Population-based Bayesian regularization for microstructural diffusion MRI with NODDIDA.

Area of Science:

• Neuroimaging
• Biophysics

Background:

• Diffusion Magnetic Resonance Imaging (dMRI) probes brain microstructure.
• Neurite Orientation Dispersion and Density Imaging with Diffusivities Assessment (NODDIDA) is a simple microstructural model.
• NODDIDA parameter estimation from clinical dMRI is ill-posed, leading to non-unique solutions.

Purpose of the Study:

• Introduce a Bayesian estimation framework for NODDIDA parameters.
• Regularize estimation using a population-based prior derived from extensive dMRI data.
• Address the ill-posed nature of NODDIDA parameter estimation.

Main Methods:

• Reformulated NODDIDA estimation as a Bayesian maximum a posteriori problem.
• Developed a population-based prior from Human Connectome Project dMRI data (35 subjects).
• Validated accuracy and robustness using subsets of the MGH dataset and an independent clinical dataset.

Main Results:

• The population-based prior significantly improved parameter estimate accuracy and robustness.
• These improvements were observed for clinically feasible dMRI protocols.
• No evident bias was introduced by the population-based prior.

Conclusions:

• The proposed Bayesian population-based prior enables clinically feasible and robust NODDIDA parameter estimation.
• This approach enhances microstructural analysis without requiring changes to dMRI acquisition protocols.
• Improves the clinical utility of dMRI for brain microstructure assessment.