Clustered health technology projects address emerging trends in imaging and language processing

Brussels 19 November 2001During the recent subcluster meeting on "Intelligent Systems for Minimally Invasive Diagnosis and Treatment Planning", held in the buildings of the European Commission's Directorate-General for the Information Society, a few remarkable novelties were presented in diagnostic medical imaging as well as in intelligent interface technologies for medical information queries in databases. Attendants were introduced to the concepts of image fusion and Molecular Medicine, to Ultra Wide Band technology, and to a new middleware for natural language understanding designed for medically-related terminology.


What is the trend in diagnostic imaging? Leonard Fass from General Electric demonstrated that we are evolving to multi-modal real time 4D imaging very soon. Data from a lot of different imaging techniques will be integrated and the accuracy will increase. In twenty years, this could enable molecular medicine. There are several barriers along the way. Reimbursement policies and legal regulations are one big category. Another problem is that very specific therapies will be developed leading to very small niche markets, and it is not clear yet how they can be served.

Because the data collection rate of imaging equipment is increasing rapidly - MRI in the lab for instance already works with a magnetic field of 8 Tesla - data acquisition rates for CT will go up to 500.000.000 bits/s in 2002, so a computer-assisted process is needed for detection and diagnosis. As an example, Mr. Fass cited ImageChecker for breast cancer which detects 85 percent of overlooked cancers. The first step to be taken is image fusion: we see for instance the rise of CT/PET, MR-PET, etc. Fusion is also the trend in imaging in the therapeutic process, for instance in image-guided surgery. Fusion images are generated from MRI/MRSI/CT scanning, and image segmentation techniques can show the response to chemotherapy.

Imaging also moves to the level of molecular imaging. Sizes are already coming down from the millimeter to micrometer range, eventually allowing for enzymes to be imaged. Enzymes are responsible for a lot of diseases. The eventual grail, Molecular Medicine, will integrate "everything". A lot of informatics is needed to get that to work. How important is this all? We live longer, Mr. Fass noted. A lot of diseases like Alzheimer's go up exponentially with age. So if we can delay the process for only a small number of years, it helps a lot. Telemedicine will become ever more important: the bathroom will turn into the most important diagnostic laboratory.

Due to some recent inventions, an already longer existing technology, Ultra Wide Band (UWB) technology, may hold some promise as an additional image collecting technology for diagnosis. Antonio Cuce from STM explained the technology and its possible uses. UWB can be applied as a kind of radar scanner. In contrast to radio wave technologies currently in use, UWB utilises a very broad frequency band. High-powered UWB cannot be used since it would give too much interference with traditional equipment.

In 1994, Lawrence Livermore National Laboratories in the United States developed a low power, inexpensive UWB device which brings the technology into reach for medical applications too. It has several advantages, such as low electromagnetic energy emmission, high miniaturisation capability, and low implementation costs. It can be operated using normal battery power supply and has the possibility to probe the motion of any internal organ. Application areas under investigation are cardiology, pneumology, ENT, emergency medicine, and more.

Avantages over current instrumentation are that it is a non-contact technology. Hence, there is no need for cleaning and it can be used on patients wearing clothes. It can be operated remotely. A problem to be studied is the inter-operation of the generated images. There is not yet sufficient experience. UWB body models will have to be developed. UWB is currrently awaiting approval by American and European regulatory authorities.

There is more to searching and presenting medical information than entering a few keywords in a search engine and getting lists of answers back, some useful, some not. Hans Cools from L&C explained the current state-of-the-art in natural language processing. Natural language processing started in the late seventies and has now reached the level of analysing sentences semanticallly, i.e. at the level of understanding. The next level, the pragmatic level which discerns whether a sentence is humorous or not, for instance, is still out of reach for computers.

The basis of computer based understanding is an ontology, a language-independent group of concepts in a field, with realationships amongst them. An example is "gaster", a synonym for "stomach", described as a "type of organ". L&C has a medical database with one million concepts and 3 million links. You can ask questions to the database in your own words. It is a kind of middleware that can be built into end-user applications. Output can be generated in for instance XML, SGML, HL7, Konna, and a lot more. The company's database is used as starting point for a number of products such as terminology management, i.e. Linkbase and Linkfactory, and semantic search like FreePharma - extraction of medication information from free text, and TeSSI - terminology supported semantic indexing.

More information on the second subcluster meeting on "Intelligent Systems for Minimally Invasive Diagnosis and Treatment Planning" is available in this VMW issue's article European medical project partners gather around Open Source software and clinical validation issues. To know more on the state-of-the-art of medical image visualisation, please also consult the VMW November 2001 article Star Trek technology inspires new ways to screen patient body.

Ad Emmen

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