288

IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 11, NO. 3, MAY 2007

Approximate Entropy-Based Epileptic EEG

Detection Using Artificial Neural Networks
Vairavan Srinivasan, Member, IEEE, Chikkannan Eswaran, Senior Member, IEEE,
and Natarajan Sriraam, Member, IEEE

AbstractThe electroencephalogram (EEG) signal plays an important role in the diagnosis of epilepsy. The EEG recordings of
the ambulatory recording systems generate very lengthy data and
the detection of the epileptic activity requires a time-consuming
analysis of the entire length of the EEG data by an expert. The
traditional methods of analysis being tedious, many automated diagnostic systems for epilepsy have emerged in recent years. This
paper proposes a neural-network-based automated epileptic EEG
detection system that uses approximate entropy (ApEn) as the input feature. ApEn is a statistical parameter that measures the predictability of the current amplitude values of a physiological signal
based on its previous amplitude values. It is known that the value
of the ApEn drops sharply during an epileptic seizure and this
fact is used in the proposed system. Two different types of neural
networks, namely, Elman and probabilistic neural networks, are
considered in this paper. ApEn is used for the first time in the proposed system for the detection of epilepsy using neural networks.
It is shown that the overall accuracy values as high as 100% can
be achieved by using the proposed system.
Index TermsApproximate entropy (ApEn), artificial neural
network (ANN), electroencephalogram (EEG), Elman network
(EN), epilepsy, probabilistic neural network (PNN), seizure.

I. INTRODUCTION
PPROXIMATELY 1% of the people in the world suffer
from epilepsy. The electroencephalogram (EEG) signal is
used for the purpose of the epileptic detection as it is a condition
related to the brains electrical activity [1]. Epilepsy is characterized by the occurrence of recurrent seizures in the EEG signal.
In majority of the cases, the onset of the seizures cannot be
predicted in a short period, a continuous recording of the EEG
is required to detect epilepsy. A common form of recording
used for this purpose is an ambulatory recording that contains
EEG data for a very long duration of even up to one week. It
involves an experts efforts in analyzing the entire length of the
EEG recordings to detect traces of epilepsy. As the traditional
methods of analysis are tedious and time-consuming, many automated epileptic EEG detection systems have been developed
in recent years [1].

Manuscript received October 20, 2005; revised January 24, 2006 and
April 11, 2006.
V. Srinivasan is with the Institute of Advanced Biomedical Techniques (ITAB), G. D. Annunzio University, 66100 Chieti, Italy (e-mail:
v.srinivasan@ieee. org).
C. Eswaran is with the Faculty of Information Technology and the Center
of Multimedia Computing, Multimedia University, 63100 Cyberjaya, Selangor,
Malaysia (e-mail: eswaran@mmu.edu.my).
N. Sriraam is with the Faculty of Information Technology, Multimedia University, 63100 Cyberjaya, Selangor, Malaysia (e-mail: eswaran@mmu.edu.my;
natarajan.sriraam@mmu.edu.my).
Digital Object Identifier 10.1109/TITB.2006.884369

