Medical Ontology Research

National Library Of Medicine

Investigators

Linked publications & trials

Abstract

Biomedical terminologies and ontologies are enabling resources for clinical decision support systems and data integration systems for translational research and health analytics. Therefore, the quality of these resources has a direct impact on healthcare and biomedical research. In the past decade, quality assurance (QA) of biomedical terminologies has become a key issue in the development of standard terminologies and has emerged as an active field of research. Given the large number of existing biomedical terminologies and ontologies, one challenge is interoperability among them, i.e., the possibility of using multiple terminologies and ontologies together for a given task (e.g., clinical analytics). Assessing the interoperability among them and developing methods for identifying correspondences among them (e.g., ontology alignment techniques) is another key objective of our project. Approaches to quality assurance and interoperability include the use of lexical, structural and semantic techniques applied to biomedical terminologies and ontologies, as well as techniques for comparing and contrasting these resources. As part of the Medical Ontology Research project, we explore quality assurance and interoperability issues in a variety of biomedical terminologies including drug terminologies, clinical terminologies, and specialized terminologies, such as HPO (Human Phenotype Ontology) and the Orphanet terminology for rare diseases. About half of our investigations have a primary focus on quality assurance, for which we have developed novel methods. In the other half, we apply existing techniques to assess interoperability among terminologies or some aspect of quality (e.g., coverage) in a terminology. In our work, we put special emphasis on the development of principled, automated, scalable methods, applied systematically to the entire content of a terminology by independent researchers, as opposed to manual review of subsets by domain experts. Finally, we also develop methods for assessing health information standards in the context of specific tasks (e.g., extraction of information from drug labels using natural language processing; clinical analytics), as well as to facilitate the adoption and implementation of these standards (e.g., development of mappings across biomedical terminologies and ontologies; development of application programming interfaces for drug terminologies). Accomplishments for this fiscal year include: - Ontology alignment: Development and evaluation of novel deep-learning methods for assessing synonymy in the UMLS Metathesaurus o Methods leveraging contextual information through knowledge graph embeddings o Application to the insertion of new terms into the UMLS Metathesaurus o Qualitative analysis of the performance of the deep learning algorithms o Deep learning approach to predicting the semantics of UMLS Metathesaurus atoms - Quality assurance: Evaluation of the International Classification of Health Interventions (ICHI) in the coding of common surgical procedures - Clinical analytics: Effect of common maintenance drugs on the risk and severity of COVID-19 in elderly patients