VARYASYONEL KIP AYRIŞIM YÖNTEMI İLE EEG SİNYALLERİNDEN EPİLEPSİNİN NÖBET ÖNCESİ TEŞHİSİ

Şentürk T., Latifoğlu F.

II. Uluslararası Bilimsel ve Mesleki Çalışmalar Kongresi, Nevşehir, Türkiye, 5 - 08 Temmuz 2018, ss.745-746

  • Yayın Türü: Bildiri / Özet Bildiri
  • Basıldığı Şehir: Nevşehir
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.745-746
  • İnönü Üniversitesi Adresli: Evet

Özet

Epilepsy is characterized by temporary and unexpected electrical deterioration in brain.

Electroencephalography (EEG) is widely used in diagnosis of epilepsy. There are many studies in the

literature to show the differences of epileptic EEG from normal EEG. However, before seizure, studies

on the prediction of epilepsy are very limited. In this study, EEG signals were analyzed using the

Variational Mode Decomposition Method (VMD) and it was aimed to determine the early phase of

epilepsy before the epileptic seizure. VMD has recently been proposed as an alternative to the

Empirical Mode Decomposition method (EMD). The VMD method can precisely separate

narrowband signals in frequency domain and can successfully apply many fields. In this study, VMD


II. International Scientific and Vocational Studies Congress

method was used in the analysis of EEG signals for detection of epilepsy before seizure, unlike

literature studies. In this proposed study, EEG signals were separated into 3 sub-bands using VMD

method. Statistical properties such as mean, variance and entropy of each subbands were analyzed. Six

epileptic EEG signals were examined in three parts: pre-seizure, post-seizure and post-seizure. Each

signal was subdivided by the VMD method. It is seemed that, in normal EEG signals, the variance and

entropy values of the first and second VMD subband signals were different from each other. That is,

one of the subbands has a higher variance while the other has a lower variance. However, a few

minutes before epilepsy and at the time of epilepsy, the situation was reversed. The variance of the

subband which was low in the previous case was significantly higher than the variance of the other

subband. In other words, before the epileptic seizure, statistical properties such as variance and

entropy of VMD subband signals were significantly changed. Based on these findings, it may be

possible to diagnose the epilepsy state before the seizures by analyzing the time-varying entropy and

variance values of the VMD subbands of the EEG signals. In this study, a pioneering approach has

been established in which epilepsy patients can acquire knowledge that they will have a seizure before

seizure.