Web Magazine for Information Professionals


Julian Cook describes a major database of medical images.

The task of MIDRIB ('Medical Images Digitised Reference Information Bank') is to create a system for the creation, storage and networked delivery of image-related information. This involves a complex chain of events that takes an image from its raw form (such as a slide on a clinician's shelf) to the point where the user can retrieve it individually from a digital networked resource of potentially hundreds of thousands of items. In between it has to be captured in digital form and described so that it can be retrieved. This sequence demands a range of processes and software tools.

Securing new sets of images for MIDRIB requires all the stages in this procedure to be in place. These include copyright, patient permission, digitisation, and cataloguing and indexing. Most of the database development work is taking place using existing medical image collections that will form part of MIDRIB, such as the Bristol Biomedical Image Archive and the TICTAC database. MIDRIB is also developing a 'shopping list' of key images which the archive should contain. At the moment this is being done for Obstetrics & Gynaecology and Cardiology.

Because the Internet is so widely accessible and its content so easily manipulated, it is important that we ensure that patients are clear about what they agreeing to, and that they understand the medium. MIDRIB has produced guidelines and information documents on this issue. These also tackle the question of patient permission relating to older images acquired before the concept and potential of the Internet were realised. The documents produced have been approved by the Bristol Ethical Committee, and are awaiting approval from St George's Ethical Committee.

The exponential growth of the Internet means that many groups are now addressing the copyright implications of this new medium with some urgency. MIDRIB's main task in this area has been to examine the work already done and apply it to our situation. We have received considerable input and help from the Wellcome Trust in this area, as well as from the Designers' and Artists' Copyright Society, and have now prepared a draft form for contributors, which has gone to our legal consultants for their opinion.

MIDRIB database records will consist primarily of images. However, it will clearly be necessary for the records also to contain some text fields, in order that the images can be searched for and retrieved. We are now in the process of finalising and implementing a provisional list of the fields for each database entry.

MIDRIB will use the MeSH (Medical Subject Headings) classification system and the UMLS (United Medical Language System) medical thesauri to assist search and retrieval from the database. One of the main advantages of using MeSH is that the system will already be familiar to anyone who regularly uses Medline. This choice has been taken with input from the Wellcome Tropical Medicine Resource and OMNI (Organised Medical Networked Information) projects, both of which have been faced with a similar decision. Subject experts will provide concise but comprehensive descriptions of the image, providing information to MIDRIB's medical indexer.

A database of images is very different from a purely text-based resource and requires specialised technical solutions. We are participating in a working collaboration with System Simulation, a company with an excellent track record in image databases, which is already involved with a number of similar projects in the Higher Education sector. MIDRIB is using the company's Index Plus software, and together we will be extending the facilities of the package.

It is likely that images will be stored in the MIDRIB database in PNG format, which is attractive because of its 'non-lossy' compression (where information is not lost during decompression) and its ability to store metadata within the file. It was developed in response to the licensing problems that have emerged with the GIF format. We are also investigating DICOM, an emerging standard in USA radiography and microbiology.

It is vital that end-users are able to find what they are looking for quickly and easily, whether they are students, teachers or researchers. MIDRIB has decided to offer several different ways of finding images in the database. Free-text searching, structured browsing through the MeSH tree and discipline-based searching will all be available.

With the record structure and searching mechanisms now broadly defined, we are making our first steps in designing a clear, usable and friendly interface which will be consistent and recognisable whether delivered over the Internet or on CD-ROM.

MIDRIB has a mailing list with over 200 members, and the strategy for evaluation is in place and underway. We will shortly be developing a set of guidelines for people wishing to contribute images. An initial service will be available by February 1997. The intention will be to use this for testing and refinement. It is likely to contain the Bristol Biomedical Archive of approximately 14,000 partially- catalogued images, the TICTAC database of prescribed and controlled drugs, a Dermatology collection of several hundred fully-catalogued images, several hundred anatomy images, and a collection of Medical Line Drawings.

The official launch of the project will be held at the Royal College of Surgeons in London on February 27 1997. It will consist of a series of demonstrations of the service, what it contains and how it can be used, as well as presentations and discussions of the technical, ethical and legal issues.