A signal-processing pipeline for magnetoencephalography resting-state networks

Area of Science:

• Neuroscience
• Biophysics
• Signal Processing

Background:

• Functional connectivity analysis using magnetoencephalographic (MEG) data requires high-quality source-level brain activity reconstruction.
• Resting-state networks (RSNs) are crucial for understanding brain function, but their accurate identification from MEG data remains challenging.

Purpose of the Study:

• To develop and validate a processing pipeline for high-quality source-level reconstruction of brain activity from MEG data.
• To assess the performance of the proposed pipeline against alternative approaches for RSN detection.

Main Methods:

• Independent Component Analysis (ICA) was used to decompose MEG recordings into artifact and brain components.
• The pipeline involved projecting channel maps to source space, combining component time courses, and comparing different reconstruction strategies.
• Performance was evaluated using synthetic data and experimental MEG recordings, testing various ICA algorithms and artifact handling techniques.

Main Results:

• FastICA with deflation emerged as the optimal ICA decomposition algorithm.
• Subtracting artifactual components at the channel level offered lower distortion compared to recombining brain components.
• Recombining brain components after source localization minimized source-level signal distortion.
• The proposed pipeline demonstrated improved specificity in retrieving RSNs from experimental data.

Conclusions:

• The developed MEG processing pipeline enables accurate source-level reconstruction of brain activity.
• This method enhances the reliable detection and characterization of resting-state networks (RSNs).
• The findings provide a valuable tool for functional connectivity studies using MEG data.