In medical decision problems it is very important to use the most relevant piece of information for decision making. We focus on a special case of diagnostic decision making when we can measure many symptoms and signs and we have to make diagnostic conclusions. We can state the problem as follows. We can measure symptoms and signs of a patient, denoted by s1, s2, ..., sk, and we have to decide about a possible diagnosis d. We know that the symptoms and signs have different costs w1, w2, ... wk when they are examined. Of course, each symptom, sign or their combination has a different predictive value for the diagnosis. Our task is to find out the combination of symptoms from given data with a sufficient informative value for diagnostic decision making. However, simultaneously we look for a combination of symptoms and signs with minimal costs among those carrying sufficient information. For that reason we will describe approaches based on information measures of statistical dependence and to show the idea of the program CORE (constitution and reduction of data) prepared for practical applications in medicine.
[1]
Jana Zvárová,et al.
On measures of statistical dependence
,
1974
.
[2]
Albert Perez.
Ε-admissible Simplifications of the Dependence Structure of a Set of Random Variables
,
1977,
Kybernetika.
[3]
Albert Perez,et al.
Information, ε-sufficiency and data reduction problems
,
1965,
Kybernetika.
[4]
I. Vajda.
Theory of statistical inference and information
,
1989
.
[5]
Milan Studený.
Asymptotic behaviour of empirical multiinformation
,
1987,
Kybernetika.
[6]
Judea Pearl,et al.
Probabilistic reasoning in intelligent systems - networks of plausible inference
,
1991,
Morgan Kaufmann series in representation and reasoning.