VARIATIONAL DENOISING OF DIFFUSION WEIGHTED MRI.

Area of Science:

• Medical Imaging
• Neuroscience
• Computational Biology

Background:

• High angular resolution diffusion imaging (HARDI) is crucial for mapping white matter tracts in the brain.
• Existing HARDI data often suffers from noise and artifacts, limiting the accuracy of tractography.
• Accurate reconstruction of the diffusion signal is essential for reliable fiber direction computation.

Purpose of the Study:

• To develop and validate a novel variational formulation for restoring HARDI data.
• To improve the accuracy of fiber direction computation from HARDI measurements.
• To demonstrate the effectiveness of the proposed method on both synthetic and real biological data.

Main Methods:

• A variational formulation incorporating spherical and spatial smoothing of HARDI signal measurements.
• Total variation (TV) norm regularization applied across the 3D image lattice.
• Finite element method (FEM) used for smoothing data on the sphere with first and second-order constraints.

Main Results:

• The proposed HARDI data restoration scheme effectively smooths signal measurements.
• Demonstrated improvement in fiber direction computation accuracy on synthetic datasets.
• Successful application to real HARDI data from rat brain and spinal cord, showing enhanced tract visualization.

Conclusions:

• The novel variational formulation provides effective HARDI data restoration.
• The method significantly improves fiber direction computation, crucial for neuroimaging analysis.
• This technique holds promise for advancing diffusion MRI-based studies of brain and spinal cord microstructure.