Constrained diffusion kurtosis imaging using ternary quartics & MLE.

Area of Science:

• Biomedical Imaging
• Diffusion MRI
• Quantitative MRI

Background:

• Diffusion Tensor Imaging (DTI) is limited in characterizing complex tissue microstructures.
• Diffusion Kurtosis Imaging (DKI) offers advanced tissue characterization by quantifying non-Gaussian diffusion.
• Existing DKI methods often overlook physical constraints and signal noise models.

Purpose of the Study:

• To develop a robust DKI estimation method that incorporates physical constraints and accounts for Rician noise.
• To improve the accuracy and efficiency of DKI parameter estimation.
• To provide a more reliable tool for quantitative MRI analysis.

Main Methods:

• Implicitly enforcing kurtosis tensor positivity using ternary quartic parameterization.
• Employing sequential quadratic programming with analytical gradients for efficient constrained optimization.
• Utilizing a maximum likelihood estimator based on Rician distribution to handle signal noise.

Main Results:

• MATLAB implementation demonstrated significantly improved estimation time and quality on synthetic data.
• Experiments on in vivo human brain data confirmed practical improvements in DKI results.
• The proposed method shows enhanced performance compared to existing DKI techniques.

Conclusions:

• The developed ternary quartic-based DKI estimation method offers substantial improvements over current approaches.
• This approach enhances the physical relevance and accuracy of DKI analysis.
• The method provides a more effective tool for quantitative assessment of tissue microstructure using diffusion MRI.