I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
4
,
A
u
g
u
s
t
20
25
,
p
p
.
3
7
7
9
~
3
7
9
4
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
4
.
pp
3
7
7
9
-
3
7
9
4
3779
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Enha
ncing
ultras
o
und ima
g
e quali
ty using
deep
stru
cture
o
f
residua
l net
wo
rk
Ade
I
ria
ni Sa
pitr
i,
S
it
i N
ur
m
a
ini
,
M
uh
a
m
m
a
d Na
ufa
l
Ra
chm
a
t
ull
a
h,
Anni
s
a
Da
rm
a
wa
hy
un
i,
F
irda
us
F
irda
u
s
,
Ang
g
un
I
s
l
a
m
i,
B
a
m
ba
ng
T
utuk
o
,
Ak
hi
a
r
Wis
t
a
Arum
I
n
t
e
l
l
i
g
e
n
t
S
y
st
e
m
R
e
s
e
a
r
c
h
G
r
o
u
p
,
F
a
c
u
l
t
y
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
U
n
i
v
e
r
s
i
t
a
s Sr
i
w
i
j
a
y
a
,
P
a
l
e
mb
a
n
g
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
2
,
2
0
2
4
R
ev
is
ed
Ap
r
5
,
2
0
2
5
Acc
ep
ted
Ma
y
2
4
,
2
0
2
5
Ultras
o
n
o
g
ra
p
h
y
,
a
m
e
d
ica
l
ima
g
in
g
tec
h
n
iq
u
e
,
is
o
f
ten
a
ffe
c
ted
b
y
v
a
rio
u
s
ty
p
e
s
o
f
n
o
ise
a
n
d
lo
w
b
r
ig
h
tn
e
ss
,
wh
ich
c
a
n
re
su
l
t
i
n
l
o
w
ima
g
e
q
u
a
li
ty
.
Th
e
se
d
ra
wb
a
c
k
s
c
a
n
si
g
n
if
ica
n
tl
y
im
p
e
d
e
a
c
c
u
ra
te
i
n
terp
re
tati
o
n
a
n
d
h
i
n
d
e
r
e
ffe
c
ti
v
e
m
e
d
ica
l
d
ia
g
n
o
se
s.
T
h
e
re
fo
re
,
imp
ro
v
in
g
ima
g
e
q
u
a
li
ty
is
a
n
e
ss
e
n
ti
a
l
a
sp
e
c
t
o
f
t
h
e
field
o
f
u
lt
ra
so
u
n
d
sy
ste
m
s.
T
h
is
st
u
d
y
a
ims
t
o
e
n
h
a
n
c
e
t
h
e
q
u
a
li
t
y
o
f
u
lt
ra
so
u
n
d
ima
g
e
s
u
si
n
g
d
e
e
p
lea
rn
i
n
g
(DL).
T
h
e
e
x
p
e
rime
n
t
is
c
o
n
d
u
c
ted
u
si
n
g
a
c
u
sto
m
d
a
tas
e
t
c
o
n
sisti
n
g
o
f
2
,
1
7
5
in
fa
n
t
h
e
a
rt
u
lt
ra
so
u
n
d
ima
g
e
s
c
o
l
lec
ted
fro
m
In
d
o
n
e
sia
n
h
o
s
p
it
a
ls,
a
n
d
t
h
e
m
o
d
e
l
is
su
b
se
q
u
e
n
tl
y
g
e
n
e
ra
li
z
e
d
u
si
n
g
o
th
e
r
d
a
tas
e
ts.
We
p
r
o
p
o
se
e
n
h
a
n
c
e
d
d
e
e
p
re
sid
u
a
l
n
e
two
r
k
c
o
m
b
in
e
d
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
s
(EDR
-
CNN
s)
t
o
imp
ro
v
e
t
h
e
ima
g
e
q
u
a
li
t
y
.
Af
ter
th
e
e
n
h
a
n
c
e
m
e
n
t
p
r
o
c
e
ss
,
o
u
r
m
o
d
e
l
a
c
h
iev
e
d
p
e
a
k
sig
n
a
l
-
to
-
n
o
ise
ra
t
io
(P
S
NR)
a
n
d
stru
c
tu
ra
l
sim
il
a
r
it
y
in
d
e
x
m
e
tri
c
s
(S
S
IM
)
sc
o
re
s
o
f
3
8
.
3
5
a
n
d
0
.
9
2
re
sp
e
c
ti
v
e
ly
,
o
u
t
p
e
rfo
r
m
in
g
o
t
h
e
r
m
e
th
o
d
s.
T
h
e
b
e
n
c
h
m
a
rk
i
n
g
wit
h
o
th
e
r
u
l
tras
o
u
n
d
m
e
d
ica
l
ima
g
e
s
in
d
ica
tes
th
a
t
o
u
r
p
r
o
p
o
se
d
m
o
d
e
l
p
ro
d
u
c
e
s
g
o
o
d
p
e
rfo
rm
a
n
c
e
,
a
s e
v
id
e
n
c
e
d
b
y
h
ig
h
e
r
P
S
NR,
l
o
we
r
S
S
IM
,
a
d
e
c
re
a
se
in
m
e
a
n
sq
u
a
re
e
rro
r
(
M
S
E),
a
n
d
a
l
o
we
r
c
o
n
tras
t
imp
r
o
v
e
m
e
n
t
i
n
d
e
x
(CII)
.
In
c
o
n
c
lu
si
o
n
,
t
h
is
stu
d
y
e
n
c
a
p
s
u
late
s
th
e
fo
rth
c
o
m
i
n
g
tren
d
s
i
n
a
d
v
a
n
c
i
n
g
l
o
w
-
il
l
u
m
in
a
ti
o
n
ima
g
e
e
n
h
a
n
c
e
m
e
n
t,
a
lo
n
g
wit
h
e
x
p
lo
ri
n
g
t
h
e
p
re
v
a
il
in
g
c
h
a
ll
e
n
g
e
s
a
n
d
p
o
ten
ti
a
l
d
ire
c
ti
o
n
s
fo
r
fu
rth
e
r
re
se
a
rc
h
.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
E
n
h
an
ce
m
e
n
t
Me
d
ical
im
ag
e
R
esid
u
al
n
etwo
r
k
Ultr
aso
n
o
g
r
ap
h
y
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
:
Sit
i N
u
r
m
ain
i
I
n
tellig
en
t Sy
s
tem
R
esear
ch
Gr
o
u
p
,
Facu
lty
o
f
C
o
m
p
u
ter
Sci
en
ce
,
Un
iv
er
s
itas
Sriwijay
a
Palem
b
an
g
,
So
u
t
h
Su
m
ater
a,
I
n
d
o
n
esia
E
m
ail: sit
i_
n
u
r
m
ain
i@
u
n
s
r
i.a
c.
id
1.
I
NT
RO
D
UCT
I
O
N
E
n
h
an
cin
g
th
e
q
u
ality
o
f
m
e
d
ical
im
ag
es,
p
a
r
ticu
lar
ly
i
n
u
ltra
s
o
u
n
d
s
y
s
tem
s
,
is
a
cr
u
c
ial
ar
ea
o
f
f
o
cu
s
in
th
e
m
ed
ical
f
ield
.
T
h
i
s
is
p
r
im
ar
ily
attr
ib
u
ted
to
th
e
ex
ten
s
iv
e
u
tili
za
tio
n
o
f
u
ltra
s
o
u
n
d
tech
n
o
lo
g
y
in
m
ed
ical
d
iag
n
o
s
is
an
d
a
wid
e
r
an
g
e
o
f
m
ed
ica
l
p
r
o
ce
d
u
r
es.
Ultr
aso
u
n
d
im
ag
es
o
f
te
n
s
u
f
f
e
r
f
r
o
m
is
s
u
es
s
u
ch
as
n
o
is
e
an
d
b
l
u
r
,
wh
ic
h
ca
n
s
ig
n
if
ican
tly
im
p
ed
e
m
e
d
ic
al
in
ter
p
r
etatio
n
a
n
d
ac
c
u
r
ate
d
iag
n
o
s
is
[
1
]
–
[
4
]
.
Sev
er
al
f
ac
to
r
s
co
n
tr
ib
u
te
to
t
h
e
lo
w
q
u
ality
o
f
u
ltra
s
o
u
n
d
i
m
ag
es.
On
e
o
f
th
ese
f
ac
to
r
s
is
ac
o
u
s
tic
v
ar
iab
ilit
y
,
wh
er
e
th
e
u
ltra
s
o
u
n
d
im
ag
e
is
af
f
ec
ted
b
y
ac
o
u
s
tic
v
ar
iatio
n
s
in
th
e
b
o
d
y
,
ca
u
s
in
g
n
o
is
e
in
th
e
im
a
g
e
[
5
]
–
[
8
]
.
I
n
ad
d
itio
n
,
s
ig
n
al
d
ef
icien
cie
s
o
r
atten
u
atio
n
s
m
ay
o
cc
u
r
a
s
th
e
u
ltra
s
o
u
n
d
s
ig
n
al
p
ass
es
th
r
o
u
g
h
d
if
f
e
r
en
t
b
o
d
y
tis
s
u
es,
an
d
r
eso
lu
tio
n
lim
itatio
n
s
m
ay
ar
is
e
d
u
e
to
p
atien
t
m
o
v
em
e
n
t,
s
u
b
o
p
tim
al
p
o
s
itio
n
in
g
,
o
r
th
e
u
s
e
o
f
less
s
o
p
h
is
ticated
u
ltra
s
o
u
n
d
tech
n
o
lo
g
y
[
9
]
–
[
1
1
]
.
