While I’m not familiar with the most recent updates to TractSeg, my understanding is that it still fundamentally operates on the FOD peaks; the bundle tractograms now output are downstream of that segmentation. So generating streamlines from those peaks, and then mapping back to fixels, feels circular to me: I’d have thought that extracting the intermediate segmentation where it’s natively in the form you want would be far more direct.
(Even if TractSeg operates on SH peaks, those are just discrete orientation estimates that are not quite equivalent to the FOD lobe mean orientations from
fod2fixel; and 4D images containing sets of XYZ 3-vector triplets, which we often refer to as “
peaks”, are just fixel data that aren’t in the fixel directory format;
fixel2peaks convert back and forth)
But if you do want to go the route of using the streamlines output, then yes,
tck2fixel is what you want.