Differences in Gaussian diffusion tensor imaging and non-Gaussian diffusion kurtosis imaging model-based estimates of...

Area of Science:

• Neuroimaging
• Diffusion-weighted imaging (DWI)
• White matter microstructure

Background:

• Diffusion tensor imaging (DTI) parameters like mean diffusivity (MD) and fractional anisotropy (FA) are widely used to assess white matter microstructure.
• Diffusion kurtosis imaging (DKI) offers complementary metrics (e.g., mean kurtosis (MK)) that may provide greater sensitivity to microstructural alterations.
• Both DTI and DKI models rely on fitting algorithms, which can influence parameter estimates and their interpretation.

Purpose of the Study:

• To investigate model- and algorithm-dependent differences in DTI- and DKI-derived white matter microstructural parameters.
• To compare estimates of mean diffusivity (MD), fractional anisotropy (FA), radial diffusivity (RD), axial diffusivity (AD), and anisotropy mode (MO) between DTI and DKI models.
• To assess the impact of different fitting routines on these parameter estimates.

Main Methods:

• Analysis of diffusion-weighted imaging data from 67 healthy subjects across two datasets (Human Connectome Project and clinical scanner).
• DTI model fitted using b=0 and b=1000 s/mm² data; DKI model fitted using higher b-values (b=1000, 3000/2500 s/mm²).
• Nonlinear and weighted linear least squares algorithms were employed for model fitting; tract-based spatial statistics used for voxelwise comparisons.

Main Results:

• DKI-derived MD, RD, and AD estimates were significantly higher than DTI-derived estimates, while FA and MO were lower.
• Voxelwise differences were widespread across the white matter skeleton, with fractional differences varying by invariant and kurtosis values.
• Differences were consistent across datasets and algorithms, but the relationships between trends, diffusion invariants, and kurtosis estimates were algorithm-dependent.

Conclusions:

• Model-dependent differences in DTI/DKI parameter estimation are substantial and can exceed disease-related changes.
• DKI-derived diffusivity estimates should not be conflated with conventional DTI metrics to prevent interpretation and multicenter comparison issues.
• Separate estimation and analysis of DTI-derived indexes and DKI-derived invariants are recommended for standardized and accurate diffusion-weighted imaging research.