International Journal of Scientific & Engineering Research, Volume 2, Issue 8, August-2011 1

ISSN 2229-5518

Review of Significant Research on EEG based Automated Detection of Epilepsy Seizures & Brain Tumor

Sharanreddy.M and Dr.P.K.Kulkarni

AbstractElectroencephalography (EEG) measures the electrical activity of the brain and represents a summation of post- synaptic potentials from a large number of neurons. Over a past few decades many researches all over the world, focusing and working to automate the analysis of EEG signals to identify and categorized the diseases. In this paper, we present a review of the significant researches associated with the automated detection of epileptic seizures and brain tumor using EEG signals.

Index TermsBrain Tumor, EEG, Epilepsy Seizures, Automated, Neural Network, Wavelet Transform, Neurological diseases

—————————— ——————————

1 INTRODUCTION

HE brain is one of the most complex organs of the human body, which involves billions of interacting physiological and chemical processes that give rise to experimental observed neuroelectrical activity. The signal electroencephalographic (EEG) is defined as a representa- tion of post-synaptic potentials that are generated at cor- tical level by synchronous activity of about 105 (10 rates to 5) neurons. The (EEG) which provides insight informa- tion representing the brain‟s electrical activity is the most utilized signal to assess and detect abnormalities in the electrical activity of the brain, The EEG signal contains the useful information along with redundant or noise infor-
mation.

2 EPLIEPSY SEIZURE DETECTION

Epilepsy is a common chronic neurological disorder. Epi- lepsy seizures are the result of the transient and unex- pected electrical disturbance of the brain. About 50 mil- lion people world wide have epilepsy, and nearly two out of every three new cases are discovered in developing countries [1]. Epilepsy is more likely to occur in young children or people over the age of 65 years; however, it can occur at any time [2].
In epilepsy, the normal pattern of neuronal activity
becomes disturbed, causing strange sensations, emotions,
and behavior, or sometimes convulsions, muscle spasms,
and loss of consciousness [1]. There are many possible
causes of epilepsy. Anything that disturbs the normal
pattern of neuron activity ranging from illness to brain damage to abnormal brain development can lead to sei-

————————————————

Sharanreddy.M, research Scholar in Electrical & Electronics engineering

P.D.A.C.E, India. E-mail: sharanreddy.m@hotmail.com

Dr.P.K.Kulkarni, Prof and HOD of Electrical & Electronics engineering in

P.D.A.C.E, India.

zures. Epileptic seizures are manifestations of epilepsy
[3].
In the last couple of years, the EEG analysis was most- ly focused on epilepsy seizure detection diagnosis. The
methodology is based on three different adroit integration of computing technologies and problem solving para- digms (e.g., neural networks, wavelets, and chaos theory).
Starting with template matching algorithm (find events that match previously selected spikes), which uses a statistical approach to compare the EEG signal with a data base of known epileptic spikes [56]. This method lacks in the accuracy to detect the epilepsy.
The other methods of automatic EEG processing were based on a Fourier transform. This approach is based on earlier observations that the EEG spectrum contains some characteristic waveforms that fall primarily within four frequency bands. Such methods have proved beneficial for various EEG characterizations, but fast Fourier trans- form (FFT), suffer from large noise sensitivity. Parametric methods for power spectrum estimation such as autore- gressive (AR), reduces the spectral loss problems and gives better frequency resolution. Since the EEG signals are non-stationary, the parametric methods are not suita- ble for frequency decomposition of these signals.

2.1 Reveiw of Neural Network based approaches

Gular et al [9] have a study the assessment of accuracy of recurrent neural networks (RNN) employing Lyapunov exponents in detection seizure in the EEG signals. Yue- dong Song, Pietro Liò [20] developed an EEG epilepsy detection scheme based on the entropy based feature ex- traction and extreme learning machine. The proposed system employed a recently-proposed statistical parame- ter re-ferred to as Sample entropy (SampEn), together with extreme learning machine (ELM) which is a recently- developed classification model, to classify subjects as normal subject, patients not having an epileptic seizure or patients having an epileptic seizure. Compare the per-

