Tensor kernels for simultaneous fiber model estimation and tractography.

Area of Science:

• Diffusion MRI
• Neuroimaging
• Computational Neuroscience

Background:

• Current diffusion MRI tractography methods often estimate fiber orientation independently at each voxel.
• This lack of continuity results in reduced confidence in signal measurements and estimated fiber orientations, particularly in complex white matter regions.

Purpose of the Study:

• To develop a novel framework for joint orientation distribution function (ODF) estimation and tractography.
• To improve the accuracy and robustness of fiber tracking in the presence of crossing and branching fibers.

Main Methods:

• Formulating fiber tracking as a recursive estimation process, where each step is guided by the previous estimate.
• Employing second- and higher-order tensor-based kernels for ODF estimation.
• Utilizing a weighted mixture of tensor kernels to represent complex fiber architectures.
• Incorporating a smoothness term to penalize deviations from previous estimates along the fiber path.

Main Results:

• Demonstrated improved angular resolution at fiber crossings in synthetic experiments using two- and three-component mixture models.
• Successfully traced the corpus callosum and corticospinal tract in vivo, confirming the method's ability to navigate regions with crossing and branching fibers.

Conclusions:

• The proposed recursive estimation framework with tensor kernels enhances ODF estimation and tractography accuracy.
• This approach effectively handles complex white matter structures, offering more reliable brain connectivity mapping.