TELK
O
M
N
I
K
A
Te
l
e
c
ommu
n
i
c
at
i
on
,
C
omp
u
ti
n
g
,
El
e
c
t
r
on
i
c
s
an
d
C
on
tr
o
l
V
ol
.
1
8
,
N
o
.
2
,
A
pri
l
2020
,
p
p.
8
0
7
~
8
1
4
IS
S
N
:
1693
-
6930
,
a
c
c
r
e
di
t
e
d
F
i
rs
t
G
r
a
de
by
K
e
m
e
n
ri
s
t
e
kdi
k
t
i
,
D
e
c
r
e
e
N
o:
21
/
E
/
K
P
T
/
2
018
D
O
I:
10
.
129
28/
T
E
L
K
O
M
N
IK
A
.
v1
8
i
2
.
1475
4
807
Jou
r
n
al
h
o
m
e
pa
ge
:
ht
t
p:
/
/
j
our
nal
.
uad
.
a
c
.
i
d/
i
nd
e
x
.
php
/
T
E
L
K
O
MNIKA
E
v
a
l
u
a
t
i
o
n
o
f
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
a
r
c
h
i
t
e
c
t
u
r
e
s
i
n
t
h
e
i
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
b
o
n
e
f
i
ssu
r
e
s
F
r
e
d
y
M
a
r
tí
n
e
z
,
C
é
s
ar
H
e
r
n
án
d
e
z
,
F
e
r
n
an
d
o
M
ar
t
í
n
e
z
F
a
c
u
l
t
a
d
T
e
c
nol
óg
i
c
a
,
U
ni
v
e
r
s
i
da
d
D
i
s
t
r
i
t
a
l
F
r
a
nc
i
s
c
o
J
o
s
é
de
C
a
l
d
a
s
,
C
o
l
o
m
b
i
a
A
r
ti
c
l
e
I
n
fo
A
BS
TR
A
C
T
Ar
t
i
c
l
e
h
i
s
t
or
y
:
Re
c
e
i
v
e
d
A
ug
16
,
201
9
Re
vi
s
e
d
J
a
n
4
,
2020
A
c
c
e
pt
e
d
F
e
b
13
,
202
0
A
ut
o
m
a
t
e
d
m
e
d
i
c
a
l
i
m
a
ge
pr
oc
e
s
s
i
ng
,
p
a
r
t
i
c
u
l
a
r
l
y
o
f
r
a
d
i
o
l
og
i
c
a
l
i
m
a
g
e
s
,
c
a
n
r
e
du
c
e
t
he
nu
m
b
e
r
o
f
d
i
a
gn
os
t
i
c
e
r
r
o
r
s
,
i
n
c
r
e
a
s
e
pa
t
i
e
nt
c
a
r
e
a
nd
r
e
d
uc
e
m
e
d
i
c
a
l
c
os
t
s
.
T
h
i
s
p
a
pe
r
s
e
e
ks
t
o
e
v
a
l
ua
t
e
t
he
p
e
r
f
or
m
a
nc
e
of
t
h
r
e
e
r
e
c
e
n
t
c
onvo
l
u
t
i
ona
l
ne
u
r
a
l
n
e
t
w
o
r
k
s
i
n
t
h
e
a
ut
o
no
m
ou
s
i
de
nt
i
f
i
c
a
t
i
on
o
f
f
i
s
s
u
r
e
s
ove
r
t
w
o
-
d
i
m
e
n
s
i
ona
l
r
a
d
i
o
l
og
i
c
a
l
i
m
a
ge
s
.
T
he
s
e
a
r
c
h
i
t
e
c
t
ur
e
s
ha
ve
be
e
n
pr
op
os
e
d
a
s
de
e
p
ne
u
r
a
l
n
e
t
w
or
k
t
ype
s
s
pe
c
i
a
l
l
y
de
s
i
gne
d
f
or
i
m
a
g
e
c
l
a
s
s
i
f
i
c
a
t
i
on
,
w
hi
c
h
a
l
l
ow
s
t
he
i
r
i
nt
e
g
r
a
t
i
o
n
w
i
t
h
t
r
a
d
i
t
i
o
na
l
i
m
a
g
e
pr
oc
e
s
s
i
n
g
s
t
r
a
t
e
g
i
e
s
f
or
a
u
t
o
m
a
t
i
c
a
na
l
y
s
i
s
o
f
m
e
di
c
a
l
i
m
a
g
e
s
.
I
n
pa
r
t
i
c
u
l
a
r
,
w
e
u
s
e
t
h
r
e
e
c
onvo
l
u
t
i
ona
l
n
e
t
w
or
k
s
:
R
e
s
N
e
t
(
r
e
s
i
d
ua
l
ne
ur
a
l
ne
t
w
or
k)
,
D
e
ns
e
N
e
t
(
de
n
s
e
c
onvo
l
u
t
i
ona
l
ne
t
w
o
r
k)
,
a
nd
N
A
S
N
e
t
(
n
e
ur
a
l
a
r
c
hi
t
e
c
t
u
r
e
s
e
a
r
c
h
ne
t
w
o
r
k
)
t
o
l
e
a
r
n
i
nf
o
r
m
a
t
i
o
n
f
r
om
a
s
e
t
of
20
0
i
m
a
g
e
s
l
a
be
l
e
d
h
a
l
f
a
s
f
i
s
s
ur
e
d
bone
s
a
nd
h
a
l
f
a
s
s
e
a
m
l
e
s
s
bon
e
s
.
A
l
l
t
h
r
e
e
n
e
t
w
or
k
s
a
r
e
t
r
a
i
ne
d
a
n
d
a
d
j
u
s
t
e
d
unde
r
t
h
e
s
a
m
e
c
o
ndi
t
i
on
s
,
a
nd
t
he
i
r
pe
r
f
o
r
m
a
nc
e
w
a
s
e
va
l
u
a
t
e
d
w
i
t
h
t
he
s
a
m
e
m
e
t
r
i
c
s
.
T
h
e
f
i
na
l
r
e
s
u
l
t
s
c
on
s
i
de
r
n
ot
o
nl
y
t
he
m
o
de
l
'
s
a
b
i
l
i
t
y
t
o
pr
e
d
i
c
t
t
he
c
ha
r
a
c
t
e
r
i
s
t
i
c
s
of
a
n
unk
now
n
i
m
a
ge
b
ut
a
l
s
o
i
t
s
i
nt
e
r
na
l
c
o
m
p
l
e
x
i
t
y.
T
he
t
hr
e
e
n
e
ur
a
l
m
ode
l
s
w
e
r
e
op
t
i
m
i
z
e
d
t
o
r
e
duc
e
c
l
a
s
s
i
f
i
c
a
t
i
on
e
r
r
o
r
s
w
i
t
h
out
pr
od
uc
i
ng
ne
t
w
or
k
ov
e
r
-
a
d
j
u
s
t
m
e
nt
.
I
n
a
l
l
t
h
r
e
e
c
a
s
e
s
,
ge
n
e
r
a
l
i
z
a
t
i
on
of
be
h
a
vi
or
w
a
s
obs
e
r
v
e
d
,
a
nd
t
he
a
b
i
l
i
t
y
of
t
he
m
ode
l
s
t
o
i
d
e
n
t
i
f
y
t
he
i
m
a
g
e
s
w
i
t
h
f
i
s
s
ur
e
s
,
how
e
ve
r
t
h
e
e
x
pe
c
t
e
d
pe
r
f
o
r
m
a
n
c
e
w
a
s
on
l
y
a
c
h
i
e
ve
d
w
i
t
h
t
he
N
A
S
N
e
t
m
o
de
l
.
Ke
y
w
or
d
s
:
Bi
om
e
d
i
c
a
l
c
o
m
pu
t
i
ng
D
e
e
p
ne
ura
l
n
e
t
w
or
k
F
i
s
s
ure
s
r
e
c
og
ni
t
i
on
Im
a
g
e
proc
e
s
s
i
ng
T
hi
s
i
s
an
op
e
n
ac
c
e
s
s
ar
t
i
c
l
e
u
nde
r
t
he
C
C
B
Y
-
SA
l
i
c
e
n
s
e
.
Cor
r
e
s
pon
di
n
g
Au
t
h
or
:
F
re
dy
M
a
rt
í
n
e
z
,
F
a
c
ul
t
a
d
T
e
c
nol
ógi
c
a
,
U
ni
ve
rs
i
da
d
D
i
s
t
ri
t
a
l
F
ra
nc
i
s
c
o
J
os
é
de
Ca
l
d
a
s
,
Bogot
á
,
Co
l
om
bi
a
.
