Popular "Emergency Room" series on television offer a biased impression on the work flow process in a hospital. Dr. Bellon stressed the necessities of the real world clinical environment where a correct and up-to-date information stream is of crucial importance. In this process, a lot of actors are involved. The electronic media can play a key role in the flexible and timely flow of data to those places and persons where it is requested. The second generation Web technology holds a tempting promise in store for more functionality and more intelligent data processing. However, Dr. Bellon is not quite convinced that the "Semantic Web" will live up to the high expectations. In his presentation he highlighted the current limitations.
Today, much of the information is still on paper or on film. Ideally, the EPR or Electronic Patient Record should become globally accessible in order to be shared by the different actors. At present, hospitals have to deal with partial records but in future, clinical staff will work with integrated systems. There are already different standards for storing and exchanging medical as well as visual data, such as DICOM for instance, but there are no standards as yet for the EPR structure, Dr. Bellon pointed out. Recently, we have witnessed the birth of eXtensible Markup Language (XML). All the former standards are now being reconsidered in the light of XML but Dr. Bellon fears that this will not solve the problem.
A user-oriented organisation can build small-scale applications to increase the general consensus between medical staff and to foster standardisation. In this regard, XML might help to design common tools for low-level aspects. Dr. Bellon illustrated the usefulness of tele-access to the EPR in a hospital by means of a pilot experience for co-operation between the hospital and a general practitioner. All up-to-date patient items are structured and made accessible to the family doctor via a Web browser which serves as a client. The application makes use of XSL (eXtensible Style Sheet Language) engines and JAVA-servlets. The JAVA code is able to generate XML and XSL style sheets are created to transform the data into "HTML". According to Dr. Bellon, a lot of navigation is involved in this operation.
The question whether we really need XML and XSL for EPR tele-interaction and if the second "Semantic Web" is of more actual help than the traditional Web, is answered with a "probably not" by Dr. Bellon. His team also experimented with providing radiological image processing and analysis services over the Internet. Here, interactive navigation is integrated into the browser with use of JAVA-applets to manipulate tens or hundreds of image slices but this is no standard application yet. It is possible to exploit the intrinsic document technologies of XML and XSL and consider the image data set as a tree to provide the viewer with a grid of images. The pages are available sequentially.
Dr. Bellon explained how the image tree is transformed into a screen tree. In turn, this screen tree is used for the "HTML" transformation. The granularity of the application becomes more interesting for development but so far it is no large-scale experience. Medical image data are being processed for pre-operative planning, for instance. Dr. Bellon indicated how the optimal position of pedicle screws in neck operations can be determined in advance. A set of 2D images is utilised for segmentation with manual correction. This results in 3D images to prepare computer-generated drilling templates which are customised for the individual patient. Also in this process, the new Web technology can pay useful services.
The communication of intelligence itself however is not that easy, Dr. Bellon stated. Distributing intelligence is not the same as combining intelligence, for that matter. This was clearly demonstrated in an experimental hospital tele-service for image matching. Dr. Bellon admitted to the audience that he has a rather pessimistic nature which made him conclude that there exists a lot of hype around the XML phenomenon. Therefore, the XML standard will not solve problems automatically. To meet the future hospital requirements, the "Semantic Web" ought to behave in an intelligent way but up to this point, Dr. Bellon does not see this happen yet. For more information on the activities of Dr. Bellon's research team, please visit the home page of the Medical Image Computing division.