The constraint principle dictates that the foundational model should represent the physical organization of the human body. How well the model can express the structural relationships between the great variety of material objects that constitute the body depends on the class structure of inheritance hierarchies or ontologies, which group together and distinguishes from one another these objects, and on the sets of relationships that specify the associations between them. There are difficulties in deriving a sufficiently expressive classification from anatomy textbooks and term lists. Only in rare instances do the texts imply a classification in terms of structure. For example, bones, muscles and joints are classified according to their shape or constituent parts (e.g., long bones and short bones, unipennate and multipennate muscles, fibrous and synovial joints). However, essentially all other anatomical entities are described sequentially in chapters that are devoted either to organ systems (e.g., cardiovascular system), or to major body parts (so called regions), such as the thorax or upper limb, without reference to a classification system and without explicit definitions. Nomina-Anatomica/Terminologia, the time-honored reference for classifying anatomical concepts, groups together anatomical entities according to the broad physiological processes in which the entities participate, regardless of their structural similarities or differences. This results, for instance, in a closer implied relationship between the liver and the appendix than between the liver and the spleen. For these reasons we found it difficult to formulate a logical ontology that would capture the physical organization of the body as long as we limited ourselves to available sources. Therefore, we proposed an inheritance hierarchy, which includes a number of classes that are new in anatomy knowledge representation. We explicitly defined all classes in accord with the definition principle: “Defining attributes of anatomical entities must be stated in terms of their constituent parts”. This Anatomy Ontology (Ao) is the fundamental component of the Foundational Model. The logical foundation for the Ao is provided by the principle that designates Organ as the organizational unit of macroscopic anatomy, and classifies other structures according to whether they constitute organs or are constituted by organs. Whenever possible we have integrated in this ontology the classification implied by Nomina/Terminologia Anatomica and the textbooks we used for reference. To date, Ao has been instantiated for the thorax, abdomen, pelvis and perineum with over 25,000 anatomical concepts, and data entry continues for the remaining body parts.
Together with its class definitions, the Ao captures a substantial amount of knowledge about the kinds of entities that constitute the body. The result is, that such organs as the liver, lung and spleen are grouped together in the Organ subclass Parenchymatous Viscus. They all satisfy the definition of this subclass:“a viscus, some organ parts of which constitute lobes, segments, lobules, acini, or cortex and medulla”. Stomach, urinary bladder, gall bladder share another set of structural attributes according to which they are assigned to the subclass Hollow Viscus: “a viscus, some organ parts of which constitute a wall; the viscus wall surrounds the organ (viscus) cavity”. Although these two sets of organs are quite unlike one another, nevertheless they share a number of other attributes: all are organs located in the body cavity and the organ parts of all are embryologically derived from endoderm and splanchnic mesoderm. These are some of the defining attributes of Viscus, the parent class of Parenchymatous Viscus and Hol-low Viscus. Although anatomy term lists and texts disregard such aspects of anatomical knowledge, the concepts of viscus, hollow and parenchymatous viscus, are taken for granted by most anatomists. Indeed, they use them in their teaching, although they may not have attempted to define them. Because of their emphasis on concrete leaf concepts and on the differences rather than the similarities between these concepts, traditional sources neglect abstract concepts. We have found, however, that abstract concepts of classes are a requirement for the logical organization of leaf concepts. The instantiation of these classes with many thousands of leaf concepts has provided an initial measure for the formative evaluation of Ao.
The amount of knowledge the foundational model can express is enhanced by its second component, the Anatomical Structural Abstraction, or ASA, which consists of a Spatial ontology (So) and a number of structural networks. A companion re-port in these Proceedings illustrates the cross map-ping of Ao concepts to the classes of So6, which are defined in terms of spatial dimensions and shape. Conflicts that arise as a consequence of such cross mapping are discussed below in reference to specificity. We have instantiated the Part-of Network of ASA for concepts that have been entered in Ao. A logical representation of -part of- relationships is critical for the description of structured objects such as the human body and its parts.
For instance, a part of the right ventricle is designated as the ‘infundibulum’, the synonyms for which include ‘outflow part of right ventricle’. However, neither Nomina/Terminologia Anatomica, nor anatomy textbooks name the remaining and larger part of the ventricle. A rule for instantiating the Part-of Network stipulates that if an entity of a given spatial dimension has one named part, its complementary remaining part or parts must also be named. In order to comprehensively instantiate the Part-of Network of the right ventricle, its part that is not the ‘infundibulum’ must be identified by a name. We have proposed naming this part the ‘in-flow part of the right ventricle’. The ‘inflow part of the right ventricle’ can be distinguished from the infundibulum not only by its function, but also by its structure and anatomical features, as well as by its embryo-logical derivation. Having made this decision, the selection of the preferred term for the ‘infundibulum’ (Latin for funnel) must be reconsidered. Although it conflicts with traditional naming conventions, the logical choice must fall on ‘outflow part of right ventricle’; relegating ‘infundibulum’ to the role of a synonym.
References:
1. Brinkley JF. Enhancements of Anatomical Information in UMLS Knowledge Sources. J Am Med Inform Assoc. AMIA’95 Symp. Suppl. 1995.
2. Rosse C, Gaddum-Rosse P. Hollinshead’s Text-book of Anatomy, 5th ed. Philadelphia:Lippin-cott-Raaven, 1997.
