Computational protocol: White matter changes and gait decline in cerebral small vessel disease

Similar protocols

Protocol publication

[…] Traditional SVD makers (WMH, lacunes and microbleeds) were rated according to the STRIVE criteria (). Lacunes were manually rated on FLAIR/T1-weighted scans and microbleeds on T2 ∗ -weighted MRI scans by raters blinded to clinical data. The follow-up FLAIR images were resliced to match the slice thickness of baseline FLAIR images, limiting the differences in partial volume effects between baseline and follow-up scans. To determine the effects of change in slice thickness of the FLAIR sequence, we calculated WMH volumes for odd and even slices separately. Intrarater and interrater reliabilities were good (for lacunes: weighted kappa values 0.87 and 0.95, respectively, and for microbleeds: 0.85 and 0.86, respectively). WMH were segmented by using an in-house developed semi-automatic detection method on baseline and follow-up FLAIR sequences (). All scans were visually checked by 1 rater and corrections were made when segmentation failures had occurred. Total WMH volume was calculated by summing all segmented areas multiplied by slice thickness.Automated segmentation on T1 images of baseline and follow-up was performed using Statistical Parametric Mapping 12 unified segmentation routines (SPM12; Wellcome Department of Cognitive Neurology, University College London, United Kingdom,, in order to obtain gray matter (GM), WM and cerebrospinal fluid (CSF) probability maps. To avoid the erroneous segmentation of WM regions with WMH as GM, the T1 images were first corrected using the binary maps of WMH by replacing the voxel intensities of WMH with the average intensity of the normal-appearing WM on the T1 images. The volumes were calculated by summing all the voxel volumes belonging to the tissue class. All images were visually checked for co-registration errors and motion and/or segmentation artifacts. Total brain volume was taken as the sum of total GM and WM volume. GM volume was composed of the volume of the neocortex, basal ganglia and thalamus.To account for inter-scan-effects, we corrected the normalized follow-up brain volumes for the difference in intracranial volume (ICV; sum of GM, WM and CSF) between baseline and follow-up by multiplying all volumes by the factor ‘ICV baseline/ICV follow-up’. Next, all volumes were normalized to the baseline ICV to adjust for head size (). Note that all brain volume represent relative volume (% of the intracranial volume). We calculated brain volume change and changes in the number of lacunes and microbleeds as the difference between follow-up and baseline. [...] Statistical analyses were performed using IBM (Armonk, NY) SPSS Statistics 20. To compare the baseline characteristics of participants who were included and those not, we used age and sex-adjusted ANCOVA or logistic regression analysis. For those included, gait and imaging characteristics at baseline and follow-up were compared by using a paired t-test, Wilcoxon signed rank test or McNemar test when appropriate. Multiple linear regression analysis was used to investigate the association between change in each gait variable and change in the different MRI and DTI measures. Since the baseline imaging markers were not associated with gait decline (), these measures were not included in the subsequent analyses. Adjustments were made for age, sex, follow-up duration (time between baseline and follow-up assessment), height and baseline gait variable. WMH volume was log transformed, because of the skewed distribution. The variance inflation factor (VIF) was calculated for all regression models to test for the presence of multicollinearity. The VIF scores were low for all models (scores were below 3, where VIF-scores > 5 are considered to reflect high multicollinearity). Regression coefficients were presented as standardized beta-values.Voxel-wise statistical analyses for changes in TBSS data and individual gait variables were performed by using permutation-based statistical interference tool for non-parametric approach as part of the FSL toolbox (randomize). The number of permutation tests was set at 5000. Significant associations were determined by using a threshold-free cluster enhancement with a p-value < 0.05, corrected for multiple comparisons. Adjustments were made for follow-up duration, age, sex, height and baseline gait variable (model A) and additionally for changes in MRI measures (including WMH volume, number of lacunes and microbleeds, WM and GM volume) (model B). […]

Pipeline specifications

Software tools SPM, SPSS
Applications Miscellaneous, Magnetic resonance imaging
Organisms Homo sapiens