how to interpret zero values in tck2connectome outputs

Hi everyone,

I am interested to know how to interpret zero values in the connectomes.

I used the tutorial from Introduction to MRtrix — Andy's Brain Book 1.0 documentation

I have 29 patients and 29 controls – analyzed separately as 58 subjects, not as two groups, but that is a different story – and when I look at connections between subcortical regions and frontal+temporal regions I find that 47 out of 58 connectomes have zero tracts for at least one of the streamline/connection of interest.

Any ideas if this is normal ? Should the output of tck2connectome always be positive ?

This is the exact command used:

tckgen -act 5tt_coreg.mif -backtrack -seed_gmwmi gmwmSeed_coreg.mif -nthreads 10 -maxlength 250 -cutoff 0.06 -select 10000000 wmfod_norm.mif tracks_10M.tck


Welcome Adrian!

Here’s one way of thinking about it:

  • Generate one streamline for the participant. Almost all connections will have zero streamlines.

  • Generate an infinite number of streamlines for the participant. Based on what we know about the default iFOD2 algorithm and its behaviour under default parameters, almost certainly no connection will not have any streamlines.

So the question is where along that spectrum your data resides. For “typical” reconstructions (~10M streamlines, ~100 parcels), connectome “densities” (fraction of all possible edges that are non-zero) of 70-90% are I think typical. So having most subjects containing at least one empty edge within that sub-network is not at all surprising to me.

But that applies specifically to iFOD2; for other probabilistic algorithms that number is likely lower, and for deterministic algorithms there will be many edges that will never have a streamline assigned no matter how many you produce.

Whether or not the numbers of axons connecting those sub-cortical nuclei to the segments of cortex represented by your parcellation are all non-zero is a different question. But regardless of the answer to that question, it’s still possible for one estimate within such to be zero given the noise inherent in the imaging & reconstruction process; there’s certainly no explicit mechanism in place within the reconstruction ensuring that all edges are non-zero.

Personally, though, I find the distinction relatively uninformative. For sufficiently large total streamline counts, the distinction between an edge having 0 streamlines vs. 1 streamline should be pretty much inconsequential for any stable analysis.