# Validity of DEXA-Derived Thigh Muscle Quantification Against AI-Assisted CT: Inter-Limb Asymmetry Provides Superior Agreement over Absolute Values

## TL;DR

This study compares DEXA and AI-assisted CT for measuring thigh muscle and finds that inter-limb asymmetry improves DEXA accuracy for lean mass but not for fat.

## Contribution

The study introduces inter-limb asymmetry as a novel approach to improve DEXA's agreement with CT for lean mass quantification.

## Key findings

- DEXA lean mass correlated with CT pure muscle volume, and inter-limb asymmetry improved this correlation.
- Inter-limb asymmetry showed narrower agreement limits compared to absolute DEXA measurements.
- DEXA fat mass asymmetry did not align with CT intramuscular adipose tissue asymmetry.

## Abstract

Background/Objectives: This study evaluated the validity of DEXA-derived muscle quantification by assessing its agreement with AI-assisted CT measurements of muscle volume and intramuscular adipose tissue. It also examined whether inter-limb asymmetry improves DEXA–CT agreement beyond absolute DEXA values. The influence of lower-limb rotation on DEXA measurements was assessed, and the study aimed to clarify how DEXA should be obtained and interpreted to more accurately reflect true muscle status. Methods: Fifty-two patients who completed CT and DEXA within 14 days were included. CT was used to obtain pure muscle volume and intramuscular adipose tissue (IMAT) using a standardized AI segmentation protocol, and corresponding DEXA thigh segmentation provided lean mass and fat percentage. Position-specific correlation analysis, regression, and Bland–Altman agreement testing were performed for 104 limbs. The same analyses were applied to inter-limb differences to isolate within-person asymmetry and reduce between-person variance. Results: DEXA lean mass correlated with CT pure muscle volume (r = 0.776, p < 0.001), and inter-limb asymmetry further improved alignment with CT (r = 0.857, p < 0.001). However, DEXA fat mass asymmetry demonstrated no association with CT IMAT asymmetry (r = −0.004, p = 0.979). When results were stratified by the recorded rotational groups, the highest correlation was observed in the neutral position (r = 0.900, p < 0.001). Bland–Altman analyses showed wide limits of agreement for all absolute measurements, whereas inter-limb asymmetry demonstrated markedly narrower limits of agreement, indicating superior numerical consistency. Conclusions: Absolute DEXA estimates showed limited agreement with CT and varied with limb position. Inter-limb asymmetry improved lean mass assessment, whereas fat mass and percentage did not correspond to CT-based IMAT. DEXA may therefore be used as a complementary tool for evaluating regional muscle quantity, but not for assessing muscle quality.