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 8, August-2011 2

ISSN 2229-5518

formance of ELM classifiers with a back propagation neural network (BPNN) based on a Levenberg-Marquardt back-propagation (LMBP) learning algorithm. Results show that the proposed scheme achieves an excellent per- formance with not only the accuracy as high as 95.67% but also with very fast learning speed (0.0250 seconds), which demonstrates its potential for real-time implemen- tation in an epilepsy diagnosis support system.
Kezban Aslan et al. [10] have conducted a study to ex- amine epileptic patients and perform classification of epi- lepsy groups. The classification process groups into par- tial and primary generalized epilepsy by employing Radial Basis Function Neural Network (RBFNN) and Multilayer Perceptron Neural Network (MLPNNs). The parameters acquired from the EEG signals and clinic properties of the patients are used to train the neural networks. The experimental results obtained, depicted that the predictions corresponding to the learning data sets were convincing for both neural network models. It would be stated from the results that RBFNN (total classi- fication accuracy = 95.2%) produced better classification than MLPNN (total classification accuracy = 89.2%). From the results, it is determined that the RBFNN model can be used as a decision support tool in clinical studies to vali- date the epilepsy group classification after the develop- ment of the model.
M. Kemal Kiymik et al. [11] have examined the per- formance of the periodogram and autoregressive (AR) power spectrum methods. Owing to the automatic com- parison of epileptic seizures in EEG a method is offered by them, which allows the combining of seizures that have alike overall patterns. Every channel of the EEG was first broken down into segments having comparatively stationary characteristics. For each segment the features are calculated, and all segments of all channels of the sei- zures of a patient are combined into clusters of same morphology. With the examination of 5 patients with scalp electrodes that demonstrated the capability of the method to cluster seizures of alike morphology and ob- served that ANN categorization of EEG signals with AR preprocessing gave improved outcome, and those out- come could also used for the deduction of epileptic sei- zure.
The use of autoregressive (AR) model is examined by Abdulhamit Subasi et al. [12] by using utmost likelihood estimation (MLE) also interpretation together with the performance of this method to dig out classifiable fea- tures from human EEG by means of Artificial Neural Networks (ANNs). It is noticed that; ANN classification of EEG signals with AR produced noteworthy results. Their approach is on the basis of the earlier where the EEG spectrum enclosed a few characteristic waveforms which fall primarily within four frequency bands-delta (<
4 Hz), theta (4–8 Hz), alpha (8–14 Hz), and beta (14–30
Hz). For the automatic classification of seizures a method
is offered as well as attained a classification rate of 92.3%
by means of a neural network with a single hidden unit as
a classifier. The classification percentages of AR with MLE on test data are over 92%. As a result of employing FFT as preprocessing in the neural net an average of 91% classification is attained.
A wavelet-chaos-neural network methodology for classification of electroencephalograms (EEGs) into healthy, ictal, and interictal EEGs has been offered by Samanwoy Ghosh-Dastidar et al. [13]. In order to decom- pose the EEG into delta, theta, alpha, beta, and gamma sub-bands the wavelet analysis is utilized. Three parame- ters are used for EEG representation: standard deviation (quantifying the signal variance), correlation dimension, and largest Lyapunov exponent (quantifying the non- linear chaotic dynamics of the signal). The classification accuracies of the following techniques are compared: 1) unsupervised - means clustering; 2) linear and quadratic discriminant analysis; 3) radial basis function neural net- work; 4) Levenberg–Marquardt back propagation neural network (LMBPNN). The research was carried out in two phases with the intention of minimizing the computing time and output analysis, band-specific analysis and mixed-band analysis. In the second phase, over 500 dif- ferent combinations of mixed-band feature spaces com- prising of promising parameters from phase one of the research were examined. It is decided that all the three key components the wavelet-chaos-neural network me- thodology are significant for enhancing the EEG classifi- cation accuracy. Judicious combinations of parameters and classifiers are required to perfectly discriminate be- tween the three types of EEGs. The outcome of the me- thodology clearly let know that a specific mixed-band feature space comprising of nine parameters and LMBPNN result in the highest classification accuracy, a high value of 96.7%.
To categorize the types of epileptic seizures a simple approach is offered by Najumnissa and Shenbaga Devi [14]. Their concentration is on the detection of epileptic seizures from scalp EEG recordings. On the basis of two stages seizures are categorized: Stage I was a set of neural network-based epileptic seizure detector and stage II was a neural network, which classifies the abnormal EEG from, stage I. From 34 patients 436 features have been chosen. In order to train the neural network out of 436 feature sets, 330 feature sets from 26 patients are utilized and the remaining 106 feature sets of eight patients were kept for testing. By means of the wavelet transform tech- nique the features are pulled out. Two networks are used by them one is for detecting normal and abnormal condi- tions, the second one for classification. The onset of the seizure was continuously moving by the window and the time of onset was recognized. In the tests of the system on EEG denoted a success rate of 94.3% was obtained. The system was made as a real-time detector by their method and it enhanced the clinical service of Electroencephalo- graphic recording.
An automated epileptic system, which uses interictal
EEG data to categorize the epileptic patients, was devel-

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 8, August-2011 3

ISSN 2229-5518

oped by Forrest Sheng Bao etal. [15].The diagnostic sys- tem was used to detect seizure activities for additional examination by doctors and impending patient monitor- ing. They have built a Probabilistic Neural Network (PNN) fed with four classes of features extracted from the EEG data. Their approach was more efficient when com- pared to the present conventional seizure detection algo- rithms because they are seizure independent i.e. doesn‟t necessitate the seizure activity attained from the EEG re- cording. This feature shuns intricacy in the EEG collection as interictal data was much easier to be collected than ictal data. In their work, the PNN was employed to classi- fy 38 extracted EEG features. During cross validation their interictal EEG based diagnostic approach achieved a
99.5% overall accuracy. The classification based on ictal data also showed a high (98.3%) degree of accuracy. The- reby, with both interictal and ictal data their algorithm worked well. The function of the classifier was further extended to achieve patient monitoring and focus locali- zation. An accuracy of 77.5% stated impending focus loca- lization. The speed of the classifier was good classifying an EEG segment of 23.6 seconds in just 0.01 seconds.
The efficiency of utilizing an ANN is assessed by Ste- ven Walczak and William J. Nowack [16] in order to de- termine epileptic seizure occurrences for patients with lateralized bursts of theta (LBT) EEGs. By means of the examination of records of 1,500 successive adult seizure patients training and test cases are obtained. Owing to the development of the ANN categorization models the small resulting pool of 92 patients with LBT EEGs requisite the usage of a jackknife procedure. Evaluations of the ANNs are for accuracy, specificity, and sensitivity on categoriza- tion of each patient into the correct two-group categoriza- tion: epileptic seizure or non-epileptic seizure. By means of eight variables the original ANN model generated a categorization accuracy of 62%. Consequently, a modified factor analysis, an ANN model using just four of the orig- inal variables attained a categorization accuracy of 68%.
Subasi et al [17] compared the traditional method of logistic regression to the more advanced neural network techniques, as mathematical tools for developing classifi- ers for the detection of epileptic seizure in multi-channel EEG.
Kiymik et al. presented time–frequency analysis of EEG signals for detecting the information on alertness and drowsiness using spectral densities of DWT coeffi- cients as an input to ANN [18]. As compared to the con- ventional method of frequency analysis using Fourier transform or short time Fourier transform, wavelets ena- ble analysis with a coarse to fine multi-resolution pers- pective of the signal [19].The detection methods which use the characteristics of the EEG seizure in time or fre- quency domain are based on the assumption that the segments of the signal are quasi stationary. However re- cent works shows that the EEG signals exhibit non- stationary behavior. For analyzing such signals, time scale and time frequency methods have proved that the most
suitable tools.
Gabor and Seyal [21] introduce a neural network algo-
rithm that relies primarily on the spike field distribution. MLP networks with the number of input and hidden
nodes equal to the number of channels in the record and a single output node are used. Five bipolar 8 channel records from the EMU with durations ranging from 7.1 to
23.3 min are used for training and testing. Two networks are trained on only the slopes of the spike‟s half-waves, and there is no notion of background context. The first uses the slope of the half-wave before the spike‟s apex for all 8 channels as inputs, and the second uses the slope after the apex. The output of the algorithm is a weighted combination of the two network outputs with a value near 1.0 indicating a spike has been found. The duration (not specified) of the spike half waves is fixed so that no waveform decomposition is required. The algorithm slides along the data one sample at a time and identifies a spike when the output is greater than a threshold (e.g.
0.9). The method requires a distinct network for each pa- tient and spike foci, so 7 networks were trained because two of the patients had independent foci. The training required 4–6 example spikes and the non spikes were generated by statistical variation resulting in 4 times more non-spikes. Although this method does not seem to be well suited for general detection, it might be a promising
method for finding „similar‟ events.