With the advent of technology, it is possible to store and

process the EEG data digitally. The digital EEG data can be fed
to an automated seizure detection system in order to detect the
seizures present in the EEG data. Hence, the neurologist can
treat more patients in a given time as the time taken to review
the EEG data is reduced considerably due to automation.
Automated diagnostic systems for epilepsy have been developed using different approaches. In 1982, Gotman [2] presented
a computerized system for detecting a variety of seizures. In
1991, Murro et al. [3] developed an automated seizure detection
system based on the discriminant analysis of the EEG signal
recorded from the intracranial electrodes. In 1997, Qu and
Gotman [4] proposed the use of the nearest-neighbor classifier
on EEG features extracted in both the time and frequency
domains to detect the onset of the epileptic seizures. In 2004,
Gigola et al. [5] used a method based on the evolution of the
accumulated energy using wavelet analysis for the prediction
of the epileptic seizure onset from the intracranial epileptic
EEG recordings.
Artificial neural network (ANN) based detection systems for
epileptic diagnosis have been proposed by several researchers.
The method proposed by Weng and Khorasani [6] uses the
features proposed by Gotman and Wang [7], namely, average
EEG amplitude, average EEG duration, coefficient of variation,
dominant frequency, and average power spectrum as inputs to
an adaptive structured neural network. The method proposed
by Pradhan et al. [8] uses a raw EEG signal as an input to a
learning vector quantization (LVQ) network. In 2004, Nigam
and Graupe [9] proposed a new neural network model called
LAMSTAR network, and two time-domain attributes of EEG,
namely, relative spike amplitude and spike rhythmicity have
been used as inputs for the purpose of the detection of epilepsy.
The method proposed by Kiymik et al. [10] uses a back propagation neural network (BPNN) with periodogram and autoregressive (AR) features as the input for the automated detection of
epilepsy.
This paper discusses an automated epileptic EEG detection
system using two different neural networks, namely, Elman
network (EN) and probabilistic neural network (PNN) using
a time-domain feature of the EEG signal called approximate
entropy (ApEn) that reflects the nonlinear dynamics of the
brain activity [11].
ApEn is a recently formulated statistical parameter to quantify the regularity of a time series data of physiological signals [12]. It was first proposed by Pincus in 1991 [13] and has
been predominantly used in the analysis of the heart rate variability [14][17] and endocrine hormone release pulsatility [18],

1089-7771/$25.00 2007 IEEE

SRINIVASAN et al.: APPROXIMATE ENTROPY-BASED EPILEPTIC EEG DETECTION USING ARTIFICIAL NEURAL NETWORKS

Fig. 1.

Block diagram of the proposed system.

Fig. 2.

Intracranial electrode placements.

289

estimation of regularity in epileptic seizure time series data [19],

and in the estimation of the depth of anesthesia [20]. Diambra
et al. [21] have shown that the value of the ApEn drops abruptly
due to the synchronous discharge of large groups of neurons
during an epileptic activity. Hence, it is a good feature to make
use of in the automated detection of epilepsy. In this paper, this
feature is applied, for the first time, in the automated detection
of epilepsy using neural networks.
II. MATERIALS AND METHOD
Fig. 1 shows the block diagram of the proposed neural network based automated epileptic detection system.
A. EEG Data Acquisition
Two sets of EEG data [22] corresponding to the normal and
epileptic subjects are used as the experimental data set for the
proposed neural network based detection system. Each data set
contains 100 single-channel EEG segments, with a segment duration of 23 s. These segments are selected and cut out from the
continuous multichannel EEG recordings after visual inspection
for artifacts, e.g., due to muscle activity or eye movements. The
first set of the EEG data corresponding to the normal subjects is
taken from the surface EEG recordings of five healthy subjects
using the standardized electrode placement technique. The subjects are relaxed in an awaken state with eyes open. The second
set of the EEG data consists of epileptic EEG signals obtained
from five different epileptic patients, recorded during the occurrence of the epileptic seizures from intracranial electrodes.
The schematic of the intracranial electrode placements is shown
in Fig. 2 [22]. The depth electrodes are implanted symmetrically into the hippocampal formations and strip electrodes are
implanted onto the lateral and basal regions (middle and bottom) of the neocortex. The epileptic EEG segments are selected
from all the recording sites exhibiting ictal activity [22]. The
intracranial epileptic EEG has been chosen for this classification system as the intracranial recordings offer the most precise
access to the emergence of seizure [5]. It has also been recently
reported that changes in the scalp electrical activity are similar
to those detected from the intracranial recordings [23]. Hence,

Fig. 3.

Specimens of the EEG signal. (a) Normal EEG. (b) Epileptic EEG.

analysis of the epileptic EEGs from depth electrodes gives important information about the onset of the seizures that could
be applied to scalp records. The EEG signals are recorded with
128-channel amplifier system, using an average common reference. After a 12-bit analog-to-digital conversion, the data are
written continuously onto the disk of a data acquisition computer system at a sampling rate of 173.61 Hz with bandpass
filter settings at 0.5340 Hz (12 dB/octave) [22]. Fig. 3(a) and
(b) shows specimens of the normal and epileptic EEG signals,
respectively.
B. Feature Extraction
The proposed system makes use of a single feature called
ApEn for the epileptic detection. The ApEn is a time-domain
feature that is capable of classifying complex systems [12]. The
value of the ApEn is determined as shown in the following
steps [11], [16].

