The decision support system for telemedicine based on multiple expertise
Introduction
The rapid progress in telecommunications, the Internet, WWW and other telemedicine-related topics has afforded new opportunities to collect medical expertise from several knowledge sources to support medical decision-making. This has raised new and interesting problems such as: how to collect different opinions from knowledge sources; how to handle inconsistent and incomplete knowledge; how to find consensus between different opinions and how to support the interface between individual and collective knowledge 2, 7, 9, 13, 14.
We have grouped the research goals behind this article into the following four groups:
- 1.
Development of methods and intelligent software to support medical expert knowledge acquisition and representation
- ◦
acquisition of patterns from a source of medical knowledge,
- ◦
representing medical patterns and
- ◦
processing the presented medical patterns.
- ◦
- 2.
Development of methods and tools to support decision-making using multilevel and multicriteria statistical processing of medical data and biosignals:
- ◦
development of multilevel, recursive mathematical models for representation and processing of quasi-periodical biosignals and
- ◦
development of methods and tools for decision-making based on a metalevel classification technique with multiple statistical methods and multilevel representation of medical data.
- ◦
- 3.
Support for cooperative decision-making in medical diagnostics with multiple experts:
- ◦
the development of tools to represent different methods of medical diagnostics,
- ◦
the development of voting-type techniques for computer-based cooperative decision-making and
- ◦
the development of tools to derive the diagnosis from multiple opinions of experts-physicians.
- ◦
- 4.
The development of applications based on the research results obtained:
- ◦
the development of an expert system for telediagnostics and
- ◦
the development of an intelligent system to support medical teleconsulting between multiple experts-physicians.
- ◦
In Section 2we discuss multilevel signal processing and in Section 3, methods and tools for multilevel processing of medical data using statistical analysis. Section 4discusses briefly a medical knowledge refinement technique. In Section 5we present a common structure of an expert system for telediagnostics. Section 6briefly discusses a system that supports medical teleconsulting and we end with some conclusions in Section 7.
Section snippets
Multilevel signal processing
Multilevel signal processing is composed of several levels of signal processing. Each level includes first a semantic analysis of the signal, breaking it into segments and then some statistical processing of those segments. The goal of this processing is to acquire the most essential features of the signal in each level using some semantic property as a basis for segmentation. Multilevel signal processing can be described in the following way.
Let χ(t) be a signal defined in the time interval [t1
A method for selecting a statistical method
Real-world medical data includes many cases with typically multiple attributes forming a huge amount of dimensions. It is often so that the separate areas of the data space also require separate processing methods. Following this, the practising physician has to decide which statistical method to use when he receives a new case and tries to find previous similar ones. In order to help a physician with a minimal level of statistical background to properly use the automated diagnostics system, an
The medical knowledge refinement technique
We have used the voting-type technique 10, 11 to refine knowledge obtained from multiple knowledge sources. In this chapter we discuss shortly how it can be applied also in the medical diagnostics area.
Let us suppose that medical knowledge is represented using predicates that define relationships within three sets: domain objects (objects of diagnosis), medical concepts used to describe the domain objects and knowledge sources i.e. persons who give their statement using the medical concepts to
The telediagnostics expert system
We plan to apply the above methods in the development of two telemedicine systems. The first one is an expert system for telediagnostics (TELEMED) [6]. An example of its general structure is shown in Fig. 4, where the three countries are supposed to use the TELEMED system.
The TELEMED system will be able to support the diagnostics decision-making of hospital medical personnel. It will give advice using the knowledge base of multiple medical expertise collected from several knowledge sources. The
Medical teleconsulting support system
The second planned system is used to support teleconsulting of medical experts (TELECONS) [6]. Its general structure is shown in Fig. 8. The TELECONS system will be able to support the cooperative decision-making of a group of medical experts during their consulting via telenetwork and help them to find consensus concerning requests obtained from a medical teleconsulting centre. Decision-making in TELECONS is based on the voting-type technique of the multiple experts knowledge base and supposes
Conclusion
This paper briefly summarises the experiments made in order to apply new methods of data processing and knowledge management to telemedicine. The multilevel representation of medical data is proposed. This representation is based on collecting knowledge about statistical methods and their behaviour when semantically-essential information is acquired from the complex dynamics of quasi-periodical medical signals. We elaborate a method that helps to find the optimal statistical diagnostic method
Acknowledgements
We would like to thank the Centre for International Mobility (CIMO) for the support that made it possible for Alexey Tsymbal to work during the preparation of ths article. We would also like to thank the editors who helped us to improve readability of our article.
References (14)
- S.A. Aivazyan, Applied statistics: classification and dimension reduction, Finances & Statistics, Moscow, 1989 (in...
- A.R. Barrett, J.S. Edwards, Knowledge elicitation and knowledge representation in a large domain with multiple experts,...
- M. Bondarenko, V. Terziyan, A. Tkachuk, Multilevel models for representation of quasi-periodical temporal patterns, in:...
- V. Grebenyuk, H. Kaikova, V. Terziyan, Multilevel automatic diagnostics based on models of speech behavior, in: Proc....
- T. Hastie, R. Tibshirani, Discriminant Adaptive Nearest Neighbour Classification, Technical report, Stanford...
- H. Kaikova, V. Terziyan, S. Puuronen, The role of Ukrainian universities in the development of the global information...
- B. Mak, T. Bui, R. Blanning, Aggregating and updating experts’ knowledge: an experimental evaluation of five...