Model-based detection of partially obstructed endotracheal tube*

Abstract

A five-element lumped pulmonary model was developed to estimate respiratory mechanics automatically and noninvasively. The model was applied to diagnose obstructed endotracheal tube. Events like bronchospasm and stiff chest wall were also tested to determine the specificity of the diagnosis. Cases with positive end-expiratory pressure were also included in the analysis to see the effects of positive end-expiratory pressure on the model. Randomized controlled animal study. University department of anesthesiology. Ten anesthetized, paralyzed, and mechanically ventilated mongrel dogs (19-45 kg) of either gender. Two levels of upper airway obstruction were induced in ten dogs by partially constricting the endotracheal tube. Acute bronchial constriction was produced in five dogs by injecting methacholine through a central venous catheter. In the same five dogs, the chest wall was stiffened by wrapping a pressure cuff around the chest. Positive end-expiratory pressure was also applied as a separate event in these five animals. Airway pressure and flow were continuously recorded at the mouth. Model parameters were iteratively identified until the root mean square error between respiratory impedance (obtained from airway pressure and flow) and model-predicted impedance (calculated using Ohm's law) was minimum. The peak inspiratory pressure increased and the peak expiratory flow rate decreased with increasing levels of partial obstruction. The value of the model parameters R(1) and C(2) increased and C(1) decreased with partial obstructed endotracheal tube, whereas R(1) increased and L and C(2) decreased with bronchospasm. With stiff chest wall, R(2) increased and C(2) decreased. With positive end-expiratory pressure, the L parameter decreased and no significant change in other model parameters was observed. Obstructed endotracheal tube is indicated if R(1) increased > or =30%, C(1) decreased > or =10%, and C(2) increased > or =10% from baseline. The test results using 45 events, including control, three complications, and positive end-expiratory pressure, show that when the model is used to diagnose obstructed endotracheal tube, the method has a sensitivity of 90% and specificity of 97%. During an obstructed endotracheal tube, model parameters change such that the event can be diagnosed noninvasively, automatically, and accurately. The model differentiates between upper airway obstruction and complications like bronchospasm and stiff chest wall.