Abstract
Objective: To investigate the effect of radiomics models obtained from dual-energy CT (DECT) material decomposition images and virtual monoenergetic images (VMIs) in predicting the pathological grading of bladder urothelial carcinoma (BUC).
Materials and Methods: Preoperative Energy-Spectrum CT images were retrospectively collected from 112 pathologically confirmed cases of BUC patients, including 76 cases of high-grade urothelial carcinoma and 36 cases of low-grade urothelial carcinoma. Enhanced CT venous phase images of all patients were reconstructed at 40 to 140 keV VMIs (interval 10 keV), Iodine maps, and Water maps, and a total of 13 sets of images were obtained, and imaging features were extracted in each of the 13 sets of images. The best features related to BUC were identified by recursive feature elimination (RFE), the Minimum Redundancy Maximum Relevance (mRMR), and the Least Absolute Shrinkage and Selection Operator (LASSO) in order. A five-fold cross-validation method was used to divide the samples into training and testing sets, and models for pathological prediction of BUC grading were constructed by a random forest (RF) classifier. ROC curves were plotted to evaluate the performance of 13 models obtained from reconstructed images.
Results: There were no statistically significant differences in the area under the curve (AUC) between the training set and the testing set for all 13 models, with the AUC ranging from 0.91 to 0.96 in the training set and 0.84 to 0.90 in the testing set for each group of reconstructed images. Although the features selected for the reconstructed images were very different among the groups, all the features selected from 40 to 100 keV VMIs had dependencevariance of the GLDM feature set.
Conclusion: The variation of spectral CT parameters did no effect on the radiomics-based prediction of the pathological grading of BUC and did not affect the accuracy of the model even if the relevant features differed between reconstructed images.