Tractography-based parcellation does not provide strong evidence of anatomical organisation within the thalamus

Neuroimage. 2019 Oct 1;199:418-426. doi: 10.1016/j.neuroimage.2019.06.019. Epub 2019 Jun 8.

Abstract

Connectivity-based parcellation of subcortical structures using diffusion tractography is now a common paradigm in neuroscience. These analyses often imply voxel-level specificity of connectivity, and the formation of compact, spatially coherent clusters is often taken as strong imaging-based evidence for anatomically distinct subnuclei in an individual. In this study, we demonstrate that internal structure in diffusion anisotropy is not necessary for a plausible parcellation to be obtained, by spatially permuting diffusion parameters within the thalami and repeating the parcellation. Moreover, we show that, in a winner-takes-all paradigm, most voxels receive the same label before and after this shuffling process-a finding that is stable across image acquisitions and tractography algorithms. We therefore suggest that such parcellations should be interpreted with caution.

Keywords: Connectivity; Diffusion; Magnetic resonance imaging; Parcellation; Tractography.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Diffusion Tensor Imaging / methods
  • Diffusion Tensor Imaging / standards*
  • Female
  • Humans
  • Male
  • Thalamus / anatomy & histology*
  • Thalamus / diagnostic imaging