Hybrid Dragonfly Optimization-Based Artificial Neural Network for the Recognition of Epilepsy

Abstract

Epilepsy can well be stated as a disorder of the central nervous systems (CNS) that brought about recurring seizures owing to chronic abnormal blasts of electrical discharge on the brain. Knowing if an individual is having a seizure and diagnosing the seizure type or epilepsy syndrome could be hard. Many methods were developed to recognize this disease. But the existing techniques for detection of epilepsy are not satisfied with accuracy, and cannot identify the diseases effectively. To trounce these drawbacks, this paper proposes an approach for the recognition of Epilepsy as of the electroencephalography (EEG) signals. This is implemented as follows. Primarily, the Kalman filter (KF) is utilized for pre-processing to eradicate the impulse noise present in the EEG signals. This filtered signal is then decomposed utilizing variable modes decomposition (VMD). Feature extraction (FE) is performed by computing 7 features. The dimensionality of this signal is then lessened using Modified-Principal Components Analysis (M-PCA). Finally, classification is conducted utilizing the artificial neural networks (ANN) that is optimized using the hybrid dragonfly algorithm (HDA). Disparate performance metrics such as sensitivity, accuracy, and false discovery rates (FDR) are ascertained and as well weighted against with the existent works.

Keywords

• Electroencephalography
• Kalman filter
• Variable mode decomposition
• Modified principal component analysis
• Artificial neural network
• Hybrid dragonfly algorithm