# Radiomic Profiling of Tumor Thrombus for Predicting Recurrence in Renal Cell Carcinoma

## TL;DR

Analyzing imaging features of tumor thrombus in kidney cancer improves recurrence prediction and supports personalized treatment.

## Contribution

Radiomic profiling of tumor thrombus adds significant predictive value beyond primary tumor features and clinical models.

## Key findings

- Tumor thrombus radiomic signatures achieved an iAUC of 0.78, outperforming primary tumor features.
- Incorporating TT radiomics improved clinical models' iAUC from 0.58 to 0.83.
- Decision curve analysis confirmed the clinical utility of TT radiomic features.

## Abstract

Integration of radiomic features from the tumor thrombus improves recurrence prediction in patients with clear cell renal cell carcinoma and tumor thrombus. This approach provides valuable insights into individual patient risk profiles and can facilitate more personalized treatment strategies.

Clear cell renal cell carcinoma (ccRCC) with tumor thrombus (TT) presents a significant prognostic challenge due to its high recurrence risk. Radiomics, an imaging-based biomarker approach, has primarily focused on the primary tumor, while the prognostic potential of TT radiomics remains largely unexplored. This study aimed to assess the added value of tumor thrombus radiomic signatures (RSs) in predicting recurrence in ccRCC patients with TT.

We conducted a retrospective analysis of patients undergoing surgical resection for nonmetastatic ccRCC with TT. Preoperative contrast-enhanced computed tomography images were used to extract radiomic features from the primary tumor and TT. Features were selected using least absolute shrinkage and selection operator (LASSO) Cox regression and incorporated into predictive models. Performance was assessed using the integrated area under the curve (iAUC), calibration, decision curve analysis (DCA), and incremental value over clinical models (pTNM, UISS, and Leibovich) for the prediction of disease-free survival (DFS).

A total of 166 patients (training set: n = 117; test set: n = 49) were included. The primary tumor RS achieved an iAUC of 0.69, the TT RS achieved 0.78, and the primary tumor + TT RS achieved 0.82. Incorporation of TT radiomics enhanced the predictive accuracy of clinical models significantly, with iAUC increases from 0.58 to 0.83 for pTNM, 0.64 to 0.83 for UISS, and 0.66 to 0.83 for Leibovich scores (all p < 0.001). The DCA confirmed the clinical utility of integrating radiomic features, particularly TT radiomics, into recurrence risk assessment. The retrospective design and absence of external validation in independent, multicenter cohorts limit the generalizability of these findings.

Tumor thrombus radiomic profiling improves DFS prediction significantly and adds complementary prognostic value to established models in patients with ccRCC. Incorporation of these features into clinical workflows may enhance risk stratification and guide personalized treatment planning. Prospective validation in large, multicenter cohorts is warranted to support clinical adoption.

This study focused on kidney cancer with tumor thrombus and demonstrated that an analysis of the imaging features from the thrombus improved the prediction of cancer recurrence significantly. This approach could enhance the understanding of individual patient risks and support more personalized treatment strategies.