gms | German Medical Science

GMDS 2014: 59. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

07. - 10.09.2014, Göttingen

A web-based software architecture and framework for integration of an ophthalmologic reading center in large scale multi-center studies

Meeting Abstract

  • L. Clin - FH Reutlingen, Reutlingen
  • M.A. Leitritz - Universitäts-Augenklinik, Tübingen, Tübingen
  • J. Dietter - Universitäts-Augenklinik, Tübingen, Tübingen
  • C. Thies - FH Reutlingen, Reutlingen

GMDS 2014. 59. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS). Göttingen, 07.-10.09.2014. Düsseldorf: German Medical Science GMS Publishing House; 2014. DocAbstr. 358

doi: 10.3205/14gmds043, urn:nbn:de:0183-14gmds0434

Published: September 4, 2014

© 2014 Clin et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.



Introduction: In this work, a web-based software architecture and framework for management and diagnosis of large amounts of medical data in an ophthalmologic reading center is proposed [1]. Data management for multi-center studies requires merging of standing data and repeatedly gathered clinical evidence such as vital signs and raw data [2]. If ophthalmologic questions are involved the data acquisition is often provided by non-medical staff at the point of care or a study center, whereas the medical finding is mostly provided by an ophthalmologist in a specialized reading center. The study data such as participants, cohorts and measured values are administrated at a single data center for the entire study. Since a specialized reading center maintains several studies, the medical staff must learn the different data administration for the different data center. With respect to the increasing number and sizes of clinical studies, two aspects must be considered. At first, an efficient software framework is required to support the data management, processing and diagnosis by medical experts at the reading center. In the second place, this software needs a standardized user-interface that has not to be trained/taylored/adapted for each new study. Furthermore different aspects of quality and security controls have to be included [1], [3]. Therefore, the objective of this work is to establish a multi-purpose ophthalmologic reading center, which can be connected to different data centers via configurable data interfaces in order to treat various topics simultaneously.

Material and methods: Requirements are derived from a reference application in the context of the medical study “Nationale Kohorte”, in which the state of health of 200.000 participants (level 1) is to be comprehensively measured twice, with a time lag of five years, at 18 different study centers. The state of ophthalmologic health is collected for approximately 40,000 of the participants (level 2) and comprises the left and right ocular fundus, visual acuity, and standardized interview questions. The data center is located at the Deutsches Krebsforschungszentrum (DKFZ) in Heidelberg, whereas the reading center for eye examinations is established at the Ophthalmological Clinic of the University of Tübingen. There, each participant case is to be diagnosed independently by at least two investigators, with diagnosis conflicts to be resolved by principal investigators.

The architecture is based on the Model View Controler (MVC) pattern to provide the necessary flexibility and extensibility.

The data-model is realized in a relational database operating on a single server. It is a submodel of a typical Hospital Information System.

The controller part is separated in four components, which are implemented as Java packages within the Apache Tomcat webserver. At first, a persistency framework to access the data-model is required to avoid that modifications of the data-model entail the need for reimplementation of the actual business logic. In the second place, a thread-based service architecture is realized that allows for scheduled processes to send and receive messages and to trigger asynchronous computations such as image processing modules. Here, shares as well as sockets and e-mail are supported for message passing. By using the factory and functor pattern a general and specialized set of process parameters is configurable. The developer must respect a set of general parameters such as repetition times and connection. Specific parameters are evaluated by reflection. The third component is the message generator itself which maps the data from the database into the specific data format that is accepted by the study center. The mapping itself is freely configurable for instance to the format provided by the DKFZ for the “Nationale Kohorte”. The fourth component is the set of functions the web-server provides for the actual user interface. These are realized as web services via http and allow access to the data for processing by the medical experts.

The view component/graphical user interface (GUI) is realized as a fat client in the Qt/C++ framework to allow for user interaction with the images, such as point and click labeling of particular elements to improve or correct image-processing results. Such functionality is hard to implement using browser-based technologies such as Javascript. The GUI is being designed for an efficient workflow in the central tasks of the two user modes:

investigator mode (diagnosis):
– retrieval of participant file, display of available images
– user selection of images of sufficient quality for diagnosis – diagnosis according to specific encoding catalog
principal investigator mode (supervision):
– overview of diverging diagnoses
– step by step conflict resolution, with optional full inspection of concerned diagnoses

The GUI is adaptable to varying study designs and questions by using dynamic loading of catalogs for coding of diagnostic findings and results.

Results: The proposed architecture and framework meets all requirements of the ophthalmologic reading center. It supports the connection of the reading center to any existing data center of a clinical study. By using a general data interface, any type of digital information can be im- or exported. Even files can be used if the data center of the study does not provide web interconnectivity. The service architecture allows for background processes to fetch new data that has to be analyzed and to send back/return the results without user-interaction. Only the protocol has to be configured initially. The GUI is standardized with respect to workflow and usability, so that medical experts are not required to learn different tools to contribute to different studies.

Discussion: Medical information systems are based on typical components which are well reflected by the MVC architecture. However, the actual implementation must consider aspects such as hypothetically unlimited time of operation, application specific maintenance and continuous extension with new functions and features. These challenges of software design require to take into account both medical needs and modern development approaches.


Baltasar Sánchez A, González-Sistal A. Design of a Web-tool for diagnostic clinical trials handling medical imaging research. J Digit Imaging. 2011 Apr;24(2):196-202.
Dziuballe P, Forster C, Breil B, Thiemann V, Fritz F, Lechtenbörger J, Vossen G, Dugas M. The single source architecture x4T to connect medical documentation and clinical research. Stud Health Technol Inform. 2011;169:902-6.
Babamir SM, Arabfard M. Improving service accessibility in service-oriented HIS. J Med Syst. 2012;36(6):4021-30.