Dear Kate,
Indeed the problem is that there is no support of individual head
models for EEG. I was planning to disable the 'MRI' button for EEG
modality and it seems I'll have to do it now since that's what got you
confused. You should use the 'template' button. Here is a short
explanation of why you should not feel bad about doing so. SPM uses a
standard 3-shell BEM model for EEG. Given individual electrode
locations it scales them to this standard head which is equivalent to
scaling the head to fit the headshape/fiducials (as is done for MEG).
This takes care of all concerns regarding head size. With that out of
the way there would be no real advantage in using individual head
models as any improvement they would yield is negligible given other
sources of localization error in EEG. Critically, a major source of
error is our lack of knowledge about precise conductivity of different
parts of the skull. This would not be improved by an individual head
model as there is no information about conductivity available from
SPM-processed MRI images. Bayesian source reconstruction framework
makes it possible to do Bayesian model comparison and formally check
whether using different kinds of head models leads to higher model
evidence. Jeremie Mattout has done some work on this and as far as I
remember has not found an advantage in using individual head models
for EEG.
We are planning to have a wider choice of head models in the future
and if there is robust code for creating individual head models for
EEG we might include it as well. But for now all our experience
indicates that using the template for EEG is perfectly OK.
Best,
Vladimir
On Sat, Nov 8, 2008 at 12:01 AM, Kate Fissell <[log in to unmask]> wrote:
> Hi,
>
> I am trying to use SPM8b, r2448 updates, to work thru the multimodal
> faces SPM8b Manual ch. 33 example. I am sort of working back and forth
> between instructions in that chapter and in the Source Reconstruction
> chapter (15).
> I re-did the segmentation of the structural file in the multimodal
> tar file, just to be sure it was done in spm8b.
> I used the M/EEG Convert button to convert the e_eeg.mat,dat file from
> spm5 to spm8 format.
> I used the Prepare button to load the sensor locations and fids from
> their mat files, in case the convert to spm8 didn't get them.
> To enter fids for the structural, I used the Coregister menu item
> "type" and typed in the coordinates from the smri_fids.mat file from the
> multimodal sMRI directory.
>
> I think I got thru the co-register step OK, and I pressed Save after
> that step.
> When I press the Forward Model button I get the error below.
> What step do I need to do to get the mat info into the dataset ?
>
> Also, the start of section 15.3 in the manual seemed to indicate that
> individual head models based on subject's structural were only
> supported for MEG. Are individual models also supported for EEG ?
>
> thanks
> Kate
>
> ---------------------------------------------------------------------
>
> ??? Reference to non-existent field 'mat'.
>
> Error in ==> forwinv/private/prepare_vol_sens at 228
> if size(vol.mat,1)~=size(vol.mat,2) &&
> size(vol.mat,1)==length(sens.pnt)
>
> Error in ==> forwinv_prepare_vol_sens at 11
> [varargout{1:nargout}] = funhandle(varargin{:});
>
> Error in ==> spm_eeg_inv_forward at 46
> [vol, sens] = forwinv_prepare_vol_sens(vol, sens, 'channel',
> D.inv{val}.forward.channels);
>
> Error in ==> spm_eeg_inv_forward_ui at 28
> D = spm_eeg_inv_forward(D);
>
> Error in ==> spm_eeg_inv_imag_api>Forward_Callback at 88
> handles.D = spm_eeg_inv_forward_ui(handles.D);
>
> Error in ==> spm_eeg_inv_imag_api at 54
> feval(varargin{:}); % FEVAL switchyard
>
> ??? Error using ==>
> spm_eeg_inv_imag_api('Forward_Callback',gcbo,[],guidata(gcbo))
> Reference to non-existent field 'mat'.
>
> --------------------------------------------------------------------------
>
|