Diagnostic performance of node-RADS classification for primary lymph node assessment in rectal cancer: a modality benchmarking study.

Abstract

Purpose: To evaluate how well the Node Reporting and Data System (Node-RADS) diagnoses lymph node involvement (LNI) in the initial stages of rectal cancer, utilizing contrast-enhanced CT (CE-CT), T2-weighted MRI (T2WI) and contrast-enhanced T1-weighted MRI (T1CE).

Methods: This retrospective study included 113 rectal cancer patients who underwent radical surgery without neoadjuvant therapy. Two radiologists independently assessed regional lymph nodes using the highest NODE-RADS classification and histopathology as reference criteria. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curve analysis. Statistical analysis was performed using the McNemar test with Bonferroni correction for multiple comparisons.

Results: Node-RADS showed improved diagnostic performance over short-axis diameter (SAD) in all modalities (AUC: 0.838 vs. 0.744 for CE-CT, 0.845 vs. 0.747 for T2WI, 0.853 vs. 0.786 for T1CE; all P < 0.05). The sensitivity and specificity of Node-RADS across three modalities ranged from 76.19 - 78.57% and 91.55 - 92.96%, respectively. Pairwise comparisons of sensitivity and specificity among the three modalities showed no significant differences after Bonferroni correction (all McNemar test P = 1.0). There was no significant difference in Node-RADS performance among the three modalities (all P > 0.05). The weighted kappa values were 0.742-0.798.

Conclusion: Node-RADS demonstrated superior diagnostic performance over SAD measurements and similar diagnostic effectiveness in assessing LNI for primary rectal cancer stages across CE-CT, T2WI, and T1CE.