Journal of North Khorasan

Background & Objectives: Seizures can be noted to the main symptom of epilepsy. Seizures prediction or early diagnosis for people reduces significantly injuries of epilepsy. The main problem that related to neurological disorders is an inability to timely prediction or the occurrence of seizures. Material and Method: EEG signals are Stochastic Process that can be treated as a sequence in time or in other words can be stated time series. In this study 300 epileptic patients categorized in three groups: normal, before and during the convulsive seizures were studied. Accordingly, after receiving data, they were preprocessed, then for Prediction time occurrence extracted special features by propose Algorithm. Eventually In order to final validate the cross-evaluation method (k-fold) has been used. Result: Firstly by wavelet transforms (WT), removed possible artifacts. In the next step by Binary Particle Swarm Optimization (BPSO) the characteristics (delay) are obtained. Then SVM algorithm (SVM) was performed to dimension reduction and manage the data (delay) so final Prediction that applied by Adaptive Nero Fuzzy Inference System Based on Optimal Delay. The final evaluation and final validation were done and the algorithm accurately in predicts by 2 units in delay approved. Conclusion: The Proposed System achieved a high accurate by interaction in introduced method. Despite the high accuracy, the present methods have a little ability in predicting seizure. Comparing the current methods indicate accuracy and high efficiency of the present approach.