Anne-Marie Rassinoux

GALEN: a third generation terminology tool to support a multipurpose national coding system for surgical procedures

GALEN has developed a new generation of terminology tools based on a language independent concept reference model using a compositional formalism allowing computer processing and multiple reuses. During the 4th framework program project Galen-In-Use we applied the modelling and the tools to the development of a new multipurpose coding system for surgical procedures (CCAM) in France. On one hand we contributed to a language independent knowledge repository for multicultural Europe. On the other hand we support the traditional process for creating a new coding system in medicine which is very much labour consuming by artificial intelligence tools using a medically oriented recursive ontology and natural language processing. We used an integrated software named CLAW to process French professional medical language rubrics produced by the national colleges of surgeons into intermediate dissections and to the Grail reference ontology model representation. From this language independent concept model representation on one hand we generate controlled French natural language to support the finalization of the linguistic labels in relation with the meanings of the conceptual system structure. On the other hand the classification manager of third generation proves to be very powerful to retrieve the initial professional rubrics with different categories of concepts within a semantic network.

xml as standard for communicating in a document-based electronic patient record: a 3 years experiment

During the past few years, the eXtensible Markup Language (XML) has progressively become a gold standard for accessing, representing and exchanging information, especially in the health care environment. This paper presents an implementation of the use of XML for the electronic patient record (EPR) and discusses more specifically its growing use in two areas of the EPR: first, as a format for the exchange of structured messages, and second, as a comprehensible way of representing patient documents. These statements rely on a 3 years experiment conducted at the Geneva University Hospital as part of its document-centered EPR.

A Multilingual Analyser ofMedical Texts

In the European Union, the need for systems which are able to accept multiple European languages is of paramount interest, because language barriers can be a strong impediment for large-scale communication in Europe. The use of analysers able to accept different European languages and convert them into a single representation common to all languages would seem to be the ideal solution. The RECIT system presented in this paper, shows an original approach for analysing sentences, understanding their meaning and storing them into a deep representation, available for future querying. The chosen approach, called proximity processing, takes advantage of the typical situation of a closed domain of knowledge (i.e. medicine) and of the structured form of medical reports (discharge summaries), using proximity rules which combine in an integrated way semantic information as well as syntactic information when needed. From the recognition of meaningful components in free text sentences, a knowledge representation is built in the form of conceptual graphs. In this article, we discuss the relevant features of both proximity processing and the subsequent transformation into a language-independent representation. In particular, we highlight the characteristics that enable our system to be easily extended to other European languages, as well as other application domains when pertinent.

Knowledge Representation of Discharge Summaries

A Natural Language Processing system for discharge summaries has been implemented, and gives promising results in the restricted domain of Digestive Surgery. Syntactical analysis is strongly driven by semantical information according to the idea of “proximity processing”. From the recognition of semantical components in the free text sentences, a knowledge representation is built in the form of concepts and relationships as defined by Sowa and his conceptual graphs. The concepts are retrieved from the text according to predefined conceptual graphs, to be considered, at a first glance, as a sort of knowledge frame, specifying the mandatory and allowable relationships to other concepts, as well as default values necessary to solve possible ambiguities. The final knowledge representation is independent from the initial text, even when synonym expressions are present, in which case they are mapped into a unique representation.

An ontology driven collaborative development for biomedical terminologies: From the French CCAM to the Australian ICHI coding system

The CCAM French coding system of clinical procedures was developed between 1994 and 2004 using, in parallel, a traditional domain experts consensus method on one hand, and advanced methodologies of ontology driven semantic representation and multilingual generation on the other hand. These advanced methodologies were applied under the framework of an European Union collaborative research project named GALEN and produced a new generation of biomedical terminology. Following the interest in several countries and in WHO, the GALEN network has tested the application of the ontology driven tools to the existing reduced Australian ICHI coding system for interventions presently under investigation by WHO to check its ability and appropriateness to become the reference international coding system for procedures. The initial results are presented and discussed in terms of feasibility and quality assurance for sharing and maintaining consistent medical knowledge and allowing diversity in linguistic expressiveness of end users.

Building Medical Dictionaries for Patient Encoding Systems: A Methodology

One of the most critical problems of automatic natural language processing (NLP) is the size of the medical dictionaries. The set of compound medical words and the often used possibility to create new terms render the exhaustivity of medical dictionaries beyond question. The structure of such dictionaries is usually composed of two parts: the first one generally contains morphological and sometimes syntactical information necessary to identify, on a grapheme level, a given word in a sentence whereas the second part is often devoted to conceptual knowledge associated with the recognised word. It is only when these two prerequisites are fulfilled that an attempt to understand the meaning of a whole expression is possible. The approach developed in this paper shows the pragmatic method used to implement a powerful analyser dedicated to help physicians or coding clerks to encode medico-economic information about patients using international classifications like ICD. It describes how to build medical dictionaries that can help the application of morphological and conceptual analysers (encoders). The methods used have proved to be efficient for various classifications as well as for multiple languages as the system presently supports French, German, English and Dutch for the full ICD-10 classification.

Tuning up conceptual graph representation for multilingual natural language processing in medicine

Multilingual natural language processing (NLP), whether it concerns analysis or generation of sentences, requires a sound language-independent representation for grasping the deep meaning of narratives. The formalism of conceptual graphs (CGs), especially designed to cope with natural language semantics, constitutes a good repository for dealing with the compositionality and intricacies of medical language. This paper describes our experiment, as part of the European GALEN project, for exploiting a conceptual graph representation of medical language, upon which multilingual medical language processing is performed.

Analysis of Medical Jargon: The RECIT System

Medical language constitutes a large subset of human language. However, it presents special features which have to be taken into consideration for natural language analysis. We have observed that natural language texts dealing with a specific medical area (especially discharge summaries and reports) share a common vocabulary and common habits of word usage. Facing these specificities, we describe in this paper an original system called RECIT, designed for medical text analysis and understanding. In particular, we discuss the principles which guide the analysis of medical jargon using semantic considerations, coupled with syntactic information when needed. As a result of the flexibility of the system, a medical application developed in the context of the AIM project HELIOS-2 is presented.

Using Part-of-Speech and Word-Sense Disambiguation for Boosting String-Edit Distance Spelling Correction

We report on the design of a system for correcting spelling errors resulting in non-existent words. The system aims at improving edition of medical reports. Unlike traditional systems, both semantic and syntactic contexts are considered here. The system is organized along three steps. The first module is based on a context independent string-to-string edit distance calculus. The second module, based on the morpho-syntactic context attempts to rank more relevantly the data set provided by the first module, finally a third contextual module processes words with the same part-of-speech by applying some contextual word-sense disambiguation. Modules 2 and 3 are using both hand written rules and data-driven Markovian matrices. A final evaluation shows a significant improvement compared to context-free spelling correction.