An Unsupervised Multichannel Artifact Detection Method for Sleep EEG Based on Riemannian Geometry

Area of Science:

• Biomedical Signal Processing
• Neuroscience
• Machine Learning

Background:

• Artifacts in electroencephalography (EEG) recordings, particularly during sleep, significantly hinder signal analysis and interpretation.
• Existing artifact detection methods often focus on specific artifact types or require sequential, single-channel processing, increasing complexity and reducing efficiency.
• The need for robust, multichannel artifact detection is critical for accurate sleep EEG analysis.

Purpose of the Study:

• To develop a novel, multichannel artifact detection method for sleep EEG signals.
• To address the limitation of existing methods by enabling simultaneous rejection of diverse artifact types.
• To leverage Riemannian geometry principles for enhanced artifact detection performance.

Main Methods:

• A new multichannel artifact detection algorithm inspired by Riemannian geometry was developed.
• The proposed method was tested on real-world sleep EEG datasets.
• Performance was evaluated against expert annotations, a simpler Riemannian geometry-based method, and the FASTER algorithm.

Main Results:

• The proposed multichannel method demonstrated high effectiveness in detecting various artifact types in sleep EEG.
• Comparative analysis showed superior or comparable performance to existing state-of-the-art methods.
• The method successfully processed multichannel data simultaneously, offering an advantage over single-channel approaches.

Conclusions:

• The developed multichannel artifact detection method offers a significant advancement in sleep EEG signal processing.
• This approach provides a more efficient and effective solution for artifact removal, improving the reliability of sleep studies.
• The findings highlight the potential of Riemannian geometry in developing sophisticated biomedical signal processing tools.