Robust Construction of Diffusion MRI Atlases with Correction for Inter-Subject Fiber Dispersion

Area of Science:

• Neuroimaging
• Diffusion MRI
• Computational Neuroscience

Background:

• Brain atlas construction typically involves image registration and fusion.
• Imperfect registration and inter-subject variability, especially fiber dispersion in diffusion MRI, cause artifacts and blurring in atlases.
• Existing methods struggle with within-voxel fiber misalignment, complicating accurate atlas generation.

Purpose of the Study:

• To develop an improved method for constructing diffusion MRI brain atlases.
• To address and mitigate artifacts arising from inter-subject fiber orientation dispersion.
• To enhance the accuracy and clarity of diffusion atlas generation.

Main Methods:

• Proposed a novel q-space (wavevector space) patch matching mechanism.
• Integrated this mechanism into a mean shift algorithm for mode seeking.
• The method identifies the most probable signal profile at each voxel, robust to outliers and dispersion.

Main Results:

• Experimental results demonstrate cleaner fiber orientation distribution functions.
• The proposed method significantly reduces artifacts caused by inter-subject fiber dispersion.
• Improved accuracy in representing white matter architecture in diffusion atlases.

Conclusions:

• The novel q-space patch matching approach enhances diffusion atlas construction.
• This method effectively handles inter-subject fiber dispersion, leading to more reliable brain atlases.
• The findings contribute to more precise neuroimaging analysis and understanding of brain connectivity.