Integrating Lemurs Optimizer with Ensemble Learning Based Epileptic Seizure Recognition Using EEG Signals

Epilepsy is another life-threatening and vital intellectual disorder where affected patients undergo regular seizures, making epileptic seizure recognition (ESR) crucial for timely diagnosis and treatment. However, there is a shortage of decision-making methods to forecast these diseases in their preliminary stages. Electroencephalography (EEG) is a standard medical procedure for seizure recognition, where the brain's electrical activity is recorded as a signal. Detecting the seizure on time is essential for all patients to deliver medication and prevent adverse effects. The visual inspection of the EEG signal takes sufficient time, and it is a difficult task that might result in lower performance. Developing an automated seizure-recognition technique can help inform neurologists and patients about seizure convulsions. Many automated seizure detection techniques are based on classical and deep learning (DL) models. This manuscript proposes a novel Lemurs Optimizer with Ensemble Deep Learning Epileptic Seizure Recognition (EOEDL-ESR) using EEG signals. The EOEDL-ESR technique utilizes an ensemble of three DL models to enhance the seizure prediction process. The EOEDLESR technique initially undergoes data pre-processing and a northern goshawk optimization (NGO)-based feature selection process. Furthermore, an ensemble of three classifiers, namely a convolutional autoencoder (CAE), an extreme learning machine (ELM), and a bidirectional long short-term memory (Bi-LSTM), is employed for classification. Moreover, the LO technique is utilized for the hyperparameter tuning process. An extensive experimental analysis of the EOEDL-ESR method is accomplished under the ESR dataset. The experimentation of the EOEDL-ESR method portrayed a superior accuracy value of 9 8. 5 5 % over existing models.