Comparison of three software systems for semi-automatic volumetry of pulmonary nodules on baseline and follow-up CT examinations

Background: Early diagnosis of lung cancer in a treatable stage is the main purpose of lung cancer screening by computed tomography (CT). Accurate three-dimensional size and growth measurements are essential to assess the risk of malignancy. Nodule volumes can be calculated by using semi-automated volumetric software. Systematic differences in volume measurements between packages could influence nodule categorization and management decisions. Purpose: To compare volumetric measurements of solid pulmonary nodules on baseline and follow-up CT scans as well as the volume doubling time (VDT) for three software packages. Material and Methods: From a Lung Cancer Screening study (NELSON), 50 participants were randomly selected from the baseline round. The study population comprised participants with at least one pulmonary nodule at the baseline and consecutive CT examination. The volume of each nodule was determined for both time points using three semi-automated software packages (P-1, P-2, and P-3). Manual modification was performed when automated assessment was visually inaccurate. VDT was calculated to evaluate nodule growth. Volume, VDT, and nodule management were compared for the three software packages, using P-1 as the reference standard. Results: In 25 participants, 147 nodules were present on both examinations (volume: 12.0-436.6 mm(3)). Initial segmentation at baseline was evaluated to be satisfactory in 93.9% of nodules for P-1, 84.4 % for P-2, and 88.4% for P-3. Significant difference was found in measured volume between P-1 and the other two packages (P < 0.001). P-2 overestimated the volume by 38 +/- 24%, and P-3 by 50 +/- 22%. At baseline, there was consensus on nodule size categorization in 80% for P-1& P-2 and 74% for P-1& P-3. At follow-up, consensus on VDT categorization was present in 47% for P-1& P-2 and 44% for P-1& P-3. Conclusion: Software packages for lung nodule evaluation yield significant differences in volumetric measurements and VDT. This variation affects the classification of lung nodules, especially in follow-up examinations.