Multimodel Diffusion MRI Signatures in Atypical Parkinsonian Disorders

Distinguishing atypical parkinsonian disorders (APS) from Parkinson's disease (PD) remains challenging due to overlapping clinical features, yet accurate differentiation is critical for prognosis and treatment. Here, we employed multimodel diffusion MRI (dMRI) analysis to characterize microstructural alterations across corticobasal syndrome (CBS), progressive supranuclear palsy-Richardson syndrome (PSP-RS), and PD, with the aim of identifying which dMRI model and regional metrics show the strongest group-level separation. We analyzed 25 CBS, 42 PSP-RS, and 21 PD participants compared to 35 age and sex-matched controls. Using a clinically feasible 3-shell high angular resolution diffusion imaging (HARDI) protocol, we applied 11 metrics from five complementary dMRI models-diffusion tensor imaging (DTI), free-water-eliminated model of DTI (FWE), neurite orientation dispersion and density imaging (NODDI), tissue-weighted NODDI, and fixel density (FD) in fixel-based analysis (FBA)-to comprehensively assess regional white and gray matter integrity. Group differentiation was quantified using covariate-adjusted Cohen's d effect sizes and spearman correlations were assessed between dMRI metrics and clinical scales. Distinct microstructural signatures were observed across disorders, and the sensitivity of the dMRI models differed. In group contrasts, DTI and NODDI-derived metrics consistently captured the strongest effects in midbrain and peduncular pathways for PSP-RS, whereas precentral and corticospinal alterations in CBS were most prominent using NODDI and FBA measures. Free-water-corrected metrics showed attenuated group differences. Across clinical-diffusion analyses, NODDI metrics exhibited the most robust associations with disease severity, while DTI and FWE measures detected more limited, regionally constrained effects. Together, these findings highlight complementary yet distinct sensitivities of tensor, free-water, multicompartment, and fixel-based models to APS-related neurodegeneration.