Optimization and Validation of the DESIGNER dMRI preprocessing pipeline in white matter aging

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Medical Physics

Background:

• Diffusion MRI (dMRI) preprocessing pipelines vary, impacting diffusion parameter accuracy.
• Optimizing these pipelines is crucial for reliable analysis of white matter integrity.

Approach:

• Evaluated the revisited Diffusion parameter EStImation with Gibbs and NoisE Removal (DESIGNER) pipeline (Dv2) on a large clinical dMRI dataset and ground truth phantoms.
• Compared Dv2 against original DESIGNER (Dv1), minimal preprocessing, and no preprocessing, assessing effects on age-related white matter changes.
• Assessed specific processing steps, including denoising and Gibbs ringing removal, and their impact on Diffusion Tensor Imaging (DTI) and Diffusiometry with Kurtosis Imaging (DKI) metrics.

Key Points:

• Preprocessing pipelines significantly affect white matter age correlations and statistical significance of DTI and DKI metrics.
• DESIGNER (Dv2) demonstrated superior performance with fewer outlier voxels and improved accuracy in DTI and DKI metrics.
• Updated noise and Gibbs removal methods in Dv2 were validated using dMRI phantoms.

Conclusions:

• The DESIGNER (Dv2) pipeline provides more accurate and robust DTI and DKI parameter maps.
• This improved accuracy is attributed to enhanced noise reduction and Gibbs ringing correction.
• DESIGNER (Dv2) offers a significant advancement over minimal or no preprocessing for clinical dMRI analysis.