Biomedical Engineering Reference
In-Depth Information
where
N
iq
denotes total number of input variables;
N
mf
(
i
)
denotes the number of
membership function in ith input variables. In the case of proposed segmentation
to 42.
Note that in Eq. (
4.3
), it does not contain the total number of training samples
[the quantity denoted as
N
TI
in Eq. (
4.1
)] and the total number of targeted hand
radiographs to be segmented for computer-aided skeletal age scoring system [the
quantity denoted as
N
R
in Eq. (
4.1
)]. Another way of explaining is that the number
of user-specified parameters will not increase with the number of radiographs that
are to be segmented and it does not depends on any training samples because the
modeling phase needs only to be accomplished once by the human expert to instill
the intelligence and flexibility.
4.3.1.3 Interpretations
Figure
4.3
showed that the number parameters of the proposed segmentation
framework are invariant to the increase of the number of targeted radiographs. As
illustrated in the figure, the number of user-specified parameters increases linearly
against the number of radiographs that are to be segmented (targeted radiographs)
whereas the number of user-specified parameters of proposed segmentation frame-
work remains unchanged against the number of targeted radiographs.
This criterion is of exceptionally importance when the objective is to segment a
large amount of hand bones for computerized BAA. It means even if both segmen-
tation frameworks are assumed to have similar number of parameters, the AAM
segmentation framework is confronting with large amounts of parameters (linear
increment) as the number of targeted radiographs increases. Besides, increment
in user-specified parameters can be viewed as a sign of computational complex-
ity increments. The reason is that if the intention is to replace all the user-speci-
fied parameters decision using automated methods (if exist), higher level of image
(a)
(b)
2.033
x
10
5
110
100
2.0325
90
2.032
80
2.0315
70
2.031
60
2.0305
50
2.03
40
30
2.0295
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
70
80
90
100
N
R
N
R
Fig. 4.3
The change of total number of parameters against the increase of targeted radiographs
for
a
AAM.
b
Proposed segmentation framework
Search WWH ::
Custom Search