0% found this document useful (0 votes)
44 views8 pages

A HighPerformance Method Based On Features Fusion of EEG Brain Signal and MRIImaging Data For Epilepsy Classification

The document presents a new method for classifying epilepsy using features fusion of EEG and MRI data. EEG data was analyzed in 1D using DCT and statistics, and MRI data was analyzed in 2D using DCT and mean. Features were selected and classified, achieving 96% accuracy for EEG, 94% for MRI, and 100% for fused EEG-MRI features.

Uploaded by

si_aymen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
44 views8 pages

A HighPerformance Method Based On Features Fusion of EEG Brain Signal and MRIImaging Data For Epilepsy Classification

The document presents a new method for classifying epilepsy using features fusion of EEG and MRI data. EEG data was analyzed in 1D using DCT and statistics, and MRI data was analyzed in 2D using DCT and mean. Features were selected and classified, achieving 96% accuracy for EEG, 94% for MRI, and 100% for fused EEG-MRI features.

Uploaded by

si_aymen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

MEASUREMENT SCIENCE REVIEW, 24, (2024), No.

1, 1-8

Journal homepage: https://content.sciendo.com

A High-Performance Method Based on Features Fusion of EEG


Brain Signal and MRI-Imaging Data for Epilepsy Classification
Fatma Demirezen Yağmur1*, Ahmet Sertbaş2
1
IT Department, Bolu Abant-Izzet Baysal University, Gölköy, 14030, Bolu, Turkey, [email protected],
2
Department of Computer Engineering, Faculty of Engineering, University of Istanbul-Cerrahpasa, Avcılar, 34320, Istanbul,
Turkey, [email protected]

Abstract: A 1-dimensional (1D) and 2-dimensional (2D) biomedical signal analysis based on the Discrete Cosine Transform (DCT) feature
extraction method was performed to diagnose epilepsy disorders with high accuracy. For this purpose, Electroencephalogram (EEG) data
were used for 1D signal analysis and Magnetic Resonance Imaging (MRI) data were used for 2D signal analysis. The feature vectors were
obtained by applying 1D DCT together with statistical methods such as mean, variance, standard deviation, kurtosis, and skewness for EEG
data and by applying 2D DCT together with the statistical method of mean for MRI data. The most useful features were selected by applying
Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Forward Selection and Backward Selection methods to the
obtained feature vectors. Using EEG stand-alone features, MRI stand-alone features and EEG-MRI fused features, the classification of
healthy and epileptic subjects was performed in the form of two clusters. The result of epilepsy classification in this work is 96% success of
1D EEG data by using the features selected by the PCA method, 94% success of 2D MRI data using the selected features by applying the
Forward Method, 100% classification accuracy of 1D EEG and 2D MRI datasets by LDA method using the obtained fused features . The
article shows that the fused features of EEG-MRI can be used very effectively for the diagnosis of epilepsy.

Keywords: Epilepsy, MRI, EEG, discrete cosine transform, medical image analysis, EEG-MRI classification.

1. INTRODUCTION kurtosis, and skewness values of the signal by applying the


Epilepsy is one of the most common neurological diseases DWT to the EEG data, created a separation tree and classified
of the brain. Numerous diagnostic tests are used to identify epileptic patients using the Probabilistic Neural Networks
the type of epileptic seizure [1]. These include algorithm [6]. Krishnaveni et al. separated the artifacts
Electroencephalogram (EEG), Computed Tomography (CT), contained in the EEG signals using the ICA method and
Magnetic Resonance Imaging (MRI) and Positron Emission classified them using the Polynomial Neural Network (PNN)
Tomography (PET). From 1982 to the present day, many and Advanced Propagation Neural Network methods [7].
studies have been carried out to diagnose epilepsy. Anderson MRI studies: Loyek et al. developed a model for focal
and Sijercic trained EEG data with a feed-forward neural prediction in detecting lesioned tissues and examined MRI
network using the Auto-Regressive (AR) model [2]. Kalaycı data from epilepsy patients using SVM for focal lesion
and Özdamar classified EEG data with Artificial Neural segmentation and tissue-based classification [8]. Clarke et al.
Networks (ANN) using the Wavelet Transform (WT) [3]. proposed a segmentation method using the Fuzzy C-Means
Subasi and Gursoy split EEG signals into subbands using the algorithm to identify abnormal MRI volumes [9]. Sujitha et
Discrete Wavelet Transform (DWT); they reduced the data al. modeled the prediction of epileptic seizures from MRI
size by applying Principal Component Analysis (PCA), using SVM [10]-[12].
Independent Component Analysis (ICA) and Linear The Discrete Cosine Transform (DCT) is a method for
Discriminant Analysis (LDA) methods and classified them converting the frequency of time series into fundamental
using Support Vector Machines (SVM) [4]. frequency components [11]. This method is often used for
Ogulata et al. have classified epileptic seizures with Multi- data compression. DCT was previously used by Martius et al.
Layer Perceptron (MLP) from EEG signals [5]. Gandhi et al. [11] for audio image compression and has by Orłowski [12]
calculated the energy, entropy, standard deviation, mean, for signal filtering. Sharif and Jafari obtained the feature

