The goal of this study was to examine the utility of a radiomics model analysis based on computed tomography (CT) in distinguishing renal oncocytoma (RO) from chromophobe renal cell carcinoma and clear cell carcinoma, and in predicting the expression of Cytokeratin 7 (CK7). Patients with RO, chRCC, and ccRCC who underwent surgery between January 2013 and December 2019 constituted the training cohort, while data for the testing cohort was collected between January and October 2020. Segmentation of the corticomedullary (CMP) and nephrographic phases (NP) was performed manually, and textural parameters were retrieved using radiomics. After deciding on features to use, a support vector machine was developed using CMP and NP. The radiomics features were interpreted using Shapley additive explanations. The elements that were chosen during the first 2 stages were used to create a radiomics signature, and the radiomics nomogram was developed by combining the radiomics features with clinical parameters. To compare the models mentioned above in the 2 groups, the researchers computed their receiver operating characteristic curves. Moreover, a Rad-score correlation analysis was performed with CK7. The training cohort included 123 individuals diagnosed with RO, chRCC, or ccRCC, while the testing cohort consisted of 57 patients. In the end, 396 radiomics features were chosen throughout all stages. The area under the curve values of 0.941 and 0.935 was achieved in the training and testing sets, respectively, by the radiomics features combining 2 phases. Rad-score and CK7 were found to have a statistically significant association, as measured by Pearson’s coefficient. The researchers propose a personalized, non-invasive CT-based radiomics nomogram to differentiate preoperatively between RO, chRCC, and ccRCC and to predict immunohistochemistry protein expression for optimal clinical diagnosis and treatment.