Hi Bruce,

Thanks for the reply. Would 'bbregister' work to register the files to eachother or would there be a more useful way to do that?

Also I noticed that across many of the subjects there is a slight difference in the TR, TE, TI, and flip angle values.

For instance, one of the subjects has a TR: 9.39 and another might have a TR: 9.44. This clearly prevents me from using them to create an atlas. I am not really sure why these values are slightly off, they should be the same. Is there anyway to change those values manually? Is it because I am using the T1.mgz files?

Thank you!

- MP

On Tue, Jul 23, 2013 at 3:15 PM, Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote:

Hi Mark

have you tried using a transform? Usually with more than one subject we require that they be registered. I haven't looked at that code in a long time, so I am not positive that it would fix this problem, but it might. In any case if they aren't registered it won't generate a useful atlas, even if it does run to completion
Bruce

On Tue, 23 Jul 2013, Mark Plantz wrote:

Hello freesurfers,
I recently ran into a problem with mri_ca_train, a function that creates
a .gca atlas from a manually segmented volume and input volume(s).

My command line is:

mri_ca_train -T1 T1.mgz CMH-11460 CMH-11461 ~/Desktop/atlas.gca

The results:

processing subject CMH-11461, 1 of 1...
reading input 0:
/Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-11461/mri/norm.mgz
filling holes in the GCA...
404261 classifiers: 1.5 per node, 7.59 in brain (0 holes filled)
computing covariances for subject CMH-11461, 1 of 1...
average std = 6.1 using min determinant for regularization = 3.7
298584 singular and 4268 ill-conditioned covariance matrices regularized
writing trained GCA to /Users/IngvalsonLab/Desktop/atlas1.gca...
classifier array training took 0 minutes and 27 seconds.
dhcp-165-124-23-230:LabMRIData IngvalsonLab$
dhcp-165-124-23-230:LabMRIData IngvalsonLab$
dhcp-165-124-23-230:LabMRIData IngvalsonLab$ mri_ca_train -T1 T1.mgz
CMH-11460 CMH-11461 ~/Desktop/atlas1.gca
reading T1 data from subject's mri/T1.mgz directory
training on 2 subject and writing results to
/Users/IngvalsonLab/Desktop/atlas1.gca
***************************************************************************
processing subject CMH-11460, 1 of 2...
reading input 0:
/Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-11460/mri/T1.mgz
***************************************************************************
processing subject CMH-11461, 2 of 2...
reading input 0:
/Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-11461/mri/T1.mgz
filling holes in the GCA...
468392 classifiers: 1.8 per node, 10.41 in brain (0 holes filled)
computing covariances for subject CMH-11460, 1 of 2...
GCAupdateNodeCovariance(17, 35, 28, 4): could not find label
dhcp-165-124-23-230:LabMRIData IngvalsonLab$

According to the fswiki, the same bug was eliminated in the 2009 version.
For some odd reason, this command works with one subject, but not with
multiple subjects.

Any ideas?

Thanks in advance,

MP

The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.