DE-IDENTIFY · BASIC · RESEARCH & EDUCATION

De-identify DICOM studies into a copy. Your originals stay as they are.

OmniDICOM is a DICOM viewer for Mac and Windows. Its De-identify, what other tools call a DICOM anonymizer, works from the Local archive: select studies or series, click De-identify, and it writes a copy under the DICOM PS3.15 Basic Application Level Confidentiality Profile, with new UIDs, in a new folder. Your original files are not changed, and nothing is uploaded. Use it to prepare teaching files, share a set of studies with a collaborating lab or assemble research data.

Works in Basic, Professional and the 14-day trial; no card required. After the trial, De-identify needs a Basic or Professional subscription.

Text burned into the images is not removed.

Only the DICOM header is changed: names, dates or ID numbers drawn into the pixels, or stored inside compressed image data such as JPEG comments, stay in the copy. Faces can be rebuilt in 3D from head CT and MR. Faces are not removed. Open the copy and check every image before you share it.

Using a profile does not guarantee that every piece of identifying information is gone. Whether the copy counts as de-identified under the rules that apply to you is for you to decide.

  1. Select the study
  2. De-identify: new ID, dates shifted
  3. Copying 91 files
  4. Done: check the copy for text
Real app capture on a copy of a public CT study staged under a demo identity (DOE^JOHN, Example Hospital): De-identify with a new ID and shifted dates. The warning about text burned into the images stays in the window from start to finish. The copying wait is cut. Captured with OmniDICOM 1.1.2; the De-identify window is the same in 1.1.4.

A copy in a new folder

Your original files are not changed

DICOM PS3.15

Basic Application Level Confidentiality Profile

New UIDs

References between the copied files still match

Based on OmniDICOM 1.1.4 · Application interface: English

How to de-identify a DICOM study

  1. Put the study in the Local archive. Click Archive in the toolbar. In the archive window, click Folder to import a folder, or drag folders, files or ZIP archives onto it: Copy files stores copies in the archive, Add links leaves the files where they are. For the study open in the viewer, choose Import to database from the arrow next to Archive. De-identify is in the Local archive only; the viewer window has no De-identify entry.
  2. Select what to copy. Select studies in the study list, or series in the series pane. Click De-identify on the archive toolbar, or right-click and choose De-identify... (studies) or De-identify series... (series).
  3. Read the warning panel. It is always shown. Series that often show patient details in the image are listed and left out: ultrasound, screen and secondary captures, photos, and images the scanner marked as having text in the image. Tick Include them anyway to copy them too, and check them yourself before you share the copy.
  4. Give the patient a new ID. New patient ID replaces the patient name and patient ID in the copy. ANON-0001, ANON-0002 … are filled in, continuing from your last copy. Use letters, digits, hyphens (-) and underscores (_), up to 16; the patient’s own name or ID is refused, however it is spelt. With several patients, a table lists each one: double-click a new ID to change it.
  5. Choose the dates. Shift by a random number of days (keeps intervals) is the default; Remove and Keep unchanged are the others. Birth dates are always removed. The next section says what each one does.
  6. Choose where the copy goes. Under Save in, click Choose folder… and pick a folder you can write to, outside the study’s own folders. OmniDICOM makes a new folder named deidentified-YYYY-MM-DD-HHMM there, after checking the free space.
  7. Keep only what you need. More options ▸ Also keep offers Age, sex, height and weight; Institution name and address; Scanner identity (station name, serial numbers); and Study and series descriptions, after review. All are off by default. With the descriptions ticked, Review… lists every study description, series description and protocol name the copy would keep: edit or clear each one.
  8. De-identify, then check the copy. Click De-identify. The progress bar counts the files, and Stop removes the unfinished copy. When it is done, the window says how many files were de-identified and lists anything left out. Open the copy opens it in the viewer beside what is open; Open output folder shows it in Finder or File Explorer. Look through every series for text in the images before you share it.

Shift, remove or keep the dates

Shift by a random number of days (keeps intervals) The default. Each patient gets one random whole number of days, from 1 to 10 years, and every date of that patient moves back by it, so the intervals are kept. Times of day are kept too, so PET decay correction and time-intensity curves still work. The number is held in memory only and is not saved.
Remove Dates are removed or replaced with a placeholder. PET SUV and time-intensity curves will not work on the copy.
Keep unchanged Real dates stay in the copy. Dates can help identify a patient.

Birth dates are always removed. For PET, SUV on the copy needs the dates shifted or kept and the patient’s weight (Also keep ▸ Age, sex, height and weight); the window says what is still missing.

What the copy holds

  • A new identity. The patient name and patient ID become the new ID. Other names and IDs, birth dates, private (vendor) tags, overlays and icon images are always removed.
  • New UIDs. Every UID that identifies a study, series, image or other object is replaced with a new one; UIDs that name a definition, such as SOP classes and transfer syntaxes, are kept. Within one copy the same original UID always gets the same new UID, so references between the copied files still match. No record of which new identifier belongs to which patient is kept anywhere.
  • The original pixels. Image data is copied exactly as stored, compressed or not, and never decoded. Saved processing is not applied.
  • A mark on every file. Patient Identity Removed is YES, and De-identification Method names the profile and each option, ending in “pixel text not removed”.
  • A summary beside it. deidentification.txt records the profile and options, how the dates were changed and what was left out. It holds no names, IDs, identifiers, paths or descriptions.
  • Folders named from new values only. Patient, study, series and image, for example ANON-0001/ST01/SE003_CT/IM00001.dcm, never the original file or folder names. No DICOMDIR is written.

