QIDO-RS, WADO-RS and STOW-RS are three named transactions within the DICOMweb Studies Service, not three separate products or competing protocols. The DICOM standard assigns each transaction a distinct verb and resource: QIDO-RS searches for DICOM objects, WADO-RS retrieves them, and STOW-RS stores them [1][3]. The DICOM Standards Committee describes the Studies Resource as enabling a user agent to store, retrieve, update, and search an origin server for DICOM Studies, Series, and Instances, along with their metadata, rendered, and thumbnail variants [1]. A practical integration therefore depends less on choosing one transaction than on understanding which resource hierarchy level each request targets and what the server must return when processing succeeds or fails. The following sections separate those boundaries using the standard's own definitions and a documented vendor implementation.
The Three Transactions Are Named Operations on One Studies Service
PS3.18 defines the Studies Service as a Representational State Transfer web service containing a set of transactions. The Retrieve Transaction is also known as WADO-RS, the Store Transaction as STOW-RS, and the Search Transaction as QIDO-RS [1][3]. This naming matters because it means the transactions share a common resource model rather than existing as isolated APIs. The Studies Resource manages a collection of DICOM Study resources, with each Study organized in a hierarchy of sub-resources corresponding to the DICOM Information Model and the Study Root Query/Retrieve Information Model [3]. A client that understands the hierarchy can move from a QIDO-RS search result to a WADO-RS retrieval URL without switching to a different service definition.
The DICOMweb overview published by the DICOM Standards Committee presents the same division in tabular form. Query maps to QIDO-RS and PS3.18 section 10.6, Retrieve maps to WADO-RS and section 10.4, and Store maps to STOW-RS and section 10.5 [2]. WADO-URI, which retrieves single DICOM instances, is listed separately under PS3.18 chapter 9 [2]. This distinction is operationally significant: WADO-URI is a URI-based web service, while WADO-RS is a RESTful transaction within the Studies Service. Teams that treat the two retrieve mechanisms as identical may overlook differences in how each addresses instances and negotiates response media types.
QIDO-RS Search Is a Query Against Hierarchical or Relational Resources
The DICOMweb Cheat Sheet 2026c lists QIDO-RS as a GET request of the form GET {/resource}{?search*}, where the resource is one of relational or hierarchical paths [5]. Hierarchical examples include /studies for all studies, /studies/{study}/series for all series of a study, and /studies/{study}/series/{series}/instances for all instances of a study's series [5]. A relational path such as /studies/{study}/instances returns all instances of a study [5]. The distinction between hierarchical and relational addressing affects which attributes a server can return and how a client should interpret the result set.
PS3.17 describes QIDO use cases for what the requesting system needs. The technical scenarios ask whether the client prefers XML or JSON, whether it searches at the Series and Instance level or processes full Study metadata, and which attributes it needs for each matching Study, Series, or Composite Instance [7]. One documented use case, Study Import Duplication Check, searches at the Study level using Study Instance UID, Series Instance UID, and SOP Instance UID, then needs Study Instance UID in the response [7]. Another use case, Mobile Device Access, prefers JSON and searches at the Study, Series, and Instance level [7]. These examples show that QIDO-RS design choices follow from the consuming application's data model, not from a single universal query pattern.
A Suggested Validation Scenario for a QIDO-to-WADO-to-STOW Round Trip
A useful hypothetical validation scenario would begin with a QIDO-RS GET against /studies using a Patient ID search parameter, then parse the returned JSON or XML to obtain a Study Instance UID and a WADO-RS Retrieve URL [5][7]. The client would issue a WADO-RS GET to retrieve the study metadata, then construct a STOW-RS POST to /studies with a single instance in multipart/related format [5][6]. The validation would check that the STOW-RS response receipt reports the instance as processed and that a subsequent QIDO-RS search returns the stored instance [6].
This scenario is a suggestion for internal integration testing, not a claim about any vendor's tested behavior. The standard does not specify a universal test suite, and the sources do not report independent conformance results. A team running this scenario should record which media types the server accepts, whether the QIDO-RS response includes the WADO-RS URL, and whether the STOW-RS receipt distinguishes warnings from failures [1][3][6]. Those observations would document the deployment boundary without inferring that other DICOMweb implementations behave identically.
