Airway Transfer Network
Chronic lung diseases, like idiopathic pulmonary fibrosis (IPF) are characterised by abnormal dilatation of the airways. Quantification of airway features on computed tomography (CT) can help characterise disease progression. Physics based airway measurement algorithms have been developed, but have met with limited success in part due to the sheer diversity of airway morphology seen in clinical practice. Supervised learning methods are also not feasible due to the high cost of obtaining precise airway annotations.
We propose synthesising airways by style transfer using perceptual losses to train our model, Airway Transfer Network (ATN).
Since physical evaluation is not possible with the absense of ground truth on real airways that are imaged, we evaluate ATN against existing SOTA, simGAN  by deriving CT airway biomarkers to predict mortality in a population of 113 patients with IPF. ATN is quicker and easier to train compared to simGAN.
ATN-based airway measurements are consistently stronger predictors of mortality than simGAN-derived airway metrics on IPF CTs.
Airway synthesis by a transformation network that refines synthetic data using perceptual losses is a realistic alternative to GAN-based methods for clinical CT analyses of idiopathic pulmonary fibrosis.
Our source code is opensource and compatible with the existing open-source airway analysis framework, AirQuant.
 Nardelli, Pietro et al. “Generative-based airway and vessel morphology quantification on chest CT images”. Medical Image Analysis 63 (2020)