Information Technology Reference
In-Depth Information
25.6
Concluding Discussion and Future Development
Whenever possible, standard methodological approaches should be applied in the
design and analysis of a clinical trial that warrant adequate informative value. How-
ever, there are circumstances when the number of experimental subjects is unavoid-
ably small. In such circumstances it is justified to consider abandoning standard
statistical methodology in place of alternative approaches [6].
The problem to suspect the RD (in particular, in pediatrics) with a help of com-
puter - assisted system is an important problem and till now it was not investigated
well enough. This problem is connected with the absence, insufficiency, uncertainty
and imprecision of the information and, therefore, belongs to the category of prob-
lems, that can be solved by fuzzy logic approaches.
Our main idea to suspect the RD is based on the assumption, that the deviation
from a “normal” disease is a sign to warn the RD. Thus, adequate acquaintance with
normal diseases leads to a recognition of the RD. We suppose, that a physician has
already established a preliminary diagnosis for a patient, but s/he hesitates about its
correcteness. Then a physician, based on the patient's data, checks the deviations
from the “normal” cases, and if these deviations reach certain level, a suspicion
of the RD arises, i.e., our approach is not for final diagnostic, the main aim of
proposed algorithm is to raise suspicion on an abnormality of the common diseases,
consequently on a possibility of RD.
We assume that RDs are not in the knowledge base of a decision support com-
puter system, that we have developed. If a suspicious of the RD arises, a user can
search for in an appropriate database [5, 22, 23], where several thousand of RDs are
described. Notice that because of immense number of RDs in WWW data banks, it
seems to be reasonable to choose an appropriate subsets for certain regions/countries
(e.g., for Georgia).
“Normal” diseases is not a crisp concept. And recognition of “normal” diseases
is a subject of investigation of many researches dealing with a problem of deci-
sion making in medicine. A “normal” disease can be classified into different cat-
egories such as possible, excluded, confirm and some others. These classification
depends, in general, on the initial information available. In the paper we have in-
vestigated, under which conditions these classifications are not a case. We use an
aggregation operators based approximate reasoning mechanism, in particular, mean-
absolute difference composition. Such composition reflects our intent to deal with
deviations from normal cases. For example, if all elements from
are considered
as excluded, or the value of membership of a diagnoses in the class of possible di-
agnoses is less than the a priory defined threshold, then it is at the border of possible
and non-possible diagnoses. Thus, once again, if we know what do the “normal”
diseases imply and how they are exhibited, we have investigated in existing systems
the cases of the “abnormal” exhibition.
We illustrate our approach with examples from daily practice - more or less
sufficient. It would be better to have the full list of deviations and their evalua-
tion. But it will be rather difficult and time-consuming job because of insufficient
Δ
