What type of study is this?

September 10, 2025

A Hybrid Machine Learning and Deep Learning Framework for Real-Time Epileptic Seizure Detection Using EEG Signals

Key Points

Graph Neural Network achieved 100% accuracy in detecting epileptic seizures, demonstrating the framework's effectiveness.
Random Forest model recorded a high accuracy of 97.96%, revealing the potential of advanced machine learning methods.
Multiple machine learning and deep learning techniques were evaluated, yielding varying accuracy levels for seizure detection.
The framework allows users to upload EEG data for real-time seizure predictions, which could support clinical assessments.

Abstract

This project aims to detect epileptic seizures using brainwave signals recorded by an electroencephalogram (EEG). The goal is to create an accurate and intuitive system that can autonomously detect if someone is experiencing a seizure. To achieve this, EEG data from a public dataset was cleaned and converted into two classes: seizure and normal. Multiple models were used to find the most accurate method for detection. Machine learning models like Logistic Regression gave 82.17% accuracy, Random Forest gave 97.96%, and XGBoost achieved 97.22%. Deep learning models such as CNN and LSTM achieved 79.78% and 80.91% accuracy respectively. The best results came from the Graph Neural Network (GNN), which reached 100% accuracy. The system also includes a feature where users can upload their own EEG values to get real-time predictions and graphical output. This approach can help in supporting doctors with quick and reliable analysis of brain activity

Demander à l'IA

Bookmark