Supporting information.

Supplementary methodological details and results. We detail our manual cleanup of the upper airways in the Southampton/Air Liquide dataset. We then compare qualitative differences between a DRR and real chest X-ray to understand potential challenges working with real chest X-ray images. We provide results on the two CNN architectures tested for airway segmentation, including DICE coefficient and inference time. We provide additional results on SSAM reconstruction accuracy (reported as dimensional values). We then provide results that justify the choice of distal airway diameter model for the SSAM-reconstructed airways.

(PDF)