A Lightweight Deep Learning Model for The Early Detection of Epilepsy

Abstract

Epilepsy is a neurological disorder and non communicable disease which affects patient's health, During this seizure occurrence normal brain function activity will be interrupted. It may happen anywhere and anytime so it leads to very dangerous problems like sudden unexpected death. Worldwide seizure affected people are around 65% million. So it must be considered as serious problem for the early prediction. A number of different types of screening tests will be conducted to assess the severity of the symptoms such as EEG,MRI, ECG, and ECG. There are several reasons why EEG signals are used, including their affordability, portability, and ability to display. The proposed model used bench-marked CHB-MIT EEG datasets for the implementation of early prediction of epilepsy ensures its seriousness and leads to perfect diagnosis. Researchers proposed Various ML /DL methods to try for the early prediction of epilepsy but still it has some challenges in terms of efficiency and precision Seizure detection techniques typically employ the use of convolutional neural networks (CNN) and a bidirectional short- and long-term memory (Bi-LSTM) model in the realm of deep learning. This method leverages the strengths of both models to effectively analyze electroencephalogram (EEG) data and detect seizure patterns. These light weight models have been found to be effective in automatically detecting seizures in deep learning techniques with an accuracy rate of up to 96.87%. Hence, this system has the potential to be utilized for categorizing other types of physiological signals too, but additional research is required to confirm this.