Graphical style sheets: Declarative, constraint-based display of biomedical domain graphics

Biomedical professionals use data graphics widely in research, teaching, and patient care. There are families of specialized data graphics, which we term domain graphics, that incorporate domain-specific layout and drawing conventions to display the unique characteristics of data from a particular field of study. Examples of biomedical domain graphics include pedigree trees, clinical flowcharts, and secondary-structure diagrams. Although domain graphics are used widely in the biomedical literature, they rarely appear in user interfaces to biomedical information systems. The principal challenges to building such user interfaces lie in identifying the characteristics that make a particular family of domain graphics well-suited domain data display, and building systems that automatically generate graphics that share these characteristics. We describe a method of modeling families of domain-specific data graphics that facilitates the use of such domain graphics as visual metaphors in biomedical user interfaces.; In this dissertation, we hypothesize that the design characteristics shared by a family of data graphics can be represented as declarative, knowledge-based graphical style sheets that a general-purpose visualization system can use to generate data graphics automatically. Graphical style sheets define the layout and drawing conventions shared by members of a particular family of data graphics. We describe a conceptual framework for visualizing data using (1) a point- and path-based representation of graphics, (2) a constraint-based representation of graphical relationships, (3) a frame-based representation of data and (4) graphical style sheets that link data objects to graphical objects. We have created PALLADIO, a scripting language and visualization system for describing graphical style sheets and for visualizing data according to the design preferences declared in these graphical style sheets.; PALLADIO implements programmatically the design knowledge needed to generate domain graphics automatically. We evaluated our framework by experimentally determining the range of domain graphics that could be encoded as graphical style sheets, the extent to which these style sheets could be used to automatically generate domain graphics, and the extent to which they could be used to evaluate other, manually-generated domain graphics. We show that a system based on our framework can represent and generate a range of useful domain-specific information graphics, and may serve as a useful foundation for general-purpose information visualization systems.