Automated Classification of Focal and Non-Focal EEG Signals Based on Bivariate Empirical Mode Decomposition

Automated Classification of Focal and Non-Focal EEG Signals Based on Bivariate Empirical Mode Decomposition

Rajeev Sharma (Indian Institute of Technology Indore, India) and Ram Bilas Pachori (Indian Institute of Technology Indore, India)
Copyright: © 2018 |Pages: 21
DOI: 10.4018/978-1-5225-2829-6.ch002
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

The chapter presents a new approach of computer aided diagnosis of focal electroencephalogram (EEG) signals by applying bivariate empirical mode decomposition (BEMD). Firstly, the focal and non-focal EEG signals are decomposed using the BEMD, which results in intrinsic mode functions (IMFs) corresponding to each signal. Secondly, bivariate bandwidths namely, amplitude bandwidth, precession bandwidth, and deformation bandwidth are computed for each obtained IMF. Interquartile range (IQR) values of bivariate bandwidths of IMFs are employed as the features for classification. In order to perform classification least squares support vector machine (LS-SVM) is used. The results of the experiment suggest that the computed bivariate bandwidths are significantly useful to discriminate focal EEG signals. The resultant classification accuracy obtained using proposed methodology, applied on the Bern-Barcelona EEG database, is 84.01%. The obtained results are encouraging and the proposed methodology can be helpful for identification of epileptogenic focus.
Chapter Preview
Top

Background

Many methods are developed for assessment of the characteristic changes in EEG signals related to epileptic seizure activities. These techniques can be helpful for localization of epileptogenic zone. In (Gutiérrez et al., 2001), the electrocorticography (ECoG) recordings of the 21 patients are studied and processed with wavelet packet functions to characterize the spikes in ECoG segments. The frequency-entropy templates are computed by wavelet packet decomposition and the best basis algorithm for each electrode (Ben-Jacob et al., 2007). It is suggested that epileptogenic focus can be associated with the locations of high template similarity of ictal template. In (Panet-Raymond and Gotman, 1990, Marciani et al., 1992), delta band activities based asymmetry measures are shown as significantly reliable indicator of epileptogenic focus. The EEG recordings acquired from 23 patients are analyzed and it is found that epileptiform oscillations in high-frequency range (60-100 Hz) may be clinically important to localize seizure onset zone in patients affected by neocortical epilepsy (Worrell et al., 2004). In another study (Sabesan et al., 2009), intracranial EEG signals from four epileptic patients are analyzed using surrogate analysis of the transfer entropy to localize the epileptogenic focus. Interelectrode synchrony is studied in (Schevon et al., 2007) and it is speculated that local synchrony can indicate the epileptogenic cortex. In (Warren et al., 2010), neural synchrony measures such as linear mean phase coherence and cross-correlation of local field potentials are investigated using the intracranial EEG recordings, and it is observed that brain region affected by epilepsy is functionally isolated from the remaining brain regions in patient with partial epilepsy.

Complete Chapter List

Search this Book:
Reset