I tried ico-4 and ico-5, and still get the same results ? poor correlation between scalp and cortical sources. I have not tried the ?loose option yet, will do that this afternoon. I did try ?diagnoise, and that seemed to improve matters (though I do not know why ? the manual says most users never need to use this), but I am experiencing issues with mne_analyse in trying to display the three different estimates. As I change from dSPM, which comes up as defult, to MNE, the inflated brain pane goes white, and I cannot get it back until I exit and go back in. Ever seen this before?
I did not get any response on the issue below, which prevents me from viewing my inverse results. Anyone have any suggestions? I can post my files if necessary. Thanks,
-Jeff
I tried ico-4 and ico-5, and still get the same results ? poor correlation
between scalp and cortical sources. I have not tried the ?loose option yet,
will do that this afternoon. I did try ?diagnoise, and that seemed to
improve matters (though I do not know why ? the manual says most users never
need to use this), but I am experiencing issues with mne_analyse in trying
to display the three different estimates.
did you correct for the bad channel like matti suggested? did you take a look
at the noise covariance you estimated from the data? Ideally it should be
as diagonal as possible with similar values along the diagonal. See e.g.
As I change from dSPM, which comes
up as defult, to MNE, the inflated brain pane goes white, and I cannot get
it back until I exit and go back in. Ever seen this before?
you certainly need to adjust the scaling of the colorbar.
I used the --diagnoise switch so as not to have to worry about the
actually noise-cov matrix at this point.
The problem is not a scaling issue - it is a severe display issue/bug -
the entire pane goes completely white, no surface displayed, nothing, and
at that point I can not even get back to any of the three types of
inverses. The only way to continue is to exit mne_analyse and restart it.
Are you using OSX? I've had an issue with the display not working
appropriately sometimes (alternates between blank/maybe white and some
fixed image even when it should be updating). I chalked it up to an
improper implementation of OpenGL/XWindows on OSX since I have never
run into the same issue on linux x86_64 or i686.
I moved my files from Mac to Linux. I can open the averaged EEG, but when
I try to open the surfaces they cannot be found. I am sure this is because
their location was hard-coded in the fif files. Is there any way to change
this? Otherwise I seem to have two options.
Set up identical directory structure on Linux to match Mac folders
Or
Re-run the part of the MNE analysis that code the locations of surface
(and other?) files into the fif files.
I have set the appropriate FreeSurfer environmental variables, thinking
that would be enough, but apparently not.
That's odd---going between machines should work. I've done
parcellation and source space/BEM setup on one Linux machine, moved
the subject structural to a different linux machine with a different
directory structure, and after generating averages and inverses there,
copied everything to my OSX install (again with a different path
structure) where they load without issue.
On both my OSX and Linux installs, the variables I set are
MATLAB_ROOT, SUBJECTS_DIR, and SUBJECT, then run
$MNE_ROOT/bin/mne_setup_sh and $FREESURFER_HOME/SetUpFreeSurfer.sh. I
assume you've done this, but if you have Freesurfer installed, you can
double-check that SUBJECTS_DIR and such are set appropriately by doing
"tkmedit SUBJNAME T1" where SUBJNAME is the name of your subject will
also test to see if that subject's structural is in the expected
location. You can also then try loading the surfaces that MNE looks
for
It's important to note, however, that the Freesurfer folks (and maybe
MNE, too?) recommend against switching versions mid-analysis. If it's
not too difficult, I'd recommend re-running the analysis on Linux.
Might be the easiest solution in any case.
I may no be doing the mne_setup_sh part, will check on that tomorrow. With
all you have given me, I can now probably figure it out. I have not done
any real analysis yet, just learning the ropes. Thanks,
The subject name is coded into the source space, forward solution, and inverse operator files. This means that the directory structure under SUBJECTS_DIR has to remain intact. Otherwise, the location does not matter as long as SUBJECTS_DIR is set correctly. HOWEVER, if you have used media with a windows file system on it to transfer files, strange things may happen because symbolic links are not supported.
I have it working now. I was modifying my local bash profile, but out sysadmin told me to not do that as the "master" profile was doing things correctly. Good suggestion on Eric's part to try FreeSurfer by itself.