Improved intravoxel incoherent motion analysis of diffusion weighted imaging by data driven Bayesian modeling.

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Medical Physics
• Radiology

Background:

• Intravoxel Incoherent Motion (IVIM) modeling uses multi-b-value diffusion-weighted MRI to assess tissue microenvironment.
• Pseudo-diffusion parameters from IVIM are valuable for pathology assessment but are sensitive to noise.
• Least-squares fitting methods for pixel-wise IVIM analysis yield noisy estimates, limiting clinical utility.

Purpose of the Study:

• To introduce a Bayesian approach with a shrinkage prior model for IVIM analysis.
• To demonstrate the reduction of estimation uncertainty in IVIM parameters, particularly pseudo-diffusion measures.
• To improve the visualization of spatial features and heterogeneity in IVIM parameter maps.

Main Methods:

• Development and application of a Bayesian shrinkage prior model for IVIM analysis.
• Comparison of Bayesian fitting results with traditional least-squares methods.
• Evaluation of parameter estimation accuracy and noise reduction.

Main Results:

• The Bayesian shrinkage prior model significantly reduces estimation uncertainty compared to least-squares methods.
• Spatial features and heterogeneity in IVIM parameter maps become more apparent with the Bayesian approach.
• Bayesian approach eliminates user-defined parameters for heterogeneity assessment, relying solely on data.

Conclusions:

• The Bayesian shrinkage prior approach offers superior performance for pixel-wise IVIM modeling.
• This method enhances the assessment of spatial characteristics and tissue heterogeneity in diffusion MRI.
• The proposed Bayesian method is recommended for robust IVIM analysis in clinical and research settings.