T
h
ese
p
r
o
b
lem
s
ca
n
s
ig
n
if
ican
tly
af
f
ec
t
th
e
qua
lity
o
f
u
ltra
s
o
u
n
d
im
ag
es
an
d
m
a
k
e
m
ed
ical
i
n
ter
p
r
eta
tio
n
an
d
d
iag
n
o
s
is
ch
allen
g
in
g
.
T
o
a
d
d
r
ess
th
is
p
r
o
b
lem
,
m
an
y
r
esear
ch
s
tu
d
i
es
h
av
e
f
o
cu
s
ed
o
n
d
e
v
elo
p
in
g
m
eth
o
d
s
to
im
p
r
o
v
e
th
e
q
u
ality
o
f
u
ltra
s
o
u
n
d
im
ag
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
7
9
-
3794
3780
T
h
er
e
ar
e
two
b
r
o
ad
ca
te
g
o
r
ie
s
o
f
tr
ad
itio
n
a
l
m
eth
o
d
s
f
o
r
lo
w
-
lig
h
t
en
h
a
n
ce
m
en
t
s
u
ch
as
h
is
to
g
r
am
eq
u
aliza
tio
n
[
1
1
]
,
[
1
2
]
a
n
d
R
etin
ex
m
o
d
els
[
1
3
]
,
[
1
4
]
.
T
h
e
h
is
to
g
r
am
tech
n
iq
u
e
aim
s
to
en
h
an
ce
th
e
co
n
tr
ast
o
f
an
im
a
g
e
b
y
r
ed
is
tr
ib
u
tin
g
th
e
in
ten
s
ity
v
alu
es o
f
t
h
e
p
ix
el
s
,
r
esu
ltin
g
in
a
b
r
i
g
h
ter
an
d
m
o
r
e
v
ib
r
a
n
t im
ag
e
.
W
h
er
ea
s
a
ty
p
ical
ap
p
r
o
ac
h
b
ased
o
n
R
etin
ex
m
o
d
els
d
ec
o
m
p
o
s
es
a
lo
w
-
lig
h
t
im
ag
e
in
to
tw
o
co
m
p
o
n
en
ts
:
th
e
r
ef
lectio
n
co
m
p
o
n
en
t
a
n
d
th
e
illu
m
in
atio
n
co
m
p
o
n
e
n
t
u
s
in
g
p
r
io
r
s
o
r
r
eg
u
lar
izatio
n
s
.
T
h
e
esti
m
ated
r
ef
lectio
n
c
o
m
p
o
n
en
t
is
co
n
s
i
d
er
ed
t
h
e
e
n
h
an
ce
d
r
esu
lt.
Ho
wev
er
s
u
ch
two
m
et
h
o
d
s
h
av
e
s
ev
er
al
lim
itatio
n
s
[
1
4
]
,
i)
B
o
th
h
is
to
g
r
am
eq
u
aliza
tio
n
an
d
R
etin
ex
m
o
d
els
ca
n
b
e
p
ar
ticu
lar
l
y
p
r
o
b
lem
atic
wh
en
d
ea
lin
g
with
im
ag
es c
o
n
tain
in
g
s
h
ar
p
ed
g
es
an
d
tex
tu
r
es,
as th
ey
m
ay
b
ec
o
m
e
ex
ag
g
e
r
ated
an
d
ap
p
ea
r
a
r
tific
ial; ii)
I
n
s
u
ch
ap
p
r
o
ac
h
es,
th
e
ex
is
tin
g
n
o
i
s
e
in
th
e
im
a
g
es
is
o
f
te
n
d
is
r
eg
ar
d
ed
,
lead
in
g
to
eith
er
its
p
er
s
is
ten
ce
o
r
am
p
lific
atio
n
in
th
e
e
n
h
an
ce
d
r
esu
lts
;
iii)
C
o
m
p
lex
o
p
tim
izatio
n
p
r
o
ce
s
s
es
in
v
o
lv
e
d
in
R
etin
ex
m
o
d
els
an
d
th
e
co
m
p
u
tatio
n
o
f
im
ag
e
h
is
to
g
r
am
s
an
d
cu
m
u
lativ
e
d
i
s
tr
ib
u
tio
n
f
u
n
ctio
n
s
in
h
is
to
g
r
am
eq
u
aliza
tio
n
in
cr
ea
s
in
g
th
e
co
m
p
u
tatio
n
al
t
im
e,
an
d
iv
)
B
o
th
ap
p
r
o
ac
h
es n
ec
ess
itate
m
an
u
al
ad
ju
s
tm
en
ts
an
d
in
ter
v
en
tio
n
s
to
ac
co
m
m
o
d
ate
s
p
ec
if
ic
s
ce
n
ar
io
s
o
r
ed
g
e
ca
s
es.
I
n
r
ec
en
t
tim
es,
p
r
o
f
o
u
n
d
a
d
v
an
ce
m
e
n
ts
h
av
e
b
ee
n
m
ad
e
in
th
e
r
ea
lm
o
f
im
a
g
e
en
h
an
ce
m
en
t
tech
n
iq
u
es,
s
p
ec
if
ically
lev
er
ag
in
g
DL
tech
n
i
q
u
es
s
u
ch
as
co
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs)
[
1
5
]
,
[
1
6
]
.
T
h
ese
r
em
ar
k
ab
le
d
ev
elo
p
m
e
n
ts
h
av
e
s
h
o
wn
g
r
ea
t
p
o
ten
ti
al
in
r
e
d
u
cin
g
n
o
is
e,
au
g
m
en
tin
g
r
eso
lu
tio
n
,
a
n
d
m
itig
atin
g
ar
tifa
cts.
As
a
r
esu
lt,
C
NN
s
h
av
e
em
er
g
ed
as
h
i
g
h
ly
p
r
o
m
is
in
g
s
o
lu
tio
n
s
to
t
ac
k
le
th
e
p
er
s
is
ten
t
ch
allen
g
e
o
f
en
h
an
cin
g
u
ltr
aso
u
n
d
im
a
g
e
q
u
ality
,
u
ltim
ately
lead
in
g
t
o
a
s
u
b
s
tan
ti
al
im
p
r
o
v
e
m
en
t
in
d
iag
n
o
s
tic
ac
cu
r
ac
y
.
T
h
e
r
e
h
as
b
ee
n
r
esear
ch
f
o
cu
s
ed
o
n
d
ev
elo
p
in
g
m
o
r
e
s
p
ec
ialized
an
d
ef
f
ec
tiv
e
DL
tech
n
iq
u
es
to
im
p
r
o
v
e
u
ltra
s
o
u
n
d
im
ag
e
q
u
ality
,
s
u
ch
as
u
s
i
n
g
tr
a
n
s
f
er
lea
r
n
in
g
o
n
e
x
is
tin
g
DL
ar
ch
itectu
r
es
[
1
2
]
–
[
1
4
]
,
[
1
7
]
–
[
2
0
]
.
T
h
ese
a
d
v
an
ce
s
co
n
tin
u
e
to
im
p
r
o
v
e
t
h
e
q
u
ality
o
f
u
ltra
s
o
u
n
d
im
ag
es
an
d
im
p
r
o
v
e
th
e
ac
cu
r
ac
y
o
f
m
ed
ical
d
iag
n
o
s
es
an
d
in
ter
v
en
tio
n
s
.
T
h
is
r
esear
ch
co
n
tr
ib
u
tes as f
o
llo
ws:
a.
Pro
p
o
s
in
g
a
cu
s
to
m
d
ee
p
r
esid
u
al
C
NNs
ar
ch
itectu
r
e
to
im
p
r
o
v
e
th
e
q
u
ality
o
f
u
ltra
s
o
u
n
d
i
m
ag
es
in
ter
m
s
o
f
b
o
t
h
p
er
f
o
r
m
a
n
ce
an
d
ef
f
ici
en
cy
.
b.
Valid
atin
g
th
e
ef
f
ec
tiv
en
ess
o
f
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
u
s
in
g
r
ig
o
r
o
u
s
ex
p
er
im
en
ts
an
d
b
en
c
h
m
ar
k
s
ag
ain
s
t
estab
lis
h
ed
s
tate
-
of
-
th
e
-
ar
t D
L
m
o
d
els.
c.
Dem
o
n
s
tr
atin
g
th
e
p
r
o
p
o
s
ed
cu
s
to
m
ar
ch
itectu
r
e
with
f
o
u
r
u
ltra
s
o
u
n
d
d
atasets
i.e
.
,
i
n
f
an
t
h
ea
r
t,
f
etal
h
ea
r
t,
f
etal
h
ea
d
,
an
d
ab
d
o
m
en
.
d.
I
m
p
lem
en
tin
g
th
e
p
r
o
p
o
s
ed
m
o
d
el
f
o
r
p
r
e
d
ictin
g
th
e
f
etal
h
e
ar
t
s
u
b
s
tr
u
ctu
r
e
ab
n
o
r
m
ality
.
2.
M
AT
E
R
I
AL
S AN
D
M
E
T
H
O
D
T
h
e
g
en
e
r
al
m
eth
o
d
o
lo
g
y
o
f
o
u
r
ex
p
er
im
en
t
is
d
e
p
icted
in
Fig
u
r
e
1
,
c
o
n
s
is
tin
g
o
f
f
o
u
r
m
ain
s
tag
es.
T
h
e
f
ir
s
t
s
tag
e,
d
ata
ac
q
u
is
itio
n
is
ca
r
r
ied
o
u
t.
