To translate DICOM UIDs, StudyInstanceUID, SeriesInstanceUID and SOPInstanceUID information accurately, a translator must separate medical-imaging language from the identifiers that bind images, studies, series, objects and storage classes together. Radiology reports, PACS exports, imaging research datasets, conformance statements and DICOMweb documentation can place patient-facing prose, modality labels and machine-readable UIDs in the same record. The explanatory language may need translation; the UIDs normally must remain exact.
This guide explains how to translate DICOM metadata and medical-imaging documentation without changing Study Instance UIDs, Series Instance UIDs, SOP Instance UIDs, SOP Class UIDs or Transfer Syntax UIDs. It solves one distinct search intent: what does each UID identify, which fields may be localized, how instance identity differs from class identity, why a UID is not a patient identifier, and how multilingual imaging systems can preserve the same study, series and object after translation.
The working rule is translate the human-readable metadata, preserve the DICOM UID, keep the UID type visible, and verify every relationship after export. DICOM uses globally unique identifiers for many objects and classes; StudyInstanceUID identifies a study, SeriesInstanceUID identifies a series, SOPInstanceUID identifies a particular SOP instance, while SOPClassUID identifies the class of object or service. These layers are related but not interchangeable. This article is educational translation and data-integrity guidance, not clinical interpretation.
DICOM UID is structured machine identity
DICOM Unique Identifiers are machine-readable values used to identify objects, classes and other entities across systems. Treating them as prose can break interoperability. Preserve every UID exactly. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
UID syntax is numeric and dotted
DICOM UIDs are represented as numeric components separated by periods under the standard’s UID value representation rules. Locale punctuation or typographic substitution can corrupt the string. Keep periods and digits unchanged. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
UID is not a decimal number
A UID may look like a long dotted number but it is an identifier, not a mathematical quantity. Numeric parsing can remove components or formatting. Store UID fields as text. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
UID root carries organizational allocation context
Many UIDs are generated beneath an assigned organizational root. The root should not be translated or simplified. Preserve the complete UID. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
UID suffix must remain unique
Organizations generate suffixes beneath a root according to their own controlled method. Recreating suffixes during translation risks collisions. Never generate replacement UIDs for language reasons. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
StudyInstanceUID identifies one imaging study
DICOM uses StudyInstanceUID to identify the study grouping. Replacing it with a patient number changes the data model. Keep study and patient identity separate. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
SeriesInstanceUID identifies one series within a study
A study can contain multiple series such as different sequences or reconstructions. Using StudyInstanceUID where SeriesInstanceUID is required collapses hierarchy. Preserve series-level identity. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
SOPInstanceUID identifies an individual SOP instance
Images, structured reports and other DICOM objects have instance identity. A series can contain many SOP instances. Keep object identity distinct from series identity. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
SOPClassUID identifies object class
The SOP Class UID identifies the standardized class of the DICOM object or service. It does not identify one image or one instance. Keep class and instance UIDs separate. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Study and series labels are human-readable
Study Description and Series Description fields can contain clinical or workflow text. These fields may be translated according to project policy. Do not use translated descriptions as identity keys. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Patient name is not StudyInstanceUID
A patient can have multiple studies over time. Using patient identity as study identity breaks chronology. Preserve both layers. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Patient ID is not a DICOM UID
PatientID is an institution-defined identifier rather than the global UID mechanism. A patient identifier can change or differ by institution. Keep patient and DICOM object identity separate. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Accession number is another workflow identifier
Radiology orders often have accession numbers linking imaging to scheduling or order workflows. Accession number does not replace StudyInstanceUID. Keep order and imaging identities separate. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Order number is not instance identity
Requested Procedure ID and other order fields can coexist with DICOM UIDs. Those belong to workflow rather than object uniqueness. Translate labels, preserve source values. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Modality is a coded attribute
CT, MR, US and other modality values are controlled data rather than free-form descriptions. Changing them to target-language initials can break systems. Translate explanatory prose, not modality codes. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
SOP Class can reflect modality or document type
Different DICOM SOP Classes represent image storage, structured reports and other objects. A human description can be localized while the class UID remains fixed. Keep both. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Transfer Syntax UID identifies encoding rules
Transfer Syntax UIDs define how DICOM datasets are encoded, including byte ordering and compression-related syntax. They are not file names or image IDs. Preserve the UID exactly. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Changing Transfer Syntax UID without re-encoding is invalid
The UID must match how the dataset is actually encoded. Translation cannot justify editing this field. Leave transport/encoding metadata to imaging software. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Media Storage SOP Class UID is another context
