Stanford
STAnford medicine Research data Repository

Stanford Medicine Research Data Repository (STARR) is integrated with hospital systems that store radiology, cardiology and retinal fundus imaging DICOM data.

Imaging

About imaging data

Radiology DICOMs: Radiology data from both hospitals are stored in a common Sectra PACS. The Sectra PACS replaced the shared GE PACS. All data from GE were migrated to Sectra.  Data in Sectra PACS is replicated in the Vendor Neutral Archive (VNA). For the last several years, Research Technology has built up the radiology data lake by query/retrieve (Q/R) from GE and Sectra PACS. STARR data lake now has historical radiology data, since 2011.  Earlier in 2020, we achieved a complete integration with VNA and now new radiology data collected daily is pushed nightly to STARR from the VNA.

Cardiology DICOMs and Syngo Database: The two hospital systems have two distinct Cardiology PACS systems, children's hospital Syngo application and adult hospital Syngo application. The DICOM data stored in Syngo are not available in the VNA. STARR has query/retrieve (Q/R) access to these two cardiology PACS. At this time, the cardiology imaging data is procured when there is an active consultation request from the researcher. 

Fundus DICOMs: SHC uses Zeiss application in ENT for retinal fundus imaging. Some of this data is in VNA and is accessible via Query/Retrieve. A subset of the retinal imaging devices don't have integration with VNA at the time of writing. At this time, the retinal imaging data is procured when there is an active consultation request from the researcher. 

STARR imaging ingestion

At Stanford, the radiology PACS is Sectra (previously GE), the Cardiology PACS is Syngo, the ENT mini-PACS is Zeiss and the VNA is Fuji. Bulk of the VNA data is Radiology. While X-rays form bulk of the studies (accession IDs), the bulk of the images are CT or MRI. Adhoc pull implies query/retrieve and is launched when researchers request data. STARR also receives nightly data from the VNA, the new daily data is batched up and sent to STARR every night. 

Access to imaging data

Researchers are encouraged to try out the self-service DICOM download via STARR Tools. Following  services are  available via STARR Tools with an approved eProtocol:

  1. PHI scrubbed DICOMs from radiology imaging data (X-rays, CTs, MRIs, Ultra sounds, ...) from both hospitals. At a time the researchers can download DICOMs for a few accession numbers. There is no limit to the number of downloads.
    • The DICOM metadata is scrubbed for PHI using Safe Harbor practices. Within the IRB, the patient's event dates are consistently jittered to preserve their timeline.
    • The DICOM pixels are scrubbed for PHI using a highly sophisticated ruleset.
    • Only the primary images are processed in the download. Derived images and non-images are redacted from the download.
    • If you wish to share these PHI scrubbed images with non-Stanford entities or need an expert determination for DICOM de-identification, please make sure that you have completed a Data Risk Assessment.
  2. Cohort identification using STARR Tools
  3. Linked EHR data from Stanford data model (STARR Tools data model).

If the self-service is not sufficient for the purpose, researchers are encouraged to submit a consultation request. Following services are available via consultation request:

  1. DICOMs from following systems
    • Large scale multi-modal radiology imaging data (X-rays, CTs, MRIs, Ultra sounds, ...) from both hospitals
    • Echocardiograms from both hospitals
    • Retinal images from the adult hospital
  2. Identified or de-identified DICOMs (using accession_ids or MRNs if available)
    • A human subject data (either identified or de-identified) is accessible with an approved IRB.
    • A limited data set is accessible with an approved eProtocol.
  3. Cohort identification using STARR Tools or OMOP or Montage (SHC Radiology only).
  4. Linked EHR data from Common Data Models (e.g. OMOP, PEDSNet or PCORNet) or Stanford data model (STARR Tools data model).
  5. If a radiology study needs access to Quality and Research (QR) PACS, it is possible to request delivery to QR PACS. Our general practice is to deliver the data to Box (small number of studies) or Nero (large number of studies).
  6. Content from LPCH Cardiology PACS Syngo database. The database contains labels and annotations such as left and right ventricle heart rate, ejection fraction, diastolic area etc.

DICOM Safe Harbor PHI removal

Research Technology has developed a highly sophisticated petascale DICOM PHI scrubbing pipeline that leverages and extends the MIRC CTP.  The pipeline uses a deterministic approach to remove PHI from DICOM metadata and pixels. At this time, the UPO determines that the PHI scrubbing pipeline produces High Risk confidential data.  

Learn more about our methods.