Interdiscip Sci Comput Life Sci (2012) 4: 209–214
DOI: 10.1007/s12539-012-0129-6
Epilepsy Diagnosis Based on Generalized Feed
Forward Neural Network
P.A. KHARAT
1∗
, S.V. DUDUL
2
1
(Department of Information Techanology, Anuradha Engineering College, Chikhli 443201, India)
2
(Department of Applied Electronics, S.G.B. Amaravati University, Amaravati 444602, India)
Received 22 November 2011 / Revised 11 March 2012 / Accepted 19 March 2012
Abstract: Epilepsy is a common neurological disorder that is characterized by recurrent unprovoked seizures.
Epilepsy can develop in any person at any age. 0.5% to 2% of people will develop epilepsy during their lifetime.
This paper aims to develop the clinical decision support system (DSS) for the diagnosis of epilepsy. In this paper
a simple, reliable and economical Neural Network (NN) based DSS was proposed for the diagnosis of epilepsy. The
generalized feed forward neural network (GFFNN) was designed for the diagnosis. Eleven statistical parameters
along with the 64 FFT were extracted for the electroencephalogram (EEG) signal. Data used for the experimen-
tation purpose wa s obtained from the University of Bonn. The classification rate of GFFNN was 100 % for the
training data and 86.67% for the cross validation data.
Key words: generalized feed forward neural networks (GFFNN), decision support system (DSS), electroen-
cephalogram (EEG), epilepsy, approximate entropy (ApEn).
1 Introduction
Epilepsy is a brain disorder in which clusters of nerve
cells, or neurons in the brain sometimes have abnor-
mal signal. In the epilepsy, the normal pattern of neu-
rons activity becomes disturbed causing strange sensa-
tion, emotion, behavior and loss of consciousness (En-
gel, 1989; Robert, 2005). Epilepsy is a disorder due
to many possible causes. Anything that disturbs the
normal pattern of neuron activity may result in ill-
ness, brain damage, or abnormal brain development.
EEG scan is a common diagnostic test for epilepsy and
can detect abnormalities in the brain electrical activ-
ity. People with epilepsy frequently have changes in
their normal pattern of brain wave, even though they
are not experiencing a seizure. EEG plays an important
role in the diagnosis of epilepsy.
Researchers proposed many automatic systems for
the diagnosis of epilepsy. Akin et al. (2001) devel-
oped a classification method for the diagnosis. Another
work (Szilagyi et al., 2001) recommended the recogni-
tion of epileptic waveform by using the multi-resolution
wavelet decomposition of EEG signal. Shrinivasan de-
signed the approximate entropy based Elman neural
network and probabilistic neural network for detection
of epilepsy (Shrinivasan, 2007). The method proposed
∗
Corresp onding author.
E-mail: pra vinakharat82@gmail.com
by Sriraam et al. (Shrinivasan, 2004) uses recurrent
neural network classifier with wavelet entropy and spec-
tral entropy features as the input for the automated
detection of epilepsy.
This paper discusses the automated DSS for the
epilepsy diagnosis using the GFFNN. Eleven statistical
and 64 FFT were input to the GFFNN. Fig. 1 shows
the block diagram of proposed DSS. As compare to the
existing system the proposed DSS is simple and eco-
nomical.
EEG data
acquisition
and processing
Feature
extraction
using FFT
GFNN
based DSS
Diagnosis
decision
Fig. 1 Block diagram of proposed DSS.
2 Benchmark EEG data set
The EEG data considered for this work was extracted
from EEG database of University of Bonn which is
available in public domain (Ralph, 2001). The com-
plete database was comprised of five sets of dataset re-
ferred as A-E. Each dataset contained 100 single chan-
nel EEG segment without any artifacts with 23.6 sec.
Sets A and B contained recording obtained from sur-
face EEG recording that was carried out on five healthy
volunteers using a standardized 10-20 electrode place-