What question did this study set out to answer?

To develop a robust model for recognizing epileptic seizures using EEG signals in a cloud computing framework.

March 29, 2026Open Access

Seizing the future: a robust epileptic seizure recognition model-based cloud computing platform and advanced deep learning model

Puntos clave

To develop a robust model for recognizing epileptic seizures using EEG signals in a cloud computing framework.
Designed an IoT-based cloud platform for real-time monitoring.
Applied the gray wolf optimization algorithm for feature selection from EEG data.
Utilized an adaptive deep learning model for seizure classification.
Incorporated security measures for data protection and patient privacy.
Achieved an accuracy of 0.9773 in seizure recognition.
Demonstrated a classification error of only 0.0227.
Reported recall and precision rates of 0.9591 and 0.9592, respectively.
Showed effectiveness in classifying complex EEG signals.

Resumen

Abstract Epileptic seizure recognition (ESR) is vital for diagnosing and managing epilepsy, a disorder marked by recurrent seizures. Real-time detection through an Internet of Things (IoT)-based cloud platform enables prompt alerts and enhances patient safety. This study proposes a novel architecture that leverages EEG signals analyzed within an IoT-based cloud computing framework to enable remote monitoring of patients. The model utilizes a metaheuristic gray wolf optimization (GWO) algorithm for selecting an optimal subset of features from the EEG data, combined with an adaptive deep learning method for robust seizure classification. The IoT infrastructure is carefully designed to closely monitor patients by capturing EEG signals through SDKs managed by Greengrass and mobile devices, ensuring seamless coordination among IoT devices surrounding each patient. The system also incorporates advanced security and availability measures, such as device shadows, certificate-based security, and identity management. Cloud services are integral to this architecture, efficiently handling large volumes of data, supporting the generation and continuous improvement of the recognition model, and facilitating communication between patients, healthcare providers, and consultants. Experiments conducted on an online benchmark EEG dataset of 500 subjects showed superior performance: accuracy of 0.9773, classification error of 0.0227, recall of 0.9591, and precision of 0.9592. These results demonstrate the model’s effectiveness in understanding complex EEG signals and accurately classifying epileptic seizure states, supporting its suitability for clinical and remote patient management.

Preguntar a la IA

Me gusta

Guardar

Ver artículo completo