I am working on a large dataset of participants with eyes-closed resting-state MEG recordings. While preprocessing a small subset for trial, I detected events using both VEOG & HEOG and performed SSP EOG removal. Not enough events are marked for some participants, and SSP EOG throws the error "Not enough time samples to compute projectors," and thus all subsequent processing steps fail. Even in cases where it doesn't fail, manual inspection of detected events makes them seem arbitrary to my untrained eyes.
My question is, is it even appropriate to perform eye movement corrections here if there are no obvious blink or saccade artifacts? (At least, in the 2 participants I saw it for, they may be present for others). If yes, is using SSP the right way to go about it? Could ICA approaches be any better and possible to do on brainstorm? Also, is this step crucial to perform, as my recordings don't have any epochs?
Sorry for the many questions; any insight would be appreciated.
Removing artifacts with SSP relies in the identification of stereotypical artifacts, as it is the case of blinks and heartbeats. In the eye-closed recording, there artifact may not be as frequent, nor as similar between them, so computing the projector to project-out the artifacts.
This may be as the process is trying to detect blinks, rather than any eye movement.
SSP works better when artifacts are repeated enough times, thus, it does not seem appropriated in this case. However, there are enough saccades that are repetitive, it may help to remove them.
In non-repetitive artifacts, you can try ICA. Moreover, try to limit the time window to compute the IC components to a segment of the signal that is heavily influenced by artifacts. As performed in here: