Classification Of Eeg Signals Using Computational Intelligence Techniques To Diagnose The Epileptic Conditions

Abstract

Epilepsy is the one of the most neurological disorder in our day newlineto day life. It affects more than seventy million people throughout the newlineworld and becomes second neurological diseases after migraine. Manual newlineinspection of seizures is time consuming and laborious task. Nowadays newlineautomated techniques are evolved for detection of seizures by means of newlinesignal processing or through machine learning techniques. In this article, newlinesupervised learning algorithms are applied to the EEG dataset and newlineperformance are measured in terms of Accuracy, precision and few more. newlineMachine learning algorithm plays a vital role in classification and newlineregression problem in the past few decades. The most important reason newlinefor this is a large set of signal or data are trained and the test signals are newlineevaluated using training network. To get the better accuracy, the input newlinedata are first normalized carefully. The various normalization techniques newlineapplied in this article are Z-Score, Min-Max, Logarithmic and Square newlineRoot Normalization. For simulation purpose, Electroencephalography newline(EEG) signal from UCI machine learning repository are used. Dataset newlineconsists of 11500 patient details with 5 different cases and each signal are newlinerecorded for the duration of 23 seconds. Spider chart is used to show the newlinemetric value in detail. It is observed from the result that supervised newlinelearning algorithm yields a better result compared to logistic and KNN newline(K-Nearest Neighbor) algorithm at high iteration. And also Convolutional newlineneural network (CNN) is one of the sub-category of neural network and newlinewidely used in the various field such as weather forecasting, signal newlineprocessing and medical applications. In this article, the University of newlineCalifornia Irvine (UCI) respiratory EEG signals are used to analyse the newlinex newlineproposed hybrid CNN and results are compared to the pre-trained newlineGoogleNet Network. EEG signals are initially converted into three newlinedifferent forms such as scalogram, spectrogram and time domain images newlineand classification of images are carried out by the pre-trained Google