ICD-11 · BFO ontology

BFO-aligned formal ontology · generated by BFO-Agent

A BFO Ontology of ICD-11

The disease categories of the WHO's International Classification of Diseases, 11th revision, rendered as a formal ontology, typed under the Basic Formal Ontology, and measured for logical coherence.

5,281
classes
165
individuals
43,290
triples
Consistent
HermiT model exists
87
unsatisfiable classes

What this is

This is a formal ontology of the disease categories named in ICD-11. Each category is expressed as an OWL class, typed under the Basic Formal Ontology as a disposition, a process, a material entity, a quality, and so on, and related to the others through BFO and Relation Ontology relations. It was produced by BFO-Agent, an automated pipeline that reads a source text and emits a BFO-grounded ontology, then validates it with the HermiT reasoner.

It is one of a family of BFO-aligned artifacts hosted under this domain, alongside ontologies of legal structure at sool and of philosophical texts at spinoza, leibniz, and geometry of the good. Unlike those, its distinguishing property is its measured coherence: the reasoner reports a consistent ontology in which 87 named classes are nonetheless unsatisfiable.

Browse the ontology   Download OWL

On coherence, read honestly

Consistency and coherence are different properties. The ontology is consistent: a logical model exists. It is not coherent: 87 named classes are equivalent to owl:Nothing and can have no instances (the mark throughout the viewer flags them).

These 87 are translation artifacts of this draft extraction, not claims that ICD-11 is logically incoherent. The dominant cause is a recurring modeling slip in which a realizable — a role or a disposition — is attached to a class as if the class bore it, colliding with a BFO disjointness axiom and collapsing the class. A vitamin, a bacterial species, and an organelle are among the classes affected, which no reading of ICD-11 would call impossible. The About page names the mechanism, and the defect is under repair in the pipeline.

Contents

  1. i

    Browser

    Interactive viewer with a BFO category tree, full-text search, per-class restrictions and ancestry, and a coherence view that groups the 87 unsatisfiable classes by cascade.

  2. ii

    Coherence view

    The unsatisfiable classes, grouped by the root they cascade from: autoimmune, streptococcal, myocarditis, ocular, pain, perinatal, and isolated collapses.

  3. iii

    About & method

    How the ontology was generated and validated, what the coherence numbers mean, and the failure mode behind the unsatisfiable classes.

  4. iv

    Download

    The OWL artifact itself. Inspect with Protégé, query with rdflib, reason with HermiT.

What this is not

This is not the ICD-11 itself, nor an official WHO product, nor a clinical tool. ICD-11 is maintained by the World Health Organization; this is an independent formal rendering of its category structure, produced automatically and offered for research into ontology extraction and the measurement of classificatory coherence.

It is a draft. The reasoner checked formal consistency, not fidelity to ICD-11 or clinical defensibility, and the unsatisfiable classes show that the extraction is not yet coherent. Many formalization choices could be made differently. The artifact is open to inspection, criticism, and replacement by better renderings.

Acknowledgements

The Basic Formal Ontology, on which this work depends, was developed by Barry Smith and colleagues and is documented in Building Ontologies with Basic Formal Ontology (Arp, Smith, and Spear, MIT Press, 2015). The HermiT reasoner is the work of Birte Glimm, Ian Horrocks, Boris Motik, Giorgos Stoilos, and Zhe Wang. ICD-11 is a product of the World Health Organization.

The extraction and validation pipeline is BFO-Agent, an architecture by the author. Source at github.com/dkoepsell/bfo-agent. The ontology and this site are released under CC-BY 4.0.