Determination of fiber orientation in MRI diffusion tensor imaging based on higher-order tensor decomposition

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Computational Anatomy

Background:

• High Angular Resolution Diffusion Imaging (HARDI) enables the resolution of multiple fiber directions within a single voxel.
• Generalized Diffusion Tensor Imaging (GDTI) can derive high-order tensors from HARDI data.

Purpose of the Study:

• To present a mathematical technique for accurately resolving multiple, randomly-oriented fiber tracts within tissue.
• To leverage the decomposition of high-order diffusion tensors for improved tractography.

Main Methods:

• A novel mathematical technique based on the decomposition of high-order diffusion tensors.
• Derivation of pseudo-eigenvalues and pseudo-eigenvectors from diffusion tensors using successive best least-square rank-1 tensor approximation.
• Application of derived pseudo-eigenvalues and pseudo-eigenvectors to identify major fiber directions within individual voxels.

Main Results:

• The presented mathematical technique allows for accurate resolution of multiple fiber directions.
• Numerical simulations demonstrate the effectiveness of the proposed method in identifying complex fiber architectures.
• Pseudo-eigenvalues and pseudo-eigenvectors successfully pinpoint major fiber orientations within voxels.

Conclusions:

• The proposed tensor decomposition method offers a robust approach for resolving complex fiber tracts in diffusion imaging.
• This technique enhances the accuracy of fiber tractography by precisely identifying multiple fiber directions within voxels.
• The method shows promise for advancing the analysis of neural pathways in various neurological conditions.