DICOM C-FIND vs DICOMweb QIDO-RS: Querying for Studies

  • C-FIND uses a stateful binary TCP association, while QIDO-RS uses stateless HTTP GET requests returning JSON.
  • Both protocols return metadata only, such as Patient IDs and Study Dates, never pixel data.
  • QIDO-RS supports standard limit and offset pagination, while classic C-FIND has no standardized pagination.
  • Retrieving actual image instances requires a separate operation: C-MOVE in DIMSE or WADO-RS in DICOMweb.
  • Hospital firewalls typically block inbound queries, so cloud pipelines usually rely on hospitals pushing studies instead.

Both C-FIND and QIDO-RS search for medical imaging metadata across PACS archives and DICOM servers. C-FIND uses a stateful binary session over TCP, while QIDO-RS uses standard HTTP GET requests returning JSON. Neither protocol returns image pixels. They return metadata only, such as Patient IDs, Study Dates, and Modality types.

How do DICOM C-FIND and QIDO-RS query for studies?

In healthcare software, systems often locate studies before taking clinical action. For example, a workstation searches for prior exams for a patient. Classic DICOM uses the C-FIND service element from the DICOM Message Service Element (DIMSE) protocol family. Modern DICOMweb systems use QIDO-RS (Query based on ID for DICOM Objects by RESTful Services).

Both protocols query standard hierarchical levels: Patient, Study, Series, and Instance. Both filter on specific attributes, such as Study Date or Patient ID. Yet their network models, query formats, and response payloads differ completely.

C-FIND: the classic DIMSE query protocol

C-FIND is part of the original DICOM network standard defined in PS3.4 and PS3.7. It does not operate as an isolated request. Instead, it requires a negotiated TCP session called an association.

To run a C-FIND query, your client acts as a Service Class User (SCU). Your client connects to the archive, which acts as a Service Class Provider (SCP). First, both systems negotiate the association using their Application Entity (AE) titles. They agree on the Query/Retrieve Information Model SOP Class, such as Study Root Query/Retrieve Information Model - FIND.

Next, the client sends a C-FIND-RQ message with an identifier dataset. This dataset holds two kinds of attributes. Matching keys contain values you filter on, such as PatientID = "12345". Return keys contain empty values, signaling which attributes the server must return.

The server evaluates the query against its archive index. For each match, the server returns a separate C-FIND-RSP message with status 0xFF00 (Pending) containing the requested tags. Once all matches are sent, the server sends a final C-FIND-RSP with status 0x0000 (Success). If 50 studies match, your client receives 51 separate network messages on that association.

QIDO-RS: RESTful search over HTTP

QIDO-RS is the web-native query protocol defined in DICOM PS3.18. Instead of binary datasets inside stateful associations, QIDO-RS uses standard HTTP GET requests.

A client queries an HTTP endpoint using standard URL paths and query parameters:

GET /studies?PatientID=12345&includefield=StudyDate&includefield=ModalitiesInStudy HTTP/1.1
Host: pacs.hospital.example
Accept: application/dicom+json

The URL path /studies targets the study level. You can also query /studies/{studyUID}/series or /studies/{studyUID}/series/{seriesUID}/instances. URL query parameters represent matching keys, while includefield specifies return keys. You can also pass includefield=all to request all populated attributes.

The server responds with a single HTTP response carrying a JSON array:

HTTP/1.1 200 OK
Content-Type: application/dicom+json

[
  {
    "00100020": { "vr": "LO", "Value": ["12345"] },
    "0020000D": { "vr": "UI", "Value": ["1.2.840.113619.2.55.3.1234"] },
    "00080020": { "vr": "DA", "Value": ["20261001"] },
    "00080061": { "vr": "CS", "Value": ["CT"] }
  }
]

QIDO-RS supports native pagination through standard limit and offset query parameters. Classic C-FIND does not support standard pagination. In C-FIND, the server sends every matching record until the stream ends or the client cancels the operation.

Key differences: matching keys, return keys, and wire formats

Comparing C-FIND and QIDO-RS reveals how modern web tooling simplifies data parsing.