Data
f
o
r
th
e
a
cq
u
is
itio
n
was
o
b
tain
ed
f
r
o
m
s
ev
er
al
p
r
ev
io
u
s
ly
co
llected
u
ltra
s
o
u
n
d
v
id
eo
s
.
T
h
e
d
ata
is
th
en
ex
tr
ac
ted
b
ased
o
n
th
e
f
r
am
e
a
n
d
cr
o
p
p
ed
th
e
h
ea
r
t
o
b
ject
d
ir
e
ctly
.
T
h
e
s
ec
o
n
d
s
tag
e
in
v
o
lv
es
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
,
wh
er
e
lo
w
-
r
eso
lu
tio
n
d
ata
u
s
in
g
n
o
is
e
in
jectio
n
an
d
h
ig
h
r
eso
lu
tio
n
ar
e
g
e
n
er
ated
as
tar
g
ets.
T
h
is
p
r
o
ce
s
s
in
v
o
lv
es
lab
elin
g
th
e
d
ata
f
o
r
f
u
r
t
h
er
an
aly
s
is
.
I
n
th
e
th
ir
d
s
tag
e,
th
e
p
r
o
p
o
s
ed
d
ee
p
lear
n
in
g
alg
o
r
ith
m
p
r
o
p
o
s
es
an
en
h
an
ce
d
r
esid
u
al
C
NN
ar
c
h
itectu
r
e.
T
h
e
f
in
al
s
tag
e
in
v
o
lv
es
p
er
f
o
r
m
an
ce
m
ea
s
u
r
em
en
t,
wh
e
r
e
th
e
p
ea
k
s
ig
n
al
-
to
-
n
o
is
e
r
atio
(
PS
NR
)
,
s
tr
u
ctu
r
al
s
im
ilar
ity
in
d
ex
m
etr
ics
(
SS
I
M)
,
m
ea
n
s
q
u
ar
e
er
r
o
r
(
MSE
)
,
an
d
c
o
n
tr
a
s
t
im
p
r
o
v
em
en
t
in
d
ex
(
C
I
I)
v
a
lu
es
ar
e
ca
lcu
lated
to
ev
alu
ate
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
d
ee
p
-
lear
n
in
g
m
o
d
el.
T
h
e
wh
o
le
s
tep
s
co
n
s
titu
te
th
e
g
en
er
al
m
eth
o
d
o
l
o
g
y
o
f
o
u
r
e
x
p
er
im
e
n
t.
Fig
u
r
e
1
.
T
h
e
g
e
n
er
al
m
eth
o
d
o
lo
g
y
o
f
o
u
r
ex
p
er
im
en
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
n
h
a
n
cin
g
u
ltr
a
s
o
u
n
d
ima
g
e
q
u
a
lity u
s
in
g
d
ee
p
s
tr
u
ctu
r
e
o
f
r
esid
u
a
l n
etw
o
r
k
(
A
d
e
I
r
ia
n
i
S
a
p
itr
i
)
3781
2
.
1
.
Da
t
a
a
cquis
it
io
n
T
h
e
d
ataset
u
s
ed
in
th
is
r
ese
ar
ch
co
n
s
is
ts
o
f
m
ed
ical
im
ag
es
ac
q
u
ir
ed
f
r
o
m
u
ltra
-
s
o
u
n
d
d
ev
ices,
in
clu
d
in
g
im
ag
es
o
f
in
f
a
n
t
h
ea
r
ts
,
f
etal
h
ea
r
ts
,
f
etal
h
ea
d
s
,
an
d
ab
d
o
m
en
s
.
W
h
ile
in
f
an
t
an
d
f
etal
h
ea
r
t
d
atasets
ar
e
cu
s
to
m
-
m
ad
e
f
r
o
m
I
n
d
o
n
esian
h
o
s
p
itals
,
th
e
r
em
ain
in
g
d
atas
ets
ar
e
s
o
u
r
ce
d
f
r
o
m
p
u
b
lic
r
ep
o
s
ito
r
ies.
T
h
e
d
ata
d
is
tr
ib
u
tio
n
o
f
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
ca
n
b
e
f
o
u
n
d
in
T
ab
le
1
.
T
h
e
s
tu
d
y
was
ca
r
r
ied
o
u
t
in
two
m
ain
s
tag
es:
i)
I
n
th
e
f
i
r
s
t
s
tag
e,
we
f
o
cu
s
ed
o
n
g
e
n
er
atin
g
an
im
a
g
e
en
h
a
n
ce
m
en
t m
o
d
el
u
s
in
g
im
ag
es
o
f
in
f
a
n
t
h
ea
r
ts
,
an
d
ii)
Su
b
s
e
q
u
en
tly
,
t
h
e
en
h
a
n
ce
m
en
t
m
o
d
el
th
at
d
em
o
n
s
tr
ated
th
e
b
es
t
p
er
f
o
r
m
an
ce
was
s
elec
ted
an
d
ap
p
lied
t
o
th
e
o
th
er
d
atasets
f
o
r
b
en
c
h
m
ar
k
in
g
p
u
r
p
o
s
es.
T
ab
le
1
.
Data
d
is
tr
ib
u
tio
n
f
o
r
t
h
e
lear
n
in
g
p
r
o
ce
s
s
.
U
S
G
O
b
j
e
c
t
Tr
a
i
n
i
n
g
V
a
l
i
d
a
t
i
o
n
Te
st
i
n
g
U
n
see
n
To
t
a
l
I
n
f
a
n
t
's
h
e
a
r
t
1
1
7
7
7
3
2
2
4
4
60
2
2
1
3
F
e
t
a
l
h
e
a
r
t
4
1
0
78
26
28
5
4
2
F
e
t
a
l
h
e
a
d
9
1
5
2
7
9
84
56
1
3
3
4
A
b
d
o
me
n
4
7
2
1
3
2
84
23
7
1
1
T
h
e
cu
s
to
m
o
f
th
e
in
f
a
n
t
h
ea
r
t
d
ataset
is
o
b
tain
ed
f
r
o
m
a
g
en
er
al
h
o
s
p
ital,
Dr
.
Mo
h
am
m
ad
Ho
esin
,
I
n
d
o
n
esia.
Vid
eo
s
ar
e
ca
p
tu
r
ed
u
s
in
g
a
Ph
ilip
s
E
PIQ
E
lite u
ltra
s
o
u
n
d
m
ac
h
in
e
an
d
v
id
eo
f
ile
s
ize
s
r
an
g
e
f
r
o
m
30
-
5
0
MB.
T
o
g
en
er
ate
a
r
elia
b
le
im
ag
e
en
h
an
ce
m
en
t
m
o
d
el,
we
u
s
ed
1
,
1
7
1
tr
ain
in
g
d
ata
an
d
9
7
6
v
alid
atio
n
d
at
a.
All
th
e
i
m
ag
e
d
ata
u
s
e
d
to
b
u
ild
th
e
im
a
g
e
en
h
an
ce
m
en
t
m
o
d
el
o
r
ig
in
ated
f
r
o
m
im
ag
es
o
f
a
ch
ild
'
s
h
ea
r
t.
I
t
ca
n
b
e
s
ee
n
a
s
am
p
le
im
ag
e
o
f
a
n
in
f
a
n
t’
s
h
ea
r
t
u
s
ed
f
o
r
m
o
d
el
en
h
a
n
ce
m
e
n
t
d
ev
elo
p
m
en
t
in
Fig
u
r
e
2.
Fig
u
r
e
2
.
A
s
am
p
le
im
a
g
e
o
f
a
n
in
f
an
t'
s
h
ea
r
t a
s
r
aw
d
ata
f
o
r
th
e
d
ev
elo
p
m
en
t
o
f
th
e
m
o
d
el
en
h
an
ce
m
e
n
t
2
.
2
.
Da
t
a
pre
-
pro
ce
s
s
ing
B
ased
o
n
th
e
d
ata
ac
q
u
is
itio
n
p
r
o
ce
s
s
,
p
r
ev
io
u
s
ly
co
llected
s
ev
er
al
u
ltra
s
o
u
n
d
v
i
d
eo
s
.
Data
is
ex
tr
ac
ted
f
r
a
m
e
b
y
f
r
am
e
,
a
n
d
all
d
ata
is
cr
o
p
p
e
d
to
8
0
0
×
6
0
0
s
ize.
T
h
en
,
th
e
d
atas
et
is
p
r
o
ce
s
s
ed
b
y
p
r
o
d
u
cin
g
h
ig
h
-
r
eso
lu
tio
n
a
n
d
lo
w
-
r
eso
lu
tio
n
im
ag
es
as
in
p
u
t
f
o
r
m
o
d
el
tr
ain
in
g
as
s
h
o
w
n
in
Fig
u
r
e
3
.
Hig
h
-
r
eso
lu
tio
n
im
ag
es
ar
e
th
e
tar
g
et
o
u
tp
u
t
f
r
o
m
th
e
im
a
g
e
p
r
o
ce
s
s
in
g
p
r
o
ce
s
s
an
d
lo
w
-
r
e
s
o
lu
tio
n
im
ag
es
is
g
en
er
ated
as
th
e
i
n
p
u
t
f
o
r
th
e
im
ag
e
en
h
an
ce
m
e
n
t
m
o
d
el.
L
o
w
-
r
eso
lu
tio
n
im
ag
es
as
in
p
u
t
m
o
d
els
u
s
e
n
o
is
e
in
jectio
n
wh
ich
o
f
ten
o
cc
u
r
s
in
m
ed
ical
u
ltra
s
o
u
n
d
d
ata
th
at
ca
n
g
iv
e
th
e
r
aw
im
a
g
e
a
s
m
o
o
th
er
o
r
m
o
r
e
b
lu
r
r
y
ap
p
ea
r
an
ce
.
