Reducing Sensor Noise in MEG and EEG Recordings Using Oversampled Temporal Projection

Area of Science:

• Neuroscience
• Biophysics
• Signal Processing

Background:

• Sensor-specific noise in electroencephalography (EEG) and magnetoencephalography (MEG) arrays can obscure neural signals.
• Existing noise suppression methods may introduce spatial artifacts or bias, complicating source localization.

Purpose of the Study:

• To review the theory of noise suppression in EEG and MEG arrays.
• To introduce a novel method, oversampled temporal projection (OTP), for suppressing spatially uncorrelated, sensor-specific noise.
• To highlight the advantages of OTP over existing techniques.

Main Methods:

• The study reviews the theory of noise suppression in EEG and MEG.
• A novel method, oversampled temporal projection (OTP), is introduced.
• OTP utilizes a leave-one-out procedure within overlapping temporal windows to project noise.

Main Results:

• OTP suppresses sparse, channel-specific artifacts without spreading them to other channels.
• The method minimizes distortion of spatial data configurations, eliminating the need for forward model modification during source localization.
• Noise suppression benefits are maintained during source localization, and a parameter controls the trade-off between noise reduction and adaptation to changing sensor characteristics.

Conclusions:

• Oversampled temporal projection (OTP) efficiently optimizes noise suppression while controlling for spatial signal bias.
• This method is particularly valuable for applications with high sensor noise, such as analyzing high-frequency brain oscillations.
• OTP offers significant advantages for improving signal-to-noise ratio in EEG and MEG data.