Registration of high angular resolution diffusion MRI images using 4th order tensors

Area of Science:

• Medical Imaging
• Neuroimaging
• Diffusion MRI Analysis

Background:

• Current diffusion MRI registration methods using scalar or 2nd-order tensors struggle with complex local tissue structures like fiber crossings.
• Inaccurate registration limits the analysis of diffusion MRI datasets with intricate fiber architectures.

Purpose of the Study:

• To develop and validate a novel non-rigid registration method for diffusion-weighted MRI (DW-MRI) datasets using 4th-order tensors.
• To improve the accuracy of DW-MRI registration, particularly in the presence of complex fiber crossings.

Main Methods:

• Utilized 4th-order tensors to represent DW-MRI data, capturing more complex local tissue information.
• Employed Hellinger distance between normalized 4th-order tensors (as distributions) for registration.
• Introduced a novel 4th-order tensor re-transformation operator, outperforming existing DTI registration operators.

Main Results:

• The proposed 4th-order tensor registration method demonstrated superior performance compared to scalar and 2nd-order tensor-based methods.
• Validated the technique on simulated diffusion MR data and real high-angular-resolution diffusion imaging (HARDI) datasets.
• The novel re-transformation operator significantly improved registration accuracy.

Conclusions:

• 4th-order tensor representation and Hellinger distance provide a more robust framework for DW-MRI registration.
• The developed method accurately registers datasets with complex fiber architectures, overcoming limitations of previous techniques.
• This advancement holds potential for more precise analysis in neuroimaging research.