An Adaptive Thresholding and Fog-enabled Remote Healthcare Monitoring System

Real-time health monitoring plays a crucial role in continuous patient care, as well as early disease detection. However, traditional IoT systems in the cloud have difficulties with real-time applications due to latencies or network bandwidth limitations, which can impact patient safety. Furthermore, fixed thresholds for patient alarms can lead to false alarms or missing them completely. Addressing these challenges, this paper introduces and compares three different adaptive thresholding algorithms that use low computing power and can run on low-end edge devices close to the sensors themselves. This enables dynamic patient-specific alerts directly from the fog layer of the proposed system. Different sampling rates of the sensors have been compared to optimize for storage and computational power on the edge device as well as network traffic. This was done to identify the optimal interval time to send data without compromising the quality of the data and patient safety. Experiments were performed using an ESP32C3 as the fog layer. For the sensor itself, a smartwatch was used as this type of device became very affordable, can send data over Wi-Fi, and is easy to attach to the patient. For simulation and testing, the eICU Collaborative Research Database Demo has been used. It contains heart rate readings from 1564 patients. The system was also briefly tested with data from a personal smartwatch to simulate different sensor types sending at different intervals.