I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
,
p
p
.
1
661
~
1
6
7
2
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
8
.
i
3
.
pp
1
6
6
1
-
1
6
7
2
1661
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
Weierstras
s
sca
le
spa
ce r
eprese
ntat
io
n and
com
po
sit
e dilated
U
-
net
bas
ed conv
o
lution for ea
rly
g
la
uco
ma
diag
no
s
is
Abdu
l
B
a
s
it
h
Z
a
hir H
us
s
a
in
1
,
Su
lt
ha
n Ib
ra
h
im
M
o
ha
m
ed
Su
la
im
a
n
2
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
A
p
p
l
i
c
a
t
i
o
n
,
M
a
d
u
r
a
i
K
a
mara
j
U
n
i
v
e
r
si
t
y
(
M
K
U
)
,
M
a
d
u
r
a
i
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
G
o
v
e
r
n
m
e
n
t
A
r
t
s
a
n
d
S
c
i
e
n
c
e
C
o
l
l
e
g
e
,
Th
e
n
i
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
May
14
,
2
0
2
4
R
ev
is
ed
No
v
12
,
2
0
2
4
Acc
ep
ted
No
v
24
,
2
0
2
4
G
lau
c
o
m
a
is
o
n
e
o
f
th
e
c
o
m
m
o
n
c
a
u
se
s
o
f
b
li
n
d
n
e
ss
in
th
e
c
u
rre
n
t
wo
rl
d
.
G
lau
c
o
m
a
is
a
b
li
n
d
in
g
o
p
ti
c
n
e
u
ro
p
a
th
y
c
h
a
ra
c
teriz
e
d
b
y
t
h
e
d
e
g
e
n
e
ra
ti
o
n
o
f
re
ti
n
a
l
g
a
n
g
li
o
n
c
e
ll
s
(RG
Cs
).
Ac
c
u
ra
te
d
iag
n
o
sis
a
n
d
m
o
n
i
to
rin
g
o
f
g
lau
c
o
m
a
a
re
c
h
a
ll
e
n
g
in
g
tas
k
th
ro
u
g
h
e
y
e
e
x
a
m
in
a
ti
o
n
s
a
n
d
a
d
d
it
i
o
n
a
l
tes
ts.
To
a
c
h
iev
e
a
c
c
u
ra
te
d
ia
g
n
o
sis
o
f
g
la
u
c
o
m
a
wit
h
h
ig
h
e
r
se
n
si
ti
v
it
y
a
n
d
sp
e
c
ifi
c
it
y
,
n
o
v
e
l
m
e
th
o
d
c
a
ll
e
d
Weie
rstra
ss
sc
a
l
e
sp
a
c
e
re
p
re
se
n
t
a
ti
o
n
a
n
d
c
o
m
p
o
site
d
il
a
ted
U
-
n
e
t
b
a
se
d
c
o
n
v
o
l
u
ti
o
n
(W
S
S
R
-
CDC)
is
in
tr
o
d
u
c
e
d
.
At
first,
th
e
Weie
rstra
ss
tran
sfo
rm
sc
a
le
sp
a
c
e
re
p
re
se
n
tatio
n
is
e
m
p
lo
y
e
d
to
e
n
h
a
n
c
e
ima
g
e
stru
c
tu
re
s
a
t
v
a
rio
u
s
sc
a
les
with
h
ig
h
e
r
a
c
c
u
ra
c
y
o
f
re
g
io
n
o
f
in
tere
st
(ROI)
d
e
tec
ti
o
n
u
sin
g
Eu
ler’s
i
d
e
n
t
it
y
.
Ne
x
t
,
CDC
m
o
d
e
l
is
u
ti
li
z
e
d
with
se
v
e
ra
l
lay
e
rs.
I
n
i
n
p
u
t
lay
e
r,
p
re
p
r
o
c
e
ss
e
d
in
p
u
t
ima
g
e
s
a
r
e
tak
e
n
a
s
in
p
u
t.
F
ra
g
m
e
n
t
d
e
riv
a
ti
v
e
a
re
f
o
rm
u
late
d
fo
r
e
v
e
ry
p
re
p
ro
c
e
ss
e
d
i
n
p
u
t.
L
o
g
c
o
sh
d
ice
lo
ss
fu
n
c
ti
o
n
a
n
d
o
p
ti
c
c
u
p
t
o
d
isc
ra
ti
o
a
re
c
o
m
p
u
ted
f
o
r
se
g
m
e
n
ted
g
la
u
c
o
m
a
d
e
tec
te
d
re
su
lt
s.
W
it
h
th
is,
t
h
e
a
c
c
u
ra
te
d
ia
g
n
o
sis
o
f
g
lau
c
o
m
a
is
m
a
d
e
with
m
in
im
a
l
e
rro
r.
Th
e
WS
S
R
-
CDC
m
e
th
o
d
wa
s
e
v
a
lu
a
ted
u
sin
g
t
h
e
g
lau
c
o
m
a
f
u
n
d
u
s
ima
g
i
n
g
d
a
tas
e
t
wit
h
se
v
e
r
a
l
fa
c
to
rs.
Th
e
re
su
lt
s
sh
o
w
th
a
t
th
e
WS
S
R
-
CDC
m
e
th
o
d
o
u
t
p
e
rfo
rm
s
c
o
n
v
e
n
t
io
n
a
l
tec
h
n
i
q
u
e
s,
imp
ro
v
in
g
a
c
c
u
ra
c
y
b
y
2
4
%
a
n
d
se
n
siti
v
it
y
b
y
1
8
%
.
It
d
e
m
o
n
stra
tes
p
r
o
m
isin
g
re
su
lt
s
in
fa
st,
a
c
c
u
ra
te,
d
ia
g
n
o
sis o
f
g
lau
c
o
m
a
.
K
ey
w
o
r
d
s
:
C
o
m
p
o
s
ite
d
ilated
U
-
n
et
b
ased
co
n
v
o
l
u
tio
n
Dee
p
lear
n
in
g
Glau
co
m
a
d
etec
tio
n
Scale
s
p
ac
e
r
ep
r
esen
tatio
n
W
eier
s
tr
as
s
tr
an
s
f
o
r
m
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
:
Ab
d
u
l Bas
ith
Z
ah
ir
Hu
s
s
ain
D
ep
ar
tm
en
t o
f
C
o
m
p
u
ter
A
p
p
licatio
n
,
Ma
d
u
r
ai
Kam
ar
aj
Un
iv
er
s
ity
(
MK
U)
Ma
d
u
r
ai,
6
2
5
0
2
1
T
a
m
il
Nad
u
,
I
n
d
ia
E
m
ail:
ab
d
b
asit
h
9
3
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
On
e
o
f
th
e
p
ar
am
o
u
n
t
s
o
u
r
ce
s
o
f
b
lin
d
n
ess
cr
ea
tin
g
o
p
tic
n
e
r
v
e
d
am
a
g
e
is
g
lau
co
m
a
.
I
f
u
n
d
iag
n
o
s
e
d
tim
ely
,
g
lau
co
m
a
b
r
i
n
g
s
ab
o
u
t
ir
r
ev
er
s
ib
le
d
estru
ctio
n
to
th
e
o
p
tic
n
er
v
e
r
esu
ltin
g
in
b
lin
d
n
ess
.
Usu
ally
,
o
p
h
th
alm
o
lo
g
is
ts
co
n
d
u
ct
d
i
s
ea
s
e
d
iag
n
o
s
is
th
r
o
u
g
h
r
eti
n
al
s
cr
u
tin
y
o
f
b
r
o
ad
en
e
d
p
u
p
ils
b
y
m
ea
n
s
o
f
s
eg
m
en
tatio
n
.
Ma
ch
in
e
lear
n
i
n
g
tech
n
iq
u
es,
h
ith
er
t
o
n
o
w
b
ein
g
s
teer
ed
an
d
v
ig
o
r
o
u
s
n
ec
ess
itate
s
au
to
m
ated
s
o
lu
tio
n
s
.
T
h
e
o
p
tic
n
er
v
e
h
ea
d
in
s
p
ec
tio
n
th
at
n
ec
ess
itat
es
q
u
an
tific
atio
n
o
f
cu
p
-
to
-
d
i
s
c
r
atio
(
C
D
R
)
i
s
co
n
tem
p
lated
as
th
e
m
o
s
t
p
er
tin
en
t
tech
n
iq
u
e
o
f
g
lau
co
m
a
d
is
ea
s
e
d
iag
n
o
s
is
.
Owin
g
to
th
is
,
d
ee
p
lear
n
in
g
(
DL
)
tech
n
iq
u
es
as
a
co
n
s
eq
u
en
ce
o
f
s
elf
-
lear
n
in
g
ca
n
in
s
titu
te
au
to
m
ated
d
iag
n
o
s
tic
co
u
r
s
es
o
f
ac
tio
n
s
in
m
in
im
al
tim
e.
C
o
n
tr
ast
lim
ited
ad
ap
tiv
e
h
is
to
g
r
am
eq
u
aliz
atio
n
(
C
L
AHE
)
was
u
tili
ze
d
as
a
p
r
ep
r
o
ce
s
s
in
g
s
tep
in
m
u
lti
-
f
ea
tu
r
e
d
ee
p
lear
n
in
g
(
MFDL
)
was
d
ev
elo
p
ed
b
y
Xu
e
et
a
l.
[
1
]
to
ca
r
r
y
o
n
c
o
m
p
ar
is
o
n
im
ag
es
co
m
p
atib
le
as
well
as
r
eg
u
lar
i
ze
clar
if
icatio
n
ac
r
o
s
s
im
ag
es
in
th
e
d
atab
ase
th
er
e
f
o
r
e
im
p
r
o
v
in
g
tr
u
e
p
o
s
itiv
e
r
ate
o
f
g
lau
c
o
m
a
d
etec
tio
n
.
Ho
wev
er
,
im
ag
e
s
tr
u
ctu
r
es
ar
e
d
if
f
er
e
n
t
s
ca
les
was
n
o
t
f
o
cu
s
ed
,
th
er
e
f
o
r
e
co
m
p
r
o
m
i
s
in
g
th
e
ac
cu
r
ac
y
o
f
r
eg
io
n
o
f
in
ter
est
(
R
OI
)
d
etec
tio
n
.
A
n
ew
m
u
lti
-
task
DL
m
eth
o
d
wh
ich
in
f
lu
en
ce
s
co
r
r
elatio
n
s
o
f
ass
o
ciate
d
ey
e
-
f
u
n
d
u
s
task
s
an
d
q
u
an
tific
atio
n
s
u
tili
ze
d
in
g
la
u
co
m
a
an
aly
s
is
was
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
6
6
1
-
1
6
7
2
1662
p
r
esen
ted
b
y
Pas
ca
l
et
a
l.
[
2
]
.
W
ith
th
is
th
e
p
r
o
b
ab
ilit
y
th
at
t
h
e
m
eth
o
d
r
an
k
e
d
a
r
a
n
d
o
m
p
o
s
itiv
e
s
am
p
le
was
m
o
r
e
th
an
a
r
a
n
d
o
m
n
eg
ativ
e
s
am
p
le,
th
er
ef
o
r
e
im
p
r
o
v
e
t
h
e
ar
ea
u
n
d
er
t
h
e
r
ec
eiv
er
o
p
er
atin
g
ch
ar
ac
ter
is
tic
(
R
OC
)
cu
r
v
e
s
ig
n
if
ican
tly
.
Ho
wev
er
,
with
th
e
s
p
ar
s
e
p
o
ten
tiality
in
s
eg
m
en
tin
g
e
d
g
e
f
ea
t
u
r
es
an
d
in
ad
e
q
u
ate
ex
tr
ac
tio
n
o
f
p
o
s
itio
n
in
f
o
r
m
at
io
n
,
th
er
e
f
o
r
e
ca
u
s
in
g
im
ag
e
d
iag
n
o
s
tic
er
r
o
r
s
.
T
h
e
ab
o
v
e
co
n
v
en
tio
n
al
m
et
h
o
d
s
ar
e
illu
s
tr
ated
in
m
ajo
r
p
r
o
b
lem
s
in
clu
d
in
g
m
in
im
u
m
ac
cu
r
ac
y
,
h
ig
h
er
er
r
o
r
,
a
n
d
f
ailu
r
e
to
p
r
o
v
id
e
ac
cu
r
ate
g
lau
c
o
m
a
d
e
tectio
n
an
d
s
en
s
itiv
ity
an
d
s
p
ec
if
icity
wer
e
n
o
t
co
n
s
id
er
ed
.
Fo
r
e
m
en
tio
n
e
d
is
s
u
es
an
d
r
esu
lts
f
ac
to
r
,
in
th
i
s
wo
r
k
a
m
eth
o
d
ca
lled
,
W
eier
s
tr
ass
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
an
d
c
o
m
p
o
s
ite
d
ilated
U
-
Net
b
ased
co
n
v
o
l
u
tio
n
(
W
SS
R
-
C
DC
)
f
o
r
g
lau
co
m
a
d
etec
tio
n
is
in
tr
o
d
u
ce
d
.
T
o
im
p
r
o
v
e
s
en
s
itiv
ity
as
well
as
s
p
ec
if
icity
o
f
g
lau
c
o
m
a
r
ec
o
g
n
itio
n
,
to
ai
d
in
d
if
f
e
r
en
tiatin
g
tr
u
e
p
ath
o
lo
g
y
f
r
o
m
n
o
r
m
al
v
ar
iab
ilit
y
,
th
e
W
SS
R
-
C
DC
t
ec
h
n
iq
u
e
is
d
esig
n
ed
.
W
SS
R
-
C
DC
tech
n
iq
u
e
is
d
ev
elo
p
e
d
th
r
o
u
g
h
p
r
e
p
r
o
ce
s
s
in
g
an
d
s
eg
m
en
tatio
n
o
n
c
o
n
tr
ar
y
to
co
n
v
en
ti
o
n
al
wo
r
k
wh
ich
em
p
lo
y
s
n
o
r
m
alizin
g
t
h
e
illu
m
in
atio
n
a
cr
o
s
s
th
e
im
ag
es.
T
h
e
r
est
o
f
t
h
e
p
ap
e
r
is
o
r
d
er
ed
as
f
o
llo
ws.
Sectio
n
2
p
o
r
tr
ay
s
th
e
r
elate
d
wo
r
k
s
.
I
n
s
ec
tio
n
3
,
t
h
e
m
eth
o
d
o
l
o
g
y
o
f
r
esear
c
h
is
d
etailed
.
T
h
e
ex
p
er
im
en
tal
s
ettin
g
s
ar
e
p
r
o
v
id
ed
an
d
im
p
lem
en
tatio
n
d
etails
ar
e
p
r
esen
ted
in
s
ec
tio
n
4
.
I
n
s
ec
tio
n
4
,
th
e
r
esu
lt
an
aly
s
is
is
d
is
cu
s
s
ed
.
Fin
ally
,
s
ec
tio
n
5
d
escr
ib
e
s
th
e
co
n
clu
s
io
n
o
f
th
e
p
ap
er
.
2.
L
I
T
E
R
AT
U
RE
SU
RVE
Y
A
DL
alg
o
r
ith
m
em
p
lo
y
in
g
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
b
ased
g
lau
co
m
a
d
etec
tio
n
to
f
o
c
u
s
o
n
th
e
d
ia
g
n
o
s
tic
er
r
o
r
was
p
r
esen
ted
b
y
Kim
et
a
l.
[
3
]
.
Ho
wev
er
,
t
h
e
m
e
m
o
r
y
co
n
s
u
m
ed
in
s
to
r
in
g
th
e
in
ter
m
ed
iate
class
if
ied
r
esu
lts
was
f
o
u
n
d
to
b
e
h
ig
h
e
r
.
T
o
f
o
cu
s
o
n
t
h
is
is
s
u
e,
a
C
NN
-
b
ased
f
u
lly
au
to
m
ated
m
ec
h
an
is
m
f
o
r
g
la
u
co
m
a
d
et
ec
tio
n
was
p
r
esen
ted
b
y
Sah
a
et
a
l.
[
4
]
.
Desp
ite
im
p
r
o
v
em
en
t
in
ter
m
s
o
f
g
lau
co
m
a
d
etec
tio
n
,
th
e
tim
e
co
n
s
u
m
p
tio
n
was
n
o
t
f
o
cu
s
e
d
.
Ho
wev
er
,
ea
r
ly
g
lau
c
o
m
a
d
etec
tio
n
ca
n
ev
en
s
to
p
th
e
v
is
io
n
l
o
s
s
.
W
ith
th
is
o
b
jectiv
e
two
p
h
ases
f
o
cu
s
in
g
lo
ca
lizatio
n
v
ia
o
p
tic
d
is
c
an
d
ac
co
r
d
in
g
l
y
d
iag
n
o
s
in
g
g
lau
co
m
a
v
ia
n
etwo
r
k
m
o
d
el
was
p
r
esen
t
ed
b
y
L
atif
et
a
l.
[
5
]
.
T
h
r
o
u
g
h
th
is
k
in
d
o
f
lo
ca
lizatio
n
m
eth
o
d
en
s
u
r
e
d
ac
cu
r
ac
y
b
u
t
also
m
in
im
ized
th
e
co
m
p
u
tatio
n
co
s
t
in
an
ex
ten
s
iv
e
m
an
n
er
.
Dee
p
n
eu
r
al
n
etwo
r
k
h
as
b
r
o
u
g
h
t
a
b
o
u
t
e
n
co
u
r
a
g
in
g
r
esu
lts
f
o
r
d
etec
ti
o
n
o
f
g
lau
c
o
m
a
in
a
n
au
to
m
atic
m
an
n
er
f
u
n
d
u
s
im
ag
es.
Ho
wev
er
,
th
e
in
h
er
en
t
in
co
n
s
is
ten
cy
ac
r
o
s
s
g
lau
co
m
a
d
atasets
is
d
em
an
d
in
g
f
o
r
d
ata
-
d
r
i
v
en
n
eu
r
al
n
etwo
r
k
m
ec
h
an
is
m
s
.
T
h
is
in
co
n
s
is
ten
cy
r
esu
lts
in
t
h
e
d
o
m
ain
g
a
p
th
at
in
f
lu
e
n
ce
s
m
o
d
el
p
er
f
o
r
m
an
ce
an
d
d
ec
r
ea
s
es
m
o
d
el
g
en
e
r
aliza
tio
n
p
o
te
n
tiality
.
Yan
et
a
l.
[
6
]
,
a
m
ix
u
p
d
o
m
ain
ad
ap
tati
o
n
m
ec
h
a
n
is
m
was
d
esig
n
ed
th
at
tr
av
er
s
es’
d
o
m
ain
ad
ap
tatio
n
with
d
o
m
ai
n
m
ix
u
p
with
th
e
p
u
r
p
o
s
e
o
f
e
n
h
an
cin
g
th
e
o
v
er
al
l
m
o
d
el
p
er
f
o
r
m
an
ce
ac
r
o
s
s
d
if
f
er
en
t
g
lau
co
m
a
d
atasets
.
Ho
wev
er
,
an
o
th
e
r
tech
n
iq
u
e
to
f
o
cu
s
o
n
s
e
n
s
itiv
ity
as
well
as
s
p
ec
if
icity
em
p
lo
y
i
n
g
lo
g
is
tic
r
eg
r
ess
io
n
-
b
ased
m
o
d
el
was
d
esig
n
ed
b
y
T
h
an
k
i
[
7
]
f
o
r
ef
f
icien
t
r
etin
al
f
u
n
d
u
s
class
if
icatio
n
.
C
o
n
v
en
tio
n
al
d
iag
n
o
s
tic
m
eth
o
d
s
ar
e
f
o
u
n
d
to
b
e
lab
o
r
io
u
s
an
d
tim
e
co
n
s
u
m
i
n
g
a
n
d
f
r
eq
u
e
n
tly
in
ac
cu
r
ate,
h
en
ce
m
ak
in
g
g
lau
co
m
a
d
iag
n
o
s
is
in
an
ac
cu
r
ate
m
an
n
er
.
T
o
b
r
id
g
e
t
h
is
g
ap
an
au
to
m
ated
g
lau
co
m
a
s
tag
e
c
lass
if
icatio
n
m
eth
o
d
e
m
p
lo
y
i
n
g
p
r
e
-
tr
ain
e
d
d
ee
p
C
NN
m
o
d
el
an
d
class
if
ier
f
u
s
io
n
.
W
ith
th
is
m
o
d
el
n
o
t
o
n
ly
ac
cu
r
ac
y
was
en
s
u
r
ed
b
u
t
also
r
esu
lted
in
ea
r
ly
r
ec
o
g
n
it
io
n
.
R
ev
iew
o
f
DL
m
eth
o
d
s
was
ex
am
in
ed
b
y
Velp
u
la
an
d
Sh
ar
m
a
[
8
]
f
o
r
ea
r
ly
g
lau
co
m
a
d
etec
tio
n
.
Glau
co
m
a
d
etec
tio
n
m
an
u
ally
is
d
em
an
d
in
g
p
ar
t
wh
ich
n
ee
d
s
p
r
o
f
icien
c
y
as
well
as
y
ea
r
s
o
f
ex
p
er
ien
ce
.
Ajith
a
et
a
l.
[
9
]
,
a
d
o
m
in
an
t
a
n
d
p
r
ec
is
e
alg
o
r
i
th
m
em
p
lo
y
i
n
g
a
C
NN
f
o
r
a
u
t
o
m
atic
d
iag
n
o
s
is
o
f
g
lau
co
m
a
was
p
r
o
p
o
s
ed
.
B
y
em
p
lo
y
in
g
th
is
Dl
tech
n
iq
u
e
im
p
r
o
v
e
d
th
e
s
en
s
itiv
ity
an
d
s
p
ec
if
icity
in
an
ex
ten
s
iv
e
m
an
n
er
.
Per
f
o
r
m
an
c
e
ass
es
s
m
en
t
o
f
s
ev
er
al
DL
tec
h
n
iq
u
es
in
p
r
e
d
ictin
g
g
lau
co
m
a
v
ia
th
r
ee
d
is
tin
ct
o
p
tim
iza
tio
n
alg
o
r
ith
m
s
was
p
r
o
p
o
s
ed
b
y
Sin
g
h
et
a
l.
[
1
0
]
.
Ho
wev
er
,
p
r
ev
ailin
g
m
eth
o
d
s
ch
ief
ly
d
ep
en
d
o
n
a
s
u
b
s
tan
tial
am
o
u
n
t
o
f
lab
eled
d
at
a
th
at
is
a
d
em
an
d
in
g
c
o
n
s
tr
ain
t
f
o
r
g
lau
co
m
a
d
etec
tio
n
.
T
o
a
d
d
r
ess
o
n
th
is
asp
ec
t,
tr
an
s
f
er
i
n
d
u
ce
d
atten
tio
n
n
etwo
r
k
(TIA
-
Net)
was
p
r
esen
ted
b
y
Xu
et
a
l.
[
1
1
]
f
o
r
au
t
o
m
at
ic
g
lau
co
m
a
d
etec
tio
n
.
Peo
p
le
ag
o
n
izin
g
as
o
f
g
lau
co
m
a
f
r
e
q
u
en
tly
n
o
t
o
b
s
er
v
e
s
o
m
e
m
o
d
if
y
in
v
is
io
n
at
p
r
em
atu
r
e
p
h
ases
.
Nev
er
th
eless
,
with
it
s
p
r
o
g
r
ess
io
n
,
g
lau
co
m
a
s
p
ec
if
ically
r
esu
lts
in
v
is
io
n
lo
s
s
th
at
is
also
f
o
u
n
d
to
b
e
ir
r
ev
e
r
s
ib
le
in
s
ev
er
al
ca
s
es.
As
a
r
esu
lt,
ea
r
ly
d
iag
n
o
s
is
is
o
f
cr
itical
im
p
o
r
tan
ce
.
Als
o
o
b
tain
in
g
ac
cu
r
ate
in
s
ig
h
ts
ar
e
also
f
o
u
n
d
to
b
e
a
tim
e
-
co
n
s
u
m
in
g
p
r
o
ce
s
s
.
D’
So
u
za
et
a
l.
[
1
2
]
,
p
a
r
am
eter
-
e
f
f
ec
tiv
e
Alter
Net
-
K
m
eth
o
d
tak
in
g
in
to
co
n
s
id
er
atio
n
alter
n
atin
g
d
esi
g
n
p
atter
n
in
teg
r
atin
g
r
esid
u
a
l
n
etwo
r
k
s
(
R
esNets
)
as
well
as
m
u
lti
-
h
ea
d
s
elf
-
atten
tio
n
(
MSA
)
was
p
r
o
p
o
s
e
d
.
