Hi,
See responses below:
> Here's a question that came up today and that may change completely the way I've been analyzing my fMRI data with FEAT. Considering an analysis pipeline in which a 2nd-level analysis is *never* performed (e.g. individual differences study), we know that:
>
> i. After completing the 1st-level FEAT analysis, a filtered 4D file is created. This file is in the same (typically low-res, 3x3x3mm for example) space as the original data, so no registration is applied at this point. (by the way, is the filtered data motion corrected?)
Yes, the filtered data is motion corrected.
> ii. Registration files are saved inside the /reg directory. If you manually apply these transformations to the filtered data, it will be registered to the space chosen as "Standard space" within the Registration tab in FEAT. If the high-resolution anatomical volume was chosen as the Standard Space, the filtered data will be upsampled to a higher resolution.
That's correct.
> iii. Stat files are saved inside the /stats directory. Take tstat, for example. This is a 3D file, where each voxel is aligned with the motion-corrected filtered data, but still in the low-resolution space
Correct.
> Now, let's suppose you define some masks in the low-resolution space, based on tstat maps you got from one scan. You want to apply this mask onto tstat maps from other scans.
I don't really follow why you start off saying that you never want to do second level analysis, but now you want to transfer images from one scan to another (where I assume "scans" are separate experiments - that is, separate first-level analyses). But I'll answer as best I can.
> My questions, then, are:
>
> 1) Can you apply the same mask to tstat maps from different scans? In other words, are tstat maps voxels, from different scans, aligned between each other?
Scans within one analysis, in the low-res space, are all aligned with the example_func (which is the target of motion correction). Between different first-level analyses there is no alignment of low-res scans or anatomical scans. Only standard space is common across different first-level analyses.
> 2) If not, what do I need to do if I want to apply a single mask to 1st-level tstat maps from different scans? I don't want to upsample my data to the standard (in my case, anatomical) space (in which case I know I should be using the flirt command with the -applyxfm option), which would guarantee that every map is registered to the standard space. Is that the only option I have?
You should combine the transformations from the example_func where the mask is defined, to the standard space and then from standard space to the new example_func space. That is: example_func2highres_warp (or mat, if you did not use any fieldmap correction) and highres2standard_warp plus the inverses of these for the other first-level analysis. You can combine these with convertwarp, and then apply the combined one (that goes from one example_func to the other) with a single call to applywarp. This will avoid creating any upsampled versions.
Mind you, it would be simpler if you just had a mask in standard space, but as I don't really understand what you are trying to do then I'm not sure what is possible or not.
Anyway, I hope this is of some help.
All the best,
Mark
|