Modeling and Analysis of EEG Signals and its Application to Characterize Transient Effects

Abstract

In this research a novel method for analysing the transient behaviours of EEG signals newlineis proposed. Transient behaviours found in EEG signals obtained from seizure events newlineof epileptic patients are right choice to study the EEG patterns. The choice of epileptic newlineseizure as a case, can not only test our algorithms on modelling and analysing the newlinetransient nature of EEG signals, but also help in improving the clinical diagnosis of newlinethe epilepsy. Identification of epileptic seizures of various classes can help improving newlinediagnosis of Epilepsy which is a neurological disorder affecting about 50 million newlineindividuals across the globe. It is estimated that more than 80 percent of individuals newlinewith epilepsy are not receiving treatment at any given moment around the world. The newlineexistence of such a treatment gap indicates either a failure to discover instances or a newlinefailure to offer adequate treatment newline newlineFour image database is created for training the convolutional neural networks. The newlineimages thus created served as the input of convolutional neural networks which are newlinetrained for classification of the EEG signals into multiple classes. Three convolutional newlineneural network architectures were used to train the images. They are AlexNet, newlineGoogLeNet and ResNet18. These three networks were trained for four different set of newlineimages separately. The proposed techniques are applied to seven class problem to newlinedetect seven different types of seizures and classification accuracy of up to 97.89% newlinehave been achieved using this technique