E
m
a
i
l
:
fhm
a
rt
i
n
e
z
s
@
udi
s
t
ri
t
a
l
.
e
du.
c
o
1.
I
N
TR
O
D
U
C
TI
O
N
In
r
e
c
e
nt
y
e
a
rs
t
h
e
re
h
a
v
e
b
e
e
n
m
or
e
a
nd
m
ore
a
dva
nt
a
ge
s
of
t
he
us
e
of
d
i
gi
t
a
l
i
m
a
ge
pr
oc
e
s
s
i
ng
i
s
us
e
d
a
s
a
t
ool
t
o
s
uppor
t
t
h
e
di
a
gn
os
i
s
from
m
e
di
c
a
l
i
m
a
g
e
s
[1,
2]
a
nd
i
t
’s
e
v
e
n
prov
e
d
v
e
ry
v
a
l
u
a
bl
e
t
o
t
ra
c
k
t
hrough
out
i
m
a
ge
s
(
t
e
m
por
a
l
q
ua
n
t
i
f
i
c
a
t
i
on
a
nd
grow
t
h)
bo
t
h
d
a
m
a
ge
a
nd
b
e
h
a
vi
o
r
of
t
i
s
s
ue
s
[3
,
4]
.
A
n
a
u
t
o
m
a
t
e
d
s
ys
t
e
m
ha
s
t
h
e
a
dv
a
nt
a
ge
of
qu
i
c
k
l
y
i
de
nt
i
fy
i
ng
s
p
e
c
i
fi
c
pa
t
t
e
rns
i
n
l
a
rg
e
vol
u
m
e
s
of
i
m
a
g
e
s
w
i
t
h
a
hi
gh
d
e
gre
e
o
f
r
e
l
i
a
b
i
l
i
t
y
.
A
s
a
s
u
pport
t
o
ol
for
s
p
e
c
i
a
l
i
z
e
d
m
e
d
i
c
a
l
pe
rs
o
nne
l
,
t
h
i
s
t
ool
c
a
n
not
onl
y
re
duc
e
d
i
a
g
nos
t
i
c
t
i
m
e
b
ut
a
l
s
o
r
e
du
c
e
s
c
onfus
i
ng
or
m
i
s
di
a
gnos
e
s
[
5].
T
h
e
gr
e
a
t
a
dva
n
t
a
g
e
of
di
a
gnos
t
i
c
i
m
a
gi
ng
i
s
i
t
s
non
-
i
nv
a
s
i
v
e
c
ha
r
a
c
t
e
r
s
i
nc
e
m
os
t
of
t
h
e
s
e
i
m
a
ge
s
a
r
e
c
a
p
t
ure
d
by
r
e
s
on
a
nc
e
or
ra
d
i
ogr
a
phy
[6].
In
g
e
ne
ra
l
t
e
r
m
s
,
t
he
s
e
t
ool
s
us
e
a
c
e
rt
a
i
n
a
l
gor
i
t
h
m
of
c
l
a
s
s
i
fi
c
a
t
i
on
on
t
he
i
m
a
ge
t
o
d
e
t
e
r
m
i
n
e
i
f
i
t
pos
s
e
s
s
e
s
or
not
a
c
e
rt
a
i
n
c
h
a
r
a
c
t
e
r
i
s
t
i
c
,
a
n
d
t
h
us
t
o
c
l
a
s
s
i
fy
i
t
[7
,
8]
.
I
m
a
ge
s
a
r
e
no
rm
a
l
l
y
pr
e
-
pro
c
e
s
s
e
d
t
o
m
a
x
i
m
i
z
e
t
h
e
a
bi
l
i
t
y
t
o
de
t
e
c
t
t
he
c
h
a
ra
c
t
e
ri
s
t
i
c
s
of
i
nt
e
re
s
t
[9
-
11]
.
T
he
c
l
a
s
s
i
fi
c
a
t
i
on
a
l
gori
t
h
m
,
i
n
g
e
ne
ra
l
,
i
s
no
t
a
pp
l
i
e
d
t
o
a
n
i
m
a
g
e
i
nd
i
s
c
ri
m
i
na
t
e
l
y,
on
t
he
c
ont
r
a
ry
,
a
r
e
gi
on
o
f
i
nt
e
r
e
s
t
(RO
I)
i
s
i
d
e
nt
i
fi
e
d
on
t
h
e
i
m
a
g
e
,
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
1
8
,
N
o.
2
,
A
pri
l
2
020:
8
0
7
-
8
1
4
808
w
hi
c
h
c
a
n
b
e
d
one
m
a
nu
a
l
l
y,
or
e
ve
n
a
ut
o
m
a
t
i
c
a
l
l
y
i
n
m
ore
a
d
va
n
c
e
d
s
c
h
e
m
e
s
,
for
e
x
a
m
p
l
e
,
t
hrou
gh
s
e
gm
e
nt
a
t
i
on
s
t
r
a
t
e
gi
e
s
[
12
-
14]
.
It
i
s
a
l
s
o
pos
s
i
b
l
e
t
o
us
e
i
t
e
r
a
t
i
ve
s
e
a
rc
he
s
on
c
e
rt
a
i
n
i
m
a
g
e
s
t
ru
c
t
ur
e
s
t
o
de
t
e
rm
i
ne
c
h
a
ra
c
t
e
ri
s
t
i
c
s
a
nd
RO
I
.
S
t
a
t
i
s
t
i
c
a
l
m
e
t
hods
a
re
a
l
s
o
us
e
d
i
n
w
hi
c
h
t
he
i
m
a
ge
i
s
na
v
i
ga
t
e
d
f
rom
pre
vi
ous
i
nfor
m
a
t
i
on
of
c
ha
r
a
c
t
e
r
i
s
t
i
c
be
h
a
vi
ors
[1
5].
D
e
e
p
c
on
vol
u
t
i
o
na
l
n
e
ura
l
ne
t
w
orks
h
a
ve
b
e
c
o
m
e
a
pow
e
rful
t
oo
l
for
i
m
a
ge
c
l
a
s
s
i
f
i
c
a
t
i
on,
w
i
t
h
pa
rt
i
c
u
l
a
r
a
ppl
i
c
a
t
i
on
t
o
m
e
d
i
c
a
l
i
m
a
ge
s
[16
,
17]
.
T
h
e
s
e
c
or
re
s
pond
t
o
r
e
gu
l
a
r
i
z
e
d
ve
rs
i
ons
of
t
h
e
t
ra
di
t
i
on
a
l
m
ul
t
i
l
a
y
e
r
pe
r
c
e
p
t
rons
(fu
l
l
y
c
o
nne
c
t
e
d
forw
a
r
d
l
a
y
e
rs
)
[
18,
19]
.
T
ha
nks
t
o
t
hi
s
re
gul
a
ri
z
a
t
i
o
n
pr
oc
e
s
s
,
c
onvo
l
ut
i
ona
l
n
e
t
w
o
rks
a
c
hi
e
v
e
c
om
p
l
e
x
s
t
ru
c
t
u
re
s
w
i
t
h
s
i
m
pl
e
p
a
t
t
e
rns
t
h
a
t
r
e
du
c
e
t
h
e
pr
obl
e
m
of
n
e
t
w
o
rk
ove
r
-
a
dj
us
t
m
e
nt
[20
,
21
].
W
e
t
ra
i
ne
d
t
hre
e
m
ode
l
s
of
de
e
p
ne
ura
l
n
e
t
w
o
rks
t
o
i
de
n
t
i
fy
fi
s
s
ur
e
s
on
di
g
i
t
i
z
e
d
ra
di
o
l
og
i
c
a
l
i
m
a
g
e
s
.
T
he
i
m
a
ge
s
us
e
d
f
or
t
h
e
t
ra
i
ni
ng
c
orr
e
s
pond
t
o
s
e
c
t
i
ons
o
f
bone
s
i
n
w
h
i
c
h
RO
I
h
a
s
be
e
n
pre
vi
ous
l
y
i
de
n
t
i
f
i
e
d
,
bu
t
no
m
or
phol
ogi
c
a
l
ope
r
a
t
i
on
i
s
a
p
pl
i
e
d
t
o
t
he
m
[22
,
23]
.
T
h
e
t
yp
e
s
of
de
e
p
ne
t
s
s
e
l
e
c
t
e
d
c
orr
e
s
pond
t
o
t
he
s
t
a
t
e
of
t
he
a
r
t
i
n
c
onvo
l
ut
i
ona
l
n
e
t
s
for
i
m
a
ge
c
l
a
s
s
i
fi
c
a
t
i
o
n
[24
,
2
5].
