Optimal Health Monitoring via Wireless Body Area Networks

We consider the use of a wireless body area network for remote patient health monitoring applications. Our proposed network consists of a controller and multiple sensors, whose signals provide information on the health state of a patient. We model this patient-sensor network as a partially observable Markov decision process. The sensor outputs are used by the controller to update the patient's health-state belief probabilities and select a subset of sensors to be activated at the next decision epoch. We propose two operational algorithms that allow accurate monitoring of a patient's health state while minimizing operational and misclassification costs: i) a greedy algorithm, which applies a one-step look-ahead approach, and ii) a dynamic programming-based algorithm which yields the optimal policy. We provide a numerical example which demonstrates the applicability of the suggested methods and provides insights.