|MUG > IMI > Research > Projects - Medical Informatics|
PROJECTS - MEDICAL INFORMATICS
During the introduction of a new HIS and the installation of a new departmental Information System for the Institute of Pathology, several older systems (AURA - developed in the 1970s and 1980s) were gradually replaced. However, access to the medical data stored in these systems had to be ensured. In order to avoid a complicated and expensive transfer of several million records into the new system, we developed AURAWeb, a web-based Meta Search Engine for diagnostic findings.
AURAWeb has been in routine use since 2002 and enables access to approximately 12 million records from 33 different medical information systems and - in the context of radiological findings - digital images from 14 PACS. The scope of application is no longer limited to legacy systems. In different fields - for example, pathology - the volume of data of all documentation systems relevant to the specific search are scanned, thereby providing a comprehensive and complete view of the available findings of a patient.
The search for medical findings is carried out directly in the databases (or associated search engines) of the participating medical information systems, which are accessed via different interfaces - SOAP, REST, DICOM, JDBC™, etc. AURAWeb is responsible for the authentication and authorization of the user, the selection of the information systems to be contacted, the transformation and co-ordination of the retrieval queries, as well as the collection, preparation and presentation of the result. The specific characteristics of the individual information systems are not of relevance to the user. Apart from the browser based user interface, AURAWeb also makes a direct access to the search engine available, via a Web service interface.
The Confocal Laser Scanning Microscopy (CLSM) makes a fast, in-vivo screening of alterations in dermal tissue possible (3-5 minutes per lesion). Since this is a new procedure and the pictorial material won from it distinctly differs from pictures created by the conventional Transmission Light Microscopy, the diagnostic parameters have not been completely ascertained. In this study, conducted in collaboration with the Department of Internal Medicine, Division of Oncology and the Department of Dermatology and Venereology, diagnostically highly significant regions in CLSM-images of Melanoma are determined by means of automatic image analysis and emphasized within the image.
Since, with connective tissue, an automatic segmentation of individual cells is not generally possible, the images are divided into individual quadrants and the characteristics, attained from a Wavelet Transformation in different resolutions, are computed within the quadrants. As a further consequence, classification rules, generated by means of a CART analysis, are applied as a set of diagnostic rules and quadrants, classified by these rules, are accordingly marked within the image.
By these means, diagnostically highly significant regions in Melanoma images can automatically be emphasized. The results agree very well with the results of the purely visual evaluation. In this way, an evaluation "benign versus malignant", can be made accessible for an objective valuation - as opposing a subjective estimate of the diagnostician.
The development of brain death is a resulting patient status, which can be established through the use of modern resuscitation techniques, organ replacement function and function support on an equipmental and pharmacological basis. The confirmation of brain death provides the foundation for discussion on the termination of treatment, as well as being the prerequisite of an organ donation for the purpose of transplantation. The diagnosis of "brain death" presents a particular challenge to the medical obligation to provide care, especially outside of clinical centers, due to the small number of cases, the complexity of the diagnosis and the resulting clinical consequences. A corresponding curriculum and postgraduate education and training are therefore required.
In this context BRAINDEXweb, a teaching and learning system for the diagnosis of brain death, according to the Austrian "Brain Death Code" (Empfehlungen zur Durchführung der Hirntoddiagnostik bei einer geplanten Organentnahme, OSR Austria; 2013) has been developed in an interdisciplinary project between the Division of Special Anesthesiology, Pain Medicine and Intensive Care Medicine of the Department of Anesthesiology and Intensive Care Medicine and the IMI. The project was kindly supported by Novartis Pharma GmbH.
BRAINDEXweb was implemented as a web application, which guides the users in a dialog through all the diagnostic steps relevant for the determination of brain death. The user answers questions posed by BRAINDEXweb and enters their perceptions and measurements. The selection and order of the questions is controlled by a set of rules, which reflect the progression of the diagnosis according to the above recommendations for diagnosis of brain death (OSR Austria; 2013) procedure. In addition, features to help review, particularly in connection with spinal reflexes, have been implemented. A lexicon, consisting of some 200 entries, and a comprehensive list of references with more than 325 publications, rounds out the system. As a result, BRAINDEXweb creates a protocol in which details of everything executed, a review of the status of the patient in accordance with the recommendations for the diagnosis of brain death and explanations of this assessment are included. The protocol complies, to a great extent, with the OSR-recommendations; however, in some areas it exceeds this in its level of detail.
BRAINDEXweb is operated by the Medical University of Graz and is available under http://braindex.medunigraz.at/. The access is personalized; users are authenticated via user name and password. Registration is required to use the service. The system is available in German only.
One consequence of the increase in networking and the integration of medical information systems is that a patient medical documentation search sometimes produces an overwhelming long list of documents. The goal of the project COMET was to facilitate the selection, by the doctors at the Institute of Pathology, of relevant previous medical findings during the diagnostic process, by making additional information, extracted from the medical findings texts, available directly in the document list. For this purpose, the diagnosis texts of pathological findings are automatically analyzed and classified with regard to the question of "Text contains evidence of inflammation (-itis) or neoplasia", and then appropriately annotated (Tags).
