Tractography from HARDI using an intrinsic unscented Kalman filter

Area of Science:

• Medical Imaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Diffusion MRI enables in vivo imaging of white matter neural pathways.
• Fiber tractography reconstructs these neural pathways but faces challenges in accuracy and computational efficiency.
• Existing unscented Kalman filter (UKF) adaptations for tractography lack intrinsic properties of diffusion tensor spaces, potentially causing inaccuracies.

Purpose of the Study:

• To introduce a novel Intrinsic Unscented Kalman Filter (IUKF) for diffusion MRI.
• To enhance simultaneous multitensor estimation and fiber tractography accuracy.
• To maintain the computational efficiency of existing UKF methods.

Main Methods:

• Developed an IUKF operating intrinsically within the space of diffusion tensors (symmetric positive definite matrices).
• Applied the IUKF for simultaneous recursive estimation of multiple diffusion tensors.
• Utilized the IUKF for propagating directional information essential for fiber tractography.

Main Results:

• The proposed IUKF demonstrates improved accuracy in multitensor estimation compared to non-intrinsic UKF variants.
• IUKF effectively propagates directional information for more precise fiber tractography.
• Experiments on phantom data and human/rat in vivo scans validate the method's effectiveness and accuracy.

Conclusions:

• The IUKF provides a more accurate and robust approach to diffusion MRI analysis.
• This method preserves the computational advantages of UKF while addressing its limitations.
• IUKF shows significant potential for advancing neuroimaging and understanding white matter structure.