Hi,
>I am a beginner in TBSS , and learn TBSS pipeline following user guide.
>However, it is not clear for me that what fsl can do , as As mentioned
>in this paper <Widespread reductions of white matter integrity in
>patients with long-term remission of Cushing's disease>.
>
>1. Data process
>
> How to implemented a ROI-based TBSS ?
If you have an ROI mask then you multiply this with the skeleton mask and
use the result for the mask in the call to randomise (the -m option).
>2. Statistical analysis
>
> Age, gender and education (demeaned across groups) were included in
>the analysis as nuisance regressors to correct for between group
>variances.
This is easily done by including these regressors (or EVs, which is an
equivalent term) in the design matrix.
> How it could be done ? what do "demeaned" and "nuisance regressors"
>mean?
demeaned is a term that indicates that the overall mean has been
subtracted from the set of values (i.e. the values now have zero mean)
nuisance regressors are just regressors (or EVs) that are not included in
the contrast of interest and therefore are used to explain away signals
relating to effects of no interest.
The GLM page in the FSL wiki shows examples of these kinds of designs and
has more information about demeaning.
>3. Post-hoc analyses
>
> i) Clinical characteristics (disease duration inyears, duration of
>remission in years, and CSI scores), and scores on psychological
>measures found to be significantly different between the patients and the
>controls (MADRS and AS scores), were fed into FSL's Randomise Tool along
>with the FA values of the voxels within regions of significant group
>differences resulting from the ROI analysis.
>
> How the behavioral variables were fed into FSL's Randomise ?
By putting them into a regressor (EV) in the design matrix. See the FSL
Wiki page on TBSS and the page on GLM.
> ii) A mask was created of the voxels that were found to differ
>significantly between groups on FA resulting from the exploratory
>whole brain analysis. Along with this mask, information on each
>individuals' AD (the 1st eigenvalue), RD (the average of the 2nd and 3rd
>eigenvalues), and MD was fed into FSL's Randomise Tool using
>permutation-based inferences with TFCE.
>
> Why and how a mask was created?
The mask was created so that the analysis was restricted to the areas of
interest (within the mask). It would have been created by thresholding
the output of the whole brain analysis, probably using the fslmaths tool.
I advise you to go through the FSL Course practical material if you are
new to FSL as many of these concepts are present in the examples used
there.
All the best,
Mark
|