I'll try your suggestion. Thank you very much.

Hi Gabor

you don't want to specify -noaseg. That tells recon-all not to use the aseg. I think autorecon2-cp and autorecon3 should do the trick

cheers Bruce On Tue, 26 Sep 2017, Gabor Perlaki wrote:

Dear Bruce,

I've corrected the lateral ventricles in the aseg.presurf.mgz and ran "recon-all -autorecon2-noaseg -autorecon3 -subjid". Although the labels in aseg.mgz is fine the surfaces (and midline cut) remained bad. Any other idea how to correct this type of error?

Best Regards, Gabor

no, I don't think that is the case. -noaseg sets UseAseg=0, which stops things like mri_fill from using the aseg, which is not what you want

cheers Bruce On Tue, 26 Sep 2017, Gabor Perlaki wrote:

Dear Bruce,

From the recon-all script it seems to me that autorecon2-noaseg, autorecon2-wm and autorecon2-cp are totally equivalent (the same flags are defined to be true in all three cases). Could you confirm this?

Best regards, Gabor

oh, sorry, you are correct. I guess you need to edit the aseg.presurf.mgz since that is what mri_fill is given as a segmentation. Sorry for the mistake Bruce

On Tue, 26 Sep 2017, Gabor Perlaki wrote:

If I accurately understand the recon-all script, then -noaseg sets UseAseg=0, but "autorecon2-noaseg" is a different flag, which only sets the following things.

case "-autorecon2-cp":
case "-autorecon2-noaseg":
case "-autorecon2-wm":
  set DoNormalization2 = 1;
  set DoSegmentation   = 1;
  set DoFill           = 1;
  set DoTessellate     = 1;
  set DoSmooth1        = 1;
  set DoInflate1       = 1;
  set DoQSphere        = 1;
  set DoFix            = 1;
  set DoMaskBFS        = 1;
  set DoWhiteSurfs     = 1;
  set DoSmooth2        = 1;
  set DoInflate2       = 1;
  set DoSegStats       = 0;
  set DoCurvStats      = 1;
  breaksw

Best Regards,

Gabor

hmmm, if you tar, gzip and upload the subject I'll take a look.

Bruce On Tue, 26 Sep 2017, Gabor Perlaki wrote:

I've already corrected aseg.presurf.mgz and the left and right lateral ventricles are now correct in the final aseg.mgz. However the pial surface is almost the same as earlier:

https://mega.nz/#!XAZkWLZa!a6NgkEtiHzEqUkhZ6WimPgUKxytsGHKrgOLfyHdUuEM

https://mega.nz/#!XRRHna7Q!vaeSLG5Cv-fGW-0xdEJlgFTNRPm7qGHOlShYh7y5z1g

great, we will take a look and get back to you Bruce On Wed, 27 Sep 2017, Gabor Perlaki wrote:

Dear Bruce,

I sent you a link for our subject's data by email.

Best regards, Gabor

Hi Gabor

if you fix the aseg labels I think it should correct the midline problem (rerun autorecon2-wm autorecon3 I think)

cheers Bruce

On Mon, 25 Sep 2017, Gabor Perlaki wrote:

Here are links to images showing the bad midline cut. Could anybody help me how to solve this issue? The main cause seems to be asymmetric lateral ventricles.

Best, Gabor

2017-09-23 8:18 GMT+02:00 Gabor Perlaki petzinger.gabor@gmail.com: Dear all, Here is my question again. Somehow, it didn't appear on the Freesurfer forum. I'm resending it without the attachments.

I've a subject, where the midline cut is bad (see the attached pictures), therefore the segmented left lateral ventricle also includes considerable number of voxels from the right one. Is it possible to correct it somehow?

-- Gabor Perlaki research associate Diagnostic Center of Pécs H-7623 Pécs, Rét str. 2. Tel.: 0036-30-2084367 E-mail: petzinger.gabor@gmail.com