Estimating Uncertainty in White Matter Tractography Using Wild Non-local Bootstrap.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biostatistics

Background:

• Diffusion MRI statistics, particularly for tractography, exhibit complex, non-linear, and non-Gaussian distributions.
• Existing methods often assume normality, which is frequently violated in the presence of multiple noise sources like physiological variation and scanner instability.
• The bootstrap offers a distribution-independent approach for estimating statistical variability, but traditional methods have limitations.

Purpose of the Study:

• To evaluate the efficacy of a novel bootstrap scheme, the wild non-local bootstrap (W-NLB), for estimating uncertainty in diffusion MRI tractography data.
• To assess W-NLB's performance compared to conventional bootstrap methods and Monte Carlo simulations.
• To validate W-NLB using both in silico and in vivo neuroimaging data.

Main Methods:

• The study introduces and applies the wild non-local bootstrap (W-NLB), a method that utilizes the self-similarity of local imaging information.
• W-NLB does not require a predetermined data model or repeated measurements, distinguishing it from residual, wild, and repetition bootstrap methods.
• In silico evaluations compared W-NLB with the conventional residual bootstrap using Monte Carlo simulations. In vivo data were used for further validation.

Main Results:

• In silico evaluations demonstrated that W-NLB generates distribution estimates more closely aligned with Monte Carlo simulations than the conventional residual bootstrap.
• Evaluations using in vivo data showed that W-NLB results are consistent with established knowledge of white matter connection architecture.
• W-NLB proved effective in estimating uncertainty without relying on restrictive data models or time-consuming multiple acquisitions.

Conclusions:

• The wild non-local bootstrap (W-NLB) is a robust and effective method for estimating uncertainty in diffusion MRI tractography statistics.
• W-NLB offers advantages over traditional bootstrap methods by not assuming data normality or requiring repeated acquisitions.
• This approach holds significant potential for improving the reliability and interpretability of diffusion MRI tractography analyses in neuroscience research.