Correct. Under standard space I selected nonlinear and put in the required '_brain' files under "main structural image".

Looking in the "reg" subdirectory of the resultant .feat directory I found no highres2standard_jac.nii.gz file. What's extra funny, though, is that neither the filtered_func_data.nii.gz or the melodic_IC.nii.gz files are registered to MNI152. They're all still in subject space. I don't know if this means something failed, but I don't see any striking error messages in the overall FEAT reports.

The only "error" I can find under the 'Registration' portion is:
did not find file: example_func2highres.mat. Generating transform.
did not find file: highres2standard.mat. Generating transform.

I've since found a way (through various google searches) to produce the highres2standard_jac.nii.gz. However, that doesn't fix the other problem I have of all my output files being in subject space rather than MNI152.

With respect to that problem, I've tried using featregapply, as per Steve Smith's suggestion earlier, but I'm getting some error message I don't understand:

echo 'set fmri(inmelodic) 1' >> RestState_1_FSL_LAS.feat/design.fsf
featregapply RestState_1_FSL_LAS.feat -f
wrong # args: should be "set varName ?newValue?"
    while executing
"set fmri(inmelodic) 1 featregapply denoised_data.nii.gz -f"
    (file "design.fsf" line 916)
    invoked from within
"source ${filename}"
    (procedure "feat5:load" line 58)
    invoked from within
"feat5:load -1 1 design.fsf"
    (file "/extern/soft/tools/fsl-5.0.4/bin/featregapply" line 164)






On Thu, Sep 11, 2014 at 3:14 PM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Paul,

Are you sure you are looking in the right place or in the outputs for subject that had indeed undergone non-linear registration? By this I mean: in the 'Registration' tab in FEAT or in Melodic GUI you specified the 'Main structural image' (you T1 in native space), and you marked the checkbox 'Non-linear' in the 'Standard space' box. Then you made sure that it ran successfully for that subject, and you are looking into the subdirectory 'reg' of the *.feat or *.ica directory of that same subject. There should be a highres2standard_jac.nii.gz file there. That is the Jacobian that you'd use for the modulation of GM.

All the best,

Anderson


On 11 September 2014 16:20, Paul Beach <[log in to unmask]> wrote:
Hi Anderson,

It looks as though I actually don't have the highres2standard_jac.nii.gz file in my reg directory. How do I go about making this file?

Thanks


On Thu, Sep 4, 2014 at 10:37 AM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Paul,

Have a look if you don't have already the file highres2standard_jac.nii.gz in the 'reg' directory (along with the highres2standard_warp.nii.gz). It's probably there, then you won't need to run the fnirtfileutils, and it's one step less to do. :-)

The highres2standard_pve_1.nii.gz is the GM partition. This is the file you need to modulate.

All the best,

Anderson



On 3 September 2014 16:50, Paul Beach <[log in to unmask]> wrote:
Anderson,

After getting all my subjects re-ran through FEAT and using non-linear registration. I then ran a test subject's MNI152 transformed brain through FAST, as we previously discussed. 
/extern/soft/tools/fsl-5.0.4/bin/fast -t 1 -n 3 -H 0.1 -I 4 -l 20.0 -o /extern/research/PI/training/beachpau/FSL/SUBJECTS/HS_001/RestState_1_FSL_LAS.feat/reg/highres2standard /extern/research/PI/training/beachpau/FSL/SUBJECTS/HS_001/RestState_1_FSL_LAS.feat/reg/highres2standard

I obtained the following files:
highres2standard_pve_0.nii.gz
highres2standard_pve_1.nii.gz
highres2standard_pve_2.nii.gz
highres2standard_pveseg.nii.gz
highres2standard_seg.nii.gz
highres2standard_mixeltype.nii.gz

Looking at the fnirtfileutils command, which I need to use to produce the jacobian map, I see that I need to have an input fnirtcoefs file as well as a refvolume. I'm not really sure which of the above output qualifies for either of those. Is it one of these or could they include one of the previously existing files in the 'reg' directory - such as the highres2standard_warp.nii.gz file?

Thanks for your help


On Thu, Aug 28, 2014 at 10:40 AM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Paul,
Yes... it looks like it run with linear (affine) only...
All the best,
Anderson



On 28 August 2014 14:18, Paul Beach <[log in to unmask]> wrote:
Thank you Anderson.

A quick follow-up as I start the process. The files in the "reg" subdirectory have the following contents: 
example_func.nii.gz  highres2example_func.mat  highres.nii.gz

I'm thinking these (mainly the middle .mat file) might not be the correct files for step 2 as, looking at my design.fsf file, I have "set fmri(regstandard_nonlinear_yn)" at '0'.

