Creating my own template for split brain

[Olivier Coulon]

Hello,
when processing brains that are not conventional in size or/and shape (e.g. infants, or non-human primates), one common fail is the split brain procedure. I suspect this is because the template closedvoronoi.ima does not match enough the shape of the brains we are trying to process... So I am wondering of it is a good idea to create my own model and how. For instance we are processing a set of baboon images, and we made an average template using all our subjects. This template was then processed in the morphologist pipeline and the split brain was successfull. So I have now a image of the brain mask, splitted in three different regions. Can I use that as a model or do I have to process it a little before that ?

Thank you in advance,

Olivier

[riviere]

Hi Olivier,
Interesting question...
The split template allows to help detecting correct seeds for the split, but there are other hard-coded constraints in the process: to allow splitting the corpus callosum, some points are removed from it using fixed coordinates in the Talairach space. For non-human brains, this step will probably not work optimally either.
We plan to include these points in the template to allow changing them with the species.
So yes, using a custom template may help, but will maybe not be enough right now.
The group of Thierry Delzescaux (thesis of Yaël Balbastre) had started to work on this aspect (they are working on macaques), Denis so it would be a good idea to synchronize this with them also.
Cheers
Denis

[Olivier Coulon]

Thank you for your answer. I understand why it is more complicated than I thought. For our data we specify AC-PC so the corpus callosum points in Talairach space should not be too wrong, no ?

So, with respect to my orgininal question, and let apart the problems you mention, if I want to try to build my template from my own splitted brain, what do I have to do to it. Closing or dilating it a little ?

Olivier

[riviere]

Hi Olivier,
the corpus callosum points will not be "too wrong" for species "not too different" from humans - you can try anyway, and it actually works more or less for primates.
For the template, you will probably have to close it and ensure it has the expected 3 seeds. I don't think it needs dilation, the current template is very basic and has been made from just one subject, it doesn't need much precision, it's just to ensure the split seeds are in the correct brain parts. The template has to be resampled in Talairach space, and probably for now have the same number of voxels as the human template (I'm not sure of that but I suspect the transformation from this template voxels to the subject space is more or less hard-coded in the program - to be checked).

Denis