B
y
em
p
lo
y
in
g
th
is
in
teg
r
ati
o
n
m
o
d
el
r
esu
lted
in
th
e
o
v
er
all
im
p
r
o
v
em
e
n
t
in
g
en
e
r
aliza
tio
n
.
Desp
ite
im
p
r
o
v
em
en
t
in
g
en
er
aliza
tio
n
ac
cu
r
ac
y
was
n
o
t
f
o
cu
s
ed
.
An
en
s
em
b
le
o
f
s
elec
tio
n
m
eth
o
d
s
was
p
r
o
p
o
s
ed
b
y
Path
an
et
a
l
.
[
1
3
]
u
s
in
g
d
ir
ec
tio
n
al
f
ilter
a
n
d
d
y
n
a
m
ic
s
elec
tio
n
tech
n
iq
u
es.
I
t
en
h
an
ce
d
in
g
e
n
er
al
s
en
s
itiv
ity
in
an
ex
ten
s
iv
e
m
e
th
o
d
.
Pre
ce
d
in
g
r
esear
ch
wo
r
k
s
h
av
e
s
h
o
wn
th
at
o
win
g
to
m
is
s
ed
d
iag
n
o
s
is
th
e
lik
elih
o
o
d
o
f
p
r
o
g
r
ess
io
n
f
r
o
m
o
cu
lar
h
y
p
er
ten
s
io
n
t
o
u
n
ilater
al
v
is
io
n
lo
s
s
is
in
cr
ea
s
in
g
g
r
ad
u
ally
.
D
u
e
to
th
is
,
ea
r
ly
g
lau
co
m
a
d
iag
n
o
s
is
is
es
s
en
tial
to
war
d
o
f
f
d
is
ea
s
e
p
r
o
g
r
e
s
s
io
n
an
d
v
is
io
n
lo
s
s
.
A
co
m
b
in
ed
C
NN
an
d
r
ec
u
r
r
e
n
t
n
eu
r
al
n
etwo
r
k
(
R
NN
)
wer
e
d
esig
n
ed
b
y
Gh
eisar
i
et
a
l.
[
1
4
]
th
at
th
r
o
u
g
h
aid
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
Weiers
t
r
a
s
s
s
ca
le
s
p
a
ce
r
ep
r
esen
ta
tio
n
a
n
d
co
mp
o
s
ite
d
ila
ted
U
-
n
et
…
(
A
b
d
u
l B
a
s
ith
Za
h
i
r
Hu
s
s
a
in
)
1663
o
f
b
o
th
s
p
atial
a
n
d
tem
p
o
r
al
f
ea
tu
r
es
en
h
a
n
ce
d
ea
r
ly
g
lau
co
m
a
d
etec
tio
n
.
Yet
an
o
th
er
au
t
o
m
ated
m
ec
h
an
is
m
em
p
lo
y
in
g
C
NN
was
p
r
o
p
o
s
ed
b
y
Sch
u
s
ter
et
a
l.
[
1
5
]
with
im
p
r
o
v
e
d
s
en
s
itiv
ity
an
d
s
p
ec
if
icity
.
Ho
wev
er
d
iag
n
o
s
tic
er
r
o
r
was
n
o
t
co
n
s
id
er
ed
.
Hu
m
an
v
is
io
n
h
as
m
o
tiv
ated
n
o
tab
le
d
ev
elo
p
m
en
ts
in
co
m
p
u
ter
v
is
io
n
h
o
wev
er
,
th
e
h
u
m
an
is
s
aid
to
b
e
h
ig
h
ly
s
u
s
ce
p
tib
le
to
s
ev
er
al
s
ilen
t
ey
e
d
is
ea
s
e
s
.
W
ith
t
h
e
ev
o
lu
tio
n
o
f
DL
tech
n
iq
u
es,
co
m
p
u
ter
v
is
io
n
f
o
r
h
u
m
a
n
ey
e
d
is
ea
s
e
d
etec
tio
n
h
as r
ec
eiv
ed
im
p
o
r
tan
ce
b
u
t m
o
s
t r
esear
ch
wo
r
k
h
as
co
n
ce
n
tr
ated
o
n
a
co
n
s
tr
ain
ed
n
u
m
b
e
r
o
f
ey
e
d
is
ea
s
es.
A
two
-
p
h
ase
lo
ca
lizatio
n
v
ia
ODGN
et
wa
s
d
esig
n
ed
L
atif
et
a
l.
[
1
6
]
.
Als
o
,
by
em
p
lo
y
in
g
th
e
s
alien
cy
m
ap
r
esu
lted
in
th
e
m
in
im
izat
io
n
o
f
co
m
p
u
tatio
n
co
s
t sig
n
if
ican
tly
.
C
DR
was
m
ea
s
u
r
ed
b
y
Sev
asto
p
o
ls
k
y
et
a
l.
[
1
7
]
em
p
lo
y
i
n
g
U
-
n
et
C
NN
th
at
in
tu
r
n
r
ed
u
ce
d
th
e
p
r
ed
ictio
n
tim
e
c
o
n
s
id
er
ab
ly
.
Yet
an
o
th
er
m
eth
o
d
to
b
o
o
s
t
th
e
lear
n
in
g
e
f
f
icien
cy
em
p
l
o
y
in
g
D
-
S
ev
id
e
n
ce
th
eo
r
y
was
p
r
esen
ted
b
y
Du
e
t
a
l.
[
1
8
]
t
h
at
in
t
u
r
n
i
m
p
r
o
v
e
d
r
ec
o
g
n
itio
n
ca
p
ab
ilit
y
.
A
m
u
lti
-
f
ea
tu
r
e
a
n
aly
s
is
was
p
er
f
o
r
m
ed
b
y
Ak
ter
et
a
l.
[
1
9
]
u
s
in
g
lo
g
is
tic
r
eg
r
ess
io
n
to
f
o
c
u
s
o
n
th
e
ac
cu
r
ac
y
asp
e
ct.
Glau
co
m
a
is
th
e
p
r
in
cip
al
a
lead
in
g
in
d
u
ce
m
en
t
o
f
ir
r
e
v
er
s
ib
le
b
lin
d
n
ess
g
lo
b
ally
,
in
f
lu
en
cin
g
m
illi
o
n
s
o
f
p
eo
p
le.
E
ar
ly
d
iag
n
o
s
is
is
cr
u
cial
to
m
i
n
im
ize
v
is
u
al
lo
s
s
an
d
n
u
m
er
o
u
s
m
eth
o
d
s
ar
e
u
tili
ze
d
f
o
r
d
ete
ctio
n
o
f
g
la
u
co
m
a.
Pu
ch
aice
la
-
L
o
za
n
o
et
a
l.
[
2
0
]
,
h
y
b
r
id
tech
n
iq
u
e
f
o
r
g
lau
c
o
m
a
f
u
n
d
u
s
im
ag
e
lo
ca
lizatio
n
em
p
lo
y
in
g
p
r
e
-
tr
ain
ed
R
-
C
NN
as
well
as
s
eg
m
en
tatio
n
em
p
lo
y
in
g
C
2
D
ar
ea
was
p
r
esen
ted
.
B
y
e
m
p
lo
y
in
g
th
e
c
u
p
-
to
-
d
is
k
ar
ea
f
o
r
s
eg
m
en
tatio
n
r
esu
lted
in
an
im
p
r
o
v
em
en
t
o
f
ac
c
u
r
ac
y
.
Yet
a
n
o
th
er
t
r
an
s
f
o
r
m
ativ
e
ap
p
r
o
ac
h
t
o
gl
au
co
m
a
d
etec
tio
n
em
p
lo
y
i
n
g
C
NN
was
in
v
esti
g
ated
b
y
Haja
an
d
Ma
h
ad
e
v
ap
p
a
[
2
1
]
.
As
g
lau
co
m
a
m
ater
ializes
in
later
s
tag
es
an
d
it
is
a
s
lo
w
d
i
s
ea
s
e,
d
etailed
s
cr
ee
n
in
g
an
d
d
etec
tio
n
is
ess
en
tial
to
k
ee
p
awa
y
f
r
o
m
v
is
io
n
f
o
r
f
eitu
r
e.
Ma
h
u
m
et
a
l.
[
2
2
]
,
f
o
r
d
etec
tin
g
g
lau
co
m
a
at
ea
r
ly
s
tag
es
u
s
in
g
DL
-
b
ased
f
ea
tu
r
e
ex
tr
ac
tio
n
was
p
r
esen
ted
.
C
h
allen
g
es
in
ar
tific
ial
in
tellig
e
n
c
e
f
o
r
g
la
u
co
m
a
d
etec
tio
n
wer
e
in
v
esti
g
ated
b
y
Hu
an
g
et
a
l.
[
2
3
]
.
Hu
m
an
’
s
g
r
ad
in
g
was
s
im
u
lated
with
DL
by
L
in
et
a
l.
[
2
4
]
e
m
p
lo
y
in
g
au
to
m
ated
d
iag
n
o
s
in
g
m
ec
h
an
is
m
.
W
ith
th
is
ty
p
e
o
f
s
im
u
latio
n
en
h
an
ce
d
clin
ical
g
lau
co
m
a
d
ia
g
n
o
s
is
.
Mu
ltimo
d
al
d
ataset
was e
m
p
lo
y
ed
b
y
L
i
et
a
l.
[
2
5
]
.
3.
M
E
T
H
O
D
Glau
co
m
a
is
a
d
is
ea
s
e
wh
ich
co
n
ce
r
n
o
p
tic
n
e
r
v
e
ca
u
s
ed
t
h
r
o
u
g
h
a
b
n
o
r
m
ally
h
ig
h
p
r
es
s
u
r
e
at
th
e
ey
e
an
d
is
also
co
n
s
id
er
ed
as
o
n
e
o
f
th
e
m
ajo
r
s
o
u
r
ce
s
o
f
b
lin
d
n
ess
f
o
r
p
eo
p
le
ir
r
esp
ec
tiv
e
o
f
th
e
ag
e,
m
o
r
e
f
r
eq
u
e
n
t
in
o
ld
er
ad
u
lts
.
Glau
co
m
a
in
cr
ea
s
es
C
DR
,
ex
er
tin
g
in
f
lu
en
ce
o
n
p
er
ip
h
er
al
v
is
io
n
lo
s
s
.
Acc
u
r
ate
an
d
p
r
ec
is
e
g
lau
co
m
a
d
etec
tio
n
i
n
d
ig
ital
f
u
n
d
u
s
im
ag
es
is
h
o
wev
er
an
o
p
en
to
p
ic
as
f
ar
as
b
io
m
ed
ical
im
ag
e
p
r
o
ce
s
s
in
g
is
co
n
ce
r
n
ed
.
H
en
ce
,
ea
r
ly
g
lau
co
m
a
d
etec
tio
n
in
r
etin
al
f
u
n
d
u
s
im
ag
e
is
cr
u
cial
f
o
r
cir
cu
m
v
en
tin
g
f
r
o
m
th
e
v
is
io
n
lo
s
s
.
I
n
th
is
wo
r
k
a
m
eth
o
d
ca
ll
ed
,
W
SS
R
-
C
D
C
is
d
esig
n
ed
.
As
illu
s
tr
ated
in
Fig
u
r
e
1
,
th
e
p
r
o
p
o
s
ed
W
SS
R
-
C
D
C
m
eth
o
d
.
T
h
e
in
p
u
t
im
a
g
es
ar
e
o
b
t
ain
ed
f
r
o
m
th
e
g
lau
c
o
m
a
f
u
n
d
u
s
im
a
g
e
d
a
taset.
T
h
e
s
am
p
le
im
ag
es
ar
e
t
h
en
s
u
b
jecte
d
to
p
r
ep
r
o
ce
s
s
in
g
an
d
s
eg
m
e
n
tatio
n
em
p
lo
y
in
g
W
SS
R
-
C
D
C
.
I
n
itially
,
p
r
e
p
r
o
ce
s
s
in
g
is
p
er
f
o
r
m
ed
b
y
ap
p
ly
in
g
W
SS
R
to
g
en
er
ate
s
ca
le
-
in
v
ar
ian
t
p
r
ep
r
o
ce
s
s
ed
im
ag
es
with
h
i
g
h
er
s
en
s
itiv
ity
.