DOI: 10.2478/msr-2024-0001 *Corresponding author: [email protected] (F. DemirezenYagmur)

1
MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 1, 1-8

space by Genetic Algorithm (GA) and Shannon Entropy (SE)


calculation and then tested the sensitivity with Gaussian
RBF-SVM by selecting features with PCA [13].
Mutlu classified EEG signals with a least-squares support
vector machine for epileptic seizure detection based on
Hilbert Vibrational Discrimination (HVD) [14]. Kaleem et al.
present a novel patient-specific approach for seizure detection
using wavelet decomposition of multi-channel EEG data and
handcrafted features from the decomposed data [15]. Gupta
and Pachori proposed a new method for classification of
epileptic seizures based on Weighted Multiscale Renyi
Permutation Entropy (WMRPE) and rhythms obtained using
Fig. 1. The flow chart shows steps regarding EEG data analysis
the Fourier-Bessel Series Expansion (FBSE) of EEG signals
from EEG acquisition to data classification.
[16]. Serna et al. Taylor-Fourier classified seizures based on
EEG-band energy features using classifiers such as K-Nearest C. Raw data input and output
Neighbor (KNN) and the Least-Squares Support Vector The raw data was uploaded to the computer using the EEG
Machine (LS-SVM) [17]. reading program Nihon Kohden EEG-1000. Raw data
1-dimensional (1D) and 2-dimensional (2D) biomedical consists of both brain signals and artifacts, such as seizures,
signal analysis based on the DCT feature extraction method contractions, eye movements, swallowing, coughing, and
was performed to diagnose epilepsy with high-accuracy. For body movements. Fig. 2 shows the transfer of the data to the
the diagnosis of epilepsy, EEG data were used for 1D signal tool. This tool has been prepared to read data in .m00 format.
analysis and MRI data for 2D signal analysis. Subsequently,
a new approach based on the fusion of features from EEG-
MRI data was proposed as a high-accuracy solution to the
problem of epilepsy classification. The 2nd section of the
study highlights EEG analysis, the 3rd section highlights MRI
analysis and the 4th section highlights the new approach,
EEG-MRI Fusion of Features.

2. EEG ANALYSIS
The EEG is a test in which the electrical activity in the
brain is measured via electrodes attached to the scalp. The
EEG is used to diagnose certain diseases that affect the brain. Fig. 2. Nihon Kohden tool.
Epilepsy is the most common one. There are many steps from
D. Pre-processing
obtaining the EEG data to classifying it. Classification is
about whether the person has epilepsy based on the EEG data. The EEGLab14_1_2b program was used to examine and
pre-process the EEG data and remove artifacts. The data then
A. Materials went through many stages such as feature extraction,
selection and classification.
The Neurofax EEG-1000 was used to record the EEGs of
the patients diagnosed with epilepsy in the Neurology 1. Data reading
Department of Bolu Abant-Izzet Baysal University Hospital The plugin 1.1.0 from Nihon Kohden is added to EEGLab
in Bolu. In total, the EEG data of 88 epilepsy patients and 40 to solve this problem (due to the unsupported data format).
healthy individuals were recorded. The EEG cap has 16 The EEG data (.m00 format) was transferred to the system
channels. The recording takes an average of 18 minutes for via the EEGLab installed plugin. The 16-channel EEG signal
each patient. The computer system used for the application is received in the system was terminated in 1080 seconds with
a transfer rate of 200 Hz in a matrix of 16x21600, and the size
an Intel (R) Core (TM) i5-4210U CPU 1.70 GHz, x64 based
of the data set was determined to be 15.6 Mb.
processor.
2. Filtering
B. Method
When raw data is read, this data also contains unwanted
The EEG data is read out with the Nihon Kohden EEG- noise. A high-pass filter with a value between 0.5 – 1 Hz was
1000 fax program, the software of the EEG device. Fig. 1 applied. In the study, the unwanted noise contained the raw
shows the stages of the EEG from the analysis of the first raw data is removed by a high-pass filter when the data is read.
data input to the classification step. The noise is removed from the data using the filter.

