From source diversity to flexible medical data presentation and user interaction in MediMedia

Brussels 28 March 2000Dr. Thomas Berlage from the German National Institut für Angewandte Informationstechnik talked about the concept of integration in the MediMedia search engine during the workshop on medical imaging. The project partners introduced two solutions to present the diversity of medical images in a functional and flexible way to the user. The application of the XML data format and the


Visual Integrator software module have considerably facilitated both the query and manipulation of medical images within the framework of the MediMedia architecture.

The MediMedia project has addressed the issue of interoperability between the medical image databases of seven European projects through the design of a multimedia search engine with a universal and user-friendly interface. The fundamental problem in the process of data integration is the fact that you have four types of diversity, according to Dr. Berlage. The user is confronted with diverse data, such as static and dynamic images; accepted standard images, like DICOM, and non-accepted standard ones. In second place, the images come from multiple sources, created for different purposes. Third, the data is used in a variety of context-specific scenarios ranging from diagnosis to education, training, and research. Finally, the great number of medical domains, including radiology, cardiology, pathology, gastrointestinal endoscopy, hematology, and surgical planning each have other data needs.

The first solution, applied in MediMedia, to bring a workable homogeneity in this rich diversity constitutes the use of XML to describe different data in a similar way. Since XML is just a standard syntax, the semantics have to be defined. Dr. Berlage explained how the MediMedia team has determined a standard format, called the data type definition (DTD) document, in order to record the specific tags and the semantics associated with them, to generate the document's syntactical outlook. For simplicity, there is only one single DTD covering all the data structures encountered in the project, and which consists of two parts, a mandatory and an optional one.

The mandatory DTD item describes the common parts between the different image sources whereas the optional DTD part contains specific data relating to a particular source. The varied database contents is automatically being converted into valid XML documents and referenced via links. In such a way, the most complex cases consisting of multiple images, can be described and made available to the user's Web browser via the common search engine. Dr. Berlage in this regard spoke of the challenge to adapt novel databases to the system but in the XML community, a host of tools are being developed at the moment with rapid changes being introduced every six months. MediMedia can profit from this evolution but the project needs specific mechanisms too.

Once the data is gathered in a DTD format, the Visual Integrator forms the second solution to present the information in a universal context. This software has a double task. On the one hand, it allows the user to send and formulate queries to the search engine in the easy or expert mode. Easy mode means that the user just has to type some words to receive the related images. Expert mode enables the user to get specified details from different systems. Dr. Berlage stated that you need some sort of query refinement to reduce the large number of hits, in case there are too many images. On the other hand, the Visual Integrator is responsible for the adequate presentation of the results, transforming the XML documents into HTML files and converting the images.

Each query session consists of different states. Associated with every state is an HTML template file which creates the HTML page, allowing the user to adapt the template to his personal needs and to different environments and situations, and interaction scenarios, such as training. Also, the Visual Integrator incorporates a set of media converters to translate special image formats into Web-based formats, like JPG, and to apply compression and resolution reduction in case of low bandwidth availability. Via specific Java applets, the templates can be extended in order to integrate advanced interaction techniques, such as 3D data sets and virtual models.

Dr. Berlage summarised the specified integration concept of the MediMedia search engine as a bi-directional process. From the database sources' site, a decontextualisation removes the images from their specific context to present them in more general terms by transforming and referencing the data, which is called normalisation. From the client's site, you need a recontextualisation since the user expresses a query, that has to be refined to get an appropriate presentation. The interaction is adapted depending on the client's context. In the middle, the process requires a physical organisation which translates between the two sites. This broker fulfils the normalisation task among the different sources as a mission. For a complete project description, we refer to the VMW April 2000 article MediMedia provides common gateway to multi-modal medical images in European databases.

Leslie Versweyveld

[Medical IT News][Calendar][Virtual Medical Worlds Community][News on Advanced IT]