A random effects modelling approach to the crossing-fibre problem in tractography

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biostatistics

Background:

• Diffusion-weighted magnetic resonance imaging (dMRI) is crucial for mapping brain white matter architecture.
• The presence of multiple fibre orientations within a single voxel (crossing fibres) poses a significant challenge for traditional tractography methods.
• Existing methods for analysing dMRI data, particularly in regions with complex fibre architecture, require robust statistical modelling.

Purpose of the Study:

• To investigate a Bayesian random effects modelling approach for analysing multiple-direction diffusion-weighted MR data.
• To specifically address the challenge of crossing fibres in neuroimaging.
• To evaluate the performance of different random effects models, including spatial and exchangeable terms, in capturing fibre angular distributions.

Main Methods:

• Employed a Bayesian random effects modelling framework.
• Investigated spatial (Markov random field) and exchangeable models, including the Besag-York-Mollie model.
• Utilized Markov chain Monte Carlo simulation for model analysis.
• Applied the models to diffusion data from two distinct brain regions (corpus callosum, corona radiata, superior longitudinal fasciculus; and pons) known for crossing fibres.

Main Results:

• Achieved convincing fibre angular distributions even with limited dMRI data (low b-value, 20 directions, two acquisitions per direction).
• Demonstrated the capability of the Bayesian random effects models to accurately represent complex fibre architecture in crossing-fibre voxels.
• The investigated models successfully captured fibre orientation information in challenging anatomical regions.

Conclusions:

• Bayesian random effects modelling presents a promising and effective alternative to current methods in MR tractography.
• This approach offers improved analysis of dMRI data, particularly in the presence of crossing fibres.
• The findings support the utility of this modelling strategy for detailed brain white matter mapping.