|
nif:isString
|
-
For this study, written informed consents as approved by the research ethics committee of the Montreal Neurological Institute and Hospital (MNI/H) were obtained from all the subjects. At its full board meeting of May 26, 2006, the Research Ethics Board (REB) of the MNI/H has endorsed the review of the project entitled: “Electrical, Metabolic and Structural Analysis of Human Epileptogenic Lesions” (Dr. J. Gotman being the principal investigator for this project). The REB of the MNI/H found this research to be acceptable for continuation at the McGill university healthcare centers (MUHC) and specifically approved this study. The REB of the MNI/H acts in conformity with standards set forth in the (US) Code of Federal Regulations governing human subjects research and functioning in a manner consistent with internationally accepted principles of good clinical practice.
Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0050359.t003 Summary of the duration-weighted group differences in resting-state FC. Full results of seeds that showed significant functional connectivity change between the two groups (Z-threshold 2.7, p<0.05/18 corrected) for the contrast of patients minus controls. In cases where there were several peaks in each cluster, we only reported the highest one. Star sign indicates those seeds that not only their corresponding functional connectivity was significantly different between the two groups, but also the average functional connectivity, taken over significant clusters, was highly correlated with duration factor in patients (|r| >0.75, p<0.05/18). R: Right, L: Left, Med: Medial, Inf: Inferior, Sup: Superior, G: Gyrus.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0050359.g001 Results of significant duration-weighted group differences in functional connectivity (ΔFC).Left: seeds in purple and their names. Right: some selected slices illustrating group differences. The color-coded Z-score maps (p<0.05/18 corrected) show the results of alterations in functional connectivity in IGE patients compared to controls (for the contrast of patients minus controls). Positive functional connectivity is coded in red to yellow, while negative in blue to white. Note that Figure 1 only shows those clusters whose functional connectivity in IGE patients significantly correlates with the duration factor (|r| >0.75, p<0.05/18).
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0050359.g002 Results of average functional connectivity within each group of subjects for the seeds reported in Figure 1.Left: seeds names. Right: The same selected slices illustrating average functional connectivity within controls (A–C), and average functional connectivity within patients (A′–C′), Color-coded statistical Z-score maps (p<0.05/18 corrected) showing a positive (coded in yellow to red) and a negative (coded in blue to white) functional connectivity.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0050359.g003 Results of correlation analysis for the seeds reported in Figure 1. (A–D): Bar diagrams illustrating average functional connectivity within each group over the clusters of significant differences reported in Figure 1. Figure 3 (A, B) corresponding to the seed in medial superior frontal gyri, respectively illustrate the average functional connectivity within each group over clusters in the right and left premotor area. Figure 3 C corresponding to the seed in right precentral gyrus, illustrates the average functional connectivity within each group over a cluster in the left dorsal premotor area. Figure 3 D corresponding to the seed in the left medial prefrontal area, illustrates the average functional connectivity within each group over a cluster in the left precentral gyrus and the supplementary motor area. (A′–D′): average functional connectivity within patients for the clusters discussed in (A–D) as a function of the duration factor. The black line shows the fitted line for correlation analysis. Correlation coefficient r and p-value are shown for each correlation analysis.
Our EEG-fMRI dataset of patients scanned at 3 Tesla contained 14 patients with IGE (aged 33+/−9 years, 7 males, and 7 females). All patients were taking medication at the time of study and they did not stop it for the purpose of scanning. The study was approved by the research ethics board of the Montreal Neurological Institute & Hospital (“MNI/H REB”) and subjects participated in the research after giving written informed consent. Patients inclusion criteria were: a) having at least two runs without GSW during the scan (fMRI recording included 6–14 six-minute runs), b) wakefulness proven by EEG recording during these runs, c) motion of less than 1 mm as determined by the realignment of the preprocessing (see section 2.5, preprocessing step 4). Table 1 gives the demographic and clinical characteristics of all patients. Fourteen age and sex-matched healthy controls (aged 33+/−8 years) were scanned using the same EEG-fMRI protocol, fulfilling inclusion criteria b) and c). There was no significant difference between the age distributions of the two groups (sign test, p>0.05).
The EEG acquisition was performed using 25 MR compatible electrodes (Ag/AgCl) placed on the scalp using the 10–20 (21 usual electrodes without Fpz and Oz, reference at FCz) and 10–10 (F9, T9, P9, F10, T10 and P10) electrode placement systems. Two electrodes located on the back recorded the electrocardiogram (ECG). To minimize movement artifacts and for the patient’s comfort, the head was immobilized using a pillow filled with foam microspheres (Siemens, Germany). Data were transmitted from a BrainAmp amplifier (Brain Products, Munich, Germany, 5 kHz sampling rate) via an optic fiber cable to the EEG monitoring computer located outside the scanner room.
Functional images were continuously acquired using a 3T MR scanner (Siemens Trio, Germany). A T1-weighted anatomical acquisition was first done (1 mm slice thickness, 256×256 matrix; TE = 7.4 ms and TR = 23 ms; flip angle 30°) and used for superposition with the functional images and inter-subject group co-registration. IGE patients had 7–14 and the healthy controls had 4 runs of 6 min each. However, only a subset of those runs was selected based on the mentioned inclusion criteria. In addition, to have the same number of runs for all subjects, only 2 runs were included at the end for each subject. The functional data were acquired using a T2*-weighted EPI sequence (5×5×5 mm voxels, 25 slices, 64×64 matrix; TE = 30 ms and TR = 1750 ms; flip angle 90°). No sedation was given and resting-state was defined as a state of relaxed wakefulness when subjects have their eyes closed and are instructed to refrain from any structured thoughts.
The Brain Vision Analyzer software (Brain Products, Munich, Germany) was used for off-line correction of the gradient artifact and filtering of the EEG signal. This software uses the method described by Allen and colleagues [22]. A 50 Hz low-pass filter was also applied to remove remaining high-frequency artifact. The ballistocardiogram (BCG) artifact was removed by independent component analysis [23], [24]. A neurologist reviewed the EEG recordings and made sure that the selected runs in IGE patients were GSW free and that the patients and controls were awake during these runs. Figure S1 provides a sample of such data in selected patients.
Seed-based Functional Connectivity Group Analysis: Data processing was carried out using FMRIB Software Library (FSL), www.fmrib.ox.ac.uk, Oxford U.K., FSL version 4.1 [25], [26]. The following preprocessing steps were applied: (1) removal of the first two volumes of each scan to allow for equilibrium magnetization, (2) slice timing correction using Fourier-space time-series phase-shifting, (3) non-brain tissue removal [27], (4) motion correction using a 6-parameter linear transformation using a maximization of the correlation ratio (default settings of FSL) [28], (5) intensity normalization of all volumes of each run as implemented in FSL (6) spatial smoothing using a Gaussian kernel with 6 mm full width at half maximum (FWHM), (7) high-pass temporal filtering with cut off frequency of 0.01 Hz. To achieve the transformation between the low-resolution functional data and the average standard space (MNI152: average T1 brain image constructed from 152 normal subjects [29]), we performed two transformations. The first was from the low resolution EPI image to the T1-weighted structural image (using a 7 degrees of freedom affine transformation), and the second was from T1-weighted structural image to the average standard space (using a 12 degrees of freedom linear affine transformation, voxel size = 2×2×2 mm). Removing the physiological noise (cardiac and respiratory related signals) has a great impact on the results of functional connectivity analysis at rest [30], [31], [32]. We used a method similar to that of Shehzad et al. [33], [34], by employing the average signals taken over sections of white matter (WM), cerebro-spinal fluid (CSF), and the whole brain (the so-called global signal) as nuisance regressors. In order to do this, we segmented each individual’s high-resolution T1-weighted image, using an automatic segmentation software [35] providing tissue probabilistic maps for Grey Matter, White Matter and CSF. The resulting probabilistic tissue maps of WM and CSF images were then thresholded to identify masks of voxels showing more than 80% probability of belonging to the corresponding tissue class. Each thresholded mask was then applied to that individual’s time series and the mean time series was calculated by averaging the time series from all the voxels within the mask. The global signal accounts for several potential sources of physiological noise assuming that fMRI experiments are concerned with local changes in neuronal activity and that global signals represent uninteresting sources of noise [36], [37]. Nine nuisance regressors were thus considered for the first general linear model (GLM) (WM, CSF, global signal and 6 motion parameters: x, y, and z translations and rotations obtained