T
E
L
K
O
M
NIKA
T
elec
o
mm
un
ica
t
io
n,
Co
m
pu
t
ing
,
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
18
,
No
.
3
,
J
u
n
e
2
0
2
0
,
p
p
.
130
2
~
13
09
I
SS
N:
1
6
9
3
-
6930
,
ac
cr
ed
ited
First Gr
ad
e
b
y
Ke
m
e
n
r
is
te
k
d
i
k
ti,
Dec
r
ee
No
: 2
1
/E/KPT
/2
0
1
8
DOI
: 1
0
.
1
2
9
2
8
/
T
E
L
KOM
NI
KA
.
v
1
8
i3
.
1
4
7
5
1
1302
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//jo
u
r
n
a
l.u
a
d
.
a
c.
id
/in
d
ex
.
p
h
p
/TELK
OM
N
I
K
A
Cla
ss
ificatio
n of p
neu
m
o
nia
fro
m
X
-
ra
y
i
m
a
g
es
using
sia
m
ese
con
v
o
lutiona
l net
w
o
r
k
K
enna
rd
Alca
nd
er
P
ra
y
o
g
o
,
Alet
hea
Su
ry
a
dib
ra
t
a
,
J
uli
o
Chris
t
ia
n Yo
un
g
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
c
s,
Un
iv
e
rsitas
M
u
lt
im
e
d
ia Nu
sa
n
tara
,
In
d
o
n
e
sia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
1
5
,
2
0
1
9
R
ev
i
s
ed
J
an
2
0
,
2
0
2
0
A
cc
ep
ted
Feb
2
3
,
2
0
2
0
P
n
e
u
m
o
n
ia
is
o
n
e
o
f
t
h
e
h
i
g
h
e
s
t
g
l
o
b
a
l
c
a
u
s
e
s
o
f
d
e
a
t
h
s
e
s
p
e
c
i
a
l
l
y
f
o
r
c
h
i
l
d
r
e
n
u
n
d
e
r
5
y
e
a
r
s
o
l
d
.
T
h
i
s
h
a
p
p
e
n
e
d
m
a
i
n
l
y
b
e
c
a
u
s
e
o
f
t
h
e
d
i
f
f
i
c
u
l
t
i
e
s
i
n
i
d
e
n
t
i
f
y
i
n
g
t
h
e
c
a
u
s
e
o
f
p
n
e
u
m
o
n
i
a
.
A
s
a
r
e
s
u
l
t
,
t
h
e
t
r
e
a
t
m
e
n
t
g
i
v
e
n
m
a
y
n
o
t
b
e
s
u
i
t
a
b
l
e
f
o
r
e
a
c
h
p
n
e
u
m
o
n
i
a
c
a
s
e
.
R
e
c
e
n
t
s
t
u
d
i
e
s
h
a
v
e
u
s
e
d
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
e
s
t
o
o
b
t
a
i
n
b
e
t
t
e
r
c
l
a
s
s
i
f
i
c
a
t
i
o
n
w
i
t
h
i
n
t
h
e
c
a
u
s
e
o
f
p
n
e
u
m
o
n
i
a
.
I
n
t
h
i
s
r
e
s
e
a
r
c
h
,
w
e
u
s
e
d
s
i
a
m
e
s
e
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
(
S
C
N
)
t
o
c
l
a
s
s
i
f
y
c
h
e
s
t
x
-
r
a
y
p
n
e
u
m
o
n
i
a
i
m
a
g
e
i
n
t
o
3
c
l
a
s
s
e
s
,
n
a
m
e
l
y
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
s
,
b
a
c
t
e
r
i
a
l
p
n
e
u
m
o
n
i
a
,
a
n
d
v
i
r
a
l
p
n
e
u
m
o
n
i
a
.
S
i
a
m
e
s
e
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
i
s
a
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
t
h
a
t
l
e
a
r
n
s
s
i
m
i
l
a
r
i
t
y
k
n
o
w
l
e
d
g
e
b
e
t
w
e
e
n
p
a
i
r
s
o
f
i
m
a
g
e
i
n
p
u
t
s
b
a
s
e
d
o
n
t
h
e
d
i
f
f
e
r
e
n
c
e
s
b
e
t
w
e
e
n
i
t
s
f
e
a
t
u
r
e
s
.
O
n
e
o
f
t
h
e
i
m
p
o
r
t
a
n
t
b
e
n
e
f
i
t
s
o
f
c
l
a
s
s
i
f
y
i
n
g
d
a
t
a
w
i
t
h
S
C
N
i
s
t
h
e
a
v
a
i
l
a
b
i
l
i
t
y
o
f
c
o
m
p
a
r
a
b
l
e
i
m
a
g
e
s
t
h
a
t
c
a
n
b
e
u
s
e
d
a
s
a
r
e
f
e
r
e
n
c
e
w
h
e
n
d
e
t
e
r
m
i
n
i
n
g
c
l
a
s
s
.
U
s
i
n
g
S
C
N
,
o
u
r
b
e
s
t
m
o
d
e
l
a
c
h
i
e
v
e
d
8
0
.
0
3
%
a
c
c
u
r
a
c
y
,
7
9
.
5
9
%
f
1
s
c
o
r
e
,
a
n
d
a
n
i
m
p
r
o
v
e
d
r
e
s
u
l
t
r
e
a
s
o
n
i
n
g
b
y
p
r
o
v
i
d
i
n
g
t
h
e
c
o
m
p
a
r
a
b
l
e
i
m
a
g
e
s
.
K
ey
w
o
r
d
s
:
C
h
e
s
t X
-
R
a
y
I
m
ag
e
cla
s
s
i
f
icatio
n
P
n
eu
m
o
n
ia
Sia
m
e
s
e
C
o
n
v
o
l
u
tio
n
al
Net
w
o
r
k
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r
th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
A
let
h
ea
S
u
r
y
ad
ib
r
ata,
Dep
ar
t
m
en
t o
f
I
n
f
o
r
m
atic
s
,
Un
i
v
er
s
ita
s
Mu
l
ti
m
ed
ia
N
u
s
a
n
tar
a,
Scien
tia
B
o
u
le
v
ar
d
St.
,
Gad
in
g
Ser
p
o
n
g
,
T
an
g
er
a
n
g
,
B
an
te
n
1
5
8
1
1
,
I
n
d
o
n
esia
E
m
ail: a
let
h
ea
@
u
m
n
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
P
n
e
u
m
o
n
i
a
i
s
o
n
e
o
f
t
h
e
l
e
a
d
i
n
g
c
a
u
s
e
s
o
f
m
o
r
t
a
l
i
t
y
i
n
c
h
i
l
d
r
e
n
u
n
d
e
r
5
y
e
a
r
s
o
l
d
b
e
s
i
d
e
s
p
r
e
t
e
r
m
b
i
r
t
h
c
o
m
p
l
i
c
a
t
i
o
n
s
[
1
]
.
I
n
2
0
1
5
,
p
n
e
u
m
o
n
i
a
i
s
r
e
s
p
o
n
s
i
b
l
e
f
o
r
a
r
o
u
n
d
1
5
%
o
f
a
l
l
d
e
a
t
h
s
o
r
a
l
m
o
s
t
9
2
0
,
1
3
6
n
u
m
b
e
r
o
f
c
h
i
l
d
r
e
n
i
n
t
h
is
ag
e
g
r
o
u
p
[
2
]
.
P
n
eu
m
o
n
ia
i
s
a
n
i
n
f
la
m
m
atio
n
o
f
l
u
n
g
ti
s
s
u
e
a
s
a
r
es
u
lt o
f
i
n
f
ec
tio
u
s
a
g
en
t
s
[
3
]
.
T
h
e
m
o
s
t
co
m
m
o
n
t
y
p
e
o
f
p
n
eu
m
o
n
ia
i
s
b
ac
ter
ial
p
n
eu
m
o
n
ia
an
d
v
ir
al
p
n
eu
m
o
n
ia
[
4
]
.
B
ac
ter
ial
p
n
eu
m
o
n
ia
is
ca
u
s
ed
b
y
b
ac
ter
ia
s
u
ch
as
Stre
p
to
co
cc
u
s
p
n
e
u
m
o
n
i
ae
Hae
m
o
p
h
il
u
s
in
f
l
u
e
n
za
e
t
y
p
e
b
(
Hib
)
,
an
d
M
y
co
p
las
m
a
p
n
eu
m
o
n
iae
[
5
]
.
Vir
al
p
n
eu
m
o
n
ia
i
s
o
f
te
n
c
au
s
ed
b
y
r
esp
ir
ato
r
y
v
ir
u
s
i
n
clu
d
in
g
i
n
f
lu
e
n
za
,
p
ar
ain
f
l
u
en
za
v
ir
u
s
,
an
d
ad
en
o
v
ir
u
s
[
6
]
.
T
h
e
m
ai
n
p
r
o
b
lem
o
f
p
n
eu
m
o
n
ia
tr
ea
t
m
e
n
t
is
th
e
d
i
f
f
ic
u
lt
y
to
m
a
k
e
a
clin
ical
d
ec
is
io
n
as
a
r
esu
lt
o
f
th
e
i
n
ab
ilit
y
to
id
en
tify
th
e
i
n
f
ec
tio
u
s
o
r
g
an
i
s
m
[
7
]
.
T
o
o
v
er
co
m
e
t
h
is
p
r
o
b
le
m
,
p
n
e
u
m
o
n
i
a
d
is
ea
s
e
is
u
s
u
a
ll
y
tr
ea
ted
u
s
i
n
g
a
n
tib
io
tics
.
Ho
wev
er
,
th
e
u
s
e
o
f
a
n
tib
io
tics
to
t
r
ea
t v
ir
al
p
n
eu
m
o
n
ia
is
i
n
e
f
f
ec
tiv
e
[
4
]
an
d
m
i
s
u
s
e
o
f
an
tib
io
tics
ca
n
in
cr
ea
s
e
t
h
e
r
is
k
o
f
an
t
ib
io
tic
r
esis
tan
ce
[
7
]
.
T
h
u
s
,
it
is
i
m
p
o
r
tan
t
to
id
en
ti
f
y
th
e
t
y
p
e
o
f
m
icr
o
o
r
g
an
i
s
m
o
n
th
e
d
ia
g
n
o
s
is
o
f
p
n
e
u
m
o
n
ia
to
r
ec
eiv
e
s
u
i
tab
le
tr
ea
t
m
en
ts
.
