Hi,
In this situation we strongly recommend that people acquire a single whole-brain EPI volume that matches your reduced FOV acquisition in terms of resolution and distortion. This normally just requires setting up the same sequence with more slices and can be acquired in 10-20 seconds. Having such an image hugely helps the registration, as then it can register the reduced FOV to the whole-brain EPI and then the whole-brain EPI to the structural. However, I assume that you do not have this, in which case things are a lot harder.
Another thing is whether you have fieldmaps or not, as the distortions and signal loss will also be making the registration harder and we strongly recommend that people acquire fieldmaps in all fMRI (or dMRI) experiments. A blip-up-blip-down pair can also be used to derive a fieldmap, in case you have this with some dMRI data from the same session.
If you have neither fieldmaps nor a whole-brain EPI, then the biggest difficulty will be the SBRef to structural registration. I would recommend that you: (a) use fslreorient2std on both fMRI and structural images; (b) use robustfov on your structural image; (c) use -nosearch option in FLIRT; (d) generate an initial matrix using the -usesqform option with a command like:
flirt -in SBRef -ref structural_brain -usesqform -applyxfm -omat usesqform_init.mat
(e) try a 3 DOF (translation only) alignment (using the $FSLDIR/etc/flirtsch/xyztrans.sch schedule file) and -init with the useqform_init.mat; (f) also try a 6 DOF registration with BBR using the epi_reg script.
If the result from (e) looks better then you can try to refine this with a BBR registration using:
flirt -in SBRef -ref structural_brain -dof 6 -cost bbr -wmseg structural_fast_wmseg -init result_3dof.mat -omat bbr2.mat -out bbr2 -schedule ${FSLDIR}/etc/flirtsch/bbr.sch
Note that whenever using BBR you need to supply a white matter segmentation of the structural image.
Hopefully this will then give you a good result.
All the best,
Mark
> On 30 Jun 2017, at 16:30, Usman Sheikh <[log in to unmask]> wrote:
>
> Dear FSL Experts,
>
> I am working on fMRI data collected with a reduced FOV (24 slices, 3.5 mm each), multiband factor of 6 and TR of 265ms. I have collected a structural scan and five functional 4D images from 5 sessions of the experiment. Now I am using FLIRT to register those functional images to standard space. But the problem is, that the results from those registrations are very poor. I am attaching one example with this mail. It's an extreme case where both translation and rotation are completely wrong but in 90% of them, there is orientation issue, where the bottom edge of the reduced-FOV-image has been matched with the bottom edge of the standard image.
>
> My workflow goes as follows:
> Say, I have run1.nii, run2.nii, and so on and their corresponding high contrast reduced FOV reference images SBRefs i.e. run1_SBRef.nii, run2_SBRef.nii and so on. And I also have a structural scan i.e. structural.nii. First, I register run1.nii to run1_SBRef.nii and then to structural_brain.nii (brain image extracted from structured.nii) and then to standard space.
>
> Now, I am going to try a few options in FLIRT but it will be great if you could guide me in the right direction.
>
> Thanks alot,
> Usman Ayub Sh,
> PhD Researcher @ BCBL
> <run2-cropped.jpg>
|