The attached picture shows a problem with the refacing ear mask clipping into the brain. In the screenshot, the grayscale image is the original MPRAGE (without refacing), the white outline is the difference between the original and _deFaced.nii (essentially where the refacing mask ends up), the red outline is a quick brainmask using Freesurfer's synthstrip, and the green is where the refacing mask and brainmask overlap.

This is just one example, but with this particular population, it happens roughly 10% of the time, with varying degree of severity. I'm wondering if this is because the population is kids, and so the distance between the surface of their brain and skin is a lot of times less than 10 1mm voxels, and maybe the ear mask was generated with adults? I also wonder if wearing the standard Siemens headphones, which are maybe even compressing what little tissue is there even a few more voxels is making the problem even a little worse.

So, I would like to at least try the refacing again, but tweaking the ear mask. I understand that I need to provide an image in the space of MCALT_FaceTemplate_T1.nii (which I extracted from the zip file), but I'm not clear exactly what the image needs to be. What I want to do is NOT replace the ears, but I don't know what I need to do exactly. Do I make a mask on MCALT_FaceTemplate_T1.nii where the general face area is 1, and the back of the head is 2? Is there a mask like this that already exists somehwere in the zip file that I can edit?

Thanks for your help!

NOTE: This was using version 0.3.5 (docker) on DICOM data from a 3T Siemens Trio.