Me
d
ical
i
m
a
g
in
g
tec
h
n
iq
u
es
h
as
b
ee
n
u
s
ed
to
s
u
p
p
o
r
t
d
iag
n
o
s
is
r
es
u
lt
s
u
c
h
as
u
ltra
s
o
n
o
g
r
ap
h
y
,
r
ad
io
g
r
ap
h
y
,
an
d
t
o
m
o
g
r
ap
h
y
.
I
m
a
g
e
p
r
o
ce
s
s
in
g
a
n
d
m
ac
h
in
e
lear
n
in
g
al
g
o
r
ith
m
s
al
s
o
p
la
y
ed
i
m
p
o
r
tan
t
r
o
le
to
ass
is
t
d
o
cto
r
in
m
a
k
i
n
g
f
as
ter
an
d
m
o
r
e
a
cc
u
r
ate
d
iag
n
o
s
i
s
[
8
]
.
So
m
e
r
esear
ch
er
s
h
a
v
e
co
m
b
in
ed
t
h
ese
ap
p
r
o
ac
h
es
o
n
m
ed
ical
i
m
ag
i
n
g
r
esu
lt
i.e
.
[
9
]
u
s
ed
p
o
s
s
ib
ili
s
ti
c
c
-
m
ea
n
s
f
o
r
u
ltra
s
o
u
n
d
i
m
a
g
es,
[
1
0
]
u
s
ed
w
av
ele
t
d
ec
o
m
p
o
s
itio
n
f
o
r
ch
e
s
t
r
ad
io
g
r
ap
h
,
an
d
[
1
1
]
u
s
ed
d
ee
p
n
eu
r
al
n
et
w
o
r
k
f
o
r
MRI
i
m
a
g
es se
g
m
e
n
tatio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
C
la
s
s
i
fica
tio
n
o
f p
n
eu
m
o
n
ia
fr
o
m
X
-
r
a
y
ima
g
es u
s
in
g
s
ia
mese
co
n
vo
lu
tio
n
a
l
…
(
K
en
n
a
r
d
A
lca
n
d
er P
r
a
yo
g
o
)
1303
As
t
h
er
e
ar
e
n
o
test
th
at
ca
n
f
u
ll
y
id
en
tify
s
p
ec
if
ic
ca
u
s
e
o
f
p
n
eu
m
o
n
ia,
a
test
to
d
is
ti
n
g
u
is
h
b
et
w
ee
n
b
ac
ter
ial
an
d
v
ir
al
p
n
eu
m
o
n
ia
w
o
u
ld
b
e
a
m
aj
o
r
ad
v
an
ce
[
1
2
]
.
T
h
e
m
o
s
t
co
m
m
o
n
test
i
s
u
s
i
n
g
h
o
s
t
b
io
m
ar
k
er
s
(
i.e
.
p
r
o
ca
lcito
n
in
(
P
C
T
)
,
C
r
ea
ctiv
e
p
r
o
tein
(
C
R
P
)
,
w
h
ite
b
lo
o
d
ce
ll
in
d
icato
r
s
)
,
b
u
t
th
ese
test
s
r
esu
lti
n
g
in
co
m
p
le
x
o
u
tco
m
e
an
d
is
lack
o
f
an
ac
c
u
r
ate
r
ef
er
en
ce
co
m
p
ar
ato
r
test
[
1
2
]
.
A
cc
o
r
d
in
g
l
y
,
c
h
es
t
x
-
r
a
y
(
C
XR
)
,
a
m
ed
ical
i
m
a
g
i
n
g
tech
n
iq
u
e,
is
u
s
ed
to
p
r
o
v
i
d
e
m
o
r
e
co
m
p
r
eh
e
n
s
i
v
e
cli
n
ic
al
s
i
g
n
s
.
I
n
cli
n
ical
d
ec
is
io
n
,
C
X
R
s
u
p
p
o
r
ts
b
ac
ter
ial
o
r
v
ir
al
etio
lo
g
y
a
n
d
ev
e
n
f
o
r
co
m
p
licatio
n
s
[
7
]
.
A
s
a
r
esu
lt,
i
n
th
i
s
w
o
r
k
,
d
ee
p
lear
n
in
g
m
et
h
o
d
w
i
ll
b
e
u
s
ed
to
cla
s
s
i
f
y
in
g
t
h
e
r
esu
lt
o
f
C
X
R
i
n
to
n
o
r
m
al
co
n
d
itio
n
,
b
ac
ter
ial
p
n
eu
m
o
n
ia,
a
n
d
v
ir
al
p
n
e
u
m
o
n
ia.
D
e
e
p
l
e
a
r
n
i
n
g
m
e
t
h
o
d
t
h
a
t
m
a
i
n
l
y
u
s
e
d
f
o
r
i
m
a
g
e
p
r
o
b
l
e
m
i
s
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
tw
o
r
k
(
C
N
N
)
.
S
e
v
e
r
a
l
r
e
s
e
a
r
c
h
e
r
s
h
a
v
e
u
s
e
d
C
N
N
t
o
c
l
a
s
s
i
f
y
m
e
d
i
c
a
l
i
m
a
g
i
n
g
r
e
s
u
l
t
s
s
u
c
h
a
s
[
1
3
]
f
o
r
c
l
a
s
s
i
f
y
i
n
g
s
t
r
o
k
e
,
[
1
4
]
f
o
r
class
i
f
y
in
g
t
y
p
e
o
f
m
u
s
cle,
a
n
d
[
1
5
]
f
o
r
class
i
f
y
in
g
ab
d
o
m
i
n
al
u
ltra
s
o
u
n
d
i
m
a
g
es.
T
h
e
u
s
ed
o
f
C
NN
f
o
r
class
i
f
y
in
g
C
X
R
h
a
s
also
b
ee
n
p
er
f
o
r
m
ed
in
[
1
6
]
an
d
[
1
7
]
f
o
r
class
i
f
y
in
g
t
w
o
class
e
s
,
n
o
r
m
al
co
n
d
itio
n
an
d
p
n
eu
m
o
n
ia.
T
h
e
r
esu
lt
s
h
o
w
s
h
ig
h
ac
cu
r
ac
y
i
n
b
o
th
w
o
r
k
s
.
Ho
w
ev
er
,
th
e
p
r
ed
icted
r
esu
lt
s
till
lack
s
o
n
r
ea
s
o
n
in
g
p
ar
t th
at
s
u
p
p
o
r
ts
th
e
r
esu
lts
.
An
o
th
er
v
ar
iatio
n
o
f
C
NN
is
s
ia
m
e
s
e
co
n
v
o
lu
tio
n
al
n
et
w
o
r
k
(
S
C
N)
.
SC
N
u
s
i
n
g
t
w
o
id
en
tical
C
N
N
w
it
h
t
h
e
s
a
m
e
ar
c
h
itect
u
r
e
an
d
th
e
s
a
m
e
w
ei
g
h
t
[
1
8
]
.
T
h
is
n
et
w
o
r
k
r
ec
ei
v
es
a
p
air
o
f
i
m
ag
e
s
a
s
in
p
u
t
an
d
r
etu
r
n
s
a
s
i
m
ilar
it
y
s
co
r
e
b
et
w
ee
n
t
h
e
m
.
S
C
N
h
as
b
ee
n
u
s
ed
b
y
[
1
9
]
o
n
tr
ac
k
in
g
p
r
o
b
le
m
s
an
d
is
ab
le
to
p
er
f
o
r
m
w
ell
i
n
t
h
o
s
e
p
r
o
b
lem
s
.
A
n
o
t
h
er
w
o
r
k
is
d
o
n
e
b
y
[
2
0
]
f
o
r
o
n
e
s
h
o
t
c
h
ar
ac
ter
r
ec
o
g
n
itio
n
.
T
h
e
y
f
o
r
m
a
p
air
f
r
o
m
a
s
in
g
le
d
ata
i
m
a
g
e
to
s
u
it
t
h
e
S
C
N
i
n
p
u
t.
Fo
r
t
h
e
o
n
e
-
s
h
o
t
lear
n
i
n
g
p
r
o
b
lem
,
th
e
y
u
s
ed
d
i
f
f
er
e
n
t
class
co
llectio
n
s
o
f
alp
h
ab
ets
f
o
r
test
in
g
an
d
class
i
f
ied
ea
ch
o
f
th
e
alp
h
ab
ets
u
s
i
n
g
SC
N.
W
h
er
ea
s
o
u
r
w
o
r
k
w
il
l
u
s
e
s
i
m
ilar
t
y
p
e
o
f
ap
p
r
o
ac
h
b
u
t
w
it
h
th
e
s
a
m
e
c
lass
o
f
d
ata.
Usi
n
g
t
h
i
s
ap
p
r
o
ac
h
,
w
e
co
u
ld
also
u
s
e
th
e
co
m
p
ar
ed
i
m
ag
e
s
to
s
u
p
p
o
r
t
class
if
icat
io
n
r
esu
lts
,
co
n
s
i
d
er
in
g
th
e
f
ac
t
th
a
t
s
i
m
ilar
m
e
th
o
d
s
s
till
d
ef
icien
t
in
th
e
r
ea
s
o
n
i
n
g
p
ar
t.
2.
RE
S
E
ARCH
M
E
T
H
O
D
2
.
1
.
Da
t
a
des
cr
iptio
n
T
h
e
u
s
ed
d
ataset
is
L
ab
eled
Op
tical
C
o
h
er
en
ce
T
o
m
o
g
r
ap
h
y
(
OC
T
)
an
d
C
h
est
X
-
R
a
y
I
m
ag
e
s
f
o
r
C
las
s
i
f
icatio
n
[
1
6
]
.
T
h
is
d
atas
et
co
n
tain
s
5
8
6
3
ch
est
x
-
r
a
y
(
C
XR
)
i
m
a
g
es
f
r
o
m
p
atien
t
s
o
n
e
to
f
i
v
e
y
ea
r
s
o
f
ag
e
at
Gu
a
n
g
z
h
o
u
W
o
m
e
n
an
d
C
h
ild
r
en
’
s
Me
d
ical
C
e
n
ter
[
1
6
]
.
I
t
also
d
iv
id
es
th
e
d
ata
in
to
3
class
es,
n
o
r
m
a
l
co
n
d
itio
n
,
b
ac
ter
ial
p
n
e
u
m
o
n
i
a,
an
d
v
ir
al
p
n
e
u
m
o
n
ia.
