From NAMIC Wiki
Jump to: navigation, search
Home < CTSC:ARRA.011910

Back to CTSC:ARRA supplement


i2b2 to XNAT to DICOM architecture and workflow

Harvard Catalyst Medical Informatics group Meeting Minutes January 19, 2010

In attendance:

  • Valerie Humblet
  • Mike Mendis
  • Shawn Murphy
  • Bill Tellier
  • Mark Anderson
  • Randy Gollub
  • Yong Gao
  • Paul Lamonica
  • Steve Pieper
  • Wendy Plesniak
  • Alex Zeitsev
  • Charles McGow
  • Paul Lamonica
  • Gordon Harris
  • Dan Marcus
  • Charles Guttmann
  • Bruce Rosen
  • Ron Kikinis
  • Diane Keough
  • Jesse Wei
  • Bill Hanlon

  • Summary of last week meeting:
    • Our goal is to create a new cell for i2b2 to allow the PACS of each institution to communicate with i2b2 and XNAT.
    • This new application is called the STAR-D software.

  • i2b2 to XNAT to DICOM:
    • i2b2 can generate a list of patients who have images that a PI wants to acquire from the PACS (and then send to XNAT)
    • The jobs of the STAR-D software are:
      • understand how to communicate using DICOM messages.
      • understand the restrictions to the PACS so it does not bring down the PACS.
      • to satisfy Radiology department apprehensions about a software getting access to their PACS.
      • make sure only authorized people can get access to the data.

  • Workflow:
    • How do we get images out?
      • From an accession number (also known by others systems of hospitals), one can get access to a study.
      • When we don't have the accession number but we know that there is a patient with images that are interested. The STAR software will interact with the PACS to get the list of images, then i2b2 will communicate with STAR.
    • Dan mentioned that one challenge will be to implement STAR-D so that there is portability to the different institutions (security etc vary by institution)
    • dmc4chee:
      • The first goal is to try to get the STAR-D software to work through the PACS-dcm4chee model: DICOM-PACS query-retrieve system but without any sensitive patient data.
      • Having DICOM data and clinical data together will be good as a test.
      • RSNA has a couple of data sets with different modalities and vendors, so does ACRIN. There are also DICOM data on Osirix. Steve already tried to send them form Osirix to dcm4chee but the system crashed, it needs to be repeated.
      • In i2b2 there is a fake dataset, we could match images with it for example by taking good data sets and replicate them. Dan mentioned a useful software: scriptable DICOM browser to replicate studies.
      • Another possibility would be to use phantoms from the different modalities at the different institutions, that we we avoid any HIPAA problem. Each institution has phantoms or other QC data stored in the PACS (ex: BIDMC has frogs pictures).

  • Actions for next week meeting:
    • Get a couple of phantoms and push them to dmc4che (through Steve or by yourselves)
    • Learning how the users of i2b2 communicate with the STAR software
    • User's permission structure in i2b2, STAR-D and XNAT
    • Reporting use to the Radiology departments
    • Throttling volume (model of archival system of PACS)
    • What kind of data are useful (ex: what is good enough to be used with Freesurfer)