Parameter estimation and control for a neural mass model based on the unscented Kalman filter

Area of Science:

• Computational Neuroscience
• Nonlinear Dynamical Systems
• Biomedical Engineering

Background:

• Kalman filters are advanced algorithms for state and parameter estimation in complex systems.
• Nonlinear system dynamics, particularly in neural networks, present significant estimation challenges.
• Electroencephalography (EEG) signal analysis benefits from accurate neural system modeling.

Purpose of the Study:

• To apply nonlinear Unscented Kalman Filters (UKFs) for state and parameter estimation in neural mass models.
• To develop and evaluate an UKF-based control strategy for modulating neural system dynamics.
• To demonstrate the feasibility of UKF control for suppressing epileptiform activity.

Main Methods:

• Implementation of Unscented Kalman Filters (UKFs) for nonlinear system observation.
• Development of a control law derived from UKF-estimated states.
• Simulation of a neural mass model exhibiting distinct EEG rhythms and epileptic dynamics.

Main Results:

• Demonstrated the efficiency of UKFs in accurately estimating states and parameters of the neural mass model.
• Successfully developed and applied an UKF-based control strategy.
• Showcased the suppression of epileptiform spikes using the UKF control method.

Conclusions:

• UKFs are highly effective for state and parameter estimation in nonlinear neural systems.
• UKF-based control strategies offer a viable method for modulating neural dynamics.
• This approach holds promise for mitigating epileptic seizures by controlling neural activity.