Seco
n
d
,
th
e
p
r
e
p
r
o
ce
s
s
ed
s
am
p
le
im
ag
es
ar
e
s
u
b
jecte
d
t
o
a
s
eg
m
en
tatio
n
m
o
d
el
ca
lled
,
C
DC
with
v
ar
io
u
s
lay
e
r
s
s
u
ch
as
in
p
u
t,
h
id
d
e
n
,
a
n
d
o
u
t
p
u
t
l
ay
er
s
.
Pre
p
r
o
ce
s
s
ed
in
p
u
t
im
ag
es
ar
e
co
n
s
id
er
ed
a
s
in
p
u
t
in
th
e
in
p
u
t
lay
er
.
T
h
ese
lay
er
s
ar
e
s
en
t
to
th
e
h
id
d
en
lay
er
.
Fra
g
m
en
t
r
ec
tifie
d
lin
ea
r
u
n
it
(
FR
eL
U
)
ac
tiv
atio
n
an
d
s
ig
m
o
id
f
u
n
ctio
n
ar
e
em
p
lo
y
ed
in
th
e
h
id
d
e
n
lay
er
.
Nex
t,
in
th
e
o
u
tp
u
t
la
y
er
,
d
iag
n
o
s
tic
er
r
o
r
is
m
in
im
ized
v
ia
th
e
lo
g
c
o
s
h
d
ice
lo
s
s
f
u
n
ctio
n
.
T
h
e
o
p
tic
C
DR
is
d
eter
m
in
ed
f
o
r
s
eg
m
en
ted
g
lau
c
o
m
a
d
etec
ted
r
esu
lts
.
W
ith
th
is
c
o
n
v
o
lu
t
io
n
m
o
d
el
i
m
p
r
o
v
ed
R
OI
d
ete
ctio
n
is
en
s
u
r
ed
in
an
ac
cu
r
ate
an
d
p
r
ec
is
e
m
an
n
e
r
.
T
h
is
p
r
o
c
ess
o
f
th
e
W
SS
R
-
C
DC
i
s
ex
p
lain
ed
in
th
e
f
o
llo
win
g
s
u
b
s
ec
tio
n
s
.
Fig
u
r
e
1
.
B
lo
ck
d
iag
r
am
o
f
W
SS
R
-
C
DC
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
6
6
1
-
1
6
7
2
1664
3
.
1
.
Weier
s
t
ra
s
s
t
ra
ns
f
o
rm
s
ca
le
s
pa
ce
re
presenta
t
io
n
-
ba
s
ed
prepr
o
ce
s
s
ing
I
n
th
e
r
eg
i
o
n
o
f
im
ag
e
an
a
ly
s
is
as
wel
l
as
d
is
ea
s
e
d
iag
n
o
s
is
,
th
e
co
n
ce
p
tio
n
o
f
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
(
SSR
)
is
u
tili
ze
d
f
o
r
p
r
o
ce
s
s
in
g
im
ag
e
in
f
o
r
m
at
io
n
at
n
u
m
er
o
u
s
s
ca
les
an
d
t
o
b
e
m
o
r
e
s
p
ec
if
ic
en
h
an
ce
im
ag
e
asp
ec
ts
.
Sp
ec
i
al
k
in
d
o
f
SS
R
is
g
iv
en
th
r
o
u
g
h
W
eier
s
tr
ass
ap
p
r
o
x
im
atio
n
,
wh
er
ev
er
im
a
g
e
in
f
o
r
m
atio
n
is
s
u
b
jecte
d
to
c
o
n
v
o
lu
ti
o
n
th
r
o
u
g
h
g
au
s
s
ian
f
u
n
ctio
n
.
T
h
e
m
ajo
r
ity
o
f
th
e
o
r
y
f
o
r
W
eier
s
tr
ass
ap
p
r
o
x
im
atio
n
s
ca
le
s
p
ac
e
co
n
tr
ac
t
th
r
o
u
g
h
co
n
tin
u
o
u
s
i
m
ag
es,
co
n
s
id
er
in
g
th
at
s
in
g
l
e
as
ex
ec
u
tin
g
th
i
s
co
n
tain
to
f
ac
e
d
etail
wh
ic
h
m
ain
ly
m
ea
s
u
r
em
en
t
in
f
o
r
m
atio
n
is
d
is
cr
ete.
T
h
er
e
f
o
r
e,
th
is
W
eier
s
tr
ass
a
p
p
r
o
x
im
atio
n
s
ca
le
s
p
ac
e
co
n
tr
ac
t so
lv
es th
e
is
s
u
e
to
d
is
cr
et
ize
co
n
tin
u
o
u
s
im
ag
es a
s
p
r
eser
v
in
g
wh
ich
lead
s
to
th
e
s
elec
tio
n
o
f
W
eier
s
tr
ass
tr
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
-
b
ased
p
r
e
p
r
o
ce
s
s
in
g
m
o
d
el.
T
h
e
W
eier
s
tr
as
s
tr
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
n
o
t
o
n
ly
i
m
p
r
o
v
es
t
h
e
im
ag
e
s
tr
u
ctu
r
es
at
d
if
f
e
r
en
t
s
ca
les
b
u
t
a
ls
o
en
h
an
ce
s
th
e
ac
cu
r
a
cy
o
f
R
OI
d
etec
tio
n
co
n
s
id
e
r
ab
ly
v
ia
E
u
ler
’
s
i
d
en
tity
.
Fig
u
r
e
2
s
h
o
ws
th
e
s
tr
u
ctu
r
e
o
f
t
h
e
W
eier
s
tr
ass
tr
a
n
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
-
b
ased
p
r
ep
r
o
ce
s
s
in
g
m
o
d
e
l.
Fig
u
r
e
2
.
Stru
ctu
r
e
o
f
W
eier
s
tr
ass
tr
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
-
b
ased
p
r
e
p
r
o
ce
s
s
in
g
m
o
d
el
As
illu
s
tr
ated
in
th
e
F
ig
u
r
e
2
,
let
u
s
co
n
s
id
er
th
e
r
aw
im
a
g
es
o
b
tain
ed
f
r
o
m
th
e
g
lau
c
o
m
a
f
u
n
d
u
s
im
ag
in
g
d
atasets
.
T
h
e
in
p
u
t
im
ag
es
as
illu
s
tr
ated
in
Fig
u
r
e
2
ar
e
s
u
b
jecte
d
s
ep
ar
at
ely
to
d
is
cr
ete
an
d
co
n
tin
u
o
u
s
s
ca
le
s
p
ac
e
r
ep
r
es
en
tatio
n
s
.
Fin
ally
,
th
e
v
alu
es
ar
e
av
er
ag
e
d
u
s
in
g
W
eier
s
tr
ass
t
r
an
s
f
o
r
m
f
u
n
ctio
n
th
er
ef
o
r
e
f
o
r
m
in
g
s
ca
le
-
in
v
ar
i
an
t
p
r
e
p
r
o
ce
s
s
ed
im
a
g
e
s
tr
u
ct
u
r
es
as
o
u
tp
u
t.
L
et
u
s
co
n
s
id
er
a
g
au
s
s
ian
SSR
of
N
-
d
im
en
s
io
n
al
s
am
p
le
im
ag
e
r
ep
r
esen
ted
as
(
1
)
an
d
(
2
)
.
,
(
1
,
2
,
…
,
,
)
(
1
)
,
(
1
,
2
,
…
,
,
)
(
2
)
Fro
m
th
e
(
1
)
an
d
(
2
)
g
au
s
s
ian
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
o
f
a
s
am
p
le
im
ag
e
‘
’
is
o
b
tain
ed
b
y
co
n
v
o
l
v
in
g
b
o
th
th
e
c
o
n
tin
u
o
u
s
an
d
d
is
cr
ete
r
ep
r
esen
tatio
n
o
f
im
ag
es
‘
,
’
b
ased
o
n
t
h
e
s
ca
le
p
ar
am
eter
‘
’
.
As
th
e
(
1
)
an
d
(
2
)
th
o
u
g
h
f
o
r
m
u
lated
b
o
th
f
o
r
co
n
tin
u
o
u
s
an
d
d
is
cr
ete
f
o
r
m
s
o
f
im
ag
e
r
ep
r
esen
tatio
n
s
h
o
wev
er
in
p
r
ac
ticality
is
n
o
t
p
o
s
s
ib
le
to
ap
p
ly
s
im
ilar
s
ca
le
s
p
ac
e
f
o
r
b
o
t
h
ty
p
es
o
f
im
ag
e
r
ep
r
esen
tatio
n
s
.
Hen
ce
,
b
ased
o
n
th
e
s
ep
ar
ab
i
lity
ch
ar
ac
ter
is
tics
o
f
g
au
s
s
ian
SS
R
,
d
is
cr
ete
an
d
co
n
tin
u
o
u
s
r
ep
r
esen
tatio
n
o
f
im
ag
es o
f
im
ag
e
s
tr
u
ct
u
r
es a
t
d
if
f
er
en
t scale
s
is
f
o
r
m
u
lated
as
(
3
)
to
(
5
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
Weiers
t
r
a
s
s
s
ca
le
s
p
a
ce
r
ep
r
esen
ta
tio
n
a
n
d
co
mp
o
s
ite
d
ila
ted
U
-
n
et
…
(
A
b
d
u
l B
a
s
ith
Za
h
i
r
Hu
s
s
a
in
)
1665
(
,
)
=
∑
(
−
)
(
,
)
(
3
)
(
,
)
=
∑
1
√
2
=
1
(
4
)
=
c
os
(
)
+
s
in
(
)
(
5
)
Fro
m
th
e
(
3
)
an
d
(
4
)
,
u
s
in
g
E
u
ler
’
s
id
en
tity
o
b
tain
ed
f
r
o
m
(
5
)
tr
u
n
ca
tes
at
th
e
e
n
d
to
g
e
n
e
r
ate
a
f
ilter
ed
r
esu
lt
with
f
in
ite
im
p
u
ls
e
r
esp
o
n
s
e
t
h
er
ef
o
r
e
g
en
e
r
atin
g
co
n
tin
u
o
u
s
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
r
e
s
u
lts
.
(
,
)
=
∑
(
−
)
(
,
)
(
6
)
(
,
)
=
(
)
(
7
)
(
)
=
2
2
2
+
+
(
2
−
)
=
0
(
8
)
Fro
m
th
e
(
6
)
an
d
(
7
)
,
u
s
in
g
th
e
B
ess
el
f
u
n
ctio
n
o
b
tai
n
ed
f
r
o
m
(
8
)
f
o
r
d
is
cr
ete
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
.
W
eier
s
tr
ass
t
r
an
s
f
o
r
m
f
u
n
ctio
n
is
ap
p
lied
t
o
av
er
ag
e
th
e
v
alu
es
o
f
co
n
tin
u
o
u
s
s
ca
le
s
p
ac
e
an
d
d
is
cr
ete
SS
R
r
esu
lt
s
.
T
h
is
is
f
o
r
m
u
lated
as
(
9
)
.
=
(
)
=
1
√
4
∫
(
)
−
(
−
)
2
4
+
∞
−
∞
,
ℎ
=
(
,
)
.
(
,
)
(
9
)
Fro
m
th
e
r
esu
lts
(
9
)
im
ag
e
s
tr
u
ctu
r
es
at
d
i
f
f
er
en
t
s
ca
les
ar
e
en
h
an
ce
d
th
er
e
f
o
r
e
en
s
u
r
in
g
t
h
e
ac
cu
r
ac
y
o
f
R
OI
d
etec
tio
n
in
ex
te
n
s
iv
e
way
.
Alg
o
r
ith
m
1
d
escr
ib
es
th
e
s
tep
-
by
-
s
tep
p
r
o
ce
s
s
o
f
W
eier
s
tr
ass
tr
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
e
p
r
esen
tatio
n
-
b
ased
p
r
ep
r
o
ce
s
s
in
g
.
I
n
th
e
alg
o
r
ith
m
,
th
r
o
u
g
h
s
am
p
le
im
ag
e
o
b
tain
ed
f
r
o
m
t
h
e
r
a
w
g
lau
co
m
a
f
u
n
d
u
s
im
ag
in
g
d
ataset
g
au
s
s
ian
SS
R
o
f
‘
N
-
d
im
en
s
io
n
’
s
am
p
le
im
ag
e
is
in
itially
f
o
r
m
u
lated
.
Seco
n
d
,
c
o
n
tin
u
o
u
s
SS
R
an
d
d
is
cr
ete
SS
R
v
ia
E
u
ler
’
s
id
en
tity
a
n
d
W
eier
s
tr
ass
tr
an
s
f
o
r
m
f
u
n
ctio
n
s
ep
ar
atel
y
.
