Dynamical models reveal anatomically reliable attractor landscapes embedded in resting-state brain networks

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Functional connectivity (FC) in resting-state brain networks (RSNs) is crucial for cognition, but its mechanistic basis remains unclear.
• Current understanding debates whether resting-state activity reflects noise or a nonlinear dynamical system with attractors.

Purpose of the Study:

• To investigate the mechanistic underpinnings of resting-state brain activity and RSNs.
• To provide evidence for resting-state activity as a nonlinear dynamical system.

Main Methods:

• Constructed whole-brain dynamical systems models using the Mesoscale Individualized NeuroDynamic (MINDy) framework from resting-state fMRI (rfMRI) data.
• Models comprised neural masses with individualized connection weights, trained on subject-specific rfMRI recordings.
• Induce bifurcations by creating convex combinations of models to explore the full spectrum of dynamics.

Main Results:

• MINDy models exhibited diverse attractor landscapes, including multiple equilibria and limit cycles.
• These attractors mapped reliably to canonical RSNs like the default mode network (DMN) and frontoparietal control network (FPN) in anatomical space.
• Model manipulations recapitulated the full range of observed dynamics, supporting the nonlinear dynamical system hypothesis.

Conclusions:

• Resting-state brain activity is best characterized as a nonlinear dynamical system with rich attractor properties, not merely noise-driven fluctuations.
• Conventional FC analysis of rfMRI may overlook critical intrinsic brain structures and dynamics.
• Neural dynamical modeling offers a more comprehensive approach to understanding the generative mechanisms and spatiotemporal organization of brain networks.