Classification of Normal and Epileptic Seizure EEG Signals Based on Empirical Mode Decomposition - Complex System Modelling and Control Through Intelligent Soft Computations - page 365

cation
of normal and epileptic seizure EEG signals. The area measures namely area of
analytic signal representation of IMFs and area of ellipse from SODP of IMFs have
been used as an input feature set for LS-SVM classi

In this work, we propose a method based on the EMD process for classi

er.
The rest of the chapter has been organized as follows. In Sect. 2 , proposed
methodology has been described which includes dataset, EMD method, feature
extraction and LS-SVM classi er. Feature extraction section consists of two further

subsections: one of which is analytic signal representation and area computation of
circular region and other is second-order difference plot and area computation of
elliptical region. Results of experimental analysis and comparison with other
methods have been discussed in Sect. 3 . Finally, conclusion has been provided in
Sect. 4 .

2 Methodology

2.1 Dataset

In this work, the online publicly available EEG dataset as described in Andrzejak
et al. ( 2001 ) has been used. Recordings in this dataset include EEG signals which
have been acquired for both healthy and epileptic subjects. This dataset contains

five subsets denoted as Z, O, N, F, and S, each of which having 100 single-channel
EEG signals of duration 23.6 s. The first two subsets Z and O are surface EEG
recordings of five healthy volunteers. These subsets contain EEG recordings with

eyes open and closed, respectively. The subset F have been recorded in seizure-free
intervals from

five patients in the epileptogenic zone and the subset N has been
acquired from the hippocampal formation of the opposite hemisphere of the brain.
The subset S contains seizure activity selected from all recording sites exhibiting
ictal activity. The subsets Z and O have been recorded extracranially using standard
electrode placement scheme (according to the international 10

20 system (And-
rzejak et al. 2001 ), whereas the subsets N, F, and S have been recorded intracra-
nially using depth electrodes implanted symmetrically into the hippocampal
formations. Subsets N and F have EEG signals which were taken from all contacts
of the relevant depth electrode (Andrzejak et al. 2001 ). The strip electrodes were
implanted onto the lateral and basal regions (middle and bottom) of the neocortex.
The EEG signals of the subset S contains segments taken from contacts of all
electrodes (depth and strip). Set N and F contain only activity measured during
seizure free intervals, while set S only contains seizure activity. The data were
digitized at a sampling rate of 173.61 Hz using 12-bit analog-to-digital (A/D)
converter. Bandwidth range of bandpass

-

40 Hz. More detail about
this dataset can be found in Andrzejak et al. ( 2001 ). In this study, we have used
subsets Z and S of the dataset to evaluate performance of, proposed method which
consists of EMD, feature extraction and classi

filter were 0.53 -

cation using LS-SVM classi er.