2
MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 1, 1-8

3. Channel cleaning 2
𝑢 ≠ 0 𝑤ℎ𝑒𝑛 𝐶(𝑢) = √
As the filtered data was limited to the frequency values 𝑁

specified for each channel, the sections outside this limitation


were cleaned by removing the data. 2. Statistical feature extraction
Statistical methods such as mean, variance, standard
4. Data cleaning deviation, skewness and kurtosis were used for feature
When raw data is read, there are unwanted artifacts in this extraction. Various feature extractions were performed by
data. These artificial effects are the involuntary movements applying these methods to EEG signals, Chandaka et al. [21].
of the person during physical processes such as eye blinking,
eye movements, body movements, swallowing, coughing, a. Mean: is the basic statistical quantity and is calculated
contractions, seizures. To remove the noise in the channels, using the following equation, where i = 1,2,3,... and D is
you must first calculate the distances. In the EEGLab the signal.
program, this distance calculation is important for the interval 1
in which the artifact begins and ends. As a result of this 𝜇𝑖 = ∑𝑁
𝑗=1 𝐷𝑖𝑗 (3)
𝑁
calculation, the artifacts caused by movements on the
channels were made visible and deleted manually. b. Variance: is obtained by taking the standard deviation
squared.
5. Determination of channel locations
EEGLab has various channel layouts that differ in the 𝜗 = 𝜎2 (4)
number of channels of the EEG cap, e.g., 19, 25, 33, 47, 81.
c. Standard deviation: is the mean value of the EEG signal
In EEGLab, the 19-channel layouts that come closest to our
data are read into the system. Since the Fz, Cz and Pz channel and is calculated using the equation, where D is the
signal and N is the number of samples. µ is the square
data in the 19-channel system do not match the channel
root of the variance.
coordinates, the 16 channels are obtained by subtraction from
the standard channel layout.
1
𝜎 = √𝑁−1 ∑𝑁
𝑖=1(𝐷𝑖 − 𝜇)
2 (5)
E. Artifact cleaning
ICA is the most commonly used method for removing d. Kurtosis: measures the height of the Probability Density
artifacts that occur on single channels. The ICA method Function (PDF) of the time series.
enables better detection of artifacts in pre-processing. The
ICA method is also useful in detecting muscle artifacts in 𝐸(𝑥−𝜇)4
𝑘= (6)
disintegrated components, Delorme et al. [18]. 𝜎4

