# Standardizing cervical lymph node evaluation in papillary thyroid carcinoma: diagnostic accuracy of Node-RADS on ultrasound

## TL;DR

This study evaluates a new ultrasound scoring system (Node-RADS) for identifying cancer spread to neck lymph nodes in thyroid cancer patients, finding it moderately accurate and more reliable than traditional methods.

## Contribution

The study introduces Node-RADS as a standardized system for ultrasound-based lymph node evaluation in thyroid cancer, demonstrating its diagnostic accuracy and reproducibility.

## Key findings

- Node-RADS scores correlated with metastasis likelihood, with higher scores indicating greater malignancy rates.
- Node-RADS outperformed individual ultrasound features in diagnostic accuracy (AUC of 0.720).
- Combining Node-RADS with peripheral vascularity improved diagnostic performance (AUC of 0.825).

## Abstract

To assess the diagnostic performance of the Node Reporting and Data System (Node-RADS) in ultrasonographic evaluation of cervical lymph node (LN) metastasis in papillary thyroid carcinoma (PTC) and compare it with conventional suspicious sonographic features.

This retrospective study included 227 PTC patients (269 LNs) who underwent preoperative cervical LN ultrasound and biopsy (September 2022–December 2023). Two radiologists independently assigned Node-RADS scores (1–5) based on LN size and morphologic features (shape, border, echogenicity). Diagnostic performance was evaluated using ROC analysis, with histopathology as the reference standard. Node-RADS was compared to five American Thyroid Association (ATA)-defined suspicious features (microcalcifications, cystic changes, etc.). Multivariable analysis assessed an integrated model combining Node-RADS with peripheral vascularity.

Metastatic LNs (63.9%, 172/269) exhibited higher Node-RADS scores (malignancy rates: 35.7% for score 1 vs. 88.2% for score 5). Node-RADS achieved an AUC of 0.720 (95% CI: 0.662–0.773), surpassing individual features (best AUC: 0.703 for peripheral vascularity). Inter-reader agreement was good (κ = 0.67). A combined Node-RADS and peripheral vascularity model significantly improved diagnostic performance (AUC: 0.825, p < 0.01).

Node-RADS offers standardized, reproducible risk stratification for cervical LN metastasis in PTC, with moderate diagnostic accuracy. Integration with peripheral vascularity enhances performance, supporting its potential clinical utility in preoperative assessment.