eduKateSG · Why Science?
Move medical images with their identity and acquisition context intact—while preserving objects, metadata, transfer syntax, services, conformance and privacy
Choose the closest route, then return to the full index whenever you need the wider evidence chain.
Science learning becomes powerful when students can tell a measured pattern from a model, a statistical decision from a mechanism, and a useful lead from a finished conclusion. DICOM—Digital Imaging and Communications in Medicine—is the international standard for medical images and related information. Its current edition is a maintained multi-part standard covering information objects, services, data structures, encoding, media storage, networking, web services, display and security profiles. A DICOM object can keep pixels or other measurements beside patient, study, series, acquisition and device context. A file that parses is not automatically assigned to the correct patient, displayed safely, de-identified adequately or clinically interpreted correctly; conformance statements, identity checks, transfer syntax, calibrated display and governance remain essential. This guide supports Primary Science, PSLE Science, Secondary Science, O-Level Science and STEM education as a longform evidence-reading route; it is not medical advice, a laboratory protocol, an admission promise or a career guarantee.
Related eduKate reading: medical imaging, ultrasound and X-ray evidence; X-ray computed tomography and 3D internal evidence; DICOM UIDs and medical-imaging identity; GA4GH Service Info and self-describing APIs; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 12 October 2026: DICOM Standards Committee current edition index, checked 12 October 2026; DICOM PS3.10 current data-format layer and file-format specification, checked 12 October 2026; DICOM PS3.5 current transfer-syntax specification, checked 12 October 2026; DICOM PS3.15 current security and system-management profiles, checked 12 October 2026; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science teaching and learning syllabus; 2026 MOE G2 Computing syllabus. Standards, terminologies, implementation guides, tools and recommendations can change, so preserve dated documentation, exact source data and software versions.
A deeper reading route
A medical image is more than a picture. A radiograph, CT series, ultrasound study or magnetic-resonance acquisition needs patient and study identity, modality, geometry, timing, acquisition parameters and sometimes structured measurements or reports. DICOM packages this context into defined information objects and specifies services that systems use to store, query, retrieve, print, process or display them.
The standard is organised into many parts. PS3.3 defines information object definitions, PS3.4 service classes, PS3.5 data structures and encoding, PS3.6 the data dictionary, PS3.7 message exchange, PS3.10 media storage and file format, PS3.14 grayscale display, PS3.15 security profiles and other parts cover web services, application hosting and specialised domains. Implementers must cite the relevant edition and parts.
Service-Object Pair Classes combine an information object with a service. A CT Image Storage SOP Class, for example, gives interoperable meaning to both the object and the storage operation. SOP Class and SOP Instance UIDs are not decorative strings; they tell a receiver what object rules apply and which specific instance is being handled.
Study, Series and SOP Instance UIDs form a crucial identity chain. A study groups related imaging work, a series groups instances with shared context and each SOP instance identifies one object. UIDs are globally unique identifiers, not patient-friendly labels. Systems should preserve them through copying and transformation unless a documented derivation creates a new instance.
Data elements use tags, value representations, lengths and values. The data dictionary defines standard attributes. Private elements can carry vendor-specific information, but a receiver may not understand them. Removing all private tags blindly can also delete information needed to interpret pixel data. De-identification requires a profile-aware, modality-aware process and verification.
Transfer syntax defines how the dataset is encoded, including byte order, explicit or implicit value representations and, for image data, possible compression. A reader must negotiate or decode the declared syntax. Lossless and lossy transfer syntaxes have different implications; the standard does not decide whether lossy compression is clinically acceptable for a particular use.
Pixel values need geometry and transformation context. Rows, columns, pixel spacing, orientation, position, photometric interpretation, bit allocation, rescale slope and intercept, windowing or modality-specific transforms can change what a viewer displays. Extracting only the pixel array can produce an image that looks plausible while measurements or orientation are wrong.
DICOM networking historically uses association negotiation and DIMSE services; DICOMweb adds web-based services such as store, search and retrieve patterns. Conformance statements explain supported SOP Classes, transfer syntaxes, roles and options. Two products both labelled DICOM can still fail to interoperate if their supported contracts do not overlap.
