Abstract
BACKGROUND: Insulin resistance (IR) plays a critical role in the musculoskeletal metabolic disorders associated with type 2 diabetes mellitus (T2DM). AIM: To develop multiparametric magnetic resonance imaging (MRI)-derived biomarkers and diagnostic models for non-invasive identification and stratification of IR. METHODS: Parameters of paravertebral muscles and vertebra were evaluated using quantitative chemical shift-encoded MRI and diffusion tensor imaging protocols. Tripartite cohort analyses were conducted through Kruskal-Wallis H tests with post hoc Dunn-Bonferroni correction for MRI-derived metrics. Diagnostic performance for T2DM-IR was assessed after selecting the most significant features through Z-score standardization and multinomial logistic regression models. RESULTS: This study evaluated 97 subjects (control: 39 subjects, T2DM-IR: 18 subjects, T2DM patients without IR: 40 subjects) using multiparametric MRI protocols. Significant intergroup differences were observed in the cross-sectional area (P = 0.047) and apparent diffusion coefficient (P = 0.027) of the psoas, and the cross-sectional area (P = 0.042) of the erector. More intramyocellular lipid (IMCL) in the psoas (P = 0.001) and erector (P = 0.004) were found in the T2DM-IR group. Multinomial receiver operating characteristic curve analysis demonstrated that IMCL of the erector performed better (area under the curve = 0.838, sensitivity: 0.800, specificity: 0.938) in the diagnosis of T2DM-IR. CONCLUSION: IMCL in erector emerges as a highly discriminative metric for T2DM-IR diagnosis. Multiparametric MRI enables non-invasive quantification of early musculoskeletal metabolic injury, providing reliable biomarkers for IR identification and stratification.