Print

Print
Hi Ellen,

Yes, the images need to be all in a common space, which means using the *_to_target images.

There are indeed 2 ways of doing a paired t-test: one is using subtractions, then sign-flippings in randomise as a 1-sample t-test (option -1), which is how you are doing, and the other is putting a full paired t-test in the design matrix and using exchangeability blocks (option -e). Both ways will give the same result, and regardless, both need to have the images in the standard space.

All the best,

Anderson



On 31 July 2014 07:32, Ellen H <[log in to unmask]> wrote:
Dear FSL community
I have a single group of individuals with baseline and follow-up DTI scans (varying interscan intervals). I'm using fsl maths to calculate the difference between each individual's two scans (fslmaths subject1_rA -sub subject1_rB sub1_diff) and merging to create a 4D image, as suggested in the wiki page.

Which two files should I be using as the input though?

When I try to use individuals' baseline and follow-up FA files, a warning message comes up saying "Inconsistent orientations for individual images in pipeline! Will use voxel-based orientation which is probably incorrect". Presumably, this is because they haven't been normalised to the same template yet? So does that mean I should be using target files instead e.g. BL0005AS_FA_to_target.nii.gz -sub FP0005AS_FA_to_target.nii.gz sub1_diff?

Also, the FSL wiki also has a FEAT setup which seems quite a different approach to the above. Could anyone explain which one is appropriate when?

Thanks very much for your help!
Cheers, Ellen.