Doğrusal ve doğrusal olmayan yöntemler ile epileptik aktivite tespiti

In this study, it was aimed to detect the seizures of epileptic patients whose seizures could not be prevented despite the use of high-dose medication. For this purpose, electroencephalography (EEG) recordings used in the diagnosis and treatment of epilepsy were analyzed. Linear and nonlinear analysis methods were investigated for EEG analysis. In the University of Bonn data for the time domain, the number of peaks crossing the threshold, and the power spectral density were investigated. However, the fact that healthy and epileptic data are in different files made the analysis insufficient. In CHB-MIT data, the mean of the signal and root mean square of the signal (RMS) from linear analysis methods and Shannon entropy, sample entropy, permutation entropy, approximate entropy, and spectral entropy from nonlinear signal analysis methods were calculated. In these calculations, the signal mean and RMS properties through the seizure period were determined with %58.4, %75 accuracies, respectively. In entropy methods, when considered respectively, the seizure period was determined with %75, %66.6, %66.6, %79.2 and %62.5 accuracy. In the detection of seizures in EEG signals, the accuracy has increased to %79.2, based on the increase in the sample entropy value or the decrease in the permutation entropy value. In addition, the accuracy value increased to %83.3 based on the decrease in the approximate entropy value or the decrease in the spectral entropy value. This is an indication that when the onset of a seizure cannot be detected with one entropy method, it can be detected with another entropy method. In addition, there are some changes detected before the seizure in the EEG signal analysis. It has been predicted that epileptic seizures can be detected beforehand if these changes are examined and detected by different analysis methods. Keywords: Epilepsy, Signal analysis, Signal mean value, Signal mean square root, Shannon entropy, Sample entropy, Permutation entropy, Approximate entropy, Spectral entropy.