PS3.15 specifies security and system-management profiles and explicitly notes that security policy remains outside the standard. Profiles can support secure transport, digital signatures, de-identification, audit trails and other mechanisms. A DICOM connection is not private merely because it uses a medical standard; deployment controls and institutional policy remain necessary.
The optimistic value is an imaging ecosystem in which scanners, archives, workstations and analysis tools can exchange rich objects without reducing them to screenshots. DICOM helps preserve scientific and clinical context; careful identity, encoding, display and privacy validation makes that interoperability trustworthy.
Multi-frame objects can store many frames with shared and per-frame functional groups. Treating them as a stack of unrelated pictures can lose timing, position, diffusion, cardiac phase or other dimensions. Test frame indexing and functional-group inheritance on known datasets before quantitative analysis.
Derived objects require provenance. A reconstruction, segmentation, structured report, secondary capture or presentation state may depend on source images and algorithms. Preserve derivation descriptions, referenced instances, software and parameters so the result is not mistaken for direct scanner output.
Burned-in annotations are pixel content, not ordinary header fields. Removing patient tags does not remove names embedded in ultrasound or screenshots. De-identification testing should combine tag-profile checks, private-tag review, pixel inspection and modality-specific rules, followed by a disclosure assessment.
Image calibration can be modality-specific. CT stored values may need rescale parameters; nuclear-medicine objects may need decay and dose context; ultrasound regions can encode measurement calibration. Generic viewers may display an image while ignoring quantitative meaning. Measurement software should declare supported IODs and transformations.
Presentation states and segmentations should be kept separate from source pixels. A user may see shutters, annotations, colour maps or measurements layered onto the image. Exporting only a rendered screenshot prevents later review of which layer created a visible feature.
Duplicate detection cannot rely on filenames. Files may have no extension, arbitrary names or several copies with identical SOP Instance UID and different bytes. Compare UIDs, transfer syntax, checksums and selected metadata; quarantine conflicting duplicates instead of choosing the newest-looking file.
Conformance testing needs both normal and adversarial fixtures. Include alternative transfer syntaxes, missing optional attributes, long sequences, multi-frame objects, private elements, large pixel data and malformed lengths. A parser that succeeds on one CT sample has not demonstrated DICOM interoperability.
Research export should define a derivative contract: which tags are retained, changed or removed; whether pixel data are compressed; whether UIDs are remapped; how dates are shifted; and how source links are protected. Validate counts, geometry and representative measurements after export.
| Stage | Question | Evidence | Responsible output |
|---|---|---|---|
| Identify | Which patient, study, series and instance is this? | Identifiers, UIDs, modality and timestamps | A traceable imaging object |
| Encode | How are attributes and pixels stored? | Tags, VRs, transfer syntax, geometry and compression | A decodable dataset |
| Exchange | Which service contract is supported? | SOP Class, association or web service and conformance statement | An interoperable transaction |
| Interpret | What can safely be concluded? | Calibrated display, source context, validation, privacy and professional review | A bounded imaging result |
Inside this guide
1–12 · Foundations and design
- 1. Start with the imaging question
- 2. Pin the DICOM edition
- 3. Identify the SOP Class
- 4. Identify the SOP Instance
- 5. Read Study Instance UID
- 6. Read Series Instance UID
- 7. Read patient attributes
- 8. Read modality and equipment
- 9. Read acquisition timing
- 10. Read tags and value representations
- 11. Use the data dictionary
- 12. Handle private elements carefully
13–24 · Models and estimation
- 13. Read transfer syntax
- 14. Distinguish lossless and lossy
- 15. Read pixel geometry
- 16. Read stored pixel values
- 17. Read photometric interpretation
- 18. Read display transforms
- 19. Use conformance statements
- 20. Test DIMSE exchange
- 21. Test DICOMweb exchange
- 22. Validate file structure
- 23. Validate identity chains
- 24. Validate image orientation
25–36 · Diagnostics and interpretation
- 25. Protect privacy
- 26. Implement security policy
- 27. Separate image transport from diagnosis
- 28. Connect to Primary Science
- 29. Connect to Secondary Science
- 30. Connect to mathematics
- 31. Connect to computing
- 32. Connect to pathways
- 33. Did you know?