Dif
f
er
en
t t
y
p
es o
f
n
o
is
e
ar
e
ad
d
ed
,
s
u
ch
as p
o
is
o
n
,
s
alt
-
an
d
-
p
ep
p
er
n
o
i
s
e,
o
r
s
p
ec
k
le
n
o
is
e.
W
e
co
n
tr
o
l th
e
in
ten
s
ity
an
d
t
y
p
e
o
f
n
o
is
e
to
ad
d
b
ased
o
n
t
h
e
d
if
f
er
e
n
t settin
g
s
to
ac
h
iev
e
th
e
d
esire
d
lev
el
o
f
n
o
is
e.
Hig
h
-
r
eso
lu
tio
n
im
ag
es
s
er
v
e
as
th
e
tar
g
et
r
esu
lt
o
f
im
ag
e
p
r
o
ce
s
s
in
g
,
wh
ile
lo
w
-
r
eso
lu
tio
n
im
ag
es
ar
e
th
e
in
p
u
t
n
ee
d
ed
to
tr
ain
th
e
m
o
d
el
to
lear
n
p
atter
n
s
in
l
o
w
-
r
eso
lu
tio
n
d
ata
a
n
d
im
p
r
o
v
e
im
ag
e
q
u
ality
.
T
h
e
r
esu
l
ts
o
f
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e
ar
e
v
is
u
alize
d
in
Fig
u
r
e
4
.
Af
ter
lo
w
-
r
eso
lu
tio
n
im
a
g
es
ar
e
en
ter
ed
in
to
t
h
e
m
o
d
el,
a
m
u
ltis
ca
le
in
f
o
r
m
ati
o
n
r
e
p
r
esen
tatio
n
p
r
o
ce
s
s
is
p
er
f
o
r
m
ed
,
co
m
b
i
n
in
g
in
f
o
r
m
atio
n
f
r
o
m
v
ar
io
u
s
s
p
atial
s
ca
les
an
d
ex
tr
ac
tin
g
r
elev
an
t
f
ea
t
u
r
es
to
im
p
r
o
v
e
im
ag
e
q
u
ality
.
B
y
tak
i
n
g
a
d
v
an
tag
e
o
f
lo
w
a
n
d
h
ig
h
-
r
eso
lu
tio
n
im
ag
e
ty
p
es,
th
e
en
h
an
ce
m
en
t
m
o
d
el
ca
n
ef
f
ec
tiv
ely
im
p
r
o
v
e
i
m
ag
e
q
u
ality
an
d
p
r
o
d
u
ce
h
ig
h
-
q
u
ality
im
ag
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
7
9
-
3794
3782
Fig
u
r
e
3
.
Data
p
r
e
-
p
r
o
ce
s
s
in
g
p
r
o
ce
s
s
O
r
i
g
i
n
a
l
i
ma
g
e
(
h
i
g
h
r
e
s
o
l
u
t
i
o
n
)
Th
e
l
o
w
-
r
e
so
l
u
t
i
o
n
i
ma
g
e
b
y
a
d
d
i
n
g
a
c
o
n
t
r
o
l
l
e
d
a
m
o
u
n
t
o
f
n
o
i
s
e
P
o
i
ss
o
n
S
p
e
c
k
l
e
G
a
u
ss
i
a
n
S
a
l
t
P
e
p
p
e
r
-
L
am
=
3
.
0
,
s
i
z
e
=
(
5
,
5
)
M
e
d
i
a
n
B
l
u
r
(
i
mg
,
5
)
M
e
a
n
=
0
,
v
a
r
i
a
n
c
e
=
5
0
,
a
n
d
s
i
g
m
a
=
v
a
r
i
a
n
c
e
*
0
.
5
S
a
l
t
=
1
(
2
5
5
,
2
5
5
,
2
5
5
)
a
n
d
P
e
p
p
e
r
=
2
(
0
,
0
,
0
)
Fig
u
r
e
4
.
Hig
h
-
r
eso
lu
tio
n
an
d
lo
w
-
r
eso
lu
tio
n
2
.
3
.
I
ma
g
e
enha
ncem
ent
m
o
del
T
h
is
s
tu
d
y
p
r
o
p
o
s
ed
cu
s
to
m
i
ze
d
en
h
an
ce
d
d
ee
p
r
esid
u
al
n
etwo
r
k
co
m
b
in
ed
co
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
s
(
E
DR
-
C
NNs)
ar
ch
itectu
r
e
as
s
h
o
wn
in
Fig
u
r
e
5
.
T
h
e
s
tr
u
ctu
r
e
is
m
o
d
if
ied
b
y
em
u
latin
g
r
esid
u
al
n
etwo
r
k
(
R
esNet)
wh
ich
is
d
esig
n
ed
to
ad
d
r
ess
th
e
ch
allen
g
e
o
f
im
ag
e
n
o
is
e
[
1
4
]
,
[
1
5
]
.
T
h
e
E
DR
-
C
NN
s
tr
u
ctu
r
e
f
o
c
u
s
es
o
n
co
n
v
o
l
u
tio
n
wid
en
i
n
g
,
en
ab
lin
g
n
e
two
r
k
s
to
e
x
p
an
d
th
eir
v
is
u
al
r
an
g
e
with
o
u
t
co
m
p
r
o
m
is
in
g
im
ag
e
r
eso
lu
t
io
n
.
I
t
u
s
es
m
u
lti
-
r
eso
lu
tio
n
b
r
an
ch
es
with
d
if
f
er
e
n
t
r
eso
lu
tio
n
s
to
h
an
d
le
in
f
o
r
m
atio
n
at
v
ar
io
u
s
s
ca
les.
T
h
e
n
etwo
r
k
b
y
p
a
s
s
co
n
n
ec
tio
n
s
en
ab
le
th
e
lear
n
in
g
o
f
m
o
r
e
co
m
p
lex
f
ea
tu
r
es
an
d
ac
ce
ler
ate
n
etwo
r
k
co
n
v
er
g
en
ce
.
I
n
o
r
d
e
r
to
im
p
r
o
v
e
m
o
d
el
p
er
f
o
r
m
an
ce
an
d
ef
f
icien
cy
,
th
e
m
o
d
if
ied
s
tr
u
ctu
r
e
in
clu
d
es
d
ee
p
ly
s
ep
ar
ab
le
atten
tio
n
an
d
co
n
v
o
l
u
tio
n
m
o
d
u
les.
T
h
e
n
etwo
r
k
s
tr
u
ctu
r
e
is
s
h
o
wn
in
T
ab
le
2
.
Dee
p
co
n
v
o
l
u
tio
n
is
u
s
ed
to
r
ed
u
ce
th
e
co
m
p
u
tatio
n
al
co
m
p
lex
ity
in
th
e
c
o
n
v
o
lu
tio
n
al
lay
er
an
d
r
e
d
u
ce
th
e
n
u
m
b
er
o
f
p
ar
am
eter
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
n
h
a
n
cin
g
u
ltr
a
s
o
u
n
d
ima
g
e
q
u
a
lity u
s
in
g
d
ee
p
s
tr
u
ctu
r
e
o
f
r
esid
u
a
l n
etw
o
r
k
(
A
d
e
I
r
ia
n
i
S
a
p
itr
i
)
3783
r
eq
u
ir
ed
b
y
th
e
m
o
d
el.
I
n
ad
d
itio
n
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
h
as
b
ee
n
m
o
d
if
ied
with
a
s
am
p
li
n
g
tech
n
iq
u
e
ca
lled
s
u
b
-
p
ix
el
c
o
n
v
o
lu
tio
n
to
in
cr
ea
s
e
th
e
im
ag
e
r
eso
lu
tio
n
af
ter
s
u
p
er
-
r
eso
lu
tio
n
p
r
o
ce
s
s
in
g
.
T
h
is
tech
n
iq
u
e
esti
m
ates n
ew
p
ix
els
an
d
s
tu
d
ies ex
is
tin
g
s
p
atial
p
atter
n
s
in
l
o
w
-
r
eso
lu
tio
n
im
ag
e
q
u
ality
.
T
h
e
s
tr
u
ctu
r
e
o
f
th
e
E
DR
-
C
NN
s
m
o
d
el
co
m
p
r
is
es
co
n
v
o
lu
tio
n
s
with
3
×
3
f
ilter
s
f
o
llo
wed
b
y
th
e
L
ea
k
y
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
an
d
C
h
ar
b
o
n
n
ier
L
o
s
s
,
as
(
1
)
:
(
,
)
=
(
(
−
2
+
2
)
2
(
1
)
wh
er
e
an
d
ar
e
th
e
in
p
u
t a
n
d
t
ar
g
et
im
ag
es,
r
esp
ec
tiv
ely
.
is
a
s
m
all
co
n
s
tan
t to
av
o
id
d
iv
is
io
n
b
y
ze
r
o
.
Fig
u
r
e
5
.
T
h
e
p
r
o
p
o
s
ed
E
DR
-
C
NNs a
r
ch
itectu
r
e
T
ab
le
2
.
