Wireless medical sensor network for blood pressure monitoring based on machine learning for real-time data classification

Abstract

Blood pressure issues are related to many illnesses threatening human health and require continuous control and monitoring. Health telemonitoring is an innovative solution allowing wellbeing and increasing autonomy of patients. Moreover, machine learning algorithms are a viable method used in recent studies for analyzing, predicting, and classifying health data while improving the health conditions of telemonitoring and telediagnosis. The main purpose of this article is to employ machine learning algorithms for Blood Pressure (BP) measurement classification in real-time. Data is gathered from the human body through a multilevel Wireless Medical Sensors Network (WMSN) architecture that is deployed to acquire, analyze, and monitor BP remotely. The first layer of the proposed architecture performs BP measurement, classification, and transmission to the second layer using a wearable sensor. The learning of classifiers is established with Cross-Validation to avoid over-fitting and to improve the performance while comparing the Decision Tree, K-Nearest-Neighbors, and Naive Bayes algorithms. Afterward, the best classifier is implemented in the BP sensor node to identify the BP status. The second layer is responsible for data aggregation in the Cloud and alerting when an anomaly occurs in BP measurement. The third layer is configured to present BP information continuously to the health professionals and patients using the Internet of Things (IoT) platform that retrieves data from the cloud. The evaluation results show better accuracy, i.e. 97.9% using the decision tree classifier. An experimental trial is carried out and the timing of 49 seconds is reached between BP measurement and the display of data on the IoT platform. Besides, the system was tested in real-time trials and produced an accurate classification of each measured BP. The obtained results approve the feasibility and effectiveness of the proposed approach in terms of BP measurement, analysis, transmission, and supervision in real-time.