A Hybrid CNN–LSTM and TCN Models for Automated Epileptic Seizure Detection from EEG Signals

Abstract

A need for automatic seizure detection system is pressing, as more than 65 million people worldwide suffer from Epileptic Seizures, which need to be treated and better cared for. Medical events, such as Epileptic Seizures, have special features in their Electroencephalography (EEG) that allow researchers to have a lot of confidence in them. In this research, the comparative performance of 2 types of Neural Network models for the detection of epileptic seizures in EEG is investigated, one is a CNN-LSTM hybrid model, and other is Temporal Convolutional Network (TCN). We performed experimental testing on UCI Epileptic Seizure Recognition dataset, and achieved excellent results with TCN (98.96%) and CNN-LSTM (98.74%). We compare our results with the most recent research in the field (2020-2025) indicating the shift in trends from traditional spatial-temporal models towards hybrid models and time-frequency based feature extraction and ensemble modelling. Our results provide important clinical insights, and provide a holistic standard for the advancement of seizure detection systems based on EEG.