Hyperedge bundling: Data, source code, and precautions to modeling-accuracy bias to synchrony estimates.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Magnetoencephalography (MEG) and electroencephalography (EEG) are crucial for studying brain activity.
• Functional connectivity (FC) analysis using these methods is challenged by spurious interactions (SI).
• SI arise from linear mixing of source signals, creating false positives in connectivity graphs.

Purpose of the Study:

• To demonstrate the confounding effect of SI in MEG/EEG FC graphs.
• To introduce and explain the hyperedge bundling method for attenuating SI.
• To provide a simulation dataset and software for understanding SI and the proposed solution.

Main Methods:

• Utilized simulations to illustrate the generation and detection of SI.
• Developed and applied the hyperedge bundling technique to mitigate SI.
• Incorporated supplementary material detailing MEG source model properties and FC graph characteristics.

Main Results:

• SI were confirmed to be a significant confound in zero-lag-free interaction metrics.
• The accuracy of source modeling impacts interaction estimates, biasing connectivity results.
• Conventional graph thresholding methods are suboptimal due to geometric biases in the source model.

Conclusions:

• SI pose a severe challenge to the validity of MEG/EEG connectivity analysis.
• Hyperedge bundling offers a novel approach to reduce SI and improve connectivity estimation.
• Findings highlight the need for advanced methods beyond simple thresholding for interpreting brain connectivity.