Investigation of the performance of an automatic arterial oxygen controller

Abstract

Premature infants often require respiratory support with a varying concentration of the fraction of inspired oxygen (FiO2) to keep the arterial oxygen saturation (SpO2) within the desired range to avoid both hypoxemia and hyperoxemia. Currently, manual adjustment of FiO2 is the common practice in neonatal intensive care units (NICUs). The automation of this adjustment is a topic of interest. The research team, at University of Missouri-Columbia (UMC), has developed a novel automatic arterial oxygen saturation controller. In this study, a systematic approach has been developed to investigate both non-clinical and clinical performance of this device. The non-clinical investigation of the performance was performed using a neonatal respiratory model (hardware-in-the-loop test). A factorial experimental design was utilized to generate challenging model responses of SpO2, which were addressed by the controllers. With this study, we demonstrate the stability and ability of the adaptive PI-controller to improve oxygen saturation control over manual control by increasing the proportion of time where SpO2 of the neonatal respiratory model was within the desired range and by minimizing the variability of the SpO2. In addition, the controller ability to significantly reduce the number of hypoxemic events of the neonatal respiratory model was reported. Results of this investigation show the competence of the controller estimation system for estimating neonatal respiratory model parameters while the adaptive PI-controller was in use. Also, the functionality of the controller with no mechanical or communication failure was validated non-clinically before heading forward to the clinical trial. The clinical investigation of the performance was performed by conducting a clinical trial at the NICU of the MU Women's and Children's Hospital. The crossover design was used for the clinical trial to allow within-subject comparison and to eliminate interpatient variability. Two human subjects, with two different target ranges of SpO2, were enrolled in the study. The adaptive automatic PI-controller shows clinical feasibility to improve the maintenance of SpO2 within the intended range. With this study, we demonstrate the potential of the automatic controller to minimize the variability of SpO2. In addition, the controller shows the ability to reduce the bradycardia and the hypoxemia. Moreover, the hardware and software of the controller show an ability to transition from manual to automatic mode, and vice versa with no pronounced “bump” or step variation in the control signal, and stability and performance were not adversely affected during the transitions.