D
iag
n
o
s
es
o
f
all
C
X
R
i
m
ag
e
s
h
av
e
also
b
ee
n
r
ev
ie
w
ed
b
y
t
w
o
r
ad
io
lo
g
is
t
s
[
1
6
]
.
2
.
2
.
P
ro
po
s
ed
m
o
del
2
.
2
.
1
.
Sia
m
ese
co
nv
o
lutio
na
l
net
w
o
rk
a
rc
hite
ct
ure
Fig
u
r
e
1
d
escr
ib
ed
s
iam
e
s
e
co
n
v
o
l
u
tio
n
al
n
et
w
o
r
k
(
S
C
N)
ar
ch
itect
u
r
e
th
at
w
ill
b
e
u
s
ed
in
th
i
s
w
o
r
k
.
T
h
is
n
et
w
o
r
k
r
ec
eiv
ed
a
p
air
o
f
2
2
4
x
2
2
4
ch
est
x
-
r
a
y
i
m
a
g
es
as
i
n
p
u
t.
T
h
en
,
in
p
u
t
w
i
ll
b
e
f
ed
in
to
co
n
v
o
lu
tio
n
al
n
et
w
o
r
k
to
ex
t
r
ac
t
f
ea
tu
r
es
f
r
o
m
ea
ch
i
m
a
g
e.
C
o
n
n
ec
tio
n
f
u
n
ctio
n
w
a
s
u
s
ed
f
o
r
co
m
b
in
i
n
g
ea
ch
o
u
tp
u
t
o
f
co
n
v
o
lu
tio
n
a
l
n
et
w
o
r
k
.
W
e
u
s
ed
co
s
i
n
e
d
is
tan
ce
as
a
co
n
n
ec
tio
n
f
u
n
ctio
n
to
e
v
alu
ate
s
i
m
ilar
it
y
b
et
w
ee
n
ea
ch
in
p
u
t.
C
o
s
i
n
e
d
is
ta
n
ce
h
as
b
ee
n
u
s
ed
f
o
r
p
atter
n
r
ec
o
g
n
i
ti
o
n
p
r
o
b
lem
s
a
s
th
is
f
u
n
ctio
n
is
i
n
v
ar
ia
n
t to
t
h
e
m
a
g
n
i
tu
d
es o
f
s
a
m
p
les [
2
0
]
.
T
h
is
f
u
n
ctio
n
f
o
r
m
u
la
i
s
[
2
1
]
.
Fig
u
r
e
1
.
Ov
er
all
s
ia
m
ese
co
n
v
o
lu
tio
n
al
n
e
t
w
o
r
k
ar
ch
itec
tu
r
e
=
.
‖
‖
‖
‖
=
∑
=
1
√
∑
2
√
∑
2
=
1
=
1
(
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
18
,
No
.
3
,
J
u
n
e
2
0
2
0
:
1
3
0
2
–
13
09
1304
w
h
e
r
e
A
a
n
d
B
a
r
e
t
h
e
o
u
t
p
u
t
o
f
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
a
s
a
v
e
c
t
o
r
w
i
t
h
c
o
r
r
e
s
p
o
n
d
i
n
g
t
o
=
1
,
2
,
…
,
a
n
d
=
1
,
2
,
…
,
n
w
i
l
l
b
e
t
h
e
n
u
m
b
e
r
o
f
v
e
c
t
o
r
a
t
t
r
i
b
u
t
e
s
.
T
h
e
o
u
t
p
u
t
o
f
t
h
i
s
c
o
n
n
e
c
t
e
d
f
u
n
c
t
i
o
n
w
i
l
l
b
e
f
o
r
w
a
r
d
e
d
i
n
t
o
f
u
l
l
y
c
o
n
n
e
c
t
e
d
l
a
y
e
r
w
i
t
h
2
h
i
d
d
e
n
l
a
y
e
r
s
.
D
r
o
p
o
u
t
l
a
y
e
r
w
a
s
a
d
d
e
d
b
e
tw
e
e
n
e
a
c
h
h
i
d
d
e
n
l
a
y
e
r
a
s
a
r
e
g
u
l
a
r
i
z
a
t
i
o
n
.
I
n
t
h
e
l
a
s
t
l
a
y
e
r
,
w
e
u
s
e
d
s
i
g
m
o
i
d
a
c
t
i
v
a
t
i
o
n
t
o
p
r
o
d
u
c
e
s
i
m
i
l
a
r
i
t
y
r
e
s
u
l
t
t
h
a
t
i
s
b
o
u
n
d
t
o
[
0
,
1
]
.
Mo
s
t
o
f
th
e
SC
N
m
o
d
el
u
s
ed
p
ar
am
eter
s
h
ar
in
g
o
n
its
t
w
o
co
n
v
o
lu
tio
n
al
n
et
w
o
r
k
s
w
ei
g
h
t
an
d
b
ias.
Ac
co
r
d
in
g
to
[
2
2
]
,
r
em
o
v
in
g
th
is
co
n
s
tr
ai
n
t
ca
n
d
ea
l
w
it
h
m
o
r
e
s
p
ec
if
ic
m
a
tch
i
n
g
tas
k
th
a
n
g
e
n
er
al
tas
k
.
As
o
u
r
p
r
o
b
le
m
u
s
ed
t
h
e
s
a
m
e
cla
s
s
f
o
r
test
in
g
,
w
e
a
ls
o
d
id
n
o
t
u
s
e
p
a
r
a
m
e
t
e
r
s
h
a
r
i
n
g
i
n
t
h
e
n
e
t
w
o
r
k
.
F
u
r
t
h
e
r
m
o
r
e
,
t
o
e
n
h
a
n
c
i
n
g
m
a
t
c
h
i
n
g
t
a
s
k
f
o
r
e
v
e
r
y
c
l
a
s
s
,
w
e
u
s
e
d
s
e
p
a
r
a
t
e
m
o
d
e
l
f
o
r
e
a
c
h
c
l
a
s
s
.
U
s
i
n
g
t
h
i
s
a
p
p
r
o
a
c
h
,
e
a
c
h
m
o
d
e
l
l
e
a
r
n
e
d
t
o
g
a
i
n
b
e
s
t
f
e
a
t
u
r
e
s
f
o
r
c
o
m
p
a
r
i
s
o
n
b
e
t
w
e
e
n
c
o
r
r
e
s
p
o
n
d
i
n
g
c
l
a
s
s
w
i
t
h
o
t
h
e
r
c
l
a
s
s
e
s
.
2
.
2
.
2
.
Co
nv
o
lutio
na
l
net
w
o
r
k
a
rc
hite
ct
ure
I
n
co
n
v
o
lu
t
io
n
al
n
et
w
o
r
k
,
w
e
u
s
ed
1
4
la
y
er
s
t
h
at
co
n
tai
n
9
co
n
v
o
lu
tio
n
al
la
y
er
s
a
n
d
5
m
a
x
-
p
o
o
lin
g
la
y
er
s
a
s
d
escr
ib
ed
in
F
i
g
u
r
e
2
.
T
h
e
n
u
m
b
er
o
f
ch
a
n
n
els
f
r
o
m
ea
c
h
la
y
er
ar
e
3
2
,
6
4
,
6
4
,
6
4
,
6
4
,
1
2
8
,
1
2
8
,
1
2
8
,
1
2
8
,
1
2
8
,
2
5
6
,
2
5
6
,
2
5
6
,
an
d
5
1
2
.
E
ac
h
co
n
v
o
lu
t
io
n
al
la
y
er
u
s
ed
a
3
x
3
f
ilter
w
it
h
t
w
o
s
tr
id
es
a
n
d
ze
r
o
p
ad
d
in
g
s
.
O
n
ea
ch
co
n
v
o
lu
tio
n
al
la
y
er
,
R
e
L
u
ac
ti
v
atio
n
s
w
ill
b
e
u
s
ed
to
r
e
m
o
v
e
n
e
g
ati
v
e
v
al
u
e.
T
h
e
m
a
x
-
p
o
o
lin
g
f
ilter
s
ize
i
s
2
x
2
w
it
h
t
w
o
s
tr
id
es
an
d
ze
r
o
p
a
d
d
in
g
s
.
T
h
is
w
a
y
,
th
e
i
m
ag
e
w
i
ll
b
e
d
o
w
n
s
a
m
p
led
a
n
d
f
o
cu
s
i
n
g
o
n
i
m
p
o
r
ta
n
t
f
ea
t
u
r
es.
T
o
r
ed
u
ce
o
v
er
f
it
tin
g
,
r
e
g
u
lar
izatio
n
tech
n
iq
u
es
s
u
c
h
a
s
d
r
o
p
o
u
t
[
2
3
]
an
d
b
atch
n
o
r
m
al
izatio
n
w
ill
b
e
u
s
ed
[
2
4
]
.
T
h
e
o
u
tp
u
t
o
f
t
h
is
n
et
w
o
r
k
h
as
a
s
i
ze
o
f
1
x
1
x
5
1
2
an
d
w
il
l b
e
f
latte
n
ed
in
to
v
ec
to
r
1
x
5
1
2
b
ef
o
r
e
f
o
r
w
ar
d
ed
to
f
u
ll
y
co
n
n
ec
ted
n
et
w
o
r
k
.
Fig
u
r
e
2
.
Selecte
d
co
n
v
o
lu
tio
n
al
n
et
w
o
r
k
f
o
r
SC
N
ar
ch
i
tectu
r
e
2
.
3
.
Sp
litt
ing
da
t
a
s
et
I
n
th
is
w
o
r
k
,
w
e
d
iv
id
ed
t
h
e
d
ataset
in
to
tr
ain
i
n
g
(
7
0
%),
v
alid
atio
n
(
1
0
%),
an
d
test
i
n
g
(
2
0
%)
s
ets.
th
is
tr
ai
n
i
n
g
a
n
d
v
alid
atio
n
d
a
ta
w
ill
b
e
u
s
ed
f
o
r
p
air
cr
ea
tio
n
.
As
a
r
esu
lt,
w
e
u
s
ed
d
if
f
er
e
n
t
a
m
o
u
n
t
s
o
f
d
ata
f
o
r
tr
ain
i
n
g
a
n
d
v
alid
ati
n
g
th
e
n
et
w
o
r
k
.
2
.
4
.
