Optimization of a free water elimination two-compartment model for diffusion tensor imaging.

Area of Science:

• Neuroimaging
• Biophysics
• Medical Physics

Background:

• Diffusion tensor imaging (DTI) measures water diffusion to infer tissue microstructure.
• Single diffusion tensor models are inaccurate in voxels with mixed tissue and free water.
• Existing two-tensor models require spatial constraints for free water elimination.

Purpose of the Study:

• To implement a novel, spatially unconstrained two-tensor free water elimination model.
• To determine optimal diffusion acquisition parameters for this model.
• To assess the model's accuracy and bias compared to single-tensor models.

Main Methods:

• Monte Carlo simulations and human brain imaging studies were used.
• Investigated optimal acquisition parameters, including b-values and number of directions.
• Implemented a two-tensor model without spatial constraints.

Main Results:

• Optimal parameters: two shells (500×32 and 1500×32) at SNR 40 with 64 encodings.
• The two-tensor model significantly reduced bias in fractional anisotropy compared to single-tensor models.
• Bias reduction was over an order of magnitude greater than single-tensor model errors.

Conclusions:

• The novel two-tensor model provides accurate water diffusion estimates without spatial constraints.
• This method enhances characterization of tissues adjacent to CSF and those affected by edema.
• The strategy effectively removes CSF-related artifacts like blurring and ghosting.