Seizure onset patterns in EEG and their detection using statistical measures

Diagnosis and treatment of epileptic seizures requires analysis of neuronal activity of the brain for detection of seizures at the onset. Present work for seizure onset includes extraction of various statistical and nonlinear features including a newly proposed statistical feature i.e. Modified Semi variance. Bhattacharyya distance as a separability measure for seizure and non-seizure activity gives fair justification for the new feature when compared with other traditional features. Selected features are then classified using a quadratic classifier. K-fold cross validation technique is used to validate the algorithm. Satisfactory results are obtained with average sensitivity and latency of 96.59% and 2.75 seconds respectively.