P
a
ir
cr
ea
t
io
n
As
SC
N
i
n
p
u
t
co
m
p
r
is
es
a
p
air
o
f
i
m
ag
e
s
,
a
p
r
ep
r
o
ce
s
s
in
g
m
et
h
o
d
to
f
o
r
m
a
p
air
f
r
o
m
t
h
e
d
ataset
m
u
s
t
b
e
p
er
f
o
r
m
ed
.
Fi
g
u
r
e
3
s
h
o
w
s
s
et
o
f
p
air
s
th
a
t
w
il
l
b
e
f
ed
i
n
to
t
h
e
n
et
w
o
r
k
.
Fo
r
s
a
m
e
clas
s
p
air
o
r
g
en
u
i
n
e
p
air
,
w
e
p
air
ed
d
ata
f
r
o
m
t
h
e
s
a
m
e
clas
s
.
W
h
ile
o
n
th
e
d
if
f
er
e
n
t
class
p
air
o
r
im
p
o
s
to
r
p
ai
r
,
th
e
f
ir
s
t
in
p
u
t
al
w
a
y
s
co
m
es
f
r
o
m
t
h
e
s
a
m
e
clas
s
th
at
co
r
r
esp
o
n
d
s
to
th
e
m
o
d
el.
T
h
e
s
ec
o
n
d
in
p
u
t
co
n
s
i
s
ts
o
f
d
ata
f
r
o
m
d
if
f
er
en
t
cla
s
s
es
t
h
an
t
h
e
f
ir
s
t
i
n
p
u
t.
I
n
o
u
r
w
o
r
k
,
w
e
u
s
ed
an
eq
u
al
n
u
m
b
er
o
f
g
en
u
in
e
p
air
s
an
d
i
m
p
o
s
to
r
p
air
s
to
av
o
id
clas
s
i
m
b
ala
n
ce
.
T
h
e
a
m
o
u
n
t
o
f
cr
ea
ted
g
en
u
in
e
p
air
s
ca
n
b
e
ca
lcu
lated
u
s
in
g
co
m
b
i
n
ato
r
ics.
Fo
r
n
n
u
m
b
er
o
f
d
ata
f
r
o
m
t
h
e
s
a
m
e
cla
s
s
,
t
h
er
e
ar
e
(
2
)
a
m
o
u
n
t
o
f
p
air
d
ata
av
ailab
le
to
u
s
e.
Hen
ce
,
th
e
s
p
lit
d
ata
tr
ain
i
n
g
an
d
v
alid
atio
n
w
ill
b
e
p
r
o
ce
s
s
ed
in
to
a
f
o
r
m
o
f
d
ata
p
air
.
Us
in
g
t
h
ese
p
air
d
ata,
w
e
r
an
d
o
m
l
y
p
i
c
k
ed
1
0
0
0
0
p
a
ir
o
f
tr
ain
i
n
g
d
ata
an
d
2
0
0
0
p
a
ir
o
f
v
alid
atio
n
d
ata
f
o
r
tr
ain
in
g
p
h
ase.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
C
la
s
s
i
fica
tio
n
o
f p
n
eu
m
o
n
ia
fr
o
m
X
-
r
a
y
ima
g
es u
s
in
g
s
ia
mese
co
n
vo
lu
tio
n
a
l
…
(
K
en
n
a
r
d
A
lca
n
d
er P
r
a
yo
g
o
)
1305
Fig
u
r
e
3
.
Set o
f
p
air
s
f
o
r
ea
ch
class
m
o
d
el
2
.
5
.
T
ra
ini
ng
W
e
tr
ain
ed
o
u
r
SC
N
u
s
i
n
g
m
in
i
b
atch
w
it
h
b
atch
s
ize
1
6
.
T
o
g
en
er
ate
ea
ch
b
atch
,
th
e
p
air
ed
d
ata
w
er
e
p
u
t
i
n
to
g
en
er
ato
r
.
I
n
ad
d
itio
n
,
g
e
n
er
ato
r
also
p
er
f
o
r
m
ed
ad
d
itio
n
al
p
r
ep
r
o
ce
s
s
in
g
to
t
h
e
i
m
ag
e
s
.
First,
i
m
a
g
e
s
w
er
e
r
esized
i
n
to
2
2
4
x
2
2
4
.
T
h
en
,
f
lip
h
o
r
i
zo
n
tal
au
g
m
en
ta
tio
n
w
as
p
er
f
o
r
m
ed
r
an
d
o
m
l
y
to
th
e
i
m
a
g
e
s
.
W
e
also
n
o
r
m
al
i
ze
d
th
e
i
m
a
g
es
b
ased
o
n
th
e
m
ea
n
a
n
d
s
tan
d
ar
d
d
ev
iatio
n
f
r
o
m
1
0
0
0
s
am
p
le
s
tr
ain
i
n
g
d
ata.
Fi
g
u
r
e
4
s
h
o
w
s
p
r
ep
r
o
ce
s
s
in
g
f
lo
w
i
n
t
h
e
g
e
n
er
ato
r
.
T
o
tr
ain
th
e
n
et
w
o
r
k
,
A
d
a
m
o
p
ti
m
izer
w
as
u
s
ed
to
o
p
ti
m
ize
t
h
e
n
et
w
o
r
k
.
T
h
e
ch
o
s
en
h
y
p
er
p
ar
a
m
eter
s
a
r
e
s
h
o
w
n
i
n
T
ab
le
1
.
Fig
u
r
e
4
.
P
r
ep
r
o
ce
s
s
in
g
i
n
t
h
e
g
en
er
ato
r
T
ab
le
1
.
SC
N
H
y
p
er
p
ar
am
eter
Settin
g
s
H
y
p
e
r
p
a
r
a
me
t
e
r
V
a
l
u
e
L
e
a
r
n
i
n
g
R
a
t
e
0
.
0
0
1
D
r
o
p
o
u
t
R
a
t
e
0
.
5
L
e
a
r
n
i
n
g
D
e
c
a
y
0
β
1
0
.
9
β
2
0
.
9
9
9
A
s
t
h
e
o
u
t
p
u
t
o
f
t
h
i
s
n
e
tw
o
r
k
i
s
a
b
i
n
a
r
y
c
l
a
s
s
i
f
i
c
a
t
i
o
n
,
w
h
i
c
h
s
i
m
i
l
a
r
i
t
y
r
e
s
u
l
t
w
a
s
m
e
a
s
u
r
e
d
,
w
e
u
s
e
d
b
i
n
a
r
y
c
r
o
s
s
e
n
t
r
o
p
y
(
B
C
E
)
f
o
r
t
h
e
l
o
s
s
f
u
n
c
t
i
o
n
.
B
C
E
i
s
u
s
u
a
l
l
y
u
s
e
d
b
y
S
C
N
w
i
t
h
s
i
g
m
o
i
d
a
c
t
i
v
a
t
i
o
n
s
o
n
t
h
e
l
a
s
t
l
a
y
e
r
[
25
]
.
A
s
s
h
o
w
n
i
n
(
2
)
s
h
o
w
s
t
h
e
l
o
s
s
f
u
n
c
t
i
o
n
f
o
r
m
u
l
a
w
h
i
c
h
g
u
i
d
e
o
u
r
S
C
N
l
e
a
r
n
i
n
g
p
r
o
c
e
s
s
[
2
6
],
=
1
∑
[
+
(
1
−
)
(
1
−
)
]
=
1
(
2
)
w
h
er
e
C
is
t
h
e
B
C
E
lo
s
s
f
u
n
c
t
io
n
v
al
u
e
f
r
o
m
d
ata
k
to
m
,
t
k
an
d
y
k
i
s
an
ac
tu
a
l
tar
g
et
an
d
p
r
ed
ictio
n
r
esu
lt
o
f
d
ata
k
.
Net
w
o
r
k
f
r
o
m
ea
c
h
cla
s
s
w
as
tr
ain
ed
f
o
r
2
0
ep
o
ch
s
,
as
t
h
e
l
o
s
s
h
as
b
ee
n
i
n
cr
ea
s
i
n
g
ar
o
u
n
d
th
e
la
s
t
5
ep
o
ch
s
.
T
h
e
b
est
w
ei
g
h
ts
co
n
f
i
g
u
r
atio
n
w
a
s
s
a
v
ed
as
s
o
ciat
es
to
t
h
e
lo
w
est
v
al
id
atio
n
lo
s
s
.
S
u
ch
w
eig
h
t
w
ill
b
e
lo
ad
ed
to
th
e
n
et
w
o
r
k
b
ef
o
r
e
u
s
i
n
g
i
t f
o
r
test
i
n
g
d
a
ta.
2
.
6
.
T
esting
I
n
itiall
y
,
w
e
ch
a
n
g
ed
t
h
e
te
s
ti
n
g
d
ata
i
n
to
a
p
air
.
Gen
er
ato
r
w
a
s
u
s
ed
to
cr
ea
te
a
p
air
b
et
wee
n
te
s
ti
n
g
d
ata
w
it
h
r
an
d
o
m
l
y
p
ick
ed
d
at
a
tr
ain
i
n
g
f
o
r
ea
ch
r
ep
r
ese
n
tat
iv
e
clas
s
.
Fi
g
u
r
e
5
d
escr
ib
es
th
e
i
n
p
u
t
cla
s
s
f
r
o
m
ea
ch
p
air
.
T
h
e
f
ir
s
t
in
p
u
t
al
w
a
y
s
co
n
s
is
t
s
o
f
th
e
d
ata
co
r
r
esp
o
n
d
in
g
to
th
e
class
m
o
d
el.
Fo
r
th
e
s
ec
o
n
d
in
p
u
t
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
18
,
No
.
3
,
J
u
n
e
2
0
2
0
:
1
3
0
2
–
13
09
1306
w
e
p
u
t
t
h
e
te
s
tin
g
d
ata,
w
h
ic
h
ca
n
b
e
th
e
d
ata
f
r
o
m
a
n
y
class
.
Usi
n
g
th
e
s
e
r
u
le
s
,
th
er
e
w
as
n
o
alter
atio
n
b
et
w
ee
n
th
e
p
air
ed
d
ata
in
p
u
t
u
s
ed
f
o
r
tr
ain
i
n
g
a
n
d
test
i
n
g
.
T
h
u
s
,
th
e
m
o
d
el
w
a
s
ab
le
to
f
in
d
s
i
m
ilar
it
y
f
ea
t
u
r
es
f
o
r
ea
ch
clas
s
.
W
e
d
ef
i
n
ed
r
an
k
a
s
t
h
e
n
u
m
b
er
o
f
r
ep
ea
ted
test
in
g
o
n
t
h
e
s
a
m
e
ca
teg
o
r
y
.
