Click here to


Are you sure ?

Yes, do it No, cancel

CT-Based Radiomics Analysis: A New Imaging Biomarker in Chronic Obstructive Pulmonary Disease?

R Au1*, V Liu1, M Koo1, W Tan2, J Bourbeau3, J Hogg2, H Coxson2, M Kirby1, (1) Ryerson University, Toronto, Ontario, Canada, (2) Centre For Heart Lung Innovation, University Of British Columbia, Vancouver, British Columbia, Canada, (3) McGill University, Montreal, Quebec, Canada


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: Chronic obstructive pulmonary disease (COPD) is a progressive disease that causes emphysematous tissue destruction. Existing computed tomography (CT) emphysema measurements use simple density thresholds. Radiomics is an emerging method that quantifies spatial relationships between image voxels. We hypothesize CT radiomics features can be developed to quantify lung tissue destruction, will be more robust than simple CT density measurements, and significantly correlate with lung function in COPD.

Methods: Spirometry and CT images were obtained in never-smokers, smokers with normal lung function (at-risk), mild COPD (GOLD I), and moderate/severe COPD (GOLD II+) from the CanCOLD study. CT density measurements include the low-attenuation-areas-below -950HU (LAA950). An image analysis pipeline was developed to calculate 32 grey level co-occurrence matrix radiomics features. Methods included whole lung segmentation, image resampling/no resampling, airway removal/no removal, and thresholding (-1000HU to 0HU). Unpaired t-tests were used for method comparisons. Pearson correlation coefficients were used to determine association between radiomics features and lung function. Linear regression models were constructed for the forced-expiratory-volume-in-1s/forced-vital-capacity (FEV1/FVC) with radiomics features and LAA950.

Results: Never-smokers (n=270), at-risk (n=387), GOLD I (n=360), and GOLD II+ (n=270) participants were evaluated. CT radiomics features were significantly different when generated using no resampling versus resampling (19/32, p<0.004) and no airway removal versus removal (4/32, p<0.001) methods. Using resampled images with airway removal, 6/32 features were correlated with lung function (p<0.001). These six radiomics features significantly differentiated between all COPD groups (p<0.02), while LAA950 did not (p>0.3). In a multivariable regression model that included radiomics features and LAA950, radiomics features were significantly and independently associated for FEV1/FVC (p<0.05).

Conclusion: CT radiomics features were significantly and independently correlated with lung function, and better differentiated COPD disease severities than conventional CT measures. These findings indicate that CT radiomics features add complementary information that can aid in severity assessment in COPD patients.

Funding Support, Disclosures, and Conflict of Interest: This research was funded by NSERC. Dr. Kirby gratefully acknowledges support from the Parker B. Francis Fellowship Program and the Canada Research Chair Program (Tier II). Dr. Kirby is a consultant for Vida Diagnostics Inc.


Lung, Quantitative Imaging, CT


IM- CT: Radiomics

Contact Email