Provider for integrated medical information

Beispiel digitaler pathologischer Bilddaten

The aim of the MedInfoGRID project is to develop a distributed documentation and information system for medical images, diagnostic findings, research and therapy information in addition to providing high-performance computing resources for biomedicine. MedInfoGRID is a multi-disciplinary project that provides services and showcase solutions for existing grid projects.

The project is being carried out in conjunction with the medical faculties of the universities of Magdeburg and Mainz, the Konrad-Zuse Institute Berlin, OFFIS, (Oldenburg Research and Development Institute for Information Technology Tools and Systems) and the radiology software producer CHILI GmbH. The project also cooperates intensively with the predecessor project MediGRID and the Services@MediGRID project, which is being carried out in parallel. The MedInfoGRID project aims to allow doctors, researchers and other interested parties to exchange images and diagnostic findings and make them available for the purposes of research without breaching patient confidentiality or data protection laws.

The open source software to be integrated for knowledge representation as a collection of teaching materials and cases will be tested in areas within clinics. These areas are radiology, pathology and gastroenterology. In order to provide services such as these, methods for compressing and streaming large image files are required in development and implementation. The integration of distributed, digital pathological image data as virtual, online microscopy is particularly innovative. This was developed at the Otto von Guericke University of Magdeburg Institute of Pathology under the direction of Prof. A. Roessner and Dr. Th. Kalinski (SpR) and in conjunction with R. Zwoenitzer (Dipl.-Inform.). The terabyte-sized data volumes place particularly high demands on the solutions to be developed.

The security concept of MedInfoGRID pseudonymises patient-related data in accordance with the laws on data protection. For this part of the project, the MediGRID Secured DICOM functionality was extended to DICOM Structured Reports (DICOM SR).

The development of grid-based services during the course of the MedInfoGRID project involves realising a grid infrastructure for distributed computing for biomedical applications. This allows time-intensive applications that carry out statistical evaluations of brain scan image data to be parallelised and ported to the grid. Brain scan image data obtained from an MRI scanner is loaded layer by layer into the data management system. These layers are then segmented via a workflow process and processed simultaneously in the available processor cores. The potential of this type of grid service was demonstrated in the example of diffusion tensor imaging. Here, the speed of the evaluation was increased significantly, correlating almost linearly with the number of processors. This was developed with the partner Charité and the Konrad Zuse Institute Berlin. Further functions based on the same principle are currently being realised for the evaluation of image data. These aim to facilitate the real-time analysis of brain activity (functional brain imaging).

Web-based access to these applications will be provided via a portal. This will make it possible to load various medical images to the grid and administer them there from any location. Various calculations can be carried out with this data via the portal, in which workflow-based jobs are started online and monitored.

A cluster is provided for this purpose. It is equipped with HP Blade technology, 152 processor cores, fast storage discs for intermediate results, as well as slower, but larger, storage media for archiving medical image data. Approximately 24 terabytes are available in total.

As part of the special D-Grid investment in 2008, additional hardware resources were provided that are used in conjunction with the other cooperation and associate partners for the implementation of computationally intensive visualisation techniques in the grid. These resources are Apple¬ X Servers with 64 processor cores and fast storage media for the explicit purpose of visualisation calculations and applications. An inter-project grid consortium was developed for this reason.
The HP cluster and storage media will be physically separated into one publicly accessible area and one internal area in Magdeburg. This is to guarantee the security of sensitive patient data. The internal area will be used to pre-process the patient data and to develop new, web-based information services.

After the data protection mechanisms have been tested and the internal cluster has been released, the data objects will be transferred anonymously to the public area. All calculation and information services that become available in the future will be accessible via the web portal.

The overall objective is the integration and use in the D-Grid of prepared, anonymous medical information and services related to it. The platform is to be accessible to other interested parties.