Information Technology Reference
In-Depth Information
16
Fuzziness in Medical Measurement
and Approximate Reasoning
Ernesto Araujo
16.1
Introduction
Finding out autonomous mechanisms and structures that emulate the human reason-
ing is an important objective that has been pursuit in diverse areas [8, 10, 11, 18].
Making computers think like people is a challenging objective [39], especially in
medicine and health care.
The field of artificial intelligence is, so far, an alternative for dealing with assess-
ment, diagnosis, and therapeutic conduct. Probability is another approach largely
employed in medicine and health care. A more recent alternative concerns the field
of computational intelligence. Computational intelligence and artificial intelligence
present differences and similarities. In the traditional perspective, the latter is un-
derstood as a top-down symbolic approach composed of case-based reasoning, de-
ductive reasoning, expert systems, logic and symbolic machine learning systems.
The former comprises neural networks, fuzzy systems, evolutionary computing,
swarm intelligence, and immune systems mimicking nature for problem solving in
a bottom-up approach [32]. The manner to select among these techniques relies on
the sort of available information and the way to represent knowledge. Information
that is (
a
) perfect, certain and precise; (
b
) imperfect, certain but imprecise; (
c
)im-
perfect, precise but uncertain, (
d
) imperfect, uncertain and imprecise - concerning
vagueness, similarity, approximation, and truth - from the field of possibility theory;
and (
e
) imperfect, uncertain and imprecise - related to occurrence, and confidence
- from the field of probability theory are depicted in Figure 16.1.
Medicine and health care are a fertile environment for vague, conflicting, and not
definitive decisions, and therefore diagnosis, assessment, and therapeutic conduct.
When dealing with this kind of information, that is simultaneously uncertain and im-
precise, then the decision is considered to be under approximation. In order to deal
with this kind of problem, a general theory of approximate reasoning was proposed
in [38]. This reasoning methodology addresses the interface between numbers and
symbols by using the fuzzy set theory approach. In this case, the theory of fuzzy
sets [35] is appropriate to express vague (approximate) information, since vagueness
