Abstract
Graphical Abstract
STEM 1 occurs from the beginning of the year (~August) to about March. It is an independent research project where you can pick any project you are interested in as long as it is in the field of STEM. You research this topic, form your hypothesis, conjecture, or engineering need, and then go forward with your project. You present preliminary results in December, and your final results in February with check-in meetings with Dr. Crowthers throughout the year- formal and informal.
Epileptic seizures, specifically tonic-clonic seizures, are an immense issue to those with epilepsy. Seizure detection models have been gaining popularity in scientific literature, however, there is still a lot of room for improvement. The goal of this project is to make a Long Short-term Memory (LSTM) network that addresses these problems, the main one being false negatives- missing a seizure, which could be life threatening.
Seizure detection is a critical challenge, as EEG monitoring is costly, inconvenient, and impractical for daily use. Current methods lack accessibility and real-time response, limiting timely intervention. A more efficient, cost-effective, and user-friendly solution is needed to improve monitoring, enhance patient safety, and support the development of wearable health technologies.
The goal of this project is to design and develop an AI-driven seizure detection system that uses ECG data. This involves building a deep learning model capable of accurately identifying seizures in real time while ensuring efficiency, scalability, and reliability. The system will incorporate techniques such as data preprocessing, class balancing, hyperparameter tuning, and model explainability to enhance performance. Ultimately, the project aims to create a cost-effective, accessible, and user-friendly solution suitable for integration into wearable health devices.
