A mutual information measure of phase-amplitude coupling using gamma generalized linear models

Area of Science:

• Neuroscience
• Computational Biology
• Signal Processing

Background:

• Cross-frequency coupling (CFC) and phase-amplitude coupling (PAC) are crucial in brain function, with abnormalities linked to neurological disorders.
• PAC's role in gut-brain axis communication is an emerging area, but current detection methods have limitations in capturing statistical relationships.

Purpose of the Study:

• To develop and validate a novel, flexible parametric method for detecting phase-amplitude coupling (PAC) by modeling joint statistics of phase and amplitude.
• To establish mutual information as a canonical measure of coupling and information density as a time-resolved PAC indicator.

Main Methods:

• Utilized a gamma-distributed generalized linear model (GLM) with a Fourier basis to model conditional amplitude distributions given phase.
• Employed the minimum description length (MDL) principle for model selection and developed a goodness-of-fit (GOF) assessment.
• Leveraged mutual information to quantify coupling strength based on the joint distribution of phase and amplitude.

Main Results:

• The proposed method demonstrated superior performance over existing gold-standard techniques in detecting low-level PAC using ROC analysis on synthetic data.
• Validation on invasive EEG and simultaneous EEG-electrogastrography recordings showed comparable performance to the Modulation Index and reproduced key findings in gut-brain PAC.
• The method effectively tracked time-varying PAC in various datasets, including sleep spindles and mismatch negativity.

Conclusions:

• The new PAC measure, integrating Gamma GLMs and mutual information, offers a robust approach to analyzing the full joint distribution of amplitude and phase.
• This method surpasses existing measures in performance and shows promise for identifying time-varying PAC in electrophysiological recordings, including gut-brain data.
• The approach potentially offers greater statistical power and improved multiple comparison handling for complex electrophysiological analyses.