There is a known problem with FNIRT/the template that affects these areas for some scans. It has to do with there being relatively low signal from fat inside the skull bone in the template, but sometimes we see quite high signal from these same areas in newer scans. This high signal is typically present (when it is present) in the topmost slices.

There are two reasons why we see this signal these days. One is that T1 sequences these days tend to have shorter TE than at the time that the data for the template was acquired. The other is that the fat in the skull bone tends to increase with age and the scans for the template comes from young(ish) subjects. So if you scan older subjects there is a greater risk you will come across this problem.

I am afraid that at present there isn’t much I can do to help if this is indeed the problem. We have started a project for improving FNIRT to hopefully be better able to handle this (and other) problem. But that is till a little time off.

On 1 Dec 2017, at 11:59, Giulia Forcellini <[log in to unmask]> wrote:

Dear FSL experts,

I'm experiencing some problems with non linear registrations that I really can't figure out.

After using FLIRT, registering the structural brain to the MNI template, I apply FNIRT for the non linear registration. However, I have few subjects in which FNIRT always seems to fail, as in the final outputs I get some very weird warps on the brain (Looks a lot like a badly deformed brain). I tried to change all the possible cost functions, to increase the warping resolution to 6 and change Lambda to 400, 200,150, 75, 60, 45, but the problem is always there.

Does someone have any idea of what can be causing these deformations? Any suggestion on how to solve this?

Thank you so much!!

All the best,

Giulia