Dear John,
Fiducials are any 3 points whose locations are known in both the structural image space and in the sensor space. If you don't have fiducials you can take any 3 points whose locations you know in both coordinate systems and use them as such. Custom meshes can be specified in the batch tool for head model specification in SPM12. They need to be stored in GIfTI format and be in the same space as your structural.
For SPM12 I wouldn't recommend you to try to hand-craft your own sensor representations. In the EEG case you can put your electrode locations and fiducials in the .sfp format and add them to your dataset using 'Prepare' tool. For MEG if your simulation is based on some raw MEG dataset you can read the sensors and fiducials from this dataset also via Prepare.
Finally, if you want to use your own lead fields you can let SPM generate a gain matrix .mat file and then replace the gain matrix there with your own. That would work for imaging. For dipole models I don't think there is an easy way to do that.
Best,
Vladimir
On 1 Oct 2014, at 19:54, "John D. Griffiths" <[log in to unmask]> wrote:
> Dear List,
>
> I would like some advice on creating and 'analyzing' synthetic M/EEG datasets in SPM.
>
> I am running simulations with neural mass models and EEG + MEG projections outside of SPM, and using spm_eeg_ft2spm to produce a D obj that I want to run through the M/EEG machinery as if it were real data.
>
> I an uncertain about two things:
>
> 1) How can enter an already-computed forward model into the D obj?
>
> 2) How can I supply the exact locations of sources (mesh surface points) and sensors, without having to estimate coregistration parameters?
>
> I am ok supplying cortical, inner, outer, and scalp surfaces, and the locations of channel positions.
>
> The problems seem to start at the coregistration stage. SPM doesn't seem to want to let me do anything unless fiducials are defined. I have by construction the locations of the sensors and mesh points and if needed also T1 image points, but I don't have any separate fiducial points per se.
>
> I can see that the info appears to be stored in our around D.inv{1}.datareg.sensors.chanpos, D.inv{1}.forward.sensors.chanpos, but could really do with some specific example code of how to supply this information manually. Could anyone help me out with this?
>
>
> Thanks,
>
>
> john
|