T
h
is
ter
m
w
a
s
also
u
s
ed
in
[
2
1
]
to
ev
alu
ate
th
e
n
et
w
o
r
k
.
T
h
en
,
class
i
f
icatio
n
r
esu
lt
w
a
s
b
ased
o
n
m
ax
i
m
u
m
s
i
m
ilar
it
y
f
r
o
m
C
n
u
m
b
er
o
f
clas
s
es a
s
s
h
o
w
n
i
n
3
[
2
0
]
.
=
(
)
(
3
)
w
h
er
e
C
*
d
ef
i
n
es c
la
s
s
i
f
icatio
n
r
esu
lt f
r
o
m
al
l c
lass
,
c
a
s
lis
t
o
f
all
class
,
a
n
d
(
)
as
m
ea
n
p
r
ed
ictio
n
r
esu
lt
b
ased
o
n
n
u
m
b
er
o
f
r
an
k
s
.
Fig
u
r
e
5
.
Ov
er
all
test
i
n
g
f
lo
w
3.
RE
SU
L
T
S
A
ND
AN
AL
Y
SI
S
3
.
1
.
Co
m
pa
ring
a
rc
hite
ct
ure
First,
w
e
w
a
n
ted
to
co
m
p
ar
e
o
u
r
ar
ch
itectu
r
e
w
it
h
th
e
o
th
er
o
n
es
th
at
h
a
v
e
b
ee
n
u
s
ed
f
o
r
p
n
eu
m
o
n
ia
class
i
f
icatio
n
.
T
h
e
co
m
p
ar
ed
n
et
w
o
r
k
i
s
I
n
ce
p
tio
n
V3
n
ets
w
h
ic
h
ar
e
u
s
ed
b
y
[
1
6
]
an
d
n
et
w
o
r
k
f
r
o
m
[
1
7
]
.
As
b
o
th
ar
c
h
itect
u
r
es
ar
e
C
NN,
w
e
d
r
o
p
th
e
f
u
ll
y
co
n
n
ec
ted
la
y
er
o
n
th
o
s
e
n
e
t
w
o
r
k
s
a
n
d
u
s
e
th
e
m
a
s
a
co
n
v
o
lu
tio
n
al
n
et
w
o
r
k
f
o
r
o
u
r
SC
N.
T
o
k
n
o
w
a
g
li
m
p
s
e
o
f
ea
ch
n
et
w
o
r
k
p
er
f
o
r
m
a
n
ce
,
w
e
u
s
ed
2
0
0
0
tr
ain
i
n
g
d
ata
p
air
an
d
4
0
0
v
alid
atio
n
d
ata
p
air
.
T
ab
le
2
s
h
o
w
s
th
e
ac
c
u
r
ac
y
f
o
r
ea
ch
ar
ch
ite
ctu
r
e.
T
ab
le
2
.
C
o
m
p
ar
is
o
n
o
f
Dif
f
er
en
t
A
r
ch
i
tectu
r
e
A
r
c
h
i
t
e
c
t
u
r
e
A
c
c
u
r
a
c
y
T
h
i
s Pa
p
e
r
7
7
.
0
5
%
(
S
a
r
a
i
v
a
e
t
a
l
.
)
[
1
7
]
7
3
.
4
6
%
I
n
c
e
p
t
i
o
n
V
3
[
2
7
]
7
2
.
1
8
%
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
i
n
t
h
is
p
ap
er
ac
h
iev
es
h
i
g
h
er
ac
cu
r
ac
y
t
h
an
th
e
o
t
h
er
n
e
t
w
o
r
k
w
h
e
n
u
s
ed
f
o
r
SC
N.
I
f
w
e
co
m
p
ar
e
th
e
n
u
m
b
er
o
f
la
y
er
s
o
n
t
h
o
s
e
n
et
w
o
r
k
s
,
Sar
aiv
a
m
o
d
el
co
n
s
is
t
o
f
9
lay
er
s
a
n
d
I
n
ce
p
tio
n
V3
h
a
s
4
8
n
u
m
b
er
o
f
la
y
er
s
,
w
h
er
ea
s
o
u
r
ar
ch
it
ec
tu
r
e
is
f
o
r
m
ed
b
y
1
4
la
y
e
r
s
.
T
h
er
ef
o
r
e,
o
u
r
ar
ch
itect
u
r
e
h
as d
ee
p
er
la
y
er
th
an
Sar
aiv
a
m
o
d
el,
b
u
t
n
o
t a
s
d
ee
p
as I
n
ce
p
tio
n
V3
n
ets.
3.
2
.
E
v
a
lua
t
io
n
T
ab
le
3
d
etails
th
e
tr
ain
in
g
r
esu
lt
s
th
at
co
n
s
is
t
o
f
lo
w
est
l
o
s
s
v
alid
atio
n
,
v
al
id
atio
n
ac
cu
r
ac
y
,
an
d
ep
o
ch
p
o
s
itio
n
f
o
r
ea
ch
class
.
B
ased
o
n
th
e
lo
s
s
v
alid
atio
n
,
ea
ch
m
o
d
el
p
er
f
o
r
m
s
q
u
ite
w
ell
to
lear
n
s
im
ilar
it
y
b
et
w
ee
n
ea
ch
class
.
T
h
u
s
,
th
es
e
m
o
d
els
ar
e
u
s
ed
to
p
r
ed
ict
te
s
tin
g
d
ata.
W
h
en
p
r
ed
ictin
g
d
ata,
o
n
e
b
en
ef
its
o
f
u
s
i
n
g
SC
N
f
o
r
class
i
f
icatio
n
is
th
e
av
ai
lab
ilit
y
o
f
o
th
er
im
ag
e
s
f
o
r
co
m
p
ar
is
o
n
.
T
h
e
am
o
u
n
t
o
f
co
m
p
ar
ed
i
m
a
g
es
i
s
eq
u
al
to
t
h
e
n
u
m
b
er
o
f
r
an
k
s
.
B
y
d
o
in
g
s
o
,
h
i
g
h
s
i
m
ilar
it
y
i
m
a
g
es
ca
n
b
e
u
s
ed
f
o
r
ca
s
e
r
ef
er
en
ce
s
.
P
ar
ticu
lar
l
y
i
n
m
ed
ical
i
m
a
g
i
n
g
d
ata,
t
h
e
r
o
le
o
f
th
e
s
e
i
m
a
g
es
i
s
ess
e
n
tial
to
s
u
p
p
o
r
t
th
e
class
i
f
icatio
n
r
esu
lt.
Fig
u
r
e
6
s
h
o
w
s
t
h
e
ill
u
s
tr
atio
n
o
f
co
m
p
ar
is
o
n
i
m
ag
e
s
w
it
h
it
s
co
r
r
esp
o
n
d
in
g
s
i
m
i
lar
it
y
v
al
u
e.
T
ab
le
3
.
T
r
ain
in
g
r
es
u
lt f
o
r
ea
ch
class
C
l
a
ss
L
o
w
e
st
L
o
ss V
a
l
i
d
a
t
i
o
n
V
a
l
i
d
a
t
i
o
n
A
c
c
u
r
a
c
y
Ep
o
c
h
P
o
si
t
i
o
n
N
o
r
mal
0
.
1
0
8
9
7
9
6
.
0
5
%
8
B
a
c
t
e
r
i
a
l
0
.
3
9
4
4
1
8
3
.
2
0
%
13
V
i
r
u
s
0
.
4
3
9
1
7
8
3
.
1
0
%
5
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
C
la
s
s
i
fica
tio
n
o
f p
n
eu
m
o
n
ia
fr
o
m
X
-
r
a
y
ima
g
es u
s
in
g
s
ia
mese
co
n
vo
lu
tio
n
a
l
…
(
K
en
n
a
r
d
A
lca
n
d
er P
r
a
yo
g
o
)
1307
T
est
Da
t
a
#
2
(
L
a
bel:
B
a
ct
er
i
a
,
P
re
dict
ed:
B
a
c
t
er
ia
)
Co
m
pa
re
w
it
h:
B
a
ct
er
ia
Fig
u
r
e
6
.
I
llu
s
tr
atio
n
o
f
co
m
p
a
r
is
o
n
v
i
s
u
a
lizatio
n
f
o
r
s
ia
m
ese
co
n
v
o
lu
t
io
n
al
n
et
w
o
r
k
As
t
h
e
p
r
ed
ictio
n
r
es
u
lt
an
d
ac
tu
al
tar
g
et
ar
e
al
s
o
b
ac
ter
ia
class
,
w
h
ic
h
is
a
co
r
r
ec
t
class
i
f
icatio
n
,
ea
ch
co
m
p
ar
is
o
n
m
o
s
tl
y
s
h
o
ws
h
i
g
h
s
i
m
i
lar
it
y
s
co
r
es.
T
ab
le
4
d
escr
ib
es
th
e
r
es
u
lt
o
f
tes
ti
n
g
d
ata
p
r
ed
ictio
n
u
s
i
n
g
d
if
f
er
en
t
n
u
m
b
er
o
f
r
an
k
s
.
T
h
e
h
i
g
h
er
n
u
m
b
er
o
f
r
an
k
s
,
in
ac
co
r
d
an
ce
w
i
th
h
ig
h
er
n
u
m
b
er
o
f
co
m
p
ar
ed
d
ata,
r
esu
ltin
g
in
h
ig
h
er
ac
cu
r
ac
y
.
T
h
e
b
est
ac
cu
r
ac
y
th
a
t
w
e
ac
h
iev
e
is
8
0
.
0
3
%
u
s
i
n
g
r
an
k
-
20.
T
h
u
s
,
w
e
ev
alu
a
te
th
i
s
m
o
d
el
u
s
i
n
g
co
n
f
u
s
io
n
m
a
tr
ix
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
f
1
s
co
r
e.
T
ab
le
5
s
h
o
w
s
t
h
e
co
n
f
u
s
io
n
m
atr
i
x
f
o
r
th
i
s
clas
s
i
f
icatio
n
r
e
s
u
lt.
T
ab
le
4
.
T
esti
n
g
r
es
u
lt
b
ased
o
n
n
u
m
b
er
o
f
r
an
k
R
a
n
k
A
c
c
u
r
a
c
y
1
7
4
.
9
1
%
5
7
7
.
0
5
%
10
7
8
.
3
3
%
20
8
0
.
0
3
%
T
ab
le
5
.
