Foundations and Advancement of Subontology Generation for Clinically Relevant Information

Project description

A critical tool for ensuring consistent data entry into electronic health care records and supporting meaningful data sharing and analysis of clinical information is the medical ontology SNOMED CT. SNOMED CT captures the agreed intended meaning of more than 300,000 medical codes and concepts over 19 major topics, covering a wide range of information from diseases, causes of diseases, medical procedures, descriptions of the anatomy of the human body and doctor's observations. Since not all this information is needed in every use case, in collaboration with SNOMED International, researchers in the Department have developed bespoke subontology generation software to extract definitions of user-specified concepts from SNOMED CT and create a reusable stand-alone subontology. This equips clinicians, domain experts, system suppliers and other users with means to create subontologies with a focus to specific specialities, e.g., heart diseases or dentistry.

As ability to create stand-alone subontologies is a new functionality and a holy-grail in the area, many research questions remain to be investigated. For instance:
(i) investigation of the foundations of subontologies to characterise their mathematical properties, to gain understanding into knowledge preservation properties in relation to the source ontology (what types of information are preserved, what are not), comparison and combination with other content and definition extractions methods (modularisation, forgetting, uniform interpolation, Craig interpolation), combination with logical difference computation and establishing the computational complexity of computing subontologies,
(ii) Exploitation and fine-tuning of subontologies to support different applications and particular tasks using SNOMED CT, e.g., data entry, data analysis, where currently flat lists of concepts (refsets) are used.
(iii) Sharing of content between different national extensions of SNOMED CT, sharing of content with other medical ontologies, and management of versioning of subontologies.
(iv) Content aggregation and creation of a superontology with a specific focus using subontologies from different medical ontologies as building blocks, or other use cases.

This describes a broad research programme which can accommodate several projects with different emphases. For instance, the research may be focussed on theoretical work, it may be focussed on enhancing the technology for different tasks or ontology languages, it may be focussed on developing and evaluating new use cases, or (most likely) study and develop a combination of some of these aspects.

Expected knowledge and skills:
(a) Background and keen interest in as many of these topics as possible: ontologies, classical logic, description or modal logic, automated reasoning, knowledge representation and reasoning, symbolic AI, clinical information and applications.
(b) Excellent mathematical/analytical skills.
(c) Programming skills to adapt and extend the subontology generation software and the SNOMED CT subontology browser, to use these in conjunctions with Protege, the OWL API and other tools, and carry out evaluations.
(d) Excellent communication and technical writing skills.
(e) Willingness to interact with clinical domain experts, SNOMED CT users, modellers and technical architects.