- 34. Write a reader-ready methods note
- 35. Build a validation ladder
- 36. Finish with a reproducible bundle
Section 1 of 36
1. Start with the imaging question
Name modality, object type, sender, receiver and the decision or research task being supported.
Working checkpoint — DICOM: Write the intended use before choosing services. Keep original object, UIDs, SOP Class, transfer syntax, metadata, pixel bytes, provenance and access history together.
For a classroom model, build a fictional study-series-instance tree.
Section 2 of 36
2. Pin the DICOM edition
The current standard is maintained as a multi-part publication that receives corrections and supplements.
Working checkpoint — medical imaging: Record retrieval date and relevant part versions. Read every image beside geometry, modality, acquisition, compression and display context.
For future study, connect physics to biomedical software while checking pathways separately.
Section 3 of 36
3. Identify the SOP Class
A SOP Class combines an information-object definition with a service such as storage.
Working checkpoint — SOP Class: Validate the class UID against the object. Preserve rejected instances and conversion receipts; a screenshot is not the source object.
For writing, cite DICOM edition, SOP Class, transfer syntax, source device and validation.
Section 4 of 36
4. Identify the SOP Instance
Each stored object uses a SOP Instance UID to distinguish it from every other instance.
Working checkpoint — DICOM file: Detect duplicates and accidental UID reuse. Separate file validity from patient identity, acquisition quality and clinical interpretation.
For students, ask what information an image needs before it can be measured.
Section 5 of 36
5. Read Study Instance UID
Study identity groups related imaging work under a globally unique identifier.
Working checkpoint — transfer syntax: Verify patient and order context separately. Treat tags, UIDs, orientation and transfer syntax as scientific inputs: one mismatch can move, mirror or misassign an image.
For parents, ask whether privacy checks include both headers and visible pixels.
Section 6 of 36
6. Read Series Instance UID
Series identity groups instances that share defined acquisition or processing context.
Working checkpoint — PACS: Do not infer clinical completeness from series count. Keep original object, UIDs, SOP Class, transfer syntax, metadata, pixel bytes, provenance and access history together.
For a classroom model, build a fictional study-series-instance tree.
Section 7 of 36
7. Read patient attributes
Name, identifier, birth details and other fields can be sensitive and may be inconsistent across systems.
Working checkpoint — DICOM: Use identity reconciliation and minimum necessary disclosure. Read every image beside geometry, modality, acquisition, compression and display context.
For future study, connect physics to biomedical software while checking pathways separately.
Section 8 of 36
8. Read modality and equipment
Modality and device attributes describe how data were produced but do not certify performance.
Working checkpoint — medical imaging: Link to quality-control records separately. Preserve rejected instances and conversion receipts; a screenshot is not the source object.
For writing, cite DICOM edition, SOP Class, transfer syntax, source device and validation.
Section 9 of 36
9. Read acquisition timing
Study, series, acquisition and content times answer different questions and can lack timezone context.
Working checkpoint — SOP Class: Preserve precision and declared offsets. Separate file validity from patient identity, acquisition quality and clinical interpretation.
For students, ask what information an image needs before it can be measured.
Section 10 of 36
10. Read tags and value representations
Each data element has a tag and a value representation that constrains encoding and interpretation.
Working checkpoint — DICOM file: Validate VR, multiplicity and length. Treat tags, UIDs, orientation and transfer syntax as scientific inputs: one mismatch can move, mirror or misassign an image.
For parents, ask whether privacy checks include both headers and visible pixels.
Section 11 of 36
11. Use the data dictionary
Standard attributes have defined names, VRs and value multiplicities.
Working checkpoint — transfer syntax: Do not guess meaning from a display label. Keep original object, UIDs, SOP Class, transfer syntax, metadata, pixel bytes, provenance and access history together.
For a classroom model, build a fictional study-series-instance tree.
Section 12 of 36
12. Handle private elements carefully
Private tags may be vendor-specific, sensitive or necessary for a specialised reconstruction.
Working checkpoint — PACS: Inventory owners and meaning before removal. Read every image beside geometry, modality, acquisition, compression and display context.
For future study, connect physics to biomedical software while checking pathways separately.
Section 13 of 36
13. Read transfer syntax
The transfer syntax declares byte encoding and possible pixel compression.
