Characterizing uncertainty in tractography: parametric and nonparametric methods

White matter tractography is a noninvasive method for estimating the white matter connectivity pathways using diffusion tensor imaging. Experimental noise may induce errors in the measured fiber directions and affect the accuracy and precision of the estimated trajectories. Both model-based (parametric) and model-free (non-parametric) probabilistic tractography methods have been proposed to account for the uncertainty in the fiber direction estimation. The non-parametric methods give an unbiased estimation of the data variability but require long imaging times and are computationally intensive. In this study we evaluate the behavior of a parametric algorithm the random vector perturbation (RAVE) in comparison with the non-parametric bootstrap tractography (BOOT-TRAC) method. The RAVE algorithm appears to generate fiber distributions similar to the BOOT-TRAC algorithm for trajectories situated in homogeneous white matter regions and might be a feasible substitute for BOOT-TRAC in cases when multiple measurements of the diffusion-weighted images are not obtainable