Biomedical Engineering Reference
In-Depth Information
shape point, in AAM, is a parameter. Furthermore, each shape point required human
operator to manually mark each of the location. Undoubtedly, each shape point location
in this context is considered as a parameter as well. One should notice that the enor-
mous number of decisions required to be made by human expert had not included the
type of model, location of model and number of iteration to be determined during the
execution phase because it has been assumed to be fixed which is not practical and not
applicable. Therefore, generally, if it is assumed that only assumptions 1-6 above are
acceptable, the total number of user-specified parameters N TNP that is practical can be
roughly be computed as follows:
N TI
(4.1)
N TNP =
( N SP ( i )) +[( N SP ( i )) + ( N SPL ( i ))]+ 3 ( N R ) + 10
i = 1
where N SP ( i ) denotes the total number of shape points in ith training sample;
N SPL ( I ) denotes the ith shape points location of N SP ( i ); N TI denotes the total
number of training samples, N R denotes the total number of targeted hand radio-
graphs to be segmented for computer-aided skeletal age scoring system. According
to Eq. ( 1.1 ) , if 100 hand bones from 100 radiographs are to be segmented, and
the total shape points in all the training samples are identical, then the total user-
specified parameters amounts to 3, 99,112. The type of parameters and number of
parameters of each type is shown in Table 4.2 .
Table 4.2 User-specified parameters used in shape-driven AAM
Type of parameters
Parameters
Training phase
(1) Number of bones to model
1
(2) Number of epochs to represent the age range
1
(3) Division of epochs
1
(4) Number of AAM
1
(5) Number of training images
1
(6) Number of shape points, n
n for each training sample
(7) Locations of shape points
1 for each shape point for each
training sample
(8) Number of principal shape models
1
(9) Number of sampling point in intensity model
1
(10) Number of principal intensity models
1
(11) Number of pose parameters
1
(12) The number of iterations before the trained AAM model
fit to the hand bone
1
Execution phase
(1) Choose which of the trained AAM models is to be fitted
on the targeted hand bone
1 for each executed radiograph
(2) Place the location of the chosen trained AAM model
1 for each executed radiograph
(3) The number of iterations before the trained AAM model
fit to the hand bone on each radiograph
1 for each executed radiograph
Total parameters (100 executed radiographs)
3,99,112
Search WWH ::




Custom Search