𝑋 = 𝐴𝑆 → 𝑆̂ = 𝑊𝑋 (1) e. Skewness: represents the PDF symmetry of the


amplitude of the time series.
By decoupling this unnecessary information with the ICA
𝐸(𝑥−𝜇)4
method, artificial factors in the data were reduced. S in (1) 𝑠= 𝜎4
(7)
denotes the original statistical source mark; A shows the
combined matrix of MxM. W, calculated by the inverse of the The statistical properties of the DCT coefficients were
matrix A, enables the separation of the signals. extracted. In the study, the 1x206900 matrix at time t is not
calculated for the entire signal of a channel, but the time is
F. Feature extraction subsampled by 4. Feature extraction was performed by
Feature extraction methods are needed to reduce data size. shifting the 60-second window over the EEG signal in 15-
second increments. If the calculations are performed at 4
1. 1D Discrete Cosine Transform (DCT) different time points for each channel, 64 features are
The raw data were pre-processed by filtering and channel extracted for 16 channels. For each statistical feature
cleaning, and then the feature was extracted by applying 1D extraction, 64 features are obtained.
DCT (2) to the 16-channel brain signal. In (2), u is the EEG
G. Feature selection
signal value and N is the signal length.
The DCT coefficients of the EEG data were extracted and
𝜋(2𝑥+1)𝑢 t (time) was divided into 4 parts, and the statistical
𝐹(𝑢) = 𝐶(𝑢) ∑𝑁−1
𝑥=0 𝑓(𝑥) cos [ 2𝑁
] (2)
calculations were performed in separate sections. As a result
𝑈𝑝 𝑡𝑜 𝑢 = 0,1, … , 𝑁 − 1 of each statistical calculation, 64 features were extracted. As
one feature, a total of 320 features were obtained from a total
of 1 sample, including 64 means, 64 variances, 64 standard
1 deviations, 64 skewness values, and 64 kurtosis values. The
𝑢 = 0 𝑤ℎ𝑒𝑛 𝐶(𝑢) = √
𝑁 algorithms PCA, LDA [4], Forward Selection and Backward
Selection were implemented. PCA helps to manage high-

3
MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 1, 1-8

dimensional data sets by extracting essential information and 3. MRI ANALYSIS


discarding less relevant features, which simplifies the MRI is one of the most useful imaging techniques for
analysis. PCA performs a linear transformation of the data
detecting a symptom that causes epilepsy in the brain.
and searches for directions with maximum variance. The
principal components are ranked according to the variance
A. Materials
they explain, enabling effective feature selection. LDA is
used to find a linear combination of features that characterizes MRI data were obtained from the Radiology Department
or separates two or more classes of a categorical variable. of Bolu Abant-Izzet Baysal University Hospital. The MRI
Forward Selection is an iterative method where we start with images acquired with the 1.5 Siemens Magnetom Symphony
having no feature in the model. In each iteration, we add the scanner included 88 MRIs of epilepsy patients and 40 of
feature that best improves our model until adding a new
healthy individuals. A total of 115 t2_tse_tra_512 MRI slices
variable no longer improves the performance of the model. In
backward elimination, we start with all features and at each were used for the epilepsy classification method.
iteration we remove the least significant feature that improves
the model’s performance. We repeat this until no more B. Method
improvement can be observed by removing features. The steps of the MRI data from the first reading section to
the classification step are shown in Fig. 3.
H. Classification
In Tanagra (data mining tool), classification was
performed by using the Tutorial Learning Methods ANN,
SVM and KNN. The purpose of classification is whether the
EEG signal indicates epilepsy or health. The classification
results were compared with the results obtained using each of
the feature selection methods. The EEG signal was used as
input for classification after undergoing certain pre-
processing, feature extraction and selection. The ANN
architecture was developed using middleware. The network Fig. 3. Flow chart of EEG data.
system consists of an information input (64 features), a
hidden layer (number of neurons 20) and an output layer C. MRI reading
(healthy-epileptic). SVM is a classifier to create hyperplanes
that maximize the distance to the nearest training set. The MRI data in Dicom format is read in Matlab. The MRI
kernel method was set as Radial Based Function (RBF) and consists of a 384x306 matrix.
our model was created. KNN is a simple model that assigns a
feature vector to a class relative to its nearest neighbor. The D. Segmentation
Euclidean distance measure was used to calculate the distance The purpose of image segmentation is to summarize pixels
between the input data point and all training data. 70% of the
according to image regions, separate surfaces, or objects. The
dataset was used for training, 30% was also used for testing.
A 9-layer cross-validation method was applied to the data. In Region-Based segmentation method was used in the study.
the architecture thus prepared, the data is trained separately This method is based on the logic that neighboring pixels are
according to the data results in different feature selections and clustered according to common density values. The purpose
the results are shown in Table 1. of this algorithm is to group the regions anatomically and
according to their functions, Gumaste and Jadhav [23]. Fig. 4
Table 1. Different classification algorithms' overall accuracy
according to feature selection methods of EEG data. and Fig. 5 show the original and the segmented MRI.

