gms | German Medical Science

Introduction: Clinical data warehouses (DWHs) may be used to provide researchers with a structured and simple access to pseudonymized clinical routine data in a homogenized format. Common data domains of DWHs are patient demographics, diagnosis codes, laboratory values, procedures, disease-related groups (DRGs), and other domains available in structured format or narrative texts. Images from picture archiving and communication systems (PACS), however, are usually only available from selected patients in dedicated Imaging warehouses [1], [2] or research image archives and PACS, as e.g. dcm4chee [3] and XNAT [4].

A specific task for a currently running study at our institution required to retrieve X-Ray images of 592 patient cases from the productive PACS in a pseudonymized format. It was unknown whether such a task may be executed (semi-)automatically.

Methods: We developed an application as middleware between our clinical DWH, an identity management system, and the productive PACS. The system provides a REST-style interface to our DWH that takes requests containing the pseudonyms used within our DWH. Towards the PACS it provides a standard DICOM query/retrieve interface. Both interfaces are internally connected via a module that pseudonymizes/de-pseudonymizes DWH identifiers to and from PACS identifiers. DICOM images are anonymized by erasing all respective headers.

Results: We utilized the prototype system without integration into the graphic use interface (GUI) of our DWH. The system provides a simple GUI, but can also be used directly via the REST-style interface. In order to extract the images for the study introduced above, we called the system with the following: /pacs?task=image&desttype=dir&skey=&idtype=patient& modality=cr,dx&id=. X-ray images may include the modalities Computed Radiography (CR) and Digital Radiography (DX). The identifier list is a comma-separated list of a patient pseudonym (used within the DWH), a case start/end date, and a replacement string with a study pseudonym in the format “^^”. The system required 40 min to query all 592 patient cases with the respective dates and to store 1,549 single images (an X-ray series may include multiple images) with the study-pseudonym as filename to a pre-specified directory. The data was directly stored with anonymized headers and pseudonymized file names.

Discussion: We presented the feasibility of a system that connects a productive PACS with a pseudonymized DWH to extract pseudonymized images for a single study use case. The approach performed well in its prototype status. However, prior to a productive deployment with our DWH more work on preventing a misuse of the interface is necessary. In addition, the integration of the REST-style interface with the GUI of the DWH is necessary.

The authors declare that they have no competing interests.

The authors declare that a positive ethics committee vote has been obtained.