Evaluation of subcortical grey matter abnormalities in patients with MRI-negative cortical epilepsy determined through structural and tensor magnetic resonance imaging

From 2012 May to December, we conducted in this neuroimaging study. We studied 24 patients (15 females and 9 males, mean age = 25.6 ± 12.9 years) with chronic partial epilepsy. All patients had had MRI scans and had long-term EEG records. We first selected epileptic patients with regional epileptiform discharges using a data set on patients at the Buddhist Tzu Chi Epilepsy Center. We termed patients “MRI-negative” if radiologists did not identify any lesions, including neoplasms, traumatic lesions, vascular anomalies, well-defined developmental abnormalities or hippocampal atrophy, in their routine brain MRIs. To completely exclude mesial temporal lobe epilepsy, we did not include patients with maximal ictal/interictal epileptiform discharges at T3, T4 or sphenoid electrodes. We also determined the location of the seizure focus or foci in individual patients through ictal video-EEG recording. We termed a focus “undetermined” if seizure activity arose on the EEG in bilateral frontal regions simultaneously or if there was a diffuse epileptiform discharge with asymmetric body posturing at seizure onset. All of the enrolled patients had seizure manifestations with the subsequent development of generalised convulsions and postictal psychomotor depression. Patient demographic information is shown in Table 1. Twenty-nine age-matched healthy volunteers (14 females and 15 males with a mean age of 27.5 ± 4.2 years) were recruited as the control group. The consent in which informed the research methodology and for publication of data and images was obtained from each participant and/or his/her parents. The study protocol was approved by the Research Ethics Committee at Buddhist Tzu Chi General Hospital (IRB 101–32 and IRB101-99).

All subjects were scanned in a 3T MRI scanner (General Electric, Waukesha, WI, USA). Anatomic T1-weighted images were acquired using a high-resolution, axial, three-dimensional, T1-weighted, fast spoiled gradient recalled echo (3D T1-FSPGR) sequence. Congruent slices with a thickness of 1 mm were generated with a repetition time (TR) of 11.812 ms, an echo time (TE) of 5.036 ms, a field of view (FOV) of 22 × 22 cm, a flip angle of 15 degrees and a 512 × 512 matrix. The DTI protocol consisted of a single-shot-spin-echo planar-imaging sequence. Thirty-four contiguous slices were acquired with a matrix size of 256 × 256, a voxel size of 1 mm × 1 mm, a slice thickness of 3 mm, a TR of 8,000 ms, a TE of 82.4 ms, a number of excitation of 2 and a FOV of 25 × 25 cm. Diffusion-weighted images were acquired in 25 directions (b = 1000 s/mm²), as was a null image (b = 0 s/mm²).

VBM was carried out using the FSL-VBM v1.1 software tool included in the FSL (FMRIB Software Library; the University of Oxford). The VBM analysis procedure comprised the following steps [19]. First, 3D T1-FSPGR images were brain-extracted and GM -segmented before being registered to the Montreal Neurological Institute (MNI) 152 standard space using non-linear registration. The resulting images were averaged and flipped along the x-axis to create a left-right symmetric, study-specific GM template. Second, all native GM images were non-linearly registered to this study-specific GM template and “modulated” to correct for local expansion (or contraction) due to the non-linear component of the spatial transformation. The modulated GM images were then smoothed with an isotropic Gaussian kernel with a sigma of 3 mm for the TFCE-based analysis [20]. Finally, differences in cerebral GM density between the patient and control groups were evaluated using the voxelwise generalised linear model applied using permutation-based non-parametric testing (5000 permutations) [21]. We identified the regions with significant differences in GM density between the patient and control groups using these postprocessing methods and a cluster-size threshold of p < 0.05.

