Cortical Changes in Epilepsy Patients With Focal Cortical Dysplasia: New Insights With T2 Mapping

In epilepsy patients with focal cortical dysplasia (FCD) as the epileptogenic focus, global cortical signal changes are generally not visible on conventional MRI. However, epileptic seizures or antiepileptic medication might affect normal‐appearing cerebral cortex and lead to subtle damage.

patients, suggesting that patients with FCD exhibit focal rather than global pathological changes.
However, volumetric magnetic resonance imaging (MRI) studies have indicated global structural abnormalities in epilepsy patients, demonstrating multifocal gray matter (GM) volume changes in patients with malformations of cortical development other than FCD 4 and generalized volume loss in 41% of the patients with chronic epilepsy. 5 Apart from the adverse effects of antiepileptic medication such as neurotoxic side effects, 6,7 long-standing epileptic activity might affect the cortex, 8,9 leading to microstructural damage. Such cortical changes might, in addition to incomplete resections, at least in part, explain why approximately half of patients are not seizure-free after FCD resection. 10 Previous volumetric studies in epilepsy 4,5,11 employed conventional MRI techniques that allow for quantification of volume changes such as atrophy but do not assess the underlying microstructural abnormalities. In contrast, quantitative MRI (qMRI) techniques measure actual tissue parameters, such as the T 2 relaxation time, reducing effects of the scanner hard-and software. 12 T 2 relaxometry allows for the assessment of diffuse or inconspicuous changes in tissue architecture, such as abnormalities in relative myelin, iron, or free water content. 12, 13 Bernasconi et al observed hippocampal T 2 changes in temporal lobe epilepsy (TLE), even in patients showing no signs of atrophy. 14 In addition, qMRI data can help to distinguish between patients with TLE and healthy subjects. 15,16 Furthermore, hippocampal profiling using volumetry and T 2 values aids to spatially localize hippocampal MRI abnormalities. 17 Reeves et al. 18 described T 1 and T 2 * differences between the FCD region of interest in white matter (WM) and normal-appearing WM and magnetization transfer ratio changes in cortical GM inside FCDs, indicating microstructural abnormalities; however, the normal-appearing cortex was not investigated for abnormalities.
The main purpose of our study was to investigate T 2 values in normal-appearing cortical tissue in patients with FCD, hypothesizing that pathologic tissue changes exceed FCD areas in epilepsy patients.

Participants
This study was approved by the local Ethics Committee, and participants gave written informed consent before participation. The study was conducted according to the principles expressed in the Declaration of Helsinki.
Inclusion criteria of the study were: patients with epilepsy and neuroradiologically diagnosed FCD and age-and sex-matched healthy subjects. Exclusion criteria were: other neurological or psychiatric disorders, MRI contraindications, uncontrolled arterial hypertension, or diabetes mellitus.
Sixteen patients (13 male, age: mean AE SD: 27.6 AE 10.3 years) and 16 matched healthy subjects (13 male, age: mean AE SD: 27.3 AE 9.4 years) were investigated. Recruitment was performed in 2018 and 2019 at Goethe University Hospital. The presence or absence of the typical transmantle sign was evaluated based on conventional fluid-attenuated inversion recovery (FLAIR) data (acquisition parameters as detailed below in the section "MRI acquisition and T 2 mapping") for each subject by an experienced neurologist (10 years of experience) and by a senior neuroradiologist specialized in epilepsy imaging (15 years of experience), making decisions by consensus. The number of seizures during a period of 3 months before data acquisition and the number of anticonvulsive drugs in the medical history were obtained anamnestically. In case patients underwent resection, pathology reports were reviewed. In addition, the report of the last electroencephalography (EEG) recording before MRI data acquisition was taken into account for each patient in order to correlate EEG findings with FCD locations. The study presented here is part of a larger prospective scientific project addressing different research questions. Accordingly, the group of patients and healthy control subjects and the acquired data overlap in part with those of previous studies with different aims, presenting novel methods for FCD detection or creating improved synthetic T 1 -weighted datasets. [19][20][21]