Method ANN SVM KNN


PCA 95% 81% 81%
LDA 92% 83% 84%
Forward Selection 86% 81% 83%
Backward Selection 91% 81% 82%

Consequently, it was found that selected feature vectors


using PCA increased the percentage of accuracy in
classification. The high success rate of classification using
ANN is shown in Table 1. It is evident form this study that
the classifications performed in the literature with ANN
have a higher success rate compared to other classification
algorithms, Kumar et al. [22]. Fig. 4. Original MRI.

4
MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 1, 1-8

3. 2D DCT mean
This method consists of using the aforementioned two
methods as a hybrid. The size of the obtained DCT
coefficients is 240x240 pixels. The pixels are divided
vertically and horizontally into eight equal, small parts. This
results in 64 subsections. The size of each subsection is 30x30
pixels. The arithmetic mean is calculated for each subsection.
Thus, the size of the new feature vector is 64 as described by
Demirezen-Yagmur and Sertbas [24].

G. Feature selection
The feature vector extracted using the DCT_mean method
is 64. The PCA, LDA, Forward Selection and Backward
Fig. 5. Segmentation [24].
Selection algorithms were applied to select the most
E. Pre-processing compelling feature.
Certain pre-processing was performed on the segmented H. Classification
MRI data. The matrices obtained as a result of the
The classification settings for EEG data also apply to MRI
segmentation are transformed to 240x240 and resized.
data. The data was trained separately according to the results
Dimensional MRI matrices are values ranging from 0 to 999,
of the different feature selections and the results are shown in
which were then converted to normalized values between 0
Table 2.
and 1.
Table 2. Different classification algorithm validation results
F. Feature extraction according to the feature selection methods of the MRI data.
In pattern recognition, feature extraction methods are
Method ANN SVM KNN
needed to reduce the size of the pattern. These methods are
mathematical functions that perform feature extraction and PCA 90% 81% 81%
selection. New features were extracted from MRI images by LDA 81% 81% 82%
the hybrid use of DCT and Arithmetic Mean methods, the Forward Selection 81% 81% 83%
transformed structural feature extraction methods known in
Backward Selection 81% 81% 81%
the literature, Meyer-Bäse [25].

1. 2D Discrete Cosine Transform (DCT) The success of the feature vectors selected by the PCA
method in classification is therefore high. The added metric
DCT is particularly useful for easy and fast feature
was needed because other recall and precision metric values
extraction from large data sets. It also plays a vital role in
were very close to each other. If such metric values are
image compression. After pre-processing, segmentation and
critical, the F1-score metric result is also examined. The PCA
normalization of the MRI data, the 2D (3) was applied and
method Backward Selection is shown in Table 5.
feature extraction was performed. In (8), u, v is the value of
the MRI matrix and N, M is the length of the matrix. 4. EEG-MRI FEATURES FUSION APPROACH
𝑁−1 𝑀−1
𝜋(2𝑥 + 1)𝑢 𝜋(2𝑦 + 1)𝑣 A new set of features is created by combining biomedical
𝐹(𝑢, 𝑣) = 𝐶(𝑢)𝐶(𝑣) ∑ ∑ 𝑓(𝑥, 𝑦) cos [ ] cos [ ] signals of different sizes. This feature vector consists of a
2𝑁 2𝑁
𝑥=0 𝑦=0
combination of 1D EEG and 2D MRI features of one and the
𝑈𝑝 𝑡𝑜 𝑢 = 0,1, … 𝑁 − 1 𝑣 = 0,1, … 𝑀 − 1 (8) same person.

A. Material
1 After applying the 1D DCT with Mean method to EEG
𝑢, 𝑣 = 0 𝑤ℎ𝑒𝑛 𝐶(𝑢), 𝐶(𝑣) = √
𝑁 data, 64 features were obtained. Similarly, 64 features were
obtained after applying the Mean method with 2D DCT on
MRI data. In the application, 1D and 2D biomedical signals
2
𝑢, 𝑣 ≠ 0 𝑤ℎ𝑒𝑛 𝐶(𝑢), 𝐶(𝑣) = √ of different sizes are fused, and a total of 128 feature data are
𝑁 obtained. A vector of 128 data features contains both EEG
and MRI features. In our study, a 128-feature data set was
2. Arithmetic mean obtained form 128 samples.
The mean is calculated by dividing the sum of all elements
in the series by the number of elements in the series. The B. Method
arithmetic mean is also referred to as the "mean" by The combination of EEG and MRI data, the feature
Abramowitz and Stegun [26]. selection and the classification flowchart are shown in Fig. 6.