DICOM file meta information can repeat class identity in file-level metadata. A translator should not normalize away apparently duplicate values. Preserve file meta fields. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Media Storage SOP Instance UID ties file meta to object
The file meta information can reference the same SOP instance stored in the dataset. A mismatch is a technical error, not a translation opportunity. Verify equality where expected. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Implementation Class UID identifies software implementation
DICOM applications can identify the implementation that wrote a file or association. It does not identify the study or patient. Keep software and clinical-object identity separate. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Implementation Version Name is human-readable technical metadata
Software may include a version-name field beside an Implementation Class UID. The version string should normally remain exact. Translate documentation around it, not the version. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Frame of Reference UID identifies spatial frame
Imaging objects can share a Frame of Reference UID to establish spatial relationship. It is not a study UID. Keep spatial-reference identity distinct. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Referenced SOP Instance UIDs link objects
DICOM objects can reference other instances by SOP Class and SOP Instance UIDs. Editing a referenced UID can break object relationships. Protect cross-reference fields. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Referenced Series Sequence preserves hierarchy
Structured metadata can point to entire series or studies. A translated display tree should not rewrite underlying UID links. Verify target navigation against source relationships. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Structured Reports can contain many references
DICOM SR objects link observations to images and other objects. The report text can be translated while referenced UIDs remain stable. Keep evidence links intact. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Presentation States reference images
Presentation State objects can point to images whose display should be controlled. Changing SOPInstanceUID references disconnects presentation from image. Preserve all references. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Segmentation objects reference source images
DICOM Segmentation can reference underlying series or images. A target label can be translated while source-instance UIDs remain unchanged. Protect source relationship. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Parametric maps and derived objects have their own UIDs
Derived images are new DICOM objects with distinct SOPInstanceUIDs. They can still refer back to source data. Do not collapse source and derived identities. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Enhanced multi-frame objects remain single SOP instances
One SOP instance can contain many frames. Frame number is not SOPInstanceUID. Keep object and frame coordinates separate. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Concatenations add another DICOM relationship layer
Large multi-frame objects can be represented through concatenation-related attributes in applicable IODs. Those relationships are not translation targets. Preserve structured identifiers. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
DICOMweb uses UIDs in URLs and requests
RESTful imaging services often place Study, Series and SOP UIDs in path or query components. Translating or URL-encoding them incorrectly breaks retrieval. Keep route identity stable. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
WADO requests depend on exact UIDs
DICOMweb and WADO-style retrieval can use study, series and object identifiers. A one-digit error retrieves nothing or another object. Verify links and visible labels together. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
QIDO query parameters are machine syntax
DICOMweb query names and UID values form API contracts. Translating keys inside a request breaks interoperability. Translate API documentation, not production requests. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
STOW requests create or store objects
Store operations carry DICOM instances with existing UIDs or system-generated identity under workflow rules. Translation should not invent identifiers. Leave UID creation to the authorized application. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
DICOM JSON representation still carries UIDs
JSON encoding changes transport representation, not the meaning of UID attributes. The UID string remains protected data. Do not translate JSON keys or UID values. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
DICOM XML representation has the same principle
XML element structure and attribute tags are machine syntax. Only explicitly translatable descriptive content should move across languages. Preserve identifiers and tags. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Tag numbers are another machine layer
DICOM attributes are identified by group and element tags such as hexadecimal tag numbers. Tag numbers are not translated terminology. Keep tags unchanged. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Attribute names can be translated in documentation
Human-facing labels such as Study Instance UID can be localized in manuals or interfaces. The actual DICOM keyword and tag should remain available for technical users. Separate display label from schema identity. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
DICOM keywords are API-facing names
Keywords such as StudyInstanceUID are used in standards, APIs and software. Translating the keyword into another language inside code breaks integrations. Translate documentation around the keyword. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
UIDs should never contain localized numerals
Some languages use different numeral glyphs in ordinary text. DICOM UID syntax expects ASCII-style digits. Keep raw UID digits unchanged. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Right-to-left interfaces need directionality testing
Long dotted UIDs can visually reorder or wrap awkwardly in RTL layouts. The underlying UID may be correct while copying becomes unreliable. Test real UI and PDF rendering. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Line wrapping can damage manual copy