T
h
e
fo
l
l
ow
i
ng
pa
rt
of
t
h
e
pa
p
e
r
i
s
a
rr
a
ng
e
d
i
n
t
hi
s
w
a
y
.
S
e
c
t
i
on
2
pr
e
s
e
nt
s
pre
l
i
m
i
na
ry
c
onc
e
pt
s
a
nd
probl
e
m
f
orm
u
l
a
t
i
o
n.
S
e
c
t
i
on
3
i
l
l
us
t
r
a
t
e
s
t
he
d
e
s
i
gn
pro
fi
l
e
a
nd
d
e
ve
l
op
m
e
n
t
m
e
t
hodol
og
y.
S
e
c
t
i
on
4
w
e
pre
s
e
nt
t
h
e
pre
l
i
m
i
na
ry
re
s
ul
t
s
.
A
n
d
f
i
na
l
l
y
,
i
n
S
e
c
t
i
on
5,
w
e
pr
e
s
e
n
t
ou
r
c
on
c
l
us
i
o
ns
.
2.
P
R
O
BL
EM
F
O
R
M
U
LA
TI
O
N
W
e
e
v
a
l
u
a
t
e
m
ode
l
s
ba
s
e
d
on
de
e
p
n
e
ur
a
l
n
e
t
w
or
ks
by
i
de
n
t
i
fyi
ng
c
h
a
ra
c
t
e
ri
s
t
i
c
s
i
n
bone
s
t
ru
c
t
ur
e
s
a
s
s
how
n
i
n
F
i
g
ure
1
.
In
p
a
rt
i
c
ul
a
r,
w
e
l
oo
k
for
m
ode
l
s
t
ha
t
i
d
e
nt
i
fy
a
nd
c
l
a
s
s
i
fy
bone
s
w
i
t
h
f
i
s
s
ure
s
a
nd
fr
a
c
t
ur
e
s
i
n
one
c
a
t
e
gory
,
a
nd
t
hos
e
he
a
l
t
hy
b
one
s
i
n
a
s
e
c
on
d
c
a
t
e
gory
.
I
n
d
e
e
p
l
e
a
rn
i
ng
,
a
c
o
nvol
u
t
i
o
na
l
ne
ura
l
n
e
t
w
or
k
(CN
N
)
i
s
a
c
l
a
s
s
of
d
e
e
p
ne
u
ra
l
ne
t
w
or
ks
c
o
m
m
o
nl
y
a
pp
l
i
e
d
t
o
a
n
a
l
y
z
i
ng
i
m
a
ge
s
.
T
h
e
y
h
a
v
e
t
he
gre
a
t
a
dva
n
t
a
ge
t
h
a
t
t
h
e
y
re
qu
i
re
m
uc
h
l
e
s
s
i
m
a
ge
pr
e
-
p
roc
e
s
s
i
n
g
t
o
i
de
nt
i
fy
t
he
f
e
a
t
ure
s
of
i
nt
e
r
e
s
t
t
ha
n
a
ny
ot
h
e
r
d
i
gi
t
a
l
p
roc
e
s
s
i
ng
s
t
ra
t
e
gy.
T
h
e
y
op
e
ra
t
e
a
s
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
l
go
ri
t
hm
s
i
n
w
hi
c
h
a
n
a
dj
us
t
a
bl
e
w
e
i
gh
t
va
l
u
e
i
s
a
s
s
i
gn
e
d
t
o
t
h
e
c
ha
r
a
c
t
e
r
i
s
t
i
c
s
of
t
he
i
m
a
ge
t
h
a
t
m
a
ke
i
t
d
i
s
t
i
ngui
s
h
a
bl
e
fro
m
o
t
he
rs
.
W
i
t
h
prop
e
r
t
ra
i
ni
ng
a
nd
a
dj
us
t
m
e
n
t
,
a
c
o
nvol
ut
i
ona
l
n
e
t
w
o
rk
c
a
n
r
e
p
l
i
c
a
t
e
t
he
b
e
ha
vi
or
of
a
s
ophi
s
t
i
c
a
t
e
d
fi
l
t
e
r
o
n
t
he
i
m
a
g
e
.
B
e
s
i
d
e
s
,
unl
i
k
e
t
r
a
di
t
i
on
a
l
ne
ur
a
l
ne
t
w
orks
,
a
c
o
nvol
u
t
i
ona
l
ne
t
w
ork
c
a
n
i
d
e
n
t
i
fy
s
p
e
c
i
a
l
a
nd
t
e
m
por
a
l
d
e
pe
n
de
n
c
i
e
s
i
n
i
m
a
g
e
s
.
T
he
hi
gh
pe
rf
orm
a
nc
e
of
c
on
vol
u
t
i
on
a
l
ne
t
w
orks
i
s
d
ue
t
o
t
he
de
s
i
gn
of
t
h
e
i
r
ne
t
w
ork
a
rc
h
i
t
e
c
t
ure
.
W
hi
l
e
t
he
i
r
ope
r
a
t
i
on
i
s
s
t
i
l
l
a
bl
a
c
k
bo
x,
t
he
i
r
hi
gh
p
e
rfor
m
a
n
c
e
i
s
a
t
t
r
i
but
e
d
t
o
c
h
a
ra
c
t
e
ri
s
t
i
c
s
s
uc
h
a
s
ne
t
w
ork
de
pt
h
,
ne
t
w
ork
w
i
dt
h
(gr
e
a
t
e
r
nu
m
b
e
r
o
f
p
a
ra
m
e
t
e
rs
)
,
a
nd
s
k
i
p
c
onn
e
c
t
i
ons
(w
he
t
he
r
d
e
ns
e
or
r
e
s
i
d
ua
l
,
w
hi
c
h
i
nc
r
e
a
s
e
s
t
h
e
c
o
m
pl
e
xi
t
y
o
f
t
he
ne
t
w
ork
a
nd
i
t
s
a
b
i
l
i
t
y
t
o
re
p
re
s
e
n
t
i
nfor
m
a
t
i
on)
.
Cons
e
q
ue
n
t
l
y
,
t
he
ne
t
w
orks
s
e
l
e
c
t
e
d
f
or
t
he
vi
s
u
a
l
c
a
t
e
g
ori
z
a
t
i
on
t
a
s
k
of
t
h
i
s
pe
rfor
m
a
nc
e
t
e
s
t
a
r
e
R
e
s
N
e
t
(
r
e
s
i
du
a
l
ne
ur
a
l
ne
t
w
ork
)
,
D
e
ns
e
N
e
t
(
de
ns
e
c
onvo
l
ut
i
on
a
l
ne
t
w
ork
)
,
a
nd
N
A
S
N
e
t
(
n
e
ur
a
l
a
r
c
hi
t
e
c
t
ur
e
s
e
a
r
c
h
ne
t
w
ork
).
F
i
gure
1
.
S
a
m
p
l
e
da
t
a
ba
s
e
of
c
r
a
c
ke
d
bon
e
i
m
a
g
e
s
us
e
d
f
or
m
ode
l
t
r
a
i
n
i
ng
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
E
v
a
l
uat
i
on
of
d
e
e
p
n
e
ur
al
ne
t
wor
k
ar
c
hi
t
e
c
t
ur
e
s
i
n
….
(
F
r
e
d
y
M
ar
t
í
n
e
z
)
809
3.
M
ET
H
O
D
O
L
O
G
Y
T
he
t
ra
i
ni
ng
o
f
t
h
e
t
hr
e
e
m
ode
l
s
i
s
pe
rfo
rm
e
d
w
i
t
h
t
h
e
s
a
m
e
d
a
t
a
s
e
t
,
a
c
us
t
o
m
s
e
t
of
X
-
ra
y
i
m
a
ge
s
s
e
pa
r
a
t
e
d
i
n
t
o
t
w
o
c
a
t
e
gor
i
e
s
(fi
s
s
ur
e
d
a
n
d
s
e
a
m
l
e
s
s
).
T
h
e
i
m
a
g
e
s
c
orr
e
s
pond
i
ng
t
o
a
c
a
t
e
gory
a
r
e
s
t
or
e
d
i
n
t
he
s
a
m
e
fol
d
e
r
for
e
a
s
y
i
de
n
t
i
f
i
c
a
t
i
on
by
c
a
t
e
gory
(t
h
e
na
m
e
of
t
h
e
fol
d
e
r
i
s
t
he
n
a
m
e
o
f
t
he
c
a
t
e
gory)
.
W
e
us
e
1000
i
m
a
g
e
s
for
e
a
c
h
c
a
t
e
gor
y
ke
e
pi
ng
t
h
e
b
a
l
a
nc
e
of
c
l
a
s
s
e
s
t
o
a
vo
i
d
b
i
a
s
e
s
i
n
t
he
m
od
e
l
.
