Optimal Model Parameter Estimation from EEG Power Spectrum Features Observed during General Anesthesia.

Area of Science:

• Computational neuroscience
• Mathematical biology
• Systems biology

Background:

• Mathematical modeling is crucial for understanding complex systems.
• Parameter estimation for these models, known as inverse problems, is computationally intensive and challenging.
• Accurate parameter estimation is vital for reproducing experimental data and validating models.

Purpose of the Study:

• To evaluate frequentist and Bayesian inference methods for parameter estimation in complex dynamical systems.
• To apply these methods to synthetic data and a neural mass model of EEG activity during anesthesia.
• To assess the practical identifiability of model parameters and computational efficiency.

Main Methods:

• Employed optimization algorithms based on frequentist approaches.
• Utilized Monte Carlo Markov Chain (MCMC) methods for Bayesian inference.
• Analyzed models including linear differential equations and a thalamo-cortical neural mass model fitted to EEG spectral power.

Main Results:

• The neural mass model accurately fitted EEG spectral power, particularly in delta and alpha frequency bands during anesthesia.
• Practical identifiability analysis, including confidence regions and correlation/sensitivity matrices, was performed for each case study.
• Parameter estimation using analytically computed spectral power proved accurate and computationally efficient, avoiding numerical integration.

Conclusions:

• Frequentist and Bayesian methods are effective for solving inverse problems in mathematical modeling.
• The proposed neural mass model provides a good fit to EEG data during anesthesia.
• Estimating parameters from spectral power offers an efficient and accurate approach, reducing computational burden.