Assessing Measurement Repeatability of a Novel Anisotropic Phantom for Advanced Diffusion MRI Models

Diffusion MRI is widely used to characterize tissue microstructure, but standardization remains challenging, particularly for advanced models or regions with crossing fibers. Phantoms provide controlled environments to assess measurement repeatability independent of biological variability. This study evaluated the repeatability of higher-order diffusion tensor metrics using a novel anisotropic diffusion phantom designed to mimic white matter tract geometry.

The phantom, containing linear, crossing (30°, 45°, 90°), and bifurcating synthetic fiber bundles, was scanned seven times using a GE Healthcare 3.0 T MRI system. Four acquisition protocols were evaluated: 30-direction DTI (b = 1000s/mm²), 60 and 90-direction High Angular Resolution Diffusion Imaging (HARDI; b = 1300s/mm²), and 30-direction Diffusion Kurtosis Imaging (DKI; b = 250, 500, 750, 1000, 1500, 2000, 2500, 3000 s/mm²). Repeatability was quantified using coefficient of variation (CoV) and intraclass correlation coefficient (ICC) for scalar diffusion metrics across six regions of interest. Fiber orientation distribution functions (fODFs) were analyzed to assess crossing fiber resolution accuracy.

DTI-derived metrics demonstrated excellent repeatability, with fractional anisotropy (FA) CoV < 10% and mean, axial, and radial diffusivities < 3%. DKI-derived metrics exhibited greater variability, though kurtosis FA remained stable (CoV ∼7%). Generalized FA showed improved reliability with increased angular resolution (ICC = 0.8445 for 90-direction HARDI). fODFs accurately resolved crossing fibers at 90° (RMSE = 3.49°) and 45° (RMSE = 8.92°) but failed at 30° separation.

The phantom provides reliable repeatability for standard DTI metrics and demonstrates utility for quality assurance of advanced diffusion models with high angular resolution protocols.