W
e
us
e
T
e
ns
orF
l
ow
a
s
t
h
e
f
ra
m
e
w
ork
on
w
hi
c
h
w
e
r
un
K
e
r
a
s
.
N
u
m
py
,
S
c
i
ki
t
L
e
a
rn
,
P
a
nda
s
,
O
pe
n
CV
a
n
d
M
a
t
pl
o
t
l
i
b
w
e
r
e
a
l
s
o
us
e
d
a
s
s
upp
ort
l
i
bra
r
i
e
s
.
T
he
i
m
a
g
e
s
w
e
re
ra
n
dom
l
y
m
i
xe
d
i
n
t
h
e
da
t
a
l
i
s
t
t
o
i
m
pro
ve
ne
t
w
ork
p
e
rfor
m
a
n
c
e
.
Be
s
i
de
s
,
t
he
y
a
r
e
a
l
l
r
e
s
i
z
e
d
t
o
t
h
e
s
a
m
e
s
i
z
e
(
256*25
6
p
i
x
e
l
s
)
w
i
t
h
t
he
s
a
m
e
go
a
l
.
W
e
do
n
ot
c
o
ns
i
de
r
t
he
a
s
p
e
c
t
r
a
t
i
o
of
t
he
i
m
a
g
e
s
w
h
e
n
re
s
i
z
i
n
g
t
he
m
.
In
a
l
l
t
hr
e
e
c
a
s
e
s
,
t
he
da
t
a
s
e
t
w
a
s
di
v
i
d
e
d
i
nt
o
t
w
o
gro
ups
,
a
t
r
a
i
n
i
ng
group
,
a
nd
a
t
e
s
t
g
roup
.
W
e
us
e
d
7
0%
of
t
he
d
a
t
a
f
or
t
ra
i
ni
n
g
a
nd
3
0%
for
p
e
rfor
m
a
nc
e
e
va
l
ua
t
i
o
n.
T
he
t
hr
e
e
m
o
de
l
s
a
r
e
c
o
m
pi
l
e
d
s
pe
c
i
fy
i
ng
t
he
op
t
i
m
i
z
a
t
i
on
fun
c
t
i
on,
t
h
e
c
os
t
or
l
os
s
fun
c
t
i
on,
a
nd
t
he
m
e
t
r
i
c
s
.
W
e
us
e
t
h
e
s
t
oc
h
a
s
t
i
c
gra
d
i
e
nt
de
s
c
e
nt
op
t
i
m
i
z
a
t
i
on
fun
c
t
i
on,
t
h
e
c
a
t
e
gori
c
a
l
c
ros
s
-
e
n
t
ropy
func
t
i
on
,
w
h
i
c
h
c
a
n
b
e
us
e
d
t
o
r
e
fl
e
c
t
t
he
a
c
c
ura
c
y
of
t
h
e
p
re
di
c
t
i
ons
,
a
nd
for
t
h
e
m
e
t
r
i
c
s
,
a
c
c
ur
a
c
y
(or
h
i
t
ra
t
e
)
a
nd
m
s
e
(
m
e
a
n
of
t
he
qu
a
dr
a
t
i
c
e
rr
ors
).
3.
1
.
R
e
s
N
e
t
(
r
e
s
i
d
u
al
n
e
u
r
a
l
n
e
tw
o
r
k
)
T
hi
s
ne
t
w
ork
m
i
m
i
c
s
t
he
s
t
ru
c
t
ur
e
of
py
ra
m
i
d
a
l
c
e
l
l
s
i
n
t
h
e
c
e
re
br
a
l
c
ort
e
x.
T
hi
s
s
t
ruc
t
ure
i
s
a
c
hi
e
ve
d
by
j
u
m
pi
ng
(d
oubl
e
or
t
r
i
pl
e
)
o
ve
r
s
o
m
e
of
t
h
e
l
a
ye
rs
,
w
h
i
c
h
us
e
Re
L
u
(Re
c
t
i
fi
e
d
L
i
n
e
a
r
U
ni
t
s
)
a
c
t
i
v
a
t
i
on
func
t
i
on
a
s
s
how
n
i
n
F
i
gur
e
2.
F
i
gure
2
.
B
ui
l
di
ng
b
l
oc
k
(R
e
s
N
e
t
)
3.
2
.
D
e
n
s
e
N
e
t
(
d
e
n
s
e
c
on
vol
u
t
i
on
al
n
e
tw
o
r
k
)
T
he
D
e
ns
e
N
e
t
s
t
ru
c
t
u
re
a
l
s
o
h
a
s
s
i
m
i
l
a
r
j
u
m
ps
t
o
t
he
Re
s
N
e
t
,
bu
t
e
a
c
h
l
a
y
e
r
re
c
e
i
ve
s
i
npu
t
fr
om
t
he
pr
e
vi
o
us
l
a
y
e
rs
,
a
nd
c
onn
e
c
t
s
t
o
t
h
e
s
u
bs
e
qu
e
nt
l
a
y
e
rs
(
e
a
c
h
l
a
y
e
r
r
e
c
e
i
ve
s
know
l
e
dg
e
fro
m
t
he
pr
e
vi
ous
l
a
y
e
rs
a
s
s
how
n
i
n
F
i
g
ure
3
.
F
i
gure
3
.
B
ui
l
di
ng
b
l
oc
k
(D
e
ns
e
N
e
t
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
1
8
,
N
o.
2
,
A
pri
l
2
020:
8
0
7
-
8
1
4
810
3.
3
.
N
A
S
N
e
t
(
n
e
u
r
al
a
r
c
h
i
t
e
c
tu
r
e
s
e
a
r
c
h
n
e
tw
o
r
k
)
T
he
N
A
S
N
e
t
ne
t
w
ork
c
ons
i
s
t
s
of
a
s
p
e
c
i
fi
c
b
l
o
c
k,
t
h
e
be
s
t
c
on
vol
u
t
i
o
na
l
s
t
ru
c
t
ure
for
CIF
A
R
-
10
,
w
hi
c
h
i
s
t
h
e
n
g
e
n
e
ra
l
i
z
e
d
for
Im
a
g
e
N
e
t
,
a
nd
fi
na
l
l
y
re
p
l
i
c
a
t
e
d
a
s
a
bl
oc
k
for
l
a
rge
d
a
t
a
s
e
t
s
a
s
s
h
ow
n
i
n
F
i
gure
4
.
F
i
gure
4
.
I
m
a
ge
N
e
t
a
r
c
hi
t
e
c
t
ur
e
(N
A
S
N
e
t
)
4.
F
I
N
D
I
N
G
S
T
o
e
v
a
l
u
a
t
e
t
h
e
pe
rfor
m
a
n
c
e
of
t
he
t
hr
e
e
m
ode
l
s
,
i
n
a
dd
i
t
i
on
t
o
l
os
s
a
nd
a
c
c
ura
c
y
w
i
t
h
t
r
a
i
n
i
ng
a
nd
v
a
l
i
da
t
i
on
da
t
a
,
w
e
h
a
v
e
us
e
d
pre
c
i
s
i
o
n,
r
e
c
a
l
l
,
a
nd
F
1
-
s
c
ore
a
s
pe
rf
orm
a
n
c
e
m
e
t
ri
c
s
.
T
he
re
s
u
l
t
s
s
how
s
upe
ri
or
N
A
S
N
e
t
p
e
r
for
m
a
nc
e
ov
e
r
Re
s
N
e
t
a
nd
D
e
ns
e
N
e
t
.
Re
s
N
e
t
ha
d
t
he
poor
e
s
t
p
e
rfor
m
a
n
c
e
,
no
t
on
l
y
a
re
i
t
s
m
e
t
ri
c
s
v
e
ry
l
ow
,
b
ut
i
t
s
a
c
c
u
ra
c
y
d
oe
s
not
i
n
c
r
e
a
s
e
s
i
gn
i
fi
c
a
nt
l
y
w
i
t
h
l
os
s
re
du
c
t
i
on
,
a
nd
t
h
e
m
od
e
l
i
s
t
h
e
m
os
t
c
o
m
pl
e
x
(
ove
r
23
m
i
l
l
i
on
pa
r
a
m
e
t
e
rs
)
.
D
e
ns
e
N
e
t
h
a
s
s
i
m
i
l
a
r
p
e
rfor
m
a
n
c
e
bu
t
w
i
t
h
o
nl
y
7
m
i
l
l
i
on
pa
ra
m
e
t
e
rs
,
b
ut
w
i
t
h
s
t
i
l
l
ve
ry
l
ow
m
e
t
r
i
c
s
.
