Next: References Up: Title Page Contents: Conference Page

Structuring Metainformation for Anatomy Images: XML as the Preferred Metainformation Interchange Language

Interactive images are an essential component of multimedia in medical education. The information that is added to an image to make it interactive is termed “metainformation”. Metainformation can include outlines of interactive regions, a name or label for the region, a link to additional information or a hyperlink to another image or document.

Early forms of metainformation embedded the information directly into the image. An image with outlined structures was simply another image. More recently, each outline is stored as separate data, and any outline can be displayed at will on top of the original image.

Today, a common form of metainformation for interactive images is the “imagemap”, available in HTML documents on the Web. The imagemap format provides a few basic types of metainformation, but allows for more sophisticated interaction through inclusion of customized, proprietary software [2].

Aside from the imagemap format, numerous formats have been developed for image metainformation. Each such format requires unique software to decode the format and to present the interactive image to the user. This applies to the Visible Human data and other image collections that are expected to remain valuable for long periods of time. While the images themselves are available in a standardized form, readable at any site around the world using numerous common image reading programs, the same is not true for the additional metainformation that is being created at all these sites.

If metainformation were to be standardized, or if there were a process to interpret metainformation in a standardized form, the information would be open to processing by any program, and not just the program that created it. This could even include programs that have yet to be conceived.

The value of standardized metainformation is illustrated with the examples below. Figure 1 and Figure 2 show two different viewers, both displaying images and information stored in Stanford’s Interactive Image Tool (IIT) format (see Figure 4). In both images, the stored metainformation is used to highlight one structure and to present textual information about that structure. Figure 1 is a full featured viewer, with complex navigational elements and extensive graphic design. Figure 2 is a very simple viewer, with a full list of structures visible in the image, usually used in conjunction with a larger program that presents instructional content. The IIT specification and some examples are available on the Web [3, 4].

A need exists to define a method of creating maps from one representation of metainformation to another, such that the metainformation can be usable at other sites. We propose the Extensible Markup Language (XML) as the common language for representing metainformation, and the Document Type Definition (DTD) (see Figure 3) as the map that makes the metainformation readable by any program able to use and present such metainformation [1].

Why XML?

XML is derived from the Standard Generalized Markup Language (SGML), of which HTML is another derivative. The power of XML (and SGML) is that each site can choose its own representation of image metainformation as long as they also provide a specification, a DTD, stating the “meaning” of each element, and how it is represented. Figure 2 indicates the type of information that may be appropriate in one XML representation of image metainformation. Using the relevant DTD, another site is able to search this collection of images for all images of, say, the heart, by searching in the “name” elements. A third site may wish to download the images and outlines alone, without also getting the links or text, into their own simplified XML representation. Use of XML allows all three sites to maintain their preferred representations, and yet be able to exchange information.

Investigation of existing image metainformation formats shows considerable similarity in the type of information stored and the protocol used for storage. In particular, many of these formats use the HTML tag method. Consequently, a move to XML would be relatively easy for most groups.

In conclusion, with the wide availability of image collections such as the Visible Human data, there will be an increasing amount of additional information generated using numerous proprietary or limited capability public formats. We propose the use of XML as a publicly available metainformation representation language. At the conference, we will present an overview of image metainformation formats in use at sites such as the University of Washington (Digital Anatomist), Johns Hopkins, and Stanford, and will follow this with a brief review of XML as an image metainformation language.

Next: References Up: Title Page Contents: Conference Page