290

IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 11, NO. 3, MAY 2007

1) Let the data sequence containing N data points be X =

[x(1), x(2), x(3), . . . , x(N )].
2) Let x(i) be a subsequence of X such that x(i) =
[x(i), x(i + 1), x(i + 2), . . . , x(i + m 1)] for 1 i
N m, where m represents the number of samples used
for the prediction.
3) Let r represent the noise filter level that is defined as
r = k SD for k = 0, 0.1, 0.2, 0.3, . . . , 0.9

(1)

where SD is the standard deviation of the data sequence X.

4) Let {x(j)} represent a set of subsequences obtained from
x(j) by varying j from 1 to N . Each sequence x(j) in the
set of {x(j)} is compared with x(i) and, in this process,
two parameters, namely, Cim (r) and Cim+1 (r) are defined
as follows:

N m
j=1 kj
(2)
Cim (r) =
N m
where
k=

1,
0,

if |x(i) x(j)| for 1 j N m

otherwise

and

N m
Cim+1 (r)

kj
N m
j=1

(3)

with conditions depicted by (A) as shown at the bottom of

the page.
5) We define m (r) and m+1 (r) as follows:

N m
ln(Cim (r))
(4)
m (r) = i=1
N m

N m
ln(Cim+1 (r))
m+1 (r) = i=1
.
(5)
N m

Fig. 4.

Flowchart for ApEn(m, r, N ).

6) ApEn(m, r, N ) is calculated using m (r) and m+1 (r)

as follows:
ApEn(m, r, N ) = m (r) m+1 (r)

N m

N m
m
ln(Cim+1 (r))
i=1 ln(Ci (r))
i=1
=
N m
N m
N m

N
m


1
m+1
m
=
ln(Ci (r))
ln(Ci
(r))
N m i=1
i=1
N m 


Cim (r)
1
=
ln
.
N m i=1
Cim+1 (r)

C. Neural Network Classifier

(6)
(7)
(8)

(9)

The flowchart for the calculation of ApEn(m, r, N ) is shown in

Fig. 4. Small values of ApEn imply strong regularity in a data
sequence and large values imply substantial fluctuations [17].

k=

1,
0,

if |x(i) x(j)| r
otherwise

and

ANNs are considered to be good classifiers due to their

inherent features such as adaptive learning, robustness, selforganization, and generalization capability. ANNs are particularly useful in situations where enough data are available for
training and where the simpler classification algorithms fail. The
results obtained for the epileptic EEG detection using several
types of ANNs have been reported [6], [8][10]. However, to the
best of our knowledge, the performances of the EN and PNNs
for the epileptic EEG detection have not been investigated so
far. Two different types of ANNs, namely, EN and PNNs are
employed in this paper for the detection of epilepsy. A brief description of the configuration, target, and threshold values used
for the two neural networks are given as follows.

|x(i + 1) x(j + 1)| r for 1 j N m

(A)

SRINIVASAN et al.: APPROXIMATE ENTROPY-BASED EPILEPTIC EEG DETECTION USING ARTIFICIAL NEURAL NETWORKS

291

TABLE I
EN TRAINING PARAMETERS

Fig. 5.

Architecture of a recurrent neural network.

