An Unsupervised Method for Artefact Removal in EEG Signals.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) signals are vital for brain activity monitoring but often contaminated by artefacts.
• Current artefact removal methods, frequently employing Independent Component Analysis (ICA), necessitate manual intervention or complex learning processes.
• Developing an automated, unsupervised approach for EEG artefact removal is crucial for efficient and reliable brain signal analysis.

Purpose of the Study:

• To develop a generic and unsupervised algorithm for removing artefacts from EEG data using ICA.
• To automate the selection of artefact-related independent components (ICs) without human intervention.

Main Methods:

• The algorithm integrates unsupervised artefact detection, ICA, and a statistical criterion for automatic IC selection.
• Evaluation involved both simulated and real-world EEG datasets (SEEG and AEEG) containing artefacts.
• A comparative analysis was performed against supervised IC selection methods.

Main Results:

• A novel unsupervised ICA-based algorithm was developed for effective EEG artefact filtering.
• The algorithm automatically identifies and selects artefact-related ICs, suitable for online applications.
• It successfully removes diverse artefact types while preserving significant amounts of original EEG information.

Conclusions:

• The developed unsupervised ICA-based method offers a significant advancement in automated EEG artefact removal.
• It eliminates the need for manual selection or prior learning, making it broadly applicable.
• The algorithm's generic nature and ability to retain data integrity address key limitations in existing ICA-based artefact removal techniques.