# Super-resolution ultrasound quantifying microvascular alterations for early detection of metastatic cervical lymph nodes: a prospective diagnostic study

## TL;DR

This study shows that super-resolution ultrasound can detect blood vessel changes in lymph nodes to help identify cancer spread more accurately and reduce unnecessary biopsies.

## Contribution

The study introduces a multivariate model using SRUS parameters to improve early detection of metastatic cervical lymph nodes.

## Key findings

- Metastatic lymph nodes showed significantly higher vascular density and fractal dimension compared to reactive nodes.
- A multivariate model combining SRUS parameters achieved an AUC of 0.813 for distinguishing metastatic from reactive lymph nodes.
- Velocity entropy and perfusion index were identified as independent predictors of metastasis.

## Abstract

To evaluate super-resolution ultrasound (SRUS) for characterizing microvascular morphology and hemodynamics in metastatic versus reactive cervical lymph nodes (LNs), with the aim of improving metastatic detection and reducing unnecessary biopsies. In this prospective study, 166 patients with histopathologically confirmed cervical LNs (77 metastatic, 89 reactive) underwent conventional ultrasound and contrast-enhanced SRUS (CE-SRUS) using a commercial US system and SonoVue® microbubbles. Quantitative SRUS parameters vascular density (VD), fractal dimension (FD), flow-weighted vascular density (FWVD), perfusion index (PI), velocity entropy (Vel Entropy), minimum velocity (Vmin) were extracted from whole-LN ROIs. Diagnostic performance was assessed via receiver operating characteristic (ROC) analysis and multivariate logistic regression. Metastatic LNs showed significantly higher VD (0.482 ± 0.073 vs. 0.405 ± 0.168, p < 0.001), FD (1.678 ± 0.070 vs. 1.626 ± 0.098, p < 0.001), FWVD (1.784 ± 0.592 vs. 1.495 ± 0.813, p = 0.013), PI (12.617 ± 2.563 vs. 10.369 ± 5.006, p < 0.001), and Vel Entropy (0.922 ± 0.092 vs. 0.796 ± 0.199, p < 0.001), but lower Vmin (2.572 ± 2.200 mm/s vs. 2.645 ± 2.800 mm/s, p = 0.017) compared to reactive LNs. Univariate ROC top performers included Dir Entropy (AUC = 0.723) and VD (AUC = 0.689). Multivariate analysis identified VD (OR = 1.046, p = 0.001), Vmin (OR = 0.525, p = 0.003), Velocity Variance (Vel Var) (OR = 1.973, p = 0.016), Vel Entropy (OR = 4.674, p = 0.042), and PI (OR = 2.481, p = 0.018) as independent predictors. The combined model achieved superior diagnostic performance (AUC = 0.813, 95% CI: 0.748–0.879; sensitivity = 76.6%, specificity = 79.8%; p < 0.001). SRUS enables non-invasive, high-resolution quantification of microvascular alterations in metastatic LNs. A multivariate model demonstrates excellent discriminative power, demonstrating significant potential to improve preoperative assessment and biopsy guidance in head and neck cancer.