. D., Holliday, I. E., Furlong, P. L. & Barnes, G. R. A new approach to neuroimaging with magnetoencephalography. Hum. Brain Mapp. 25, 199–211. https://doi.org/10.1002/hbm.20102 (2005).
Hillebrand, A., Barnes, G. R., Bosboom, J. L., Berendse, H. W. & Stam, C. J. Frequency-dependent functional connectivity within resting-state networks: an atlas-based meg beamformer solution. Neuroimage 59, 3909–3921. https://doi.org/10.1016/j.neuroimage.2011.11.005 (2012).
López, M. E. et al. Meg beamformer-based reconstructions of functional networks in mild cognitive impairment. Front. Aging Neurosci. 9, 107. https://doi.org/10.3389/fnagi.2017.00107 (2017).
Rolls, E. T., Joliot, M. & Tzourio-Mazoyer, N. Implementation of a new parcellation of the orbitofrontal cortex in the automated anatomical labeling atlas. Neuroimage 122, 1–5. https://doi.org/10.1016/j.neuroimage.2015.07.075 (2015).
Holm, S. A simple sequentially rejective multiple test procedure. Scand. J. Stat. 65–70 (1979).
Acknowledgements
This work was supported by sponsored by Beijing Nova Program [No. 20230484469]; Beijing Nova Program (CN) [No. XX2016120]; National Natural Science Foundation of China [No.31640035], [No.82172022], [No.82471496], [No.12402350]; China Postdoctoral Science Foundation [No.2022M720326]; Jiangsu Province Social Development Project [No.BE2023795]; Beijing Municipal Commission of Education [KM201710005013]; and Natural Science Foundation of Beijing [No. 4162008]. Thanks for Intelligent Physiological Measurement and Clinical Translation, Beijing International Base for Scientific and Technological Cooperation.
Author information
Authors and Affiliations
Contributions
Conceptualization, C.Y. and T.W.; methodology, C.Z. and Y.W.; resources, C.Y., T.W., W.H. and G.L.; writing-original draft preparation, C.Z.; writing-review and editing, Y.W. and C.Y.; visualization, C.Z. and Y.W. All authors have read and agreed to the published version of the manuscript.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Zhang, C., Wu, Y., Hu, W. et al. Frequency-band specific directed connectivity networks reveal functional disruptions and pathogenic patterns in temporal lobe epilepsy: a MEG study. Sci Rep 15, 12326 (2025). https://doi.org/10.1038/s41598-025-90299-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-025-90299-3