Working checkpoint — DICOM: Confirm the decoder supports the exact UID. Preserve rejected instances and conversion receipts; a screenshot is not the source object.
For writing, cite DICOM edition, SOP Class, transfer syntax, source device and validation.
Section 14 of 36
14. Distinguish lossless and lossy
Compression history can affect measurements and suitability even when images look similar.
Working checkpoint — medical imaging: Preserve lossy flags and ratios. Separate file validity from patient identity, acquisition quality and clinical interpretation.
For students, ask what information an image needs before it can be measured.
Section 15 of 36
15. Read pixel geometry
Pixel spacing, slice spacing, orientation and position place image samples in physical space.
Working checkpoint — SOP Class: Reconstruct coordinates with a known phantom. Treat tags, UIDs, orientation and transfer syntax as scientific inputs: one mismatch can move, mirror or misassign an image.
For parents, ask whether privacy checks include both headers and visible pixels.
Section 16 of 36
16. Read stored pixel values
Bits stored, signedness, rescale transforms and modality rules turn encoded values into quantities.
Working checkpoint — DICOM file: Test minimum, maximum and representative pixels. Keep original object, UIDs, SOP Class, transfer syntax, metadata, pixel bytes, provenance and access history together.
For a classroom model, build a fictional study-series-instance tree.
Section 17 of 36
17. Read photometric interpretation
Monochrome and colour models control how stored components become visible intensities.
Working checkpoint — transfer syntax: Verify inversion and colour conversion. Read every image beside geometry, modality, acquisition, compression and display context.
For future study, connect physics to biomedical software while checking pathways separately.
Section 18 of 36
18. Read display transforms
Windowing, lookup tables and grayscale display functions influence presentation without changing raw measurements.
Working checkpoint — PACS: Separate display preset from source data. Preserve rejected instances and conversion receipts; a screenshot is not the source object.
For writing, cite DICOM edition, SOP Class, transfer syntax, source device and validation.
Section 19 of 36
19. Use conformance statements
A product conformance statement describes supported SOP Classes, roles, syntaxes and options.
Working checkpoint — DICOM: Compare both sides before deployment. Separate file validity from patient identity, acquisition quality and clinical interpretation.
For students, ask what information an image needs before it can be measured.
Section 20 of 36
20. Test DIMSE exchange
Association negotiation and DIMSE services require compatible presentation contexts.
Working checkpoint — medical imaging: Capture success and rejected contexts. Treat tags, UIDs, orientation and transfer syntax as scientific inputs: one mismatch can move, mirror or misassign an image.
For parents, ask whether privacy checks include both headers and visible pixels.
Section 21 of 36
21. Test DICOMweb exchange
Web store, search and retrieve services still depend on DICOM identities, media types and access controls.
Working checkpoint — SOP Class: Test query and object fidelity end to end. Keep original object, UIDs, SOP Class, transfer syntax, metadata, pixel bytes, provenance and access history together.
For a classroom model, build a fictional study-series-instance tree.
Section 22 of 36
22. Validate file structure
PS3.10 file meta information and the encoded dataset should agree on SOP Class and transfer syntax.
Working checkpoint — DICOM file: Run a standards-aware parser and keep warnings. Read every image beside geometry, modality, acquisition, compression and display context.
For future study, connect physics to biomedical software while checking pathways separately.
Section 23 of 36
23. Validate identity chains
Patient, order, study, series and instance links can be structurally valid yet assigned to the wrong person.
Working checkpoint — transfer syntax: Use workflow reconciliation and exception review. Preserve rejected instances and conversion receipts; a screenshot is not the source object.
For writing, cite DICOM edition, SOP Class, transfer syntax, source device and validation.
Section 24 of 36
24. Validate image orientation
A mirrored, rotated or reordered display can be clinically dangerous despite valid pixels.
Working checkpoint — PACS: Use orientation fixtures and visible markers. Separate file validity from patient identity, acquisition quality and clinical interpretation.
For students, ask what information an image needs before it can be measured.
Section 25 of 36
25. Protect privacy
Headers, burned-in pixels, private tags and derived objects can all retain identifiers.
Working checkpoint — DICOM: Apply a defined confidentiality profile and inspect output. Treat tags, UIDs, orientation and transfer syntax as scientific inputs: one mismatch can move, mirror or misassign an image.
