Segmentation of thalamic nuclei

In this project, the nuclei of the human thalamus are dissociated by applying the hierarchical watershed transform to the tensorial morphological gradient (TMG) map computed from the diffusion tensor field. It has the advantage of using all of the information in the diffusion tensor for the segmentation.

In conclusion, our experiment confirmed that the proposed algorithm is able to delineate the main nuclei of the thalamus. As opposed to other DTI-based segmentation algorithms, the proposed method requires no manual seed placement and/or initial surface delineation: the only parameter is the number of regions into which the image should be segmented. This independance from user input could be a major advantage in the application of this technique to large scale studies of normal anatomy, development, or disease.


Figura 1: 3D segmentation result

Main challenges:

  • Due to the anatomical inter-subject variability, it is difficult to draw meaningful conclusions based on comparisons with histological atlases and there is no subject-specific gold standard. Qualitative and quantitative analyses should be performed in a larger group of subjects.
  • Markers selection for the watershed transform play a huge role in this application, since the resolution of DTI is low, compared to the size of the nuclei. A distinct markers selection method could not only lead to better results, but also could make the method more robust.

Related publications: