Cross walk

  • In imaging the ID is embedded in the data, it is difficult to identify
  • Need to store the mappings for deidented radiology images
  • Primarily maintain a master list
    • sometimes has an accession number
  • Have requests from researchers that are using research scans but want to join to clinical data
  • Need to interface programmatically
  • Needs to be performant as it would be part of their workflow
  • How does the crosswalk get updated
    • radiology pushes to us when the labels get generated each night
    • we will need identified info to do validation/matching on I2DB end
      • need to have a needs review bucket for conflicts or no matches
      • can handle issues with patient IDs being changed/corrected
    • we can then store the output in the radiology schema
  • When processing the data
    • They will need to quickly query the crosswalk

Flywheel

  • session level details
    • metadata
    • series level details
      • has metadata with each entry
      • maps to session level
  • what format is the data
    • comes from open search, probably JSON
      • can be accessed via a REST API
    • other data is in psql
    • map the series to the accession number

Follow Ups

  • Resend the BYOB? forms to get Jenny's access setup
  • Snehil work on mapping discovery
  • Chris & Ian do discovery on API
  • Attend follow up meeting
  • How is it funded?
    • Does radiology need to pay for the data storage and ingestion?
    • Do they want to pay for Jenny to be a BYOB?

Table of Contents


Updated on August 7, 2025