CTSC:ARRA.072710

Back to CTSC:ARRA supplement

Agenda

• Update and demo on Phase 2 mi2b2 software
• uxp and UI design: workflow analysis
• Miscellaneous announcements: CHB i2b2 update, ACRIN presentation, etc

Harvard Catalyst Medical Informatics group Meeting Minutes July 27, 2010

In attendance:

• Bill Wang
• Wendy Plesniak (not in attendance but sent updated slides to use for uxp and UI design discussion
• Valerie Humblet
• Chris Herrick
• Charles McGow
• Darren Sack
• Shawn Murphy
• Cynthia Lee
• Steve Pieper
• Randy Gollub
• Bill Tellier
• Jesse Wei
• Kathy Andriole (phone)
• Paul Lamonica (phone)

Meeting Minutes

mi2b2 software

• Chris and Darren installed the Phase 2 software at MGH last week. Overall, the install was a success. They were able to query up the studies on a test patient, drill down to the series level on a study, and then the image level on a series. They were also able to click on an individual image and return the image for viewing.

They did find a couple of hiccups along the way:

1. In general, image retrieval performance was very slow. It would often take 1-2 minutes before the image would show in the workbench. It's hard to tell what exactly might be going on. At times, we could see the CMOVE request get initiated, but the image wouldn't start downloading until 1-2 minutes later. We're wondering if this may have to do with some kind of assigned priority on the PACS side (our application is getting bumped to the end of the list). Additionally, there were other times when the image would download in its entirety, but Star-D wouldn't get the response/indication that the image was fully downloaded until several seconds (up to a minute) later. Darren confirmed that this was not likely to be a node, PACS or network performance issue or priority level issue since it was not busy in terms of system activity and simultaneously he was able to pull images out of deep archive with only seconds delay, quick for anytime, at that same node.
2. They ran into problems displaying some Ultrasound images. This could be due to the compression type, but unclear at this time.
3. All the images appeared reversed and upside down. They will have to look into options for displaying correctly - this may be some kind of configuration that will have to be built into the workbench.

• Further discussion regarding Phase 2 performance issues focused on whether images were compressed before being transferred. Each site, MGH, CHB, BWH and BIDMC all confirmed that their PACS supports full (non-compressed) DICOM transfers. We will write up our own DICOM conformance statement for mi2b2 that will explicitly specify that our software requires all PACS interfaces to support full DICOM (uncompressed) transfer.
• Long call with Mike D'Arcy (BIRN-CC collaborator) to discuss software development for Phase 3 that requires robust image cache management. He is using PixelMed not DMC4Chee (Mike is less familiar with DMC4Chee) and uses PixelMed exclusively for this work. He is going to send some software tools for caching images he has made with PixelMed for us to learn from. GRIDftp has been developed in an integrated way with PixelMed so that their group is invested in that technology. DMC4CHee is more robust and hence complex, but does have a fairly large user/developer community. PixelMed is a smaller community but importantly has a key developer, Dave Cluney, who is responsible and active in his support of bug fix requests and is high profile in the DICOM community. Dave Cluney recently wrote a ~700 page book on DICOM structured reporting. DICOM structured reporting is just starting to be used in our medical centers. Agree that the field is moving this way. BWH for instance is using it for fetal ultrasound, breast imaging and few other modalities. This is different from annotation (image mark-up). Group agreed to come back to this issue in the future.
• Also discussed what viewer is being used (ImageJ) and how that might be responsible for the display problem. Kathy suggested that there is a single field that can be used (Yes/No toggle) to invert the display. She is going to try to find the info and send to Chris.

Workflow analysis

• Discussed the next new revision based on input from last week. Focused on the details of Image Request Queue. Homework for all, think about what fields you want to be able to filter by and info fields that are needed for verifying image contents (E.g. modality, study type/name, date

Miscellaneous announcements: CHB i2b2 update, ACRIN presentation, etc

1. Charles McGow has i2b2 up and running at CHB. Dan Nigrin demonstrated a query in it for Randy and Valerie last week. Attempting to query for patients with CNS concussion yielded n = thousands, when AND Procedures>Radiology>MRI>CNS the yield dropped to 0. Randy will follow up with Charles on this.
2. Chris Herrick received the official invitation to present our mi2b2 project at ACRIN along with an offer to cover travel expenses.
3. Shawn and Randy will present to Catalyst COI August 24th instead of July 27 due to the Committee members being away.