Quantitative CT features for COPD classification and GOLD spirometric severity assessment: a single-center model-development study.

Background: Chronic obstructive pulmonary disease (COPD) is diagnosed primarily by post-bronchodilator spirometry, but quantitative CT can characterize emphysema and visible-airway structural abnormalities that are not fully captured by airflow measurements alone. Because CT-derived emphysema and airway features have been extensively studied, this single-center study aimed to develop and describe quantitative CT-based classification models and to explore bronchial-tree morphology as a hypothesis-generating imaging phenotype, rather than to establish a clinically validated diagnostic replacement for pulmonary function testing. Methods: We enrolled 229 participants (138 with COPD and 91 control participants) at Fudan University Zhongshan Hospital. HRCT was used to quantify visible-airway dimensions at a prespecified eighth-generation airway, emphysema extent quantified as low attenuation area percentage (LAA%), and bronchial-tree morphology. Clinical variables including smoking history, modified Medical Research Council (mMRC) dyspnea score, and pulmonary function tests were recorded. Multivariable logistic regression was used to develop single-center models for COPD classification and for classification of severe airflow obstruction (FEV1%pred < 50%, corresponding to GOLD spirometric grades 3-4). Model performance is reported as apparent performance because internal and external validation were not performed. Results: Compared with control participants, COPD participants had smaller airway lumen area, higher wall-to-diameter ratio (T/D), higher wall area percentage (WA%), and higher LAA%. Mean wall thickness did not differ significantly, indicating that the group difference mainly reflected lumen narrowing rather than wall thickening. Emphysema extent was associated with lower FEV1/FVC and FEV1%pred. Sparse-type bronchial morphology was associated with COPD in multivariable analysis (OR 28.6, 95% CI 6.7-149.8), but the wide confidence interval and the need for reproducible operational validation require cautious interpretation. The combined clinical-CT development models achieved apparent AUCs of 0.906 for COPD classification and 0.859 for classification of severe airflow obstruction. Conclusion: Quantitative HRCT emphysema and visible-airway measurements were associated with COPD status and spirometric airflow obstruction. The models should be regarded as single-center development models, and the bronchial-tree morphology finding should be interpreted as hypothesis-generating until reproducibility, calibration, incremental value, and external validation are demonstrated.