Linear Registration of "V1 mask in MNI space" into "Functional Space" in FSL

Software Support
,
1

Summary of what happened:

I have a one hemisphere V1 mask in MNI space (1x1x1), and I try to do linear registration into functional space. The functional scans are 3x3x3 and they are not whole-brain scans. I tried using FLIRT as follows:

Command used :

flirt -in V1-LH.nii.gz -ref example_func.nii.gz -applyxfm -init MNI2func.mat -interp nearestneighbour -out myMask_in_func.nii.gz

or in two steps (first to anatomical then to functional space):

flirt -in V1-LH.nii.gz -ref highres.nii.gz -applyxfm -init standard2highres.mat -out MNI_mask_in_anat.nii.gz -interp nearestneighbour
flirt -in MNI_mask_in_anat.nii.gz -ref example_func.nii.gz -applyxfm -init highres2example_func.mat -out MNI_mask_in_func.nii.gz -interp nearestneighbour

I get a mask that is partially out of boundaries of brain and partially extend to other hemisphere. If the approach is not correct, how can we register this mask into functional space (linear reg.) using FSL?
(We want to use the mask in “Pre-threshold masking” option in FSL)

2

Hi @Tutku_Karahan and welcome to neurostars!

How did you create your transformation matrix, MNI2Func.mat?

Best,
Steven

3

Hi, thank you for your reply and thanks for the kind welcome.
From the previous FSL analysis I have done with the same functional data and MNI 1mm registration , there exists a feat file which has a reg file and inside that I get all of the transformation matrices. In that file, the naming was different (not MNI2func.mat), it was standard2example_func, and I also tried the same step with that file.
flirt -in V1-LH.nii.gz -ref example_func.nii.gz -applyxfm -init standard2example_func.mat -interp nearestneighbour -out myMask_in_func.nii.gz

I also tried to use convertxfm as another way of achieving MNI2Func.mat from highres2standard.mat and example_func2highres.mat :
convert_xfm -inverse highres2standard.mat -omat standard2highres.mat
convert_xfm -inverse example_func2highres.mat -omat highres2example_func.mat
convert_xfm -omat MNI2func.mat -concat highres2example_func.mat standard2highres.mat
The resulting masks were identical for both of the approaches as expected.

4

Hi @Tutku_Karahan, have you verified that the registration which was run as part of the FEAT analysis was successful, i.e. good alignment between the functional and standard space images?

Also, did you provide a field map to perform B0 unwarping? If so, the linear functional<->standard space transformation will only be approximate - you would need to use the inverse of the non-linear example_func2standard_warp.nii.gz transformation.

5

Hi,
Thank you for the answer. I checked the alignment between functional and standard space images, it was successful.
For the B0 unwarping, as a previous step before analysis I did magnetic field distortion correction with topup, unwarping the functional images. I used fslmerge to create a merged file using the two ‘blipped’ images (blip-up and blip-down) and generate a topup file by using topup command and apply this topup to functional files using applytopup command. So after these steps I input the unwarped functional file into FEAT analysis, i did not utilize field map option or provide a field map in FEAT. So, do I still need to use nonlinear transformation?
I also tried nonlinear transformation and it worked as a map, there was no problem. But our study requires avoiding nonlinear transformations as much as possible.

6

Would you be able to post some screenshots, of the original mask overlaid on the MNI152, and of the transformed mask overlaid on your functional data? Or if you could share your data using e.g. dropbox/google drive, I can have a look.