Work

Novel AI-derived muscle ultrasound biomarkers linking nutritional status and prognosis in cancer patients

Year: 2026

Type: article

Abstract: Introduction: Malnutrition and sarcopenia worsen prognosis in patients with cancer and require accessible diagnostic methods for assessing muscle quantity and quality. Artificial intelligence (AI)-assisted muscle ultrasound may improve the standardization and efficiency of body-composition assessment. This study evaluated the prognostic relevance of AI-assisted rectus femoris ultrasound biomarkers within a comprehensive nutritional assessment of patients with cancer. Methods: This prospective longitudinal observational cohort study included 389 patients with cancer, most of whom had tumors of digestive origin. Participants underwent longitudinal and transverse rectus femoris ultrasonography. Images were analyzed using PIIXMED™ MSK, which applies U-Net based anatomical segmentation followed by deterministic image-processing procedures. The evaluated variables included subcutaneous adipose tissue thickness (SAT), rectus femoris thickness (RFT) rectus femoris cross-sectional area (RFA), pennation angle, mean echo intensity, and the percentages of low- and medium-echogenicity pixels within the segmented region of interest, reported as the Muscle Index (Mi) and Fat Index (FATi), respectively. The muscle-area index was calculated as AreaMi = RFAxMi. Results: = 0.01) independently predicted an increased risk of mortality. Conclusion: In cancer patients, AI-assisted rectus femoris ultrasound variables effectively predict mortality. Greater muscle area and higher muscle index (Mi > 47%) indicate a favorable prognosis. Early ultrasound assessment detects structural alterations before overt deterioration, providing a crucial window for targeted nutritional interventions to preserve muscle integrity and prevent complications.

Source: Frontiers in Nutrition

DOI: 10.3389/fnut.2026.1961328