MRI Acquisition and T 2 Mapping
Data acquisition was performed using a 3T MR scanner (Magnetom Trio, Siemens Healthineers, Erlangen, Germany). The system utilizes a body coil for radiofrequency (RF) transmission and an 8-channel phased-array head coil for signal reception.
Custom-built programs were used for data analysis, employing functions included in MatLab (MathWorks, Natick, MA), FreeSurfer (Athinoula A. Martinos Center for Biomedical Imaging, Boston, MA), 22 and the FMRIB Software Library (FSL, Oxford, UK). 23 For voxelwise measurement of the T 2 relaxation time, four fast spin echo datasets with different echo times (TEs) were acquired using the following acquisition parameters: matrix size = 256 × 176, number of axial slices = 69, slice thickness = 2 mm, no interslice gap, slice coverage of 138 mm (whole brain), spatial resolution = 1 × 1 × 2 mm 3 , resulting field of view = 256 × 176 mm 2 , TE = 13, 67, 93, and 106 msec, repetition time (TR) = 10 seconds, bandwidth = 176 Hz/pixel, refocusing angle = 160 , turbo factor = 13, parallel imaging with a reduction factor of 2, and acquisition time for each dataset = 1 minute 32 seconds. Each of the datasets was acquired twice for averaging, resulting in a total acquisition time of 12 minutes 16 seconds.
To correct for subject motion, all datasets were first coregistered to a common target. For each TE, the two respective datasets were then averaged. Exponential fitting of the dependence between TE and the signal intensities in the averaged datasets yielded apparent T 2 values (T 2 app). It has been shown that T 2 app can deviate considerably from the true T 2 value if the actual refocusing angle deviates from the ideal value of 180 , yielding stimulated or secondary echoes. 24 This can be due to B 1 inhomogeneities, deviations of the slice profile from a perfectly rectangular shape or deliberate choice of a reduced refocusing angle to reduce the RF power, as in this study. Thus, a B 1 -dependent correction was performed, converting T 2 app into actual T 2 values as described previously. 25 To allow for whole-brain coverage at a feasible imaging time, only four TE values were chosen. However, a previous study showed that a similar protocol yielded accurate T 2 mapping with a scan-rescan deviation of about only 1.5%. 26 For FCD definition, conventional FLAIR datasets were obtained with the following acquisition parameters: matrix size = 256 × 220 × 160, isotropic resolution = 1 mm, field of view = 256 × 220 × 160 mm 3 , TE = 353 msec, TR = 5000 msec, inversion time (TI) = 1800 msec, bandwidth = 930 Hz/pixel, duration = 7 minutes 12 seconds.

Segmentation and Data Analysis
The cerebral cortex and the cortical and WM surfaces were identified, and the cortical thickness was determined vertex-wise by applying the "recon-all" stream implemented in the Freesurfer toolbox to the synthetic MP-RAGE datasets. Boundary-based coregistration of the T 2 and FLAIR datasets to the synthetic MP-RAGE datasets was performed with BBRegister. 27 Based on the FLAIR datasets, masks of FCD areas were manually defined for each subject. Voxels were selected that appeared abnormal in conventional FLAIR datasets and which were part of or associated with a neuroradiologically diagnosed FCD. A systematic standardized assessment of FLAIR data was performed by three independent raters to exclude subtle morphologic cortical and juxtacortical changes outside the FCD areas. This analysis was performed by an experienced neurologist (10 years of experience) and by two senior neuroradiologists specialized in epilepsy imaging (15 years and 18 years of experience). A confidence score was obtained from each rater for each dataset. The raters stated whether morphologic cortical / juxtacortical changes outside the FCD areas were definitely or probably present or absent (1, definitely absent; 2, probably absent; 3, equivocal; 4, probably present; or 5, definitely present). 28 The FCD voxels were removed from the T 2 maps to exclude FCD-associated changes from the analysis. To reduce partial volume effects with cerebrospinal fluid (CSF) and WM, T 2 values were sampled in the central 20% of the cortex 29 and saved in surface datasets.

Statistical Analysis
Mean cortical T 2 values were determined across all non-zero vertices for each subject, as discussed in a previous study, 30 and compared between groups with an unpaired t-test. The Pearson correlation coefficients between these values and clinical parameters (number of seizures in the previous 3 months and number of anticonvulsive drugs in the medical history, including current treatment) were calculated. Additionally, T 2 values were averaged separately for each hemisphere and compared via paired t-tests between the hemisphere where the FCD was located and the contralateral side across the patient cohort. Compensation for multiple comparisons was performed via the Benjamini-Hochberg / False Discovery Rate (FDR) method for the respective statistical tests (comparing average cortical T 2 values between all patients and healthy subjects, estimating the correlation of T 2 and clinical parameters, and comparing T 2 between the hemispheres). A secondary comparison of mean cortical T 2 values was performed between the subgroup with a low seizure rate (≤12 seizures in the 3 months before data acquisition) and the control group to eliminate the effects of patients with aggressive disease.
Surface-based group analysis was performed for cortical T 2 values and for the cortical thickness. The T 2 surface data and cortical thickness maps were normalized and smoothed with a Gaussian kernel (full-width at half-maximum of 1 cm). A General Linear Model (GLM) analysis was performed for group comparisons. For a given region, only the data from patients without a local FCD in this region were included in the GLM calculation. Permutation simulations were performed for vertices with significant P values (<0.05) to detect clusters indicating group differences and to correct for multiple comparisons (5000 simulations, clusterwise threshold = 0.05).
An FDR of 0.05 was chosen for the Benjamini-Hochberg procedure. Corrected P values below 0.05 were considered significant for the surface-based analysis.

