Node-RADS 1.0 for lymph node involvement detection in renal cell carcinoma: A study on 2D and 3D CT imaging

Abstract

Objectives

To evaluate the diagnostic accuracy of the node reporting and data system 1.0 (Node-RADS) score in predicting lymph node (LN) involvement (LNI) in renal cell carcinoma (RCC) patients and to compare the diagnostic performance of Node-RADS score based on two-dimensional (2D) and three-dimensional (3D) modality.

Methods

From January 2012 to September 2024, patients with RCC and histologically confirmed LN status who underwent preoperative abdomen computed tomography imaging were retrospectively enrolled in the study. Radiological assessments were performed independently and blinded by three readers according to the Node-RADS, and both Node-RADS scores based on 2D and 3D modality were recorded. The gold reference of LNI was determined by the histological results after surgery. The sensitivity, specificity, area under receiver operating characteristic curve (AUC) and inter-agreements for different features were calculated.

Results

Finally, 375 patients (median age 55  years; interquartile range (IQR), 45–63 years), including 270 negative LNI and 105 positive LNI, were enrolled in the study. Younger patients showed higher LNI prevalence (p < 0.001). Node-RADS score ≥ 3 yielded a sensitivity of 0.876, a specificity of 0.926 and an AUC of 0.901 (95% CI 0.866–0.937) on 2D, which was comparable to Node-RADS score on 3D (Delong test: p = 0.651). An excellent inter-reader agreement was observed for Node-RADS score between three readers based on 2D and 3D modality (Kendall’s W 0.889 and 0.869).

Conclusions

The Node-RADS score demonstrates high overall accuracy in identifying LNI in patients with RCC. Node-RADS score based on 2D and 3D modality demonstrate comparable diagnostic performance.