We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Physics > Medical Physics

Title: Resolving orientation-specific diffusion-relaxation features via Monte-Carlo density-peak clustering in heterogeneous brain tissue

Abstract: Characterizing the properties and orientations of sub-voxel fiber populations, although essential to study white-matter architecture, microstructure and connectivity, remains one of the main challenges faced by the MRI microstructure community. While some progress has been made in overcoming this challenge using models, signal representations and tractography algorithms, these approaches are ultimately limited by their key assumptions or by the lack of specificity of the diffusion signal alone. In order to alleviate these limitations, we combine diffusion-relaxation MR acquisitions drawing from tensor-valued diffusion encoding, Monte-Carlo signal inversions that extract non-parametric intra-voxel distributions of diffusion tensors and relaxation rates, and density-based clustering techniques into a novel method called "Monte-Carlo density-peak clustering" (MC-DPC). This procedure first delineates clusters in the diffusion-orientation subspace of the distributions outputted by Monte-Carlo signal inversions and then draws from the statistical aspect of these inversion algorithms to compute statistics over orientation-resolved means of diffusivities and relaxation rates. Evaluating MC-DPC on tensor-valued diffusion- and $T_2$-weighted datasets in silico and in vivo, we demonstrate its ability to simultaneously capture sub-voxel fiber orientations and fiber-specific diffusion-relaxation properties that are consistent with the known anatomy and existing literature. Straightforwardly translatable to other diffusion-relaxation correlation experiments probing $T_1$ and $T_2^*$, MC-DPC shows potential in tracking bundle-specific patient-control group differences and longitudinal microstructural changes, developing new tools for microstructure-informed tractography, and mapping tract-specific myelination states.
Comments: 18 pages, 10 figures
Subjects: Medical Physics (physics.med-ph)
Cite as: arXiv:2004.08626 [physics.med-ph]
  (or arXiv:2004.08626v2 [physics.med-ph] for this version)

Submission history

From: Alexis Reymbaut [view email]
[v1] Sat, 18 Apr 2020 14:05:20 GMT (5478kb,D)
[v2] Tue, 21 Apr 2020 07:56:02 GMT (5478kb,D)
[v3] Mon, 11 May 2020 15:48:54 GMT (5303kb,D)

Link back to: arXiv, form interface, contact.