Registration-based lung mechanical analysis of chronic obstructive pulmonary disease (COPD) using a supervised machine learning framework

Abstract

Rationale and objectives: This study evaluated the performance of computed tomography (CT)-derived biomechanical based features of lung function and the presence and severity of chronic obstructive pulmonary disease (COPD). It performed well when compared to CT-derived density and textural features of lung function and the presence and severity of COPD.

Materials and methods: A total of 162 subjects (Global Initiative for Chronic Obstructive Lung Disease [GOLD] stages 0-4 and nonsmokers) subjects with CT scan performed at total lung capacity or expiration to functional residual capacity were evaluated. CT-derived biomechanical, density, and textural feature sets were compared to forced expiratory volume in 1 second (FEV1)%, FEV1/forced vital capacity, and total St. George's respiratory questionnaire scores. The ability of these feature sets to assess the presence and severity of COPD was also evaluated. Optimal features are selected by linear forward feature selection and the classification is done using k nearest neighbor learning algorithm.

Results: The proposed biomechanical features showed good correlations with the pulmonary function tests and health status metrics. In COPD versus non-COPD classification, biomechanical feature set achieved an area under the curve (AUC) of 0.85 performing well in comparison to density (AUC = 0.83) and texture (AUC = 0.89) feature sets. Classifying the subjects into the severity of GOLD stage using biomechanical features (AUC = 0.81) performed better than the density- and texture-based feature sets, AUC = 0.76 and 0.73, respectively. The biomechanical features performed better alone than in combination with the other two feature sets.

Conclusion: This study shows the effectiveness of CT-derived biomechanical measures in the assessment of airflow obstruction and quality of life in subjects with COPD. CT-derived biomechanical features performed well in assessing the presence and severity of COPD.

Figure 1

Flow chart explaining the steps involved from image acquisition to classification experiments

Figure 2

Axial slices of a GOLD1 (Left column in each filter section) and GOLD4 (Right column in each filter section) COPD subject. a, d) Convolution with Gaussian at 1.2mm standard deviation b, e) Convolution with Gaussian at 2.4mm standard deviation c, f) Convolution with Gaussian at 4.8mm standard deviation g, j) Gradient magnitude of Gaussian at 1.2mm standard deviation h, k) Gradient magnitude of Gaussian at 2.4mm standard deviation i, l) Gradient magnitude of Gaussian at 4.8mm standard deviation m, p) Laplacian of Gaussian at 1.2mm standard deviation n, q) Laplacian of at 2.4mm standard deviation o, r) Laplacian of Gaussian at 4.8mm standard deviation

Figure 3

Axial slices of a mild COPD (GOLD1) and severe COPD (GOLD4) subject. a) Jacobian map of GOLD1 subject b) Jacobian map of GOLD4 subject c) Strain map of GOLD1 subject d) Strain map of GOLD4 subject e) Anisotropic deformation index map of GOLD1 subject f) Anisotropic deformation of GOLD4 subject

Figure 4

K nearest neighbor algorithm