NeuraNest: EEG-based Smart Seizure Detection and Alert System

Epileptic seizures are dangerous to health because they occur suddenly and unpredictably. Here, we introduce a seizure prediction system consisting of a deep learning model integrated with an interactive web interface. The underlying model is a hybrid architecture that combines Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNN), trained on the publicly available CHB-MIT EEG dataset. This model effectively detects both spatial features and temporal relationships over extended periods in EEG activity to distinguish seizure from non-seizure activity. The web application enables the upload of EEG recordings in European Data Format (EDF), which are automatically preprocessed and analysed using a trained model. There is also an SOS alarm facility in the system that alerts all registered caregivers in case a seizure is detected. This project aims to deliver an effective, practical, and responsive seizure prediction system that is beneficial to epilepsy patients and families in real-world settings.