Dimension DIMSE C-FIND DICOMweb QIDO-RS
Protocol Raw TCP binary protocol HTTPS (RFC 9110)
Session Model Negotiated stateful association Stateless request and response
Query Format Binary DICOM dataset with empty return keys URL path and query string parameters
Response Format Multiple binary datasets (status 0xFF00), closing with 0x0000 Single JSON array (application/dicom+json)
Pagination Not standardized; server dumps all matches Standard limit and offset query parameters
Network Fit Designed for hospital LANs; port 104 or 11112 Designed for web and cloud infrastructure; port 443

In classic C-FIND, forgetting to include an empty return key means the server will not return that attribute. In QIDO-RS, query parameters map directly to DICOM keywords or 8-character hexadecimal tags. Web developers parse the resulting JSON array using standard JSON parsers built into every language. You do not need a specialized DICOM binary parser to inspect query results.

Neither protocol returns pixels: the retrieval step

Engineers often assume that finding a study also downloads its images. Neither C-FIND nor QIDO-RS returns pixel data or complete DICOM files. Both protocols return metadata only.

To retrieve image instances, you must initiate a secondary retrieval operation. In classic DIMSE, you run C-MOVE, which instructs the archive to push images to your system using C-STORE. In DICOMweb, you run WADO-RS (Web Access to DICOM Objects by RESTful Services). WADO-RS downloads instances or bulk data over HTTP as multipart/related payloads.

This separation keeps search fast. Scanning thousands of metadata records requires minimal bandwidth. Transmitting thousands of multi-frame image slices requires gigabytes of transfer capacity. Keeping query and retrieval separate prevents metadata searches from blocking on heavy network transfers.

Do you actually need to query hospital archives?

When software companies design cloud medical imaging products, their first instinct is often to pull data from hospitals. They plan an architecture where a cloud worker queries an on-premises PACS using C-FIND or QIDO-RS, finds recent exams, and pulls them.

In production, this pull pattern breaks down against hospital network security. Hospital firewalls block inbound connection attempts from the internet. Direct queries require open inbound firewall ports or permanent VPN tunnels. Our guide on push vs pull in medical imaging workflows explains why hospital security teams favor outbound push instead.

Furthermore, clinical systems already push images automatically. When a technologist completes an imaging exam, the scanner or local PACS pushes studies immediately to configured destinations using DICOM C-STORE or DICOMweb STOW-RS. If your application processes incoming exams, you do not need to query for them. You only need an endpoint ready to receive incoming pushes.

Building and operating your own internet-facing DICOM listener requires managing public TLS certificates, handling association handshakes, and scaling concurrent transfers. Alternatively, a managed gateway such as Dicomly accepts inbound C-STORE transfers over mutual TLS and delivers each instance as an HTTPS STOW-RS request to your configured webhook receiver. This removes the operational burden of hosting on-premises listeners or managing VPNs. The trade-off is architectural: your application receives studies as senders push them, rather than querying historical records on demand.

Next steps for testing your query pipeline

If your application builds a clinical viewer or search interface, test both query protocols locally before deploying to production.

To test classic C-FIND against a local PACS or archive, use the DCMTK findscu tool. For example, query for all studies matching a patient ID:

findscu -v -S -k 0008,0052=STUDY -k 0010,0020="12345" -k 0008,0020="" -k 0020,000D="" pacs.local 11112 -aet MY_SCU -aec ARCHIVE

This command sets the query level to STUDY (0008,0052) and filters on Patient ID 12345. It also passes empty tags for Study Date (0008,0020) and Study Instance UID (0020,000D) as return keys.

To test QIDO-RS against a DICOMweb server, use curl. For example:

curl -s -G "https://pacs.example.com/dicomweb/studies" \
  -H "Accept: application/dicom+json" \
  --data-urlencode "PatientID=12345" \
  --data-urlencode "includefield=StudyDate"

Once test queries return valid results, verify how your application handles empty result sets, missing tags, and network timeouts. If your platform only needs to receive new studies, review our quickstart guide. You can also read how endpoints manage sender certificates to set up an event-driven pipeline.

Dicomly builds infrastructure for healthcare software teams that receive DICOM studies from hospitals. Why we built Dicomly →

Sources

  1. DICOM Standard Part 4 (PS3.4, Annex C — Query/Retrieve Service Class) (dicom.nema.org)
  2. DICOM Standard Part 7 (PS3.7, Chapter 9 — DIMSE-C Service Protocol) (dicom.nema.org)
  3. DICOM Standard Part 18 (PS3.18, Section 10.6 — Search Transaction / QIDO-RS) (dicom.nema.org)
  4. RFC 9110 — HTTP Semantics (www.rfc-editor.org)

Start receiving DICOM today.

Early access. No credit card. First endpoint free.