Modeling Brain Desynchronization by EEG Sensor Variance in Epileptic Patients

• Activity Number: 374
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Craig Krebsbach* and Gavino Puggioni
• Companies: University of Rhode Island and University of Rhode Island
• Keywords: Markov Switching ; Regime Switching ; Bayesian ; Gibbs Sampling ; Seizure Prediction ; EEG

Abstract

Modern approaches to electroencephalogram (EEG) analysis of epileptic seizures have thus far helped to increase the capability of predicting seizures through multivariate sensor analysis of the brain. Various areas of the brain tend to go into a state of desynchronization preceding and during seizures, thus offering a strong rationale for using the time referenced variance between simultaneously recorded EEG channels as an interesting indicator. We propose a Bayesian Markov regime switching model with two states (non seizure and seizure) that offers simplicity and interpretability advantages over other approaches. The estimation is implemented using an MCMC algorithm. The method is illustrated with an application to the CHB-MIT scalp database from PhysioNet, consisting of measurements from several patients with clinically marked seizures.