The diagnostic efficacy of venous-phase spectral CT combined with Node-RADS for differentiating enlarged lymph nodes...

To investigate the value of combining quantitative parameters from spectral CT with the Node Reporting and Data System 1.0 (Node-RADS) in distinguishing between benign and malignant lymph nodes enlarged due to solid tumors, and to explore a combined multiparametric approach.

We retrospectively analyzed data from patients with primary solid tumors accompanied by lymphadenopathy who underwent venous-phase spectral CT scans with energy-resolved imaging. A total of 243 lymph nodes were included in this study. Based on pathological findings, clinical observations, and CT follow-up, the patients were divided into malignant (n = 124) and benign (n = 119) groups. The groups were compared for general characteristics. For each lymph node, we measured the venous-phase spectral CT parameters, including iodine concentration (IC), normalized iodine concentration (NIC), effective atomic number (Z_eff), and CT values of virtual monoenergetic images (VMI) at different energy levels (40 keV,70 keV, 100 keV). We also calculated the spectral curve slope (λ) for different energy ranges. A node-RADS score was assigned to each lymph node. Variables were screened using univariate and multivariate logistic regression analyses and applied to develop a combined exploratory approach. The model's diagnostic performance was evaluated using receiver operating characteristic (ROC) curves. The differences in the area under the curve (AUC) values between the two models were assessed using the DeLong test. Calibration analysis and bootstrap internal validation were also performed.

The malignant and benign groups showed significant differences (P < 0.05) in Node-RADS scores, Z_eff, CT values for VMI (40 keV, 70 keV, 100 keV), λ, and IC. Based on the multivariate logistic regression analysis, the final multivariable regression model included the CT value at 100 keV on VMI (OR = 0.970, 95% confidence interval [CI]: 0.949-0.991, P = 0.005) and Node-RADS score (OR = 3.856, 95% CI: 2.666-5.576, P = 0.001). At a cutoff value of 0.495, the AUC, sensitivity, specificity, and accuracy of the combined model for identifying metastatic lymph nodes were 0.819 (95% CI: 0.766-0.872), 72.6% (95% CI: 0.638-0.802), 78.2% (95% CI: 0.696-0.853), and 75.3% (95% CI: 0.694-0.806), respectively. The DeLong test revealed significant differences in the AUCs between the combined model and the single Node-RADS model for differentiating the benign or malignant nature of lymph nodes (Z = 2.138, P = 0.033). Bootstrap validation confirmed model stability, and the combined model demonstrated good calibration (Brier score = 0.173).

An exploratory approach combining quantitative venous-phase spectral CT (VMI _{100 keV}) parameters with Node-RADS scoring may provide a modest statistical improvement when identifying whether enlarged lymph nodes are benign or malignant. These preliminary exploratory observations suggest an objective imaging approach that warrants further prospective and multicenter validation, alongside formal clinical utility analyses, before considering routine clinical implementation.