Machine Learning Approaches for Detecting Epileptic Seizures Based on Eeg Signals

Epilepsy is a prevalent neurological illness that affects an estimated 70 million people across the world. Seizures must be tracked if epilepsy is to be monitored and treated effectively. Caretaker seizure diaries are used for current epilepsy treatment; nevertheless, the use of clinical seizure detection may overlook episodes. In the long run, wearable gadgets may be more comfortable and easier to use for ambulatory monitoring. Custom-built machine learning (ML) algorithms are evaluated for the ability to identify seizures across a wide range of epileptic episodes using wrist and ankle-worn biosensors. In this paper, an automated technique a new wireless sensor based system is developed for detecting and monitoring epileptic patients in a non-clinical environment and to reduce the amount of time neurologists spend diagnosing seizures. Here a mechanism was developed for capturing multi-modal data such as Electroencephalogram (EEG) using wrist-worn biosensor. But exclusion of noise and feature extraction are two major encounters for forecasting of epileptic seizures. In order to detect epileptic seizure in a shorter period of time, statistical values and Lyapunov features were obtained, instead of raw data utilizing Support Vector Machine (SVM) as a classifier and obtained a substantial enhancement over the state-of-the-art methods.