2.2 Review of Wavelet Transform Based approaches

For the detection of seizure and epilepsy Hojjat Adeli et al. [22] have offered a wavelet chaos methodology for analysis of EEGs and delta, theta, alpha, beta, and gamma sub-bands of EEGs. In the form of the correlation dimen- sion (CD, representing system complexity) and the largest Lyapunov exponent (LLE, representing system chaoticity) the nonlinear dynamics of the original EEGs are quanti- fied. The new wavelet-based methodology isolated the changes in CD and LLE in specific sub-bands of the EEG. The methodology was applied to three diverse groups of EEG signals: healthy subjects, epileptic subjects during a seizure-free interval (interictal EEG), and epileptic sub- jects during a seizure (ictal EEG).The effectiveness of CD and LLE in distinguishing between the three groups is examined based on statistical importance of the varia- tions. It has been noted that in the values of the parame- ters acquired from the original EEG there may not be noteworthy differences, differences may be recognized when the parameters were employed in conjunction with particular EEG sub-bands and concluded that for the higher frequency beta and gamma sub-bands, the CD distinguished between the three groups, in disagreement to that the lower frequency alpha sub-band, the LLE dis- tinguished between the three groups.
Subasi [25] deals with a novel method of analysis of EEG signals using discrete wavelet transform, and classi- fication using ANN. In this work the signal decomposed

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 8, August-2011 4

ISSN 2229-5518

in 5 levels using DB4 wavelet filter .The energy of details and approximation were used as the input features.
M.Akin, M.A.Arserim, M.K.Kiymik, I.Turkoglu [26] have tried to find a new solution for diagnosing the epi- lepsy. For this aim, the Wavelet Transform of the EEG signals have taken, and the δ, θ, α, and β sub frequencies are extracted. Depending on these sub frequencies an ar- tificial neural network has been developed and trained. The accuracy of the neural network outputs is too high (97% for epileptic case, 98% for healthy case, and 93% for pathologic case that have been tested). This means that this neural network identifies the health conditions of the patients approximately as 90 of 100. From this point we can say that an application of this theoretical study will be helpful for the neurologists when they diagnose the epi- lepsy.
Xiaoli Li [35] proposed an approach based on multi –
resolution analysis to automatically indicate the epileptic
seizures or other abnormal events in EEG. The energy of
EEG signals at the different frequency bands is calculated
for detecting the behaviors of brain during epileptic sei- zures. The energy change of each frequency band is indi-
cated as a feature by calculating the Euclidean distance between a reference segment and the segments extracted in real time. The selection of wavelet functions, scale pa- rameters, width of wavelet function, and sample sizes (segment length) are emphasized. Then, the features go through a recursive in-place growing FIR-median hybrid (RIPG-FMH) filter. The results suggest that wavelet trans- form is a useful tool to analyze the EEG signals with the epileptic seizures.
Ganesan.M, Sumesh.E.P, Vidhyalavanya.R [36] pro- posed a technique for the automatic detection of the spikes in long term 18 channel human electroencephalo- grams (EEG) with less number of data set. The scheme for detecting epileptic and non epileptic spikes in EEG is based on a multi resolution, multi-level analysis and Ar- tificial Neural Network (ANN) approach. The signal on each EEG channel is decomposed into six sub bands using a non-decimated WT. Each sub band is analyzed by using a non-linear energy operator, in order to detect spikes. A parameter extraction stage extracts the parameters of the detected spikes that can be given as the input to ANN classifier. The system is evaluated on testing data from 81 patients, totaling more than 800 hours of record- ings.90.0% of the epileptic events were correctly detected and the detection rate of non epileptic events was 98.0%.

