Frontal lobe epilepsy (FLE) is the second common type of focal epilepsy behind temporal lobe epilepsy, accounting for ~20%–30% of sufferers (Manford et al., 1996). As the largest lobe of neocortex, the highly interconnected nature of the frontal lobe allows for a quick and widespread propagation of epileptic activity to the other brain regions, which may give rise to the perplexing clinical and electrophysiological finding of FLE. A variety of semiologies are common for FLE patients, such unilateral clonic seizures, tonic asymmetric seizures with preserved consciousness, hypermotor seizures, and secondary generalized seizures. The ambiguity in localization and lateralization of deficits of EEG is well known, even false negative EEG is not uncommon (Beleza and Pinho, 2011). Seizures in FLE are mostly likely to be associated with multi-cognitive defects and motor-related abnormality networks (Kellinghaus and Lüders, 2004; Beleza and Pinho, 2011), and the treatment outcome is disappointing, as only 20%–30% of patients achieve seizure freedom with medication (Regesta and Tanganelli, 1999). Of all patients with refractory focal epilepsies referred to epilepsy surgery, 25% have FLE, and only 30%–50% achieve seizure freedom with surgery (Bagla and Skidmore, 2011). As epileptogenic zone may be subtle and not obvious in routine MRI examination (Bonini et al., 2014), it presents challenges in FLE patients with ambiguous EEG pattern, and calls for better clarification of the underlying neuroanatomic characteristic of FLE, especially for patients with normal routine MRI examination.

As a brain network disorder, epilepsy has been widely studied using quantitative neuroimaging data, which supported that epileptogenic network are involved in the generation and expression of seizures, and to the maintenance of the disorder (Spencer, 2002). In contrast to make low-level regional and connectional alterations with conventional approach, graph-theory analysis provides a correlational framework to reveal the persistent functional-trophic cross-talk, maturational inter-change, as well as common developmental and pathological influences (Bullmore and Sporns, 2012; Alexander-Bloch et al., 2013; Bernhardt et al., 2013), and can be used for various modalities of neuroimaging data. Normal topological network is characterized by high clustering coefficients and short average path lengths. According to previous graph-theory based fMRI and diffusional MRI (dMRI) studies (Vaessen et al., 2013, 2014; Gleichgerrcht et al., 2015; Zhou et al., 2019; Lin et al., 2020; Togo et al., 2022), rearrangement of global and local topological parameters has been found in patients of focal epilepsy (including FLE and TLE), such as diminished network strength, clustering coefficient, path length, and global efficiency, both within and beyond the epileptogenic zone.

Unlike fMRI and dMRI, structural T1-weighted images is a standard component of every clinical imaging protocol with short acquisition time. These images are generally unaffected by distortion and signal drop out artifacts in orbitofrontal and temporo-basal regions which often occur in echo-planar functional and diffusion MRI sequences (Bernhardt et al., 2013; Gleichgerrcht et al., 2015). Graph theory based structural covariance network (SCN) of cortical thickness directly seeds from cortical gray matter regions in a high-resolution space, which is not limited by the imaging voxels but by the sampling density of the points on the cortical mesh (Hosseini et al., 2012; Alexander-Bloch et al., 2013). SCN analysis has been applied to investigate in various central nervous system disorders (Hosseini et al., 2012; Singh et al., 2013), and considered a promising tool in investigating the brain network alterations in epilepsy (Li et al., 2020), but it was rarely used in FLE patients. In order to investigate the topological property alteration of cortical thickness in FLE, SCN analysis was performed between a group of clinically diagnosed FLE patients with negative MRI and a age-, gender-, education-matched healthy control (HC) group in the present study.

This study investigated the topological property alteration of cortical thickness based SCN in FLE patients without structural abnormality on conventional MR. Both the global and regional network parameters were significantly different between FLE and HC groups. The results revealed that the structural network properties were significantly changed in MRI-negative FLE patients, suggesting higher sensitivity of graph-theory based analysis in detecting the neurobiological injury in FLE patients. To the best of our knowledge, the present study reported for the first time the cortical-thickness related topological property alteration in MRI-negative FLE patients and it may be a valuable tool in future clinical practice.

The structural alterations of FLE were not consistent in previous studies. In the study of Widjaja et al. (2011), widespread cortical thinning was found, while in other studies, there was no significant difference in cortical /GM volume between FLE and HC (Lu et al., 2022). In the present study, cortical thickness was not significantly different between the two groups. However, our study demonstrated that FLE patients had significantly different topological change for both global and regional network parameters, though both groups showed a small-world topology. Consistent with previous functional graph-theory based neuroimaging studies of epilepsy (Vaessen et al., 2013), the path length and node betweenness were significantly increased in FLE, indicating the global network topology was altered toward a regularized pattern. However, variations in structural network were also found. In the study of Vaessen et al. (2013), the structural path length and clustering remained normal in children with FLE, structural modularity was different between two groups and it was increased with stronger cognitive impairment. Thus modularity may be associated with cognitive status, which was not included in present study and warrants further investigation. In addition to methodological approach difference (e.g., different atlases and morphometric features), patient selection variation may also contribute to the discrepancy. For the study of Vaessen et al. (2013, 2014) and Widjaja et al. (2011), the patients recruited were children, but the patients in the present study were adults older than 14 years old. As cortical structure should be closely associate with age, and the variation during development and maturation of frontal lobe may be huge even in normal population (Silver et al., 2021; Herring et al., 2022), future studies with larger sample size is warranted to clarify the influence of age.

