Tck2connectome: Missing nodes 36, 43

Dear experts,

during the tck2connectome step of our preprocessing timeline, we faced an issue we cannot explain.

Blockquote
tck2connectome: [WARNING] The following nodes are missing from the parcellation image:
tck2connectome: [WARNING] 36, 43
tck2connectome: [WARNING] (This may indicate poor parcellation image preparation, use of incorrect or incomplete LUT file(s) in labelconvert, or very poor registration)

According to the fs_default.txt LookUp table, these nodes belong to the left & right thalamus. Our pipeline follows the tutorial of Andy’s brain book and our data is single shell if that makes a difference (see: Error in mtnormalise with problems with gmfod in dwi2fod).
Interestingly, nodes without streamlines assigned are differing between subjects.
Before this warning and afterwards, the pipeline runs without any problems.

We are a bit out of ideas what could’ve caused this, the visual QC looked fine.
The issue became apparent when it caused NaNs in the statistical analysis of brain graph metrics, done using GraphVar. All of our networks seem to be fragmented.
Any help or ideas are greatly appreciated! I can provide more details if necessary.

Thanks a lot,
Joshua

1 Like

Hi,

Sorry, I’m a bit confused. Are you saying that for the different subjects, the unassigned nodes are different? If that’s the case, I suspect the problem is not with the parcellation, but with the tractography itself.

Best regards,

Manuel

1 Like

Hi Manuel,

sorry for being imprecise.
The issue we found, is that for ALL subjects, the warning above pops up.
So nodes 36 & 43 are missing in the parcellation image for ALL subjects.

A thing I noticed while checking this (and probably caused the confusion): In the step afterwards: Nodes without streamlines change from subject to subject. I just wanted to add it to not withhold information that might be useful.

Best wishes
Joshua

Hi,

I think this is related with the command labelconvert and some disagreements between the lit tables used. Thalamus vs thalamus-proper. Check the FS labels and then the output of labelconvert, if the thalamus is still there.

Best regards,

Manuel

Hi Manuel,

thank you so much, this appears to have solved the issue.
If anyone is looking for details: The recon-all to create the input parcellation was done by other people with an older FreeSurfer version. Hence, the input LUT had to be of this older version, too. With a bit of digging I was able to find it on their website. I tested it on some images, and no nodes are missed in the parcellation anymore.

Best,
Joshua