A Machine Learning Framework to detect epileptic seizures and prioritize data transmission in Internet of Medical Things
Wearable Devices; Machine Learning; Communication
Internet of Medical Things (IoMT) relies on wearable devices to sense, collect, and transmit biometry data continuously, leveraging for this on-board radio wireless facility to communicate with a gateway, which in turn delivers the collected data across the Internet to an IoMT system. In this context, medical diagnosis and treatment for syndromes such as epilepsy require data availability, transmission reliability, low network latency, and fast seizure detection. Thus, proceeding with seizure detection in-device or edge using Machine Learing (ML) algorithms fulfil such requirements. In addition, IoMT applications have strict Quality of Service (QoS) requirements to ensure accurate information for medical staff, where it is important to classify (in a non-intrusive manner) and give priority to IoMT traffic flows transmitted from wearable devices in a shared wireless infrastructure. In this thesis proposal, we present a ML framework to identify wearable medical device traffic, detect epileptic seizures, and prioritize data transmission to hospital cloud servers. Addressing the issues above would push IoMT applications to a higher level of user experience. In addition, our proposal will minimize the risk of data interception as it reduces data exposure in wireless medium access.