The Foundational Model of Anatomy ontology (FMA) is OPEN SOURCE and available for general use. To obtain a copy of the FMA database either click the "Get the FMA" link to the left or see: FMA Licenses. You can also investigate the FMA knowledge base using a web browser, the Foundational Model Explorer (FME), which can be found at FME (or by selecting "FM Explorer" from the navigation bar on the left).
What is the FMA Ontology?
The Foundational Model of Anatomy ontology is one of the information
resources integrated in the distributed framework of the Anatomy Information
System developed and maintained by the Structural Informatics
Comparisons with Other
Both atlases and textbooks target a particular user group. Different anatomy atlases and textbooks are published, for example, for nursing and medical students, or surgeons and radiologist, even though the same information is presented in each publication, albeit at different levels of granularity and from different viewpoints. The FMA, by contrast, is designed to provide anatomical information needed by any user group and is intended to accommodate any viewpoint.
Atlases represent anatomical information primarily through annotated images. Although images can be linked to the FMA, the essence of the FMA is that it uses non-graphical symbols (terms and numerical identifiers) for representing anatomical realities that are graphically represented in volumetric data sets, such as the Visible Human, or as images, such as those in atlases of macroscopic and microscopic anatomy.
Anatomy textbooks include non-anatomical information (e.g., physiologic function, pathological lesions, clinical cases) in order to illustrate the relevance of anatomical knowledge to clinical practice. The FMA, by contrast, is strictly constrained to “pure” anatomy, i.e., the structural organization of the body. This means that the modeling is in a deliberately structural context. Non-anatomical information may be associated with the anatomical content of the FMA in applications developed for specific user groups (e.g., medical students).
Dictionaries are term oriented and compile their content in alphabetical order, regardless of the meaning of their terms. Moreover, in a dictionary, different terms are defined in different contexts, i.e., from different viewpoints. The FMA, by contrast, is class or type rather than term oriented (for explanation see Foundational Model Explorer/Conducted Tour/Terms and Concepts) and arranges its classes in an inheritance hierarchy or taxonomy in a strictly structural context. The classes or types of this taxonomy are established on the basis of the structural properties shared by members of a class.
Thesauri organize their content according to the meaning of their terms. However, since these terms are not explicitly defined, the meanings have to be implied by each user on the basis of perceived similarities and differences between terms. The FMA, by contrast explicitly defines the classes of its taxonomy, and links all these classes through an inheritance hierarchy to a single root: Anatomical Entity.
The time-honored term list for anatomy is Terminologia Anatomica (a successor of Nomina Anatomica published more than a hundred years ago). Its purpose is to standardize anatomical terminology, rather than represent knowledge. Its potential and short comings with regard to serving as a basis for a computer-based knowledge source are discussed in Publications/Terminologia. Unlike Terminologia Anatomica, the intent of the FMA is to accomodate all current naming conventions, rather than attempting to standardize terminology. Therefore, the FMA incorporates all of, but is not limited to, the terms of Terminologia Anatomica.
Summary: Textbooks and atlases are similar to computer applications in that they target specific user groups. Dictionaries, thesauri and term lists are intended for general audiences. Of the latter resources only dictionaries represent knowledge explicitly through natural language definitions; any knowledge in thesauri and term lists is implied by the grouping of the terms. All these traditional sources have been conceived of with the printed page in mind as a medium for their dissemination. Although their content may be transferred to electronic media, these traditional sources are unsuitable for supporting machine-based intelligence (i.e., inference). The FMA, by contrast, is a hybrid between these traditional sources of anatomical information: its intent is to encode anatomical knowledge that can be reused for any application to serve the needs of any user group. Moreover, it is qualitatively distinct from traditional sources in that it encodes anatomical knowledge in a way that can support machine-based inference, a requirement for the development of next-generation, “smart” applications in education, clinical practice and research.
1. Anatomy taxonomy (At),
2. Anatomical Structural Abstraction (ASA),
3. Anatomical Transformation Abstraction (ATA),
4. Metaknowledge (Mk),
Thus, the Foundational Model of Anatomy ontology may be represented by the abstraction:
FMA = (At, ASA, ATA, Mk)
The most comprehensive component of the FMA is the Anatomy taxonomy (At). The dominant class in the At is Anatomical Structure. Anatomical structures include all material objects generated by the coordinated expression of groups of the organism’s own structural genes. Thus, they include biological macromolecules, cells and their parts, portions of tissues, organs and their parts, as well as organ systems and body parts (body regions).
