This has been reported before.
It isn't a problem with the spherical convolution it is just very, very
tiny inaccuracies in the FFT implementation for larger spheres.
You only see the problem if you follow it with "-bin" in fslmaths.
If you instead follow it with "-thr 1e-10 -bin" then it should work fine.
All the best,
On 5 Aug 2011, at 15:41, Andres Roman wrote:
> Hi FSL crew,
> I am trying to create a ROI sphere in the DLPFC of 10mm but it seems to
> crash everytime I do it.
> I used the following comands to create a point first,
> fslmaths MNI152_T1_2mm_brain_mask -roi 128 1 162 1 100 1 0 1 mask -odt float
> This works perfect. But then when I try to expand this to a 10mm sphere I
> get a very weird result (which I attach as a pic from fslview). I used the
> following command to expand it:
> fslmaths mask.nii.gz -kernel sphere 10 -fmean -bin spheremask_10mm -odt
> After this I used a smaller size of sphere (4mm) and it worked fine as you
> can see in the other pic I attach. I was wondering why this could happen.
> Thank you once more for your help.
> Best wishes
> Andres Roman
> <10mm ROI sphere.gif>