Figure 8 (colorplate) illustrates the action of hue-balls. Figure 9 shows all the components of diffusion tensor volume rendering that have been described so far. The diffusion tensor dataset being rendered (the same as in Figure 6) is a 128 by 128 by 60 voxel scan of a live brain. Our volume renderer is a simple ray-caster which uses one ray per pixel. When rendering a tensor volume, it calculates the eigensystem at each point along the ray, then uses the eigenvalues to determine opacity (according to the anisotropy opacity map), and the eigenvectors to determine lighting (according to lit-tensors).
Figure 9(a) is a rendering of a scalar dataset which was generating by mapping the tensor dataset through an anisotropy opacity map similar to the third map in Figure 6. The general structure of the anisotropic region as a whole is clear, but the distinct white matter fiber tracts are not easily distinguishable. This is remedied in Figure 9(b), where the volume is colored by the hue-ball, but the surface shading is the same as before (using only one white light). Now the regions with distinct anisotropy directions attain distinct colors. For example, the cingulum bundle (blue) is now clearly visible as a distinct tract from the corpus collosum (orange) below it. Some other major features are labeled. Figure 9(c) shows the result of using lit-tensors without a hue-ball. The light positions and colors are the same as in 9(a). The dark regions are mainly planarly anisotropic; they need to be correctly aligned to reflect light, whereas linearly anisotropic regions are inherently more reflective. Figure 9(d) shows the use of a hue-ball with lit-tensors. With one fixed rendering, the effect of lit-tensor highlights is not as clear as in the accompanying video (see slides here and here), in which the highlights are seen to travel across the features with viewpoint motion.