3D Reconstruction Methods for MCell Simulations
Traditionally, 3D reconstructions have
been used in Biological Sciences to extract morphometric data from a volume
of tissue, or simply as a means of visualizing the structure. For
these purposes, surprisingly large imperfections can be tolerated, such
as cracks, holes, self-intersections, or degenerate polygons. For
an MCell simulation, on the other hand, the goal is 3D reconstruction
of the actual topology. Therefore, an algorithm used to generate
surfaces for MCell simulations must meet stringent criteria:
-
it must correctly reconstruct even highly complex, convoluted
topology, e.g. a sponge-like surface (an orientable 2D manifold of arbitrary
genus embedded in 3D with possible boundaries).
-
all polygons in a surface must be consistently oriented to
define inside and outside faces of the surface.
-
the polygons should generally be nearly equilateral triangles
(for efficient mapping of effector sites).
-
other artifacts and noise should be minimal.
 |
Rat diaphragm (30000 triangles) |
 |
Mouse sternomastoid (original mesh: 2.4 million triangles, reduced to 80000 as shown here) |
|
Wireframe views of neuromuscular junction postsynaptic membrane reconstructions used in MCell simulations.
|
We have extensively evaluated many
available surface reconstruction tools (e.g. "surface
reconstruction from unorganized points", Nuages,
IBM DataExplorer, VTK),
and thus far have found that only VTK's marching cubes algorithm meets
all of the above criteria. Even so, highly convoluted surfaces at
ultrastructural resolution require extensive interpolation between contours
obtained from electron micrographic datasets, and this step requires additional
software tools. Once a surface is generated, it is necessary to edit
properties of the triangles that comprise the surface, e.g. add particular
types of effector sites at different densities, or specify different permeabilities
for different ligands. Much additional development of reconstruction
and computer graphics tools is required to simplify and accelerate all
of the above steps.
|
