Voxelwise approach (volume) with reduce coverage from white matter only

[j-bourque]

Dear brainSMASH community,

This tool is awesome, and we've used it previously in our lab. Now we want to use it for a white matter voxel map from a white matter tract atlas. We are now following the whole-brain volume example for this white matter analysis. However, we were wondering whether the brainsmash code for the voxelwise approach would still work as intended with our reduced coverage of the brain (only white matter voxels from the tract atlas). I am sorry if this is a really beginner's question. Thank you for your time, Josiane

[jbburt]

Hi Josiane,

Thanks for the question! This isn't a beginner's question at all.

My initial reaction is that BrainSMASH may not be an appropriate tool for white matter analysis. For axonal fiber tracts in white matter, the orientation of the tracts is likely as important as (if not more important than) spatial proximity. For instance, you might expect two neighboring voxels to be far more similar if they belong to the same tract than if they belong to neighboring parallel tracts. Information about the orientation of fiber tracts could be gleaned from eg diffusion tension imaging, but it's not at all obvious how to incorporate this information into the BrainSMASH framework.

It might be possible to use probabilistic tractography to compute a measure of "closeness" within and across white matter voxels by computing statistics on the resulting streamlines; this measure could then be used in lieu of Euclidean distance to form the distance matrix. But I'm really not sure how one would do this in practice.

What do you think?

Josh