5
MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 1, 1-8

The second important factor in the diagnosis of epilepsy is


the 2D MRI data of the patients. This data is used to find
conditions that cause epilepsy in the brain. In our study, the
MRI data was segmented, certain pre-processing was
performed and classified as Epileptic/Healthy. The 2D MRI
data was selected using different feature selection methods
shown in Table 4. Among these methods, the most successful
classification accuracy of 90% was achieved by using the
features selected by the PCA method. The F1-score metric
(Table 5), the other analysis of classification estimation, was
more successful with ANN.

Table 5. F1-score results.


Fig. 6. EEG-MRI combination flow.
Method ANN SVM KNN
C. Feature selection EEG 96% 90% 96%
The feature vector formed using the combination of EEG MRI 96% 89% 90%
and MRI is 128. The PCA, LDA, Forward Selection and EEG-MRI 100% 94% 90%
Backward Selection algorithms are used to increase the
accuracy of these feature vectors.
In our study, it was claimed that the combination of EEG
D. Classification and MRI features is more effective in the diagnosis of
The classification settings for EEG and MRI also apply to epilepsy. As can be seen from the results of our application,
the EEG-MRI combination. The data was trained separately 100% accuracy (Table 4) and F1-score (Table 5) were
according to the results of the different feature selections and obtained by applying LDA to the data set obtained from the
the results are shown in Table 3. EEG-MRI feature fusions. As shown in the results of our
application, by applying LDA (to the data set obtained from
Table 3. Different classification algorithm validation results the EEG-MRI feature fusions), 100% validation of the
according to feature selection methods of EEG-MRI feature fusion. selected features (Table 4) and F1-score (Table 5) were
obtained. Many studies have been carried out in the literature
Method ANN SVM KNN on the use of EEG data in the diagnosis of epilepsy. The most
PCA 96% 87% 84% recent studies are summarized and compared in Table 6.
LDA 100% 81% 84%
Forward Selection 99% 83% 85% Table 6. Classification comparison of data in recent years.
Backward Selection 97% 90% 83% Researcher Feature Classifier Result
Extraction
5. CONCLUSION AND DISCUSSION Sharif and Jafari [13] GA-SE RBF-SVM 91.8%
In this article, high-precision diagnosis of epilepsy is Mutlu [14] HVD* LS-SVM 97.6%
realized by 1D and 2D biomedical signal analysis based on Kaleem et al. [15] DWT SVM, KNN 99.6%
the DCT feature extraction method.
In the study, artificial factors were removed by performing Gupta et al. [27] DCT SVM 97.7%
certain pre-processing on the EEG data obtained from Bolu Gupta and Pachori [16] WMRP* LS-SVM, RF* 97.3%
Abant-Izzet Baysal University Faculty of Medicine Serna et al. [17] TF* LS-SVM 96.7%
Department of Neurology, and the data was classified using Deivasigamani [28] CWT* ANFIS* 96.7%
various algorithms. The 1D EEG data were selected using
Current study DCT ANN, SVM, 100%
various feature selection methods. Among the methods listed
KNN
in Table 4, the most successful classification accuracy of 95% Note: * RF: Random Forest, TF: Taylor Fourier, CWT: Continuous Wavelet
was obtained by ANN using the features selected by PCA. In Transform, ANFIS: Adaptive Neuro Fuzzy Inference System.
the results of the F1-score metric (Table 5), the other analysis
of classification estimation, the success of the ANN and the 6. FINAL RESULT
KNN method was higher. Various feature extraction methods and classifiers were
Table 4. Accuracy results. used in the studies. This resulted in high performance. In the
study with DCT, the highest success rate so far was 97.7%;
Method ANN SVM KNN however, in our study on EEG_MRI fusion, a higher success
EEG 95% 83% 84% rate of 100% was achieved. We plan to continue our work by
MRI 90% 81% 83% expanding the data set and applying deep learning methods.
We therefore believe that our study with the combination
EEG-MRI 100% 90% 85% method can greatly contribute to the diagnosis of epilepsy.

6
MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 1, 1-8

ACKNOWLEDGMENT [11] Martisius, I., Birvinskas, D., Jusas, V., Tamosevicius,


All procedures performed in studies with human Z. (2011). A 2-D DCT hardware codec based on
participants comply with the ethical standards of the Loeffler algorithm. Electronics and Electrical
institutional and/or national research board (Bolu Abant-Izzet Engineering, 113 (7), 47-50.
Baysal University Ethics Committee Ethics Approval http://dx.doi.org/10.5755/j01.eee.113.7.611
Certificate dated 25/05/2017 and numbered 2017/45). We [12] Orlowski, P. (2010). Simplified design of low-pass
thank the Neurology and Radiology Department for their linear parameter-varying, finite impulse response
support in data collection. filters. Information Technology and Control, 39 (2),
130-137.
REFERENCES https://itc.ktu.lt/index.php/ITC/article/view/12301
[1] World Health Organization. (2005). Atlas: Epilepsy [13] Sharif, B., Jafari, A. H. (2017). Prediction of epileptic
Care in the World. ISBN 92-4-156303-6. seizures from EEG using analysis of ictal rules on
[2] Anderson, C. W., Sijercic, Z. (1996). Classification of Poincaré plane. Computer Methods and Programs in
EEG signals from four subjects during five mental Biomedicine, 145, 11-22.
tasks. In Proceedings International Conference on https://doi.org/10.1016/j.cmpb.2017.04.001
Engineering Applications of Neural Networks [14] Mutlu, A. Y. (2018). Detection of epileptic
(EANN'96). Systems Engineering Association, 407- dysfunctions in EEG signals using Hilbert vibration
414. decomposition. Biomedical Signal Processing and
[3] Kalaycı, T., Özdamar, O. (1995). Wavelet Control, 40, 33-40.
preprocessing for automated neural network detection https://doi.org/10.1016/j.bspc.2017.08.023
of EEG spikes. IEEE Engineering in Medicine and [15] Kaleem, M., Guergachi, A., Krishnan, S. (2018).
Biology Magazine, 14 (2), 160-166. Patient-specific seizure detection in long-term EEG
https://doi.org/10.1109/51.376754 using wavelet decomposition. Biomedical Signal
[4] Subasi A., Ismail Gursoy, M. (2010). EEG signal Processing and Control, 46, 157-165.
classification using PCA, ICA, LDA and support vector https://doi.org/10.1016/j.bspc.2018.07.006
machines. Expert Systems with Applications, 37 (12), [16] Gupta, V., Pachori, R. B. (2019). Epileptic seizure
8659-8666. https://doi.org/10.1016/j.eswa.2010.06.065 identification using entropy of FBSE based EEG
[5] Oğulata, S. N., Şahin, C., Erol, R. (2009). Neural rhythms. Biomedical Signal Processing and Control,
network-based computer-aided diagnosis in 53, 101569.
classification of primary generalized epilepsy by EEG
https://doi.org/10.1016/j.bspc.2019.101569
signals. Journal of Medical Systems, 33 (2), 107-112.
[17] de la O Serna, J. A., Paternina, M. R. A., Zamora-
https://doi.org/10.1007/s10916-008-9170-8
[6] Gandhi, T. K., Chakraborty, P., Roy, G. G., Panigrahi, Méndez, A., Tripathy, R. K., Pachori, R. B. (2020).
B. K. (2012). Discrete harmony search based expert EEG-rrhythm specific Taylor–Fourier filter bank
model for epileptic seizure detection in implemented with O-splines for the detection of
electroencephalography. Expert Systems with epilepsy using EEG signals. IEEE Sensors Journal, 20
Applications, 39 (4), 4055-4062. (12), 6542-6551.
https://doi.org/10.1016/j.eswa.2011.09.093 https://doi.org/10.1109/JSEN.2020.2976519
[7] Krishnaveni, V, Jayaraman, S., Gunasekaran, A., [18] Delorme, A., Sejnowski, T., Makeig, S. (2007).
Ramadoss, K. (2008). Automatic removal of ocular Enhanced detection of artifacts in EEG data using
artifacts using JADE algorithm and neural network. higher-order statistics and independent component
International Journal of Information, Control and analysis. Neuroimage, 34 (4), 1443-1449.
Computer Sciences, 1.0 (4). https://doi.org/10.1016/j.neuroimage.2006.11.004
https://doi.org/10.5281/zenodo.1333386 [19] Krishnaveni, V., Jayaraman, S., Manoj Kumar, P. M.,
[8] Loyek, C., Woermann, F. G., Nattkemper, T. W. Shivakumar, K., Ramadoss, K. (2005). Comparison of
(2008). Detection of focal cortical dysplasia lesions in independent component analysis algorithms for
MRI using textural features. In Bildverarbeitung für die removal of ocular artifacts from electroencephalogram.
Medizin 2008. Springer, 432-436. Measurement Science Review, 5 (2), 67-78.
https://doi.org/10.1007/978-3-540-78640-5_87 https://www.measurement.sk/2005/S2/krishnaveni.pdf
[9] Clark, M. C., Hall, L. O., Goldgof, D. B., Clarke, L.P., [20] Demirezen Yağmur, F., Sertbaş, A. (2020). Automatic
Velthuizen, R. P., Silbiger, M. S. (1994). MRI diagnosis of epilepsy from EEG signals using Discrete
segmentation using fuzzy clustering techniques. IEEE
Cosine Transform. In 2020 28th Signal Processing and
Engineering in Medicine and Biology Magazine, 13 (5),
Communications Applications Conference (SIU). IEEE.
730-742. https://doi.org/10.1109/51.334636
[10] Sujitha, V., Sivagami, P., Vijaya, M. S. (2011). https://doi.org/10.1109/SIU49456.2020.9302300
Predicting epileptic seizure from MRI using fast single [21] Chandaka, S., Chatterjee, A., Munshi, S. (2009). Cross-
shot proximal support vector machine. In ICWET '11: correlation aided support vector machine classifier for
Proceedings of the International Conference & classification of EEG signals. Expert Systems with
Workshop on Emerging Trends in Technology. ACM, Applications, 36 (2), 1329-1336.
525-529. https://doi.org/10.1145/1980022.1980135 https://doi.org/10.1016/j.eswa.2007.11.017

