Medical Imaging concepts
Background
Hospitals examine a large number of patients every day, often involving refined medical imaging techniques. The richness of the resulting plethora of medical imaging datasets is just beginning to be explored. Image datasets, including their metadata, are often stored aligned with the international Digital Imaging and Communications in Medicine (DICOM) data model, making it the de facto standard (https://www.dicomstandard.org/) [1].
A set of medical imaging concepts in the SPHN Schema enables the representation of key metadata related to imaging modalities and the corresponding images produced. It supports the structured description of imaging data acquired during imaging examinations, possibly in the context of clinical trial studies, from high-level procedures down to individual image slices.
A simplified DICOM entity-relationship (ER) model for imaging modalities (Figure 1, left part) illustrates the DICOM real-world entities:
Study: Representing a medical (imaging) exam performed on a single patient. It groups together series of images, structured reports or other instances
Series: Representing and grouping together logically related images, captured by a single imaging modality using a predetermined set of acquisition parameters
Image: Representing and grouping together one or more frames (2D medical image slices)
DICOM defines these real-world entities by templates of attributes. These templates are called (composite) information object definitions (IODs) and the attributes are also known as data elements or tags [2]. In a medical context, clinical IODs (CIODs) are commonly used. The SPHN Medical Imaging Concepts are closely aligned with the DICOM data model (Figure 1, right part).
Figure 1: Simplified side-by-side comparison of the DICOM ER model (left part) and its alignment with the SPHN Schema (right part).
Concept design
Overview
The SPHN Schema features concepts to represent various facets of imaging modalities, including Clinical Trial Study, Imaging Procedure, Imaging Series, and Imaging Frame (see Figure 2). Their hierarchical structure is aligned with the DICOM ER model. Correspondences are summarized in Table 1.
Table 1. Levels of the DICOM ER model and corresponding concepts in the SPHN Schema (release 2025.2).
DICOM |
SPHN |
Comment |
|---|---|---|
Study |
Clinical Trial Study |
Medical imaging data may be collected as part of a clinical trial study protocol. This likely involves multiple imaging sessions of multiple patients. |
Imaging Procedure |
An imaging procedure used for examination of a body site or function, may entail multiple imaging series. |
|
Series |
Imaging Series |
A set of medical image slices or frames obtained by a single imaging modality using a single protocol step with predetermined acquisition parameters. |
Image |
Imaging Frame |
A medical image slice or frame in an imaging series, i.e., an image. |
Figure 2 illustrates the hierarchical relationships between the general imaging concepts (Imaging Series and Imaging Frame) and their inherited specialized imaging concepts (CT, MR, PET and X-Ray Imaging Series and Frame). The medical imaging concepts are designed to represent imaging procedures and related metadata. Each Imaging Procedure can be associated with a Clinical Trial Study and may include zero or more Imaging Series, which are a grouping of related image datasets. Each imaging series may in turn consist of zero or more Imaging Frames, representing individual image slices within the series.
Figure 2. Imaging procedure pattern implemented in the SPHN Schema.
In addition to imaging modalities, supporting concepts describe contrast agents, administration procedures, synchronization parameters, and algorithms used in image processing. Descriptions of all imaging concepts are provided in the next section.
Individual concept details
Clinical Trial Study
”describes a study, research or clinical trial”
The WHO defines a clinical trial as “any research study that prospectively assigns human participants or groups of humans to one or more health-related interventions to evaluate the effects on health outcomes.” (https://www.who.int/news-room/questions-and-answers/item/clinical-trials)
In the context of the SPHN medical imaging concepts, the concept Clinical Trial Study can group imaging-related interventions (see Imaging Procedure).
Imaging Procedure
”imaging procedure used for examination of a body site or function”
The Imaging Procedure concept reflects the overall imaging examination performed on a subject and links to administrative context (e.g. clinical trial, case, source system). It captures subject-level and procedural metadata such as intent, demographics, body measurements, body site, and start/end time, and groups one or more imaging series.
An imaging procedure can - but does not have to - be part of a clinical trial study.
Currently, SPHN supports four types of imaging procedures: Computed Tomography (CT), Magnetic Resonance (MR), Positron Emission Tomography (PET), and X-Ray. Each of these procedures is defined using specific concepts that follow the series-frame modeling pattern described above (see Figure 2).
A complete example with all attributes is shown for the Magnetic Resonance imaging type in Figure 3.
Figure 3. Magnetic resonance imaging represented in the SPHN Schema (including all attributes and directly linked relevant concepts).
Imaging Series
”a set of medical image slices or frames obtained by a single imaging modality using a single protocol step with predetermined acquisition parameters”
The Imaging Series concept represents a logical subset of images acquired with a specific modality and consistent acquisition settings. It includes modality code, body position of the subject, imaging device used, number of frames, timing, and references to the produced data files.
The generic Imaging Series concept has four specialized child concepts, Computed Tomography Imaging Series, Magnetic Resonance Imaging Series, Positron Emission Tomography Imaging Series, and X Ray Imaging Series.
See example in Figure 3.
Imaging Frame
”a medical image slice or frame in an imaging series”
The Imaging Frame concept represents information about an individual image. It contains fine-grained technical and contextual details such as image type (e.g., 2D, 3D), dimensions, anatomical projection, metrics, compression algorithm, contrast agent phase, synchronization, and content qualification.
The generic Imaging Frame concept has four specialized child concepts, Computed Tomography Imaging Frame, Magnetic Resonance Imaging Frame, Positron Emission Tomography Imaging Frame, and X Ray Imaging Frame.
