| Summary: | Epilepsy is a common neurological disorder characterized by the recurrence of seizures, which can significantly impact the lives of patients. Electroencephalography (EEG) can provide important physiological information on human brain activity which can be useful to diagnose epilepsy. However, manual analysis and visual inspection of many EEG signals can be time-consuming and may lead to contradictory diagnoses by doctors. EEG signals play an important role in the diagnosis of epilepsy, as the quantification of cerebral signal anomalies may indicate the condition and the pathology of the cerebral signal. In this study, we attempted to develop a two-step process for the automated diagnosis of epilepsy using EEG signals. In the first step, we applied a low-pass filter and designed three intermediate filters for different frequency bands, employing multi-layer neural networks. In the second step, we used a wavelet transform method to process the data. The characteristics of the local brain are the distribution of epileptic EEG activity in the wavelet model across the whole brain surface. We also evaluated the use of two different classifiers, an artificial neural network (ANN) and a support vector machine (SVM), for the diagnosis of epilepsy. These classifiers were trained on normal and epileptic data and were able to accurately distinguish between normal and epilepsy as well as other conditions. We also found that the use of the wavelet transform did not significantly affect the classification performance but using a multi-layer neural network provided better precision. In this study, we developed a two-step automated process; incorporating low-pass filters, intermediate filters, multi-layer neural networks, and wavelet transform led to an accurate and efficient diagnosis of epilepsy. The results of this paper show high accuracy rates for both the artificial neural network (92.38%) and the support vector machine (95.5%) classifiers. Moreover, the study highlighted the effectiveness of utilizing a multi-layer neural network for improved precision. These findings contribute to the ongoing efforts in developing automated methods for epilepsy diagnosis, offering the potential for faster and more reliable detection techniques that can enhance patient care and outcomes.
|
|---|