Hi,
Have you tried it with brain extraction run on both images?
We generally recommend that. I suspect that this will fix it.
Also, for within-subject registration you should use 6 dof, not 12.
If this still doesn't fix it, and if you have good information from
your DICOM conversion, then try -usesqform as that can sometimes
be a good option to try.
All the best,
Mark
On 14 Mar 2011, at 18:53, Lazar Fleysher wrote:
> Hello everybody,
>
> I am sure, this question have been answered somewhere, but for some
> reason I cannot find a similar topic in the forum archive.
>
> Basically, I have a high resolution T1-weighted image (MPRAGE) and
> would like to coregister a low resolution T2-weighted image to T1.
> This is a standard problem.
> I use flirt to do it and the result I get is T2 image rotated by 90
> degrees around LR axis.
> (see the attached image) Grayscale is T1 and color is T2.
>
>
> The command which was used to attempt the coregistration is this
>
> flirt -in t2image.nii.gz -ref t1image.nii.gz -out coregimg2 -omat
> transformation2.mat -bins 256 -cost mutualinfo -searchrx -18 18
> -searchry -18 18 -searchrz -18 18 -dof 12 -interp trilinear
> -searchcost mutualinfo
>
> Because the images intially a almost "coregistered" I have restricted
> the angular search to speed up the calculation.
>
> Could some one suggest what I did wrong or maybe some options which I
> should use to attempt to resolve
> the problem?
>
> Thank you very much for your help
>
> Lazar
> <Screenshot-5.jpg>
|