Updated: Feb 1, 97
KAVAS

Project Code:   A2019 Project value:   5509 KECU EC contribution:   3000 KECU No of partners:   5 No of countries:   5 Duration:   36 months

Contact: Mrs. Jytte Brender Medical Informatics Laboratory ApS Stengaards Alle 33D DK-2800 Lyngby, Denmark Tel.: +45-42-88.55.98 Fax: +45-31-75.09.77

Overview

Nevertheless, health care of European citizens will probably improve, when to the expertise of their doctors, one could add the gain of computers. This project aims to investigate the process of knowledge acquisition and data assessment, and to produce a tool apt for assisting physicians in their daily navigation on the ocean of clinical information. The prototype being constructed brings together the methodology of both knowledge elicitation and machine learning techniques. The aspects of human-computer interactions are also addressed, including the control by the user of the level of deepness reached at a certain point of consultation. In the consortium one can find a good combination of software companies, universities and hospitals.

Purpose and objectives

• reduce the data and information generation
• improve the quality of patient management
• reduce the costs in health care service.

The general objective of KAVAS-2 is to develop tools and methodologies that help medical domain experts to develop the protocols that will meet the above mentioned requirements. KAVAS-2 has realised this objective by the development of

• KAVIAR - Knowledge Acquisition, Visualisation, Assessment and Refinement - with which the domain expert can formalise his domain knowledge and derive practical (clinical) classification procedures from data bases. It also provides the means for the assessment of the quality of the derived models
• The refinement and assessment of a methodology for the user-centred evaluation of medical decision support systems
• The analysis and formulation of user requirements for teleconsultancy issues, including the elucidation of cultural aspects of transferability of medical knowledge, information and data?

Results

• Defines a goal for the modelling process
• Selects appropriate tools and techniques, based on the nature of the problem and the goals set in the previous steps
• Applies the selected tools
• Measures and explores different aspects of the quality of the solution(s)
• Validates the results against the defined goals and either reformulates the goals of implicitly or explicitly acknowledges or rejects the solutions.

KAVIAR supports this process by providing tools for goal formulation, explicitation of domain knowledge in the form of conceptual graphs and state transition diagrams, extraction and structuring of knowledge from text sources, the formulation of cost models and quality claims, the development of classification models on the basis of date using either an induction, neural net or density inference tool, assessment of the quality of the derived models.

Evaluation methodology

List of Deliverables

Year 1

• MLE-Induce prototype, 1 (9/P/I)
• System for interfacing to well established data base management systems (12/P/I)
• Epistemological analysis of deep knowledge in selected medical domains (12/R/P)
• UIMS and the global VISUALIZER functionality (12/P/I)

Year 2

• Demonstrator system of the Monitor Functionality (5/P/I)
• MLE-Pruning (6/P/I)
• Data types and scopes in induction of classification trees (12/A/I)
• Ecological Interface Design: Visualisation of domain knowledge & mental strategies based on cognitive analysis of expert tasks (12/R/I)

Year 3

• MLE-Quality and Conditioning (3/P/I)
• Methods and techniques for deep knowledge acquisition and representation (4/R/I)
• Implementation of Visualisation Manager and component graphical mediating techniques for visualisation within the KAVIAR framework (6/P/I)
• The framework for quality assessment of semantic aspects of knowledge (6/R/I)
• A DSS for medical audit in surgery (6/P/I)
• Integrated KAVIAR (8/P/I)
• The Monitor Toolkit (8/P/I)
• Prototype Knowledge Capturer Toolkit (9/P/I)
• Modelling of cultural components of knowledge-based activities in a medical domain
• Prototype KBS for fluid / electrolyte / acid-base metabolism (10/P/I)
• Minimum Requirements Specification for the assessment of tele-medical consultancy service systems (12/R/P)
• Progress in the development of the KAVIAR and its component parts (12/R/I)
• Evaluation of the KAVIAR, - approach, results and conclusion (12/R/I)
• Domain knowledge and pre prototype KBS generated by KAVIAR and/or its components with accompanying evaluation reports (ongoing/O/I)

List of Participants

	Copyright 1997 © EHTO All rights reserved This server is the only official EHTO WWW knowledge repository. Mail suggestions to: webmaster@ehto.org