Mining non-lattice subgraphs for detecting missing hierarchical relations and concepts in SNOMED CT

Abstract

Objective: Quality assurance of large ontological systems such as SNOMED CT is an indispensable part of the terminology management lifecycle. We introduce a hybrid structural-lexical method for scalable and systematic discovery of missing hierarchical relations and concepts in SNOMED CT. Material and Methods: All non-lattice subgraphs (the structural part) in SNOMED CT are exhaustively extracted using a scalable MapReduce algorithm. Four lexical patterns (the lexical part) are identified among the extracted non-lattice subgraphs. Non-lattice subgraphs exhibiting such lexical patterns are often indicative of missing hierarchical relations or concepts. Each lexical pattern is associated with a potential specific type of error. Results: Applying the structural-lexical method to SNOMED CT (September 2015 US edition), we found 6801 non-lattice subgraphs that matched these lexical patterns, of which 2046 were amenable to visual inspection. We evaluated a random sample of 100 small subgraphs, of which 59 were reviewed in detail by domain experts. All the subgraphs reviewed contained errors confirmed by the experts. The most frequent type of error was missing is-a relations due to incomplete or inconsistent modeling of the concepts. Conclusions: Our hybrid structural-lexical method is innovative and proved effective not only in detecting errors in SNOMED CT, but also in suggesting remediation for these errors.

A review of auditing techniques for the Unified Medical Language System

“…Hybrid techniques hold promise for better performance. For example, Cui et al 126 combined the structural technique of nonlattice with the natural language processing technique. Others combine structural and lexical techniques.…”

Matching biomedical ontologies based on formal concept analysis

“…DCM, for example, takes co-occurrence patterns as suggestive complex matches, where a large-sized data is required for mining such patterns. Additionally, complex mappings combing multiple classes and properties across ontologies may indicate the absence of a class representing that complex semantic meaning within ontology, thus can be used for quality assurance of large, real-world biomedical ontologies [ 43 ].…”

Auditing SNOMED CT hierarchical relations based on lexical features of concepts in non-lattice subgraphs

“…A lattice is a specific type of DAG such that any two nodes (or concepts) have a unique maximal shared descendant and a unique minimal shared ancestor. A pair of concepts is called a non-lattice pair , if the two concepts have more than one maximal shared common descendant [13, 14, 15]. For example, in Fig.…”

“…1), separately analyzing each such non-lattice pair would be redundant. Therefore, a notion of non-lattice subgraph is further introduced to avoid redundant analysis [15]. Given a non-lattice pair p = ( c 1 , c 2 ) and its maximal common descendants mcd ( p ), the corresponding non-lattice subgraph can be obtained by first computing the minimal common ancestors of the maximal common descendants, mca ( mcd ( p )); then aggregating the concepts and the IS-A edges between (including) any concept in mca ( mcd ( p )) and any concept in mcd ( p ).…”

“…In previous work [15], we introduced a hybrid structural-lexical approach based on the lexical patterns of concept names in non-lattice subgraphs to automatically suggest missing hierarchical relations and concepts in SNOMED CT. However, the predefined lexical patterns only covered 4% of non-lattice subgraphs in SNOMED CT.…”