1) EN: It is a special type of recurrent neural network. It is

a two-layered backpropagation network with a feedback
connection from the output of the hidden layer to its input.
This feedback connection allows EN to recognize and
generate temporal patterns, as well as spatial patterns [24].
Fig. 5 illustrates the architecture of a recurrent neural
network. The ApEn values corresponding to the normal
and epileptic EEG signals are used as the inputs for the
neural network. For the two-layered EN, the activation
functions used are tan-sigmoidal and log-sigmoidal for
the hidden and output layers, respectively. The number of
neurons used in the hidden layer and the output layer are
90 and 1, respectively. Gradient descent algorithm with
an adaptive learning rate is used for training the EN [24].
The network is trained with a target value of 0 for normal
EEG and 1 for epileptic EEG. The range of the output
values used for the classification are 00.3 for the normal
EEG and 0.71 for the epileptic EEG. The neural network
training parameters are as shown in Table I.
2) PNN: It is a type of radial basis network. It is a feedforward
neural network with two middle layers called radial basis
and competitive layers [24], [25]. The two layers employ
radial basis and compete activation functions, respectively.
Fig. 6 shows the architecture of a PNN. The ApEn values
corresponding to the normal and epileptic EEG signals are
used as the input for the neural network. The networks
target values correspond to a value of 1 for normal EEG
and 2 for epileptic EEG.

Fig. 6.

Architecture of a PNN.

D. Performance Evaluation Parameters

The performances of EN and PNN are evaluated by using
three parameters, namely, sensitivity (SE), specificity (SP), and
overall accuracy (OA), which are defined in (10), (11), and (12),
respectively [1]
SE(%) =

TNCP
100
TNAP

(10)

where TNCP represents the total number of correctly detected

positive patterns and TNAP represents the total number of actual positive patterns. A positive pattern indicates a detected
seizure.
TNCN
SP(%) =
100
(11)
TNAN
where TNCN represents the total number of correctly detected
negative patterns and TNAN represents the total number of actual negative patterns. A negative pattern indicates a detected
nonseizure.
TNCDP
OA(%) =
100
(12)
TNAPP
where TNCDP represents the total number of correctly detected
patterns and TNAPP represents the total number of applied
patterns. A pattern indicates both seizure and nonseizure.

292

IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 11, NO. 3, MAY 2007

TABLE II
APEN VALUES USED FOR TRAINING AND TESTING THE NEURAL NETWORKS

Fig. 8. Sample plots of ApEn showing partial overlap between normal and
epileptic EEG signals.

Fig. 7. Sample plots of ApEn showing clear discrimination between the normal and epileptic EEG signals.

III. RESULT
ApEn values are computed for selected combinations of m, r,
and N . The values of m, r, and N that are used for the experiments are as follows:

1) m = 1, 2, 3;
2) r = 0%90% of SD of the data sequence in increments of
10%;
3) N = 173, 256, 512, 1024, 2048.
ApEn values are computed for both normal and epileptic
EEG signals and are fed as inputs to the two neural networks.
Among the available 100 EEG data sets, 60 data sets are used for
training and the remaining are used for testing the performance
of the neural networks. This choice is made arbitrarily keeping in
mind that enough datasets are provided for the neural network
to understand the inherent structure of the data so that it can
classify the unknown datasets properly. Thus, the total number
of data points used for training and testing are 238 740 and 159
160, respectively. Using these data points, frames of different
sizes (173, 256, 512, 1024, and 2048) are formed. The ApEn
values are calculated for each data frame and the number of
ApEn values used for training and testing the neural networks
are shown in Table II.

SRINIVASAN et al.: APPROXIMATE ENTROPY-BASED EPILEPTIC EEG DETECTION USING ARTIFICIAL NEURAL NETWORKS

Fig. 10.
Fig. 9.

293

Overall accuracy obtained using ANNs with m = 2.

Overall accuracy obtained using ANNs with m = 1.

