// blog
DICOM Study, Series and SOP Instance UIDs: How a Scan Is Organized
Study Instance UID (0020,000D) identifies one exam, Series Instance UID (0020,000E) one set of images within it, and SOP Instance UID (0008,0018) one image or object. UIDs are dot-separated numbers of at most 64 characters, rooted in an organization’s registered prefix, and must be globally unique. Group files by Series Instance UID, not by folder, and give any image you change or derive a new SOP Instance UID.

Open a folder of DICOM files and the file names tell you nothing. The structure is inside the files, in three unique identifiers.
Study, series, instance
| Level | Tag | What it is |
|---|---|---|
| Study | (0020,000D) Study Instance UID | One exam: “CT chest, Tuesday” |
| Series | (0020,000E) Series Instance UID | One set of images in it: one reconstruction, one MR sequence |
| Instance | (0008,0018) SOP Instance UID | One object: usually one image |
A CT exam is typically one study with a few series (soft tissue and lung reconstructions, a scout image), each with hundreds of instances. An MR study has a series per sequence. A scan CD’s DICOMDIR indexes exactly this tree.
Two more you’ll meet: SOP Class UID (0008,0016) says what kind of object it is (CT Image Storage, say), and Frame of Reference UID (0020,0052) says which images share a coordinate system, so they can be overlaid.
What a UID looks like
From our chest CT sample:
(0020,000d) UI [1.3.6.1.4.1.14519.5.2.1.9999.103.2778462321678208806502460428055] StudyInstanceUID
The rules (PS3.5, chapter 9): only digits and dots, each part without leading zeros, at most 64 characters. The beginning is a root registered to an organization; the rest it makes unique. 1.2.840.10008 is DICOM’s own root, used for things the standard defines, like transfer syntaxes.
When you must make a new one
If you change the pixels, you’ve made a new image, and it needs a new SOP Instance UID. When we scaled our X-ray samples down for the web, DCMTK’s dcmscale gave them new instance UIDs under DCMTK’s own root, 1.2.276.0.7230010.3. The study and series UIDs stayed the same.
The rule of thumb: same image, same UID; different pixels, new UID. Reusing a UID for a different image makes systems think they already have it.
Grouping files correctly
- Group by Series Instance UID, not by folder. A folder can mix series, and one series can span folders.
- Sort within a series by position, not by file name: Image Position (Patient) (0020,0032) along the slice direction, or Instance Number (0020,0013) as a fallback.
Our viewer does both when you open a folder or a ZIP: it keeps the series with the most images and sorts it from head to feet.
Key takeaways
- Study (0020,000D) → Series (0020,000E) → Instance (0008,0018).
- UIDs: digits and dots, at most 64 characters, under a registered root (or 2.25.).
- New pixels need a new SOP Instance UID.
- Group by series UID, sort by position.
Open a folder and see the series it finds →
We’re glad to share what we know. When you want a viewer, portal or imaging integration of your own, we’ll put all of it to work on your project, and because we build with AI-assisted development, it costs far less than you might expect. Let’s talk.
Frequently asked questions
Can I use the Study Instance UID as a database key?
Yes, it's meant to be globally unique. Store it as text, not as a number: it's up to 64 characters of digits and dots.
Is the Study ID the same as the Study Instance UID?
No. Study ID (0020,0010) is a short, local identifier from the scanner or hospital, and Accession Number (0008,0050) is the order number. Only the Instance UIDs are globally unique.
How do I make my own UIDs?
Under a root your organization owns, or with the 2.25. prefix followed by a UUID turned into a decimal number, which the standard allows. Libraries such as DCMTK generate them for you.
































































































































