Visualization in Connectomics - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 212

built up showing the 3D water diffusion at all points in the volume, which in turn is
related to the directed structures running through those points.
Diffusion tensor imaging. When at least six directions are acquired, a 3 3sym-

metric diffusion tensor can be derived, in which case the modality is described as
Diffusion Tensor Imaging (DTI). Per voxel DTI, often visualized with an ellipsoid,
is not able to represent more than one major diffusion direction through a voxel.
If two or more neural fibers were to cross, normal single tensor DTI would show
either planar or more spherical diffusion at that point. The left image of Fig. 21.2
shows a 3-D subset of such a dataset, where each tensor has been represented with
a superquadric glyph [ 50 ].
DTI visualization techniques can be grouped into the following three classes [ 102 ]:
Scalar metrics reduce the multi-valued tensor data to one or more scalar values such
as fractional anisotropy (FA), a measure of anisotropy based on the eigenvalues of the
tensor, and then display the reduced data using traditional techniques, for example
multi-planar reformation (slicing) or volume rendering. An often-used technique is
to map the FA to intensity and the direction of the principal tensor eigenvector to
color and then display these on a slice. Multiple anisotropy indices can also be used
to define a transfer function for volume rendering, which is then able to represent
the anisotropy and shape of the diffusion tensors [ 49 ].
Glyphs can be used to represent diffusion tensors without reducing the dimen-
sionality of the tensor. In its simplest form, the eigensystem of the tensor is mapped
directly to an ellipsoid. More information can be visually represented by mapping
diffusion tensors to superquadrics [ 50 ] (see Fig. 21.2 ).
Vector- and tensor-field visualization techniques visualize global information of
the field. The best known is probably fiber tractography, where lines are reconstructed
that follow the tensor data in some way and hence are related to the major directions
of neural fibers. In its simplest form, streamlines, tangent to the principal eigenvec-

×

Fig. 21.2 On the left , superquadric glyphs have been used to represent the diffusion tensors in a
3-D region of a brain dataset (image courtesy of Gordon Kindlmann, University of Chicago). On
the right , the cingulum neural fiber bundle has been highlighted in a full-brain tractography [ 8 ]