Elsevier

IRBM

Volume 41, Issue 6, December 2020, Pages 331-353
IRBM

Original Article
Epileptic Seizure Detection Based on New Hybrid Models with Electroencephalogram Signals

https://doi.org/10.1016/j.irbm.2020.06.008Get rights and content

Highlights

  • In the method, EEG signals classification has been carried out.

  • The SVM models are Linear SVM, Cubic SVM, and Medium Gaussian SVM.

  • The best model is the combination of cubic SVM and MAD normalization.

Abstract

Objectives: Epileptic seizures are one of the most common diseases in society and difficult to detect. In this study, a new method was proposed to automatically detect and classify epileptic seizures from EEG (Electroencephalography) signals.

Methods: In the proposed method, EEG signals classification five-classes including the cases of eyes open, eyes closed, healthy, from the tumor region, an epileptic seizure, has been carried out by using the support vector machine (SVM) and the normalization methods comprising the z-score, minimum-maximum, and MAD normalizations. To classify the EEG signals, the support vector machine classifiers having different kernel functions, including Linear, Cubic, and Medium Gaussian, have been used. In order to evaluate the performance of the proposed hybrid models, the confusion matrix, ROC curves, and classification accuracy have been used. The used SVM models are Linear SVM, Cubic SVM, and Medium Gaussian SVM.

Results: Without the normalizations, the obtained classification accuracies are 76.90%, 82.40%, and 81.70% using Linear SVM, Cubic SVM, and Medium Gaussian SVM, respectively. After applying the z-score normalization to the multi-class EEG signals dataset, the obtained classification accuracies are 77.10%, 82.30%, and 81.70% using Linear SVM, Cubic SVM, and Medium Gaussian SVM, respectively. With the minimum-maximum normalization, the obtained classification accuracies are 77.20%, 82.40%, and 81.50% using Linear SVM, Cubic SVM, and Medium Gaussian SVM, respectively. Moreover, finally, after applying the MAD normalization to the multi-class EEG signals dataset, the obtained classification accuracies are 76.70%, 82.50%, and 81.40% using Linear SVM, Cubic SVM, and Medium Gaussian SVM, respectively.

Conclusion: The obtained results have shown that the best hybrid model is the combination of cubic SVM and MAD normalization in the classification of EEG signals classification five-classes.

Introduction

EEG (Electroencephalography) is the evaluation of brain electrical activity by recording on the computer using small metal electrodes placed on the scalp. Thanks to EEG, the electrical activity of the brain can be evaluated, and the answer to these questions can be found. In the recording of EEG signals, electrodes are placed on the scalp during application. The international system (10-20) method is used for electrode placement. Fig. 1 shows a demo repreparation of EEG recording to PC [1]. With the help of these electrodes, the signals received from the scalp are transmitted to the device and saved on the computer. The EEG device is a device that records the signals from the brain as in the ECG examination.

The normal electrical activity of the brain is disturbed in many cases, especially epilepsy (epilepsy). By evaluating the brain waves that make up the EEG, information about the location and shape of this disorder is obtained. Examination methods, which have been developed later than EEG, such as computed tomography (CT) and magnetic resonance imaging (MR), do not provide information about the electrical activity of the brain. EEG is the examination method that will decide treatment, especially in the diagnosis and types of epilepsy disease [1], [13], [14], [15], [30], [31], [32].

EEG disorder in the public is up to 30%. These data show the importance of clinical findings and seizure history in the relationship between epilepsy EEG discharge. There is a small chance of future seizures in these people. Changes in EEG are seen in brain infections such as encephalitis. EEG is illuminating in this regard, especially in patients with fever and impaired consciousness. In herpes encephalitis, specific features can be seen in EEG. Some changes in EEG also occur in patients with dementia (dementia). Fig. 2 denotes the epilepsy case in the brain with EEG signals [2]. Fig. 3 shows the EEG signal recording process. In order to diagnose epilepsy disease clinically, it is necessary to determine what type of seizure the patient has had. EEG takes the most crucial place in the examinations for the diagnosis of epilepsy. This examination, which controls the electrical activity of nerve cells in the brain, takes place by sticking the electrodes to the scalp and recording movements [2], [15], [17], [18], [27], [28], [29], [30].

