Biomedical Engineering Reference
In-Depth Information
motivation for the model, the degree of realism that is desired, the amount of
detailed physiological and behavioral information that is available for the system
under study, whether simulated sensory stimuli can adequately capture the real
experimental situation, and the extent of the computing resources and program-
ming skills that are available.
2.
FUTURE WORK AND RELEVANCE TO BIOMEDICINE
Most of the work to date in functional neuronal modeling, as summarized
above, has totally ignored the geometry of the neural network. This is adequate
when the timing of axonal conduction can be neglected, when the supply of
metabolic substrates (glucose, oxygen) to neurons can be taken as always ade-
quate, and when development of the adult, functioning system is not at issue.
However, questions involving the functional consequences of the embodiment
of the nervous system in an organism (some would say the necessity of em-
bodiment to the ontogeny of function) are increasingly of interest and will re-
quire the development of a new integrative style of modeling that includes
relevant aspects of the geometry of the nervous system and geometric con-
straints on network connectivity and complexity, as well as constraints on meta-
bolic resources. To take into account more of the relevant mechanical and
physicochemical factors in integrated models will require new combinations of
tools. In particular, to take locally discontinuous obstacles, forces, and signals
into effect, it will be necessary to go beyond differential equations and work
with networks of interacting cells. Models of this kind will require, at a mini-
mum, more detailed representations of cell geometry, diffusion equations to
handle small-molecule signaling, something like cellular automata rules
1
to
model contact signaling, and, especially, finite-element modeling to deal with
the operation of mechanical forces across tissue volumes.
2.1. Modeling the Geometry of Nervous System Structures
2.1.1. Computer Representation of Neural Tissue
Reasonably accurate computer representations of brain tissue structure at
the cellular level will be essential if growth models are to be integrated with
functional models. A key problem is that space must be filled, yet growth and
cell division must be accommodated. Local growth has long-range effects be-
cause existing nearby structures must be pushed aside, and these in turn push
aside more distant structures, and so on until a hard boundary (the cranium), or
the edge of the model volume is reached. Space can be filled with hexahedral
bricks, but these introduce spurious anisotropies (71). In addition, vasculature
