Dear Ozge,
Since you are not actually using SPM methods, I don't see any
motivation for us to debug your code. The BEM method used in SPM can
be inaccurate for points located close to the boundary and while it's
not a big problem for source reconstruction in SPM it might be for
some other methods. You can try the 3-spheres model or the more
accurate OPENMEEG BEM, which is supported in the code of
spm_eeg_inv_forward but not in the GUI because the required openmeeg
library can be tricky to install and the running time can be quite
long. See http://www-sop.inria.fr/athena/software/OpenMEEG/
Best,
Vladimir
On Fri, Dec 21, 2012 at 1:10 PM, Ozge Batu <[log in to unmask]> wrote:
> Dear SPM users,
>
> I am trying to solve the EEG inverse problem by using the leadfield matrix obtained by using the SPM toolbox. Even in case of a simple simulated data which consists of a randomly generated sparse vector of unknowns, EEG leadfield matrix and l1-norm loss function; I cannot obtain a successful reconstruction. I couldn't figure out why the leadfield matrix results in such inaccurate solutions. Could it be related to the spark, nullity or any other property of the matrix? The leadfield matrix and the sample script that I used can be downloaded from
> https://dl.dropbox.com/u/19664203/simulated_problem.7z
> I'd appreciate if you could help me to understand the reasons behind this. Thank you for your time.
>
> Best,
>
> Ozge
|