A Personalized Spring Network Representation of Emphysematous Lungs From CT Images

Emphysema is a progressive disease characterized by irreversible tissue destruction and airspace enlargement, which manifest as low attenuation area (LAA) on CT images. Previous studies have shown that inflammation, protease imbalance, extracellular matrix remodeling and mechanical forces collectively influence the progression of emphysema. Elastic spring network models incorporating force-based mechanical failure have been applied to investigate the pathogenesis and progression of emphysema. However, these models were general without considering the patient-specific information on lung structure available in CT images. The aim of this work was to develop a novel approach that provides an optimal spring network representation of emphysematous lungs based on the apparent density in CT images, allowing the construction of personalized networks. The proposed method takes into account the size and curvature of LAA clusters on the CT images that correspond to a pre-stressed condition of the lung as opposed to a naïve method that excludes the effects of pre-stress. The main findings of this study are that networks constructed by the new method 1) better preserve LAA cluster sizes and their distribution than the naïve method; and 2) predict different course of emphysema progression compared to the naïve method. We conclude that our new method has the potential to predict patient-specific emphysema progression which needs verification using clinical data.