Robust EEG/MEG Based Functional Connectivity with the Envelope of the Imaginary Coherence: Sensor Space Analysis

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Functional connectivity (FC) from EEG/MEG provides insights into cognition and brain disorders.
• Volume conduction (VC) poses a challenge in sensor-level FC analysis, causing spurious correlations.
• Existing methods like imaginary coherence (iCOH) are robust to VC but may miss certain true connections.

Purpose of the Study:

• To introduce a novel method, envelope of imaginary coherence (EIC), to improve sensor-level FC estimation.
• To compare EIC with established FC methods in simulated neural activity.
• To assess the feasibility of sensor-level FC analysis under varying noise conditions.

Main Methods:

• Developed EIC by applying an envelope to the imaginary part of coherence.
• Utilized bivariate autoregressive and stochastic neural mass models for simulations.
• Employed realistic cortical surface mapping and synthetic MEG signal generation with leadfield computation.
• Performed receiver operating curve analysis to evaluate performance across noise levels.

Main Results:

• EIC effectively approximates coherence from underlying sources, addressing limitations of iCOH.
• EIC and iCOH demonstrated superior performance, yielding the most accurate functional connectivity maps.
• The study confirmed the feasibility of sensor-level FC analysis, with performance varying by noise level.

Conclusions:

• EIC offers a valuable advancement for sensor-level functional connectivity analysis.
• Both EIC and iCOH are effective, complementing each other in different scenarios for studying brain activity.
• These methods enhance the understanding of both normal and disordered brain function.