Differentiation of venous thrombus consistency in patients with renal cell carcinoma using volumetric analysis of computed tomography first-order radiomic features

Renal cell carcinoma (RCC) has a unique tendency to invade the inferior vena cava through the renal vein, forming a venous tumor thrombus (VTT). Precisely determining the solidity of the VTT before surgery is crucial for appropriate treatment planning and optimal outcomes, as the perioperative spread of friable thrombus can lead to life-threatening complications like pulmonary embolism. The aim of the study was to assess the applicability of CT volumetric analysis to differentiate VTT consistency in patients with RCC. Twenty-eight patients with RCC and VTT were enrolled in the study, in whom nephrectomy and thrombectomy were performed. Preoperative abdominal multiphasic contrast-enhanced CT was executed. Patients were categorized into solid or friable groups according to the consistency of the thrombus. For the volumetric texture analysis, the first-order radiomic features were calculated from the entire thrombus volume separately for each CT phase. A significant association between the percentage of RCC cells in VTT and the radiodensity values of mean, median, 10th percentile of corticomedullary phase (CMP), mean and median of nephrographic phase (NP) was found. The percentage of RCC cells within the tumour thrombus showed a significant correlation with the CMP median (Pearson’s r = 0.748), CMP mean (Pearson’s r = 0.731), CMP 90th percentile (Pearson’s r = 0.638), CMP 10th percentile (Pearson’s r = 0.622) and NP median (Pearson’s r = 0.450). The highest performance in differentiating between solid and friable VTT demonstrated the CMP median, with an area under the curve (AUC) of 0.934 (95% CI 0.846-1.0), sensitivity of 90% and specificity of 62%. The highest sensitivity and specificity in differentiating between solid and friable VTT showed CMP 90th percentile: 90% and 75%, respectively, AUC 0.928 (95% CI 0.830-1.0). The highest specificity (100%) had CMP mean, while the sensitivity and AUC were 85% and 0.925 (95% CI 0.827-1.0), respectively. First-order radiomic features of CT images, such as the mean, median and 90th percentile of CMP of the whole-volume VTT had the potential to predict VTT consistency in RCC.