"Compute Brain Mask" check

[joost]

Hi,

using BV 3.2, "T1 MRI" -> "Components" -> "Compute Brain Mask" with default input parameters.

my input images are skull stripped, bias corrected, talairach-ed, resolution (1x1x1) T1's. _attached_ are two snapshots of the "Compute Brain Mask" output.

my question: does this look ok? i can send more information but i didn't want to clog the forum's mailbox.

thanks,
-joost

[Jean-Francois Mangin]

Looks perfectly OK. Be sure that the histogram analysis is OK
before triggering further processing. You can mimic the analysis
file if it is easier for you. The only thing required is average
and standard deviation of grey and white matter. Mimic any
existing file you find, the rest of the stuff inside is not used further.

[joost]

Hi,

thanks! a further question: my goal is to create high quality .arg files for later use in automatic recognition. sometimes the .arg file looks bad, e.g. i see 'knots' in the sulci skeleton when i visualize them through the mesh.

my question: what are _essential_ things i can modify to improve the creation of the .arg files (before running the automatic recognition)? i ask this because you guys probably processed tons of images and may know typical things that go awry during processing. i realize that at some point poor image quality (not solvable) also plays a part...

thanks,
-joost

[Jean-Francois Mangin]

It is not clear to me what problem you are talking about.
If you have skull striped images, and robust grey/white matter segmentation
coming from another software, I suggest to inject the brain mask
at the right place (or even the hemisphere/cerebellum split).
The grey/white matter mask however should be used
more carefully. Put it at the right place to get access to
things like cortical thickness averaged on each sulcus (the right
place is brainVISA grey/white classification). In return
do not interact with the process generating the folds.
It means the images called cortex_something have to be computed
by BrainVISA because they include the spherical topology constraints.
Use your classification to mimic the histogram analysis file.