N
A
S
N
e
t
i
s
t
h
e
on
l
y
on
e
t
ha
t
g
e
t
s
a
n
a
c
c
e
pt
a
bl
e
p
e
rfor
m
a
nc
e
a
nd
w
i
t
h
a
l
ow
e
r
nu
m
b
e
r
o
f
p
a
ra
m
e
t
e
rs
(a
l
i
t
t
l
e
ov
e
r
4
m
i
l
l
i
o
n
).
S
u
m
m
a
ry
of
t
he
m
ode
l
:
Re
s
N
e
t
(
re
s
i
d
ua
l
n
e
ur
a
l
ne
t
w
ork
a
s
s
how
n
i
n
F
i
g
ure
s
5,
6
a
nd
7)
:
−
T
ot
a
l
pa
r
a
m
s
:
23,
5
91,
810
−
T
ra
i
na
b
l
e
p
a
ra
m
s
:
23
,
538
,
690
−
N
on
-
t
ra
i
n
a
bl
e
p
a
r
a
m
s
:
53
,
120
S
um
m
a
ry
of
t
h
e
m
od
e
l
:
D
e
ns
e
N
e
t
(D
e
ns
e
Co
nvol
u
t
i
o
na
l
N
e
t
w
ork
a
s
s
how
n
i
n
F
i
gu
re
s
8,
9
a
nd
10)
:
−
T
ot
a
l
pa
r
a
m
s
:
7,
03
9,
5
54
−
T
ra
i
na
b
l
e
p
a
ra
m
s
:
6,
955
,
906
−
N
on
-
t
ra
i
n
a
bl
e
p
a
r
a
m
s
:
83
,
648
S
um
m
a
ry
of
t
h
e
m
od
e
l
:
N
A
S
N
e
t
(N
e
ura
l
A
rc
h
i
t
e
c
t
ure
S
e
a
rc
h
N
e
t
w
ork
,
F
i
gure
s
11
,
1
2
a
nd
13)
:
−
T
ot
a
l
pa
r
a
m
s
:
4,
27
1,
8
30
−
T
ra
i
na
b
l
e
p
a
ra
m
s
:
4,
235
,
092
−
N
on
-
t
ra
i
n
a
bl
e
p
a
r
a
m
s
:
36
,
738
F
i
gure
5
.
T
ra
i
ni
ng
l
os
s
a
nd
a
c
c
ur
a
c
y
(R
e
s
N
e
t
)
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
E
v
a
l
uat
i
on
of
d
e
e
p
n
e
ur
al
ne
t
wor
k
ar
c
hi
t
e
c
t
ur
e
s
i
n
….
(
F
r
e
d
y
M
ar
t
í
n
e
z
)
811
F
i
gure
6
.
C
onfus
i
o
n
m
a
t
ri
x
(R
e
s
N
e
t
)
(a
)
(b)
F
i
gure
7
.
P
e
rfor
m
a
n
c
e
m
e
t
ri
c
s
(R
e
s
N
e
t
)
:
(a
)
Cl
a
s
s
i
fi
c
a
t
i
on
re
port
(R
e
s
N
e
t
),
(b)
RO
C
c
ur
ve
a
n
d
RO
C
a
re
a
(R
e
s
N
e
t
)
F
i
gure
8
.
T
ra
i
ni
ng
l
os
s
a
nd
a
c
c
ur
a
c
y
(D
e
ns
e
N
e
t
)
F
i
gure
9
.
C
onfus
i
o
n
m
a
t
ri
x
(D
e
ns
e
N
e
t
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
1
8
,
N
o.
2
,
A
pri
l
2
020:
8
0
7
-
8
1
4
812
(a
)
(b)
F
i
gure
10
.
P
e
rfor
m
a
n
c
e
m
e
t
ri
c
s
(D
e
ns
e
N
e
t
)
:
(a
)
Cl
a
s
s
i
fi
c
a
t
i
on
re
port
(D
e
ns
e
N
e
t
)
,
(b
)
RO
C
c
urv
e
a
nd
RO
C
a
r
e
a
(D
e
ns
e
N
e
t
)
F
i
gure
11
.
T
r
a
i
n
i
ng
l
os
s
a
nd
a
c
c
ura
c
y
(N
A
S
N
e
t
)
F
i
gure
12
.
Confus
i
on
m
a
t
ri
x
(N
A
S
N
e
t
)
(a
)
(b)
F
i
gure
13
.
P
e
rfor
m
a
n
c
e
m
e
t
ri
c
s
(N
A
S
N
e
t
)
:
(a
)
Cl
a
s
s
i
fi
c
a
t
i
on
re
port
(N
A
S
N
e
t
),
(b)
RO
C
c
urve
a
nd
RO
C
a
re
a
(N
A
S
N
e
t
)
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
E
v
a
l
uat
i
on
of
d
e
e
p
n
e
ur
al
ne
t
wor
k
ar
c
hi
t
e
c
t
ur
e
s
i
n
….
(
F
r
e
d
y
M
ar
t
í
n
e
z
)
813
T
he
c
onvo
l
ut
i
on
a
l
n
e
t
w
or
k
m
od
e
l
s
us
e
d
ha
v
e
a
n
opt
i
m
i
z
e
d
s
t
r
uc
t
ure
f
or
i
m
a
g
e
c
l
a
s
s
i
f
i
c
a
t
i
on
.
T
he
R
e
s
N
e
t
n
e
t
w
or
k
-
ba
s
e
d
m
od
e
l
a
c
h
i
e
v
e
s
a
c
ons
i
de
ra
b
l
e
r
e
duc
t
i
o
n
c
o
m
pa
r
e
d
t
o
t
he
nu
m
b
e
r
of
a
dj
us
t
a
bl
e
pa
ra
m
e
t
e
rs
for
a
d
e
e
p
n
e
t
w
ork
,
how
e
v
e
r
t
h
e
n
um
b
e
r
of
pa
r
a
m
e
t
e
rs
r
e
m
a
i
ns
hi
g
h,
a
n
d
t
he
be
s
t
o
pt
i
m
i
s
a
t
i
on
s
t
i
l
l
s
how
s
a
n
und
e
r
-
a
dj
us
t
m
e
nt
of
t
he
d
a
t
a
(
50%
a
c
c
u
ra
c
y).
T
he
D
e
ns
e
N
e
t
m
od
e
l
w
i
t
h
a
m
uc
h
d
e
ns
e
r
a
rc
h
i
t
e
c
t
ure
a
c
h
i
e
v
e
s
h
i
ghe
r
a
c
c
ura
c
y
t
ha
n
R
e
s
N
e
t
(58
%),
but
w
i
t
h
a
m
u
c
h
h
i
gh
e
r
nu
m
b
e
r
o
f
p
a
ra
m
e
t
e
rs
.
F
i
na
l
l
y
,
t
h
e
op
t
i
m
i
z
e
d
N
A
S
N
e
t
a
rc
h
i
t
e
c
t
ur
e
a
c
hi
e
v
e
s
t
h
e
hi
g
he
s
t
m
e
t
ri
c
va
l
u
e
s
(75%
a
c
c
ur
a
c
y
)
w
i
t
h
a
m
u
c
h
l
ow
e
r
nu
m
be
r
of
p
a
ra
m
e
t
e
rs
,
be
c
om
i
ng
t
h
e
r
i
gh
t
s
o
l
ut
i
on
t
o
t
he
p
robl
e
m
.
5.
C
O
N
C
LU
S
I
O
N
In
t
hi
s
p
a
p
e
r,
w
e
h
a
ve
e
va
l
u
a
t
e
d
t
h
e
p
e
rfor
m
a
n
c
e
of
t
h
re
e
c
onvo
l
ut
i
on
a
l
n
e
ur
a
l
n
e
t
w
or
ks
i
n
t
he
i
de
nt
i
fi
c
a
t
i
on
of
f
i
s
s
ure
s
on
bon
e
s
.
T
he
a
i
m
of
t
he
re
s
e
a
rc
h
i
s
t
o
f
i
nd
a
n
a
u
t
om
a
t
i
c
m
od
e
l
t
h
a
t
i
s
c
a
pa
bl
e
of
p
roc
e
s
s
i
ng
r
a
di
ol
og
i
c
a
l
i
m
a
ge
s
a
nd
g
i
vi
n
g
a
p
re
l
i
m
i
n
a
ry
di
a
gn
os
i
s
of
pos
s
i
b
l
e
bone
fi
s
s
ure
s
,
i
n
t
h
e
hop
e
of
r
e
duc
i
ng
t
he
prob
a
bi
l
i
t
y
of
m
i
s
di
a
gnos
i
s
,
i
nc
r
e
a
s
i
ng
t
h
e
pe
r
c
e
nt
a
ge
of
p
a
t
i
e
n
t
s
a
t
t
e
nd
e
d
a
nd
i
m
pr
ovi
n
g
t
he
qua
l
i
t
y
of
m
e
di
c
a
l
s
e
r
vi
c
e
.
