Transcript Relations in Image Ontologies

Relations in Anatomy and Image Ontologies
Dirk Marwede
Institute for Formal Ontology and Medical Information Science,
Saarland University, Saarbruecken, Germany
Department of Diagnostic Radiology, Leipzig University,
Germany
Overview
• OBO relation ontology
• Anatomical Entities (Foundational Model of Anatomy, FMA)
– Relations in Anatomy Ontology
• Diagnostic Domain of Medial Imaging
– Image Entity Types
– Relations between Image Entity Types
– Building an Imaging Ontology (RadiO)
OBO relation ontology
• Three types of binary relations
– < class, class >: e.g. is_a
lung is_a lobular organ
– < instance, class >: instance_of
This particular „lung“ instance_of class lung
– < instance, instance >: instance-level relation, e.g. part_of
This particular instance of „right lower lobe of lung“ part_of this
particular instance of „right lung“
• Relations between classes represent what is general in reality.
– class level:
• lung is_a lobular organ
– instance level:
• this particular instance of lung is_a particular instance of lobular organ
Anatomical Ontologies (FMA)
• Class level-relations
– Structural Relationships between Anatomical Entities
• Boundary (bounded_by)
• Partonomy (part_of, regional_part_of, constitutional_part_of,…)
• Spatial Association
– Location (located_in, contained_in, adjecent_to)
– Orientation (coordinate, laterality)
– Connectivity (continuous_with, attached_to)
– Properties of Anatomical Entities:
• Dimension
• Physical Properties
Canonical Anatomical Entity - Lung
Image Ontologies
• Medical Imaging is concerned with diagnosing diseases.
• How we come from an image to a diagnosis?
– What kind of entities exist on the image and how do they relate to each
other ?
• Image Entity Types
– Anatomical Image Entities
– Pathological Image Entities
– Image Features
dependent entities
Imaging body entites
Image Entity Types –
Anatomical Image Entities
Image Entity Types – Pathological Image
Entity
• Which diseases can be inferred from images?
Pathological
Image Entities
• What kind of image features do diseases have?
• Does a disease have in any case identical image features?
• If not, what are criteria which give evidence for a
disease?
Features
Class-Level Relations between Image Entity
Types
• C image_of C1
– basic relation holding between two continuants.
– C is an anatomical or pathological image entity, C1 is an anatomical or
pathological entity.
• C has_feature C1
– a property relation holding between two continuants.
– C is an anatomical or pathological image entity, C1 is a feature attribute.
• C has_location C1
– basic location relation holding between two continuants.
– C is a visual feature or pathological image entity and C1 is an anatomical
image entity
Class-level relations between Image Entity
Types and the FMA
Anatomical Image Entity
image_of
Anatomical
Entity (FMA)
has_location
has_feature
has_location
has_feature
Feature
Pathological Image Entity
Subrelations of has_feature
Relations to annotate properties to anatomical and pathological image entities
Visual features
Morphology features
c has_shape c1 at time of examination.
Adrenal gland [Anatomic Image Entity] has_shape round [Attribute:Shape] at time of examination
c has_size c1 at time of examination.
T1 vertebral body [Anatomical Image Entity] has_size decreased in height [Attribute:Size] at time of examination
c has_composition c1 at time of examination.
Tumor [Pathological Image Entity] has_composition cystic [Attribute:Composition] at time of examination
Signal features
c has_density c1 at time of examination.
Liver [Anatomical Image Entity] has_density hypodense [Attribute:Density] at time of examination (or contrast
phase).
General features
c has_amount c1 at time of examination.
Pulmonary nodule [Pathological Image Entity] has_amount multiple [Attribute:Amount] at time of examination.
Use of Relations in Medical Imaging
[Pathological Image Entity] has_feature [GeneralFeature]
Pulmonary embolism
has_timing acute
Disease Ontology
[Pathological Image Entity] has_location [Anatomical Image Entity]
has_feature [MorphologyFeature]
Mass
has_location upper right lobe of right lung
has_margin spiculated
has_shape round
has_composition solid
Imaging Ontology
evidence for malignant neoplasm
[Anatomical Image Entity]
Thyroid gland
has_feature [MorphologyFeature]
has_size enlarged
Imaging Ontology ?
[Anatomical Image Entity] has_feature [MorphologyFeature]
[Pathological Image Entity] has_location [Anatomical Image Entity]
Hilar lymph node
Granuloma
has_composition calcified
has_location upper lobe of left lung
Imaging Ontology
evidence for tuberculosis has_timing old
Disease Ontology
Conclusions
•
Entities of two domains
– Body Entities
– Image Entities (dependent entities)
•
Construction of an Imaging Ontology
– Image Entity Types
• Anatomical Image entities
• Pathological Image Entities
• Features
– Class level relations between Image Entity Types (and the FMA)
•
Application Ontology (RadiO) for Diagnostic Domain
– Annotating image features to Anatomical and Pathological Image entities.
– Criteria for Pathological Image Entities: Tracking the use of relations between image entity
types to discriminate image features of diseases.
•
Separating/linking of an Imaging Ontology from/to other Ontologies of
Diseases/Diagnosis.
Thank you!
Matthew Fielding
Barry Smith
Daniel Rubin
RadLex Committee