For my research internship I would like to extract FA values of the corpus callosum. I have came to the point where whole brain tractography is performed, and now I would like to perform tractography for my ROI (corpus callosum) only. As I am a beginner, I could use some advice on how to approach this. I tried some things with manual drawing of ROIs and applying those in the tckedit -include and -exclude options and that works fine. However, I am wondering whether this is the right approach, and if it would be better to use tckgen instead of tckedit. I am a bit overwhelmed by the amount of options that tckgen offers, and would not know which options (and required seeding option) to choose.
Excuse me for my basic question and thanks a lot in advance,
Firstly, I am presuming that the purpose behind the use of tractography (whether whole-brain or targeted) is to use such in the determination of those voxels to consider as belonging to the corpus callosum.
I tried some things with manual drawing of ROIs and applying those in the
tckedit -include and
-exclude options and that works fine. However, I am wondering whether this is the right approach …
Is there any specific reason by which you think this may not be the right approach; or is it purely an issue of uncertainty?
If you don’t have the prior experience with software tools or anatomy with which to reconstruct specifically the corpus callosum in a blind / data-driven way, then starting with a whole-brain tractogram and progressively adding and refining ROIs to get the result you want is absolutely the right way to go about it, at least for an N=1 experiment.
I would quickly summarise the justifications for not using this approach as:
If you have to process a large number of subjects, and therefore don’t want the time investment or rater variance in doing it manually (but doing so for one or a small number of subjects will still give you an idea of how to go about the next step);
If you require a very dense reconstruction, where targeted streamlines reconstruction using
tckgen will produce a larger number of corpus callosum streamlines per unit time than will performing whole-brain tractography and then selecting only corpus callosum streamlines afterwards;
If you have or are establishing a pipeline for reconstruction of this specific structure, and therefore have or will be establishing the set of criteria by which this structure is reconstructed / extracted.
The danger of jumping straight to targeted tracking with
tckgen is that while it’s easy to identify false positive streamlines and subsequently refine your rules to exclude them, you may be entirely oblivious to false negative streamlines, i.e. plausible streamlines trajectories that could be reconstructed, and would be considered a part of the corpus callosum, but are silently excluded due to your criteria. With the
tckedit approach you at least have some capacity to identify these.
Thank you for your clear explanation!
It was indeed an issue of uncertainty and I was also thinking about the false negatives. I think I will play around with different ROIs by using the
tckedit method to get the desired result and leave it there for this project.
Thank you for your help!