T
he
s
e
l
e
c
t
e
d
n
e
t
w
o
rks
w
e
re
:
R
e
s
N
e
t
(
re
s
i
du
a
l
ne
ur
a
l
n
e
t
w
ork
)
,
D
e
ns
e
N
e
t
(
d
e
ns
e
c
onvo
l
ut
i
ona
l
ne
t
w
ork
)
,
a
nd
N
A
S
N
e
t
(
ne
ura
l
a
rc
hi
t
e
c
t
ur
e
s
e
a
rc
h
n
e
t
w
o
rk
).
T
he
pe
rfor
m
a
n
c
e
of
e
a
c
h
of
t
h
e
m
od
e
l
s
w
a
s
e
v
a
l
u
a
t
e
d
b
y
c
a
l
c
u
l
a
t
i
ng
t
h
e
pr
e
c
i
s
i
on
,
r
e
c
a
l
l
,
a
n
d
F
1
-
s
c
ore
m
e
t
r
i
c
s
.
T
he
m
od
e
l
s
w
e
re
a
l
s
o
us
e
d
t
o
e
v
a
l
ua
t
e
l
os
s
a
n
d
a
c
c
ur
a
c
y
w
i
t
h
t
ra
i
ni
ng
a
nd
v
a
l
i
d
a
t
i
on
d
a
t
a
.
D
e
t
a
i
l
s
of
t
he
nu
m
be
r
of
p
a
r
a
m
e
t
e
rs
of
e
a
c
h
m
od
e
l
,
c
onfus
i
on
m
a
t
r
i
c
e
s
a
nd
RO
C
c
ur
ve
w
e
r
e
a
l
s
o
s
how
n
.
A
f
t
e
r
a
na
l
yz
i
ng
t
h
e
be
h
a
v
i
or
of
t
h
e
m
ode
l
s
,
i
t
w
a
s
found
t
ha
t
o
nl
y
t
h
e
N
A
S
N
e
t
n
e
t
w
ork
pr
oduc
e
s
a
n
a
c
c
e
p
t
a
b
l
e
c
l
a
s
s
i
fi
c
a
t
i
on
for
t
he
pro
bl
e
m
.
T
h
e
pre
c
i
s
i
on
v
a
l
u
e
s
of
t
he
N
A
S
N
e
t
m
ode
l
w
e
r
e
h
i
gh
e
r
t
ha
n
t
he
ot
he
r
t
w
o
m
od
e
l
s
.
S
i
m
i
l
a
r
b
e
ha
vi
or
w
a
s
obs
e
rve
d
i
n
t
h
e
o
t
h
e
r
c
a
l
c
ul
a
t
e
d
m
e
t
r
i
c
s
.
I
n
a
dd
i
t
i
on,
t
he
N
A
S
N
e
t
m
o
de
l
i
s
t
he
s
m
a
l
l
e
s
t
of
t
he
t
hre
e
,
r
e
qu
i
ri
n
g
a
l
i
t
t
l
e
m
ore
t
ha
n
4
m
i
l
l
i
on
t
ra
i
na
b
l
e
p
a
ra
m
e
t
e
rs
,
c
o
m
pa
r
e
d
t
o
7
m
i
l
l
i
on
i
n
t
h
e
D
e
ns
e
N
e
t
m
ode
l
a
nd
m
ore
t
ha
n
23
m
i
l
l
i
o
n
i
n
t
h
e
Re
s
N
e
t
m
ode
l
.
T
he
s
e
re
s
u
l
t
s
a
re
i
m
por
t
a
n
t
for
t
he
c
o
rre
c
t
s
e
l
e
c
t
i
on
of
a
n
a
ut
om
a
t
e
d
d
i
a
g
nos
t
i
c
m
ode
l
,
a
nd
s
how
t
ha
t
i
t
i
s
pos
s
i
bl
e
t
o
i
m
pro
ve
t
he
p
e
rfor
m
a
nc
e
of
t
hi
s
m
ode
l
t
hr
ough
a
l
a
r
ge
r
s
e
t
of
t
r
a
i
n
i
ng
i
m
a
ge
s
a
nd
b
e
t
t
e
r
t
un
i
ng
of
pa
r
a
m
e
t
e
rs
.
F
ut
ure
w
ork
w
i
l
l
f
oc
us
on
i
m
provi
ng
t
he
f
i
t
of
t
hi
s
ne
t
w
ork
by
a
l
t
e
ri
ng
i
t
s
d
e
pt
h
a
nd
us
i
ng
i
m
a
ge
s
w
i
t
h
m
ore
v
i
s
ua
l
i
nfor
m
a
t
i
on
.
A
C
K
N
O
WL
ED
G
E
M
EN
TS
T
hi
s
w
ork
w
a
s
s
u
pport
e
d
by
t
h
e
U
ni
v
e
rs
i
d
a
d
D
i
s
t
r
i
t
a
l
F
ra
n
c
i
s
c
o
J
os
é
d
e
Ca
l
da
s
,
i
n
pa
r
t
t
hrou
gh
CID
C
,
a
nd
p
a
rt
l
y
by
t
he
F
a
c
u
l
t
a
d
T
e
c
n
ol
óg
i
c
a
.
T
h
e
vi
e
w
s
e
xpr
e
s
s
e
d
i
n
t
h
i
s
pa
p
e
r
a
re
n
ot
n
e
c
e
s
s
a
ri
l
y
e
n
dors
e
d
by
D
i
s
t
r
i
c
t
U
n
i
ve
rs
i
t
y
.
T
he
a
ut
hors
t
h
a
nk
t
h
e
r
e
s
e
a
rc
h
gr
oup
A
RM
O
S
fo
r
t
he
e
v
a
l
u
a
t
i
on
c
a
r
ri
e
d
ou
t
o
n
prot
ot
ype
s
of
i
d
e
a
s
a
nd
s
t
r
a
t
e
gi
e
s
.
R
EF
ER
EN
C
ES
[
1]
M
.
H
u
s
s
a
i
n
,
A
.
B
hui
ya
n
,
A
.
T
ur
p
i
n
,
C
.
L
uu,
R
.
S
m
i
t
h
,
R
.
G
u
y
m
e
r
,
a
nd
R
.
,
“
K
o
t
a
g
i
r
i
.
A
ut
o
m
a
t
i
c
i
de
n
t
i
f
i
c
a
t
i
on
o
f
pa
t
h
ol
ogy
-
di
s
t
or
t
e
d
r
e
t
i
na
l
l
a
ye
r
boun
da
r
i
e
s
u
s
i
ng
s
d
-
oc
t
i
m
a
gi
n
g
,”
I
E
E
E
T
r
ans
a
c
t
i
on
s
on
B
i
om
e
di
c
al
E
ngi
ne
e
r
i
ng
,
vol
.
64
,
n
o.
7
,
pp.
1
638
–
1649
,
201
7.
[
2]
O
.
Z
e
nt
e
no
,
F
.
Z
v
i
e
t
c
ov
i
c
h,
D
.
Z
a
p
a
t
a
,
H
.
M
a
r
ue
n
da
,
B
.
V
a
l
e
nc
i
a
,
A
.
L
l
a
n
os
,
J
.
A
r
e
va
l
o
,
M
.
M
on
t
e
r
o,
R
.
L
a
va
r
e
l
l
o
,
a
nd
B
.
C
a
s
t
a
ne
d
a
,
“
A
n
i
n
t
e
g
r
a
t
e
d
pr
o
t
o
c
ol
f
o
r
t
he
r
e
s
e
a
r
c
h
a
n
d
m
on
i
t
o
r
i
ng
o
f
c
ut
a
ne
ous
l
e
i
s
h
m
a
n
i
a
s
i
s
,”
I
E
E
E
L
at
i
n
A
m
e
r
i
c
a
T
r
a
ns
a
c
t
i
on
s
,
v
ol
.
15
,
n
o.
1
1
,
pp
.
2164
–
217
0,
20
17
.
[
3]
S
.
M
a
kr
og
i
a
nn
i
s
,
K
.
F
i
s
hbe
i
n
,
A
.
M
oo
r
e
,
R
.
S
pe
n
c
e
r
,
a
nd
L
.