So I guess I have to re-run things from the beginning, feat-wise?




On Thu, Aug 28, 2014 at 7:29 AM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Paul,

Almost:
1. Yes, that's it.
2. Take the warps that should be in the "reg" subdirectory of the .feat directory, and use fnirtfileutils to produce the Jacobian map. Of course, they will only be there if you used non-linear registration when running the preprocessing.
3. Use fslmaths to modulate the GM maps by the Jacobian, that is, multiply the GM maps by the Jacobian.
4. Take the modulated maps and merge them into a 4D file.

That's it. This 4D output goes as a voxelwise EV in randomise.

All the best,

Anderson



On 27 August 2014 17:52, Paul Beach <[log in to unmask]> wrote:
Anderson,

I think I understand (some of) what you're saying here.

1. Take T1 images for each subject that are already MNI152 registered and run them through FAST.
2. take the resultant GM masks and run them through fslmaths to produce the Jacobian map ... so, fnirtfileutils --cout on each of those masks?
3. do "something" with those jacobian maps to produce a final 4D file...this is the most confusing part. Are you referring to the final FSLVBM 4D file? I'm not really sure what I would be doing with fslmerge, the jacobian maps, and whatever other inputs to get the final 4D file.
4. Would this be the final 4D file I run through randomise? or would step 3 be performed after completing FSLVBM steps?

Thanks,
Paul


On Tue, Aug 26, 2014 at 5:37 PM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Paul,

It's probably a bit more complicated than that. To have the GM in the same common space as the functionals you need to use the same warps that were used to align the fMRI maps to that space, to then align the GM. To align the fMRI, the default in FEAT and in the Melodic GUI is a T1-w template, whereas for VBM, it's a GM template. This means you'd have to tweak the alignment of one of these.

Maybe a quick strategy could be to get, for each subject, the T1-w image that is already in the common space (it's used to drive the registration of the fMRI in the 1st level FEAT and in the Melodic GUI) and segment that image using FAST, then modulate the resulting GM partition with fslmaths (you can produce the Jacobian map with fnirtfileutils) and finally produce the 4D file with fslmerge.

All the best,

Anderson




On 26 August 2014 20:44, Paul Beach <[log in to unmask]> wrote:
Thanks for your response, Anderson.

So basically just the output file from fslvbm_3_proc?


On Mon, Aug 25, 2014 at 4:04 PM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Paul,

For this, you'd need the GM maps (modulated and merged as a 4D file) in the same common space as the functional maps used with randomise in the dual regression. In randomise, use then the options --vxl and --vxf.

All the best,

Anderson



On 25 August 2014 17:10, Paul Beach <[log in to unmask]> wrote:
FSL folks,

I'm interested in using VBM-based results as a regressor of no interest in my ICA analyses (such as in: Werner et al., 2013. Human Brain Mapping) . I want to make sure I have the process straight, though.

Specifically, my questions are: which VBM output file should be used as a voxel-wise EV? Then, how do I set that as a confound regressor, as voxel-wise EVs don't have the typical EV box for each subject?


Thanks
--
Paul Beach
DO/PhD candidate - Year VI
Michigan State University
- College of Osteopathic Medicine
- Neuroscience Program
 - MSU Cognitive and Geriatric Neurology Team (CoGeNT)




--
Paul Beach
DO/PhD candidate - Year VI
Michigan State University
- College of Osteopathic Medicine
- Neuroscience Program
 - MSU Cognitive and Geriatric Neurology Team (CoGeNT)




--
Paul Beach
DO/PhD candidate - Year VI
Michigan State University
- College of Osteopathic Medicine
- Neuroscience Program
 - MSU Cognitive and Geriatric Neurology Team (CoGeNT)




--
Paul Beach
DO/PhD candidate - Year VI
Michigan State University
- College of Osteopathic Medicine
- Neuroscience Program
 - MSU Cognitive and Geriatric Neurology Team (CoGeNT)




--
Paul Beach
DO/PhD candidate - Year VI
Michigan State University
- College of Osteopathic Medicine
- Neuroscience Program
 - MSU Cognitive and Geriatric Neurology Team (CoGeNT)




--
Paul Beach
DO/PhD candidate - Year VI
Michigan State University
- College of Osteopathic Medicine
- Neuroscience Program
 - MSU Cognitive and Geriatric Neurology Team (CoGeNT)




--
Paul Beach
DO/PhD candidate - Year VI
Michigan State University
- College of Osteopathic Medicine
- Neuroscience Program
 - MSU Cognitive and Geriatric Neurology Team (CoGeNT)