Macroscopic anatomical structures are most comprehensively represented, whereas biological molecules have been entered mainly to illustrate the structural continuum from major body parts, such as the thorax, to biological macromolecules, such as myosin.
Portions of body substances, such as blood, CSF, intercellular matrix, and cytoplasm, are defined in terms of their relationship to anatomical structures, and so are spaces, surfaces, lines and points that are associated with anatomical structures.
An objective of Foundational Model Explorer/Conducted Tour is to give an appreciation of the classes that subsume all these diverse classes within a continuous taxonomy.
Traditional sources do not include such a comprehensive taxonomy and do not explicitly define classes or types. Therefore we had to propose and define a number of classesor types of anatomical entities in the FMA for which there is no precedence in the traditional literature. There is an increasing correspondence between the FMA and traditional sources as one proceeds from the root of the At (Anatomical entity) toward the leaf classes of the FMA (e.g., heart, hepatocyte). For example, the FMA contains essentially all the terms of Terminologia Anatomica with an appropriate record of their derivation from this source.
Implementation of FMA Ontology
Protégé represents information about an anatomical class in an object called frame. In the FME, for each class selected in the left pane of the display, the corresponding frame will be displayed in the right pane. Frames that represent an anatomical class include a set of slots, and these slots are filled with one or more values. For example, because physical state is an attribute of Heart, the value for the physical state slot in the frame Heart is solid, whereas for the same slot in the frame Blood, is liquid. The set of slots that constitute the frame of a particular class fully describe or characterize that class.
A key feature of Protégé is the discipline and system it imposes on the configuration of these frames. The frames constitute the mechanism for inheritance, and hence, the creation of the class hierarchy; i.e., the taxonomy. By extension, the frames also serve to support inference and can therefore be used to enforce the economy and consistency of the knowledge entered.
Canonical types (e.g., head, heart, right ventricle, left inguinal canal) are the leaf classes of At. These canonical types may, in turn, serve as classes when application developers wish to associate with the FMA the anatomy of instances or individual subjects (e.g., heart of John Doe). All canonical types must be assigned to a semantic class through the -is a- relationship before ASA and ATA relations of the class can be entered in the model.
We are currently developing the Anatomical Structural Abstraction (ASA) Components of the FMA. These components include a Dimensional Ontology (Do), Boundary Network (Bn), Part-of Network (Pn), and Spatial Association Network (SAn).
Thus, the ASA may be represented by the equation:
ASA = (Do, Bn, Pn, SAn)
Part/whole relations have been populated extensively, though not comprehensively for microscopic entities. We have defined different types of part-of relations (regional, constitutional, systemic, membership), which are described elsewhere. Such attributed part-of relationships have been entered for a substantial number of organs. Similarly, branch-of and tributary-of relationships have been entered for the majority of nerves, arteries, veins and lymphatic vessels. We are in the process of entering various spatial association relationships such as location, containment, and orientation (see for example ‘Esophagus’). We are developing a tool that will provide semi-automatic entry of the numerous adjacency relationships of organs and organ parts (see for example ‘T8 part of esophagus’ in Foundational Model Explorer), thus speed up the population of this relationship.
Essentially very little data have been entered as yet for the Anatomical Transformation Abstraction (ATA), which is designed to model embryonic as well as postnatal development. We have proposed a representation scheme for the ATA in a research grant application.
1. The Human Brain Project, spearheaded by the National Institutes of Mental Health, has identified as one of its goals the development of an ontology for neuroanatomy. The FMA has been considered as a prototype for such an ontology and our preliminary work demonstrates the scalability of the FMA’s semantic structure and implementation schema to the neuroanatomy domain (Publications/NeuroAnatomy-FMA). In the process we have also identified some of the challenges in reconciling differences in representing neuroanatomical views from the different domains of neurosciences and clinical practice.
2. A critical requirement for the validation of animal models of human disease is the verification of anatomical equivalences ranging from genes to cells, tissues, organs, organ systems and body parts. In response to a need by the National Cancer Institue’s consortium for mouse models of human cancer (MMHCC), we have developed, as a demonstration project, symbolic representations of the anatomy of the mouse prostate and mammary gland, mirroring human anatomy in the FMA (Publications/ Mouse-FMA.).
3. An ontology of physics in biology (OPB) is being developed by Dan Cook, John Gennari, Onard Mejino and Adriana Emmi, and is based on the ontological approach used for the FMA.
