EEG Data Preprocessing and Artifact Removal

Electroencephalography data were pre-processed and analyzed using EEGLAB (Delorme and Makeig, 2004 (link)) and ERPLAB (Lopez-Calderon and Luck, 2014 (link)) for MATLAB^® (Mathworks). Continuously recorded data, re-referenced to average mastoids at import, was filtered to 256 Hz, for subsequent downsampling at 512 Hz, and high-pass filtered at 0.16 Hz to remove slow drifts. Channels whose amplitude exceeded 150 μV for more than 20% of the recording duration or whose power spectral density was 5 dB larger than the average power over all channels, were automatically labeled as bad. Channels marked as bad were visually inspected and confirmed for rejection. To further remove remaining artifacts, the EEG was filter to 0.16 and 100 Hz and decomposed by applying ICA to good channels only. ICs components with a 90% chance of being an artifact (Muscle, Eye, Heart, Line Noise, and Channel Noise) were automatically identified and removed using ICLabel¹ (Pion-Tonachini et al., 2019 (link)) and eyeCatch² (Bigdely-Shamlo et al., 2013 (link)). The signal was then reconstructed without the artefactual components and channels marked as bad were interpolated.