Diffusion Capillary Phantom vs. Human Data: Outcomes for Reconstruction Methods Depend on Evaluation Medium.

Area of Science:

• Neuroimaging
• Biomedical Engineering

Background:

• Diffusion MRI enables non-invasive estimation of brain structural connectivity.
• Diffusion phantoms are commonly used to benchmark tractography algorithms, with results often extrapolated to in vivo human studies.

Purpose of the Study:

• To investigate if quality metrics from a standard diffusion phantom can accurately predict tractography performance in human white matter.
• To compare the efficacy of different tractography reconstruction models using both phantom and in vivo data.

Main Methods:

• Compared fiber length and crossing estimates from diffusion tensor imaging (simple tensor model), ball-and-sticks (complicated model), and model-free methods (diffusion spectrum imaging, generalized q-sampling imaging).
• Evaluated methods using a capillary phantom and in vivo human brain data (N=14).

Main Results:

• Tractography evaluation outcomes differed significantly between phantom and human data.
• The diffusion phantom favored complex models for resolving crossing fibers, whereas human data showed model-free methods performed better.
• This discrepancy was consistent across fiber length estimation and fiber crossing resolution in white matter pathways.

Conclusions:

• Current capillary diffusion phantoms may not accurately reflect in vivo human white matter tractography performance.
• Phantom-based evaluations may favor complex reconstruction models, contrary to findings in human data.
• Development of more realistic phantoms or simulations is needed for reliable tractography method evaluation.