Third order spectral analysis robust to mixing artifacts for mapping cross-frequency interactions in EEG/MEG

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Estimating functional connectivity from EEG and MEG data is challenging due to signal mixing from underlying brain sources.
• Existing methods struggle with these mixing artifacts, confounding the detection of brain source interactions.
• Previous work introduced metrics based on the imaginary part of coherency to address linear mixing artifacts.

Purpose of the Study:

• To generalize existing metrics to the nonlinear domain for improved bispectral analysis of EEG/MEG data.
• To introduce a novel metric robust to mixing artifacts in functional connectivity analysis.
• To investigate cross-frequency functional brain connectivity and phase relationships between sources.

Main Methods:

• Proposed a novel metric based on an antisymmetric combination of cross-bispectra for nonlinear spectral analysis.
• Validated the method on simulated EEG data, comparing its robustness to mixing artifacts against traditional bispectral metrics.
• Applied the method to real resting-state EEG data and utilized a fit-based procedure for source-level projection.

Main Results:

• The proposed metric demonstrated reduced sensitivity to mixing artifacts compared to traditional bispectral methods.
• The new approach showed superior performance in extracting phase relationships for delayed interactions over the imaginary part of the cross-spectrum.
• Identified a significant cross-frequency interaction (10-20Hz) between alpha and beta rhythms in resting-state EEG data.

Conclusions:

• The novel bispectral analysis approach effectively overcomes mixing artifacts in EEG/MEG data.
• The method enables robust investigation of nonlinear cross-frequency brain connectivity and phase relationships.
• The study identified a specific occipito-parieto-central network interaction at 10-20Hz during resting state.