A crucial step in image processing, as well as in other types of pattern research such as genome analysis, is the creation of pattern-theoretic knowledge representations that are both realistic and computationally tractable. We seek to propagate this view by conducting a workshop in which the central issue would be knowledge representations but would necessarily also include the mathematical analysis of such representations and their implementation into feasible algorithms and efficient computer programs tailored to the new high-performance computing architectures. Participants would include not only mathematicians interested in the analytical challenges but also those actively conducting research in relevant areas such as anatomy, radiology, bioengineering, genomics, computer architectures, multidimensional visualization, and software engineering. The issue of knowledge representation has profound consequences for the research strategy one adopts. For example; for the Visible Human Project (National Library of Medicine) one could build upon conventional image- processing algorithms for visualization, calibration, registration, segmentation, etc., however, with the knowledge representations that have been developed for structure and biological function and their variability (both normal and abnormal), much more can be done. With appropriate representations it is possible to transform the morphology being created in anatomical atlases so that it corresponds to any arbitrary observed instance of the morphology, and in so doing enable transfer to the arbitrary case all of the relevant knowledge included in the annotated standard atlas, as is now being done in brain research. Another example would be the identification of abnormal patterns such as in cancer cytology or in calcification patterns in mammography. In another category would be the application of pattern-theoretic methods to the analysis of DNA sequences now being generated by the Human Genome Project. Such analyses would include the study of intra-species sequence patterns, phylogenetic comparisons, and archeo-biological research. Specific objectives are 1) To foster interaction among key individuals conducting interdisciplinary research in the theory, implementation, and application of knowledge representations based on the emerging field of pattern theory, with emphasis on relevance to biomedicine; 2) To identify key research topics important to progress in the field, including theoretical extensions, algorithm developments, practical implementations, and biomedical applications; and 3) To stimulate awareness of the field by making available to the broad research community, a published Proceedings reporting the current state-of-the-art of pattern theory, its importance to knowledge representation, and its realized and promised contributions to the solutions of problems in biomedicine.
