Hi,

On 15 Jan 2009, at 16:40, Cibu Thomas wrote:

> Hello there
> I am using TBSS to examine whether differences in FA can account for
> differences in % signal change in specific functionally defined ROIs.
> Based on some of the postings I saw on the forum, I followed a
> series of
> steps that I thought makes sense given my objectives.The outcome
> didn't
> look very exciting and I am writing to check whether, atleast, what I
> have done so far is correct. So here goes:
> 1. I ran the following commands for a data set comprising 8 subjects
> tbss_1_preproc *.nii.gz
> tbss_2_reg -T
> tbss_3_postreg -S
> tbss_4_prestats 0.2
>
> 2. For the correlation analysis, I used the GLM gui as follows
> In the FSL main console:
> Misc -> GLM setup - > Higher level non-time series design - > # time
> points (inputs) = 8
> In the EVs tab:
> # of EVs = 1, Group = 1, Paste the EV1 (% signal change in ROI)
>
> In the contrasts & F-tests tab:
> Contrast = 1; F-test = 0
> C1 enabled, Title = Group Mean, EV1 = 1
> I saved this design in the stats folder and ran the following
> randomise
> command:
> randomise -i all_FA_skeletonised -o RH_OFC_M_BOLD_FA_results -m
> mean_FA_skeleton_mask -d RH_OFC_M.mat -t RH_OFC_M.con -D -n 250 -T -V

For TBSS you should use --T2 not -T

For initial analyses -n 250 is ok, but Tom would say it's much too low
for real final accuracy (he says 5000).

You probably want to look for negative as well as positive
correlations, so maybe add a second contrast with -1.

Make sure the values in EV1 are zero-mean - i.e. you've subtracted the
mean of all values.

Cheers.

> To view the results:
> fslview $FSLDIR/data/standard/MNI152_T1_1mm mean_FA_skeleton -l
> Green -b
> 0.2,0.8 RH_OFC_M_BOLD_FA_results_max_tstat1 -l Red-Yellow -b 3,6
>
> My expectation was to see small patches of red-yellow in interesting
> tract regions, which was not the case, all I could see was the mean FA
> skeleton.
>
> So it looks like either there were no correlations or I might have
> done
> something wrong especially in setting up the stats design.
> I would greatly appreciate it if someone could help me rule out the
> latter possibility.
>
> Many thanks in advance
>
> Cibu Thomas
>


---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------