C
o
n
f
u
s
io
n
m
atr
i
x
u
s
i
n
g
r
a
n
k
-
2
0
class
i
f
icat
io
n
C
o
n
f
u
s
i
o
n
M
a
t
r
i
x
P
r
e
d
i
c
t
e
d
N
o
r
mal
B
a
c
t
e
r
i
a
V
i
r
u
s
T
o
t
a
l
A
c
t
u
a
l
N
o
r
mal
2
9
7
8
12
3
1
7
B
a
c
t
e
r
i
a
15
4
4
0
1
0
1
5
5
6
V
i
r
u
s
15
83
2
0
1
2
9
9
T
o
t
a
l
3
2
7
5
3
1
3
1
4
1
1
7
2
Usi
n
g
t
h
i
s
co
n
f
u
s
io
n
m
atr
i
x
,
w
e
ca
lc
u
late
o
t
h
er
ev
al
u
atio
n
m
etr
ics
s
u
c
h
a
s
p
r
ec
is
io
n
,
r
ec
all,
an
d
f
1
s
co
r
e.
T
h
e
co
m
p
lete
e
v
al
u
atio
n
m
etr
ics ar
e
d
e
s
cr
ib
ed
in
T
ab
le
6
.
T
o
m
ak
e
s
u
r
e
t
h
at
o
u
r
m
o
d
el
is
n
o
t b
ia
s
ed
to
s
p
ec
if
ic
cla
s
s
,
w
e
co
m
p
ar
ed
t
h
e
o
v
er
all
ac
c
u
r
ac
y
w
it
h
t
h
e
av
er
ag
e
f
1
s
co
r
e
,
as
th
i
s
v
al
u
e
tak
es
r
ec
all
a
n
d
p
r
ec
is
io
n
in
to
ac
co
u
n
t.
I
f
w
e
co
m
p
ar
ed
th
ese
v
alu
e
s
,
th
er
e
is
n
o
s
ig
n
i
f
ica
n
t
d
if
f
er
e
n
ce
.
Ho
w
e
v
er
,
if
w
e
co
m
p
ar
ed
f
1
s
co
r
e
b
et
w
ee
n
n
o
r
m
al
cla
s
s
a
n
d
v
ir
u
s
clas
s
,
th
e
d
i
f
f
er
en
ce
i
s
q
u
ite
b
ig
.
W
h
en
w
e
lo
o
k
at
th
e
co
n
f
u
s
io
n
m
at
r
i
x
,
m
o
s
t
o
f
th
e
w
r
o
n
g
class
if
icatio
n
s
o
n
v
ir
u
s
clas
s
,
ar
e
class
i
f
ied
as
b
ac
ter
ia.
I
n
tu
iti
v
el
y
,
th
er
e
ar
e
s
o
m
e
ca
s
es
o
n
v
ir
u
s
clas
s
t
h
at
ar
e
also
s
i
m
ilar
w
it
h
th
e
b
ac
ter
ia
ca
s
e
s
.
T
h
is
m
a
k
es
t
h
e
m
o
d
el
p
r
ed
ictio
n
w
as
m
i
x
ed
b
et
w
ee
n
th
o
s
e
t
w
o
cla
s
s
e
s
.
Nev
er
t
h
e
less
,
as
t
h
er
e
is
n
o
s
ig
n
i
f
ica
n
t
d
if
f
er
e
n
ce
b
et
w
ee
n
av
er
ag
e
f
1
s
co
r
e
w
i
th
o
v
er
all
ac
cu
r
ac
y
,
o
u
r
m
o
d
el
i
s
n
o
t
b
i
ased
to
s
p
ec
if
ic
clas
s
an
d
p
er
f
o
r
m
s
q
u
ite
w
e
ll
to
class
i
f
y
d
ata
i
n
g
e
n
er
al.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
18
,
No
.
3
,
J
u
n
e
2
0
2
0
:
1
3
0
2
–
13
09
1308
T
ab
le
6
.
C
o
m
p
lete
li
s
t o
f
ev
a
l
u
atio
n
m
etr
ics
M
e
t
r
i
c
s
N
o
r
mal
B
a
c
t
e
r
i
a
V
i
r
u
s
A
v
e
r
a
g
e
R
e
c
a
l
l
9
3
.
6
9
%
7
9
.
1
3
%
6
7
.
2
2
%
8
0
.
0
3
%
P
r
e
c
i
si
o
n
9
0
.
8
2
%
8
2
.
8
6
%
6
4
.
0
1
%
7
9
.
2
3
%
F
1
S
c
o
r
e
9
2
.
2
4
%
8
0
.
9
6
%
6
5
.
5
8
%
7
9
.
5
9
%
O
v
e
r
a
l
l
A
c
c
u
r
a
c
y
8
0
.
0
3
%
3.
3
.
F
ilte
r
v
is
ua
liza
t
io
n
Fig
u
r
e
7
s
h
o
w
s
f
ilter
v
i
s
u
al
izatio
n
f
r
o
m
t
h
e
f
ir
s
t
co
n
v
o
l
u
tio
n
al
la
y
er
.
I
n
th
e
f
ir
s
t
i
n
p
u
t,
t
h
e
ex
tr
ac
ted
f
ea
t
u
r
es in
t
h
i
s
la
y
er
ce
n
ter
in
g
o
n
th
e
r
ig
h
t a
n
d
lef
t
s
id
e
o
f
th
e
lu
n
g
s
.
Fo
r
th
e
b
ac
ter
ial
class
,
th
e
r
esu
l
t d
etailed
m
o
r
e
o
n
t
h
e
r
ig
h
t
s
id
e
o
f
th
e
l
u
n
g
.
On
t
h
e
o
p
p
o
s
ite,
th
e
f
ilte
r
r
esu
lt
o
n
s
ec
o
n
d
in
p
u
t
m
o
s
tl
y
e
x
tr
ac
ts
t
h
e
s
a
m
e
f
ea
t
u
r
es o
n
f
ir
s
t c
o
n
v
o
lu
t
io
n
al
f
o
r
ea
ch
class
.
Fig
u
r
e
7
.
Fil
ter
v
is
u
aliza
tio
n
f
r
o
m
t
h
e
f
ir
s
t c
o
n
v
o
lu
t
io
n
al
la
y
er
4.
CO
NCLU
SI
O
N
I
n
th
i
s
w
o
r
k
,
w
e
p
r
o
p
o
s
e
x
-
r
a
y
i
m
ag
in
g
cla
s
s
i
f
icatio
n
u
s
i
n
g
s
ia
m
e
s
e
co
n
v
o
lu
tio
n
al
n
et
w
o
r
k
(
SC
N)
w
h
ic
h
is
o
f
te
n
u
s
ed
f
o
r
s
i
m
il
ar
it
y
lear
n
i
n
g
.
Ou
r
m
o
d
el
als
o
d
r
o
p
s
th
e
co
n
s
tr
ain
t
o
f
p
ar
a
m
eter
s
h
ar
in
g
a
n
d
en
h
a
n
ci
n
g
f
ea
t
u
r
es
o
n
ea
ch
c
lass
u
s
i
n
g
m
o
d
el
s
ep
ar
atio
n
.
T
h
e
SC
N
ar
ch
itect
u
r
e
w
e
ar
e
u
s
ed
co
n
s
i
s
ts
o
f
2
co
n
v
o
lu
tio
n
al
n
e
t
w
o
r
k
s
w
i
t
h
1
4
la
y
er
s
(
9
co
n
v
o
l
u
tio
n
al
la
y
er
s
an
d
5
m
a
x
p
o
o
lin
g
l
a
y
er
s
)
r
esp
ec
tiv
el
y
,
co
s
in
e
d
i
s
tan
ce
as
co
n
n
ec
tio
n
f
u
n
ctio
n
,
a
n
d
f
u
l
l
y
co
n
n
ec
ted
la
y
er
w
i
th
2
h
id
d
en
la
y
er
s
.
W
h
en
u
s
ed
f
o
r
S
C
N,
o
u
r
ar
ch
itectu
r
e
ac
h
ie
v
e
s
b
etter
r
esu
lts
th
a
n
o
th
er
ar
ch
itect
u
r
e
s
u
c
h
as
in
ce
p
tio
n
v
3
an
d
m
o
d
el
f
r
o
m
[
1
7
]
.
T
h
is
ar
ch
itect
u
r
e
w
a
s
ab
le
to
ac
h
iev
e
8
0
.
0
3
%
ac
cu
r
ac
y
a
n
d
7
9
.
5
9
%
f
1
s
co
r
e.
W
e
also
s
h
o
w
t
h
e
co
m
p
ar
i
s
o
n
i
m
a
g
e
w
h
ic
h
u
s
ed
to
s
u
p
p
o
r
t
th
e
d
ec
is
io
n
f
r
o
m
cla
s
s
i
f
icatio
n
r
esu
lt.
T
h
er
ef
o
r
e,
o
u
r
m
o
d
el
h
as
a
b
etter
r
esu
lt
r
ea
s
o
n
in
g
an
d
d
etail
s
co
m
p
ar
in
g
to
t
h
e
s
i
m
ilar
m
eth
o
d
s
li
k
e
C
NN.
Fo
r
f
u
tu
r
e
w
o
r
k
s
,
as
th
e
a
m
o
u
n
t
o
f
p
air
r
es
u
lt
is
co
m
b
in
at
o
r
ics,
th
er
e
is
p
len
t
y
o
f
tr
ai
n
i
n
g
d
ata
th
a
t
ca
n
b
e
ad
d
e
d
to
im
p
r
o
v
e
th
e
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
Mo
r
e
ex
p
er
im
e
n
t
s
o
n
h
y
p
er
p
ar
a
m
eter
s
etti
n
g
s
also
h
elp
to
i
m
p
r
o
v
e
t
h
e
m
o
d
el.
Fi
n
all
y
,
w
e
p
lan
to
i
m
p
r
o
v
e
th
e
r
es
u
lt
v
i
s
u
al
izatio
n
u
s
in
g
c
lass
ac
ti
v
atio
n
m
ap
(
C
A
M)
th
at
i
s
s
u
itab
le
f
o
r
S
C
N.
I
t
w
o
u
ld
m
ax
i
m
ize
th
e
p
u
r
p
o
s
e
o
f
co
m
p
ar
is
o
n
i
m
a
g
es
an
d
g
i
v
e
a
b
ette
r
u
n
d
er
s
ta
n
d
in
g
o
f
t
h
e
r
es
u
lt.