from motion correction step in preprocessing). Thus, for each run of each individual, a GLM analysis was carried out using the time series of nuisance signals [38]. The purpose of this first regression analysis was to model the nuisance signals as much as possible so that these sources of artifact can be removed from the data. To extract the seeds’ time series, the MNI coordinates of both activation and deactivation peaks reported in the study of Ogg et al. [19] were considered. The activation peaks correspond to regions activated more during CPT than fixation (CPT>Fixation), and the deactivation peaks correspond to regions with a higher activity in fixation compared to CPT (Fixation>CPT). In cases where they had reported several clusters in one region, we considered the peak corresponding to the largest cluster. Overall, 11 activation-related and 7 deactivation-related seeds were selected (Table 2). In order to obtain BOLD time-series for each seed in every run, first a spherical mask (radius = 6 mm) around the seed in the standard space was defined. Then this mask was resampled into the low-resolution functional space and finally, the mean time-series over the mask was calculated. For each run of each individual, the mean time-series of the seed was then used as a regressor in a GLM model to find voxels that show correlation with that seed. For each seed, the two runs of each individual were combined using fixed effects model. The resulting functional connectivity map for each individual was then transformed to the MNI space. This analysis was followed by between-group analysis performed using a mixed-effects model [39]. This GLM analysis included the duration factor, explained below, as a regressor. The duration of epilepsy was obtained for each patient as a clinical measure [20]. Since the original duration values (Table 1) are count data in terms of number of years of having IGE, the square root transformation was applied to make their distribution closer to a Gaussian distribution [40]. We then scaled these values to range between 0 and 1 (note that scaling by a constant value does not change the GLM results). We named these normalized values duration factors. We performed between-groups analyses using the mixed-effects model implemented in FSL [39]. To incorporate the clinical measures in the group-level analysis we used a similar approach as explained in Vahdat et al. [41]. We considered two regressors in group-level GLM: one modeling the average of healthy control subjects, and the other one modeling a weighted average of epileptic patients based on duration factor. Then, the contrast of interest was defined as the difference between these two regressors. This analysis increases the sensitivity for detecting changes between the two groups, which are related to the clinical measures of epilepsy duration. Corrections for multiple comparisons at the cluster level was carried out based on random field theory as implemented in FSL [39] (threshold |Z| >2.7; cluster significance: p<0.05, corrected). To correct for multiple seeds, we identified as statistically significant the clusters with a probability level lower than p = 0.05/18 = 0.0028 (Bonferroni correction, 18 being the number of seeds). This between-subjects analysis produced thresholded Z score maps of activity associated with each seed. The positive values in the Z-map corresponded to areas where functional connectivity was elevated in patients compared to controls. The negative values corresponded to areas where functional connectivity was reduced in patients compared to controls. The anatomical regions were labeled according to the Harvard-Oxford cortical and subcortical [42], and Juelich histological atlases [43]. In the patient group, we calculated the Pearson correlation between the average functional connectivity taken over each significant cluster and the duration factor. Correlation coefficients larger than 0.75 (|r| >0.75, corresponding to p<0.05/18, corrected for multiple comparisons) were considered as statistically significant in our analysis. In addition, we extracted seizure frequency (number of seizures in the last 2 years before the scan date), and time interval from the last ictal event as two extra clinical measures. Similar to the approach used to calculate the duration factor from duration of epilepsy, seizure frequency values were converted into a frequency factor calculated for each patient. However, only 8 patients had non-zero values for this measure. Consequently, they were compared with 8 age- and gender-matched healthy controls in the subsequent GLM analysis. This regression analysis was similar to the one previously proposed, replacing duration factor by frequency factor. As seizures were at least 1 month away for all our patients that we had information about, we did not include time from last ictal event as a factor in our analysis (see File S1 and Table S1).
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}