Pragmatic Analysis of IoMT Network Modelling Techniques from a Statistical Perspective

Internet of Medical Things (IoMT) networks are high precision wireless interfaces that require design of accurate sensing, actuation, and processing devices. These devices include wireless electrocardiograph (ECG) sensors, electroencephalograph (EEG) sensors, wireless blood pressure monitoring devices, etc. Due to direct patient interface, these devices are required to have superior performance in terms of sensing accuracy, processing efficiency, and communication quality of service (QoS) parameters. A wide variety of models are proposed by IoMT researchers to perform this task, but each of these models vary in terms of network size, deployment complexity, cost of deployment, processing delay, etc. These models include machine learning routing techniques, blockchain based security methods, privacy preservation methods, high-precision sensor design methods, high-performance communication interfaces, etc. Due to such a wide variation in performance, IoMT network design requires continuous validation, which increases time-to-market, thereby increasing deployment cost. In order to reduce number of validations, a statistical survey of models for IoMT network design is discussed in this text. This discussion is focussed towards evaluating various characteristics, advantages, limitations, and future research scopes in existing models. readers would be able to identify best performing model(s) for a given IoMT application. It is followed by a statistical analysis of the reviewed IoMT network design models in terms of end-to-end delay, communication QoS, network security, scalability, computational complexity, cost of deployment, and application of use. This statistical performance evaluation will further assist readers to statistically compare the reviewed methods, and identify best performing model(s) & their combinations for context-specific network deployments. Due to this, readers would be able to reduce network design & validation delay, which will further assist in reducing design costs, thereby facilitating faster deployments. The paper text also recommends various fusion, and transfer learning-based techniques which can be used for improving performance of the reviewed models.