Reference data for longitudinal assessment of bilateral tracts

Hello fellow mrtrix users!

Another question related to this topic. We are analysing data from two time-points in a cohort of ischemic stroke patients and we want to predict what white matter fibers will be affected. Then we would like to measure changes along these predicted fibers, such as changes in FA, FOD and fiber integrity between the two time points (an earier measurement post-stroke vs. a later measurement post-stroke). When we perform that, we would like to compare the observed changes to those in control areas and I have a bit of a conceptual question here: what areas would best represent control areas if the subjects have a stroke in a streamline that projects bilaterally? A lot of studies (in fact, most studies I read on the topic), typically measure asymmetry indexes on the ipsi- and contra-lateral side, i.e. a lesioned CST on one side, compared to the intact cortico-spinal tract in the other hemisphere. Certainly, for bilateral streamlines, using the contra-lateral (and contra-lesional side) won’t be that useful as a control area if we expect distal degenerative changes there too, over time. Would you have any recommendations?

Thank you very much in advance.

Best regards,
Ivana

Hi @Ivana,

I moved this post to a new topic since the core question didn’t really fit with the title of the previous topic; it’s still possible to link to other topics if there may be relevant information.

… we want to predict what white matter fibers will be affected. Then we would like to measure changes along these predicted fibers, such as changes in FA, FOD and fiber integrity between the two time points (an earier measurement post-stroke vs. a later measurement post-stroke)

I’m not sure if this is just an issue of linguistics, but given the current state of science, the word “predict” should be used only when it is accurate to do so. If I take your current wording literally, it sounds like there are two separate experiments here: one building some sort of predictive model that learns from longitudinal data how to identify tracts likely to be affected in the future based on acute scan data only, and a second that uses the results of this predictive model to extract quantitative measures specifically only from those tracts promoted by the predictive model. If that is in fact the case, then there isn’t enough information provided to address specifically the first experiment. If not, I’d suggest avoiding the term “predict” in such a way, as otherwise you get perfectionists like myself jumping up and down :crazy_face:

… and fiber integrity …

:grimacing:

We have a long history of disdain for this term… You might find some discussions of such in older topics on this forum, or pretty much any diffusion MRI review article from the last decade will explain why use of this term is discouraged.

what areas would best represent control areas if the subjects have a stroke in a streamline that projects bilaterally?

This probably goes all the way to the heart of the “What is your hypothesis?” question, specifically because a proper understanding of such intrinsically provides what the null hypothesis must be. So it’s more of a science question than a software question.

Do you want to know if quantitative values for a specific tract change over time … :

  • … More than other tracts? Then segment lots of other tracts and use their longitudinal values.

  • … More than the rest of the brain white matter? Then compare it to the rest of the brain white matter.

  • … More than other bilateral tracts? Then segment all other bilateral tracts.

  • … More than the same tract in healthy controls? Then segment that tract in healthy controls.

Rob