Cerebellum mesh

[chrysa]

hi to all,

I have run an experiment and I very much expect some activation in the cerebellum. I tried the Brain Visa tool 'Ana Get whole brain volume' and I merged the brain_mesh with the Lhemi_mesh and Rhemi_mesh in Brainstorm to proceed with an accurate head model (I tried with only the brain_mesh and the alignment with MRI is much worse).
However, it is written in that BrainVisa tool that the spatial resolution for the cerebellum may be bad in the brain_mesh.
Is there any other way to define the cerebellum in a mesh more accurately?

thanks in advance,
Chrysa

[riviere]

Hi,
When we say that the resolution in cerebellum is not good enough, it is mainly that we are working on sulci, and sulci in the cerebellum are too thin to be correctly extracted on 1mm-resolution images. But having a mesh of the cerebellum global shape is completely possible. However as we are not really working on the cerebellum, such a mesh is not produced by the standard morphologist pipeline. Getting a segmentation of the cerebellum + brainstem is easy. Separating cerebellum hemispheres and brainstem will perhaps require to play a little bit with low-level commands.
Denis

[chrysa]

Hi Denis,

thanks for your reply. Could you 'redirect' me to a tutorial with low-level commands?

Thanks,
Chrysa

[riviere]

Hi Crysa,

Well, I could redirect you if such a tutorial existed...
There is not much doc on the use of low-level tools. We have several sets of commands to manipulate brain images: Aims* commands, Vip* commands etc. All of them have a help when you use them with the -h option. But there are many many of them... There is a list of commands with the help here which allows searching, not not much more...
You can try the VipSplitBrain command, disabling all template/position information options, to avoid assumptions on where and what is split, but I'm not sure it will work as it has been specialized for brain hemispheres.
Otherwise the idea is to roughly segment/threshold the white matter, erode it progressively until it breaks into a sufficient number of pieces (3 mainly, for 2 hemispheres and brainstem), then use these disconnected pieces as seeds in a voronoi growing into the initial segmentation. You can try using the commands AimsThreshold, AimsErosion, AimsVoronoi (or Vip alternatives), and then AimsMeshBrain. But not sure it will work so easily...

Denis