Q: help on 'too many neighbors bug'

[kingjr]

Hey,

I'm encountering an error on some of my subjects when running mne.bem.make_watershed_bem and I'm not sure what to do here. Do you think it's related to a non-closed mesh?

make_watershed_bem(subject='subject05', subjects_dir='data',
                   overwrite=True, volume='T1', atlas=False,
                   gcaatlas=False, preflood=None)

leads to

3. Creating medium tessellation...
3.1 Decimating the dense tessellation...
3.2 Creating data/subjects/subject05/bem/subject05-head-medium.fif
Running subprocess: mne_surf2bem --surf /tmp/tmpU6HNDB/tmp-surf.surf --id 4 --fif data/subjects/subject05/bem/subject05-head-medium.fif
stderr:

mne_surf2bem version 1.8 compiled at Aug  9 2015 04:22:26

input  file #   1 : /tmp/tmpU6HNDB/tmp-surf.surf / id = 4 / sigma N/A
output file       : data/subjects/subject05/bem/subject05-head-medium.fif

Triangle file :  nvert = 14703 ntri = 29998
    Triangle and vertex normals and neighboring triangles...[done]
    Vertex neighbors...Too many neighbors for vertex 10884.

(u'', u'\nmne_surf2bem version 1.8 compiled at Aug  9 2015 04:22:26\n\ninput  file #   1 : /tmp/tmpU6HNDB/tmp-surf.surf / id = 4 / sigma N/A\noutput file       : data/subjects/subject05/bem/subject05-head-medium.fif\n\nTriangle file :  nvert = 14703 ntri = 29998\n\tTriangle and vertex normals and neighboring triangles...[done]\n\tVertex neighbors...Too many neighbors for vertex 10884.\n')

which ends up crashing

---------------------------------------------------------------------------
CalledProcessError                        Traceback (most recent call last)
scripts/run_mne_anatomy.py in <module>()
     44     # Make scalp surfaces
     45     make_scalp_surfaces(subjects_dir, subject, force='store_true',
---> 46                         overwrite='store_true', verbose=None)
     47 
     48     # Make morphs to fsaverage

/home/jrking/anaconda/lib/python2.7/site-packages/mne/commands/mne_make_scalp_surfaces.pyc in _run(subjects_dir, subject, force, overwrite, verbose)

/home/jrking/anaconda/lib/python2.7/site-packages/mne/utils.pyc in verbose(function, *args, **kwargs)
    538         finally:
    539             set_log_level(old_level)
--> 540     return function(*args, **kwargs)
    541 
    542 

/home/jrking/anaconda/lib/python2.7/site-packages/mne/commands/mne_make_scalp_surfaces.pyc in _run(subjects_dir, subject, force, overwrite, verbose)
    136         try:
    137             run_subprocess(['mne_surf2bem', '--surf', surf_fname, '--id', '4',
--> 138                             '--fif', other_fname], env=this_env)
    139         finally:
    140             del tempdir

/home/jrking/anaconda/lib/python2.7/site-packages/mne/utils.pyc in run_subprocess(command, verbose, *args, **kwargs)

/home/jrking/anaconda/lib/python2.7/site-packages/mne/utils.pyc in verbose(function, *args, **kwargs)
    538         finally:
    539             set_log_level(old_level)
--> 540     return function(*args, **kwargs)
    541 
    542 

/home/jrking/anaconda/lib/python2.7/site-packages/mne/utils.pyc in run_subprocess(command, verbose, *args, **kwargs)
    806         err_fun = subprocess.CalledProcessError.__init__
    807         if 'output' in inspect.getargspec(err_fun).args:
--> 808             raise subprocess.CalledProcessError(p.returncode, command, output)
    809         else:
    810             raise subprocess.CalledProcessError(p.returncode, command)

CalledProcessError: Command '['mne_surf2bem', '--surf', '/tmp/tmpU6HNDB/tmp-surf.surf', '--id', '4', '--fif', 'data/subjects/subject05/bem/subject05-head-medium.fif']' returned non-zero exit status 1

Thanks a lot!

PS: shall I use the mailing list rather than github for this kind of questions?

[agramfort]

yes use the mailing list.

have you tried to use the atlas or gcaatlas options which help if you have a crappy MRI.

[larsoner]

I usually get that error when there is noise at the bottom of the MRI. 99% of the time, zeroing out the bottom slice in T1.mgz and rerunning the surface generation fixes it for me.

[kingjr]

@Eric89GXL Thanks, I never manipulated mgz data, do you have a script that check this noise and zero it?

[larsoner]

No, we do it manually using FreeView because it's so rare, but come to think of it, it should be almost trivial with nibabel to do the zeroing. We never check the noise, we just try the zeroing fix and it almost always works.

[kingjr]

It was a libgfortran lib issue missing from the os environment.

Thanks for your help!