With the software we have developed, a very good detection rate (F1-Measure of 0.98) can be achieved with regard to the classification of findings with "inflammation" and "neoplasia" characteristics. COMET has been integrated into the findings search engine AURAWeb, and has been in routine use at the Medical University of Graz since 2009. Histological findings are classified immediately after retrieval, so that no change is required in the information systems data bases. In each medical finding, the generated tags are shown directly in the resulting document list and offered as additional sorting criteria.
With the introduction of the Austrian Data Protection Act 2000, a set of changes and reforms concerning the use of data came into force. Thus
In order to provide support during the practical realization of this legal norm in research and education, in particular within the field of the human sciences, we compiled a Data Protection Policy [PDF, German only, 302 kB] on behalf of the Federal Ministry for Education, Science and Culture and the civic authorities of Vienna (Gemeinde Wien). A significant result, thereof, was the creation of a set of rules, which met the Data Protection Basic Regulations without causing a high increase in the time and effort required for bureaucracy within the scientific activities.
Text documents are fundamental parts of the electronic medical record. Current clinical information systems are not specifically designed to support quick and keyword-centred search and navigation within clinical text collections. This is further complicated by the variety of medical language with its many synonyms, spelling variations, broader and narrower terms.
EPA-Navi has the objective to develop a content-based (semantic) navigation in clinical text collections together with an industrial partner (Averbis GmbH) and to test it in a clinical context.
Knowledge of the macromolecular interfaces of protein structures plays a substantial role in understanding mutual reaction and biological functions. Examples of this are immune complexes (anti-bodies - antigen) and receptor factor compounds in clinical medicine.
Our software tool, which we developed in co-operation with the Jean Dausset Laboratory - Clinical Immunology (Professor G. P. Tilz), uses protein complexes from the PDB database, in order to determine the Interface Contact Matrix between 2 proteins. This is defined as a plot of paired interactions between individual amino acids of the two polypeptide chains. The amino acid chains are laid on two axles (horizontal and vertical) and an entry in the ICM is made in those places, where at least two of the atoms of both amino acids are within a certain distance. The entries are also annotated with physicochemical characteristics (such as hydrogen bonds, hydrophobic / hydrophilic characteristics etc.). The ICM thus generated is connected to the 3D-Visualisation of the macromolecular interface, so that, by means of mouse clicks in the appropriate place in the matrix, the appropriate 3D-Struktur is accentuated.
This technology enables the identification of Hot Spots on the contact surface between interacting proteins and common contact patterns at different complexes. The 3D-Visualisation also permits realistic views of the macromolecular interface structures and, by computation of the molecular surfaces of the amino acids, involved complementary surfaces can be made visible.
Within the framework of research projects, a service of the IMI that is frequently in demand is the retrieval of medical findings and the creation of reports from both the central hospital information system openMEDOCS and from the databases of different departmental systems.
We developed the *Med-Report Portal in order to provide scientists with reports and evaluations. The export assistant was designed by us to select and export master data and medical documents from openMEDOCS for further evaluation and analysis. Together, these two applications form a suitable solution to take advantage of the increased value for scientific work resulting from the documentation work. Thereby, special attention is paid to the quality and analyzability of the data, as well as retaining data protection.
PACSview and Cardiac@View are applications developed by us for the display of medical pictures.
PACSview is the current standard application in the Styrian Hospitals for the display of radiological pictures conforming to DICOM. It is integrated into the PACS infrastructure and thus available in 23 Styrian Hospitals.
Cardiac@View was developed during a Usability Study, in co-operation with the Department for Internal Medicine, Division of Cardiology and the Department for Cardiology and Intensive Care of the LKH Graz/West, and is characterized by large variety of presentable image formats; the possibility of producing and presenting video sequences and its availability on various prevalent system platforms. Cardiac@View can also be extended, very simply, to include image processing functionalities by integrating modules developed in IDL.
The Randomizer is a Web application, developed by the IMI, for randomization in clinical trials. The Application is designed for multi-center trials and provides - apart from different trial management functions - six different randomization methods and a simulator for testing study designs. The GCP-Compliance was affirmed by the Austrian Agency for Health and Food Safety (AGES). The software has been available since 2003 as a service of the IMI and is constantly being developed further. It has been successfully used in more than 250 clinical studies in Austria, Germany, Switzerland and the USA. Further information about the Randomizer, including a demo version, is available under https://www.randomizer.at/.
Unlike creating free text the collection of structured data in clinical documentation requires considerable efforts. Standardized terminologies and clinical information models have long been an important topic in medical informatics. Unfortunately, their use does not automatically improve semantic interoperability, as the same piece of clinical information can be represented by a variety of combinations of different terminologies and information structures.
SemanticHealthNet is a European Network of Excellence (http://www.semantichealthnet.eu/). Its aim is to build a sustainable organizational structure involving standards bodies, industry, government and research institutes, in order to improve the semantic interoperability of patient-related data within and across institutions and countries. The task of IMI - as a work package leader - is to develop and prototypically implement a conceptual framework in which clinical information with the same clinical meaning but represented by using different terminologies and information models can be detected and retrieved.