Predicting sinus fat invasion in clear cell RCC: A combined model of inflammatory markers, spectral CT, and ECV

Objective To evaluate the diagnostic value of combining preoperative inflammatory markers, spectral CT quantitative parameters, and extracellular volume fraction (ECV) for predicting sinus fat invasion (SFI) in clear cell renal cell carcinoma (ccRCC), and to construct and validate an optimal noninvasive preoperative prediction model. Methods This retrospective study analyzed clinical and imaging data from 380 patients with pathologically confirmed ccRCC who underwent preoperative spectral CT between 2020 and 2025. Patients were divided into SFI and non-SFI groups based on postoperative pathology. Inflammatory markers (SII, NLR, PLR, SIRI), spectral CT parameters (NIC, K, Z-eff, 70 keV CT value), and ECV were collected and compared. Independent predictors were identified via univariate and multivariate logistic regression. A combined prediction model was developed, internally validated, and assessed using ROC curves, calibration curves, decision curve analysis (DCA), and SHAP analysis. Results Multivariate analysis retained eight independent predictors: SII (OR = 1.131), NLR (OR = 1.294), SIRI (OR = 1.072), NIC (OR = 0.369), K (OR = 0.149), Z-eff (OR = 0.703), 70 keV CT value (OR = 0.917), and ECV (OR = 1.141). PLR and LMR were not independent predictors (P > 0.05). The combined model demonstrated outstanding diagnostic performance in the training set (AUC = 0.966, sensitivity = 93.90%, specificity = 97.83%) and maintained robust performance in the validation set (AUC = 0.900, sensitivity = 89.13%, specificity = 85.29%). Validation set specificity was 85.29% (95% CI 76.88%-93.71%). The difference from training specificity (97.83%, 95% CI 95.72%-99.93%) was not statistically significant by overlapping CIs, though the point estimate suggests some overfitting. External validation is needed. Calibration and DCA indicated good model fit and clinical utility. SHAP analysis highlighted NIC as the most influential feature. Conclusion A multimodal model integrating inflammatory markers, spectral CT parameters, and ECV effectively and noninvasively predicts SFI in ccRCC patients preoperatively, offering superior diagnostic performance compared to single parameters, thereby supporting personalized surgical planning. For patients predicted to be at high risk for SFI, radical nephrectomy or extended partial nephrectomy with intraoperative margin assessment should be considered, though partial nephrectomy may remain feasible in select cases by experienced surgeons. The model should complement, not replace, intraoperative judgment.