The potentiality of the ApEn to discriminate the two signals, namely, normal and epileptic EEG signals depends on the
values of m, r, and N . Fig. 7 shows the sample plots of the
ApEn that have clear discriminations between the normal and
epileptic EEG signals. The ApEn values shown in Fig. 7(a) and
(b) are calculated with N = 256 and 1024, respectively. Fig. 8
shows the sample plots of the ApEn corresponding to N = 512
and 1024 that have a partial overlapping between normal and
epileptic EEG signals. Figs. 911 show the overall detection accuracies (%) obtained by the two neural networks using ApEn
as the input feature.
IV. DISCUSSION
It can be observed from Figs. 911 that high overall accuracy
values in the range 99.35%100% are achieved by using EN for
some specific combinations of m, r, and N (e.g., r = 0 SD for
all the values of m and N ; m = 2, r = 0.2 SD, N = 1024 and

m = 3, r = 0.2 SD, N = 1024). The specific combinations

of m, r, and N that yield high-accuracy values substantiate the
conclusions reported in [12][15]. The overall accuracy values
obtained by EN for other combinations of the values of m, r, and
N range from 95.45% to 100%, which are also in the acceptable
range of the clinical tests [8].
PNN gives good overall accuracy values in the range 98%
100%, only for a few combinations of m, r, and N (e.g., m =
1, r = 0 SD for all the values of N and m = 3, r = 0.2 SD,
N = 1024).
It can be concluded from the results that, in general, EN performs better than PNN for almost for all the combinations of
m, r, and N . The feedback configuration and the incremental
training process associated with the EN may be considered as
possible reasons for its superior performance. The overall detection accuracy values obtained by the proposed method are also
better than the value of 99.6% obtained with EN using other
input features [26].

294

IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 11, NO. 3, MAY 2007

Our experimental results are based on data sets corresponding

to five different subjects only. The optimum ApEn parameter
values obtained based on this data may not hold good for a
general case. Hence, using a linear separator with known ApEn
parameter values may not give good results in situations where
a large number of different subjects are involved. This problem
will not arise in the proposed ANN-based method as it has
performed well irrespective of the ApEn parameter values used.
It is known [21] that ApEn possesses good characteristics
such as robustness in the characterization of the epileptic patterns and low computational burden. Hence, an automated system using ApEn as the input feature is best suited for the realtime detection of the epileptic seizures. The proposed system is
based on two types of EEG, namely, EEG signals of awake and
epileptic subjects. It can be made more robust by acclimatizing
it to the other manifestations of EEG like sleep EEG.
V. CONCLUSION
In this paper, two different types of neural networks, namely,
ENs and PNNs, have been employed for the automated detection
of epilepsy. A robust and computationally low-intensive feature,
namely, ApEn has been used for the proposed epileptic detection
system. Experimental results show that overall accuracies as
high as 100% can be achieved by this system. As the proposed
system is based on a single feature that has a low computational
burden, it is best suited for the real-time detection of epileptic
seizures from ambulatory recordings.
ACKNOWLEDGMENT

Fig. 11.

Overall accuracy obtained using ANNs with m = 3.

Though the use of ANNs increases the computational

complexity, the high overall detection accuracies achieved with
this system surpasses its disadvantage as in any automated
seizure detection system, the detection of the seizure with
high accuracy is of primary importance. It may be noted that
using a simple threshold system would prove to be inefficient,
as ApEn shows clear discrimination between the normal and
epileptic EEG signals (e.g., Fig. 7) only for some specific
combinations of the three parameters, namely, m, r, and N . On
the other hand, the ANNs perform well for almost for all the
combinations of m, r, and N . For instance, a normal threshold
system would be inefficient for the ApEn values, as shown
in Fig. 8, since one could clearly see the partial overlapping
between the normal and epileptic EEG signals. On the contrary,
the overall accuracy obtained in this case by EN is 100% for
both the cases (N = 512 and N = 1024) while the overall
accuracy of PNN is 84.52% and 77.42% for N = 512 and
1024, respectively. These results support the use of ANNs.