T
h
e
p
r
o
p
o
s
ed
E
DR
-
C
NNs n
etwo
r
k
C
o
n
v
o
l
u
t
i
o
n
La
y
e
r
(
t
y
p
e
)
O
u
t
p
u
t
S
h
a
p
e
N
u
mb
e
r
o
f
P
a
r
a
m
e
t
e
r
s
I
n
p
u
t
I
mag
e
I
n
p
u
t
_
1
(
I
n
p
u
t
L
a
y
e
r
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
3
)
]
0
C
o
n
v
1
C
o
n
v
2
d
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
1
7
9
2
l
e
a
k
y
_
r
e
_
l
u
(
Le
a
k
y
R
e
LU
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
0
C
o
n
v
2
c
o
n
v
2
d
_
1
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
3
6
9
2
8
l
e
a
k
y
_
r
e
_
l
u
_
1
(
Le
a
k
y
R
e
LU
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
0
C
o
n
v
3
c
o
n
v
2
d
_
2
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
3
6
9
2
8
l
e
a
k
y
_
r
e
_
l
u
_
2
(
Le
a
k
y
R
e
LU
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
0
C
o
n
v
4
c
o
n
v
2
d
_
3
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
3
6
9
2
8
l
e
a
k
y
_
r
e
_
l
u
_
3
(
Le
a
k
y
R
e
LU
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
0
C
o
n
v
5
c
o
n
v
2
d
_
4
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
3
6
9
2
8
l
e
a
k
y
_
r
e
_
l
u
_
4
(
Le
a
k
y
R
e
LU
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
0
R
e
si
d
u
a
l
a
d
d
_
1
(
A
d
d
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
3
)
]
0
C
o
n
v
6
c
o
n
v
2
d
_
5
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
3
6
9
2
8
l
e
a
k
y
_
r
e
_
l
u
_
5
(
Le
a
k
y
R
e
LU
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
0
C
o
n
v
7
c
o
n
v
2
d
_
6
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
3
6
9
2
8
l
e
a
k
y
_
r
e
_
l
u
_
6
(
Le
a
k
y
R
e
LU
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
0
C
o
n
v
8
c
o
n
v
2
d
_
7
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
3
6
9
2
8
l
e
a
k
y
_
r
e
_
l
u
_
7
(
Le
a
k
y
R
e
LU
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
0
C
o
n
v
9
c
o
n
v
2
d
_
8
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
3
6
9
2
8
l
e
a
k
y
_
r
e
_
l
u
_
8
(
Le
a
k
y
R
e
LU
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
6
4
)
]
0
C
o
n
v
1
0
c
o
n
v
2
d
_
9
(
C
o
n
v
2
D
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
3
)
]
1
7
3
1
R
e
si
d
u
a
l
a
d
d
_
1
(
A
d
d
)
[
(
N
o
n
e
,
N
o
n
e
,
N
o
n
e
,
3
)
]
0
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
7
9
-
3794
3784
Af
ter
s
ev
er
al
co
n
v
o
lu
tio
n
lay
er
s
th
er
e
is
r
esid
u
al
lin
k
in
g
wh
ich
is
d
o
n
e
b
y
a
d
d
in
g
th
e
o
u
tp
u
t
o
f
th
e
f
ir
s
t
lay
er
to
t
h
e
o
u
tp
u
t
o
f
t
h
e
last
lay
er
b
ef
o
r
e
t
h
e
n
e
x
t
co
n
v
o
lu
tio
n
lay
er
.
T
h
is
ar
ch
itectu
r
al
d
esig
n
h
elp
s
in
s
tu
d
y
in
g
m
o
r
e
d
etailed
f
ea
tu
r
es
an
d
p
r
eser
v
in
g
th
e
o
r
i
g
in
al
in
f
o
r
m
atio
n
o
f
th
e
in
p
u
t.
T
h
e
f
in
al
o
u
tp
u
t
g
o
es
th
r
o
u
g
h
t
h
e
last
co
n
v
o
lu
tio
n
la
y
er
an
d
th
e
s
ig
m
o
id
ac
tiv
atio
n
f
u
n
ctio
n
to
p
r
o
d
u
ce
th
e
f
in
al
o
u
tp
u
t w
h
ich
is
th
e
en
h
an
ce
d
im
a
g
e
p
r
e
d
ictio
n
.
T
h
e
to
tal
n
u
m
b
er
o
f
p
ar
am
eter
s
in
th
e
m
o
d
el
is
2
6
3
,
9
7
1
wh
ic
h
also
in
clu
d
es
th
e
weig
h
ts
an
d
b
iases
in
ea
ch
co
n
v
o
lu
tio
n
lay
er
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
is
g
en
er
ated
o
n
h
ar
d
war
e
eq
u
ip
p
e
d
with
T
en
s
o
r
Flo
w
2
.
7
an
d
NVI
DI
A
R
T
X
2
0
8
0
ti
GPU.
All
d
ata
h
as
an
in
itial
in
p
u
t
s
ize
o
f
1
2
8
×
1
2
8
f
o
r
th
e
tr
ain
in
g
an
d
v
alid
ati
o
n
p
r
o
ce
s
s
.
T
h
e
m
o
d
el
was
tr
ain
ed
u
s
in
g
Ad
am
'
s
lo
s
s
f
u
n
ctio
n
an
d
o
p
tim
izer
with
a
lear
n
in
g
r
ate
o
f
0
.
0
0
0
1
.
Mo
d
el
en
h
an
ce
d
d
ee
p
r
esid
u
al
C
NNs
ar
e
b
ased
o
n
m
u
ltip
le
r
esid
u
al
b
lo
ck
s
u
s
in
g
a
b
atch
n
o
r
m
aliza
tio
n
lay
e
r
f
o
r
co
n
s
i
s
ten
t
r
esu
lts
.
Ou
r
im
p
lem
en
tatio
n
in
clu
d
es
o
n
l
y
1
6
r
em
ain
i
n
g
b
lo
c
k
s
with
6
4
ch
an
n
els.
2
.
4
.
E
v
a
lua
t
i
o
n o
f
i
m
a
g
e
qu
a
lity
m
e
t
rics
T
h
e
ev
alu
atio
n
s
tep
is
u
s
ed
to
co
m
p
a
r
e
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
with
th
e
im
ag
e
q
u
ality
im
p
r
o
v
em
e
n
t
m
eth
o
d
.
T
h
e
p
ea
k
s
ig
n
al
-
to
-
n
o
is
e
r
ati
o
(
PS
NR
)
is
u
s
ed
to
m
ea
s
u
r
e
th
e
q
u
ality
o
f
t
h
e
r
ec
o
n
s
tr
u
cted
im
a
g
e.
Ma
th
em
atica
lly
[
2
1
]
,
PS
NR
is
ex
p
r
ess
ed
b
y
(
2
)
,
=
10
.
10
(
m
ax
(
)
2
)
(
2
)
wh
er
e
r
ep
r
esen
ts
ea
ch
i
m
ag
e
an
d
MSE
is
th
e
m
ea
n
s
q
u
a
r
e
er
r
o
r
.
T
h
e
lo
wer
t
h
e
MSE
,
th
e
h
ig
h
e
r
th
e
PS
NR
.
SS
I
M
is
an
im
ag
e
p
r
o
ce
s
s
in
g
tech
n
i
q
u
e
th
at
m
ea
s
u
r
es
th
e
s
tr
u
ctu
r
al
s
im
ilar
ity
b
etwe
e
n
th
e
ac
tu
al
im
a
g
e
an
d
th
e
r
ec
o
n
s
tr
u
cted
im
ag
e
,
s
u
ch
as
im
ag
e
co
m
p
r
ess
io
n
o
r
im
ag
e
u
p
-
s
am
p
lin
g
/d
o
wn
-
s
am
p
lin
g
[
1
7
]
.
Ma
th
em
atica
lly
,
SS
I
M
ca
n
b
e
r
ep
r
esen
ted
b
y
(
3
)
,
(
,
̅
)
=
(
2
̅
+
1
)
(
2
̅
+
2
)
(
2
+
̅
2
+
1
)
(
2
+
̅
2
+
2
)
(
3
)
T
h
e
eq
u
atio
n
(
4
)
r
e
p
r
esen
ts
th
e
p
er
p
ix
el
co
m
p
ar
is
o
n
b
etwe
en
th
e
two
im
a
g
es
f
o
r
SS
I
M.
MSE
h
as
a
d
u
ty
as
im
ag
e
r
ec
o
v
er
y
o
r
im
a
g
e
en
h
a
n
ce
m
en
t.
MSE
m
ea
s
u
r
es
th
e
av
er
ag
e
o
f
th
e
s
q
u
a
r
ed
d
if
f
er
en
ce
b
etwe
en
th
e
o
r
ig
in
al
p
ix
el
v
alu
e
an
d
t
h
e
p
r
ed
icted
p
ix
el
v
alu
e
in
th
e
e
n
h
an
ce
d
im
ag
e.
=
1
∑
(
−
̂
)
2
=
1
(
4
)
T
h
e
ca
lcu
latio
n
p
r
o
ce
s
s
in
v
o
l
v
es
ca
lcu
latin
g
th
e
s
q
u
ar
ed
d
if
f
er
en
ce
b
etwe
en
t
h
e
o
r
ig
i
n
al
p
ix
els
an
d
th
e
p
r
e
d
icted
p
i
x
els
wh
er
e
is
t
h
e
d
ata
p
o
i
n
t
n
u
m
b
e
r
ar
e
th
e
o
r
ig
in
al
p
ix
els
o
f
th
e
im
ag
e
an
d
̂
is
th
e
p
ix
el
p
r
ed
ictio
n
f
r
o
m
th
e
m
o
d
el.
T
h
e
C
I
I
ev
alu
ates
th
e
co
m
p
etitiv
en
ess
o
f
v
ar
io
u
s
co
n
tr
ast
en
h
an
ce
m
en
t
tech
n
iq
u
es,
th
e
m
o
s
t
well
-
k
n
o
wn
b
e
n
ch
m
ar
k
b
ein
g
th
e
im
ag
e
en
h
an
ce
m
en
t
m
ea
s
u
r
e.
