Connectivity matrix size bigger than atlas node numbers

Hi all,
I have recently run a QSIPrep recon pipeline and wanted to use its output atlas based connectivity matrix. As far as I understand, the atlas based connectivity matrices should be of size n of nodes of atlas x n of nodes of atlas, however mine all seem to be bigger, regardless of the atlas. For example Gordon333 connectivity matrix is 394x394, Brainnetome246 is 256x256 and so on. Is this normal? How should I use them? Could somebody help?

Hi @Marie_Santillo,

And welcome to neurostars!

This is probably because QSIPrep appends subcortical and cerebellar regions to atlases that do not have them (a lot of atlases only contain neocortex).


Hi Steven,

Thank you so much for your warm welcome!

That makes sense, I must’ve missed that in the release notes :sweat_smile:!
Would you happen to know if there is a way for me to access the MNI coordinates of the harmonized atlases?

Thank you so much for your help,

Hi @Marie_Santillo,

The atlases are in the docker/singularity container in /atlas/qsirecon_atlases/

You can copy the MNI atlases .nii.gz file to your home directory with the following

singularity shell $qsiprep_IMG
cp /atlas/qsirecon_atlases/* $HOME/

Otherwise, you can warp the T1w atlases in the qsirecon output dir back to MNI using the MNI-to-T1w transform file in the qsiprep anatomical output directory.


1 Like