Assessing osteosarcopenia using automated multi-structure-segmentation and Hounsfield unit-based analysis of lumbar muscles and vertebral bodies on computed tomography

Sarcopenia and osteoporosis are prevalent age-related conditions that substantially increase the risk of falls, fractures, and functional decline. This study aimed to develop and clinically validate a fully automated computed tomography (CT)-based framework for simultaneous assessment of lumbar skeletal muscle and vertebral trabecular bone using a single axial image at the L3 level. We retrospectively analyzed 1,100 L3 CT images and implemented an automated segmentation pipeline with optimized pre- and post-processing, followed by anatomically standardized Hounsfield unit-based quantification of muscle and bone compartments. The framework demonstrated high segmentation accuracy across lumbar muscle groups and the vertebral body. CT-derived skeletal muscle indices showed moderate to strong correlations with established skeletal muscle indices and remained independently associated with functional impairment after adjustment for relevant clinical covariates. Vertebral trabecular attenuation values were significantly correlated with bone mineral density-derived T-scores. Sensitivity analyses using alternative index definitions yielded consistent associations with functional impairment and osteopenia in cross-sectional analyses. This integrated CT-based approach enables reproducible, simultaneous evaluation of muscle and bone from routine imaging and demonstrates meaningful associations with established clinical reference standards, supporting its potential utility for opportunistic screening of osteosarcopenia.