Brain structural differences in temporal lobe and frontal lobe epilepsy patients: A voxel-based morphometry and vertex-based surface analysis

Abstract

Purpose

Exploration of the effect of chronic recurrent seizures in focal epilepsy on brain volumes has produced many conflicting reports. To determine differences in brain structure in temporal lobe epilepsy (TLE) and extratemporal epilepsy (using frontal lobe epilepsy (FLE) a surrogate) further, we performed a retrospective analysis of a large cohort of patients with seizure-onset zone proven by intracranial monitoring.

Methods

A total of 120 TLE patients, 86 FLE patients, and 54 healthy controls were enrolled in this study. An analysis of variance of voxel-based morphometry (VBM) was used to seek morphometric brain differences among TLE patients, FLE patients, and healthy controls. Additionally, a vertex-based surface analysis was utilized to analyze the hippocampus and thalamus. Significant side-specific differences in hippocampal gray matter volume were present between the left TLE (LTLE), right TLE RTLE (RTLE), and control groups (p<0.05, family-wise error (FWE) corrected).

Results

Vertex analyses revealed significant volume reduction in inferior parts of the left hippocampus in the LTLE group and lateral parts of the right hippocampus in the RTLE group compared to controls (p<0.05, FWE corrected). Significant differences were also detected between the LTLE and control group in the bilateral medial and inferior thalamus (p<0.05, FWE corrected). FLE patients did not exhibit focal atrophy of gray matter across the brain.

Conclusion

Our results highlight the variation in morphometric lateralized changes in the brain between different epilepsy onset zones, providing critical insight into the natural history of people with drug-resistant focal epilepsies.

Introduction

Temporal lobe epilepsy (TLE) is the most common drug-resistant epilepsy in adults, and surgery can potentially be curative, with higher seizure-free rates in patients who have hippocampal sclerosis. ¹ In addition to reduced hippocampal volume, extrahippocampal gray matter (GM) volume changes have also been reported in TLE patients, but the results have been inconsistent.^2,3 A meta-analysis of TLE patients found consistent volume loss in parts of the thalamus and pallidum, but pointed out that studies with larger cohorts and focus on extrahippocampal regions are needed.^4–6 With increasing utilization of minimally invasive techniques in the treatment of drug-resistant epilepsy, such as thalamic deep brain stimulation (DBS), responsive neurostimulation, and laser interstitial thermal therapy (LITT), an improved understanding of how epilepsy affects volume and shape changes in the brain is of paramount importance. For instance, Grewal et al. ⁷ showed that the degree of inaccuracy in traditional indirect targeting of the anterior nucleus of the thalamus in DBS was related to regional atrophy and morphometric changes in patients with epilepsy.

Magnetic resonance imaging (MRI) voxel-based morphometry (VBM) studies have been widely used to investigate the regional brain volume alterations in TLE patients.^8–10 These studies suffer from limitations of small sample size and have produced conflicting results. For example, some studies found more extensive gray matter volume loss or ipsilateral hippocampus atrophy in left TLE (LTLE) patients,^11–13 while other studies reported the opposite results.^14,15 In addition to VBM analysis, surface-based vertex analysis can localize volumetric reduction along the surface of the segmented subcortical nuclei to characterize their subcomponents that contributed most to the group volume difference. This method has previously been used to evaluate morphometric differences in other neurologic conditions and provides complementary information regarding volumetric changes in subcortical nuclei compared to VBM alone.^16–18 Vertex-based shape analysis studies in TLE patients are needed to understand the subcortical structural changes in epilepsy patients better. ⁵

While TLE has been widely studied, the literature on brain morphometry studies in patients with extratemporal epilepsy (ETLE) is limited. An early study comparing 31 ETLE patients and 15 healthy controls found no difference in volumes of the subcortical nuclei between the two groups. ¹⁹ Moreover, few studies have investigated the brain morphometry differences between TLE and ETLE patients. ²⁰ Given the small number of ETLE studies and relatively small sample size of these studies, MRI-based brain morphometry studies of ETLE patients with larger sample sizes are warranted.

