Abstract:

The abnormal activities of the human brain are recorded by the Electroencephalogram (EEG) signal. The EEG signal has a wide range of applications in the healthcare field. During the diagnosis of any disease, theEEG signal plays an important role. In this work, we discussed the epilepsy seizure detection methods from the EEG signal dataset. Various machine learning approaches are used to detect epileptic seizures from the recorded EEG signal dataset. The three main steps followed for the epilepsy seizure detection data availability, pre-processing, and features learning and classification. The epilepsy seizure EEG signal shows the interrupted EEG signal means brain disorder functionality is demonstrated via the EEG signal. The seizure detection is performed for medical treatment. The pre-processing and features extraction step affected the seizure detection process execution time and prediction accuracy. State of the art methods use the Bonn university EEG dataset and CHB-MIT dataset for the classification of epileptic signal from recorded EEG signal. The machine learning classifiers provided improvement in the classification accuracy and effectively utilized. Some deep learning methods like CNN are also used, which provided better classification accuracy than the machine learning approach. After studies various machine and deep learning approaches, we get an idea to design a Bayesian Optimization-based Long Short Term Memory (LSTM) model for epilepsy seizure detection.

Keywords: