Dear Bruce,

I downloaded the recent build from 3rd January 2017 and tried to re-process the problematic subject.

Unfortunately, the posterior part of lateral ventricle is still not in large extent not correctly labeled.
Moreover, I noticed that the T2 normalization in ventricle part contains artificially low values. See the screenshot of aparc.a2009s+aseg, raw T2 and errors in normalized T2.

I have uploaded the updated reconstruction at your server with name aseg_error_fs_20170103.tar.gz so you can take a closer look.

Regards,

Antonin


Hi Antonin,

yes, we came across that bug and have fixed it (it's one of the reasons 
that we didn't release before Christmas)

Thanks for catching it thouggh
Bruce

On Wed, 4 Jan 2017, Antonin Skoch wrote:

> Dear experts,
> 
> while testing V6beta version I encountered problem with aseg. In one subject
> (with quite large ventricles) posterior part of right lateral ventricle is
> not labeled as brain at all and part is labeled as WM-hypointensities and
> nonWM-hypointensities. Moreover, in aparc.a2009s+aseg there are
> periventricular regions with value -1 (not present in aseg.presurf).
> 
> These negative values I observed in one more subject.
> 
> I am uploading the problematic subject named aseg_error.tar.gz to your
> server.
> 
> Linear Talairach registration seems OK in this subject.
> 
> Could you please comment on, how to correct this error?
> 
> My second question, is it possible to check non-linear Talairach
> registration, i.e. talairach.m3z?
> 
> Regards,
> 
> Antonin Skoch
> 
> 
>