MODELING BRAIN DESYNCHRONIZATION BY EEG SENSOR VARIANCE IN EPILEPTIC PATIENTS

摘要: The use of electroencephalogram (EEG) for predictive purposes seizures in epileptic patients has grown steadily with the access to greater computing power. Methods seizure analysis date have focused on modeling and computer aided machine learning help increase sensitivity specificity detection. Brain synchronization between various areas brain at onset tends be a common feature seizures, followed by resynchronization end seizures. While previous methods looked cross-correlation or lag-correlation only two brain, most EEG data these days vast array sensors that can easily exceed 15-20 brain. goal this research is take relatively new approach statistical multivariate data, variance multiple as an extended measure desynchronization time series format. Use Children’s Hospital Boston Massachusetts Institute Technology (CHB-MIT) scalp database from PhysioNet used demonstrate potential e↵ectiveness assessing overall three young intractable DCC-GARCH model, Bayesian regime switching mixture change-point model. ACKNOWLEDGMENTS I would like thank my wonderful advisor Dr. Gavino Puggioni all his patience over past years. Without insights wouldn’t been possible. Thanks also committee members Natallia Katenka Kunal Mankodiya their resourcefulness. Special thanks Walter Besio knowledge continual assistance. Behavioral Science Lisa Harlow her years support being Statistics defense chair. Last, but certainly not least, you Liliana Gonzalez accepting me into department invaluable input support.