The authors would like to thank Dr. J. Bruhn of the

Department of Anesthesiology and Intensive Care Medicine,
University of Bonn, Germany, Dr. R. G. Andrzejak of the Computational Neuroscience Group, Departament de Tecnologia,
Universitat Pompeu Fabra, Barcelona, Spain, and Dr. P. G. Larsson of the National Centre for Epilepsy, Norway, for the useful
discussions on approximate entropy and epilepsy.
REFERENCES
[1] N. Mc Grogan (1999). Neural network detection of epileptic seizures
in the electroencephalogram, [Online]. Available: http://www.new.
ox.ac.uk/nmcgroga/work/transfer
[2] J. Gotman, Automatic recognition of epileptic seizures in the
EEG, Electroencephalogr. Clin. Neurophysiol., vol. 54, pp. 530540,
1982.
[3] A. M. Murro, D. W. King, J. R. Smith, B. B. Gallagher, H. F. Flanigin, and
K. Meador, Computerized seizure detection of complex partial seizures,
Electroencephalogr. Clin. Neurophysiol., vol. 79, pp. 330333, 1991.
[4] H. Qu and J. Gotman, A patient-specific algorithm for the detection of
seizure onset in long-term EEG monitoring: Possible use as a warning
device, IEEE Trans. Biomed. Eng., vol. 44, no. 2, pp. 115122, Feb.
1997.
[5] S. Gigola, F. Ortiz, C. E. D Attellis, W. Silva, and S. Kochen, Prediction of epileptic seizures using accumulated energy in a multiresolution
framework, J. Neurosci. Methods, vol. 138, pp. 107111, 2004.
[6] W. Weng and K. Khorasani, An adaptive structure neural network with
application to EEG automatic seizure detection, Neural Netw., vol. 9,
pp. 12231240, 1996.
[7] J. Gotman and L. Wang, State-dependent spike detection: Concepts and
preliminary results, Electroencephalogr. Clin. Neurophysiol., vol. 79,
pp. 1119, 1991.

SRINIVASAN et al.: APPROXIMATE ENTROPY-BASED EPILEPTIC EEG DETECTION USING ARTIFICIAL NEURAL NETWORKS

[8] N. Pradhan, P. K. Sadasivan, and G. R. Arunodaya, Detection of seizure

activity in EEG by an artificial neural network: A preliminary study,
Comput. Biomed. Res., vol. 29, pp. 303313, 1996.
[9] V. P. Nigam and D. Graupe, A neural-network-based detection of
epilepsy, Neurol. Res., vol. 26, pp. 5560, 2004.
[10] M. K. Kiymik, A. Subasi, and H. R. Ozcalik, Neural networks with
periodogram and autoregressive spectral analysis methods in detection of
epileptic seizure, J. Med. Syst., vol. 28, pp. 511522, 2004.
[11] J. Bruhn, H. Ropcke, and A. Hoeft, Approximate entropy as an electroencephalographic measure of anesthetic drug effect during desflurane
anesthesia, Anesthesiology, vol. 92, pp. 715726, 2000.
[12] S. M. Pincus, I. M. Gladstone, and R. A. Ehrenkranz, A regularity
static for medical data analysis, J. Clin. Monit., vol. 7, pp. 335345,
1991.
[13] S. M. Pincus, Approximate entropy as a measure of system complexity,
Proc. Natl. Acad. Sci. USA, vol. 88, pp. 22972301, 1991.
[14] S. M. Pincus, T. R. Cummins, and G. G. Haddad, Heart rate control in
normal and aborted SIDS infants, Amer. J. Physiol., vol. 264, no. 3,
pp. R638R646, 1993.
[15] S. M. Pincus and A. L. Goldberger, Physiological time series analysis:
what does regularity quantify?, Amer. J. Physiol., vol. 266, pp. H1643
Hl656, 1994.
[16] J. S. Richman and J. R. Moorman, Physiological time-series analysis using approximate entropy and sample entropy, Amer. J. Physiol., vol. 278,
pp. H2039H2049, 2000.
[17] W. S. Kim, Y. Z. Yoon, J. H. Bae, and K. S. Soh, Nonlinear characteristics
of heart rate time series: Influence of three recumbent positions in patients
with mild or severe coronary artery disease, Physiol. Meas., vol. 26,
pp. 517529, 2005.
[18] S. M. Pincus, Older males secrete luteinizing hormone and testosterone
more irregularly and joint more asynchronously, than younger males,
Proc. Natl. Acad. Sci. U.S.A., vol. 93, pp. 1410014105, 1996.
[19] N. Radhakrishnan and B. N. Gangadhar, Estimating regularity in epileptic
seizure time-series data. A complexity-measure approach, IEEE Eng.
Med. Biol., vol. 17, no. 3, pp. 8994, May./Jun. 1998.
[20] X. S. Zhang and R. J. Roy, Derived fuzzy knowledge model for estimating
the depth of anesthesia, IEEE Trans. Biomed. Eng., vol. 48, no. 3,
pp. 312323, Mar. 2001.
[21] L. Diambra, J. C. Bastos de Figueiredo, and C. P. Malta, Epileptic activity
recognition in EEG recording, Phys. A: Stat. Mech. Appl., vol. 273,
pp. 495505, 1999.
[22] R. G. Andrzejak, K. Lehnertz, F. Mormann, C. Rieke, P. David, and
C. E. Elger, Indications of nonlinear deterministic and finite-dimensional
structures in time series of brain electrical activity: Dependence on recording region and brain state, Phys. Rev. E, Stat. Nonlinear Soft Matter Phys.,
vol. 64, pp. 061907-1061907-8, 2001.
[23] M. Le Van Quyen, J. Martinerie, V. Navarro, P. Boon, M. D Have, and
C Adam et al., Anticipation of epileptic seizures from standard EEG
recordings, Lancet, vol. 357, pp. 83188, 2001.
[24] H. Demuth and M. Beale, Neural Network Toolbox.(for use with Matlab).
Natick, MA: MathWorks, 2000.
[25] C. J. Huang and W. C. Liao, A comparative study of feature selection
methods for probabilistic neural networks in cancer classification, in
Proc. 15th IEEE Intern. Conf. Tools with Artif. Intell. (ICTAI03), Sacramento, CA, Nov. 35, 2003, pp. 451458.
[26] V. Srinivasan, C. Eswaran, and N. Sriraam, Artificial neural network
based epileptic detection using time-domain and frequency-domain features, J. Med. Syst., vol. 29, pp. 647659, 2005.

