RESUMO
SNOMED CT is the most comprehensive clinical terminology employed worldwide and enhancing its accuracy is of utmost importance. In this work, we introduce an automated approach to identifying erroneous IS-A relations in SNOMED CT. We first extract linked concept-pairs from which we generate Term Difference Pairs (TDPs) that contain differences between the concepts. Given a TDP, if the reversed TDP also exists and the number of linked-pairs generating this TDP is less than those generating the reversed TDP, then we suggest the former linked-pairs as potentially erroneous IS-A relations. We applied this approach to the Clinical finding and Procedure subhierarchies of the 2022 March US Edition of SNOMED CT, and obtained 52 potentially erroneous IS-A relations and a candidate list of 48 linked-pairs. A domain expert confirmed 41 out of 52 (78.8%) are valid and identified 26 erroneous IS-A relations out of 48 linked-pairs demonstrating the effectiveness of the approach.
RESUMO
Biomedical terminologies play a vital role in managing biomedical data. Missing IS-A relations in a biomedical terminology could be detrimental to its downstream usages. In this paper, we investigate an approach combining logical definitions and lexical features to discover missing IS-A relations in two biomedical terminologies: SNOMED CT and the National Cancer Institute (NCI) thesaurus. The method is applied to unrelated concept-pairs within non-lattice subgraphs: graph fragments within a terminology likely to contain various inconsistencies. Our approach first compares whether the logical definition of a concept is more general than that of the other concept. Then, we check whether the lexical features of the concept are contained in those of the other concept. If both constraints are satisfied, we suggest a potentially missing IS-A relation between the two concepts. The method identified 982 potential missing IS-A relations for SNOMED CT and 100 for NCI thesaurus. In order to assess the efficacy of our approach, a random sample of results belonging to the "Clinical Findings" and "Procedure" subhierarchies of SNOMED CT and results belonging to the "Drug, Food, Chemical or Biomedical Material" subhierarchy of the NCI thesaurus were evaluated by domain experts. The evaluation results revealed that 118 out of 150 suggestions are valid for SNOMED CT and 17 out of 20 are valid for NCI thesaurus.
Assuntos
Systematized Nomenclature of Medicine , Terminologia como Assunto , Vocabulário Controlado , LógicaRESUMO
The Unified Medical Language System (UMLS), a large repository of biomedical vocabularies, has been used for supporting various biomedical applications. Ensuring the quality of the UMLS is critical to maintain both the accuracy of its content and the reliability of downstream applications. In this work, we present a Graph Convolutional Network (GCN)-based approach to identify misaligned synonymous terms organized under different UMLS concepts. We used synonymous terms grouped under the same concept as positive samples and top lexically similar terms as negative samples to train the GCN model. We applied the model to a test set and suggested those negative samples predicted to be synonymous as potentially misaligned synonymous terms. A total of 147,625 suggestions were made. A human expert evaluated 100 randomly selected suggestions and agreed with 60 of them. The results indicate that our GCN-based approach shows promise to help improve the synonymy grouping in the UMLS.