External Email - Use Caution
Hello FreeSurfer Experts,
I’m attempting to generate inner/outer skull and skin surfaces using mri_watershed in order to compute a forward operator in MNE (don’t have FLASH for mri_make_bem).
The issue I have is that the outer skull surface is incorrect and looks more like the skin surface, including a large amount of neck. Upon closer inspection, this seems to be due to large chunks of dura included in the brainmask (mostly in the area superior to the interhemispheric fissure), pushing the surface out too far.
To fix this, I tried running recon-all -skullstrip with various flags, and found that a lower watershed threshold with the -gcut flag worked really well to remedy this problem, producing an accurate brainmask and brain.mgz files (after re-running -autorecon2). My understanding was that this change to brain.mgz should have then improved the surfaces generated when subsequently running the watershed algorithm- however a new brain.mgz file that looked like the original incorrect volume (before using gcut) was instead generated.
I’ve searched the list and similar issues have been reported (see ”e.g.”, below), but I have not yet found a solution - is it possible to use the brainmask or brain.mgz file that I already have when running mri_watershed to generate skin and skull surfaces? There is a -mask flag to “mask a volume with the brain mask”- does this mean that it finds the existing brainmask.mgz file generated by autorecon1? I tried using it, but the brain.mgz file generated was not restricted to the voxels included in the previously generated brainmask. Please clarify the usage of this flag.
Alternatively, is it possible to somehow use gcut when running mri_watershed on its own (as opposed to being run as part of recon-all -skullstrip)? Or to create the skin and skull surfaces when running recon-all -skullstrip with the -gcut flag?
I am using freesurfer 7.1.1 on OSX Mojave
E.g. - others reporting similar issue: https://secure-web.cisco.com/1xwUD8H2ELqmT-XJg96ZsminARQ0lHfB96Ll9HKpRLV0qUJ... https://secure-web.cisco.com/1JkDmZCv6uauCpr2tI2Yaw1OLcLrdPSvI9AMarDaaIy20tW... https://secure-web.cisco.com/1cACd5dT5U95tvMeIiQb6mqBCDQWypNk0zlZRiSVFkBzr1l...
Many thanks in advance, Danielle