I
t
is
u
s
ed
to
c
o
m
p
ar
e
th
e
r
esu
lts
o
f
co
n
tr
ast
en
h
an
ce
m
en
t
m
eth
o
d
s
.
C
o
n
tr
ast
en
h
an
ce
m
en
t
ca
n
b
e
m
ea
s
u
r
ed
u
s
in
g
C
I
I
as
a
r
atio
.
C
I
I
is
d
ef
in
ed
in
(
5
)
as:
=
(
5
)
wh
er
e
C
is
th
e
av
er
ag
e
v
alu
e
o
f
th
e
lo
ca
l c
o
n
tr
ast m
ea
s
u
r
ed
with
a
3
×3
win
d
o
w
as:
−
+
(
6
)
d
an
is
th
e
av
er
ag
e
v
alu
e
o
f
th
e
lo
ca
l
co
n
tr
ast
in
th
e
o
u
tp
u
t
an
d
o
r
ig
in
al
im
ag
e
s
,
r
esp
ec
tiv
ely
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
M
o
del e
nh
a
ncem
ent
wit
h inf
a
nt
hea
rt
o
bje
ct
I
n
th
is
s
ec
tio
n
,
ex
p
er
im
e
n
ts
ar
e
co
n
d
u
cted
to
cr
ea
te
an
i
m
ag
e
en
h
an
ce
m
e
n
t
m
o
d
el
th
at
y
ield
s
s
atis
f
ac
to
r
y
p
er
f
o
r
m
a
n
ce
in
ter
m
s
o
f
PS
NR
,
SS
I
M,
MSE
,
a
n
d
C
I
I
.
T
o
b
u
ild
s
u
ch
a
m
o
d
e
l,
o
n
ly
in
f
an
t
h
ea
r
t
o
b
jects a
r
e
u
s
ed
.
C
u
r
atin
g
a
n
i
n
f
an
t h
ea
r
t
d
ataset
f
o
r
co
n
s
t
r
u
ctin
g
h
ig
h
-
f
id
elity
DL
m
o
d
els tailo
r
ed
f
o
r
clin
ical
d
ep
lo
y
m
e
n
t
n
ec
ess
itates
im
a
g
es
o
f
s
u
f
f
icien
tly
h
ig
h
q
u
al
ity
.
T
h
is
e
n
ab
les
th
e
alg
o
r
ith
m
to
d
is
ce
r
n
b
o
th
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
n
h
a
n
cin
g
u
ltr
a
s
o
u
n
d
ima
g
e
q
u
a
lity u
s
in
g
d
ee
p
s
tr
u
ctu
r
e
o
f
r
esid
u
a
l n
etw
o
r
k
(
A
d
e
I
r
ia
n
i
S
a
p
itr
i
)
3785
co
m
m
o
n
an
d
d
is
tin
ct
p
atter
n
s
am
o
n
g
im
a
g
es
an
d
f
ac
ilit
ates
p
r
ec
is
e
m
ea
s
u
r
em
en
ts
.
Fiv
e
D
L
m
o
d
el
s
tr
u
ctu
r
es
h
av
e
b
ee
n
d
ev
el
o
p
ed
,
an
d
we
will
a
s
s
e
s
s
th
e
b
e
s
t
p
er
f
o
r
m
a
n
ce
o
f
th
e
s
elec
ted
m
o
d
el
in
T
ab
le
3
.
T
h
e
en
tire
im
ag
e
was
tr
ain
ed
b
y
u
tili
zin
g
s
m
all
2
5
6
×
2
5
6
p
atch
es
ex
tr
ac
ted
f
r
o
m
in
f
an
t
h
ea
r
t
i
m
ag
es.
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
s
p
an
n
ed
5
0
ep
o
ch
s
,
em
p
lo
y
in
g
a
b
a
tc
h
s
ize
o
f
4
.
N
o
tab
ly
,
th
e
E
DR
-
C
NNs
ar
ch
itectu
r
e
was
im
p
lem
en
ted
,
i
n
co
r
p
o
r
atin
g
th
e
C
h
ar
b
o
n
n
ier
lo
s
s
as
th
e
lo
s
s
f
u
n
ctio
n
.
Fo
r
o
p
tim
izatio
n
,
th
e
Ad
am
Op
tim
izer
was u
tili
ze
d
,
with
a
lear
n
in
g
r
ate
s
et
at
0
.
0
0
0
1
.
Tab
le 3
.
Th
e
fi
v
e
m
o
d
e
ls
g
e
n
e
ra
t
e
d
to
se
lec
t
t
h
e
b
e
st stru
c
t
u
re
C
N
N
s
R
e
sN
e
t
M
o
d
e
l
1
M
o
d
e
l
2
ED
R
-
C
N
N
s
1
6
l
a
y
e
r
5
0
l
a
y
e
r
1
2
l
a
y
e
r
1
0
l
a
y
e
r
1
0
l
a
y
e
r
I
n
p
u
t
I
mag
e
I
n
p
u
t
I
mag
e
I
n
p
u
t
I
mag
e
I
n
p
u
t
I
mag
e
I
n
p
u
t
I
mag
e
C
o
n
v
1
C
o
n
v
1
C
o
n
v
1
C
o
n
v
1
C
o
n
v
1
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
C
o
n
v
2
.
x
C
o
n
v
2
C
o
n
v
2
Le
a
k
y
R
e
LU
R
e
LU
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
C
o
n
v
3
C
o
n
v
3
C
o
n
v
2
C
o
n
v
n
R
e
LU
C
o
n
v
4
C
o
n
v
4
C
o
n
v
3
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
R
e
LU
C
o
n
v
5
Le
a
k
y
R
e
LU
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
C
o
n
v
3
.
x
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
C
o
n
v
4
R
e
LU
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
C
o
n
v
5
C
o
n
v
6
Le
a
k
y
R
e
LU
C
o
n
v
n
R
e
LU
C
o
n
v
6
C
o
n
v
7
C
o
n
v
5
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
C
o
n
v
7
C
o
n
v
8
Le
a
k
y
R
e
LU
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
C
o
n
v
4
.
x
C
o
n
v
8
C
o
n
v
9
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
R
e
LU
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
R
e
LU
C
o
n
v
1
0
C
o
n
v
6
C
o
n
v
n
R
e
LU
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
Le
a
k
y
R
e
LU
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
C
o
n
v
9
A
c
t
i
v
a
t
i
o
n
S
i
g
m
o
i
d
C
o
n
v
7
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
C
o
n
v
5
.
x
C
o
n
v
1
0
Lo
ss
c
h
a
r
b
o
n
n
i
e
r
Le
a
k
y
R
e
LU
R
e
LU
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
C
o
n
v
1
1
O
p
t
i
mi
z
e
r
A
d
a
m
C
o
n
v
8
C
o
n
v
n
R
e
LU
C
o
n
v
1
2
-
Le
a
k
y
R
e
LU
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
R
e
LU
-
C
o
n
v
9
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
A
c
t
i
v
a
t
i
o
n
S
o
f
t
ma
x
A
d
d
i
t
i
o
n
(
R
e
s
i
d
u
a
l
)
-
Le
a
k
y
R
e
LU
R
e
LU
Lo
ss
M
S
E
A
c
t
i
v
a
t
i
o
n
R
e
LU
-
C
o
n
v
1
0
A
c
t
i
v
a
t
i
o
n
R
e
LU
O
p
t
i
mi
z
e
r
A
d
a
m
Lo
ss
M
S
E
-
R
e
si
d
u
a
l
Lo
ss
M
S
E
-
O
p
t
i
mi
z
e
r
A
d
a
m
-
A
c
t
i
v
a
t
i
o
n
S
i
g
m
o
i
d
O
p
t
i
mi
z
e
r
A
d
a
m
-
-
-
Lo
ss
C
h
a
r
b
o
n
n
i
e
r
-
-
-
-
O
p
t
i
mi
z
e
r
A
d
a
m
T
h
r
o
u
g
h
th
e
a
n
aly
s
is
o
f
th
e
f
iv
e
m
o
d
el
s
tr
u
ct
u
r
es,
we
o
b
s
er
v
e
d
iv
er
s
e
p
er
f
o
r
m
a
n
ce
o
u
tco
m
es
in
PS
NR
,
SS
I
M,
MSE
,
an
d
C
I
I
v
alu
es.
As
f
o
u
n
d
in
th
e
r
esu
lt
,
th
e
cu
s
to
m
E
DR
-
C
NNs
s
tr
u
ctu
r
e
s
tan
d
s
o
u
t
b
y
s
u
r
p
ass
in
g
th
e
p
e
r
f
o
r
m
an
ce
o
f
th
e
o
th
er
m
o
d
els.
T
h
e
PS
NR
v
alu
es
ap
p
r
o
ac
h
4
0
d
B
,
a
n
d
it
d
e
m
o
n
s
tr
ates
co
m
p
ar
ab
le
p
er
f
o
r
m
an
ce
in
o
t
h
er
m
etr
ics
as
well
in
Fig
u
r
e
6
.
T
h
is
co
m
p
ellin
g
p
er
f
o
r
m
an
ce
s
u
g
g
ests
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
war
r
an
ts
f
u
r
t
h
er
ev
alu
atio
n
to
ascer
tain
its
r
esil
ien
ce
an
d
ab
ilit
y
to
g
e
n
er
alize
ac
r
o
s
s
v
ar
io
u
s
m
ed
ical
o
b
jects.
Fig
u
r
e
6
.
