group comparison after sift2, tck2connectome

Hello MRtrix3 community,

Apologies for rudimentary questions from someone new to diffusion analysis!

After using tckgen to perform tractography and then SIFT2 and tck2connectome, the output looks to be an roi x roi matrix populated by the sum of the streamline weights connecting each pair of rois.

  1. Say I was interested in comparing thalamus-prefrontal connectivity in alcoholics vs controls. Is this sum-of-track-weights metric immediately usable for group comparisons? Is it the best indicator available for overall structural connectivity (i.e. capacity of information flow)?

  2. I understand that I could use tcksample and tck2connectome to generate a matrix populated by some other scalar metric like (e.g. mean FA) for each roi x roi pair. I hope this isn’t overly broad, but what scalar metric(s) are currently considered the most biologically relevant, meaningful measures of connection strength?

Basically, if you were reviewing a cross sectional study investigating group differences in connection strength for a pair of ROIs, what metric(s) would you want/expect to see used? Is there a gold standard here?

Thanks!
Donovan

Hi Donovan,

Is this sum-of-track-weights metric immediately usable for group comparisons?

@jdtournier: :pleading_face: :clown_face:

The preferable way of making this metric comparable across subjects is demonstrated in this letter. Unfortunately the manuscript properly describing and justifying this has been delayed by a half decade or so, and hypothesized on this forum about 427 times :man_facepalming: But I will hopefully post a preprint in the coming month or two: I need to get everybody else’s manuscripts off my desk (plus nag Donald some more) before I can finish this off.

Is it the best indicator available for overall structural connectivity (i.e. capacity of information flow)?

From a theoretical physical perspective, assuming one is interested in endpoint-to-endpoint connectivity, I’d personally say yes. This will again be explained in my manuscript. The pragmatics of data variance is however an outstanding question, since this quantification relies moreso than some other approaches on the outcomes of tractography, which is known to be pretty junk.

… what scalar metric(s) are currently considered the most biologically relevant, meaningful measures of connection strength?

I’m not sure that I can really advocate for any specific metrics here; you’ll find the full gamut having been proposed / utilized somewhere in the literature. Plus I’m already in a bit of controversy in this context, so don’t want to say anything that could be used against me in a court of law. :upside_down_face:

Basically, if you were reviewing a cross sectional study investigating group differences in connection strength for a pair of ROIs, what metric(s) would you want/expect to see used? Is there a gold standard here?

If you were to replace “pair of ROIs” with “a white matter pathway of interest”, you could consider quantifying the FBA metrics, but rather than performing statistical inference at a per-fixel granularity, calculating the mean within a fixel mask corresponding to a pathway of interest. You would no longer be “extracting connectome properties”, but this approach has been used in multiple studies, and it does mean that you get fixel specificity.

Cheers
Rob