Two-tensor tractography using a constrained filter.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Diffusion Tensor Imaging

Background:

• Current diffusion MRI tractography methods often estimate fiber orientation independently per voxel, lacking confidence assessment.
• This limitation hinders accurate reconstruction of complex white matter structures with crossings and branchings.

Purpose of the Study:

• To develop a novel technique for simultaneous estimation of multi-tensor fiber models and tractography.
• To improve the accuracy and robustness of fiber tracking, particularly in regions with complex fiber architectures.

Main Methods:

• Formulated fiber tracking as a recursive estimation process using a weighted mixture of Gaussian tensors.
• Employed an unscented Kalman filter to simultaneously fit the local fiber model and propagate tractography.
• Modified the Kalman filter to enforce model constraints, ensuring positive eigenvalues and convex weights.

Main Results:

• Demonstrated significant improvement in angular resolution at fiber crossings and branchings using synthetic data.
• Showcased consistent estimation of mixture weights even in the presence of noise and uncertainty.
• Validated in vivo tracing of complex fiber pathways with inherent path regularization.

Conclusions:

• The proposed recursive Kalman filter-based approach enables simultaneous multi-tensor fiber modeling and tractography.
• This method offers improved accuracy and robustness for reconstructing white matter pathways, especially in complex anatomical regions.
• The technique provides causal estimates of local structure, enhancing confidence in tractography results.