Creating On-Line Graphical Illustrations for Medical Texts

We present a new approach for connecting textual and graphical information within interactive systems for medical illustration. In contrast to earlier approaches where text is used to label a predeened image, we focus on interaction based on long given texts. Thus our approach resembles work with medical textbooks rather than anatomical atlases. We aim at the creation of images and animations based on geometric models which illustrate the written contents of a given text. Within printed teaching and reference materials, the connection between text and images is essential for the understanding of the concepts being explained. However, due to the physical properties of books, particularly medical texts with many pictures and interspersed text, the integration of textual and pictorial information is conceptually diicult for readers, as images and text referring to them are often on diierent pages (Waldeyer and Mayet, 1992). Indeed, often materials from diierent books must be integrated. Multimedia systems try to overcome this problem by directly linking images and text. Thus a user may request an image as the illustration of a speciic fact by activating a link. But also here, predeened images or animations are shown and a user generally has no possibility to request an illustration which exactly ts his or her intentions or reading history. As in textbooks, images accompany the text and are often displayed with no further direct connection to the text. An analysis of some relatively new systems in the area of anatomy reveals that their focus is almost exclusively on the graphical representation. VoxelMan (HH ohne et al., 1994) includes diierent views on the model and images created with diierent techniques (rendered, CT, MRT) in a single illustration which can be labeled upon the user's demand. Longer explanations can be shown, though no attempt is made so far to couple these directly with the illustration. By contrast, the Zoom Illustrator (Preim et al., 1996) pays special attention to the consistency of rendered images and their textual labels. The textual information displayed in the label is typically rather short (up to 30 words). The labels are directly attached to the graphical objects and thus the connection between graphics and text is more local on a per object basis.