WADO-RS Retrieval Returns Objects, Metadata, Rendered Frames and Thumbnails
WADO-RS is the Retrieve Transaction of the Studies Service [1][3]. The Cheat Sheet shows that retrieval can target the Study, Series, or Instance resource and can request variants including /metadata, /rendered, /renderedMPR, /rendered3D, and /thumbnail [5]. A GET to a bulkdataURI retrieves a single data set [5]. This resource structure means a WADO-RS request is not limited to returning a binary DICOM object; the same transaction can return a JSON or XML metadata representation or a rendered image in an alternative media type.
The Springer article by Genereaux and Dennison describes WADO-RS as retrieving DICOM persistent objects using RESTful services [6]. It states that images can be requested at a particular size, cropped to a specific region, or have a presentation state applied [6]. The article attributes these capabilities to the WADO-RS retrieve operation, but it does not provide independent test results or claim that every server implements every rendering option. The standard's own content negotiation query parameters are designed for hyperlinks where header fields are not accessible [1]. A client that relies on query parameters for media type negotiation should verify that the origin server supports the same parameters, since the standard distinguishes parameters the user agent must supply from those the origin server must support [1].
STOW-RS Storage Uses POST and Returns a Processing Receipt
STOW-RS is the Store Transaction of the Studies Service [1][3]. The Cheat Sheet lists POST /studies to store instances and POST /studies/{study} to store study instances [5]. The Springer article describes STOW-RS as invoked to create instances by uploading instances to the study resource with pixel data and a properly formed set of metadata in either XML or JSON [6]. It also states that a set of instances can be uploaded and that the server returns a receipt of what was processed and what failed processing [6]. This receipt behavior is a key difference from a simple HTTP 201 response: the client must inspect the payload to determine per-instance outcomes.
The standard indicates that a payload may contain a Status Report describing any failures, warnings, or other useful information [3]. This means a STOW-RS client cannot assume that a successful HTTP status code implies every instance in a multipart request was stored. The Springer article gives an example of uploading a standard JPEG image from a consumer camera or mobile phone to a DICOM archive with an HTTP POST and a few key metadata elements in XML [6]. That example is presented as an application capability, not as a claim about clinical suitability or validation. Teams designing STOW-RS ingestion should treat the response receipt as the authoritative record of per-instance processing.
Proxy and Deployment Boundaries Separate DICOMweb from DIMSE
The DICOMweb overview states that DICOMweb can be implemented directly or as a proxy to the DIMSE services to offer modern web-based access to DICOM-enabled systems [2]. It adds that image-producing modalities do not all need to be retrofitted to support DICOMweb [2]. PS3.1 makes the proxy relationship more explicit: the HTTP requests defined within the standard are sufficient for the HTTP server to act as a DICOM SCU to retrieve or store the requested objects from an appropriate DICOM SCP using baseline DICOM functionality [8]. In other words, the HTTP server can act as a proxy for the DICOM SCP [8].
The Cheat Sheet's operational differences table contrasts DICOMweb and DIMSE. DICOMweb is resource-oriented with stateless requests, while DIMSE is service-oriented with association-oriented session state [5]. HTTP is synchronous, with long-running operations typically handled through polling and large data sets possibly using pagination or chunking, whereas DIMSE has native support for multi-response and callback-based workflows such as C-FIND, C-MOVE, and N-EVENT-REPORT [5]. The table also notes that DICOMweb has no standardized cancellation mechanism; clients may abandon requests or stop polling [5]. These differences affect how a proxy must translate between a stateless HTTP request and a stateful DIMSE association.
Sources & references
- PS3.18 - DICOM Standard - NEMA · dicom.nema.org
- DICOMweb™ · www.dicomstandard.org
- PS3.18 - DICOM - NEMA · dicom.nema.org
- Access DICOMweb APIs to manage DICOM data in Azure Health Data Services | Microsoft Learn · learn.microsoft.com
- Cheat Sheet · www.dicomstandard.org
- DICOMweb™: Background and Application of the Web Standard for Medical Imaging | Journal of Imaging Informatics in Medicine | Springer Nature Link · link.springer.com
- HHH.4.3 Description of The Use Cases · dicom.nema.org
- PS3 - DICOM · dicom.nema.org