Objects

Objects at different levels of deidentifying a DICOM dataset. Deidentification comes down to changing the right parts of a DICOM dataset. The change to each DICOM element in a single dataset can be characterized as a delta set. A DeltaSet is effected by a deidentifier. A deidentifier is a concrete implementation of a Protocol.

Deidentification objects from abstract (top) to concrete (bottom)

Protocol

Defines how to handle the deidentification of any incoming dataset using four Components: Filters, Tags, Pixel and Private. It does not say anything about implementation, it only prescribes what should be done to each part of a dataset and under which circumstances to reject it outright.

A single protocol can be implemented by many deidentifiers.

Deidentifier

A piece of software that takes a DICOM Dataset and removes PHI from it. It does this via four Components: Filters, Tags, Pixel and Private.

A deidentifier can do one of two things with an incoming dataset:

1. Reject the dataset trough triggering one of the filters

2. Apply a transformation to the dataset. The transformation is defined in the Tags, Pixel and Private components. The observed changes in the tags form a Deltaset

A deidentifier implements a deidentification protocol. Multiple deidentifiers can implement the same protocol.

Contrary to a Protocol, a deidentifier is a concrete implementation. It will have to actually implement a protocol’s abstract Action Codes. For action codes like REMOVE this is trivial, just remove the dicom element. But for CLEAN many different operations might be said to implement ‘cleaning’. It is up to the creators of a deidentifier to defend the choice for an implementation in a given context.

Deltaset

A set of observed changes to dataset elements. See the Spaces and Codes page for a full description

Like this:

Tag Name	Value Before	Value After	Delta
PatientName	SMITH^JOHN	Patient01	CHANGED
Modality	CT	CT	UNCHANGED
Study Date	20240315	<tag not found>	REMOVED
Manufacturer	Company A	<empty>	EMPTIED
De-identification Method	<tag not found>	deidentifier B	CREATED

Dataset

A standardized container that stores a medical image along with associated metadata. Each dataset contains both pixel data (the actual medical image) and a comprehensive set of information tags that describe patient details, acquisition parameters, and clinical context. Each element in a dataset consists of a tag, tag description and value. For example:

Tag	Description	Example Value
(0010,0010)	Patient’s Name	SMITH^JOHN
(0010,0020)	Patient ID	MRN12345678
(0010,0030)	Patient’s Birth Date	19700101
(0008,0020)	Study Date	20240315
(0008,0060)	Modality	MR
(0008,0070)	Manufacturer	Medical systems LTD
(0008,0090)	Referring Physician’s Name	JONES^SARAH^M.D.
(0020,000D)	Study Instance UID	1.2.840.10008.1.2.3.4

DICOM datasets can be stored as files, in databases or in memory.