From NAMIC Wiki
Jump to: navigation, search
Home < CTSC:ARRA.010510

Back to CTSC:ARRA supplement


  1. DICOM query retrieve tool

Harvard Catalyst Medical Informatics group Meeting Minutes January 5, 2010

In attendance:

  • Valerie Humblet
  • Mike Mendis
  • Shawn Murphy
  • Bill Tellier
  • Mark Anderson
  • Bill Hanlon
  • Todd Perlstein
  • Jesse Wei

  • The goals of the meetings were:
    • to review where we are in the mi2b2 architecture
    • to go through the workflow
    • to get a couple of assignments and output samples

  • Architecture:
    • i2b2: PI makes a query, requests images with medical record number (MRN) or accession number. We need to define what is a study at the different institutions. The MRN will probably not be enough, one will need to add the date or the modality etc.
    • PACS: Images are extracted and sent to XNAT.
    • XNAT: new derived data sent back to i2b2.

  • Accession number: if we send an accession number (acc.#), will we get a study or an image?
    • BWH and MGH: it leads to a study. For billing purposes, there could be multiple different acc.#s for the same study (example: if one orders an MRI/MRA, 2 acc #s are generated but they lead to the same study).
    • CHB: acc.# leads to a study.
    • BIDMC: one acc.# per patient encounter.
    • At each site, one is able to send a C-find to the PACS with an acc.#.

  • UID:
    • In theory, no 2 images will ever have the same UID.
    • Each manufacturer is assigned numbers.
    • UID is created by the modality.
    • You don't need to know it to get data from the PACS, PACS works with acc.#
    • UID is for internal accounting and unique identification. If you derived data, you need another UID.
    • XNAT organizes images by UID.

  • Navigation client:
    • When requesting images, once the transfer is specified, the images will be sent to the target location. XNAT will be set up to receive images.
    • We need a queue mechanism on the model we implement to lower the stress on the PACS (transfer at night, very slowly etc).
    • dcm4chee:it will be used to simulate a PACS. Its biggest advantage is that we can use it outside the firewall with the fake i2b2 instance also outside the firewall. But we have to make sure that it behaves like a PACS. We don't want to invest to much time working with it turns out to be very different.
    • Jesse mentioned that the clinical workstations also support query/retrieve functions; raising the possibility that we could also use one instead of a PACS.

  • Action items:
    • Understand the output of a general request for each site.
    • Take a look at the navigation clients made to identify a study or an image.
    • Design the client that will request the images. Look at dcm4chee and the clients at the different institutions.