Laboratoire d'Informatique Médicale et d'Ingénierie des Connaissances en e-Santé


Logo UPMC Logo université Paris 13


Image décorative

Optimizing semantic ingegration of healthcare data thanks to multilingual vocabulary resources and collaborative matching services

Nicolas Paris

Le 25/03/2019 de 10:10 à 11:00

Description :

Background: In hospitals worldwide, local medical dataset are built. Common Data Models are emerging and enable to reproduce medical studies, algorithms, and compare practices. However care sites often rely on different terminologies and few methods and tools are efficient to federate those heterogeneous hospital informations.

Methods: OMOP was chosen as a common data model to support multilingual interoperability. The deidentified US MIMIC database was transformed into OMOP. The french AP-HP database is also being transformed. A pipeline was built on top of OMOP for french notes deidentication. A collaborative web application was built on top of OMOP for international terminology mapping. It takes advantages of bilingual synonyms, translations and statistical values derived from the patient data itself.
Results: The transformation of MIMIC into OMOP is over and its quality is high. The deidentication pipeline performances are in the state of the art and enable ethical medical research. The terminology mapping tool outperforms the other tested tools (RELMA & USAGI).

Conclusion: OMOP provides a medical data framework which is being extended to faster data standardization. Upcoming work on the terminology-mapping tool includes the integration of bi-lingual informations derived from the note content of both US and FR databases.

Traitement en cours ...