Results
The number of seizures of the 16 FCD patients during a period of 3 months before the investigation was 130 AE 293 (mean AE standard deviation, SD). However, 11 of the patients had equal to or less than 12 seizures during this time. The number of anticonvulsive drugs in medical history including current treatment was 4.1 AE 2.2 (range 1-7). Ten of the 16 patients had an FCD with a positive transmantle sign. Figure S1 in the Supplemental Material shows the FCDs in FLAIR and MP-RAGE datasets for three representative patients. The three independent raters found that morphological cortical and juxtacortical changes outside the FCD areas were definitely absent for all patients. For three patients, the FCD type was histopathologically confirmed after data acquisition and resection (1x type IIa and 2x type IIb). EEG recordings revealed abnormalities at the FCD locations in 12 patients (findings indicating structural abnormalities in seven patients, epileptiform activity in one patient, both in four patients).
A normalized map of cortical T 2 values for a single representative subject is shown in Figure 1. Mean cortical T 2 values were significantly increased (P = 0.01) in the patient group (group mean AE SD: 83.4 AE 2.1 msec) compared to the control group (81.4 AE 2.1 msec). The cortical T 2 values in the patient group were neither correlated with the number of seizures in 3 months (r = 0.16, P = 0.55), nor the number of anticonvulsive drugs in medical history (r = 0.09, P = 0.74). Cortical T 2 values were also increased in the subgroup with 11 patients with a low seizure rate (82.9 AE 1.3 msec, P = 0.026). Figure 2 presents regions with cortical T 2 differences between groups. The spatial distribution of the clusters in Fig. 2 after correction for multiple comparisons demonstrates that, for the investigated cohort of FCD patients, cortical T 2 increases were mainly observed for the frontal and parietal lobes. However, the uncorrected maps suggest that a T 2 increase might also be present in some temporal regions. Surface-based analysis of the cortical thickness after correction for multiple comparisons revealed no clusters of a significant increase or decrease in cortical thickness. Mean unilateral cortical T 2 values in the patient group did not differ between the hemisphere where the FCD was located (83.5 AE 2.6 msec) and the contralateral hemisphere (83.2 AE 1.9 msec, P= 0.67). Figure 3 demonstrates surface maps of the FCD locations of all patients, the red/yellow color indicating the presence of an FCD in one/two patients at the respective locations.

