Using the MRtrix toolbox, would it be possible to restrict fiber tracking to ONE hemisphere only?
For example, if I would like to perform tracking and connectome generation only in the left unaffected hemisphere due to the presence of a large structural pathology in the right hemisphere.
I am aware that I could first generate a whole-brain structural connectome and then “delete” all values of connections that I am not interested in (i.e. hemisphere with structural pathology and inter-hemispheric connections). However I am not sure if that would generate similiar results to restricting connectome generation to one hemisphere only.
Yes there is, but I’m not sure what you’re suggesting is the right way to address your research question…
First off: to restrict tracking to one hemisphere, you can supply a brain mask (using the -mask option to tckgen) encompassing the hemisphere of interest only. This should simply terminate tracks as they leave the mask. I’m not sure how ACT would behave in this case - it may be that you need to edit the 5TT image in this case to ensure the pathological hemisphere is labelled as non-brain. No doubt @rsmith will clarify in due course.
However, you might find you’re better off just leaving the tracking to do its thing, apply SIFT/SIFT2 (optionally restricting processing to a mask of just the hemisphere of interest), and then generating your connectome from that as usual, but subsequently discarding all rows / columns corresponding to the pathological hemisphere. Note that this immediately discards all interhemispheric connections, i.e. corpus callosum & anterior commissure…
Better still, you could use the 5th tissue type in the 5TT image to mark out the lesion (this is what it’s designed for, after all ). This turns off all ACT priors in these regions, relying purely on the FOD information. As an extreme version, you could mark out the whole hemisphere as pathological… Something like this might be a better way to address your research question…?
This should simply terminate tracks as they leave the mask. I’m not sure how ACT would behave in this case …
Exactly the same as non-ACT: If a streamline leaves the tracking mask, this is treated as an acceptable termination; similar to when travelling down the spinal column.
However, you might find you’re better off just leaving the tracking to do its thing, apply SIFT/SIFT2 (optionally restricting processing to a mask of just the hemisphere of interest), and then generating your connectome from that as usual, but subsequently discarding all rows / columns corresponding to the pathological hemisphere.
If you’re wanting to use SIFT / SIFT2, it’s essential that the tracking experiment be done in this way. To demonstrate: Imagine that you have a healthy brain, have done whole-brain tracking, have run SIFT/SIFT2, and have a perfect match between fixel fibre densities and streamlines densities. Now you remove a subset of streamlines from the tractogram; e.g. inter-hemispheric connections. Now, you run SIFT / SIFT2 again; it’s clear at this point that you no longer have a perfect model fit (since you’ve removed streamlines from the perfect fit!), so SIFT / SIFT2 is going to perturb the tractogram in a manner that tries to “boost” the streamlines density in those fixels that were traversed by those streamlines you removed. So by removing streamlines before SIFT / SIFT2, your reconstruction is being pushed toward a non-optimal solution, because in some fixels, some of the fibres underlying that fixel explicitly do not exist within your reconstruction.
Better still, you could use the 5th tissue type in the 5TT image to mark out the lesion …
Depends on the rationale for excluding the hemisphere. If 5TT image generation works as expected, I’d suggest just letting whole-brain tracking run its course. If the pathological nature of the disorder breaks the translation from T1 intensities to tissues, but is only localised, you could try just adding the pathological areas to the 5th 5TT volume. Highlighting the entire hemisphere as pathological, then ignoring any connections involving that hemisphere, wouldn’t really provide anything over just using a healthy-hemisphere brain mask.
). This turns off all ACT priors in these regions, relying purely on the FOD information. As an extreme version, you could mark out the whole hemisphere as pathological… Something like this might be a better way to address your research question…?