A head-to-head comparison of MRI and CT using Node-RADS for rectal cancer lymph nodes: Observer agreement and...

To compare the assessment agreement and diagnostic performance of CT and MRI using the Node Reporting and Data System (Node-RADS) for preoperative regional lymph node metastasis (LNM) in rectal cancer.

This retrospective study included 97 rectal adenocarcinoma patients who underwent preoperative contrast-enhanced CT and MRI. A total of 148 lymph nodes (LNs) were evaluated. Postoperative histopathology served as the reference standard. CT-based and MRI-based Node-RADS scores, as well as the ESGAR category, were assessed. Diagnostic performance was evaluated by receiver operating characteristic (ROC) curve analysis, including area under the curve (AUC), sensitivity, and specificity. Inter- and intra-observer agreement for Node-RADS was evaluated using weighted kappa statistics. A subgroup analysis was further conducted on 27 image-pathology matched LNs.

Node-RADS scores showed a strong positive correlation with pathological N stage (MRI: r = 0.724; CT: r = 0.654; both p < 0.01). MRI-based Node-RADS achieved the highest diagnostic performance $\text{(AUC =}0.883\text{)}$ , which was significantly higher than CT-based Node-RADS and the ESGAR . Inter-reader agreement for Node-RADS was good to excellent for both modalities, and was higher for MRI ( $\text{κ}=0.921$ ) than CT ( $\text{κ =}0.785$ ). For small LNs (<5 mm in short axis diameter), MRI-based Node-RADS demonstrated a higher AUC than CT-based Node-RADS (0.895 vs. 0.760). In the imaging-pathology matched subgroup, MRI-based Node-RADS $\text{(AUC =}0.793\text{)}$ also outperformed CT-based Node-RADS $\text{(AUC =}0.712\text{)}$ .

Node-RADS provides a standardized tool for preoperative regional lymph node assessment in rectal cancer, showing better diagnostic performance than ESGAR criteria. MRI demonstrates higher diagnostic performance and inter-observer agreement than CT, supporting MRI as the preferred modality.