Probabilistic estimation of respiratory rate from wearable sensors

Abstract

Respiration rate (RR) is a physiological parameter that is typically used in clinical settings for monitoring patient condition. Consequently, it is measured in a wide range of clinical scenarios, notably absent from which is measurement using wearable sensors. With increasing numbers of patients being monitored via wearable sensors, as described below, there is an urgent need to be able to estimate RR from such sensors in a robust manner. In this chapter, we describe a novel technique for measuring RR using waveform data acquired from wearable sensors. The technique derives RR from a physiological signal which is routinely acquired by many mobile sensors: the photoplethysmogram (PPG). Each RR measurement from the proposed method is accompanied by a confidence measure, providing estimates of clinical quality that will allow the system to, for example, only report RR values when they exceed some probabilistic level of certainty. The goal of this method is to improve upon existing methods, which simply report RR values without probabilistic estimation, and which therefore suffer the lack of robustness that prevents their use in clinical practice.