EPILEPTIC SEIZURE DETECTION USING ARTIFICIAL NEURAL NETWORK AND A NEW FEATURE EXTRACTION APPROACH BASED ON EQUAL WIDTH DISCRETIZATION


Orhan U., Hekim M., Ozer M.

JOURNAL OF THE FACULTY OF ENGINEERING AND ARCHITECTURE OF GAZI UNIVERSITY, vol.26, no.3, pp.575-580, 2011 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 26 Issue: 3
  • Publication Date: 2011
  • Journal Name: JOURNAL OF THE FACULTY OF ENGINEERING AND ARCHITECTURE OF GAZI UNIVERSITY
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.575-580
  • Keywords: EEG signals, equal width discretization, biomedical signal processing, epileptic seizure detection, multilayer perceptron neural network, histogram, FUZZY INFERENCE SYSTEM, EEG-SIGNALS, WAVELET TRANSFORM, AUTOMATIC RECOGNITION, CONTINUOUS-VARIABLES, ALERTNESS LEVEL, CLASSIFICATION, PREDICTION, MODEL
  • Çukurova University Affiliated: No

Abstract

In this study, we proposed a new feature extraction approach based on equal width discretization (EWD) method and used the statistical features obtained by means of this approach as the inputs of multilayer perceptron neural network (MLPNN) model in the detection of epileptic seizure from Electroencephalogram (EEG) signals. For this aim, EEG signals were discretized by EWD method, histograms of the signals were obtained according to the density of each discrete interval, and finally these histograms were used as the inputs of MLPNN models both without any hidden layer and with a hidden layer which has 5 neurons. Both of them detected epileptic seizures from EEG signals with high classification success ratios. This result showed that a linear classifier can also solve the problem of epileptic seizure detection by means of the offered feature extraction approach. Consequently, EWD approach may be used as a new feature extraction method in the biomedical signal processing.