Channel type 'csd' not accepted by autoreject

Hi all, I’m using MNE version 0.24.

Trying to use the autoreject package for pre-processing clinical EEG acquisitions. Autoreject works fine with the epoched data when channel types are ‘eeg’ but if I compute current source density prior to autoreject, the channel type changes from ‘eeg’ to ‘csd’. Using ar.fit_transform(epochs) then throws the following error: ValueError: No channels match the selection. Any advice?

Hello @kartavya15 and welcome to the forum!

@mainakjas is the creator of autoreject, maybe he knows more about this limitation?

cc @adam2392

Replied on the github issues page: :slight_smile: