Click here to

Session: Quantitative Imaging [Return to Session]

Principal Component Analysis of Quantitative Computed Tomography Features and Visual Emphysema Scores: Association with Lung Function Decline

M Koo1*, W Tan2, J Hogg2, J Bourbeau3, C Hague2, J Leipsic2, and M Kirby1,2 for the CanCOLD Collaborative Research Group and The Canadian Respiratory Research Network, (1) Ryerson University, Toronto, ON, CA, (2) Centre For Heart Lung Innovation, St. Paul's Hospital, Vancouver, BC, CA, (3) McGill University Health Centre Research Institute, McGill University, Montreal, QC, CA


MO-B-TRACK 6-4 (Monday, 7/26/2021) 11:30 AM - 12:30 PM [Eastern Time (GMT-4)]

Purpose: Quantitative computed tomography (QCT) provides fully-automated measurements reflecting disease features in chronic obstructive pulmonary disease (COPD), yet visual scoring of emphysema remains the clinical standard and outperforms QCT for prediction of longitudinal outcomes, such as mortality. However, previous studies largely focus on only single QCT features. Our objective is to determine if multiple QCT features in combination are associated with COPD disease progression, independent of visual scoring.

Methods: Participants with and without COPD from the Canadian-Cohort-Obstructive-Lung-Disease (CanCOLD) study were evaluated. CT images were visually scored by a radiologist for emphysema severity. 14 QCT features were extracted (VIDA software), including: low-attenuation-area<-950 HU (LAA₉₅₀), wall-area-percent, total-airway-count, low-attenuation-cluster slope, disease-probability-map measures of emphysema and small airway disease, and vessel-volume. Principal component analysis (PCA) was performed for dimension reduction. Multiple linear regression models for annualized change in forced-expiratory-volume-in-one-second (ΔFEV₁) were constructed with linear combinations of features from each component along with visual emphysema score (VES), adjusted by confounding variables.

Results: We evaluated 588 participants; never-smoker: n=145, ever-smoker: n=161; mild COPD: n=171; moderate-severe COPD: n=111. Three PCA components were extracted that explained 68% of the variance in ΔFEV₁; these components reflected airway, emphysema/vessels, and emphysema clustering/air trapping measurements. In a multiple linear regression model for ΔFEV₁, VES was significant (standardized estimate (SE)=-0.098, p=0.04). In a model for ΔFEV1 with LAA₉₅₀ added, VES remained significant (SE=-0.094, p=0.04) but LAA950 was not. In a model with the three PCA components added, the emphysema clustering/air trapping component was significant (SE=-0.233, p<0.0001), while VES and all other components were not.

Conclusion: Linear combinations of QCT features increased the relative importance of QCT compared to visual scoring in terms of the association with FEV1 change. These findings will motivate future studies to develop and investigate QCT measurements in COPD.

Funding Support, Disclosures, and Conflict of Interest: Dr. Kirby gratefully acknowledges support from the Parker B. Francis Fellowship Program and the Canada Research Chair Program (Tier II). We also acknowledge funding support from the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant.



    Quantitative Imaging, Lung, CT


    IM/TH- Image Analysis (Single Modality or Multi-Modality): Quantitative imaging

    Contact Email