Tissue Study

Goal

To realistically render scenes involving internal human organs.  The scene may specifically focus on organs located in the abdomen and chest so that large arteries can be visualized in the midst of other tissue types.  As an additional requirement, the scene will include light emissive blood as the result of a possible bioluminescent agent being injected into the bloodstream.

Materials

The description of the anatomy will consist of a multi-detector spiral CT scan taken at 0.5mm resolution in plane and through plane. (Voxel size 0.5mm per side).  The blood will be contrast enhanced during the scanning so that the density level of the blood is higher than the surrounding tissue.  The volume can be manually labeled per voxel based on the scalar information in the volume data and anatomical knowledge.

Proposed Rendering Algorithm

In order to render the tissue realistically, I need to use some version of radiative transport that includes the effects of scattering.  Intuitively, I am worried about using the diffusion approximation to multiple scatter (first introduced to computer graphics by Stam) that is used in most of the papers described below.  Although this makes the transport equation much easier to solve, it may not be accurate enough to represent the varying thickness of the arteries that are filled with light emitting blood as the thickness variations will occur at much smaller than one mean-free-path of scattering for photons in the tissue.  This leads me to believe that a more accurate representation of multiple scattering, as used in the exhaustive Monte-Carlo approach described in the Pharr paper, would be worthwhile.

The next issue which also affects the scattering calculation methodology is whether to represent the scene as a participating media or to represent the scene as the isosurfaces of the tissue boundaries which can then be solved with a much faster ray tracing approach.  I am leaning towards attempting the full volume rendering solution to the problem as that will allow me to play with the accuracy of the multiple scattering terms, however, the BSSRDF technique utilized by the papers below is so powerful it might be accurate enough and much faster.  With the volume rendering approach, I would also utilize a volume photon map to reduce the convergence time of a volumetric monte-carlo approach.  This may take a very large amount of photons to accurately render the scene.

I plan to animate the flow of the bioluminescent material in the blood by doing a particle simulation of advection and diffusion relative to the boundary conditions imposed by the arterial walls.  This will not be a major emphasis of my work, but should add more visual appeal and also reality to the stills and animations of the scene.

Inspiration