Nick,
I tried my best to align the monkey brain with Talairach brain. The midline cut looks much better than the previous one. But the horizontal cutting plane does not look good. I repeated the process several times and it is always at the level of STG/STS as attached. Is there any way to get around this problem? It does not have to be automatic. I don't mind intensive manual labor.
Thanks, Kazu
-----Original Message----- From: Nick Schmansky [mailto:nicks@nmr.mgh.harvard.edu] Sent: Fri 1/26/2007 2:26 PM To: Tanji, Kazuyo (NIH/NIMH) [F] Cc: Freesurfer Mailing List Subject: RE: [Freesurfer] -noaseg
Kazu,
I should have mentioned this earlier, but when using the -noaseg flag, the talairach transform needs to be correct in order for mri_fill to properly cut the hemispheres vertically and the pons and cerebellum horizontally. It is obvious from your images that mri_fill is not cutting in this direction (and yes, the cerebellum should not appear in the filled.mgz, but it should appear in brainmask.mgz). I'm guessing that the transform, which is one of the early steps in recon-all, is off.
Follow the instructions here:
http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/Talairach
to align the monkey brain with the (human) talairach brain, as best as you can. Use this command:
tkregister --mgz -s <subjid> --fstal
which excludes the '--surf orig' option (since you don't yet have the orig surface).
Once you feel you have a proper alignment (it can take awhile playing around with tkregister2 to get this, but it doesnt have to be perfect), then you can restart recon-all:
recon-all -s <subjid> -autorecon2-wm -noaseg
The cc and pons points were saved from the prior run, so they dont need to be specified. You will see them in the recon-all.log file.
Actually, once the mri_fill step of recon-all creates the new filled.mgz, you may want to inspect it to make sure it cuts correctly, otherwise the topology fixing step will get stuck again on the huge defect.
Nick
On Fri, 2007-01-26 at 10:04 -0500, Tanji, Kazuyo (NIH/NIMH) [F] wrote:
Nick,
Thanks for the suggestion. I saw 2 abnormal perpendicular lines inside white matter volume (as attached). Does it result from failure in defining cutting planes? Also I thought at this stage the cerebellum should not be displayed anymore. Is there ways to correct those manually?
Best, Kazu
-----Original Message----- From: Nick Schmansky [mailto:nicks@nmr.mgh.harvard.edu] Sent: Thu 1/25/2007 5:12 PM To: Tanji, Kazuyo (NIH/NIMH) [F] Subject: RE: [Freesurfer] -noaseg
Kazu,
In looking at the log, it appears to be stuck on correcting a defect which has 27024 vertices, which is a huge number. This indicates something is wrong with an earlier stage.
Type:
tkmedit <subj> orig.mgz -aux filled.mgz
and make sure that the filled.mgz volume looks like it is properly representing the white matter voxels (use Ctrl 1 and Ctrl 2 to flip between volumes).
Also:
tksurfer <subj> lh orig
will bring-up the tessellated 'filled' volume. Look for obvious problems, like dimples or holes.
You may need to edit the white matter to fix problems. See:
http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/WhiteMatterEdits
Also make sure your skull-strip worked properly:
http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/SkullStripFix
That is, make sure there are not bits of skull in brainmask.mgz volume.
Nick
On Thu, 2007-01-25 at 14:45 -0500, Tanji, Kazuyo (NIH/NIMH) [F] wrote:
Nick,
Thanks for offering to take a look. I am attaching the log file.
Best, Kazu
-----Original Message----- From: Nick Schmansky [mailto:nicks@nmr.mgh.harvard.edu] Sent: Thu 1/25/2007 2:04 PM To: Tanji, Kazuyo (NIH/NIMH) [F] Subject: RE: [Freesurfer] -noaseg
-----Original Message----- From: Nick Schmansky [mailto:nicks@nmr.mgh.harvard.edu] Sent: Thu 1/25/2007 2:04 PM To: Tanji, Kazuyo (NIH/NIMH) [F] Subject: RE: [Freesurfer] -noaseg
Kazu,
No it should not take so long...it usually takes no more than 10 hours. So this indicates something is unusual about the surface. Can you send me the recon-all.log file (in the subjects scripts dir)? I can check for problems in that.
Nick
Hi Nick,
Thanks so much for the suggestion. I specified the volume index and ran recon-autorecon2-wm -noaseg. I started 5pm on Januray 22 (Monday) and it has not finished processing. It is still at 'mris_fix_topology' (left hemisphere). Could you suggest whether it is unusual for this step to take this long or it is understandable given the non-standard method I am trying? My CPU is quite new (Pentium D) and RAM is 2G.
Thanks, Kazu
From: Nick Schmansky [mailto:nicks@nmr.mgh.harvard.edu] Sent: Mon 1/22/2007 2:03 PM To: Tanji, Kazuyo (NIH/NIMH) [F] Cc: Freesurfer Mailing List Subject: RE: [Freesurfer] -noaseg
Kazu,
This is for a non-human primate, right? If so, you will need to manually define seed points for the corpus callosum and pons. To do so, open the white matter volume in tkmedit:
tkmedit <subjid> wm.mgz
and using the cursor, identify the 'volume index' of the point where you think the corpus callosum is located and make a note of it (for instance, in an example subject 'bert', its around 128 117 123). Also make a note of where the pons is located.
Then, restart recon-all from mri_fill, passing those coordinates, for example:
recon-all -s bert -autorecon2-wm -noaseg \ -cc-crs 128 117 123 \ -pons-crs 128 154 108
Nick
ps, i've posted this to the list as others may have a better suggestion.
On Mon, 2007-01-22 at 12:55 -0500, Tanji, Kazuyo (NIH/NIMH) [F] wrote:
Hi Nick,
Thank you so much. It worked. Now I got an error message at 'mri_fill'.
The message says: mri_fill: could not find corpus callosum INFO: Modifying dst c_(r,a,s), using the transform dst find_cutting_plane:seed point not in structure! Searching neighborhood... recon-all exited with ERRORS at Mon Jan 22 10:57:43 EST 2007
Is this error due to -noaseg flag? What should I do next?
Kazu -----Original Message----- From: Nick Schmansky [mailto:nicks@nmr.mgh.harvard.edu] Sent: Sun 1/21/2007 6:38 PM To: Tanji, Kazuyo (NIH/NIMH) [F] Cc: Freesurfer Mailing List Subject: Re: [Freesurfer] -noaseg
Kazu,
It looks like the -noaseg option does make use of norm.mgz (a change that seems to have occurred since the option was created and tested).
As a work-around, you should be able to do this:
ln -s brainmask.mgz norm.mgz
and then run the -autorecon2 -noaseg flags. brainmask.mgz is a close enough stand-in for norm.mgz.
Nick
On Sun, 2007-01-21 at 17:45 -0500, Tanji, Kazuyo (NIH/NIMH) [F] wrote:
Hi,
I have a question on how to use -noaseg option.
I tried a script: recon-all -autorecon2 -noaseg -subjid xxx.
This seems to skip several steps of autorecon2 and start from
'mri_normalize' . This results in an error because norm.mgz is not existing.
Could you suggest better ways to use -noaseg?
Thanks, Kazu
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