See example in Figure 3.
Imaging Device
”medical imaging equipment acquiring image datasets”
The Imaging Device-concept is a child of the generic Medical Device concept, featuring model name and device name in addition.
Radiopharmaceutical
”pharmaceutical containing radionuclides used to diagnose or treat certain diseases”
The Radiopharmaceutical-concept follows the ‘Drug’ representation. It references in addition the Radionuclide to specify properties of this core active component.
Radionuclide
”an unstable nuclide, also known as radioisotope, that can undergo radioactive decay”
Radionuclides (also called radioisotopes) are the core components of radiopharmaceuticals. The Radionuclide-concept describes the properties of a radionuclide including half life, positron fraction, and radioactivity. While half life and positron fraction are provided as quantities with specific units, the radioactivity references the specific Radioactivity-concept.
Radioactivity
”physical phenomenon of emitting energy and subatomic particles in the form of radiation when a nucleus of an atom spontaneously disintegrates”
The Radioactivity-concept describes the magnitude of nuclear activity, together with the time when and the method by which it was determined (e.g., measured or calculated).
Synchronization
”intentional temporal alignment of events to ensure coherence across interacting processes”
The Synchronization-concept represents the deliberate coordination of timing between events or processes so that they occur in a controlled and consistent temporal relationship. This is particularly important when timing an imaging procedure with external factors like motion caused by respiration or cardiac activity.
The Synchronization-concept therefore has two specialized child concepts, Cardiac Synchronization and Respiratory Synchronization.
Image Dimension
”number and size of axes that define the structure of an image”
The Image Dimensions-concept describes the dimensions of an image dataset/volume, for example the number of rows and columns, slice thickness and spacing etc. As there can be multiple image datasets/volumes in an Image Series, this concept groups together the dimension properties of each image dataset/volume in it.
Pixel Dimension
”size of an image pixel”
The Pixel Dimensions concept describes the physical dimensions of a pixel in an image dataset/volume (not the number of pixels though). This concept groups together the dimensions in row and column direction. It does NOT refer to the number of pixels along the width and height of a 2D image! This information would be available via “number of rows” and “number of columns” of concept Image Dimensions. It also does NOT refer to the physical pixel dimensions in a detector/camera!
Guidelines for data delivery
The DICOM Series properties shall be stored in Imaging Series.
These are the properties that are always the same for all image slices/frames in the imaging series
This includes, for example, modality, imaging device, and/or protocol.
The DICOM Frame properties (functional groups) shall be stored in Imaging Frame.
These are the properties that can vary in a multi-frame image series
This includes, for example, orientation, time (series), location (e.g. stereo), and/or dimension.
In case there are two or more DICOM Frames that would result in the same Imaging Frame, only one should be stored as an Imaging Frame instance and linked to the Imaging Series. Or, in other words, the Imaging Frames should be unique in an Imaging Series .
The attribute “slice thickness” of Image Dimensions defines the thickness of a slice. Slices can overlap or there can be space between them. The dimensions are physical distance in the imaging target (patient, specimen, or phantom), not detector sizes.
Pixels, in contrast, are on a 2D grid and do neither overlap nor have spaces between them.
The Pixel Dimensions-concept DOES reflect
physical distances in the imaging target (patient, specimen, or phantom) represented by a pixel.
The Pixel Dimensions-concept DOES NOT reflect
detector sizes
number of pixels along the width and height of an image. This information is represented via “number of rows” and “number of columns” of concept Image Dimensions.
The DICOM Image can come in one of two versions as it can describe one 2D image or contain a list of frames, where each frame again relates to a 2D image:
Many imaging modalities DICOM specifies two kinds of CIODs (two versions for each imaging modality).
Legacy format (example for CT: https://dicom.nema.org/medical/dicom/current/output/chtml/part03/sect_A.3.3.html#table_A.3-1)
Only a single 2D image is defined.
The legacy format is also known as the Single-Frame format.
Enhanced format (example for CT: https://dicom.nema.org/medical/dicom/current/output/chtml/part03/sect_A.38.html#table_A.38-1)
Multiple 2D images can be defined.
The enhanced format is also known as the Multi-Frame format.
All data items in the DICOM Image are represented by the SPHN Imaging Frame concept. The SPHN Imaging Frame instances can therefore be directly linked to an SPHN Imaging Series instance.
The attributes in DICOM Image and DICOM Frame should be stored in SPHN Imaging Frame. Only unique SPHN Imaging Frame instances are linked to an SPHN Imaging Series instance.
Explanatory example: A CT scan with 400 2D image slices would in principle result in 400 instances of SPHN Frame. Practically, most of these 400 SPHN Frame instances would have the same metadata in it. Therefore there is no need to store all 400 SPHN Imaging Frame instances, but only the unique ones.
The cardinality of the body site-attribute of Imaging Procedure is 0:n since an examination can entail multiple body sites, ,e.g., multiple part of a limb like forearm, elbow, and upper arm.
Imaging modalities for which no specific concept exists, e.g. Ultrasound, Microscopy, or SPECT (single photon emission computed tomography), can be added as an instance of the general SPHN Imaging Series (and SPHN Imaging Frame) concepts.
Further details on the individual concepts can be found in the concept documentation available in the SPHN Semantic Interoperability Framework Git-repository (https://git.dcc.sib.swiss/sphn-semantic-framework/sphn-schema/-/tree/master/dataset/documentation-2025.2)