Chapter 2: Background Information
3.2 Methods
Fifty-six children aged 3.0-4.7 years (3.6 ± 0.4 years, 22 females and 34 males) were recruited from an ongoing prospective study (Kaplan et al., 2012) and underwent diffusion tensor imaging and language assessment. 21 of these children (3.8 ± 0.5 years, 11 females, 10 males) also completed an ihMT scan. Children were excluded from the study if they had been diagnosed with a neurological or developmental disorder, did not speak English as a primary language, or had contraindications to MRI scanning (e.g.
metal implants). Participants who met inclusion criteria were invited to participate by mail or phone, and underwent telephone screening to confirm eligibility and to be provided with study information. At the time of the study, we collected information on total years of maternal post secondary education, which has been shown to be a good proxy for socioeconomic status, and is also predictive of children’s educational outcomes (Davis-Kean, 2005). Total years of maternal post-secondary education was used as a control variable in all analysis. For the entire sample, the mean number of years of post secondary education was 4.0 ± 3.0, and ranged between 0-10 years (all mothers had completed high school). The University of Calgary Conjoint Health Research Ethics Board approved this study. Informed consent was obtained from each participant’s legal guardian and verbal assent was obtained from participants.
3.2.2 Language assessment
Each participant was assessed using the Phonological Processing and Speeded Naming subtests from the NEPSY-II. The NEPSY-II is a development
neuropsychological assessment, which has been standardized in over 1000 American children aged 3-16 years (Korkman et al., 2007). The Phonological Processing subtest assesses phonemic awareness, while the Speeded Naming subtest measures rapid semantic access to, and production of names of colors and shapes. The language assessment took approximately 10 minutes to complete.
3.2.3 MRI scanning
All imaging took place at the Alberta Children’s Hospital on the same General Electric 3T MR750w system using a 32-channel head coil (GE, Waukesha, WI). Children were scanned while awake and watching a movie, or naturally asleep. Whole-brain diffusion weighted images were acquired using a single shot spin echo echo-planar imaging protocol, with 30 gradient encoding directions at b=750s/mm2 and 5 images without gradient encoding (b=0s/mm2). The DTI sequence had a TR of 6750 ms, a TE of 79 ms, spatial resolution of 1.6x1.6x2.2 mm3, and lasted 4:03 minutes. The ihMT
sequence used a whole brain 3D spoiled gradient echo (SPGR) sequence with a 5 ms Fermi pulse with peak B1 of 45 mG and ± 5 kHz offset prior to excitation, TR of 10.18ms, TE of 2.04ms, 2.4 mm3 isotropic resolution, and total scan time of 5:04 minutes.
3.2.4 Image analysis
3.2.4.1 Tract based spatial statistics
DTI data was quality checked and processed through in house, Matlab-based software designed to detect and remove motion-corrupted volumes. Only children who
had fewer than 10 volumes of motion-corrupted data were included in the study (i.e. at least 25 good volumes). Data was then processed using FSL’s diffusion pipeline (Jenkinson et al., 2012), including eddy current correction and simple head motion correction using an affine registration to a reference volume, fitting of a diffusion tensor model at each voxel, and calculation of fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD). Whole-brain voxel-wise analysis was performed using FSL’s Tract Based Spatial Statistics (TBSS) (Smith et al., 2006). Each subject’s FA measurements were aligned into a common space using a non-linear registration method, which applies a b-spline representation of the registration warp field. Before registration, all participants’ FA images were compared against each other, and the subject who required the least amount of warping during fitting was chosen as the target image for registration. A mean white matter tract skeleton, which represents the center of all tracts common to the group, was created using an FA threshold value of 0.3. Subjects’ FA data was projected onto the skeleton and analyzed using voxel wise cross-subject statistics with family wise error rate correction for multiple comparisons.
This was repeated for MD, AD, and RD. For this study, each DTI parameter was independently compared with age-standardized Phonological Processing and Speeded Naming scaled scores, while controlling for child’s sex and maternal years of post-secondary education. After analysis, significant clusters were thresholded at p<0.05 and labeled using the Johns Hopkins University-International Consortium for Brain Mapping White Matter Atlas (Mori, 2005).
3.2.4.2 Deterministic tractography
To corroborate findings from the whole brain analysis, a post hoc analysis using semi-automated deterministic tractography was performed. Tracts containing significant clusters from the TBSS analysis were identified, then isolated using Trackvis (Wang et al., 2007) and in-house custom software. Using a priori information on tract location (Wakana et al., 2004), regions of interest (ROIs) were drawn on the most representative subject from the TBSS analysis. All subjects were registered to this target subject using FSL’s linear image registration tool (Jenkinson and Smith 2001) and the inverse
normalization was calculated. The inverse normalization was used to warp the target ROIs to each subject’s native DTI space for automated tractography in native space.
Tracts were then generated using an angle threshold of 30o, quality checked, and corrected manually if necessary (approximately 15% of subjects required some form of manual correction). Mean values of FA, MD, AD and RD were extracted for each tract and using SPSS v21, Pearson’s correlations were run between diffusion parameters and Phonological Processing and Speeded Naming standard scores. The workflow for the tractography analysis is visually summarized in Figure 10.
Figure 10. Workflow for post-hoc tractography analysis.
3.2.4.3 ihMT
Quantitative ihMT (qihMT) images were calculated and output directly from the MRI scanner. Magnetization transfer images were first skull stripped and binarized using FSL’s BET, and then applied as a mask to perform brain extraction on the corresponding qihMT images (poor results were obtained from directly skull stripping the qihMT images). Using FSL’s FLIRT tool, participants skull stripped qihMT images were registered to their FA maps. Tracts generated from deterministic tractography were
applied to the qihMT data, and mean values for each tract were extracted. Mean qihMT values for each tract were individually entered into a general linear model in SPSS, and compared to aged-standarized Phonological Processing and Speeded Naming scores (controlling for child’s sex and maternal years of post-secondary education). For any significant tract, a lateralization index was calculated (left-right/left+right), and compared to the age-standardized language scores, again controlling for sex and maternal years of post-secondary education.