DETECTION OF PRE-EPILEPTIC SEIZURE BY USING WAVELET DECOMPOSITION AND ARTIFICIAL NEURAL NETWORKS

Epilepsy is one of the most frequent disease all over the world and can give valuable information about the functional situation of the brain. This information is clear when EEG signals are collected and analyzed in an appropriate, accurate and reliable way and they make detection of epileptic seizures is possible. It is difficult to interfere with epileptic seizures because they occur suddenly and randomly thus, detection of epilepsy in the early stage is vital for both patients and medical officials. In this study, firstly both 100 interictal (EEG signals which occur between two epileptic seizures) and 100 healthy EEG signals are normalized in order to detect the pre-epileptic seizure. After the normalization process, each EEG signals are decomposed by performing Wavelet Decomposition (WD) and classified with ANN (Artificial Neural Networks). 4-fold cross-validation technique is used to evaluate the classification performance. LogEn entropy is used to extract features from both interictal and healthy EEG signals and classification accuracy is achieved 97,5%.