[Mne_analysis] Watershed problem: inner skull surface excludes cerebellum

Eric Larson larson.eric.d at gmail.com
Mon Feb 4 10:39:34 EST 2019
Search archives:

        External Email - Use Caution        

Have you looked at `brainmask.mgz` to see if it's correct? I suspect that
it is not.

MNE's `mne watershed_bem` is just a convenience wrapper
<https://github.com/mne-tools/mne-python/blob/maint/0.17/mne/bem.py#L1082-L1105>
for `mri_watershed` in Freesurfer. Thus you can try running `mri_watershed`
directly with different options to see what helps. Typically the `-atlas` /
`-gcaatlas` options are useful, as is the `-h` (what we rename `--preflood`
in MNE) option. Sometimes tweaking these can help.

However, sometimes this is not enough to make things work, and you'll need
to dig into why Freesurfer is not working properly -- probably starting
with where it fails first. You can see what Freesurfer's recon_all actually
does, in order, here
<https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllDevTable>. In your case,
if you had done `recon_all -autorecon1`, it would have run just up through
the skull stripping step, which should have been enough to reproduce the
problem if your brainmask.mgz is indeed wrong. You can try manually setting
the control points and/or using -gentle
<https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg11658.html>.
You can look at the talairach transformation to see if it's not quite
right, and if it's not, adjust it manually
<https://surfer.nmr.mgh.harvard.edu/fswiki/Edits>. But before trying those,
one thing I have found helpful is to have an expert options `xopts.txt`
file containing the line:

mri_normalize -mprage -b 20 -n 5

I would actually try that in a clean directory with `recon_all -autorecon1
-xopts xopts.txt` first to see if it magically fixes your `brainmask.mgz`,
and resort to manual control point setting and/or talairach adjustment only
if necessary.

Once everything looks good at the end of `-autorecon1`, you can try running
the `mne watershed_bem` command to see if the output is good. Once it is
(and once brainmask.mgz is correct), you can then proceed with `recon_all
-autorecon2` and `recon_all -autorecon3` to complete your `recon_all` run.

Those are all the tips I can think of offhand. If you need further
assistance, try to isolate your problem to a single Freesurfer command
(e.g., "`recon_all -autorecon1` fails to produce `brainmask.mgz` properly")
and post further questions to the Freesurfer listserv, since that's where
the skull-stripping and watershed experts are more likely to reside. But
please do let us know if you find a solution in the end.

HTH,
Eric


On Mon, Feb 4, 2019 at 5:15 AM Broman Emilia <emilia.broman at aalto.fi> wrote:

>         External Email - Use Caution
>
> Hi,
>
> I have been using mne_watershed_bem to generate BEM surfaces from T1
> images. For 9 subjects (out of 50) the inner skull surface excludes the
> cerebellum. I found an old thread from 2014 where you suggested using the
> gcatlas option and changing the preflood number. I have tried playing with
> these, but the results are still quite bad. Do you have any new ideas on
> how to deal with this problem?
>
> Thanks,
> Emilia
>
> _______________________________________________
> Mne_analysis mailing list
> Mne_analysis at nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190204/58e26c74/attachment-0001.html 


More information about the Mne_analysis mailing list