Multiple head models for multiple subjects?

Discussions
1

Hi,
I am analysing EEG data from 18 subjects from a visual oddball task. After creating a protocol, I have created one Subject and named it ‘Subject Group’. I treat this as my data bucket into which I import files from each subject (cleaned, epoched standard and deviant files). I treat the files from each subject as one run withn the ‘Subject Group’ (I have a total of 18 standard and 18 deviant ‘runs’ now)
Now, I have computed Noise Covariance matrix taking all the epoched files together. Can I use the averaged standard and deviant file as inputs while calculating just one OpenMEEG BEM head model under the ‘Subject Group’? Or do i need to compute a head model separately for each subject? I have done this to avoid having to copy the head model to all conditions (I am using a script) .

My next question is, do I calculate both data and noise covariances for the sLORETA source imaging (I am using unconstrained method), or is the noise covariance enough (I could not get a clear picture of this from the tutorial)?

Thanks,
Swathi

2

Hello,

You need one noise covariance and one separate inverse model per subject. The noise covariance can be evaluated from the individual trials for each subject separately, as suggested here:
http://neuroimage.usc.edu/brainstorm/Tutorials/NoiseCovariance#Variations_on_how_to_estimate_sample_noise_covariance

If you are using the same anatomy, same selection of channels and same electrode positions for all the subjects, then the head model is the same for all the subjects. You can compute the model once and copy it to the other subjects. If you are looking for how to do this in a script, you can maybe find some inspiration in the script for the group analysis tutorial:

For the minimum norm solution (and its normalized versions sLORETA and dSPM), you need only a noise covariance matrix. The data covariance is needed only for the beamformer.

Francois

1 Like
3

Thanks, Francois! Will make the changes…