Pulmonary adenocarcinomas with ground-glass attenuation on thin-section CT: Quantification by three-dimensional image analyzing method

Purpose: The purpose of this study was to evaluate software designed to calculate whole tumor volumes and the ratio of the solid component to whole volume (%solid) in pulmonary nodules with ground-glass opacity in three dimensions. Methods: The study included 49 patients with histologically diagnosed adenocarcinomas smaller than 2 cm in diameter. The %solid was calculated both automatically using new software, and by manual measurement of the following four parameters by two observers: the ratio of the largest diameter (a) and the area (b) at the mediastinal window to those at the lung window, and the ratio of the largest diameter (c) and the area (d) of the solid component to those of the ground-glass component at the lung window. Agreement of intra- and inter-observer data by both Spearman's rank correlation test and Bland-Altman's method, and a comparison by Spearman's rank correlation test of the %solid in both Noguchi sub-classifications and vessel invasion in histologic specimens, between the software and manual methods, were assessed. Results: Of the 49 nodules, 48 were successfully measured and assessed. The agreement of the observers with the software was better (Bland-Altman's method; mean difference, -0.3%; 95% limits of agreement, -3.1 to 2.5%) than with the manual measurements (a: 5.3%, -17.6 to 28.3%; b: 8.3%, -10.6 to 26.9%; c: 10.7%, -17.6 to 39%; d: 6.4%, -22 to 34.8%). The correlation between %solid and the histological group was worse with the software (Spearman's rank correlation test; r = 0.487, p < 0.001) than with the manual method (a, r = 0.534; b, r = 0.557; c, r = 0.552; d, r = 0.545). Conclusion: Although the software requires improvement in the calculation of %solid with volumetric analysis, this is a reproducible and promising quantitative method for determining the grades of malignancy of small lung cancers. (c) 2007 Elsevier Ireland Ltd. All rights reserved.