Scholarlink

<p>For monitoring physiological indicators outside of the traditional clinical setting, Remote Patient Monitoring (RPM) has become more relevant. The present IoT-based monitoring solutions are usually based on threshold or deep learning models, both of which are either unresponsive or computationally intensive. Hence, they are not suitable for lightweight implementation and scalability. To address these problems, this paper introduces a simulation-based IoT remote patient monitoring framework, called SmartHealth-IoT, with software-generated virtual wearable physiological streams. The proposed framework combines virtual physiological stream generation, statistical-temporal feature extraction using a sliding window, lightweight classification with an optimised LightGBM model (VitalBoostNet), and a hybrid rule-model alert generation mechanism. VitalBoostNet is not a new deep learning topology but rather an optimised gradient-boosted (GB) classifier integrated into the SmartHealth-IoT software pipeline for low-latency multi-class risk classification. The study is conducted under controlled virtual patient conditions rather than using real patient data obtained from smartwatches or clinical wearable recordings. The experiments demonstrated that VitalBoostNet achieved 96.73% accuracy and 96.33% F1 score while maintaining low inference latency. The proposed configuration has been compared with Logistic Regression, SVM, Random Forest, XGBoost, LSTM and rule-only classification and was found to offer a good accuracy-latency balance. With software-emulated communication, scalable tests with up to 100 virtual patients were conducted, resulting in increased throughput and reduced latency to up to 158ms. The results show that SmartHealth-IoT provides a simulation environment to test lightweight RPM workflows under controlled conditions in a reproducible manner. Before any clinical or deployment-oriented claims can be made, however, validation using actual wearable datasets, clinical-grade sensors, patient-specific variability and hardware-level communication is needed.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

SmartHealth-IoT: a simulation-based IoT remote patient monitoring framework using virtual wearable physiological data streams

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月31日

查看全文
  • A. Bharath,
  • G. Merlin Sheeba

摘要

For monitoring physiological indicators outside of the traditional clinical setting, Remote Patient Monitoring (RPM) has become more relevant. The present IoT-based monitoring solutions are usually based on threshold or deep learning models, both of which are either unresponsive or computationally intensive. Hence, they are not suitable for lightweight implementation and scalability. To address these problems, this paper introduces a simulation-based IoT remote patient monitoring framework, called SmartHealth-IoT, with software-generated virtual wearable physiological streams. The proposed framework combines virtual physiological stream generation, statistical-temporal feature extraction using a sliding window, lightweight classification with an optimised LightGBM model (VitalBoostNet), and a hybrid rule-model alert generation mechanism. VitalBoostNet is not a new deep learning topology but rather an optimised gradient-boosted (GB) classifier integrated into the SmartHealth-IoT software pipeline for low-latency multi-class risk classification. The study is conducted under controlled virtual patient conditions rather than using real patient data obtained from smartwatches or clinical wearable recordings. The experiments demonstrated that VitalBoostNet achieved 96.73% accuracy and 96.33% F1 score while maintaining low inference latency. The proposed configuration has been compared with Logistic Regression, SVM, Random Forest, XGBoost, LSTM and rule-only classification and was found to offer a good accuracy-latency balance. With software-emulated communication, scalable tests with up to 100 virtual patients were conducted, resulting in increased throughput and reduced latency to up to 158ms. The results show that SmartHealth-IoT provides a simulation environment to test lightweight RPM workflows under controlled conditions in a reproducible manner. Before any clinical or deployment-oriented claims can be made, however, validation using actual wearable datasets, clinical-grade sensors, patient-specific variability and hardware-level communication is needed.