Elliptic Seizure Detection on EEG Signals Using Bidirectional Long Short-Term Memory Model

Many screening methods have been analyzed by employing the modalities of Magnetic Resonance Imaging (MRI) and Electroencephalography (EEG) to diagnose epileptic seizures. Deep Learning (DL) is a branch of artificial intelligence, which is a broad field. Conventional DL could only perform as well as people who made the features by hand. On the other hand, DL fully automates both feature extraction and categorization. EEG is an often employed and noteworthy method for assisting in the diagnosis of epilepsy and examining the movement of the human brain. Since EEG signals are non-stationary, a patient's seizure patterns will change during various recording sessions. This study uses a Bidirectional Long Short-Term Memory (Bi-LSTM) model to detect seizures and brain tumours. Preprocessing the EEG signal, extracting preictal features, hyper-optimization using Mayfly Optimization (MO). Long-term EEG recordings from the EEG and ABIDE fMRI datasets are used in the evaluation. Many returned features are used by the Bi-LSTM model, such as frequency domain and temporal information between extracted EEG channels prior to classification. The proposed Bi-LSTM accomplishes a 99.12% accuracy rate, 98.25% precision, 99.06% recall, and 98.53% f1-measure which is superior when related to existing models.