We have been using EEGLAB to run preprocessing over our EEG data before analysing it with Python, and recently we began looking into MNE as an alternative for the preprocessing step. We were using ICA to detect and automatically remove EOG components using EEGLAB's Binica, along with FASTER and ADJUST. From a quick read through the MNE docs it looks like we could get similar behaviour if we ran the ICA in MNE, then used the find_bads_eog or detect_artifacts function to identify noisy components. Is this the correct approach? Note that we do not have any dedicated EOG channels in our data, would either approach work fine just using a forehead electrode location (like Fpz)?
This is correct. You can also use the corrmap approach to match EOG
components. We have some local Python implementations of ADJUST and FASTER
that we would not mind sharing if there is interest in this and would
facilitate adopting MNE-Python.
Thanks for the quick reply, so would you say any of those approaches would yield fairly similar results?
We could also certainly be interested in having a look at Python implementations of ADJUST and FASTER if that was convenient to share as we've already done some testing with them.
we can open a PR for ADJUST too. You could then try. I personally have not
compared ADJUST or FASTER to the more MEG-driven temporal ways of dealing
with that we show in our examples. Maybe Jona can say something.