Sequential method for fast neural population activity reconstruction in the cortex from incomplete noisy measurements

Area of Science:

• Computational neuroscience
• Biomedical engineering
• Applied mathematics

Background:

• Stochastic dynamic neural fields are increasingly used for modeling complex systems.
• Accurate state estimation is crucial for understanding neural population activity.
• Incomplete and noisy sensor data pose challenges for traditional modeling approaches.

Purpose of the Study:

• To develop a computationally efficient state estimation method for restoring neural membrane potential.
• To address challenges posed by incomplete and noisy sensor data in stochastic neural field models.
• To enhance the real-time prediction and analysis capabilities in biomedical and technical systems.

Main Methods:

• A novel state estimation technique combining Galerkin-type spectral approximation with a state-space approach.
• Utilizing the Amari equation for simulating neural population activity in a stochastic setting.
• Implementing a sequential extended Kalman filter (EKF) for incremental data processing and efficient computation.

Main Results:

• The proposed method significantly reduces computational cost compared to batch filtering by minimizing grid nodes.
• The sequential approach demonstrates increased robustness against round-off errors by eliminating matrix inversion.
• Superior performance confirmed through comparative analysis on established dynamic neural field modeling scenarios.

Conclusions:

• The developed sequential state estimation method offers a computationally efficient and robust solution for membrane potential restoration.
• This technique is well-suited for large datasets and real-time online computations in dynamic neural field modeling.
• The findings have implications for neural network training and advanced data-driven modeling in neuroscience and engineering.