ImagingSelection — Subset of DICOM instances
A new resource introduced in FHIR R5. It identifies a precise subset of a DICOM study — series, instances, frames, geometric regions. The bridge between the global ImagingStudy and the radiologic content relevant to a clinical report.
Purpose of the resource
A DICOM study may contain thousands of images. For a radiology report, the radiologist
points to only a small subset that justifies the interpretation: 3 axial slices showing a
suspicious lesion, two frames of a coronary angiography showing a stenosis.
ImagingSelection is the container that unambiguously identifies that subset.
Before R5, FHIR used ImagingManifest (deprecated in R4 and removed). The new
resource is explicitly aligned with DICOM PS3.4 KOS (Key Object Selection) and with TID
1500 Structured Reports, which allows direct convertibility between the two worlds.
Key fields
| Field | Type | Cardinality | Role |
|---|---|---|---|
identifier | Identifier[] | 0..* | External identifiers for the selection. |
status | code | 1..1 | State: available, entered-in-error, unknown. |
subject | Reference(Patient | Group | Device | Location) | 0..1 | Patient or group the selection refers to. |
issued | dateTime | 0..1 | Date the selection was issued (= report date). |
performer | BackboneElement[] | 0..* | Author(s) of the selection with their role (radiologist, technologist…). |
basedOn | Reference(CarePlan | ImagingStudy | ServiceRequest)[] | 0..* | Originating request. |
category | CodeableConcept[] | 0..* | Classification (key image, teaching file…). |
code | CodeableConcept | 0..1 | Selection type (DICOM TID 1500 concept names). |
studyUid | id | 0..1 | DICOM UID of the source study. |
derivedFrom | Reference(ImagingStudy | DocumentReference)[] | 0..* | Originating ImagingStudy or DICOM KOS. |
endpoint | Reference(Endpoint)[] | 0..* | WADO-RS, WADO-URI, QIDO-RS from which to fetch pixels. |
seriesUid | id | 0..1 | DICOM UID of the selected series. |
seriesNumber | unsignedInt | 0..1 | DICOM series number. |
frameOfReferenceUid | id | 0..1 | DICOM Frame of Reference (shared 3D space). |
bodySite | BackboneElement | 0..1 | Anatomic site (SNOMED CT or DICOM CID). |
focus | Reference(ImagingSelection)[] | 0..* | Parent or sibling selections (e.g. before/after comparison). |
instance | BackboneElement[] | 0..* | List of selected DICOM instances with UID, SOP class, frames, regions. |
instance.imageRegion2D | BackboneElement[] | 0..* | 2D region inside the image (point, polyline, rectangle, ellipse). |
instance.imageRegion3D | BackboneElement[] | 0..* | 3D region inside the frameOfReference (point, sphere, ellipsoid). |
JSON example
A radiologist marks a roughly 4×4 cm lesion on axial slice 42 of an abdominal
CT series. The selection references the study by its DICOM studyUid, points
at the WADO-RS server hosting the pixels, and carries an imageRegion2D of
polyline type.
{
"resourceType": "ImagingSelection",
"id": "example-keyimage",
"status": "available",
"subject": { "reference": "Patient/example" },
"issued": "2026-05-16T10:25:30+02:00",
"performer": [{
"function": {
"coding": [{
"system": "http://terminology.hl7.org/CodeSystem/v2-0912",
"code": "REF",
"display": "Referring"
}]
},
"actor": { "reference": "Practitioner/dr-jones" }
}],
"studyUid": "1.2.840.113619.2.176.2025.1499492.7040.1171286242.109",
"derivedFrom": [{ "reference": "ImagingStudy/example" }],
"endpoint": [{ "reference": "Endpoint/dicom-wado-rs" }],
"seriesUid": "1.2.840.113619.2.176.2025.1499492.7040.1171286242.111",
"seriesNumber": 3,
"frameOfReferenceUid": "1.2.840.113619.2.176.2025.1499492.7040.1171286242.219",
"bodySite": {
"concept": {
"coding": [{
"system": "http://snomed.info/sct",
"code": "12921003",
"display": "Anatomic site of pelvis"
}]
}
},
"instance": [{
"uid": "1.2.840.113619.2.176.2025.1499492.7040.1171286242.345",
"number": 42,
"sopClass": {
"system": "urn:ietf:rfc:3986",
"code": "urn:oid:1.2.840.10008.5.1.4.1.1.2"
},
"subset": ["1", "2"],
"imageRegion2D": [{
"regionType": "polyline",
"coordinate": [100.0, 110.0, 130.0, 110.0, 130.0, 150.0, 100.0, 110.0]
}]
}]
} Typical use cases
- Radiology report: each assertion of a DICOM TID 1500 SR can be converted into an
ImagingSelectionattached to anObservation. - Key images: flag the key images of a study for another clinician or an MDT (Multi-Disciplinary Team).
- Teaching files: case libraries with annotations.
- Annotations: lesion contour, 2D/3D ROI for longitudinal follow-up (RECIST).
- Radiation therapy: identify target structures inside a dosimetric plan.
- AI / CADx: traceability of the region analysed by an algorithm and its output.
Common pitfalls
- Duplicate DICOM UIDs:
studyUid,seriesUidandinstance.uidmust exactly match those of the associated ImagingStudy. Any mismatch breaks the WADO-RS resolution. - Selection without
endpoint: the consumer will not know where to fetch the pixels. Always provide at least an Endpoint typeddicom-wado-rsor equivalent. - Misaligned 2D coordinates:
imageRegion2D.coordinateis expressed in the pixel space of the instance. For physical 3D coordinates, useimageRegion3Dtogether withframeOfReferenceUid. - Confusion with
ImagingStudy: ImagingStudy models the whole study; ImagingSelection only references a subset. One does not replace the other. - Missing SOP class:
instance.sopClassis required so a consumer can tell what kind of DICOM object is referenced (CT, MR, US, segmentation, structured report…).
Related resources
- ImagingStudy — the global DICOM study referenced via
derivedFrom. - DiagnosticReport — the radiology report that may embed one or more ImagingSelection in
media. - Observation — the structured clinical assertion carried by the selection (lesion measurement, BI-RADS, TI-RADS…).
- Endpoint — the WADO-RS / QIDO-RS / STOW-RS service endpoint.
- BodyStructure — anatomy pointed to by the selection.
See also: ImagingStudy — the global DICOM study and DiagnosticReport — radiology report.