Combined WB & targeted tracking with waypoints

1

Hi MRtrix team,

Thank you for developing this great software package & for providing such an excellent discussion forum.

I would like to check that my tractography approach is valid, as I haven’t come across any examples of cases exactly like mine in the literature or on the forum.

My aim is to calculate normalised connection density within a specific pathway defined between two spherical ROIs, but with additional inclusion and exclusion waypoints applied. These are the steps I have taken:

  1. Whole-brain tracking:

    tckgen -act {5tt.mif} -backtrack -seed_gmwmi -select 10000000 wb.tck

  2. Targeted tracking:

    tckgen -seed_image {ROI1} -include {ROI2} -include {Waypoint_ROI1} -exclude {Waypoint_ROI2} -select 500 -act {5tt.mif} -seeds 0 {wmfod.mif} targeted.tck

  3. Combine tracking:

    tckedit {wb.tck} {targeted.tck} {combined.tck}

  4. Run SIFT2 to get streamline weights

  5. Re-extract targeted pathway with 500 streamlines:

    5a. Firstly, get pathway with streamlines ending in spherical ROIs:

    tckedit {combined.tck} -include {ROI1} -include {ROI2} -ends_only -tck_weights_in {SIFT2_weights.txt} -tck_weights_out weights_out_endsonly.txt targeted_endsonly.tck

    5b. Then, re-apply waypoint criteria & select 500 streamlines:

    tckedit {targeted_endsonly.tck} -include {Waypoint_ROI1} -exclude (Waypoint_ROI2} -number 500 -tck_weights_in {weights_out_endsonly.txt} -tck_weights_out weights_waypoints_500streamlines.txt

(The reason I split step 5 into two parts is because I wanted to include the -ends_only criterion only for ROI1 and ROI2 which define the ends of the pathway).

(Also, I firstly tried the more ‘standard’ method of performing the wholebrain tracking followed by tckedit –include, however my pathway is not large enough to extract enough streamlines with this approach).

A couple of specific questions…

  • I have seen that this combined approach has been used in a handful of previous papers (e.g. most recently here) (in addition to a few posts on this forum, e.g. here), however I haven’t seen any examples where waypoints are also incorporated into this – is my assumption correct that it is valid to do this?

  • Regarding the number of streamlines generated during the targeted tracking step, my understanding is that for this kind of experiment it is important to generate the same number of streamlines (e.g. in this example, 500) for each subject, is that correct? I previously ran this step without specifying the –select option, with some subjects having very few (<10) streamlines and other subjects having hundreds. Ultimately, across my subjects, there was a strong correlation between normalised connection density (ssw x proportionality coefficient) and number of streamlines generated – this makes sense, however my understanding is that the number of streamlines should not be a factor when using SIFT2 instead of SIFT. Any clarification / confirmation of this would be really helpful!

Thanks in advance,

Emily

2 Likes