During the acquisition of EEG with the BioSemi device (64 electrodes), we noticed a strange artifact with the following features:
- Sinusoidal type (sum of sinusoids), with dominant frequencies from (15 - 30 Hz).
- Mainly present in the peripheral electrodes as Fp1, Fp2, T3, T4, and others. Even in some files, the artifact is present in motor cortex electrodes as C3 and C4.
- Across the full length of the EEG file, from the beginning to the end and no-event related.
- Present in the majority of components, therefore brain activity is eliminated in the process of rejecting this artifact.
Below a description of the preprocessing steps that are used in our lab for data cleaning with eeglab:
- Data import (Cz as reference)
- Downsampling (2048 Hz to 256 Hz)
- Filtering 1 - 35 Hz - pop_eegfilt( EEG, 1, 35, [], [0], 0, 0, 'fir1', 0);
- Visual inspection and rejection of artifacts in time
- Interpolation of channels (when it is necessary)
- Re-reference (Average referencing)
- Decomposing data with PCA (we use 24 components for faster processing)
- Rejection of noisy components
Does anyone know how we can prevent the artifact, or how we can remove it in preprocessing?