Findings, CDEs, and Observations

Radiology findings can be represented in a standard format based on FHIR Observations semantically labeled with ACR/RSNA Common Data Element identifiers.

The core of the ORDM model is the Observation structure, which represents a single observation or finding. An Observation represents an imaging finding (broadly construed; for examples, see the gamuts.net) such as “the patient has a left-sided pleural effusion”. Each Observation is labeled with a ACR/RSNA Common Data Element (CDE) Set ID to indicate what kind of finding it is, and the value for each attribute of the finding is labeled with a CDE Element ID. For example, the finding...

“solid nodule measuring 6 mm in the right lower lobe”

...is labeled with the CDE Set ID RDES195, which identifies the observation as a “Pulmonary Nodule” and provides a standard tags for the nodule attributes, such as “6 mm”, “solid” and “right lower lobe”.

CDE Set as a data model for an Observation
Schematic of a finding encoded as FHIR Observation referencing a CDE Set.

The formal FHIR Observation object could look like:

{
  "resourceType": "Observation",
  // Cross-system ID for this nodule in the current report
  "id": "1d9a6d1796fc",
  // CDE Set ID specifies what kind of observation this is
  "code": {
    "system": "https://radelement.org",
    "code": "RDES195",
    "display": "Pulmonary Nodule"
  },
  // Standard anatomy; see https://anatomiclocations.org
  "bodySite": {
    "code": {
      "coding": [
        {
          "system": "https://anatomiclocations.org/",
          "code": "RID1315",
          "display": "lower lobe of right lung"
        }
      ]
    }
  },
  // Attributes are represented as components
  "component": [
    // Each component consists of a CDE Element code and a value
    {
      // Presence
      "code": {
        "coding": [
          {
            "system": "https://radelement.org",
            "code": "RDE1300",
            "display": "Presence"
          }
        ]
      },
      "valueCodeableConcept": {
        "coding": [
          {
            "system": "https://radelement.org",
            "code": "RDE1300.0",
            "display": "present"
          }
        ]
      }
    },
    // Composition
    {
      "code": {
        "coding": [
          {
            "system": "https://radelement.org",
            "code": "RDE1301",
            "display": "Composition"
          }
        ]
      },
      "valueCodeableConcept": {
        "coding": [
          {
            "system": "https://radelement.org",
            "code": "RDE1301.0",
            "display": "solid"
          }
        ]
      }
    },
    // Size
    {
      "code": {
        "coding": [
          {
            "system": "https://radelement.org",
            "code": "RDE1302",
            "display": "Size"
          }
        ]
      },
      "valueString": "6"
    },
    // Location 
    {
      "code": {
        "coding": [
          {
            "system": "https://radelement.org",
            "code": "RDE1304",
            "display": "Location"
          }
        ]
      },
      "valueCodeableConcept": {
        "coding": [
          {
            "system": "https://radelement.org",
            "code": "RDE1304.9",
            "display": "right lower lobe"
          }
        ]
      }
    }
    // ... Other properties of the nodule would go here
  ]
}

The deep technical work on this concept was led by Brian Bialecki of ACR; Ali Tejani is leading an effort to write up a white paper for publication soon!