CAT12 - cerebellum extraction


I have recently been interested in nirs and cerebellum and wanted to try our optimal montage on the cerebellum :slight_smile:

I just saw that cerebellum extraction was deactivated by default when using cat12? Is there any reason for that? I was wondering if you were having any suggestions to extract the cerebellar surface similarly to the cortical surface?


1 Like

I added initially the option to compute the cerebellum surface when looking at the CAT12 options.
Then decided to hide it because @CGaser explained it was an experimental feature, too unstable to be used in production at that moment. But this was a while ago, maybe there were improvements on this front.

@CGaser: Is it safer now to use this option?
matlabbatch{1} = 2; % 2: lh+rh+cerebellum

If you want to try it, select the process "Import anatomy > Segment MRI with CAT12", generate the Matlab script, and change the option cerebellum to 1. You should get something like this:

% Process: Segment MRI with CAT12
sFiles = bst_process('CallProcess', 'process_segment_cat12', sFiles, [], ...
    'subjectname', SubjectNames{1}, ...
    'nvertices',   15000, ...
    'tpmnii',      {'', 'Nifti1'}, ...
    'sphreg',      1, ...
    'vol',         1, ...
    'extramaps',   0, ...
    'cerebellum',  1);

Please let us know how this goes, and post screen captures of the result.

Hello Francois,

I just tried on an MRI; the result isn't really convincing :smiley: I tried with 'CAT12.8-Beta'. I am going to update it and try again:
Capture d’écran 2021-11-17 à 14.29.47 Capture d’écran 2021-11-17 à 14.28.57

Edit: in case that's helpful, here is the report provided by CAT12 (I modified the code so it doesn't delete the Cat12 output at the end)

The extraction of the cerebellum is still experimental and should currently not be used. However, our results never looked so poor like yours....



1 Like

Interestingly, if we overlay the cerebellum with the MRI it seems that it's capturing well the bottom of the cerebellum but the algorithm is having a hard time closing the surface :
I can share the MRI file with you if you want.

Here is the surface produced by Freesurfer:


Here is the surface produced by Freesurfer:

Note that this very smooth cerebellum surface is not produced by FreeSurfer, it is meshed from the ASEG volume parcellation by Brainstorm using Matlab function isosurface.m, called in function in_tess_mrimask.m.

1 Like