Identifying spatio-temporal seizure propagation patterns in epilepsy using Bayesian inference.

Area of Science:

• Neurology
• Computational Neuroscience
• Biomedical Engineering

Background:

• Focal drug-resistant epilepsy necessitates surgical intervention to remove the epileptogenic zone.
• Accurate identification of the epileptogenic zone is crucial for successful epilepsy surgery.
• Current methods using stereotactic EEG recordings face challenges due to spatial sparsity.

Purpose of the Study:

• To develop and validate a probabilistic hierarchical model for identifying the epileptogenic zone in focal drug-resistant epilepsy.
• To assess the model's accuracy in predicting surgical outcomes.
• To provide a tool to assist clinicians in pre-surgical planning.

Main Methods:

• Developed a probabilistic hierarchical model based on the Epileptor phenomenological model of seizure dynamics.
• Utilized Bayesian inference to optimize patient-specific virtual models using intracranial EEG data.
• Validated the model using synthetic data and a retrospective cohort of 25 epilepsy patients.

Main Results:

• The model demonstrated accurate predictions of surgical outcomes in a retrospective patient cohort.
• Model predictions aligned well with clinical hypotheses in patients who became seizure-free post-surgery.
• A significant mismatch was observed between model predictions and outcomes in patients with failed surgeries.

Conclusions:

• The proposed probabilistic model shows promise as a valuable tool for identifying the seizure focus in drug-resistant epilepsy.
• The model's ability to predict surgical success aids in pre-operative decision-making.
• This approach can potentially improve surgical planning and patient outcomes in epilepsy treatment.