I’m working on HCP data and I have a few methodological questions. I want to find connectivity of some Freesurfer ROIs (I’m using a2009s atlas). For Tckgen I’m not able to run global tractography for my subjects so I’m using nodes_fixSGM.mif file (output of labelsgmfix command) as -seed_image for my tractography (tckgen). First of all, I have no idea about the -select number that I need to use for tckgen and tcksift after that? How can I find a good estimation of those numbers?
Should I edit fs_a2009s.txt file based on my ROIs for labelconvert/labelsgmfix?
and the last question, the output matrix of tck2connectome command is 153*153 but fs_a2009s contain 165 ROI. Is something wrong with my output?
So you’re seeding exclusively from subcortical GM? That’s not necessarily a problem, but it’s not what we’d typically recommend. I guess it’ll depend on what you’re trying to achieve.
Nonetheless:
This is always a tricky question to answer, and will be somewhat specific to your data and experiment. In general, the aim is to ensure you have sufficient numbers that your results are reproducible from one run to the next. On other words, that the variability due to the stochastic nature of the algorithm is small relative to the effect you’re trying to characterise. We recommend you experiment with your data to figure out a number that yields a stable outcome. You can always add more, but then computation times become longer and longer for diminishing returns.
I’m not sure about that one, and why you’d need to do that. Maybe others have an idea about that…(?)
That step doesn’t depend on the fs_a2009s file. What matters is the maximum node index in the parcellation image you provided to the command (i.e. the maximum value in the image). This determines the matrix size that tck2connectome will produce. I’m not sure exactly how you obtained your parcellation, but if you were genuinely expecting it to contain all of the 165 labels in your parcellation, then this may indicate that something has gone wrong somewhere – it looks like there are no voxels corresponding to the last 12 labels…
so I’m using nodes_fixSGM.mif file (output of labelsgmfix command) as -seed_image for my tractography (tckgen).
So you’re seeding exclusively from subcortical GM?
No; but also yes.
That’s the full parcellation image, which has been fed through labelsgmfix, being provided as the seed image. tckgen will interpret this as a binary mask when provided to -seed_image. So it’s basically interpreted as a GM binary mask. The trouble with that if you are using ACT is that because FreeSurfer only labels cortical voxels in cases where the GM partial volume fraction is large (don’t know the exact logic used there), basically all streamline seeds drawn from within cortical voxels will be rejected as plausible streamline seeds, because they will have a GM > 0.5 partial volume fraction and therefore can’t be used as a seed, because -seed_gmwmi was not used.
Should I edit fs_a2009s.txt file based on my ROIs for labelconvert/labelsgmfix?
and the last question, the output matrix of tck2connectome command is 153*153 but fs_a2009s contain 165 ROI. Is something wrong with my output?
The only reason why you would potentially need to use an edited version of that file is if you want your connectome construction to deviate from the provided file in some way; e.g. changing the order of nodes, or excluding nodes, or the like.
If the output of tck2connectome gives a different number of rows / columns than what is expected, then that is an issue with your data or processing that needs to be diagnosed; it’s not simply a matter of editing the lookup table file, as there’s no understanding of what modification would even need to be done. I can only suggest manually inspecting the outputs of your various processing steps and trying to diagnose the first point at which the data deviates from what you expected.