4. A collaboration with the Mayo Clinic through NCBO for using the Foundational Model of Anatomy ontology as the anatomic basis for clinical indexing and as a bridge to indexing the output of biomedical experiments in order to advance translational research.
5. Under the title ‘Digital Human’, the Federation of American Scientists, in association with NSF and DARPA, is spearheading an initiative for “unifying” biomedical ontologies in order to support the modeling of cellular and higher level physiological processes. This initiative has singled out the Digital Anatomist Foundational Model as the candidate anatomy ontology to serve as a bridge and reference for aligning existing and evolving ontologies, recognizing that the structural context and scope of the FMA make it the best candidate for this role.
7. The FMA has been selected as a candidate ontology to be incorporated into the OBO Foundry, which is a collaborative experiment, involving a group of ontology developers who have agreed to the adoption of a set of principles specifying best practices in ontology development..
8. A number of projects have utilized the FMA as a reference background source for aligning different and disparate anatomical ontologies. Zhang and Bodenreider at NIH have shown that indirect mappings between Adult Mouse Anatomical Dictionary (MA) and the NCI Thesaurus through the FMA is an efficient and reliable method (MA-NCI alignment). Aleksovski, ten Kate and Harmelen matched the anatomy part of CRISP with the anatomy part of MeSH using the FMA ontology as the background knowledge (CRISP-MeSH alignment). Their results showed that indirect matching using a background ontology produced significantly more matches than direct matches.
9. As a reference ontology, the FMA can provide both content and ontological framework to any application that requires for its specific use only a subset of the ontology. This is subset, extract or "view" is called an application ontology. As proof of concept, we have derived from the FMA the ontological framework to reorganize the anatomy axis of RadLex, a terminology project in Radiology.
1. Users who have obtained a copy of the FMA database (FMA License) can use the Protégé knowledge base system to access the FMA. Protégé provides a built in interface for model access as well as an API upon which new applications can be constructed.
2. The Anatomy taxonomy (At) component of the FMA has been incorporated in the National Library of Medicine’s Unified Medical Language System (UMLS) as one of its constituent vocabularies. In UMLS the FMA is known as the University of Washington Digital Anatomist (UWDA) vocabulary. The UWDA consist of the At and selected structural relationships (part-of, branch-of, tributary of). The UWDA can be accessed through the UMLS knowledge server. Current efforts are underway to incorporate the entire and latest version of the FMA into the UMLS.
3. We developed the Foundational Model Explorer (FME) as a web browser, through which we provide access to the FMA. The FME is freely available to the general public for browsing the FMA knowledge base.
4. The FMA can be accessed through a programming language independent (ascii-based) server which processes StruQL queries and returns XML result sets For more information on this access mechanism see: http://sig.biostr.washington.edu/projects/oqafma/
5. The FMA is also available in OWL at http://www.bioontology.org/wiki/index.php/FMAInOwl.
If you would like to obtain a copy of the Foundational Model of Anatomy database you can find licensing information here: FMA License
The Foundational Model of Anatomy ontotogy is being developed as an integral component of the Anatomy Information System through which all projects pursued by the Structural Informatics Group (SIG) are coordinated. Members of the SIG are involved to different extents and in different capacities in the development of the FMA.
Cornelius Rosse, M.D., D.Sc.,
Principal investigator on projects relating to
Jose Leonardo (Onard)
V. Mejino Jr., M.D.,
Richard F. Martin, Ph.D.
Salvacion Nance Quimosing-Madarang, M.D.,
Landon (Todd) Detwiler, B.Sc., MS (Computer
Joshua Daniel Franklin
Former members and contributors:
Kurt L. Rickard, Ph.D.,
Peter Mork, M.S. (Computer Science),
Nathaniel Robinson, B.Sc. (Computer Science),
Ravensara S. (Raven) Travillian, M.A
Adriana Emmi, MD. PhD
Franz Calvo MD
James F. Brinkley,
L Cook, M.D., PhD.,
Mark Musen and the Knowledge Modeling Group from Stanford Center for Biomedical Informatics Research have provided invaluable support in the development of the Foundational Model of Anatomy Ontology. Their knowledge representation expertise and their continued support and improvement of the Protégé knowledge modeling system have contributed significantly to the success of the FMA project.
Barry Smith from the
The development of the Foundational Model of Anatomy ontology has been supported by the National Library of Medicine through research grants LM006822 and LM06316, and research contract LM03528.
Additional support has been received from the