F
e
r
r
u
c
c
i
,
“
I
m
a
ge
-
ba
s
e
d
t
i
s
s
u
e
d
i
s
t
r
i
bu
t
i
on
m
ode
l
i
ng
f
or
s
ke
l
e
t
a
l
m
u
s
c
l
e
qua
l
i
t
y
c
ha
r
a
c
t
e
r
i
z
a
t
i
o
n
,”
I
E
E
E
T
r
ans
a
c
t
i
on
s
on
B
i
om
e
di
c
al
E
n
gi
n
e
e
r
i
n
g
,
vo
l
.
63
,
n
o.
4
,
pp.
80
5
–
81
3,
20
16.
[
4]
A
.
A
na
n
d,
I
.
M
oo
n,
a
nd
B
.
J
a
vi
di
,
“
A
u
t
o
m
a
t
e
d
di
s
e
a
s
e
i
de
nt
i
f
i
c
a
t
i
o
n
w
i
t
h
3
-
d
op
t
i
c
a
l
i
m
a
g
i
n
g:
A
m
e
d
i
c
a
l
di
a
gno
s
t
i
c
t
oo
l
,”
P
r
oc
e
e
d
i
ng
s
of
t
he
I
E
E
E
,
vo
l
.
1
05
,
n
o.
5
,
pp
.
9
24
–
9
46
,
2
017
.
[
5]
J
.
Z
hou
,
T
.
Z
hong
,
a
nd
X
.
H
e
,
“
A
ux
i
l
i
a
r
y
di
a
gno
s
i
s
o
f
b
r
e
a
s
t
t
u
m
or
ba
s
e
d
on
pn
n
c
l
a
s
s
i
f
i
e
r
op
t
i
m
i
z
e
d
b
y
p
c
a
a
nd
ps
o
a
l
go
r
i
t
h
m
,”
I
n
9
t
h
I
nt
e
r
n
at
i
ona
l
C
on
f
e
r
e
n
c
e
o
n
I
n
t
e
l
l
i
g
e
nt
H
um
an
-
M
ac
hi
n
e
Sy
s
t
e
m
s
a
nd
C
y
b
e
r
n
e
t
i
c
s
(
I
H
M
SC
201
7)
,
vol
.
2,
p
p.
222
–
227
,
201
7.
[
6]
N
.
T
s
a
i
,
J
.
G
oodw
i
n
,
M
.
S
e
m
l
e
r
,
R
.
K
ot
h
e
r
a
,
M
.
V
a
n
H
o
r
n
,
B
.
W
o
l
f
,
a
nd
D
.
G
a
r
n
e
r
,
“
D
e
v
e
l
op
m
e
n
t
of
a
no
n
-
i
nv
a
s
i
ve
bl
i
nk
r
e
f
l
e
x
o
m
e
t
e
r
,”
I
E
E
E
J
our
nal
of
T
r
an
s
l
a
t
i
ona
l
E
ng
i
n
e
e
r
i
ng
i
n
H
e
al
t
h
an
d
M
e
d
i
c
i
ne
,
vo
l
.
5
,
n
o.
1
,
p
p.
1
–
4,
2017
.
[
7]
O
.
B
e
r
t
e
l
,
C
.
M
o
r
e
no
,
a
nd
E
.
T
or
o
,
“
A
p
l
i
c
a
c
i
ó
n
de
l
a
t
r
a
ns
f
or
m
a
d
a
w
a
v
e
l
e
t
pa
r
a
e
l
r
e
c
ono
c
i
m
i
e
n
t
o
de
f
o
r
m
a
s
e
n
vi
s
i
ón
a
r
t
i
f
i
c
i
a
l
,”
T
e
k
hnê
,
vo
l
.
6
,
no
.
1
,
pp
.
3
–
8
,
2009
.
[
8]
Y
.
G
uo
,
L
.
J
i
a
o
,
S
.
W
a
ng,
S
.
W
a
ng
,
F
.
L
i
u,
a
n
d
W
.
H
ua
,
“
F
uz
z
y
s
u
pe
r
pi
xe
l
s
f
o
r
p
ol
a
r
i
m
e
t
r
i
c
s
a
r
i
m
a
ge
s
c
l
a
s
s
i
f
i
c
a
t
i
on
,”
I
E
E
E
T
r
an
s
ac
t
i
ons
o
n
F
uz
z
y
Sy
s
t
e
m
s
,
vo
l
.
26
,
no
.
5
,
pp
.
2846
–
286
0,
20
18
.
[
9]
J
.
C
a
s
t
a
ñe
d
a
a
nd
Y
.
S
a
l
gue
r
o
,
“
A
dj
us
t
m
e
n
t
of
v
i
s
u
a
l
i
de
n
t
i
f
i
c
a
t
i
on
a
l
go
r
i
t
h
m
f
or
u
s
e
i
n
s
t
a
nd
-
a
l
one
r
obo
t
na
vi
g
a
t
i
on
a
ppl
i
c
a
t
i
on
s
,”
R
e
v
i
s
t
a
T
e
k
hnê
,
vo
l
.
14
,
no
.
1
,
p
p.
73
–
86
,
2017
.
[
10]
P
.
J
or
i
s
,
W
.
D
e
ve
l
t
e
r
,
W
.
V
oo
r
d
e
,
P
.
S
u
e
t
e
ns
,
F
.
M
a
e
s
,
D
.
V
a
nde
r
m
e
u
l
e
n
,
a
nd
P
.
C
l
a
e
s
,
“
P
r
e
pr
oc
e
s
s
i
ng
of
he
t
e
r
o
s
c
e
da
s
t
i
c
m
e
d
i
c
a
l
i
m
a
ge
s
,”
I
E
E
E
A
c
c
e
s
s
,
vo
l
.
6
,
no
.
1
,
pp
.
26
047
–
2
6058
,
201
8.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
1
8
,
N
o.
2
,
A
pri
l
2
020:
8
0
7
-
8
1
4
814
[
11]
V
.
J
a
o
ue
n
,
J
.
B
e
r
t
,
N
.
B
ous
s
i
o
n,
H
.
F
a
y
a
d
,
M
.
H
a
t
t
,
a
n
d
D
.
V
i
s
vi
ki
s
,
“
I
m
a
g
e
e
n
ha
n
c
e
m
e
n
t
w
i
t
h
P
D
E
s
a
nd
non
c
on
s
e
r
va
t
i
ve
a
dv
e
c
t
i
on
f
l
ow
f
i
e
l
ds
,”
I
E
E
E
T
r
an
s
ac
t
i
ons
on
I
m
age
P
r
oc
e
s
s
i
ng
,
vo
l
.
28
,
no
.
6
,
pp.
30
75
–
3
088
,
2019
.
[
12]
B
.
L
a
s
s
e
n,
E
.
va
n
R
i
kx
oor
t
,
M
.
S
c
h
m
i
dt
,
S
.
K
e
r
ks
t
r
a
,
B
.
va
n
G
i
nn
e
k
e
n,
a
nd
J
.
K
uhn
i
gk
,
“
A
ut
o
m
a
t
i
c
s
e
g
m
e
nt
a
t
i
on
of
t
he
pu
l
m
ona
r
y
l
o
be
s
f
r
o
m
c
he
s
t
c
t
s
c
a
n
s
b
a
s
e
d
o
n
f
i
s
s
u
r
e
s
,
ve
s
s
e
l
s
,
a
nd
br
onc
hi
,”
I
E
E
E
T
r
ans
a
c
t
i
on
s
on
M
e
di
c
al
I
m
ag
i
ng
,
vol
.
32
,
no
.
2
,
pp
.
210
–
222
,
2013
.
[
13]
P
.
M
o
e
s
k
ops
,
M
.
V
i
e
r
ge
ve
r
,
A
.
M
e
nd
r
i
k,
L
.
d
e
V
r
i
e
s
,
M
.
B
e
nde
r
s
,
a
nd
I
.
I
s
gu
m
,
“
A
u
t
o
m
a
t
i
c
s
e
gm
e
n
t
a
t
i
o
n
o
f
m
r
br
a
i
n
i
m
a
ge
s
w
i
t
h
a
c
onvo
l
u
t
i
ona
l
ne
u
r
a
l
ne
t
w
or
k
,”
I
E
E
E
T
r
a
n
s
ac
t
i
o
ns
on
M
e
d
i
c
a
l
I
m
a
gi
n
g
,
vo
l
.
35
,
n
o.
5
,
pp.
12
52
–
1
261
,
2016
.
[
14]
G
.
W
a
ng
,
W
.
L
i
,
M
.
Z
u
l
ua
g
a
,
R
.
P
r
a
t
t
,
P
.
P
a
t
e
l
,
M
.
A
e
r
t
s
e
n
,
T
.
