I am examining the connectivity matrix outputs from QSIPrep reconstruction, and am wondering if there is more information on what the differences are between the metrics output per atlas (“sift_invnodevol_radius2_count_connectivity”, “radius2_meanlength_connectivity”, “radius2_count_connectivity”, “sift_radius2_count_connectivity”). In particular, is one of these recommended over another? I assume measures including “sift” should be used over those without as MRTrix3 recommends performing SIFT prior to connectome construction? Thanks so much for additional insight!
This will ultimately depend on what your questions are.
invodevol scales the connectivity by the inverse of the node volumes, accounting for the innate bias that larger nodes have more streamlines emanating from them.
meanlength scales connectivity by the mean length of the streamlines connecting them, accounting for biases that DWI tends to make more long-range connections.
sift means connectivity edges are modulated by the coefficients from MrTrix SIFT2, which makes it so the weighted streamline densities match what it is expected based on the fiber density values.
I cannot say what is best for you, but recommend you take a look at these two papers (at least) to get a better sense of biases in tractography and what kind of questions you can answer with it.