Bringing intelligence to Edge/Fog in Internet of Things‐based healthcare applications: Machine learning/deep learning‐based use cases

• Published in: International journal of communication systems 2023-06, Vol.36 (9), p.n/a
• Main Authors: Makina, Hela, Ben Letaifa, Asma
• Format: Article
• Language: English
• Subjects: cloud computing; Deep learning; Edge/Fog Computing; eHealth; Energy consumption; Internet of Things; IoT; Machine learning
• ISSN: 1074-5351; 1099-1131
• DOI: 10.1002/dac.5484

Summary Internet of things and smart medical applications are deeply changing the way healthcare is delivered worldwide. A typical Internet of Things (IoT)‐based eHealth system includes medical sensors for data collection, access network to transmit data, and Cloud servers for data processing and storage. Machine learning/deep learning (ML/DL) have proven to be a powerful tool for data classification, disease prognosis, and diagnosis and even medication prescription. ML/DL models need a large amount of data and significant computational and storage capacities especially for the training phase. Deploying ML/DL algorithms in the Cloud can be effectively done due to the processing and storage power of Cloud data centers. However, it raises many issues related to the availability, latency, energy consumption, bandwidth, security, and privacy. Recently, there is a growing interest to run as much processing as possible nearer the data sources, in the Edge, to compensate the Cloud‐based solutions limitations. In this paper, we propose to investigate the benefits of using IoT, ML/DL, and Edge Computing to enhance healthcare applications. Then, we are going to review the main approaches and trends for executing ML/DL in the Edge, to give their benefits and limitations, and draw finally conclusions about existing research issues. This paper aims to provide a comprehensive survey on intelligent techniques applied in the Edge network for eHealth applications. Through different implementations on the Edge (on‐device, server‐based, collaborative computation), ML/DL‐based solutions offer a great opportunity for IoT‐based healthcare to offer a trade‐off between accuracy, latency, and energy consumption.