This Small Business Innovation Research Phase I project from Visible Productions L.L.C. seeks to develop the initial version of a software system that can efficiently create 3D polygonal mesh surface models of anatomical structures from 2D section data such as CT, MRI and Visible Human Project images. The software systems will consist of refining and integrating the following existing modules: 1) Semi-automated tracing module, which utilizes neural networks to perform user-assisted and semi-automated tracing of the perimeters of anatomical structures on 2D section images; 2) 3D model-based registration module which transforms adjacent section images and individual structures as necessary to ensure proper alignment and registration; 3) Automatic connectivity triangulation module, an improved triangulation software which creates 3D polygon surfaces from the traced 2D perimeters with little user intervention; and 4) Post-processing system, an improved software for editing, decimating, and reshaping polygonal mesh surfaces for customizing the models. The Phase I will concentrate on development and testing of the semi-automated tracing module since this is the least developed component in Visible Productions' current system. A concept and plan for a model-based registration system will be generated. Visible Productions' existing triangulation and post-processing modules will be improved and the project team will begin the integration effort with other modules. There is a very large market for 3D polygonal mesh models of anatomical structures, but the tools available to create these models are very limited. The users of such models include researchers, educators, medical professionals, diagnosticians, multimedia publishers, simulation design specialists, and entertainment and gaming industry engineers. Visible Productions' software would allow these diverse groups to efficiently create their own models at a reasonable price. Also, the system will allow the creation of custom models for customers at a much faster rate than is currently possible.
