Quick contact info

Calle Pollensa, 6 2ª Planta, Edificio ECU, 2, 28290 Las Rozas de Madrid, Madrid

+ (34) 916 49 22 92 icon_widget_image info@medlabmg.com

Snomed2Vec: A New Approach To Semantic Search

Ignacio Martínez ha presentado Snomed2Vec en el IEEE CBMS2019

Snomed2Vec is a new approach to semantic search. Backing it is Ignacio Martínez, Data Scientist at MMG, who has applied the Word2Vec algorithm to the SNOMED CT ontology to find similar concepts. The analyst presented his idea at the 32nd International Symposium on Computer-Based Medical Systems (IEEE CBMS2019), held this week, at the Maimonides Institute for Biomedical Research in Cordoba.

The scientific committee of the event accepted his paper in early spring, written in collaboration with Ignacio San Román and Adrián Alonso, members of the Research and Development department of MMG as well as Martínez. In his case, he combines his work in the Spanish start-up with his work in the Rafael Méndez de Lorca Hospital (Murcia). He also collaborates with the universities of Granada and Murcia.

As the Data Scientist explained in its presentation, hospital information systems use electronic health records to store heterogeneous patient data. In most cases, the corpus of the texts is very long. Therefore, it is necessary to standardize them to identify the clinical concepts within it. For this, it is necessary to map each text using ontologies, such as SNOMED-CT. 

What Snomed2Vec provides  

To facilitate the search for concepts, Martínez has designed an Artificial Intelligence tool called Snomed2Vec. Basically, it is based on the application of the Word2Vec algorithm on the SNOMED CT ontology. Right now, this is the most accurate terminology, a codified and multilingual product worldwide. It is a concept recognition tool that suggests which is the most similar to SNOMED-CT that appears in a text.

The tool uses the Word Embedding representation, which is based on the idea that contextual information alone constitutes a viable representation of linguistic terms. Word2Vec is a shallow neural network to produce the representation of words by vectors. Snomed2Vec applies this network to this medical ontology in order to offer a new approach to semantic search, one of MMG’s fields of work.

The 32nd International Symposium on Computer-Based Medical Systems was held in Cordoba between 5th and 7th June. It is the most important event in medical computer systems in the world. Its objective is to give space to high-quality articles on industry and real case applications. In addition, it seeks to allow researchers to show their results to the scientific community.