What is left out of the copy

Structured reports (SR) and key object selections, PDFs and other encapsulated documents, presentation states, RT objects, raw data, registrations, segmentations, surfaces, waveforms and video are not copied, as a whole series or as one file inside an image series: this tool does not clean their content.

Each file is checked again as it is copied, so a screen saved inside a series that looked like plain images, such as a summary screen at the end of a CT series, is left out too and listed with the files that were not copied.

  1. Original and copy
  2. Scroll together
  3. Tags of the copy
Real app capture: the original staged under the demo identity (left) and its de-identified copy (right), scrolled together, then the copy’s DICOM tags. The pixels are copied as stored, so text burned into an image would still be there. Captured with OmniDICOM 1.1.2.

Where it fits in research and teaching

  • Teaching files. A copy of a study for a lecture or a case collection, without the patient’s details in the header, once you have checked its images for text.
  • Sharing with a collaborating lab. One copy of a set of studies, with each patient’s dates shifted by the same number of days so the intervals between the studies stay.
  • Research datasets. De-identify the studies, open the copy, then save each series as a NIfTI volume. Convert DICOM to NIfTI →

What you need

  • Basic, Professional or the 14-day trial. Every computer starts with 14 days of Professional, no card required. After the trial, or once a licence has lapsed, the window explains that De-identify needs a subscription and offers Subscribe and Enter licence…; nothing is written.
  • The studies in the Local archive, copied or linked. Linked files no longer on disk are counted and skipped.
  • A folder for the copy that you can write to, outside the study’s own folders, with room for the copy. The deepest file’s path must stay within 259 characters, the longest path Windows writes, on a Mac as well.
  • Mac: Apple silicon and macOS 13 or later. Windows: 64-bit Windows 10 or 11. The same window and the same steps on both.
  • An internet connection to start the trial and for licensing and update checks. The copy itself is made on your computer.

Common questions

Does De-identify change my original files?

No. It writes a copy into a new folder, and your original files are not changed. Stopping a run, closing the window or quitting OmniDICOM during a run removes the unfinished copy.

Does it remove names or dates burned into the images?

No. Only the DICOM header is changed. Text drawn into the pixels, or stored inside compressed image data such as JPEG comments, stays in the copy, and faces can be rebuilt in 3D from head CT and MR. Series that often carry such text are left out unless you include them. Open the copy and check every image before you share it.

Is De-identify a DICOM anonymizer?

It does what many tools call anonymizing: it applies the DICOM PS3.15 Basic Application Level Confidentiality Profile to a copy, with the options you choose. Using a profile does not guarantee that every piece of identifying information is gone, and whether the copy counts as de-identified under the rules that apply to you is for you to decide.

Can I keep the intervals between studies?

Yes. The default, Shift by a random number of days (keeps intervals), moves every date of a patient back by the same random number of days and keeps the times of day. Keep unchanged keeps the real dates. Birth dates are always removed.

Will the series of a study still fit together?

Yes, within one copy: the same original UID always gets the same new UID, so references between the copied files still match. Another run gives different UIDs, so de-identify a whole study at once rather than series by series.

Can I use it during the 14-day trial?

Yes. De-identify works in Basic, Professional and the 14-day trial. After the trial it needs a Basic or Professional subscription; Basic is US$10 a month or US$99 a year per computer.

Does it work on Mac and Windows?

Yes. Mac and Windows run the same app, with the same De-identify window and the same steps. Open output folder shows the copy in Finder on a Mac and in File Explorer on Windows.

Can the copy be used for diagnosis?

No. OmniDICOM is for research and education only. It is not a medical device and is not intended for diagnosis or clinical decision-making.

TRY IT ON YOUR OWN STUDIES

Try De-identify for 14 days.

The download needs no card. The 14-day Professional trial starts on first launch with an internet connection, and De-identify and NIfTI saving work in it. After the trial, Basic, which includes both, is US$10 a month per computer. Pictures and movies exported during the trial carry a watermark; de-identified copies and NIfTI volumes do not.

OmniDICOM runs on Mac and Windows computers. Open this page on your computer to download it. Meanwhile, read the how-to guides or watch OmniDICOM at work on YouTube.

macOS 13+ · Apple silicon · signed, notarized PKG
Windows 10/11 · 64-bit · signed EXE

After you download

  1. Mac: open the signed, notarized PKG and choose Continue → Install, then open OmniDICOM DICOM Viewer from Applications.
  2. Windows: run the signed setup (.exe) and follow the installer, or install from the Microsoft Store. If Windows shows “Windows protected your PC”, follow these steps.
  3. The 14-day Professional trial starts on the first launch with an internet connection. For research and education only.

Compare plans · Installer details and verification

More about De-identify

Image credits

The clips are captures of the OmniDICOM app; pointers are drawn in and waits are cut. The study is a public CT study, copied and given a demo identity (DOE^JOHN, DEMO-0042, Example Hospital) for the capture. LDCT-and-Projection-data: McCollough C, et al. (2020), The Cancer Imaging Archive, doi:10.7937/9npb-2637, CC BY 4.0, modified; NIBIB grants EB017095 and EB017185.