Related Experiment Video
Updated: Aug 28, 2026
A Novel Application of Musculoskeletal Ultrasound Imaging
Published on: September 17, 2013
Artificial Intelligence-Based Muscle Ultrasound: A Novel Approach to Nutritional Evaluation Beyond Quantity in
Juan José López-Gómez1,2,3, Lucía Estévez-Asensio1,2,3, Elena Santos-Pascual4
1Servicio de Endocrinología y Nutrición, Hospital Clínico Universitario de Valladolid, 47003 Valladolid, Spain.
Background:
Neurological disease may lead to malnutrition through disease-related complications, underscoring the need for accurate muscle assessment. This study aims to evaluate an AI-based tool for quantifying and characterizing muscle ultrasound images, comparing its performance with the usual techniques of muscle mass and function.
Methods:
This was a prospective, open-label, longitudinal observational study of 117 adults with neurological disorders at high nutritional risk, designed to evaluate nutritional status and clinical evolution. The clinical assessment integrated anthropometry, bioelectrical impedanciometry, handgrip strength, dysphagia testing, and rectus femoris quadriceps ultrasound. Ultrasound images were evaluated through an AI-based platform to extract muscle quantity (rectus femoris muscle area (RFMA) and rectus femoris muscle thickness (RFMT) and quality biomarkers (percentage of low-echogenicity areas (Mi), interpreted as muscle; percentage of medium-echogenicity areas (FATi), interpreted as intramuscular fat). Patients were followed for two years to record mortality.
Results:
The sample included 117 adults with neurological disorders (52.1% women), with a mean age of 63.01 (16.14) years. A total of 77 patients (65.8%) had a condition with direct neuromuscular involvement. According to Global Leadership Initiative on Malnutrition (GLIM) criteria, 73 patients (62.4%) had malnutrition, while 30 patients (25.6%) had severe malnutrition. There were no differences in muscle mass parameters, but patients with neuromuscular involvement (NM) had lower values of percentage of Mi (NM: 42.44 (9.09%) vs. 47.36 (7.74)%; p < 0.01), and higher values of FATi (41.91 (5.71)% vs. 39.15 (4.89)%). The prevalence of mortality was 26 patients (22.2%). In the multivariate analysis, FATi (above median) (OR = 5.11 (IC95%: 1.26-20.67)) increased risk of death, adjusted by age, neuromuscular involvement, sex, and Mi.
Conclusions:
Patients with neuromuscular disorders showed a markedly lower proportion of Mi and a higher presence of FATi compared to those with non-neuromuscular conditions. Mortality was associated with greater FATi on AI-based ultrasound analysis. These findings suggest AI-enhanced imaging captures clinically relevant tissue alterations with potential prognostic value; however, given the observational data and heterogeneity of neurological conditions, these implications should be interpreted cautiously.