Given the paucity of comprehensive studies evaluating both TLE and ETLE patients compared to healthy controls, we combined VBM analysis and surface-based vertex analysis among a large cohort of TLE patients, FLE patients, and healthy controls to gain more insight into brain structural changes in these groups. In order to create a homogenous group of ETLE patients, we focused specifically on frontal lobe epilepsy (FLE), the second most common form of adult epilepsy. ²¹

Methods

Participants

This population-based observational study used a retrospective design and was approved by the Mayo Clinic Institutional Review Board. We reviewed the Mayo Clinic electronic medical record to identify patients with a diagnosis of epilepsy who had undergone evaluation by a multidisciplinary epilepsy team and had Phase II intracranial monitoring to confirm the seizure-onset zone. Additional inclusion criteria included (a) unilateral seizure onset in the mesial temporal lobe or frontal lobe, (b) the presence of a high-resolution volumetric T1-weighted MRI before surgery, and (c) age between 15 and 60 years. The age range was chosen to minimize the effect of aging and development, as well as the increasing incidence of nonepilepsy-related hippocampal pathology in older adults. Patients with major anatomical distortions, such as prior brain surgery or other significant brain abnormality (e.g., large infarction, tumor, etc.), were also excluded. A total of 206 epilepsy patients (120 TLE patients and 86 FLE patients) and 54 healthy controls were enrolled. To maintain homogeneity, only ETLE patients with frontal seizure onset were included due to the higher incidence, as well as the greater likelihood of the frontal lobe being the extratemporal region to affect the limbic network. Demographic and clinical data, such as duration of epilepsy, seizure frequency, seizure type, seizure onset zone, and pathology of epilepsy, were extracted from the medical record. Since FLE patients had a lower mean age than TLE patients (Supplemental Tables S1 and S2), TLE patients and controls > 50 years of age were not included in the statistical analysis when comparing to FLE patients (age balanced). The TLE and FLE groups were each divided into two subgroups according to the laterality of seizure onset: right TLE (RTLE), LTLE, right FLE (RFLE), and left FLE (LFLE). A cohort of age- and sex-matched healthy controls were selected from the Autism Brain Imaging Data Exchange (ABIDE)^22,23 database (http://fcon_1000.projects.nitrc.org/indi/abide/) and Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (http://adni.loni.usc.edu).

MRI preparation

T1-weighted volumetric spoiled gradient-recalled or magnetization prepared rapid gradient echo images with were used for VBM analysis. All images were acquired with a voxel size <1 mm×1 mm×1.2 mm.

Imaging data processing

The DARTEL method in Statistical Parametric Mapping v12 (http://www.fil.ion.ucl.ac.uk/spm) was used for the segmentation and normalization of the T1 images. The T1 images of the head were first checked for sample homogeneity to discard poor quality images, and then an interactive reorientation was applied. The bias-corrected images were segmented into GM, white matter (WM), and cerebrospinal fluid, and were normalized to standard Montreal Neurological Institute space. The modulated GM images (gray matter volume/amount) were smoothed with a Gaussian kernel of 8 mm full width at half maximum and selected for the VBM statistical analysis.

Volumetric and vertex-based analyses were performed using FMRIB’s Software Library (FSL v6.0; http://www.fmrib.ox.ac.uk/fsl). Automated segmentation of the subcortical nuclei was conducted using the “fsl_anat” command in FSL, which provides a general pipeline for processing T1-weighted brain images (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/fsl_anat). The volumes of subcortical structures were measured using the “fslstats” command. Two versions of surface-based vertex analysis (new and traditional) were adopted in this study to characterize the portion of the subcortical nuclei that significantly contributed to the global volume alteration. The “randomise” function was used in new surface-based vertex analysis to calculate vertex-wise F statistics in the subcortical nuclei. In traditional surface-based vertex analysis, false discovery rate correction is the only available method of multiple-comparison correction for surface-based statistical output.

Statistical analysis

A voxel-wise analysis of variance (ANOVA) of the GM volume difference across the whole brain was performed using the smoothed GM volume among TLE, FLE, and control groups in REST toolkit v1.8 (http://www.restfmri.net). The same analysis was also performed to detect GM volume difference among subgroups of epilepsy groups and the control group. Multiple-comparison corrections were performed using family-wise error (FWE) rate correction with a cluster defining threshold of p=0.001 (two tail), corresponding to a cluster level of p=0.05. ²⁴ Post hoc t-tests with Bonferroni correction were used to compare the intergroup differences within the mask of the ANOVA results. The average GM volume of the hippocampus and thalamus were extracted for region of interest (ROI) analysis using the mask of hippocampus and thalamus in the automated anatomical atlas. ²⁵

Demographic data were analyzed using PASW Statistics for Windows v18.0 (SPSS, Inc., Chicago, IL). Paired t-tests were used to detect the volume and GM volume difference between the left and right side of the hippocampus and thalamus in epilepsy groups, as well as the control group. Intergroup comparison of volume and GM volume were also performed using ANOVA. To assess the volume asymmetry of the hippocampus and thalamus caused by side-specific atrophy, we calculated the laterality index (LI), defined as the ratio {[(left–right)/(left+right)]/2} for the hippocampus and thalamus. ¹⁷ The LIs were also compared among TLE subgroups and the control group using ANOVA. p-Values <0.05 were considered statistically significant.

