How to anonymize or de-identify DICOM studies with PACScenter?

PACScenter has been used for the most diverse scenarios. Despite the traditional production environments, there are also other scenarios that PACScenter has been used along the years, mainly to support research and work as a “PACS for research”. Despite the true advantages in place, namely extension capabilities of Dicoogle, the RESTful and real time APIs available in PACScenter, there is also a module to extract and de-identify DICOM studies. 

PACScenter de-identification module allows to configure profiles, with the following capabilities:

  • Anonymize the metadata according with a pre-defined profile (by default remove all tags)
  • Possibility to retain device, institutional, longitudinal temporal information
  • Possibility to retain original Unique Identifiers or generate random identifiers, such as PatientID
  • Possibility to anonymize Pixel Data (burned in annotations)

Confidentiality Profiles: 

The profile capability is available at the facility configuration according to DICOM Attribute Confidentiality Profiles (DICOM PS3.15 – Security and System Management Profiles). It enables defined settings that can be used for different research projects. For instance, there may be projects that we may need to retain device identifiers, while others may not require it. Moreover, for each DICOM tag it is possible to create defined strategies and rules: Retain, Replace, Remove, Randomize and Replace UID. While the retain will force to keep a specific DICOM tag value, the replace allows it to be replaced by a specific value. To make user life easier it is also possible to randomize these values or UIDs (these to keep Study-Series-SOPInstance relation. 

De Identification wrappers architecture (burned in annotations and others):

The PACScenter deidentification architecture allows it to be extensible and easily replaced by different modules to manipulate pixel data. While the main target of this module is for research projects we realize that this flexibility will allow us to keep interfaces open to anonymize further information in pixel data or other. Our current production use cases are based on AWS Rekognization and Comprehend Medical targeting burned in annotation, but current interfaces does not close doors to other adoptions and allows to be extensible by developing a single micro-service. 

If some of these scenarios catch your attention, contact us at

Stay tuned!



Comments are closed.