Information Technology Reference
In-Depth Information
at least
2
of
(
φ
1
,
φ
2
,
φ
3
)
can be rewritten as
(
φ
1
∧
φ
2
)
∨
(
φ
1
∧
φ
3
)
∨
(
φ
2
∧
φ
3
)
,
for
φ
1
,
φ
2
,
φ
3
some arbitrary medical entities in CADIAG2.
The Knowledge Base.
The knowledge base of CADIAG2, to which we will com-
monly refer as
KR
, consists of a collection of approximately 40.000
IF-THEN
rules
that express possibly uncertain relationships among distinct medical entities.
Rules in CADIAG2 can be characterized by triples of the form
θ
,
φ
,
η
,where
θ
is the
antecedent
,
φ
the
consequent
and
η
is the degree to which
θ
confirms
φ
(i.e., the degree of confirmation), for
.
In some literature about CADIAG2 -see for example [1] or [7]- rules are defined
as 4-tuples of the form
θ
,
φ
medical entities and
η
∈
[
0
,
1
]
θ
,
φ
,
η
,
ζ
,for
θ
,
φ
medical entities and
η
,
ζ
∈
[
0
,
1
]
,where
η
stands for the degree to which
θ
(the antecedent) confirms
φ
(the consequent) and
ζ
, sometimes referred to in the corresponding
literature as the
strength of confirmation
and the
frequency of occurrence
of the
rule respectively. The 4-tuple
for the degree to which
φ
confirms
θ
θ
,
φ
,
η
,
ζ
corresponds in our notation to the pair of
triples
.
We can distinguish among three different types of rules by considering their form
and how they are used along the inference process in relation to the values assigned
to their antecedent or consequent.
θ
,
φ
,
η
and
φ
,
θ
,
ζ
•
Type
confirming to the degree
η
(
c
η
). A rule of this type is of the general form
θ
,
φ
,
η
.Itis
triggered in a run of the inference mechanism in CADIAG2 for strictly positive
values or grades of its antecedent (in a way that will be made clear below, where
we describe the inference engine of CADIAG2). We will generally refer to rules
of this kind as rules of type
c
.
,for
θ
,
φ
medical entities (
φ
a basic medical entity) and
η
∈
(
0
,
1
]
A rule of type
c
formalizes (possibly) uncertain interrelations among medical enti-
ties, the bigger the degree of confirmation
η
the more certain the presence of the
consequent given the antecedent of the rule.
What follows is an example of a rule of type
c
, taken from [1]:
1
IF
suspicion of liver metastases by liver palpation
THEN
pancreatic cancer
with degree of confirmation 0
.
3.
•
Type
mutually exclusive
(
me
). A rule of type
me
is of the form
θ
,
φ
,
0
,for
θ
,
φ
a basic medical entity). It is only triggered in a run
of the inference engine of the system when there is certainty about the truth or
occurrence of
medical entities (
φ
θ
.
A rule of type
me
expresses mutual exclusiveness between antecedent and conse-
quent (i.e., the presence of one of them excludes the other).
1
The subsequent examples in this subsection are also taken from [1].
