Fig 1.
Flow diagram illustrating the inclusion/exclusion process of this study.
Fig 2.
The screen-capture image shows the brief process of texture analysis using the in-house software program. The segmentation of pulmonary nodules is manually performed, and texture features of the nodules are automatically extracted by the software program.
Table 1.
Intraclass correlation coefficients for inter-observer reproducibility of texture features.
Table 2.
Histographic, Volumetric, and morphologic features of pulmonary metastases and non-metastatic lesions in all the pediatric patients with osteosarcoma.
Table 3.
Histographic, Volumetric, and morphologic features of pulmonary metastases and non-metastatic lesions in small non-calcified pulmonary nodules.
Fig 3.
Correlation matrix of texture features.
The correlation matrix was created with 10 of 15 texture features those showed statistically significant differences in independent t-test. The standard deviation and variance showed significant correlation with mean attenuation. The surface area, volume and GLCM entropy showed significant correlation with effective diameter. Pearson correlation coefficient of each pair of texture features was shown in the upper right corner of the scatter plot. *P < 0.05; **P < 0.01.
Fig 4.
Texture analysis of pulmonary nodules: Metastatic nodules versus non-metastatic nodules.
(A) CT scan shows a 12.8-mm solid pulmonary nodule (arrow) with calcification. This nodule shows high mean attenuation (123.6 ± 289.6 HU). The pulmonary nodule was confirmed as a pulmonary metastasis. (B) CT scan shows a 2.9-mm small non-calcified pulmonary nodule (arrow). This nodule shows relatively high mean attenuation (-8.8 ± 255.8 HU). It was confirmed as a pulmonary metastasis. (C) CT scan shows a 2.4-mm small non-calcified pulmonary nodule (arrow). This nodule has relatively low mean attenuation (-229.9 ± 212.3 HU). It was confirmed as an intrapulmonary lymph node.
Table 4.
Results of logistic regression analysis for predictors of pulmonary metastases and non-metastatic lesions in pulmonary nodules.
Fig 5.
Receiver operator characteristic (ROC) curve analysis.
Receiver operator characteristic (ROC) curve for mean attenuation and effective diameter obtained by texture analysis and conventional measurement for differentiating pulmonary metastases and non-pulmonary metastatic lesions. ROC was performed for mean attenuation (A) and effective diameter (B) in the total group, and for mean attenuation in the small non-calcified nodule group (C).