Joint reconstruction of white-matter pathways from longitudinal diffusion MRI data with anatomical priors

- A method for reconstructing white-matter pathways in longitudinal diffusion MRI data
- We reconstruct pathways using a subject's data from all time points jointly
- We represent a path in the space of a robust within-subject template
- We compute the posterior probability of the path given each time point's image data
- More reliable and more sensitive than analyzing each time point independently

We consider the problem of reconstructing white-matter pathways in a longitudinal study, where diffusion-weighted and T1-weighted MR images have been acquired at multiple time points for the same subject. We propose a method for joint reconstruction of a subject's pathways at all time points given the subject's entire set of longitudinal data. We apply a method for unbiased within-subject registration to generate a within-subject template from the T1-weighted images of the subject at all time points. We follow a global probabilistic tractography approach, where the unknown pathway is represented in the space of this within-subject template and propagated to the native space of the diffusion-weighted images at all time points to compute its posterior probability given the images. This ensures spatial correspondence of the reconstructed pathway among time points, which in turn allows longitudinal changes in diffusion measures to be estimated consistently along the pathway. We evaluate the reliability of the proposed method on data from healthy controls scanned twice within a month, where no changes in white-matter microstructure are expected between scans. We evaluate the sensitivity of the method on data from Huntington's disease patients scanned repeatedly over the course of several months, where changes are expected between scans. We show that reconstructing white-matter pathways jointly using the data from all time points leads to improved reliability and sensitivity, when compared to reconstructing the pathways at each time point independently.