Fin
ally
,
b
o
th
th
e
s
ca
le
r
ep
r
esen
tatio
n
r
esu
lts
ar
e
co
m
b
i
n
ed
to
o
b
tain
p
r
e
p
r
o
c
ess
ed
r
esu
lts
th
at
in
tu
r
n
en
s
u
r
e
th
e
ac
cu
r
ac
y
o
f
R
OI
d
etec
tio
n
in
a
s
ig
n
if
ican
t
m
an
n
er
.
Alg
o
r
ith
m
1
.
W
eier
s
tr
ass
tr
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
-
b
ased
p
r
ep
r
o
c
ess
in
g
Input
: Dataset ‘
’, Sample Image ‘
=
{
1
,
2
,
…
,
}
’
Output
: scale space
-
efficient preprocessed results ‘
’
1:
Initialize
‘
’, scale parameter ‘
’
2:
Begin
3:
For
each Dataset ‘
’ with Sample Image ‘
’
4:
Fo
rm
ul
at
e
Gaussian
Sc
al
e
Sp
ac
e
Re
pr
es
en
ta
ti
on
o
f
an
N
-
dimensional
sample
image
as
given
in equations (1) and (2)
5:
Me
as
ur
e
co
nt
in
uo
us
sc
a
le
sp
ac
e
re
pr
es
en
ta
ti
on
r
es
ul
ts
as
gi
ve
n
in
eq
ua
ti
on
s
(3
),
(4
)
an
d
(5)
6:
Me
as
ur
e
di
sc
re
te
sc
al
e
sp
ac
e
re
pr
es
en
ta
ti
on
re
su
lt
s
as
gi
ve
n
in
eq
ua
t
io
ns
(6
),
(7
)
an
d
(8)
7:
Ge
ne
ra
te
pr
ep
ro
ce
ss
ed
re
su
lt
s
by
ap
pl
yi
ng
We
ie
rs
tr
as
s
Tr
an
sf
or
m
fu
nc
ti
on
as
gi
ve
n
in
equation (9)
8:
Return
preprocessed results ‘
’
9:
End for
10:
End
3
.
2
.
Co
m
po
s
it
e
d
ila
t
ed
U
-
net
co
nv
o
lutio
n
-
ba
s
ed
s
eg
m
ent
a
t
io
n
Seg
m
en
tatio
n
o
f
r
etin
al
b
lo
o
d
v
ess
els
is
r
eg
ar
d
ed
as
a
n
ef
f
icien
t
m
ec
h
an
is
m
f
o
r
d
iag
n
o
s
in
g
o
c
u
lar
d
is
ea
s
es
to
lar
g
e
ex
ten
t
g
lau
co
m
a
d
is
ea
s
e
d
etec
tio
n
.
Se
g
m
en
tatio
n
o
f
b
lo
o
d
v
ess
els
is
p
er
f
o
r
m
ed
b
y
em
p
lo
y
in
g
C
DC
m
o
d
el.
Her
e
s
eg
m
en
tatio
n
is
p
er
f
o
r
m
ed
f
o
r
th
e
p
r
ep
r
o
ce
s
s
ed
s
am
p
le
in
p
u
t
im
ag
e
to
i
d
en
tify
g
lau
co
m
a
b
y
th
e
C
DR
ev
alu
atio
n
.
W
ith
th
e
p
r
ep
r
o
ce
s
s
ed
im
ag
e
r
esu
lts
as
in
p
u
t,
au
to
m
atic
o
p
tic
d
is
c
as
well
as
cu
p
s
eg
m
en
tatio
n
d
ep
en
d
o
n
DL
s
u
ch
as
C
D
C
m
o
d
el
is
d
esig
n
ed
.
B
y
u
s
in
g
C
D
C
m
o
d
el
ed
g
e
f
ea
tu
r
es
ar
e
r
etain
ed
an
d
also
im
ag
e
d
iag
n
o
s
tic
er
r
o
r
is
im
p
r
o
v
ed
b
y
m
ea
n
s
o
f
lo
g
co
s
h
d
ice
l
o
s
s
f
u
n
ctio
n
.
T
h
e
b
lo
ck
d
iag
r
am
o
f
th
e
C
DC
-
b
ased
s
eg
m
e
n
tatio
n
m
o
d
el
is
d
escr
ib
ed
in
Fig
u
r
e
3
.
Fig
u
r
e
3
d
em
o
n
s
tr
ates
th
e
b
lo
ck
d
iag
r
am
o
f
th
e
C
DC
-
b
ased
s
eg
m
en
tatio
n
m
o
d
el.
T
h
is
m
o
d
el
in
clu
d
es
in
p
u
t
lay
e
r
,
h
id
d
en
lay
er
,
an
d
o
u
tp
u
t
lay
er
.
At
f
ir
s
t,
th
e
p
r
ep
r
o
ce
s
s
ed
s
am
p
le
in
p
u
t
im
ag
e
(
i.e
.
,
p
r
ep
r
o
ce
s
s
ed
r
esu
lt
‘
’
)
is
tak
en
th
r
o
u
g
h
th
e
co
n
v
o
lu
tio
n
al
lay
er
.
Mo
r
eo
v
er
,
FR
eL
U
ac
tiv
atio
n
is
u
tili
ze
d
in
h
id
d
en
lay
e
r
wh
er
ea
s
th
e
s
ig
m
o
id
f
u
n
ctio
n
is
u
s
ed
to
o
u
tp
u
t
lay
er
.
T
wo
p
r
o
ce
s
s
es
ar
e
ca
r
r
ied
o
u
t
s
u
ch
as
l
o
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
6
6
1
-
1
6
7
2
1666
co
s
h
d
ice
lo
s
s
f
u
n
ctio
n
a
n
d
o
p
tic
cu
p
to
d
is
c
r
atio
in
t
h
e
o
u
tp
u
t
lay
er
.
T
h
e
g
la
u
co
m
a
d
iag
n
o
s
is
is
m
ad
e
in
an
ac
cu
r
ate
m
an
n
e
r
.
Fig
u
r
e
3
.
B
lo
ck
d
iag
r
am
o
f
C
DC
-
b
ased
s
eg
m
en
tatio
n
m
o
d
e
l
FR
eL
U
b
ein
g
a
lin
ea
r
f
u
n
cti
o
n
with
t
h
e
in
p
u
t
b
ein
g
p
o
s
itiv
e,
th
e
o
u
tp
u
t
v
alu
e
is
th
e
s
am
e
as
th
e
in
p
u
t v
alu
e
.
On
co
n
tr
ar
y
,
th
e
o
u
tp
u
t o
f
ze
r
o
is
p
r
o
d
u
ce
d
an
d
i
s
m
ath
em
atica
lly
ex
p
r
ess
ed
as
(
1
0
)
.
1
(
)
=
ma
x
(
0
,
)
(
1
0
)
Fro
m
th
e
(
1
0
)
,
‘
’
f
o
r
m
s
th
e
in
p
u
t
o
r
th
e
p
r
ep
r
o
ce
s
s
ed
r
esu
lts
p
r
o
v
id
e
d
as
in
p
u
t.
T
h
e
n
,
th
e
ab
o
v
e
f
u
n
cti
o
n
s
atis
f
y
in
g
d
u
ality
co
n
d
itio
n
is
s
tated
as
(1
1
)
.
1
(
)
=
{
,
>
0
0
,
≤
0
(
1
1
)
T
h
is
ar
ch
itectu
r
e
i
n
clu
d
ed
tw
o
p
ath
s
,
n
am
ely
,
co
n
t
r
ac
tin
g
p
ath
a
n
d
ex
p
a
n
s
io
n
p
at
h
r
e
s
p
ec
tiv
ely
.
B
o
th
o
f
th
em
ap
p
lied
th
e
FR
eL
U
as
th
e
ac
tiv
atio
n
f
u
n
ctio
n
.
B
y
ap
p
ly
in
g
th
is
f
u
n
ctio
n
h
a
s
th
e
ad
v
an
tag
e
o
f
d
eter
m
in
in
g
th
e
s
eg
m
en
te
d
p
o
r
tio
n
s
in
an
ac
cu
r
ate
an
d
p
r
ec
i
s
e
m
an
n
er
f
o
r
g
lau
c
o
m
a
d
etec
tio
n
.
On
o
n
e
h
an
d
,
th
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
is
p
e
r
f
o
r
m
ed
b
y
th
e
co
n
tr
ac
tin
g
p
at
h
wh
er
ea
s
th
e
s
eg
m
e
n
tatio
n
m
ap
p
ed
r
esu
lts
ar
e
o
b
tain
ed
b
y
s
y
n
th
esizin
g
t
h
e
s
p
atial
in
f
o
r
m
atio
n
with
h
i
g
h
r
eso
lu
tio
n
f
ea
t
u
r
es.
I
n
co
n
tr
ac
tin
g
p
ath
,
two
co
n
v
o
l
u
tio
n
lay
er
s
o
f
‘
3
×
3
’
ar
e
r
e
p
ea
ted
.
Mo
r
eo
v
er
,
m
a
x
p
o
o
lin
g
o
f
‘
2
×
2
’
is
ca
r
r
ied
o
u
t.
I
n
ea
ch
s
tep
in
c
o
n
s
tr
ictin
g
p
ath
,
n
u
m
b
er
o
f
asp
ec
t
ch
an
n
els
is
en
h
an
ce
d
m
o
d
er
at
ely
f
r
o
m
‘
16
256
’
.
C
o
n
v
er
s
ely
,
in
t
h
e
ex
p
a
n
s
io
n
p
ath
,
t
h
e
n
u
m
b
e
r
o
f
f
ea
tu
r
e
ch
an
n
els
is
r
ed
u
ce
d
f
r
o
m
‘
256
16
’
.
Ad
d
itio
n
ally
,
th
e
m
a
x
p
o
o
lin
g
lay
er
o
f
‘
2
×
2
’
an
d
two
c
o
n
v
o
lu
ti
o
n
lay
er
s
o
f
‘
3
×
3
’
ar
e
p
er
f
o
r
m
ed
c
o
n
s
ec
u
tiv
ely
.
Nex
t,
th
e
f
r
a
g
m
en
t
d
er
iv
ativ
e
f
o
r
p
r
e
p
r
o
ce
s
s
ed
r
esu
lt
f
o
r
th
e
f
ir
s
t
d
er
iv
ativ
e,
s
ec
o
n
d
d
er
iv
ativ
e
‘
2
’
an
d
‘
−
ℎ
’
d
er
iv
ativ
e
to
in
cr
ea
s
e
th
e
r
eso
lu
tio
n
o
f
th
e
o
u
tp
u
ts
v
ia
co
m
p
o
s
ite
d
ilatio
n
is
s
tated
as
(
1
1
)
to
(
1
3
)
.
(
)
=
→
(
)
=
(
)
=
−
1
(
1
1
)
2
(
)
2
2
(
)
=
(
−
1
)
−
2
(
1
2
)
(
)
(
)
=
(
−
1
)
−
(
1
3
)
T
h
en
th
e
FR
eL
U
is
s
tated
as
(
1
4
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
Weiers
t
r
a
s
s
s
ca
le
s
p
a
ce
r
ep
r
esen
ta
tio
n
a
n
d
co
mp
o
s
ite
d
ila
ted
U
-
n
et
…
(
A
b
d
u
l B
a
s
ith
Za
h
i
r
Hu
s
s
a
in
)
1667
=
=
{
!
(
−
)
!
−
;
>
0
0
;
≤
0
(
1
4
)
Nex
t,
with
o
b
tain
ed
ac
tiv
ati
o
n
r
esu
lts
in
h
id
d
en
lay
e
r
v
ia
‘
’
as
p
r
o
v
id
e
d
(
1
4
)
,
to
o
b
tain
h
ig
h
e
r
s
eg
m
en
tatio
n
p
er
f
o
r
m
a
n
ce
r
es
u
lts
,
lo
g
co
s
h
d
ice
l
o
s
s
f
u
n
ctio
n
is
f
o
r
m
u
lated
as
(
1
5
)
.
c
os
ℎ
(
)
=
l
og
c
os
(
+
−
2
)
(
1
5
)
Fin
ally
,
th
e
m
o
s
t
ex
ten
s
iv
ely
u
tili
ze
d
m
ea
s
u
r
ed
f
o
r
g
lau
co
m
a
d
etec
tio
n
,
n
a
m
ely
,
u
s
in
g
o
p
tic
C
DR
is
m
o
s
t
wid
ely
u
s
ed
f
ea
tu
r
e
f
o
r
g
lau
co
m
a
d
etec
tio
n
.
T
h
e
r
ea
s
o
n
b
eh
in
d
th
e
em
p
l
o
y
m
en
t
o
f
C
DR
is
th
at
th
e
p
h
en
o
m
en
o
n
o
f
cu
p
p
in
g
is
s
aid
to
o
cc
u
r
u
p
o
n
p
r
ev
alen
ce
o
f
ce
r
tain
co
n
s
id
er
a
b
le
am
o
u
n
t
o
f
s
tr
ain
ch
u
r
n
e
d
o
u
t
in
th
e
r
etin
a.
T
h
e
C
DR
h
er
e
is
m
ea
s
u
r
ed
tak
in
g
in
to
co
n
s
id
er
atio
n
s
th
e
ar
e
a
o
f
OC
an
d
OD.
I
t m
ath
em
atica
lly
f
o
r
m
u
lated
as
(
1
6
)
.
=
2
∗
[
[
]
[
]
]
(
1
6
)
Fro
m
(
1
6
)
is
f
o
r
m
u
lated
d
ep
en
d
o
n
a
r
ea
o
f
c
u
p
‘
’
an
d
ar
e
a
o
f
d
is
c
‘
’
with
r
esp
ec
t
to
th
e
r
esu
ltan
t
im
ag
es
o
b
tain
ed
in
‘
’
r
esp
ec
tiv
ely
.
Acc
o
r
d
in
g
to
t
h
e
r
esu
ltan
t
v
al
u
es
o
b
tain
ed
in
(
1
6
)
,
th
e
o
u
tp
u
t
in
th
e
o
u
tp
u
t
l
a
y
er
is
g
en
er
ated
b
y
eith
e
r
g
lau
c
o
m
a
to
u
s
o
r
h
ea
lth
y
im
a
g
e
r
esu
lts
.
T
h
e
p
s
eu
d
o
co
d
e
r
ep
r
esen
tatio
n
o
f
C
DC
-
b
ased
s
eg
m
en
tatio
n
is
g
iv
en
i
n
Alg
o
r
ith
m
2
.
Alg
o
r
ith
m
2
.
C
DC
-
b
ased
s
eg
m
en
tatio
n
f
o
r
g
lau
c
o
m
a
d
etec
t
io
n
Input
: Dataset ‘
’
Output
: Early glaucoma detection
1:
Initialize
‘
’, preprocessed results ‘
’
2:
Begin
3:
For
each Dataset ‘
’ with preprocessed results ‘
’
//Input layer
4: Provide preprocessed results ‘
’ as input
//Hidden layer
5:
Fo
rm
ul
at
e
Re
LU
ac
ti
va
t
io
n
fo
r
ea
ch
pr
ep
ro
ce
ss
ed
re
su
lt
s
‘
’
as
gi
ve
n
in
eq
ua
ti
on
s
(1
0)
and (11)
6:
Fo
rm
ul
at
e
fragment
d
e
rivative
for
preprocessed
result
as
given
in
equa
tions
(11),
(12)
and (13)
7: Formulate FReLU activation function as given in equation (14)
//Output layer
8: Me
asure log cosh dice loss function as given in equation (15)
9: Measure optic cup to disc ratio as given in equation (16)
10:
If
‘
[
]
≥
0
.
5
’
11:
Then
glaucomatous image
12:
End if
13:
If
‘
[
]
<
0
.
5
’
14:
Then
healthy image
15:
End for
16:
End
T
h
e
C
DC
-
b
ased
s
eg
m
en
tatio
n
is
illu
s
tr
ated
in
A
lg
o
r
ith
m
2
.
As
g
iv
en
in
alg
o
r
ith
m
2
,
to
en
s
u
r
e
ea
r
ly
g
lau
co
m
a
d
etec
tio
n
with
m
in
i
m
al
er
r
o
r
,
f
ir
s
t,
th
e
p
r
e
p
r
o
ce
s
s
ed
in
p
u
t
im
a
g
es
ar
e
g
i
v
en
as
in
p
u
t
to
th
e
in
p
u
t
lay
er
.
Seco
n
d
,
in
h
id
d
en
lay
e
r
f
r
ag
m
en
t
d
er
i
v
ativ
e
f
o
r
ea
c
h
p
r
ep
r
o
ce
s
s
ed
in
p
u
t
ar
e
o
b
ta
in
ed
.
L
astl
y
,
in
th
e
o
u
tp
u
t
lay
er
two
p
r
o
ce
s
s
es
ar
e
ca
r
r
ied
o
u
t,
f
i
r
s
t,
lo
g
co
s
h
d
ic
e
lo
s
s
f
u
n
ctio
n
is
a
p
p
lied
to
r
e
d
u
ce
th
e
d
ia
g
n
o
s
tic
er
r
o
r
an
d
th
en
o
p
tic
C
DR
is
ev
alu
ated
t
o
o
b
tain
t
h
e
s
eg
m
e
n
ted
g
lau
c
o
m
a
d
e
tecte
d
r
esu
lts
in
an
ac
cu
r
ate
an
d
p
r
ec
is
e
m
an
n
er
.
4.
RE
SU
L
T
S
AND
D
I
SCU
SS
I
O
N
T
h
e
p
r
o
p
o
s
ed
W
SS
R
-
C
DC
f
o
r
g
lau
c
o
m
a
d
etec
tio
n
is
ev
alu
ated
u
s
in
g
Py
th
o
n
h
ig
h
-
lev
el
p
r
o
g
r
a
m
m
in
g
-
lan
g
u
a
g
e
an
d
r
e
s
u
lts
ar
e
co
m
p
ar
ed
with
th
e
p
r
ev
io
u
s
wo
r
k
s
s
u
ch
as,
m
u
lti
-
f
ea
tu
r
e
MFDL
[
1
]
an
d
m
u
lti
-
task
DL
[
2
]
.
T
h
e
ai
m
o
f
th
e
p
r
o
p
o
s
ed
W
SS
R
-
C
D
C
is
to
ac
h
iev
e
ac
cu
r
ate
g
lau
c
o
m
a
d
etec
tio
n
with
m
a
x
i
m
u
m
a
c
c
u
r
a
c
y
a
n
d
l
e
s
s
e
r
d
i
a
g
n
o
s
t
i
c
e
r
r
o
r
.
B
a
s
e
d
o
n
t
h
e
o
b
j
e
c
t
i
v
e
,
t
h
e
e
x
i
s
t
i
n
g
m
e
t
h
o
d
s
s
u
c
h
a
s
M
F
D
L
[
1
]
an
d
m
u
lti
-
task
DL
[
2
]
a
r
e
ta
k
en
as
b
ase
p
a
p
er
.
T
h
ese
two
b
ase
p
ap
e
r
s
ar
e
ex
p
lain
ed
to
u
n
d
e
r
s
tan
d
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
T
h
e
ex
is
tin
g
DL
m
eth
o
d
s
wer
e
d
esig
n
ed
f
o
r
g
la
u
co
m
a
d
etec
t
io
n
.
Ho
w
ev
er
,
t
h
e
ac
c
u
r
ac
y
was
n
o
t
en
h
an
ce
d
,
d
iag
n
o
s
tic
er
r
o
r
was
n
o
t
r
ed
u
ce
d
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
co
n
ce
p
t
is
d
er
i
v
ed
b
y
co
n
s
id
er
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
6
6
1
-
1
6
7
2
1668
th
e
p
r
o
b
lem
s
o
f
th
ese
b
ase
p
ap
er
s
.
T
h
e
d
r
awb
ac
k
s
o
f
th
ese
m
eth
o
d
s
ar
e
ef
f
ec
tiv
ely
co
n
v
in
ce
d
b
y
im
p
lem
en
tin
g
th
e
p
r
o
p
o
s
ed
m
e
th
o
d
.
I
n
ad
d
itio
n
,
th
e
r
es
u
lts
ar
e
ev
alu
ated
b
ased
o
n
th
e
m
etr
ics
s
u
ch
a
s
s
en
s
itiv
ity
,
s
p
ec
if
icity
,
im
ag
e
d
iag
n
o
s
tic
er
r
o
r
a
n
d
ac
cu
r
a
cy
u
s
in
g
th
e
g
lau
c
o
m
a
f
u
n
d
u
s
im
ag
in
g
d
ataset
ex
tr
ac
ted
f
r
o
m
h
ttp
s
://www.
k
ag
g
le.
co
m
/d
atasets
/ar
n
av
jain
1
/g
lau
co
m
a
-
d
atasets
.
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
W
SS
R
-
C
D
C
m
eth
o
d
is
co
m
p
ar
ed
with
th
e
o
th
er
co
m
p
etin
g
m
eth
o
d
s
,
MFDL
[
1
]
an
d
m
u
lti
-
task
DL
[
2
]
an
d
v
alid
ated
.
4
.
1
.
I
m
ple
m
ent
a
t
io
n det
a
ils
W
e
d
ev
elo
p
ed
a
n
ea
r
ly
g
lau
c
o
m
a
d
etec
tio
n
m
et
h
o
d
ca
lled
W
SS
R
-
C
DC
wi
th
im
p
r
o
v
e
d
p
r
e
cisi
o
n
an
d
ac
cu
r
ac
y
:
−
T
h
e
W
SS
R
-
C
DC
m
eth
o
d
co
m
p
r
is
es two
s
ec
tio
n
s
,
n
am
ely
,
p
r
ep
r
o
ce
s
s
in
g
an
d
s
eg
m
en
tatio
n
.
−
T
h
e
W
SS
R
-
C
DC
m
eth
o
d
is
c
o
m
p
ar
ed
with
two
ex
is
tin
g
m
e
th
o
d
s
,
MFDL
[
1
]
a
n
d
m
u
lti
-
ta
s
k
DL
[
2
]
u
s
in
g
a
g
lau
co
m
a
f
u
n
d
u
s
im
ag
in
g
d
a
taset to
v
alid
ate
th
e
r
esu
lts
.
−
I
n
itially
,
th
e
f
u
n
d
u
s
im
ag
es
ar
e
o
b
tain
e
d
as
in
p
u
t
f
r
o
m
th
e
d
ataset.
T
h
e
im
ag
es
wer
e
s
u
b
jecte
d
to
p
r
ep
r
o
ce
s
s
in
g
an
d
s
eg
m
e
n
tatio
n
f
o
r
ea
r
ly
g
lau
co
m
a
d
etec
tio
n
.
−
I
n
th
e
f
ir
s
t
p
ar
t,
th
e
W
eier
s
tr
ass
t
r
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
-
b
ased
p
r
ep
r
o
ce
s
s
in
g
m
o
d
el
is
em
p
lo
y
ed
to
p
r
o
ce
s
s
im
ag
e
s
tr
u
ctu
r
es a
t d
if
f
er
e
n
t scale
s
v
ia
E
u
ler
’
s
i
d
en
tity
.
−
Seco
n
d
,
a
DL
m
o
d
el
em
p
lo
y
in
g
C
DC
m
o
d
el
is
u
tili
ze
d
to
p
r
ep
r
o
ce
s
s
ed
n
o
is
e
im
ag
e.
T
h
e
p
r
o
ce
s
s
u
n
d
er
g
o
es
co
n
tr
ac
tin
g
an
d
ex
p
an
s
io
n
s
ep
ar
ately
.
Als
o
to
m
in
im
ize
d
iag
n
o
s
tic
er
r
o
r
,
lo
g
co
s
h
d
ice
lo
s
s
f
u
n
ctio
n
is
ap
p
lied
.
Fin
ally
,
u
s
in
g
o
p
tic
C
DR
g
lau
co
m
a
d
etec
tio
n
is
m
ad
e
in
ex
te
n
s
iv
e
m
a
n
n
er
.
Acc
o
r
d
in
g
to
th
e
ab
o
v
e
im
p
l
em
en
tatio
n
p
atter
n
s
,
f
o
u
r
d
if
f
er
en
t
ev
alu
atio
n
m
etr
ics
ar
e
d
etailed
in
th
e
n
ex
t
s
ec
tio
n
.
4
.
2
.