T
h
e
p
er
f
o
r
m
a
n
ce
v
al
id
atio
n
o
f
t
h
e
s
elec
ted
m
o
d
el
0
10
20
30
40
50
O
r
ig
in
a
l E
D
R
Mo
d
el
1
Mo
d
el
2
Mo
d
el
3
C
NNs
EDR
-
C
NNs
(
o
u
r
s
)
CII
M
S
E
S
S
IM
P
S
N
R
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
7
9
-
3794
3786
3
.
2
.
M
o
del
ev
a
lua
t
i
o
n
T
h
i
s
s
e
c
ti
o
n
p
e
r
t
ai
n
s
t
o
t
h
e
e
x
p
e
r
i
m
e
n
t
a
l
r
e
p
o
r
t
t
h
a
t
f
o
c
u
s
es
o
n
t
h
e
e
v
a
l
u
at
i
o
n
a
n
d
i
n
t
e
r
p
r
e
ta
t
i
o
n
o
f
o
u
r
m
o
d
e
l
'
s
p
e
r
f
o
r
m
a
n
c
e
.
W
e
b
e
n
c
h
m
a
r
k
a
t
h
o
r
o
u
g
h
a
n
a
l
y
s
i
s
o
f
t
h
e
E
D
R
-
C
N
Ns
o
b
t
a
i
n
e
d
r
e
s
u
l
t
s
a
g
a
i
n
s
t
v
a
r
i
o
u
s
m
e
t
h
o
d
s
,
i
n
c
l
u
d
i
n
g
c
o
n
t
r
a
s
t
lim
i
t
e
d
a
d
a
p
t
i
v
e
h
i
s
t
o
g
r
a
m
e
q
u
a
l
i
z
a
ti
o
n
(
C
L
A
HE
)
[
2
1
]
,
b
i
l
a
t
e
r
a
l
f
i
l
t
e
r
[
2
2
]
,
[
2
3
]
,
M
i
r
N
e
t
,
l
o
w
-
l
i
g
h
t
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
(
L
L
C
N
N
)
[
2
4
]
,
a
n
d
A
u
t
o
e
n
c
o
d
e
r
[
2
5
]
a
s
s
ee
n
i
n
T
a
b
le
4
.
F
o
u
r
p
e
r
f
o
r
m
a
n
c
e
v
a
l
u
e
s
w
e
r
e
u
s
ed
t
o
i
n
d
i
c
at
e
t
h
e
b
e
s
t
i
m
a
g
e
e
n
h
a
n
c
e
m
e
n
t
a
r
c
h
it
e
c
t
u
r
e
.
Hi
g
h
e
r
P
SNR
v
a
l
u
es
i
n
d
i
c
a
t
e
b
et
t
e
r
i
m
a
g
e
q
u
a
li
t
y
,
S
S
I
M
v
a
l
u
e
s
r
a
n
g
e
f
r
o
m
-
1
t
o
1
,
w
h
e
r
e
1
i
n
d
ic
a
t
es
i
d
e
n
t
i
c
al
im
a
g
e
s
,
l
o
w
e
r
MS
E
v
a
l
u
e
s
i
n
d
i
c
at
e
b
e
t
t
e
r
i
m
a
g
e
q
u
a
l
i
t
y
a
n
d
h
i
g
h
e
r
C
I
I
v
a
l
u
e
s
i
n
d
i
c
a
t
e
b
e
t
t
e
r
c
o
l
o
r
i
m
a
g
e
q
u
a
l
it
y
.
W
e
c
a
n
o
b
s
e
r
v
e
t
h
a
t
t
h
e
E
DR
-
C
NN
s
p
r
o
d
u
c
e
o
u
t
s
t
a
n
d
i
n
g
r
es
u
l
ts
,
w
it
h
3
8
.
3
5
P
S
NR
,
0
.
9
2
SS
I
M
,
9
.
6
6
M
S
E
,
a
n
d
1
.
9
7
C
I
I
.
A
l
l
t
h
e
t
e
s
t
v
al
u
e
s
d
e
m
o
n
s
t
r
a
t
e
s
a
ti
s
f
ac
t
o
r
y
r
e
s
u
l
ts
i
n
c
o
m
p
a
r
is
o
n
t
o
t
h
e
o
t
h
e
r
f
i
v
e
m
e
t
h
o
d
s
.
T
ab
le
4
.
Mo
d
el
im
ag
e
e
n
h
an
c
em
en
t p
er
f
o
r
m
an
ce
with
C
L
AHE
,
B
ilat
er
al
f
ilter
,
Mir
Net,
L
L
C
NN,
Au
to
en
co
d
e
r
,
an
d
E
DSR
-
C
N
Ns
M
e
t
r
i
c
s
C
LA
H
E
B
i
l
a
t
e
r
a
l
F
i
l
t
e
r
M
i
r
N
e
t
LLC
N
N
A
u
t
o
e
n
c
o
d
e
r
ED
R
-
C
N
N
s
P
S
N
R
3
1
.
9
2
3
7
.
7
6
3
5
.
8
1
3
7
.
7
8
3
6
.
1
0
3
8
.
3
5
S
S
I
M
0
.
5
5
0
.
9
4
0
.
7
6
0
.
9
1
0
.
8
9
8
0
.
9
2
M
S
E
4
2
.
0
1
1
1
.
0
6
2
4
.
5
8
1
1
.
0
1
1
6
.
1
4
9
.
6
6
C
I
I
1
.
5
8
0
.
9
9
0
.
9
8
1
.
3
6
0
.
9
4
1
.
9
7
T
h
e
g
r
ap
h
s
f
o
r
lo
s
s
an
d
PS
NR
d
u
r
in
g
th
e
tr
ain
in
g
an
d
v
alid
atio
n
p
r
o
ce
s
s
,
in
d
icate
a
co
n
s
is
ten
t
im
p
r
o
v
em
e
n
t
in
th
e
q
u
ality
o
f
m
ed
ical
im
ag
es
th
r
o
u
g
h
o
u
t
th
e
en
h
an
ce
m
e
n
t
p
r
o
ce
s
s
.
Fig
u
r
e
7
(
a)
to
7
(
d
)
v
is
u
alize
s
th
e
r
esu
lts
o
f
th
e
m
o
d
el
co
m
p
ar
is
o
n
u
s
in
g
Mir
Net,
L
L
C
NN,
Au
to
en
co
d
er
,
a
n
d
E
DR
-
C
NNs.
T
h
e
wh
o
le
g
r
a
p
h
ical
r
e
p
r
esen
tatio
n
h
ig
h
lig
h
ts
a
n
o
tab
le
t
r
en
d
in
th
e
PS
NR
v
alu
es,
d
e
m
o
n
s
tr
atin
g
an
u
p
war
d
tr
ajec
to
r
y
s
ig
n
if
y
in
g
en
h
an
ce
d
im
ag
e
q
u
ality
.
I
t
ca
n
b
e
o
b
s
er
v
ed
th
at
th
e
PS
NR
v
alu
e
v
ar
ies
with
ea
ch
iter
atio
n
,
b
u
t
it
s
h
o
u
ld
b
e
n
o
ted
th
at
th
e
PS
NR
v
alu
e
g
i
v
en
to
th
e
E
DR
-
C
NNs
m
eth
o
d
d
o
es
n
o
t
s
h
o
w
s
ig
n
if
ican
t
f
lu
ctu
atio
n
s
in
Fig
u
r
e
7
(
d
)
.
As
th
e
m
o
d
el
im
p
r
o
v
es,
th
e
PS
N
R
v
alu
es
s
h
o
u
ld
in
cr
ea
s
e,
in
d
icatin
g
b
etter
im
ag
e
f
id
elity
,
as
well
a
s
lo
s
s
.
T
h
e
lo
s
s
cu
r
v
e
d
is
p
lay
s
a
g
r
ad
u
al
d
ec
lin
e,
ev
en
t
u
ally
co
n
v
er
g
in
g
to
ze
r
o
with
a
s
m
o
o
th
p
r
o
g
r
ess
io
n
.
A
lo
wer
lo
s
s
in
E
DR
-
C
NNs
r
e
s
p
o
n
s
e
in
d
icate
s
th
at
th
e
m
o
d
el's
p
r
ed
ictio
n
s
ar
e
g
ettin
g
clo
s
er
to
th
e
ac
tu
al
v
alu
es.
I
t
ca
n
b
e
o
b
s
er
v
ed
th
at
th
e
E
DR
-
C
N
Ns m
o
d
el
ef
f
ec
tiv
ely
lear
n
s
an
d
ad
ap
ts
its
p
ar
am
eter
s
,
lead
in
g
to
a
d
e
s
ir
ab
le
d
ec
r
ea
s
e
in
th
e
lo
s
s
.
Fig
u
r
e
8
(
a)
to
8
(
g
)
,
s
h
o
wca
s
ed
th
e
h
eig
h
ten
e
d
im
ag
e
q
u
ality
ac
h
iev
ed
th
r
o
u
g
h
th
e
im
p
lem
e
n
tatio
n
o
f
s
ev
er
al
im
ag
e
e
n
h
an
ce
m
en
t
m
o
d
els.
I
t
ca
n
b
e
s
ee
n
f
r
o
m
th
e
f
ig
u
r
e
th
at
all
lo
w
-
illu
m
in
atio
n
im
ag
e
en
h
an
ce
m
e
n
t
m
eth
o
d
s
b
ased
o
n
d
ee
p
lea
r
n
in
g
ef
f
ec
tiv
ely
i
m
p
r
o
v
e
th
e
ac
cu
r
ac
y
o
f
a
n
in
f
an
t'
s
h
ea
r
t.
B
u
t
th
e
v
is
u
al
ef
f
ec
t
is
d
if
f
er
e
n
t.
