Skip to content
SimpleTools24

What Patient Information Is Stored in DICOM Files?

By SimpleTools24Published 4 min read

DICOM files store a lot of personal information alongside the image: typically the patient's name, ID number, date of birth and sex, plus the dates of the examination, the hospital or clinic, referring and performing staff, and unique identifiers that can link the file back to records. Some images also have identifying text burned into the pixels. Removing all of this safely is called de-identification (often loosely called anonymization), and the DICOM standard defines detailed rules for doing it.

In this guide

This article explains what's in the files and what de-identification involves, so you can make informed decisions before sharing images. It isn't legal advice: organisations handling patient data must follow their own policies and applicable law.

Where Personal Data Lives in a DICOM File

Patient Tags

Common DICOM tags that identify a person
TagNameWhy it matters
(0010,0010)Patient's NameDirect identifier
(0010,0020)Patient IDMedical record number, links to hospital records
(0010,0030)Patient's Birth DateCombined with other data, highly identifying
(0010,0040)Patient's SexQuasi-identifier
(0010,1010)Patient's AgeQuasi-identifier
(0010,1040)Patient's AddressDirect identifier, when present

Study and Visit Information

  • Accession Number (0008,0050) links the images to the order in hospital systems.
  • Study Date and Time reveal when the person was examined.
  • Institution Name and Address narrow down where.
  • Referring and performing physician names add context.

Unique Identifiers

Study, series and instance UIDs are globally unique. They don't spell out a name, but whoever holds the original records can use them to re-identify the images. De-identification normally replaces them with new, consistent UIDs so the study still loads correctly.

Free-Text and Private Tags

Fields such as Study Description or Image Comments sometimes contain names or notes typed by staff. Manufacturers also add private tags with vendor-specific data, which may include identifiers in undocumented places.

Burned-In Annotations

Ultrasound images, screenshots, scanned documents and some X-rays have text drawn directly into the pixels: names, dates, hospital logos. Removing tags doesn't touch these. DICOM includes a Burned In Annotation tag (0028,0301), but it isn't always set reliably, so images need to be checked visually.

Faces in 3D Data

High-resolution head CT or MRI can be reconstructed into a recognisable 3D face. Research datasets often apply "defacing" tools to remove facial features for this reason.

What De-identification Involves

The DICOM standard's Basic Application Level Confidentiality Profile, in PS3.15 Annex E, lists hundreds of attributes and what to do with each: remove it, empty it, replace it with a dummy value, or replace UIDs consistently. Options cover cases like keeping dates relative to each other, or retaining device information needed for research.

A reasonable de-identification process includes:

  1. Applying a confidentiality profile with a dedicated tool, rather than editing tags by hand.
  2. Replacing UIDs consistently, so images still group into the right series.
  3. Removing or cleaning private tags.
  4. Checking pixels for burned-in text, and masking it where necessary.
  5. Reviewing a sample of the output in a viewer to confirm nothing identifying remains.

Open-source tools used for this include DCMTK's command-line utilities, pydicom-based scripts, and dedicated anonymizer applications. Which one fits depends on volume and whether you need research-grade auditing.

In the United States, HIPAA describes two routes to de-identification, Safe Harbor and Expert Determination; other regions have their own rules, such as GDPR in Europe.

Checking a File Before You Share It

Even if you aren't de-identifying anything, it's worth knowing what a file contains before you send it. Open it in a DICOM viewer and look at the tag list. The SimpleTools24 viewer:

  • Lists every data element, filterable by tag number or name.
  • Hides identifying values by default, so you can browse metadata on a shared screen without exposing a name or ID, and reveal them only when you choose to.
  • Decodes files in your browser, so checking a file doesn't upload it anywhere.

The viewer doesn't modify or de-identify files. Treat it as an inspection step before and after using a proper de-identification tool. What a DICOM file is explains how tags and values are structured if the list looks unfamiliar.

Practical Advice for Individuals

If you have your own scans on a disc or download:

  • Sharing with another clinic: send the original files through the channel the clinic recommends. They need the identifiers to match records.
  • Asking for an opinion online: don't post original DICOM files publicly. Screenshots with names cropped out reveal much less, but check the corners of the image for burned-in text.
  • Uploading to websites: prefer tools that process files on your device, and read their privacy information first. Copying files from a disc is covered in how to open DICOM images from a CD.

Frequently Asked Questions

Does exporting a DICOM image to JPEG remove the patient data?

It removes the metadata tags, but not burned-in text in the pixels. Check the image visually.

Is removing the name enough?

No. Birth date, IDs, dates, institution and UIDs can still identify someone, especially in combination.

Can I anonymize DICOM files online?

Some services offer it, but that means uploading identifiable health data. For anything beyond your own files, use approved local tools.

Why do de-identified files still have UIDs?

They're replaced with new ones, so images stay grouped correctly without linking back to the original records.

Summary

DICOM files carry identity in tags, UIDs, private elements and sometimes the pixels themselves. Proper de-identification follows a standard profile and includes a visual check; for everyday sharing, know what's in the file before it leaves your hands.

Sources