• Where to put the pictures?
  • Sven Lüttmann - VISUS
  • Where to put the pictures?
  • Sven Lüttmann - VISUS

When it comes to the structure, use, and success of national records such as the ePA in Germany, ELGA in Austria, or the EPD in Switzerland, one topic is generally given a wide berth: the handling of image data. In particular because of their immense data volume, they do not fit very well into the structure of the individual solutions. This article addresses three approaches to solutions for integrating image data in patient records which have, to some extent, already been tested in practice. The basis for all three is the IHE-XDS-I profile.

If medical documents or also other data are first consolidated and their metadata standardized, they can be handled very easily in communication with patient records. The individual documents are not very large; they are generally self-contained as an information unit and have titles which are more or less unambiguous. Necessary attributes can additionally be derived using smart algorithms. Let’s take the discharge summary as an example: Both the patient and the doctor providing further care know that they will find information in this document about the diagnosis, treatment and medication, and possibly relevant excerpts from image examinations. In the form of a PDF, the file size will likely be manageable and clearly in the kilo- to megabyte range which can be easily transported and saved.

Making documents out of images

In the case of radiological image data, the situation is unfortunately quite different. Here it is possible to quickly reach several hundred megabytes, and in some instances even gigabytes of data which form a bundle of several hundred individual files, such as in the case of thin-layer CT studies. Document-centered file systems based on IHE XDS would be overwhelmed in the communication of these data, since each image would have to be configured as a separate document. Thus a patient’s record becomes extremely overloaded and unwieldy.

The solution for all three approaches is to use what is known as an imaging document source (IDS). In IHE-XDS-I structures, such an IDS is the counterpart to the document repository from the IHE-XDS profile which keeps the medical documents available for retrieval. While the repository handles the documents, the IDS is responsible exclusively for the storage of DICOM image data. The data also remain here and are not individually transferred to the document repository. Rather – and this is the highlight – a so-called key object selection (KOS) document or also a DICOM manifest is created in which all references to the image data are stored via unique IDs. This document, which also includes the necessary patient data and a reference to an image source (IDS), is then transferred to the repository and registered in the registry as only an entry (= document).

Decentralized, privately centralized, or entirely centralized

The three approaches which VISUS is actively working on differ in where the IDS is located, who is responsible for it, and who takes over the creation of the KOS documents. The first possibility is decentralized storage of the data, thus each institution establishes its own IDS which is maintained and operated by its own IT department. The use of an existing PACS which supports the IHE-XDS-I profile is also possible, but because of data privacy and data security, this approach is not considered further.

It may, however, be entirely practical for large institutions with a corresponding IT infrastructure and human resources to manage image data themselves. Nonetheless, the capacities needed should not be underestimated in any case. Memory space as well as service and operation of the IDS require money and time and must be permanently guaranteed.

For this reason, it may make sense for the hospitals to outsource IDS as a SaaS (Software-as-a-Service) concept – VISUS is currently taking this approach in a prototypical manner in Switzerland. This means that hospitals send the image data to a central computing center in which VISUS operates scalable products which permanently ensure memory space management and the management and provision of data. VISUS acknowledges the receipt of data, takes charge of the creation and filing of the reference document in the XDS repository, ensures retrieval, and keeps the image data until the patient deletes them from his/her record.

Finland is taking an even more centralized approach. Here it is not private providers which offer the IDS service. Rather there is a national image source in which the image data are required to be included. From here, the reference documents are created for all images nationally and stored in the national XDS repository.

For the ePA, which is required to be running on the telematics infrastructure starting on January 1, 2021, there is still no solution for the handling of images. However, the three examples show that there is no lack of technical possibilities and solutions can also be found for Germany, with its high standards regarding data privacy and security.