Discussion
Our study used T 2 relaxometry and surface-based analysis techniques to investigate normal-appearing cortical tissue in epilepsy patients with FCD. Widespread cortical T 2 increases in frontal, parietal, and some temporal regions were observed in the patient group, suggesting effects of the disease in cortical regions beyond FCD areas.
The underlying microstructural changes for the T 2 differences we observed are not known. However, a previous investigation reported that gliosis correlated with a hippocampal T 2 increase in temporal lobe epilepsy. 31 These results indicate that gliosis might either cause an increase of the tissue parameter T 2 or affect the T 2 measurement. Although pathological immunoreactivity outside the FCD was not observed in an immunohistochemical study by Rossini et al., 3 gliosis on a microstructural level is a candidate for a remodeling process which might explain the observed T 2 increases. Importantly, clusters indicating atrophy were not observed in the investigated cohort, which is in line with a previous study demonstrating hippocampal T 2 changes in temporal lobe epilepsy, even in patients showing no signs of atrophy. 14 Furthermore, it has been reported that hippocampal gliosis may occur without atrophy. 32 Since T 2 depends on the free water content in tissue, 12,13 cortical reconstruction characterized by tissue damage and replacement of cells in nervous tissue by water on a microstructural level could potentially be another mechanism leading to increased cortical T 2 values in epilepsy patients with FCD.
However, factors driving these tissue changes in epilepsy are not yet fully understood. Cortical T 2 changes in epilepsy patients with FCD might be caused by seizure activity. It should be noted that, to the best of our knowledge, none of the investigated patients had experienced a status epilepticus at the time of this study. Since seizures might contribute to hippocampal gliosis in temporal lobe epilepsy, 9 the pathogenesis of the observed tissue changes might also be related to seizure activity. However, both in our study and in a previous investigation by Liu et al, 5 which evaluated atrophy in patients with epilepsy longitudinally, no significant relationship between structural parameters and seizure recurrence was observed. Furthermore, in the present study T 2 changes were also observed in the subgroup with a lower seizure rate. Additionally, multiple antiepileptic drug exposure might be another risk factor for neocortical damage in epilepsy. 5 However, a significant correlation of cortical T 2 values in the patient group and the number of anticonvulsive drugs in the medical history was not observed in our study. As it is likely that multiple factors contribute to the widespread T 2 increases, more extensive relaxometry studies might better characterize and separate the different effects.
It should be noted that, although cerebellar atrophy is a common finding in patients with long-standing epilepsies, particularly under treatment with anticonvulsive drugs, 7 the cerebral cortex was deemed a better target for qMRI analysis in epilepsy patients with FCD for the following reason: The cerebellar cortical layer is relatively thin and tightly folded. Therefore, analysis of cerebellar T 2 values would result in strong partial volume effects, and thus increase the variability of average T 2 values in cerebellar WM and GM across the groups, rendering group comparisons difficult. For the analysis of the cerebral cortex, T 2 values were read in the central 20% of the cortical layer to reduce partial volume effects. This approach would not be feasible for the thin cerebellar cortex, considering the given resolution. Several volumetric studies have demonstrated cortical changes, 4,5,11 suggesting that epilepsy might systemically affect the brain beyond underlying focal lesions. Observations included multifocal abnormalities in GM volume in patients with malformations of cortical development, 4 focal neocortical volume loss in 14% and generalized volume loss in 41% of patients with chronic epilepsy, 5 and extratemporal cortical atrophy in temporal lobe epilepsy. 11 In our study, cortical atrophy was not observed, indicating that T 2 relaxometry might be more sensitive for the assessment of cortical tissue changes in epilepsy, particularly in smaller cohorts. Additionally, a diffusion tensor imaging study by Rugg-Gunn et al. reported abnormal anisotropy and diffusivity in areas that appeared to be normal on conventional images in patients with epilepsy and cerebral lesions. 33 In our study, cortical T 2 values in the patient group did not differ between the hemisphere where the FCD is located and the contralateral side, also supporting a global character of tissue changes. It is unclear so far whether tissue abnormalities outside focal lesions exhibit an epileptogenic potential. However, in a previous study regions with increased diffusivity matched areas of epileptiform EEG activity in some patients with normal conventional MRI findings. 33 Limitations A limitation of the study is the potential risk of underestimating cortical T 2 increases in patients in the surface-based analysis. When analyzing data obtained from patients with a cerebral lesion, the question arises how to exclude the lesion and how to handle this region in the further steps of the analysis. A potential procedure would be to fill the lesion with average values taken from the surrounding cortex ("lesion filling"). However, since T 2 values are increased in normal-appearing cortex in patients, this approach may yield a focal overestimation of the T 2 increase in the epilepsy group. Here, a more conservative approach was followed, by calculating the GLM analysis for each vertex only for the patients for whom cortical T 2 values outside the FCD were available, ie, excluding patients who showed an FCD in the respective region. Further limitations are the small sample size and the relatively long MRI acquisition time.

Conclusion
The observed widespread cortical T 2 increases suggest cortical remodeling on a global level in normal-appearing cortex in epilepsy patients with FCD. However, the etiology of these cortical changes is not fully understood, and it is likely that a combination of multiple factors contributes to cortical abnormalities in the tissue composition in patients with epilepsy and FCD.

Conflicts of Interest
The authors report no conflicts of interest relevant to this study. Dr. E. Hattingen has received speaker's honoraria from BRACCO. Dr. F. Rosenow has received honoraria for presentations and consultations from EISAI, UCB Pharma, Desitin Arzneimittel, Hexal, Novartis, Medtronic, GW-Pharma, Shire, Sandoz, and Cerbomed, as well as research grants from UCB, European Union, Deutsche Forschungsgemeinschaft, European Science Foundation and the Hessonian Ministries of Science and Arts and of Social Affairs and Integration. Dr. H. Steinmetz has received speaker's honoraria from Bayer, Sanofi, and Boehringer Ingelheim. The remaining authors have no conflicts of interest.