Computational protocol: The Impact of Cognitive Training on Cerebral White Matter in Community-Dwelling Elderly: One-Year Prospective Longitudinal Diffusion Tensor Imaging Study

Similar protocols

Protocol publication

[…] All diffusion data was processed using the University of Oxford’s Center for Functional Magnetic Resonance Imaging of the Brain (FMRIB) Software Library (FSL) 5.0.8 ( First, each participant’s data was pre-processed through an automated pipeline consisting of: (1) head motion and eddy current correction; (2) extraction of brain tissue from the b = 0 volumes using the Brain Extraction Tool (BET); (3) smoothing of the DTI image using the command “fslmaths” with a 1-voxel box kernel and the f-median flag; (4) computation of the DTI scalar indices (FA, MD, AD and RD images). Next, to avoid removal of within-subject longitudinal differences, a nonbiased subject-specific template was used, as recommended by Engvig et al.. Structural Image Evaluation, using Normalization, of Atrophy (SIENA) from FSL was run with option “-B”; the fractional intensity threshold parameter (f) was set to 0.2. After registering FA maps from baseline to 12 months post-scan (tp1 halfway registered maps) and vice versa(tp2 halfway registered maps), both volumes were resampled into the space at the halfway point between. Thereafter, a subject-specific template (base FA template) was created by averaging the two time points’ halfway registered FA maps. Next, all base FA templates were aligned into the Montreal Neurological Institute (MNI) template space using FMRIB’s nonlinear registration tool (FNIRT), and a mean FA map was created. Next, the mean FA map was thinned to create an average white matter tract skeleton, representing the centre of the tracts shared by all participants; the skeleton threshold was set at 0.2 as default to exclude grey matter and cerebral spinal fluid. Finally, each participant’s FA data from the base FA template was projected onto the mean skeleton, adopting the FA value from the local centre of the nearest relevant tract. After creation of the FA mean skeleton, the MD, AD and RD values for each participant were produced in a similar manner–halfway registration, warping and projecting the analogous data onto the mean FA skeleton. Changes of FA, MD, AD and RD maps from 12 months post-scan to baseline were calculated by tp1 halfway registered maps, subtracting from tp2 halfway registered maps using the command “fslmaths”. [...] One-way analysis of variance (ANOVA), the nonparametric Kruskal-Wallis test and the Chi-squared test were performed to compare baseline characteristics between participants among the three groups, using IBM SPSS Statistics 22.0 (IBM Corporation, Somer, NY, USA). Independent sample t-tests, nonparametric Mann-Whitney tests and Chi-squared tests were conducted to compare characteristics between participants included in the final analysis and those who were excluded. The general linear model (GLM) repeated measure was used to investigate the training’s effect on the behavioural level. The model included the main effect for time and group and a time × group interaction term. Each cognitive measure was tested separately with covariates of the baseline RBANS total score.Cohen’s d, as a measure of effect sizes of the t test for means, was calculated using G*Power (© Franz Faul, Edgar Erdfelder, Albert-Georg Lang, and Axel Buchner, 2006, 2009). The NES was used to compare cognitive measures at 12 months post-test to baseline scores and control group scores. Bias-corrected NES of training was defined as: [(trained mean at 12-month post-test − trained mean at baseline) − (control mean at 12-month post-test − control mean at baseline)] ÷ pooled standard deviation at baseline, before applying a bias correction factor 1–3 ÷ [4 (nT + nc − 2) − 1]. An operational definition of “small”, “medium” and “large” effect sizes were offered with d value equalling 0.2, 0.5 and 0.8, respectively. Measures of percentage change (%change) in cognitive measures for each participant were derived by calculating the difference between variables at the two time points, then divided by variable at baseline [i.e. %change = (variable at 12 post-test − variable at baseline)/variable at baseline × 100].Voxel-wise statistical analysis of the DTI data was carried out using the threshold-free cluster enhancement (TFCE) option in “randomise” in FSL by applying appropriate contrasts with GLM; p < 0.05, corrected for multiple comparisons (FWE), was considered significant. 5000 permutations were performed for each contrast and statistical inference.First, we investigated the hypothesis that long-term changes of white matter diffusivity are detectable across groups as FA decreases, and/or MD increases, and/or RD increases and/or AD decreases. A whole-brain voxel-wise analysis was performed using two-sample paired t test on DTI indices between baseline and 12 months post-scan.Second, we tested the hypothesis that the effect of cognitive intervention on white matter would last for as long as 12 months after training cessation. First, a whole-brain voxel-wise analysis testing for group differences in FA, MD, RD and AD at baseline was performed to exclude baseline differences. Next, a whole-brain voxel-wise F-test on changes of DTI indices among the three groups was conducted, controlling for age, sex, years of education and baseline DTI-indices values. A two-way mixed-effect ANOVA for repeated measures was conducted to further examine post hoc time × group interactions. The John Hopkins University (JHU) ICBM-DTI-81 White-Matter labels atlas and the JHU White Matter Tractography Atlas were used for anatomical labelling of regional differences of DTI indices from TBSS.Finally, mean FA, MD, AD and RD values of all subjects were extracted from the brain regions, with significant group differences generated in the voxel-wise analysis, before being exported to SPSS using Pearson’s or two-tailed Spearman’s nonparametric correlation analysis according to data type. The aim was to test the relationships between change of DTI indices and age, sex, education and training-related changes in cognitive function. […]

Pipeline specifications

Software tools FSL, BET, SPSS, G*Power
Applications Miscellaneous, Magnetic resonance imaging, Functional magnetic resonance imaging
Organisms Homo sapiens
Diseases Lymphoma, Non-Hodgkin, Radial Neuropathy