For parents, ask whether privacy checks include both headers and visible pixels.
Section 26 of 36
26. Implement security policy
DICOM provides mechanisms and profiles, while organisations define access, retention and incident policy.
Working checkpoint — medical imaging: Use secure transport, authentication, audit and monitoring. Keep original object, UIDs, SOP Class, transfer syntax, metadata, pixel bytes, provenance and access history together.
For a classroom model, build a fictional study-series-instance tree.
Section 27 of 36
27. Separate image transport from diagnosis
A correctly exchanged image does not guarantee acquisition quality or interpretation.
Working checkpoint — SOP Class: Keep qualified clinical review and modality QA visible. Read every image beside geometry, modality, acquisition, compression and display context.
For future study, connect physics to biomedical software while checking pathways separately.
Section 28 of 36
28. Connect to Primary Science
A fictional fictional imaging-object card helps younger learners separate a real-world observation from the record used to describe it.
Working checkpoint — DICOM file: Use invented data and define every label. Preserve rejected instances and conversion receipts; a screenshot is not the source object.
For writing, cite DICOM edition, SOP Class, transfer syntax, source device and validation.
Section 29 of 36
29. Connect to Secondary Science
DICOM supports lessons about measurement, models, evidence, systems and uncertainty.
Working checkpoint — transfer syntax: Trace one statement back to its source field. Separate file validity from patient identity, acquisition quality and clinical interpretation.
For students, ask what information an image needs before it can be measured.
Section 30 of 36
30. Connect to mathematics
Counts, dates, units, intervals, codes and probabilities require compatible definitions before comparison.
Working checkpoint — PACS: Recalculate one small example by hand. Treat tags, UIDs, orientation and transfer syntax as scientific inputs: one mismatch can move, mirror or misassign an image.
For parents, ask whether privacy checks include both headers and visible pixels.
Section 31 of 36
31. Connect to computing
Tags, coordinates, compression, identifiers and services show how software keeps pictures connected to scientific context.
Working checkpoint — DICOM: Break one field or reference, validate it and repair it transparently. Keep original object, UIDs, SOP Class, transfer syntax, metadata, pixel bytes, provenance and access history together.
For a classroom model, build a fictional study-series-instance tree.
Section 32 of 36
32. Connect to pathways
DICOM connects physics, anatomy, medical imaging, software, cybersecurity and biomedical engineering and can illuminate STEM education and career pathways.
Working checkpoint — medical imaging: Check current official education and qualification routes separately. Read every image beside geometry, modality, acquisition, compression and display context.
For future study, connect physics to biomedical software while checking pathways separately.
Section 33 of 36
33. Did you know?
A DICOM object can contain much more than pixels: it can carry the identity, geometry, acquisition and processing context needed to interpret them.
Working checkpoint — SOP Class: Explain the fact in plain language, then name what it does not prove. Preserve rejected instances and conversion receipts; a screenshot is not the source object.
For writing, cite DICOM edition, SOP Class, transfer syntax, source device and validation.
Section 34 of 36
34. Write a reader-ready methods note
State standard version, source system, transformation, terminology, validation, privacy controls and supported claim.
Working checkpoint — DICOM file: Keep the note beside the machine-readable record. Separate file validity from patient identity, acquisition quality and clinical interpretation.
For students, ask what information an image needs before it can be measured.
Section 35 of 36
35. Build a validation ladder
Check bytes, syntax, identifiers, references, terminology, scientific plausibility, privacy and a known fixture in sequence.
Working checkpoint — transfer syntax: Stop at the earliest unresolved layer. Treat tags, UIDs, orientation and transfer syntax as scientific inputs: one mismatch can move, mirror or misassign an image.
For parents, ask whether privacy checks include both headers and visible pixels.
Section 36 of 36
36. Finish with a reproducible bundle
Deliver specifications, source receipts, mappings, software versions, tests, hashes and limitations without exposing governed records.
Working checkpoint — PACS: Make every published claim traceable. Keep original object, UIDs, SOP Class, transfer syntax, metadata, pixel bytes, provenance and access history together.
For a classroom model, build a fictional study-series-instance tree.
Contents · Previous section · Continue to the Science Learning Hub