Using SIFT2 for cross-species comparison of connectomes

Hi MRtrix Community,

I am recently very interested in cross-species connectomics (from macaque to human).

Based on my understanding, MRtrix uses the SIFT2 algorithm to assign a weight to each streamline. Under certain conditions (possibly AFD normalization?), a connectivity matrix generated with tck2connectome (using the -tck_weights_in option), multiplied by each subject’s “subject-specific proportionality coefficient \mu” (obtained from the file generated by tcksift2 with the -out_mu option), results in a new matrix. In this new matrix, the connectivity value between brain region A and brain region B represents the “actual axonal cross-sectional area” of the white matter bundles connecting A and B. This value should have a biophysical interpretation, with units in mm^2. Could this cross-sectional area perhaps be understood as a kind of “bandwidth,” measuring the capacity for information transfer?

  • First, I would like to confirm whether my understanding is correct.

  • Second, if my understanding is correct, I am curious whether we can directly use this value to compare the so-called “connection strength” (or, the white matter bundle’s cross-sectional area? or, its information bandwidth?) between homologous white matter bundles in humans and macaques. While I can imagine that the cross-sectional area of human white matter bundles would typically be larger than that of macaques, it might be more intriguing to investigate if we account for brain size as a covariate. Could this reveal more interesting findings?

I would greatly appreciate it if someone could shed light on this.