External Email - Use Caution
Hello FS Community,
I'm working on a study about left-right asymmetry and one of the experiments I'd like to do is to directly compare the average percent change between sides for each ROI when processed using a volumetric registration and when processed using FreeSurfer's surface-based approaches (same set of images). I'm hoping to characterize the degree to which both software produce consistent patterns of LR asymmetry for volume. However, I'm not sure what the best way is to implement a symmetrical registration with ROIs in FreeSurfer and I'm wondering what is recommended?
For starters, the DK atlas/template is not feasible due to its inherent asymmetry, and I've considered a number of approaches but it feels like a challenge prohibits each one. What is the best way to go about doing this?
I've considered the following ideas:
1. (Most ideally) I could import my already obtained symmetrical volumetric template/atlas into FreeSurfer, but my understanding is that making the .gca file requires multiple annotated subjects rather than a single annotated atlas, since FS uses probabilistic segmentations. 2. I've looked at the xhemi approach from Greve et al. 2013 but my impression from reading the documentation is that this gives results that differ based on whether the left hemisphere or right hemisphere is used and also that the final results are described for each vertex, rather than for each parcel which I need (although the original paper did do some ROI-analysis). 3. I've also thought about performing recon-all on each sample twice: once in the proper orientation and once more after flipping the LR orientation of each sample. From there I could compare the left hemi and the flipped-right hemi to each other which I think avoids the DK atlas bias. This feels similar to option 2 and so I'm not sure if this would give results that vary based on which hemisphere is used for analysis. Also the documentation says I would need to manually change lines in the text file, which is not an ideal solution for several hundred subjects, although this could be scripted. (https://secure-web.cisco.com/1b40CbQ8lToemytkQqJc49_lPSzSyb_qOkrmyKts2VtGlgw...).
Maybe I'm overthinking it. What is the best way to go about doing this?
Thank you, Andrew