=Paper=
{{Paper
|id=Vol-1692/abstractJ
|storemode=property
|title=MeTMapS – Medical Terminology Mapping System
|pdfUrl=https://ceur-ws.org/Vol-1692/abstractJ.pdf
|volume=Vol-1692
|authors=Shao Fen Liang,Jean-François Ethier,Talya Porat,Archana Tapuria,Brendan C. Delaney,Vasa Curcin
|dblpUrl=https://dblp.org/rec/conf/odls/LiangPTDC16
}}
==MeTMapS – Medical Terminology Mapping System==
https://ceur-ws.org/Vol-1692/abstractJ.pdf
MeTMapS - Medical Terminology Mapping System
Shao Fen Liang 1,*, Jean-Francois Ethier2, Talya Porat1, Archana Tapuria1, Brendan
C. Delaney 3 and Vasa Curcin 1
1
Division of Health and Social Care Research, King’s College London, London, UK
2
French Institute of Health and Medical Research, Centre de Recherche des Cordeliers, Paris, France
3
Department of Surgery and Cancer, Imperial College London, London, UK
ABSTRACT select a term from the full UMLS collection tree to get results if
Motivation: Medical terminology mapping is a long-standing they have the sufficient knowledge of UMLS. However, the results
challenge for projects requiring retrieval, querying and integra- are displayed in a long list so as to cover all mappings of the
tion of heterogeneous patient data. Current tools fail to fully uti- UMLS collection. The system provides good completion and de-
lise the richness of the underlying coding systems, and can be tailed information but it is also likely to overload users with infor-
difficult to install and maintain. For example, National Library of
mation. LexEVS2 is another system for terminology mapping and
has been used in TRANSFoRm (Ethier, 2013) and BioPortal (Sal-
Medicine’s UMLS provides a rich collection of terminology map-
vadoresa, 2013). However, setting up the LexEVS platform and
ping, however, its search results are displayed in a simplistic
training the users are non-trivial tasks because LexEVS requires
general purpose interface that cannot easily be navigated and users to set up a server and load data into its database so the data
results filtered according to user’s preferences. Specifically, can be transformed into LexEVS data objects. We developed the
returned results cannot be visualised in a tree to show positions Medical Terminology Mapping System (MeTMapS) with the aim
and relationships. BioPortal offers a large number of terminolo- of addressing these concerns and producing a usable system for
gies and ontologies, each of which can be viewed in a tree struc- clinical researchers.
ture, however it does not allow for multiple ontologies to be
viewed and compared on a single page. Our work aims to ad- 2 METHOD
dress these issues and provide a simple and easy to use termi-
In the construction of MeTMapS (see Figure 1) we have utilised
nology mapping software.
UTS APIs and BioPortal Widgets, some of which have been modi-
Results: MeTMapS was evaluated with academic and clinical fied to achieve better performance, e.g. the tree widget used to
research users. The users have tested the mapping between display hierarchical terminologies. Any terminology not covered
ICD10, Read CTV2, V3 in Hypertension. It was also tested on a by UMLS can be added into MeTMapS via BioPortal, if the paired
list of clinical terms from the inclusion and exclusion criteria of mapping file and ontology are ready.
the INFORM clinical trial protocol. Our initial evaluation pro-
duced positive results.
Availability: We are currently in the process of updating the
design based on some improvements suggested by the partici-
pants. MeTMapS is developed under Apache V2 license and is
currently hosted at KCL for internal use and will shortly be
opened to the public once the internal security concerns are
resolved. In the meantime, the tool is available from the author
upon request.
