Detection of Epileptic EEG Signal Using Wavelet Transform and Adaptive Neuro-Fuzzy Inference System

• Main Author: Khosropanah, Pegah
• Format: Thesis
• Language: English; English
• Published: 2011
• Online Access: http://psasir.upm.edu.my/id/eprint/20055/1/ITMA_2011_13_ir.pdf; http://psasir.upm.edu.my/id/eprint/20055/

Epilepsy is a chronic brain disorder that is characterized by abrupt discharge of neurons. Epilepsy has two main classes: generalized and focal epilepsy. In focal epilepsy source of the seizure within the brain is localized but in generalized epilepsy, it is distributed. About 1% of world populations suffer from epilepsy and one third of them have intractable seizure by medicine. Epileptics tolerate many difficulties due to seizure. Most of them also live in social seclusion. In addition, because of the medicine side effects and treatments, they may have troubles such as: double vision, fatigue, sleepiness, unsteadiness, as well as stomach upset. An effective treatment for epileptics in some rare cases with focal epilepsy (usually in median-temporal lobe) is by operation to separate a huge part of the brain tissue which has no essential function. Consequently, most of these patients need permanent care and treatment and 25% of them have to receive high dose of drugs and laboratory treatments. Therefore, diagnostic and warning algorithms for epilepsy infinite recognition, controlling seizure (to prepare for seizure e.g., pull over if driving) and organizing medicine schedule (to reduce unwanted side effects of not on time medication) will be useful. Such algorithms use brain electrical activity signals called electro encephalography (EEG) and have 2 methods of detection: visual (by specialist inspection) and automatic (by using signal processing knowledge). There are some problems faced by a neurologist in the inspection of long term EEG such as; being too time consuming, analytical precision requirement, similarity of epileptic spikes with artifacts like eye blinking, and too slight epileptic spikes nature to be detected in time domain. Proposing an automatic system to reduce time for epilepsy detection has been interesting field in recent decades. Most epilepsy types, even in inter-ictal (between two seizure) period, have transient signs in EEG called as spike and sharp waves (SSWs) that represent epilepsy disorder and its category. Most important signs are spikes. In this thesis an automated system has been developed to detect spikes from EEG to increase diagnosis speed, inspection precision and accuracy by applying some preprocessing such as filtering and artifact removing. Wavelet is applied as a feature extraction method and adaptive neuro-fuzzy inference system (ANFIS) is used for classification. Total accuracy of 97.5% has been obtained.