Dual tree complex wavelet transform based analysis of epileptiform discharges

Diagnosis of epileptic seizures entails visual inspection of complex seizure patterns which is a tedious task. Development of automated systems for analysing brain activity would significantly minimise the epilepsy treatment gap by providing assistance to neurophysiologists. Present research work is intended to provide insight to the epileptiform discharges during the seizures using dual tree complex wavelet transform. Algorithm is developed using publically available data from Bonn University. Statistical and nonlinear features, selected on the basis of Bhattacharyya distance, are extracted from EEG segments to demarcate the seizure and nonseizure EEG boundaries. Quadratic classification of EEG features followed by k-fold cross validation with varying train to test ratios is employed to develop a generalised robust model. Performance of classifier is accessed in terms of statistical parameters.