Hi,
Just to make sure things are clear - the applywarp that
Jesper suggests below and the output I talked about (using
the --iout option in fnirt) should be the same. Each one is
a warped version of the input image which should match
the reference image (MNI152_T1_2mm). You should always
look at this output and compare it with the reference image
(load both into fslview) in order to know if it has worked.
Having checked your original email again, I wonder how long
you waited before deciding that it was "stuck forever". The
sequence of output with the -v and --debug options starts
with the invwarp:: lines and will have min/max of 0/0, but
it should then show the same lines again, but with non-zero
values for min/max. There are then many output lines
when it is doing the optimisation.
If your original image is very large (including lots of
non-brain areas) then it can take quite a lot of time
and memory to calculate the inverse warp. So you
might need to just either wait longer or find a faster
machine, or one with more memory. It is probably
worth monitoring the memory usage with "top" to
see how it is going. Also, if you need to, then you
can also speed invwarp up and make the memory
usage better by cropping the original image (prior to
bet, flirt and fnirt) to a smaller FOV (just include the
brain plus enough around it to get to the outer
scalp). Your current FOV at 250x250x256mm
is likely to be more than you need (as the MNI152
images are 180x220x180mm).
All the best,
Mark
On 18 Oct 2008, at 09:05, Jesper Andersson wrote:
> Hi Anthony,
>
>
>>
>> I am not sure how fnirted_A (output image from fnirt) should look
>> like for it appears as a mixture of squares of different
>> intensities. It doesn't look like a brain image but does seem to
>> cover the shape of a brain.
>
> It is not see easy to interpret fnirted_A, since it is a 4D file of
> spline coefficients. If you want to check if fnirt did OK you are
> better off looking at what the warped image looks like. You can do
> that with
>
> applywarp --ref=MNI152_T1_2mm --in=A --warp=fnirted_A --
> out=A_InSpaceOf_MNI
>
> and then use fslview to look at A_InSpaceOf_MNI. If that looks good
> fnirt probably did alright.
>
> Good luck Jesper
>
|