IoT-driven wearable devices enhancing healthcare: ECG classification with cluster-based GAN and meta-features

Abstract

Wearable devices in medical technology promise advancements in healthcare but face challenges like limited data use and delayed analysis, hindering their real-time effectiveness. Enabling wearable devices with edge computing maximizes their potential, allowing real-time tasks like ECG classification to be performed intelligently at the device level. We propose the Wearable IoT Edge, a computing device that empowers wearable health devices with real-time data insights and IoT capabilities, facilitated by the Wearable Interworking Proxy and compliant with oneM2M standard-based server. We demonstrate the application of a proposed Wearable IoT Edge by addressing ECG classification challenges. Our approach addresses data imbalance by integrating a Cluster-Based Generative Adversarial Network (GAN) with meta-features derived from Convolutional Neural Networks (CNNs) and Transformers to enhance ECG classification accuracy. Experimental results demonstrate a 3.18% improvement in the F1 score for ECG classification validating the effectiveness of the approach. These findings highlight the Wearable IoT Edge’s potential to improve real-time healthcare monitoring and diagnostics.