Tract-Based Probability Densities of Diffusivity Measures in DT-MRI

Abstract

Abstract. We evaluate probability density functions of diffusivity measures in DTI fiber tracts as biomarkers. For this, we estimate univariate and bivariate densities, such as joint probability densities of the tract arc length and FA, MD, RD, and AD, in the transcallosal fibers in the brain. We demonstrate the utility of estimated densities in hypothesis testing of differences between a group of patients with VCI and a control group. We also use the estimated densities in classifying individual subjects in these two groups. Results show that these estimates and derived quantities, such as entropy, can detect group differences with high statistical power as well as help obtain low classification errors.

Uncertainty visualization in HARDI based on ensembles of ODFs

“…A different approach studies DTI (and HARDI) data sets, where several measurements in multiple imaging directions are available. These multiple readings allow one to construct a distribution of possible values, and an estimation of noise related to the choice of b value [13, 46, 12, 17, 38], the registration [41, 23], the model fitting [1, 39, 30, 51, 24], the maximum finding, decomposition, or enhancement [25, 43, 22, 38, 42], or the fiber tracking algorithms [18, 10, 14, 36, 54, 26, 12]. …”