Modelling and fuzzy logic control of neonatal oxygen therapy.

摘要

Adequate tissue oxygenation can be impaired in neonates who have under-developed lungs. Systemic arterial blood oxygen saturation (SaO{dollar}sb2{dollar}) is an indicator of tissue oxygenation and its low level can result in tissue damage associated with hypoxemia. Retinopathy of prematurity and bronchopulmonary dysplasia are diseases associated with exposure to high inspired oxygen levels. Oxygen therapy is the administration of an air/oxygen gas mixture in an effort to regulate arterial oxygen levels and thus control tissue oxygenation. The blood oxygen saturation level from a pulse oximeter, SpO{dollar}sb2{dollar}, indicates SaO{dollar}sb2{dollar} and an automatic SpO{dollar}sb2{dollar} controller is recommended as an assistant to nursing staff. A computer driven model of the neonate's oxygen transport system, accommodating bi-directional gas flow in the pulmonary system and pulmonary/cardiac shunts in the circulatory system, was developed for testing several algorithms. Programs modelling one hour of physical behavior of three types of neonates are produced. Manual control is based on clinical observations and a PID controller's gain values are selected from empirical data. Two fuzzy logic based controllers are designed--non-linear (FL) and non-linear with artifact detection (FLA). All four algorithm were run through the neonate models and distributions of time spent by the models at various SpO{dollar}sb2{dollar} levels were used as a comparison method. The PID algorithm provided better control than manual mode and the FL controller was better than both. The FLA controller was slightly better than FL control. In summary, a fuzzy logic controller with artifact detection provided better oxygen therapy for software neonatal models than standard fuzzy logic, PID or manual control.