To follow up, it makes sense that the noise might be an issue, but it seems like version 5.3 creates the surfaces just fine regardless of the noise, so I'm wondering if perhaps there has been a chance in preprocessing before the end of the T1.mgz computation that solves the problem.

Would there happen to be any changes that were made to the normalization step that might be the reason why version 5.1 fails to create complete surfaces for these subjects while version 5.3 is working?

Thanks again for your help!

On Fri, Apr 25, 2014 at 12:49 PM, Maia Pujara <maiapujara@gmail.com> wrote:

Hi all,

After running 200+ subjects through recon-all using version 5.1, all but 30 came back with complete surfaces. The 30 that failed seem to have the same problem with normalization.

After re-running these 30 subjects with v5.3, we were able to get a good quality brainmask.mgz output, which was not the case when we ran these subjects through v5.1. See attached for comparisons of v5.1 and v5.3 outputs.

Would we need to be adding additional flags to get 5.1 to produce usable results for these subjects? Or might there be some other workaround for this?

Thanks in advance!

Maia