CTSC:ARRA.101309

Back to CTSC:ARRA supplement

Agenda

1. Review specific use cases to define the exact scope of the project (what is in and what is out of our bucket) and:
   1. articulate questions that require policy level discussion, review and decision (to be addressed in subsequent subgroup meetings of site leaders).
   2. specify technical requirements for what is "in the bucket" to be addressed in subsequent implementation planning meetings

Harvard Catalyst Medical Informatics group Meeting Minutes October 13, 2009

In attendance:

• Valerie Humblet
• Randy Gollub
• Yong Gao
• Shawn Murphy
• Diane Keogh
• Mike Mendis
• Steve Pieper
• Richard Robertson
• Kathy Andriole
• Alexander Zaitsev
• Charles Guttmann
• Dan Nigrin

Uses cases

Randy presented a couple of slides with different use cases. The goal of the meeting was to discuss the first ones and learn what are the mechanisms in place at each institutions.

Use case 1

The PI wants to use his own clinical images, with the clinical tools. He is not leaving his institution.

• At CHB and BWH, a PI can not have direct access to his data in PACS, he would need an IRB for research purpose. Some institutions have an easier IRB process in place for this type of case. We need to educate ourselves about the process at BIDMC and MGH.
• Tools that are in place for the PI to search his data.
  • At MGH you can use a physician's name to do a query in RPDR (The RPDR -Research Patient Data Registery- is a warehouse of clinical data from Massachusetts General Hospital (MGH), Brigham and Women's Hospital (BWH), Faulkner Hosptial (FH), Spaulding Rehabilitation Hosital (SRH), and Newton Wellesley Hospital (NWH) available for research purpose).
  • CHB and BIDMC are getting searching capabilities but they need to add radiology data.
  • At CHB, the search for radiology data is not automated, the PI must do it manually.
  • At CHB, surgeon have access to the PACS system and it is not reviewed by the Radiology Department. They search patient from their medical record number.
• Red flag rose by Rick Robertson from CHB:
  • The Radiology department already has problems with groups who access medical imaging data without asking the radiologists. It leads to different groups working on the same subject without collaborating with each others. He feels that by making the retrieval of clinical data easier, it will make the situation even worst.

Use case 2

Same as case 1 but the PI wants to use a research tool for post-processing. He needs to download the data.

• At CHB they can not do it without a review from Radiology. The request goes to the Informatics group. From the workstations, they can extract data from the PACS (as TIF for example).
• At BWH the PI needs to request a CD to be burned, he must have an IRB approved and Radiology has to give its approval. When the whole project is approved, the PI then gets DICOM access. He can use any approved DICOM viewer.
• In general, Osirix can connect directly to the PACS, it just needs to be configured the right way.
• Problem with heavy download: what happens if a PI decides to download 6,000 images?
  • As Shawn mentioned, we learned that at BWH and CHB, Pi's are able to get request through the PACS, it is a good thing but we have to make sure it does not bring down the system.
  • At CHB some PI's tried before and of course it crashed the system.
  • At BWH the workflow is controlled.

Discussion

• Shawn pointed that the group must be educated on how Radiology at the different institutions identify a specific image session. The accession number is an unique identifier but some image sessions might contain several accession numbers.
• Charles mentioned the query by imaging parameters. As Kathy noticed, it is not easily done with the commercial systems but the info are in the DICOM header. Right now building a new function to the PACS is out of the scope of the project. After a query (for example by patient ID and scan date), the data can be sent to XNAT where a more refined search can be done.
• Rick would like to see a mechanism created that would allow one to check if people are using the data for the same purpose. It will make the search project driven, one would be able to know who has access the data, when and for what approved IRB. The goal is to get an opportunity to raise a flag early in the process so people do not waste time on research projects already developed by others.
  • Shawn and Diane warned the group that this might lead to some groups having exclusive access to data. It is probably an issue that needs to be address at another level.
  • It comes down to who is the owner of the images (patient, hospital). The patient owns his data but at the same time one is required to store the images for 7 years where it was acquired. It is a field that we will have to investigate.

Some conclusions

• The goal of the mi2b2 project is to create an infrastructure that will make search easier but will have a layer of oversight and access control. To do so we want to build tools in parallel to what is already in place and respect each institutions way to do things so far.
• For the next meeting:
  • Get a definition of what a study is at each institution.
  • How can we get information about coding?
  • How can a PI query the PACS system at MGH and BIDMC?
  • What is the IRB process for case study 1 at the different institutions?