Longitudinal changes in radiomic features of skeletal muscle on CT images with changes in muscle strength and physical performance in older adults
Megan Hetherington-Rauth, Tyler A Mansfield, Ashley A Weaver, Leon Lenchik, Peggy M CawthonAbstract
Background
Skeletal muscle health is a key determinant of aging, independence, and disease outcomes. Traditional computed tomography (CT)-derived measures of muscle cross-sectional area (CSA) and density reflect muscle quantity and quality but may not capture structural heterogeneity relevant to functional decline. Radiomics, a CT image analysis approach, enables extraction of high-dimensional texture features, offering additional insight into muscle quality. While muscle radiomics have been associated with strength and mobility in older adults, longitudinal applications remain understudied. We examined whether 5-year changes in thigh muscle radiomic features were associated with changes in muscle strength and performance in the Health, Aging, and Body Composition Study.
Methods
Thigh CT scans from 1321 older adults were analyzed using automated muscle segmentation to quantify CSA, density, and extract radiomic features. Strength measures included grip strength and maximal isokinetic knee extension; performance measures included 20-m walking speed, 5-time chair stands, and a modified performance score. Factor analysis reduced radiomic features to seven latent factors. Nested linear mixed-effects models tested associations of muscle CSA and density with outcomes, with and without radiomic factors.
Results
Factors 2 and 5 were consistently associated with 5-year changes across muscle outcomes. Factor 2 (high pixel gray-level intensity and uniformity) was positively associated, whereas Factor 5 (pixel clustering asymmetry and heterogeneity) was inversely associated. Adding radiomic factors modestly improved model fit (p < 0.05), explaining 1–2% additional variation in performance.
Conclusion
Longitudinal changes in thigh muscle radiomics capture subtle compositional characteristics beyond muscle CSA or density, enhancing understanding of functional decline with aging.