Collective almost synchronization-based model to extract and predict features of EEG signals

Area of Science:

• Neuroscience
• Computational modeling
• Complex systems

Background:

• Understanding brain function is crucial across science, medicine, and engineering.
• Computational models offer a promising avenue for brain research.
• Electroencephalography (EEG) is a key neuroscience tool for recording brain activity.

Purpose of the Study:

• To develop and validate a computational model for reproducing and predicting EEG signals.
• To assess the model's performance on both healthy and epileptic EEG data.
• To evaluate the model's ability to forecast future EEG characteristics.

Main Methods:

• A computational model based on complex networks of weakly connected dynamical systems (Hindmarsh-Rose neurons or Kuramoto oscillators).
• The model was configured to operate in a Collective Almost Synchronization (CAS) dynamic regime.
• Model performance was evaluated by reproducing EEG data and predicting features like the Hurst exponent and power spectrum.

Main Results:

• The proposed model successfully reproduced both healthy and epileptic EEG signals.
• The model accurately predicted EEG features, including the Hurst exponent and power spectrum.
• The model demonstrated predictive capability for EEG signals up to 5.76 seconds in the future with an average error of 9.22%.

Conclusions:

• The developed complex network model effectively simulates and forecasts EEG signals.
• The model shows potential for advancing the analysis and prediction of neurological conditions like epilepsy.
• The study highlights the utility of computational neuroscience models in interpreting complex brain activity.