The Value of Intratumoral and Peritumoral Radiomics Based on Contrast-Enhanced Ultrasound for Predicting WHO/ISUP...

To develop a multimodal predictive model that integrates intratumoral and peritumoral radiomic features, contrast-enhanced ultrasound (CEUS) quantitative parameters, and clinical characteristics to enhance preoperative World Health Organization/International Society of Urological Pathology (WHO/ISUP) grading accuracy for clear cell renal cell carcinoma (ccRCC).

This retrospective study analyzed preoperative CEUS data from 186 histopathologically confirmed ccRCC patients, who were randomly divided into training (n = 148) and testing (n = 38) cohorts. Radiomic features were extracted and selected from intratumoral regions and 5-mm peritumoral regions on CEUS images, and 6 logistic regression (LR)-based predictive models were subsequently constructed: 5 standalone models (Intra, Peri5mm, ImageFusion5mm, IntraPeri5mm, and C-CEUS) and 1 combined model that integrated features from the best-performing radiomic model and C-CEUS. Additionally, subgroup analyses based on tumor size and CEUS wash-in rate were performed to verify the combined model's stability. Finally, a nomogram derived from the combined model was established for intuitive preoperative prediction of WHO/ISUP grades.

The radiomic model of IntraPeri5mm demonstrated the highest discriminative performance among the 4 radiomic features. The area under the curve (AUC) reached 0.785 in the testing cohort. Multivariate analysis identified delta perfusion index (dPI) and Maximum diameter on the largest cross-section (sizemax) as independent predictors of the WHO/ISUP grading (all p < .05). The combined model incorporating IntraPeri5mm radiomic features, clinical variables (sizemax), and CEUS parameters (dPI) demonstrated improved predictive accuracy, with AUC of 0.852 (0.706-0.998) in the testing cohort and an accuracy of 0.842 (95% CI: 0.687-0.940). Moreover, the AUC values for all subgroups exceeded 0.80.

The combined model which may enhance personalized risk stratification outperformed single-modality approaches in preoperative WHO/ISUP grading of ccRCC.