T
h
es
e
tech
n
iq
u
es
s
h
o
wca
s
e
th
e
a
d
v
an
ce
m
en
ts
in
u
s
in
g
d
ee
p
lear
n
in
g
to
im
p
r
o
v
e
th
e
q
u
ality
o
f
im
a
g
es
b
y
en
h
a
n
c
in
g
d
etails,
r
ed
u
cin
g
n
o
is
e,
a
n
d
m
itig
atin
g
o
t
h
er
im
ag
e
-
r
e
lated
is
s
u
es.
E
ac
h
ap
p
r
o
ac
h
h
as
its
u
n
i
q
u
e
c
h
a
r
ac
ter
is
tics
an
d
ap
p
licatio
n
s
with
in
th
e
r
ea
lm
o
f
im
ag
e
en
h
an
ce
m
e
n
t
an
d
r
esto
r
atio
n
.
Nev
e
r
th
el
ess
,
wh
e
n
it
co
m
es
to
PS
NR
,
SS
I
M,
a
n
d
MSE
p
er
f
o
r
m
an
ce
,
E
DR
-
C
NNs
s
u
r
p
ass
o
th
er
ap
p
r
o
ac
h
es.
I
n
th
is
s
ec
tio
n
,
we
u
n
d
e
r
to
o
k
a
co
m
p
r
e
h
en
s
iv
e
b
en
c
h
m
ar
k
i
n
g
an
aly
s
is
u
s
in
g
th
e
latest
ad
v
an
ce
m
en
ts
in
th
e
f
ield
.
T
h
is
an
aly
s
is
in
v
o
lv
es
a
m
eticu
lo
u
s
ev
alu
atio
n
an
d
co
m
p
ar
is
o
n
o
f
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
ag
ain
s
t
well
-
estab
lis
h
ed
d
ee
p
-
lear
n
in
g
ap
p
r
o
ac
h
es
with
in
th
e
d
o
m
ain
o
f
en
h
a
n
ce
m
en
t.
T
o
f
ac
ilit
ate
th
is
as
s
ess
m
en
t,
we
em
p
lo
y
ed
f
o
u
r
d
is
tin
ct
d
at
asets
:
th
e
in
f
an
t
h
ea
r
t
d
ataset
f
o
r
th
e
d
ev
el
o
p
m
en
t
o
f
o
u
r
e
n
h
an
ce
m
en
t
m
o
d
el,
an
d
th
r
ee
ad
d
itio
n
al
d
atasets
—
f
etal
h
ea
r
t,
f
etal
h
ea
d
,
an
d
ab
d
o
m
en
—
to
th
o
r
o
u
g
h
ly
e
v
alu
ate
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
.
T
o
e
n
s
u
r
e
th
e
f
air
n
ess
o
f
th
e
test
,
all
th
e
m
eth
o
d
s
ar
e
test
ed
u
n
d
e
r
th
e
s
am
e
h
ar
d
war
e
en
v
ir
o
n
m
en
t
a
f
ter
r
ea
ch
in
g
t
h
e
o
p
tim
al
lev
el
o
f
tr
ain
i
n
g
.
T
h
e
ex
p
er
im
en
tal
o
u
tc
o
m
es
y
ield
ed
n
o
tewo
r
th
y
in
s
ig
h
ts
,
s
p
ec
if
ically
in
ter
m
s
o
f
PS
NR
,
S
SIM
,
an
d
MSE
m
etr
ics
ac
r
o
s
s
all
f
o
u
r
d
atasets
i
n
d
etail
f
o
r
v
ar
io
u
s
tech
n
iq
u
es
r
e
f
er
t
o
Fig
u
r
es
9
(
a
)
to
9
(
c)
.
I
t
o
b
s
er
v
es
t
h
at
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
E
DR
-
C
NNs
co
n
s
is
ten
tly
d
eliv
er
s
s
atis
f
ac
to
r
y
r
esu
lts
ac
r
o
s
s
th
e
en
tire
d
ataset,
ac
h
iev
i
n
g
h
ig
h
er
PS
NR
,
lo
wer
SS
I
M,
an
d
d
ec
r
ea
s
ed
MSE
v
alu
es.
T
h
is
in
d
icate
s
th
at
o
u
r
s
u
g
g
ested
m
o
d
el
c
o
n
s
is
ten
tly
p
r
o
d
u
ce
s
f
av
o
r
ab
le
p
er
f
o
r
m
a
n
ce
.
B
y
ev
alu
atin
g
th
e
en
h
an
ce
d
im
ag
es u
s
in
g
d
if
f
er
en
t te
ch
n
i
q
u
es a
n
d
d
atasets
,
it c
an
b
e
s
tated
th
at
th
e
r
esu
lts
ar
e
p
r
o
m
is
in
g
T
o
en
s
u
r
e
th
e
p
e
r
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
ed
im
ag
e
en
h
an
ce
m
en
t
tech
n
i
q
u
e
u
s
in
g
E
DR
-
C
N
Ns,
s
u
ch
a
m
o
d
el
h
as
b
ee
n
im
p
lem
en
te
d
f
o
r
s
eg
m
e
n
tin
g
an
d
d
etec
tin
g
th
e
1
0
o
b
jects
o
f
t
h
e
f
e
tal
h
ea
r
t
s
u
b
s
tr
u
ctu
r
e,
in
clu
d
in
g
th
e
lef
t
atr
iu
m
(
L
A
)
,
r
ig
h
t
atr
iu
m
(
R
A)
,
lef
t
v
en
t
r
icle
(
L
V)
,
r
ig
h
t
v
en
tr
icle
(
R
V)
,
tr
icu
s
p
id
v
alv
e
(
T
V)
,
p
u
lm
o
n
ar
y
v
alv
e
(
PV)
,
m
itra
l
v
alv
e
(
MV
)
,
in
ter
v
e
n
tr
i
cu
lar
s
ep
tu
m
(
I
VS)
,
ao
r
ta
(
Ao
)
an
d
h
o
le
(
d
ef
ec
t)
.
T
h
e
ex
p
er
im
en
t
s
in
v
o
lv
e
d
th
e
u
tili
za
tio
n
o
f
th
e
y
o
u
o
n
ly
l
o
o
k
at
o
n
ce
with
m
ask
in
g
(
YO
L
AC
T
)
f
r
am
ewo
r
k
.
YOL
AC
T
i
s
an
alg
o
r
ith
m
d
esig
n
ed
f
o
r
r
ea
l
-
tim
e
o
b
ject
d
etec
tio
n
th
at
co
m
b
in
es
o
b
ject
d
etec
tio
n
with
in
s
tan
ce
s
eg
m
en
tatio
n
.
I
ts
g
o
a
l
is
to
co
n
cu
r
r
en
tly
i
d
en
tify
a
n
d
s
eg
m
en
t
o
b
jects
p
r
esen
t
in
a
n
im
ag
e.
Dete
ctin
g
th
e
f
etal
h
ea
r
t
s
u
b
-
s
tr
u
ctu
r
e
o
b
ject
p
r
o
v
es
to
b
e
h
ig
h
ly
ch
al
len
g
in
g
d
u
e
to
its
s
m
all
s
ize,
th
e
p
o
o
r
q
u
ality
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
n
h
a
n
cin
g
u
ltr
a
s
o
u
n
d
ima
g
e
q
u
a
lity u
s
in
g
d
ee
p
s
tr
u
ctu
r
e
o
f
r
esid
u
a
l n
etw
o
r
k
(
A
d
e
I
r
ia
n
i
S
a
p
itr
i
)
3787
im
ag
es,
an
d
its
s
h
ap
e
th
at
d
y
n
am
ically
ch
an
g
es
d
u
e
to
f
etal
m
o
v
em
en
ts
[
5
]
.
W
h
ile
it
o
f
ten
p
er
f
o
r
m
s
well
d
u
r
in
g
test
in
g
u
s
in
g
v
alid
ati
o
n
d
ata,
its
ef
f
ec
tiv
en
ess
ten
d
s
to
d
ec
lin
e
o
r
ev
en
r
esu
lt
in
f
ailu
r
e
wh
en
f
ac
ed
with
u
n
s
ee
n
d
ata
[
6
]
,
[
7
]
.
T
h
e
im
ag
e
s
am
p
le
as v
is
u
alize
d
in
Fig
u
r
e
10
.
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
7
.
Gr
a
p
h
ic
tr
ain
in
g
an
d
v
alid
atio
n
o
f
lo
s
s
an
d
PS
NR
o
v
er
ep
o
ch
s
f
o
r
in
f
a
n
t h
ea
r
t
o
b
ject
in
clu
d
in
g
(
a)
Mir
Net
,
(
b
)
L
L
C
NN
,
(
c)
A
u
to
en
co
d
er
,
an
d
(
d
)
E
DR
-
C
NNs st
r
u
ctu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
7
9
-
3794
3788
(
a)
(
b
)
(
c)
(
d
)
(
e)
(f)
(
g
)
Fig
u
r
e
8
.
T
h
e
s
am
p
le
o
f
in
f
an
t
h
ea
r
t im
ag
e
e
n
h
an
ce
m
en
t w
ith
s
ix
o
f
th
e
ex
is
tin
g
m
o
d
el
(
a)
tar
g
et,
(
b
)
C
L
AHE
,
(
c)
b
ilater
al
f
ilter
,
(
d
)
Mir
Net,
(
e)
L
L
C
NN,
(
f
)
a
u
to
en
co
d
er
,
an
d
(
g
)
E
DSR
-
C
NNs
Evaluation Warning : The document was created with Spire.PDF for Python.