IEEE Access · 2025 · 10 citations · 53 references
ComputingComputer ScienceRobotics EngineeringInternet Of ThingsMachine LearningMedical RoboticsDigital Health SystemsHealth Monitoring SystemsMedical Cyber-physical SystemsPredictive Health ModelingDigital Twin ApplicationsDigital Twin ArchitectureDigital Twin ModelingDigital TwinsMachine Learning DiagnosticsMachine Learning For Health MonitoringMachine Learning For HealthcareRobotic Patient Monitoring
The digital twins (DTs) paradigm creates a virtual replica of a physical artifact that represents its real-world status in a virtual space. When implemented in the Internet of Robotic Things (IoRT), it develops a virtual model of the physical robot that receives real-time sensor data, acting like an actual robot. The DTs-based IoRT allows remote patient monitoring (RPM) that improves treatment plans and reduces the burden of gathering vital signs on health carers by automating repeated tasks. However, RPM procedures generate large amounts of sensor data, which are difficult to examine. Furthermore, health carers cannot forecast abnormalities based on health data. Machine Learning (ML) can analyze massive amounts of data and perceive patterns to anticipate anomalous health conditions. This research leverages the advantages of ML and DTs-based IoRT to remotely monitor chronic or infectious patients and predict their future status. The proposed system is an extension of our previously published work. It uses the virtual twin (VT) to navigate the physical twin (PT) for collecting data from the patient-mounted sensors and applies ML techniques to predict health anomalies. It evaluated six ML algorithms to determine the most accurate model. The models were trained and tested by using a dataset combining two public datasets and real-world data collected by healthcare providers. Results confirmed that the K Nearest-Neighbor (KNN) has the best accuracy i.e. 97%. The system was also tested in the clinical setting to collect patient data and the best-performing algorithm (KNN) was used for status prediction, obtaining 98% accuracy.
53
Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik et al. · Communications of the ACM · 2021 · 5.8K citations · Full text
Improvements to Platt's SMO Algorithm for SVM Classifier Design
S. Sathiya Keerthi, Shirish Shevade, Chiranjib Bhattacharyya et al. · Neural Computation · 2001 · 1.8K citations
A survey on missing data in machine learning
Tlamelo Emmanuel, Thabiso Maupong, Dimane Mpoeleng et al. · Journal Of Big Data · 2021 · 975 citations · Full text