@article{article_1882022, title={Integration of radiomics with MRI clear cell likelihood score for classification of clear cell renal cell carcinoma}, journal={Ege Tıp Dergisi}, volume={65}, pages={313–320}, year={2026}, DOI={10.19161/etd.1882022}, url={https://izlik.org/JA29RR25WE}, author={Karabulut, Ahmet Kasım and Koska, İlker Özgür and Turgut, Ali Çağlar and Sarsik Kumbaraci, Banu and Kızılay, Fuat and Güler, Ezgi}, keywords={renal cell carcinoma, magnetic resonance imaging, clear cell, radiomics, machine learning}, abstract={Aim: To compare the diagnostic performance of the MRI-based clear cell likelihood score (ccLS), radiomics-based machine learning models, and their combination for differentiating clear cell renal cell carcinoma (ccRCC) from other renal tumor subtypes. Materials and Methods: This single-center retrospective study included patients with solid renal masses who underwent multiparametric MRI and had histopathologic confirmation. Lesions were evaluated using ccLS by two independent readers. Radiomic features were extracted from T1-weighted and T2-weighted images following standardized preprocessing. Multiple machine learning pipelines combining different feature selection methods and classifiers were evaluated using stratified 10-fold cross-validation with four repetitions. Using multivariable logistic regression, the radiomics score together with clinical factors and semantic imaging features were evaluated, and a nomogram was constructed based on the selected variables. Results: For differentiating ccRCC from other renal tumor subtypes, the ccLS model achieved an area under the receiver operating characteristic curve (AUC) of 0.832 (0.759–0.899). The best-performing radiomics-based machine learning model achieved a mean AUC of 0.89 ± 0.10. The combined model demonstrated higher diagnostic performance, with a mean AUC of 0.96 ± 0.01. Conclusions: Combining radiomics-based machine learning with established semantic MRI assessments was associated with improved differentiation of ccRCC from other renal tumor subtypes compared with individual approaches. Further validation in larger, multicenter cohorts is warranted before broader clinical application.}, number={2}