Global nonlinear approach for mapping parameters of neural mass models

Area of Science:

• Computational Neuroscience
• Neuroimaging Analysis
• Systems Biology

Background:

• Neural mass models (NMMs) are crucial for interpreting brain dynamics by simulating synaptic interactions.
• Accurate parameter estimation in NMMs is challenging due to model complexity and limited prior knowledge.
• Existing methods often simplify NMMs or restrict parameter exploration, hindering a full understanding of model behavior.

Purpose of the Study:

• To develop a global nonlinear approach for simultaneously recovering all parameters of NMMs from data.
• To explore the parameter space of NMMs in a minimally constrained manner.
• To improve the practical identifiability of NMMs and aid in understanding brain dynamics in health and disease.

Main Methods:

• Utilized a multi-objective evolutionary algorithm (MOEA) for global optimization of NMM parameters.
• Employed the weighted horizontal visibility graph (wHVG) to quantify multiple features of time series data.
• Applied the MOEA and wHVG to analyze resting-state EEG alpha activity in healthy individuals and patients with epilepsy.

Main Results:

• The MOEA demonstrated superior performance compared to single-objective optimization methods for NMM parameter recovery.
• Integrating wHVG improved model output constraints, leading to more restricted parameter value regions and enhanced identifiability.
• Analysis of epilepsy EEG data revealed that a reduced mean excitatory synaptic gain parameter differentiates patients from controls.

Conclusions:

• The global nonlinear approach with MOEA and wHVG offers a powerful tool for parameter recovery in complex NMMs.
• This method advances the understanding of brain dynamics and facilitates the identification of neural correlates of neurological disorders like epilepsy.
• The MOEA framework shows potential for detecting pathological brain rhythms, such as epileptiform discharges.