I am trying to run tractography to find connections from a specific seed region and targets, and I have been using the MRtrix3 Connectome BIDS app for running my analysis. This involves using the seed_dynamic option to generate tracts throughout the whole brain.
How do I ensure that, using this option, it seeds from all voxels within a specific region and that it effectively captures those streamlines?
How would using the seed_image option, and specifying the mask to be used, change the produced results?
Thanks so much.
-seed_dynamic, there isn’t really a way to guarantee that it will seed from all voxels within a specified region. However if in the process of trying to reconstruct the whole-brain tractogram, the algorithm observes that there are an inadequate number of streamlines reconstructed within specific voxels, it will place streamline seeds in such a region with greatly increased frequency in an attempt to correct the discrepancy.
-seed_dynamic fails to generate streamline seeds (that go on to produce acceptable streamlines) within a particular voxel, there’s a good chance that manually placing streamline seeds in such a voxel using some other seeding mechanism will similarly fail to reconstruct acceptable streamlines from such. But as I say, there’s no mathematical guarantee of such.
Certainly if you were to use
-seed_image to perform whole-brain tractography, I would expect there to be less seeds placed in such problematic voxels that there would be using
Thanks a lot for the explanation, it’s very helpful!
I have noted that -seed_dynamic can not be used with other seeding mechanisms. Can I use “-seed_dynamic some_WM_FODs.mif” to tckgen streamlines between one cerebral hemisphere and one cerebellar hemisphere without using “-include the cerebral hemisphere” and “-include the cerebellar hemisphere”? To use WM_FODs.mif of the cerebral hemisphere and cerebellar hemisphere?
Thank you@rsmith very much!
It sounds like you’re trying to use dynamic seeding, but in a targeted tracking experiment. Personally I don’t see this working out. Dynamic seeding only makes sense if the fibre density of all fixels in the segmented FOD image reflects the density of streamlines you expect to see. So there’s an implicit assumption that all FODs within the input image should get reconstructed by streamlines, and that all fibre density within those fixels should be attributed to the set of streamlines reconstructed. With targeted tracking these are broken: the whole point of targeted tracking is to reconstruct only a subset of all possible connections, so that mapping between fibre density and streamlines density can’t be reasonably enforced. As soon as you have a fixel where some of the fibre density should be attributed to the reconstruction and some of it should be attributed to something outside of the reconstruction, you can no longer use the difference between streamlines and fibre density in the same way.
I get it. Thank you very much!