We introduce a method for automatic segmentation of the thalamic nuclei. The method uses diffusion-weighted MR images to subdivide the thalamus in a predefined set of nuclei. The method relies on two basic assumptions. First, it assumes that white-matter fascicles’ orientation is homogeneous within each thalamic nucleus (Wiegell et al., 1999). Second, it assumes that nuclei are spatially homogeneous (Wiegell et al., 2003). Based on these assumptions thalamic nuclei are segmented by a weighted combination of local tensor information (Basser et al., 1994) and spatial coordinates. The method avoids manual intervention by computing in each subject the dissimilarity (Pekalska and Duin, 2005) among voxels in the thalamic region. Dissimilarity represents the difference of each voxel from the rest of the voxels along each dimension; spatial position and diffusion tensor. We use multidimensional scaling (Bronstein et al., 2005) to produce a compact encoding of the dissimilarity information. This reduces processing time for the clustering algorithm. Finally, we use k-means clustering (Hartigan and Wong, 1979) to automatically segment the thalamic nuclei.

Automated identification of the human thalamic nuclei using local white-matter properties

PIZZORNI FERRARESE, Francesca;MENEGAZ, Gloria;
2012

Abstract

We introduce a method for automatic segmentation of the thalamic nuclei. The method uses diffusion-weighted MR images to subdivide the thalamus in a predefined set of nuclei. The method relies on two basic assumptions. First, it assumes that white-matter fascicles’ orientation is homogeneous within each thalamic nucleus (Wiegell et al., 1999). Second, it assumes that nuclei are spatially homogeneous (Wiegell et al., 2003). Based on these assumptions thalamic nuclei are segmented by a weighted combination of local tensor information (Basser et al., 1994) and spatial coordinates. The method avoids manual intervention by computing in each subject the dissimilarity (Pekalska and Duin, 2005) among voxels in the thalamic region. Dissimilarity represents the difference of each voxel from the rest of the voxels along each dimension; spatial position and diffusion tensor. We use multidimensional scaling (Bronstein et al., 2005) to produce a compact encoding of the dissimilarity information. This reduces processing time for the clustering algorithm. Finally, we use k-means clustering (Hartigan and Wong, 1979) to automatically segment the thalamic nuclei.
diffusion weighted MRI; clustering; thalamus
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11562/430333
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact