Dear Diederick,

I am afraid it is impossible to say anything from that description. You say that the functional->MNI is particularly bad? Does that mean that the structural->MNI looks ok?

I have seen this on occasion when fnirt have "found" large warps right outside the edge of the brain and where slight functional->structural misregistration have moved voxels from the edge of the cortex into those large warps. It creates a characteristic pattern which looks almost like rays of light emerging from the subjects brain. When this happens it is typically caused by high intensity values in the "meningies" in your structural and is often a consequence of very short TE.

If this is what your images look like you might try to change

--applyrefmask=1,1,1,1,1,1

to

--applyrefmask=0,0,0,0,1,1

or possibly evel

--applyrefmask=0

in your local T1_2_MNI152_2mm.cnf file. This means that it will match up also the scalp which will help keeping the warps within bounds.

If the description above doesn't pertain to you case then it is probably best you upload the relevant parts of the reg directory and I'll have a look at it.

Jesper


On 22 Nov 2010, at 14:36, Diederick Stoffers wrote:

Hi guys,

I am working with some old fMRI data that I am reanalysing using the latest available tools including FNIRT, I expected this might yield far better results as most of our group differences are located subcortically (periventricular).

I was disappointed with the registration results in feat, especially the registration of the functional to MNI is pretty bad. Normal two-step linear registration with 6 and 12 DOF does just fine. I have so far been unable to pinpoint the exact cause. I have had great results with fnirt in the past (that were 3T data that had been B0 corrected, these are 1.5T data with no B0 map available).

Could anyone give me some pointers on what is causing this unsatisfactory registration and, more important, what I can do to remedy it? I can upload an overview of a particularly bad registration

Cheers,

Diederick