Advances in Respiratory Impedance Predictions using Pulmonary Functional Imaging Models of Asthma

Abstract

Previous asthma pulmonary functional modelling used the locations of ventilation defects observed in inhaled gas imaging to implement binary airway closures and required additional random narrowing in ventilated regions to describe measured impedance, compromising predicted ventilation. Here we used gradations in intensity discretized using k-means clustering, and simulated annealing to choose degrees of narrowing within these regions to minimize the difference between measured and predicted ventilation. We found that the resistance (p < 0.005), reactance (p < 0.05), and ventilation (p < 0.005) predicted by the graded narrowing model was closer to subjects’ measurements compared to the binary model. The graded approach did not fully account for the frequency dependence of resistance known to be indicative of heterogeneity. Thus, while the modelled airway narrowing predicted ventilation and impedance closer to subjects’ measurements than binary closures, other factors or unobserved heterogeneity are needed to account for additional frequency dependence of resistance.