An IoT framework for bio-medical sensor data acquisition and machine learning for early detection

Abstract

Internet of things in clinical domain has opened up new possibilities in remote monitoring of patients by connecting healthcare bio-sensor systems over the internet. This paper has proposed a working prototype of a real-time health monitoring system, which collects sensor data from body area network and communicates the data to a predictive model that is trained on historical clinical data. The prototype is equipped with Analog DeviceTM AD 8232 module for electrocardiogram and heart rate monitoring. CYPRESS CY8CKIT-042-BLE-A PSoC® 4 Bluetooth® Low Energy Pioneer Kit is used for implementation of a body area network, which collects patient’s vitals and communicates the sensor data to a Raspberry Pi3. The gateway device between WPAN (Bluetooth® Low Energy) and WLAN (IEEE 802.11n) is implemented using Raspberry Pi3. The gateway device collects the sensor data over a Bluetooth personal area network coming from all the connected devices and the data is acquired over internet server. ECG- ST wave and heart rate data are sent to the cloud server from the sensors. On the server, a machine learning model is deployed to predict any malfunctions based on sensor readings posted from the real-time health monitoring system and generate early alerts. We have obtained >90% prediction accuracy with random forest classifier using the UCI heart diseases repository.