Information Technology Reference
In-Depth Information
(i.e., its degree of presence). These values are, in most of the literature on CA-
DIAG2, interpreted as membership degrees in the context of fuzzy set theory and
respond to the possibly vague nature of medical entities in CADIAG2.
Values assigned to compound medical entities in the system (in principle only for
those that are relevant for the inference) are generated according to the following
rules, for
θ
,
φ
any medical entities:
•
The assignment to
θ
∧
φ
is obtained as the minimum between the corresponding
assignments to
θ
and
φ
.
•
The assignment to
θ
∨
φ
is obtained as the maximum between the corresponding
assignments to
θ
and
φ
.
•
The assignment to
∼
θ
is obtained as the difference between 1 and the assign-
ment to
θ
.
After the initial medical information about the patient is obtained and entered into
the system the rules in the knowledge base come into play. All the rules triggered
by the initial information about the patient are used during the inference process. At
each step in the inference process a rule of type
c
,
me
or
ao
is applied (that is done,
in principle, in no particular order). Rules of these types are triggered as follows,
for general entities
θ
,
φ
:
•
Arule
of type
c
can be triggered at some step in the inference process
if a strictly positive value has been previously assigned to
θ
,
φ
,
η
θ
.
The use of the
rule
θ
,
φ
,
η
will generate a new assignment for
φ
, calculated as the minimum
between the value assigned to
θ
that triggers the rule and
η
.
•
Arule
of type
me
can be triggered during the inference process if cer-
tainty about the presence of
θ
,
φ
,
0
in the patient (i.e., the assignment 1) has been
previously concluded. The application of
θ
θ
,
φ
,
0
allows us to conclude cer-
tainty about the absence of
φ
in the patient (i.e., the assignment 0).
•
Arule
θ
,
φ
,
1
of type
ao
can be triggered if certainty about the absence of
φ
has been previously concluded. The application of the rule
θ
,
φ
,
1
will allow
us to conclude certainty about the absence of
θ
in the patient.
The inference process goes on until the system comes to the stage where neither new
medical entities nor new assignments for those already generated can be inferred.
CADIAG2 yields as outcome of the inference the set of diagnoses generated during
the inference process along with the maximal value (with respect to the ordering
defined above) assigned to them during the inference.
It has to be mentioned that, according to part of the literature on CADIAG2 -
for example [1]-, the original inference process in CADIAG2 works in a slightly
different way. The update in the value of the distinct sentences involved in the
inference is done as soon as two different values for the same sentence are produced
by the system. The value chosen in the update for atomic sentences in
L
is the
maximal one (with respect to the ordering
). Notice though that this feature has
a highly undesirable result (unless further restrictions on the rules or on the order
in which the rules are used are imposed), which is that the outcome of a run of the
