Dear Siamak,
>
> I have a problem with fnirt when using it to register two anatomical
> images
> with small volume for refmask such as hippocampus mask. When I run the
> command with the following options:
>
> fnirt in image1 ref image2 ioutimage3 refmask=ref.nii.gz config
> file T1_2mm
> and affine initial matrix.
> It sticks in a loop and doesn't register two images even after a
> day. I'm not
> sure that this problem occurs when using small refmasks since I had
> the same
> problem sometime using larger refmasks.
You will need to change your config-file to take the high resolution
of your --ref image into account. I would suggest using something like
--subsamp=8,8,4,4,2,2
for your sub-sampling scheme. As a first attempt I would suggest
leaving the other parameters as they are.
When trying to use the full resolution of a high-res structural fnirt
will use lots of RAM, and if/when your machine starts swapping it will
appear very slow. N.B. that the final registered image will still have
the full resolution of the --ref image.
Also, I am not sure why you would want to use a mask for hippocampus.
Unless your brains are very different (e.g. lots of atrophy in one and
not in the other) registering also the surrounding structures will not
affect the registration of the hippocampus negatively.
Good luck Jesper
|