ACK
NO
WL
E
D
G
E
M
E
NT
S
Th
i
s
r
e
s
e
a
r
c
h
w
a
s
s
u
p
p
o
r
t
e
d
b
y
t
h
e
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
L
a
b
o
r
a
t
o
r
y
i
n
U
n
i
v
e
r
s
i
t
a
s
M
u
l
t
i
m
e
d
i
a
N
u
s
a
n
t
a
r
a
.
RE
F
E
R
E
NC
E
S
[1
]
F
o
ru
m
o
f
In
tern
a
ti
o
n
a
l
Re
sp
irat
o
ry
S
o
c
ieties
,
“
T
h
e
G
lo
b
a
l
I
m
p
a
c
t
o
f
Re
sp
irato
ry
Dis
e
a
se
,”
S
e
c
o
n
d
E
d
it
i
o
n
.
S
h
e
ff
iel
d
:
E
u
ro
p
e
a
n
Res
p
ira
t
o
ry
S
o
c
iety
,
2
0
1
7
.
[2
]
Cil
lo
n
iz
C
.
,
M
a
rti
n
-
L
o
e
c
h
e
s
I
.
,
G
a
r
c
ia
-
V
id
a
l
C
.
,
S
a
n
Jo
se
A
.
,
T
o
rre
s
A
.
,
“
M
icro
b
ial
Et
io
l
o
g
y
o
f
P
n
e
u
m
o
n
ia:
Ep
id
e
m
io
lo
g
y
,
Dia
g
n
o
sis
a
n
d
Re
sista
n
c
e
P
a
tt
e
rn
s
,”
I
n
ter
n
a
ti
o
n
a
l
jo
u
rn
a
l
o
f
m
o
lec
u
la
r
sc
ien
c
e
s,
v
o
l.
1
7
,
n
o
.
1
2
,
De
c
e
m
b
e
r
2
0
1
6
.
[3
]
W
o
js
y
k
-
Ba
n
a
sz
a
k
I,
Br
b
o
ro
w
ic
z
A
.
P
n
e
u
m
o
n
ia
in
Ch
il
d
re
n
.
I
n
:
M
a
h
b
o
u
b
B
.
Ed
it
o
rs.
Re
sp
irato
ry
Dise
a
se
a
n
d
In
f
e
c
ti
o
n
.
Rij
e
k
a
:
In
tec
h
Op
e
n
;
2
0
1
3
:
1
3
7
-
1
7
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
C
la
s
s
i
fica
tio
n
o
f p
n
eu
m
o
n
ia
fr
o
m
X
-
r
a
y
ima
g
es u
s
in
g
s
ia
mese
co
n
vo
lu
tio
n
a
l
…
(
K
en
n
a
r
d
A
lca
n
d
er P
r
a
yo
g
o
)
1309
[4
]
T
o
n
g
N.
Up
d
a
te o
n
2
0
0
4
b
a
c
k
g
ro
u
n
d
p
a
p
e
r
6
.
2
2
p
n
e
u
m
o
n
ia
, “
P
ri
o
r
it
y
m
e
d
icin
e
s f
o
r
Eu
ro
p
e
a
n
d
t
h
e
w
o
rld
A
p
u
b
li
c
h
e
a
lt
h
a
p
p
ro
a
c
h
to
i
n
n
o
v
a
ti
o
n
,
"
Ge
n
e
v
a
,
W
HO
,
2
0
1
3
.
[5
]
M
a
th
u
r
S
.
,
F
u
c
h
s
A
.
,
Bielick
i
J
.
,
V
a
n
De
n
A
n
k
e
r
J
.
,
S
h
a
rlan
d
M
.
,
“
A
n
ti
b
i
o
ti
c
u
se
f
o
r
c
o
m
m
u
n
it
y
-
a
c
q
u
ired
p
n
e
u
m
o
n
ia
i
n
n
e
o
n
a
tes
a
n
d
c
h
il
d
r
e
n
:
W
HO
e
v
id
e
n
c
e
re
v
ie
w
,”
Pa
e
d
ia
trics
a
n
d
I
n
ter
n
a
ti
o
n
a
l
Ch
i
ld
He
a
lt
h
,
v
o
l.
3
8
,
p
p
.
S
6
6
-
S
7
5
,
No
v
e
m
b
e
r
2
0
1
8
.
[6
]
Kim
J.
E
.
,
Ki
m
U.
J
.
,
Ki
m
H.
K
.
,
Ch
o
S
.
K
.
,
A
n
J.
H
.
,
Ka
n
g
S
.
J
.
,
P
a
rk
K.
H
.
,
Ju
n
g
S
.
I
.
,
Ja
n
g
H.
C.
,
“
P
re
d
icto
rs
o
f
v
iral
p
n
e
u
m
o
n
ia i
n
p
a
ti
e
n
ts w
it
h
c
o
m
m
u
n
it
y
-
a
c
q
u
ired
p
n
e
u
m
o
n
ia
,”
Pl
o
S
o
n
e
,
v
o
l.
9
,
n
o
.
1
2
,
De
c
e
m
b
e
r
2
0
1
4
.
[7
]
M
a
c
k
e
n
z
ie,
G
.
,
“
T
h
e
d
e
f
in
it
io
n
a
n
d
c
las
sif
ica
ti
o
n
o
f
p
n
e
u
m
o
n
ia
,”
Pn
e
u
mo
n
ia
,
v
o
l.
8
,
n
o
.
1
2
,
De
c
e
m
b
e
r
2
0
1
6
.
[8
]
Ra
z
z
a
k
M
.
I
.
,
Na
z
S
.
,
Zai
b
A
.
,
“
De
e
p
lea
rn
in
g
f
o
r
m
e
d
ica
l
ima
g
e
p
ro
c
e
ss
in
g
:
Ov
e
rv
ie
w
,
c
h
a
ll
e
n
g
e
s
a
n
d
t
h
e
f
u
tu
re
,”
Cla
ss
if
ica
ti
o
n
in
Bi
o
A
p
p
s
,
v
o
l.
2
6
,
p
p
.
3
2
3
-
3
5
0
,
N
o
v
e
m
b
e
r
2
0
1
7
.
[9
]
S
u
ry
a
d
ib
ra
ta
A
.
,
Ki
m
K.
B.
,
“
Ga
n
g
li
o
n
c
y
st
re
g
io
n
e
x
trac
ti
o
n
f
ro
m
u
lt
ra
so
u
n
d
im
a
g
e
s
u
si
n
g
p
o
ss
ib
il
isti
c
c
-
m
e
a
n
s
c
lu
ste
rin
g
m
e
th
o
d
,”
Jo
u
rn
a
l
o
f
In
f
o
rm
a
ti
o
n
a
n
d
C
o
mm
u
n
ica
t
io
n
Co
n
v
e
rg
e
n
c
e
En
g
in
e
e
rin
g
,
v
o
l
.
15
,
pp.
49
-
52
,
2
0
1
7
.
[1
0
]
A
id
o
o
A
.
Y
.
,
W
il
so
n
M
.
,
Bo
tch
w
a
y
G
.
A
.
,
“
Ch
e
st
r
a
d
io
g
ra
p
h
im
a
g
e
e
n
h
a
n
c
e
m
e
n
t
w
it
h
w
a
v
e
let
d
e
c
o
m
p
o
siti
o
n
a
n
d
m
o
rp
h
o
lo
g
ica
l
o
p
e
ra
ti
o
n
s
,”
T
EL
KOM
NIKA
T
e
lec
o
mm
u
n
ica
ti
o
n
Co
mp
u
t
in
g
El
e
c
tro
n
ics
a
n
d
Co
n
t
ro
l
,
v
o
l
1
7
,
n
o
.
5
,
p
p
.
2
5
8
7
-
2
5
9
4
,
Oc
to
b
e
r
2
0
1
9
.
[1
1
]
Al
-
Ka
f
ri
A
.
S
.
,
S
u
d
irm
a
n
S
.
,
Hu
ss
a
in
A
.
J
.
,
A
l
-
Ju
m
e
il
y
,
D
.
,
Na
ta
li
a
F
.
,
M
e
id
ia
H
.
,
A
f
ril
ian
a
N
.
,
A
l
-
Ra
sh
d
a
n
W
.
,
Ba
sh
taw
i
M
.
,
A
l
-
Ju
m
a
il
y
M
.
,
“
Bo
u
n
d
a
ry
De
li
n
e
a
ti
o
n
o
f
M
RI
Im
a
g
e
s
f
o
r
L
u
m
b
a
r
S
p
in
a
l
S
ten
o
sis
De
tec
ti
o
n
T
h
ro
u
g
h
S
e
m
a
n
ti
c
S
e
g
m
e
n
tatio
n
Us
in
g
De
e
p
Ne
u
ra
l
Ne
tw
o
rk
s
,”
in
IEE
E
Acc
e
ss
,
v
o
l.
7
,
p
p
.
4
3
4
8
7
-
4
3
5
0
1
,
2
0
1
9
.
[1
2
]
Zar H.
J
.
,
S
a
v
v
a
s A
.
,
M
a
rk
P
.
N.
,
“
A
d
v
a
n
c
e
s in
th
e
d
iag
n
o
sis o
f
p
n
e
u
m
o
n
ia i
n
c
h
il
d
re
n
,”
BM
J
,
v
o
l.
3
5
8
,
J
u
ly
2
0
1
7
.
[1
3
]
Ma
rb
u
n
J.
T
.
,
A
n
d
a
y
a
n
i,
U.
,
“
C
las
sif
ic
a
ti
o
n
o
f
stro
k
e
d
ise
a
se
u
s
in
g
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
,”
J
o
u
rn
a
l
o
f
Ph
y
sic
s: Co
n
fer
e
n
c
e
S
e
rie
s
,
v
o
l.
9
7
8
,
n
o
.
1
,
M
a
rc
h
2
0
1
8
.
[1
4
]
Ka
tak
is S
.
,
Ba
ro
tsis N
.
,
Ka
sta
n
io
ti
s D
.
,
T
h
e
o
h
a
ra
to
s C
.
,
T
so
u
ro
u
n
is
D
.
,
F
o
t
o
p
o
u
l
o
s S
.
,
P
a
n
a
g
io
to
p
o
u
l
o
s E
.
