how did you change the type to float? Make sure it doesn't scale the inputs if that is the segmentation volume. rebuild_gca_atlas.csh uses our compute cluster queueing system which is what the pbsubmit script is. You'll need to modify it to use whatever cluster you are using or to run them in series
cheers Bruce On Tue, 28 Aug 2012, Jeff Thompson wrote:
Hello, When running the mri_ca_train I am getting a Bus error near the beginning of it.
'bash-3.2$ mri_ca_train -parc_dir Alloybrain_mask_LUTComplete_float.mgz -T1 norm.mgz Alloy single_oneAlloy.gcareading segmentation from subject's mri/Alloybrain_mask_LUTComplete_float.mgz directory reading T1 data from subject's mri/norm.mgz directory training on 1 subject and writing results to single_oneAlloy.gca gcaAllocMax: node dims 64 64 64 gcaAllocMax: prior dims 128 128 128 gcaAllocMax: max_labels 4
processing subject Alloy, 1 of 1... reading input 0: Desktop/freesurfer/subjects/Alloy/mri/norm.mgz Bus error ' Previously it was giving an error saying that :"mri_ca_train: segmentation file Desktop/freesurfer/subjects/Buddha/mri/Alloybrain_mask_LUTComplete.nii.gz is not type UCHAR or FLOAT"
So I changed the file type to float to get rid of that. Any ideas?
When I run rebuild_gca_atlas.csh it gives a pbsubmit command not found error. If I have a cluster to use, but am able to just load everything onto it and then run it, can I comment out all of the pbsubmit lines?
Thank you for all the help,
Jeff
Date: Mon, 27 Aug 2012 12:00:13 -0400 From: fischl@nmr.mgh.harvard.edu To: jeff_rthompson@hotmail.com CC: freesurfer@nmr.mgh.harvard.edu Subject: RE: [Freesurfer] creating a gca file from .mgz files
Hi Jefff you can give mri_ca_normalize a flag to tell it to use a manual segmentation instead of the gca. Try -seg <manual seg volume>. Then the gca name won't matter - it should ignore it.
You could use label2label I guess, although if it's accurate across brains then you probably don't need it and can just build the gca from a handful of brains instead of the dozens we need for humans
cheers Bruce
On Mon, 27 Aug 2012, Jeff Thompson wrote:
Hello, Okay, but from looking at the recon-all pipeline it looks like to
generate
the norm.mgz file the .gca atlas is needed first. If I create the seg_edited.mgz file from my manually labelled file can I use that in rebuild_gca_atlas.csh to generate the .gca atlas? This would incorporate
the
mri_ca_train, is that correct?
The other question I have is regarding mri_label2label. If I have
manually
labelled one brain from my set, can I use mri_label2label to convert
these
labels to other dog brains in my set if they have been registered to the first brain used?
Thank you,
Jeff Thompson
Date: Sat, 25 Aug 2012 11:11:46 -0400 From: fischl@nmr.mgh.harvard.edu To: jeff_rthompson@hotmail.com CC: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] creating a gca file from .mgz files
Hi Jeff
it's more complicated than that. You need a .mgz volume for every
subject
you want to train on that contains the labels, and an accompanying intensity volume (we usually use the norm.mgz for this purpose). Then
we
extract statistics to create the .gca across these subjects using mri_ca_train
cheers Bruce
On Thu, 23 Aug 2012, Jeff Thompson wrote:
From the gcaFormat freesurfer wiki page I know that: The .gca file
is
composed of three volumes, which can be extracted like this:
mri_convert atlas.gca -nth 0 means.mgz mri_convert atlas.gca -nth 1 labels.mgz mri_convert atlas.gca -nth 2 priors.mgz
where:
means --> the mean intensity of the most likely label
labels --> indices of the most likely label at each voxel (indices
refer
to '#No.' column of $FREESURFER_HOME/FreeSurferColorLUT.txt)
priors --> the probability that that label occurs at that voxel
Is it possible to go the other way? I want to create the labels.mgz
file
and
the others manually, and turn that into the .gca file with
mri_convert.
Can this be done?
Thank you,
Jeff Thompson
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
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.