A data-driven framework for neural field modeling

Area of Science:

• Computational Neuroscience
• Electrophysiology
• Mathematical Modeling

Background:

• Neural field models describe large-scale brain activity.
• Estimating parameters from electrophysiological data is challenging.

Purpose of the Study:

• To develop a framework for creating parametric neural field models from electrophysiological data.
• To enable the estimation of neural field states and parameters.

Main Methods:

• Utilized Wilson and Cowan or Amari style neural field equations.
• Employed basis function decomposition for model reduction to a finite-dimensional state-space model.
• Implemented a two-stage iterative algorithm with unscented Rauch-Tung-Striebel smoother and least squares for estimation.

Main Results:

• Demonstrated the theoretical possibility of reconstructing neural fields from data.
• Successfully estimated intracortical connectivity structure and synaptic dynamics.
• Validated the framework using simulated data with modeled sensors.

Conclusions:

• The proposed framework offers a viable method for analyzing electrophysiological data.
• This approach advances the understanding of neural dynamics and connectivity.
• Facilitates the study of brain function through data-driven modeling.