I've made the fix (added -T 1 to the invokations of mri_mask). Locally at the NMR center, it will be live tomorrow. Not sure what the best way to get the change to you. I can post the new script, but I don't know what version you are using. A new version might not be compatible with what you've done already. You can also edit your version to add -T 1.
doug
Bruce Fischl wrote:
Hi Yuhong,
thanks for catching this - it is indeed a bug that we (meaning Doug :>) are fixing.
Bruce On Mon, 9 Jan 2006, Yuhong Jiang wrote:
Hi Doug,
For mri_mask's threshold,
"mri_mask -T 1 T1.mgz brain.mgz brain.mgz" works, but "mri_mask -T 0 T1.mgz brain.mgz brain.mgz" does not work, and "mri_mask T1.mgz brain.mgz brain.mgz" does not work either.
So one should specify the threshold to be at least 1. The default used by -autorecon2 (or normalization2) is not sufficient.
--Yuhong
Yuhong Jiang yuhong@MIT.EDU 617-876-2054 http://web.mit.edu/yuhong/www=============================================================================
From: greve@nmr.mgh.harvard.edu greve@nmr.mgh.harvard.edu To: Yuhong Jiang yuhong@mit.edu Cc: freesurfer@nmr.mgh.harvard.edu, fischl@nrm.mgh.harvard.edu Date: Jan 7, 2006 2:43 PM Subject: Re: [Freesurfer] Re: mri_mask bug Reply | Reply to all | Forward | Print | Add sender to Contacts list | Delete this message | Report phishing | Show original | Message text garbled?
it sounds like there's something that went wrong with the skullstripping if you have to use a threshold of 20. Have you looked at brain.mgz immediately after skullstripping? The background voxels should be 0.
On Fri, 6 Jan 2006, Yuhong Jiang wrote:
Hi Bruce,
The exact input is:
mri_mask T1.mgz brain.mgz brain.mgz
This command was invoked by "recon-all -subjid subject -normalization2", which is part of "recon-all -subjid subject -autorecon2". Because of the bug, after this command, the manual edited brain.mgz went back to pre-manual editing state.
It appears that by adding a threshold (mri_mask -T 20 T1.mgz brain.mgz brain.mgz) the mask is properly executed. However, this does mean that one cannot run autorecon2 or normalization2, which calls for mri_mask without specifying a threshold. Perhaps someone should fix this command (as well as autorecon2 and normalization2)??
--Yuhong
=============================================================================
Yuhong Jiang yuhong@MIT.EDU 617-876-2054 http://web.mit.edu/yuhong/www=============================================================================