Entropy features for focal EEG and non focal EEG

https://doi.org/10.1016/j.jocs.2018.02.002Get rights and content

Highlights

  • This article takes a new database for classification study of brain signal classification related to focal and non focal EEG.

  • A maximum accuracy of 99% is achieved.

  • This paves the way for real time identification of focal and non focal EEG signals, which is a landmark in brain signal processing.

Abstract

Electroencephalogram (EEG) is the recording of the electrical activity of the brain which can be used to identify different disease conditions. In the case of a partial epilepsy, some portions of the brain are affected and the EEG measured from that portions are called as Focal EEG (FEEG) and the EEG measured from other regions is termed as Non Focal EEG (NFEEG). The identification of FEEG assists the doctors in finding the epileptogenic focus and thereby they can plan for surgical removal of those portions of the brain. In this work, a classification methodology is proposed to classify FEEG and NFEEG. The Bern Barcelona database was considered and entropies such as Approximate entropy (ApEn), Sample entropy (SampEn) and Fuzzy entropy (FuzzyEn) as features which are fed into several classifiers. It was found that Non Nested Generalized Exemplers (NNge) classifier gave the highest classification accuracy of 99%, sensitivity of 99% and specificity of 99%, which is good comparing to proposed methods in the literature. In addition to the above, the maximum computation time of our features is 1.14 s which opens the window towards real time processing.

Introduction

Epilepsy is a neurological disorder in the present world today. This causes involuntary convulsion to the patient’s muscles and at times lead to loss of consciousness. In this world about 50 million people are living with epilepsy [[1], [2], [3], [4]].

Electroencephalogram (EEG) is widely used for various analysis of brain activity [[1], [2], [3], [5], [6]]. Epilepsy is clinically analysed using EEG. Some people with epilepsy become resistant to drugs and thus they need surgical removal of those parts of the brain which causes epilepsy to get rid of this disease. That portion of the brain which causes epileptic seizures is called as the epileptogenic foci. Such surgery is common in the present society now. The outcome of the surgery has successfully removed or significantly reduced the occurrence of the epileptic seizure in the patients [7]. Hence there is a precise need to find the exact region of the brain causing the epileptic seizures for surgical planning [8]. Presently locating the epileptogenic foci is being performed manually by the physician by clinical procedure which is subjective. This type of treatment is done in the case of partial epilepsy or partial seizures. In partial seizures, some portions of the brain are affected by the epileptic seizures and other portions are normal. In this context, FEEG is the EEG that is recorded from the brain areas where the first ictal EEG (seizure) changes were detected. And NFEEG is that EEG that is recorded from the brain areas that were not involved at the seizure onset [9].

Hence an automatic identification between these two signals – FEEG and NFEEG will assist doctors in identification of the epileptogenic foci for their surgical evaluation of the regions of the brain. In attempt to this, Bern Barcelona database [9] is considered and used it towards building algorithms for automatic classification of FEEG and NFEEG. In the literature, there are only a very few works who used this database for this purpose. The highest accuracy reported in the literature is 87% [10]). Here a simple method is presented in comparison to the existing methods that achieved the highest accuracy of 100%, 100% sensitivity and 100% specificity.

The paper is organized as follows. Section 2 presents the details of the data, brief information about the entropies used as features and the various classifiers which have used in this work. Section 3 presents the results obtained in this work. Section 4 presents a discussion on related studies of this database and compares our results with other methods and results in the literature. The conclusion is given in Section 5.

Section snippets

Methods and materials

This section describes the methods employed for automatic classification of FEEG and NFEEG.

Results

In this work, the features considered namely ApEn, SampEn, RE and FuzzyEn were applied directly to the raw EEG data from the database considered for the full length of samples i.e., 10, 240. Table 1 shows the mean and variance of these features calculated for 100 FEEG and 100 NFEEG data from the database. Table 1 also shows Student’s t-test p value for each of the features considered is very less than 0.0001 which shows that these features has strong discrimination for this dataset for

Discussion

Earlier also some of these entropy features have been applied to another famous database i.e., Bonn University database by few researchers [[1], [2], [3]] and they achieved accuracies as high as 98% and 99.7% [[1], [2], [3], [5]] for that database. In this work, a different combination of various entropy features was applied towards classification of FEEG and NFEEG for the Berlin Barcelona database and achieved the highest possible accuracy of 99%, sensitivity of 99% and specificity of 99%.

Conclusions

Epilespy is one of the common disorder of the population at large and a lot of people undergo surgical procedures in the brain to get rid of epilepsy − particularly for those who have drug resistant epilepsy. Locating the epileptogenic focus is one of the challenging task for the doctors which is very important as removing those portions of the brain helps the patient in totally recovering from the epilepsy disorder. In this work, a method using only simple entropy calculation measures is

N. Arunkumar is working as a faculty in SASTRA University since 2005. He specialization is in Biomedical Signal Processing. He has published several papers in International Peer reviewed journals. He has excelled in academics. His core area of research includes brain signals, clinical research and ayurvedic pulse signal processing.

References (37)

  • Y. Song et al.

    Automatic epileptic seizure detection in EEGs based on optimized sample entropy and extreme learning machine

    J. Neurosci. Methods

    (2012)
  • N. Kannathal et al.

    Entropies for detection of epilepsy in EEG

    Comput. Methods Programs Biomed.

    (2005)
  • S.-H. Lee et al.

    Measure of certainty with fuzzy entropy function

    Computational Intelligence

    (2006)
  • M. Czajkowski et al.

    Multi-test decision tree and its application to microarray data classification

    Artif. Intell. Med.

    (2014)
  • U.R. Acharya et al.

    Use of principal component analysis for automatic classification of epileptic EEG activities in wavelet framework

    Expert Syst. Appl.

    (2012)
  • U.R. Acharya et al.

    Application of non-linear and wavelet based features for the automated identification of epileptic EEG signals

    Int. J. Neural Syst.

    (2012)
  • R.G. Andrzejak et al.

    Nonrandomness, nonlinear dependence, and nonstationarity of electroencephalographic recordings from epilepsy patients

    Phys. Rev. E

    (2012)
  • R. Sharma et al.

    Application of entropy measures on intrinsic mode functions for the automated identification of focal electroencephalogram signals

    Entropy

    (2015)
  • Cited by (71)

    • Alcoholic EEG signals recognition based on phase space dynamic and geometrical features

      2022, Chaos, Solitons and Fractals
      Citation Excerpt :

      There can be numerous behavioral, emotional, and physiological effects in the individual due to excessive consumption of alcohol. Uncontrolled alcohol intake is responsible for liver cirrhosis, heart diseases, brain disorders, cancers, road accident deaths and injuries [3,4]. The early detection of alcoholism can prevent its irremediable consequences, such as disability and mortality [3,5].

    View all citing articles on Scopus

    N. Arunkumar is working as a faculty in SASTRA University since 2005. He specialization is in Biomedical Signal Processing. He has published several papers in International Peer reviewed journals. He has excelled in academics. His core area of research includes brain signals, clinical research and ayurvedic pulse signal processing.

    K. Ramkumar is working as faculty in SASTRA University for about 18 years. He research area is control systems, robotics and electric cars modeling. He has excelled in his academics and has been a pioneer in his research and published several research papers.

    V. Venkatraman is a faculty of department of mathematics in SASTRA University. He has about 17 years of experience and excelled in academics and research in 3D modeling of wave equations. He has published several research papers in his area and excelled in academics.

    View full text