Biophysical compartment models for single-shell diffusion MRI in the human brain: a model fitting comparison

Area of Science:

• Neuroimaging
• Diffusion MRI (dMRI)
• Biomedical Engineering

Background:

• Clinically standard diffusion MRI (dMRI) often uses single-shell data (b=1000 s mm^-2).
• The diffusion tensor model is commonly used for single-shell data but has limitations.
• Compartment models show promise for multi-shell dMRI but their utility in single-shell data is unclear.

Purpose of the Study:

• To investigate the optimization of compartment model fitting for single-shell dMRI data.
• To compare the effectiveness of various compartment models against the standard tensor model.
• To maximize information extraction from limited single-shell dMRI datasets.

Main Methods:

• Fitted various compartment models (e.g., ball and stick, ball and zeppelin) to single-shell dMRI data.
• Employed Markov chain Monte Carlo (MCMC) and non-linear least squares fitting techniques.
• Validated findings across multiple subjects, including comparisons with multi-shell data.

Main Results:

• Markov chain Monte Carlo (MCMC) outperformed non-linear least squares for model fitting.
• The 2-fibre-orientation mono-exponential ball and stick (BSME2) model yielded stable, artifact-free results efficiently.
• Compartment models (BZ2, BSME2) better characterized complex white matter microstructures than the tensor model, avoiding FA confounding.

Conclusions:

• Compartment models, particularly BSME2 and BZ2, are superior for extracting detailed microstructural information from single-shell dMRI.
• MCMC is an effective fitting technique for optimizing compartment model performance on single-shell data.
• These models offer significant advantages over the traditional tensor model for analyzing clinically acquired dMRI data.