The influence of CT acquisition and reconstruction parameters on the stability of radiomic features of pure ground-glass nodules

Abstract

Objective: To investigate the influence of CT reconstruction algorithm, radiation dose, and contrast agent on the stability of radiomic features of pure ground-glass density pulmonary nodules. Methods: A total of 50 pure ground-glass density pulmonary nodules in 35 patients were prospectively selected from Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College in 2018. After reconstructing the original image of the same patient's pulmonary nodules, six sequences of different parameters were obtained. ITK-SNAP software was used to segment different sequences of pure ground-glass density pulmonary nodules. All scanning data were extracted by A. K. software. The radiomic features with good retest reliability were selected by the intraclass correlation coefficient. The statistical software of R language was used to analyze the characteristic parameters. All the radiomic feature values of different sequences were paired and compared. The number of radiomic features changed by acquisition and reconstruction parameters was counted. The influence of different parameters on the reproducibility of pure ground-glass density pulmonary nodules was compared. Results: A total of 391 radiomic features were extracted from 50 cases of pure ground-glass density pulmonary nodules. 320 features with an intraclass correlation coefficient greater than 0.75 were selected for further analysis. By changing the three parameters of CT reconstruction algorithm, radiation dose, and contrast agent simultaneously, the changed radiomic features of the pure ground-glass density pulmonary nodules reach 60.9% (195/320), including 6.7% (1/15) morphological feature, 100.0% (40/40) histogram features, and 58.1% (154/265) texture features. When only one parameter was changed (keeping the other two parameters unchanged), changing the CT reconstruction algorithm, radiation dose, and contrast agent respectively, the changed radiomic features of pure ground-glass density pulmonary nodules were 10.6% (34/320), 30.9% (99/320) and 50.6% (162/320), the difference was statistically significant (P<0.05). When the radiation dose and contrast agent were changed, the radiomic features obtained by the FBP reconstruction algorithm had smaller changes than the features obtained by the 50% ASiR-V algorithm (P=0.001). Conclusions: CT reconstruction algorithm, radiation dose, and contrast agent will affect the radiomic features of pure ground-glass density pulmonary nodules. The contrast agent has the most significant influence on the radiomic features, followed by radiation dose and CT reconstruction algorithm minimum. Compared with morphological features, histogram features and texture features are more likely to be affected by CT reconstruction algorithms, radiation doses, and contrast agents. Compared with the 50% ASiR-V algorithm, the radiomic features obtained by the FBP reconstruction algorithm are less affected by the radiation dose and contrast agent. The influence of these parameters should be fully considered in the radiomic analysis.