Pearson correlations were utilized to explore the relationship between the clinical data (duration of epilepsy and seizure frequency) and volumetric data. Multiple correlations were corrected by the Bonferroni correction.

Results

VBM analysis of TLE, FLE, and control groups

The demographic and clinical data of the three groups are summarized in Supplemental Table S1. There were no significant differences in age or sex among the three groups. There was no difference in the volume of total GM among the three groups. Patients in the TLE group, however, had a significantly lower volume of total WM and brain parenchyma compared to controls.

Using a FWE rate correction with a cluster defining threshold of p=0.001 (two tail, corresponding to a cluster level of p=0.05) as a multiple-comparison correction method, no significant difference was found in the voxel-wise analysis of regional GM volume among the three groups.

VBM analysis of LTLE patients, RTLE patients, and controls

The demographic and clinical data of the three groups are summarized in Supplemental Table S2. Compared to the control group, decreased GM volume was present in the left hippocampus in the LTLE group and right hippocampus in the RTLE group, but decreased GM volume also extended to the parahippocampal regions (p<0.05, corrected, Figure 1(a)–(c)). Meanwhile, the RTLE group had a lower GM volume in the right hippocampus and a higher GM volume in left hippocampus compared to the LTLE group (Figure 1(d)). There was also a more extensive GM volume reduction in the hippocampus and parahippocampal areas of the RTLE group than the LTLE group (Figure 1).

Figure 1.

Analysis of variance (ANOVA) of left temporal lobe epilepsy (LTLE), right temporal lobe epilepsy (RTLE), and control groups. (a) ANOVA among LTLE, RTLE, and control groups. Red area indicates significant difference of gray matter volume among three groups (p<0.05, GRF corrected). (b) Post hoc analysis of RTLE and control groups. Blue area indicates significant lower gray matter volume in the RTLE group compared to the control group (p<0.05, corrected). (c) Post hoc analysis of LTLE and control groups. Blue area indicates significant lower gray matter volume in the LTLE group compared to the control group (p<0.05, corrected). (d) Post hoc analysis of RTLE and LTLE groups. Red area indicates significant higher gray matter volume in the RTLE group compared to the LTLE group, and blue area indicates significant lower gray matter volume in the RTLE group compared to the LTLE group (p<0.05, corrected).

Analysis of hippocampal and thalamic volume showed significant volume differences between the right and left side within all three groups (Figure 2(a) and (b)). The right hippocampal volume is larger than that of the left side in the control group (p<0.05). The RTLE group has a smaller right hippocampus than left hippocampus, while the LTLE group has a smaller left hippocampus than right hippocampus (p<0.001). There was a significantly decreased right hippocampus volume in the RTLE group and a decreased left hippocampus volume in the LTLE group compared to the control group (p<0.001; Supplemental Table S2). The left thalamus was significantly larger than the right in all three groups. The LTLE group had a smaller left thalamic volume compared to controls (p<0.05).

Figure 2.

Volumetric analysis of the hippocampus and thalamus of LTLE, RTLE, left frontal lobe epilepsy (LFLE), right frontal lobe epilepsy (RFLE), and control groups. (a) Thalamic volume difference within and among LTLE, RTLE, and control groups. (b) Hippocampal volume difference within and among LTLE, RTLE, and control groups. (c) Difference of laterality index (LI) of thalamus among LTLE, RTLE, and control groups. (d) Difference of LI of the hippocampus among LTLE, RTLE, and control groups. (e) Thalamic volume difference within and among LFLE, RFLE, and control groups. (f) Hippocampal volume difference within and among LFLE, RFLE, and control groups. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001.

There were significant differences in the LI of the hippocampus between any two groups among RTLE, LTLE, and control groups (p<0.05; Figure 2(c)). Positive and negative values of the hippocampal LI showed a rightward and leftward asymmetry of the hippocampal volume in the RTLE and LTLE groups, respectively. A smaller thalamic LI in the LTLE group was also observed compared to the control and RTLE groups (p<0.05; Figure 2(d)). Receiver operating characteristic curve analysis showed an area under the curve of 0.79 for LI to differentiate the RTLE and LTLE patients (Supplemental Figure S1).

VBM analysis of LFLE patients, RFLE patients, and controls

The demographic and clinical data of the three groups are summarized in Supplemental Table S3. No significant difference was found in the voxel-wise analysis of regional GM volume among the three groups. In the volumetric analysis, no significant difference of thalamic volume or hippocampal volume was found between any two groups. The left thalamic volume was significantly larger than the right side in all three groups (p<0.05; Figure 2(e)). No significant volume difference was found between the left and right hippocampus in the RFLE and LFLE groups (p<0.05; Figure 2(f)).

