Hello,
Could you please clarify the default endpoint assignment settings for tck2connectome? I was under the assumption, that the default is a simple voxel lookup (-assignment_end_voxels), however, I was not able to find any explicit documentation of the default settings.
Thank you for your help,
Franziska
Hi Franziska,
While not explicitly stated in the documentation (I’ll add that), the information is kind of there in the References section, albeit poorly phrased:
If using the default streamline-parcel assignment mechanism (or -assignment_radial_search option):
That was presumably supposed to mean “applicable when that mechanism is used, whether by default or because of manually selecting in order to override the maximal distance parameter”, but as currently phrased erroneously implies that the default may be different to -assignment_radial_search.
Regards
Rob
Hi Rob,
Great, thank you for the clarification.
If I understand correctly, the choice between -assignment_radial_search and -assignment_end_voxels affects only fibers terminating outside of an overlaid parcellation atlas, such as the AAL-atlas? Fibers ending within an AAL-parcel are assigned to that specific parcel regardless of which option is used, correct?
Since I am using probabilistic tractography with dynamic seeding and ACT, fibers are restricted to terminate in grey matter. This would mean that fibers terminating in areas outside of the AAL-atlas would be caused by registration inaccuracies from the segmented T1-image (for ACT) to DWI-space?
Thank you for your help,
Franziska
If I understand correctly, the choice between
-assignment_radial_searchand-assignment_end_voxelsaffects only fibers terminating outside of an overlaid parcellation atlas
If the voxel in which a streamline terminates (beware using “fibre” to refer to streamlines) contains a label, then that is the first voxel that will be checked by the radial search as it is the closest, and so the ascribed label will indeed be identical to that of -assignment_end_voxels. While in the case of AAL this may well be the case more often than not, and therefore one might question the added complexity, I needed a software default that would behave somewhat sensibly regardless of what input data are provided, and not all parcellations are so spatially generous.
Since I am using probabilistic tractography with dynamic seeding and ACT, fibers are restricted to terminate in grey matter. This would mean that fibers terminating in areas outside of the AAL-atlas would be caused by registration inaccuracies from the segmented T1-image (for ACT) to DWI-space?
Some streamlines terminate due to leaving the image FoV down the spinal column, and there may not be any parcel there.
Depending on how the 5TT image was derived, some voxels may be ascribed to the fifth volume (erroneously termed “pathology”), which permits streamlines to terminate if they choose to for reasons other than the tissue segmentation.
Streamlines can terminate wherever is designated as grey matter in the 5TT image. This is not equivalent to the biological reality of what has a biological constitution aptly described as “grey matter”. Misregistration is not the only potential inadequacy that could lead to such. It could be erroneous tissue segmentation (eg. partial volume between WM and CSF can have equivalent intensity in a T1-weighted image to GM), or it could be some GM nuclei that is visible in the anatomical image and comes out of tissue segmentation but doesn’t have a corresponding node in the parcellation.
Regards
Rob
Hi Rob,
This is very helpful. Thank you again for the clarification.
Best,
Franziska