Long UIDs can wrap across narrow columns. Inserted spaces or hidden line breaks can create invalid copies. Use layout that preserves the raw string. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Spreadsheets can truncate long UIDs
Long values exceed ordinary numeric precision in many tools. Scientific notation or truncation can silently corrupt identity. Import UID columns as text. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
CSV requires quoting and field discipline
UIDs themselves do not contain commas, but surrounding metadata often does. A malformed CSV round trip can shift UID columns. Validate row structure after export. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
OCR is unsafe for long UID strings
Scanned reports and screenshots can confuse digits or lose periods. A visually plausible UID can be wrong. Use structured source data whenever possible. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Smart punctuation is dangerous
Editing software can replace periods or introduce nonbreaking spaces around long strings. Machine equality can fail even when appearance seems unchanged. Protect raw UID text. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Machine translation should lock UIDs
MT can insert spaces, punctuation or explanatory words. A fluent target can destroy DICOM links. Mask UIDs before automated translation. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Translation memory can carry another study’s UIDs
Radiology templates and conformance documents are highly repetitive. A high match can preserve stale identifiers. Populate UIDs from the current source. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
De-identification can replace UIDs deliberately
Research de-identification workflows may remap DICOM UIDs to protect privacy while preserving internal relationships. That is a controlled data-transformation task, not translation. Do not remap UIDs during ordinary localization. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
UID remapping must preserve referential integrity
When de-identification changes UIDs, all references must be updated consistently. Changing only visible occurrences breaks datasets. Use imaging-aware tooling. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Clinical text translation is separate from DICOM object identity
Study descriptions, protocol names and report text can be localized. The same study and image objects should remain traceable through UID relationships. Keep semantic and identity layers separate. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Worked example: study-series-instance hierarchy
A fictional study contains two series, each with many SOP instances. The target translates series descriptions while preserving StudyInstanceUID, each SeriesInstanceUID and every SOPInstanceUID. Verify hierarchy after export. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Worked example: structured report references
A fictional DICOM SR references three source images by class and instance UID. The report narrative is translated but the evidence references stay exact. Test image links. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Worked example: DICOMweb path
A fictional viewer URL contains study, series and instance UIDs. The interface label is localized while route components remain unchanged. Verify retrieval in the target build. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Error clinic
Common failures include Study/Series UID swaps, translated DICOM keywords, UID truncation, stale TM values, localized digits and broken referenced-instance links. These are interoperability errors rather than prose errors. Run dedicated DICOM identity QA. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Release checklist
Study, Series, SOP Instance, SOP Class, Transfer Syntax, Frame of Reference and referenced UIDs all need relationship checks. Checking only visible labels is insufficient. Validate token equality and object hierarchy. Treat descriptive clinical or workflow text as translatable, the DICOM UID or keyword as protected, and the hierarchy or reference relationship as something to verify after export. This keeps the same study and object graph intact across languages.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Frequently asked questions
DICOM UIDs should not be translated. Users often ask whether StudyInstanceUID is the patient ID or whether SOPClassUID identifies one image. Explain study, series, instance and class identity separately. The practical test is retrieval and referential integrity. A reviewer using only the target should be able to open the same study, series and instance and follow the same referenced-object links without guessing which UID was intended.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Connection to eduKateSG architecture
This specialist owner protects medical-imaging identity rather than broad medical translation or device labeling. Existing health-related owners cover other domains. Use this page for DICOM object identity and references. For structured files, store UIDs as text and exclude them from locale formatting, numeric conversion, spell-checking and generative rewriting. Translation can then change the human-facing fields around stable imaging identity.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Authoritative sources
The DICOM Standard defines UID syntax and all UID-bearing attributes. Vendor documentation can explain implementation but should not override the standard. Verify production behavior against the current DICOM Standard. A strong reviewer checks both directions: start from the target series or report and confirm its UIDs, then start from the UIDs and confirm the same study, series, object class and referenced evidence.
Before release, ask what the field identifies: patient, order, study, series, SOP instance, SOP class, transfer syntax, frame of reference or software implementation. The target should make that layer clear instead of treating every long number as one DICOM ID.
Reference verification route
For production work, use the current DICOM Standard, especially PS3.3 for information objects and attributes and PS3.5 for data structures, value representations and UID rules. DICOMweb behavior should be checked against the current web-services part of the standard. Translation should preserve every UID and machine keyword while localizing only the intended human-readable content.