Functional connectivity-based attractor dynamics of the human brain in rest, task, and disease

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Neuroimaging Analysis

Background:

• Functional brain connectivity is key to understanding brain organization and its relation to behavior and disease.
• Linking brain architecture to dynamic activity is crucial for developing interpretable generative brain models.

Purpose of the Study:

• To propose and validate a novel computational framework, functional connectivity-based Attractor Neural Networks (fcANNs), for modeling macro-scale brain dynamics.
• To investigate the organizational principles of large-scale brain attractors and their relationship to brain activity.

Main Methods:

• Developed fcANNs, a theoretically inspired model simulating recurrent activity flow based on self-organization principles.
• Analyzed brain dynamics in relation to attractor states, defined as neurobiologically meaningful configurations minimizing system free energy.
• Utilized seven distinct human neuroimaging datasets for model validation.

Main Results:

• Demonstrated that large-scale brain attractors reconstructed by fcANNs exhibit an approximately orthogonal organization.
• Provided evidence for the self-orthogonalization mechanism within free-energy-minimizing attractor networks.
• Showed that fcANNs accurately reconstruct and predict brain dynamics across various conditions (resting, task, disorders).

Conclusions:

• fcANNs establish a formal link between brain connectivity and dynamic activity.
• The orthogonal organization of brain attractors supports the underlying theoretical framework.
• fcANNs offer a computationally simple and interpretable alternative to conventional descriptive analyses for understanding brain function.