We have developed a computational model of gas mixing and ventilation in the human lung represented as a bifurcating network. We have simulated multiple-breath washout (MBW), a clinical test for measuring ventilation heterogeneity in patients with obstructive lung conditions. By applying airway constrictions inter-regionally, we have predicted the response of MBW indices to obstructions and found that they detect a narrow range of severe constrictions that reduce airway radius to between 10% -30% of healthy values. These results help to explain the success of the MBW test to distinguish obstructive lung conditions from healthy controls. Further, we have used a perturbative approach to account for intra-regional airway heterogeneity that avoids modelling each airway individually. We have found, for random airway heterogeneity, that the variance in MBW indices is greater when largemagnitude constrictions are already present, and that the indices become more sensitive to structural heterogeneity when already elevated. This method is a computationally efficient way to probe the lung's sensitivity to structural changes, and to quantify uncertainty in predictions due to random variations in lung mechanical and structural properties.
Author summaryThe multiple-breath washout (MBW) test is a clinical test of lung function that measures the efficiency of gas transport and mixing within the lung, and which has proven very sensitive in detecting early disease in cystic fibrosis (CF). In this paper we have developed a computational model of lung function to simulate air movement and gas transport in the lungs and generate MBW outcomes. We have used this to show why MBW is so sensitive to airway blockages similar to those encountered in CF. Importantly, the model incorporates a new and computationally-efficient method that also allows us to account for uncertainty and randomness in lung structure and mechanical properties. This has been used to show how variability of MBW outcomes increases in disease states. The model provides a framework for modelling clinical data, where accounting for uncertainties in inputs is crucial in making clinically meaningful predictions.