A CT-based radiomics model was proposed to increase the accuracy of preoperative noninvasive differentiation of fp-AML from ccRCC. There were 774 three-dimensional radiomics features extracted from CT images. The feature selection was carried by three steps: Pearson’s correlation matrices were calculated to remove redundant features; Welch’s t-test was used to determine the statistically significant features; and sequential forward floating selection method was utilized to select the discriminative features. The sparse radial basis function neural network was employed for classification. Results show that the radiomics model yields accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curves of 90.00%, 66.67%, 100.0%, and 0.9173, respectively. The reliability of model was assessed by probabilistic outputs of classifiers. When the probability threshold is 0.95, the model obtains confident classification accuracy, undecided rate, and misclassification rate of 71.67%, 25.00%, and 3.33%, respectively. Results demonstrate that the proposed radiomics model can achieve reliable discrimination of fp-AML from ccRCC.