Improved Epileptic Seizure Identification Using Machine Learning Classifiers In Conjunction With EEG Signals

Abstract

The neurological condition known as epilepsy is marked by recurring seizures, making prompt diagnosis and treatment extremely difficult. Signals from the electroencephalogram (EEG) are essential for both diagnosing and treating epilepsy. This work suggests a novel method, HYRLD-Hybrid RNN(LSTM) with CNN(Densenet201) and Support Vector Machine that can identify and diagnose seizures in humans in order to get around these restrictions for improving the diagnosis of epileptic seizures by using machine learning classifiers on EEG signals. For the study EEG datasets are collected from various sources which are openly available on the internet. For feature selection, a hybrid PSO-Whale optimization method is employed.

Metrics including sensitivity (93.9%), specificity (93.1%), accuracy (98.38%) and precision (95.1%) are used to assess each classifier's performance on a dataset of epileptic patient recordings. Pre-processing raw EEG data to obtain pertinent features that capture the unique patterns linked to seizures is part of the methodology. The study shows how machine learning methods combined with EEG signals can improve the diagnosis of epileptic seizures and open up new possibilities for the creation of sophisticated diagnostic model and individualized treatment plans for the management of epilepsy.