NODDI-derived U-Fiber MRI Markers for Risk Stratification of Progression Independent of Relapse Activity in Relapsing-remitting Multiple Sclerosis.
Abstract
Rationale
AND
Objectives
Progression independent of relapse activity (PIRA) contributes to disability accumulation in relapsing-remitting multiple sclerosis (RRMS), but magnetic resonance imaging (MRI) markers of its microstructural substrate remain unclear. We evaluated whether U-fiber metrics derived from neurite orientation dispersion and density imaging and diffusion tensor imaging are associated with subsequent PIRA and may support MRI-based risk stratification.
Materials And Methods
This single-center longitudinal study included 138 patients with RRMS who underwent 3 T brain MRI and were followed for a median of 3.0 years. PIRA required an Expanded Disability Status Scale (EDSS) worsening confirmed over 6 months; events preceded by a relapse within 90 days or followed by a relapse before the 6-month confirmation assessment were excluded. Metrics were extracted from 16 atlas-based superficial U-fiber bundles. Candidate variables underwent univariable screening, variance inflation factor filtering, and exploratory multivariable logistic regression to construct a reduced model. Performance was assessed using five-fold cross-validation, calibration, Brier score, decision curve analysis, and continuous net reclassification improvement (NRI).
Results
PIRA developed in 35 patients (25.4%). Symbol Digit Modalities Test (SDMT), EDSS, right occipitotemporal neurite density index (NDI), right parietotemporal NDI, and right occipitotemporal FA were retained. The reduced model achieved an AUC of 0.793 (95% confidence interval [CI], 0.693-0.893) and a Brier score of 0.143. Adding U-fiber metrics improved reclassification beyond SDMT and EDSS (continuous NRI, 0.518; 95% CI, 0.138-0.882; p = 0.006).
Conclusion
Advanced diffusion MRI-derived U-fiber metrics may complement clinical measures for exploratory PIRA risk stratification in RRMS and warrant validation in independent multicenter cohorts.