Diffusion MRI: Assessment of the Impact of Acquisition and Preprocessing Methods Using the BrainVISA-Diffuse Toolbox

Area of Science:

• Neuroimaging
• Diffusion MRI analysis
• Medical physics

Background:

• Diffusion MRI is susceptible to geometric distortions from motion, eddy currents, and magnetic susceptibility variations.
• Existing correction methods vary in acquisition settings and data quality outcomes.
• The trade-off between scan time and preprocessing data quality remains unevaluated.

Purpose of the Study:

• To investigate the impact of six preprocessing pipelines on diffusion MRI analysis outcomes.
• To evaluate the ratio between scan time and preprocessed data quality.
• To develop a toolbox (Diffuse) guiding users to optimal preprocessing pipelines.

Main Methods:

• Comparison of six preprocessing pipelines using diffusion MR images from 20 HCP dataset subjects.
• Evaluation criteria included geometric accuracy, tensor model estimation, white matter indices, and pathway dispersion.
• A pipeline with reversed phase-encoding (RPE) acquisition served as the quality reference.

Main Results:

• Eddy-current correction enhances tensor fitting, particularly in the corpus callosum.
• Field maps alone for susceptibility correction require smoothing, artificially reducing tensor-fitting error.
• Acquiring a b0 volume with RPE at equivalent scan times yields superior geometry reconstruction and local tensor quality without smoothing.

Conclusions:

• Diffusion MRI metrics are highly dependent on acquisition settings and preprocessing methods.
• Acquiring b0 with RPE presents the optimal scan time/data quality compromise for diffusion MRI.
• The Diffuse toolbox aids in selecting appropriate preprocessing strategies based on input data.