the ontology of the radiographic image: from radlex to radio

The Ontology of the Radiographic Image:

From RadLex to RadiO

What do these two things have in common?

What do these two things have in common?

Both of these two things are images…

What is RadLex?

•RadLex is a controlled vocabulary for reporting radiographic image observations – RadLex is not an ontology (includes term categories such as uncertainty and teaching attributes).

•It’s purpose: „to provide a uniform structure for capturing, indexing, and retrieving a variety of radiology information sources, such as teaching files, research data, and radiology reports. RadLex will unify and supplement radiology terms in other lexicons, such as the ACR Index, SNOMED, the Unified Medical Language System, the Fleischner Society Glossaries, and DICOM.“

•In keeping with this statement of purpose, we aim to expand this web and further supplement RadLex:

•Foundational Model of Anatomy

•Ontology of Biomedical Reality

What is RadiO?

•Ontologically sound framework for electronic radiology reporting

•Interfaces RadLex, via an intermediary ontology of the image, to the FMA (achieved), and the Ontology of Biomedical Reality (planned).

•For any application ontology dealing with imaging techniques, it is important that bodily entities and their appearances are kept in separate (but interrelated) ontological domains.

•Not all pathologies are susceptible to imaging, thus the assertions in radiology reports refer primarily to the images themselves and not the organs that these images are images of (think of cancer on the cellular level)

•Aim: to build a knowledge base of imaging ‘findings’ and their use to form concrete diagnoses.

What is RadiO?

•The application ontology is implemented in Protégé and currently consists of three layers:

•1. an electronic reporting layer,

•2. an upper-level imaging domain ontology, and

•3. the FMA, for anatomical references (w/ OBR, for pathological references).

Upper-Level RadLex Term Categories

Medical Image Domain Ontology

Biomedical Domain

Anatomic Location

Findings

Morphologic and Physiologic Processes

Visual Features

Diagnoses and Etiologies

Image Entity

Anatomical Image Entity

Image Feature

Pathological Image Entity

term_for image_of

Foundational Model of Anatomy

Ontology of Biomedical Reality

is_a is_a

is_a

is_a

is_a

is_a

Aligning RadLex Upper-Level Term Categories to The Domain of the Body, via the Imaging Domain

Anatomic Location

• RadLex Documentation: – “This category specifies the body part or

other anatomic region… [and] are arranged hierarchically according to sub-part relationships.”



Biomedical Domain

Anatomic Location

Findings

Visual Features

Image Entity


Image Feature


term_for image_of



is_a is_a

is_a

is_a

is_a

is_a




Anatomic LocationThe posterior wall of the trachea (T) and the

anterior wall of the esophagus (E) almost contact and, because they contain air and the space between them does not, form the posterior tracheal stripe (arrow) in the lateral view.

There are four essential elements here:1. the posterior wall of the tracheal,2. the anterior wall of the esophagus,3. the space between them,4. and the lateral view (since it is only from

this view that the various anatomical entities are superimposed (or in this case, not superimposed, since the esophagus is behind the trachea from the front view) in such a way to produce it.)

NO FMA CORRELATE!!Not a genuine universal (like the thing of which my left arm, this table and my favourite restaurant are parts)



Biomedical Domain

Anatomic Location

Findings

Visual Features

Image Entity


Image Feature


term_for image_of



is_a is_a

is_a

is_a

is_a

is_a




Visual Features

• RadLex Documentation: “There terms describe features on the image that can be described without reference to specific physical, anatomic, or pathological processes or structures.– These refer only to the image!

Image feature includes image entity attributes such as:

• shape,

• size,

• density,

• as well as patterns that have no direct correlate on the side of the patient, such as image artifacts.

CT of the head, showing a star-like artifact

CT of the liver, demonstrating intense focal enhancement

Visual Features



Biomedical Domain

Anatomic Location

Findings

Visual Features

Image Entity


Image Feature


term_for image_of



is_a is_a

is_a

is_a

is_a

is_a




Morphologic and Physiologic Processes/Diagnoses and

Etiologies RadLex Documentation:• Morphologic and Physiologic Processes:

“These terms describe gross morphologic and physiologic processes, but which do not relate directly to specific physical structures or proven diagnoses.”

• Diagnoses and Etiologies: – Inferred causes– Proven causes

i.e., specific structures

The Natural Question:

What‘s the difference between terms of the category morphologic and physiologic processes and diagnoses and etiologies?

The Difficult Answer:

No hard and fast rule… but consider the following…


Morphologic and Physiologic Processes/Diagnoses and

Etiologies

Findings


term_for image_of



bronchioaveolar carcinoma

tumour

RadLex to Ontology of Biomedical Reality



Thank You

the ontology of the radiographic image: from radlex to radio

Documents

imaging domain slide

images slide

supplement radlex

application ontology

radio slide

intermediary ontology

imaging domain morphologic

imaging techniques