Dear FreeSurfer experts,
I'm running longitudinal recon-all (2 timepoints/subject) in FreeSurfer 8.1 on pooled adult MPRAGE and MP2RAGE (MPRAGEised UNI) data, and would appreciate your recommendation on a few points.
1. Custom masking (-xmask) is overwritten by SynthStrip's internal call in FS8.1, so I inject a custom brainmask.mgz manually after -autorecon1 at the cross-sectional stage - this works. At the base stage, mri_robust_template correctly combines both sessions' brainmasks into brainmask_template.mgz (confirmed visually), but recon-all then runs SynthStrip fresh on the template and uses that instead. Is brainmask_template.mgz meant to propagate forward as the template's brainmask.mgz but gets overrun by the automated SynthStrip call? And is global-expert-options.txt the recommended way to enforce consistent SynthStrip settings across cross, base, and long, as an alternative to manual injection at each stage?
2. Default SynthStrip is sometimes overinclusive around the eye/optic nerve on our MPRAGEised UNI images. We're considering --no-csf (with -b 2 to offset its known underinclusiveness in MPRAGE images) to address this. Are there downsides to --no-csf we should be aware of for FreeSurfer's own downstream processing specifically - e.g., anything in intensity normalization, Talairach registration, or pial surface placement (mris_make_surfaces) that relies on some CSF margin being present in the mask?
3. For the MP2RAGE/MPRAGEised data: is it preferable to submit the full (non-skull-stripped) MPRAGEised UNI image as input, letting recon-all handle stripping (or interject with custom mask), or to submit a pre-stripped image plus mask (as presurfer-style pipelines recommend)? The recommendations I'm aware of for this are based on older FreeSurfer versions, so I'm wondering whether the same guidance still holds for 8.1, or whether its updated components (e.g. SynthStrip/SynthMorph) change that recommendation. The full image's non-brain regions have atypical, non-anatomical UNI intensities, and I'm unsure whether that helps or hinders FreeSurfer's intensity normalization versus supplying an already-stripped image.
Thank you very much for your time.
All the best, Ann-Christin
-- Ann-Christin Kimmig, Ph.D. Visiting Postdoctoral Research Fellow Psychiatry Neuroimaging Laboratory Department of Psychiatry Brigham and Women's Hospital Harvard Medical School