Consistent patient ID across studies
Maintain consistent pseudonyms for the same patient across multiple studies.
Problem
You’re de-identifying multiple DICOM studies from the same patient for a research dataset. You need to replace the original Study Instance UIDs (which are globally unique and traceable) with pseudonyms, but you must ensure that the same original UID always maps to the same pseudonym across all studies. Without this consistency, a single patient’s studies cannot be linked together in the de-identified dataset, breaking longitudinal analysis.
Additionally, distinct source UIDs must never collide (map to the same pseudonym). If Study UID A and Series UID B both mapped to 2.25.123, the dataset would have corrupted referential integrity.
Steps
-
Open your DICOM file (⌘O) and switch to Edit mode (Ctrl+2 / ⌘2 in the desktop app).
-
Note the original Study Instance UID. Filter for
(0020,000D)and record the current value. -
Switch to Anon mode (Ctrl+3 / ⌘3 in the desktop app, Ctrl+Shift+A / ⌘⇧A in the browser), select the Basic profile, and click Apply.
-
Verify the remapped UID. Switch back to Edit mode and check
(0020,000D). The UID should:- Start with
2.25.(the OID root used for derived UIDs) - Be different from the original UID
- Stay the same each time you anonymize this file on this device (same input → same output)
- Start with
-
Verify distinct inputs stay distinct. Filter for both
(0020,000D)(Study UID) and(0020,000E)(Series UID). These must be different original values, and after anonymization, their pseudonyms must also be different.
Result
Each UID is replaced with a deterministic 2.25.* pseudonym derived from a keyed hash of the original, using a secret kept on this device. The same input always yields the same pseudonym on this device, distinct inputs never collide, and someone holding an original UID cannot check it against your output. Another device has another secret and gives different UIDs, so anonymize a dataset that must stay linked on one device. Studies that originally shared a UID still share one afterward, keeping referential integrity across the dataset.