Information Technology Reference
In-Depth Information
The one that follows is an example of a rule of this type:
IF
positive rheumatoid factor
THEN NOT
seronegative rheumatoid arthritis
•
Type
always occurring
(
ao
). A rule of type
ao
is of the form
θ
,
φ
medical entities. It can be triggered by the system only when there is certainty
about the falsity or non-occurrence of
θ
,
φ
,
1
,for
φ
.
A rule of type
ao
expresses the fact that the antecedent implies the consequent. It
follows that, if the consequent is excluded, the presence of the antecedent is also
excluded.
Notice that a rule
θ
,
φ
,
1
of type
ao
can be alternatively formalized by the triple
∼
φ
,∼
θ
,
1
and that it is not a special case of a rule of type
c
due to the fact that
∼
θ
is not a basic medical entity.
Next we give an example of a rule of this type:
IF NOT (
rheumatoid arthritis
AND
splenomegaly
AND
leukopenia
≤
4000
l
)
THEN NOT
Felty's syndrome
/
μ
There are other typologies for the rules in
KR
that will prove useful in further sec-
tions in this paper. A very general typology is the one that follows:
•
Binary
rules.
Rules of the form
θ
,
φ
,
η
where
θ
,
φ
are basic entities.
•
Compound
rules.
Rules of the form
θ
,
φ
,
η
where
θ
is a compound medical
entity and
φ
is a basic entity.
The vast majority of rules in
KR
are binary. There are less than one hundred com-
pound rules in
KR
yet, despite the number, they are important for the functioning of
the system.
We have a further distinction among binary rules of use in further sections:
•
Symptom-symptom
rules.
Rules of the form
θ
,
φ
,
η
where both
θ
,
φ
are symp-
toms and
η
∈{
0
,
1
}
.
•
Diagnose-diagnose
rules.
Rules of the form
θ
,
φ
,
η
where both
θ
,
φ
are diag-
noses and
η
∈{
0
,
1
}
.
•
Symptom-diagnose
rules.
Rules of the form
θ
,
φ
,
η
where
θ
is a symptom,
φ
a diagnose and
η
∈
[
0
,
1
]
.
•
Diagnose-symptom
rules.
Rules of the form
θ
,
φ
,
η
where
θ
is a diagnose,
φ
is a symptom and
η
∈
[
0
,
1
]
.
Most rules of type
diagnose-symptom
in
KR
are not used by the inference engine,
only those of type
ao
areusedbyit.
The Inference Engine.
CADIAG2 gets started with medical information about
the patient. Such information is formally given by a set of basic medical entities
present in the patient, each one together with a number in the interval
which,
in principle, is intended to represent the degree to which such entity is present
[
0
,
1
]