Robust, Standardized Quantification of Pulmonary Emphysema in Low Dose CT Exams

Abstract

RATIONALE AND OBJECTIVES: The aim of this study was to present and evaluate a fully automated system for emphysema quantification on low-dose computed tomographic images. The platform standardizes emphysema measurements against changes in the reconstruction algorithm and slice thickness. MATERIALS AND METHODS: Emphysema was quantified in 149 patients using a fully automatic, in-house developed software (the Robust Automatic On-Line Pulmonary Helper). The accuracy of the system was evaluated against commercial software, and its reproducibility was assessed using pairs of volume-corrected images taken 1 year apart. Furthermore, to standardize quantifications, the effect of changing the reconstruction parameters was modeled using a nonlinear fit, and the inverse of the model function was then applied to the data. The association between quantifications and pulmonary function testing was also evaluated. The accuracy of the in-house software compared to that of commercial software was measured using Spearman's rank correlation coefficient, the mean difference, and the intrasubject variability. Agreement between the methods was studied using Bland-Altman plots. To assess the reproducibility of the method, intraclass correlation coefficients and Bland-Altman plots were used. The statistical significance of the differences between the standardized data and the reference data (soft-tissue reconstruction algorithm B40f; slice thickness, 1 mm) was assessed using a paired two-sample t test. RESULTS: The accuracy of the method, measured as intrasubject variability, was 3.86 mL for pulmonary volume, 0.01% for emphysema index, and 0.39 Hounsfield units for mean lung density. Reproducibility, assessed using the intraclass correlation coefficient, was >0.95 for all measurements. The standardization method applied to compensate for variations in the reconstruction algorithm and slice thickness increased the intraclass correlation coefficients from 0.87 to 0.97 and from 0.99 to 1.00, respectively. The correlation of the standardized measurements with pulmonary function testing parameters was similar to that of the reference (for the emphysema index and the obstructive subgroup: forced expiratory volume in 1 second, -0.647% vs -0.615%; forced expiratory volume in 1 second/forced vital capacity, -0.672% vs -0.654%; and diffusing capacity for carbon monoxide adjusted for hemoglobin concentration, -0.438% vs -0.523%). CONCLUSIONS: The new emphysema quantification method presented in this report is accurate and reproducible and, thanks to its standardization method, robust to changes in the reconstruction parameters.