Colm--- it requires a bit of scripting--- I did it a while ago, but
can't find the code.
Basically each atlas has a corresponding XML file ... and each region
within an atlas is basically a separate volume...
Basically I just did an iterative approach... where I read out the
labels in order (may even just do a copy/paste in EXCEL and save your
self some parsing)....
basically in pseudo code..
for($i=0;$i< # of labels; $i++)
{
fslroi <atlasfiletosplit.nii,gz> <name_you_want_to_call_given_volume>
<volume_to_pull> 0
}
So for me:
/usr/local/fsl/data/atlases
contains the following atlases... so for example:
Cerebellum_MNIflirt.xml HarvardOxford-Cortical.xml JHU-labels.xml
Juelich.xml Talairach.xml
Cerebellum_MNIfnirt.xml HarvardOxford-Subcortical.xml JHU-tracts.xml
MNI.xml Thalamus.xml
The harvardoxford-cortical.xml contains many lines that look like :
<label index="23" x="47" y="23" z="40">Intracalcarine Cortex</label>
fslroi HarvardOxford/HarvardOxford-cort-prob-2mm.nii.gz
Intracalcarine_cortex 23 1
will extract the 23rd volume of the input atlas... and call it
intracalcarine_cortex.nii.gz
If this is useful I may write my script and stick all the volumes on
one of my web servers... I already have most of the code written...
One thing to keep in mind is that these are UNTHRESHOLDED prbability
maps... so basically each output_volume is a 3-d volume (in either 2mm
or 1mm MNI space)... where each voxel represents the probability the #
of subjects that shared that label in common space...
Basically beacuse of registration/the fact that not all of our brains
are the same... when you take some sort of ground truth estimation in
native space of the subject, assign a given voxel a label (like
amygdala), and then transform that voxel/image to a standard space,
there is some probability a given voxel will be consistently labeled
across subjects...
Anyway let me know if me sticking this somewhere would be useful...
it's too much data to simply post as an email...
dg
On Mon, Jan 31, 2011 at 6:21 PM, Connolly, Colm <[log in to unmask]> wrote:
> Hi everybody,
>
> Is there an easy to extract each of the regions labeled in (for example) the T&T atlas included in FSL? Is the only way to do this to use FSLview and generate mask datasets by hand?
>
> I ask as I'd like to generate a series of masks (using as automated a procedure as possible) so that the individual ROIs in the atlas could be saved to individual files so that they could be used as masks by me and others in our lab.
>
> Regards,
> --
> Colm G. Connolly, Ph. D.
> Dept of Psychiatry
> University of California, San Diego
> 9500 Gilman Dr. #0738
> La Jolla, CA 92093-0738
> Tel: +1-858-246-0684
>
--
David A Gutman, M.D. Ph.D.
Center for Comprehensive Informatics
Emory University School of Medicine
|