D
oe
l
,
A
.
D
a
v
i
d
,
J
.
D
e
pr
e
s
t
,
S
.
O
ur
s
e
l
i
n
,
a
nd
T
.
V
e
r
c
a
ut
e
r
e
n
,
“
I
n
t
e
r
a
c
t
i
ve
m
e
di
c
a
l
i
m
a
g
e
s
e
g
m
e
nt
a
t
i
on
us
i
ng
de
e
p
l
e
a
r
n
i
n
g
w
i
t
h
i
m
a
ge
-
s
p
e
c
i
f
i
c
f
i
ne
t
uni
ng
,”
I
E
E
E
T
r
ans
ac
t
i
ons
on
M
e
d
i
c
a
l
I
m
a
gi
n
g
,
vo
l
.
37
,
n
o.
7
,
pp.
1
562
–
1573
,
2
018.
[
15]
J
.
S
e
dl
a
r
,
M
.
B
a
j
ge
r
,
M
.
C
a
on
,
a
n
d
G
.
L
e
e
,
“
M
o
de
l
-
g
ui
de
d
s
e
g
m
e
n
t
a
t
i
on
o
f
l
i
ve
r
i
n
C
T
a
nd
P
E
T
-
C
T
i
m
a
ge
s
of
c
hi
l
d
pa
t
i
e
n
t
s
ba
s
e
d
on
s
t
a
t
i
s
t
i
c
a
l
r
e
gi
on
m
e
r
g
i
ng
,”
I
n
I
nt
e
r
n
at
i
on
al
C
o
nf
e
r
e
n
c
e
on
D
i
g
i
t
.
I
m
ag
e
C
om
p
ut
.
T
e
c
h
.
A
pp
l
.
(
D
I
C
T
A
20
16)
,
p
p.
1
–
8
,
2016
.
[
16]
H
.
T
a
ng
,
C
.
Z
ha
n
g,
a
n
d
X
.
X
i
e
,
“
A
ut
o
m
a
t
i
c
pu
l
m
on
a
r
y
l
obe
s
e
g
m
e
nt
a
t
i
on
us
i
ng
d
e
e
p
l
e
a
r
ni
ng
,”
2019
I
E
E
E
1
6t
h
I
nt
e
r
n
at
i
on
al
Sy
m
pos
i
um
on
B
i
om
e
d
i
c
a
l
I
m
a
gi
n
g
(
I
SB
I
20
19)
,
vo
l
.
1
,
no
.
1
,
pp
.
122
5
–
12
28,
2
019
.
[
17]
J
.
M
i
ng
,
W
.
K
i
m
,
a
nd
K
.
R
yo
ung
,
“
F
i
nge
r
-
ve
i
n
r
e
c
ogn
i
t
i
o
n
ba
s
e
d
o
n
d
e
e
p
D
e
n
s
e
N
e
t
us
i
ng
c
om
pos
i
t
e
i
m
a
ge
,”
I
E
E
E
A
c
c
e
s
,
vo
l
.
7
,
no
.
1
,
pp
.
668
45
–
6
6863
,
201
9.
[
18]
V
.
N
e
a
go
e
,
A
.
C
i
o
t
e
c
,
a
nd
G
.
C
u
c
u
,
“
D
e
e
p
c
onv
ol
u
t
i
ona
l
ne
u
r
a
l
ne
t
w
or
ks
v
e
r
s
us
m
ul
t
i
l
a
ye
r
pe
r
c
e
p
t
r
on
f
o
r
f
i
na
nc
i
a
l
pr
e
d
i
c
t
i
on
,”
I
n
I
nt
e
r
n
at
i
on
al
C
on
f
e
r
e
n
c
e
on
C
om
m
un
i
c
a
t
i
on
s
(
C
O
M
M
2
018)
,
pp
.
201
–
206
,
201
8.
[
19]
S
.
C
ui
,
Y
.
L
u
o,
H
.
T
s
e
n
g,
R
.
H
a
ke
n,
a
nd
I
.
E
l
N
a
qa
,
“
A
r
t
i
f
i
c
i
a
l
n
e
ur
a
l
ne
t
w
o
r
k
w
i
t
h
c
o
m
pos
i
t
e
a
r
c
h
i
t
e
c
t
ur
e
s
f
o
r
pr
e
d
i
c
t
i
on
of
l
oc
a
l
c
on
t
r
ol
i
n
r
a
d
i
o
t
he
r
a
p
y
,”
I
E
E
E
T
r
a
ns
a
c
t
i
on
s
on
R
adi
a
t
i
on
and
P
l
as
m
a
M
e
d
i
c
al
Sc
i
e
n
c
e
s
,
vo
l
.
3
,
no.
2
,
pp.
2
42
–
2
49,
2
019
.
[
20]
H
.
T
ong
a
nd
Z
.
a
nd
Z
hi
,
“
B
a
g
of
t
r
i
c
k
s
f
or
i
m
a
ge
c
l
a
s
s
i
f
i
c
a
t
i
on
w
i
t
h
c
onvo
l
u
t
i
ona
l
ne
ur
a
l
n
e
t
w
o
r
k
s
,”
I
n
I
E
E
E
C
onf
e
r
e
n
c
e
on
C
om
p
ut
e
r
V
i
s
i
on
and
P
a
t
t
e
r
n
R
e
c
og
ni
t
i
on
(
C
V
P
R
2
019)
,
p
p.
558
–
567
,
2019
.
[
21]
X
.
Z
ha
o
,
T
.
Z
h
a
ng
,
H
.
L
i
u
,
G
.
Z
hu
,
a
nd
X
.
Z
ou
,
“
A
ut
o
m
a
t
i
c
w
i
ndo
w
i
ng
f
or
m
r
i
w
i
t
h
c
on
vol
ut
i
on
a
l
ne
u
r
a
l
ne
t
w
o
r
k
,”
I
E
E
E
A
c
c
e
s
s
,
vo
l
.
7
,
no
.
1
,
pp
.
685
94
–
6
8606
,
201
9.
[
22]
H
.
M
o
nt
i
e
l
,
E
.
J
a
c
i
n
t
o
,
a
nd
F
.
M
a
r
t
í
n
e
z
,
“
R
e
c
ogn
i
t
i
o
n
of
f
i
s
s
u
r
e
s
i
n
bo
ny
s
t
r
uc
t
ur
e
s
t
hr
ough
i
m
a
g
e
p
r
oc
e
s
s
i
ng
,”
I
nt
e
r
n
at
i
on
al
J
o
ur
n
al
o
f
E
ngi
ne
e
r
i
ng
and
T
e
c
hno
l
o
gy
,
v
ol
.
10
,
no.
4
,
pp
.
1223
–
122
9,
20
18.
[
23]
T
.
S
a
v
i
t
hr
i
a
nd
S
.
D
e
v
i
,
“
N
odu
l
e
de
t
e
c
t
i
on
f
r
o
m
p
os
t
e
r
i
o
r
a
n
d
a
n
t
e
r
i
or
c
he
s
t
r
a
d
i
og
r
a
phs
w
i
t
h
di
f
f
e
r
e
nt
m
e
t
h
ods
,”
I
n
F
ut
ur
e
T
e
c
hn
ol
o
gi
e
s
C
on
f
e
r
e
nc
e
(
F
T
C
20
16)
,
p
p.
504
–
515
,
2016
.
[
24]
G
.
H
ua
ng
,
S
.
L
i
u,
L
.
M
a
a
t
e
n
,
a
n
d
K
.
W
e
i
nbe
r
g
e
r
,
“
C
on
de
n
s
e
n
e
t
:
A
n
e
f
f
i
c
i
e
n
t
de
n
s
e
n
e
t
u
s
i
ng
l
e
a
r
ne
d
gr
o
up
c
onvo
l
u
t
i
ons
,”
I
n
I
E
E
E
/
C
V
F
C
on
f
e
r
e
nc
e
on
C
om
pu
t
e
r
V
i
s
i
on
an
d
P
at
t
e
r
n
R
e
c
og
n
i
t
i
on
, p
p.
2
752
–
2761
,
201
8.
[
25]
J
.
S
e
n
a
n
d
B
.
N
e
i
l
,
“
E
m
l
-
ne
t
:
A
n
e
xp
a
nd
a
b
l
e
m
u
l
t
i
-
l
a
y
e
r
n
e
t
w
or
k
f
o
r
s
a
l
i
e
nc
y
pr
e
d
i
c
t
i
o
n
,”
A
r
X
i
v
1
805
(
010
47)
,
pp.
1
–
10,
2
018
.
Evaluation Warning : The document was created with Spire.PDF for Python.