# Differentiating lymph node status in malignant melanoma: the role of apparent diffusion coefficient ratio in diffusion-weighted MRI – a prospective diagnostic study

## TL;DR

This study explores using MRI-based apparent diffusion coefficient ratios to detect melanoma lymph node metastasis, offering a non-invasive diagnostic improvement.

## Contribution

The study introduces a standardized ADC ratio cut-off in diffusion-weighted MRI for non-invasive detection of metastatic lymph nodes in melanoma.

## Key findings

- ADC ratios outperformed morphological criteria like short-axis diameter and Node-RADS in diagnosing metastatic lymph nodes.
- Optimal cut-offs for cADC and mADC showed high AUC values in both sentinel and metastatic lymph node groups.
- Machine learning models confirmed the diagnostic patterns observed in logistic regression analyses.

## Abstract

This study aimed to investigate a standardized apparent diffusion coefficient (ADC) ratio cut-off for identifying lymph node (LN) metastasis in malignant melanoma (MM), addressing the challenges of current non-invasive LN status diagnostics and dependences of absolute ADC values.

This prospective, single-center study evaluated consecutive patients using diffusion-weighted MRI (DWI-MRI). We included 52 patients with early-stage MM who underwent sentinel lymph node (SLN) extraction, and 12 patients with advanced-stage MM and newly confirmed or progressive metastatic lymph nodes (MLNs). ADC values for positive and negative SLNs as well as MLNs were measured. Ratios of the ADC of SLNs or MLNs were calculated relative to benign contralateral LNs (cADC) and adjacent muscle tissue (mADC). ROC analysis identified cut-offs for absolute ADC values and relative ADC ratios. Diagnostic performance gained by regression was validated via machine learning (ML) classifiers, to evaluate robustness.

A total of 64 patients (median age 68.5; IQR 61–77; 46.9% female) were included. Significant differences were observed in all ADC measurements using single-shot-echo-planar-imaging (EPI): median cADC was 0.76 (IQR 0.69–0.93) for metastatic SLNs versus 1.02 (IQR 0.91–1.08) for benign SLNs (p = 0.03), mADC was 0.48 (IQR 0.45–0.53) versus 0.61 (IQR 0.56–0.71, p = 0.001). Optimal cut-offs for cADC and mADC in the SLN group were 0.81 (AUC 0.75; 95%-CI 0.47-1) and 0.49 (AUC 0.9; 95%-CI 0.78-1). These closely matched the optimal cut-offs found in the MLN set: cADC 0.83 (AUC 0.98; 95%-CI 0.96-1) and mADC 0.49 (AUC 0.99; 95%-CI 0.98-1). In both sets, these ratios outperformed morphological criteria, such as short-axis diameter (AUC 0.61; 0.38–0.84 in SLN- and 0.86; 0.8–0.92 in MLN set) and Node-RADS (AUC 0.5; 0.27–0.72 and 0.89; 0.84–0.94). ML models reproduced the diagnostic patterns observed in logistic regression.

The introduction of a standardized ADC ratio cut-off in DW-MRI to differentiate benign from metastatic lymph nodes in MM offers a valuable, normalized, non-invasive addition to traditional diagnostics, though the findings require confirmation in larger, multicenter and multi-scanner cohorts.

The online version contains supplementary material available at 10.1186/s12880-026-02205-6.