Hyperedge bundling: A practical solution to spurious interactions in MEG/EEG source connectivity analyses

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Inter-areal functional connectivity (FC), particularly neuronal synchronization, is crucial for brain communication.
• Invasive electrophysiology provides evidence, but human studies rely on non-invasive magneto- and electroencephalography (MEG/EEG).
• Signal mixing (source leakage) in MEG/EEG data creates artificial (AI) and spurious interactions (SI), confounding FC analysis.

Purpose of the Study:

• To address the intractable problem of spurious interactions (SI) in source-reconstructed MEG/EEG connectivity analysis.
• To introduce and validate a novel approach for correcting SIs using hyperedge bundling.
• To enhance the accuracy of brain network localization and visualization in MEG/EEG studies.

Main Methods:

• Developed a novel approach to bundle observed functional connectivity (FC) connections into hyperedges based on signal mixing adjacency.
• Utilized realistic simulations to evaluate the performance of hyperedge bundling in separating true positives from false positives.
• Applied the hyperedge bundling method to real MEG data for visualizing large-scale cortical networks.

Main Results:

• Hyperedge bundling demonstrated good separability of true positives and minimal loss in the true positive rate in simulations.
• The method significantly reduced graph noise by minimizing the false-positive to true-positive ratio.
• Demonstrated improved visualization of large-scale cortical networks using real MEG data, highlighting the advantage of edge bundling.

Conclusions:

• Hyperedge bundling offers a robust solution for correcting spurious interactions in MEG/EEG functional connectivity analysis.
• This novel approach enhances the reliability of detecting and dissociating true cortical interactions from artifactual signals.
• The proposed hypergraph representation accurately reflects detectable and dissociable cortical interactions in MEG/EEG connectivity studies.