137 White Matter Connectivity in Narcolepsy Type 1 replaced using non-parametric, Gaussian process modelled predictions [223, 224]. All ROIs were subsequently resampled to the corresponding pre-processed b0 images using linear coregistration of the T1-weighted image, to precisely identify subject-unique ROIs for local and tractography analyses. Incomplete imaging of an ROI due to incorrect field-of-view alignment resulted in the exclusion of this subject from the analyses including this ROI. The posterior-toanterior-encoded DTI scan of one control subject was missing and therefore only the anterior-to-posterior-encoded images were used for this subject. Tract-based spatial statistics (I) In FSL’s TBSS processing stream [225], the two DTI scans were averaged and projected onto an alignment-invariant skeleton image and thresholded at a > 0.2 FA level. All individual FA, MD, AD and RD maps were subsequently projected onto the skeleton, smoothed (5 mm) and fed into permutation-based voxelwise cross-subject statistics with 10,000 permutations and threshold-free cluster enhancement to identify between-group differences. The FA results were masked with FSL’s JHU white matter tractography atlas’ tracts [226] and presented as absolute and relative significant voxel (p < .05) counts with the corresponding minimum p-value. To verify robustness of the results, an additional TBSS FA analysis was done evenly mixing patients and controls by creating two randomly composed groups with a random number generator. Region-of-interest analyses (II) Pre-processed DTI images by FSL and predefined ROIs served as input for the volume-based pipeline of ExploreDTI’s MATLAB-based MR diffusion toolbox [227]. B-matrix calculation and weighted linear least-squares estimators were used to calculate individual diffusion measure maps (FA and MD), that were subsequently masked with the personal ROIs in DTI subject space. The generated outcome measures were averaged for each subject’s two DTI series and served as input for statistical analysis. ROI analyses were limited to only include FA and MD as we hypothesized that these summary measures would be most sensitive to detect underlying white matter differences and to limit the total number of comparisons. In case mean FA or MD were found to be significantly different in an ROI, post-hoc analyses were performed on AD and RD in these regions. 5
RkJQdWJsaXNoZXIy MjY0ODMw