I’m working with a 128 channel EEG dataset (geodesic EGI) with individual MRI. In a handful of subjects, the the EEG channel positions in the coregistration EEG/MRI step is bad - the whole net seems to be shifted forward so that the ‘nasion’ ends up on the tip of the nose and some sensors over the eyes.
I have tried refining registration followed by projecting points on surface. I have also tried editing the nasion fiducial point in case I placed it too far down. None of these options have helped.
When I open my head model in the anatomy tab, it doesn’t seem so shrunken into the neck, and this head model is definitely selected in green, so I don’t know why it looks like this on the EEG view.
I just tested removing all EEG points below the nasion and it seems to have fixed the problem.
My questions are:
Is this an appropriate fix?
Would you recommend any other solutions to this problem?
What is the utility of keeping the usually noisy facial channels in the source reconstruction process in general?
Below I have pictures of:
one view of all electrodes (white) with bad alignment
another view of all electrodes (white) with bad alignment
head with the below nasion electrodes removed, showing good alignment
Yes it is. With the scalp surface and the digitized points, the idea is to find a rigid-body transformation for the digitized points to match the best the scalp surface. However, the scalp surface (from MRI) and the digitized points sometimes differ, specially in "soft" structures: cheeks, chin, lips, nose tip, ears. This is caused as the there are differences in the acquisition of the MRI and the digitized points e.g., subject position (thus, gravity effects on those soft structures), facial expression, etc. The top part of the head is less prone to differences.
From your images, it seems a good result.
There is no removal of electrodes, what is removed is the digitized points.
Due to their location the contribution of the facial electrodes to the source estimation would be little or none. However, because the place they are located, there are more prone to movement artifacts. You can always change the type of those electrodes, to be (or not) considered when generating the forward and inverse models.