The algorithm FIRST (FMRIB’s Integrated Registration and Segmentation Tool) was applied to separately evaluate the left and right volumes of seven subcortical regions: hippocampus, caudate nucleus, putamen, globus pallidus, nucleus accumbens, thalamus and amygdala [22, 23]. During registration, the 3D T1-FSPGR images were transformed to the MNI 152 standard space using affine transformations with 12 degrees of freedom. A subcortical mask was applied to locate the different subcortical structures, followed by segmentation based on shape models and voxel intensities after subcortical registration. Finally, a boundary correction was used to determine which boundary voxels belong to a given structure. In this study, a Z-value of 3 was used, corresponding to a structure. After the registration and segmentation of all MRI images, all segmented subcortical regions were visually checked for errors in registration and segmentation (Figure 1). The acquired volume of each subcortical structure was normalised to the whole brain volume without cerebrospinal fluid to obtain a volume-ratio value.

All diffusion-weighted images were corrected for eddy current distortion and head motion using the FDT v2.0 software package (FMRIB's Diffusion Toolbox). The preprocessed DTI data were fit to a diffusion tensor model to generate the mean diffusivity (MD) and fractional anisotropy (FA) maps. To obtain transformation parameters, the individual T1-FSPGR image was registered to the null image to fit the DTI resolution using a 12-parameter rigid body transformation. We applied these parameters to transform the segmentation mask to the DTI space using a rigid registration and a nearest neighbour interpolation based on the normalised mutual information method. For each subject, the corresponding values of the MD and FA were calculated for each automatically segmented region.

Three clusters exhibited significant decreases in GM density in the whole brain VBM comparison. Within these clusters, the right putamen, the bilateral nucleus accumbens and the right caudate nucleus were involved (Table 2, Figure 2). In addition, an increase in GM density was also observed over the bilateral paracentral gyri in the patient group (Additional file 1: Figure S1).

The total brain volume was not a confounding factor for the true brain volume (excluding the volume of cerebrospinal fluid), and the true brain volumes of our patients were not different from those of the controls (t = 2.009, p = 0.615). In general, the studied subcortical structures showed different degrees of volume reduction. The volumes of the left caudate nucleus (2.848 ± 0.469 vs. 3.143 ± 0.506, t = 0.430, p = 0.034), left amygdala (0.749 ± 0.176 vs. 0.868 ± 0.214, t = -0.661, p = 0.033) and right putamen (4.002 ± 0.334 vs. 4.297 ± 0.548, t = 1.836, p = 0.025) were reduced significantly in the patients compared with the controls (Table 3).

In general, the MD values for the subcortical structures studied were higher in our patients. The MD value was increased in the left thalamus (0.894 ± 0.050 vs. 0.863 ± 0.059, t = -2.188, p = 0.045), right globus pallidus (0.979 ± 0.046 vs. 0.771 ± 0.041, t = -0.777, p = 0.035) and right hippocampus (1.101 ± 0.102 vs. 1.042 ± 0.053, t = -1.801, p = 0.009). The differences in the FA values were minimal and inconsistent. The FA values were reduced in the bilateral nucleus accumbens in the patients, compared with the controls (left nucleus accumbens, 0.265 ± 0.055 vs. 0.295 ± 0.049, t = -1.991, p = 0.038; right nucleus accumbens, 0.240 ± 0.047 vs. 0.285 ± 0.053, t = -1.555, p = 0.002).

The age at seizure onset positively correlated with the volume ratio of the bilateral nucleus accumbens (regression for right nucleus accumbens: r = 0.523, p = 0.013; regression for left nucleus accumbens: r = 0.386, p = 0.076) (Figure 3A). The disease duration significantly negatively correlated with the volume ratio of the left thalamus (r = 0.598, p = 0.003) (Figure 3B), the mean FA of the bilateral hippocampus (left: r = 0.459, p = 0.032; right: r = 0.463, p = 0.030) (Figure 3C) and the mean FA of the left putamen (r = 0.435, p = 0.043). The clinical-MD correlations with epilepsy duration showed a positive trend, but only that for the left putamen reached statistical significance (r = 0.428, p = 0.047) (Figure 3D). The Additional file 2: Table S1 shows the linear regression relations among normalised volume, the DTI parameters of the subcortical structures and either age at seizure onset or disease duration. The diffusion parameters did not show significant correlations with age at seizure onset.