Diagnostic performance of CT-based node-RADS for detecting metastatic lymph nodes in melanoma and comparison with
Renato Fabrizio1, Letizia Di Meglio1, Maurizio Cè1
1Department of Radiology, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico di Milano, Via Francesco Sforza 28, 20122, Milan, Italy.
Objective:
To evaluate the diagnostic performance of the CT-based Node-RADS score for assessing metastatic lymph node involvement in patients with cutaneous melanoma and to compare it with short axis and roundness index.
Methods:
This retrospective study included patients with histologically confirmed melanoma who underwent contrast-enhanced CT before sentinel lymph node biopsy. A control cohort of trauma patients free from any condition potentially affecting lymph nodes was included to represent physiological lymph nodes and to address class imbalance. Lymph nodes were independently assessed by two radiologists and re-evaluated after one month; inter- and intra-observer agreement were evaluated using Cohen's kappa. Diagnostic performance was assessed using ROC curve analysis, and AUCs were compared with the DeLong test.
Results:
A total of 123 lymph nodes were analyzed, of which 55 (45%) were metastatic. Inter- and intra-observer agreement was excellent (κ = 0.84-0.97 and 0.85-0.98, respectively). All imaging features differed significantly between metastatic and non-metastatic lymph nodes (p < 0.001). Node-RADS yielded an AUC of 0.72 (95% CI 0.65-0.79), not superior to the short axis (AUC 0.80; p = 0.13) or long axis (AUC 0.71; p = 0.76), but outperforming the roundness index (AUC 0.60; p = 0.008). The short axis showed the highest discriminatory performance. Node-RADS showed high specificity (0.96) but low sensitivity (0.47), with 40% of low-risk nodes resulting metastatic. On multivariable analysis, only shape was independently associated with metastatic involvement (OR 6.69; p = 0.006).
Conclusion:
Node-RADS does not outperform short-axis measurement and shows limited sensitivity in subclinical nodal disease, highlighting the need for advanced detection approaches.