2.3 Review of Other approaches

The detection of epileptic seizures from scalp EEG re- cordings was the area of focus for McSharry et al [23]. A synthetic signal was created by merging a linear random process and a non-linear deterministic process. They in- troduced a multidimensional probability evolution (MDPE) statistic capable of detecting faint variations in the underlying state space that were associated with mod- ifications in the dynamical equations used in production
of synthetic signal.. F-tests were used to calculate the sig- nificance of the observed difference between the va- riances of the recording, all through the learning period and testing the window. Moreover, the significance of the observed difference between the multidimensional distri- butions observed in the state space all through those pe- riods are attained using tests and also the linear statistics and the MDPE statistics were used by them to analyze the database of scalp EEG recordings. The MDPE and va- riance were utilized for seizure detection but the MDPE offered better accuracy for seizure onset detection in re- cordings E/1, E/2, and F/1. Nonlinear statistics largely augmented the scope of automatic detection, but its utili- zation has justified on a case-by-case basis.
Forrest Sheng Bao et al. [28] have developed a diagnos- tic system that can employ interictal EEG data to auto- matically diagnose epilepsy in humans. The system could also detect seizure activities for preceding examination by doctors and impending patient monitoring. The system was developed by extracting three classes of features from the EEG data. These features were fed up with to build a Probabilistic Neural Network (PNN). Leave-one- out cross-validation (LOO-CV) on an extensively used epileptic-normal data set reveals a striking 99.3% accura- cy of the system on distinguishing normal people's EEG from patients' interictal EEG. Moreover, it was found that the system can be used in patient monitoring (seizure detection) and seizure focus localization, with 96.7% and
76.5% accuracy respectively on the data set.
G.R. de Bruijne et al. [24] have proposed a patient
monitoring system based on audio classification for de-
tecting the epileptic seizures. The system facilitated an
automated detection of the epileptic seizures which is likely to have a significant positive impact on the daily
care of epilepsy patients. Their system comprised of three stages. First, the signal was improved by means of a mi- crophone array, followed by a noise subtraction proce- dure. Secondly, the signal was evaluated by audio event detection and audio classification. The characteristics were extracted from the signal on detection of an audio event. Bayesian decision theory was used to categorize the feature vector on the basis of discriminate analysis. At last, it decides whether to activate an alarm or not. With the help of the audio signals obtained from the measure- ments with the epileptic patients the performance of the system was tested. They have achieved better classifica- tion results with a limited set of features.
Sivasankari N and Dr. K. Thanushkodi [27] purposed method for epileptic seizure detection from the recorded EEG brain signals using Fast Independent Component Analysis and ANN. To begin with, independent subcom- ponents are separated from the recorded signals with the aid of Fast Independent Component Analysis. Further, the signals are trained using ANN (Artificial Neural Networks) technique namely Back propagation algo- rithm. The exertion of FastICA and ANN proffered en- couraging results in the detection of epileptic seizure

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 8, August-2011 5

ISSN 2229-5518

from the recorded EEG signals. The accuracy results of the proposed approach (76.5% for epileptic case, 66% for healthy case) for the EEG data 200 EEG signals each.
Chua K. C, Chandran V, Rajendra Acharya [31], Lim C. M. proposed nonlinear approach motivated by the higher order spectra (HOS) to differentiate between normal, background (pre-ictal) and epileptic EEG signals. In the work, the features are extracted from the power spectrum and the bispectrum. Their performance is studied by feeding them to a Gaussian mixture model (GMM) clas- sifier. Results show that with selected HOS based fea- tures, achieve 93.11% accuracy compared to classification accuracy of 88.78% as that of features derived from power spectral density (PSD).
T. Tzallas, M. G. Tsipouras, and D. I. Fotiadis [32] ex- plored the ability of the Time Frequency analysis to classi- fy EEG segments which contain epileptic seizures. They extracted several time-frequency features and examined the effect of the parameters entering the problem, that is, the frequency resolution of the time-frequency analysis and the number of time windows and frequency sub bands used for feature extraction. Promising results have been reported after the evaluation of the proposed me- thod in four different classification problems, derived from a well-known database. They achieved accuracy of (97.72 – 100%) after testing on different datasets.
Suparerk Janjarasjitt [33] a member of International Journal of Applied Biomedical Engineering proposed method using wavelet transform as a primary computa- tional tool for extracting characteristics of the epileptic EEG signals at various scales (resolutions). The wavelet- based scale variance defined as log-variance of wavelet coefficients of the epileptic EEG signal is used as a feature vector for the classification. The k-means clustering is then used to classify the epileptic EEG data from the cor- responding wavelet-based scale variance features. The accuracy for the classification of the epileptic EEG signals for the different set of dataset with variable accuracy from
95.00% - 99.00%
Ralph Meier, Heike Dittrich, Andreas Schulze-Bonhage
and Ad Aertsen [34] proposed a method for generic, on-
line, and real-time automatic detection of multi-
morphologic ictal-patterns in the human long-term EEG
and its validation in continuous, routine clinical EEG re-
cordings from 57 patients with a duration of approx- imately 43 hours and additional 1,360 hours of seizure-
free EEG data for the estimation of the false alarm rates. They Analyzed 91 seizures (37 focal, 54 secondarily gene- ralized) representing the six most common ictal morphol- ogies (alpha, beta, theta, and delta- rhythmic activity, amplitude depression, and polyspikes). And found that taking the seizure morphology into account plays a cru- cial role in increasing the detection performance of the system. Moreover, besides enabling a reliable (mean false alarm rate < 0.5/h, for specific ictal morphologies <
0.25/h), early and accurate detection (average correct de- tection rate > 96%) within the first few seconds of ictal
patterns in the EEG, this procedure facilitates the auto- matic categorization of the prevalent seizure morpholo- gies without the necessity to adapt the proposed system to specific patients.
Alexandros T. Tzallas, Markos G. Tsipouras, and Dimi- trios I. Fotiadis, members of IEEE [37] proposed method of analysis of EEG signals using time-frequency analysis, and classification using artificial neural network, is intro- duced. EEG segments are analyzed using a time- frequen- cy distribution and then, several features are extracted for each segment representing the energy distribution over the time frequency plane. The features are used for the training of a neural network. Short-time Fourier trans- forms and several time-frequency distributions are com- pared. The proposed approach is tested using a publicly available database and satisfactory results are obtained (89-100% accuracy).

