Effects of lipid-induced magnetic microstructure on fat fraction quantification in muscular dystrophies

PURPOSE: To study the impact of mesoscopic magnetic susceptibility heterogeneity on chemical shift-encoded (CSE) proton density fat-fraction (PDFF) quantification in muscular dystrophies, a subgroup of neuromuscular disorders. THEORY AND METHODS: In MRI, extramyocellular lipid deposits induce orientation-dependent Larmor frequency variations due to microstructural anisotropy, resulting in spatially varying frequency shifts between fat and water and increased transverse relaxation rates. A newly developed PDFF quantification method accounting for resonance shifts and dual R(2)* rates was applied on standard 6-point CSE acquisitions of Duchenne (n = 15), Becker (n = 31), and facioscapulohumeral (n = 30) muscular dystrophy patients, and control subjects (n = 40). The impact of frequency shifts, decay functions, and lipid models on PDFF estimation was systematically assessed. RESULTS: Accounting for resonance shifts resulted in large PDFF quantification differences compared to a reference method (−3.8% [−14.8%, 7.2%]), significantly improved fitting quality (Bayesian Information Criterion (BIC) difference ≥ 10), and reduced fat/water separation artifacts, confirming predictions by numerical simulations. Bias and variability due to the lipid model were reduced to less than 1%. Fitting quality in high R(2)* regions was further improved using a dual relaxation model with linear/quadratic decay (BIC difference ≥ 2). Sensitivity to change was improved on the tested cohorts (SRM increased by 0.18). DTI-estimated angular dependencies reflected theoretical and numerical predictions for elongated axially symmetric lipid deposits. CONCLUSION: The proposed approach improvements could enhance the PDFF quantification reliability in neuromuscular disorders studies and support more accurate monitoring of myosteatosis.