295

Vairavan Srinivasan (M04) received the B.E. degree from the University of Madras, Chennai, India,
in 2003, and the Masters degree in biomedical signal
processing from the Faculty of Information Technology, Multimedia University, Cyberjaya, Malaysia, in
2006.
He is currently a Researcher of fetal magnetocardiogram (fMCG) signal processing at the Institute of
Advanced Biomedical Techniques (ITAB), G. D. Annunzio University, Chieti, Italy. His current research
interests include investigation of parameters of the
chaos theory in the nonlinear characterization of the fMCG source signals and
in the automated detection system for fetal arrhythmias, EEG-based detection
of epilepsy, anesthetic depth, and magnetoencephalogram (MEG).

Chikkannan Eswaran (SM81) received the

B. Tech, M. Tech, and Ph. D. degrees from the Indian
Institute of Technology Madras, Chennai, India, all
in electrical engineering.
He was a Visiting Faculty/Research Fellow at Ruhr
University, Germany, Concordia University, Canada,
University of Victoria, Canada, and Nanyang Technological University, Singapore. He has also been a
Professor in the Department of Electrical Engineering, Indian Institute of Technology Madras. He is currently a Professor at the Faculty of Information Technology, Multimedia University, Cyberjaya, Malaysia, where he is the Chairman
of the Center of Multimedia Computing. He is the author or coauthor of more
than 100 journal and conference papers in the areas of digital signal processing,
biomedical engineering, and information systems.

Natarajan Sriraam (M01) received the B.E. degree

(ECE) from the National Engineering College, India,
and the M. Tech. degree in biomedical engineering
from Manipal Institute of Technology, Manipal, India. He is currently working toward the Ph.D. degree
at the Multimedia University, Cuberjaya, Malaysia.
He is currently a Lecturer in the Faculty of Information Technology, Multimedia University. His
current research interests include biomedical signal
and image processing, bioinformatics, data mining,
and neural networks.
Mr. Sriraam is a member of the Institute of Electrical Engineers, the IEEE
Engineering in Medicine and Biology Society, and the IEEE Signal Processing
Society.