Slow-Fast Duffing Neural Mass Model.

Area of Science:

• Computational Neuroscience
• Neurophysics
• Biophysics

Background:

• Epileptic seizures can be triggered by random neuronal activity or alterations in ion channel dynamics.
• Existing neural mass models require enhancements to accurately capture seizure dynamics.

Purpose of the Study:

• To extend the Jansen and Rit neural mass model (JRNMM) for simulating paroxysmal transitions observed in intracranial electroencephalogram (iEEG) recordings.
• To investigate seizure etiologies linked to stochastic fluctuations and pathological ion current dynamics.

Main Methods:

• Introduction of the Duffing neural mass model (DNMM) to simulate seizure generators via perturbations.
• Development of a slow-fast DNMM incorporating slow dynamics of internal parameters to model ion current evolution.
• Utilizing a continuous-discrete unscented Kalman filter (CD-UKF) and genetic algorithm (GA) for parameter estimation and optimization of a log-likelihood function (LLF).

Main Results:

• The slow-fast DNMM successfully replicates transitions to and from seizure states.
• Simulations demonstrate that seizure origins can be attributed to random input fluctuations or pathological slow state evolution.
• The CD-UKF and GA effectively captured iEEG data dynamics with paroxysmal transitions.

Conclusions:

• Extended neural mass models can effectively simulate seizure dynamics and their underlying mechanisms.
• The developed slow-fast DNMM provides a framework for understanding seizure initiation.
• Advanced filtering and optimization techniques are crucial for analyzing complex iEEG data.