A Hybrid Deep Learning Model for Precise Epilepsy Detection and Seizure Prediction

Abstract

Manually classifying brain activity linked to epilepsy can be a lengthy, expensive process that varies depending on the observer. This study utilizes deep learning, particularly Convolutional Neural Networks (CNNs), to streamline this task using the Epileptic Seizures dataset. The CNN models are trained on spectrograms derived from brain signal plots, enabling the automatic detection of brain activity associated with epilepsy, such as generalized rhythmic delta activity, lateralized rhythmic delta activity, and epileptic seizures. This automated approach aims to lower diagnostic costs, reduce variability between observers, and lessen the manual workload, providing neurologists with a more reliable and efficient tool for diagnosing epilepsy. With an impressive accuracy of 97.11%, the proposed model shows its capability to effectively classify epilepsy-related brain activity. Additionally, this research includes a comparative analysis of different deep learning models, assessing their performance and appropriateness for automatic epilepsy detection. The insights gained from this analysis will contribute to the development of dependable classification systems, facilitating earlier diagnoses and enhancing patient outcomes.

Keywords: Epilepsy Detection, Image Processing, Machine Learning