Following the described methodology, it was determined that the GLSZM, GLDM, and NGTDM groups do not contain any features that meet the established robustness criteria. In contrast, the First-Order and GLRLM groups each had two features that met the criteria in some of the analyses. Overall, radiomic features demonstrated specific repeatability and reproducibility when extracted from a phantom using fsCT. The results obtained are consistent with the reliability of these features in pCT. Additionally, this methodology can be applied to dual-energy CT using other technologies.

Limitations

This study presents several limitations that should be considered. First, all acquisitions were performed using a single CT model with a specific technology. Additionally, only one platform was analyzed for radiomic feature extraction, which could influence the robustness of these features, as indicated in previous studies such as Zhong et al. [10].