Non-rigid Registration of High Angular Resolution Diffusion Images Represented by Gaussian Mixture Fields.

Area of Science:

• Medical Imaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Non-rigid registration is crucial for analyzing complex medical imaging data.
• High angular resolution diffusion weighted imaging (HARDI) provides rich information about white matter microstructure.
• Accurate registration of HARDI data is challenging due to complex deformations.

Purpose of the Study:

• To develop a novel algorithm for non-rigidly registering two HARDI datasets.
• To represent HARDI data using Gaussian mixture fields (GMFs) for registration.
• To improve the accuracy and efficiency of HARDI data alignment.

Main Methods:

• Modeling non-rigid warp using a thin-plate spline.
• Formulating registration as L2 distance minimization between GMFs.
• Deriving closed-form expressions for objective function derivatives.
• Implementing a novel re-orientation scheme based on "Preservation of Principle Directions".

Main Results:

• The algorithm demonstrates effective registration of synthetic and real HARDI datasets.
• The closed-form derivatives simplify and potentially speed up the optimization process.
• The new re-orientation scheme offers a more straightforward approach to registration refinement.

Conclusions:

• The proposed algorithm provides a robust and efficient method for non-rigid HARDI registration.
• The mathematical contributions enhance the theoretical foundation of diffusion MRI registration.
• This work advances the analysis of white matter structure and integrity from HARDI data.