3 BRAIN TUMOR DETECTION

The brain is an incredibly complex organ. Like a true res- ident in an Ivory Tower, the brain lives apart from and quite differently than the rest of the body. The brain con- tains about 10 Billion (10,000,000,000) working brain cells. They are called neurons and make over 13 Trillion (13,000,000,000,000) connections with each other to form the most sophisticated organic computer on the planet -- maybe even the universe. By today's computer standards, the brain far exceeds any network of linked state-of-the- art computers [39]. Although cells in different parts of the body may look and work differently, most repair them- selves in the same way, by dividing to make more cells. Normally, this turnover takes place in an orderly and controlled manner. If, for some reason, the process gets out of control, the cells will continue to divide, develop- ing into a lump, which is called a tumor.
Clinical neurologists use Computer Tomography (CT)
imaging techniques for diagnosis of brain tumors because
of it high accuracy in initial diagnosis of the primary pa-
thology (96% of cases). Such scans stand short, however
of analyzing the physiological functioning of the brain as
a whole both at the time of initial diagnosis or as part of a long term management of the patient. For such purpose,
EEG has been used to render a clearer overall view of the brain functioning at initial diagnosis stages. In brain Tu- mor diagnostics, EEG is most relevant in assessing how the brain responds to treatments (e.g. post operative)
Being a non-invasive low cost procedure, the EEG is an attractive tumor diagnosis method on its own. It is a reli- able tool for the glioma tumor series. The EEG in vascular lesions is abnormal form the onset of symptoms where as a CT only become abnormal on the third or fourth day or after week. The EEG is, however less successful in detect- ing brain stem tumors and meningioma series.
Murugesan, M. Sukanesh, R [40] proposed a method for automated system for efficient detection of brain tu-

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 8, August-2011 6

ISSN 2229-5518

mors in EEG signals using ANN. The ANN employed in the proposed system is feed forward back propagation neural network. Generally, the EEG signals are bound to contain an assortment of artifacts from both subject and equipment interferences along with essential information regarding abnormalities and brain activity (responses to certain stimuli). Initially, adaptive filtering is applied to remove the artifacts present in the EEG signal. Subse- quently, generic features present in the EEG signal are extracted using spectral estimation. Specifically, spectral analysis is achieved by using Fast Fourier Transform that extracts the signal features buried in a wide band of noise. The clean EEG data thus obtained is used as training in- put to the feed forward back propagation neural network. The trained feed forward back propagation neural net- work when fed with a test EEG signal, effectively detects the presence of brain tumor in the EEG signal. The expe- rimental results demonstrate the effectiveness of the pro- posed system in artifacts removal and brain tumor detec- tion.
Seenwasen Chetty, Ganesh K. Venayagamoorthy [41] proposed The ANN based EEG classifier to distinguish between the EEG signal of a normal patient and that of a brain tumor patient. The results show that an artificial neural network is able to distinguish between an abnor- mal and normal EEG signal, and classify them correctly as brain tumor and healthy patient respectively. This is possible with ANNs since they are able learn the patterns in a normal and abnormal EEG signal. ANN gives a 100% classification success rate with both normal and abnor- mal EEG.
Fadi N. Karameh, Munther A.Dahleh [42] focused on developing an automated system to identify space occu- pying lesions on the brain using EEG signals. EEG fea- tures are extracted using wavelet transform for different tumor classes and classification by self-organizing maps.
M. Murugesan and Dr. (Mrs.).R. Sukanesh [43] pro-
posed a technique for classification of electroencephalo-
gram (EEG) signals that contain credible cases of brain tumor. The classification technique support vector ma-
chine is utilized in the proposed system for detecting brain tumors. The artifacts present in the EEG signal are removed using adaptive filtering. Then the spectral anal- ysis method is applied for extracting generic features em- bedded in an EEG signal. Precisely, Fast Fourier Trans- form for spectral analysis is used to separate the signal features which are buried in a wide band of noise. The radial basis function-support vector machine is trained using the clean EEG data obtained. With proper testing and training, they effectively classify the EEG signals with brain tumor.
Rosaria Silipo, Gustavo Deco and Helmut Bartsch [44]
proposed a brain tumor classification method on EEG
signals. The classification done by applying a nonlinear
analysis to the hidden dynamic of the F3 and F4 EEG
leads, that describe the electrical activity of the left and
right brain hemisphere, respectively. The hidden dynamic
of the pair (F3, F4) is tested against a hierarchy of null hypotheses, corresponding to one- and two-dimensional nonlinear Markov models of increasing order. An appre- ciative measure of information flow, based on higher or- der cumulates, quantifies the hidden dynamic of each time series and is used as a discriminating statistic for testing the null hypotheses. The minimum order of the accepted Markov models represents a measure of the in- trinsic nonlinearity of the underlying system. Rest EEG records of 6 patients with evidence of meningeoma or malignant glioma in lead F4, or without any pathology, are investigated. A high order hidden dynamic is de- tected in normal EEG records, confirming the very com- plex structure of the underlying system. Different inter- dependence degrees between the hidden dynamics of leads (F3, F4) discriminate meningeoma, malignant gli- oma, and no pathological status, while loss of structure in the hidden dynamic can represent a good hint for glioma
/ meningeoma localization.
Habl, M. and Bauer, Ch. and Ziegaus, Ch., Lang, Elmar
and Schulmeyer, F [45] presented a technique to detect and characterize brain tumors. They removed location
arifactual signals, applied a flexible ICA algorithm which does not rely on a priori assumptions about unknown source distribution. They have shown that tumor related EEG signals can be isolated into single independent ICA components. Such signals where not observed in corres- ponding EEG trace of normal patients.
Lawrence J. Hirsch [46] suggested the use of conti- nuous EEG monitoring (CEEG), which refers to pro- longed (hours, days, or weeks) continuously recorded digital EEG in critically ill patients with altered mental status or with a significant risk for acute brain ischemia. Use of CEEG is rapidly expanding, largely due to the widespread availability of digital video/EEG, advances in computer memory storage capabilities, and the ability to review studies remotely via computer networking. He reviewed experience with CEEG in 570 patients who were monitored to detect or rule out Non Convulsive Seizures (NCSzs), or for unexplained decrease in level of con- sciousness (Claassen et al. 2004). The mean age was 52 years, and 75 patients were younger than 18 years. Over- all, seizures were detected in 110 (19%) patients. Impor- tantly, 101 of these 110 patients had exclusively NCSzs; thus, without EEG monitoring, the diagnosis would have been missed. The diagnoses most associated with NCSzs were prior epilepsy (31% had NCSzs), CNS infection (26%), brain tumor (23%), or a recent neurosurgical pro- cedure (23%). Other statistically significant predictors of seizures on CEEG from this population were coma at time of initiation of CEEG (56% of 97 comatose patients had seizures on CEEG), convulsive seizures before monitoring (43% of 134 patients), history of epilepsy (41% of 68 pa- tients), age younger than 18 years (36% of 75 patients), and periodic epileptiform discharges (focal or genera- lized) on EEG.
Small, Joyce Graham Bagchi, Basu K.Kooi, Kenneth A