In FLE, the clinical symptoms of these seizures are variable and dependent on the brain regions, which may include peri-rolandic, supplementary sensorimotor area, dorsolateral frontal, orbitofrontal, anterior frontopolar, opercular, and cingulate types (Bagla and Skidmore, 2011). But most FLE patients usually are recognized as multi-cognitive defects and motor-related networks defects (Kellinghaus and Lüders, 2004; Beleza and Pinho, 2011). In the present study, all the patients were clinically diagnosed FLE, with the prominent seizure characteristics of hypermotor attacks and secondary generalized clonic–tonic seizure. This is in agreement with previous report on FLE (Bagla and Skidmore, 2011; Bonini et al., 2014) and we speculated that motor-related network abnormality should be found. In agreement with this speculation, both normalized clustering and local efficiency of right precentral gyrus were significantly increased, indicating an abnormal increase in network segregation in sensorimotor network. Similar findings were reported in previous study of FLE (Woodward et al., 2014).

Connections between temporal lobe and sensorimotor cortex have been repeatedly reported in previous neuroimaging studies of MRI-negative TLE patients, especially for right MTS to influence ipislateral thalamus and temporal pole (Coan et al., 2014; Roggenhofer et al., 2020). In addition to TLE, temporal pole and precentral gyrus involvement is also reported for patients with GTCS (Liu et al., 2017; Li et al., 2020). Thus, it is not surprising that regional topological parameters were also compromised for right temporal pole in the study. As a seizure is typically the result of the networks that are recruited or traveled by the epileptiform discharges, therefore seizure-onset localization can be misled by clinical manifestations that may arise from recruited networks that are remotely located from the ictal source. The decreased clustering and local efficiency of right temporal pole may reflect the intrinsic network pattern of fronto-limbic system.

Similarly, network hub analysis showed that there were significant difference between FLE and HC groups. In HC group, the hubs were evenly distributed around the neocortex, including bilateral precuneus, left inferior temporal gyrus and left medial orbitofrontal gyrus, which was consistent with DMN network hubs (Zhou et al., 2019). However, for FLE patients, the hubs were reduced and shifted posterior and none was found in frontal lobes. The hubs identified in FLE group, the inferior parietal lobule which is usually closely connected with precuneus (Togo et al., 2022). Together with supramarginal gyrus, IPL is a region implicated in a diverse range of higher cognitive functions and may be associated with multiple brain networks, including DMN, Frontoparietal control network, and cingulo-opercular network (Igelström and Graziano, 2017). This hub alteration pattern may reflect the DMN abnormality in FLE, which was also reported in previous fMRI studies of epilepsy (Lin et al., 2020; Togo et al., 2022). The mechanism was unknown, but it may result from the interrupted small-world properties within frontal lobe in FLE, as noted by the disappearance of network hub in frontal lobe in FLE.

In addition, the hubs found in FLE, were among regions of reduced cortical thickness reported by Widjaja et al. (2011), in a study of FLE children. Similarly, in the functional connectivity study of FLE patients by Wu et al. (2019), increased ALFF in the precuneus of FLE patient, despite their response to antiepileptic medication, and these were the hubs for HC group in the present study. Inferior parietal lobules were also found to be affected in patients with hyperkinetic seizures (Sasagawa et al., 2021). The clinical significance of the network hub alteration remains unknown and warrants further investigation.

The present study has several limitations. Firstly, the FLE patients were clinically diagnosed, which was based on seizure semiology and EEG; while magneto encephalography, FDG-PET, and invasive intracranial monitoring were unavailable for the patients. Thus the possibility of multiple origin other than frontal lobe could not be completely excluded. In addition, most of the patients could not be lateralized, thus it is not clear whether the laterality affected the cortical thickness and brain volume distribution. Also, the sample size was relatively small with various illness duration, and the medicine used was not the same among subjects. These would have potential effects on the topological results, and future studies with larger sample size and better homogeneity of FLE patients may provide further insights. Secondly, cross-sectional design was used in the present study which brings difficulty to get a causal conclusion. Longitudinal design may be needed in the future to further confirm our finding and assess whether the changes of network graph properties is the consequence of seizures.

In summary, the present study investigated topological properties of cortical thickness covariance network alteration in patients with FLE using the graph theory method. Both global and regional network parameters were significantly different between patients with FLE and normal controls, despite the surface-based cortical thickness was not significantly different between two groups. These results indicated that graph-theory based structural covariance network analysis may provide clues to reveal the structural alterations in MRI-negative FLE.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Ethics Committee of the Third Xiangya Hospital, Central South University. The patients/participants provided their written informed consent to participate in this study.

YiL, WW, DL, and ZS designed the study. QL, DY, QZ, JD, YH, and HH collected the clinical and imaging data. DY, QL, and LD analyzed the data. YiL, QL, DY, DL, LD, and WW co-wrote the manuscript. All authors contributed to the article and approved the submitted version.

This study was supported by Hunan Provincial Natural Science Foundation (2022JJ30890) and Natural Science Foundation of Changsha City (No. kq2208363).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.