Information Technology Reference
In-Depth Information
To continue to discuss a question about a threshold, let extend our example and
assume, that a description of a rare disease Primary Ciliary Diskinesia is available
and differs from the normal Bronchitis, Pneumonia (see Table 25.8) linguistically
and numerically as presented in the Table 25.10. We choose this example only for
illustration, to show deviations of a “normal” case from a rare case, recalling, that
our approach is to suspect the RD.
Due to the
mean
-absolutedifference composition a possible degree for
d
3
under
the presence of
S
p
is 0
44 and under the presence of
S
p
is 0
45.
Let us see values for
d
1
,
d
2
if a patient has the complete picture of a rare disease
Primary Ciliary Diskinesia
,i.e.,
.
.
S
p
=
{
(
s
1
,
0
.
75
)
,
(
s
2
,
0
.
75
)
,
(
s
3
,
0
.
75
,
(
s
4
,
0
.
75
)
,
(
s
5
,
0
.
8
)
,
(
s
6
,
0
.
75
)
}
The possible degrees for
d
1
,
d
2
under
S
p
are 0
28 correspondingly.
The possible degrees for
d
1
,
d
2
and
d
3
under
S
p
,
S
p
,
S
p
are presented in the Table
25.11 and Table 25.12.
.
44 and 0
.
Table 25.11
Possible degrees for “normal”
diseases
Table 25.12
Possible degrees for the RD
d
3
d
1
d
2
S
p
0
.
47
S
p
0
.
0 0
.
18
S
p
0
.
45
S
p
0
.
31 0
.
24
S
p
0
.
0
S
p
0
.
44 0
.
28
25 can be chosen as a threshold to point a suspicion
of the RD, i.e., if all numbers in one string in the Table 25.11 are above 0
For this particular example 0
.
25 a
physician should think about “non-normality” of a possible diagnosis for a patient
at hand. It means, that we found an approach to suspect the RD if the diagnoses from
the initially defined list seems to be possible. The choice of a threshold remains on
experts. There are many ways to formalize this procedure [7, 11, 20], and what
approach to use is a subject of a separate investigation.
It can be argued that one number, as a threshold, does not reflect a flexibility
in diagnostics: interval- and fuzzy sets estimations of linguistic expressions show
more flexibility for the RD diagnostic problem.
Several additional ways to warn non-normality of a diseases can be added. In
some systems a possibility of diagnostic contradictions and not generated diagnoses
are predefined: in the first case a diagnosis is confirmed and excluded at the same
time, whereas in the second case the membership value of each disease in the set
of confirmed diagnosis is below the predefined threshold: 0
.
<
D
p
(
d
j
)
<
ε
.Ifa l
investigated diseases
d
j
from
are classified as diagnostic contradictions and/or
not generated diagnoses, then a suspicion of the RD exists.
Δ