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 8, August-2011 7

ISSN 2229-5518

[47] studied 117 patients with verified deep cerebral tu- mor, qualitatively and statistically in the relation to 60 clinical and EEG variable, s.92 patients had moderate to profound EEG abnormality. They stressed need for ade- quacy off EEG technique and analyzed factors causing distant rhythms.

4 CONCLUSION

The EEG signals are commonly utilized to clinically as- sess brain activities. The detection of epileptic seizures and brain tumor from the EEG signals is a significant process in the diagnosis of epilepsy seizures and brain tumor. More precisely, parameters extracted from EEG signals are greatly valuable for diagnostics. In this paper a lietrature survey of the significant and recent researches that are concerned with effective detection of Epileptic seizures and brain tumor using EEG signals are pre- sented. The main goal behind this review is to assist the researchers in the field of EEG signal analysis to under- stand the available methods and adopt the same for the detection of neurological disorders associated with EEG.

REFERENCES

[1] J.D. Bronzino, “Biomedical Engineering Handbook”, New

York: CRC Press LLC, Vol. I, 2nd edition 2000

[2] D.Bradbury, “Volunteer‟s Guide to an EEG”, Institute Of Neu- rology Wellcome Department Of Imaging Neuroscience, http://www.fil.ion.ucl.ac.uk/EEGvolunteerguide.pdf

[3] Hojjat Adeli, Samanwoy Ghosh and Dastidat “Autoamted EEG Based Diagnosis of Neurological Disorders”, CRC Press; 1 edi- tion, 2010

[4] Usha K. Misra and Jayantee Kalita, “Clinical Electroencephalo-

graphy”, Journal of Pediatrics Neurology, IOS Press, Vol 4, Nov

2006

[5] Smith,J. “Automatic and Detection of EEG Spikes” ,IEEE trans- actions on Bio medical Engineering ,Vol. BMF-21:1-7,Jan,1974

[6] S.G. Mallat, “A Theory for Multiresolution Signal Decomposi-

tion: The Wavelet Representation”, IEEE Trans. On Pattern

Analysis and Machine Intell, Vol. 2, No. 7, pp: (674-693), 1989.

[7] Maan M. Shaker, “EEG Waves Classifier using Wavelet Trans- form and Fourier Transform”, International Journal of Biomedical Sciences, Vol 1, No 2, 2006, Issn 1306-1216.

[8] M. Teplan “Fundamentals of Eeg Measurement” Measurement

Science Review, Vol 2, Sec 2, 2002

[9] N.F. Gulera, E.D. Ubeylib, I. Guler, "Recurrent neural networks employing Lyapunov Exponents for EEG signals classification" Expert Systems with Applications, Vol. 29, pp. 506–514, 2005

[10] Kezban Aslan, Hacer Bozdemir, Cenk Sahin, Seyfettin Noyan

Ogulata and Rizvan Erol, "A Radial Basis Function Neural Network Model for Classification of Epilepsy Using EEG Sig- nals", Journal of Medical Systems, Vol. 32, pp. 403 – 408, 2008

[11] M. Kemal Kiymik, Abdulhamit Subasi, and H. Rıza Ozcalık, "Neural Networks With Periodogram and Autoregressive Spec- tral Analysis Methods in Detection of Epileptic Seizure", Journal

of Medical Systems, Vol. 28, No. 6, Dec 2004

[12] Abdulhamit Subasi, M. Kemal Kiymik, Ahmet Alkan, and Etem Koklukaya, "Neural Network Classification Of EEG Signals By Using AR With MLE Preprocessing For Epileptic Seizure Detec- tion", Journal of Mathematical and Computational Applications, Vol.

10, No. 1, pp. 57-70, 2005

[13] Samanwoy Ghosh-Dastidar, Hojjat Adeli, and Nahid Dadmehr, "Mixed-Band Wavelet-Chaos-Neural Network Methodology for Epilepsy and Epileptic Seizure Detection", IEEE Transactions On Biomedical Engineering, Vol. 54, No. 9, pp. 1545-1551, September

2007

[14] D. Najumnissa and S. Shenbaga Devi, "Intelligent identification and classification of epileptic seizures using wavelet trans- form", International Journal of Biomedical Engineering and Technol- ogy, Vol. 1, No. 3,pp. 293-314, 2008

[15] Forrest Sheng Bao, Donald Yu-Chun Lie, and Yuanlin Zhang,

"A New Approach to Automated Epileptic Diagnosis Using

EEG and Probabilistic Neural Network", in Proceedings of the

2008 20th IEEE International Conference on Tools with Artificial In- telligence, Vol. 02, pp. 482-486, 2008.

