Multi-modal and multi-model interrogation of large-scale functional brain networks.

Area of Science:

• Computational neuroscience
• Systems neuroscience
• Neuroimaging

Background:

• Existing whole-brain models are often modality-specific, limiting cross-modal integration.
• Neural activity measured by fMRI and MEG/EEG, despite differences, may arise from shared network dynamics.
• Understanding these shared dynamics is key to unifying brain modeling approaches.

Purpose of the Study:

• To develop and test large-scale models capable of jointly predicting features from distinct brain signal modalities (fMRI, MEG/EEG).
• To investigate if universal principles of self-organizing delay-coupled nonlinear systems can link macroscopic structural connectomes to observed neural activity.
• To assess the performance of the Stuart Landau (SL) and Wilson-Cowan (WC) models in capturing functional connectivity and metastable oscillatory modes (MOMs) across modalities.

Main Methods:

• Utilized Stuart Landau and Wilson-Cowan models to simulate 40 Hz oscillations.
• Measured functional connectivity (FC), functional connectivity dynamics (FCD), and metastable oscillatory modes (MOMs) in empirical fMRI and MEG data.
• Compared simulated data features against empirical measurements, systematically adjusting global coupling, mean conduction time delay, and excitation-inhibition balance (for WC model).

Main Results:

• Both SL and WC models, with optimized parameters (especially delays), could represent MEG FC, FCD, and generate MOMs comparably.
• Omitting delays significantly degraded model performance for both modalities.
• The SL model showed poorer performance for fMRI FCD and MOMs, suggesting the importance of balanced dynamics for ultra-slow activity patterns.
• Optimal model parameters varied across fMRI and MEG modalities.
• Neither model achieved high cross-modal correlation (>0.4) for empirical FC with identical parameters.
• Both models generated FC patterns extending beyond anatomical constraints and empirical-like MOMs (size, duration).

Conclusions:

• Static and dynamic properties of neural activity across timescales emerge from delay-coupled oscillator networks at 40 Hz.
• Mesoscale heterogeneities in neural circuitry are likely critical for generating parallel, cross-modal functional networks.
• Future whole-brain modeling should incorporate these mesoscale details for improved cross-modal predictions.