Physiologically Guided Modeling for EEG Multichannel Signals

Area of Science:

• Neuroscience
• Biophysics
• Signal Processing

Background:

• Electroencephalographic (EEG) recordings show strong interchannel correlations due to scalp potentials from neural sources.
• Current data-driven methods like independent component analysis offer limited interpretability and physiological assumption testing.
• Neural sources are often modeled as current dipoles, but their orientation is not always fixed.

Purpose of the Study:

• To develop a physiologically guided multichannel model for EEG analysis.
• To improve the interpretability and physiological validity of EEG source modeling.
• To enable statistical testing of the fixed dipole-orientation assumption in EEG.

Main Methods:

• A parametric Frequency-Modulated Möbius (FMM) formulation was used to model EEG signals.
• Each electrode signal was represented as a linear combination of latent dipole-related sources.
• A reduced-rank formulation enabled likelihood-based testing of the fixed dipole-orientation assumption.

Main Results:

• The proposed FMM model accurately captures temporal and spatial structure in EEG data.
• The model explains interchannel dependence using a small number of latent sources.
• Statistical evidence was found supporting the plausibility of fixed dipole orientation.

Conclusions:

• The physiologically guided FMM model offers a more interpretable and physiologically grounded approach to EEG analysis.
• This framework effectively models interchannel correlations and allows for testing of dipole orientation.
• The method provides a valuable tool for analyzing EEG data with enhanced physiological insight.