# Imaging-derived neuromuscular ultrasound phenotypes are associated with functional status in amyotrophic lateral sclerosis

## TL;DR

This study shows that neuromuscular ultrasound can identify two distinct groups of ALS patients based on muscle and nerve features, which correlate with their functional abilities.

## Contribution

The study introduces a novel method for stratifying ALS patients using NMUS features combined with clinical data.

## Key findings

- Two reproducible NMUS-based subgroups were identified: a Mild cluster and a Severe cluster.
- The Severe cluster showed reduced muscle thickness and higher echogenicity, along with lower ALSFRS-R scores.
- Cluster membership correlated with muscle strength and electrophysiological measures, confirming functional validity.

## Abstract

Amyotrophic lateral sclerosis (ALS) presents with marked clinical heterogeneity, complicating diagnosis and management. Neuromuscular ultrasound (NMUS) provides a non-invasive means to visualize peripheral nerve and muscle integrity, but its potential to delineate ALS subtypes has not been systematically explored.

To identify clinically meaningful ALS subgroups through unsupervised clustering of NMUS features integrated with clinical and electrophysiological data.

A total of 454 ALS patients (August 2024–December 2025) underwent standardized NMUS assessment, including muscle thickness, echogenicity, and nerve cross-sectional area, alongside ALSFRS-R, manual muscle testing (MMT), and compound muscle action potentials (CMAPs). K-means clustering was applied to standardized NMUS variables, with cluster stability assessed using silhouette coefficients, sensitivity analyses (k = 2–5), and resampling-based adjusted Rand indices. Multivariable regression examined associations between cluster membership and ALSFRS-R.

Two reproducible NMUS-based subgroups were identified: a Mild cluster (n = 288, 63.4%) and a Severe cluster (n = 166, 36.6%). The Severe cluster showed reduced muscle thickness and higher echogenicity across multiple sites, together with lower ALSFRS-R scores (adjusted β = − 3.84, 95% CI − 5.41 to − 2.27, P < 0.001). Cluster membership correlated negatively with MMT and CMAP amplitudes, supporting functional and electrophysiologic validity. Stability metrics confirmed robustness of the two-cluster solution.

Integrating NMUS with clinical data enables objective, imaging-derived stratification of ALS patients into biologically and functionally distinct subgroups. This approach offers a pragmatic framework for phenotypic characterization and may inform personalized monitoring and trial design in ALS.

The online version contains supplementary material available at 10.1007/s00415-026-13705-4.