You can try ROI analysis.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Dana
Sent: Tuesday, May 22, 2012 11:13 AM
To: [log in to unmask]
Subject: [FSL] DTI randomise
Hi Everyone,
I have been busy with DTI for a while comparing 2 groups on their FA, and it seems to be after all that there are no significant differences in the whole brain analysis.
Does anyone know if it it useful to continue with ROI analysis anyway?
Beneath, I typed the exact steps we followed for the outcome. Perhaps we did something wrong what explains our outcome, but I'm afraid we should face the truth that there are just no differences.
- We did the Eddy Current correction in FDT
- We created a brain mask for each participant with Bet brain extraction
- We did a DTI-fit adding our created brain mask, and created files for bvals and bvecs.
- We made a new directory with all the patients and controls listed as CON_***.nii.gz and PAT_***.nii.gz .
Then runned the script:
tbss_1_preproc *.nii.gz
which created new directories, FA, and origdata We got a slicedir-link, which showed us all the participants, so we could remove one problem-case to prevent a gap in the front of the brain.
- We registered the outcomes aligning all FA images to a 1x1x1mm standard space. For this we used the FMRIB58_FA standard-space image as the target in TBSS.
The code:
tbss_2_reg -T.
-The next TBSS script applies the nonlinear transforms found in the previous stage to all subjects to bring them into standard space.
We used the code :
tbss_3_postreg -S
-Then we tresholded the mean skeleton using:
tbss_4_prestats 0.2
- Next, after making design matrixes by using : design_ttest2 design 39 49 we tried the randomise in many ways :
TFCE :
randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d design.mat -t design.con -n 500 - - T2
Clusterwise:
randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d design.mat -t design.con -n 500 C 3
Voxel-based:
randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d design.mat -t design.con -n 500 -x
For the Clusterwise, we also used C 4 and C 2 as options.
All the outputs for both tstats1 and tstats2 didn't have significant differences (displaying p-values 0.95 to 1) Only around 0.7- 1, we saw some differences, and that is our main problem.
(a code for viewing)
fslview MNI152_T1_1mm mean_FA_skeleton -l Green -b 0.2,0.8 tbss_tstat1 -l Red-Yellow -b 0.95,1 tbss_tstat2 -l Blue-Lightblue -b 0.95,1
|