Estimating short-run and long-run interaction mechanisms in interictal state

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Epilepsy analysis relies on understanding electroencephalogram (EEG) states.
• Distinguishing between interictal, pre-ictal, and ictal states is crucial for seizure prediction and management.
• Current time series analysis methods may not fully capture the complex dynamics of EEG during epileptic activity.

Purpose of the Study:

• To develop and apply novel time series analysis methods for EEG data from seizure patients.
• To identify statistical properties that differentiate between interictal, pre-ictal, and ictal EEG states.
• To investigate causal relationships between brain areas and epileptic states.

Main Methods:

• Statistical analysis to detect non-stationary behavior in short EEG intervals.
• Autoregressive Integrated Moving Average (ARIMA) modeling for non-stationary intervals.
• Granger-causality analysis for short-term interactions between EEG channels.
• Cointegration analysis for long-term interactions and causal links between EEG channels.

Main Results:

• Identified causal relationships between neuronal assemblies based on duration and direction of influence.
• Observed bidirectional causality in short intervals, with positive effects between neuronal ensembles.
• Found no significant effect of short-term causality on long-term interactions (increasing amplitudes).
• Demonstrated that Cointegration analysis can identify a causal link from the interictal to the ictal state.

Conclusions:

• Novel time series methods effectively analyze EEG dynamics in epilepsy.
• Causal relationships in neuronal activity vary significantly with time scale (short vs. long intervals).
• A causal link from the interictal to the ictal state was identified, offering insights into seizure progression.