Original Article
Quantitative features can predict further growth of persistent pure ground-glass nodule
Abstract
Background: To evaluate whether quantitative features of persistent pure ground-glass nodules (PGGN) on the initial computed tomography (CT) scans can predict further nodule growth.
Methods: This retrospective study included 59 patients with 101 PGGNs from 2011 to 2012, who received regular CT follow-up for lung nodule surveillance. Nineteen quantitative image features consisting of 8 volumetric and 11 histogram parameters were calculated to detect lung nodule growth. For the extraction of the quantitative features, semi-automatic GrowCut segmentation was implemented on chest CT images in 3D slicer platform. Univariate and multivariate analyses were performed to identify risk factors for nodule growth.
Results: With a median follow-up of 52 months, nodule growth was detected in 10 nodules by radiological assessment and in 16 nodules by quantitative features. In univariate analysis, 3D maximum diameter (MD), volume, mass, surface area, 90% percentile, and standard deviation value (SD) of PGGN on the initial CT scan were significantly different between stable nodules and nodules with further growth. In multivariate analysis, MD [hazard ratio (HR), 3.75; 95% confidence interval (CI), 2.14−6.55] and SD (HR, 2.06; 95% CI, 1.35−3.14) were independent predictors of further nodule growth. Also, the area under the curve was 0.896 (95% CI: 0.820−0.948) and 0.813 (95% CI: 0.723−0.883) for MD with a cut-off value of 10.2mm and SD of 50.0 Hounsfield Unit (HU). Besides, the growth rate was 55.6% (n=15) of PGGNs with MD >10.2 mm and SD >50.0 HU.
Conclusions: Based on the initial CT scan, the quantitative features can predict PGGN growth more precisely. PGGN with MD >10.2 mm and SD >50.0 HU may require close follow-up or surgical intervention for the high incidence of growth.
Methods: This retrospective study included 59 patients with 101 PGGNs from 2011 to 2012, who received regular CT follow-up for lung nodule surveillance. Nineteen quantitative image features consisting of 8 volumetric and 11 histogram parameters were calculated to detect lung nodule growth. For the extraction of the quantitative features, semi-automatic GrowCut segmentation was implemented on chest CT images in 3D slicer platform. Univariate and multivariate analyses were performed to identify risk factors for nodule growth.
Results: With a median follow-up of 52 months, nodule growth was detected in 10 nodules by radiological assessment and in 16 nodules by quantitative features. In univariate analysis, 3D maximum diameter (MD), volume, mass, surface area, 90% percentile, and standard deviation value (SD) of PGGN on the initial CT scan were significantly different between stable nodules and nodules with further growth. In multivariate analysis, MD [hazard ratio (HR), 3.75; 95% confidence interval (CI), 2.14−6.55] and SD (HR, 2.06; 95% CI, 1.35−3.14) were independent predictors of further nodule growth. Also, the area under the curve was 0.896 (95% CI: 0.820−0.948) and 0.813 (95% CI: 0.723−0.883) for MD with a cut-off value of 10.2mm and SD of 50.0 Hounsfield Unit (HU). Besides, the growth rate was 55.6% (n=15) of PGGNs with MD >10.2 mm and SD >50.0 HU.
Conclusions: Based on the initial CT scan, the quantitative features can predict PGGN growth more precisely. PGGN with MD >10.2 mm and SD >50.0 HU may require close follow-up or surgical intervention for the high incidence of growth.
