KHRESMOI: Knowledge Helper for Medical and Other Information users (KHRESMOI)

Published on: 31/07/2013
Document

Khresmoi automatically collects the most up-to-date medical and health information from various sources (journals, websites, books, images), automatically makes sense of all of this information and makes it searchable so that it is easily and quickly available to the general public, to doctors and to radiologists according to their expertise. The system is multilingual and includes a search for images so that, for example, radiologists can find X-ray images similar to those of a particular patient and then find what the diagnoses for these similar images were.

The system allows access to health and medical data:

  • from many sources,
  • analysing and indexing multi-dimensional (2D, 3D, 4D) medical images,
  • with improved search capabilities due to the integration of technologies to link the texts and images to facts in a knowledge base,
  • in a multi-lingual environment,
  • providing trustable results at a level of understandability adapted to the users.

KHRESMOI combines multiple data sources and knowledge derived from various heterogeneous knowledge sources. This includes text sources such as online journals and books, and trusted websites; and image sources, including images from journals and images from Picture Archiving and Communication Systems (PACS) at radiology departments.

Image removed.

Policy Context

The project operates in the following context: Members of the general public frequently seek medical information online. This process is currently inefficient, unreliable and potentially dangerous. It is thus important that they are provided with reliable and understandable medical information in their own language. Medical doctors need rapid and accurate answers – a search of MEDLINE takes on average 30 minutes, while doctors have on average 5 minutes available for such a search. Furthermore, over 40 % of searches do not yield the information required. Radiologists are drowning in images – at larger hospitals around 100GB of new images are created per day.

Description of target users and groups

  • Members of the general public, including people who are ill, people with a relative/friend who is ill and non-professional seekers of medical information.
  • Medical doctors (MDs), in particular general practitioners and various health professionals.
  • Radiologists in university or non-university hospitals.

Description of the way to implement the initiative

The aim of the project will be achieved by:

  • Effective automated information extraction from biomedical documents, including improvements using crowd sourcing and active learning, and automated estimation of the level of trust and target user expertise.
  • Automated analysis and indexing for medical images in 2D (X-Rays), 3D (MRI, CT), and 4D (MRI with a time component).
  • Linking information extracted from unstructured or semi-structured biomedical texts and images to structured information in knowledge bases.
  • Support of cross-language search, including multi-lingual queries, and returning machine-translated pertinent excerpts.
  • Adaptive user interfaces to assist in formulating queries and display search results via ergonomic and interactive visualisations.

Technology solution

The research will flow into several open source components, which are integrated into an innovative open architecture for robust and scalable medical and health information search. Standard medical terminologies are used and the LinkedLifeData is a source combining these semantic resources. Collaboration of the project with health informatics standards is also active.

Technology choice: Proprietary technology, Open source software

Main results, benefits and impacts

The expected results include:

  • Medical Impact: Improve the access to medical information for doctors, so that they have more time to talk to and to treat patients, having all the information required for doing so more effectively. Convert the flood of radiological image data into a boon instead of a curse.
  • Scientific Impact: Address the lack of publicly available large-scale data sets and realistic task-based scenarios on which to assess new technologies. Make available cutting edge techniques implemented in open source software.
  • Industrial Impact: Improve existing open source products’ stability, features and performance, and hence their attractiveness and suitability for wider deployment.
  • Public Impact: Members of the public will be using the Health on the Net search engine, improved by the KHRESMOI technology, relatively early in the project.

Return on investment

Return on investment: €5,000,000-10,000,000

Track record of sharing

As several of the tools have been developed as open source there has been sharing from the very start of the project and it will continue well beyond the end of the project.

Several annotated data sets have been created in the project and share with other scientists and this is also an important impact that will remain well beyond the end of the project.

There were also several direct collaborations with projects by inviting representatives to the project meetings to exchange ideas; several formal memorandums of understanding were signed with other projects.

Lessons learnt

  1. Management structure based on fully open information in a wiki accessible to all project members is an important part of effective and efficient management in the project.
  2. Combining many existing tools and extending them in an integrated project is a good approach as the development does not start from zero but many existing parts can be reused making the project operational from the very start.
  3. Combining several complex technologies has a large potential and such integrated projects have the possibility to combine techniques across domains with a critical mass that many other instruments do not have.
Scope: Pan-European