Proposing a two-level stochastic model for epileptic seizure genesis

Area of Science:

• Computational Neuroscience
• Epilepsy Research
• Biomedical Signal Processing

Background:

• The transition from normal brain function to epileptic states is complex and debated.
• Existing models often simplify the underlying physiological dynamics of seizure genesis.
• Understanding seizure onset mechanisms is crucial for developing effective prediction and treatment strategies.

Purpose of the Study:

• To propose a novel, unified stochastic model for spontaneous seizure generation.
• To model key physiological parameters, specifically excitatory and inhibitory synaptic gains, influencing electroencephalogram (EEG) activity.
• To provide a robust tool for validating and comparing seizure prediction algorithms.

Main Methods:

• A two-level spontaneous seizure generation model was developed.
• The first level employs a hidden Markov process to model physiological parameters, with transition matrices derived from real seizure onset data.
• The second level integrates a depth-EEG model using excitatory and inhibitory synaptic gains as adjustable parameters.
• Parameter identification algorithms were used to estimate real parameter sequences from depth-EEG signals.

Main Results:

• The proposed stochastic model successfully simulates seizure genesis consistent with various theoretical scenarios.
• Short-term and long-term validations confirmed the model's efficacy.
• Synthetic depth-EEG signals generated by the model demonstrated realistic characteristics.

Conclusions:

• The developed two-level stochastic model offers a comprehensive approach to understanding seizure genesis.
• The model provides a valuable platform for the comparative analysis of diverse seizure prediction algorithms.
• This work contributes to advancing computational models in epilepsy research and EEG signal analysis.