[caret-users] Hand Cleaning Cerebellar Extractions

Donna Dierker donna at brainvis.wustl.edu
Fri Feb 26 08:59:26 CST 2010

Hi Jessica,

See inline replies below.


On 02/25/2010 03:50 PM, Jessica Bernard wrote:
> Hello,
> I have been working with images of the cerebellum (and just stared 
> working with CARET) that have been extracted from the whole brain, and 
> I need to hand clean them to remove some of the occipital cortex to 
> complete my analyses. I've searched through the archives and website 
> to find the best way to do this, but I have not had much luck. Would I 
> do this by hand drawing borders? 
No, your problem must be addressed in the segmentation volume.  It is 
possible that creating a surface of the existing segmentation might help 
you visualize and pinpoint "bridge" and "neck" structures that are 
holding non-cerebellar tissue onto your cerebellum.  Clicking on those 
lets you focus on the smallest region of voxels to erase, so that then 
you can do disconnect isllands, e.g.:

File: Open Data File: Volume Anatomy file - load your T1 anatomical
File: Open Data File: Volume Segmentation file - load your cerebellum
Toolbar: D/C: Overlay/Underlay Volume: Anatomy underlay, Segmentation 
Volume: Segmentation: Reconstruct into Surface
    select hypersmooth surface
Switch to inflated surface -- look for cinch-like structures, where 
there is a narrow bridge connecting occipital cortex to cerebellum; 
click on the surface as close to the cinch as you can get
Switch to Volume and select view All
Volume: Segmentation: Edit voxels
When done, File: Save data file: Volume Segmentation File -- save as 
different filename
Volume: Segmentation: Disconnect Islands
After making sure results are sane, File: Save data file: Volume 
Segmentation File (overwrite)

I must say, I do not envy your task.  I don't know anyone trying to do 
that.  A masochistic streak might come in handy. ;-)
> Could anyone recommend a method or a useful tutorial? 
Nothing specifically cerebellum oriented.  General patching tips are here:

> Thank you very much, and I apologize for what might be a completely 
> obvious question.
There is nothing obvious about how to segment the cerebellum accurately.

It depends on how you are using your cerebellar segmentations, but to 
get one that yielded a surface of the quality that David obtained with 
the Colin cerebellum, you need cubic 0.5 voxels scanned 20 or so times, 
coregistered, and averaged.  Most anatomical MRI is not of that 
quality.  But I am guessing your analyses do not need surface 
representations of the cerebellum (i.e., the segmentations need not be 
that fine-grained).  My reading of your needs is that you just want to 
shear off occipital cortex that is hanging onto your cerebellum.
> Best Wishes,
> Jessica
> -- 
> Jessica Bernard
> PhD Candidate
> Department of Psychology
> Cognition & Cognitive Neuroscience
> University of Michigan
> 4056 East Hall, 530 Church St.
> Ann Arbor, MI 48109
> Email: jessbern at umich.edu <mailto:jessbern at umich.edu>
> ------------------------------------------------------------------------
> _______________________________________________
> caret-users mailing list
> caret-users at brainvis.wustl.edu
> http://brainvis.wustl.edu/mailman/listinfo/caret-users

More information about the caret-users mailing list