Voxelwise regularisation of high angular resolution diffusion imaging data.

Area of Science:

• Medical Imaging
• Neuroscience
• Biophysics

Background:

• Diffusion MRI (dMRI) data analysis is challenged by noise.
• High angular resolution diffusion imaging (HARDI) requires robust noise suppression techniques.
• Existing methods may introduce spatial dependencies, invalidating statistical analyses.

Purpose of the Study:

• To develop a novel noise suppression algorithm for HARDI data.
• To improve the accuracy of apparent diffusion coefficient (ADC) profile estimation.
• To preserve the validity of region-of-interest statistical testing in dMRI.

Main Methods:

• A Bayesian framework was employed for direct regularization of ADC profiles.
• A novel Markov random field (MRF) model was applied within each voxel across gradient directions.
• Anisotropic smoothing was utilized, exploiting the spherical distribution of gradient directions.

Main Results:

• The proposed anisotropic MRF algorithm demonstrated superior noise suppression compared to isotropic MRF and maximum likelihood estimators.
• Simulated and experimental HARDI datasets showed significant improvements with the new method.
• The algorithm effectively smoothed dMRI data without introducing invalid spatial dependencies.

Conclusions:

• The developed anisotropic MRF approach offers a significant advancement in HARDI noise suppression.
• This method enhances the reliability of diffusion MRI data analysis and statistical testing.
• The technique holds promise for more accurate characterization of brain microstructure.