Robust detrending, rereferencing, outlier detection, and inpainting for multichannel data

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalography (EEG) and magnetoencephalography (MEG) data are susceptible to artifacts like glitches and drifts.
• These artifacts hinder accurate data analysis and interpretation in neuroscience research.
• Current artifact removal methods can be wasteful, often requiring data discarding.

Purpose of the Study:

• To present a suite of robust, automated techniques for artifact removal in EEG and MEG data.
• To offer effective alternatives to data discarding, thereby preserving valuable experimental information.
• To enhance the overall quality and reliability of neurophysiological recordings.

Main Methods:

• Robust detrending to remove slow drifts and common mode signals.
• Robust rereferencing to mitigate artifact impact on the reference signal.
• Outlier detection and data interpolation (inpainting) for corrupt data segments.
• Specific artifact removal techniques including step removal and filter ringing suppression.

Main Results:

• Demonstrated effectiveness of the proposed methods on both synthetic and real EEG/MEG data.
• Techniques are largely automatic, requiring minimal user tuning.
• Significant improvement in data quality, facilitating more reliable analysis.

Conclusions:

• The presented artifact removal techniques offer a less wasteful and more robust approach to EEG and MEG data preprocessing.
• These methods can substantially improve data quality, leading to more accurate scientific interpretations.
• The automated nature of these techniques makes them highly practical for routine use in neurophysiological research.