Up: Pierre Zweigenbaum's Home Page
An Access System for Medical Records using Natural Language
European Programme
``Telematics Systems of General Interest''
Advanced Informatics in Medicine
A2023, jan 1992-mar 1995
Experience of existing Hospital Informations Systems (HISs) highlights
the fact that most relevant medical information is stored in narrative
form, in Patient Discharge Summaries (PDSs). These PDSs aim at the
transmission of the minimal but sufficient set of data to be used for
the next visit of the patient.
The two main goals MENELAS contributes to are to
(i) Provide better account of and better access to medical information
(PDSs) through natural languages in order to help
physicians in their daily practice, and to
(ii) Enhance European cooperation by multilingual access to standardised
medical nomenclatures.
The major achievements of MENELAS are the realisation of its two
functional systems:
- The Document Indexing System
-
encodes free text PDSs into both an internal
representation (a set of Conceptual Graphs) and
international nomenclature codes
(ICD-9-CM). Instances of the Document Indexing System have been
realised for French,
English and Dutch.
- The Consultation System
- allows users to
access the information contained in PDSs previously indexed by the
Document Indexing System.
The test domain for the project was coronary diseases.
The existing prototype shows promising
results for information retrieval from natural
language PDSs and for automatically encoding
PDSs into an existing classification such as
ICD-9-CM. Providing a better access to patient medical information
has an obvious impact on the quality of health care. Facilitating the
production of a normalised encoding of this information from national
languages is a key enabling factor for harmonisation across Europe.
A set of components, tools, knowledge bases and methods has also been
produced by the project. These include:
- Language analysis components
- for French, English and Dutch;
language-generic pragmatic analyser; conceptual graph toolbox.
- Index generator component
- parameterised
for ICD-9-CM.
- Information retrieval components
- including a User Interface.
- Linguistic knowledge bases:
- domain-specific syntactic and
semantic lexicons for French, English and Dutch.
- Medical knowledge bases:
- language-independent ontology
and models for the domain of coronary diseases;
conceptual description of the relevant ICD-9-CM
codes.
- A methodology
- for developing ontologies.
- System administrator tools
- to help modify the linguistic and
medical knowledge bases.
The following demonstrators were presented on the MENELAS stand
at the AIM Final Conference (Lisbon, Dec 1994):
- The Document Indexing Systems
for French, English and Dutch. Given a PDS in the corresponding
language, the Document Indexing System proposes a set of ICD-9-CM
nomenclature codes for the diagnoses and procedures mentioned in the
text. It also builds a more elaborate, canonical, internal
representation of the contents of the text. This representation is
generic, i.e., it is abstracted away from the source language.
- The Consultation System. Through a graphical User
Interface, this system allows a user to formulate a query and
retrieve the PDSs that answer this query.
Retrieval is performed on the generic representation of the PDS
contents.
- The domain ontology and knowledge bases can be browsed through
the graphical Knowledge Management Tool.
The pilot system will offer the following services:
- Faced with an atypical patient, clinicians can search the stored
reports for patients with similar characteristics, and find the
treatments they received [diagnosis aid].
- Nomenclature codes are automatically produced for each report,
and can be included in standardised discharge summaries
[medico-administrative management].
- Public health researchers can test clinical research hypotheses
on the basis of the available patient sample [clinical research and
public health].
The project gives primacy to the user of the technology. MENELAS
lets the user keep with usual practice (writing PDSs): the PDSs are
produced in a natural way using medical language, no manual encoding
of the data is imposed on the health care professionals.
In the short term (1994-5), a first reuse of the MENELAS prototype
took place at Lille, in Northern France, within the framework of
the AIM integration project ISAR. Apart from this, a core
of project partners, in collaboration with partners from other
consortia of both the health care (AIM) and the language
engineering (LRE, MLAP) Telematics Community programmes, have
jointly launched a feasibility study, including a market analysis,
which identified how MENELAS-like systems can be integrated into
the electronic document processing flow of hospitals (MLAP
project DOME).
- Main Project publications:
-
Edited Final Project Report
Final Report
See also http://www.biomath.jussieu.fr/~pz/Publications/biblio-menelas/.
- U194
- INSERM U.194, Paris, F
- TCD
- Trinity College Dublin, Dublin, IRL
- IMS
- Irish Medical Systems, Dublin, IRL
- KUL
- Katholieke Universiteit Leuven (see also
Peter Spyns' home page), Leuven, B
- ERLI
- GSI/ERLI, Charenton, F
- IBM
- IBM Paris Scientific Center, Paris, F
- LTG
- Human Communication Research Center, Edinburgh, UK
- GLIM (assoc. to U194)
-
Groupe Linguistique, Informatique et Médecine, Montpellier, F
- CERTIM (assoc. to U194)
-
C.E.R. Traitement de l'Information Médicale, Marseille, F
- LIM (assoc. to U194)
-
Laboratoire d'Informatique Médicale, Rennes, F
- SIM (assoc. to U194)
-
Service d'Informatique Médicale, AP-HP, Paris, F
- RVH (subcont. to IMS)
-
Royal Victoria Hospital, Belfast, UK
Pierre Zweigenbaum
DIAM -- SIM/AP-HP
91, boulevard de l'Hôpital
F-75634 Paris Cedex 13
France
Tel.: +33 1 45 83 67 28
Fax: +33 1 45 86 80 68
e-mail: pz@biomath.jussieu.fr
WWW: http://www.biomath.jussieu.fr/~pz/
Up: Pierre Zweigenbaum's Home Page
Pierre Zweigenbaum (ed.)
MENELAS Home Page
(http://www.biomath.jussieu.fr/~pz/Menelas/)