A Method for Epileptic Seizure Detection in EEG Signals Based on Tunable Q-Factor Wavelet Transform Method Using Grasshopper Optimization Algorithm With Support Vector Machine Classifier

Background: Epilepsy is a Brain disorder disease that affects people's quality of life. If it is diagnosed at an early stage, it will not be spread. Electroencephalography (EEG) signals are used to diagnose epileptic seizures. However, this screening system cannot diagnose epileptic seizure states precisely. Nevertheless, with the help of computer-aided diagnosis systems (CADS), neurologists can diagnose epileptic seizure stages properly. Objective:The aims of this study are to epileptic seizures diagnosis by using EEG signals and distinguish its different stages. Using various statistical and non-linear features, the CADS proposed in this study is capable of diagnosing epilepsy seizures precisely and quickly. Therefore, this system can help the neurologists to diagnose more accurately. Material and methods: This paper focuses on a novel method for epileptic seizure detection based on EEG signals using artificial intelligence (AI) techniques. First, the Bonn dataset is used for trials, and the EEG signals are divided into 5-second intervals. Then, tunable q-factor wavelet transform (TQWT) was utilized to analyze EEG signals into various sub-bands. Several statistical and nonlinear features (fractal dimensions (FDs) and entropies) of TQWT sub-bands are extracted in the following sections. In the next procedure, the Autoencoder (AE) method with proposed layers is applied to reduce features, and finally, different classification algorithms such as Support Vector Machine (SVM), SVM with grasshopper optimization algorithm (SVM-GOA), K-Nearest Neighbors (KNN), and Random Forest (RF) are employed. The employment of AE for feature reduction and SVM-GOA for classification is the novelty of this study. Results: According to results, the proposed method of epileptic seizure detection demonstrated better performance compared to related works. The proposed SVM-GOA classification method has a higher accuracy rate as much as 99.42% and 99.23% for two-class and multi-class classification problems, respectively. Conclusion: The combination of effective features in diagnosing periods of epileptic seizure along with appropriate classification methods increases the accuracy of the CADS. Considering the importance of diagnosing various epileptic seizures, a high-precision CADS is introduced in this work. High accuracy, using different methods for extracting features and classification are among the advantages of our proposed method.