* Contact: fennie.liang@kcl.ac.uk
1 INTRODUCTION Fig. 1. MeTMapS architecture shows a user inputs a search term
to the system. The system requests the term from UMLS and gets a
Medical terminology mapping is a long-standing challenge when CUI returned. The relevant ontology, which contains the search
designing observational studies from Electronic Health Records, term is then requested from BioPortal and returned to the system.
particularly when working with multiple databases employing The returned ontology and the mapped results can then be viewed
different coding systems. Mapping from one terminology to anoth- by the user.
er can rarely be done automatically due to many-to-many map-
pings that frequently occur between terminologies, and it typically
requires a user with medical knowledge and a good understanding
For example, we have generated a Read CTV2 (used by most of
of each terminology to manage the cardinality issues. This is one of
the primary care systems in the UK) ontology and uploaded it into
the reasons that the US National Library of Medicine has built the
BioPortal. We have also produced a mapping file, which contained
UMLS Metathesaurus with over a hundred national and interna-
Concept Unique Identifiers (CUI) from UMLS for MeTMapS.
tional terminologies in different languages and their mappings. The
UTS Metathesaurus browser1 is provided as an interface for navi-
gating the mappings. Users can either simply search for a term or
1 2
https://uts.nlm.nih.gov///metathesaurus.html https://wiki.nci.nih.gov/display/LexEVS/LexEVS
1
Liang et al
The idea behind the MeTMapS user interface (Figure 2) is to focus contain the same CUI as Hypertensive disorder from UMLS). The
the users only on the terminologies they need. The search results correct term always appears first with the most relevant results
are organised into tree structures, offering a better view of relation- being at the top, and suggestions provided while typing. The search
ships among the selected terms, their parents, siblings and de- terms are not order- or case-sensitive and handles terms separated
scendants. To minimise the setup needed and the learning curve for by hyphen. (For example, search term ‘sugarfree’ includes ‘sugar-
the users, the system is designed as a web application with a step- free’ and search term ‘breast-cancer includes ‘breast cancer’) It
by-step workflow guide. The interface comprises three main sec- handles exact term matches for different concepts such as ‘Fundus
tions: 1) Search of terms and selection of coding systems; 2) coloboma’ and ‘Gastric fundus structure’ and multiple terminolo-
browsing of results and selection of mappings from structured gies can be simultaneously selected for mapping. We have found
trees; and 3) removal of unwanted mappings and storage of results that MeTMapS could be improved by providing suggestions on
for further use. Each section has a guide for users to follow, requir- mis-spelled terms, and also enhanced on handling partial words
ing no previous training. With all relevant information displayed such as ‘Myo inf’ as opposed to ‘myocardial infarction’ and on
on one page, it facilitates navigation and orientation. Auto- handling known synonyms such as ‘kidney’ and ‘renal’.
completion is also provided to help users quickly find their desired
term.
ACKNOWLEDGEMENT
This research is supported by the National Institute for Health Re-
search (NIHR) Biomedical Research Centre at Guy’s and St
3 RESULTS Thomas’ NHS Foundation Trust and King’s College London.
The MeTMapS system was evaluated by seven academic and clini-
cal research users including GPs, clinical informaticians and IT
REFERENCES
specialists at KCL in the last two months. A total of 54 sessions
were completed, where the system was used by Ethier, J-Fo and Dameron, O. et al. (2013) A unified structur-
the researchers. Participants were asked to map the required clini- al/terminological interoperability framework based on LexEVS:
cal terms in the 'Hypertension' clinical domain either from ICD10 application to TRANSFoRm. Journal of the American Medical
to Read CTV2 or from ICD10 to Read CTV2 and V3. Participants Informatics Association, 2013(0) 1-9.
also tested a list of clinical terms extracted from the inclusion and Salvadoresa M. and Alexandera PR. Et al. (2013) BioPortal as a
exclusion criteria of the INFORM clinical trial protocol (Wil- Dataset of Linked Biomedical Ontologies and Terminologies in
kinson, 2016). The clinical terms listed were mapped to the clinical RDF. Semantic Web, 4(3) 277-84.
terminologies Read CTV2 and V3 that are used by most of the GP Wilkinson Ian (2016) Randomised cross-over trial in a multi-ethnic
systems recruiting patients for the trial. A screen shot of a search cohort: AIM HY – INFORM. Work Strand 3,
for Hypertensive disorder and the mapped results from Read http://www.aimhy.org.uk/our-research/ws3/.
CTV2, Read V3 and ICD10 is shown in Figure 2 (the listed results
Fig. 2. MeTMapS Interface
2