Correlation analysis

The mean GM volumes of the hippocampus and thalamus were extracted to perform a correlation analysis with the clinical data in the TLE groups. Both the total volume and GM volume of the right hippocampus were negatively correlated with the disease duration of the RTLE patients (p<0.05, corrected; Figure 3(a) and (b)). Likewise, both the volume and GM volume of the left hippocampus were negatively correlated with disease duration of the LTLE patients (p<0.05, corrected; Figure 3(c) and (d)). After controlling for age and seizure frequency, these correlation coefficients remained statistically significant (r=–0.328, p=0.026; r=–0.301, p=0.036; r=–0.306, p=0.011; and r=–0.302, p=0.012; uncorrected, for right hippocampal volume, right hippocampal GM volume, left hippocampal volume, and left hippocampal GM volume, respectively).

Figure 3.

Correlation between hippocampus volumetric data and disease duration. (a) and (b) Both volume and gray matter volume of the right hippocampus were correlated with the disease duration of epilepsy patients in the RTLE group. (c) and (d) Both volume and gray matter volume of the left hippocampus were correlated with the disease duration of epilepsy patients in the LTLE group.

Vertex analysis

Surface-based vertex analyses (new version in FSL v5.0.0 and traditional) were performed between TLE groups and the control group (Supplemental Table S2). New vertex analyses revealed significant volume reduction in inferior parts of the left hippocampus in the LTLE group and lateral parts of the right hippocampus in the RTLE group compared to the control group (p<0.05, FWE corrected; Figure 4(a) and (b)). Meanwhile, significant differences of the vertex analysis were mainly detected in certain medial and inferior parts of both sides of thalamus between the LTLE and control groups (p<0.05, FWE corrected; Figure 5). The traditional vertex analyses showed significant group differences in lateral parts of right hippocampus in the RTLE group (Figure 6), and no difference was detected after multiple-comparison correction in the LTLE group. Traditional vertex analyses also showed significant group differences in both sides of the thalamus between the TLE and control groups (Figure 6). The same analysis when performed between FLE groups and the control group revealed no significant differences after multiple-comparison correction.

Figure 4.

Surface-based vertex analysis of hippocampus between TLE groups and the control group. (a) Significant group difference between the LTLE group and the control group (orange area) was found on the inferior parts of the hippocampus. (b) Significant group difference between the RTLE group and the control group (orange area) was found on the lateral parts of the hippocampus.

Figure 5.

New surface-based vertex analysis of thalamus between the RTLE group and control group. Significant group differences between the RTLE group and control group were found on both sides of the thalamus. The statistical difference was observed mainly on the medial and inferior parts of the left thalamus and anterior, medial, superior, and inferior portions of the right thalamus.

Figure 6.

Traditional surface-based vertex analysis results of the thalamus and hippocampus. Axial projection of shape alteration in LTLE (a) and RTLE (b) patients. Coronal projection of shape alteration in LTLE (c) and RTLE (d) patients.

Discussion

We present the largest single-center cohort with combined voxel-wise analysis, ROI volumetric analysis, and vertex analysis of TLE, FLE, and control patients to assess morphometric brain changes comprehensively. Our study is also unique in that our patients had confirmation of seizure-onset zone with a combination of standard presurgical evaluation and subsequent intracranial monitoring, which allows more accurate classification of patients with “MRI-negative epilepsy.” In the TLE cohort, significant side-specific differences in hippocampal GM volume were observed between the LTLE and RTLE groups and the control group. Vertex analysis shows that regional shape changes primarily affected the CA1 and subiculum regions of the hippocampus on the ipsilateral side at seizure onset. Meanwhile, significant alterations in thalamic shape change were identified, consistent with regional atrophy, which was not reflected in the total thalamic volume, as only LTLE patients had a significant total volume loss compared to controls in the ipsilateral thalamus. These volumetric and shape changes may have critical implications for diagnosis and surgical treatment of epilepsy.

Side-specific atrophy of the medial temporal lobe in TLE patients has been extensively explored. While atrophy ipsilateral to the side of onset has been a common feature, as in our results, there have been other inconsistencies.^2,5,6,14 For instance, studies have shown greater ipsilateral hippocampal atrophy in LTLE patients compared to RTLE,^11–13 while others have found the opposite.^14,15 However, one of the largest previous studies from the Enhancing Neuro Imaging Genetics through Meta-Analysis (ENIGMA) Consortium, showed a larger effect size in ipsilateral hippocampus in the RTLE group than the LTLE group, ⁴ which is consistent with our findings. In contrast, the results of these studies may be driven by the small sample size, lack of intracranial electroencephalogram localization of seizure onset, different statistical approaches, or variation in patient cohorts due to local referral and management patterns. However, larger studies seem to concur on greater ipsilateral atrophy in RTLE compared to LTLE.

