Dear Jerome and other SPM MEG users,

Today we had a discussion with some much better experts in head
modelling than I and the bottom line is that for MEG (and MEG only) it
does not matter whether the cortical mesh sticks out of volume
conductor mesh or not. So you can use either SPM or FT version of the
meshes and it should have little effect on your results (apart from
possible differences in coregistration).

'Sticking out' is indeed a problem for EEG, but presently this is also
not an issue in SPM8b because for EEG we only use template meshes
which are always the same and they are OK.

Best,

Vladimir

On Tue, Nov 25, 2008 at 7:06 PM, Vladimir Litvak
<[log in to unmask]> wrote:
> Dear Jerome,
>
> SPM does not actually segment the cortical surface to create the mesh.
> This is too complex to be done automatically in a good way. It always
> uses the template mesh using inverse normalization to fit it into the
> subject-specific inner skull mesh. This is not guaranteed to work as
> you indeed experienced. As I told you before, this  is unlikely to
> happen if you use 'SPM' version of the meshes which are smoother. You
> should examine every subject's meshes to see that everything is OK. If
> you still get the problem for some subjects there is no choice but use
> the template for those subjects. Since you have a polhemus head shape
> the template is scaled quite precisely to match your subjects' head
> size so perhaps you should just use the template for everybody.
>
> The problem with EEGLAB import has something to do either with setting
> of the path or with some functions missing from EEGLAB distribution.
> What you can do is copy all the files from SPM/external/fileio/private
> to a separate directory (do not call it 'private') and put it in your
> Matlab path with EEGLAB. This should make things work for you but in
> principle this problem should be solved by the EEGLAB team.
>
> Best,
>
> Vladimir
>
>
>
> On Tue, Nov 25, 2008 at 6:22 PM, Jérôme Courtemanche
> <[log in to unmask]> wrote:
>> Hi Dr Litvak,
>>
>>  1)  When i found out that my results were much better using the template
>> instead of my subject's MRI, I took a close look at the segmentation of the
>> scans. There is some regions were the cortex springs out of the CSF so this
>> is bad (as you warned me). I thought it was because of inhomogeneities
>> (known to be severe with 3T data). I used the MNI N3 tool to correct it, and
>> with different degree of non-uniformity correction, it changed nothing in
>> the problem. image attached.
>>
>> a) Theorically, are the results supposed to be better if one uses the
>> anatomic instead of the template-polhemus duo?
>> b) If so as I hope, what is your advice in my situation?
>>
>> 2) The EEGLAB people didn't reply my question in a week:
>>
>>  I tried to convert SPM8b MEG file to EEGLAB using the fileio (even tried
>> with the updated one ftp://ftp.fcdonders.nl/pub/fieldtrip/modules/).
>>
>>  But after trying to load either the .dat file or the .mat file
>>  I get the message:
>>
>> eeglab error in function read_header() at line 311
>> Undefined  command/function 'keyval'
>>
>> Have you ever succeeded converting spm2eeglab
>>
>> Very best regards,
>>
>> Jerome
>>
>