Predicting multiple breath washout outcomes from hyperpolarised gas magnetic resonance imaging (MRI)

Abstract

Background: We have developed a lightweight computational model to relate the distribution of ventilation in the lung, as measured by hyperpolarised gas MRI, to multiple breath washout (MBW) measurements on the same patients. Furthermore we have fitted the model to MBW measurements to predict the ventilation distribution, and compared to the MRI measurements. Method: We used MRI and MBW data from 9 healthy children and 12 children with mild cystic fibrosis (normal FEV1, max LCI 11.1, 4/12 had LCI > 7), with both experiments performed in supine position. Breath volumes and Fowler dead space measured by MBW were used as parameters in the model. First, the distribution of ventilation rates from each voxel from the MRI data and MBW measured functional residual capacity (FRC) were used to parametrise the MBW model. Second, a lognormal distribution of ventilation was assumed and the FRC and distribution width Dv were fitted to MBW data using a constrained non-linear least squares method. Results: The model fits to MBW data were robust and repeatable, showing stong agreement with measured FRC, and fitted Dv correlated strongly with LCI. The model agreement was good for subjects with normal LCI (5-7). 3 of the 4 cases of abnormal LCI were not predicted from the MRI ventilation data, and fits to MBW data from these subjects over-predicted the ventilation heterogeneity observed in MRI. Conclusions: We saw limited agreement between our data-based model and experimental outcomes for patients with mild CF. This suggests more detailed models of the lung physiology, which take into account differences between the test protocols, may be required to translate between the two measurements.