Selecting a pre-processing pipeline for the analysis of EEG event-related rhythms modulation

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Electroencephalography (EEG) data preprocessing involves multiple methodological choices.
• These choices can influence the reliability and quality of the final cleaned EEG data.
• Action Observation and Motor Imagery (AO/MI) protocols are sensitive to preprocessing artifacts.

Purpose of the Study:

• To investigate the impact of different EEG preprocessing strategies on data quality.
• To quantify the effects of segmentation, Independent Component Analysis (ICA) algorithms, and re-referencing techniques.
• To optimize EEG data cleaning for AO/MI tasks.

Main Methods:

• EEG data from an AO/MI protocol were preprocessed using varying segmentation approaches.
• Two ICA algorithms, Signal-Space Projection (SSP) and Extended Infomax, were evaluated for artifact removal.
• Three re-referencing methods were compared: Common Average Reference (CAR), robust-CAR, and Reference Electrode Standardization Technique (REST).

Main Results:

• Data segmentation demonstrated a significant impact on the effectiveness of the artifact removal process.
• Extended Infomax was superior in identifying and removing artifactual components compared to SSP.
• CAR and REST re-referencing techniques produced comparable and high-quality results.

Conclusions:

• EEG data segmentation is a critical preprocessing step influencing overall data quality.
• Extended Infomax is recommended for artifact detection in AO/MI EEG analysis.
• Both CAR and REST are effective re-referencing strategies for improving EEG data quality.