[16] Steven Walczak and William J. Nowack,"An Artificial Neural

Network Approach to Diagnosing Epilepsy Using Lateralized Bursts of Theta EEGs", Journal of Medical Systems, Vol. 25, No. 1, pp. 9-20, February 2001

[17] A. Subasia, and E. Ercelebi, "Classification of EEG signals using neural network and logistic regression," Computer Methods and Programs in Biomedicine, Vol. 78, pp. 87-99, 2005

[18] M.K. Kiymik, M. Akin, and A. Subasi, "Automatic recognition

of alertness level by using wavelet transform and artificial neu- ral network," Journal of Neuroscience Methods, Vol. 139(2), pp.

231– 240, 2004

[19] E. Haselsteiner, and G. Pfurtscheller, "Using time-dependent neural networks for EEG classification," IEEE Transactions on Rehabilitation Engineering, Vol. 8, pp 457–463, 2000

[20] Yuedong Song, Pietro Liò, “A new approach for epileptic sei-

zure detection: sample Entropy based feature extraction and ex- treme learning machine“, J. Biomedical Science and Engineering,

2010, 3, 556-567

[21] Gabor AJ, SeyalM. “Automated interictal EEG spike detection using artificial neural networks”. Electroencephalography and Clinical neurophysiology, 1992; 83: 271–280

[22] Hojjat Adeli, Samanwoy Ghosh – Dastidar and Nahid Dad- mehr, “A Wavelet-Chaos Methodology for Analysis of EEGs and EEG Sub-Bands to Detect Seizure and Epilepsy”, IEEE Transactions on Biomedical Engineering, Vol. 54, No. 2, pp. 205-

211, February 2007

[23] P.E. McSharry, T. He, L.A. Smith and L. Tarassenko, "Linear and non-linear methods for automatic seizure detection in scalp electroencephalogram recordings", Journal of Medical and Biolog- ical Engineering and Computing, Vol. 40, No. 4, pp. 447-461, July

2002.

[24] G.R. de Bruijne, P.C.W. Sommen and R.M. Aarts, "Detection of Epileptic Seizures Through Audio Classification", IFMBE Pro- ceedings, Vol. 22, pp. 1450–1454, 2008

[25] A. Subasi, "Epileptic seizure detection using dynamic wavelet network," Expert Systems with Applications, Vol. 29, pp 343–355,

2005

[26] M.Akin, M.A.Arserim, M.K.Kiymik, I.Turkoglu “A New Ap-

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 8, August-2011 8

ISSN 2229-5518

proach for Diagnosing Epilepsy by using Wavelet Transform and Neural Networks” Proceedings – 23rd Annual Conference – IEEE/EMBS Oct.25-28, 2001, Istanbul, Turkey

[27] Sivasankari. N and Dr. K. Thanushkodi “Automated Epileptic

Seizure Detection in EEG Signals Using FastICA and Neural

Network” Int. J. Advance. Soft Comput. Appl, Vol. 1, No. 2, Nov

2009 ISSN 2074-8523; ICSRS Publication, 2009

[28] Forrest Sheng Bao, Donald Yu-Chun Lie and Yuanlin Zhang, "A New Approach to Automated Epileptic Diagnosis Using EEG and Probabilistic Neural Network", in Proc. of IEEE Intl. Conf. on Tools with Artificial Intelligence, Vol. 2, Nov. 3-5, Dayton, Ohio,

2008, pp: 482- 486

[29] Hamid R. Mohseni1, A. Maghsoudi and Mohammad B. Sham- sollahi, “Seizure Detection in EEG signals: A Comparison of Different Approaches” in proceedings of 28th Annual IEEE Inter- national Conference on Engineering in Medicine and Biology Society, Aug. 30-Sept. 3, 2006, pp: 6724-6727

[30] Chen, H., Zhong, S., and Yao D. “Detection singularity value of

character wave in Epileptic EEG by wavelet”, in proceedings of IEEE International Conference on Communications, Circuits and Systems and West Sino Expositions, 2002, Vol. 2, 29 June-1 July

2002, pp:1094-1097.

[31] Chua K. C, Chandran V, Rajendra Acharya, Lim C. M. “Auto- matic Identification of Epilepsy by HOS and Power Spectrum parameters using EEG Signals: A Comparative study” 30th An- nual International IEEE EMBS Conference Vancouver, British Co- lumbia, Canada, August 20-24, 2008

[32] A. T. Tzallas, M. G. Tsipouras, and D. I. Fotiadis “Automatic

Seizure Detection Based on Time-Frequency Analysis and Ar- tificial Neural Networks” Hindawi Publishing Corporation Com- putational Intelligence and Neuroscience, Vol 2007, Article ID

80510

[33] Suparerk Janjarasjitt “Classification of the Epileptic EEGs Using the Wavelet-Based Scale Variance Feature” International Journal of Applied Biomedical Engineering, Vol.3, No.1, 2010

[34] Ralph Meier, Heike Dittrich, Andreas Schulze-Bonhage and Ad Aertsen “Detecting Epileptic Seizures in Long-term Human EEG: A New Approach to Automatic Online and Real-Time De- tection and Classification of Polymorphic Seizure Patterns” Journal of Clinical Neurophysiology, Vol 25, No 3, June 2008

[35] Xiaoli Li “EEG Analysis with Epileptic Seizures using Wavelet Transform” Dept. of Automation and Computer-Aided Engineer- ing, Chinese University of Hong Kong, Shatin, Hong Kong. 28, Nov. 2002

[36] Ganesan.M, Sumesh.E.P, Vidhyalavanya.R “Multi-Stage, Multi-

Resolution Method for Automatic” International Journal of Sig- nal Processing, Image Processing and Pattern Recognition, Vol 3, No 2, June, 2010

[37] Alexandros T. Tzallas, Markos G. Tsipouras, and Dimitrios I.

