Multivariate EEG signal analysis for early prediction of epileptic seizure

Abstract

Epilepsy is the most common neurological disorder. It will be of great help for the epileptic patient if the seizure can be predicted in advance as timely prediction of the epileptic seizure reduces the risk of injury and timely medication can be given to the patient. In this paper multivariate analysis of EEG signals is proposed to predict an epileptic seizure and identify an epileptogenic region. Brain region is divided into five different zones to localize epileptogenic region. Two linear parameters, variance and complexity and one nonlinear parameter correlation is explored for prediction of an epileptic seizure. Correlation has been computed using a sliding window to predict the seizure. An increase in variance and complexity is observed during pre-ictal period. Results show that seizure can be predicted in advance and also the epileptogenic region can be localized and identified.