Previous studies have also assessed for volume loss in TLE patients beyond the mesial temporal lobe but have found inconsistent results.^4–6 Importantly, a prior meta-analysis of extrahippocampal gray matter volume loss failed to find significant regional cerebral gray matter atrophy beyond the mesial temporal lobe, with the exception of the thalamus. ⁶ In agreement with this meta-analysis, we failed to find regional gray matter atrophy in RTLE or LTLE cohorts beyond the mesial temporal lobe and thalamus. Differences in results compared to a few prior studies may be due to their smaller sample sizes or due to a relatively rigorous multicomparison correction performed in our study. Alternatively, selection bias in prior studies may also be an important difference. More rigorous seizure-onset zone determination was employed in our cohort through intracranial monitoring, which allowed inclusion of many MRI-negative subjects. Existing large studies, such as the ENIGMA study, ⁴ had the inclusion criteria of radiographically confirmed mesial temporal sclerosis for TLE, which leads to considerable selection bias for cases that may be more advanced or unrepresentative of the true spectrum of TLE patients.

In the FLE group, VBM and volumetric analysis showed no group difference in volume measurements. In this study, we only included ETLE patients with ictal onset zone confirmed by intracranial monitoring in the frontal lobe to create a homogenous group. However, we still obtained negative results across the cerebral cortex and subcortical nuclei between the FLE and control groups. These results are in agreement with previous literature and suggest that extratemporal lobe epilepsy does not have a consistent network involvement across patients, even with relatively similar lobar-onset zones. ¹⁹

In the correlation analysis, side-specific atrophy of the hippocampus correlated with the disease duration of the TLE groups and remained statistically significant after controlling for age and seizure frequency. This result has also been reported by many studies ⁶ and indicates the correlation is a consistent finding in TLE. Meanwhile, the thalamus measurements correlated neither with the disease duration nor with seizure frequency. We also performed a voxel-wise correlation analysis between GM volume and disease duration, and the results showed the hippocampus was the only region correlated with disease duration (Supplemental Figure S5).

Both traditional VBM and new vertex-based analysis were adopted to separate regional shape changes and atrophy from differences in total volume changes. Since these two vertex-based analyses use different approaches to statistical analysis, some differences in the final results are anticipated, as also shown in the study by Mao et al. ¹⁷ Both traditional and new vertex-based analyses demonstrate a regional surface contraction in the CA1 region (lateral parts) of the ipsilateral hippocampus in RTLE patients. Additionally, significant regional shape changes were identified in the thalamus of both RTLE and LTLE patients, despite significant total volume difference from controls existing only in the left thalamus in the LTLE group. These findings suggest the vertex-based shape analysis is more sensitive than VBM for detecting thalamus structural change in TLE patients. Such regional shape changes have important implications for interventions such as indirect targeting of the thalamus for DBS in people with drug-resistant focal epilepsy. ⁷

Several limitations of our study are noteworthy. First, by including only patients with intracranial monitoring to increase the accuracy of seizure localization, this may create a selection bias toward more patients with atypical semiology or exclude a greater number of patients with advanced MTS. Second, the heterogeneity in imaging sequences, field strength, and vendor may introduce additional bias. In order to reduce this bias in our cohort, multiple controls cohorts were used to reduce between-group bias. Additionally, volumetric differences across different vendors and field strengths have been shown to be comparable to within-scanner variance. ²⁶ Given the small effect size of these scanner factors, our large cohort of patients should result in negligible scanner-related bias.

Conclusions

From voxel-wise analysis, ROI volumetric analysis, and vertex analysis, our study provides a comprehensive overview of morphometric brain change in TLE and FLE patients. We show side-specific atrophy of the hippocampus in TLE that is more pronounced in RTLE patients. Additionally, shape alterations in the thalamus—indicative of regional atrophy—were more apparent than changes in total thalamic volume in TLE patients, which may have ramifications for subcortical targeting of intracranial electrodes for neuromodulation. Lastly, FLE patients do not exhibit prominent side-specific atrophy of gray matter across the whole brain as seen in TLE patients. The variable atrophy and shape alterations in the hippocampus and thalamus in various epilepsy subtypes are critical to recognize, as they may have substantial impact on diagnosis and treatment, such as with stereotactic targeting in neuromodulation or LITT.