European medical project partners gather around Open Source software and clinical validation issues

Brussels 19 November 2001On November 19, 2001, the second subcluster meeting "Intelligent Systems for Minimally Invasive Diagnosis and Treatment Planning" took place in Brussels. The subcluster consists of medical projects in this field funded by the European Commission under the IST programme. The subcluster is part of the cluster "Intelligent Systems for Health Professionals", which includes a second subcluster called "Intelligent Systems for mobility of Health Professionals". The main goal of the meeting was to create synergy amongst the projects that are working in the same area. This starts, with knowing in detail what research is done in the other projects. Hence, the new project OTELO had a longer presentation. Furthermore, common problems and common opportunities were identified. The two major ones were the possible creation of an Open Source library of imaging algorithms and software, and experiences and procedures for clinical validation.


In the opening presentation, the European Commission representatives for the Directorate-General of Information Society Jean-Claude Healy and Ilias Iakovidis explained the objectives of the current IST programme, the cluster, which comprises approximately 52 projects, involving 348 organisations, with a total budget of 84,5 million euro, and the subcluster, and provided a brief sketch of the future, the Sixth Framework Programme which starts end of next year. Networks of excellences and clustered projects will play an important role there. The HealthGrid forms an example of such a large collaboration. Within the Commission, Andreas Lymberis is responsible for the day to day operation of the subcluster "Intelligent Systems for Minimally Invasive Diagnosis and Treatment Planning", which currently includes 12 projects, involving 95 organisations, with a total budget of 37 million euro.

New kind on the block is the Otelo project. It started on September 1st, 2001. Otelo combines image processing and telemedicine techniques to create a fully integrated end-to-end mobile tele-echography system that can be used by people who cannot go in person to a medical expert. An OTELO echographic system will consist of a master unit in a medical centre, and a portable echographic slave unit. The remote system can be operated by any person. No medical training is needed. It consists of a probe which can read six degrees-of-freedom haptic information. Image compression techniques are used before the data are sent to the master unit. To correctly position the probe, robotics technology is used. The medical expert has a virtual probe which is linked to the real one. The project will produce three generations of master-slave prototypes. The Otelo system does not rely on high-bandwidth, as this may not be available in remote areas. Communication links can range from 56 Kbps up to 256 Kbps.

Lasse Jyrkinen from Oulu University in Finland explained why he thinks an Open Source initiative for medical image processing software is needed. Two of the main problems today are that first, there are many algorithms described in the literature which are hardly used because no easily accessible implementation exists, and second, that there is no general easy to use framework to test implementations and co-operations between different software modules. First, Mr. Jyrkinen tried to take away the scepticism that exists in commercial software companies against the Open Source model where you "give away" software that takes a long time to develop.

Lasse Jyrkinen sees several usable business models: support provider, loss leader, hardware vendor, or accessoring. As an example, he presented the business case of a start-up company that develops a brain simulator system. In their product, 75 percent of the code is open source, and 25 percent proprietary code. And there are more examples, for instance a major MRI manufacturer who needed state-of-the-art 3D visualisation into its MRI intervention product. The main goals of the new initiative should be to build a software package, a transfer of selected algorithms, and documenation. Additionally, the results should be disseminated and evaluation of conformance could take place. Partners in such an initiative can be algorithm providers, industrial providers, implementors, and disseminators.

According to Mr. Jyrkinen, the most important success factor is critical mass. There must be sufficient algorithms, sufficient providers, and a large enough number of quality users. Lasse Jyrkinen does not start this initiative empty-handed. His institute has algorithms and implementations representing around 30 person years of work in DICOM connectivity, system libraries, image processing, and algorithms. There is already a significant activity around Open Source at an international and European level. An example of an Open Source European initiative is the project Spirit, funded under the 5th Framework IST programme: Applications relating to Health. The goal of Spirit is to create a community around medical open source software and to provide link exchanges and download of medical projects.

Dr. Nikolaos Zamboglou from the Offenbach Clinic in Germany described his experience with going through the clinical validation phase. This is one of the necessary steps in the process from prototype to product in the medical market. Phase 1 in this process started with an EC-funded project called Virtuoso. The idea was to build a virtual simulator for irradiation treatment dose. This was done and a prototype was created. Nucletron performed a market analysis. Clinical evaluation was done by the clinical project partners. Functionality and user interface were tested. The evaluation revealed several shortcomings. The prototype was Unix based while most clinicians were accustomed to Windows. The speed and connectivity were too low. The user interface needed improvement and some functionality was still missing. But the basic working, especially the telemedicine part was great, the evaluators concluded.

In phase 2, product development was entered. It started with a second market analysis. The prototype was demonstrated to several users including physicians, physicists and companies. From this evaluation, a list of necessary improvements was compiled. This was the basis for a new specification, with redefined GUI, working procedures, and DICOM RT connectivity. This all was written down in a functional requirements specification (FRS) document which also included risk analysis and test procedures. Phase 3 was the implementation process. The code was redesigned, CE and FDA requirements were taken into account, everything was tested and documentation was designed and developed.

Then everything was ready for phase 4, the second clinical validation, which consisted of doing a number of test functions in Offenbach and consultation of additional users at early stage. This led to some changes in the specification. Hence, there were several loops between phase 3 and 4. In the second clinical validation, everything had to be checked, from the Dicom 3 interface to CT/MR scanners, the other interfaces, several patient positioning set-ups in the CT, several beam configurations, and many items more. All according to the test plan. Clinical validation is hard, boring work. A prototype is necessary for a real evaluation. But Dr. Zamboglou estimated that from prototype to product, two to three times the effort of creating a prototype in the first place is needed.

Christian Huberson reported on clinical trials within the MI3 and Virtual Orthopaedic European University (VOEU) projects. New methods for placing pedicle screws were tested in these trials. Standard methods result in 10 to 40 percent of misplaced screws. Questions to be answered in this clinical trial, in fact in any clinical trial are:

  1. What do we want to demonstrate?
  2. How do we demonstrate it?

In this case for instance, a criterum for misplaced screws was defined: 0.2 mm. Misplaced screws can also lead to secondary complications. This should decrease but how can this be measured? That proved to be difficult. An outcome measure must then be defined. It has to be demonstrable, reliable, and valid. Hence, you have to make assumptions about the results you expect. When setting up comparisons, Mr. Huberson noted that it is best to concentrate first on cases where the old method shows its poorest results.

Concerning ethical issues, one has to carefully study and follow those defined by the ethical committee in the country. Beware, the rules differ from country to country. Mr. Huberson also mentioned some pitfalls to avoid:

  • lack of planning
  • missing or incomplete data
  • exclusion of criteria that are too restrictive
  • it is always necessary to undertake pilot sudies of the logistics
  • if you have little experience, then collaborate with people who have
In the discussion afterwards, it was stretched that clinical trials are easier than clinical validation and that there exists a clear difference between the two.

Ad Emmen

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