In the literature, many works regarding the seizure epileptic diagnosis and detection from EEG signals using machine learning and deep learning. Some works have been given as follows. Mohammed Diykh et al. used the complex weighted networks to classify the EEG signals to diagnose epilepsy [4]. Also, they used some classification algorithms, including the least support vector machine, k-means, Naïve Bayes, and K-nearest in work [4]. In the work of Rubén San-Segundo et al. [5], the authors used deep learning for classification of epileptic EEG signals. In their method, for feature extraction, they used the convolutional neural network (CNN), and for the classification of EEG signals, three fully-connected layers CNN has been used. They obtained very good results, thanks to the proposed method in the classification of epileptic EEG signals [5]. Lasitha S. Vidyaratne et al. proposed a system to detect the seizure epileptics from EEG signals as online. They obtained 99.8% classification accuracy in the onset of epileptic detection using their method [6]. In the work of Ozan Kocadagli et al. [7], they proposed a different approach based on the combination of ANN (artificial neural network), wavelet transform, and fuzzy relations to classify the epileptic EEG signals and then achieved the promising results utilizing their method [7]. Yang Li et al. proposed a hybrid approach based on the combining of multiscale radial basis function (MRBF) networks and the Fisher vector (FV) encoding to detect the seizure epileptic in EEG signals, and then they obtained good results in their work [8]. As for the work of Ali Yener Mutlu, the Hilbert vibration decomposition has been used to decompose the EEG signals and then used the least squares support vector machine (LS-SVM) classification algorithm [9]. Ahnaf Rashik Hassan et al. proposed complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to classify the epileptic seizure EEG signals and then achieved very promising results in the detection of EEG signals by their method [10]. In the work of Hafeez Ullah Amin et al. [11], they proposed a hybrid model combining the discrete wavelet transform (DWT) and arithmetic coding to classify the EEG epileptic seizure signals. Poomipat Boonyakitanont et al. compared the various feature extraction, feature selection, and classification algorithms to detect the EEG epileptic signals [12]. Md. Faizul Bari et al. proposed a hybrid model combining the statistical and spectral features of max normalized intrinsic mode functions to detect the EEG epileptic seizure [16].

Apart from the literature works, we have done the classification of multi-class EEG epileptic signals having eyes open, eyes closed, healthy, from the tumor region, and epileptic seizures using the combinations of support vector machine (SVM) and normalization methods including z-score, minimum-maximum, and MAD.

The novelty and the contributions of this work are:

  • Three different SVM models have been used for multi-class EEG signals classification (five classes). In the literature, generally, two different EEG classes have been classified.

  • We have firstly combined the SVM and MAD normalization to classify the multi-class EEG signals classification with high performance.

The rest of this study is organized as follows: the details of the dataset and methods are given in Section 2. The experimental results and discussion are presented in Section 3. Lastly, concluding remarks are given in Section 4.

Section snippets

Dataset

In this paper, the Epileptic Seizure Recognition Data Set was used. This dataset is taken from the UCI machine learning repository [19], [20]. There are five classes EEG signals, including the eyes open, eyes closed, with tumor region, healthy brain, and epileptic seizure in the dataset. Also, there are 2300 samples for each class in the dataset. Totally, we have 11500 samples in the dataset, including all the classes. The EEG signals have been recorded and then digitized to the computer by an

Experimental results and discussion

In this paper, hybrid models combining the support vector machine classifiers and three different normalization methods, including z-score, minimum-maximum, and MAD, to classify the multi-class EEG epileptic signals automatically. In the training and testing of four different SVM models, including the linear, cubic, quadratic, and medium Gaussian kernel, the 5-fold cross-validation method has been used. Also, in order to evaluate the performance of the proposed methods, the confusion matrix,

Conclusions

In this paper, EEG signals classification five-classes including the cases of eyes open, eyes closed, healthy, from the tumor region, an epileptic seizure, has been conducted using the support vector machine (SVM) and the normalization methods comprising the z-score, minimum-maximum, and MAD normalizations. In order to classify the EEG signals, the support vector machine classifiers having different kernel functions including Linear, Cubic, and Medium Gaussian, have been used. Also, time,

Human and animal rights

This article does not contain any studies with human participants or animals performed by any of the authors. The used data in this paper is taken from the UCI Machine Learning repository. It is a public dataset.

CRediT authorship contribution statement

K. Polat: Formal analysis, Writing - original draft, Writing - review & editing. M. Nour: Formal analysis, Writing - original draft, Writing - review & editing.

Declaration of Competing Interest

The authors of the paper declare that they have no conflict of interest.

Acknowledgements

This project was funded by the Deanship of Science Research (DSR), at King Abdulaziz University, Jeddah, Saudi Arabia, under grant No. (D-661-135-1441). The authors, therefore, gratefully acknowledge DSR technical and financial support (Grant No. D-661-135-1441).

References (37)

Cited by (27)

View all citing articles on Scopus
View full text