基于体素的COPD表型研究

Objective: To screen phenotypes of chronic obstructive pulmonary disease (COPD) using the voxel-based quantitative CT. Methods: COPD patients who had received inspiratory and expiratory CT scanning were consecutively recruited from the multi-center study named Digital Lung. Quantitative parametres of functional small-airway disease (fSAD) and emphysema (Emph) were measured. The k-means clustering method was used to analyze the different phenotypes using the paramenters of emphysema and small-airway disease. Fisher discriminatory analysis was used to get the classification function, and then make initial and cross validation. Results: Totally 50 COPD patients were recuited. We obtained three phenotypes in the whole cohort and the The Global Initiative for Chronic Obstructive Lung Disease (GOLD) II patients with the similar lung function, namely, emphysema-dominant type, small airway disease-dominant type, and the mixed type. Fisher linear classification function of the three phenotypes were obtained. In the initial verification and cross validation, the correct rate of the whole COPD subjects were 92.0% and 90.0%, the correct rate of GOLD 2 grade patients were 100.0% and 95.2%, respectively. Conclusion: The voxel-based quantitative CT can be used to divide COPD patients into three imaging phenotypes, namely, emphysema-dominant type, small airway disease-dominant type, and mixed type.