Dis
cus
s
io
n
First,
s
en
s
it
iv
ity
test
is
p
er
f
o
r
m
ed
to
m
ea
s
u
r
e
its
ab
ilit
y
to
d
eter
m
in
e
th
e
p
atien
t
ca
s
es
co
r
r
ec
tly
(
i.e
.
,
g
lau
co
m
a
as
g
lau
c
o
m
a
an
d
h
e
alth
y
as
h
ea
lth
y
)
.
T
o
m
ea
s
u
r
e
th
e
s
en
s
itiv
ity
,
r
ate
th
e
p
r
o
p
o
r
tio
n
o
r
r
atio
o
f
tr
u
e
p
o
s
itiv
e
in
-
p
atien
t
ca
s
es
h
as
t
o
b
e
an
aly
ze
d
.
T
o
b
e
m
o
r
e
s
p
ec
if
ic,
s
en
s
itiv
ity
in
d
icate
s
th
e
r
atio
o
f
p
o
s
itiv
es
wh
ich
ar
e
p
r
o
p
er
l
y
h
y
p
o
th
esiz
ed
.
I
t is ex
p
r
ess
ed
as
(
1
7
)
.
=
+
(
1
7
)
Fro
m
th
e
(
1
7
)
,
s
en
s
itiv
ity
r
ate
‘
’
,
is
ca
lcu
lated
d
ep
en
d
o
n
tr
u
e
p
o
s
itiv
e
ca
s
es
‘
’
(
i.e
.
,
h
ea
lth
y
p
atien
t
d
etec
ted
as
h
ea
lth
y
)
as
well
as
f
alse
n
eg
ativ
e
ca
s
es
‘
’
(
i.e
.
,
h
ea
lth
y
p
atien
t
d
etec
te
d
as
g
lau
co
m
a)
r
esp
ec
tiv
ely
.
Seco
n
d
,
s
p
ec
if
ic
ity
o
r
th
e
p
r
o
b
a
b
ilit
y
o
f
n
e
g
ativ
e
test
r
esu
lts
i
s
ev
alu
ated
.
Sp
ec
if
icity
in
d
icate
s
th
e
f
r
ac
tio
n
o
f
n
e
g
ativ
es th
at
a
r
e
ac
cu
r
ately
in
f
er
r
e
d
an
d
ex
p
r
ess
ed
as
(
1
8
)
.
=
+
(
1
8
)
Fro
m
(
1
8
)
,
s
p
ec
if
icity
‘
’
,
r
at
e
is
ca
lcu
lated
,
d
ep
en
d
o
n
tr
u
e
n
eg
ativ
e
r
ate
‘
’
(
i.e
.
,
g
lau
co
m
a
p
atien
t
d
etec
te
d
as
g
lau
co
m
a)
a
n
d
th
e
f
alse
p
o
s
itiv
e
‘
’
(
i.e
.
,
g
lau
co
m
a
p
atien
t
d
etec
ted
as
h
ea
lth
y
)
r
ate
r
esp
ec
tiv
ely
.
T
h
ir
d
,
f
o
r
ass
ess
in
g
th
e
s
ig
n
if
ican
ce
o
f
g
lau
co
m
a
d
etec
tio
n
o
n
e
o
f
th
e
im
p
o
r
tan
t
p
er
f
o
r
m
an
ce
m
etr
ics
is
ac
cu
r
a
cy
.
Acc
u
r
ac
y
is
r
ef
er
r
e
d
th
e
r
a
tio
o
f
p
r
o
p
er
f
o
r
ec
ast
to
to
tal
n
u
m
b
er
o
f
s
am
p
les.
I
t is f
o
r
m
u
lated
as
(
1
9
)
.
=
+
+
+
+
(
1
9
)
Fro
m
th
e
(
1
9
)
,
ac
cu
r
ac
y
r
ate
‘
’
,
is
esti
m
ated
b
y
tr
u
e
p
o
s
iti
v
e
r
ate
‘
’
,
,
f
alse
p
o
s
itiv
e
r
ate
an
d
.
I
t
is
ca
lcu
lated
in
p
er
ce
n
tag
e
(
%).
Fin
ally
,
d
iag
n
o
s
tic
er
r
o
r
o
r
m
ea
s
u
r
e
to
v
alid
ate
th
e
ef
f
ec
tiv
en
ess
o
f
tech
n
i
q
u
e
is
f
o
r
m
u
lated
as
(
2
0
)
.
=
∑
=
1
∗
100
(
2
0
)
Fro
m
th
e
(
2
0
)
,
th
e
d
iag
n
o
s
tic
er
r
o
r
‘
’
i
s
ev
alu
ated
b
y
s
a
m
p
les
‘
’
as
well
as
s
am
p
les
wr
o
n
g
ly
d
etec
ted
‘
’
with
h
ea
lth
y
as g
la
u
co
m
a
an
d
g
lau
c
o
m
a
as h
ea
lth
y
.
T
ab
le
1
co
m
p
ar
es
th
e
o
u
tc
o
m
es
o
f
th
e
W
SS
R
-
C
DC
tec
h
n
iq
u
e
o
f
s
en
s
itiv
ity
with
t
h
o
s
e
o
th
er
m
eth
o
d
s
,
MFDL
[
1
]
an
d
m
u
lti
-
task
DL
[
2
]
u
s
in
g
th
e
g
la
u
co
m
a
f
u
n
d
u
s
im
a
g
in
g
d
atas
et.
T
h
e
r
ea
s
o
n
f
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
Weiers
t
r
a
s
s
s
ca
le
s
p
a
ce
r
ep
r
esen
ta
tio
n
a
n
d
co
mp
o
s
ite
d
ila
ted
U
-
n
et
…
(
A
b
d
u
l B
a
s
ith
Za
h
i
r
Hu
s
s
a
in
)
1669
en
h
an
cin
g
th
e
s
en
s
itiv
ity
is
t
o
ap
p
ly
g
a
u
s
s
ian
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
.
Her
e,
th
e
d
is
c
r
ete
an
d
co
n
tin
u
o
u
s
r
ep
r
esen
tatio
n
o
f
im
a
g
e
s
tr
u
ct
u
r
es
at
d
if
f
e
r
en
t
s
ca
les
was
o
b
tain
ed
u
s
in
g
E
u
ler
'
s
id
en
tity
a
n
d
B
ess
el
f
u
n
ctio
n
.
T
h
is
in
tu
r
n
im
p
r
o
v
ed
th
e
o
v
er
all
s
en
s
itiv
ity
o
f
W
SS
R
-
C
D
C
tech
n
iq
u
e
b
y
1
2
% a
n
d
2
4
% th
an
th
e
[
1
]
,
[
2
]
.
Fig
u
r
e
4
illu
s
tr
ates a
g
r
ap
h
ical
d
ep
ictio
n
o
f
s
p
ec
if
icity
f
o
r
2
,
0
0
0
d
if
f
er
en
t sam
p
le
im
ag
es p
r
o
v
id
ed
as
in
p
u
t.
Fin
ally
,
co
n
tin
u
o
u
s
s
ca
le
s
p
ac
e
an
d
d
is
cr
ete
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
s
wer
e
co
m
b
in
ed
u
s
in
g
W
eier
s
tr
as
s
tr
an
s
f
o
r
m
f
u
n
ctio
n
th
at
in
tu
r
n
en
s
u
r
es
ac
c
u
r
ac
y
o
f
R
OI
d
etec
tio
n
i
n
a
s
ig
n
if
ican
t
m
an
n
er
a
n
d
th
er
ef
o
r
e
im
p
r
o
v
in
g
t
h
e
s
p
ec
if
icity
u
s
in
g
W
SS
R
-
C
DC
tech
n
i
q
u
e
b
y
8
% a
n
d
1
4
% th
an
th
e
[
1
]
,
[
2
]
.
T
ab
le
2
co
m
p
ar
es
th
e
o
u
tco
m
es
o
f
th
e
p
r
o
p
o
s
ed
W
SS
R
-
C
DC
m
eth
o
d
in
ter
m
s
o
f
ac
c
u
r
ac
y
with
th
o
s
e
o
th
er
m
eth
o
d
s
,
MFDL
[
1
]
an
d
m
u
lti
-
task
DL
[
2
]
u
s
in
g
th
e
g
lau
co
m
a
f
u
n
d
u
s
im
ag
in
g
d
ataset.
T
h
e
two
d
if
f
er
en
t
ac
tiv
atio
n
f
u
n
ctio
n
s
ar
e
u
tili
ze
d
in
th
e
h
id
d
en
lay
er
.
T
h
e
f
ir
s
t
ac
tiv
atio
n
f
u
n
ctio
n
was
th
e
em
p
lo
y
m
e
n
t
o
f
FR
eL
U
wh
er
e
with
th
e
p
r
ep
r
o
ce
s
s
ed
r
esu
lt
im
ag
es
as
in
p
u
t,
th
e
f
r
ag
m
en
t
d
er
iv
ativ
e
f
o
r
th
e
f
ir
s
t
d
er
iv
ativ
e
was
o
b
tain
e
d
an
d
th
e
s
ec
o
n
d
d
er
iv
ativ
e
was
m
ea
s
u
r
ed
.
Fin
ally
,
‘
−
ℎ
’
d
e
r
iv
ativ
e
was
f
o
r
m
u
lated
with
th
e
p
u
r
p
o
s
e
o
f
in
cr
ea
s
in
g
th
e
r
eso
lu
tio
n
o
f
th
e
o
u
tp
u
ts
v
ia
co
m
p
o
s
ite
d
ilatio
n
.
T
h
is
in
tu
r
n
r
ed
u
ce
d
FP
an
d
FN
r
ate
an
d
th
er
ef
o
r
e
im
p
r
o
v
in
g
o
v
e
r
all
ac
cu
r
ac
y
o
f
W
SS
R
-
C
D
C
m
eth
o
d
b
y
8
%
u
p
o
n
co
m
p
ar
is
o
n
t
o
[
1
]
an
d
1
6
% u
p
o
n
co
m
p
ar
is
o
n
to
[
2
]
.
T
ab
le
1
.
C
o
m
p
a
r
is
o
n
o
f
th
e
p
e
r
f
o
r
m
a
n
ce
o
f
s
en
s
itiv
ity
o
f
W
SS
R
-
C
DC
m
eth
o
d
with
ex
is
ti
n
g
MFDL
[
1
]
a
n
d
m
u
lti
-
task
DL
[
2
]
S
a
mp
l
e
i
ma
g
e
s
S
e
n
s
i
t
i
v
i
t
y
WSSR
-
C
D
C
M
F
D
L
mu
l
t
i
-
t
a
s
k
DL
2
0
0
0
.
9
0
.
8
4
0
.
7
7
4
0
0
0
.
8
7
0
.
8
2
0
.
7
4
6
0
0
0
.
8
5
0
.
8
0
.
7
1
8
0
0
0
.
8
4
0
.
7
8
0
.
6
9
1
,
0
0
0
0
.
8
3
0
.
7
5
0
.
6
6
1
,
2
0
0
0
.
8
1
0
.
7
3
0
.
6
3
1
,
4
0
0
0
.
7
8
0
.
7
0
.
6
1
,
6
0
0
0
.
7
8
0
.
7
0
.
6
1
,
8
0
0
0
.
8
2
0
.
7
2
0
.
6
5
2
,
0
0
0
0
.
8
3
0
.
7
4
0
.
6
7
Fig
u
r
e
4
.
Sp
ec
if
icity
v
er
s
u
s
s
am
p
le
im
ag
es
T
ab
le
2
.
C
o
m
p
a
r
is
o
n
o
f
ac
cu
r
ac
y
o
f
th
e
W
SS
R
-
C
DC
m
eth
o
d
,
MFDL
[
1
]
a
n
d
m
u
lti
-
task
DL
[
2
]
S
a
mp
l
e
i
ma
g
e
s
A
c
c
u
r
a
c
y
(
%)
WSSR
-
C
D
C
M
F
D
L
mu
l
t
i
-
t
a
s
k
DL
2
0
0
0
.
9
4
0
.
8
8
0
.
8
3
4
0
0
0
.
9
2
0
.
8
4
0
.
8
1
6
0
0
0
.
8
8
0
.
8
2
0
.
7
8
8
0
0
0
.
8
5
0
.
8
1
0
.
7
5
1
,
0
0
0
0
.
8
3
0
.
7
8
0
.
7
1
1
,
2
0
0
0
.
8
2
0
.
7
5
0
.
6
9
1
,
4
0
0
0
.
