has anyone of you any experience with trying to separate saccades from eye blinks purely based on vertical and horizontal EOG-channels as typically recorded in the Elekta Neuromag MEG system? My main concern currently is whether it would be viable at all to try to do that.
As I interpret MNE's log-messages in find_eog_events ("... help distinguish blinks from saccades" - "Now detecting blinks and generating corresponding events"), it is possible to filter out blinks from the EOG. So I'm wondering how reliably one could identify saccade events. I don't need to know where a subject looked, but knowing when the eyes moved and perhaps in what direction would be great.
I'd appreciate any experience reports and literature pointers.
I don't think MNE has a direct implementation of such a procedure, but
there are few things you could do:
1. Use ICA -- blinks and saccades usually have quite different
components (both in time-course and topographies). There's an
ica_find_eog_events() function in MNE that may be helpful with this.
3. Take advantage of the very different time-courses of blinks and
saccades to some type of rejection. If I recall correctly, Luck
discusses this in his ERP book when evaluating different
(semi)-automatic rejection techniques.
The direction of the saccade may be apparent from the polarity of EOGH
signal or the artefact in the other channels, although I don't know off
the top of my head if e.g. progressive/right saccades would have
positive or negative polarity for a mastoid/average/whatever reference.