Foundational Model of Anatomy

The Foundational Model of Anatomy, usually abbreviated FMA, is a reference ontology for human anatomy that represents anatomical entities and their spatial-structural relations in machine-readable form.

It was developed by the Structural Informatics Group at the University of Washington and has become a standard resource for anatomical terminology and knowledge representation.

The FMA is not a dictionary or a textbook but a formal ontology, meaning that its classes and relations are defined by logical axioms that a computer can reason with.

Definition and scope

• An ontology is a formal, explicit specification of a shared conceptualisation, and the FMA applies this idea to the domain of human anatomy.
• The FMA represents the canonical, phenotypic structure of the human body at multiple levels of granularity, from molecules and organelles to organs and body parts.
• Its scope is human anatomy rather than comparative anatomy, so it does not attempt to represent the anatomy of other species in the same detail.
• The ontology includes both anatomical entities and the relations that connect them, such as part-of, branch-of and tributary-of.
• FMA identifiers are used as stable references, and the FMA ID for an anatomical structure can be cited independently of its English or Latin name.
• The FMA is one of the reference ontologies used by the Open Biomedical Ontologies community, and it is aligned with several other anatomical ontologies.

Size and structure

• The FMA contains approximately 75,000 classes and more than 120,000 terms, which makes it one of the largest anatomical ontologies ever built.
• It contains more than 2.1 million relationship instances drawn from more than 168 relationship types.
• The relationships are not limited to part-of, because the FMA also models spatial adjacency, branching, supply and drainage.
• The ontology is organised around a small set of upper-level categories, including anatomical structure, body part, organ and cell.
• Each class can carry synonyms, definitions and references, and these annotations make the ontology usable for text mining as well as for reasoning.
• The distinction between anatomical structure and body part allows the same entity to be modelled at different granularities without duplication.
• The class hierarchy supports deductive reasoning, so a query for an organ can retrieve its subtypes even when no subtype is named in the query.
• The FMA is distributed through web browsers and through ontology repositories such as BioPortal, which provide programmatic access and SPARQL endpoints.
• Its canonical labels and definitions are in English, and multilingual use depends on mappings rather than on native translations.

Applications

• In biomedical informatics, the FMA provides the anatomical layer for electronic health records and clinical decision support.
• In medical imaging, FMA terms can annotate structures in computed tomography, magnetic resonance and histology images.
• In natural language processing, the FMA supplies synonyms and relations that help extract anatomical mentions from clinical text.
• In surgical planning, FMA-based models can link preoperative images to anatomical knowledge and to expected resection margins.
• In education, the ontology supports anatomy browsers and question banks that need consistent structure names and relations.
• In artificial intelligence, FMA is used as a knowledge graph for reasoning about anatomy and for grounding language models.
• FMA terms are used in electronic health records to encode organ location, laterality and surgical site when a controlled anatomical vocabulary is required.
• Knowledge graphs that combine FMA with genomics and imaging can support queries that move from organ to cell to molecule.

Relationship to other standards

• Terminologia Anatomica is a terminology standard, while the FMA is an ontology, so the two serve different purposes even when they share names.
• UBERON is a cross-species anatomy ontology, and it imports or aligns with FMA terms for human anatomy.
• MeSH is an indexing vocabulary for biomedical literature, and FMA identifiers are often used alongside MeSH terms in annotations.
• SNOMED CT includes anatomical concepts and has been mapped to FMA in several projects, although the mappings are incomplete.
• The FMA influenced the design of other anatomical ontologies, and its relation types are reused in several biomedical ontologies.
• Mapping FMA to clinical terminologies is labour-intensive, and the maps must be maintained as both ontologies evolve.

History

• The FMA was initiated in the 1990s by Cornelius Rosse and colleagues at the University of Washington, who wanted a principled representation of anatomy.
• The ontology grew out of earlier work on anatomical knowledge bases and on the digital anatomist project.
• By the 2000s the FMA had become a reference resource, and its identifiers appeared in imaging, informatics and anatomy education systems.
• In the 2010s the FMA was aligned with UBERON and other OBO ontologies, which improved cross-species interoperability.
• Maintenance of the FMA has depended on grant funding and on a small team, and this fragility is a recurring concern for its users.

Research and controversies

• A central criticism is that the FMA is human-specific, so it cannot represent comparative anatomy without extension or alignment.
• The granularity of the ontology is uneven, with some regions modelled in great detail and others represented only superficially.
• Whether an ontology should model canonical anatomy or population variation is a philosophical and practical tension in the FMA.
• The FMA's treatment of anatomical variation is limited, so patient-specific anatomy is usually represented outside the ontology.
• Version control and provenance are essential for clinical use, because a change in definition can alter decision-support outputs.
• Sustaining a large ontology requires governance, versioning and community input, and these processes are less developed than the ontology itself.
• Quality assurance relies on expert review, logical consistency checking and comparison with Terminologia Anatomica.
• The FMA is distributed under an open licence for research and education, although commercial use may require separate permission.
• Its identifiers are persistent in principle, but revisions can change class membership, so users should record the release version.
• As anatomy becomes more computational, the FMA is likely to remain a backbone for anatomical knowledge graphs even when other ontologies handle specific tasks.
• Its development illustrates the tension between expressive logic and practical usability, a tension that continues in biomedical ontology design.
• For a student, the practical value of the FMA is the discipline of asking what a structure is, what it is part of and how it relates to its neighbours.
• The FMA is therefore best understood as a tool for reasoning and integration rather than as a complete description of the body.
• Its continuing influence is visible in the many anatomical ontologies that reuse its relation types and identifier patterns.
• Periodic releases incorporate corrections, new relations and alignment updates, so the ontology is maintained rather than frozen.
• The FMA thus illustrates both the promise and the cost of formalising anatomical knowledge.