8
2
0
.
7
5
0
.
6
9
1
,
6
0
0
0
.
8
4
0
.
7
7
0
.
7
1
1
,
8
0
0
0
.
8
6
0
.
7
9
0
.
7
3
2
,
0
0
0
0
.
8
8
0
.
8
1
0
.
7
5
0
0
.
2
0
.
4
0
.
6
0
.8
1
2
0
0
4
0
0
6
0
0
8
0
0
1
0
0
0
1
2
0
0
1
4
0
0
1
6
0
0
1
8
0
0
2
0
0
0
S
p
e
c
i
f
i
c
i
t
y
S
a
m
p
l
e
i
m
a
g
e
s
W
S
S
R
-
C
D
C
M
F
D
L
mu
l
t
i
-
t
a
sk
d
e
e
p
l
e
a
r
n
i
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
8
,
No
.
3
,
J
u
n
e
20
2
5
:
1
6
6
1
-
1
6
7
2
1670
Fig
u
r
e
5
d
ep
icts
a
g
r
a
p
h
ical
d
ep
ictio
n
o
f
th
e
e
r
r
o
r
r
ate
wh
en
s
u
b
s
titu
ted
in
(
2
0
)
f
o
r
th
r
e
e
m
eth
o
d
s
W
SS
R
-
C
D
C
,
MFDL
[
1
]
,
an
d
m
u
lti
-
task
DL
[
2
]
.
T
h
e
r
ea
s
o
n
was
th
at
b
y
a
p
p
ly
in
g
th
e
lo
g
co
s
h
d
ice
lo
s
s
f
u
n
ctio
n
h
i
g
h
er
s
eg
m
en
tatio
n
p
er
f
o
r
m
an
ce
r
esu
lts
wer
e
o
b
ta
in
ed
.
Nex
t,
b
ased
o
n
th
e
o
p
t
ic
C
D
R
s
eg
m
en
ted
p
o
r
tio
n
s
wer
e
a
n
aly
ze
d
f
o
r
g
l
au
co
m
a
an
d
h
ea
lth
y
im
ag
es.
T
h
is
in
tu
r
n
r
ed
u
ce
d
th
e
w
r
o
n
g
ly
d
etec
ted
r
esu
lts
a
n
d
t
h
e
r
e
f
o
r
e
r
e
d
u
c
e
d
t
h
e
o
v
e
r
a
l
l
d
i
a
g
n
o
s
t
i
c
e
r
r
o
r
u
s
i
n
g
t
h
e
W
SS
R
-
C
D
C
m
e
t
h
o
d
b
y
2
6
%
a
n
d
3
7
%
t
h
a
n
t
h
e
[
2
]
.
T
h
e
ea
r
ly
s
tag
e
o
f
g
la
u
co
m
a
id
en
tific
atio
n
is
a
cr
u
cial
task
to
av
o
id
b
lin
d
n
ess
.
Mo
s
t
co
n
v
o
lu
ti
o
n
tech
n
iq
u
es
ar
e
d
ev
elo
p
ed
to
d
eter
m
in
e
ey
e
d
is
o
r
d
e
r
s
th
r
o
u
g
h
f
u
n
d
u
s
im
ag
es.
T
h
e
ex
is
tin
g
m
eth
o
d
s
ar
e
d
escr
ib
ed
in
th
e
m
ajo
r
is
s
u
es
s
u
ch
as
less
er
ac
cu
r
ac
y
,
h
ig
h
er
er
r
o
r
,
m
in
im
u
m
s
en
s
itiv
ity
,
th
e
g
lau
c
o
m
a
id
en
tific
atio
n
p
e
r
f
o
r
m
an
ce
wa
s
n
o
t
im
p
r
o
v
ed
,
an
d
f
ailu
r
e
t
o
p
r
o
v
id
e
ac
cu
r
ate
r
esu
lts
.
T
o
o
v
er
co
m
e
t
h
e
is
s
u
es,
in
o
r
d
er
to
s
o
lv
e
th
is
is
s
u
e,
a
n
o
v
el
W
SS
R
-
C
DC
m
eth
o
d
i
s
n
ee
d
ed
f
o
r
ea
r
l
y
d
is
ea
s
e
d
etec
tio
n
.
T
h
e
m
ajo
r
f
in
d
in
g
s
an
d
o
u
tc
o
m
e
o
f
th
e
p
r
o
p
o
s
ed
W
SS
R
-
C
D
C
m
eth
o
d
o
b
s
er
v
ed
f
r
o
m
t
h
e
ab
o
v
e
r
esu
l
ts
ar
e
as f
o
llo
ws:
−
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
ad
d
r
e
s
s
es
th
e
ea
r
ly
g
lau
co
m
a
d
e
tectio
n
in
r
etin
al
f
u
n
d
u
s
im
ag
es
b
y
u
s
in
g
W
eier
s
tr
as
s
t
r
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
a
n
d
C
DC
m
o
d
el.
−
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
u
s
es
th
e
W
eier
s
tr
as
s
tr
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
f
o
r
p
er
f
o
r
m
i
n
g
p
r
ep
r
o
ce
s
s
in
g
to
cr
ea
te
g
e
n
er
a
te
s
ca
le
-
in
v
ar
ian
t p
r
e
p
r
o
ce
s
s
ed
im
ag
es
with
h
ig
h
e
r
s
en
s
itiv
ity
.
−
T
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
e
m
p
lo
y
s
C
D
C
m
o
d
el
f
o
r
i
d
en
tify
in
g
t
o
d
etec
tin
g
g
lau
co
m
a
.
−
L
o
g
co
s
h
d
ice
lo
s
s
f
u
n
ctio
n
is
u
tili
ze
d
to
d
ec
r
ea
s
e
th
e
d
iag
n
o
s
tic
er
r
o
r
.
−
T
h
e
o
p
tic
cu
p
to
d
is
c
r
atio
is
m
ea
s
u
r
ed
to
g
et
th
e
s
eg
m
en
te
d
g
lau
co
m
a
d
etec
ted
r
esu
lts
.
−
T
h
e
o
u
tco
m
e
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
ac
h
iev
es
2
4
%
o
f
ac
c
u
r
ac
y
,
1
8
%
o
f
s
en
s
itiv
ity
,
1
1
%
o
f
s
p
ec
if
icity
,
an
d
3
2
%
o
f
er
r
o
r
as c
o
m
p
ar
ed
to
ex
is
tin
g
wo
r
k
s
.
Fig
u
r
e
5
.
Diag
n
o
s
tic
er
r
o
r
s
v
e
r
s
u
s
s
am
p
le
im
ag
es
5.
CO
NCLU
SI
O
N
I
n
o
u
r
wo
r
k
,
th
e
o
b
jectiv
e
o
f
th
e
p
r
o
p
o
s
ed
W
SS
R
-
C
D
C
f
o
r
g
la
u
co
m
a
d
etec
tio
n
u
s
in
g
g
lau
co
m
a
f
u
n
d
u
s
r
etin
al
im
ag
es
is
to
o
b
tain
ac
cu
r
ate
an
d
p
r
ec
is
e
g
lau
co
m
a
d
etec
ted
r
esu
lts
.
First,
p
r
ep
r
o
ce
s
s
in
g
u
s
in
g
r
aw
im
ag
es
was
p
er
f
o
r
m
ed
u
s
in
g
W
eier
s
tr
ass
tr
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
es
en
tatio
n
to
o
b
tai
n
p
r
o
ce
s
s
ed
r
esu
lts
at
d
if
f
er
en
t
s
ca
les.
Nex
t,
wit
h
th
e
p
r
ep
r
o
ce
s
s
ed
im
ag
e
r
e
s
u
lts
,
s
eg
m
en
tatio
n
f
o
r
g
lau
co
m
a
d
etec
tio
n
was
p
er
f
o
r
m
ed
b
y
m
ea
n
s
o
f
C
DC
m
o
d
el.
Her
e
also
co
m
p
o
s
ite
d
ilated
r
esu
lts
wer
e
s
u
b
jecte
d
t
o
lo
g
co
s
h
d
ice
lo
s
s
f
u
n
ctio
n
with
th
e
o
b
jectiv
e
o
f
r
etain
in
g
th
e
e
d
g
e
f
ea
tu
r
es
with
m
in
im
al
d
iag
n
o
s
tic
er
r
o
r
.
B
y
in
co
r
p
o
r
atin
g
th
ese
f
ea
tu
r
es
in
to
FR
eL
U
,
ex
ce
llen
t
s
eg
m
en
tatio
n
ac
cu
r
ac
y
was
ac
h
iev
ed
c
o
m
p
ar
e
d
to
p
r
ec
ed
in
g
tech
n
iq
u
es.
T
h
e
p
r
o
p
o
s
ed
W
SS
R
-
C
D
C
is
to
p
r
o
v
id
e
p
r
ec
is
e
g
la
u
co
m
a
r
esu
lts
em
p
lo
y
in
g
W
eier
s
tr
ass
t
r
an
s
f
o
r
m
s
ca
le
s
p
ac
e
r
ep
r
esen
tatio
n
b
ased
p
r
ep
r
o
ce
s
s
in
g
m
o
d
el
an
d
C
DC
m
o
d
el
f
o
r
s
eg
m
en
tatio
n
with
m
in
im
u
m
er
r
o
r
a
n
d
m
ax
im
u
m
ac
c
u
r
ac
y
a
n
d
s
p
ec
if
icity
.
T
h
ese
f
in
d
in
g
s
h
av
e
im
p
licatio
n
s
f
o
r
id
en
tif
y
in
g
th
e
g
l
au
co
m
a.
E
x
p
er
im
en
ts
wer
e
p
er
f
o
r
m
ed
o
n
th
e
g
lau
co
m
a
f
u
n
d
u
s
r
etin
al
im
ag
e
d
atab
ase
to
test
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
tech
n
iq
u
e
a
n
d
ex
is
tin
g
m
eth
o
d
s
.
T
h
e
p
r
o
p
o
s
ed
W
SS
R
-
C
D
C
is
co
m
p
ar
ed
with
t
h
e
tw
o
e
x
is
tin
g
m
eth
o
d
s
(
i.e
.
MFDL
an
d
m
u
lti
-
task
DL
)
.
T
h
e
r
esu
l
ts
o
f
th
e
W
SS
R
-
C
D
C
p
r
o
v
id
e
b
etter
p
er
f
o
r
m
a
n
ce
with
an
im
p
r
o
v
em
e
n
t
o
f
ac
cu
r
ac
y
b
y
2
4
%,
s
en
s
itiv
ity
b
y
1
8
%
s
p
ec
if
icity
b
y
1
1
%,
a
n
d
r
ed
u
ctio
n
o
f
e
r
r
o
r
b
y
3
2
%
as
co
m
p
a
r
ed
t
o
ex
is
tin
g
wo
r
k
s
.
T
h
e
p
r
o
p
o
s
ed
W
SSR
-
C
D
C
m
eth
o
d
ac
h
iev
es
b
etter
ac
cu
r
ac
y
an
d
s
en
s
itiv
ity
with
m
in
im
al
d
iag
n
o
s
tic
er
r
o
r
th
an
th
e
co
n
v
en
tio
n
al
m
eth
o
d
s
.
I
n
f
u
tu
r
e
wo
r
k
,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
is
f
u
r
th
er
e
x
ten
d
e
d
to
u
s
e
n
ew
r
esear
ch
wo
r
k
f
o
r
d
etec
tin
g
t
h
e
g
lau
c
o
m
a
r
esu
lts
b
y
u
s
in
g
v
ar
io
u
s
co
n
v
o
l
u
tio
n
tech
n
iq
u
es.
I
n
ad
d
itio
n
,
th
e
f
e
atu
r
e
s
elec
tio
n
is
p
er
f
o
r
m
ed
to
ex
tr
ac
t th
e
f
ea
tu
r
es with
less
t
im
e.
0
2
4
6
8
10
12
14
16
18
20
2
0
0
4
0
0
6
0
0
8
0
0
1
0
0
0
1
2
0
0
1
4
0
0
1
6
0
0
1
8
0
0
2
0
0
0
D
i
a
g
n
o
st
i
c
e
r
r
o
r
(
%
)
S
a
m
p
l
e
i
m
a
g
e
s
W
S
S
R
-
C
DC
M
F
DL
m
u
l
t
i
-
t
as
k
dee
p
l
ea
r
ni
ng
Evaluation Warning : The document was created with Spire.PDF for Python.