7
MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 1, 1-8

[22] Kumar, Y., Dewal, M. L., Anand, R. S. (2014). [26] Abramowitz, M., Stegun, I. A. (1972). Handbook of
Epileptic seizures detection in EEG using DWT-based Mathematical Functions with Formulas, Graphs, and
ApEn and artificial neural network. Signal, Image and Mathematical Tables; 9th printing. Dover Publications,
Video Processing, 8 (7), 1323-1334. ISBN 0-486-61272-4.
https://doi.org/10.1007/s11760-012-0362-9 [27] Gupta, A., Singh, P., Karlekar, M. (2018). A novel
[23] Gumaste, P. P., Jadhav, D. V. (2015). Image signal modeling approach for classification of seizure
segmentation techniques for brain MRI images: A and seizure-free EEG signals. IEEE Transactions on
survey. International Journal of Modern Trends in Neural Systems and Rehabilitation Engineering, 26 (5),
Engineering and Research (IJMTER), 2 (7). 925-935.
[24] Demirezen Yağmur, F., Sertbaş, A. (2018). Diagnosis https://doi.org/10.1109/TNSRE.2018.2818123
of epilepsy in brain MR images using DCT-means. In [28] Deivasigamani, S., Senthilpari, C., Yong, W. H. (2021).
2018 26th Signal Processing and Communications Machine learning method based detection and diagnosis
Applications Conference (SIU). IEEE, 2439-2443. for epilepsy in EEG signal. Journal of Ambient
https://doi.org/10.1109/SIU.2018.8404758 Intelligence and Humanized Computing, 12, 4215-
[25] Meyer-Bäse, A. (2004). Pattern Recognition in Medical 4221. https://doi.org/10.1007/s12652-020-01816-3
Imaging. Academic Press, 22-35.
https://doi.org/10.1016/B978-0-12-493290-6.X5000-7
Received May 10, 2023
Accepted January 09, 2024

You might also like