Enhanced Epilepsy Sensitivity and Detection Rate With Improved Specificity by Integration of Modified LSTM Networks

Abstract

Epileptic seizures represent a critical concern in mental health, arising from abnormal synchronization and rapid neuronal activity in the brain. The consequences of such seizures, including loss of consciousness and cognitive impairment, have significant psychological, social, and cognitive implications. Electroencephalography (EEG) and sensor data play a pivotal role in epilepsy detection, leveraging machine learning techniques to analyze extensive datasets. This article presents a novel approach that combines long short-term memory (LSTM) networks and wavelet transform (WT) techniques to enhance seizure detection accuracy. The proposed multimodule approach, featuring residual neural networks and convolutional neural networks (CNNs), coupled with k-fold validation, achieves an accuracy of 98.5%, sensitivity of 99.0%, specificity of 98.0% and an AUC-receiver operating characteristic (ROC) value of 0.95 for classifying interictal epileptiform discharges. This underscores the model’s efficacy in addressing the complexities of EEG signal analysis.