مقاله انگلیسی کاربردهای یادگیری عمیق برای IoT در مراقبت های بهداشتی
Abstract
Keywords
Introduction
Review of related studies
Research method
Organization of deep learning applications for IoT in healthcare
Discussion and comparison
Conclusion
Declaration of competing interest
Acknowledgement
Abbreviations
References
ABSTRACT
In machine learning, deep learning is the most popular topic having a wide range of applications such as computer vision, natural language processing, speech recognition, visual object detection, disease prediction, drug discovery, bioinformatics, biomedicine, etc. Of these applications, health care and medical science-related applications are dramatically on the rise. The tremendous big data growth, the Internet of Things (IoT), connected devices, and high-performance computers utilizing GPUs and TPUs are the main reasons why deep learning is so popular. Based on their specific tasks, medical IoT, digital images, electronic health record (EHR) data, genomic data, and central medical databases are the primary data sources for deep learning systems. Several potential issues such as privacy, QoS optimization, and deployment indicate the pivotal part of deep learning. In this paper, deep learning for IoT applications in health care systems is reviewed based on the Systematic Literature Review (SLR). This paper investigates the related researches, selected from among 44 published research papers, conducted within a period of ten years – 2010 to 2020. Firstly, theoretical concepts and ideas of deep learning and technical taxonomy are proposed. Afterwards, major deep learning applications for IoT in health care and medical sciences are presented through analyzing the related works. Later, the main idea, advantages, disadvantages, and limitations of each study are discussed, preceding suggestions for further research.
Introduction
In machine learning, deep learning is considered a new area and, consequently, an essential subset of artificial intelligence (AI). Several definitions are offered for deep learning, but the following definition is the most comprehensive of all. Deep learning is a series of algorithms founded on artificial neural networks having multiple layers [1]. Artificial neural networks were first proposed in the 1940s by McCulloch and Pitts [2]. Biological neural systems inspired the main idea of neural networks for information processing [3]. A biological neuron consists of several entities, but investigating the role of the following elements is substantial in studying the functionality of artificial neural networks: - Synapse: Input signals receiver. - Dendrite: Weight assignments. - Cell body: Summation and integration. - Axon: Signal transportation. - Axon terminal: Output result. As Fig. 1 shows, deep neural networks are artificial neural networks with several layers. Each layer is responsible for extracting some information represented by a score times weight, and forwarding them to the next layer. The sum of all the values related to a specific input makes the output. The input layer collects the input data, hidden layers are in charge of storing the corresponding weights, and the output layer yields the output results [4].