Getting partially clean epochs from autoreject

Support & Discussions
,
1

I’m dealing with long continuous data, and I split them into around epochs of 30s-1min. Autoreject will remove epochs that are bad, I assume these are epochs that can’t be repaired?

To save as much data as possible, I’m wondering if there’s a way to not drop these bad epochs, and only take the clean channels in those epochs instead? (My analysis can handle variable number of sensors)

2

Hi, you can retrieve the reject log:

ar = autoreject.AutoReject(…)
ar.fit(epochs)
epochs_ar, reject_log = ar.transform(epochs, return_log=True)

Then, use reject_log.bad_epochs to list bad epochs and reject_log.labels to retrieve which channels are bad in those epochs.

Best wishes,
Richard

3

Thank you so much for the quick answer! This is exactly what I need.