Trouble with mne_watershed_bem

✉️ Mailing List Archive (read-only)
1

Hi MNEers,

I am using mne_watershed_bem to create boundary element models (my MRI data
do not seem to be in the right format to use FLASH images or seglab, and
so far I haven't had much luck getting it done in BrainSuite); the results
for the brain surface and inner skull look good, but the results for the
outer skull and outer skin do not (that is to say, those boundaries go way
off the top of the head, making what looks like a mushroom above the head).
However, my functional data is all MEG data (I'm not planning on doing any
source localization of EEG data), and my understanding of the materials
I've looked at is that MEG solutions just need the inner skull; thus, would
it be ok for me to go ahead with poor segmentations of the outer skull and
scalp?

Also, if it turns out I do need the outer skull and scalp, are there any
options in mne_watershed_bem that I can use to improve the segmentation?
I've tried running it both with and without --atlas, but I only saw a
little bit of improvement. I noticed that the corresponding Freesurfer
routine, http://surfer.nmr.mgh.harvard.edu/fswiki/mri_watershed, has many
parameters that can be specified, but I don't know which of these (if any)
might improve the segmentation.

Thanks,
Steve Politzer-Ahles

2

Hi Stephen,

You can go ahead as you suggest (being sure to use --homog). For the purposes of registering your polhemus data to your MRI, I do suggest that you fix the outer skin layer. For fixing the outer skin there is not really any tweaking to be done to the call to mri_watershed, because for fitting the outer skin: a surface is just fit around the first high voxel intensities. I recommend you open your T1.mgz file in tkmedit and erase the voxels near your mushroom shape. mri_watershed is detecting these high intensity voxels and placing the surface around them (erasing the high intensity voxels, will fix the problem). You should be able avoid the problem in future participants through a slightly different MR acquisition (speak to your local MRI experts and show them the artifact, otherwise, e-mail me offlist, and I might be able to tell).
HTH
D