Fotiadis “The Use of Time-Frequency Distributions for Epilep- tic Seizure Detection in EEG Recordings” Proceedings of the 29th Annual International Conference of the IEEE EMBS Cité Internatio- nale, Lyon, France August 23-26, 2007

[38] Pedram Ataee, Alireza Nasiri Avanaki, Hadi Fatemi Shariatpa-

nahi, Seyed Mohammadreza Khoee “Ranking Features of Wavelet-Decomposed Eeg Based on Significance in Epileptic Seizure Prediction” 14th European Signal Processing Conference

(EUSIPCO 2006), Florence, Italy, September 4-8, 2006 [39] http://www.brain-surgery.com/primer.html

[40] Murugesan, M. Sukanesh, R “Automated Detection of Brain

Tumor in EEG Signals Using Artificial Neural Networks”, IEEE Conference on Advances in Computing, Control, & Telecommunica- tion Technologies, 2009. Act 09. Trivandrum, Kerala

[41] Seenwasen Chetty, Ganesh K. Venayagamoorthy “An investi-

gation into the Detection of brain tumours using electroence- phalography (EEG) signals with Artificial neural networks” Computational Intelligence Group, Department of Electronic En- gineering M L Sultan Technikon

[42] Fadi N. Karameh, Munther A.Dahleh “Automated Classifica-

tion of EEG Signal in Brain tumour diagnostics” IEEE Proceed- ing of the American Control Conference, Chiago, llinois. June 2000.

[43] M. Murugesan and Dr. (Mrs.).R. Sukanesh “Towards Detection

of Brain Tumor in Electroencephalogram Signals Using Support Vector Machines” International Journal of Computer Theory and Engineering, Vol. 1, No. 5, December, 2009

[44] [44] Rosaria Silipo, Gustavo Deco and Helmut Bartsch “Brain

Tumour classification Based on EEG hidden dynamics” Elsevier Inc journal Intelligent Data Analysis, Vol 3, Issue 6, Pages 413-514 (December 1999)

[45] [45] Habl, M. and Bauer, Ch. and Ziegaus, Ch., Lang, Elmar and Schulmeyer, F. “Can ICA help identify brain tumor related EEG signals?” Proceedings / ICA 2000, Second International Work- shop on Independent Component Analysis and Blind Signal Separa- tion: Helsinki, Finland, June 19 - 22, 2000. Unspecified, pp. 609-

614. ISBN 951-22-5017-9.

[46] Lawrence J. Hirsch “Continuous EEG Monitoring in the Inten- sive Care Unit: An Overview” Journal of Clinical Neurophysi- ology, Volume 21, Number 5, October 2004

[47] Small, Joyce Graham Bagchi, Basu K.Kooi, Kenneth A. “Electro- clinical profile of 117 deep cerebral tumors” Elsevier Inc Electro- encephalography and Clinical Neurophysiology, Vol 13, Issue 2 , Pages 193-207, April 1961

[48] I.Omerhodzic, S. Avdakovic, A. Nuhanovic, K. Dizdarevic “En-

ergy Distribution of EEG Signals: EEG Signal Wavelet-Neural Network Classifier” International Journal of Biological and Life Sci- ences, 6:4 2010

[49] David A. Peterson, James N. Knight, Michael J. Kirby, Michael

H. Thaut and CharlesW. Anderson “Feature Selection and Blind Source Separation in an EEG-Based Brain-Computer In- terface” EURASIP Journal on Applied Signal Processing, 2005:19,

3128–3140

[50] K Vijayalakshmi and Appaji M Abhishek “Spike Detection in Epileptic Patients EEG Data using Template Matching Tech- nique” International Journal of Computer Applications (0975 –

8887) Vol 2, No.6, June 2010

[51] Bagchi, Basu K. Kooi, Kenneth A. Selving, Benjamin T. Cal- houn, Hazel D. “Subtentorial tumors and other lesions: An elec- troencephalographic study of 121 Cases” Elsevier Inc Electroe- ncephalography and Clinical Neurophysiology, Vol 13, Issue 2 , Pages 180-192, April 1961

[52] Lee Suan Yew and Gwee Ah Leng “The Diagnosis of brain

tumour” Singapore Medical Jounal, Vol 9, No 4 Dec 1968.

[53] Murphy, J P M.D. “The role of the EEG in the differential diag- nosis of brain tumor”, Southern Medical Jounrel, Vol 50, Issue 8,

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 8, August-2011 9

ISSN 2229-5518

Aug 1957

[54] Rosniwati Ghafar, Aini Hussain, Salina Abdul Samad and Nooritawati Md Tahir “Umace Filter for Detection of Abnormal Changes In Eeg: A Report of 6 Cases” World Applied Sciences Journal 5 (3): 295-301, 2008

[55] Robert Fisher, M.D., Ph.D “Overview of Epilepsy” Book of Com-

prehensive Epilepsy Center.

[56] G. Filligoi, M. Padalino, S. Pioli “ A Matlab software for detec- tion & counting of epileptic seizures in 72 hours Holter-EEG” Multidisciplinary Journals in Science and Technology, Journal of Se- lected Areas in Bioengineering (JSAB), January Edition, 2011

[57] Yuedong Song, Pietro Liò “A new approach for epileptic sei-

zure detection: sample entropy based feature extraction and ex-

treme learning machine” J. Biomedical Science and Engineering,

2010, 3, 556-567

[58] Marzia De Lucia, Juan Fritschy, Peter Dayan, David S. Holder “A novel method for automated classification of epileptiform activity in the human electroencephalogram-based on inde- pendent component analysis” International Federation for Medical

and Biological Engineering, 2007

IJSER © 2011

http://www.ijser.org