Re: [Freesurfer] Issues with skull stripping

Can you send us the 1mm data so we can take a look?

Bruce On Wed, 15 Apr 2009, Falk Lüsebrink wrote:

Hi Freesurfers,

I’m currently working on four different subjects and I’m having problems with the skull stripping for three of them. The four subjects I’m trying to process are:

1.  0.6mm iso voxel size, being the same subject as 4)
   
   

2.  0.5mm iso voxel size
   
   

3.  0.8mm iso voxel size
   
   

4.  1mm iso voxel size
   
   
   
   

The only subject which works nearly perfectly fine is subject 1.

In case of 2 and 3 there are missing parts of the brain or even a whole hemisphere. I tried using the –wsgthresh switch following the tutorial on the wiki. In case of subject 2 the brainmask.mgz looks fairly good while setting the threshold to 80, but still some parts of the brain are missing and setting the threshold to 100 or 150 doesn’t change much. There is not even more skull left.

For subject 3 a threshold of 30 is sufficient and the whole brain is being shown, though some bits of skull and dura are not removed. I got a good result for the skull stripping with v4.0.5 by not conforming the data, but I haven’t been able to get the same result with v4.3.0.

In case of subject 4 autorecon1 exists with an error during the talairach check. I tried to follow the wiki for that issue, but haven’t been able to even roughly match the movable with the target. So I skipped the tal-check running into the same problem as subject 2 and 3.

The command lines I used were for subject 1 to 3 recon-all –autorecon1 –hires and for subject 4 just recon-all –autorecon1. I tried to use recon-all –autorecon1 –cm for subject 2 and 3, as this resulted in good skull stripping using v4.0.5. But recon-all exists with an error during mri_watershed: “GLOBAL region of the brain is empty!” I get the same error if I don’t use the –no-wsgcaatlas switch during recon-all –skullstrip, even if I didn’t use the –cm or –hires switches before.

As I’m not able to reproduce the same results between v4.0.5 and v4.3.0, have there been made changes to the processes of autorecon1 which I haven’t read of affecting the skull stripping? And does anyone have an idea besides manually doing the skull stripping?

Thanks in advance,

Regards,

Falk