What is the weight of the edge reported by qsiprep? and Any recommand metrics to be use?

1

Hi,
I found the qsiprep will give various connetome after connectome construction, including radius2_count_connectivity, meanlength_connectivity, sift_invnodevol_radius2_count_connectivity,radius2_count_connectivity, etc. What is their meaning? How were they calculated? and which one is the more recommand?

here is a example output of msmtconnectome. Seem no any intro on qsiprep’s wiki, most part is about atlas, custome pipeline.

aal116_radius2_count_connectivity 116x116 107648 double
aal116_radius2_meanlength_connectivity 116x116 107648 double
aal116_region_ids 1x116 928 int64
aal116_region_labels 116x20 4640 char
aal116_sift_invnodevol_radius2_count_connectivity 116x116 107648 double
aal116_sift_radius2_count_connectivity 116x116 107648 double
aicha384_radius2_count_connectivity 384x384 1179648 double
aicha384_radius2_meanlength_connectivity 384x384 1179648 double
aicha384_region_ids 1x384 3072 int64
aicha384_region_labels 384x24 18432 char
aicha384_sift_invnodevol_radius2_count_connectivity 384x384 1179648 double
aicha384_sift_radius2_count_connectivity 384x384 1179648 double
brainnetome246_radius2_count_connectivity 246x246 484128 double
brainnetome246_radius2_meanlength_connectivity 246x246 484128 double
brainnetome246_region_ids 1x246 1968 int64
brainnetome246_region_labels 246x13 6396 char
brainnetome246_sift_invnodevol_radius2_count_connectivity 246x246 484128 double
brainnetome246_sift_radius2_count_connectivity 246x246 484128 double
gordon333_radius2_count_connectivity 333x333 887112 double
gordon333_radius2_meanlength_connectivity 333x333 887112 double
gordon333_region_ids 1x333 2664 int64
gordon333_region_labels 333x27 17982 char
gordon333_sift_invnodevol_radius2_count_connectivity 333x333 887112 double
gordon333_sift_radius2_count_connectivity 333x333 887112 double
power264_radius2_count_connectivity 264x264 557568 double
power264_radius2_meanlength_connectivity 264x264 557568 double
power264_region_ids 1x264 2112 int64
power264_region_labels 264x33 17424 char
power264_sift_invnodevol_radius2_count_connectivity 264x264 557568 double
power264_sift_radius2_count_connectivity 264x264 557568 double
schaefer100x17_radius2_count_connectivity 100x100 80000 double
schaefer100x17_radius2_meanlength_connectivity 100x100 80000 double
schaefer100x17_region_ids 1x100 800 int64
schaefer100x17_region_labels 100x25 5000 char
schaefer100x17_sift_invnodevol_radius2_count_connectivity 100x100 80000 double
schaefer100x17_sift_radius2_count_connectivity 100x100 80000 double
schaefer100x7_radius2_count_connectivity 100x100 80000 double
schaefer100x7_radius2_meanlength_connectivity 100x100 80000 double
schaefer100x7_region_ids 1x100 800 int64
schaefer100x7_region_labels 100x27 5400 char
schaefer100x7_sift_invnodevol_radius2_count_connectivity 100x100 80000 double
schaefer100x7_sift_radius2_count_connectivity 100x100 80000 double
schaefer200x17_radius2_count_connectivity 200x200 320000 double
schaefer200x17_radius2_meanlength_connectivity 200x200 320000 double
schaefer200x17_region_ids 1x200 1600 int64
schaefer200x17_region_labels 200x25 10000 char
schaefer200x17_sift_invnodevol_radius2_count_connectivity 200x200 320000 double
schaefer200x17_sift_radius2_count_connectivity 200x200 320000 double
schaefer200x7_radius2_count_connectivity 200x200 320000 double
schaefer200x7_radius2_meanlength_connectivity 200x200 320000 double
schaefer200x7_region_ids 1x200 1600 int64
schaefer200x7_region_labels 200x27 10800 char
schaefer200x7_sift_invnodevol_radius2_count_connectivity 200x200 320000 double
schaefer200x7_sift_radius2_count_connectivity 200x200 320000 double
schaefer400x17_radius2_count_connectivity 400x400 1280000 double
schaefer400x17_radius2_meanlength_connectivity 400x400 1280000 double
schaefer400x17_region_ids 1x400 3200 int64
schaefer400x17_region_labels 400x25 20000 char
schaefer400x17_sift_invnodevol_radius2_count_connectivity 400x400 1280000 double
schaefer400x17_sift_radius2_count_connectivity 400x400 1280000 double
schaefer400x7_radius2_count_connectivity 400x400 1280000 double
schaefer400x7_radius2_meanlength_connectivity 400x400 1280000 double
schaefer400x7_region_ids 1x400 3200 int64
schaefer400x7_region_labels 400x27 21600 char
schaefer400x7_sift_invnodevol_radius2_count_connectivity 400x400 1280000 double
schaefer400x7_sift_radius2_count_connectivity 400x400 1280000 double

2

Hi,

Here’s my understanding (no particular order):

  1. You see radius2 in all of these. This means that a 2mm radial search was used to assign streamline endpoints to a gray matter region. So, if no gray matter was within 2mm of a streamline endpoint, it does not get included in the connectome.
  2. count vs meanlength is denoting the metric in the connectome. That is, it is either the number of streamlines connecting regions, or the mean length of the streamlines connecting regions.
  3. The beginning of the file name tells you what atlas the regions come from
  4. The ids and labels tell you the names of the regions in the atlas
  5. invnodevol is short for “inverse node volume”, which means the metric is factored by the inverse of the average node volume (that is, the beginning and end region volumes). This is one way to control for the intuition that larger regions will innately have more streamlines run through them.
  6. sift means that the connectome weights are modulated by SIFT2 outputs.

Which connectome you choose will be entirely dependent on your analysis. There is not a gold standard.

More info: tck2connectome — MRtrix 3.0 documentation
tcksift2 — MRtrix 3.0 documentation

edited to add that using SIFT is generally a good idea to weigh against streamlines that are less biologically plausible given the calculated FODs.

2 Likes
3

Got it. Thank you very much.