Reproducibility of MRI-based white matter tract estimation using multi-fiber probabilistic tractography: effect of user-defined parameters and regions

Abstract

Objective There is a pressing need to assess user-dependent reproducibility of multi-fibre probabilistic tractography in order to encourage clinical implementation of these advanced and relevant approaches. The goal of this study was to evaluate both intrinsic and inter-user reproducibility of corticospinal tract estimation. Materials and methods Six clinical datasets including motor functional and diffusion MRI were used. Three users performed an independent tractography analysis following identical instructions. Dice indices were calculated to quantify the reproducibility of seed region, fMRI-based end region, and streamline maps. Results The inter-user reproducibility ranged 41–93%, 29–94%, and 50–92%, for seed regions, end regions, and streamline maps, respectively. Differences in streamline maps correlated with differences in seed and end regions. Good inter-user agreement in seed and end regions, yielded inter-user reproducibility close to the intrinsic reproducibility (92–97%) and in most cases higher than 80%. Discussion Uncertainties related to user-dependent decisions and the probabilistic nature of the analysis should be considered when interpreting probabilistic tractography data. The standardization of the methods used to define seed and end regions is a necessary step to improve the accuracy and robustness of multi-fiber probabilistic tractography in a clinical setting. Clinical users should choose a feasible compromise between reproducibility and analysis duration.

Regional cortical thinning, demyelination and iron loss in cerebral small vessel disease

“…The connectivity map obtained from the probability tractography above was normalized by dividing the total number of streamlined samples and then thresholded at 2 × 10 –4 to achieve the best balance between removing potentially noisy fibre tracts and minimizing any false negative pathways. 41 The thresholded tracts were visually inspected to ensure their accuracy and specificity. Then, mean MD values for the tracts connecting WMH and the cortex were extracted, and were used as a proxy of white matter microstructural integrity.…”