Increasing and validating the contents of medical image databases at a European level

Brussels 28 March 2000Co-ordinator of the MediMedia initiative, Dr. Michel Delvaux from the Hôpitaux de Toulouse, described the past efforts of the MediMedia consortium as an attempt and an ongoing challenge to increase and validate the contents of medical image databases in Europe. Today, the search engine technology is in place, as Dr. Delvaux stated in his talk during the workshop on medical imaging, but the user is now asking for valuable content. The fact that we are able to access in a "one-step" query medical images which are scattered over the Internet, using the search engine is not enough. We also need to implement services which can provide specific content to the user according to a set of quality criteria.


Basically, there are two types of content, namely general content covering a large area of interest, and specialised content which only targets a narrowed field. A database with general content, for instance, can consist of a number of images showing several simple diseases, that will be helpful for a general practitioner or a patient documentation. However, this type of large content databases frequently is not exhaustive whereas the technical quality remains quite poor. Specialised content overcomes this bias since databases of this type are exhaustive but they are often limited to one speciality so you have to address different databases to find the answer to your query.

At the moment, the user is faced with databases which are either limited to one technique, to one type of disease or which are fragmented in their user interface, in their construction, in the services they provide, and devoted so specifically to one scenario that you can not apply them to another scenario. How can we generate a content which can be used in a variety of scenarios? According to Dr. Delvaux, it should be possible to provide access to personal collections of selected cases for the benefit of the community. Second, there is the systematic collection of cases for clinical use. Such case collections are already stored by specialised university units, so we can implement a system to make these collections available retrospectively or choose for a prospective submission of newly generated collections.

There are also local clinical databases with case collections that however are not publicly available. The Gastroenterology Department of the Hospital in Toulouse, for instance, since three years hosts a collection of 20.000 cases of endoscopic procedures. An important drawback of such a local system is its suffering from a lack of standardisation, its incompatibility with other coding systems, and the requirement for privacy of data. The patient data therefore can not be transferred together with the image. In addition, the images have been indexed according to the patient so the structure of the database has to be applied when the patient data are used. The liability of these systems can not be ensured because the databases do not evolve over time, thus making it extremely difficult to scientifically exploit the material.

To overcome these problems, the GASTER project started with eleven centres spread over different countries in the European Union to collect cases which were already stored in the endoscopy units of the participating members. By now, over 3000 images in DICOM format are retrievable via the Web but the certified access to this retrospective collection has been granted to scientific communities only. In the EURORAD project, a prospective collection of cases has been implemented with peer review. This probably constitutes a way to be used in the future to increase the content of medical image databases but the peer review takes time, as Dr. Delvaux pointed out. It is not sufficient to submit rare cases because we need normal cases as well in order to enhance the educational value of the database.

The databases can be accessed by individuals, user groups in associations or in universities, or academics, who already have made some efforts to provide images for the databases. Especially scientific societies have organised this work in order to obtain content. Today's problem with this content is that it lacks common access, liability, and self-sustainability. A solid quality control process needs to be established to link these databases to our network in the future, as stated by Dr. Delvaux. The control process includes both medical and technical quality. Medical quality comes down to clinical relevance of the content. This can be achieved by peer review or by a content quest to detect the content and bring it to the large network that is implemented.

We also need to put more effort into the standardisation of the indexing and coding systems. Text information has to be assessed as well and adapted to the different users and to the purpose of the database which is either clinical or educational. Here, standardisation is mandatory, as well as for the privacy of data. As far as the technical quality is concerned, elements such as basic quality of the material, accessibility through the user interface and over time via the Internet, flexibility via XML, liability over a period of time, compliance to standards, security for the patient, and protection against unintentional mistakes or deliberate sabotage, are required. The quality control has to be performed by a number of actors, such as the scientific societies, regulatory authorities, standardisation bodies, and the industrial providers.

The scientific societies and the industry are the basic actors which will have to interact together to define the quality criteria that should be applied to all the databases. These quality criteria need to rely on the rules issued by the standardisation bodies and the regulatory authorities. On the one hand, the basic actors have to take into account all the rules for implementation of the databases, but on the other hand, only their experience is able to provide the necessary feedback for the standardisation bodies and regulatory authorities to refine the rules over time. Hopefully, this will lead to a long-standing co-operation in a certification structure in which we will need the active support of the regulatory authorities to turn this dream into reality.

In conclusion, Dr. Delvaux expects a network of distributed databases will be built to increase the content that we intend to have. The content comes from the users but these users must be organised. The scientific societies and the publishers are the major actors who must collaborate to provide the content. On the other hand, equipment manufacturers, software developers as well as service providers are responsible for the technical environment but they have to interact with the other parties. The quality of the content that we want to reach, can only be achieved if we can certify and standardise that content in an institute or a permanent structure. For a detailed description of the MediMedia project, 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]