,
“
M
u
sc
le
Ty
p
e
Clas
si
f
ica
ti
o
n
o
n
Ultras
o
u
n
d
Im
a
g
in
g
Us
in
g
De
e
p
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
s
,”
2
0
1
8
IE
EE
1
3
th
Ima
g
e
,
Vi
d
e
o
,
a
n
d
M
u
l
ti
d
ime
n
si
o
n
a
l
S
i
g
n
a
l
Pro
c
e
ss
in
g
W
o
rk
sh
o
p
(
IVM
S
P)
,
Zag
o
ro
c
h
o
ria,
p
p
.
1
-
5
,
2
0
1
8
.
[1
5
]
Ch
e
n
g
P
.
M
.
,
M
a
lh
,
H.
S
.
,
“
T
ra
n
sf
e
r
lea
rn
in
g
w
it
h
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
s
f
o
r
c
las
si
f
ica
ti
o
n
o
f
a
b
d
o
m
in
a
l
u
lt
ra
so
u
n
d
im
a
g
e
s
,
”
J
o
u
rn
a
l
o
f
d
i
g
it
a
l
ima
g
in
g
,
v
o
.
3
0
,
n
o
.
2
,
p
p
.
2
3
4
-
2
4
3
,
A
p
ril
2
0
1
7
.
[1
6
]
Ke
r
m
a
n
y
D.
S
.
,
G
o
ld
b
a
u
m
M
.
,
Ca
i
W
.
,
V
a
len
t
im
C.
C
.
,
L
ian
g
H
.
,
Ba
x
ter
S
.
L
.
,
M
c
Ke
o
w
n
A
.
,
Ya
n
g
G
.
,
W
u
X
.
,
Ya
n
F
.
,
Do
n
g
J.
,
“
Id
e
n
ti
f
y
in
g
m
e
d
ica
l
d
iag
n
o
se
s
a
n
d
trea
tab
le
d
ise
a
se
s
b
y
i
m
a
g
e
-
b
a
se
d
d
e
e
p
lea
rn
in
g
,”
Ce
ll
,
v
o
l.
1
7
2
,
n
o
.
5
,
p
p
.
1
1
2
2
-
1
1
3
1
,
F
e
b
ru
a
ry
2
0
1
8
.
[1
7
]
S
a
ra
iv
a
A
.
,
F
e
rre
ira
N
.
,
L
o
p
e
s
d
e
S
o
u
sa
L
.
,
Ca
rv
a
lh
o
d
a
Co
sta
,
N
.
,
S
o
u
sa
,
J
.
,
S
a
n
to
s
D
.
,
V
a
len
te
,
A
.
,
S
o
a
re
s,
S
.
,
“
Clas
sifi
c
a
ti
o
n
o
f
Im
a
g
e
s
o
f
Ch
il
d
h
o
o
d
P
n
e
u
m
o
n
ia
u
si
n
g
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
s
,”
i
n
Pro
c
e
e
d
in
g
s
o
f
th
e
1
2
th
I
n
ter
n
a
t
io
n
a
l
J
o
in
t
C
o
n
f
e
re
n
c
e
o
n
Bi
o
me
d
ica
l
En
g
in
e
e
ri
n
g
S
y
ste
ms
a
n
d
T
e
c
h
n
o
l
o
g
ies
(
B
IOS
T
EC
2
0
1
9
)
,
p
p
.
1
1
2
-
1
1
9
,
2
0
1
9
.
[1
8
]
Ba
ra
k
a
t
B.
K
.
,
A
l
a
sa
m
R
.
,
El
-
S
a
n
a
J.
,
“
W
o
rd
S
p
o
tt
in
g
Us
in
g
Co
n
v
o
lu
ti
o
n
a
l
S
iam
e
se
Ne
t
w
o
rk
,”
2
0
1
8
1
3
t
h
IAP
R
In
ter
n
a
t
io
n
a
l
W
o
rk
sh
o
p
o
n
Do
c
u
me
n
t
An
a
lys
is S
y
ste
ms
(
DAS
)
,
V
i
e
n
n
a
,
p
p
.
2
2
9
-
2
3
4
,
2
0
1
8
.
[1
9
]
L
e
a
l
-
T
a
i
x
é
,
L
.
,
Ca
n
to
n
-
F
e
rre
r,
C.
,
&
S
c
h
in
d
ler,
K.,
“
L
e
a
rn
in
g
b
y
trac
k
in
g
:
S
ia
m
e
se
C
NN
f
o
r
ro
b
u
st
targ
e
t
a
ss
o
c
iatio
n
,
”
Pro
c
e
e
d
in
g
s
o
f
th
e
IEE
E
Co
n
fer
e
n
c
e
o
n
Co
m
p
u
te
r
Vi
sio
n
a
n
d
Pa
tt
e
rn
Rec
o
g
n
it
i
o
n
W
o
rk
sh
o
p
s
,
p
p
.
3
3
-
4
0
,
2
0
1
6
.
[2
0
]
Ko
c
h
G
.
,
Zem
e
l
R
.
,
S
a
lak
h
u
t
d
in
o
v
R.
,
“
S
iam
e
se
n
e
u
ra
l
n
e
tw
o
rk
s
f
o
r
o
n
e
-
sh
o
t
im
a
g
e
re
c
o
g
n
it
io
n
,”
ICM
L
d
e
e
p
lea
rn
in
g
w
o
rk
sh
o
p
,
L
il
le
,
V
o
l.
2
,
2
0
1
5
.
[2
1
]
Ra
h
u
to
m
o
F
.
,
Kitas
u
k
a
T
.
,
A
rit
su
g
i
M
.
,
“
S
e
m
a
n
ti
c
c
o
sin
e
si
m
il
a
rit
y
,”
T
h
e
7
th
In
ter
n
a
ti
o
n
a
l
S
t
u
d
e
n
t
Co
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
d
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
lo
g
y
ICAS
T
,
K
u
m
a
m
o
to
,
2
0
1
2
.
[2
2
]
Yi
D
.
,
L
e
i
Z
.
,
L
iao
S
.
,
L
i
S
.
Z.
,
“
De
e
p
m
e
tri
c
lea
rn
in
g
f
o
r
p
e
r
so
n
re
-
id
e
n
ti
f
ica
ti
o
n
,”
2
0
1
4
2
2
n
d
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Pa
tt
e
rn
Rec
o
g
n
it
i
o
n
,
S
to
c
k
h
o
lm
,
p
p
.
3
4
-
39
,
2
0
1
4
.
[2
3
]
S
riv
a
sta
v
a
N
.
,
Hin
to
n
G
.
,
Kriz
h
e
v
sk
y
A
.
,
S
u
tsk
e
v
e
r
I
.
,
S
a
lak
h
u
t
d
i
n
o
v
R.
,
“
Dro
p
o
u
t
:
a
sim
p
le
w
a
y
t
o
p
re
v
e
n
t
n
e
u
ra
l
n
e
tw
o
rk
s
f
ro
m
o
v
e
r
f
it
ti
n
g
,”
T
h
e
j
o
u
rn
a
l
o
f
ma
c
h
in
e
le
a
rn
i
n
g
re
se
a
rc
h
,
v
o
l.
1
5
,
n
o
.
1
,
p
p
.
1
9
2
9
-
2
9
5
8
,
Ju
n
e
2
0
1
4
.
[2
4
]
Io
f
fe
S
.
,
S
z
e
g
e
d
y
C.
,
“
Ba
tch
No
r
m
a
li
z
a
ti
o
n
:
A
c
c
e
lera
ti
n
g
De
e
p
N
e
tw
o
rk
T
ra
in
in
g
b
y
Re
d
u
c
in
g
In
t
e
rn
a
l
Co
v
a
riate
S
h
if
t
,”
ICM
L
'
1
5
:
Pro
c
e
e
d
i
n
g
s
o
f
th
e
3
2
n
d
I
n
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
M
a
c
h
in
e
L
e
a
rn
in
g
,
v
o
l.
3
7
,
p
p
.
4
4
8
-
4
5
6
,
Ju
ly
2
0
1
5
.
[2
5
]
Ok
i
H.,
M
iy
a
o
J.,
Ku
rit
a
T
.
,
“
S
i
a
m
e
se
N
e
t
w
o
rk
f
o
r
Clas
si
f
ica
ti
o
n
w
it
h
Op
ti
m
iza
ti
o
n
o
f
AU
C
,
”
.
In
:
G
e
d
e
o
n
T
.
,
W
o
n
g
K.,
Lee
M
.
(e
d
s)
Ne
u
ra
l
In
f
o
r
m
a
ti
o
n
P
ro
c
e
ss
in
g
.
ICONIP
2
0
1
9
.
L
NCS
,
S
p
ri
n
g
e
r,
Ch
a
m
,
v
o
l
1
1
9
5
4
,
p
p
.
3
1
5
-
3
2
7
,
2
0
1
9
.
[2
6
]
Na
sr,
G.
E
.
,
Ba
d
r
E
.
,
Jo
u
n
C.
,
“
Cro
ss
En
tro
p
y
Err
o
r
F
u
n
c
ti
o
n
in
Ne
u
ra
l
Ne
t
w
o
rk
s:
F
o
re
c
a
stin
g
Ga
s
o
li
n
e
De
m
a
n
d
,”
Pro
c
e
e
d
in
g
s
o
f
t
h
e
Fi
ft
e
e
n
t
h
In
ter
n
a
t
io
n
a
l
Fl
o
ri
d
a
Arti
fi
c
i
a
l
In
tell
ig
e
n
c
e
Res
e
a
rc
h
S
o
c
i
e
ty
Co
n
fer
e
n
c
e
,
pp.
3
8
1
-
384
,
Ja
n
u
a
ry
2
0
0
2
.
[2
7
]
S
z
e
g
e
d
y
,
C.
,
V
a
n
h
o
u
c
k
e
,
V
.
,
Io
f
fe
,
S
.
,
S
h
len
s,
J.,
W
o
jn
a
,
Z.
,
“
Re
th
in
k
in
g
th
e
in
c
e
p
ti
o
n
a
rc
h
it
e
c
tu
re
f
o
r
c
o
m
p
u
ter
v
isio
n
,”
2
0
1
6
IEE
E
C
o
n
fer
e
n
c
e
o
n
Co
m
p
u
ter
V
isio
n
a
n
d
P
a
tt
e
rn
Rec
o
g
n
it
i
o
n
(
CVP
R),
L
a
s
V
e
g
a
s,
NV
,
p
p
.
2
8
1
8
-
2
8
2
6
,
2
0
1
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.