Enhanced firefly optimizer with deep neural network for the detection of epileptic seizures using EEG signals

Currently, Electroencephalogram (EEG) is extensively used for diagnosing the epilepsy. The objective of this research is to investigate the changes in epilepsy frequency by proposing a new optimization based deep learning model. At first, the EEG recordings were acquired from two online databases; Bern Barcelona (BB), and Bonn University (BU). Then, Chebyshev type two filter was implemented to remove the unwanted artifacts from the acquired EEG signals. Further, Multivariate Variational Mode Decomposition (MVMD) methodology was applied to decompose the denoised EEG signals. The signal decomposition helps in finding the necessary information, which required to model the complex time series data. Then, the features were extracted from decomposed signals by using fifteen entropy, linear and statistical features. In addition, enhanced firefly optimization technique was proposed for optimizing the extracted features. In the enhanced firefly optimizer, a crossover operator of genetic algorithm was added for enhancing the local convergence rate that gives better classification. At last, the optimized feature vectors were classified by Deep Neural Network (DNN) that includes two circumstances (seizure and healthy), and (Interictal, ictal, and normal). From the experimental simulation, the proposed model improvement maximum of 1.4%, and 8.82% of accuracy in BU and BB EEG datasets, respectively related to the existing models.