I’ve run FBA on my midlife population sample (N~850) with the hypothesis that it would be more sensitive to early aging-related changes in WM than the traditional DTI models. I’m seeing some evidence that this might be true to a small extent for FD, but FC doesn’t really map onto any of my aging measures once I control for ICV, and i more often see associations in unexpected directions (ie faster aging associated with increased FC)
Since FC is calculated from the warp to the template, i started to wonder if something about the warp was introducing some sort of artifact in my data (admittedly i don’t have any ideas on specifics here, just an appreciation of the influence of the warp and how that differs from the microstructural measures i’m more familiar with). This led to me testing associations between ventricle volume and FBA measures, since I know that’s influential for the warp, and the thought is that there shouldn’t be any real associations outside of the fact that healthier brains should tend to have both smaller ventricles and higher FD/FC (so a potential negative relationship).
As such i was surprised to find that ventricle volume was strongly positively associated with FC in the fornix (segmented with TractSeg; std beta ~.7, controlling for sex and ICV), which incidentally was where we saw the strongest unexpected positive associations with aging. There wasn’t a strong positive association between FC and ventricle volume for any other tracts (there were some with negative associations up to B~-.3). FD in the fornix was strongly negatively associated with ventricle volume (B ~ .8) along with many other tracts (no positive associations).
Certainly the size and proximity of the fornix wrt to the lateral ventricles would make it more susceptible to artifacts from the warp, and maybe I shouldn’t let the finding raise doubt in my mind for the FC measure across the board, but I wanted to see if anyone might have any thoughts on why I’m seeing this or suggestions on resources I might be missing to help me better understand the FC measure and what I’m seeing in my data. I suspect that our large-ish voxels (2.5mm) aren’t doing us many favors either.
Thanks!