Linking a relational database of biological features to computer-aided reconstruction of tissue.

The many branches of biological science, from basic to clinical, use anatomical illustration as a daily work tool. Often the rendering is used as a substrate for other information such as functional mapping as in brain research or chemical uptake as in PET scans. An improvement over current anatomical visualization systems would allow for the storage and manipulation of biological features within a relational database (RD) framework. With current systems, adding new features requires large expenditures of time and effort, since they are usually built to display one biological feature. Workstation based, mechanical computer-aided design (MCAD) tools often have the facility to calculate internal variables such as temperature or stress. Methods already exist within these tools for impressing these data onto the image of an object for illustration. The present project centers around determining the elements necessary to harness the power of a standard MCAD system for anatomical illustration and linking ancillary biological information to this display using a commercial RD. A visualization system has been developed which demonstrates the ability to query and display features, found within tissue, as pseudo-colored distributions without the need for reprogramming. Development of this system has shown that the solids modeler does not provide an adequate algorithm for computer-aided reconstruction of tissue from serial sections. We also have found that the complexity of biological structures mandates automatic triangulation of the interior of reconstructed solids, if development is to proceed to three-dimensional displays of distributions. Furthermore, contour point decimation is essential to confine computational loading to interactive time spans. We also note that a relational record structure does not lend itself to manipulation of multidimensional information and does not provide the necessary tools to maintain database consistency. Encapsulation of data features with their linked geometry along with development of an appropriate operator space is necessary to obtain consistency.PhDApplied SciencesBiological SciencesBiologyBiomedical engineeringBiophysicsNeurosciencesUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/128647/2/9116235.pd