Abstract Introduction Chronic obstructive pulmonary disease (COPD) patients show remodeling and loss of distal airways and arteries, disrupting airway-vascular coupling even in early stages. Previous studies have quantified this airway-vascular coupling on computed tomography (CT) images as the ratio of bronchial lumen diameters to adjacent arteries (BA) in individuals with COPD, and showed associations increased exacerbations and emphysema progression. However, previous studies have not quantified heterogeneity of the BA ratio in early / at risk COPD. We hypothesize the mean BA ratio will decline with increasing COPD severity, and that the variability in BA ratio will peak in the ‘at-risk’ group, reflecting early heterogeneous loss of coupling. Therefore, our objective was to develop a fully-automated artery-vein separation method pipeline, and to apply this across a large community-based cohort to quantify BA pair heterogeneity in individuals with and at risk of COPD. Methods Participants from CanCOLD with full-inspiration CT images were selected for analysis. Arteries were segmented using a deep learning segmentation pipeline, trained on 250 labelled chest CT images using 3D MedNeXt architecture. The arteries were spatially matched with the airways by determining centroids for each segment, ensuring parallel segment orientation and matching each airway segments to nearest arterial segments using minimum centroid-to-centroid distances. BA pairs were determined for five major airway segments (RB1, RB4, RB10, LB1 and LB10). For each of the segments, bronchial lumen diameter (Bin) and adjacent artery diameter (A) was calculated. Mean (mean Bin/A), standard deviation (Std Bin/A) and coefficient of variation (CV Bin/A) of the BA ratios were determined. Analysis of covariance was used for comparison of mean, standard deviation and coefficient of variation of BA ratios between never-smokers, at-risk, GOLD I and GOLD II+ COPD, adjusted for covariates (age; sex; pack-years; smoking status; body mass index; low attenuation area less than 910 Hounsfield units; CT lung volume and CT model). Results A total of 1082 CanCOLD participants having all BA pairs were evaluated: n = 221 never-smokers; n = 306 at-risk; n = 324 GOLD I; and n = 231 GOLD II+. Mean Bin/A was significantly reduced in GOLD II+ as compared to the other groups (p 0.05). Std Bin/A was increased in at-risk groups compared to never-smokers and GOLD II+ COPD (p 0.05). CV Bin/A was increased in at-risk groups compared to never-smokers (p 0.05). Conclusion These findings indicate that airway-to-artery ratio heterogeneity is an early marker of airway-vascular uncoupling and highlights its potential utility as a quantitative imaging endpoint for intervention studies. This abstract is funded by: Natural Sciences and Engineering Research Council of Canada and Canadian Institutes of Health Research
Singh et al. (Fri,) studied this question.