Dealing with Incomplete Data Records in Qualitative Modeling and Simulation of Biomedical Systems

Keywords

Abstract

In the biomedical domain, it is very common to find that the internal structure of the systems under investigation are totally or partially unknown, making it impossible to use analytical models. Therefore, qualitative modeling and simulation techniques with their inherent tolerance for uncertainty and ambiguity provide an excellent platform for the analysis of biomedical systems that may be difficult to model in a more precise fashion. Moreover, even where quantitative models are available, qualitative models may constitute an important complement to the more classical quantitative models.

One of the major problems in biomedical qualitative modeling is the lack of information. Inductive, pattern-based modeling techniques are extremely data hungry. It is therefore essential for behavioral qualitative methodologies to have available a large amount of rich data to work with. Unfortunately, in biomedical applications, this is hardly ever the case.

The lack of information may have several different causes, all of them related to acquisition difficulties. The problems are further amplified when the data records obtained from medical experiments are incomplete.

In this paper, a technique called missing data option is proposed that allows to work with incomplete medical data records. This technique represents an enhancement to a previously introduced qualitative modeling and simulation methodology entitled fuzzy inductive reasoning.

Interested in reading the full paper? (6 pages, 219,700 bytes, pdf)

Homepage

Last modified: June 17, 2005 -- © François Cellier