HealthAIoT: AIoT-driven smart healthcare system for sustainable cloud computing environments

Abstract

The evolution of Artificial Intelligence of Things (AIoT), coupled with advancements in edge and cloud computing, has enabled the development of real-time IoT-based applications. Integrating Internet of Things (IoT) and AI-driven edge–cloud services can address challenges such as early disease detection, system performance, data management and environmental sustainability in cloud-centric healthcare environments. To address these challenges, we propose HealthAIoT, a new architecture that utilises AIoT with cloud computing services to create a smart healthcare system. In our current implementation, HealthAIoT assesses the risk of developing diabetes mellitus in healthy individuals based on their personal health metrics and medical history; however, the proposed framework is fundamentally designed to be disease-agnostic and can be extended to incorporate detection and monitoring for other diseases. The HealthAIoT architecture consists of two main modules: a diabetes predictor and a cloud scheduler. The diabetes prediction module and cloud scheduler both utilise Multilayer Perceptron (MLP) models. The cloud scheduler manages health-related data and application requests from IoT devices, optimising resource utilisation and minimising the environmental impact of cloud services. The performance of the HealthAIoT framework is tested using the realistic testbed CloudAIBus. Experimental results demonstrate that the MLP-based diabetes predictor achieves 78.30% accuracy and an F1-score of 0.7719 on unseen patient data while cloud scheduler achieves 93.6% accuracy. Further, system performance is evaluated using metrics including energy consumption, carbon-free energy usage, cost, execution time, and latency. By identifying individuals at the highest risk of developing diabetes, the framework enables targeted preventative interventions, optimises resource usage and maximises impact, while also serving as a foundational framework for broader healthcare applications.